Effect of the processing route on the microstructure and mechanical behavior of superlight Mg-9Li-1Zn alloy via friction stir processing

Mengrn Zhou,Zhuorn Zeng,Chun Cheng,Yoshiki Morisd,Qingyu Shi,Jin-Yih Wng,Hidetoshi Fujii,∗

a Joining and Welding Research Institute (JWRI),Osaka University,Osaka,Japan

b Department of Mechanical Engineering,Tsinghua University,Beijing 100084,China

c State Key Laboratory of Tribology,Tsinghua University,Beijing,China

d Key Laboratory for Advanced Materials Processing Technology Ministry of Education of China,Tsinghua University,Beijing 100084,China

e College of Engineering and Computer Science,Australian National University,ACT 2601,Australia

f Department of Materials Science and Engineering,National Dong-Hwa University,Hwa-Lian,Taiwan,China

Abstract In this study,the effect of the processing route using a friction stir processing (FSP) method on the microstructure and mechanical behavior of a Mg-9Li-1Zn alloy was systematically investigated.In the FSP method,the odd-numbered (1st and 3rd) process directions and even-numbered (2nd and 4th) passes were alternated to distribute the strain throughout the whole processed zone uniformly.Consequently,the processed zone had a much more uniform microstructure and hardness distribution than the processed zone obtained using the conventional FSP method.Using this method,the grain size of a Mg-9Li-1Zn sheet alloy was refine from ∼31 μm to ∼0.21 μm with uniformly distributed α and β phases.The processed alloy exhibited a high strength-ductility synergy with an ultimate tensile strength (UTS) of 220.1 MPa and total elongation of 70.0% at a strain rate of 10−3 s − 1,overwhelmingly higher than those of the base metal,155.6 MPa in UTS and 36.0%in elongation.The in-situ SEM-DIC analysis and TEM observation demonstrated that such an outstanding ductility with moderate strength is caused by grain boundary sliding,the dominant deformation mechanism of the ultra-fine-graine sample after FSP.The processing route with reverse processing direction was proven to be efficien in producing the ultrafin grain size microstructure and improving the mechanical properties of superlight Mg-9Li-1Zn alloy.

Keywords: Mg-Li-Zn alloy;Friction stir processing;Microstructure; In-situ SEM-DIC;Mechanical properties.

1.Introduction

The use of lightweight materials in the transportation fiel is one of the most practical methods to reduce the energy consumption of vehicles [1–3].It is reported that at least a 10% fuel efficien y increase in vehicles can be achieved by replacing aluminum (Al) alloys with much lighter Mg alloys [3].Due to their much lower density and excellent ductility,magnesium-lithium (Mg-Li)-based alloys have recently attracted tremendous attention [4–7];however,Li addition decreases the strength of Mg alloys because it introduces a much softer body-centered cubic (BCC)βphase when its content exceeds 5 wt%.To expand the applications of Mg-Li alloys,a way to increase their strength must be developed.

In recent years,compositional alteration methods have been introduced,which have focused on improving the strength of Mg-Li alloys [6–22].Compositional methods involve adding elements such as Al,calcium (Ca),copper (Cu),and rare earth (RE) elements [6–13];however,the addition of these elements inevitably increases the density of Mg-Li alloys,and introduced some metallurgical problems.Tang.Sreported the remarkable hardness over 150 HV of Al-added Mg-Li alloy but extremely poor ductility [7].Wang.Tadded Ca into the Mg-Li-Al system and observed the simultaneous deterioration of strength and ductility due to the excessive(Mg,Al)2Ca eutectic phase forming at grain boundary [9].N.Saito observed the natural aging response of Mg-Li-Cu alloys that its mechanical properties were varied dramatically with time [11].Consequently,the compositional method not only increased the density of Mg-Li alloy that deteriorated its specifi strength,but also introduced some inevitable metallurgical issues that fetter its further application.Therefore,microstructural methods are preferred for improving the strength of Mg-Li alloys to maintain their low density while avoiding those metallurgical problems that preferably occurred during alloying.

Previous studies have reported improved the strength of Mg-Li alloys using severe plastic deformation (SPD) methods,such as accumulative roll-bonding (ARB) [14,15],equal channel angular extrusion (ECAE) [16–18],high-speed-ratio differential speed rolling (HRDSR) [19],and high-pressure torsion (HPT) [20–22].The ARB method was reported to refin the grain size of a hot-rolled Mg-5Li-1Al alloy from 30 μm to less than 100 nm while also increasing the ultimate tensile strength (UTS) from 214 MPa to 316 MPa but suffered over 60% loss in ductility [14].ECAE was used to obtain a grain size of around 1 μm and a strength of over 210 MPa [18].Kim et al.reported the refinemen of Mg-9Li-1Zn grains from 16 μm to less than 1 μm using HRDSR and achieved an improved UTS from 160 MPa to 217 MPa [19].Through HPT,the grain size of a Mg-8Li alloy was refine from ∼100 μm to less than 500 nm while also achieving an increased hardness from 48 HV to 63 HV [21].These previous studies demonstrated the ability of microstructural methods for improving the mechanical properties of Mg-Li alloys;however,these methods face the same critical problem of the size limitation and the potential reduction of ductility of the workpiece being processed,which greatly restricts the application of mechanical methods for improving the strength of Mg-Li alloys.

Friction stir processing (FSP) is a solid-state material processing method originated from friction stir welding (FSW)that has been used for the grain size refinemen of metals over a wide range of melting points [22–28].By applying a large load and low tool rotation speed,the processing temperature was dramatically decreased,and ultra-fin sub-micron size grains in the stir zone were obtained [29,30].At lowtemperature,FSP can improve the strength of metals with almost no size limitations [31].Zhou et al.reported the use of low-temperature FSP to obtain an ultra-fin grain Mg-Li alloy with not only enhanced strength,but also superior superplasticity [23].Multi-pass FSP has been performed on an Al alloy,and is capable of substantially refinin the grain size and improving its mechanical properties [32–35].However,the unique nature of FSP leads to a severe asymmetry in microstructure and mechanical property if the FSP is unidirectional [36].During the FSP,the material at the advancing side (AS,the material fl w direction went opposite with the tool rotation direction) and the retreating side (RS,the material fl w direction went along with the tool rotation direction)in the stir zone suffered completely different magnitudes in strain and strain rate [37–39],resulting in significan difference on the microstructure after FSP [36].

The aim of the present study is to develop a method that can alleviate the microstructure asymmetry and adopt this method to fabricate Mg-Li alloy with outstanding strength and ductility simultaneously.In this study,a comprehensive investigation of the FSP of Mg-Li alloys is performed to have an in-depth understanding of how the FSP route affects the microstructure and mechanical behavior of a hot-rolled Mg-9Li-1Zn alloy,allowing the deformation modes that result in the outstanding properties to be discussed.

2.Experimental procedures

2.1.Processing parameters

Hot-rolled Mg-9wt%Li-1wt%Zn alloy LZ91 plates were used in this study.Fig.1 shows (a) the typical microstructure and (b) X-ray diffraction (XRD) spectrum of the as-received material.The base metal (BM) has a typical bimodal lamellar structure comprised of 33%αphase (white),66%βphase(dark gray),and a small amount of Mg2Zn11intermetallic compound.Theαphase is the Mg-rich phase with a hexagonal close-packed (HCP) structure,while theβphase is a Li-rich phase with a body-centered cubic (BCC) structure.The plates were prepared with a length of 200 mm,a width of 70 mm,and a thickness of 3 mm using an electrical discharge machine (EDM,Sodick AG360L).Before the FSP,the plate surfaces were carefully polished with#4000 grit abrasive paper and subsequently cleaned with ethanol.The FSP was conducted with a cemented tungsten carbide (WC) tool [23].The tool had a shoulder diameter of 15 mm,a probe diameter of 6 mm,and a probe length of 2.8 mm.Four total FSP passes were applied to the material via two processing routes:(I)four unidirectional processing passes were overlapped with the same processing direction;(II)the odd-numbered processing(1st and 3rd)passes were crossed with the even-numbered(2nd and 4th) passes.Fig.2 schematically illustrates the processing details of(a)routeIand(b)routeII.The tool rotation speed was 30 rpm,and the moving speed was 10 mm/min for all the processes to maintain an extremely low processing temperature.

2.2.Mechanical property evaluation

After processing,the hardness profil of the stir zone (SZ)was measured by a Vickers hardness test machine (Future-Tech,FM-800) using a load of 50 gf with a dwelling time of 15 s.Tensile tests were carried out using a universal tensile test machine (SHIMAZU,AGX-10kNX) at the center of the SZ.The gage had a length of 4 mm,a width of 2 mm,and a thickness of 2 mm.Details of the small tensile specimens and the geometrical relation with the SZ are described inSup-plementary Fig.S1.The tensile tests were carried out with crosshead speeds of 24 mm/min,2.4 mm/min,0.24 mm/min,0.024 mm/min,and 0.0024 mm/min to obtain tensile strain rates of 10−1/s,10−2/s,10−3/s,10−4/s,and 10−5/s,respectively.

Fig.1.(a) SEM image and (b) XRD spectrum of the BM.The Mg-rich α phase and the Li-rich β phase are indicated by the white and black arrows,respectively.

Fig.2.Schematic illustration of the processing details of Route I and Route II.The tool rotation direction at different processing stages is illustrated by the magenta arrows.The AS and RS aside the 1 Pass represent the advancing side and the retreating side,respectively.

2.3.Microstructure characterization

An optical microscope (Olympus,BM-51X) was used to macroscopically characterize the SZ after the FSP.Before observation,the samples were carefully polished by a 1 μm alumina suspension (Baikalox,1.0 CR) and 0.04 μm colloidal silica suspension (Struers,OP-S) and subsequently cleaned in high purity ethanol.The chemical etching was applied to the samples after polishing with a solution containing 10 g picric acid,175 ml ethanol,25 ml acetic acid,and 25 ml distilled water.Scanning electron microscopy (SEM,JEOL-7001F),scanning transmission electron microscopy (TEM,JEOLJEM-2010) equipped with energy dispersive X-ray spectrometry (EDX),and X-ray diffraction (XRD,Bruker D8 DISCOVER) were used for the microstructure characterization.XRD analysis was carried out with an X-ray beam width of 300 μm at the center of the SZ.SEM samples followed the same preparation sequence as the optical microscopy.Focused ion-beam (FIB,JEOL-JIB-4500) was used to prepare the TEM samples.The ex-situ TEM observation was conducted on the 2 Cross-pass sample before and after 80% tensile test.Image J was used for the calculation of grain size in this study.

2.4.In-situ digital image correlation (DIC) analysis

In-situtensile tests in the SEM combined with digital image correlation (DIC) analysis was used to evaluate the grainscale strain distribution evolution in the BM and processed zones obtained via different processing routes.Thein-situtensile tests were performed by a tensile test stage (TSL-TS-1500-II) equipped in the SEM (JEOL-7001F) with a control system (TSL-Tensile Test V6.57).Thein-situtensile specimen from the processed zone has a gage size of 10 mm in length,2 mm in width,and 1 mm in thickness.For strain tracing,a nanoindentation machine was used to deploy an indentation matrix with a spacing of 80 μm on the surface of thein-situtensile test specimen.The optical image of the tensile test stage in the SEM,the geometric details of thein-situtensile test specimen,and the indentation matrix are illustrated inSupplementary Fig.S2.To establish a precise grain-scale correlation during thein-situtensile test,a certain number of oxidization particles were manually deployed on the surface by setting the material in the air for 12 h before placing it into the SEM.Depending on the grain size at different processing parameters,magnification of 500x,1000x,and 3000x were used for thein-situtensile tests in the SEM.The images were obtained at a constant working distance of 15 mm using an acceleration voltage of 15 kV and a beam current of 12 nA.Local strains of up to 40% and 80% were used in thein-situtensile tests of the BM and the processed zone,respectively.The tensile test was performed at a strain rate of 10−5/s,and the stage was automatically stopped for every 1% of global strain for the SEM image acquisition.Every SEM image contained 2500 × 2000 pixels.

Vic-2D Digital Image Correlation Ver.6.0.6 software was used for DIC analyses and strain calculations.For the DIC analysis of the BM and the processed zone,the subunit size was set as 30×30 and 15×15 pixels,respectively.The subunit matrix produced in-plane displacement coordinate maps on the WD-TD plane using the Gaussian weight method.The in-plane deformation and strain were calculated from the coordinate maps using the Lagrangian finit strain method.The global certainty during the DIC analysis was measured,the global certainty was above 98%,and its evolution during thein-situtensile test for every experiment was recorded and the mean global certainty of allin-situtensile is over 98.5% (see Supplementary Fig.S3–5 for the evolution of global certainty during thein-situtensile tests of the BM,the 1 Crosspass and 2 Cross-pass,respectively).

3.Results

3.1.Microstructure symmetry

Fig.3 shows the OM and SEM images,and the local grain size (both theαandβphases) of the areas subjected to (a-d)1 pass,(e-h) 2 unidirectional passes,and (i-l) 4 unidirectional passes via processRoute Iand (m-p) 1 cross pass,and (q-t) 2 cross passes viaRoute II.Fig.4 summarized the evolution of local mean grain size distribution in the processed zone viaRoute IandRoute II.Fig.3b-d shows the typical SEM images of the microstructure from 1 pass on the AS (b area in Fig.3a),the center of the SZ (c area in Fig.3a),and the RS (c area in Fig.3a) in the SZ,respectively.Microstructural asymmetry can be observed.On the AS,both theαandβphases show sub-micron sizes and equiaxed grains,and the smallest grain size with 1.28 μm was confirmed In the center of the SZ,the grain size was coarsened to 1.92 μm,while some of theαphase grains agglomerated to form a band-like structure.Less grain refinemen (with the mean grain size of 5.68 μm) was observed on the RS compared with the BM,and only a slight change in the morphology of both theαandβphases.Besides at the AS,the grain size of theαphase is much smaller than that of theβphases.At the processed zone center of 1 pass,the grain size of theαphase was significantl refine to 0.91 μm but that of theβwas 2.44 μm.However,as the number of processing passes increased viaRoute I,the asymmetrical distribution of the microstructure,and the great difference in grain refinemen ofαandβphases were not optimized (Fig.3a-i).After overlapping another 1 pass on the initial 1 pass,the asymmetrical distribution of the microstructure was still observable (Fig.3e).As shown in Fig.3f-h,the overlap of the second pass further refine the grains in the SZ,the mean grain size on the AS,the center of the SZ,the RS were refine to 1.06 μm,1.42 μm,3.73 μm,respectively.Whilst,the asymmetry in microstructure was still obvious in the SZ,while the band-like structure is still observable at the center of the SZ and the RS (Fig.3g,h).Even after overlapping 4 total unidirectional passes viaRoute I,the asymmetry still existed,while the significan difference in grain size between theαandβphases still exists.The grain size at the AS,the center of the SZ and the RS was refine to 0.56 μm,0.74 μm and 1.75 μm,respectively (Fig.3j-l).The center of the SZ(Fig.3k)firstl exhibited equiaxed grain inRoute Ibut the band-like structure still exists at the RS (Fig.3l).However,the difference in the grain size and morphology between the AS (0.56 μm,Fig.3j) and RS (1.75 μm,Fig.3l) was not negligible.The conventional overlapping method proved the capability of refinin the grains in the SZ,but the difference in total strain between the AS and the RS caused a significan asymmetry in the microstructure.This asymmetric distribution of the microstructure would inevitably cause homogeneity in the mechanical properties,which will deteriorate the tensile behavior.

In contrast to the unidirectional FSP (Route I),a highly symmetric microstructure was obtained throughRoute IIand the grain size and morphology were similar throughout the entire SZ usingRoute II.After 1 Cross-pass,the RS (Fig.3p)showed refine grain size of 1.21 μm and equiaxed grains(the RS of 2 Unidirectional-passes viaRoute Iwas nonequiaxed with the grain size of 3.73 μm,Fig.3h).The entire SZ shows homogeneous grain size around 1.20 μm after 1 Cross-pass.After 2 Cross-passes (four total passes,the same as 4 Unidirectional-passes),the entire SZ showed uniformly refine grains,and the difference in grain size between theαandβphases was vanished.The mean grain size at the AS,the center of the SZ and the RS was remarkably refine to 0.58 μm,0.21 μm and 0.62 μm,respectively(Fig.3r-t),while the SZ of 2 Cross-passes owns an ultra-fine (mean grain size less than 0.3 μm) area with the width over 8 mm(Fig.4b).

It was demonstrated thatRoute IIproduced a better grain refinemen effect thanRoute Iafter four total passes.At the same time,the problem of microstructural asymmetry inRoute Iwas overcome usingRoute II.In order to obtain a refine and homogeneously distributed microstructure in LZ91,FSP viaRoute IIwas the ideal method.

3.2.Mechanical properties

3.2.1.Evolution of hardness

Fig.3.Optical photos and SEM images of Route I: (a-d) 1 Pass,(e-h) 2 Unidirectional-pass,and (i-l) 4 Unidirectional-pass and of Route II: (a-d) 1 Pass,(m-p) 1 Cross-Pass,and (q-t) 2-Cross Pass.The normal direction,the processing direction,and the transverse direction are abbreviated as ND,PD,and TD,respectively.The observation regions of each SEM image are marked individually in the optical images.The mean grain size,the grain size of the α and β phases were marked at the top of the SEM images.

Fig.4.The mean grain size distribution at the stir zone via (a) Route I and (b) Route II.

Since severe inhomogeneity was observed in the microstructure via Route I and was optimized via Route II(Fig.3),the reason beneath the process history needs to be revealed.Fig.5 shows the optical microscopy images of the mid-thickness horizontal layer (1.5 mm beneath the surface)at the keyhole of (a) 1 Pass,(b) 2 Unidirectional-passes,and(c) 1 Cross-pass samples,along with the hardness distribution before and after the process.The hardness test results and keyhole optical images were used to demonstrate the formation of the asymmetry in microstructure and property viaRoute Iand how it was overcome viaRoute II.Using a f ow path trace method [38,39],three lines were drawn to trace the fl w path to illustrate the grain refinemen effect on the AS(L1),the center of the SZ (L2),and the RS (L3).The color scheme of white to red was used to represent the degrees of grain refinement where the white indicates coarsened,and the red represents refine grains.In the 1 Pass sample(Fig.5a),the material exhibited a similar hardness before entering the SZ.The grain size of the AS (along with L1)showed the highest degree of refinemen because the material along fl w path L1 experienced the most significan total strain[37].Therefore,the maximum hardness was observed in the region in which the material originated from L1.From L1 to L3,the total strain gradually decreased,and the grain size gradually increased (Fig.3b-d).The material fl w through L3 experienced only a minor strain and showed the lowest hardness within the SZ.When another pass with the same processing direction was overlapped (2 Unidirectional-passes,Fig.5b),the total strain difference further accumulated.The material initially present on the AS (originating from L1 in Fig.5a) once again experienced the highest total strain (L1 in Fig.5b),but the material on the RS experienced another period of minor strain (L3 in Fig.5b).

Fig.5.Optical microscopy images of the mid-thickness layer and hardness variation at the keyhole of(a)1 Pass,(b)2 Unidirectional-pass,and(c)1 Cross-pass samples.The tool rotation direction is indicated by the blue arrows in the keyhole.The material fl w traces on the AS,the center of the stir zone,and the RS are represented by lines L1,L2,and L3,respectively.The color scheme of L1,L2,and L3 is white (coarsened grains) to red (refine grains),where a deeper red indicates a smaller grain size.

By changing the FSP direction (thus reversing the AS and the RS),the cross pass method placed the“less deformed”RS side on the most severely deformed side (the AS) during the next pass (Fig.5c).The “mostly deformed” AS side was also placed on the least deformed side (the RS) during the next pass.Consequently,the SZ of the 1 Cross-Pass sample had a homogenous and symmetric hardness and microstructure distribution (Fig.3n-p).Consequently,the SZ of the 2 Crosspass sample had the most uniform grain refinement

Fig.6 shows the Vickers hardness distribution of the SZ obtained via (a)Route Iand (b)Route II.The grain refine ment by the 1 Pass immediately increased the hardness from 52 HV to 62.4 HV (Fig.6a),and further increased to 67.5 HV (2 Unidirectional-pass) and 69.6 HV (4 Unidirectionalpass).However,all the regions with a peak hardness in the SZ obtained viaRoute Iwere located at the edge of the AS(Fig.3b,Fig.3f,and Fig.3j).The difference in hardness between the AS and RS was more evident as the number of processing passes increased.The difference in hardness between the AS and the RS in the 1 Pass sample was 10.4 HV(blue triangles in Fig.6a).After the 4 Unidirectional-pass,the difference increased to 19.6 HV (red circles in Fig.6a).ThroughRoute II,a symmetric hardness distribution was obtained (Fig.6b) due to the homogeneous microstructure distribution.The hardness difference across FSP zone was 8.1 HV after 1 Cross-pass,and 8.8 HV after 2 cross pass,which was much smaller than that of the 4 Unidirectional-passes throughRoute I.

Fig.6.The hardness distribution in the stir zone in the samples obtained via (a) Route I and (b) Route II.The range of the SZ is shown by two magenta dashed-dotted lines.

3.2.2.Tensile properties

The tensile test results of the SZ obtained viaRoute IandRoute IIare shown in Table 1-3 and Fig.7.The ultimate tensile strength (UTS,Fig.7a,b) and plastic elongation (EI,Fig.7c,d) were selected to schematically illustrate the tensile properties.The UTS of the BM increased with the tensile strain rate.The UTS of the BM was 141.6±3.3 MPa at a tensile strain rate of 10−5/s,and increased to 175.3 ± 2.9 MPa at a tensile strain rate of 10−1/s.The EI of the BM decreased with an increase in the tensile strain rate.The EI of the BM was 55.1 ± 9.2% at a tensile strain rate of 10−5/s,which decreased to 28.2 ± 8.4% at a tensile strain rate of 10−1/s.The BM showed a minor sensitivity withmvalue of 0.04.

Table 1 The results of yield strength (YS) of Mg-Li alloy LZ91 (MPa).

Table 2 The results of ultimate tensile strength (UTS) of Mg-Li alloy LZ91 (MPa).

Table 3 The results of plastic elongation (EI) of Mg-Li alloy LZ91 (%).

After being processed viaRoute I,an apparent change in the UTS and EI is confirmed Similar to the BM,the SZ of the 1 Pass sample showed an increase in the UTS with the tensile strain rate.The UTS of the 1 Pass sample was 124.1 ± 2.6 MPa at a tensile strain rate of 10−5/s,which increased to 228.1 ± 4.6 MPa at a tensile strain rate of 10−1/s.The EI of the 1 Pass sample was 142.1 ± 14.6% at a tensile strain rate of 10−5/s,which decreased to 32.5 ± 3.3%at a tensile strain rate of 10−1/s.Them-value of the 1 Pass sample is 0.09.After 4 Pass unidirectional FSP,The UTS(EI)of sample at the tensile strain rates of 10−5/s and 10−1/s were,90.7±2.1 MPa(194.3±22.9%)and 246.1±4.8 MPa(19.8 ± 4.8%),respectively,with an even higherm-value of 0.14.

After being processed viaRoute II,Due to the homogeneous microstructure distribution and smaller grain size,the SZ of the 2 Cross-pass sample obtained via Route II exhibited the most evident tensile strain rate sensitivity.The UTS of the 2 Cross-pass sample was 69.8 ± 3.6 MPa at a tensile strain rate of 10−5/s,which remarkably increased to 286.7 ± 2.9 MPa at a tensile strain rate of 10−1/s.The EI of the 2 Cross-pass sample was 11.9 ± 1.3% at a tensile strain rate of 10−1/s,which significantl increased to 274.7 ± 26.6% at a tensile strain rate of 10−5/s.Them-value of the sample after 2 Cross-Pass was 0.21.At the strain rate of 10–3/s,the UTS and elongation of 2 Cross-Pass sample was 220.1 ± 6.3 MPa and 70.0 ± 9.7%,respectively,overwhelming higher than those of 4 Unidirectional-pass sample (155.6 ± 3.4 MPa and 36.0 ± 5.5%) and of base metal(206.0 ± 5.8 MPa and 38.2 ± 4.5%).

Fig.7.Tensile test results of the stir zone obtained via (a,c) Route I and (b,d) Route II.Changes in the (a,b) UTS and (c,d) EI along with different initial tensile strain rates were selected to represent the tensile properties.

3.3.Deformation behavior

In order to understand the origin of the outstanding mechanical properties after 2 Cross-pass FSP,in-situSEM-DIC was used to analyze the deformation behavior of samples after 2 Cross-pass and in comparison with those of BM sample and sample 2 unidirectional Pass FSP.For the BM sample(Fig.8),typical undeformed state (Fig.8a) and global strains of 10% (Fig.8b),20% (Fig.8c),30% (Fig.8d),and 40%(Fig.8e) were selected for DIC analysis.Fig.8f-j show the evolution of the linear local tensile strain distribution along with L1 in Fig.8a-e,respectively.Fig.8k-o illustrate thein-situevolution of crack formation and propagation in area A in Fig.8a-e,respectively.It is confirme that the local strain distribution was non-uniform,even at the initial stage of deformation (10%,Fig.8b).The results of the linear tensile strain exhibited that certain areas along L1 reached a local tensile strain close to 0.2 at a global strain of only 10%(Fig.8g).The morphology analysis confirme the initiation of the crack between the grain boundary of theαandβphases(red arrow in Fig.8i).Upon increasing the global strain,the local strain was initially concentrated in areas (like the orange area A in Fig.8b),which produced a much higher local strain compared to the adjacent material.The maximum local strain in area A reached 0.38,0.43,and 0.55 at global strains of 20%,30%,and 40%,respectively.Meanwhile,certain areas (like at 150 μm and 300 μm in L1) stopped at a local strain less than 0.1,regardless of the increase in global strain.At a global strain of 40%,the maximum local strain was more than 600% higher than that of the minimum local train(Fig.8j).The crack not only propagated along the grain boundary of the α and β phases,but it also invaded the α phase and increased in length and quantity upon increasing the global strain (blue arrows in Fig.8m-o).

Fig.8.The in-situ SEM-DIC analysis during the tensile test of the BM.(a-e) The local strain distribution maps at global tensile strains of 0%,10%,20%,30%,and 40%.(f-j) the linear local tensile strain analyzed along line L1 in (a-e).(k-o) High-resolution in-situ crack propagation process in area A in (a-e).The color scheme of the local tensile strain in (a-e) is shown on the right side with the local tensile strain scale from 0 to 0.8.The red and blue arrows in(k-o) indicate the crack at the α/βboundary and inside the α grain,respectively.

Fig.9.The in-situ SEM-DIC analysis during the tensile test of the 1 Cross-pass sample.(a-e) The local strain distribution maps at global tensile strains of 0%,20%,40%,60%,and 80%.(f-j) The linear local tensile strain was analyzed along the line L2 in (a-e).(k-o) High-resolution in-situ crack propagation process in area A in (a-e).The color scheme of the local tensile strain in (a-e) is shown on the right side with the local tensile strain scale from 0 to 1.0.The red arrows in (k-o) indicate the crack at the α/β boundary.

Fig.9 shows thein-situSEM-DIC deformation behavior analysis during the tensile test of the 1 Cross-pass sample.Fig.9f-j show the evolution of the linear local tensile strain distribution along L2 in Fig.9a-e,respectively.Fig.9k-o illustrates thein-situevolution of crack formation and propagation in area B in Fig.9a-e,respectively.In contrast to severe local strain in BM,the strain localization in 1 Crosspass sample showed was alleviated.At 0 μm to 120 μm on L1,the local tensile strain showed onlya± 15.1% value of the global strain during the whole tensile test.However,Fig.9a-e shows a site of strain localization (area B).At a global strain of 20%,the maximum true strain was 0.23(orange area Fig.9g).Upon increasing the global strain,the maximum local strain reached 1.36 at a global strain of 80% (Fig.14j).Similar crack initiation and propagation phenomena were observed at the edge of the relatively largeαphase grains (Fig.9k-o).The crack propagated in length and quantity upon increasing the global strain,but this occurred after a much significan strain than that at the BM(Fig.9k-o)

Fig.10 shows thein-situSEM-DIC deformation behavior analysis during the tensile test of the 2 Cross-pass sample.A remarkable difference in the deformation behavior can be observed in thein-situtensile process of the 2 Crosspass sample.A significantl uniform strain distribution was achieved during the tensile test over a wide range of SEMDIC analysis areas (Fig.10a-e).The color scheme of local tensile strain implied that the strain instability was tremendously weak,and the linear local strain analysis demonstrated that the local tensile strain only showed a negligible difference of less than ±5.6% compared with the global strain.

To support the DIC analysis,a higher resolutionquasi-insituobservation of the inner part (0.5 mm beneath the surface) of the 2 Cross-pass sample was carried out at different deformation states.At the initial stage of deformation,although both theαand theβgrains were slighted horizontally elongated,the micromorphology was significantl changed.The initial microstructure showed completely and uniformly mixed ultra-finαandβgrains (Fig.11a),in which theαgrains gradually gathered during deformation(red dotted areas in Fig.11b) and eventually formed hugeαgrain clusters (yellow dotted areas Fig.11c).No evidence of cracks in the inner part of the specimen was observed at any stage during tensile deformation.After tilting the 80% deformed sample 45°,the surface morphology became rough,and many deep dimples were observed (red arrows in Fig.11d).

The microstructure of the 2 Cross-pass sample before and after the tensile tests is further characteristerized by TEM(Fig.12).Fig.12a and Fig.12b confir the ultra-fin grain in the 2 Cross-pass sample Fig.12c typically shows theα/βboundary without any deformation,in which the vague boundary with a less-ordered diffraction pattern indicated an incoherent boundary state.After 80% tensile strain at a strain rate of 10−5/s,the grains were only slightly elongated after 80%global strain tensile deformation (Fig.12d,e).Fig.12f shows the high-resolution TEM (HRTEM) image of the typicalα/βboundary after tensile deformation.A sharp boundary with an ordered diffraction pattern indicated a coherent boundary state compared with the undeformed state.

3.4.Fracture behavior

As the significan difference between the tensile behavior of samples via different processing routes,the observation toward the fracture surface was conducted to explain the effects of the processing route on the fracture behavior and its relation to the tensile behavior.Fig.13 shows the SEM images of the fracture surface of the 2 Cross-pass sample at tensile strain rates of (a) 10−1/s,(b) 10−2/s,(c) 10−3/s,(d) 10−4/s,and (e) 10−5/s.An evident change in the fracture behavior was observed among the different tensile strain rates.The fracture surface at a tensile strain rate of 10−1/s shows a completely brittle transgranular fracture (EI of only 11%,Fig.7c).The strain rate was much higher than the diffusion of the boundary,which restrained the induction of the GBS and returned the primary strengthening mechanism to grain boundary strengthening.As the tensile strain rate decreased to 10−2/s,in addition to transgranular fracture,some small dimples began to appear (green arrows in Fig.13b).The fracture behavior became a mixture of brittle and ductile.Decreasing the strain rate allowed some grains to deform by the movement of dislocations,which resulted in an increase in the EI from 11.9% (10−1/s) to 22.5% (10−2/s).When the tensile strain rate was further decreased to 10−3/s,a completely ductile fracture with lots of dimples was observed.The slope of the UTS versus the strain rate increased rapidly,and the dimples became deeper as the tensile strain rate decreased to 10−4/s.Due to the GBS and extremely slow deformation at a tensile strain rate of 10−5/s,the fracture surface formed by deep dimples showed a surprising ductility (EI close to 300%) (Fig.13e).

4.Discussion

4.1.Microstructure dependence on processing route

Fig.10.The in-situ SEM-DIC analysis during the tensile test of the 2 Cross-pass sample.(a-e) The local strain distribution maps at global tensile strains of 0%,20%,40%,60%,and 80%.(f-j) The linear local tensile strain analyzed along the line L3 in (a-e).(k-o) High-resolution local strain distribution maps at area C in (a-e).

After a total of four passes of FSP,a significan difference in microstructure can be observed between the Route I and the Route II.Previous studies of the SPD process toward the duplex Mg-Li alloys inferred that the complexity of material fl w during the process had a much visible influ ence on the grain refinemen even the total strain is smaller[14–22].R.Wuet al.reported the application of ARB on the Mg-8Li-3Al-1Zn alloy and the grain size was only refine from ∼40 μm to less than 10 μm by 2 passes [15].M.Furuiet al.conducted 8 passes of ECAP to the Mg-9Li-1Al and only refine the grain size from ∼20 μm to ∼3 μm [16].Though the above-mentioned methods were capable of introducing large total strain to the material,the relatively simple deformation pattern fettered its capacity further to reduce the grain size of duplex Mg-Li alloy.In our previous study,only 1 pass of FSP could refine the grain size of Mg-9Li-1Zn alloy from ∼20 μm to around 1 μm due to the much complicated material fl w during the deformation [23].Comparing to the ARB and ECAP,the material fl w in FSP contains multiple deformation pattern simultaneously that induced severe local location redistribution of two phases,which further promote the grain refinemen [39].Besides,the grain refinemen and spacing distance of theαphase proved its remarkable influ ence on theβphase.At the early stage of the process,like 1 Pass,the grain refinemen of theαphase was much evident than that of theβphase (Fig.3b,c).In the further process,theαphase grains act as the “pinning particles” to prevent the grain growth of theβphase.The coarsenβgrains in Fig.3g,o,k only existed between the agglomerated band-likeαphase.With the complete elimination of the agglomerated band-likeαgrains,the uniform distribution of the ultra-fineαphase (Fig.3s) significantl reduced the spacing distance that provided a strong “pinning” effect towards theβphase.The specifi material fl w via Route II offered this ability to uniformly refin and distribute theαandβphase,resulting in the best grain refinemen effect.In this study,the material fl w inRoute IIwas much complicated than theRoute Idue to the reverse setting of the processing direction.The deformation pattern,especially the shearing deformation introduced to the material was kept the same direction in every passes inRoute I.However,in the 1 Cross-pass and the 2 Cross-pass,the shearing direction was altered reversely that completely changed the material fl w during the deformation.Consequently,the reverse setting of the processing direction inRoute IIinduced a much complicated material fl w that resulted in the smaller grain size at the stir zone.

Fig.11.The quasi-in-situ observation of the microstructure 1 mm beneath the surface of the 2 Cross-pass sample in the (a) undeformed state,(b) global strain of 30%,(c) global strain of 80%,and (d) the surface morphology tilted 45° at a global strain of 80%.

Fig.12.(a) BF image,(b) HAADF images,(c) the typical high-resolution TEM image of the α/β boundary of the 2 Cross-pass sample before deformation.(d) BF image,(e) HAADF images,and (f) typical high-resolution TEM image of the alpha/beta boundary of the 2 Cross-pass sample after 80% global tensile strain was applied.The orange arrows indicate that the particles contain a heavy element.The green dotted area indicates the cluster of α phase grains after tensile tests.

Fig.13.The fracture surface of the 2 Cross-pass sample at initial tensile strain rates of (a) 10−1 /s,(b) 10−2 /s,(c) 10−3 /s,(d) 10−4 /s,and (e) 10−5/s.The dimples corresponding to ductile fracture are indicated by the green arrows.The coordinate system for the images is shown on the top left.

4.2.Strengthening mechanism

Due to the remarkable differences in strength between theαgrains andβgrains at the BM,the BM exhibited severe local strain inhomogeneity during tensile deformation.This local strain inhomogeneity is reportedly associated with pileup geometric necessary dislocations (GNDs) to maintain geometrical integrity [40].Meanwhile,the grain morphologies were also shown to play a vital role in determining the local strain distribution.Concerning the hot-rolled Mg-9Li-1Zn BM used in this study,mostαgrains showed a compressed morphology (Fig.1),which led to a weak strain contribution during tensile deformation (Fig.8).To accomplish this deformation,the adjacentβgrains contributed most of the tensile strain,and a crack eventually formed on theβphase side due to concentrated pile-up GNDs [40].Rarely strain-rate sensitivity was observed at the BM (Fig.7),which implies that conventional grain boundary strengthening was the main strengthening mechanism.

The SZ (Fig.7) showed a dramatic change in its mechanical behavior compared with the BM.The strength and elongation changed significantl upon changing the tensile strain rate.At a relatively high strain rate such as 10−1/s,the 2 Cross-pass sample with the fines grain size showed the highest UTS,while the lowest UTS was obtained at a relatively low strain rate like 10−5/s.This change indicated that the conventional grain boundary strengthening mechanism was no longer the main strengthening mechanism,especially at low tensile strain rates (<10−3/s) as the GBS happened.For most metals,the most effective strengthening method is grain boundary strengthening,which is described by the Hall-Patch relation in Eq.(1):

whereσyrepresents the yield strength,σ0is the lattice friction coefficientkyis a material-related constant,anddis the mean grain size.Fig.14 summarized the results of the Hall-Patch relation and revealed the critical of grain size of LZ91 to induce GBS at different strain rates.From the viewpoint of Eq.(1),the 2 Cross-pass sample with the fines grain size of 210 nm should exhibit the highest strength,regardless of the deformation state;however,the SZ after FSP,especially the 2 Cross-pass sample,exhibited a significan variation in its UTS upon changing the strain rate.The UTS of the 2 Crosspass sample significantl decreased from 220.1 ± 6.3 MPa to 134.4 ± 2.7 MPa,even though the tensile strain rate only slightly decreased from 10−3/s to 10−4/s.This showed the highest slope of UTS variation along with changing the tensile strain rate (Fig.7b).In Fig.14,the slop change of Hall-Patch relation illustrated the critical grain size of GBS.The slop of the Hall-Patch relation at the strain rate of 10−1/s and 10−2/s were similar because the critical grain size for GBS is much smaller than 210 nm at this high strain rate.As the strain rate reduced to 10−3/s,the critical grain size for GBS was revealed to be between 1.29 μm and 1.42 μm.Though the slop of the Hall-Patch relation was decreased after the critical grain size for GBS at the strain rate of 10−3/s,the values remain positive,which infers the grain boundary strengthening still dominates the deformation behavior.However,as the strain rate decrease to 10−4/s,the slop of the Hall-Patch relation turned negative (−28.36 MPa/μm−1/2) after the critical grain size for GBS,befittin from the low strain rate of 10−4/s,the critical grain size for GBS increased to over 1.92 μm.At the strain rate of 10−5/s,no critical grain size for GBS can be observed and the slop of the Hall-Patch relation further decreased to −36.99 MPa/μm−1/2.This indicated the GBS dominated the deformation behavior at the strain rate of 10−5/s.

The strain sensitivity indexmis widely used as an indicator of deformation behavior [20,21,41] and is given by the following Eq.(2):

where theσ1and theσ2are the UTS of the tensile tests,andε1andε2are the strain rates of the tensile tests.Using this equation,mof the 2 Cross-pass at tensile strain rates from 10−3/s to 10−4/s was calculated to be 0.21.The value ofmof BM is near 0 because the grain boundary strengthening acts as the dominant strengthening mechanism.Themvalue of 0.21 of the 2 Cross-pass sample strongly indicates that a certain amount of grain boundary sliding (GBS) contributed to the tensile deformation.GBS is a diffusion-controlled phenomenon,and its initial strain rate is define by Eq.(3) [20]:

whereAis a constant (∼10),Dis the diffusion coefficien(D=D0exp(-Q/RT),D0is a frequency factor,Qis the activation energy for grain-boundary diffusion,Ris the ideal gas constant,Gis the shear modulus,bis the Burgers vector,kis Boltzmann’s constant,Tis the deformation temperature,dis the average grain size,σis the tensile stress,andpandnare the grain size and stress exponents,respectively.This equation implies that decreasing the grain size (d) will increase the initial strain rate for the GBS.Since the grain size (d) is in the denominator of this equation,and the exponentp(usually determined as 2) is larger than 1,decreasing the grain size would exponentially increase the initial strain rate for the GBS.Based on Eq.(3),it can be concluded that decreasing the value ofd(by grain refinement increases the initial strain rate of the GBS.The grain size of the 2 Cross pass sample was approximately 210 nm,which was sufficientl small to induce GBS at a tensile strain rate less than 10−3/s [20].In addition to the grain size,changing the diffusion coefficien(D,Eq..(2)) also increased the initial strain rate of the GBS.The high-resolution TEM image (Fig.12c) of theα/βboundary showed an incoherent structure because the severe deformation and low temperature provided a weaker driving force for grain boundary reconstruction to form a coherent boundary.Moreover,the incoherent structure of theα/βboundary also decreased the activation energy for grain boundary diffusion (Q,Eq.(3)).A lowerQimplies an increase in the diffusion coefficien (D,Eq.(3)),which further increased the strain rate necessary to initiate GBS.The transformation of theα/βboundary from incoherent to coherent after tensile deformation demonstrated the occurrence of GBS that was driven by grain boundary diffusion (Fig.13c,f).After FSP,the strengthening mechanism of Mg-9Li-1Zn changed from boundary strengthening to GBS,which produced a superplasticity in the 2 Cross-pass sample at a tensile strain rate of 10−5.This was attributed to the low activation energy for grain boundary diffusion (Q) from the incoherentα/βboundary.The dramatic change in the mechanical behavior of the 2 Cross-pass sample at different strain rates was due to the evolution of mechanical strengthening.At a relatively high strain rate (like 10−1/s),grain boundary strengthening was the main mechanism because the diffusion velocity of GBS was much lower than the deformation rate.The Hall-Patch relation worked at this state,and the 2 Cross pass-sample had the smallest grain size that contributed to the highest UTS during tensile tests at a high strain rate of 10−1/s.In contrast,the grain boundary did not act to impede the movement of dislocation but diffused in the 2 Cross-pass sample at relatively low strain rates(like 10−5/s).The 4 Unidirectional-pass sample fromRoute Ihas a coarse grain size than that from theRoute II,which fettered its deformation ability at low strain rate and resulted in lower elongation.The low deformation rate permitted boundary diffusion-induced GBS during the tensile tests,which provided the sample with an extraordinary ductility but the lowest UTS.

Fig.14.The Hall-Patch relation of LZ91 under different processing states at different strain rates.The results at the strain rate of 10−1 /s were performed as the index,assuming that the strengthening mechanism deviate from pure grain boundary strengthening.The comparison between the strain rate of 10−1 /s with the strain rate of (a) 10−2 /s,(b)10−3 /s,(c)10−4 /s and (d)10−5 /s were performed,and the related Hall-Patch relations were calculated.

4.3.Effect of processing route on the deformation mode

Regarding the strengthening mechanism,the deformation behavior after FSP also changed dramatically when selecting different process route.Unlike the BM and that after theRoute I,the 1 Cross-pass sample already exhibited a certain level of homogeneous deformation.One main factor is the alternation of the grain morphology from a compressed state at the BM to a refine equiaxed state after FSP.As shown in Fig.3o,the 1 Cross-pass sample contained most of the refine and uniformly-mixedαandβgrains,while also contained a small number of band-likeαgrain clusters.In the areas in the 1 Cross-pass sample without these band-likeαgrain clusters,the local strain distribution was relatively uniform,and the linear strain evolution along with an increase in global strain was smooth (0–120 μm at L2 in Fig.8f-j).The residual band-likeαgrain clusters showed a remarkably strong local strain restriction capacity that was similar to the largeαgrains at the BM,which was related to the grain morphology.It can be concluded that the remarkable grain refinemen and uniform distribution of two phases fromRoute IIeradicated the microstructure inhomogeneity (especially the agglomeration of theαphase).As described,GBS mainly occurred at theα/βgrain boundary,and it was difficul for GBS to be induced within the residualαgrain clusters because the great difference in active energy.This microstructure inhomogeneity was revealed to be one of the reasons why both the 1 Pass and 1 Cross-pass samples showed lower ductility than the 2 Cross-pass sample at a strain rate less than 10−3/s.On the other hand,the induction of GBS also required a small grain size (d) and low activation energy (Q) as described in Eq.(3).The 1 Pass and 1 Cross-pass samples showed a coarser grain size than the 2 Cross sample,which somehow prevented the induction of GBS.Ultra-fin grains up to 210 nm in the 2 Cross-pass remarkably decreased the grain sized,which significantl increased the initiation strain rate for GBS.

Another factor lies in the grain boundary evolution after 2 Cross-pass.Previous reports have shown that the boundary diffusivity of ɑ/ɑ grains in a Mg-Li system is 1.61×10−16m2s−1,which is orders of magnitude lower than the roomtemperature diffusivity along Li-richβ/βandα/βboundaries,which have a diffusivity of 3.7 × 10−10m2s−1[20].After carefully analyzing the average length of theα/βboundaries within 100 μm2(10 μm × 10 μm),their values in the BM,the 1 Pass,4 Unidirectional-Pass,and the 2 Cross-pass samples were 3.1 ± 0.9 μm/μm2,121.6 ± 17.4 μm/μm2,231.9 ± 44.1 μm/μm2and 362.2 ± 31.7 μm/μm2,respectively.The 2 Cross-pass sample astonishingly showed over 100 times the value of the BM and 3 times higher than the 1 Pass,which is believed to be the key factor that induced GBS during deformation.However,many SPD methods applied to Mg-Li alloys,such as accumulative roll-bonding(ARB) [14,15],equal channel angular extrusion (ECAE) [16–18],and high-speed-ratio differential speed rolling (HRDSR)[19] could not uniformly mixα/βgrains,even though they produced similar or even smaller grain sizes.This is because of the “stir” effect by the severe shear deformation introduced by the FSP.This suggests that FSP should be considered as the most ideal processing method for metals and alloys,especially those that contain a functional second phase (or particles) because of the extraordinary dispersion effect.

The incoherence of the boundary structure between the ɑ andβgrains also contributed to the induction of GBS.GBS is a diffusion-controlled process,and its diffusion coefficienDis described by Eq.(4)

In this equation,D0is a frequency factor,Qis the activation energy for grain-boundary diffusion,andRis the ideal gas constant.As reported by the theoretical studies and atomic-scale experimental analyses,the incoherent grain boundaries in severely deformed materials have nonequilibrium states similar to those in irradiated metals.The driving force for the segregation of solute atoms along boundaries might be the reduction in the boundary energy [42–44].The coherency between theα/βgrains shown in Fig.12f illustrates that the diffusion coefficienDin this study should be greater than the undeformed state and tends to return to the equilibrium state after GBS.As shown in Fig.12c and Fig.12f,the incoherent grain boundary state between theα/βgrains was coherent after superplastic deformation,which proved the occurrence of diffusion-dominated GBS.

Consequently,the 2 Cross-pass sample subjected to the FSP method in this study exhibited high ductility,and the main reason was revealed to be the introduction of the GBS.Through the selection of FSP Route II,the remarkably uniform refine grain size,the significantl increased high-diffusivityα/βboundary length,and the incoherentα/βboundary synergistically contributed to uniform deformation by GBS.

5.Conclusion

An ultra-fin microstructure and extraordinary mechanical properties of Mg-9Li-1Zn alloy LZ91 were obtained by the FSP method viaRoute II.The effects of the processing route on the microstructure and mechanical behavior of superlight Mg-9Li-1Zn alloy via friction stir processing were systematically analyzed,and the following conclusions were obtained:

(1) The excellent strength-ductility synergy Mg-Li alloy LZ91 was produced this study.The grain size was refine from ∼31 μm to ∼0.21 μm with uniformly distributedαandβphases.It exhibited the UTS of 220.1 MPa and total elongation of 70.0% at a strain rate of 10−3/s,overwhelmingly higher than those of the base metal,155.6 MPa in UTS and 36.0% in elongation.

(2) Unidirectional FSP(Route I)leads to an asymmetric microstructure and mechanical properties.By reversing the processing direction,a highly symmetric and homogeneous SZ was obtained viaRoute II.After 2 Cross-pass,the resultant grain size was refine to 210 nm,combined with a uniform distribution ofαandβphases that has not been previously reported.

(3) Thein-situSEM-DIC analysis revealed that the BM exhibited strong local strain concentration at theα/βboundary.After 1 Cross-pass,the sample showed much better local strain homogeneity during tensile deformation than the BM,whilst the 2 Cross-pass sample exhibited an remarkable uniform local strain distribution during the tensile tests,suggesting that any strain localization is mediated by grain boundary activities.

(4) With the grain refinement the deformation mechanism gradually changed from intra-granular deformation to GBS.The 2 Cross-pass sample has a strain rate sensitivity value of 0.21,which implies the occurrence of GBS as the dominant deformation mechanism at low strain rate (10−5/s).

(5) In addition to great refinement the substantially enhanced GBS in the 2 Cross-pass sample is caused by significantl enhancedα/βphase boundary of 362.2 ± 31.7 μm/μm2,in contrast to 3.1±0.9 μm/μm2in the BM to 231.9±44.1 μm/μm2in the 4 Unidirectional-pass sample.The high-resolution TEM images revealed incoherent boundaries between theα/βphases,which further reduced the activation energy for grain boundary diffusion (Q).The decrease inQallows GBS to be able to operate at a higher strain rate,and therefore leading to a better ductility at higher strain rate (10−3/s).

Data availability

The raw/processed data required to reproduce these find ings cannot be shared at this time due to technical or time limitations.

Acknowledgments

This study was partially supported by the JST-Mirai Program Grant Number JPMJMI19E5,and a Grant-in-Aid for Science Research from the Japan Society for the Promotion of Science.Mengran Zhou would like to appreciate the Interactive Materials Science Cadet Program from Osaka University and the Ministry of Education,Culture,Sports,Science and Technology (MEXT) of Japan.

Supplementary materials

Supplementary material associated with this article can be found,in the online version,at doi:10.1016/j.jma.2021.12.002.