Japanese Society of Microscopy

Journal of Electron Microscopy 53(4): 361369 (2004) doi: 10.1093/jmicro/dfh048

...............................................................................................................................................................................................................................................................................

Full-length paper

Aluminium phosphide as a eutectic grain nucleus in hypoeutectic Al-Si alloys

Kazuhiro Nogita1,*, Stuart D. McDonald1, Katsuhiro Tsujimoto2, Kazuhiro Yasuda3 and Arne K. Dahle1
1Division

of Materials Engineering, University of Queensland, St Lucia 4072, Australia, 2Research Laboratory for High Voltage Electron Microscopy, Kyushu University, Japan and 3Department of Applied Quantum Physics and Nuclear Engineering, Kyushu University, Japan *To whom correspondence should be addressed. E-mail: k.nogita@minmet.uq.edu.au
..................................................................................................................................................................................................................

Downloaded from http://jmicro.oxfordjournals.org/ at Taylor's University College on April 13, 2013

Abstract

..................................................................................................................................................................................................................

Aluminium phosphide (AlP) particles are often suggested to be the nucleation site for eutectic silicon in Al-Si alloys, since both the crystal structure and lattice parameter of AlP (crystal structure: cubic F43m; lattice parameter: 5.421 ) are close to that of silicon (cubic Fd3m, 5.431 ), and the melting point is higher than the Al-Si eutectic temperature. However, the crystallographic relationships between AlP particles and the surrounding eutectic silicon are seldom reported due to the difficulty in analysing the AlP particles, which react with water during sample preparation for polishing. In this study, the orientation relationships between AlP and Si are analysed by transmission electron microscopy using focused ion-beam milling for sample preparation to investigate the nucleation mechanism of eutectic silicon on AlP. The results show a clear and direct lattice relationship between centrally located AlP particles and the surrounding silicon in the hypoeutectic Al-Si alloy. Al-Si alloy, eutectic, solidification, nucleation, aluminium phosphide, focused ion beam 2 March 2004, accepted 21 May 2004

Keywords Received

.................................................................................................................................................................................................................. ..................................................................................................................................................................................................................

Introduction
Aluminium-silicon (Al-Si) foundry alloys are popular because of their good castability, surface finish and resistance to corrosion, coupled with their high strength-to-weight ratio. Typically, ~50100% of the volume of an Al-Si casting alloy is a eutectic mixture of aluminium and silicon. Solidification of the eutectic is one of the last reactions to occur during cooling and this is where most casting defects, such as porosity and hot tears, are formed. These defects reduce the mechanical properties, compromise the ability of the casting to contain liquid or gas under pressure, interfere with coating and machining operations and can detract from the appearance of the product, in addition to causing a high reject rate. Although it is frequently acknowledged that the solidification of the eutectic is critical for porosity formation [14], there is currently a lack of understanding about how the eutectic evolves (particularly nucleation) and how it can be controlled. It has only recently been determined, using the technique of electron backscattered diffraction (EBSD), that there are three

eutectic solidification modes (or macroscopic growth patterns) that operate in Al-Si alloys [4]. Figure 1 shows an illustration of the three modes, with supporting evidence of their existence from the optical microscopy of samples quenched during eutectic solidification. These are (a) nucleation on or adjacent to primary aluminium dendrites, (b) independent heterogeneous nucleation of eutectic grains in interdendritic spaces and (c) growth of the eutectic solidification front opposite to the thermal gradient. More than one growth mode may operate at any given time, although the addition of certain elements can cause a preference for one of the above mechanisms. It is likely that the different eutectic solidification modes arise due to the activation and/or inactivation of various nuclei. Ideally, it is desirable to control both the silicon morphology and the eutectic growth mode independently. This will allow both the macro- and microstructures of the eutectic to be tailored as required. Figure 2 is a micrograph of an unmodified Al-Si alloy that has been quenched during eutectic solidification to identify

362

J O U R N A L O F E L E C T R O N M I C R O S C O P Y, Vol. 53, No. 4, 2004

(a)

(b)

(c)

Downloaded from http://jmicro.oxfordjournals.org/ at Taylor's University College on April 13, 2013

100 m

200 m

200 m

Al-Si eutectic

Primary dendrite

Mould wall

Fig. 1 Eutectic growth modes. (a) Nucleation on or adjacent to primary aluminium dendrites, (b) independent heterogeneous nucleation of eutectic grains in interdendritic spaces and (c) growth of the eutectic solidification front opposite to the thermal gradient.

(a)

(b)

Fig. 2 Optical micrographs of quenched sample during eutectic solidification at (a) low magnification and (b) higher magnification with nuclei particle centre of eutectic silicon.

K. Nogita et al. Aluminium phosphide as a eutectic grain

363

Counts (arbitrary units)

Si

Al

Downloaded from http://jmicro.oxfordjournals.org/ at Taylor's University College on April 13, 2013

Counts (arbitrary units)

A
0 10 20 30 40

Distance (m)
Fig. 3 The EDS line scan results showing the presence of Al, P and O in the centrally located nucleant particle.

eutectic grains. Many of the eutectic grains contain centrally located particles within the silicon phase. It is tempting to conclude that such particles may have been a nucleus for the eutectic grain. In Al-Si alloys, it has often been suggested that particles similar in appearance to this one are AlP, because AlP and Si have near identical lattice parameters, theoretically making it an ideal nucleus. Conventional energy-dispersive spectroscopy (EDS) analysis (Figure 3) does show the presence of Al and P; however, oxygen (O) is almost always present in large quantities [5], raising the possibility that the particle is actually aluminium phosphate (AlPO4), which is a common contaminant of aluminium-based melts. Unfortunately, conclusive proof of nucleation cannot be obtained by observing an intimate physical relationship between phases such as that seen in Figure 3, but is reliant on crystallographic evidence provided by transmission electron microscopy (TEM). This makes it difficult to identify a nucleus with confidence for two reasons: firstly, a great number of TEM samples
Table 1. Chemical composition of the sample (wt.%) Al Balance Si 9.77 Cu 0.83 Fe 0.11 Mg <0.01 Zn <0.01

would need to be prepared before a suitable particle could be found within the prepared region and, secondly, the sample can be damaged during conventional preparation procedures. The technique of focused ion beam (FIB) milling and in situ sample manipulation allow eutectic grains to be excavated and removed for TEM analysis of potential nuclei. This study uses this advanced analysis technique to locate and identify the composition of particles that are potential eutectic grain nuclei and test for the existence of an orientation relationship between these particles and surrounding phases.

Methods
Aluminium-silicon alloys of nominal composition (Al 10mass% Si) were used for experimentation. The composition of the base alloy is given in Table 1. The alloys were produced by placing ~1 kg commercial purity aluminium and silicon in a clay-graphite crucible and heating in an electric resistance

..................................................................................................................................................................................................................... ..........................................................................................................................................................................................................................

Cr <0.005

Ni <0.01

Mn <0.01

Ti 0.007

Sr <0.001

Zr <0.005

P 0.004

364

J O U R N A L O F E L E C T R O N M I C R O S C O P Y, Vol. 53, No. 4, 2004

(a)
Nucleant Al

(b)

Si
Downloaded from http://jmicro.oxfordjournals.org/ at Taylor's University College on April 13, 2013

(c)

(d)

Fig. 4 (ad)

furnace to a temperature of 760C. Phosphorus additions were made in the form of a Cu-P master alloy to raise the final phosphorus content to an analysed level of 30 p.p.m. Metallographic samples were taken by pouring the melt into a copper mould at room temperature. Samples were prepared for microscopy using standard techniques, with the final polishing stage being produced by 0.05 m colloidal silica. Samples were etched for 60 s at room temperature in a mixture of 60 ml water, 10 g sodium hydroxide and 5 g potassium ferricyanide (Modified Murakami reagent). Optical micrographs were taken of both unetched and etched samples. Eutectic grains were examined using scanning electron microscopy and those grains that contained visible potential nuclei (i.e. centrally located particles) were selected for further preparation using FIB milling. It is only recently that the

advanced technique of FIB milling combined with in situ sample micromanipulation has become available to selectively extract a TEM sample from any desired region within a larger sample. The FIB sample preparation was performed with a Hitachi FB-2000A with 30 kV Ga liquid metal ion source. The procedure that was used is outlined in Fig. 4ag, which shows the sample prior to milling (a), after being protected with a W coating (b), after excavation using a 30 kV Ga source (c), being tilted to 45 to allow the sample to be completely cut (d), during manipulation with a microneedle after FIB milling (e), after being attached to a TEM sample grid and further thinned (30 kV Ga) (f), and ready for TEM analysis (g). The TEM observations were conducted using a 200 kV Technai20 TEM and included the acquisition of micro-selected area diffractions (SAD) taken from an area of ~60 nm in diameter, small probe convergent-beam electron diffraction patterns

K. Nogita et al. Aluminium phosphide as a eutectic grain

365

(e)

(f)

Downloaded from http://jmicro.oxfordjournals.org/ at Taylor's University College on April 13, 2013

(g)

Fig. 4 The FIB procedure for sample preparation. (a) Prior to milling, (b) protected region of interest with W coating, (c) excavation by 30 kV Ga, (d) 45 tilt to cut a bottom part of sample, (e) being manipulated with microneedle after FIB milling, (f) attached to TEM sample grid and further thinning by 30 kV Ga, and (g) ready for TEM analysis.

(CBED) taken from an area of ~10 nm in diameter and energy-dispersive X-ray analysis.

Results and discussion

The crystal structure and lattice parameters of the closepacked planes of Si, AlP and AlPO4 are shown in Table 2. From this table it is apparent that there is minimal mismatch
Table 2. Crystallographic data for selected phases Phase

between the AlP and the Si phases (<1%). However, there are several criteria that must be satisfied if a particle is to be an effective nucleus for solidification. Favourable nuclei often have a lattice parameter with a good match with the solid to be nucleated, although this is not sufficient information to predict the composition of an effective nucleus. A potent nucleus should provide a small surface tension to the phase that is nucleating. In addition to crystallographic similarities,

.....................................................................................................................................................................................................................

Crystal structure

Lattice parameter

Silicon (Si) Cubic a = 0.5421 nm ..................................................................................... ............................................................................................................................... Aluminium phosphide (AlP) Cubic a = 0.5431 nm ..................................................................................... ............................................................................................................................... Aluminium phosphate (AlPO4) Hexagonal a = 0.439 nm, c = 1.094 nm

366

J O U R N A L O F E L E C T R O N M I C R O S C O P Y, Vol. 53, No. 4, 2004

Nucleant particle

Downloaded from http://jmicro.oxfordjournals.org/ at Taylor's University College on April 13, 2013

Si

Al
Fig. 5 The FIB sample of eutectic Si in hypoeutectic Al-Si alloy.

the nuclei must also be present as a solid (although not necessarily thermodynamically stable) at the appropriate temperature and composition of the liquid, and also with a suitable size distribution. The nuclei must also be capable of being wetted by the liquid and should not be consumed or enveloped by reactions that have occurred earlier in solidification. For these reasons it is not without pitfalls to assume AlP is the nucleus for eutectic solidification solely because of minimal lattice mismatch between AlP and Si. This is compounded by the fact that oxygen has been detected along with phosphorus in large quantities in suspected nuclei and that AlPO4 is a common refractory binder used in aluminium production. Experimental determination is therefore the best way to determine the composition of many nuclei. Figure 5 is a micrograph of a cross-sectioned sample produced using the FIB. The sample contains eutectic Al and Si and a third phase that has possibly acted as a nucleus for the surrounding silicon. From the micrograph it appears the nucleus is not a discrete, compact particle, but is probably a convoluted morphology (connected in three dimensions) or alternatively one of several particles. Figure 6 shows a slightly higher magnification image along with a micro-SAD pattern (area of ~60 nm in diameter) from the locations labelled A

(unknown particle), and small probe CBED patterns (area of ~10 nm) from locations B (unknown particle) and C (eutectic silicon). Location A lies within a particle that was exposed to preparation using conventional polishing techniques and the SAD pattern along with EDS results show that this particle has an amorphous structure and contains Al, P and O (Fig. 6a). In contrast, the CBED patterns from locations B and C, which were deep to the polished surface and exposed only by FIB milling, indicate crystalline structures. By comparison of CBED patterns (obtained using an identical angle of tilt after aligning the [011] pole in location C) for the particle (location B) and the silicon crystal (location C) there is clearly a common diamond structure. The EDS analysis confirmed that the composition of the particle at location B was AlP and no oxygen was present. The crystallographic orientation relationship between the AlP (Fig. 6b) and neighbouring eutectic Si (Fig. 6c) is identical, conclusively proving that the AlP particle acted as an epitaxial nucleation site for the eutectic silicon. The above results confirm that AlP is the nucleation site for eutectic silicon. Furthermore, it is tempting to conclude that reactions occurring during conventional sample preparation methods disrupt the crystalline nature of any exposed regions

K. Nogita et al. Aluminium phosphide as a eutectic grain

367

Downloaded from http://jmicro.oxfordjournals.org/ at Taylor's University College on April 13, 2013

Fig. 6 TEM image and corresponding (A) electron diffraction patterns from Al-P-O precipitate on polished surface showing amorphous structure, (B) small probe CBED pattern from AlP showing [011] pole, and (C) small probe CBED pattern from Si showing [011] pole. CBED patterns for (B) and (C) were obtained using an identical tilt angle.

of AlP. The oxygen that is present in exposed particles (see, for example, Fig. 3) is likely to be an artefact of the polishing processes and does not indicate the presence of AlPO4. Although FIB milling allowed the selective preparation of TEM samples and avoided damage of the sample compared with conventional procedures, it was not without disadvantages. The first problem was due to a contamination of the surface with W during the sample excavation and milling stages. The second problem arose because the minimum sample thickness achievable is limited to ~200 nm. This limited the quality of the high-resolution lattice images that were obtainable from the sample and prevented the direct measurement of lattice parameters. The identification of AlP as a nucleation site for eutectic silicon supports a model of eutectic nucleation that is strongly influenced by impurities in the melt. A probable nucleation sequence is described with the aid of Figure 7. During the first

stage of solidification in hypoeutectic Al-Si alloys, dendritic growth occurs and phosphorus and silicon will segregate from the dendriteliquid interface due to their low solubility in solid aluminium. If sufficient phosphorus is present, AlP will form at the dendriteliquid interface due to the increased solute content, as shown schematically in Fig. 7a. Alternatively, the AlP particles may be stable in the melt and simply be pushed ahead of the dendriteliquid interface during solidification. Each of these particles can act as a nucleus for a silicon crystal and the commonly observed polyhedral shape of the silicon crystal, resembling that of primary silicon crystals, may be adopted because of the locally high silicon concentration at the interface. Eutectic solidification commences from each growing polyhedral crystal (Fig. 7b). Previous results have shown that the eutectic aluminium in commercial unmodified alloys has an identical crystallographic relationship to the surrounding aluminium dendrites

368

J O U R N A L O F E L E C T R O N M I C R O S C O P Y, Vol. 53, No. 4, 2004

AlP

Si

(a)

Al dendrite

(b)

Fig. 7 (a) AlP segregate to the dendrite-liquid interface during the growth of Al dendrite. (b) AlP particles act as nucleation sites for polyhedral silicon crystals from which eutectic grains grow.

[4,6]. While this may seem in conflict with nucleation of eutectic silicon on AlP particles within the melt, this is not necessarily so. Instead, it appears that the nucleation and growth of the eutectic aluminium and silicon phases occur somewhat independent from one another. When eutectic silicon nucleates on AlP particles near the dendrite liquid interface, the adjacent melt becomes enriched in aluminium because of the almost zero solubility of aluminium in silicon. Renucleation of aluminium is not necessary to allow coupled eutectic growth, due to the proximity of the existing aluminium dendrites. Instead, the eutectic aluminium is likely to simply grow from the dendrites to surround the silicon phase. A schematic representation of the complete nucleation sequence for the Al-Si eutectic is shown in Fig. 7. Cooling curve traces of Al-Si alloys further support the above observation of the very efficient nucleation of silicon on AlP particles [5]. These cooling curves show virtually no undercooling and recalescence with the presence of phosphorous in the melt, and the effect of phosphorous is evident even at quite low concentrations (e.g. in the order of 10 p.p.m.). Observations of quenched samples have also shown that the nucleation frequency is relatively high in alloys containing added phosphorous, and that addition of elements such as Sr and Na (which are introduced to modify the silicon from coarse plates to a fine fibrous morphology) causes a dramatic decrease in nucleation frequency [4]. The exact mechanism by which these elements cause a decrease of nucleation frequency is unknown, but it has been speculated it may be a result of deactivation of the AlP nucleants through coverage by intermetallic phases and/or alterations of surface energy. The high nucleation frequency observed with AlP is further stimulated by the strong segregation of the eutectic aluminium from silicon. This constitutional effect will strongly support the stability, survival and further growth of the silicon. The success of combining the techniques in this work to study potential nucleant particles could stimulate similar studies of nucleation-related phenomena for a range of systems. For example, it would be most interesting to use a similar approach to investigate the nucleation of primary aluminium crystals, in combination with other experiments.

Such studies could investigate the structure, properties and composition of nucleant particles, perhaps adding more substantial results to resolve discussions regarding the role of borides (AlB2, TiB2 and mixed borides) and aluminides (Al3Ti) in the grain refinement of aluminium alloys. Furthermore, further evidence may be obtained to substantiate the recent model by Greer et al. [7] about the role of nucleant particle size in nucleation. Their model proposes that it is the larger range particle sizes in the nucleant particle size distribution that become activated on solidification and that it is the undercooling (through the cooling rate and recalescence) that controls the fraction of the particle size distribution that becomes activated as nucleation sites in a given alloy with a certain nucleant particle distribution. However, the approach taken in this paper is certainly not limited to the study of liquidsolid transformations. For example, this approach could prove itself as a powerful method to study lattice relationships in three dimensions for any phase transformation and nucleant. However, it is also important to realize that it is likely that the method requires further evidence, such as from combining it with other techniques, e.g. thermal analysis methods, quenching and statistical analysis, to provide for complete analyses of the phenomena governing the transformations at hand.

Downloaded from http://jmicro.oxfordjournals.org/ at Taylor's University College on April 13, 2013

Concluding remarks
The technique of FIB milling along with micromanipulation was used to create TEM samples from selected areas within the microstructure of hypoeutectic Al-Si alloys. Selective sample preparation allowed TEM analysis of suspected nuclei that would not have been possible with conventional sample preparation. Using TEM analysis techniques, including micro-SAD and CBED, it is shown experimentally that AlP particles and surrounding eutectic silicon share a common orientation and it is confirmed that AlP nucleates eutectic silicon. A likely solidification sequence for eutectic solidification in unmodified hypoeutectic Al-Si alloys is proposed. Conventional sample preparation is shown to introduce a significant amount of oxygen and damage the natural crystallography of the AlP phase. The analysis technique detailed is ideally suited to identifying potential nuclei in many alloys where a low number of particles or high reactivity of the sample may limit the use of conventional TEM samples.

Acknowledgements
This research is supported by the Kyushu University High Voltage Electron Microscopy Program under the Nanotechnology Support Project of the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan (project ID: T-Kyudai-H15-013). The authors also acknowledge financial support from the Center for Microscopy and Microanalysis (CMM) and the University of Queensland.

References ...................................................................................................................................
1

Iwahori H, Yonekura K, Yamanoto Y, and Nakamura M (1990) Occurring behavior of porosity and feeding capabilities of sodium- and

K. Nogita et al. Aluminium phosphide as a eutectic grain

strontium-modified Al-Si alloys. Trans. Am. Foundrymens Soc. (AFS Trans.) 98: 167173.

369

...................................................................................................................................

...................................................................................................................................

Laslaz G (1995) Macrostructure in hypoeutectic Al-Si alloys, dendrites and eutectic grains. Its effects on shrinkage behaviour. In: Fourth International Conference on Molten Aluminium Processing, Orland, Florida, USA, 1214 Nov. pp. 459480. Dahle A K, Taylor J, and Graham D A (2000) The role of eutectic growth mode in porosity formation in Al-Si alloys. Aluminium Trans. 3: 1730. Dahle A K, Nogita K, Zindel J W, McDonald S D, and Hogan L M (2001) Eutectic nucleation and growth in hypoeutectic Al-Si alloys at different strontium levels. Metal. Mat. Trans. A 32A: 949960.

...................................................................................................................................

...................................................................................................................................

...................................................................................................................................

McDonald S D (2002) Eutectic Solidification and Porosity Formation in Unmodified and Modified Hypoeutectic Aluminium-Silicon Alloys. (PhD thesis, University of Queensland.) Nogita K and Dahle A K (2001) Eutectic solidification in hypoeutectic Al-Si alloys: electron back-scattering diffraction analysis. Mat. Character. 46: 305310. Greer A L, Bunn A M, Tronche A, Evans P V, and Bristow D J (2000) Modelling of inoculation of metallic melts: application to grain refinement of aluminium by Al-Ti-B. Acta Materialia 48: 28232835.

...................................................................................................................................

Downloaded from http://jmicro.oxfordjournals.org/ at Taylor's University College on April 13, 2013