Accepted Manuscript

On grain coarsening and refining of the Mg-3Al alloy by Sm

Xiaoyu Hu, Penghuai Fu, David StJohn, Liming Peng, Ming Sun, Mingxing Zhang

PII: S0925-8388(15)31750-3

DOI: 10.1016/j.jallcom.2015.11.193

Reference: JALCOM 36048

To appear in: Journal of Alloys and Compounds

Received Date: 23 August 2015

Revised Date: 21 November 2015

Accepted Date: 25 November 2015

Please cite this article as: X. Hu, P. Fu, D. StJohn, L. Peng, M. Sun, M. Zhang, On grain coarseningand refining of the Mg-3Al alloy by Sm, Journal of Alloys and Compounds (2015), doi: 10.1016/j.jallcom.2015.11.193.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service toour customers we are providing this early version of the manuscript. The manuscript will undergocopyediting, typesetting, and review of the resulting proof before it is published in its final form. Pleasenote that during the production process errors may be discovered which could affect the content, and alllegal disclaimers that apply to the journal pertain.

http://dx.doi.org/10.1016/j.jallcom.2015.11.193

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

1

On grain coarsening and refining of the Mg-3Al alloy by

Sm

Xiaoyu Hua,b

, Penghuai Fua, David StJohn

b,*, Liming Peng

a,*, Ming Sun

a, Mingxing

Zhangb

a. National Engineering Research Center of Light Alloy Net Forming and State Key

Laboratory of Metal Matrix Composites, Shanghai Jiao Tong University, 800

Dongchuan Road, Shanghai 200240, China

b. Centre for Advanced Materials Processing and Manufacturing, School of

Mechanical and Materials Engineering, The University of Queensland, St Lucia,

Queensland, 4072, Australia1

Abstract: The effect of Sm on the microstructure of a Mg-3Al alloy has been

investigated where it was found that the size of the primary α-Mg grains was

significantly coarsened by the addition of 0.7 wt. % Sm, but dramatically refined by

2.1 wt. % Sm. It is proposed that the coarsening effect of Sm is due to a decrease in

the nucleation potency of the native Al-Fe-C-O particles when they transform to the

lower potency Al-Fe-Sm-C-O particles. The grain refinement observed when 2.1 wt. %

Sm is added, is caused by the formation of potent Al2Sm nucleant particles prior to

the solidification of α-Mg. The “Interdependence Theory” was then applied to

describe the nucleant selection process in the Mg-3Al-2.1Sm alloy.

Key words: Solidification; Metals and alloys; Rare earth compounds; Nucleation

* David StJohn, Tel.: +61 7 3365 3641, fax: +61 7 3365 3888. E-mail address:

d.stjohn@uq.edu.au; Corresponding author. Liming Peng, Tel. +86 21 54742911, fax:

+86 2134202794. E-mail address: plm616@sjtu.edu.cn.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

2

1. Introduction

Grain refining is one of the most important methods for improving the mechanical

properties of metallic materials [1, 2]. Mg-Al alloys are the most common and

economical Mg alloys, but it is difficult to achieve grain refinement in cast Mg-Al

alloys [3, 4]. It is now understood that two key factors promote grain refinement [5]:

Solute elements and heterogeneous nucleant particles.

The addition of solute elements with a high growth restriction factor, Q, provides

rapid development of constitutional supercooling, which restricts the growth of

existing grains allowing additional nucleation events to occur. For example, Park et al.

[6] recently reported that a Mg-8Al-2Zn alloy was significantly refined by 2 ~ 6 % Sn

(all compositions are in wt. % hereinafter). Sevik et al. [7] reported that 2 % Sr

refined a Mg-6Al-0.3Mn-0.3Ti-1Sn alloy from ~ 50 μm to 30 μm significantly

improving the mechanical properties. Choi et al [8] showed that 0.02 % Ti refined the

microstructure of a Mg-6Al-1Zn alloy and improved the corrosion properties. All

these solute elements have sufficiently high values of Q to rapidly form constitutional

supercooling in a Mg melt [9].

In most Al-containing Mg alloys, carbon-based particles, such as Al4C3, Al2CO and

Al2MgC2, are believed to be the main heterogeneous nuclei for the α-Mg phase

during solidification [10-12]. Hence, much research effort has been applied to

identify carbon-based particles which can inoculate the melt in a cost-effective and

reliable manner [13-16]. Recently, Liu et al. added a new Mg-50%Al4C3 master alloy

into an AZ91 alloy and achieved remarkable grain refinement [17]. Suresh et al [18]

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

3

found that an addition of 0.4 % charcoal reduced the grain size of an AZ91 alloy from

100 μm to 14 μm. Other particles have also been reported to refine Mg-Al alloys. For

example, Wang et al [19] found that an addition of Al-4.99Zr-1.1B master alloy which

contains ZrB2 particles, reduced the grain size of an AZ31 alloy from 170 μm to 45 μm.

Fu et al. [20] reported that adding ZnO particles to Mg produced grain refinement,

although the efficiency was affected by reaction between Mg and ZnO reducing the

ZnO to Zn in solution. However, problems of pollution and/or poor dispersion of

particles make most of the above methods inappropriate for industrial application.

Fe or Mn additions can also reduce the grain size of Mg-Al alloys by generating

compounds such as Al8(Mn,Fe)5 [13]. Crystallographic studies by Qiu et al. [21] have

shown that Al8Mn5 and Al8(Mn,Fe)5 are both potential nucleant particles. Cao et al.

[22] suggested that native Al-Fe-C-O particles naturally present in an alloy can also

act as active nucleant particles. Superheating and Elfinal methods [13] have also

been applied to refine Mg-Al alloys, but have proved to be inappropriate for

industrial application because they are not reliable or safe and react with the steel

crucibles. Achieving reliable and safe refinement of the microstructure of Mg-Al

alloys is still an unsolved issue [23].

Recently, Qiu et al. [24] concluded that Al2RE compounds (RE = Rare Earth elements,

such as Gd, Sm, Y, Dy, Nd or Ce), which have crystallographic misfits with the Mg

matrix of less than 2%, are located in a ‘fully wetting zone’ and all have the potential

to act as effective grain refiners for cast Mg alloys. This was confirmed by adding 1 %

Al to a Mg-10Y alloy reducing the grain size from ~180 to ~36 μm [25]. Similar effects

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

4

were also found by Dai et al. [26], who reported that the Mg-10Gd-1Al alloy had a

fine grain size of ~37 μm. Recently, Wang et al [27] found that the grain size of a

Mg-6Sm alloy was reduced by 3 % Al from ~ 1830 μm to ~ 30 μm. Jiang et al. [28]

reported that the grain size of a Mg-6Ce alloy was significantly reduced by adding 3 ~

5 % Al.

Since there is still no commercially efficient method for producing grain refinement

in Mg-Al alloys, Al2RE based refiners for the Mg-Al system should be considered.

However, when REs were added to Mg-Al alloys to generate Al2RE phases, different

results were obtained by researchers. For example, while Yu et al. [29] reported that

0.4 % Ce significantly refined an AZ31 alloy, Pan et al. [30] pointed out that the range

of 0.5 ~ 1.5 % Ce significantly coarsened the AZ31 alloy. Chaubey et al [31]

investigated Mg-10Al-xCe alloys and claimed that 1 % Ce had a refining effect. Tong

et al. [32] reported that 1 % Nd slightly refined the AZ91 alloy, but Li et al. [33] found

that adding a Nd-containing mischmetal coarsened the grain size which increased

further as more mischmetal was added. Coarsening also occurred when Mg-Al-Y

alloys were investigated [34]. Li et al [35] reported that an addition of 1 ~ 3 % Gd to

an AZ31 alloy produced significant grain refinement. Liu et al [36] found that a

combined addition of 1 % Gd and Ca reduced the grain size of a Mg-7Al-1Si alloy, but

the grain size became larger after more Gd-Ca was added. It seems that the

experimental results from different researchers often contradict each other. In the

present paper, we investigate the effects of Sm on the microstructure and grain

refinement of the Mg-3Al alloy. Unlike the results reported above from the literature

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

5

where either refining or coarsening of the grain size by the addition of REs to Mg-Al

alloys, the present study observed both coarsening and refining depending on the

amount of RE addition. This study was designed to reveal the mechanisms

responsible for the observed change from grain coarsening to grain refining by the

addition of Sm.

2. Experimental details

Each ~ 3kg batch of Mg-3Al-xSm alloy (x = 0, 0.7, 1.4, 2.1, wt.%) was prepared by

melting pure Mg (99.7 %), pure Al (99.9 %) and Mg-25%Sm master alloy at 1003 K in

a steel crucible heated by an electric resistance furnace under a protective gas

mixture of CO2 and SF6. The melt was stirred for 5 min to ensure complete mixing of

the elements and then held for 10 min before sampling. A sand mould, which

provided a cooling rate of ~ 5 K/s, was used. A K-type thermocouple was placed at

the center of the mould before casting in order to obtain the cooling curves. The tip

of thermocouple was located at 10 mm from the bottom of the mould. Once the

alloy melt was cast into the mould, the cooling curves were recorded by a computer.

For each test, two identical conical samples (Φ40mm-Φ20mm×h50mm) were

obtained to ensure reproducibility. Chemical analysis by ICP-AES (Inductively Coupled

Plasma-Atomic Emission Spectroscopy) revealed that the actual compositions only

deviated slightly from the nominal compositions (Table 1). The specimens for

microstructural observation were obtained by cutting along the radial direction at ~

10 mm from the bottom of the conical samples, polished with grinding and polishing

papers and MgO solution until no significant scratches could be observed, and then

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

6

etched in a picric ethanol solution (30 g picric acid, 40 ml water, 60 ml acetic acid and

360 ml ethanol) for ~ 6 second. Optical microscopy (OM) and scanning electron

microscopy (SEM) were used to observe the microstructure. The accelerating voltage

of the SEM was set as 20 KV. The grain size of each alloy was measured according to

ASTM 112-96. Energy dispersive X-ray (EDS) and Electron backscatter diffraction

(EBSD) analyses were performed to determine the compositions of the different

phases present in the microstructure. The size distribution of the Al2Sm particles in

the Mg-3Al-2.1Sm alloy was measured on the SEM and OM images using image

analysis software (Image-Pro Plus) [26-27]. More than 15 images of ~ 0.6mm2 in area

were analyzed. The “Interdependence Theory” proposed by StJohn et al. [37] was

then used to describe the nucleation selection process. For this purpose, the

characteristics of the Al2Sm particles, such as the relationship between nucleation

undercooling (ΔTn), the average particle spacing (SΩ) and the particle size distribution,

were calculated according to the methods provided in [37].

3. Results

3.1 Microstructures

The microstructure of the Mg-3Al alloy contains α-Mg and a small amount of

β-Mg17Al12 (Fig. 1(a)). When 0.7 % or 1.4 % Sm was added to the Mg-3Al alloy, an

additional phase Al11Sm3 [38, 39] was observed (Fig. 1(b-c)). Most of the Al11Sm3

particles have an irregular and needle-like morphology and are located at the grain

boundaries. When 2.1 % Sm is added, both Al11Sm3 and Al2Sm phases [38, 39] are

formed and most of the Al2Sm particles have a blocky morphology and many of them

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

7

are located in the center of the grains (Fig. 1(d)). These two phases were identified

by EDS on at least ten locations as shown in Fig. 1(e): the ratio of Al to Sm at location

“1” marked in Fig. 1(e) is 2.2:1 which is very close to 2:1, and at location “2” the ratio

is 11.9:3 which is very close to 11:3. The presence of Mg in the EDS results is due to

the influence of the α-Mg matrix since the Al:Sm ratio remains relatively constant

while the proportion of Mg varies from 10% to 80%.

Besides α-Mg and Mg17Al12, Al-Fe-C-O particles, which are reported to be

heterogeneous nucleant particles naturally present in Mg-Al alloys [22], were also

found in the Mg-3Al alloy as in Fig. 1(a). Fig. 2(a) shows the EDS line-scanning results

of a typical native Al-Fe-C-O particle. Although it is difficult to identify the exact

chemical composition because of the small size of the particles, these particles were

found to be rich in Al, Fe, C and O compared to the adjacent α-Mg matrix. Similar

particles were also observed in all of the Sm-bearing alloys (Figs. 2(b-c)). However,

the EDS line-scanning results (Fig. 2(b)) show that the particles in the Sm-bearing

alloys contain not only Al, Fe, C and O, but also a high level of Sm. These particles are

uniformly distributed throughout the microstructure. At least 15 particles were

analyzed by EDS to confirm the existence of Sm.

3.2 Grain size

The relationship between the grain size of the Mg-3Al-xSm alloys and the Sm content

is shown in Fig. 3. Figs.3(a-d) are the OM images that show the changing trend of

grain size when the Sm content varies from 0 to 2.1 %. Fig. 3(d) has a higher

magnification clearly showing that many of the Al2Sm particles are at the center of

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

8

the grains. Fig. 3(e) presents a plot of grain size versus Sm content. The grain size of

the Mg-3Al alloy is ~250 μm. With 0.7 % Sm the grain size coarsens to ~1000 μm,

while the addition of 2.1 % Sm refines it to ~70 μm (Figs. 3(a-d)). It is interesting to

note that the relationship between Sm content and grain size is non-monotonic, i.e.

small amounts of Sm (0.7 % and 1.4 %) coarsens the Mg-3Al alloy grain size while a

larger amount of Sm (2.1 %) dramatically refines it. This relationship differs from the

results in the literature [29-34], where RE addition was reported to either refine or

coarsen the grain size of the Mg-Al alloys but not both.

4. Discussion

4.1 The grain coarsening effect of 0.7 % or 1.4 % Sm

As mentioned in the introduction, it is well accepted that the two key factors

promoting grain refinement are (1) solutes that cause rapid development of

constitutional supercooling (i.e. solutes with a high growth restriction factor, Q) and

(2) heterogeneous nucleant particles of high nucleation potency [5, 37, 40].

Correspondingly the grain size (dgs) can be expressed as [40]:

dgs = a + b/Q Eq. 1

where the slope b is related to the potency of the nucleant particles and a

corresponds to the maximum number of activatable nuclei. For a binary alloy Q can

be expressed as [41]:

Q=m C0 (k-1) Eq. 2

where m is the slope of the liquidus; k is the partition coefficient; C0 is the

concentration of the solute in the alloy. If there is more than one solute element

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

9

then the total value of Q can be estimated by summing the Q values provided by

each element [42]:

Qtotal=ΣmiCi(ki-1)i Eq. 3

The Q value for the Mg-3Al-0.7Sm alloy is calculated to be ~ 14 K, which is slightly

higher than for the Mg-3Al alloy at ~ 13 K. The increase of Q caused by the addition

of 0.7 % Sm would only generate a small degree of grain refinement according to Eq.

1, but coarsening would not be expected. However, the grain size of the

Mg-3Al-0.7Sm alloy is much larger than the Mg-3Al alloy. Hence, a in Eq. 1

dramatically increases and b is not a linear relationship indicating that the potency of

the heterogeneous nucleant particles present in the Mg-3Al alloy has been reduced

when Sm is added to the binary alloy.

According to the literature, the native Al-Fe-C-O particles can act as heterogeneous

nucleant particles in Mg-Al alloys [10-12]. Although Al-C-O (or Al4C3) particles are

usually considered to be the native heterogeneous nucleant particle, Al-Fe-C-O

particles will form when steel crucibles are used [11-12]. In the present study,

Al-Fe-C-O particles were found in the Mg-3Al alloy as shown in Fig. 1(a) and Fig. 2(a).

The EDS line-scanning results show that these particles are rich in Al, Fe, C and O,

which agrees well with the literature [11-12]. However, in the Sm-bearing alloys,

almost all of the similar particles contain not only Al, Fe and C but also a high level of

Sm (Fig. 3(b)). It seems that the Al-Fe-C-O particles in the Mg-3Al alloy have

transformed to Al-Fe-Sm-C-O particles after Sm addition. Despite the crystal type and

parameters of Al-Fe-Sm-C-O being unknown, it is reasonable to conclude that Sm

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

10

combines with the Al-Fe-C-O particles producing Al-Fe-Sm-C-O particles that have a

lower nucleation potency than the native nuclei in the Mg-Al alloy, resulting in

coarsening of the grain size of the Mg-3Al-0.7Sm and Mg-3Al-1.4Sm alloys. Note that

the grain size of the Mg-3Al-1.4Sm alloy is slightly finer than that of the

Mg-3Al-0.7Sm alloy. According to Eq. 1 and assuming the number of Al-Fe-Sm-C-O

particles is similar, this decrease in grain size could be due to the slightly higher

Q-value of the Mg-3Al-1.4Sm alloy (~ 15 K) than that of the Mg-3Al-0.7Sm alloy (~ 14

K).

4.2 The grain refining effect of 2.1 % Sm

When 2.1 % Sm is added to the Mg-3Al alloy, Al2Sm particles solidify prior to α-Mg

during solidification and the alloy is significantly refined. The solidification of Al2Sm is

confirmed by the solidification cooling curves of the Mg-3Al-0.7Sm and

Mg-3Al-2.1Sm alloys (Fig.4). The peak of the first derivative of the cooling curve at

~955.5 K in the Mg-3Al-2.1Sm alloy (Fig. 4(b)) indicates that the solidification of the

Al2Sm phase occurs prior to the nucleation and solidification of the α-Mg grains. In

contrast, there is no similar peak present in the derivative curve of the Mg-3Al-0.7Sm

alloy (Fig. 4(a)), indicating that Al2Sm does not solidify prior to α-Mg.

According to the Edge-to-Edge Matching (E2EM) model [24], the Al2Sm compounds

have the potential to become heterogeneous nuclei for the α-Mg phase and the

orientation relationship (OR) between Al2Sm and α-Mg can be described as follows:

[112]Al2Sm|| [21— —— —— —

1— —— —— —

0]α-Mg,

(11———————

0) Al2Sm || (01—— —————

10)α-Mg, (1—————— —

1———————

1)Al2Sm|| (0001)α-Mg

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

11

In this case, the misfit between the Al2Sm particles and the α-Mg matrix is ~ 0.45 %,

which is considerably lower than 2% indicating that Al2Sm has the potential to be a

potent heterogeneous nucleant for α-Mg. In the Mg-3Al-2.1Sm alloy, most of the

Al2Sm particles are at the center of the grains (Fig. 2(d)), indicating they are

heterogeneous nuclei for α-Mg. EBSD analysis (Fig. 5) confirms that the OR between

Al2Sm and α-Mg presented above can be reproducibly observed in the Mg-3Al-2.1Sm

alloy. Fig. 5(a) is an SEM image of the Mg-3Al-2.1Sm alloy in which three phases were

observed: α-Mg, polygonal Al2Sm and needle-like Al11Sm3 [25]. Figs. 5(b-c) show the

Kikuchi band patterns collected from the Al2Sm particle and its nearby α-Mg,

respectively. It is clear that the [21—— —— —— —

1—— —— —— —

]Al2Sm Kikuchi pole is very close to the [12—————— —

10]α-Mg

pole, and the (111)Al2Sm band is almost parallel to the (0001) α-Mg band. These results

agree well with the literature [24-25], indicating that the grain refining effect of 2.1 %

Sm is due to Al2Sm acting as the nuclei for α-Mg.

4.3 The nucleant selection process in the Mg-3Al-2.1Sm alloy

As mentioned above, a small crystallographic misfit between particles and the matrix

is an important prerequisite for particles being potent nucleants. Besides this, the

final grain size of an alloy is also determined by how many potent nucleant particles

will become active nuclei. In alloys where master alloys are added, typically in

aluminum alloys, grains nucleate on a very small proportion (1%-2%) of the added

particles while the vast majority (≥98%) do not contribute to grain refinement.

Recently, StJohn et al [37] explained this phenomenon through two aspects. Firstly,

the nucleation undercooling (ΔTn, which represents the undercooling needed to

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

12

activate nucleation) is inversely correlated with the particle size (Ω) [43]. This analysis

shows that most of the particles were of much lower potency compared to the

largest particles. Hence, ΔTCS (the amount of constitutional supercooling) that is

needed to trigger nucleation will be less for particles with the largest size. Secondly, a

Nucleation-Free Zone (NFZ) will immediately form around each new growing grain.

Since the constitutional supercooling within NFZ is always lower than at the end of

the NFZ zone, nucleation on all the potent nucleant particles existing within NFZ will

be suppressed. Thus, once a new grain nucleates on the largest particle, all the other

particles around it will not be able to be activated because of NFZ. That is to say, the

α-Mg grains will only nucleate on the largest particles.

In the present study, the total number density of Al2Sm particles (Nv) in the

Mg-3Al-2.1Sm alloy is 8.06×106 cm

-3. By comparing the number density of Al2Sm

particles with the number of grains calculated from the measured grain size of the

Mg-3Al-2.1Sm alloy, the proportion of the effective nuclei can be calculated to be ~

40%. This proportion is much higher than the 1 ~ 2% reported for the situation where

heterogeneous nucleant particles are added to other alloys by master alloy additions

[4]. Fig. 6 presents the distribution characteristics of the Al2Sm particles in the same

Mg-3Al-2.1Sm sample. Fig. 6(a) shows the distribution NΩ of the number of potent

nucleant particles for each value of particle size Ω. It can be seen that the majority of

particles have a size of 8 ~ 9 μm. As the particle size increases or decreases from 8 ~

9 μm, the number densities of these particles decrease. Fig. 6(b) shows the

relationship between nucleation undercooling ΔTn and Ω according to the

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

13

relationship given by the following equation [43]:

∆�� =��

∆��∙

Eq.4

where σ is the Solid-Liquid interfacial energy and ΔSV is the entropy of fusion. All the

related data can be found in [24]. Fig. 6(c) represents the relationship between the

average particle spacing SΩ and Ω, which shows an opposite trend to that in Fig. 6(a).

This means that as the number density increases for particles of a particular size, the

average particle spacing SΩ becomes smaller. By bringing together the information in

Figs. 6(a-c), Fig. 6(d) plots ΔTn against SΩ.

According to the Interdependence theory [37], the grain size dgs of a cast alloy

consists of two parts:

dgs = xnfz + xSΩ Eq. 5

where xnfz represents the length of NFZ where the amount of constitutional

supercooling is always lower than ΔTn for the most potent nucleant particles, and xSΩ

represents the average particle spacing (SΩ) of operative particles. This relationship is

illustrated in Fig. 7, where TE represents the equilibrium liquidus temperature and TA

represents the actual temperature of the melt. Note that SΩ is just one component of

dgs. Hence, SΩ will be smaller than the grain size dgs. In the present study, the average

spacing SΩ between the largest particles is about 120 μm. Assuming that only the

largest particles are operative, the grain size was predicted to be larger than 120 μm.

However, the grain size of the Mg-3Al-2.1Sm alloy is ~ 70 μm which is about half of

the predicted SΩ between the largest particles. This indicates that nucleation occurs

on more particles than all of the largest particles present in the alloy. However, the

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

14

three largest particle sizes have values of ΔTn within 0.05 K of each other (Fig. 6(d)).

Since there will be a temperature gradient in the melt due to heat extraction, any of

the particles in this size range could be activated. Hence, as shown in Fig. 7, instead

of the average spacing SΩ between the largest particles being 120 μm, the average SΩ

between any of the particles whose ΔTn is less than 0.2 K is ~ 60 μm which was

calculated by adding the particle densities for the three largest particle sizes and

converting this number to grain size. According to Fig. 4b the liquidus temperature

(TE) of this alloy is estimated to be ~ 916 K. When the amount of constitutional

supercooling (ΔTCS) reaches a value between 0.05 K and 0.2 K below 916 K, a particle

whose ΔTn is lower than 0.2 K will be activated. Those smaller particles whose ΔTn is

greater than 0.2 K are unlikely to be activated because they are further away from

the end of NFZ and more constitutional supercooling will be required to activate

nucleation. From this analysis the probability of the grain size being about 70 µm due

to a higher proportion of effective nuclei is a reasonable expectation. In addition, the

low value of ΔTn of less than 0.2 K would result in a small NFZ zone which also

promotes a fine grain size. Recalling that the length of NFZ and the length of SΩ (in

this case ~60 μm) together define the grain size, the analysis to create Fig. 7 implies

that NFZ would be about 10 µm in length.

An alternative explanation may be related to the fact that the Mg-Al-Sm alloy system

is a typical peritectic system [34]. The peritectic reaction during solidification is

Al2Sm+Liquid→Mg, which may also contribute to the observed grain refinement. As

Al2Sm is one of the reactants to generate α-Mg, the chemical reaction is favorable to

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

15

the nucleation of α-Mg on the surface of Al2Sm [26, 40, 41-44]. However, in normal

cooling situations the classic peritectic reaction hardly ever occurs [45]. Hence, more

work is still needed to understand this deeply.

4.4 Implications for achieving refinement during the solidification of Mg-Al-RE alloys

As mentioned in the Introduction, previous research observed either refining or

coarsening of the grain size in a range of Mg-Al-RE systems. Considering the results of

this study, a review of this research shows that refining occurs when Al2RE is formed,

or is expected to form, before the solidification of α-Mg due to higher concentrations

of RE elements [25-26, 34]. Coarsening is observed in alloys of low RE content in

AZ31 where Al2RE is unlikely to form. In this case coarsening is caused by the

combination of the RE element with the Al-Fe-C-O phase forming an Al-Fe-RE-C-O

phase dramatically reducing the nucleation potency of this phase. It is proposed that

the findings of this work can be generalized to other Mg-Al-RE systems because most

of the Al2RE phases have very small crystallographic misfits with α-Mg as shown in

Table 2 [24]. Further, the coarsening behavior may be influenced by the impurity

level, especially the Fe and C contents, and possible differences in the affinity

between different REs and Fe or C. The only contradictory evidence is that reported

by Yu et al. [29] where a 0.4 % Ce alloy refined the AZ31 alloy. However, Pan et al. [30]

found that for a range of 0.5 to 1.5 % Ce contents coarsening occurred in an AZ31

alloy. The results given in [30] are more credible since both authors claimed that no

Al2Ce was found in the AZ31Ce alloys. However, a definitive conclusion cannot be

made because the actual chemical compositions of the nominal alloys were not

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

16

provided in [29] and [30].

5. Conclusions

The addition of Sm to a Mg-3Al alloy has either a coarsening or refining effect on the

grain size depending on the addition amount. The addition of a small amount of Sm

(0.7 and 1.4 % Sm) leads to significant grain coarsening, which is proposed to be due

to a reduction in the nucleation potency of the Al-Fe-C-O particles when they

transform to Al-Fe-Sm-C-O particles. On the other hand, 2.1 % Sm refines the Mg-3Al

alloy because Al2Sm particles solidify prior to α-Mg and, due to a favorable

orientation relationship, are potent heterogeneous nucleation sites for α-Mg during

the solidification of the Mg-3Al-2.1Sm alloy. An analysis of the distribution of Al2Sm

particle sizes showed that there is a sufficiently high number density of large

particles with lower than 0.2 K nucleation undercooling (i.e. ~ 40% of the total

number of particles) to produce the measured grain size. It is proposed that these

findings can be generalized to all Mg-Al-RE alloys of a similar range of compositions.

Acknowledgements

This work is supported by National Natural Science Foundation of China (No.

51201103), SJTU Special Funds for Science and Technology Innovation (No.

13X100030018), the Program for Outstanding Academic Leader of Shanghai

(14XD1425000) and the Australian Research Council (ARC) Discovery Grant

(DP120101672). Dr. Dong Qiu from RMIT is gratefully acknowledged for the support

for the helpful discussion.

References

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

17

[1] E.O. Hall, Proc. Phys. Soc. B 64 (1951) 747-753

[2] N.J. Petch, Iron Steel Inst. 174 (1953) 25-28

[3] L. Lu, A.K. Dahle, D.H. StJohn, Scr. Mater. 54 (2006) 2197-2201

[4] R. Gunther, C. Hartig, R. Bormann, Acta Mater. 54 (2006) 5591-5597

[5] D.H. StJohn, A. Prasad, M.A. Easton, M. Qian, Metall. Mater. Trans. A 46 (2015)

4868-4885

[6] S.H. Park, J. Jung, J. Yoon, B.S. You, Mater. Sci. Eng. A 626 (2015) 128-135

[7] H. Sevik, S.C. Kurnaz, J. Magnesium Alloys 2 (2014) 214-219

[8] J.Y. Choi, W.J. Kim, J. Alloys Compd. 614 (2014) 49-55

[9] Y. Ali, D. Qiu, B. Jiang, F. Pan, M.X. Zhang, J. Alloys Compd. 619 (2015) 639-651

[10] E.F. Emley: Principles of Magnesium Technology. 1966, London: Pergamon Press

[11] M.X. Zhang, P.M. Kelly, M. Qian, J.A. Taylor, Acta Mater. 53 (2005) 3261-3270

[12] Y. Huang, K.U. Kainer, N. Hort, Scr. Mater. 64 (2011) 793-796

[13] D.H. StJohn, Q. Ma, M.A. Easton, P. Cao, Z. Hildebrand, Metall. Mater. Trans. A 36

(2005) 1669-1679

[14] L. Lu, A.K. Dahle, D.H. StJohn, Scr. Mater. 53 (2005) 517-522

[15] M. Suresh, A. Srinivasan, U.T.S. Pillai, B.C. Pai, Mater. Sci. Eng. A 528 (2011)

8573-8578

[16] L. Wang, Y.M. Kim, J. Lee, B.S. You, Mater. Sci. Eng. A 528 (2011) 1485-1490

[17] S. Liu, Y. Chen, H. Han, J. Alloys Compd. 624 (2015) 266-269

[18] M. Suresh, A. Srinivasan, U.T.S. Pillai, B.C. Pai, Proc. Eng. 55 (2013) 93-97

[19] S. Wang, C. Gan, X. Li, K. Zheng, W. Qi, Rare Metal Mater. Eng. 43 (2014)

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

18

2567-2571

[20] H.M. Fu, D. Qiu, M.X. Zhang, H. Wang, P.M. Kelly, J.A. Taylor: J. Alloys Compd.

456 (2008) 390-394

[21] D. Qiu, M.X. Zhang, J.A. Taylor, H.M. Fu, P.M. Kelly, Acta Mater. 55 (2007) 1863–

1871

[22] P. Cao, Q. Ma, D.H. StJohn, Magnesium Technology 2005. The Minerals, Metals

and Materials Society, Warrendale, PA, 2005, pp. 297

[23] D.H. StJohn, M.A. Easton, M. Qian, J.A. Taylor, Metall. Mater. Trans. A 44 (2013)

2935-2949

[24] D. Qiu, M.X. Zhang, J. Alloys Compd. 586 (2014) 39-44

[25] D. Qiu, M.X. Zhang, J.A. Taylor, P.M. Kelly, Acta Mater. 57 (2009) 3052-3059

[26] J.C. Dai, M.A. Easton, S.M. Zhu, G.H. Wu, W.J. Ding, J. Mater. Res. 27 (2012)

2790-2797

[27] C. Wang, J. Dai, W. Liu, L. Zhang, G. Wu, J. Alloys Compd. 620 (2015) 172-179

[28] Z. Jiang, B. Jiang, Y. Zeng, J. Dai, F. Pan, Mater. Sci. Eng. A 645 (2015) 57-64

[29] K. Yu, S. Rui, J. Song, W. Li, L. Guo, T. Nonferr. Metal. Soc. 18 (2008) s39-s43

[30] F. Pan, J. Peng, M. Yang, J. Chongqing Uni. 32 (2009) 363-366

[31] A. K. Chaubey, S. Scudino, K.G. Prashanth, J. Eckert, Mater. Sci. Eng. A 625 (2015)

46-49

[32] G. Tong, H. Liu, Y. Liu, T. Nonferr. Metal. Soc. 20 (2010) s336-s340

[33] S. Li, W. Zheng, B. Tang, D. Zeng, X. Guo, J. Rare Earths. 25 (2007) 227-232

[34] H.W. Chang, D. Qiu, J.A. Taylor, M. A. Easton, M.X. Zhang, J. Magnesium Alloys. 1

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

19

(2013) 115-121

[35] W.P. Li, H. Zhou, Z.F. Li, J. Alloys Compd. 475 (2009) 227-232

[36] J. Liu, W. Wang, S. Zhang, D. Zhang, H. Zhang, J. Alloys Compd. 620 (2015) 74-79

[37] D.H. StJohn, Q. Ma, M.A. Easton, P. Cao, Acta Mater. 59 (2011) 4907-4921

[38] B.R. Jia, L.B. Liu, D.Q. Yi, Z.P. Jin, J.F. Nie, J. Alloys Compd. 459 (2008) 267-273

[39] J. Wang, L. Wang, Y. Wu, L. Wang, Mater. Sci. Eng. A 528 (2011) 4115-4119

[40] M.A. Easton, D.H. StJohn, Metall. Mater. Trans. A 36 (2005) 1911-1920

[41] Y.C. Lee, A.K. Dahle, D.H. StJohn, Metall. Mater. Trans. A 31 (2000) 2895-2906

[42] M.A. Easton, D.H. StJohn, Acta Mater. 49 (2001) 1867-1878

[43] A.L. Greer, A. M. Bunn, A. Tronche, P. V. Evans, D. J. Bristow. Acta Mater. 48 (2000)

2823-2835

[44] Y. M. Kim, C. D. Yim, B. S. You, Scr. Mater. 57 (2007) 691-694

[45] D. H. StJohn, Acta Metall. et Mater. 38 (1990) 631-636

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

1

Table 1 Actual compositions of each alloy according to the ICP-AES results.

Nominal

compositions

Al (wt.%) Sm(wt.%) Fe(wt.%) Mg(wt.%)

Mg-3Al 2.93

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

1

Fig. 1 SEM images of (a) Mg-3Al, (b) Mg-3Al-0.7Sm, (c) Mg-3Al-1.4Sm, and (d) Mg-3Al-2.1Sm alloys;

(e) shows the EDS results for location 1 and 2 in Mg-3Al-2.1Sm alloy microstructure.

Al-Fe-C-O

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

2

Fig.2 SEM images of Mg-3Al-xSm alloy: (a) Fe- rich particle (Al-Fe-C-O) in Mg-3Al alloy and (b) Sm-Fe-

rich particle (Al-Fe-Sm-C-O) in Mg-3Al-0.7Sm alloy; (c) Sm-Fe- rich particle in Mg-3Al-2.1Sm alloy.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

3

0.0 0.7 1.4 2.10

200

400

600

800

1000

1200

Gra

in s

ize

(µm

)

Sm content (wt %)

Grain size(e)

Fig. 3 OM images showing the grain size of (a) Mg-3Al, (b) Mg-3Al-0.7Sm, (c) Mg-3Al-1.4Sm, and (d)

Mg-3Al-2.1Sm alloys. Note that (d) has a higher magnification to show that most Al2Sm particles are

located in the central region of the grains. (e) the grain size plotted against Sm content for all alloys.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

4

Fig.4 Solidification cooling curves and the first derivative curves of (a) Mg-3Al-0.7Sm alloy and (b)

Mg-3Al-2.1Sm alloy.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

5

Fig. 5 (a) SEM image of Mg-3Al-2.1Sm alloy; the corresponding EBSD Kikuchi band patterns of (b) Mg

matrix obtained at the pentagon marked in (a), and (c) its nearby Al2Sm particle obtained at the

square in (a).

0 2 4 6 8 10 12

2.0x105

4.0x105

6.0x105

8.0x105

1.0x106

1.2x106

1.4x106

1.6x106

1.8x106

2.0x106

2.2x106

Num

ber

of p

artic

les

(cm

-3)

Ω (µm)

Number of particles(a)

0 2 4 6 8 10 120.0

0.2

0.4

0.6

0.8

1.0

1.2

∆Tn

(K)

Ω (µm)

Ω-∆Tn(b)

0 2 4 6 8 10 12

60

80

100

120

140

160

180

200

SΩΩ ΩΩ ( µµ µµ

m)

Ω (µm)

d-SΩ(c)

0.0 0.2 0.4 0.6 0.8 1.0 1.2

60

80

100

120

140

160

180

200

SΩΩ ΩΩ (

µµ µµm)

∆∆∆∆Tn (K)

∆Τn-SΩ(d)

Fig. 6 Characteristics of Al2Sm particles in Mg-3Al-2.1Sm alloy. (a) the distribution NΩ of the number

of potent particles for each value of diameter Ω. (b) The relationship between nucleation

undercooling ΔTn and d. (c) Relationship between the average particle spacing SΩ and Ω. (d) ΔTn

plotted against SΩ.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

6

Fig. 7 A graph that illustrates the interrelationship between the development of constitutional

supercooling (∆∆∆∆TCS) between the equilibrium liquidus temperature TE and the actual temperature of the melt TA, and the distribution of particles for the range of particle sizes converted to their

nucleation undercooling (∆∆∆∆Tn-SΩ) that together establish the grain size of Mg-3Al-2.1Sm alloy.

MAN

USCR

IPT

ACCE

PTED

ACCEPTED MANUSCRIPT

Highlights

(1) The Mg-3Al alloy was significantly coarsened by the addition of 0.7 wt. % Sm, but

dramatically refined by 2.1 wt. % Sm.

(2) Sm coarsened the Mg-3Al alloy by inhibiting the Al-Fe-C-O particles nucleant potency.

(3) 2.1% Sm significantly refined the Mg-3Al alloy because of the solidification of Al2Sm

particles prior to the solidification of α-Mg.

(4) “Interdependence Theory” was applied to describe the nucleant selection process in the

Mg-3Al-2.1Sm alloy.