Quasi-in-situ investigation into the microstructure and texture evolution of pur

Zhou Mengrn,Sun Yufeng,Morisd Yoshiki,Ushiod Kohsku,Fujii Hidetoshi,∗

a Joining and Welding Research Institute,Osaka University,Osaka 567-0047,Japan

b The State Key Laboratory of Tribology,Department of Mechanical Engineering,Tsinghua University,Beijing 100084,China

c School of materials science and engineering,Zhengzhou University,Zhengzhou 450001,China

Received 3 March 2020;received in revised form 27 May 2020;accepted 28 May 2020 Available online 11 September 2020

Abstract The quasi-in-situ microstructure and texture evolution along the real flow path of pure magnesium during friction stir welding were investigated.Five representative stages were involved from the base metal to the formation of the final stir zone.The material experienced compression,preheating,acceleration,deceleration,and subsequent annealing over the course of the welding process.A highly concentrated(0001)texture,denoted as“orientation convergence”,was initiated at the beginning of the acceleration stage(shearing deformation zone)in front of the tool.Both continuous and discontinuous dynamic recrystallization occurred simultaneously in the stir zone,and continuous dynamic recrystallization was determined to be the primary recrystallization mechanism.The marker material morphology and EBSD data were used to elucidate the in-situ evolution of the shear direction and shear plane along with the real flow path.

Keywords:Magnesium;Friction stir welding;Texture evolution;Material flow;EBSD.

1.Introduction

Friction stir welding(FSW)was invented by TWI in 1991,and has been demonstrated to be an ideal welding method for many materials[1-5].Since it is a solid-state joining method,it overcomes numerous problems associated with traditional fusion welding,such as crack formation,porosity,the need for a shielding gas,and significant energy consumption[6].To expand the applications of FSW,many efforts have been made to establish an accurate relation between the microstructure and the welding process[7-15].However,it is still challenging to conduct an in-situ investigation of the microstructure and texture evolution throughout the entire friction stir welding process because the material deforms under the tool at a considerably high strain rate and high temperature.Recently,some efforts have been made to understand the temperature and stress state during FSW using neutron diffraction[16,17].Although neutron diffraction can provide in-situ information about the temperature and stress distribution,it provides no intuitive microstructure information.Therefore,most studies concerning friction stir welding are predominantly carried out by investigating the final state of the stir zone(SZ)[2-5,7-17].

The application of the marker material combined with the quick stop(stop-action)was first proposed by Colligan[18].Liu et al.recently reported that a marker material was inserted using the stop-action technique to track the quasi-in-situ microstructure evolution during FSW[19],and the real flow path could be intuitively confirmed by observing the morphology of the marker material.Liu and Nelson successfully clarified the material flow and microstructure evolution of FSW steel utilizing the marker material and the“stop-action”technique[20,21].Furthermore,rapid cooling makes it possible to freeze the material in the state just as it experiences deformation[22].The combined use of an inserted marker material,the tool stop-action technique,and the rapid cooling method have allowed researchers to accurately track the evolution of the in-situ microstructure along the real flow path during FSW[23,24].

The stop-action method has been used to reveal the quasiin-situ microstructure and texture evolution of pure Al,Cu,Cu-Zn alloy,pure Fe,and steel[19-24].However,the materials studied in previous studies have been mainly limited to body-centered cubic(BCC)and face-centered cubic(FCC)metallic materials,while few studies have examined hexagonal close-packed(HCP)metallic materials.Magnesium(Mg),its alloys are representative HCP metals which have attracted recently significant attention due to their attractive properties,such as high specific strength,high specific stiffness,and damping capacity[25-30].Mg alloys can be used to replace the heavier Al alloys currently used in vehicles,potentially increasing fuel efficiency.Therefore,revealing the microstructure and texture evolution of Mg during FSW has both academic and industrial significance.

Previous studies have been made to explain the microstructure and texture evolution of HCP Mg alloys formed by FSW,and the development of the parallel probe’s basal texture in Mg alloys subjected to FSW has been widely reported[31-33].Sato et al.presented the final flow pattern consisting of multiple layers of onion-rings,and schematically indicated the shear plane distribution[31].However,these studies have mainly focused on the final state of the SZ,and detailed investigations into the microstructure evolution during FSW are urgently needed.Several questions remain unanswered,including the primary recrystallization mechanism during material flow and the origin of the highly-oriented(0001)texture.The latest studies have attempted to answer these questions by focusing on the keyhole left by the FSW process[34,35].By observing the microstructural development around a keyhole,it was revealed that the material undergoes several stages of deformation and metallurgical phenomena,such as compressive deformation,shear deformation,recrystallization,and grain growth before developing its final microstructure.Cooling assisted by ice water has also been used to freeze the microstructure in order to make close in-situ observations[35].It was revealed that both discontinuous dynamic recrystallization(DDRX)and continuous dynamic recrystallization(CDRX)occurred during the deformation,and CDRX was the dominant process[35].The highly-concentrated(0001)texture,denoted“orientation convergence”,was suggested to originate from shearing deformation.

However,the findings of Mironov et al.[35]focused mainly on the TD-ND and the WD-ND planes.For a further understanding of how the material flow bypasses the tool requires academic attention.Liu et al.[36]proposed that a so-called“rotation flow zone”exists,in which material remarkably close to the tool moves around the tool more than once during the FSW.Most of the material never reaches the tool surface and is instead delivered to the backside of the tool by following the flow paths.Therefore,the evolution of Mg during the FSW should be elucidated by tracing the real material flow path.

Fig.1.Schematic illustration of the experiment design.The AS and the RS mean the advancing side and the retreating side of the welding process.The welding direction,the transverse direction,and the normal direction are abbreviated as WD,TD,and ND,respectively.

In this study,the microstructure and texture evolution of pure Mg during FSW along the real deformation path were investigated by the combined use of stop action,marker insert method,and rapid CO2cooling together with an electron back-scatter diffraction(EBSD)microstructural analysis.

2.Experimental

Hot-extruded pure Mg(purity>99.95%)plates with a width of 70mm,a length of 200mm,and a thickness of 3mm were used in this study.Mg-Li-Zn alloy LZ91 foil was chosen as the marker material because of its excellent ductility at elevated temperatures.The foil was prepared using an electric discharge wire cutting machine(EDM,Sodick AG360L)into a thickness of 0.4mm,a length of 60mm,and a height of 3mm,which is equivalent to the thickness of the pure Mg plate.Before welding,the surfaces of the pure Mg plates and the LZ91 marker were carefully polished by SiC abrasive paper and cleaned with acetone to remove surface oxidization.The LZ91 marker was placed between two pure Mg plates,and FSW with rapid CO2cooling and stop-action was carried out in this study.The layout of the tool,the marker foil,and the welding method are schematically illustrated in Fig.1.Liquid CO2was ejected into the keyhole during the welding until the tool stopped and was completely pulled out.The use of rapid CO2cooling introduced a considerably high cooling rate,capable of“freezing”the microstructure at the keyhole[23].Combined with the stop action,the rapid CO2cooling can preserve the material around the tool in its deforming state.A thread-free SKD61(equal to AISI H-13 steel)tool was used for the welding(for schematic information about the tool,see Supplementary Fig.S1).The tool used in this study had a shoulder diameter of 15mm,a probe diameter of 6mm,and a probe length of 2.8mm.The tool was tilted 3° with respect to the insertion direction,and was set to weld at a rotation speed of 1700rpm and a traveling speed of 100mm/min.A K-type thermocouple was placed at the bottom of the material to record the welding temperature,and was used to confirm that the liquid CO2cooling barely affected the peak temperature,while it dramatically increased the cooling rate(see Supplementary Fig.S2).

After welding,the material around the keyhole was taken out to conduct microstructure observations(red dotted line area in Fig.1).Two representative layers were chosen to investigate the microstructure and texture evolution:(1)0.5mm from the upper face of the SZ,where the material flow was mainly affected by the shoulder of the tool,called the shoulder layer;and(2)0.5mm from the lower face of the SZ,where the probe dominantly influenced the material flow in this layer,called the probe layer.

Optical microscopy(OM,Olympus BX51M)was used to define the coordinates and to evaluate the morphology of the marker material in both layers after welding.Before OM observations,samples were chemically etched with a solution consisting of 10g picric acid,175mL ethanol,25mL acetic acid,and 25mL distilled water.A scanning electron microscope(SEM,JOEL JFM-7001FA)equipped with an EBSD and TSL software was used to characterize the microstructure and crystal orientation.The EBSD analysis was carried out at the inner side close to the marker for both the shoulder and probe layers.To evaluate the microstructure and texture change with subsequent annealing,the keyhole obtained from the rapid CO2cooling FSW was additionally annealed to 480°C for 20s to compare the microstructure in the SZ with the usual FSW sample without rapid CO2cooling.The annealing temperature of 480°C(0.81 Tm)was chosen because it was the peak temperature recorded during the FSW by the K-type thermocouple(see Supplementary Fig.S2).

The OM and EBSD samples were prepared by mechanical polishing.Since pure magnesium is prone to form mechanical twins during the polishing,the following sequence with specific precautions were carried out for the specimen preparation:(1)mechanical polishing samples obtained from EDM using #4000 grit abrasive paper for 5 min;(2)buff with 1μm and 0.1μm alumina suspensions for 10 min and subsequently cleaning in high purity ethanol with ultrasonication for 5 min;(3)buffing by a 40nm colloidal silica suspension(OP-S,Struers)for 20 min and cleaning in high purity ethanol with ultrasonication for 5 min.For EBSD,an additional electrolytic polishing at 15V for 30 s in a solution consisting of 10mL of perchloric acid and 90mL of ethanol at−30°C.

The conditions for the EBSD-SEM observation were set to 15kV accelerating voltage and an EBSD scanning step of 0.5μm.The analyzed area was 180μm×200μm for the rapidly cooled samples,while an enlarged area of 360μm×400μm was for the annealed and normal FSW samples.The average confidence index(CI)of all the EBSD scanned regions was 0.71.Grains smaller than two pixels were automatically removed using the grain-dilation function to remove scan noise.The criterion for low-angle boundaries(LABs)and high-angle boundaries(HABs)was set to 2°-15° and>15°,respectively.Black,white,and blue lines were used in the inverse pole figures(IPF)maps to represent the HABs,LABs,and{10–12}twin boundaries,respectively.The kernel average misorientation(KAM)map was created using the average misorientation angle of a given point on an EBSD map with its adjacent points and was used to estimate the local misorientation and dislocation density levels.

Fig.2.Optical micrographs showing typical morphology of the marker material and EBSD analyzed areas(blue squares)together with the coordinate x(mm)as a distance from key hole center at(a)the shoulder layer and(b)the probe layer.The critical coordinates of each stage are shown by blue dotted lines.The black dotted lines represent the edge of the shoulder.The morphology of the marker material after welding is shown by red dotted lines.

3.Results and discussion

Fig.2 shows the morphology and coordinates of the marker material after welding of(a)the shoulder layer and(b)the probe layer(see Supplementary Fig.S3 for higher resolution optical images of the marker material).The blue squares in Fig.2 indicate the EBSD scanned area.It is evident that the flow of the marker material never contacted the surface of the probe during the entire flow path.Furthermore,critical coordinates at each stage are presented by the blue dotted lines according to the definition of each stage,which will be discussed in detail later.Fig.3a reveals the change in the length of the{10–12}twin boundaries and LABs with the distance x from the keyhole based on EBSD data.In a similar manner,Fig.3b reveals the change in the average grain size,and the(0001)texture strength with the distance x.Here,the structure evolution during welding in the shoulder layer and in the probe layer are represented in each stage in Fig.3a and b,respectively.Furthermore,Fig.4 shows the characteristics of the boundary length distribution as a function of the misorientation angles in each representative coordinate of both the shoulder layer and the probe layer.Combined with the marker morphology observation and EBSD data(Figs.3 and 4),the entire welding process is divided into the following five stages:Stage I:Compression,Stage II:Preheating,Stage III:Acceleration,Stage IV:Deceleration,and Stage V:Annealing.Since the microstructure and texture of the Mg alloy after Stage V has already been extensively investigated by many researchers[4,31-33,37-39],the present study places an emphasis on Stages I to IV to reveal the microstructure and texture evolution along the real material flow path during the FSW.

Fig.3.Distribution of the length of{10–12}twinning(μm);length of LABs(μm);average grain size(μm),and(0001)texture strength determined by EBSD along the marker material of(a)the shoulder layer and(b)the probe layer.

3.1.Stage I:compression

Since both layers have similar variation tendencies in most of the stages,the focus was placed on the shoulder layer to describe the representative characteristics of the microstructure and texture evolution.With the forward movement of the tool,the material far ahead of the tool first experienced compression along the welding direction.Although the marker morphology remained nearly unchanged in Stage I(Fig.2),remarkable increases in the{10–12}twin boundary length and the LABs length were observed(Fig.3).The inverse pole figure(IPF)maps at−13mm and−10mm of the shoulder layer are shown in Fig.5a and b to describe the microstructure evolution at this stage,respectively.From−13mm to−10mm,many{10–12}twins were observed in the IPF maps,and the length of the{10–12}twins were twice as long as the other regions within this 3mm distance(Fig.3).In the hotextruded Mg plates,most of the initial grains have a typical basal texture,with most of their c-axes oriented parallel to the normal direction(ND),which is perpendicular to the welding direction(WD).The{10–12}twins in an HCP material(like pure Mg)can be formed by applying tensile deformation along the c-axis or compressive deformation perpendicular to the c-axis[40-42].Since the tensile deformation along the caxis is not physically possible in this situation,the formation of these{10–12}twins likely originates from the compression stress along the WD.The LABs length increases∼101%from−13mm to−10mm(Fig.3),which indicates that the compression strain gradually accumulates when the material moves close to the tool.

Meanwhile,the number of twins and also the area fraction of the{10–12}twins increases in Stage I.The compressionformed{10–12}twins increase its length and width until they encounter the grain boundary of the initial grain or the boundary of another{10–12}twin(for a schematic illustration of this twin encounter,see Supplementary Fig.S4).Fig.5c and d,which are the magnified IPF maps circled in Fig.5b,illustrate the interactions between the{10–12}twin boundaries from two co-zone twin variants and two non-co-zone twin variants observed at−10mm,respectively.Fig.5c shows that several LABs were parallel to each other,which does not commonly occur due to dislocation tangling or rearrangements.A linear point-to-point misorientation analysis along the line L1(Fig.5e)revealed a similar misorientation angle of 7.4°,which suggests the encountering of two co-zone twin variants.Meanwhile,the misorientation analysis along the line L2(Fig.5f)showed a misorientation of approximately 60°,which indicates the encounter of two non-co-zone twin variants.The prominent boundary length change near a misorientation angle of 60° in Fig.4 showed evidence of an encounter with the{10–12}twins.

3.2.Stage II:preheating

As the material moved closer to the tool,the temperature rapidly increased.At Stage II,the following features were observed from Figs.3 and 4 in both layers:(1)the length of the{10–12}twin boundaries remarkably decreased;(2)the LAB length declined at first,and then once again increased to a high value;and(3)the(0001)texture strength rapidly increased when Stage III was entered.

Fig.4.The misorientation-angle distribution at each critical coordinate of the shoulder layer(a-d)and the probe layer(e-h).(b-d)are partially enlarged figures of(a),and(f-h)are partially enlarged figures of(e).

Fig.6 shows the IPF maps of the shoulder layer during Stage II(−10mm to−7mm).From−10mm to−8mm,there was an apparent decrease in the{10–12}twin fraction(Fig.6a-c),and the length of the LABs declined by about∼53%(Fig.3a),which were caused by the recovery and recrystallization due to the increased temperature.Moreover,the{10–12}twins at−8mm lost their ideal twining morphology,and their misorientation angle simultaneously shifted from 86°to 83°(Fig.6e and g).It can be concluded that before entering Stage III,the material partially experienced forces from other directions rather than the simple compression.An apparent increase in the length of the LABs was also observed before entering the next stage(−8mm to−7mm at the shoulder layer(Fig.3a),and−6mm to−5mm at the probe layer(Fig.3b)).The occurrences of the CDRX(green arrows in Fig.6f)and DDRX(white arrows in Fig.6f)are frequently confirmed at−7mm.Denoted by the green arrows in Fig.6f,the continuous LAB-to-HAG variation infers the sequential occurrence of dislocation rearrangement to form LAB,dislocation wall formation,and the adjacent dislocation absorption to increase the misorientation angle to form HAB,which represents a typical occurrence of CDRX[43].The white arrows and KAM map in Fig.6f show the bulging and migration of HABs to the high dislocation density side and the retention of a dislocation free zone indicated the occurrence of strain-induced boundary migration,which is a typical feature of DDRX[44,45].Meanwhile,“orientation convergence”[35](a significant increase in the(0001)texture strength while most misorientation angles converged within 30°,Figs.3 and 6h)appeared at−7mm(the critical coordinate between Stage II and Stage III)at the shoulder layer(−5mm at the probe layer).A visible change in the grain morphology from−8mm to−7mm allows us to conclude that the new force,shearing from the tool,was introduced before entering Stage III.

Fig.5.IPF maps from(a)−13mm and(b)−10mm in the shoulder layer.The encounter of the(c)two co-zone{10–12}twin variants and(d)two nonco-zone{10–12}twin variants at Stage I are illustrated.(e)and(f)represent the linear point-to-point misorientation distribution along Line L1(in c)and Line L2(in d),respectively.

Fig.7 shows the appearance of secondary{10–12}twins(marked with yellow arrows)in the initial{10–12}twin grains from(a)−7mm to(b)−6mm,which were observed only in the probe layer.Fig.7c(enlarged figure of the encircled region in Fig.7a)describes the orientation relationship between the initial{10–12}twin(cyan)grains and two newly-formed secondary{10–12}twins(magenta and coral).In Fig.7c,the misorientation angle between the two secondary{10–12}twins was near 60°(indicated by the magenta arrow),which originated from the encountering of two non-co-zone twin variants.Two primary reasons support this hypothesis.First,the adjacent grains(yellow arrows in Fig.7a and b)of the secondary twins have a similar orientation of twins formed before this stage,which does not exist in the base metal.Moreover,geometrical analysis suggests that if the material had moved right under the tool,the vertical force introduced by the load of the tool would preferably induce the secondary twin as it has the highest Schmid factor.Therefore,it can be concluded that these twins are secondary twins instead of two different primary twin variant pairs encountering each other.Considering the orientation of the initial{10–12}twins,the newly-introduced compressive force from the ND was implied to be introduced as the vertical load by the tool.Before entering Stage III(−7mm and−6mm at the probe layer),the material at the probe layer already moved under the shoulder of the tool.Fig.8 schematically illustrates the difference in the{10–12}twin formation between the shoulder layer and the probe layer in Stage II.The{10–12}twins(blue triangles in Fig.8a)were first induced far from the tool due to compression along the WD(the c-axis of the twins turned parallel to the WD)(Fig.8b).In the shoulder layer,those{10–12}twins were directly involved in the shearing deformation zone when Stage III was entered.In the probe layer,those{10–12}twins moved under the shoulder of the tool before entering Stage III.Therefore,the vertical compression force from the tool load formed the secondary{10–12}twins(red triangles in Fig.8)from the initial{10–12}twins(Fig.8c).

The character of the material in this stage is similar to the conventional thermo-mechanical affected zone(TMAZ)in FSW.They have relatively high LAB length due to the mechanical deformation,and the relatively high temperature partially induced the recovery and the recrystallization.However,unlike the conventional TMAZ,the{10–12}twins and the relevant thermal and mechanical evolution played critical roles in the evolution of the microstructure and the texture in this stage.

3.3.Stage III:acceleration

At Stage III,the decrease in the marker thickness implied an acceleration of the material flow speed.Fig.9a schematically illustrates the deformation pattern at this stage.The sudden increase in the(0001)texture strength(Fig.3b)suggests that the massive activation of the basal slip dominated deformation during this stage(termed as the“orientation convergence”[35]).The flow speed and strain distribution at Stage III can be approximately estimated by the morphology variation of the marker material according to the continuity equation shown in Fig.9[19,23,24].It is considered that multiplying the thicknessSiat planei(at planek,Sk)by the speedViat planei(at planek,Vk)remains a constant value due to the continuous flow of material during the deformation.Fig.9 clearly reveals the reduction in the marker material thickness as it moved closer to the tool.Therefore,the flow velocity of the material accelerated fromVitoVkas it neared the tool.

Fig.6.IPF maps showing typical microstructure evolution at Stage II((a)−10mm,(b)−9mm,(c)−8mm and(d)−7mm of the shoulder layer).The morphology and change in misorientation angle of{10–12}twins are shown in(e)and(g).The occurrence of CDRX and DDRX are marked with green and white arrows respectively in(f).The change in misorientation angle distribution(h)from−8mm to−7mm shows the occurrence of“orientation convergence”.

The length of the{10–12}twins remained constant during this stage(Fig.3a).As shown in the(0001)pole figure(PF)in Fig.9c,the{10–12}twins likely originated due to the compressive force(magenta arrows in Fig.9b and c)perpendicular to the shearing force[40].Since the{10–12}twins are present during the entire deformation process(Stage III and Stage IV),the compressive force was likely generated by the restraining force of the material outside the stirring zone.

Fig.10 shows the microstructure evolution during Stage III.A similar grain morphology was observed from−6mm to 0mm.At this stage,the microstructure consisted of equiaxed grains and{10–12}twins.The(0001)PFs are shown at the corner of each IPF map.The PFs from−6mm to 0mm gradually show that the maximum concentration point moved from the right side to the lower side.Since the deformation process was dominated by basal slip,the shift in the maximum concentration point implied a change in the shearing direction and shearing plane during this stage.As shown in Fig.10e and f,the CDRX(transformation from LAB to HAB,indicated by green arrows)and DDRX(strain-induced boundary migration(SIBM),indicated by white arrows)were simultaneously confirmed as the recrystallization mechanisms involved during this stage.The noticeable increase in boundary length as the misorientation angle increased from 10° to 20°(see 0mm in Fig.4c and g)implied that CDRX was the primary recrystallization mechanism at this stage.

Fig.7.IPF maps and(0001)pole figures in Stage II of the probe layer at(a)−7mm and(b)−6mm.The secondary{10–12}twins are indicated by the yellow arrows.The encounter of secondary{10–12}twins from two non-co-zone twin variants is shown in(c).

Fig.8.Schematic illustration showing(a)the difference in compression stress and resulting mechanical twins between Stage I and Stage II in the shoulder layer and the probe layer.(b)and(c)show the{10–12}twins induced by horizontal compression and vertical compression,respectively.

Fig.9.Schematic illustration showing the deformation pattern of Stage III.IPF of−5mm is shown by(a).The continuity relation of the marker material in Stage III is illustrated by(b).Compression direction calculated from(c)(0001)PF.

3.4.Stage III:acceleration

When the material crossed 0mm,the flow speed continuously decreased and eventually stopped to form the SZ that has been reported by other researchers.Fig.11b schematically shows the deformation mode during Stage IV.The thickness of the marker gradually increased during this stage(red square area in the OM image above Fig.11b),which implies a decrease in the flow speed,according to the continuity equation mentioned at Stage III.Similar to Stage III,both the shearing force(blue arrow pairs)and compressive force(magenta arrows)were present during this stage.The shearing force was parallel to the tangent of the marker,while the compressive force was perpendicular to the shear direction.The shearing deformation dominated by basal slip activation was still regarded as the primary deformation mechanism during this stage,because the(0001)texture strength remained at a high value(usually>40,Fig.4).Similarly,the compressive force direction calculated from the(0001)pole figure agrees with the direction estimation from the marker morphology(Fig.11c).

Fig.10.Typical IPF maps and(0001)pole figures of Stage III((a)6mm,(b)4mm,(c)2mm and(d)0mm).IPF map in the dotted area(e)at−6mm with KAM map shows the occurrence of DDRX-like bulging and HAB migration(white arrows).The dotted area(f)at 0mm shows the coexistence of CDRX(green arrows)and DDRX(white arrows).

Fig.11.Schematic illustration showing the deformation pattern during Stage IV.IPF of−5mm is shown by(a).The continuity relation of the marker material in Stage III is illustrated by(b).The compression direction was calculated from(c)(0001)PF.

Typical IPF maps with(0001)PFs at Stage IV are shown in Fig.12.The highly-concentrated(0001)texture pattern gradually rotated(see the maximum concentration point in the pole figures of Fig.12)during Stage IV.This c-axis rotation causes most of the grains’c-axes to orient parallel to the WD,which is often reported in the final state of the SZ[31,32,34,35].The rotation of the pole with a maximum intensity in the(0001)PFs from−7mm to 7mm illustrates how the shearing deformation of the tool eventually transfers material from in front of the tool to the backside,to form the SZ.Fig.12e and f is enlarged IPF maps at 3mm and 7mm,respectively.The evidence for the DDRX(white arrows)and CDRX(green arrows)was confirmed.The kernel average misorientation(KAM)map in Fig.12f supports the assertion that recrystallized grains nucleated through the DDRX.From 0mm to 7mm,the boundary length with a misorientation between 2°and 10° dramatically decreased(Fig.4b and f)while the length of the 15°-30° misorientation boundary rapidly increased(Fig.4c and g).This implies the massive occurrence of the CDRX.The reduction in the boundary length with a misorientation angle between 2°-10° was due to the gradual decrease in deformation because of the material flow deceleration.The primary recrystallization mechanism at Stage IV was CDRX,which is in agreement with the slightly increased length of the boundary with a misorientation angle between 15° and 30°

3.5.Stage V:annealing

Due to rapid CO2cooling,the material which flowed out from the tool(>7mm)was maintained in a quasi-static and deformed state(similar to that at 7mm).After flowing out,the material could only experience static recrystallization and grain growth.To confirm that the freezing process had no effect on the microstructure compared with the usual FSW material,artificial post-annealing was conducted on the SZ obtained by rapid CO2cooling.The sample was annealed at 480°C for 20 s to simulate a subsequent annealing period after welding.In order to obtain reliable EBSD results due to the grain growth,the scan area of the usual FSW and annealed SZ samples were enlarged to 360μm×400μm(four times larger than that of the rapid CO2cooling).Fig.13 shows the texture and microstructure relationship of(a)the SZ from the usual FSW,(b)the SZ from rapid CO2cooling,and(c)the rapid CO2cooled SZ followed by artificial annealing.Without rapid cooling,coarsened grains with an average grain size of 33.6μm can be observed in the SZ formed by the normal FSW(Fig.13a).The grain size decreased from 33.6μm to 9.1μm when rapid CO2cooling was used(Fig.13b).The average grain sizes from the Stage I to Stage V are 19.3μm(26.8μm),15.2μm(20.7μm),11.4μm(14.6μm),9.1μm(12.6μm),33.6μm(34.1μm),respectively for the shoulder(probe)layer.Compared with other materials subjected to CO2cooling assistant FSW[22-24,35,36],grain refinement is not very obvious in this study for two reasons.First,the base material originates from hot-extrusion with a relatively high cooling rate and not through conventional hotrolling.Therefore,the initial grain size of the base metal is around∼25μm,which is much smaller than in hot-rolled pure Mg plates.Moreover,pure Mg plates were used in this study.Without solute drag and precipitation,grain coarsening happens much more rapidly than conventional Mg alloys,like AZ31(the pure Mg stir zone without CO2cooling shows even bigger grain sizes than the base metal,see Fig.13a).After artificial annealing,the grain size increased to 36.5μm(similar to the normal FSW),while also retaining the texture strength and an unchanged texture pattern.Some ternary Mg alloys(especially Mg-X-Ca alloys)show randomized crystal orientations after the FSW[37].This occurs because solute atoms affect the recrystallization by forming randomlyoriented recrystallized grains,which drag the movement of the grain boundary due to the solute drag effect[46].Therefore,it is reasonable to assume that HCP-structured Mg alloys should experience similar shear from Stage I to Stage IV.This demonstrates that the method used in this study can potentially reveal the effects of different alloying elements on recrystallization during the FSW.

Fig.12.Typical IPF maps and(0001)pole figures of Stage IV((a)1mm,(b)3mm,(c)5mm and(d)7mm).The dotted area(e)at 3mm shows the coexistence of CDRX(green arrows)and DDRX(white arrows).IPF map in the dotted area(f)at 7mm with KAM shows the existence of DDRX-like nucleation.

Fig.13.Typical IPF maps at SZ from:(a)the usual FSW,(b)the rapid cooled FSW,(c)the rapid cooled FSW after annealing.Similar(0001)textures and microstructures are confirmed.

Fig.14.Reconstruction of the in-situ material flow of pure Mg during FSW.The evolution of the crystal orientation at the(a)shoulder layer and(b)the probe layer is illustrated based on the EBSD data.The reconstruction of the in-situ shear direction and shear plane are illustrated by the red arrows(<11−20>),and the(0001)basal plane of hexagonal prisms in(c)and(d)from the view of front and backside,respectively.

Meanwhile,the occurrence of recrystallization during the subsequent annealing period reduced the total length of the LABs from 1412μm in the SZ of the normal FSW to 4626μm in the rapidly-cooled SZ(calculation area of 180μm×200μm).After annealing at 480°C for 20 s,the rapidly-cooled SZ experienced similar recrystallization and grain growth as the normal FSW.Since the texture and microstructure of the SZ of normal FSW(Fig.13a)and the SZ of the rapid cooling and annealing(Fig.13c)were profoundly similar to each other,it is reasonable to conclude that(1)rapid CO2cooling+stop action successfully froze the material as it deformed without affecting the nature of its flow;and(2)the widely observed final state of the microstructure and texture in the SZ originated from those at 7mm and then experienced recrystallization and growth during subsequent annealing.

3.6.Texture evolution of pure magnesium during FSW

Based on the results of EBSD analysis from each stage,the texture evolution of pure magnesium during FSW is systematically revealed.The character of the texture evolution of each stage is summarized.At Stage I,the WD-parallel compression introduced a large number of{10–12}twins.The texture was modified from a relatively simple orientation to a mixing of the initial orientation and the{10–12}twins(Fig.5b).Due to the mixing of two types of orientations,the(0001)texture strength decreased rapidly from∼15 to∼5 which infers the occurrence of{10–12}twins induced texture randomization in this stage.In Stage II,the increase in temperature and the approach of the tool affected the texture evolution.The increase in temperature caused a certain degree of grain growth.The initial grain with the preferential orientation grew rapidly,which significantly reduced the amount of the{10–12}twins(Fig.3).Before entering the next stage,the material partially experienced shear deformation from the tool.These two reasons contribute to the dramatically increased(0001)texture during this stage,and the partially introduced shear deformation formed the“one-point”like(0001)texture pattern.Due to the simple shear deformation in Stage III,the texture was kept in the“one-point”like(0001)texture pattern(Fig.10a-d)and a high(0001)texture strength(Fig.3).A small amount of{10–12}twins were introduced because of the compressive force originating from the restraining of the material flow,but did not significantly affect the pattern and the strength of the texture.The origin of the specific character of texture,denoted“orientation convergence”[35],in the FSW of Mg alloy was revealed to initiate at the start of Stage III and strengthened over Stages III and IV.During Stage III and Stage IV,the(0001)texture strength remained at a relatively high level(usually over 40,Fig.3)while the pattern of the(0001)texture remained the same only with changes in concentration directions(Figs.10 and 12).After moving outward from the tool,the material underwent static recrystallization and grain growth during Stage V(Fig.13a).The texture evolution during Stage V was revealed by the quasi-in-situ method,and exhibits an unchanged texture pattern and similar texture strength(Fig.13b and c).Finally,the highly concentrated and one-point like(0001)texture was formed in the SZ,which has been previously reported[4,31,33-35,37,38].

3.7.Quasi-in-situ material flow rebuild

The variation in the morphology of the marker material,combined with the EBSD analysis allows the details(such as local shearing direction and local shearing plane)to be reproduced in order to establish an in-situ understanding of the deformation of pure Mg during the FSW.Fig.14a and b illustrates the representative orientation of most grains.It can be observed that their(0001)planes and<11–20>directions are parallel to the tangent plane and direction of the marker material.By rebuilding the(10–10)pole figure to make a comparison with the EBSD data(Supplementary Fig.5),the shearing directions of the marker material morphology and the EBSD data match each other perfectly.The rebuilt material flow(local shear planes and shear directions)is shown in Fig.14c and d,which represent the backside and front view through the tool,respectively.It can be confirmed that the material flows via the path(which is rebuilt by the marker material in Fig.2)with continuous changes in the shearing plane(the(0001)planes of the hexagonal prisms in Fig.14c and d)and shearing direction(<11−20>directions of those hexagonal prisms,red arrows in Fig.14c and d).Hence,in this study,the integrated material flow and microstructure evolution were successfully quasi-in-situ rebuilt by the combined use of a marker material,CO2cooling,and the quick-stopaction technique.

4.Conclusions

Quasi-in-situ microstructure and texture evolution along the real material flow path of pure Mg during FSW were investigated.When FSW was performed with rapid CO2cooling,the material was successfully frozen as it experienced deformation.A marker material inserted between the workpieces revealed the real flow path for both the shoulder layer and the probe layer.After simultaneously examining the microstructure and texture by EBSD,it was revealed that the stir welding process of pure Mg could be divided into five stages.

Stage I(Compression):In this stage,the{10–12}twins formed due to the compressive force from the forward movement of the tool.LABs formed during this stage,but almost no recrystallization occurred due to the relatively low temperature.

Stage II(Preheating):The temperature rapidly increased near the rotating tool.The deformation of the shoulder layer underwent simple compression along the WD while the material underwent a slight shear deformation before entering Stage III.Besides the compression along the WD,the probe layer underwent vertical compression along the ND during this stage because the material in the probe layer went under the tool before it entered Stage III.Due to the increase in temperature,recovery and recrystallization occurred at this stage,resulting in a decrease in the LAB length.

Stage III(Acceleration):The material accelerated to a relatively high flow speed due to the shearing deformation of the tool.The shearing force along the flow direction and the compressive force perpendicular to the shearing force were present at this stage.Orientation convergence originated from this stage,which led to an extremely high(0001)texture strength.Both the CDRX and DDRX were observed,and the CDRX was suggested to be the main recrystallization mechanism at this stage.

Stage IV(Deceleration):The high-speed flowing material gradually decelerated and eventually stopped to form the final SZ.The shearing plane and shearing direction continually changed during this stage,and the length of the LABs gradually decreased due to the lesser deformation as the material flow decelerated.

Stage V(Annealing):After the material flow completely stopped,the material experienced static recrystallization and grain growth due to subsequent annealing.During this stage,the length of the LABs decreased to a relatively low value due to recrystallization.Since pure Mg was used in this study,the annealing process retained the texture strength,and the texture pattern was unchanged.

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.

Acknowledgements

This study was partly supported by the New Energy and Industrial Technology Development Organization(NEDO)under the“Innovation Structural Materials Project(Future Pioneering Projects)”.

Supplementary materials

Supplementary material associated with this article can be found,in the online version,at doi:10.1016/j.jma.2020.05.015.