Microstructure and mechanical properties of repair welds of low-pressure sand-cast Mg-Y-RE-Zr alloy by tungsten inert gas welding

Xin Tong, Guohua Wu, Liang Zhang, Yingxin Wang, Wencai Liu, Wenjiang Ding

National Engineering Research Center of Light Alloy Net Forming and State Key Laboratory of Metal Matrix Composites, School of Materials Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, China

Abstract The sand castings of Mg-Y-RE-Zr series alloys are widely utilized in the large scales and complex shapes in the aerospace industry,as a result of which there are always some cast defects in the products.In this study, the feasibility of repair welding of sand-cast Mg-4Y-3RE-0.5Zr alloy by tungsten inert gas (TIG) welding was scrutinized with different welding currents from 150 to 210 A.The results indicated that defect-free repaired joints with good appearance could be acquired at 170 and 190 A.Interestingly, the grain size of the fusion zone (FZ) was refined initially and then increased with the linear increment of welding current.Because at the higher heat inputs, although the cooling rate of the molten pool was reduced, substantial constitutional supercooling for the grain refinement was attained after the Zr particles were transformed into Zr solutes.The tensile strength of the repaired joint at 170 A was 195MPa with the maximum joint efficiency of 87.8%, and the elongation reached to 124.4% of the sand-cast base material (BM).However, serious grain coarsening and continuous eutectic structures generated in heat-affected zone (HAZ) above 190 A resulted in the weakening of the joint due to the brittle intergranular fracture.

Keywords: Sand-cast Mg-Y-RE-Zr alloy; Repair welding; Microstructure; Tensile properties; Fracture behavior.

1.Introduction

Developing the lightweight magnesium (Mg) alloys with excellent performances has long been a significant target of the aerospace industry, where reducing the weight of components is of crucial importance[1,2].Compared with traditional Mg alloys, Mg-rare earth (Mg-RE) alloys have attracted considerable attention for their higher specific strength, better corrosion resistance, and especially the desirable creep resistance up to the high temperature of 523K [3-5].In particular,the Mg-Y-RE-Zr series alloys have become the most famous commercial Mg-RE alloys and been widely applied in aviation and aerospace fields [6,7].

From the practical needs of aerospace engineering, the large-scale and complex-shaped components are mostly produced by low- or differential-pressure sand casting process [8-10].However, because of the slow filling rate, long solidification time, and intricate mould cavity of the sand casting process, typical defects such as sand inclusions,shrinkage porosities, and microcracks can be produced during solidification procedure [11,12].Moreover, because RE elements exhibit even stronger tendencies to be oxidized than Mg (the oxidation tendency is Y＞Gd＞Nd＞Mg) [13], it is much easier to generate oxidation slags in Mg-Y-RE-Zr alloy castings with the high Y contents.Actually, some of these defects are inevitable in complex castings even processed in the modified gating and cooling systems [14].So it is essential to develop an economical repair welding technique with easy accessibility for Mg-Y-RE-Zr series sand castings.

Because of the utility and economy, TIG welding has been widely adopted to Mg alloys at the industrial level[15].Moreover, TIG welding can be performed manually, so it exhibits unique advantages for repairing inner structures and camberedsurfaces in comparison to some advanced welding technologies such as friction stir welding (FSW) [16], diffusion welding(DW)[17],and laser beam welding(LBW)[18].Recently,Meng et al.[14]proved the feasibility of TIG welding for repairing of as-cast Mg-6Gd-3Y-0.5Zr alloy,and the maximum tensile strength of the T6 treated joint reached up to 295MPa.Turowska and Adamiec [19]analyzed the microstructure of the TIG welds of WE43 alloy just under a fixed welding parameter and evaluated their mechanical properties at elevated temperatures.

However,the existing papers primarily investigated the butt joints of Mg-RE alloys [20,21], while the repair welding is usually applied for refilling some small holes after removing the local defects.Thus there may be a more severe tensile stress field and a higher cooling rate within the weld pool of repaired joints.It is unclear whether these differences have effects on the performance of repaired casting.On the other hand, most of the published works were focused on the welding of deformed Mg-RE alloys or these series alloys fabricated by permanent mould casting [22], which are quite different from the engineering application.The coarser grain size as well as the continuous eutectic structure in the sand-cast alloy is more easily to generate overburning and hot crack under the high temperature of TIG welding [17].Thus,optimized welding parameters in the previous researches may not be suitable for the sand-cast Mg-Y-RE-Zr alloy.

Liu et al.[23]concluded that, because of the high heat input of TIG welding, the grain size of FZ in TIG welded AZ31B alloy is the largest in comparison to other advanced welding methods.But for the welded Mg-Y-RE-Zr alloys,in addition to the weld heat input, also Zr can significantly affect the grain size [24-26].In Mg alloys, there are two different existence forms of Zr: Zr solute (Zrs) and Zr particle(Zrp).Zrpprovides the heterogeneous site for nucleation of the primaryα-Mg due to their similar lattice constants, while Zrsprovides substantial constitutional supercooling to restrict grain growth.The Interdependence Model [27,28]illustrated that both Zrpand Zrsare required to achieve the grain refinement, but a major contribution was provided by Zrs[29,30].So in the casting process, long holding time and repeated stirring are necessary to dissolve Zr in the Mg melt to achieve better grain refinement [30].

Similarly, the resulting Zr phases in Mg-Y-RE-Zr alloys will be influenced by the heat input induced by the TIG welding process.In other words, the grain size of FZ in this series alloy may be determined by the combined effects of the cooling rate and the existence forms of Zr.Qin et al.[31]prepared dissimilar joints between AZ61 and ZK60 (Zr content＞0.45wt%) Mg alloys by TIG welding, and the grain size in FZ first decreased and then increased with the linear increase of the welding current.The authors attributed this finding to the synergistic effect of the cooling rate and the heat input but without consideration of the Zr.However, up to now, there is no further research on the effect of Zr existence forms on the microstructure of Mg-RE-Zr alloys during the welding process, and the repair welding of sand-cast Mg-Y-RE-Zr alloy has not been paid much attention to.

Given these facts, in the present study, the repair welding of sand-cast Mg-4Y-3RE-0.5Zr alloy was performed by TIG welding under different welding currents.The qualities of the repaired castings were evaluated in terms of the microstructure evolutions and mechanical properties of the formed joints.We believe that these results will broaden our understanding of the repair welding of sand-cast Mg-RE-Zr alloys and further their applicability to the aerospace industry.

2.Experimental procedure

2.1.Preparation of the low-pressure sand-cast alloy

In this research, the low-pressure sand-cast Mg-4Y-2Nd-1Gd-0.5Zr alloy was chosen for repair welding experiments.The base material was fabricated by melting commercial pure Mg (＞99.95%), Mg-30Y, Mg-90Nd, Mg-90Gd and Mg-30Zr (wt%) master alloys in an electric resistance furnace protected by a mixed atmosphere of CO2(99 vol%) and SF6(1 vol%).After refined at 740°C for 10min, the melt was filled into a sand mould preheated to 200°C to obtain the alloy plates.The actual composition detected by Inductively Coupled Plasma-Atomic Emission Spectroscopy (ICP-AES)was listed in Table 1.Detailed information regarding the testing of the content of dissolved Zr (Zrs) can be found elsewhere [30,32].

Table 1The actual composition of the sand-cast Mg-4Y-2Nd-1Gd-0.5Zr alloy analysed by ICP-AES (wt%).

2.2.Repair welding process

The plates were cut into rectangular specimens with dimensions of 70mm×60mm×10mm for welding experiments.Afterwards, a blind hole (Ф 15 mm×H 5 mm) was processed by a pneumatic milling cutter at the centre of the plate for AC-TIG repair welding(Rehm INVERTIG.PRO digital 450 AC/DC).The chemical composition of the filler rod is shown in Table 1.Prior to welding, all the surfaces of the workpiece and the filler rod were ground with 400 # grit SiC paper and degreased in acetone for removing the surface attachments.In order to reduce the experimental error, all the experiments were performed via a set of self-designed welding jig and wire feeder.Fig.1 illustrates the schematic of the process procedure,and detailed information about the welding parameters are provided in Table 2.

Table 2The primary process parameters of the repair welding.

2.3.Characterization

X-ray computer tomography was used to detect the welding defects within the repaired joint.Cross-sections ofthe repaired plates were mechanically ground, polished, and etched in a solution of 4.2g picric acid, 10mL acetic acid,10mL water and 70mL ethanol for 12s.Macrostructure observation of the cross-section of the joint was carried out by a stereoscopic microscope (Stemi 305).Microstructure characterization was performed by an optical microscope (OM, Carl Zeiss Axivoert 40) and scanning electron microscope (SEM,NOVA NanoSEM 230) equipped with an energy-dispersive spectroscope (EDS).The grain sizes of the repair welded alloys were calculated by the image analysis software Image-Pro Plus 6.0.The phase composition was identified by X-ray Diffraction (XRD, Rigaku Ultima IV) with Cu Kαradiation.A differential scanning calorimeter (DSC, METTLER 1100LF) was used to measure the phase transition in FZ.The range of testing temperature was from 50 °C to 650 °C with a heating rate of 10 °C/min under a shielding atmosphere of Ar gas.Corresponding DSC data were processed by a software Netzsch Proteus Thermal Analysis to calculate the temperature and heat flow of the phase transition.

Fig.1.Schematic of (a) low-pressure sand casting process, (b) repair welding process, (c) repair welded joint, and (d) sampling location and geometry of the tensile specimen.

The hardness profiles were attained through cross-sections of the joints via a Vickers hardness tester (VH, HV-300) with a load of 49N for 15 s.Tensile specimens were cut in the horizontal direction 1mm from the upper surfaces of the cast plates after welding, and Fig.1d shows the sampling location and geometry of the tensile specimen.Before the tensile test, the fusion zone was marked on the surfaces of these tensile samples for determining the fracture location after the failure.The tensile properties of the joints were measured at a constant strain rate of 1mm/min on a tensile machine(Zwick/Roell Z100) at room temperature.

Fig.2.X-ray flaw detections of the repair welded joints fabricated with a welding current of (a) 150 A, (b) 170 A, (c) 190 A, and (d) 210 A.

3.Results

3.1.Macrostructural analysis

Fig.2 shows the analyses of the X-ray computer tomography of the representative repaired joints fabricated with different welding currents.The forming defects of the repaired weld can be easily recognized by the X-ray with a penetration direction from the upper surface of the cast workpiece.From Fig.2a, there was incomplete fusion at the edge of circular weld with 150 A.With the increasing welding current, no obvious welding defects could be found at the welding current of 170 and 190 A.However, an aggregate welding porosity in size of ～2mm appeared in the weld at 210 A, as shown in Fig.2d.Moreover, it can be seen that the contrast of all the welds was darker in comparison to those of the BM under the X-ray inspection.

The macrosection of the repair welded joints are shown in Fig.3.Although the dimensions of these blind holes before repairing were the same, the shapes and sizes of the welds were significantly different after welding with different welding currents.As the welding current increased, both of the penetration depth as well as the processed width were increased due to the increasing heat input.Meanwhile, the aspect ratios between the penetration depth and processed width of the repair welds also increased with the increase of welding current.Some key transition zones such as FZ, HAZ, and BM labeled in Fig.3 are the consequence of steep temperature variation in various regions.

3.2.Microstructural analysis

The microstructure of the sand-cast Mg-4Y-2Nd-1Gd-0.5Zr alloy is shown in Fig.4.The microstructure primarily consists of equiaxedα-Mg matrix surrounded by eutectic structure (Fig.4a), which is proved to be a dendrite growth mode.A small number of divorced eutectics can also be observed in the triple junction of grain boundaries.The average grain size of theα-Mg matrix was approximately 50μm.As depicted in the SEM images and corresponding elemental distribution maps (Fig.4b and c), the lamellar eutectic structures in network shape distributing along grain boundaries were mainly rich in Y, Nd, Gd, and Zr elements.

Fig.5 exhibits the optical microstructures of FZ, HAZ of the joints performed at different welding currents.Different transition zones presented different organizational characteristics.Compared with the BM, the grain sizes in FZ were obviously refined, while grain sizes in HAZ near FZ were increased.The quantitative calculation of grain sizes of HAZ and FZ within different joints are shown in Fig.6, and the grain size (48μm) of the sand-cast alloy is shown as a horizontal black line for comparison.With the lowest welding current, the grain size of HAZ was just increased by 5μm compared with that of the BM.With the increase of the repair welding current, the grain sizes of the HAZ increased obviously, and reached the maximum value (79μm) at 210 A.In addition, the eutectic phases distributing along the grain boundaries became more and more continuous with coarser morphologies.However, the grain size of the center of FZfirstly decreased and then increased with the increasing welding current, achieving its minimum value (13μm) at 170 A.It is worth noting that at the lowest welding current of 150 A,some black particle aggregated at the grain boundaries in the FZ which were significantly different from those eutectic phases.But with the increase of the welding current, the number of black clusters decreased gradually.

Fig.3.Macrographs of the repair welded joints repaired with a welding current of (a) 150 A, (b) 170 A, (c) 190 A, and (d) 210 A.

Fig.4.(a) Optical and (b) SEM micrographs of low-pressure sand-cast alloy.(c) Corresponding EDS maps of SEM micrograph in (b).

Fig.7 displays the SEM microstructures and corresponding EDS maps of the transition interface between HAZ and FZ of these joints.As the welding current increased, the eutectics presented in HAZ in the upper-left region of these SEM images became coarser, which was consistent with the analyses in Fig.5.At the same time, some bright white particles in the FZ in the lower-right region, which have been detected as the black clusters in the optical observation in Fig.5, were mainly rich in Zr and Y elements according to the EDS maps.The number of these bright white particles in the FZ also gradually decreased with the increase of welding current under SEM mode.Compared with the elemental distribution of the transition interface of 150 A (Fig.7a), Zr and Y elements were more evenly distributed in that of 210 A, as shown in the EDS maps in Fig.7d.

Fig.8a-d show the SEM microstructures in the center of FZ.The bright white particle clusters also existed with the lowest welding current, which was in good agreement with the observation in Fig.7.From the images at high magnification and corresponding EDS analysis in Fig.8e and f, it can be concluded that these flocculent clusters should be Zr-Y phases.Except for these clusters, the eutectic structure was fewer in the FZ with the lower welding current (Fig.8a).However, with the increase of the current, semi-continuous corrosion pits appeared along the grain boundaries, which should be produced after the erosion of the eutectics during the etch process.This indicated that the eutectics were more likely to appear in FZ with a higher welding current.

XRD patterns of the sand-cast alloy and FZ of the joints repaired with different welding currents are shown in Fig.9.After repair welding process, the phases detected in FZ were the same as that in the sand-cast state.All of the samples were composed ofα-Mg solid solution, Mg24(Gd, Y)5,Mg12Nd, Mg14Nd2Y, and Mg5(Gd, Y) intermetallic compounds (Fig.9a), indicating that there were no new phases generated in the welding process.However, theα-Mg matrix in FZ formed with different welding currents exhibited different positions of diffraction peaks.Taking the three strongest peaks ofα-Mg in Fig.9a as examples, all the diffraction angles of the peaks were lower than the standard values(Fig.9b).As the welding current increased, the peak positions ofα-Mg in FZ first shifted to left gradually, reaching the minimum values at 190 A.And then the peaks shifted to the right slightly when the welding current continuously increased to 210 A.It can be known from the Bragg’s law that the interplanar spacing ofα-Mg matrix in FZ increased first and then decreased with the increasing welding current.

Fig.5.Optical microstructures of HAZ (a, c, e, g) and FZ (b, d, f, h) formed with different welding currents: (a, b) 150 A, (c, d) 170 A, (e, f) 190 A, and(g, h) 210 A.

Fig.6.The calculation results of grain sizes of the HAZ and FZ within different joints.

The DSC analyses of the BM and FZ are provided in Fig.10.According to the previous researches and Mg-REphase diagrams, the two endothermic reactions occurred at～540°C (labeled as B) and ～630°C (labeled as C) can be speculated to be the eutectic decomposition and the melting ofα-Mg matrix, respectively [3].Not only the cast alloy but also the FZ prepared at 210 A had eutectic decomposition in the DSC curves.Interestingly, another endothermic reaction with a wider temperature range started at ～350 °C (labeled as A) just in the four DSC curves of FZ.Table 3 shows the calculated heat flow value of the peaks A and B as well as the melting temperature ofα-Mg matrix in these curves.It can be seen that the melting temperature ofα-Mg first decreased and then increased with the increase of welding current, showing a minimum melting point (637.7°C) at the current of 190 A.

Table 3Calculated heat flow value of the peaks A and B as well as the melting temperature of α-Mg.

Fig.7.SEM micrographs and corresponding EDS maps of the transition zones between HAZ and FZ formed with different welding currents: (a) 150 A, (b)170 A, (c) 190 A, and (d) 210 A.

Fig.8.SEM micrographs of the FZ formed with different welding currents: (a) 150 A, (b) 170 A, (c) 190 A, and (d) 210 A.(e) SEM image and corresponding EDS maps of the particle cluster in FZ with the welding current of 150 A.(f) is the EDS point analysis in (e).

Fig.9.(a) XRD spectra of the BM and FZ in the joints with different welding currents.(b) XRD spectra with the scanning angle from 31° to 37° in (a).

Fig.10.DSC analysis of the FZ in the joints repaired with different welding currents.

Fig.11.Vickers hardness distributions of repair welded joints from FZ to BM.

3.3.Vickers hardness and tensile properties

Fig.11 shows the Vickers hardness profiles of the repaired joints.The hardness of the sand-cast alloy (69 HV) is exhibited as a horizontal black line for comparison.The Vickers hardness of the FZ in four repaired joints was higher than that of the sand-cast alloy.The average Vickers hardness of the FZ of the joint repaired with 150 A was 75 HV, and then realized the maximum value of 77HV at 170 A, which was 11.6% higher than that of the sand-cast state.However, as the current further increased to the maximum current of 210 A,the Vickers hardness was reduced apparently to 72 HV.Further, it can be seen that the widths of the FZ and HAZ were all increased gradually with the increasing current, and the larger the welding current, the more severe the softening of the HAZ.

As shown in Fig.12a, the tensile tests were conducted to investigate the effects of welding current on the tensile properties of the repair welded joints.The comparison of 0.2% tensile yield strength (YS), ultimate tensile strength (UTS), and elongation (EL) values between the repaired joints (columns)as well as the sand-cast alloy (horizontal dotted lines) are shown in Fig.12b.With the increase of the welding current,the UTS of the repaired joints first increased from 178MPa(150 A) to the maximum value of 195MPa (170 A) withthe highest joint efficiency of 87.8% and then it decreased to 149MPa at 210 A.The YS values also showed the similar trend.What is more, the EL of the joints repaired at 170 A was 5.1%, which was unexpectedly higher than that of the sand-cast alloy by 24.4%.However, as the welding current increased to 210 A, the EL of the repaired joint was sharply reduced to 1.4%.

Fig.12.(a) Engineering stress-strain curves and (b) tensile properties of the BM and repair welded joints repaired with different welding currents.

Fig.13.(a) Fracture locations of the repair welded joints.(b-h) Tensile fractures of the BM and joints repaired with different welding currents: (b) BM, (d)150 A, (e) 170 A, (f) 190 A, (g, h) 210 A.(c) is the magnified view of the square region in (b) and (i) is EDS point analysis of the cluster in (d).

3.4.Fracture behavior

Fig.13a shows the failure locations of the repaired joints.It can be observed that the joints fabricated with the minimum (150 A) and maximum (210 A) currents all failed in FZ,while the joints of 170 and 190 A failed at HAZ.Fig.13b shows the SEM image of the typical fracture features of the sand-cast alloy at room temperature.The fracture surface exhibited a mixed characteristic of transgranular and intergranular fractures with some fractured eutectics structures, coarse dimples,and cleavage planes.The magnified view of the fractured eutectics structures with a lamellar shape is presented in Fig.13c.

Due to the different failure locations, the tensile fracture surface of 150 A (Fig.13d, failed at FZ) shows much finer structure characteristics in comparison to those of 170 A and190 A (Fig.13e and f, all failed at HAZ).The fracture surface of the joint with 150 A consisted of many coarse dimples and tear ridges, with the main fracture characteristic of quasicleavage.Notably, some bright white clusters were observed at the torn grain boundary, the EDS point analysis (Fig.13i)proved that they were Zr-Y particles which have also been observed in OM (Fig.5) and SEM (Fig.8) images in the FZ repaired at 150 A.Although both the joints with 170 A and 190 A were broken at HAZ, their fracture morphologies were significantly different.The fracture surface of the joint formed with 170 A was characterized by some fractured eutectics structures, cleavage planes, and a small number of dimples, which showed a similar fracture mode with the BM in Fig.13b.However, the joint of 190 A exhibited a relatively flat fracture surface comprised of more fractured eutectic structures and river patterns, suggesting the main fracture characteristic of brittle cleavage fracture.As can be seen in Fig.13g, the macro fracture surface of the joint fabricated with 210 A could be divided into two regions with different fracture characteristics, distinguished by the yellow dotted line.The upper left region containing an aggregate welding porosity shows a finer fracture structure than the lower right region, which demonstrated that the upper left region located in FZ while the lower right region was HAZ.It can be speculated that the cracks may initiate near the welding porosity in FZ, then propagate toward HAZ in the lower right corner as indicated by the yellow arrows, and finally resulting in the fracture failure, which was in accordance with the lowest elongation of 1.4%.

Fig.14 exhibits optical micrographs of the longitudinal sections of these ruptured tensile samples.As mentioned above,the failure modes of the BM(Fig.14a)and the joint repaired at 170 A(Fig.14c)all showed the coexistence of transgranular and intergranular fracture.Fig.14b and d present the initiation and propagation process of the cracks, it can be seen that the cracks initiated in the eutectic structures along grain boundaries and then propagated toward the interior of the grain.However, because of the much coarser grains surrounded by continuous eutectics in HAZ of the joint repaired with 190 A, the crack primarily grown along the brittle eutectics, as shown in Fig.14f.Obviously, the tensile failure in FZ of the joints with 150 A (Fig.14g) and 210 A (Fig.14h)is mainly transgranular fracture.It is worth noting that some of the cracks in Fig.14g was generated near the clusters at grain boundary which have been detected in Fig 13d, indicating that these clusters may be detrimental to the mechanical properties of the repaired joint.

4.Discussion

In addition to the heat input, the grain size of FZ was also affected by the existence of Zr which has been considered as a dominant grain refiner for a range of commercial magnesium alloys [24-26].Therefore, it is necessary to detect the existence forms of Zr in FZ fabricated with different welding currents.Table 4 exhibits the compositions including the content of Zrsof FZ in the repaired joints.It can beseen that with the increase of welding current, the ratio of the content of solute Zr to the total Zr content increased and became closer to 1.This is because the temperature of the molten pool was higher with a larger welding current, so the diffusion rate of Zr atoms from the surface of Zr particles into the liquid phase was increased simultaneously.Secondly,as the heat was conducted from the molten pool to the BM during repair welding, and the larger the current, the higher the temperatures of BM and FZ, resulting in a lower cooling rate of the molten pool [31,33].After the extinction of the arc, the residence time of the molten pool at high temperature would be prolonged, therefore it provided a more sufficient time for the Zr diffusion.Thirdly,the disturbance effect of the arc also played a significant role in the dispersion of the Zr particles; the agglomerated Zr particles could be dispersed by the stronger arc force with a higher welding current, which can further facilitate the dissolution of the clusters.As a result, it can be seen from Figs.5, 7, and 8 that the number density of Zr clusters decreased apparently as the welding current increased from 150 to 210 A.That is, with the increasing welding current, the content of Zrpin FZ gradually decreased, while the content of Zrsincreased.

Table 4Actual compositions of FZ in the four joints analysed by ICP-AES (wt%).

XRD results in Fig.9 show that the three strongest peaks of theα-Mg matrix gradually shifted to lower diffraction angle values with the increasing welding current from 150 to 190 A.This suggests that the Zr clusters were dissolved to be Zrsin a higher welding current, and then Zrswas retained in the matrix under the condition of rapid cooling of the molten pool, leading to the formation of a supersaturated solid solution in FZ.According to Bragg’s law, the lattice distortion and crystal plane spacing of theα-Mg matrix increased with the increasing welding current, so the corresponding diffraction angle decreased.However, when the current was further increased to 210 A, the BM was heated to a higher temperature, thus it started to have negative impacts on the cooling rate of FZ.During the relatively slow solidification procedure with the highest welding current, theα-Mg dendrites in FZ could continuously discharge solute atoms into the intergranular residual liquid,which increased the content of alloying elements in the liquid phase [3].Eventually, the semi-continuous eutectic structures were obtained along theα-Mg grain boundaries, as confirmed by the corrosion pits at the grain boundaries in Fig.8.Due to the consumption of the solute elements in eutectic structures, the solid solubility of the matrix was reduced, leading to a decrease in the interplanar spacing and an increase of the diffraction angle value in XRD analysis.

Fig.14.Optical micrographs of the longitudinal section of the fracture surface of BM and joints repaired with different welding currents: (a) BM, (c) 170 A,(e) 190 A, (g) 150 A, (h) 210 A.(b), (d) and (f) is the magnified view of the square region in (a), (c) and (e), individually.

Based on the DSC analyses in Fig.10 and the corresponding quantitative calculation in Table 3, only the sample cut from FZ of 210 A showed an obvious heat flow at ～540°C.This indicated that except for the FZ of 210 A, the number of the eutectics was very low in the other three samples fabricated with lower currents.On the other hand, the melting point of theα-Mg matrix with the welding current from 150 to 190 A showed a decreasing tendency, indicating that the solid solubility of the matrix was gradually increased with the increasing current [34].That is, more alloying elements were dissolved into the matrix in FZ with the lower three currents rather than used to form eutectic structures, which wasconsistent with the XRD analyses.As the current further increased to 210 A, the solid solubility in the matrix was then reduced on account of the formation of eutectic structures,thus resulting in a slight increase in the melting point of the matrix.The endothermic peaks (like peak A) that occurred before eutectic decomposition also have been detected in the DSC of some solution-treated alloys, which could be ascribed to the dissolution of the precipitates [35,36].Due to the rapid cooling of the molten pool, a large number of solute elements could be dissolved into theα-Mg matrix in FZ, leading to the formation of the non-equilibrium microstructures [37].Similarly, during the heating process of the DSC test, theβ1andβphases precipitated first and then were dissolved at a higher testing temperature, resulting in the formation of endothermic peak A [35].In contrast, the relatively slow cooling rate of the sand-cast alloy resulted in the lower solid solubility of the matrix, thus there was no reaction peak before the eutectic decomposition.

Fig.15.Mechanism schematics of the microstructure formation in molten pools formed with different welding currents: (a-c) 150 A, (d-f) 170 A, (g-i)210 A.

Fig.15 illustrated the mechanism schematics of the microstructure evolution in the molten pools.Because of the minimum heat input with the welding current of 150 A, a large number of Zr particles cannot be dissolved to produce the constitutional supercooling effect due to the low temperature of the molten pool.On the other hand, the size of the Zr particle as a heterogeneous nucleus is also a key factor for the grain refinement [30,34].It was difficult for the Zr cluster with a large size to act as the effective heterogeneous nucleus in the molten pool.These above two factors all reduced the refinement effect of Zr in the repaired weld.With the growth ofα-Mg dendrites,these Zr clusters would be pushed towards the grain boundaries after solidification of FZ, as shown in Figs.8a and 15c.

As exhibited in Fig.15d-f, as the welding current increased to 170 A, some of the Zr clusters were dissolved into the liquid phase and transformed to be Zrs.A large number of previous works have demonstrated that Zrsprovided constitutional supercooling for the nucleation ofα-Mg dendrites on undissolved Zrpand actually a major contribution to grain refinement was provided by Zrs[30,32].The refinement effectiveness of solute elements is determined by its growthrestriction factorQwhich can be expressed as [32]:

where the subscriptirepresents different alloy,miandkiis the slope of the liquidus and the solute partition coefficient of the corresponding binary phase diagram, respectively.Cimeans the solute content of the binary alloy.Alloying elements with higherQvalue have better grain refinement effect through constitutional supercooling.The values ofm(k?1) for Y and Zr used for this calculation were 1.7 and 38.3[32,38],thus Zr showed a much better refinement effect than Y.As a result,the grain size in FZ decreased from 20μm to 13μm when the welding current increased from 150 to 170 A.

However, although increasing the welding current would lead to a more complete dissolution of the residual Zr particles, the higher heat input applied to FZ reduce the cooling rate of the molten pool, as mentioned above.At this point,the lower cooling rate became the dominated factor of the grain size, which led to the coarsening of the grain and even the formation of the eutectic structures along the grain boundaries at 210 A, as shown in Figs.8d and 15i.Furthermore,aggregate welding porosity also appeared in FZ, due to the air indraft in the molten pool at higher temperatures.

Due to the higher cooling rate of the weld, FZ in the repaired joint possessed a smaller grain size with a higher solid solubility of the matrix than those of the BM [19,22].The resulting strong grain boundary strengthening and solid solution strengthening lead to the higher hardness of the repaired weld in comparison to that of the BM [39,40].As the current was increased to 170 A, Zr clusters in the molten pool with a higher temperature were transformed into Zrsto refine the grain size to the minimum (13μm), so the hardness of FZ was increased from 75 HV (at 150 A) to 77 HV (at 170 A).However, the cooling rate of FZ decreased with the further increase in the current, and the grains began to coarsen.Especially in the FZ formed with 210 A, the eutectic structures formed along the grain boundaries decreased the solid solubility of theα-Mg matrix, which would weaken the above two strengthening mechanisms.Therefore, the hardness of FZ fabricated with 210 A was drastically reduced to 72 HV.

During the repair welding process, the grain coarsening occurred in the regions near the molten pool to reduce the interfacial energy, resulting in the formation of HAZ [31,41].Moreover, when the temperature of the HAZ was above the melting point of the low-melting eutectics, these eutectics along the grain boundaries were melted locally and then solidified again.In addition, the eutectic structures in HAZ would become more continuous due to the lower volume fraction of the grain boundaries after the grain coarsening, as shown in Fig.5.Previous researchers believed that this microstructure with the characteristic of coarse grains surrounded by continuous eutectics possessed lower intergranular bonding strength,weakening the hardness of HAZ [14].From the DSC analysis in Fig.10, the critical temperature for the decomposition of the eutectics was 540°C.As the welding current increased,the area of the BM affected by heating effect became larger,leading to a wider HAZ.Therefore, as the welding current was increased from 150 to 210 A, the width of the HAZ was increased from 1mm to 2.5mm, and the minimum hardness in HAZ was also reduced from 67 HV to 62 HV.

The joint of 150 A contained incomplete fusion defects in the repaired weld, as demonstrated in the X-ray flaw detections in Fig.2a.Some Zr clusters were also exposed to the fracture surface in Fig.13d, confirming the poor bonding force between the cluster and the matrix.These two factors deeply deteriorated the mechanical properties of the joint repaired with 150 A;the EL of this joint was only 2.5%,and the tensile fracture located in the FZ(Fig.13a).The joint repaired at 210 A also exhibited a similarly poor UTS of 149MPa and even lower EL of 1.4% owing to the welding porosity with a large size generated in the FZ during the welding (Fig.13g).

Well-formed repair welded joints have been obtained by proper heat inputs at 170 and 190 A; there were no obvious welding defects inspected from the X-ray detections, thus the mechanical properties of the repaired joints were higher than those of the others.Moreover,the number of Zr clusters in FZ of 170 A and 190 A were significantly reduced, resulting in the lower probability of crack initiation in FZ.To investigate the reason of different fracture locations in the repaired welds of 170 and 190 A, grain boundary strengthening in the three regions were estimated by Hall-Petch relationship [42]:

whereσ0is the friction stress free from grain boundary contributions,kyis the stress concentration factor, anddis the average grain diameter.The constantkyfor this alloy was estimated to be 250MPa μm1/2[7].Taking the joint of 170 A for an example, the contribution of grain boundary strengthening in FZ, HAZ, and BM was 35.7MPa, 30.8MPa, and 69.3MPa,respectively.It can be seen that HAZ may be the weakest region in the whole joint due to the presence of coarser grains surrounded by continuous eutectics.This is the reason why both the joints of 170 and 190 A were fractured in the HAZ and the joint of 190 A even showed a slightly lower mechanical performance than that of the 170 A.Moreover, many researches illustrated that continuous eutectics can function as crack initiation sites during tensile tests [3, 7], seriously reducing the ductility of the sample.Although the UTS of the repaired joint with 170 A was only 87.8% of the sand-cast BM, the EL of the joint was surprisingly higher than the BM by 24.4%.The improvement of the EL could be attributed to that the fine grains in FZ restrained the crack initiation during the tensile process by affording more deformation before the fracture failure.

As shown in Fig.13b and c, the sand-cast alloy exhibited fractured eutectic structures, which was consistent with the optical microstructure of the longitudinal section of the fracture surface (Fig.14a and b).This reconfirmed the poor combination between the eutectic structures and theα-Mg grains[7, 34, 43].The joints repaired with 150 and 210 A were all broken in FZ due to the welding defects.Compared with the sand-cast BM, the microstructure of FZ showed fewer eutectic structures thus stronger intercrystalline bonding force.Therefore, the longitudinal section of the fracture surface of150 and 210 A all exhibited obvious transgranular fracture characteristics (Fig.14g and h).It is worth noting that the Zr clusters, which may act as crack initiators, deteriorated the ductility of the repaired joint [30].So these cracks in the FZ of 150 A tended to initiate near the Zr clusters at the grain boundaries, and then gradually propagated into the interior of grains, which was different from the failure procedure in the joint of 210 A.

Although both the joints of 170 and 190 A were failed at HAZ, the grain coarsening occurred in HAZ of the 170 A was not as severe as that of the 190 A, and the eutectic structure in HAZ of 170 A was semi-continuous.Therefore, the failure mode in the repaired joint of 170 A was similar to that of the sand-cast alloy with a mixed feature of transgranular and intergranular fracture.When the welding current was increased to 190 A,coarser grains surrounded by more continuous eutectics in HAZ were attained by the higher heat input[31].The cracks initiated at the interface between the eutectics and matrix, and then mostly developed along the continuous eutectics, leading to the main characteristic of intergranular fracture (Fig.14e and f) [3, 7].Therefore, there were much more fractured eutectics and river patterns appeared in the fracture surface with a relatively flat morphology (Fig.13f).The continuous brittle eutectics in HAZ were also the main reason for the cracks to propagate to the HAZ from FZ in the joint of 210 A, as mentioned in Fig.13g.

5.Conclusions

This research verified the feasibility of repair welding in sand-cast Mg-4Y-2Nd-1Gd-0.5Zr alloy by TIG welding,and the qualities of the repaired joints were evaluated in terms of the microstructure evolutions and mechanical properties.The following conclusions can be drawn:

1.Some welding defects of incomplete fusion and welding porosity were generated in the fabricated welds with the current of 150 and 210 A, respectively, while the defectfree repaired joints with good appearance could be acquired at 170 and 190 A.

2.With the welding current increased of from 150 to 210 A,the mechanical properties of the TIG repair welds of sandcast Mg-4Y-2Nd-1Gd-0.5Zr alloy frist increased and then decreased, showing a maximum tensile strength of 195MPa with the joint effciiency of 87.8% at 170 A, and the elongation reached 124.4% of the original sand-cast alloy.

3.Although there was no obvious grain coarsening that occurred in the HAZ of the repaired joint fabricated with 150 A, not only incomplete fusion defects existed in the weld, but also some Zr clusters were observed along the grain boundaries, which deteriorated the mechanical properties of the repaired joint.

4.The grain size of FZ in Mg-4Y-2Nd-1Gd-0.5Zr alloy was determined by the combination of the cooling rate and the existence forms of Zr.Although increasing the welding current could reduce the cooling rate of the molten pool,it also facilitated the transformation of Zr particles to Zr solute which provided constitutional supercooling for the grain refinement.Thus the minimum grain size (13μm)was obtained in the FZ of 170 A.

5.Obvious grain coarsening and continuous eutectic structures occurred in HAZ above the welding current of 190 A,resulting in the intergranular fracture in HAZ.A small number of eutectics also be detected in the FZ as the welding current further increased to 210 A.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This research was supported by the National Natural Science Foundation of China (Nos.51775334 and 51821001),the National Key Research & Development Program of China(No.2016YFB0701205) and the National Science and Technology Innovation Special Zone Project (No.002-002-01).