Microstructure evolution in hot worked and annealed magnesium alloy AZ31

Materials Science and Engineering A 485 (2008) 318–324Microstructure evolution in hot worked and annealedmagnesium alloy AZ31A.G.Beer ∗,M.R.BarnettCentre for Material and Fibre Innovation,Geelong Technology Precinct,Deakin University,Geelong,Victoria 3217,AustraliaReceived 10February 2007;received in revised form 27July 2007;accepted 1August 2007AbstractThe present work examines the microstructure that evolves during the annealing of hot worked magnesium alloy AZ31.First,the influences of deformation and annealing conditions on the microstructures are assessed.It is found that the annealing behaviour is consistent with what one would expect for a recrystallization type reaction.Whilst both the deformation and annealing conditions influence the time required to reach a stable annealed microstructure,the grain size attained is governed solely by the prior deformation conditions employed.At the highest temperature and strain rate examined,the rate of recrystallization is quite high and the grain size was found to be approximately double when annealed for only 1s prior to quenching.Finally,semi-empirical equations are developed to predict the kinetics of recrystallization,as well as the evolution of grain size,during annealing.© 2007 Elsevier B.V. All rights reserved.Keywords:Hot working;Annealing;Dynamic recrystallization;Static recrystallization;Grain size;Magnesium alloy AZ311.IntroductionThe low density of wrought magnesium makes it an ideal choice for automotive applications,such as body panels and space frames.However,its widespread application is restricted due to its limited strength and ductility.Grain refinement is a promising technique for improving the strength and ductil-ity of wrought magnesium products due to the strong influence that grain size has on the mechanical properties.For example,reducing the grain size of extruded AZ31from 22␮m down to 3␮m has been shown to effectively double the compressive yield strength [1].In magnesium and its alloys,grain refinement is commonly achieved through the operation of dynamic recrystallization (DRX)during hot deformation [2–9].For the most common wrought alloy,AZ31,grain sizes as fine as 3–5␮m can be obtained [9].However,several authors have recently highlighted the increase in grain size that occurs following the subsequent annealing,or slow cooling,of hot-rolled [10–14]and extruded [14,15]magnesium alloys.Yamamoto and co-workers found that annealing hot-rolled AZ31at 400◦C for 1h increased the∗Corresponding author.Tel.:+61352273321;fax:+61352271103.E-mail address:aiden.beer@.au (A.G.Beer).grain size from between 5and 10␮m to over 50␮m [12].These observations illustrate the deterioration in the final mechani-cal properties that can be expected during the cooling or heat treatment of hot worked magnesium alloys.There is,however,confusion in the literature regarding the mechanism /s by which grains coarsen during annealing of hot worked magnesium alloys.In most cases it is stated that grain growth occurred during annealing (e.g.[10,16–19]),whilst oth-ers propose that static and/or metadynamic recrystallization may operate [9,15].Secondary recrystallization,or abnormal grain growth,has also been observed to operate in magnesium alloys but typically when high annealing temperatures are employed (450◦C and greater)[14,16,20].It has,however,been observed after the annealing of hot rolled AZ31at 250◦C when the anneal-ing time was 1000min [13].Factors that influence the operation of this grain coarsening in magnesium alloys have received little attention.A few stud-ies highlight the fact that the deformation conditions applied prior to annealing influences the way in which the microstruc-ture evolves [9–11,19].Yang et al.[19]found that,for the same annealing conditions,larger grain sizes were attained in AZ31samples that were compressed at a lower strain rate.In an earlier study,one of us examined the dependence of grain size on theZener–Hollomon parameter,Z (=˙εexp(Q DEF /RT )),following the annealing of AZ31samples compressed in a channel die [9].0921-5093/$–see front matter © 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.msea.2007.08.001A.G.Beer,M.R.Barnett/Materials Science and Engineering A 485 (2008) 318–324319It was found that the grain size remained small at low values of Z,abnormally large grains were developed at intermediate val-ues of Z,and the annealed grain size decreased with increasing Z at high values of Z.In this earlier study,as well as in many oth-ers,a constant annealing time,typically in the order of hours,is employed.While it has been shown that longer annealing times generate larger grain sizes[14,16],the rate at which coarsen-ing occurs immediately following hot deformation has not been systematically examined.The present work investigates the influence of deformation and annealing conditions on the degree of grain size increase during the annealing of hot worked magnesium alloy AZ31.The data thus obtained are used to develop semi-empirical equations that describe the evolution of grain size during annealing and these are then verified with results from the literature.2.Experimental procedureThe material used in the current study is magnesium alloy AZ31(Mg–3%Al–1%Zn–0.2%Mn),purchased in the form of extruded rod,with an initial grain size of22.5␮-pression samples,15mm high and10mm in diameter,were machined such that the compression direction was parallel to the extrusion direction.Teflon tape was used as a lubricant and all samples were held at the test temperature for5min prior to deformation.Uniaxial compression testing was conducted at strain rates ranging between0.01and1s−1,and at temperatures rang-ing between300and400◦C.Deformation was carried out to an equivalent strain of1.5,after which the samples were held at the deformation temperature for times ranging from 1s to10,000s.After the annealing step,a spring-loaded lever immediately ejected samples into a water bath,and it is estimated that samples were quenched in less than 0.5s.For metallographic examination,samples were mounted in epoxy resin,ground with1200grit SiC paper and polished with 6and3␮m diamond paste.Samples were then chemically pol-ished for45s in10%nital and then etched for5s with acetic picral.The average grain sizes were determined from optical micrographs using a linear intercept method.3.ResultsThe microstructural evolution during the annealing of AZ31, for samples deformed at a temperature of350◦C and at three strain rates,and subsequently annealed at350◦C for increas-ing times,is presented in Fig.1.It can be seen that that larger grains are developed with increased annealing times,with the degree of grain coarsening being greater for the samples that were deformed at the lowest strain rate.It can also be seen that, for annealing times of100s and above,the microstructure isFig.1.The microstructural evolution of wrought AZ31deformed to a strain of1.5in compression,at a temperature of350◦C and strain rates between0.01and 1s−1,and subsequently annealed at the deformation temperature for increasing times:(A)350◦C,0.01s−1,1s;(B)350◦C,0.01s−1,100s;(C)350◦C,0.01s−1, 10,000s;(D)350◦C,0.1s−1,1s;(E)350◦C,0.1s−1,100s;(F)350◦C,0.1s−1,10,000s;(G)350◦C,1s−1,1s;(H)350◦C,1s−1,100s;(I)350◦C,1s−1,10,000s.320 A.G.Beer,M.R.Barnett/Materials Science and Engineering A 485 (2008) 318–324 unchanged in the samples deformed at the highest strain rate(Fig.1H and G)whilst the microstructure is still evolving forthe samples deformed at the lowest strain rate(Fig.1B and C),i.e.a stable annealed grain size is attained at earlier times withhigher deformation rates.To quantify the effects of deformation and annealing condi-tions on the microstructure,the evolution of the average grainsize is plotted against annealing time in Fig.2for a constantdeformation and annealing temperature(Fig.2(a))and a constantstrain rate(Fig.2(b)).With an increase in strain rate from0.01to1s−1(Fig.2(a)),the stable,or plateau,grain size(d PLATEAU)attained is reducedfrom approximately40␮m down to13␮m.It can also be seenthat with an increased strain rate,the time at which coarseningbegins is reduced,from approximately100s to less than10s,asis the time to reach the plateau grain size.An increase in both thedeformation and annealing temperature results in an increaseinFig.2.The influence of(a)strain rate(with a deformation and annealing tem-perature of350◦C)and(b)deformation and annealing temperature(with a strain rate of0.1s−1)on the evolution of the average grain size with annealing time after the hot deformation of magnesium alloy AZ31to a strain of1.5.Error bars indicate the variation in grain size determined from duplicate samples,being l–2␮m at small grain sizes up to5–6␮m at larger annealed grain sizes.Very little variation was observed for the time to reach the plateau grainsize.Fig. 3.The influence of the Zener–Hollomon parameter,Z (=˙εexp(135,000/RT)),on the plateau grain size,d PLATEAU,the dynamically recrystallized grain size,d DRX(from samples that were immediately quenched), and d(t=1s)(from samples that were annealed for1s prior to quenching), developed after the hot deformation of magnesium alloy AZ31to a strain of1.5 (annealing temperature equal to the deformation temperature).the grain size developed after1s annealing and in the plateaugrain size(Fig.2(b)).The time at which coarsening begins,aswell as the time to reach the plateau grain size,is similar for thethree temperatures.The combined influence of deformation temperature andstrain rate on the plateau grain size is shown in Fig.3,whered PLATEAU is plotted against the Zener–Hollomon parameter,Z(=˙εexp(Q DEF/RT)).Here,Q DEF is taken as the activation energy for self-diffusion(135kJ/mol)[22,23].Also included isthe dynamically recrystallized grain size,d DRX,measured fromsamples that were quenched immediately after deformation,andthe grain size of samples annealed for1s,d(t=1s).Both d PLATEAU and d DRX obey a power law relationship withZ,with the degree of grain coarsening(the difference between d DRX and d PLATEAU)increasing with decreasing values of Z.The plateau grain size ranges from two tofive times d DRX over the conditions examined here.Interestingly,similar plateau grain sizes are attained under equivalent deformation conditions(i.e. similar Z values),despite the fact that different annealing temper-atures are employed.The fact that the plateau grain size depends on the applied deformation conditions and not the annealing tem-perature is indicative of a recrystallization type reaction[21].It can also be seen that the grain sizes of samples annealed for 1s(d(t=1s))prior to quenching are similar to those that were quenched immediately(d DRX),apart from two instances.For these cases,the applied strain rate was1s−1and the average grain size has approximately doubled within the1s anneal time prior to quenching.4.DiscussionOf particular importance in the presentfindings are the fol-lowing points:(i)the high rate of grain coarsening,(ii)theA.G.Beer,M.R.Barnett/Materials Science and Engineering A 485 (2008) 318–324321 influence of Z on the annealed grain size,and(iii)the influenceof Z on the time to reach the plateau grain size.4.1.The high rate of grain coarseningAn importantfinding in the current study is that the rate ofgrain size increase during annealing after hot deformation isexpected to be quite high for the conditions encountered in com-mercial processes.Fig.3reveals that in some cases,when thestrain rate is1s−1,the grain size developed when the material hasbeen annealed for1s prior to quenching is approximately twicethat of when it has been quenched immediately.During high ratedeformation processes such as extrusion,when grain coarseningrates would be comparatively higher,it is unlikely that quench-ing rates would be sufficient to retain the DRX microstructure.Inthese cases,a stable annealed microstructure would be quicklydeveloped.However,the grain sizes developed in high rate com-mercial processes can be minimised by processing at the higheststrain rate or lowest temperature(highest value of Z)possible,as indicated by Fig.3.A study by Kawalla and co-workers[11]examined themicrostructures of rolled wedge-shaped AZ31specimens thatwere either quenched immediately or held at the deforma-tion temperature for up to60s prior to quenching.At thehighest strain,immediate quenching produced grain diametersof24–30␮m(with statically recrystallized grain diameters of27–30␮m obtained after annealing).The present work suggeststhat the rate of recrystallization induced coarsening is so highthat even the“immediate”quenching employed in that studymay have been insufficient to“lock in”the dynamically recrys-tallized grain size.4.2.The influence of Z on t(dPLATEAU)It is evident from Fig.2that the time at which coarseningbegins,and the time at which the plateau grain size is reached,isa strong function of strain rate(Fig.2(a))but relatively indepen-dent of the deformation and annealing temperature(Fig.2(b)).Similar observations have been made for the kinetics of metady-namic recrystallization(also referred to as post-dynamic staticrecrystallization)in C–Mn steel[22].Many studies have employed semi-empirical equations todescribe the kinetics of recrystallization in steel,for example[22,23].A similar approach is adopted in the current studyto model the recrystallization kinetics in magnesium(a lackof physical terms for magnesium currently restricts the devel-opment of a physically based recrystallization model).Whilstthese equations are of practical use,and include likely physicaldependencies to represent the physical processes occurring,theinterpretation the results in a physical sense is difficult,e.g.theactivation energy in Eq.(1)refers to the reaction as a whole andis unlikely to be constant[24].The time to reach the plateau grain size can thus be describedusing the expression(after[22,25]):t(dPLATEAU)=BZ r expQ REXRT(1)where B and r are constants.Eq.(1)excludes terms for strain(ε)or original grain size(d o)as the material was completely dynam-ically recrystallized prior to annealing and thus their influenceon the recrystallization kinetics is expected to be negligible[22].A plot of log(t(dPLATEAU))versus Z(with Q DEF equal to135kJ/mol),for equivalent annealing temperatures,is given inFig.4(a).A linear relationship is assumed,with the average gra-dient,r,found to be−1.22.A linear relationship is also assumedbetween ln(t(dPLATEAU))and1000/RT,for average values of Z(Fig.4(b)).Horizontal bars indicate errors due to using averagevalues of Z.In this plot,the average gradient,which correspondsto the activation energy for recrystallization,Q REX,was deter-mined to be200kJ/ing the determined values of r andQ REX,constant B in Eq.(1)was found to be2.7×10−3.Fig.4(a)reveals the combined roles of deformation tempera-ture and strain rate(Z),for a constant annealing temperature,ont(dPLATEAU).It is clear that at higher values of Z,the time to reachthe plateau grain size is decreased.That is,the kinetics of recrys-tallization is enhanced at high strain rates and low temperatures.This can be related to a higher stored energy(which provides thedriving force for recrystallization)in the dynamically recrystal-lized structure at higher values of Z.It is also evident that theFig. 4.The influence of(a)Zener–Hollomon parameter,Z(=˙εexp(135,000/RT)),and(b)1000/RT,on the time to reach the plateau grain size,t(dPLATEAU),during the annealing of hot deformed magnesium alloy AZ31.322 A.G.Beer,M.R.Barnett /Materials Science and Engineering A 485 (2008) 318–324kinetics of recrystallization is enhanced with higher annealing temperatures,which is expected due to the fact that recrystalliza-tion is controlled by thermally activated processes.The fact that Fig.2(b)indicates that the kinetics of recrystallization is insen-sitive to the deformation and annealing temperature is due to the “double”effect of temperature,whereby the increase in bound-ary mobility and nucleation rate with increasing temperature is offset by a decrease in the energy stored during deformation,which drives the reaction.4.3.The Influence of Z on d PLATEAUFig.3reveals that d PLATEAU obeys a power law relation-ship with Z and does not depend on the annealing temperature.This strong Z dependence is indicative of a recrystallization type reaction and is analogous to the Z dependence of the grain size during metadynamic recrystallization in steels,e.g.[22,26,27].As recrystallization involves both nucleation and growth,factors which influence these processes may alter their relative activity and will inherently influence the final grain size.The Z depen-dency of d PLATEAU in the current study is likely to be related to the influence of Z on the density of nucleation sites.The fact that the annealing temperature influences the kinetics of recrystal-lization (t (d PLATEAU ))but not the annealed grain size (d PLATEAU )indicates that the annealing temperature does not change the relative rates of nucleation and growth during recrystallization.The mean grain size,¯d,that evolves during recrystallization can be modelled by relating it to the dynamically recrystallized grain size (d DRX ),the fully recrystallized grain size (d PLATEAU )and the volume fraction recrystallized (X )through the expression (after [23]):¯d(t )=X 4/3(t )d PLATEAU +(1−X (t ))2d DRX (2)where d DRX and d PLATEAU are determined using the power law expressions given in Fig.3.X ,the fraction recrystallized,is described by the modified Avrami equation (as proposed byparison between the predicted and measured average grain sizes thatare developed during the annealing of hot deformed magnesium alloy AZ31for times ranging from 1to 10,000s (grey lines represent ±15%of the measured grain size).Sellars [25]):X =1−exp −4t t (d PLATEAU )n(3)where t (d PLATEAU )is given in Eq.(1)and the Avrami exponent,n ,was found to be 1.3.Predictions of the annealed grain sizes are compared with the actual values in Fig.5.It can be seen that the model provides a reasonable description of the grain size during annealing,with most predicted grain sizes being within ±15%of the measured grain size.The grain sizes attained after annealing hot compressed AZ31to a given annealing time have been examined previously [9,19].Using the model developed in the current study,predictions of the annealed grain sizes are compared with the published values in Figs.6and 7,with the predicted fraction recrystallized also shown.Yang et al.[19]examined the grain sizes attained in hot compressed samples that were subsequently annealed at tem-Fig.6.Predictions of annealed AZ31grain sizes from the literature with increasing annealing temperature:samples annealed at increasing temperatures for 1000s after compression to a strain of 1.2,at a temperature of 300◦C and at strain rates of (a)0.3s −1[19]and (b)0.003s −1[19].A.G.Beer,M.R.Barnett/Materials Science and Engineering A 485 (2008) 318–324323Fig.7.Predictions of annealed AZ31grain sizes from the literature with increas-ing Z(higher strain rates,lower deformation temperatures):samples annealed at 300◦C for30min after channel die compression to a strain of0.6at temperatures between300and400◦C and at strain rates between0.001and1s−1[9]. peratures between100and400◦C for1000s.These workers concluded that grain growth occurred during annealing.How-ever,the current recrystallization model(Fig.6(a)and(b)) predicts similar grain sizes,and a similar trend in the tempera-ture dependence of the average grain size.For the higher strain rate(Fig.6(a)),the model predicts that recrystallization occurs between473and523K.This coincides with the temperature range for rapid grain coarsening given by Yang and co-authors. Whilst the grain size is closely predicted for the high strain rate, to within2␮m(Fig.6(a)),the model over-predicts the grain size for the lower strain rate(Fig.6(b)).This may be related to the fact that,after such low strain rate deformation,the stored energy,and thus the driving force for recrystallization,would be low.As a consequence,the boundary mobility during recrys-tallization would also be low and boundaries would be more influenced by factors such as composition variation or the pres-ence of particles.Further work is required to quantify these effects.As mentioned earlier,one of us has examined the Z depen-dence of the grain size following the annealing,at300◦C for 30min,of samples compressed in a channel die[9].The result-ing behaviour was ascribed to grain growth and/or metadynamic recrystallization at low values of Z and static recrystallization at high values of Z.A similar trend in the Z dependence of the average grain size is predicted using the current recrystallization model(Fig.7).At low values of Z,the kinetics of recrystalliza-tion is slow and the average grain size is equivalent to that of the DRX grain size.The degree of recrystallization increases at intermediate values of Z,as does the average grain size.At high values of Z,recrystallization is complete and the fully annealed grain size decreases with increasing Z.Although a similar Z dependence is observed,the model predicts a grain size at the lower end of the observed range.This may be attributed to the lower strain applied in this study and the fact that the percentage DRX was not100%in all samples.Interestingly,the present modelfits the data for samples that were aligned for c-axis extension and c-axis compression,both of which had a com-parably higher DRX fraction than samples aligned for c-axis constraint.5.Conclusions1.The annealing behaviour of magnesium alloy AZ31is con-sistent with what one would expect for a recrystallization type reaction.2.The rate of recrystallization,in tests where the annealing tem-perature is the same as the deformation temperature,changes with strain rate but is insensitive to temperature.This is con-sistent with metadynamic recrystallization in other metals.3.The stable annealed grain size attained is governed solelyby the prior deformation conditions employed,with smaller annealed grain sizes being attained at higher Z conditions, which is again indicative of a recrystallization type reaction.4.The rate of grain coarsening is rapid at the highest temper-ature and strain rate examined.For high rate deformation processes,such as extrusion,quenching rates are unlikely to be high enough to retain the DRX microstructure.5.Equations were developed to predict the kinetics of recrys-tallization,as well as the evolution of grain size,during the annealing of hot worked AZ31.AcknowledgementsThis research was supported under the Australian Research Council’s Linkage funding scheme(project number LP0347685)and by Advanced Magnesium Technologies. References[1]M.R.Barnett,et al.,Acta Mater.52(2004)5093–5103.[2]S.E.Ion,F.J.Humphreys,S.H.White,Acta Metall.30(1982)1909–1919.[3]A.Mwembela,et al.,Magnesium Alloys and Their Applications,Werkstoff-Informationsgesellschaft,Wolfsburg,Germany,1998,pp.215–222.[4]A.Galiyev,R.Kaibyshev,G.Gottstein,Acta Mater.49(2001)1199–1207.[5]O.Sitdikov,R.Kaibyshev,Mater.Trans.42(2001)1928–1937.[6]H.Watanabe,et al.,Mater.Trans.42(2001)1200–1205.[7]A.G.Beer,M.R.Barnett,Magnesium Technology2002,TMS,Seattle,Washington,USA,2002,pp.193–198.[8]J.C.Tan,M.J.Tan,Mater.Sci.Eng.A339(2003)124–132.[9]M.R.Barnett,Mater.Trans.44(2003)571–577.[10]A.Galiyev,R.Kaibyshev,Mater.Sci.Forum467–470(2004)1175–1180.[11]R.Kawalla,N.Coung,A.Stolnikov,6th International Conference Magne-sium Alloys and Their Applications,Wiley-VCH,2004,pp.803–810.[12]A.Yamamoto,et al.,Mater.Sci.Forum(2004)669–672.[13]M.Eddahbi,et al.,Mater.Sci.Eng.A410–411(2005)308–311.[14]E.Essadiqi,C.Galvani,V.Kao,2nd International Light Metals TechnologyConference,LKR,St.Wolfgang,Austria,2005,pp.155–160.[15]J.Bohlen,et al.,Scripta Mater.53(2005)259–264.[16]M.T.Perez-Prado,O.A.Ruano,Scripta Mater.46(2002)149–155.[17]H.K.Kim,W.J.Kim,Mater.Sci.Eng.A385(2004)300–308.[18]H.K.Kim,J.Mater.Sci.39(2004)7101–7109.[19]X.Yang,H.Miura,T.Sakai,J.Jpn.Inst.Met.68(2004)946–951.324 A.G.Beer,M.R.Barnett/Materials Science and Engineering A 485 (2008) 318–324[20]G.T.Higgins,J.Nucl.Mater.8(1963)153–159.[21]F.J.Humphreys,M.Hatherly,Recrystallization and Related AnnealingPhenomena,Pergamon,Oxford,UK,1996,pp.369–392.[22]P.D.Hodgson,R.K.Gibbs,ISIJ Int.32(1992)1329–1338.[23]J.H.Beynon,C.M.Sellars,ISIJ Int.32(1992)359–367.[24]F.J.Humphreys,M.Hatherly,Recrystallization and Related AnnealingPhenomena,Pergamon,Oxford,UK,2004,p.230.[25]C.M.Sellars,J.A.Whiteman,Met.Sci.13(1979)187–194.[26]A.R.Morgridge,Bull.Mater.Sci.25(2002)291–299.[27]A.M.Elwazri,P.Wanjara,S.Yue,ISIJ Int.43(2003)1080–1088.。