Revisiting the role of peritectics in grain refinement of Al alloys

Available online at www.sciencedirect.com

www.elsevier.com/locate/actamat

Acta Materialia 61 (2013) 360–370

Revisiting the role of peritectics in grain refinement of Al alloys

Feng Wang a, Zhilin Liu a, Dong Qiu a, John A. Taylor a,b, Mark A. Easton b,c,Ming-Xing Zhang a,⇑

a School of Mechanical and Mining Engineering, The University of Queensland, St. Lucia, QLD 4072, Australiab CAST Cooperative Research Centre, Australia

c School of Materials Science and Engineering, Monash University, Clayton, VIC 3800, Australia

Received 5 September 2012; received in revised form 29 September 2012; accepted 29 September 2012Available online 27 October 2012

Abstract

The grain refining effect of four peritectic-forming solutes (Ti, V, Zr and Nb) as well as three eutectic-forming solutes (Cu, Mg and Si)on pure Al was investigated. Significant grain refinement is observed by the addition of peritectic-forming solutes, whereas the additionof eutectic-forming solutes only slightly decreases the grain size. The mechanisms underlying the grain refinement of these alloys werethen studied by a new analytical methodology for assessing grain refinement that incorporates the effects of both alloy chemistry andnucleant potency. It is found that the low degree of grain refinement by the addition of eutectic-forming solutes is mainly attributedto the segregating power of solutes, i.e. the constitutional undercooling contribution. However, peritectic-forming solutes do not onlycause grain refinement by their segregation power but, more importantly, they introduce copious potent nuclei into the melt and promotesignificant grain refinement via heterogeneous nucleation.� 2012 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Al alloys; Grain refinement mechanism; Peritectics; Growth restriction factor; Nucleant particles

1. Introduction

Grain refinement of Al and its alloys has been an impor-tant process in commercial production for over half a cen-tury because fine grains not only deliver enhanced castingsoundness and mechanical properties of cast metals, butalso facilitate the subsequent mechanical working andforming processing [1,2]. In spite of over a half-centuryof scientific study, the details of the mechanisms of grainrefinement are still a matter of some controversy. Manytheories have been proposed to explain the phenomenaobserved in both practice and experiment [3–14]. However,none of them can fully account all the observations. Cur-rently, it is generally accepted that both the potency ofthe nucleant particles and the segregating power of the sol-ute are critical to effective grain refinement [12–16]. These

1359-6454/$36.00 � 2012 Acta Materialia Inc. Published by Elsevier Ltd. All

http://dx.doi.org/10.1016/j.actamat.2012.09.075

⇑ Corresponding author.E-mail address: [email protected] (M.-X. Zhang).

two factors are essentially based on two fundamental con-tributions, i.e. heterogeneous nucleation and solutesegregation.

Heterogeneous nucleation primarily focuses on the roleof potent nucleant particles in promoting nucleation ofgrains. It suggests that an increase in either nucleantpotency or number density of the nucleant particles consid-erably increases the nucleation rate, leading to effectivegrain refinement [3–11]. In contrast, the contribution ofsolute is related to the role of solute segregation in frontof the liquid–solid interface in restricting growth of grainsand developing constitutional undercooling providing adriving-force to activate further nucleation on the sub-strates present [12–16]. Although most observations ofgrain refinement have been successfully explained fromthese two aspects, there is still some confusion and issuesthat remain unsolved. One of the major problems is thata number of newly developed grain refiners based on thesetwo approaches do not work as effectively as expected,

rights reserved.


mailto:[email protected]


F. Wang et al. / Acta Materialia 61 (2013) 360–370 361

particularly for Mg–Al-based alloys [17–19] and Ti alloys[20,21]. Hence something further may be required for effec-tive grain refinement to be achieved.

It has long been realized that peritectics often effectivelyrefine the grains of a metal [22], e.g. Ti and Zr with Al [23],Zr with Mg [24], etc. Furthermore, the most successfulgrain refiners to date are Ti-based compounds (TiB2, TiAl3and TiC) in Al alloys and Zr in Al/Mn/Si-free Mg alloys.Although many investigators [5,6,23] have studied theeffect of peritectics on the grain refinement of Al and attrib-uted the consequent grain refinement to the peritectic reac-tion, it has been difficult to understand why peritecticelements can refine the grains of alloys at concentrationsbelow which the peritectic reaction occurs [25,26]. Evenat hyperperitectic concentrations, it has been suggestedthat the formation of the peritectic phase is the result ofdirect crystallization from the melt on the properitecticphase rather than a peritectic reaction [27,28] under normalsolidification conditions. However, using different experi-ment techniques, many investigators [10,11,29–31] haverecently obtained evidence that the properitectic TiAl3phase is formed on TiB2 prior to the nucleation of a-Al,indicating that a TiAl3 phase may be responsible for theenhanced grain refinement in this system.

For convenience, the process of direct crystallizationfrom a melt onto a properitectic phase is termed peritecticnucleation. In fact, how the peritectic phase forms, throughperitectic nucleation or peritectic reaction, would not sig-nificantly affect the grain refining efficiency. The only differ-ence between these two processes is that the peritecticreaction involves consumption of the properitectic phaseat slow cooling rate close to equilibrium, and in the peritec-tic nucleation process, the growth of the peritectic phaseonly occurs in the melt once it nucleates on the properitec-tic phase under faster cooling conditions. Both are hetero-geneous nucleation processes. However, the question stillremaining is: which factor, the heterogeneous nucleationor solute segregation contribution, governs the grain refine-ment efficiency? Therefore, it is intriguing to reinvestigatethe underlying grain refinement mechanisms operating inperitectic alloys, and the outcome may shed new light onthe factors affecting grain refinement.

One of the most important considerations is how to sep-arate the effects of nucleant particles and solute. Mostrecently, a new analytical model of grain refinement (i.e.the Q-model) has been developed by Easton and StJohn[32–34]. It has been found that the grain size, d, can beexpressed as [14,32,35,36]:

d ¼ 1ffiffiffiffiffiffiffiffiffiffiffiffif � qV

3p þ D � DT n

v � Q ; ð1Þ

where qV is the volumetric number density of nucleant par-ticles; f is the fraction of these particles that are activated;D is the diffusion coefficient; v is the growth velocity; DTn isthe undercooling required to active nucleation, and Q is thegrowth restriction factor, defined by Q = m(k � 1)�c0, in

which m is the slope of the liquidus; k is the partition coef-ficient; and c0 is the concentration of the solute. One of themajor advantages of this model lies in the simple represen-tation of the relationship between grain size and the reci-procal of growth restriction factor, Q, a measurable andreliable parameter based on factors that can be obtainedfrom the phase diagram. To investigate and compare thegrain refinement efficacy of alloys and master alloys [37],Eq. (1) is usually simplified by:

d ¼ aþ bQ; ð2Þ

where the slope, b, is related to the potency of the nucleantparticles, and the intercept, a, corresponds to the maximumnumber of activated nuclei [24,32,37]. As the potency of thenucleant particles increases, the slope, b, of the line de-creases; while as more nuclei are introduced, the term, a,decreases. It has been shown that this new analytical modelgives a reasonable description of the grain refinement phe-nomena observed across a range of Al alloys [32–34] and isalso applicable to studies of grain refinement of Mg alloys[24,37,38] and Ti alloys [20,39–41] over a wide range ofsolidification rates [34] and conditions [42].

As a result, this paper first focuses on studying the effectof a range of elements, including four peritectic-formingsolutes (Ti, V, Zr, Nb) and three eutectic-forming solutes(Cu, Mg, Si), on the grain size of Al. Using the new grainrefinement assessment approach (i.e. the Q-model), thegrain refinement due to the addition of these solutes is thenevaluated.

2. Experimental

High-purity commercial Al ingots (99.96% Al, 0.016%Si, 0.017% Fe, 0.0001% B and 0.0001% C; all the composi-tions hereinafter are in wt.% unless otherwise specified)were used as the base material in the present work. Com-mercial-purity Ti, V and Nb powders and Zr particles wereselected as peritectic-forming alloying elements. To com-pare the effect of peritectic-forming elements on grainrefinement with the eutectic-forming solutes in Al alloys,Cu, Si and Mg powders were also chosen as another testgroup of alloying elements. All the peritectic and eutecticsolute additions were added in the form of compacted pel-lets, which were made by mixing the metal powders withfreshly cut Al chips followed by compaction of the mixturein a die. Details of the solute addition levels and their cor-responding growth restriction factors are listed in Table 1.The levels of additions were determined by the value of thegrowth restriction factor, Q. To compare the influence ofQ, the actual addition levels of both peritectic-formingand eutectic-forming solutes were deliberately selected sothat the same values of Q are achieved. In the presentwork, four different Q levels, Q = 1, 2, 5 and 12, were used.

It should be noted that the Q values of various solutes inAl were calculated using the m(k � 1) values listed in thepaper of Easton and StJohn [12]. Furthermore, the Q val-

Table 1Addition levels (wt.%) and corresponding growth restriction factors of alloying elements added to aluminium.

Solute Group 1 (approx. Q = 12 k) Group 2 (approx. Q = 5 k) Group 3 (approx. Q = 2 k) Group 4 (approx. Q = 1 k)

Nominaladdition level

Growthrestrictionfactor







Ti 0.05 11.2 0.02 4.5 0.01 2.3 0.005 1.1V – – 0.15 4.5 0.07 2.1 0.04 1.0

0.40 5.10.60 5.11.00 5.1

Nb – – – – – – 0.12 0.80.20 1.00.50 1.0

Zr – – – – – – 0.10 0.70.20 0.70.30 0.70.50 0.7

Cu 4.50 12.7 1.60 4.5 – – 0.40 1.1Mg 4.00 12.2 1.50 4.6 – – 0.35 1.1Si – – 0.80 4.7 0.35 2.1 0.20 1.2

362 F. Wang et al. / Acta Materialia 61 (2013) 360–370

ues in peritectic systems were determined by assuming thatall of the excess solute above the maximum solubility, cm

(cm = 0.15% Ti, 0.17% V, 0.11% Zr, 0.15% Nb for relevantbinary Al systems [12]), is combined with Al to form prop-eritectic intermetallic particles. As a result, the c0 used inthe growth restriction factor equation is replaced by themaximum solute solubility cm. In other words, once theaddition level exceeds the cm, the growth restriction factoris assumed to be a constant value, Qmax = m(k � 1)cm.

All melts were prepared in clay-bonded graphite cruci-bles, which were coated with a wash of boron nitride.The crucibles with the weighed Al charges were placed ina resistance furnace and heated to 780 �C, at which pointthe appropriate alloying additions were made after skim-ming off the surface oxide skin. The melt was then isother-mally held for 20 min, a period determined by a simple trialexperiment as described below. After being held for 20 min,the melt was stirred immediately before dipping two graph-ite moulds, which were preheated to the same temperatureas the melt, into the melt. The moulds were 50 mm in diam-eter and 50 mm high with a wall thickness of 10 mm. Oneof the graphite mould samples was for thermal analysis,and the other was for metallography examination. Thetwo-thermocouple technique developed by Backerud andShao [7] was used for thermal analysis. The sample forthermal analysis in the graphite mould was placed on a10 mm thick piece of fibrefrax board, and then two thermo-couples were inserted into the melt, one close to the mouldwall and the other at the centre of the mould. Anotherpiece of fibrefrax was then placed on the top of the mouldso that a radial cooling rate of around 1 �C s�1 wasobtained. The sample for metallographic examinationwas cast in the same way but without inserting thermocou-ples. Each time before adding alloying elements, two pureAl samples were also cast as reference samples.

Metallographic samples were sectioned approximately10 mm from the base of the cast ingots. After mechanically

polishing, samples were anodized in a 0.5% HBF4 solutionfor about 2 min at 20 V. These samples were examined byoptical microscopy (OM) using polarized light and thegrain sizes were measured using a linear intercept technique(ASTM E112-10). Whenever possible, depending on thegrain size, at least five fields were examined and more than50 counts were obtained on each field. In order toobtain the detailed microstructure information, somesamples with pronounced grain refinement were furtherexamined by scanning electron microscopy (SEM; JEOL6460LA).

To ensure complete dissolution of the metal powdersadded into melts, appropriate holding times needed to bedetermined. Because of their high melting temperatures,the peritectic-forming solutes at the highest additions(0.05% Ti, 1.00% V, 0.50% Zr and 0.50% Nb) were usedto select the experimental holding time. After adding thecompacted pellets, the melts were held at 780 �C for differ-ent times ranging from 5 to 60 min. Then, the as-cast sam-ples were examined using both OM and SEM. The detailsof casting, sampling and examination methods are thesame as those described above. Fig. 1 shows the variationsof grain size of the four alloys (i.e. Al–0.05Ti, Al–1.0 V,Al–0.5Zr and Al–0.5Nb) with holding time. Although thegrain size initially decreases with an increase in the holdingtime, constant grain size is achieved once the holding timereaches �20 min. To further confirm that 20 min is longenough for complete dissolution of particles, the detailedmicrostructure of alloys after different holding times wasexamined. Fig. 1 also shows three typical SEM micro-graphs of the representative Al–0.5% Nb alloy after hold-ing for 5, 10 and 20 min. Although some undissolved Nbmetal powder particles can be observed in the alloy samplesafter holding for 5 and 10 min, as shown in Fig. 1a and b,no trace of Nb powder could be detected in the samplescast after holding for 20 min as shown in Fig. 1c. This indi-cates that the effect of incomplete dissolution of particles

Fig. 1. Variations of average grain size of different Al alloys containing 0.05% Ti, 1.00% V, 0.50% V and 0.50% Nb with isothermal holding time at 780 �Cbefore casting, and some typical SEM back-scattered electron micrographs as well as EDX spectra of the Al–0.5 wt.% Nb alloys. (a) SEM micrographcorresponding to point (a) (5 min holding) showing the residual large pure Nb particle; (b) SEM micrograph corresponding to point (b) (10 min holding)showing the presence of a small pure Nb particle; (c) SEM micrograph corresponding to point (c) (20 min holding) showing complete dissolution of the Nbpowder particles and the presence of only properitectic Al3Nb particles.

Table 2Targeted and measured chemical compositions of selected samples (allcompositions are in weight percent).

Sample no. Nominal addition Chemical analysis

Sample 1 0.07% V 0.084% VSample 2 0.15% V 0.14% VSample 3 0.40% V 0.36% VSample 3 0.60% V 0.58% VSample 4 0.10% Zr 0.10% ZrSample 5 0.20% Zr 0.24% ZrSample 6 0.30% Zr 0.28% ZrSample 7 0.50% Zr 0.52% ZrSample 8 0.005% Ti 0.007% TiSample 9 0.010% Ti 0.011% TiSample 10 0.020% Ti 0.021% TiSample 11 0.050% Ti 0.060% TiSample 12 0.12% Nb 0.10% NbSample 13 0.20% Nb 0.18% NbSample 14 0.50% Nb 0.43% NbSample 15 0.40% Cu 0.38% CuSample 16 0.35% Mg 0.30% MgSample 17 0.20% Si 0.22% Si


can be effectively eliminated in the present work throughthe use of a holding time of 20 min.

To determine the recovery efficiency of the alloyingadditions to the experimental melts, chemical analysiswas done on selected samples using inductively coupledplasma-atomic emission spectroscopy (ICP-AES). Thenominal chemical compositions and the measured valuesare listed in Table 2. It can be seen that the method for add-ing alloying elements is reasonably effective.

3. Results and discussion

3.1. Effects of peritectic forming elements and eutectic

forming elements on microstructures of cast Al alloys

Fig. 2 shows typical micrographs of Al alloys with theaddition of Mg solute. It is observed that the grain sizeremains almost unchanged relative to that of pure Al afterthe addition of increasing amounts of Mg solute. The addi-tions decrease the grain size slightly, but still within therange of 900–1100 lm and columnar grains are typical.The effect of increased solute is more evident in that it leadsto the grains becoming dendritic, with eutectic becomingevident within the interdendritic spaces (see Fig 2d in par-ticular). The other eutectic-forming solutes show similareffects. For the purpose of brevity, the microstructure ofevery element addition is not shown. In contrast, additions

of peritectic-forming elements not only lead to grain refine-ment, but also induce equiaxed grains. Figs. 3 and 4 showtypical micrographs of Al–Zr and Al–Nb alloy sampleswhere an evident reduction in grain size can be observedas well as an accompanying change from columnar torefined equiaxed grains.

Fig. 2. Typical micrographs of as-cast Al alloys with increasing contents of Mg solute: (a) pure Al; (b) 0.35 wt.% Mg; (c) 1.5 wt.% Mg; (d) 4.0 wt.% Mg.

Fig. 3. Typical micrographs of as-cast Al alloys with increasing contents of Zr solute: (a) pure Al; (b) 0.1 wt.% Zr; (c) 0.2 wt.% Zr; (d) 0.3 wt.% Zr.


The variations of grain size of cast Al alloys with theamount of alloying elements added are shown in Fig. 5.From Fig. 5a, it can be seen that the grain size decreasesslightly with an increase in eutectic-forming solutes (Cu,Mg and Si), even up to a Q value of 12. The Q value is afunction of the solute content when the addition is belowthe maximum solubility, cm. This result is in evident con-trast to the results reported in the literature [32,43,44],where a marked grain refinement effect was observed withsimilar additions of these solutes. The apparent discrep-ancy in grain refining effect will be discussed below.

Fig. 5b shows the variation of grain size with the addi-tion of peritectic-forming solutes. Unlike the effect of eutec-

tic-forming elements, these solutes markedly reduce thegrain size; the addition of Ti in particular. Adding 0.05%Ti, which corresponds to a Q value of approximately 12,the grain size dramatically reduces from around 1250 to280 lm. Further analysis of the grain size data shown inFig. 5b reveals that the reduction in grain size on the initialaddition of peritectic-forming solutes (except Ti) is rela-tively small. However, after the solute content exceedsthe maximum solubility, cm, there is a pronounced decreasein grain size. For instance, in the Al–Zr alloy, an initialaddition of 0.1% Zr only decreases the grain size by around50 lm and the morphology remains columnar, while a fur-ther addition of 0.2% Zr exceeding the maximum solubility

Fig. 4. Typical micrographs of as-cast Al alloys with increasing contents of Nb solute: (a) pure Al; (b) 0.12 wt.% Nb; (c) 0.2 wt.% Nb; (d) 0.5 wt.% Nb.

Fig. 5. Plots of measured grain size vs. solute content of different elements: (a) eutectic-forming elements (Cu, Mg, Si); (b) peritectic-forming elements (V,Zr, Nb, Ti).

1 Since further additions of Nb and Zr cannot achieve higher Q valuesthan 1 due to the limitation of their solubilities in Al, there are no Al–Zrand Al–Nb samples in the Q ffi 2 group.


(cZrm ¼ 0:11%) reduces the grain size by approximately

500 lm.

3.2. The effect of Q value on grain refinement in various Al

binary alloys

Fig. 6 shows the grain size variation with respect to theQ value as listed in Table 1. As previously reported [11–15],it is observed that the grain size generally decreases as Qincreases. However, at each Q level, the grain size also var-ies with different solutes. As shown in Fig. 6a, at Q ffi 1, theeffect of eutectic-forming solutes, Cu, Mg and Si, on grainsize is marginal. Because the additions of 0.04% V, 0.005%Ti, 0.12% Nb and 0.10% Zr are below their respective max-imum solubilities in Al, their grain refining efficiency at thisQ level is similar to that of the eutectic-forming solutes.However, once the additions (0.2–0.5% Nb and 0.2–0.5%Zr) exceed the maximum solubilities (cNb

m ¼ 0:15% andcZr

m ¼ 0:11%), significant grain size reduction can be

observed, even though the Q value remains constant. AtQ ffi 2,1 illustrated in Fig. 6b, as the additions of 0.07% Vand 0.01% Ti are still below their maximum solubilities,their grain refining efficiency is similar to that of the eutec-tic solute, Si. At Q ffi 5, although the grain size is notreduced much by the additions of Cu, Mg and Si, oncethe addition of V exceeds its maximum solubility(cV

m ¼ 0:17%) in Al, effective grain refinement is observed,as shown in Fig. 6c. Furthermore, adding 0.02% Ti alsonoticeably reduces the grain size of cast Al at this Q value.Increasing the Q value to 12, only Cu, Mg and Ti solutescan achieve the desired values of the growth restriction fac-tor. From Fig. 6d, one can find that, for the alloys withadditions of eutectic-forming solutes (i.e. Cu and Mg),the grain sizes are still around 900 lm and the grain mor-

Fig. 6. Average grain sizes of various Al alloys with growth restriction factor, Q: (a) approx. Q = 1 K; (b) approx. Q = 2 K; (c) approx. Q = 5 K; (d)approx. Q = 12 K.

Fig. 7. The grain size of binary Al alloys with Cu, Mg, Si, Ti, V, Nb andZr additions plotted vs. the reciprocal of growth restriction factor, 1/Q.


phology remains columnar as shown in Fig. 1. However,with the addition of 0.05% Ti, the grain size is significantlyreduced to 300 lm and fine equiaxed grains are alsoobtained.

These results indicate that the addition of eutectic-form-ing solutes marginally reduces the grain size even at high Q

values. The peritectic-forming solutes have a similar grainrefining efficiency to the eutectic elements when the addi-tion is below the maximum solubility. However, onceenough peritectic solutes were added such that the properi-tectic phases could be formed, the grain size reduces sub-stantially. Therefore, it appears that the formation ofproperitectic phase is closely related to the grain refine-ment. However, Ti additions reduce the grain size substan-tially even below its maximum solubility in Al. The reasonfor the positive effect of Ti on grain refinement does notappear to be solely due to its high Q value. Some discussionon this topic is provided in below.

To further elucidate the effect of Q on grain refinementof Al alloys, the measured grain size data in the currentwork was replotted against the inverse of growth restrictionfactor (1/Q) in Fig. 7. For reasons of clarity, the grain sizeerror bars are not shown in the figure. For quick reference,the corresponding error bars can be found in Figs. 5 and 6.As can be seen from Fig. 7, there is a very strong lineartrend in grain size vs. 1/Q for the alloys with additions ofeutectic-forming solutes (i.e. Cu, Mg and Si). Furthermore,the values of intercept, a, and slope, b, of the fitting lines

for these alloys as listed in Table 3 are very close. Accord-ing to a recent model [24,32,34,37], this suggests that thecharacteristics of the nucleant population are almost con-stant for each base alloy and remain unchanged by theaddition of eutectic-forming solutes. This indicates thatthe additions of these eutectic-forming solutes do not intro-duce a significant number of nucleant particles into themelt and that the eutectic-forming solutes cause grainrefinement primarily through the contribution of solutesegregation. The discrepancy that the grain refinementobtained by the addition of similar amounts of the same

Table 3The values of intercept a and gradient b of the fitting lines in Fig. 6.

Element Intercept a (lm) Gradient b (lm K) R2

Cu 926 232 0.83Mg 939 251 0.94Si 921 208 0.76V (below cm = 0.17% V) 941 166 0.94Ti (below 0.01% Ti) 801 413 –Ti (above 0.01% Ti) 180 1925 0.89


solutes in the literature [32,43–45] is more pronounced thanthat obtained in the current study may result from variouscauses. The base alloys used in the previous work may con-tain larger populations of nuclei since they contain moreimpurities, a potential source of natural nucleants. It mayalternatively be due to the influence of cooling rate. It iswell known that, apart from the presence of nucleantsand solute, cooling rate also plays a significant role in grainrefinement. The cooling rate in some previous work [43–45](using the Alcan test) is usually greater than that used inthe present study (1 �C s�1).

Although trend-lines cannot be fitted through the grainsize data for the alloys with additions of Zr and Nb, usefulinformation on grain refinement can still be obtained fromFig. 7. It is clear that increasing the concentration above

Fig. 8. SEM images and EDX results of selected grain-refined samples: (a) Bsample; (b) a typical EDX spectrum on the particle in (a); (c) BSE image shotypical EDX spectrum of the particle in (c).

the maximum solubility leads to a substantial reductionin grain size without increasing the Q value. This indicatesthat the grain size reduction may primarily be due to theintroduction of copious nucleant particles. It is alsoobserved that the values of a and b for the alloys with addi-tions of V at additions below the maximum solubility arevery similar to the values for the alloys with additions ofeutectic-forming solutes. This indicates that V refines grainsize in a similar manner to eutectic-forming solutes whenthe V addition is not high enough for the formation ofpro-peritectic particles. However, as seen in Fig. 7, thereis a large departure from the linear fit when the contentexceeds the maximum solubility. This suggests that, at con-tents above maximum solubility, V also has increased thepopulation of active nuclei in the melt and caused addi-tional grain refinement through heterogeneous nucleation.

In order to identify the introduced nuclei that areresponsible for the grain refinement, the Al–Zr and Al–Valloy samples displaying fine grains were selected andexamined using SEM. Fig. 8a and c are two typical SEMmicrographs, showing intermetallic particles clearlyobserved at/near the grain centres, as indicated by thearrows. Based on the corresponding energy-dispersive X-ray (EDX) spectra in Fig. 8b and d, it is found that the par-ticles in the Al–Zr sample are most likely the Al3Zr phase

SE image showing a particle at the grain centre of Al–0.2 wt.% Zr alloywing a particle near the grain centre of Al–0.4 wt.% V alloy sample; (d) a


and that the ones in the Al–V sample are the Al10V phase.According to the Al–Zr and Al–V phase diagrams, thesetwo phases are the properitectic phases.

To further confirm the presence of the properitectic par-ticles prior to the formation of Al, thermal analysis wascarried out during the solidification of the alloys with addi-tions of peritectic-forming alloying elements. Three typicalcooling curves and cooling rate curves of Al alloys with dif-ferent V additions are shown in Fig. 9. Although there is nosignificant difference in the cooling curve between thesethree alloys, the cooling rate curve of the Al–0.4 V alloydoes differ from those of the pure Al and Al–0.15 V alloy.There is a prominent peak prior to the nucleation of Al onthe dT/dt curve of the Al–0.4 V alloy as shown in Fig. 9c.However, such a peak cannot be observed on the coolingrate curves of the other two alloys, as shown in Fig. 9aand b. This evident peak on the cooling rate curve of Al–0.4 V alloy is most likely the result of a phase transforma-tion occurring prior to the nucleation of Al. The latent heatreleased during the phase transformation leads to changeof the cooling rate. The temperature corresponding to thispeak is about 671.41 �C. According to the Al–V phase dia-gram, the temperature is consistent with the formation ofproperitectic Al10V particles as confirmed by the SEMobservation and EDX analysis in Fig. 8c and d. Similarphenomena have also been observed in the Al–Zr andAl–Nb alloy systems. When the addition of the peritectic-forming elements is greater than the maximum solubility,an extra peak is observed on the cooling rate curve, indicat-ing the formation of the properitectic phase. Based on theabove results, it appears that the nuclei that are responsiblefor the grain refinement by addition of V, Nb and Zr overtheir maximum solubilities in Al are very likely to be thein situ formed properitectic particles. Although the factorsthat control the distribution of the properitectic particlesare unknown and further studies are required, increasingthe amount of additions might lead to a higher numberdensity of the properitectic particles newly formed in themelt, which promotes a greater heterogeneous nucleationrate. Hence, smaller grains are obtained, as shown inFig. 6, even though the Q value remains constant whenthe addition is over the maximum solubility.

From Fig. 7 it can be seen that the grain size data for thealloys with Ti contents of 0.01% and 0.005% is similar to

Fig. 9. Cooling curves and cooling rate curves of Al alloys with diff

that of Al alloys containing eutectic-forming solutes, eventhough the two experimental data points are insufficientto obtain a reliable fitting line. This implies that, below acertain solute content, the grain refinement produced bythe addition of Ti is essentially due to its solute segregationpower. However, as the content of Ti is increased to 0.02%and 0.05%, a disproportionally large decrease in grain sizeis observed. Note: since these Ti addition levels are still wellbelow the maximum solubility, thermal analysis was car-ried out to try and understand the underlying mechanism;however, no differences were observed between the curves(cooling and cooling rate) for Al plus 0.05% Ti and thoseof pure Al. After further investigating Fig. 7, it is observedthat the fitting line of the alloys at high Ti additions (0.02%and 0.05%) shows a surprisingly steep gradient, whichreaches a value of 1925 lm K. However, the interpretationof the grain refinement mechanism directly from the steepgradient of the fitting line can be misleading because thevalue of the gradient is overestimated if nucleant particlesare added in association with the solute [24]. The majorreason for overestimation may be that, of the seven solutesinvestigated in the present work, Ti is the most chemicallyreactive one. Therefore there is high potential that additionof Ti may result in the reaction of Ti with trace carbon,boron or oxygen in the melts [3,4,46]. It is known that a lit-tle loss of solute Ti will result in a significant decrease in thevalue of Q as Ti has an extremely large growth restrictingpower in Al. This can be inferred from the disproportion-ally large value of parameter m(k � 1) for Ti in Al whichis approximately 245, compared with the values of V andZr which are only 30 and 6.8, respectively [12]. Hence,the 1/Q value of each plotted point needs to be correctedby moving it horizontally to the right to reflect a lower Q

value than that calculated using the total Ti content. Thecorrected fitting line is schematically shown in Fig. 7. Fur-thermore, it is well documented that some of the formedcompounds have good grain refining potency (e.g. TiC[46–48] and TiB2 [1,3,4]) while others have very lowpotency (e.g. TiO2 [1,2]). As a result, each further additionof Ti means that more nuclei are introduced into the melt.In order to eliminate the influence of the increased numberof nucleant particles, the two points of larger Q values needto not only move horizontally to the right to correct for thelower Q values, but must also move vertically upward to

erent V additions: (a) pure Al; (b) 0.15 wt.% V; (c) 0.40 wt.% V.


reduce the number of particles to equal that present in thealloys with smaller Q values. This is also qualitativelyshown in Fig. 7. Thus, the real value of the fittedslope for Ti additions above 0.02% is obtained and is lowerthan that obtained by directly fitting the plotted datapoints.

Despite the correction, it is noted that the intercept ofthe corrected fitting line at high Ti additions is still smallerthan that of the fitting line at lower Ti additions. Thisimplies that the underlying reason for the dramatic grainrefinement is likely to be due to the introduction of nucleiinto the melt in conjunction with the solute. Some investi-gators [2,49] have tried to determine the exact identity ofthe nuclei at low additions of Ti in Al. However, few directobservations have been reported in the literature due to thevery small size and low particle density of the nuclei. Thus,the actual cause of the high grain refining efficiency by add-ing small amounts of Ti in Al is still unknown. The authorsplan to try and identify the nuclei that are responsible forthe grain refinement using advanced technologies, includ-ing synchrotron and atom probe.

As a result of the Q-model analysis, it is found that oneof the major mechanisms operative in the considerablegrain refinement is the increased population of activatednuclei which promotes grain refinement via enhancedheterogeneous nucleation. Addition of peritectic-formingsolutes (V, Nb, Zr and Ti) is an effective approach to intro-duce active heterogeneous nucleant particles throughin situ formation in the melt. It is well documented in theliterature [14,48,50–52] that the number density, size distri-bution and crystallographic features of the nucleant parti-cles are critical factors that control the grain refinementefficiency. It has been demonstrated that an increase ineither the number density or the crystallographic matchingof nucleant particles will promote grain refinement. Theproperitectic particles formed most likely fulfil suchrequirements as effective heterogeneous nucleation sites.Therefore, in addition to the alloying elements that resultin very high growth restricting effect (i.e. high Q value),adding peritectic-forming solutes is another effectiveapproach to refine the grains of cast metals. Further inves-tigation into the effect of peritectics in other alloying sys-tems is currently being undertaken.

4. Conclusions

The grain refining effect of a range of peritectic-formingsolutes (Ti, V, Zr and Nb) and eutectic-forming solutes(Cu, Mg and Si) on Al has been studied. It is shown thatthe addition of eutectic-forming solutes only slightlydecreases the grain size of Al while remarkable grain refine-ment is observed by the addition of peritectic-formingsolutes.

Using a new analytical methodology for grain refine-ment assessment, i.e. the Q-model, the mechanisms under-lying the grain refinement were investigated. It is concludedthat:

1. The small degree of grain refinement by additions ofeutectic-forming elements is mainly attributed to theirsegregating power. To produce effective grain refine-ment, however, further addition of nucleant particles isrequired for these alloys.

2. The significant grain refinement achieved by additions ofZr, Nb and V is not only due to their growth restrictioneffect but additionally is a result of the in situ formationof pro-peritectic particles at concentrations above theirmaximum solubilities.

3. Extremely large reduction in grain size is observed in thealloys with additions of Ti at concentrations far belowits maximum solubility. It is considered that Ti signifi-cantly refines the grain size not only through the contri-bution from its constitutional undercooling but alsofrom the introduction of potent nuclei formed throughthe reaction of Ti with impurities.

4. The experimental results in the present work indicatethat the introduction of nucleants associated with theperitectic elements plays a key role in the grain refine-ment of Al alloys.

Acknowledgements

The authors are very grateful to the Australian ResearchCouncil for funding support. F.W. would also like toacknowledge the support of the China ScholarshipCouncil.

References

[1] Murty BS, Kori SA, Chakraborty M. Int Mater Rev 2002;47:3.[2] McCartney DG. Int Mater Rev 1989;34:247.[3] Cibula A. J Inst Met 1949;76:321.[4] Cibula A. J Inst Met 1951;80:1.[5] Marcantonio JA, Mondolfo LF. J Inst Met 1970;98:23.[6] Marcantonio JA, Mondolfo LF. Metall Trans 1971;2:465.[7] Backerud L, Shao Y. Aluminium (Germany) 1991;67:780.[8] Jones GP. United Kingdom; 1983. p. 40.[9] Vader M, Noordegraaf J. In: Proceedings of the 119th TMS annual

meeting, February 18, 1990–February 22, 1990. Anaheim, CA,Materials Park, OH: TMS; 1990. p. 851.

[10] Mohanty PS, Gruzleski JE. Acta Metall Mater 1995;43:2001.[11] Mohanty PS, Gruzleski JE. Acta Mater 1996;44:3749.[12] Easton MA, St. John DH. Metall Mater Trans A 1999;30:1613.[13] Easton MA, St. John DH. Metall Mater Trans A 1999;30:1625.[14] St. John DH, Qian M, Easton MA, Cao P. Acta Mater 2011;59:4907.[15] Johnsson M, Backerud L, Sigworth GK. Metall Trans A

1993;24A:481.[16] Greer AL, Cooper PS, Meredith MW, Schneider W, Schumacher P,

Spittle JA, et al. Adv Eng Mater 2003;5:81.[17] Bamberger M. Mater Sci Technol 2001;17:15.[18] Fu HM, Qiu D, Zhang MX, Wang H, Kelly PM, Taylor JA. J Alloys

Comp 2008;456:390.[19] Fu HM, Zhang MX, Qiu D, Kelly PM, Taylor JA. J Alloys Comp

2009;478:809.[20] Bermingham MJ, McDonald SD, Dargusch MS, St. John DH.

Scripta Mater 2008;58:1050.[21] Tamirisakandala S, Bhat RB, Tiley JS, Miracle DB. Scripta Mater

2005;53:1421.[22] Eborall MD. J Inst Met 1949;76:295.


[23] Crossley FA, Mondolfo LF. Trans Am Inst Min Metall Eng1951;191:1143.

[24] St. John DH, Ma Q, Easton MA, Peng C, Hildebrand Z. MetallMater Trans A 2005;36A:1669.

[25] Jones GP, Pearson J. Metall Trans B 1976;7:223.[26] Sigworth GK. Metall Trans A 1984;15A:277.[27] St. John DH, Hogan LM. J Mater Sci 1982;17:2413.[28] St. John DH. Acta Metall Mater 1990;38:631.[29] Iqbal N, van Dijk NH, Offerman SE, Geerlofs N, Moret MP,

Katgerman L, et al. Mater Sci Eng A 2006;416:18.[30] Schumacher P, Greer AL. NATO advanced research workshop on

undercooled metallic melts: properties, solidification and metastablephases, 6–12 June 1993, vol. A178; 1994. p. 309.

[31] Schumacher P, Greer AL. In: Proceedings of the 124th TMS annualmeeting, February 12, 1995–February 16, 1995. Las Vegas, NV,Materials Park, OH: TMS; 1995. p. 869.

[32] Easton MA, St. John DH. Metall Mater Trans A 2005;36A:1911.[33] Easton MA, St. John DH. Acta Mater 2001;49:1867.[34] Easton MA, St. John DH. Mater Sci Eng A 2008;486:8.[35] Sun M, Easton MA, St. John DH, Wu G, Abbott TB, Ding W.

Metall Mater Trans A, in press.[36] Qian M, Cao P, Easton MA, McDonald SD, St. John DH. Acta

Mater 2010;58:3262.[37] St. John DH, Cao P, Qian M, Easton MA. Adv Eng Mater

2007;9:739.

[38] Cao P, Qian M, St. John DH. Scripta Mater 2007;56:633.[39] Bermingham MJ, McDonald SD, Dargusch MS, St. John DH. J

Mater Res 2008;23:97.[40] Bermingham MJ, McDonald SD, St. John DH, Dargusch MS. J

Alloys Comp 2009;481:L20.[41] Bermingham MJ, McDonald SD, St. John DH, Dargusch MS. J

Mater Res 2009;24:1529.[42] Nagasekhar AV, Easton MA, Caceres CH. Adv Eng Mater

2009;11:912.[43] Jie W, Reif W. Z Metallkde 1993;84:445.[44] Spittle JA, Sadli S. Mater Sci Technol 1995;11:533.[45] Lee YC, Dahle AK, St. John DH, Hutt JEC. Mater Sci Eng A

1999;A259:43.[46] Banerji A, Reif W. Metall Trans A 1986;17:2127.[47] Banerji A, Reif W. Metall Trans A 1985;16:2065.[48] Zhang MX, Kelly PM, Easton MA, Taylor JA. Acta Mater

2005;53:1427.[49] Quested TE. Mater Sci Technol 2004;20:1357.[50] Quested TE, Greer AL. Acta Mater 2004;52:3859.[51] Greer AL, Bunn AM, Tronche A, Evans PV, Bristow DJ. Acta Mater

2000;48:2823.[52] Zhang MX, Kelly PM, Qian M, Taylor JA. Acta Mater 2005;53:3261.

Revisiting the role of peritectics in grain refinement of Al alloys

Documents

Transcript of Revisiting the role of peritectics in grain refinement of Al alloys