A comparative study of the role of solute,potent particles and ultrasonic treatment during solidification of pure Mg,Mg–Zn and Mg–Zr alloys

Ngsivmuni Blsurmni,Gui Wng,Mrk A.Eston,Dvid H.StJohn,Mtthew S.Drgusch,?

a Centre for Advanced Materials Processing and Manufacturing(AMPAM),School of Mechanical and Mining Engineering,The University of Queensland,Brisbane 4072,Australia

bSchool of Engineering,RMIT University,Carlton,3053 VIC,Australia

Abstract Ultrasonic treatment(UST)applied during the solidification of pure Mg,eutectic(Mg–Zn)and peritectic(Mg–Zr)alloys was investigated in order to explore the grain refinement mechanisms.Temperature dependent grain refinement is observed in pure Mg where decreasing the superheat temperature(at which UST is applied from above the melting temperature,TM)from 100°C to 40°C produces significant refinement with a uniform grain structure.The presence of solute reduces the temperature dependence of the UST refinement and excellent grain refinement is obtained regardless of the superheat temperature(100°C or 40°C)and even with the use of preheated sonotrode in the Mg–6wt.% Zn alloy.A further improvement in grain refinement is achieved when the alloy contains potent particles that introduce additional nucleation of grains in Mg–0.5 and 1.0wt.% Zr alloys(producing an average grain size of≤100μm).At 40°C superheat,UST of Mg–Zn alloys produces excellent refinement(average grain size<200μm)with non-dendritic grains,which is normally achieved only with the addition of grain refining master alloy in the as-cast condition.The enhanced refinement observed in the eutectic alloy is explained through the undercooling imposed by a relatively cold sonotrode combined with high frequency vibrations and acoustic streaming.The advantages of using a cold sonotrode,a low superheat and solute are demonstrated for achieving significant refinement during solidification of Mg alloys under UST without or with a lower addition of grain refining master alloys.? 2020 Chongqing University.Publishing services provided by Elsevier B.V.on behalf of KeAi Communications Co.Ltd.This is an open access article under the CC BY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/)Peer review under responsibility of Chongqing University

Keywords:Grain refinement;Mg–Zn alloy;Mg–Zr alloy;Ultrasonic treatment;Interdependence Model;Solidification.

1.Introduction

Castings with fine equiaxed grains during solidification have the tendency to reduce macro-segregation and hot tearing,and isolated regions with large porosity thereby improving the mechanical properties of the alloy in the as-cast state and after subsequent wrought processing[1–3].During solidification of alloys,segregation of solute elements(Co)ahead of the solid–liquid interface retards the growth rate of grains and develops a constitutionally supercooled zone(ΔTcs)facilitating new nucleation events on particles within the constitutional supercooling zone.The effectiveness of solute in reducing the growth rate is quantified by the growth restriction factor,Q=mCo(k?1),wheremis the slope of the liquidus line andkis the partition coefficient on a phase diagram[4,5].Several alloying elements added to Mg alloys have been reported to produce refinement as a function ofQ[4,6,7].A better conceptual understanding of the solute-activated nucleation mechanism is provided by the Interdependence Theory of nucleation,and grain size predictions have been made successfully for a wide range of Al and Mg alloys[8].

A combination of effective substrates and adequate solute is the minimum necessary criterion to achieve significant grain size reduction during solidification in normal casting conditions under natural convection[1,5,6].Commercial Mg alloys(WE,ZE and ZK series alloys)typically contain 0.7wt.% of Zr(greater than the peritectic concentration of 0.5wt.%)that produces excellent refinement by the combination of solute Zr andα–Zr particles[9–11].The effectiveness of Zr as a grain refiner is poisoned when the alloy contains Mn,Si and Al,therefore,carbon based particles have been identified as alternative grain refiners for Mg–Al based alloys(AZ series)[1,2,4,12].Mg alloys are therefore often classified into two groups:those that are(i)refined by Zr and(ii)not refined by Zr additions[2,4,12].Although Zr offers the best refinement in most commercial Mg alloys,the efficiency of the process is low due to the challenges involved in(i)settling and loss of Zr,(ii)dispersion and size distribution of nucleants and(iii)increasing the number density of active nucleants[5,8,13].On the other hand,the carbon inoculation process in Mg–Al alloys adds more complexity due to the uncertainty about actual potent substrates(Al2OC,Al4C3,Al3MgC2)[2,4]and the mechanism of nucleation(single particle or duplex layers on a particle)[1,2,14].Alternative sources of grain refiners based on oxide particles either generated in-situ or added exsitu have been proposed for Mg alloys(MgO[15,16]CaO[17]and ZnO[18]);however,achieving consistency in refinement performance is still challenging because it relies on an effective addition method and increasing the number density of active particles[1,2,4,14,16].

Ultrasonic treatment(UST)during solidification of magnesium alloys has been shown to produce excellent refinement without addition of grain refiners and it also improves the efficiency of grain refinement when potent particles are added to the alloy[19–21].Ramirez et al.[22]and Qian et al.[23–25]have investigated a series of commercial alloys(AZ91,AJ62,AZ31)binary Mg alloys(Mg–Al and Mg–Zn)and pure Mg to demonstrate the potential of using UST to refine the grain size of magnesium.For a range of hypo-and hyper-peritectic compositions,the authors have investigated the effect of UST in binary Mg–Zr alloys and showed that(i)α–Zr particles were uniformly distributed in the melt,(ii)the settling of Zr particles is significantly reduced,(iii)improved dissolution of Zr solute from the master alloy,and(iv)an increased number of active Zr particles available for grain refinement[26].As there is a greater number of potent particles present in the peritectic alloys,UST grain refinement is expected to facilitate a higher rate of nucleation in Mg–Zr alloys[19,20,26].However,in eutectic forming Mg–Zn and Mg–Al alloys,Qian et al.[24]observed significant refinement without the addition of external inoculant particles.It is found that the grain density beneath the sonotrode in the solidified ingot is higher and decreases with the increase in distance from the sonotrode to the bottom of the crucible.This is further clarified recently by the authors where a gauze was placed below the sonotrode and captured fine,non-dendritic grains in the cavitation zone[27].The areas away from the cavitation zone have grains that are coarser and dendritic in morphology.In addition,the grains that are produced beneath the vibrating sonotrode were captured by a silica tube and quenched immediately,had grain sizes ranging from 50 to 250μm for an Al–2.0 Cu alloy[27].Interestingly,a similar grain refinement tendency is also obtained in pure metal solidification(Mg,Al,and Zn),where the final grain size ranged from 100 to 250μm(for a casting volume of 135 cm3)[20,28,29].It is important to note that this uniform,non-dendritic grain structure(100–200μm)in pure metals and alloys was obtained when using an unpreheated sonotrode over a temperature range closer to the liquidus temperature(i.e.UST turned on<40°C above the melting or liquidus temperature and continued after the onset of nucleation for a specified duration)[20,28,29].This indicates that the temperature range of UST becomes an additional variable that contributes to further refinement in addition to the solute and potent particles.

Therefore,the present work aims to systematically investigate the effect of temperature range of UST,solute and potent particles on grain refinement in two common magnesium alloying systems as well as pure Mg:

(i)Mg–Zr alloy:A peritectic alloy containing an excellent combination of potent Zr nucleant particles and solute with a very high growth restriction factor(m(k?1)=38.29/wt%))[4,6].

(ii)Mg–Zn alloy:Eutectic alloy that does not contain highly potent particles(oxide[15],Al–Fe based impotent and inactive particles[29])and weaker solute characterised by a low to moderate growth restriction factor(m(k?1)=5.32/wt%)[6].

(iii)Pure Mg:Pure metals have negligible solute and contain native particles(either oxide or impurity particles[29])

Grain refinement models have been successful in conceptualising the nucleation of grains with solute and thermal fields surrounding the grains,for example,characterization of the nucleation free zone[8]or solute suppressed zone[30],and prediction of the grain size of cast metals in normal casting conditions.The prediction of grain size during UST is difficult as the solute and temperature gradients are greatly affected[31].In addition to nucleation,grain refinement could also occur by the fragmentation of dendrites due to acoustic streaming and cavitation bubbles[32–34].As the Mg–Zr alloys contain potent Zr particles,the grain size was successfully predicted using the temperature gradient variable(z)incorporated into the Interdependence Model by assuming that UST enhances nucleation on Zr particles[26].Regardless of the origin or the location of new nucleation events,the Interdependence Theory[8]and analytical models[1,35,36]can be used to understand the role of solute and potent particles on the final grain size[37].Therefore,this work compares and investigates the origin of equiaxed grains during UST with normal casting conditions in order to explain the role of solute,potent particles and the grain refinement mechanism.

2.Experimental procedure

Magnesium ingot(99.9wt.%),high purity Zn ingot(99.995wt.%)and Mg–25 Zr master alloy were used to make Mg–Zn alloys with 0.5,3.0 and 6.0wt.% Zn and Mg–Zr alloys with 0.2,0.5 and 1.0wt.% Zr(all compositions are in wt.% unless stated otherwise).Alloys of the required nominal compositions were melted under a protective atmosphere(CO2+SF6)at 780°C and then the crucible containing the melt was removed from the furnace and cooled in the open air.One or two K-type thermocouples were either placed close to the centre of the casting or in off-centre position was used to measure the cooling curves of the respective alloys.Additional cover gas protection was provided to the solidifying alloys in order to prevent oxidation at the top surface of the casting and the schematic diagram of the experimental setup can be found in[29,37].UST was applied at a fixed power(40%)using a 1.5kW piezoelectric transducer attached to a titanium sonotrode(?19mm).The sonotrode was not preheated,however,the sonotrode was placed above the hot melt to warm the surface of the sonotrode to facilitate smooth operation,as otherwise the formation of a strong chill layer stops the functioning of UST abruptly.

Table 1UST temperature range and time duration of the alloys investigated.

Table 1 shows the superheat temperature,alloy composition,temperature range of UST applied and the time duration measured from the cooling curves.Superheat is the temperature above the melting or liquidus temperature of the alloy at which the sonotrode was turned on in air and immediately inserted into the melt from the top surface.Previous study has shown that the S1 range produces a relatively coarse and non-uniform grain structure compared to the S2 range in pure Mg[29].Therefore,based on the alloy system investigated,two solidification ranges(S1 and S2)and one liquid melt treatment(L)of UST was employed.Fig.1(a,b,c)shows the cooling curves indicating the UST range S1,S2 and L for pure Mg,Mg–6.0 Zn and Mg–Zr alloys.S1 and S2 UST were applied from above the melting or liquidus temperature at 100°C and 40°C superheat respectively until solidification was complete.The range S2P in Fig.1(b)refers to the condition where UST is applied over the range S2 using a sonotrode preheated at 600°C for 2h in a preheating furnace.Heating the sonotrode above or at this temperature for more than 2h was unsuccessful causing the device to fail which is a limitation of the current experimental setup.As the Mg–Zr alloy contains pro-peritectic Zr particles,an additional range L was investigated by applying UST from 750°C until 660°C(Fig.1(c)).The sonotrode was removed from the melt before the onset ofα–Mg nucleation to investigate the effect of UST on conditioning the Zr particles.After solidification with and without UST,the ingots(30mm bottom diameter,65mm top diameter and 70mm in height)were sliced vertically and 10×15mm samples were cut from the centre(middle region)of the casting.After standard metallographic polishing and etching with picric-acetic acid reagent[29],grain refinement before and after UST was analysed using a Leica Polyvar optical microscope(under polarised mode)and the grain size measurements were computed by the linear intercept method[26,29].A Hitachi TM303 Scanning Electron Microscope(SEM)was used to investigate the nucleant particles in the Mg–Zn and Mg–Zr alloys after UST at an operating voltage of 18keV.

Fig.1.Cooling curves of as-cast(a)pure Mg,(b)Mg–6.0 Zn alloy and(c)Mg–1.0 Zr alloy indicating the temperature range and time duration when is UST applied.The range S1 and S2 refer to UST application that includes the onset of nucleation.S2P refers to the UST applied using the sonotrode preheated to 600°C for 2h and L refers to liquid treatment without affecting the onset ofα-Mg nucleation.

Fig.2.Optical micrographs under polarised light of pure Mg(a to c)and Mg–6.0 Zn alloy(d to g)in as-cast(a,d)and after UST over the temperature range described in Table 1 as UST-S1(b,e),UST-S2(c,f)and UST-S2P(g).

3.Results

The microstructures of pure Mg and Mg–6.0 Zn alloys are shown in Fig.2(a–c)and Fig.2(d–g)respectively.The as-cast columnar grains in pure Mg(Fig.2(a))are refined after UST in both the S1(Fig.2(b))and S2 ranges(Fig.2(c)).Compared to the range S1,UST-S2 produces a more uniform grain morphology with a nearly 83% reduction in the grain size from the as-cast condition.Specifically,a 40°C superheat with UST applied for 3min produces better grain refinement than a higher superheat of 100°C with UST applied for a longer time of 4min.Fig.2(d)shows the grains of as-cast Mg–6.0 Zn alloy with well-developed dendritic arms containing secondary and tertiary branches.Excellent refinement is achieved in all the UST temperature ranges S1,S2 and S2P(Fig.2(e–g))and the grain morphology is nearly spherical without the development of any secondary dendritic branches.The grain size after UST in Mg–6.0 Zn alloys typically ranges from 120μm to 150μm.It is interesting to note that the application of UST using a preheated sonotrode(S2P)also results in significant refinement(Fig.2(g)).Unlike pure Mg,grain refinement in the Mg–6.0 Zn alloy is not dependent on the superheat(temperature range)of UST application.

Fig.3.Optical micrographs under polarised light of Mg–0.5 Zn(a,b)and Mg–3.0 Zn alloy(c,d)in as-cast(a,c)and after UST-S2(b,d).

It is found that the low temperature UST-S2 produces excellent refinement in pure Mg and Mg–6.0 Zn alloy(Fig.2(c and f)),therefore,UST-S2 was used for the investigation of the effect of Zn concentration.Fig.3 shows the microstructure refinement of 0.5,3.0% Zn in the as-cast condition and after UST-S2.The grain size in the as-cast condition is reduced with the addition of Zn(Fig.3(a and c))and becomes more dendritic in morphology for Zn>3.0wt.%.On the other hand,UST produces globular grains that are much finer and non-dendritic(Fig.3(b and d)).It should be noted that the average grain size decreases after UST with increasing Zn,however,the degree of decrease is lower with UST(22μm/wt.% of Zn,R2=0.99)compared to the as-cast condition(235μm/wt.% of Zn,R2=0.99).

Fig.4 shows the microstructure refinement obtained in Mg–Zr alloys after UST.Due to settling of Zr[26],(Fig.4(a,d,g))significant refinement in the as-cast condition is obtained only in the Mg–1.0 Zr alloy(Fig.4(g)).It has been found that the actual Zr content in the alloy is only 0.3%Zr for a nominal addition of 1.0% Zr from the master alloy[26].UST-L shows comparatively better refinement(Fig.4(b,e,h))than the as-cast condition and excellent refinement is realised when UST is applied below the liquidus temperature(UST-S2)in Fig 4.(c,f,i).It is interesting to note that grain refinement is significant when the Zr content exceeds the peritectic concentration of 0.5%,due to the increase in solute Zr,decreased settling and better distribution of Zr nucleating particles[26].This is the reason why most commercial Mg alloys contain 0.7% Zr[11].

Fig.4.Optical micrographs under polarised light of Mg–0.2 Zr(a to c),Mg–0.5 Zr(d to f)and Mg–1.0 Zr alloy(g to i)in as-cast(a,d,g),UST-L(b,e,h)and UST-S2(c,f,i)conditions.

The grain sizes of Mg–Zn and Mg–Zr alloys after UST are shown in Fig.5.As UST-L is terminated above the liquidus temperature in Mg–Zr alloys(Fig.5(a)),the grain refinement efficiency depends only on the distribution of the nucleating particles(average spacing between the potent particles)and their activation facilitated by solute Zr.Excellent refinement is achieved only at hyper-peritectic compositions of Zr>0.5%.On the other hand,the UST-S2 range produces additional refinement even at 0.2% Zr due to grain formation beneath the sonotrode during solidification(This will be detailed further in the discussion)[26–29].It is interesting to note that above 0.5% Zr,UST-L applied for 60 to 90 s brings excellent refinement(95μm)which is similar to the refinement obtained by UST-S2 for 3 min(76μm)during solidification.Regardless of the temperature range of UST(S2 or L-range),excellent refinement with the average grain size less than 100μm can be obtained for peritectic Mg–Zr alloys(for Zr>0.5%).

Fig.6.Grain size as a function of the superheat temperature at which UST was first applied.

Incremental refinement is observed with the addition of Zn in the as-cast condition in Fig.5(b).After UST-S2,the average grain size reduces from 250μm to 150μm which is considered to be good refinement when no external heterogeneous potent particles are added.Since it is well-documented in pure metals and eutectic alloys that the application of UST in the melt or terminating UST above the liquidus temperature produces insignificant refinement(<28% refinement max.)[29,38,39],UST-L was not investigated in Mg–Zn alloys in the present work.The grain size achieved in Mg–Zr alloys for Zr≥0.5% is much finer(<100μm)than for Mg–Zn alloys(Co≥3wt.%).

Fig.6 shows the grain refinement of pure Mg,Mg–6.0 Zn and Mg–1.0 Zr alloy as a function of superheat temperature.In pure Mg the grain size decreases when the temperature range is reduced close to the equilibrium temperature[29].This is because the grain size in the final microstructure of pure metals depends on the rate of grains produced and their successful survival.When Zn is added to Mg the grain refinement tendency is much less dependent on the temperature range of UST.Even with a preheated sonotrode,the grain refinement is almost the same as that of the sonotrode without preheating(Fig.2(g)).Addition of Zr particles increased the extent of refinement and it is evident that with adequate refiner additions(Zr>0.5wt.%),UST-L itself is sufficient to obtain excellent refinement(Fig.4(e)and(h)).The presence of Zn and Zr reduces the dependency of grain refinement on the temperature range over which UST is applied.

Fig.5.Grain size variation with alloying additions after UST for(a)Mg–Zr and(b)Mg–Zn alloys.

Fig.7.Grain size plotted against 1/Q in as-cast and UST-S2 conditions for(a)Mg–Zn and Mg–Zr alloys(represented by linear fitting)containing Mg alloys results(as data points)reported in the literature for similar casting conditions[25,40].An enlarged image at higher Q values in(b)shows the reduction in the nucleation free zone(vertical arrows)with respect to Zr addition and UST,assuming that a constant number density of particles(xsd)is available for nucleation.(c)Grain density measurements after UST-S2 reveals that a sharp increase for Mg–Zr alloy compared to that of the Mg–Zn alloy.(d)Grain refinement reported for Al and Mg based eutectic and peritectic alloys in which a significant refinement of grain size less than 200μm(dashed line)can be obtained at the UST-S2 temperature range[37].

The linear fit of grain sizedgsagainst 1/Qshown in Fig.7(a)provides useful information regarding the number of active substrates and the potency of refinement using a semi-empirical model[1,35,36]

In Eq.(1),the interceptaorxsdof the linear fit characterizes the distance between activatable particles.Decrease in the intercept values indicates that the number of active substrates are increased after Zr addition and UST.The slope(b)represents the refinement potency which is affected by the solute diffusion(D),nucleation undercooling(ΔTn)and growth velocity(v).The nucleation undercooling for a given particle size distribution in a constitutionally supercooled melt only needs a fraction(z)of undercoolingzΔTnto activate a new nucleation event[8].In the as-cast condition,the Mg–Zn alloy exhibits a steeper slope of 2842μm K(R2=0.69)due to the absence of potent nucleant particles and lowQvalues compared to Mg–Zr alloy which has a slope of 1008μm K(R2=0.99)containing Zr particles and a higherQvalue.It is interesting to note that after UST-S2 a similar degree of refinement(with non-dendritic grains)was obtained in both Mg–Zn and Mg–Zr alloys with slopes of 279μm K(R2=0.79)and 378μm K(R2=0.99)respectively.Fig 7(a)includes data from other Mg alloys including Mg-Al,Mg-Zn,Mg-Zn-Ca and Mg-Zn-Ca-Zr alloys at similar casting conditions[25,40].A magnified image is shown in Fig.7(b)to differentiate the grain refinement observed between Mg-Zn and Mg-Zr alloys at higherQvalues.The decrease in the intercept value from 915μm for as-cast Mg-Zn alloy to 51μm for Mg-Zr alloy after UST indicates that the number of active nucleants are increased in the alloy.Assuming that a constant nucleant particle density exists at higherQin both these alloys,the refinement produced by potent particles in peritectic Mg–Zr alloy is better than in Mg–Zn alloys.This can be understood by the decrease in the vertical distances marked as Nucleation Free Zone(NFZ)after Zr addition and UST respectively.The grain density measurements in Fig.7(c)shows a better view of the degree of refinement obtained after USTS2 when potent particles are present in the alloy.The grain density increases steeply after Zr addition and it is nearly double at 1.0% Zr compared to that of the 6.0% Zn alloy.Fig.7(d)compares the UST-S2 refinement tendency for selected Al and Mg alloys under comparable casting conditions.Similar to the present findings,both peritectic(Al3Ti1B added to Al alloys,Zr added to Mg alloys)and eutectic(Mg–Zn,Al–Cu)alloys produce excellent refinement after UST-S2[20,37,38,41,42].It should be noted that potent particles in peritectic alloys offer the best refinement,however,a significant refinement with average grain size less than 200μm can be readily obtained in eutectic alloys,which is possible only with the addition of grain refining additions in normal casting conditions.Fig.7(a)and(d)also indicate that the grain size for the UST-S2 condition is almost independent of the solute(Q)content of most alloys.

4.Discussion

In normal casting conditions under natural convection,nucleation on potent particles occurs when the amount of constitutional supercooling developed by the solute profile around an already nucleated grain reaches the nucleation undercooling of the available substrates.According to the Interdependence model[8]the grain size is

wherexcsis the amount of grain growth from the previous nucleation event andxdlis the diffusional distance ahead of the growing grain,that are both needed to establish adequate constitutional supercooling for triggering the next nucleation event.The sum of these two factors,where no nucleation is possible,is referred to as nucleation free zone(NFZ)[1,8],where

xsdis the additional distance to the next potent particle.Using this analysis,Fig.7(a and b)allows us to understand the relative effects of UST on the grain size of the alloys studied in this paper.

TheNFZis marked in Fig.7(b)after the addition of Zr and UST is represented asNFZZr as?castandNFZUST?S2(Zn/Zr)respectively.Compared to the Mg-Zn alloy,the addition of Zr to the Mg–Zr alloy in the as-cast condition reduces theNFZZr as?castby increasingQand the addition of Zr nucleant particles.There is a significant difference observed betweenNFZZn as?castandNFZZr as?castthat can be well understood and explained using Eqs.(2)and(3)[26,43].On the other hand,there is a less significant difference betweenNFZZn UST?S2andNFZZr UST?S2between these alloys.In order to explain the refinement by UST-S2,it is necessary to understand(i)what are the nucleation particles,(ii)how are they activated,and(iii)mechanisms of grain refinement in conventional and UST conditions.

Fig.8.(a and b)Schematic illustration of the grain structure in Mg–6.0 Zn and Mg–1.0 Zr alloys after UST-S2.Assuming that the grain centre particles were the actual nucleants the average distance between the nearest neighbours is measured as 130 and 100μm for Mg–6.0 Zn and Mg–1.0 Zr alloys respectively.(c)BSD image of a grain in the Mg–6.0 Zn alloy without any visible particles and(d)Mg–1.0 Zr alloy containing Zr particles(bright white particles distributed within the grain).

Let us consider Mg–6.0 Zn and Mg–1.0 Zr alloy(alloys with the largestQvalues)that have an average grain size of 120 and 76μm respectively after UST-S2.Fig.8(a and b)schematically shows a random location mapped from the microstructure images of the Mg–6.0 Zn Fig.3(f))and Mg–1.0 Zr alloy(Fig.4(i)).For simplification,it is assumed that nucleation has occurred on the grain centered particles(indicated by a dotted circle)in both these cases.According to Eqs.(2)and((3),the NFZ constitutes the growth of a grain after the previous nucleation event(small dotted circle in Fig.8(a)),the length of the diffusion zone and the spacing between the most potent particles to initiate further nucleation on an available particle(indicated by arrows to the nearest neighbors).The average length between the grain centered particles for Mg–6.0 Zn alloy and Mg–1.0 Zr alloy according to Fig.8(a and b)was measured as 134±27μm and 100±71μm respectively.Since solute with adequateQis available in both these alloys(Q=32Kfor 6wt.% Zn andQ=19Kfor 0.5wt.% solute Zr),the grain refinement in Mg–Zr alloys can be attributed to Zr particles,however,the nucleant particles are difficult to identify in Mg–6.0 Zn alloys.Fig.8(c and d)shows the backscattered image of a grain in the Mg–6.0 Zn alloy without any visible particles and Mg–1.0 Zr alloy containing potent Zr particles distributed within the grains(marked within the image)after UST-S2.A comparableNFZin the Mg–6.0 Zn alloy to that of Mg–1.0 Zr alloy is difficult to achieve with only impurity or impotent particles with a size distribution which is likely to be very fine.

Fig.9 shows the relationship between particle size(dp)andΔTncalculated for Mg alloys[44].Zr particles in the as-cast condition have been found to have a size distribution of 1.5–5μm,which is reduced to 0.5–2.5μm after UST[26].Typically larger particles in this size distribution are activated during UST-L and UST-S2 for excellent grain refinement Figs.4 and 8(d))[13].For impurity particles the size distribution is not well known,however,the commonly reported oxide particles(MgO particles in Mg alloys)have a diameter ranging from 0.02 to 0.07μm(which is much smaller than the Zr particles)with a few occasional larger particles(up to 6μm)[15].Generally,a highΔTngenerated by UST(for the oxide or impurity particles)is explained in terms of the Clapeyron criterion where the implosion of the bubble is assumed to activate nucleation by pressure pulses that lie in the diameter range of MgO particles(marked in Fig.9)[22,24,25].In recent work it was further clarified that an undercooled zone exists beneath the sonotrode(in an eutectic Al-2.0 Cu alloy)where the grains are formed and then dispersed to the other regions of the casting during USTS2[27].Fig.10(a)shows the macrostructure of the Al–2.0 Cu alloy with fine grains captured beneath the sonotrode by a gauze that prevents the transport of grains to other parts of the casting.Fig.10(b)shows the microstructure of the fine,non-dendritic grains extracted using a silica tube during UST.This indicates that the mechanism of grain formation by UST-S2 is different when compared to conventional castings similar to UST-L.Because in the as-cast or when the UST is turned off before solidification begins,nucleation is activated by constitutional supercooling described by the Interdependence model(Eqs.(2)and((3)).It is essential to understand these differences in the mechanism in order to explain the similar refinement tendency between the eutectic and peritectic alloys.

Fig.9.Nucleation undercooling vs.particle diameter calculated using the parameters reported for Mg alloys[44]for the size range of potent Zr[26]and impotent oxide particles[15].

Fig.10.(a)Macrostructure of Al–2.0 Cu alloy shows the formation of fine grains beneath the sonotrode(by arresting the motion of grains using a gauze)and(b)microstructure of a sample extracted from the melt during UST(S2 range)after 80 s,using a silica tube and quenching immediately[27].

Fig.11 explains the effect of an unpreheated sonotrode producing equiaxed grains(as shown in Fig.10)by an undercooling beneath the sonotrode in a Mg–6.0 Zn alloy.Fig.11(a)shows a schematic diagram of a two-thermocouple setup with one thermocouple attached to the sonotrode measuring the temperature close to the sonotrode(TSonotrode)and the second thermocouple measuring the instantaneous temperature of the melt(TMelt)which is placed adjacent to the mould wall.AfterTMeltreaches the required temperature,the vibrating sonotrode is immersed into the melt as illustrated in Fig.11(b).Fig.11(c and d)shows the cooling curves and the temperature gradient calculated by the difference betweenTMeltandTSonotrodeat 100°C superheat.After the crucible with melt was removed from the furnace and placed in air,the temperature was monitored by bothTSonotrodeandTmelt(Fig.11(a)).Placing the sonotrode above the melt surface(approximately 4–5cm height)warms the sonotrode and this temperature never exceeds 150°C(dotted line in Fig.11(c and d)).After immersion,the temperature of the sonotrode quickly heats to the melt temperature which is observed from the near-zero flat line of the temperature difference curve(ΔTMelt?Sonotrode).An enlarged image of Fig 11(c)is shown in Fig 11(d)at the onset of solidification.It should be noted that after immersing the sonotrode,the cooling rate ofTMeltis increased and at the same time theTSonotrodeincreases steadily to match theTmelttemperature.

A similar observation is noted in the 40°C superheat and the corresponding cooling curves are shown in Fig.11(e and f).At approximately 12s,bothTMeltandTSonotrodereach a steady-state temperature for both superheat immersion temperatures.It should be noted that the vibrating sonotrode just before establishing contact with the liquid melt is measured to be at a temperature not exceeding 150°C and the effect of the cold sonotrode is not realized by a large drop in temperature nearby the sonotrode due to the formation of a thick solidified layer of metal.This is due to the powerful acoustic streaming that stabilizes the melt temperature throughout the volume[31,37].Practically,the sonotrode at this stage would be well below the liquidus temperature of the melt by hundreds of degrees,meaning that the region beneath the sonotrode remains the coldest region of the casting compared to mould walls and grain formation is favored by a high thermal undercooling at both superheat temperatures.An approximate estimation of undercooling at the instant of sonotrode immersion(measured with respect toTmelt)shown in Fig.11(d and f)is an example of this situation.At high superheat UST,(UST-S1 in Fig.11(c and d)),theTSonotrodetemperature is similar to the melt temperature at the onset of solidification in which UST was applied almost 1.5min before solidification begins.

Fig.11.Schematics showing two thermocouples placed in the melt adjacent to the mould wall(TMelt)and attached to the sonotrode(TSonotrode)(a)before and(b)after inserting the sonotrode into the melt.Cooling curves of the Mg–6.0 Zn alloy and temperature difference(ΔT)curves during(c,d)UST-S1 and(e,f)UST-S2.(d,f)shows the enlarged views of the onset of solidification in(c,e)indicating that a significantly higher initial undercooling is provided by an unpreheated sonotrode.

For a low superheat UST(UST-S2),the onset of solidification occurs quickly after the immersion of the sonotrode(within 15s),indicating that a higher undercooling at the sonotrode surface exists for UST-S2 compared to UST-S1.This appears to be the main reason for the enhanced refinement when UST is applied at the low superheat temperatures(UST-S2,Fig.7).The reduction in undercooling due to high superheat UST-S1 could be the reason for the reduced number of grains observed in pure Mg(Fig.2(b)).Nevertheless,a similar grain size is produced irrespective of superheat or sonotrode preheating in a Mg–6.0 Zn alloy(Fig.2(f and g)).These observations indicate that the vibrating sonotrode generates substantial undercooling beneath it promoting the formation of non-dendritic grains(much larger than the undercooling described by the Clapeyron criterion)in eutectic alloys.To be noted here is the difference in the source of UST grain formation compared to conventional nucleation.Normally,in the as-cast condition under natural convection,solidification starts from the mould wall and progresses towards the thermal centre of the casting.The presence of solute activates nucleation ahead of the solid–liquid interface by constitutional supercooling.On the other hand,UST produces grains beneath the sonotrode which are dispersed by acoustic streaming to the other regions of the casting to generate an equiaxed grain structure.

Fig.12.Criteria to produce significant grain refinement in the UST-S2 range by solute addition without grain refiner and with grain refining particles.Significant refinement can be obtained without the addition of grain refining particles by an adequate amount of solute combined with a low superheat of UST.

Since the microstructure is uniform throughout the crosssection of the ingot,the average grain size after UST-S2(240μm,185μm and 120μm for 0.5% Zn,3.0% Zn and 6.0% Zn respectively)is assumed to be established by UST early in the solidification sequence.By using Table 1 for the UST-S2 time duration,the number of grains produced for 0.5% Zn,3.0% Zn and 6.0% Zn are estimated as 8.5,8.8 and 21.8 grains mm?2min?1respectively.This shows that the number density of grains produced by UST increases with solute content and when an adequate amount of solute is present,the diffusion field around the grain is more likely to assist grain survival without remelting or extensive growth leading to the finest grain structure[45].

Although the exact mechanism of non-dendritic grain formation at the sonotrode-liquid interface needs further investigation,the role of solute and potent particles during USTS2 are summarized in Fig.12.The grain size reduction obtained for Mg–0.5 Zn alloy without potent particles is much greater than Mg–0.2 Zr alloy where some grain refinement has been obtained with potentα–Zr particles.Incremental addition of Zn alone(0.5,3.0 and 6.0wt.%)decreases the grain size.Under the same solidification conditions,the Mg–Zr alloy(Zr≥0.5%)shows excellent refinement with the average grain size being less than 100μm,which is a 81%reduction from the UST refined pure Mg.The interesting observation to be noted is that when adequate solute is present in the alloy,a significant refinement of<200μm with a nondendritic microstructure can be readily obtained in eutectic alloys without the addition of grain refiner.Because,decreasing or eliminating Zr addition does not compromise the extent of grain refinement obtained from UST,this approach could be beneficial to reduce the cost of Zr containing Mg alloys(WE,ZE and ZK series alloys).For example,the effectiveness UST was demonstrated by the authors for the development of biodegradable Mg alloys in dilute Mg–0.5 Zn–0.5Ca alloys where eliminating the use of Zr produced a similar degree of grain refinement to that of the Mg–0.5 Zn–0.5Ca–0.2 Zr alloy[40].

5.Conclusions

The formation of a fine and non-dendritic grain structure in pure Mg during UST solidification using an unpreheated sonotrode depends on the superheat temperature.Applying UST at a lower superheat(40°C above melting temperature)until complete solidification produces significant refinement compared to a higher superheat(100°C above melting temperature).The formation of non-dendritic grains during UST can be explained by a higher thermal undercooling provided beneath the cold sonotrode leading to copious nucleation or grain formation which then disperses the grains into the melt.In conventional casting conditions,the addition of Zr in Mg–Zr alloys produces excellent refinement compared to the addition of Zn in Mg–Zn alloys due to the presence of potent Zr particles.However,when UST is applied at a low superheat significant refinement with an average grain size less than 200μm achieved in eutectic Mg–Zn alloys with an adequate amount of solute(>3.0wt.%Zn).This appears to be an excellent strategy for producing significant refinement in Mg alloys without or with a reduced addition of external potent particles.At a low superheat application,the dependency of UST refinement on potent particles and the type of solute(Zn or Zr)is reduced,although a solute with high Q and potent particles offer the best refinement.

Author contributions

N.B.designed and performed the experiments,analysed and characterized the results,designed the figures and drafted the manuscript.G.W.and M.D.supervise the findings of the work and contributed to the final version of the manuscript by evaluating the methods and procedures adopted.M.E.and D.S.critically reviewed the concepts of discussion,characterization,interpretation of the results and significantly contributed to improving the writing of the manuscript.All the authors contributed to the results and discussion section of the manuscript.

Acknowledgements

The authors acknowledge the funding support provided by Australian Research Council Research Hub for Advanced Manufacturing of Medical Devices IH150100024,the ARC Discovery grant DP140100702 and ARC linkage project LP150100950.The first author thanks the technical support of Tharmalingam Sivarupan during the casting process.