the grain refinement mechanism of cast aluminium by zirconium

10
The grain refinement mechanism of cast aluminium by zirconium Feng Wang a , Dong Qiu a , Zhi-Lin Liu a , John A. Taylor a , Mark A. Easton b , Ming-Xing Zhang a,a School of Mechanical and Mining Engineering, The University of Queensland, St. Lucia, QLD 4072, Australia b School of Physics and Materials Engineering, Monash University, Clayton, VIC 3800, Australia Received 14 March 2013; received in revised form 30 May 2013; accepted 30 May 2013 Abstract The mechanism underlying the grain refinement of cast aluminium by zirconium has been studied through examination of a range of Al alloys with increasing Zr contents. Pro-peritectic Al 3 Zr particles are reproducibly identified at or near the grain centres in grain-refined alloy samples based on the observations of optical microscopy, scanning electron microscopy and X-ray diffraction. From the crystal- lographic study using the edge-to-edge matching model, electron backscatter diffraction and transmission electron microscopy, it is sub- stantiated that the Al 3 Zr particles are highly potent nucleants for Al. In addition, the effects of Al 3 Zr particle size and distribution on grain refinement has also been investigated. It has been found that the active Al 3 Zr particles are bigger than previously reported other types of active particles, such as TiB 2 for heterogeneous nucleation in Al alloys. Considering the low growth restriction effect of Zr in Al (the maximum Q-value of Zr in Al is 1.0 K), it is suggested that the significant grain refinement of Al resulting from the addition of Zr can be mainly attributed to the heterogeneous nucleation facilitated by the in situ formed Al 3 Zr particles. Ó 2013 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Cast aluminium; Grain refinement; Peritectic; Edge-to-edge matching model 1. Introduction The mechanism of grain refinement through inoculation in aluminium alloys has been studied for over half a cen- tury [1–4]. Although many theories/models [5–13] have been proposed, none of them can fully elucidate all of the observations from experiment and practice. In general, it is now accepted that the presence of both potent nucleant particles and sufficient solutes is essential for effective grain refinement [2,14–17]. Despite the common recognition of these two essentials, the details of the grain refinement mechanism(s) are still of some ambiguity. One of the major problems is to determine the exact factors that control the efficiency of grain refinement. For example, the recent investigations [18–22] on a number of newly developed grain refiners for Mg–Al-based alloys and Ti alloys showed relatively low efficiencies compared to that expected from the crystallographic matching, which represents the potency of heterogeneous nucleation and the constitutional undercooling contribution of solute elements. Therefore, it appears that something important is still missing in the cur- rent understanding of grain refinement. It is worth noting that the spotlight of previous studies on grain refinement mechanism of Al alloys was primarily focused on the Al– Ti–B and Al–Ti–C systems as they are the most common grain refiners used in the foundry [3,4]. However, it is believed that the study of grain refinement resulting from other solute elements, which also produce effective grain refinement in Al may provide fresh insight into the factors that control the grain refining efficiency. It has been long realized that peritectic systems are often associated with effective grain refinement of the parent metal, such as Al–Ti, Mg–Zr and Mg–Y–Al alloys [23–26]. However, the peritectic approach has been consid- ered as incorrect because addition of far less Ti than its 1359-6454/$36.00 Ó 2013 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.actamat.2013.05.044 Corresponding author. E-mail address: [email protected] (M.-X. Zhang). www.elsevier.com/locate/actamat Available online at www.sciencedirect.com Acta Materialia xxx (2013) xxx–xxx Please cite this article in press as: Wang F et al. The grain refinement mechanism of cast aluminium by zirconium. Acta Mater (2013), http://dx.doi.org/10.1016/j.actamat.2013.05.044

Upload: independent

Post on 30-Apr-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

Available online at www.sciencedirect.com

www.elsevier.com/locate/actamat

Acta Materialia xxx (2013) xxx–xxx

The grain refinement mechanism of cast aluminium by zirconium

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

a School of Mechanical and Mining Engineering, The University of Queensland, St. Lucia, QLD 4072, Australiab School of Physics and Materials Engineering, Monash University, Clayton, VIC 3800, Australia

Received 14 March 2013; received in revised form 30 May 2013; accepted 30 May 2013

Abstract

The mechanism underlying the grain refinement of cast aluminium by zirconium has been studied through examination of a range ofAl alloys with increasing Zr contents. Pro-peritectic Al3Zr particles are reproducibly identified at or near the grain centres in grain-refinedalloy samples based on the observations of optical microscopy, scanning electron microscopy and X-ray diffraction. From the crystal-lographic study using the edge-to-edge matching model, electron backscatter diffraction and transmission electron microscopy, it is sub-stantiated that the Al3Zr particles are highly potent nucleants for Al. In addition, the effects of Al3Zr particle size and distribution ongrain refinement has also been investigated. It has been found that the active Al3Zr particles are bigger than previously reported othertypes of active particles, such as TiB2 for heterogeneous nucleation in Al alloys. Considering the low growth restriction effect of Zr in Al(the maximum Q-value of Zr in Al is 1.0 K), it is suggested that the significant grain refinement of Al resulting from the addition of Zrcan be mainly attributed to the heterogeneous nucleation facilitated by the in situ formed Al3Zr particles.� 2013 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Cast aluminium; Grain refinement; Peritectic; Edge-to-edge matching model

1. Introduction

The mechanism of grain refinement through inoculationin aluminium alloys has been studied for over half a cen-tury [1–4]. Although many theories/models [5–13] havebeen proposed, none of them can fully elucidate all of theobservations from experiment and practice. In general, itis now accepted that the presence of both potent nucleantparticles and sufficient solutes is essential for effective grainrefinement [2,14–17]. Despite the common recognition ofthese two essentials, the details of the grain refinementmechanism(s) are still of some ambiguity. One of the majorproblems is to determine the exact factors that control theefficiency of grain refinement. For example, the recentinvestigations [18–22] on a number of newly developedgrain refiners for Mg–Al-based alloys and Ti alloys showed

1359-6454/$36.00 � 2013 Acta Materialia Inc. Published by Elsevier Ltd. All

http://dx.doi.org/10.1016/j.actamat.2013.05.044

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

Please cite this article in press as: Wang F et al. The grain refinementhttp://dx.doi.org/10.1016/j.actamat.2013.05.044

relatively low efficiencies compared to that expected fromthe crystallographic matching, which represents thepotency of heterogeneous nucleation and the constitutionalundercooling contribution of solute elements. Therefore, itappears that something important is still missing in the cur-rent understanding of grain refinement. It is worth notingthat the spotlight of previous studies on grain refinementmechanism of Al alloys was primarily focused on the Al–Ti–B and Al–Ti–C systems as they are the most commongrain refiners used in the foundry [3,4]. However, it isbelieved that the study of grain refinement resulting fromother solute elements, which also produce effective grainrefinement in Al may provide fresh insight into the factorsthat control the grain refining efficiency.

It has been long realized that peritectic systems are oftenassociated with effective grain refinement of the parentmetal, such as Al–Ti, Mg–Zr and Mg–Y–Al alloys[23–26]. However, the peritectic approach has been consid-ered as incorrect because addition of far less Ti than its

rights reserved.

mechanism of cast aluminium by zirconium. Acta Mater (2013),

2 F. Wang et al. / Acta Materialia xxx (2013) xxx–xxx

maximum solubility in Al can still produce significant grainrefinement even though there is no peritectic reactioninvolved. This raises the question of whether a similar grainrefining effect can be obtained in other peritectic systems.One of the typical examples of peritectics is the Al–Zr sys-tem, in which the addition of Zr produces appreciable grainrefinement in Al. A number of researchers [7,23,24] haveinvestigated the grain refinement of Al alloys by Zr andsimply attributed the grain refinement to the peritecticreaction induced by the pro-peritectic Al3Zr phase. Never-theless, there is still lack of direct experimental evidence toconfirm the effect of the peritectic reaction. In addition, it isnoted that few of the researchers [7,27] studied the crystal-lography between the pro-peritectic Al3Zr particle and theAl matrix, which has been considered to be of great signif-icance in affecting grain refinement efficiency [25,28–30].Thus, the detailed mechanism underlying the grain refine-ment is still beyond full understanding, particularly interms of the aforementioned two essentials. The particularquestions remaining are whether the observed efficiency ofgrain refinement by Zr is mainly related to the peritecticreaction induced by the pro-peritectic Al3Zr phase (as pro-posed by previous researchers), whether it is due just to thenucleation potency of the pro-peritectic Al3Zr particles thatpromote grain refinement via heterogeneous nucleation, orboth.

Recently, the grain refining effect on pure Al of a rangeof peritectic-forming solutes including Zr was re-examinedby the authors and a Q-value model was used to elucidatethe obtained grain refinement [31]. It was suggested that theconsiderable grain refinement resulting from the additionof Zr is probably due to the introduction of copious nucle-ant particles, which promote grain refinement via enhancedheterogeneous nucleation. However, little experimental evi-dence was provided. The present work aims to clarify theactual roles/effects of Zr addition on grain refinement ofAl through (i) identification of the pro-peritectic Al3Zr par-ticles at grain centres by X-ray diffraction (XRD) and scan-ning electron microscopy (SEM); (ii) characterization ofthe crystallographic features between Al3Zr and Al matrixusing the edge-to-edge matching (E2EM) model [29,30,32–35] to evaluate the potency of Al3Zr; and (iii) verification ofthe crystallographic matching using both electron back-scatter diffraction (EBSD) and transmission electronmicroscopy (TEM). Then the mechanism of grain refine-ment of Al by Zr is discussed in terms of the nucleantpotency, solute contribution, and size and distribution ofthe active Al3Zr particles. Although the addition of Zr overthe maximum solubility may adversely affect the mechani-cal properties of Al alloys, the present work is purely a the-oretical study aiming to understand the mechanism ofgrain refinement in cast Al.

In the past decade, it has been demonstrated that theE2EM model is a powerful tool in understanding grainrefining efficiency through calculation of crystallographicmatching between inoculants and metal matrix. This modelwas originally developed by Zhang and Kelly [32–35] to

Please cite this article in press as: Wang F et al. The grain refinementhttp://dx.doi.org/10.1016/j.actamat.2013.05.044

predict the orientation relationships (ORs) and habit planebetween adjacent phases arising from phase transforma-tions in solids. Since the crystallographic characteristicsbetween the nucleation substrate and metal matrix alsoplay an important role in grain refinement, the E2EMmodel has been extended to the study of grain refinementand achieved great success in elucidating the mechanismsof grain refinement [29,30] and poisoning phenomena[36,37], evaluating the potency of current grain refiners[29,30,38] and discovering new and effective ones [26,27]in both Al and Mg alloys. The critical assumption of theE2EM model is that the crystallographic features betweenany two phases are governed by the minimization of inter-facial energy, which corresponds to the maximum match-ing of parallel atomic rows across the interface betweenphases. In general, the matching rows should be close-packed (cp) or nearly cp atom rows with small interatomicspacing misfit (fr) to maximize the atomic matching alongthe parallel rows. The matching rows can be either straightor zigzag, but the model requires that straight rows matchwith straight rows while zigzag rows match with zigzagrows. Such rows are termed matching directions. In addi-tion, the pair of matching directions should also be con-tained in a pair of cp or nearly cp planes which have asmall interplanar spacing misfit (fd). Such planes aretermed matching planes. If the values of fr and fd are suffi-ciently small (e.g. fr and fd both <10%) in a given system,an energetically favourable OR can be predicted andexpressed in terms of the parallelism of the matching rowsand near parallelism of the matching planes. The angulardeviation between the matching planes and the orientationof the interface plane can be further determined by usingthe Dg parallelism criterion [39,40]. Following the prece-dent success, the present work will use the E2EM modelto evaluate the crystallographic matching and then thepotency of Al3Zr in Al. The predications will then be veri-fied by EBSD and TEM.

In addition to the crystallographic matching betweenthe nucleant particles and matrix metal, the size of thenucleant particles also plays a crucial role in grain refine-ment, as can be well described using the free growth the-ory/model proposed by Greer and his co-workers [41–44].The theory proposes that the onset of free growth of anucleus on a nucleant particle is the controlling factor ingrain initiation, and the critical undercooling for freegrowth to be reached is related to the nucleant particle sizeby the following equation:

DT fg ¼4r

DSVdð1Þ

where r is the solid–liquid interfacial energy, DSV is the entro-py of fusion per unit volume and d is the characteristic size ofthe active nucleant particle, which is the disc diameter of aTiB2 particle in their case studies [41–44]. Based on the freegrowth theory, considerable success has been achieved in pre-dicting the grain size of Al alloys inoculated with commercialgrain refiners and in understanding the mechanism that limits

mechanism of cast aluminium by zirconium. Acta Mater (2013),

F. Wang et al. / Acta Materialia xxx (2013) xxx–xxx 3

the grain refining efficiency. In the present work, the size dis-tribution of Al3Zr particles at both the grain centres and grainboundaries will be determined.

2. Experimental

A series of aluminium alloys with increased Zr additionlevels (0.0, 0.1, 0.2, 0.3 and 0.5%) were cast by adding com-pacted Al–Zr pellets, which were made of pure Zr powdersand Al chips, to a high-purity commercial aluminium melt.(All the contents throughout the paper are in weight per-cent unless otherwise specified.) All of the aluminium meltswere prepared in clay-bonded graphite crucibles, whichwere placed in a resistance furnace and heated to 780 �C.After addition of the compacted pellets, the melt was thenisothermally held at 780 �C for 20 min and stirred immedi-ately before dipping the pre-heated graphite moulds intothe melt to collect samples. The melt was then cooledbetween two pieces of Fiberfrax in air (1 K s�1) to theambient temperature. The cooling rate in the present workis lower than that (4 K s�1) used in the standard TP-1 test[41], in order to facilitate the peritectic reaction. Details ofthe casting procedure are available in a previous paper [31].

Metallographic samples were sectioned approximately10 mm from the base of cast ingots and were mechanicallyground and polished. These samples were first examined inan optical microscope with polarized light after anodizingusing a 0.5 pct HBF4 solution for about 2 min at 20 V.The grain sizes were measured using the linear intercepttechnique (ASTM E112-10). They were then further exam-ined in a Bruker D8 diffractometer for XRD and in aJEOL-6460LA scanning electron microscope (SEM). TheXRD was performed at 40 kV with Cu Ka radiation (wave-length kja1 = 1.54056 A). Crystallographic orientationrelationships between the observed nucleant particles,Al3Zr and the Al matrix were determined using an auto-mated EBSD facility equipped with an orientation imagingsystem in the SEM.

TEM foils were prepared using the focused ion beam(FIB) method. The intermetallic Al3Zr particles at graincentres were cut out together with the Al matrix using aFEI Dual Beam FIB/SEM-Helios NanoLab. A typicalimage of the sampling process through FIB/SEM is illus-trated in Fig. 1. The foils were then examined in a JEOL2100 TEM at 200 kV.

3. Results and discussion

Fig. 2 shows the typical as-cast microstructures of Alalloys with different additions of Zr. It is evident that thegrain morphology with the addition of 0.1% Zr is similarto that of pure aluminium, i.e. large columnar grains. How-ever, once the addition level reaches 0.2%, an evidentcolumnar to equiaxed transition and significant grainrefinement in microstructure are observed. The grain sizevariation of Al alloys with increasing Zr addition is plottedin Fig. 3. It can be seen that an initial addition of 0.1% Zr

Please cite this article in press as: Wang F et al. The grain refinementhttp://dx.doi.org/10.1016/j.actamat.2013.05.044

decreases the average grain size by only around 50 lm,while a further addition of 0.1% Zr (i.e. total 0.2% Zr inthe alloy), which exceeds the maximum solubility(cm = 0.11% Zr), decreases the grain size significantly, byapproximately 500 lm. After this sharp reduction, thegrain size then decreases gradually as the addition levelof Zr further increases.

In order to understand the mechanism underlying thegrain refinement produced by the addition of Zr, the as-cast Al–Zr alloy samples were first examined by XRD todetermine the exact phase constituents in the alloys. Afterindexing the diffraction spectra shown in Fig. 4, it can beseen that spectra (a) and (b) show only the peaks of a-Al,indicating that these two samples contain no other phasesbut a-Al. However, extra peaks are reproducibly observedin spectra (c), (d) and (e), as shown in Fig. 4. After carefulcomparison with the Powder Diffraction File database [45],these extra peaks were identified to be the reflections ofAl3Zr. This implies that Al3Zr phase has formed duringsolidification in these three alloys.

The samples used to obtain the XRD spectra shown inFig. 4 were also examined by SEM to reveal the detailedmicrostructure. It is found that, in the samples with signif-icant grain refinement (Zr addition greater than 0.1%),small particles are reproducibly observed at or near thecentres of the refined grains, as shown in Fig. 5a. Theseparticles are believed to act as heterogeneous nucleationsites during solidification. In contrast, no trace of such par-ticles could be detected in the samples with the 0.1% Zraddition, which consists of coarse grains. Fig. 5b shows atypical EDX spectrum of the particles observed at or nearthe grain centres. An analysis of the EDX spectrum showsthat the particles are rich in Al and Zr, and that theapproximate atomic ratio of Al to Zr is about 3:1. Consid-ering the XRD and EDX results, as well as the equilibriumAl–Zr phase diagram, it is confirmed that the particlesobserved at or near the grain centres are Al3Zr particles.

Combining the XRD and SEM observations with thegrain size measurements, it is noted that, in the alloy sam-ples containing less than or equal to 0.1% Zr, no effectivegrain refinement can be observed, whereas pronouncedgrain refinement is achieved in the alloy samples containingmore than 0.1% Zr, with the observation of Al3Zr particlesat or near grain centres. This further verifies the hypothesisthat the Al3Zr particles observed at grain centres are prob-ably the nucleants that are responsible for the grain refine-ment obtained by the addition of Zr to Al.

In order to evaluate the nucleation potency of Al3Zrparticles, the crystallographic matching between Al3Zrand Al was first studied using the E2EM model. TheE2EM model calculation in terms of crystal structure, lat-tice parameters and atomic positions of both phasesinvolves identification of cp atomic rows and cp planes,and calculation of the interatomic spacing misfit (fr) alongthe matching directions and the interplanar spacing mis-match (fd) between the matching planes. Aluminium pos-sesses a face-centred cubic (fcc) structure, with a lattice

mechanism of cast aluminium by zirconium. Acta Mater (2013),

(a) (b)

Al3Zr particle

Al3Zrparticle

Almatrix

Fig. 1. SEM images of a TEM foil obtained using FIB: (a) before being cut out from cast sample and (b) after being transferred to the copper holder.

Fig. 2. Typical micrographs of as-cast aluminium alloys with increasing contents of Zr solute: (a) pure aluminium, (b) 0.1 wt.% Zr, (c) 0.2 wt.% Zr, and (d)0.3 wt.% Zr.

Fig. 3. Grain size variation of as-cast aluminium with increasing additionlevel of Zr.

4 F. Wang et al. / Acta Materialia xxx (2013) xxx–xxx

Please cite this article in press as: Wang F et al. The grain refinementhttp://dx.doi.org/10.1016/j.actamat.2013.05.044

parameter of a = 0.4049 nm [46]. It has one cp straight row

along the h1 1 0iSAl direction and one nearly cp zigzag row

along the h�2 1 1iZAl direction. For convenience, superscripts“S” and “Z” are used to distinguish the straight and zigzagrows, respectively. As a simple fcc structure, Al has threecp or nearly cp planes. The most highly cp plane is

{111}Al, which contains both the h1 1 0iSAl and h�2 1 1iZAl

directions. {02 0}Al is the second most highly cp plane,

but it only contains the h1 1 0iSAl direction. The third cp

plane is {220}Al, and it also contains both the h1 1 0iSAl

and h�2 1 1iZAl directions. The atomic configuration of Alwithin the most highly cp plane (111) is shown in Fig. 6a.

The Al3Zr phase has a tetragonal crystal structure, withlattice parameters a = 0.4007 nm and c = 1.7286 nm. Eachunit cell contains 12 Al atoms and 4 Zr atoms [46]. Themost highly cp plane of Al3Zr is the {114} plane, which

mechanism of cast aluminium by zirconium. Acta Mater (2013),

Fig. 4. XRD spectra of the as-cast Al alloys. From bottom to top: (a) pure Al, (b) with 0.1 wt.% Zr addition, (c) with 0.2 wt.% Zr addition, (d) with0.3 wt.% Zr addition, and (e) with 0.5 wt.% Zr addition.

Fig. 5. (a) SEM BSE micrograph showing a typical Al3Zr particle at the centre of a grain in an Al alloy with the addition of 0.2 wt.% Zr; (b) EDXspectrum taken from the particle at the grain centre.

F. Wang et al. / Acta Materialia xxx (2013) xxx–xxx 5

contains three cp rows: h2 2 �1iZAl3Zr, h1 1 0iSAl3Zr andh4 0 �1iSAl3Zr. The second most highly cp plane is the {020}plane, which contains only one cp row, h4 0 �1iSAl3Zr. Anotherpossible cp plane of the Al3Zr phase is {220}, which con-tains two cp rows: h2 2 �1iZAl3Zr and h1 1 0iSAl3Zr. The atomicconfiguration of Al3Zr within the most highly cp plane(114) is shown in Fig. 6b.

Based on the identified cp rows and cp planes, the valuesof fr and fd between Al3Zr and Al are calculated. Table 1lists the values of fr along the cp rows between Al3Zr andAl by coupling the same types of atomic row (i.e. straightrows match with straight rows, and zigzag rows match withzigzag rows). It can be seen that all the values of fr are lessthan 5%. The calculated interplanar spacing mismatch, fd,between the cp planes of Al3Zr and Al are listed in Table 2.One can see that there are only three pairs of planes with fd

below 10%. These are: f1 1 1gAljjf1 1 4gAl3Zr with

Please cite this article in press as: Wang F et al. The grain refinementhttp://dx.doi.org/10.1016/j.actamat.2013.05.044

fd = 1.34%, f0 2 0gAljjf0 2 0gAl3Zr with fd = 1.05% andf0 2 2gAljjf2 2 0gAl3Zr with fd = 1.05%.

According to the E2EM model [29,30], to form an OR,the matching rows must lie in the matching planes. There-fore, the cp row pairs with fr less than 10% and the corre-sponding cp plane pairs that contain these row pairs with fd

less than 10% are represented in Fig. 7. Since a given pairof rows always lies in one or more pair of cp planes, thearrows are used to indicate the associated plane pairs fora given row pair.

Combining the matching row pairs with the planepairs containing the matching rows, six possible ORsare obtained. After refinement of these ORs using theDg parallelism criterion [39,40], three distinguishableORs between Al and Al3Zr, which are potentiallyobserved in Al–Zr alloys, are predicted. These are listedin Table 3.

mechanism of cast aluminium by zirconium. Acta Mater (2013),

Fig. 6. Atomic configurations of Al and Al3Zr on their most highly cp planes: (a) (111)Al, (b) ð1 1 4ÞAl3Zr. Bold lines highlight the cp rows within theseplanes.

Table 1Interatomic spacing misfit, fr, along the CP matching rows between Al3Zrand Al.

h1 1 0iSAljjh1 1 0iSAl3Zr h1 1 0iSAljjh4 0 �1iSAl3Zr h�2 1 1iZAljjh2 2 �1iZAl3Zr

1.05% 1.99% 4.03%

6 F. Wang et al. / Acta Materialia xxx (2013) xxx–xxx

To experimentally verify the predictions, the ORsbetween the Al3Zr particles at or near grain centres andthe Al matrix were determined using automated EBSD.After SEM images and EBSD patterns from both Al3Zrparticles and Al grains were recorded, the Kikuchi patternswere indexed based on the lattice parameters mentioned inthe previous section. Fig. 8 shows a typical backscatteredelectron (BSE) image where an Al3Zr particle is clearlyobserved near the grain centre, and the correspondingEBSD patterns from the Al3Zr particle and the Al grain.It can be seen that the ð1 1 4ÞAl3Zr Kikuchi band is almostparallel to the ð1 1 �1ÞAl band, and the ½1 �1 0�Al3Zr and½4 0 �1�Al3Zr Kikuchi poles are very close to the ½1 0 1�Al and½1 �1 0�Al poles, respectively. Therefore, the OR shown inFig. 8 can be roughly expressed as:

½1 �1 0�Al3Zrjj½1 0 1�Al; ½4 0 �1�Al3Zrjj½1 �1 0�Al; ð1 1 4ÞAl3Zrjjð1 1 �1ÞAl

ð2ÞTo examine a large number of particle–grain pairs effi-

ciently, a simple numerical approach developed based onEuler angles [47] was employed to determine the ORsbetween the Al3Zr particles and Al grains. In the presentstudy, the Euler angles of 40 Al3Zr particles and their cor-responding Al grains were recorded by EBSD. In order tocorrelate the EBSD results with the E2EM model predic-tions, the ORs determined from EBSD and predicted bythe E2EM model are all expressed in a stereographic pro-jection in terms of [001]Al. Two directions, ½1 �1 0�Al3Zr and½4 0 �1�Al3Zr, and one plane, ð1 1 4ÞAl3Zr were selected toexpress the ORs, as shown in Fig. 9. It is important to note

Please cite this article in press as: Wang F et al. The grain refinementhttp://dx.doi.org/10.1016/j.actamat.2013.05.044

that, since the stereographic projections for direction andplane poles are the same for cubic crystal, all the planeand direction poles are expressed in a single stereographicprojection for brevity.

A comparison of the EBSD ORs with the predicted ORsin Fig. 9 shows that the experimentally determined ORs byEBSD agree very well with the predictions of the E2EMmodel, except for OR (A). In terms of theð1 1 4ÞAl3Zrjjð1 1 �1ÞAl plane pair, the experiment results are4� away from the OR (A) prediction.

In order to increase the accuracy of the experimentalresults and to verify the EBSD results, the ORs betweenthe Al3Zr particles and Al grains were further determinedusing convergent beam Kikuchi line diffraction patterns(CBKLDP) in an advanced TEM [35,48–50]. Fig. 10 showsa typical TEM image of an Al3Zr particle embedded in theAl matrix and the corresponding Kikuchi line diffractionpatterns. After indexing the Kikuchi pattern, the ORsshown in Fig. 10 can be expressed as follows:

½1 �1 0�Al3Zr 0:91� from ½1 0 1�Al; ½4 0 �1�Al3Zr 0:74� from

� ½1 �1 0�Al; ð1 1 4ÞAl3Zr 0:87� from ð1 1 �1ÞAl ð3Þ

Using the CBKLDP approach, the ORs between sevenpairs of Al3Zr particles and Al grains were determined.For comparison, these ORs are also expressed in the ste-reographic projection shown in Fig. 9. As can be seen,the ORs obtained by TEM through the CBKDP approachare highly consistent with those determined by EBSD andthose predicted by the E2EM model. This further confirmsthe predictions of the E2EM model and the favourablecrystallographic matching between Al3Zr and Al. TheOR between the Al3Zr and Al was previously studied bySchumacher et al. [27] in 2000. Using the selected area dif-fraction (SAD) technique, a cube–cube OR between theAl3Zr and Al was determined, which agreed with the ORreported by Marcantonio and Mondolfo [7]. In order to

mechanism of cast aluminium by zirconium. Acta Mater (2013),

Fig. 7. Graphic representation of the matching row pairs (see Table 1) and the related suitable matching plane pairs (see Table 2) which carry them, aspredicted by E2EM.

Table 3Refined crystallographic ORs between Al3Zr and Al.

ORs Nearly parallel directions (1) Nearly parallel directions (2) Nearly parallel plane

A ½1 �1 0�Al3Zrk½1 0 1�SAl ½4 0 �1�SAl3Zr 1:34� from ½1 �1 0�SAl ð1 1 �4ÞAl3Zr 3:50� from ð1 1 �1ÞAlB ½4 0 �1�SAl3Zrk½1 �1 0�SAl ½1 �1 0�SAl3Zr 1:26� from ½1 0 1�SAl ð1 1 4ÞAl3Zr 0:04� from ð1 1 �1ÞAlC ½1 �1 0�Al3Zr 0:68� from ½1 0 1�SAl ½4 0 �1�SAl3Zr 1:30� from ½1 �1 0�SAl ð1 1 4ÞAl3Zr 0:68� from ð1 1; �1ÞAl

]101[

]011[

)111(

]011[

]140[

)114

(

(c) Al Pattern(b) Al3Zr Pattern(a) SEM BSE image

Fig. 8. (a) A typical SEM BSE micrograph showing an Al3Zr particle near the centre of an Al grain, and the corresponding EBSD patterns from: (b) theAl3Zr particle and (c) the Al matrix.

Table 2Interplanar spacing mismatch, fd, between the CP or nearly CP planes in Al3Zr and Al.

1 1 1Aljj1 1 4Al3Zr 1 1 1Aljj0 2 0Al3Zr 1 1 1Aljj2 2 0Al3Zr 0 2 0Aljj1 1 4Al3Zr 0 2 0Aljj0 2 0Al3Zr 0 2 0Aljj2 2 0Al3Zr 0 2 2Aljj1 1 4Al3Zr 0 2 2Aljj0 2 0Al3Zr 0 2 2Aljj2 2 0Al3Zr

1.34% 17.00% 64.79% 14.77% 1.05% 42.25% 39.66% 28.50% 1.05%

F. Wang et al. / Acta Materialia xxx (2013) xxx–xxx 7

compare with the cube–cube OR determined in the previ-ous work, the OR obtained in the present study can beexpressed in a pseudo-cube–cube form as follows:

½1 0 0�Al3Zr 2:11� from ½1 0 0�Al; ½0 1 0�Al3Zr 1:91� from

½0 1 0�Al; ½0 0 1�Al3Zr 2:08� from ½0 0 1�Al ð4Þ

It can be seen that the OR obtained in the present studyis very close to the cube–cube OR. Using Euler’s theorem,the present OR can be seen to be related to the cube–cubeOR by a rotation of 0.98� about an axis that is very close to[100]Al or ½1 0 0�Al3Zr. It is considered that this deviation

Please cite this article in press as: Wang F et al. The grain refinementhttp://dx.doi.org/10.1016/j.actamat.2013.05.044

between the previous [27] and present results is probablyattributable to the different techniques used to determinethe OR. In the previous study, both the Al3Zr and Al wereproduced through crystallization of an Al-based metallicglass matrix, which was obtained by an extremely highcooling rate. In the present work, the Al3Zr particles andAl were obtained through solidification of molten Al–Zralloy at a slow cooling rate. Furthermore, in the previousstudy, the OR was determined using the SAD method,which is not as accurate as the CBKLDP technique usedhere [48,49].

mechanism of cast aluminium by zirconium. Acta Mater (2013),

(a) (b) AlZrAl ]101//[]011[3

AlZrAl ]011//[]140[3 AlZrAl )111//()114(

3

(b) AlZrAl ]101//[]011[3

2o

2o

2 o

2 o

Al]011[

Al]101[

2o

2o

Al)111(

(c) ]011//[]140[

(d) )111//()114(

OR(A)OR(B)OR(C)EBSD ORs

Al PolesTEM ORs

OR(A)OR(B)OR(C)EBSD ORs

Al PolesTEM ORs

OR(A)OR(B)OR(C)EBSD ORs

Al PolesTEM ORs

OR(A)OR(B)OR(C)EBSD ORs

Al PolesTEM ORs

(c) (d)

ZrAl 3 Al

ZrAl 3 Al

Fig. 9. Stereographic projection showing the ORs between Al3Zr particles and Al grains, plotted with fcc Al in the [001] direction.

(a) (b) (c)Al3 Zr Particle

Al Matrix

B

C

)024(

)1211(

)451

(

)2011(

)1231(

)1321(

)31

1(

)202(

)115(

)15

1(

)13

3(

)042(

Fig. 10. (a) Typical bright-field TEM image of the interface between the Al3Zr particle and the Al matrix; (b) Kikuchi pattern from Al3Zr particle; (c)Kikuchi pattern from Al matrix.

8 F. Wang et al. / Acta Materialia xxx (2013) xxx–xxx

From the E2EM model, EBSD and TEM investigationsof the crystallographic relationship between Al3Zr and Al,it is substantiated that the Al3Zr particles are indeed potentnucleants for Al. Previous crystallographic studies on grainrefinement [29,30] indicated that the grain refining effi-ciency of a grain refiner is related to the fr and fd values.Smaller misfit and mismatch values correspond to highergrain refining efficiency due to the lower interfacial energybetween the grain refiners and the solid formed on therefiners. The small fr and fd values between Al3Zr and Al

Please cite this article in press as: Wang F et al. The grain refinementhttp://dx.doi.org/10.1016/j.actamat.2013.05.044

shown in Tables 1 and 2 as well as Fig. 7 imply high grainrefining efficiency of Al3Zr for Al. However, in order toactivate the potential grain refining efficiency, the constitu-tional undercooling contribution of solutes must also beconsidered according to the current understanding of thegrain refinement mechanism [2,14–17]. In the present Al–Zr system, the constitutional undercooling contributionof Zr in Al can be quantified by the growth restriction fac-tor, Q [17,51]. This is defined as Q = m � (k � 1) � c0, wherem is the slope of the liquidus, k is the partition coefficient

mechanism of cast aluminium by zirconium. Acta Mater (2013),

0 10 20 30 40 50 60 70 80 90 100 110 1200

10203040506070 (b)

Rel

ativ

e fr

eque

ncy

(%)

Particle size ( m)

Active Al3

3

3

010203040506070 Total Al3Zr particles in Al-0.2Z ralloy

Total Al3Zr particles in Al-0.3Z ralloy

Total Al 3

(a)

Zr particles in Al-0.5Z ralloy

Zr particles in Al-0.2Z ralloy

Zr particles in Al-0.3Z ralloy

Zr particles in Al-0.5Z ralloy

Active AlActive Al

Fig. 11. The size distributions of the Al3Zr particles: (a) the total Al3Zr particles, including those at grain centres and along grain boundaries, and (b) theactive Al3Zr particles at grain centres in the alloys with three different Zr addition levels.

F. Wang et al. / Acta Materialia xxx (2013) xxx–xxx 9

and c0 is the solute content. It is worth noting that, whenthe Zr content exceeds the maximum solubility(cm = 0.11% Zr), the pro-peritectic Al3Zr forms prior tothe solidification of Al according to the equilibrium phasediagram. Therefore, the Q-value for the alloys with Zradditions above the maximum solubility was determinedby assuming that all of the excess Zr above the maximumsolubility is combined with Al to form Al3Zr and hencethe value of c0 used in calculating the growth restrictionfactor is replaced by the value of the maximum solubility,cm. Based on the parameters obtained from the phase dia-gram, the maximum growth restriction factor of Zr in Al iscalculated to be only 1 K. In comparison with the growthrestriction factor of Ti in Al (�10 K for normal additions[3,51]), the constitutional undercooling contribution of Zrin Al is very small.

In terms of the solute theory of grain refinement, thepresence of only potent Al3Zr nucleant particles in Alshould not produce effective grain refinement since it isdeficient in a significant constitutional undercooling contri-bution due to the very low growth restriction factor of Zrin Al. However, contrary to this expectation, a substantialgrain size reduction is observed when the Zr additionexceeds the maximum solubility. This pronounced grainrefinement implies that the growth restriction effectimposed by the solute segregation ahead of the growinginterface is not the limiting factor in the grain refinement;rather, something else contributes to the grain refinementobtained in the present work.

As mentioned above, the size of nucleant particles alsoplays a key role in grain refinement. Therefore, the size dis-tributions of the total Al3Zr particles (including particles atgrain centres and along grain boundaries) and of the activeAl3Zr particles at grain centres were examined using SEMBSE micrographs. The results are shown in Fig. 11. Notethat each data point in Fig. 11 represents the relative fre-

Please cite this article in press as: Wang F et al. The grain refinementhttp://dx.doi.org/10.1016/j.actamat.2013.05.044

quency of the particles within the size range of ±5 lm ofthe displayed value. It is also important to mention thatthe ratio of length to thickness of the Al3Zr particlesobserved by SEM is approximately within the range from6 to 20, as shown in Figs. 5a and 8a. In the present analysis,the length of the Al3Zr particles is used to characterize theirsize.

From Fig. 11a, one can see that the size distribution of thetotal Al3Zr particles (0–115 lm) is much broader than thatof the TiB2 particles (0–6 lm) when the commercial grainrefiners are used in Al alloys [41,43]. This substantially largesize of Al3Zr particles is probably a result of the very slowcooling rate (1 K s�1) in the present casting experiment,which allows the Al3Zr particles to grow for a much longertime during the solidification process. According to the freegrowth theory [41–44], the bigger the nucleant particles, thesmaller the undercooling required to activate the grain initi-ation on the nucleant particles. Therefore, the significantlylarge size of the Al3Zr particles that corresponds to a smallcritical undercooling barrier may explain why considerablegrain refinement was still achieved even though the Q-valueof Zr in Al is very small. Further analysis of the size distribu-tions of the total Al3Zr particles reveals that the majority(70–80%) of the total Al3Zr particles are smaller than25 lm. In contrast, examination of the size distributions ofthe active Al3Zr particles at grain centres in all three alloysshown in Fig. 11b indicates that no particles are smaller than25 lm. The size of the majority of the active Al3Zr particlesincreases as the Zr addition level is increased. In general,most of the active Al3Zr particles at grain centres are largerthan 35 lm, which lies at the upper end of the size distribu-tions of the total Al3Zr particles. This agrees well with thefree growth theory, which suggests that grain initiationoccurs first on the biggest nucleant particles, then on pro-gressively smaller particles as the undercooling is increased[41–44].

mechanism of cast aluminium by zirconium. Acta Mater (2013),

10 F. Wang et al. / Acta Materialia xxx (2013) xxx–xxx

The analysis of the effect of Al3Zr particle size on thegrain refinement of Zr in Al verifies that the size of nucleantparticles also plays a critical role in promoting grain refine-ment as postulated by the free growth theory [41–44], inaddition to the crystallography of the nucleant particlesand the matrix metal, and the constitutional undercoolingcontribution of the solute. Therefore, it is concluded thatthe significant grain refinement obtained in the Al–Zralloys results from a combined effect of the high potencyand big Al3Zr particles as nucleation sites for Al grains.

4. Conclusions

(1) From examinations using optical microscopy, XRDand SEM, the pro-peritectic Al3Zr particles arereproducibly observed at or near the grain centresin the alloy samples showing significant grain refine-ment, while no trace of such particles could bedetected in the unrefined samples.

(2) The crystallographic study on Al3Zr and Al using theE2EM model indicates that the values of interatomicspacing misfit and interplanar mismatch between Al3-

Zr and Al are very small, implying high grain refiningefficiency of Al3Zr in Al.

(3) Three orientation relationships between Al3Zr and Alhave been predicted based on the E2EM model. Theyare also experimentally verified by both EBSD andTEM. These ORs are:

OR (A): ½1 �1 0�SAl3Zrjj½1 0 1�SAl, ½4 0 �1�SAl3Zr 1.34� from

½1 �1 0�SAl, ð1 1 4ÞAl3Zr 3.50� from ð1 1 �1ÞAl;

OR (B): ½4 0 �1�SAl3Zrjj½1 �1 0�SAl, ½1 �1 0�SAl3Zr 1.26� from

½1 0 1�SAl, ð1 1 4ÞAl3Zr 0.04� from ð1 1 �1ÞAl;

OR (C): ½1 �1 0�SAl3Zr 0.68� from ½1 0 1�SAl, ½4 0 �1�SAl3Zr

1.30� from ½1 �1 0�SAl, ð1 1 4ÞAl3Zr 0.68� from ð1 1 �1ÞAl.

(4) Combining the observations of small growth restric-tion factor (Q-value) of Zr and the high grain refiningefficiency and big Al3Zr particles in Al, it is proposedthat the growth restriction activated by the underco-oling ahead of the growing solid–liquid interface isnot the limiting factor in grain refinement. Instead,it is the crystallographic matching and the size distri-bution of nucleant particles that actually govern thegrain refinement efficiency.

Acknowledgements

The authors are very grateful to the Australian Re-search Council for funding support. F.W. would also liketo acknowledge the support of China Scholarship Coun-cil. The authors acknowledge the facilities, and the scien-tific and technical assistance, of the AustralianMicroscopy & Microanalysis Research Facility at theCentre for Microscopy and Microanalysis, The Universityof Queensland.

Please cite this article in press as: Wang F et al. The grain refinementhttp://dx.doi.org/10.1016/j.actamat.2013.05.044

References

[1] McCartney DG. Int Mater Rev 1989;34:247.[2] Easton MA, StJohn DH. Metall Mater Trans A 1999;30:1613.[3] Murty BS, Kori SA, Chakraborty M. Int Mater Rev 2002;47:3.[4] Quested TE. Mater Sci Technol 2004;20:1357.[5] Cibula A. J Inst Met 1949;76:321.[6] Cibula A. J Inst Met 1951;80:1.[7] Marcantonio JA, Mondolfo LF. J Inst Met 1970;98:23.[8] Marcantonio JA, Mondolfo LF. Metall Trans 1971;2:465.[9] Backerud L, Shao Y. Aluminium (Germany) 1991;67:780.

[10] Jones GP. United Kingdom; 1983. p. 40.[11] Mohanty PS, Gruzleski JE. Acta Metall Mater 1995;43:2001.[12] Mohanty PS, Gruzleski JE. Acta Mater 1996;44:3749.[13] Vader M, Noordegraaf J. Proceedings of the 119th TMS annual

meeting, February 18–22, 1990, Anaheim, CA. Warrendale, PA:TMS; 1990, p. 851.

[14] Easton MA, StJohn DH. Metall Mater Trans A 1999;30:1625.[15] Greer AL, Cooper PS, Meredith MW, Schneider W, Schumacher P,

Spittle JA, et al. Adv Eng Mater 2003;5:81.[16] Johnsson M, Backerud L, Sigworth GK. Metall Trans A 1993;24A:481.[17] StJohn DH, Qian M, Easton MA, Cao P. Acta Mater 2011;59:4907.[18] Bamberger M. Mater Sci Technol 2001;17:15.[19] Bermingham MJ, McDonald SD, Dargusch MS, StJohn DH. Scripta

Mater 2008;58:1050.[20] Fu HM, Qiu D, Zhang MX, Wang H, Kelly PM, Taylor JA. J Alloys

Compd 2008;456:390.[21] Fu HM, Zhang MX, Qiu D, Kelly PM, Taylor JA. J Alloys Compd

2009;478:809.[22] Tamirisakandala S, Bhat RB, Tiley JS, Miracle DB. Scripta Mater

2005;53:1421.[23] Crossley FA, Mondolfo LF. Trans Am Inst Min Metall Eng 1951;191:1143.[24] Eborall MD. J Inst Met 1949;76:295.[25] StJohn DH, Ma Q, Easton MA, Peng C, Hildebrand Z. Metall Mater

Trans A 2005;36A:1669.[26] Qiu D, Zhang MX, Taylor JA, Kelly PM. Acta Mater 2009;57:3052.[27] Schumacher P, Cizek P, Bunn AM. Light metals 2000, March 12–16,

2000. Nashville, TN. Warrendale, PA: TMS; 2000. p. 839.[28] Bramfitt B. Metall Trans 1970;1:1987.[29] Zhang MX, Kelly PM, Easton MA, Taylor JA. Acta Mater 2005;53:1427.[30] Zhang MX, Kelly PM, Qian M, Taylor JA. Acta Mater 2005;53:3261.[31] Wang F, Liu Z, Qiu D, Taylor JA, Easton MA, Zhang M-X. Acta

Mater 2013;61:360.[32] Kelly P, Zhang M. Metall Mater Trans A 2006;37:833.[33] Kelly PM, Zhang MX. Mater Forum 1999;23:41.[34] Zhang MX, Kelly PM. Acta Mater 2005;53:1073.[35] Zhang MX, Kelly PM. Acta Mater 2005;53:1085.[36] Qiu D, Taylor JA, Zhang MX. Metall Mater Trans A 2010;41:3412.[37] Qiu D, Taylor JA, Zhang MX, Kelly PM. Acta Mater 2007;55:1447.[38] Qiu D, Zhang MX, Fu HM, Kelly PM, Taylor JA. Philos Mag Lett

2007;87:505.[39] Zhang WZ, Purdy GR. Philos Mag A 1993;68:279.[40] Zhang WZ, Ye F, Zhang C, Qi Y, Fang HS. Acta Mater 2000;48:2209.[41] Greer AL, Bunn AM, Tronche A, Evans PV, Bristow DJ. Acta Mater

2000;48:2823.[42] Greer AL, Quested TE. Philos Mag 2006;86:3665.[43] Quested TE, Greer AL. Acta Mater 2004;52:3859.[44] Quested TE, Greer AL. Acta Mater 2005;53:4643.[45] JCPDS-International Centre for Diffraction Data, PCPDF-WIN V.

2.3; 2002.[46] Villars P, Calvert LD. Pearson’s handbook of crystallographic data for

intermetallic phases. Materials Park, OH: ASM International; 1991.[47] Qiu D, Zhang M-X, Kelly PM. Scripta Mater 2009;61:312.[48] Zhang MX, Kelly PM. Acta Mater 1998;46:4081.[49] Zhang MX, Kelly PM. Acta Mater 1998;46:4617.[50] Zhang MX. PhD thesis, University of Queensland; 1997.[51] Easton MA, StJohn DH. Acta Mater 2001;49:1867.

mechanism of cast aluminium by zirconium. Acta Mater (2013),