E

j m a t e r r e s t e c h n o l . 2 0 1 3;2(2):100–109

www.jmrt .com.br

Original article

Effect of a rotating magnetic field at the microstructure of anA354

Bruno Fragoso ∗, Henrique Santos

Faculdade de Engenharia, Universidade do Porto, Porto, Portugal

a r t i c l e i n f o

Article history:

Received 23 October 2012

Accepted 6 December 2012

Available online 14 June 2013

Keywords:

Electromagnetic stirring

Grain refinement

Thermal analysis

a b s t r a c t

In this piece of research the authors show how to build an experiment that can impose rotat-

ing magnetic stirring (RMF) during the solidification of a cylindrical cast with thermal data

acquisition. Several tests were carried out at increasing frequency and magnetic intensity.

The discontinuous solidification curves of the castings were acquired by a thermocouple

connected to a high resolution data recording system.

Thermal analysis, EBSD, quantitative metallography and DSC were used to characterize

the influence of RMF on solidification. Important changes were found in the microstruc-

ture, namely the quantity and size of eutectic phases, grain size within the eutectic cell,

remarkable differences in crystallographic disorientations and clustering of eutectic phases.

Aluminium A354

EBSD

DSC

© 2013 Brazilian Metallurgical, Materials and Mining Association. Published by Elsevier

Editora Ltda.

lation of liquid that causes cavitation and consequently some

Este é um artigo Open Access sob a licença de CC BY-NC-ND

1. Introduction

The use of magnetic fields during solidification has beenwidely used in continuous castings with success for grainrefinement, solute redistribution and surface quality improve-ment [1].

A domain of interest is the usage of magnetic fields insolidification of functional castings because there are nocontaminants added and sometimes, due to productive con-straints, it is not possible to impose adequate cooling ratesin order to achieve a desired grain size. Besides this, if the

mould material is brittle and other techniques are applied,rather than electromagnetic stirring, the vibrations could bedestructive for the mould.

∗ Corresponding author.E-mail: bruno.fragoso@fe.up.pt (B. Fragoso).

2238-7854 © 2013 Brazilian Metallurgical, Materials and Mining Association. Phttp://dx.doi.org/10.ste é um artigo Open Access sob a licença de CC BY-NC-ND

The interaction of a rotating magnetic field with liquidmetal can be explained by the induction of Eddy currentsand interference with the free electrical charges in the metal,resulting in a Lorentz force capable of moving the liquid mass.Details of this field of knowledge can be consulted elsewhere[2–5].

The mechanisms of grain refinement due to magnetic fieldsapplied during solidification are not yet well understood, butsome authors state that they are due to fragmentation of thedendrites’ primary arms that act as nuclei [6,7]. Others statethat they are due to pressure differences caused by the oscil-

nucleation of embryos [8–10]. Yet others claim that there is anundercooling effect due to the magnetic intensity [8]. Thereis some general agreement that the mass transportation of

ublished by Elsevier Editora Ltda.1016/j.jmrt.2012.12.001

o l . 2 0 1 3;2(2):100–109 101

tiv

tseow

bf

2

2

A0wt

itfihtTmtmv

8vfis

1tad0

ntiwb

Fig. 1 – RMF experimental assembly. (a) Cooling case; (b)Inductor coil; (c) Gypsum mould; (d) Test piece location; (e)Thermocouple for thermal analysis.

300

250

200

150

150 200 250 300 350 400 450

Time (minutes)

Tem

pera

ture

(ºC

)

100

100

50

500

0

Fig. 2 – Thermal cycle for the gypsum drying: The first stephas a ramp of 5 ◦C/min. and a dwell time of 240 min; thesecond step has a ramp of 3 ◦C/min. and a dwell time of

j m a t e r r e s t e c h n

he cooler liquid from the mould surface to the hotter regionsncreases the solidification rate [8]. Mass transportation pro-ides an opportunity to solute undercooling [6,7].

Another effect is the appearance of segregation areas ashe magnetic stirring is intensified [11–15]. In cases of RMFtirring, the segregation is located in the core of the vortex,ven with higher density solutes. This has a particular effectn aluminium alloys with higher silicon and copper content,hich is the case in this experimental study [4,14].

Besides the effect on grain size and segregation distur-ances, porosity may arise due to high central shrinkage. Thiseature is associated with severe macrosegregation [16,17].

. Materials and methods

.1. Materials

n electrical three-phase motor inductor from SEW® with.5 kW and with phase angle (ϕ) of 120◦ was adapted in such aay as to supply different frequencies and magnetic intensi-

ies.The inductor was controlled by a low-voltage frequency

nverter MICROMASTER® from Siemens®. Six high-flow elec-rical fans were used to cool the inductor. The magneticeld inside the inductor was characterized vertically andorizontally in several frequencies and magnetic intensi-ies for different locations with a Hall sensor from Allegroechnologies®. Due to the Hall’s sensor’s linear and ratio-etric characteristics, which have a high response speed,

he magnetic field was determined using an oscilloscope toeasure the peak amplitude in millivolts, the latter being con-

erted afterwards to Gauss and finally to millitesla.A high resolution low-voltage data acquisition system

018R from Superlogics® was used together with DASYLab®

11 to acquire the temperature during the test piece solidi-cation at a rate of 2 Hz. The assembly of the experiment ishown in Fig. 1.

The cast specimens were 30 mm in diameter (Ø) and10 mm in height and were cooled down in such a way thathe main heat was lost through the top. This was done in

pre-dried cylindrical gypsum mould according to a cycleescribed in Fig. 2. The gypsum thermal conductivity was.18 ± 0.03 W m−1 K−1.

Four different levels of frequency with proportional mag-etic intensity were tested. Each of these batches contained

hree samples for reproducibility assessment. From here onn the text, “frequency” will mean the imposed frequency

ith a correspondent magnetic intensity. The correlationetween frequency and magnetic intensity as well as the

Table 1 – Correlation between frequency and magnetic intensity

Bath Frequency (Hz) B max (

Coil wall

1 0 02 5 73 25 244 50 39

120 min.

corresponding Taylor number (Ta) and Reynolds magnetic

number (Rew) are displayed in Table 1.

The chosen position for the thermocouple was at Ø/4 and25 mm in depth. This was to avoid mechanical vibrations

(B) for the experiments.

mT) Ta Rew

Cavity wall

0 – –5 5.01 × 101 2.16 × 10−1

19 3.20 × 103 1.08 × 100

30 1.56 × 104 2.16 × 100

102 j m a t e r r e s t e c h n o l . 2 0 1 3;2(2):100–109

Table 2 – Chemical composition in wt.% of the cast alloyA354.

Cu Mg Si Ti Al

1.62 0.43 9.21 0.07 Rest

620

DSC cycle

Area ofinterest

520

420

320

220

120

20

0 2000 4000 6000

Time (s)

Tem

pera

ture

(ºC

)

8000

interfering with the detected temperature because the ther-mocouple was 1.6 mm in diameter. With the thermocouplealready in position, the alloy A354 (composition in Table 2)was poured in an open atmosphere at 670 ± 5 ◦C and themagnetic field was activated immediately after filling. Themagnetic field was deactivated when the sample reached400 ◦C. The samples were cut longitudinally and prepared withfine polishing. In the final step of polishing a colloidal silicasuspension from Buehler® and an aqueous solution of NaOHat 8% were used to remove the deformed aluminium matrix.Please note that all the analyses were undertaken withouthomogenization heat treatment.

2.2. Characterization techniques

2.2.1. EBSDOne portion with 100 mm2 from one trial without RMF andanother with RMF at 50 Hz were removed by a hand saw fromthe cast samples.

The EBSD analysis was carried out in a FEI Quanta 400FEGESEM/EDAX Genesis X4M.

Two different observation fields were analyzed per eachsample. The first observation field was analyzed at low mag-nification for a global view of crystallographic orientation andgrain size of primary ˛. The second field was scrutinized athigher magnification and located inside a eutectic envelope.The intention was to evaluate the crystallographic orientationand thus analyze the epitaxial growth of eutectic ˛ (˛eu).

Aluminium and silicon crystallographic disorientationsand grain sizes were determined by means of an OIMTM Anal-ysis System.

2.2.2. Thermal and DSC analysisThe cooling curves were processed by means of an algorithmthat was able to reduce the data noise and also calculate thefirst derivative (T is the temperature in ◦C and t is the time inseconds). These data, associated with the studies performedby other authors [18], enabled the identification of the mainsolidification events.

Cylindrical samples with a diameter of 3.7 mm weremachined from the cast. The samples were cut and groundto 1000 mesh in water-cooled sand paper in order to provide agood surface contact between the sample and the bottom ofthe pan. The samples were 70 ± 5 mg and the DSC equipmentused was a Labsys TG/DTA from Setaram.

The thermal cycle for DSC analysis is shown in Fig. 3. Theregion of interest in DSC was determined to be between 450 ◦Cand 540 ◦C with the scan rate of 3 ◦C min−1.

The DSC cycle was adaptive with the following goals inmind: avoid diffusion of the studied phases during heat-ing cycle and aptitude to differentiate each peak during

Fig. 3 – Thermal cycle for DSC analysis.

the respective phase melting. These goals are apparentlyin conflict because non-diffusion requires a high heatingrate and peak separation requires a slow heating rate. Sothe authors have decided to impose a DSC cycle with anintermediate/high-heating rate prior to the expected incipientmelting temperatures that occurred above 480 ◦C. The authorsdecided to impose a 20 ◦C min−1 heating rate before the incip-ient melting and to reduce the heating rate to 3 ◦C min−1 at480 ◦C, temperature where incipient melting occurred. But theheating rate change from 20 ◦C to 3 ◦C min−1 caused a 30 ◦Cthermal instability; as the authors are studying transforma-tions occurring above 480 ◦C, the temperature for the heatingrate change has been decreased to 450 ◦C, allowing a clean DSCcurve above 480 ◦C.

The upper limit of the region of interest has to do with theeutectic reaction Al(eutectic) + Si(eutectic) → L. This endother-mic reaction begins to bend the heat flow curve at 535 ◦C andno other endothermic peaks are detected.

The enthalpy of the liquid phase was calculated by the inte-gration of the endothermic peaks in the software provided bySetaram (Data Processing Module v1.53e).

2.2.3. Quantitative metallographyMicro structural pictures were captured with a total of≈30 mm2 at 100× magnification.

The quantitative analysis was carried out using ImageJ andstatistical analysis was performed in MS Excel®. Two types ofparticle population were characterized: silicon particles andcopper/magnesium rich interdendritic colonies. The featuresanalyzed were: area fraction, shape factor, specific perimeter,quantity of particles per mm2 and average size.

2.2.4. Solute segregation analysisThe samples were etched with Weck’s tint etching, whichcreates microstructural contrast due to crystal orientationand solute segregation. Within a certain field of view, where

the crystal orientation is constant, contrast differences maybe interpreted as a consequence of solute microsegrega-tion.

j m a t e r r e s t e c h n o l . 2 0 1 3;2(2):100–109 103

0

8

7

6

5

4

3

2

1

0

20

40

60

80

100

−20 −15 −10 −5 0 5 10 15 20

0

205.5

4.5

2.5 11.5 2

33.5

5

3633

21 15 63

0

9 1218 2427

39

30

4

35

30

25

20

15

10

5

40

60

Horizontal |B|Vertical |B|

Distance to axis (mm)Distance to axis (mm)

Inte

nsity

|B| (

mT

)

Hei

ght (

mm

)

Hei

ght (

mm

)

Inte

nsity

|B| (

mT

)

80

100

−20 −15 −10 −5 0 5 10 15 20

Fig. 4 – Isomagnetic intensity distribution of the vertical and horizontal axes for a frequency of 50 Hz. The shaded arearepresents the cast specimen.

3

3

Amcematfia

fim

odfjtbcm

stmhw

3

Ts

. Results and discussion

.1. Magnetic field characterization

s can be seen in Fig. 4, the vertical magnetic intensity isainly concentrated on the inductor edge. In the geometrical

entre the intensity is null. On the other hand, as should bexpected from a coil designed for a rotational motor, the mainagnetic intensity was located in the middle of the height

nd near the inside surface of the stator core for the horizon-al field. It was also found that the intensity of the horizontaleld is much greater than the vertical one, although both fieldsre zero at the rotation axis.

From these results it is possible to infer that the magneticeld imposed on the casting represented by a shaded area isainly horizontal.This is supported by the fact that the liquid metal does not

ccupy the whole area left by the rotor. In fact, due to practicalesign considerations, the liquid metal is located a little awayrom the stator core and also from the top and bottom. Theustification for this is that this type of assembly may overheathe inductor and so the heat must be properly isolated. Thiseing the case, the authors decided to use a very low thermalonductivity material (gypsum) and an air pocket between theould and the stator core.It is possible to assume that the metal flow was axially

table because, according to Grants and Gerbeth [19], the rela-ion between the radius “R” and the height “h” of the casting

ust obey the criterion 0.4 ≥ R/h > 1, Rew > 1100 and Ta must beigher than 1.635 × 105 to be non-stable. All these conditionsere fulfilled, as can be seen in Table 1.

.2. EBSD

he EBSD mapping is shown in Fig. 5. The picture ordering isuch that the main crystallographic differences between the

sample without RMF and the one with RMF are easy to follow.The rows are ordered in the following way:

1st row – SEM IMAGE;2nd row – EBSD mapping for alpha phase;3rd row – EBSD mapping for Si phase.

On the other hand, the columns represent the two RMFsituations at two different locations and magnifications:

Column one and two are from the sample without RMF andcolumn three and four are from the sample with RMF.

As stated in Fig. 5, if one compares EBSD maps “e” with “g”,two main aspects must be enhanced: the grain size reductionof the primary phase with RMF; the grain orientation with RMFis more diversified. A quantitative analysis of the grain sizedistribution of primary ˛ is shown in Fig. 6, revealing that inthe situation of RMF at 50 Hz the grain size is smaller thanwithout RMF. In fact, the largest grain size with RMF at 50 Hzis about 440 �m, while without (RMF) it is about 650 �m. Inaddition to this fact, with RMF at 50 Hz, a larger fraction ofsmaller grains appear at about 200 �m. The cumulative sumof the grain fraction for the different grain diameters showsthat the main difference occurs at 200 �m, and from this pointon RMF achieves the complete area fraction for smaller grainsizes.

Further investigations in the eutectic areas show that theeutectic alpha (˛eu) at 50 Hz is approximately 50% smaller thanwithout the magnetic field. Here the crystal orientations withRMF are more diverse than those without RMF. These largeorientation differences suggest that with RMF at 50 Hz the ˛eu

grains do not have an epitaxial growth as most of the ˛eu grainswithout RMF seem to have. This statement is also valid for theeutectic silicon.

The eutectic silicon also exhibits some changes when we

compare the cases without RMF and those with RMF at 50 Hz.As shown in Fig. 5j and l, the main difference between somehigher crystal disorientation and the grain size at Fig. 5d is

104 j m a t e r r e s t e c h n o l . 2 0 1 3;2(2):100–109

a

Global view (no RMF)

400 μm

400 µm

400 µm

400 μm

400 µm

400 µm

100 μm

100 µm

100 µm

100 μm

100 µm

100 µm

α grain orientation of “a” α grain orientation of “b”

Si grain orientation of “a” Si grain orientation of “b” Si grain orientation of “c” Si grain orientation of “d”111

101001

α grain orientation of “c” α grain orientation of “d”

Global view (50Hz)Detail on eutectic region of “a” Detail on eutectic region of “c”

b c d

e f g h

i j k l

gnifications; (e), (l) EBSD crystal orientation maps.

No RMF

1,00

0,80

0,60

0,40

0,20

0,000 200 400 600 800

50 Hz

Cumulative sum (No RMF)

Cumulative sum (50 Hz)

Are

a fr

actio

n

Fig. 5 – (a), (d) SEM micrographs at several ma

finer than the one shown in Fig. 5b. No evidence was found tosupport the loss of epitaxy when RMF is applied.

3.3. Thermal analysis

There are significant changes in the first derivatives of thethermal analysis between batches.

This is particularly evident in Fig. 7 if one compares theexperiment without RMF with the one with 50 Hz.

Without RMF there is a formation of a phase while ˛ grows.This event appears after ˛ growth temperature. The authorsundertook two more tests without RMF but with interruptedsolidification at the point “Quench” in Fig. 7. The interrup-tion was performed by quenching the casting in cold waterwhile it was monitored with the same device for thermal anal-ysis. The metallographic results revealed that the precipitantphase is primary silicon on the castings without RMF. Themicrostructure on the castings with RMF did not show anyprimary silicon. The primary silicon is shown in Fig. 8.

The precipitation of primary silicon in hypoeutectic Al–Sialloys has already been reported by Wang et al. [20], whoattribute this fact to local concentrations of Si higher thanthe eutectic concentration. High local concentrations are due

Grain diameter (μm)

Fig. 6 – Primary ˛ grain diameter distribution.

j m a t e r r e s t e c h n o l . 2 0 1 3;2(2):100–109 105

0,050

0,025

0,0001000 200

a b c

1 2

Without RMF

With RMF at50 HzTime (s)

dT.d

t–1 (º

C.s

–1)

α growth temperature

300 400 500

Quench−0,025

−0,050

−0,075

−0,100

−0,125

−0,150

Fig. 7 – First derivative of the cooling curves for thee

tltbmsttat

cFa

Fq

RMF RMF off before 1st peak

RMF off after 2nd peak

50 100 150 200 250 300 400350

Time (s)

RMF off between peaks

Off

Fig. 9 – Thermal analysis for the experiments with RMFstopping.

xperiments without RMF and with RMF at 50 Hz.

o the silicon maximum solubility of 1.5 ± 0.1 at% at 5 77.6 ◦Ceading to the rejection of solute by the primary aluminium athe solid/liquid interface. The authors suppose that the inhi-ition of primary silicon precipitation is due to the intenseass transportation at the solid front, avoiding the build-up of

olute. Some authors have already reported the intense soluteransportation from the solid front [21,22]. Besides the precipi-ation inhibition of primary silicon, the experiments with RMFt 50 Hz revealed a very interesting fact, i.e. the occurrence ofwo new peaks while ˛ is growing.

For further analysis of these facts, experiments at 50 Hz

onsisted in stopping the RMF at “a”, “b” and “c” (detail inig. 7), and the resultant thermal analyses were undertakennd evaluated.

ig. 8 – Primary silicon surrounded by primary ˛ in theuenched sample casting without RMF.

The thermal analysis curves of the situations a, b and cfrom Fig. 7 are shown in Fig. 9, where stopping of the RMFbefore peak 1 leads to strong variations in dT/dt.

Some seconds after the RMF has been stopped, there is anincrease in the cooling rate followed by a decrease, increas-ing again and finally decreasing. These last two seem to bereplicates of the first ones, but much more intense.

This phenomenon becomes progressively weaker as thesolidification takes place.

The authors suppose that peaks 1 and 2 in Fig. 7 are associ-ated with the high mass/heat transfer during the RMF stirringthat may melt some dendrites, and that when the bulk reachesa certain temperature there is an abrupt nucleation of ˛, lead-ing to new peaks of recalescence. Dendrite fragmentation hasalready been reported by others in electromagnetic stirring[23].

3.4. DSC analysis

Two curve examples of distinct experiment pieces are depictedin Fig. 10. One feature that immediately stands out is the lossof peak 1 for the case with RMF. The other aspect for this case

is the drift to higher temperatures and the “cleaner” overallappearance of the curve.

480 490 500

Temperature (ºC)

Hea

t flo

w (

arbi

trar

y)

510 520 530

No RMF

50 Hz

1 2 3 4

Fig. 10 – DSC curves from the experiments without RMFand with RMF at 50 Hz. The peaks found are numbered.

106 j m a t e r r e s t e c h n o l . 2 0 1 3;2(2):100–109

Table 3 – 95% confidence DSC peak temperatures and enthalpies.

Without RMF 50 Hz

Peak top (◦C) Enthalpy (J g−1) Peak top (◦C) Enthaply (J g−1)

Peak 1 491.2 ± 0.3 0.25 ± 0.1 n.a. n.a.Peak 2 495.6 ± 0.6 3.746 ± 1.9 497.6 ± 0.2 3.731 ± 1.3

0.31.6

Peak 3 506.6 ± 0.5 0.746 ±Peak 4 523.7 ± 0.7 2.927 ±

This can be further clarified in Table 3. Peak 1 has beendetected at 491.2 ◦C without RMF. It seems that the phase thatis formed at this temperature is suppressed with RMF. Anotherpoint is that the enthalpies remain more or less the same withthe RMF imposition, except for peak 3, where the amountsof this phase more than doubles. For peak 4 there is a slightdecrease of 14%. It seems, then, that while the solidificationoccurs, the growth of some phases is promoted and at theend of the solidification there is not enough solute for otherphases to form. The peak temperatures for RMF are very closeto those without RMF, albeit being shifted to values approxi-mately 2 ◦C higher. The only exception for this is peak 4. Theauthors suppose that the particles with RMF have a slightlydifferent stoichiometry.

If one analyses the data dispersion for the tests conducted(three for each case), it is clear that with RMF the dispersionis lower for most of the data.

3.5. Quantitative metallography

The quantitative metallography analysis focused on two maingroups of particles, one group being the Si particles and theother consisting of all the others, except for the iron rich par-ticles.

Fig. 11 shows that the quantity of silicon particles increaseslinearly in line with the root of the RMF frequency. With-out RMF the number of particles is approximately 245/mm2

while at 50 Hz it is about 320/mm2, representing an increaseof 30%. On the other hand, the average size of particles also

200 350

325

300

275

250

225

2000 2 4 6

y = −8,61x + 174,63R2 = 0,92

y = 11,08x + 244,11R2 = 0,91

8

Quantity of particles per mm∧2

Ave

rage

siz

e (μ

m)

Am

ount

of p

artic

les

per

mm

2

Average size

Frequency1/2 (Hz)

180

160

140

120

100

Fig. 11 – Amount of Si particles/mm2 and average sizedistribution as a function of RMF frequency.

508.8 ± 0.3 1.624 ± 0.4524.5 ± 0.7 2.523 ± 0.4

decreases with the square root of the RMF frequency. Both ofthese attributes have a reasonable correlation coefficient.

Another interesting feature is the average specific perime-ter (ASP) of the silicon particles. ASP was calculated by dividingthe perimeter by the area of each particle. As stated in Fig. 12,ASP and its dispersion grow linearly with the root of the fre-quency. Without any RMF, ASP is about 0.43 �m−1, while at50 Hz it is 0.52 �m−1, representing an increase of 20%. Thisphenomenon might be interesting with regard to the solutionheat treatment, since the rate of dissolution depends stronglyon the specific area of particles to be solutionized.

The behaviour of the area fraction occupied by the siliconparticles is demonstrated in Fig. 13 and is shown to be morecomplex than the previous features mentioned. When an RMFof 5 Hz is applied, there is an instant reduction of area fractionthat is stable up to 25 Hz, but for 50 Hz this apparent plateauis broken, so as to further area fraction diminishment. Thearea fraction without RMF is about 5.7%, while with an RMFat 50 Hz this drops to 4.9%, representing a fall of about 15% ofthe area fraction occupied by the silicon particles. The shapefactor (SF) of silicon particles was calculated by the division ofthe largest by the smallest length of each silicon particle. SFremains unaltered throughout the RMF frequency and withvery small dispersion.

From all the previous data it is possible to conclude thatas RMF increases, the quantity and ASP of silicon particlesincrease. On the other hand, as RMF increases the size andarea fraction occupied by the Si particles decreases. The aspectof note is that SF remains the same despite all these changes.

The procedure for quantitatively analysing the remainingparticles required that they were etched with an aqueoussolution of NaOH at 10% by 70 ◦C. This etching darkened the

0,60

0,55

0,50

0,45

0,40864

Frequency1/2(Hz)

AS

P (

μm

−1 )

20

y = 0,01x + 0,44R2 = 0,94

Fig. 12 – Average specific perimeter (ASP) of the siliconparticles as a function of RMF.

j m a t e r r e s t e c h n o l . 2

4.0

3.5

3.0

2.5

2.00 2 4 6

Frequency1/2(Hz)

84.0

4.5

5.0

Area fraction

Shape factor

Sha

pe fa

ctor

Are

a fr

actio

n (%

)

5.5

6.0

Fig. 13 – Area fraction and shape factor of silicon particlesa

pmpwgia

s

lpT1rs

Ff

s function of RMF frequency.

articles enriched in copper and magnesium, thus enabling aore accurate quantitative determination of these secondary

articles. These same particles appeared as agglomerates andere considered to be colonies. Owing to the highly complex

eometry of these colonies, only their silhouette was takennto account. The analyses that followed only focused on therea fraction and shape factor.

The area fraction and shape factor from the colonies arehown in Fig. 14.

The area fraction of the colonies decreases remarkably in ainear manner as the RMF frequency increases. The data dis-ersion tends to increase in line with the increasing frequency.he area fraction occupied by these colonies begins at about

.5% in the absence of RMF, while at 50 Hz it is about 0.5%. Thisepresents a reduction of approximately 65%. Once again thehape factor seems to be unaffected by the RMF and shows the

Area fraction

2.50

2.00

1.50

1.00

0.50

0.000 10 20 30 40 50 60

Frequency (Hz)

Are

a fr

actio

n (%

); sh

ape

fact

or

Shape factor

y = −0.021×+1.4795R2 = 0,89

ig. 14 – Area fraction and shape factor of colonies as aunction of RMF frequency.

0 1 3;2(2):100–109 107

same magnitude as that of the silicon particles. In sum, thereare strong evidences that more solute elements dissolve. Inthe matrix with magnetic field and the undissolved particleshave a more suitable size and surface area for solution heattreatment. It is thus expectable that heat treatments may beshortened leading to energetic savings. More work must bedone on this field to check these expectations.

3.6. Solute segregation analysis

After etching one sample per RMF situation with Weck’setchant, it had become clear that substantial differencesare imposed by the RMF. This is supported by Fig. 15. InFig. 15a (no RMF) the interdendritic spaces show a severesegregation effect resembling a halo. This halo effect is onlyperceived in the particles rich in copper and magnesium. Onthe other hand, the silicon particles reveal a shallow haloand are somewhat uniformly distributed. Fig. 15b shows thatthe RMF begins to affect the segregation level and the dis-tribution of silicon particles. The segregation halo aroundcopper/magnesium-rich particles begins to be more discreteand also narrower. However, the silicon particles show aslightly broader halo and begin to appear in clusters. AtFig. 15c the clustering effect of the silicon particles begins tobe more evident and the parallelism between secondary den-drites begins to be less evident. At 25 Hz (Fig. 15c) the halothickness at the silicon particles is broadly the same as it isat 5 Hz, but the clustering of silicon particles evolves to widerdiameters. At this frequency the dendritic structure is almostannihilated to give place to a globular primary ˛. At Fig. 15d avery large area of clustered silicon is shown. Besides this fea-ture, the halo around the silicon particles has grown to valuesthat reveal that in some cases the halo is broader than the sil-icon particle width. The primary ˛ has now lost the dendriticstructure to assume a shape of somewhat irregular globules.The Si cluster evolution with the increase in RMF frequencyalso displays interesting behaviour at the RMF stopping casesdemonstrated in Figs. 7 and 9.

In Fig. 15e one can observe some cluster formation andpreservation of the dendritic structure as well as noting thatsegregation in the interdendritic spaces prevails. After the firstsecondary recalescence peak there is an abrupt formation oflarger Si clusters that are agglomerated between them, sepa-rated by a small amount of primary ˛.

Interesting to note is the diminution of the segregationhalos amongst the copper/magnesium particles. These fea-tures can be observed at Fig. 15f. After the second secondarypeak the clusters are fully agglomerated. The apparent maindifference between the agglomerates of Fig. 15d and g is theirsize and the slightly larger segregation halos around the cop-per/magnesium rich particles.

The authors suppose that the RMF can only affect the sizeof the silicon particles when it is activated at the nucleationand growth of the eutectic phases.

4. Conclusions

• There is grain refinement with the appliance of RMF, bothin the primary ˛ and in the eutectic ˛ and Si;

108 j m a t e r r e s t e c h n o l . 2 0 1 3;2(2):100–109

a b c d

e f g

No RMF 5 Hz

100 μm

100 μm 100 μm 100 μm

100 μm 100 μm 100 μm

50 Hz - stop “a” at figure 7 50 Hz - stop “b” at figure 7 50 Hz - stop “c” at figure 7

25 Hz 50 Hz

evera

r

Fig. 15 – Sensitive segregation tint etching micrographs at s

• The growth of eutectic ˛ is affected by the magnetic field,losing its epitaxial growth with primary ˛ and becomingfiner;

• With the use of RMF there are phases which are suppressedand in which segregation in the solid front is diminished,for instance primary Si;

• The aluminium matrix has higher content in alloyingelements because the precipitation of silicon and otherintermetallic phases is less intense;

• The size of the eutectic colonies is highly dependent on theRMF power and on the solidification stage moment whenRMF is switched off;

• At higher RMF frequencies, the liquid pockets where theeutectic reaction occurs appear to be bigger;

• The interdendritic segregation appears to be lower at higherRMF frequencies;

• The interlamellae silicon segregation increases as RMF fre-quency increases.

Conflicts of interest

The authors declare no conflicts of interest.

Acknowledgements

This work was supported by Fundacão para a Ciência e Tec-

nologia (Portuguese Foundation for Science and Technology)and Zollern & Comandita through SFRH/BDE/33673/2009 PhDgrant (financed by QREN – POPH – type 4.1).

l RMF frequencies. The arrows highlight eutectic colonies.

e f e r e n c e s

[1] Davidson PA. Magnetohydrodynamics in materialsprocessing. Annu Rev Fluid Mech 1999;31:273–300.

[2] Boden S, Eckert S, Gerbeth G. X-ray visualisation ofsolidifying GaIn-alloys in the presence of melt convection.In: EPM 2009, 6th International Conference onElectromagnetic Processing of Materials. 2009.

[3] Iwai K, Wan J, Kohama T. Solidified structure control ofmetallic alloy using electromagnetic field. In: EPM 2009, 6thInternational Conference on Electromagnetic Processing ofMaterials. 2009.

[4] Steinbach S, Wille A, Fleischer T, Ratke L. The effect of fluidflow on the dendritic microstructure in two AlSiCu alloys. In:EPM 2009, 6th International Conference on ElectromagneticProcessing of Materials. 2009.

[5] Yang Y, Wang B, Ma X, Tong W, Li Y, Luo T, et al. Crystalgrowth and grain refinement by pulsed magnetic field. In:EPM 2009, 6th International Conference on ElectromagneticProcessing of Materials. 2009.

[6] Yasuda H, Nakatsuka N, Nagira T, Sugiyama A, Yoshiya M,Uesugi K, et al. X-ray imaging study on grain refinement dueto dendrite fragmentation. In: EPM 2009, 6th InternationalConference on Electromagnetic Processing of Materials.2009.

[7] Sugiura K, Iwai K. Refining mechanism of solidified structureof alloy by electromagnetic refining process. ISIJ Int2005;45(7):962–6.

[8] Jie D, Jianzhong C, Wenjiang D. Theoretical discussion of theeffect of a low-frequency electromagnetic vibrating field onthe as-cast microstructures of DC Al–Zn–Mg–Cu–Zr ingots. JCryst Growth 2006;295(2):179–87.

[9] Yasuda H, Toh T, Iwai K, Morita K. Recent progress of EPM in

steelmaking, casting, and solidification processing. ISIJ Int2007;47(4):619–26.

[10] Robles Hernandez FC, Sokolowski JH. Comparison amongchemical and electromagnetic stirring and vibration melt

o l . 2

j m a t e r r e s t e c h n

treatments for Al–Si hypereutectic alloys. J Alloys Compd2006;426(1–2):205–12.

[11] Zhang WQ, Yang YS, Liu QM, Zhu YF, Hu Z. Structuraltransition and macrosegregation of Al–Cu eutectic alloysolidified in the electromagnetic centrifugal casting process.Metall Mater Trans A 1998;29(1):404–8.

[12] Metan V, Eigenfeld K, Räbiger D, Leonhardt M, Eckert S. Grainsize control in Al–Si alloys by grain refinement andelectromagnetic stirring. J Alloys Compd2009;487(1–2):163–72.

[13] Steinbach S, Ratke L. The influence of fluid flow on themicrostructure of directionally solidified AlSi-base alloys.Metall Mater Trans A 2007;38(7):1388–94.

[14] Ratke L, Genau A, Steinbach S. Flow effects on the dendriticmicrostructure of AlSi-base alloys. Trans Indian Inst Met2009;62(4):337–41.

[15] Yasuda H, Yamamoto Y, Nakatsuka N, Nagira T, Yoshiya M,Sugiyama A, et al. In situ observation of nucleation,fragmentation and microstructure evolution in SnBi andAlCu alloys. Int J Cast Met Res 2008;21(1–2):125–8.

[16] Kapusta A, Tilman B. Experimental study of spatial

distribution of the amplitude of pressure waves excited bypinch-effect in long conical vessels of circular and squarecross-sections. In: EPM 2009, 6th International Conferenceon Electromagnetic Processing of Materials. 2009.

0 1 3;2(2):100–109 109

[17] Dardik I, Kapusta A. Method of axial porosity eliminationand refinement of the crystalline structure of continuousingots and castings. US Patent 7661456 B2, February 16,2007.

[18] Malekan M, Shabestari SG. Computer-aided coolingcurve thermal analysis used to predict the quality ofaluminum alloys. J Therm Anal Calorim 2011;103(2):453–8.

[19] Grants I, Gerbeth G. Stability of axially symmetric flowdriven by a rotating magnetic field in a cylindrical cavity. JFluid Mech 2001;431:407–26.

[20] Wang SR, Wang YZ, Yang LY, Ma R, Wang Y. Soluteredistribution resulting in growth of primary silicon in casthypoeutectic Al–Si alloys. Appl Mech Mater2011;121–126:367–71.

[21] Boden S, Eckert S, Gerbeth G. Visualization of freckleformation induced by forced melt convection in solidifyingGaIn alloys. Mater Lett 2010;64(12):1340–3.

[22] Derebail R, Koster JN. Visualization study of melting andsolidification in convecting hypoeutectic Ga–In alloy. Int JHeat Mass Transfer 1998;41(16):2537–48.

[23] Campanella T, Charbon C, Rappaz M. Grain refinementinduced by electromagnetic stirring: A dendritefragmentation criterion. Metall Mater Trans A2004;35(10):3201–10.