Observations on the microstructure and performance of an AlTiC grain-refining master alloy

Observations on the microstructure and performance of an AlTiC grain-refining master alloy

Materials Science and Engineering, A 188 (1994) 283-290 283 Observations on the microstructure and performance of an A1-Ti-C grain-refining master a...

989KB Sizes 0 Downloads 13 Views

Materials Science and Engineering, A 188 (1994) 283-290


Observations on the microstructure and performance of an A1-Ti-C grain-refining master alloy C. D. Mayes a, D. G. M c C a r t n e y b a n d G. J. T a t l o c k a aDepartment of Materials Science and Engineering, University of Liverpool, PO Box 147, Liverpool L69 3BX (UK) bDepartment of Materials Engineering and Materials Design, University of Nottingham, University Park, Nottingham NG7 2RD (UK) (Received November 1, 1993; in revised form January 3 l, 1994)

Abstract A1-Ti-C grain-refining master alloys are of increasing importance in aluminium casting because they are believed to introduce a smaller volume fraction of insoluble particles into the melt than conventional A1-Ti-B master alloys are. The main microstructural features of an AI-6wt.% Ti-0.02wt.%C master alloy have been examined using X-ray diffraction and optical, scanning and transmission electron microscopy. In addition to the aluminium matrix, two principal intermetallic phases were identified, namely titanium aluminide and titanium carbide. The titanium aluminides were found to contain small amounts of dissolved vanadium and to possess a range of morphologies from well faceted to rough and irregular. A combination of microanalysis and electron diffraction was used to show that clusters of submicron particles, rich in Ti but with virtually no AI, were TiC crystals. Individual particles were found to be single crystals with an octahedral morphology. A simple thermodynamic analysis of the stability of TiC in the Al-rich corner of the A I - T i - C phase diagram is presented which confirms the stability of TiC during master alloy synthesis. Using standard test procedures, the A I - T i - C alloy proved to be an effective grain refiner when added to commercial purity aluminium, and it is proposed that undissolved TiC particles were most likely to have been the heterogeneous nuclei.

1. Introduction Aluminium alloys are frequently grain refined during solidification in order to promote the formation of an equiaxed grain structure in the casting and to eliminate columnar grain formation. This is commonly achieved by adding a small quantity of a master alloy to the melt just prior to casting [1, 2]. Although it is generally accepted that this addition introduces nucleation sites into the molten alloy, the precise mechanisms of grain refinement are still not well understood [1, 2]. In industrial practice, AI-Ti-B master alloys are the most widely used grain refiners with addition levels of 100 wt.ppm Ti and 20 wt.ppm B (or less) being common [3]. Insoluble titanium diboride particles, which have the form of hexagonal platelets approximately 2 /~m by 0.5 /~m thick, are introduced by A1-Ti-B master alloy additions, and much recent research has been focused upon characterizing their precise composition and morphology [1, 4-6]. It is recognized, however, that many of the added platelets are redundant, in that they do not act as nucleation centres for the solidifying aluminium and hence 0921-5093/94/$7.00 SSD10921-5093( 94 )095 39-9

become undesirable hard inclusions in the final product [7, 8]. A1-Ti master alloys are also employed as grain refiners, but at the normal addition levels of 0.01-0.02 wt% Ti (i.e. well below the peritectic composition of the AI-Ti phase diagram, which occurs at 0.15 wt.% Ti [9]) such alloys have, in the past, been found to behave inconsistently and to be much less effective than A1-Ti-B master alloys. Recently, however, enhanced performance has been achieved in AI-Ti-based master alloys containing approximately 0.02 wt.% C and low boron levels (less than 0.003 wt.% B) [10-12]. Inherently, such alloys contain very low volume fractions of insoluble particles and so are particularly attractive for use in applications where aluminium metal cleanliness and freedom from hard inclusions is of crucial importance [11]. Despite the number of papers relating to the good grain-refining performance of low carbon AITi-C master alloys, relatively little is known about their microstructure and, in particular, any low solubility particles which could act as nuclei for the solidifying aluminium in grain refined melts. Only Banerji and coworkers [13-15] have examined microstructures in © 1994 - Elsevier Sequoia. All rights reserved


C. D. Mayes et al.


Microstructure and performance of Al-Ti-C

any detail. They reported the presence of TiC in A I - T i - C master alloys but their materials had much higher carbon levels, in excess of 0.5 wt.%. The principal objective of the present study was to examine the microstructure of a commercially produced master alloy of nominal composition A1-6wt.%Ti-0.02wt.%C. A combination of optical microscopy, scanning electron microscopy (SEM), transmission electron microscopy (TEM) and X-ray diffraction (XRD) was used to identify and characterize the phases present. The main microstructural observations are reported in this paper together with brief results on the grain-refining performance of the same batch of material. Comparisons are also drawn with the performance of a conventional A1-5wt.%Ti-lwt.%B master alloy.

2. Experimentalprocedures The A l - T i - C master alloy was supplied by Anglo Blackwells Ltd., Widnes, UK. The synthesis and production of the master alloy are proprietary processes. However, it is known that the rod of 9.5 mm diameter supplied was produced by the mechanical working of a cast ingot. The alloy composition is given in Table 1. Spark emission spectrometry was used to determine the level of all elements except for carbon. Carbon analysis was performed by IR spectrometry for CO2 gas following sample combustion in high purity oxygen. Optical microscopy and XRD were performed on master alloy samples which had been ground and polished using standard procedures. Samples were deep etched for SEM using an iodine-methanoltartaric acid mixture as described previously [1, 16]. Both electropolished 3 mm disks and carbon extraction replicas were examined in a transmission electron microscope operating at 120 kV. The disks were prepared conventionally by grinding and then electropolishing in a perchloric acid-ethanol mixture (1:6 by volume) at room temperature and a 20 V electrode potential. To ensure thinning of intermetallics as well as the aluminium matrix, disks were given a short ion

milling for 2 h using Ar ions at 4 kV. To prepare the extraction replicas, the master alloy rod was fractured and a thin carbon film deposited on the fracture surfaces using a vacuum evaporator. The fracture surface replica was then released using two etches. Firstly the sample was immersed in an H 2 0 - H N O 3HC1-HF mixture ( 1 0 0 : 1 0 : 5 : 1 ) f o r 15 rain; this was followed by a 5 min immersion in an NaOH solution (10 g of NaOH in 250 ml of H20) at room temperature. The carbon film was finally washed in water and collected on a copper grid for TEM examination. In the transmission electron microscope the main techniques employed were energy-dispersive X-ray analysis (EDXA), electron energy loss spectroscopy (EELS) and selected area diffraction. Energy-dispersive X-ray and electron energy loss spectra were acquired at 120 kV, with the former being used for the analysis of elements with atomic number greater than 11 and the latter for light-element analysis. TheA1-Ti-C master alloy, whose microstructure was characterized, was also tested for grain refinement efficiency when added to commercial purity aluminium. The aluminium composition is given in Table 1. Master alloy samples for microstructural investigation and grain refinement tests were all taken from the same batch supplied by the manufacturer. Two standardized grain refining tests were employed, namely the Aluminum Association test and the Alcoa cold-finger test. The details of these tests have been given elsewhere [3, 17, 18]. Master alloy addition levels were 0.01 wt.% Ti in the Aluminum Association test and 0.02 wt.% Ti in the Alcoa test. The aluminium melts were heated to 720 °C prior to master alloy addition, held for 2 min after the addition was made and then cast according to the test procedure [3, 17, 18]. For comparative purposes, identical tests were performed, on the same batch of aluminium, using a conventional A1-5wt.%Ti-lwt.%B master alloy. The grain structures of castings were studied by macroetching. Equiaxed grain sizes were measured using a standard linear intercept method [19]. To reveal the grains for measurement purposes, samples were cut from the castings at known positions, anodized [20] and viewed under crossed polars.

TABLE 1. Alloy compositions as determined by chemical analysis Alloy

AI-Ti-C AI-Ti-B Commercial purity AI

Amount (wt.%) of the followingelements Ti








5.9 5.2 0.0046

0.001 0.91 0.0004

0.025 0.011 0.010

0.12 0.17 0.17

0.04 0.08 0.06

0.26 0.04 0.004

0.2 0.2 0.1

Balance Balance Balance

C. D. Mayes et al.


Microstructure and performance of Al-Ti-C

3. Results 3.1. Microstructure and phase identification

X-ray diffractometry of a flat polished surface of the master alloy revealed only peaks that were consistent with either a-A1 or titanium aluminide (TiA13) phases. Thus any further phases were established to be present only in small volume fractions. An optical micrograph of a transverse section of the rod material is shown in Fig. l(a). Irregular titanium aluminides 50--100 p m long are embedded in an a-Al matrix. It is probable that the cracked aluminides, observable in Fig. l(a), were produced during mechanical working from ingot to rod. Using SEM and E D X A the micron-sized particles in the aluminium matrix were generally found to be Fe rich and so probably AI-Fe intermetallics but their precise nature was not determined. At a higher magnification in the optical microscope (Fig. l(b)) the aluminide phase apepars to have within it micron-sized dark particles. These were further examined using SEM and Fig. 2 shows micrographs of deep-etched samples with a range of typical


.... z7


aluminide morphologies. Some aluminides have welldefined facets as revealed in Figs. 2(a) and 2(b) whereas others were much rougher with pitted surfaces, as shown for example in Fig. 2(c). The particles observed within the pits are consistent in size with the dark particles revealed by optical microscopy• Quantitative EDXA of the aluminide phase with a 20 keV primary electron beam revealed that it had the composition (Ti0,yV0.03)A12,. Vanadium is an impurity in the raw materials used for master alloy synthesis and its segregation to the aluminide had been noted previously [21] but not quantified. To confirm that the SEM observations were not artefacts of the specimen preparation procedure, thin foils were examined by TEM, and Fig. 3 shows an example of a 200 nm particle within the aluminide phase• E D X A indicated that this particle was Ti rich, but more detailed compositional and structural data were difficult to obtain because of the surrounding aluminide phase. To overcome this problem, carbon extraction replicas were prepared and examined in the transmission electron microscope• Replicas successfully removed many particles which E D X A also showed to be Ti rich but essentially without A1. They are seen to form micron-sized agglomerates, as shown in Fig. 4, and discrete particles within agglomerates




r ....



20t~m Fig. 1. Optical micrographs of a sample taken from the transverse section of a master alloy rod: (a) irregular titanium aluminides (TiA13) embedded in an a-Al matrix; (b) micron-sized particles within an aluminide.

Fig. 2. Scanning electron micrographs of deep-etched masteralloy samples showing the different titanium aluminide morphologies present. Micron- and submicron-sized particles are observable in surface pits. In (a) and (b), aluminides exhibit facets whereas, in (c), the morphology is non-faceted.


C. D. Mayes et aL


Microstructure and performance of Al-Ti-C

lOOnm Fig. 3. Transmission electron micrograph from a thinned foil showing a Ti-rich particle surrounded by the titanium aluminide phase.

were typically 200 nm in size. By electron diffraction analysis these individual particles were found to be single crystals. They were identified as either TiC or TiN by tilting to (101), (111) and (121) zone axes. Both these crystals have the NaCI crystal structure with lattice parameters of 0.433 nm and 0.424 nm respectively, and uncertainties inherent in T E M selected-area diffraction make unambiguous identification difficult. Therefore E E L S analysis was employed to obtain chemical information on the low atomic number elements present in the particles. A typical electron energy loss spectrum is shown in Fig. 5 and it should be noted that only Ti and C absorption edges occur. Much of the C edge of this spectrum is due to the carbon replica, but the absence of a nitrogen edge (which would have appared at 402 eV) confirms that the particles are not TiN and so must be TiC. The morphology of individual TiC particles was also examined using the extraction replicas. By tilting a replica in the transmission electron microscope the shape of individual crystals was observed to be octahedral with well-defined facet planes. In Fig. 6(a) the T E M image of a typical crystal is shown, viewed along an axis of the octahedron. The diffraction pattern which is superimposed on the image was obtained by double exposure and, for the microscope conditions used, there is a rotation of 136 ° between the pattern and the image. A schematic indexed pattern is also shown

Fig. 4. An agglomerate of small Ti-rich particles which contained essentially no A1 on a carbon extraction replica viewed by TEM.













1034 Energy loss eV.

Fig. 5. Electron energy loss spectrum taken from a sinbgle Tirich particle on a carbon extraction replica. Note the presence of only carbon and titanium absorption edges. If nitrogen were present, an edge at 402 eV would have appeared.

in Fig. 6(a) and reveals that the single crystal was oriented with its [001] zone axis parallel to the beam direction. Hence this crystallographic direction coincides with the axis along which the octahedron was viewed. When the 136 ° rotation between image and diffraction space is accounted for, it is clear that the other (001)-type directions lie along the diagonals of

C. D. Mayes et al.


Microstructureand performance of A I - E - C


TABLE 2. Grain sizes obtained from the Aluminium Association test procedure

• (020)

Master alloy type (nominal)

Addition level

Linear intercept grain size





A1-6Ti-0.02C AI-5Ti- 1B

0.01 0.01

135 _+5 120 _+5

(2~o) (2~o>

!!+~!i ~




[~oo] 11o]

[olo] Fig. 6. (a) Transmission electron micrograph showing a TiC single crystal viewed along an [001] zone axis as the superimposed diffraction pattern confirms. The pattern is indexed on the schematic diagram. Crystal orientations were determined after allowing for the 136 ° rotation between diffraction and image space. (b) Schematic representation of an octahedral TiC crystal showing crystallographic directions and facet planes as deduced from TEM imaging and diffraction.

the T E M image in Fig. 6(a). This analysis is summarized in Fig. 6(b) which is a schematic representation of the TiC crystal morphology with planes and directions indicated. 3.2. Grain-refining performance W h e n the A I - T i - C and A I - T i - B master alloys were assessed using the Aluminum Association test at the 0.1 wt.% Ti addition level, fully equiaxed grain structures were obtained. T h e measured grain sizes are listed in Table 2. Clearly the T i - C master alloy was an effective grain refiner albeit less potent than the conventional T i - B master alloy. Similarly, a difference in effectiveness was revealed through the Alcoa test procedure. To obtain a fully equiaxed structure, a 0.02

Fig. 7. Macrophotographs of etched Alcoa test castings: (a) the ingot with 0.0l wc% Ti from the Ti-B alloy; (b) the sample with 0.02 wt.% Ti from the Ti-C alloy.

wt.% Ti addition was required from the T i - C master alloy whereas a 0.01 wt.% Ti addition from the T i - B alloy was sufficient to give equiaxed grains throughout the sample. Figure 7 shows the grain structures which resulted from the latter tests, and the grain sizes measured on a transverse section 40 mm from the ingot base are also given. Without a grain refiner addition the commercial purity aluminium solidified with large columnar grains in both types of test.

4. D i s c u s s i o n

T h e results indicate that, by using a combination of optical and electron optical techniques, TiC crystals and agglomerates were unambiguously identified in the A I - T i - C master alloy. Moreover, the individual crystals were shown to have {111/facet planes. M o r e detailed compositional and structural information has thus been obtained than in other previous studies [13-15]. For example, although Banerji and Reif [14] presented electron microprobe analysis evidence for TiC formation in A I - T i - C master alloys with 0.5-2.0 wt.% C levels, they did not examine the crystal


C. D. Mayes et aL


Microstructure and performance of A l - T i - C

morphologies in detail. Furthermore, they did not confirm that the crystal structure of the titanium-carbon compound was indeed cubic with the NaCl-type structure. The extraction replica technique employed in the present study also enabled us to confirm that there was no detectable Al in the TiC crystals. Equally there was no evidence for the presence of A I 4 C 3 , although one cannot disregard the fact that, being highly reactive, it might have dissolved during the replication process. The only other phases observed, besides the carbides, aluminides and iron-rich intermetallics, were very small amounts of the oxides of aluminium (A1203) and titanium (TiO 2). With regard to the titanium-rich aluminide phase, a comparison can be made between its morphology in this A I - T i - C master alloy and in the more common A1-Ti-B master alloys. Recent work on these latter alloy types has shown that a so-called duplex type of aluminide can form in the melt during alloy synthesis [1, 5, 22] with TiB 2 particles apparently embedded in the TiA13 crystals. Here our observations, in particular in Figs. 2 and 3, show that a duplex type of aluminide has again formed during synthesis with faceted TiC particles entrained both within the aluminide crystal and at its surface. The overall aluminide stoichiometry (Ti097V0.03)A12.9 is close to the composition range of TiA1296 to TiA1308 reported by Mondolfo [9] for binary aluminides. The presence of vanadium in the alloy is attributable to its occurrence, as a known impurity, in the titanium metal used in the alloy's manufacture. Moreover, this study has confirmed and quantified a previous observation of V segregation to the aluminide phase [21]. It is important to consider how our experimental observations of TiC in the master alloy can be related to the available thermodynamic data on the A1-Ti-C system. Jones and Pearson [23] were the first to calculate the solubility product of TiC in molten aluminium using a thermodynamic database. Much more recently Jarfors and coworkers [24, 25] examined the Al-rich corner of the system theoretically and experimentally whilst Rapp and Zheng [26] reanalysed some earlier work and determined the solubility of A14C 3 and TiC

as functions of temperature. In Table 3, published solubility data for both A I 4 C 3 and TiC are summarized, in consistent units, from refs. 24-26 and it can be seen that there are significant discrepancies between the values from the different thermodynamic data sets. For the present discussion, however, our observations can best be examined by referring to both Table 3 and the schematic isothermal section of the Al-rich corner of the A1-Ti-C phase diagram shown in Fig. 8. Considering firstly the synthesis of the 6 wt.%Ti-0.02wt.%C master alloy (mole fractions of 3 . 4 x 1 0 -2 and 4.5 x 10 -4 for Ti and C respectively), it is apparent from the data in Table 3 and Fig. 8 that at a typical processing temperature of 1100 K this composition falls well outside the L+ A l 4 C 3 region. Hence, the observation of TiC, but n o t A14C3, is entirely consistent with the thermodynamic calculations. Turning now to the grain refiner tests, TEM of thinned disks was employed in an effort to identify the substrates responsible for heterogeneous nucleation. No firm evidence was obtained for the hypothesis, first put forward byCibula [27], that TiC particles act as the

.0r o

Liquid (L)


L*AI~.Ca Molar fraction carbon I

Fig. 8. Schematic illustration of an isothermal section (at 1000 or 1100 K) from the Al-rich c o m e r of the A 1 - T i - C phase diagram (taken mainly from ref. 24).

T A B L E 3. Equilibrium solubility data for TiC and AI4C 3 in molten aluminium, where T is in kelvins and X~ is the mole fraction saturation content of c o m p o n e n t i Reference

[~6] [g6] [24, 25] [24, 25]

Equations for solubilityof T i C and AI4C 3 in molten AI

X c at point A, Fig. 8

XT~ at point A, Fig. 8

1000 K

1100 K

1000 K

1100 K

X c = 294 exp( - 1 9 3 7 0 T - 1 ) XcXxi ~- 2.02 exp( - 19650 T - 1_ 0.116 In T )

1.14 x 10 -6

6.62 x 10 -6 2.33 x 10 -3

2.36 x 10 -3

X c = 3393 exp( - 2 3 7 9 7 T - J) XcXTi = 169 exp( - 26865 T - J)

1.57 x 10 -7

2 . 3 2 x 10 -3

3.06 x 10 -3

1.37 x 10 -6

C. D. Mayes et al.


Microstrucmre and [email protected])rmance of Al-Ti-C

heterogeneous nucleation centres. However, since the carbide size in the present materials was less than 1 ktm, and the average grain size around 150/~m, there must be a relatively low probability that an electrontransparent region (at 120 kV) coincides with a nucletion centre. Consequently our inability to obtain evidence for the nature of nucleating particles is not unexpected since fewer than 20 thin disks were examined. This work has therefore not been able to clarify the mechanism by which the A I - T i - C alloy grain refines. Nevertheless, the observation by Cisse and Boiling [28] that bulk TiC nucleates AI epitaxially combined with the recent T E M observations by Banerji et al. [29] showing TiC particles apparently at AI grain centres undoubtedly suggest that undissolved TiC particles from the master alloy addition acted as heterogeneous nuclei at the 0.01 wt.% Ti hypoperitectic composition where the aluminide phase is expected to dissolve [2]. The thermodynamic data in Table 3 can also be used to examine the stability of TiC in molten A1 at the 0.01 wt.% Ti (5.7 × 10 5 mole fraction) level used in grain refining. At this level, TiC is predicted to be thermodynamically unstable since the Ti composition of point A on Fig. 8 at 1000 K and 1100 K is well above 5.7 × 10 5 mole fraction (see Table 3). Therefore, depending on the base alloy carbon level, TiC is predicted either to dissolve or to transform to A14C 3. This contradiction between thermodynamic predictions and the experimentally observed grain-refining behaviour of the alloy needs to be examined in future work by performing phase stability calculations based on improved thermodynamic data. Finally, in examining the grain-refining test results it should be noted that these were not intended to be a comprehensive study of the alloy's grain-refining performance. The purpose was merely to confirm that material, from the same batch as that examined microstructurally, was indeed a potent grain refiner. It should be noted that the grain size measurements given in Table 2 and Fig. 7 are restricted to (i) short melt holding times, (ii) commercial purity A1 as the base alloy and (iii) comparisons based on equivalent weight percentage Ti additions. However, another study has confirmed that a similarly synthesized master alloy can effectively grain refine a wide range of base alloy types including systems of commercial importance such as alloy 3003 (AI-Mn-Fe-Si) and alloy 7050 (AI-Cu-Mg-Zn) [18]. A number of factors are believed to be responsible [2] for the different grain sizes produced by the Ti--C and Ti-B alloy types. These include the number and size of insoluble particles added, the extent to which agglomerated particles can be dispersed and the nucleating efficiency of the different crystal structures of the particles involved. Hence, it is impossible, on the basis of the present


study, to ascertain which factor is of principal importance. Indeed further progress in predicting the performance of grain-refining master alloys is dependent on a better understanding of the processes of particle aggregation during master alloy synthesis and particle dispersion when added to an aluminium melt.

5. Summary The microstructure of an A I - T i - C grain-refining master alloy was studied by X-ray diffraction and optical, scanning and transmission electron microscopy to determine the principal phases present. The titanium aluminide phase was found to contain small amounts of dissolved V (which had been incorporated during alloy synthesis) and to exhibit a range of morphologies. Tirich submicron-size particles, low in AI, were found to be embedded within aluminides. Using an extraction replica technique, these Ti-rich particles were further characterized in the transmission electron microscope. They were frequently observed as clusters and were identified as TiC using a combination of EDXA, EELS and electron diffraction techniques. Individual TiC particles were single crystals and were shown to have a characteristic octahedral morphology with {001) directions through the corners of the octahedron. Calculations based on available thermodynamic data indicate that TiC was stable during master alloy synthesis which is consistent with these experimental observations. The A1-Ti-C master alloy grain refined commercial purity aluminium at 0.(ll and 0.02 wt.% Ti addition levels but was less effective than an AI-5wt.%Ti-lwt.%B master alloy added to the same overall Ti addition. No firm evidence was obtained for the nature of the particles responsible for heterogeneous nucleation, but undissolved TiC from the master alloy was the most likely nucleation substrate.

Acknowledgements The authors would like to acknowledge support for this research from the Science and Engineering Research Council and Anglo Blackwells Ltd, UK, through the award of a CASE research studentship to C.D.M.

References I M. M. Guzowski, G. K. Sigworth and D. A. Sentner, Metal/. Trans. A, 18( 198716(/3-619.


C. D. Mayes et al.


Microstructure and performance of A l - T i - C

2 D.G. McCartney, Int. Metall. Rev., 34(1989) 247-260. 3 D.A. Granger, Proc. Int. Semin. on Refining andAlloying of Liquid Aluminium and Ferro-alloys, Norwegian Institute of Technology, Trondheim, 1985, pp. 231-244. 4 L. Arnberg, L. B~ickerud and H. Klang, Met. Technol., 9 (1982) 7-13. 5 R. Kiusalaas and L. B/ickerud, in J. Beech and H. Jones (eds.), Solidification Processing 1987, London, Institute of Metals, 1988, pp. 137-140. 6 C. D. Mayes, D. G. McCartney and G. J. Tatlock, in D. L. Peachey and D. B. Williams (eds.), Proc. 12th Int. Cong. on Electron Microscopy, Vol. 4, San Francisco Press, San Francisco, CA, 1990, pp. 962-963. 7 I. Maxwell and A. Hellawell, Acta Metall., 23 (1975) 229-237. 8 A. Hellawell, in Solidification and Casting of Metals, Metals Society, London, 1979, pp. 161-168. 9 L. E Mondolfo, Alurninium Alloys--Structure and Properties, Butterworth, London, 1976, p. 385. 10 S. R. Thistlewaite, D. Hadlet and J. C. Hoff, in H. Bhoner (ed.), Light Metals 1985, Metallurgical Society of AIME, Warrendale, PA, 1985, pp. 1169-1178. 11 G. T. Campbell, S. A. Danilak, S. R. Thistlewaite and P. Fisher, in E. R. Cutshall (ed.), Light Metals 1992, Metallurgical Society of AIME, Warrendale, PA, 1992, pp. 831-836. 12 G.K. Sigworth, USPatent5 041 263, 1991. 13 A. Banerji and W. Reif, Metall. Trans. A, 16 (1985) 2065-2068. 14 A. Banerji and W. Reif, Metall. Trans. A, 17 (1986) 2127-2137.

15 A. Banerji, Q. L. Feng and W. Reif, Metall. Trans. A, 20 (1989) 564-566. 16 C.D. Mayes, Ph.D. Thesis, University of Liverpool, 1992. 17 Standard Test Procedure for Aluminum Grain Refiners TP-1, Aluminum Association, Washington, DC, 1987. 18 S.M. Ahmady, D. G. McCartney and S. R. Thistlewaite, in C. M. Bickert (ed.), Light Metals 1990, Metallurgical Society of AIME, Warrendale, PA, 1990, pp. 837-843. 19 R. T. DeHoff and F. N. Rhines, Quantitative Microscopy, McGraw-Hill, New York, 1968, p. 239. 20 Metals Handbook, Vol. 9, 9th edn., American Society for Metals, Metals Park, OH, 1985, p. 354. 21 D. G. McCartney, D. Hadlett and S. R. Thistlewaite, in J. Beech and H. Jones (eds.), Solidification Processing 1987, Institute of Metals, London, 1988, pp. 141-194. 22 C.D. Mayes, D. G. McCartney and G. J. Tatlock, Mater. Sci. Technol., 9(1993) 97-103. 23 G. P. Jones and J. Pearson, Metall. Trans. B, 7 (1976) 223-234. 24 A. Jarfors, H. Fredriksson and L. Froyen, Mater Sci. Eng., A135(1991) 119-123. 25 A. Jarfors and H. Fredriksson, Microgravity Sci. Technol., 3 (4) (1991) 216-220. 26 R. A. Rapp and X. Zheng, Metall. Trans. A, 22 (1991) 3071-3075. 27 A. Cibula, J. Inst. Metals, 76(1949-1950) 321-360. 28 J. Cisse and G. F. Boiling, J. Cryst. Growth, 13-14 (1972) 777-781. 29 A. Banerji, Q. Feng and W. Reif, Acta Metall. Mater., 41 (1993) 2461-2472.