Partial modification in unidirectionally solidified AlSi eutectic alloys

Partial modification in unidirectionally solidified AlSi eutectic alloys

Acta metaU. Vol. 37, No. I, pp. 303-311, Printed in Great Britain. 1989 All rights reserved Copyright 0 OOOl-6160/89 $3.00 + 0.00 1989 Pergamon ...

2MB Sizes 0 Downloads 26 Views

Acta metaU. Vol. 37, No. I, pp. 303-311, Printed

in Great


1989 All rights reserved



OOOl-6160/89 $3.00 + 0.00 1989 Pergamon Press plc

PARTIAL MODIFICATION IN UNIDIRECTIONALLY SOLIDIFIED Al-Si EUTECTIC ALLOYS L. CLAPHAM and R. W. SMITH Department of Metallurgical Engineering, Queen’s University, Kingston, Ontario, Canada K7L 3N6 (Received 18 December


Abstract-Previous investigations on partially modified cast Al-Si alloys have indicated that the transition from an unmodified to a modified eutectic structure occurs discretely and is apparently associated with a critical level of impurity. The present work extends the partial modification study to unidirectionally solidified samples in order to more clearly define the role of impurity level and solidification rate. The values of strontium concentration and growth rate leading to the different types of Al-Si structure (i.e. unmodified, partially modified, and fully modified) have been determined. Partially modified samples are produced at relatively low growth rates (less than 25 pm/s) when the strontium level lies between 0.004 and 0.014 wt%. Their structures consist of discrete regions of flake and fibrous silicon, the distribution of which is dependent on the growth rate and the segregation characteristics of the strontium modifier. The results support the argument that impurity modification is due to the adsorption of the impurity element onto the silicon phase during growth. R&sum&Des ttudes anterieures sur les alliages Al-Si moulis, partiellement modifits, ont montrt: que la transition d’une structure eutectique non modifiee $ une structure eutectique modifiee a lieu de fagon disc&e et qu’elle semble associbe B un niveau critique d’impuretis. Dans cet article, nous &tendons I’ttude des modifications $ des 6chantillons &labor&spar solidification unidirectionnelle, dans le but de dCfinir plus nettement le r61e du taux d’impuretCs et de la vitesse de solidification. Nous avons dCtermi& les valeurs de la concentration en strontium et de la vitesse de croissance qui conduisent aux diffkrents types de structures AI-Si (non modifites, partiellement modiMes, et totalement modifi&). Les 6chantillons partiellement modifiis sont produits pour des vitesses de croissance relativement basses (infkrieures g 25 pm/s) quand la concentration en strontium est comprise entre 0,004 et 0,014% en poids. Leurs structures sont constitutes de plaquettes isolbes de silicium fibreux dont la rkpartition dbpend de la vitesse de croissance et des caracttristiques de sdgregation du strontium, tlkment modificateur. Les rksultats dbmontrent que la modification par les impure& est due i l’adsorption de I’tltment d’impurett sur la phase silicium pendant sa croissance. Zusammeofassung-Friihere Untersuchungen an teilweise modifizierten gegossenen Al-Si-Legierungen haben ergeben, daD der Ubergang von einer unmodifizierten in eine modifizierte eutektische Struktur diskret ablluft und offenbar mit einer kritischen Konzentration an Verunreinigungen zusammenhingt. In der vorliegenden Arbeit wird diese Untersuchung auf gerichtet erstarrte Proben ausgedehnt, urn die Rolle der Verunreinigungen und der Erstarrungsrate klarer zu machen. Hierzu wurden die Strontiumkonzentrationen und die Wachstumsraten, die zu den verschiedenen Typen der Al-Si-Struktur (d.h unmodifizierter, teilmodifizierter und vollstiindig modifizierter Typ) fiihren, bestimmt. Die teilmodifizierten Proben entstehen bei vergleichsweise niedrigen Wachstumssraten (kleiner als 25 pm/s), wenn die Strontiumkonzentration zwischen 0,004 und O,O14Gew.-% liegt. Die Struktur besteht aus fliichigem und fibrijsem Silizium, deren Verteilung von der Wachstumsrate und der Charakteristik der Segregation des Strontiums abhiingt. Die Ergebnisse stiitzen das Argument, daB die Modifizierung durch Verunreinigungen von der Absorption des Verunreinigungselementes in die Siliziumphase wghrend des Wachstums herriihrt.

1. INTRODUCTION Structural modification in the Al-Si eutectic can be achieved through the addition of small amounts (>O.OlO wt%) of either sodium of strontium to the melt prior to solidification. The presence of the impurity alters the silicon phase morphology from a flake to a fibrous form, resulting in an improvement in the strength of the final product. The last 12 years has witnessed a renewed interest in the study of the mechanism of impurity modification, a question which has yet to be adequately resolved. Most investigators have been concerned with studies of either the unmodified (flake silicon) or the fully modified

(fibrous silicon) state, however recent work by the present authors [l, 21 has shown that a considerable amount of information regarding modification behaviour can be obtained from a study of partially modified alloys, i.e. those in which the concentration of modifier (either sodium or strontium) is too low to produce complete modification. With these alloys it is possible to very carefully examine the exact conditions under which the unmodified (flake) to modified (fibrous) transition occurs and determine the influence of variables such as bulk impurity concenimpurity segregation and solidification tration, (growth) rate. Previously published work on partially modified 303



cast samples [l] suggested that if the growth rate (R) is low enough then a flake structure may nucleate and grow, despite the presence of strontium in the melt. In these specimens fibrous eutectic is often observed between the flake colonies, suggesting that strontium is pushed ahead of the moving flake interface and concentrates in the terminal liquid. The transition from the flake to the fibrous structure is always discrete, with no gradual transition region observed. The work reported here represents an extension of the study on partially modified castings to unidirectionally solidified samples. With unidirectional solidification it is possible to carefully examine the flake to fibrous transition and to more precisely define the role of impurity concentration and growth rate in the modification process. 2. EXPERIMENTAL All samples in the present study were prepared to the eutectic composition (12.7 wt% Si) from “superpurity” (99.994 wt%) aluminium and semiconductor grade silicon. Strontium was used as a modifying agent and was added in the form of a Al-10 wt% Sr alloy. Master alloys of appropriate composition were melted under an argon atmosphere and subsequently cast into cylindrical 6mm diameter graphite split moulds. Chemical analysis using an atomic absorption spectroscopy (AAS) technique (outlines in Ref. [3]) indicated that these master alloy rods were of homogeneous composition along their length. Rods were swaged and placed into 5mm ID x 300mm long alumina tubes which were sealed under argon. The samples were then solidified unidirectionally using a Bridgman (solid-down) technique. In the preliminary stage of the work an accelerated growth technique (AGTdescribed in Ref. [l]) was utilised to determine the ranges of growth velocity and strontium concentration likely to result in a partially modified structure. Once these ranges were established only constant velocity experiments were carried out. Samples for optical microscopic investigation were prepared using standard polishing techniques and were examined in the unetched condition. The silicon phase morphology was examined in more detail with scanning electron microscopy (SEM) following deep-etching in dilute (3-10%) HCl or HF solutions to preferentially remove the aluminum phase. 3. RESULTS 3.1. UnmodiJied and fully modt$ed structures In order to establish a basis for microstructural comparison a number of pure binary samples (no strontium added) were solidified at various growth rates. The lowest growth rate (R = 1.25 pm/s) produced a microstructure containing a mixture of complex-regular, flake and a coarser plate-like silicon, as shown in Fig. l(a). This is consistent with

results of other workers [4-6] for samples produced under similar growth conditions (although in references 4 and 5 and the plate-like silicon is referred to as “angular silicon”). At R > 10 pm/s complexregular silicon is no longer observed; rather the structure consists of flake silicon which is present in colonies and plate-like silicon which is dispersed randomly between these colonies, Fig. l(b). Quenched interfaces on longitudinal sections reveal that the plate-like silicon grows ahead of the flake colonies in a similar manner to a primary phase, Fig. l(c). The individual flake colonies form at what appears to be a more closely coupled interface; the flakes themselves being relatively parallel to one another. Further increases in growth rate up to about 70pm/s do not lead to morphological changes but merely result in a refinement of the type of structure seen in Fig. l(b) and (c). Samples containing relatively high levels of strontium (> 0.040 wt%) were also solidified over a range of growth rates; this produced a typical modified fibrous structure as shown in Fig. l(d). 3.2. Partially modljied structures Results obtained with the AGT suggested that partially modified structures could be obtained at strontium levels less than 0.014 wt% and solidification rates between 1 and 25 pm/s. Approximately 30 different specimens were subsequently solidified at constant growth rates, with the resulting structures classified as a function of both strontium level and growth rate, Fig. 2. The strontium concentration for each sample was determined exactly at the point of structural observation. This is important to note since the present study found that the strontium level often varied considerably along the length of a unidirectionally solidified rod, despite the fact that the master alloys were always uniform with respect to strontium concentration. This variation could be attributed in part to the loss of strontium at the top of the rod, which occurred despite the presence of an inert atmosphere. A more important factor, however, was the tendency for significant impurity segregation to occur, particularly at low growth rates. This often precluded the possibility of obtaining certain points on the graph (Fig. 2) for example, numerous attempts were made to grow a sample containing 0.010 wt% Sr at 1.5 pm/s, however despite the use of master alloy levels up to 0.014 wt% Sr, the resulting unidirectionally solidified alloy always contained less than 0.005 wt% Sr due to segregation ahead of the unmodified interface. Figure 2 indicates that, for the growth rates under consideration, the unmodified eutectic can accommodate up to 0.004 wt% Sr with no structural changes apparent (Region A). The transition to a fully modified state (Region C) can be seen to be more sensitive to the growth rate, e.g. 0.006 wt% Sr can produce a fibrous structure at 27 /*m/s, yet twice that amount is required to achieve the same result at


and SMITH:



Fig. 1. Microstructures of unidirectionally solidified Al-Si eutectic specimens (please note that silicon is the dark phase in all optical micrographs): (a) unmodified, R = 1.25 pm/s, transverse section; (b) unmodified, R = 28 pm/s, transverse section; (c) unmodified, R = 28 pm/s, longitudinal section including the quenched interface (i.e. the liquid region has been quenched in order to delineate the structure at the solid/liquid interface). Direction of interface movement is bottom to top; (d) fully modified with 0.040 wt% Sr, R = 5 pm/s, transverse section. I


A. unmodified







modified -mixed 0 -banded I -colony. C. fully modihed +

7 pm/s. In all of the fully modified samples the fibrous morphology appeared identical in form to that shown in Fig. l(d). Partially modified structures occupy the region between the modified and unmodified bounda~es in Fig. 2. (Region B). The partially modified eutectic was observed to assume three different forms depending on the growth rate; in the present study these have been termed:

mixed, banded, colony,

analysed strontium concentration


Fig.2. AI-Si eutectic structural types corresponding to various strontium concentrations and growth velocities.

R < 5 pm/s, 5 G R < 15 pm/s, and, R > 15 pm/s.

Each of these structural types will now be considered separately. 3.2. I. Mixed partially modtjied. In the mixed partially modified state the flake colony silicon is no longer present; rather the structure consists of a mixture of the more massive plate-like silicon and



and SMITH:





alloys containing very high levels of sodium [8,9]. Although no quantitative measurement was undertaken, in general these bands were larger and less numerous at lower growth rates, with thinner bands occurring with increased frequency at higher rates. The banded microstructure itself is best described as one in which the modified structure is continuous and the unmodified discontinuous. Figure 4(b) is a typical micrograph from a banded partially modified sample. In order to ensure that these bands were not an artifact structure resulting from the Bridgman growth technique, samples were prepared and solidified using a three-zone temperature gradient furnace. This unidirectional solidification method, unlike the Bridgman technique, utilises no mechanical movement, with solid/liquid interface advance induced by the migration of a furnace-controlled temperature gradient along the sample. Bands identical to those shown in Fig. 4(a) and (b) were observed following the solidfication of alloys containing appropriate

Mixed partially modified AI-B eutectic structure, R = 3 pm/s: (a) transverse section, (b) Longitudinal section including the quenched interface, growth direction bottom to top. Fig. 3.

modified silicon, as shown in Fig. 3(a). The complexregular form of silicon was also observed occasionally along the length of these samples, although not as frequently as in the pure binary eutectic specimens at low growth rates. Hellawell also noted the coexistence of complex-regular and modified fibrous silicon in sodium-modified alloys [7]. Figure 3(b) illustrates a typical quenched interface from one of these samples; it is clear from this that the unmodified structure forms some distance ahead of the main interface, which is fibrous in nature. 3.2.2. Banded partially modijied. Banded partially modified microstructures are generally observed when the growth rate exceeds 5 pm/s but is less than about 15 pm/s. These bands are unlike mechanicallyinduced artifact bands, in that they are often concave upwards in appearance and do not extend across the entire cross-section of the sample, Fig. 4(a). Their distribution is, in fact, similar to the “overmodification” bands typically observed in Al-Si


Fig. 4. Banded partially modified AI-Si eutectic structure, R = 15 pm/s, growth direction bottom to top: (a) macrophotograph of longitudinal section showing band distribution and shape; (b) micrograph, longitidunal section.


and SMITH:


Fig. 5. Partially modified AlmSi casting. exhibiting banded region between adjacent flake colonies. Growth direction is from bottom to top. amounts of strontium. This result indicates that the bands are not unique to the Bridgman method of growth. Examination of partially modified castings also revealed that similar banding could be produced in cast microstructures. These bands were most frequently observed in modified regions between two adjacent flake colonies growing away from the casting wall. An example of these can be seen in Fig. 5. The presence of partially modified bands in castings and in samples produced by stationary and moving unidirectional solidification techniques virtually eliminates the possibility that they may be an artifact structure, however it does not explain why they form. According to the literature [lo], the modified structure should grow at a slightly higher undercooling than the unmodified. It is therefore feasible that, if the growth mode is alternating, then the interface may be hesitating periodically as it advances. If this is the case then it should be possible to detect interface hesitation through the measurement of temperature fluctuations in the vicinity of the interface during solidification. A fine chromelalumel thermocouple was inserted into an alloy conwas subsequently taining 0.009 wt% Sr, which solidified at lOpm/s. In order to differentiate legitimate temperature variations from system noise, the direction of sample movement was reversed after the interface had moved 2 cm past the thermocouple tip. This resulted effectively in the sample being “melted back” at the same rate as it was originally solidified. Since both the flake and fibrous structures melt at the same temperatures the resulting thermocouple trace



could be compared to the one produced during solidification, with any differences likely to be a result of the banding behaviour. Comparison of the solidification and melting curves, however, revealed no differences which could be attibutable to banding. This result indicated that, if the interface was hesitating, then the temperature variations associated with the fluctuations were too minute to be detected using a conventional thermocouple which is accurate to f0.5’C. Scanning electron microscopy was utilised to more carefully examine the transition behaviour between the flake and fibrous structure at the band boundary. Observations of a number of these regions indicated that the two types of silicon are interconnected, however details were difficult to determine because the silicon fibres could seldom be followed for more than short distances. In general the fibres were observed to branch into a different growth direction from that of the flake, Fig. 6. It is also interesting to note that the flakes and fibres in these samples have approximately the same width, which seems to facilitate the transformation from one structural form to the other. The evidence presented thus far regarding partially modified banded microstructures indicates that the transition between flake and fibrous growth occurs without significant difficulty. It does not, however, suggest a reason why the banding occurs, nor why it is only observed over a limited range of growth rates. Further consideration of this behaviour will be undertaken in the section 4. 3.2.3. Colony partially modified. The banded microstructure gives away to a colony form when the freezing front velocity exceeds about 15 pm/s. These structures are characterised by relatively large flake silicon colonies which grow along the sample and are separated by smaller regions of modified structure. The flake and fibrous colonies form adjacent to one another: as a result longitudinal sections generally

Fig. 6. Scanning electron micrograph of a silicon flake transforming to a number of fibres at a flakeifibre boundary in a banded sample. Growth direction top right to bottom left. the bar reoresent IO urn.



and SMITH:


exhibit extended axial regions of the two different structures, although both change in size and position along the sample length. Quenched interfaces reveal that the flake colonies grow slightly ahead of the modified structure and have a rounded “cap-like” cross-section similar to that observed in castings, Fig. 7. The distance between the flake and fibrous growth fronts is approximately I50 pm which, at this temperature gradient (G, = 230”C/cm), corresponds to about 3.5”C. The experimental results of Glenister and Elliott [lo] suggest that at this growth velocity the di~eren~ in under~ooling ~tw~n the two growth structures should be 2-3”C, in reasonable agreement with this result. The shape of the interface in these samples is similar to that of cellular lamellar eutectics (111, where the addition of an impurity which has limited solubility in both of the solid eutectic phases leads to constitutional undercooling and cellular instability at the solid/liquid interface. However, a more careful investigation of the partially modified samples revealed the following: (A) The “cell” size in these partially modified samples is much larger than that of other cellular eutectics at similar growth rates 11I], with the average cross sectional area of these regions being approximately five times that which is commonly observed. In addition, eutectic cells which form as a result of constitutional undercooling establish a relatively uniform cell size which is constant for a particular set of growth conditions [I I], The eutectic colonies in these partially modified Al-B alloys are not of uniform size, i.e. in any particular cross section the area of individual cells may vary by as much as an order of magnitude. (B) The strontium levels present in these alloys (-0.006 wt%) are too low to be able to produce cellular growth according to theoretical constitutional undercooling arguments [ 121. Calculations for this system (see Appendix) indicate that in order for cellular growth to occur the strontium concen-

Fig. 7. Quenched interface (longitudinal section) from a colony partially modified sample, showing a modified region between two flake colonies. R = 20 pm/s, temperature gradient GL = 230”C/cm.


tration would have to exceed approximately 0.025 wt% (even if this level were exceeded, the scale the microstructure would be approximately 20% of that actually observed). The evidence indicates that these structures are not a direct rest& of const~tutionai undercooling, but rather that they may be similar to the type of colony structure seen in partially modified castings [I]. Each colony is itself a eutectic grain, having nucleated from a single site. In unmodified AI-Si eutectic samples it is generally impossible to differentiate between individual grains because of the irregular nature of the grain boundary. In partially modified cast samples the strontium accumulation in the terminal liquid regions results in a modified silicon morphology at the flake colony boundaries, therefore it is possible to easily identify these colonies. In the case of unidir~tional solidification the interface moves vertically upwards, with the flake coionies growing along the length of the sample. Strontium rejected into the liquid ahead of the growing flake structure increases the local liquid density. This liquid then tends to flow laterally into any chance depression in the solid~liquid interface. When the s~ontium level becomes large enough in these depressions a modified structure will form. However, since the local modified eutectic front requires a slightly higher undercooling than the flake eutectic front at these growth rates (2-3”(Z), the depression tends to retire even further from the macroplanar flake eutectic interface, thus providing a larger degree of stability to the depression. 3.3. interphase spacing measurements in modified and unmodified samples ~eas~ements of eutectic interphase spacings (2) appear frequently in the literature, however as yet no standardized method has been developed which can be applied to give a representative 1 value for all of the different structural types. Interphase spacing measurements are most easily obtained on uni~r~tionally grown lamellar eutectics, where phases are arranged in a regular manner with a common growth direction. In these samples the line intercept method is generally applied within single eutectic grains (i.e. a region of parallel lamellae) on a section transverse to the growth direction. Irregular eutectics such as AI-Si present a problem in that the structure is much less regular, making exact dete~ination of the interphase spacing difficult. As discussed previously, the characteristic microstructure of unidirectionally grown Al-Si eutectics (R > 10 @m/s) consists of colonies of semi-regular silicon flakes, between which are often present large plate-like silicon particles [Fig. l(b) and (c)J. In other studies, interphase spacing measurements on unidirectionally grown, unmodified AI-Si samples have been determined by applying the line intercept method randomly across a transverse sample section, including


and SMITH:


flake colonies and also plate-like silicon particles [13]. In the present study the line intercept method was used to determine silicon phase spacings in unmodified Al-Si, however, measurements were taken within individual flake eutectic colonies rather than across random sections. This method is believed by the present authors to produce a more realistic result, since the flake colony structure is more representative of coupled eutectic growth than plate-like silicon, which grows ahead of the interface in a manner similar to that of a primary phase [Fig. l(c)]. As in other studies [6, lo], fibre spacings were determined using a standard method, i.e. in a transverse section the number of particles N in an area A was counted, with the interfibre spacing (A) calculated according to: 1 =JA/N.



I ,

100 VELOCITY, urn ,-’

Fig. 9. Graph of undercooling versus growth rate for flake and fibrous Al-Si eutectic samples, reproduced from Glenister and Elliott [lo]. The open circles represent flake silicon data and the filled circles fibrous silicon data. Note that at R = 10 pm/s the undercoolings for the two different

types of structure are nearly identical.

Figure 8 shows the interphase spacing measurements for flake and fibrous silicon solidified at various growth rates. It should be noted that measurements were taken from steady state regions of unidirectionally solidified samples (transverse sections), with flake spacings measured from pure binary samples and fibrous spacings from fully modified samples (i.e. no partially modified samples were used for spacing measurements). The most notable feature of this graph is that the lines for flake and fibrous spacings cross at a growth rate of approximately 8 pm/s (log R = 0.9). This suggests that, at this solidification rate, the fibrous and flake eutectic modes experience similar conditions for growth. It also indicates that below R = 8 pm/s the fibrous structure is actually more coarse than that of the flake. It can be argued that a




flake silwn

direct comparison of the relative refinement of the two different structures is not possible, and the authors recognise the problems involved in such a comparison. A graph of undercooling vs growth rate would yield a more accurate result because no structural interpretation is necessary, i.e. if the two different structures were found to have the same undercooling at this growth rate (N 8 pm/s) then this would confirm that growth conditions are indeed similar. Although no undercooling values were measured in the present study, Glenister and Elliott [IO] very carefully determined the growth undertoolings for both flake and fibrous structures at a number of different growth rates. The graph representing the results of their work (reprinted as it appears in Ref. [lo]) is shown in Fig. 9. The results of particular interest to the present study lie in the low velocity region of this graph. The data points indicate that as the growth rate is lowered the measured undercoolings of the two structural types converge, becoming almost identical at 10 pm/s. This is very similar to the growth rate at which the lines for the particle spacing measurements intersect in Fig. 8. Thus, these two different sets of results, from independent studies, support the proposal that the flake and fibrous forms of the eutectic both require similar conditions for growth at solidification rates of approximately 8-10 pm/s.







log R

Fig. 8. Interphase spacing (A) of fibrous and flake silicon as a function of growth rate R (the units of h are pm; R is measured in urn/s. flake silicon = oDen circles fibrous &con = solid circle&

The Al-Si eutectic can apparently accommodate up to 0.004 wt% strontium without exhibiting any apparent structural changes. Concentrations higher than 0.012 wt% Sr, however, are sufficient to produce a fully modified morphology. Between these two limits (at growth rates less than 30pm/s) lies the region of partial modification, in which the structure is very sensitive to the imposed growth conditions. Concurrent work [3] on the segregation behaviour of




strontium in the Al-Si eutectic has indicated that in this growth rate range the effective distribution coefficient (keK)for strontium lies between 0.2 and 0.3 for unmodified alloys, compared to a value of 0.9-l .O for modified alloys. This information, combined with the results shown in figure 8 and 9 is sufficient to provide a logical explanation for the different types of structure observed in partially modified, unidirectionally grown samples. As illustrated in Fig. 2, banded partially modified microstructures are observed when R = 5-15 pm/s. This corresponds to log R = 0.7&l. 18, which can be seen to span across the region of Fig. 8 where the intersection of the two lines is observed. Within this range the two different structures have similar requirements for growth and as a result interphase spacings and undercoolings are similar. This may explain why no temperature fluctuations could be detected during banded sample growth, and also why the widths of the flakes and fibres in the bands are similar when viewed in the SEM. It appears, therefore, that when the strontium level lies in the “critical” range, which from Fig. 2 is seen to be 0.005-0.012 wt% Sr, the growth mode may fluctuate from modified to unmodified depending on the local concentration of strontium immediately ahead of the solid/liquid interface. At growth rates between 5 and 15 pm/s this fluctuation can occur quite readily because growth requirements (e.g. diffusion considerations) are nearly identical under these solidification conditions. The evidence suggests, therefore, that the banded structure observed in the partially modified alloys is the structural response of the system to slight local variations in the strontium level. These variations may be the result of convection within the liquid region of the sample, since even minor currents may be sufficient to deplete the strontium boundary layer in localised areas. Alternatively, the structure may be a result of the difference in keRbetween the flake and fibrous structures. During flake growth significant levels of strontium are rejected ahead of the interface into the liquid, which enhances the strontium boundary layer. If the strontium concentration directly ahead of the interface becomes too high to permit further growth of flake silicon, then the eutectic will change to a fibrous mode. The fibrous structure, however, has a higher k,, value than the flake, and will therefore likely deplete the boundary layer of strontium; if this depletion is severe enough then fibre growth may no longer be possible and the system will revert back to a flake growth mode. At solidification rates of R < 5 pm/s the conditions for flake and fibrous growth are not similar (Fig. 8), as a result an alloy containing low levels of strontium does not have the option to “flip” from one growth form to the other and banding is not observed. Mixed structures are usually seen in these samples as shown in Fig. 3(a) and (b). Although the exact characteristics of the strontium boundary layer are not known

under these conditions, the level of strontium will be higher directly at the interface than it is in the bulk liquid further ahead of the interface. Complexregular and plate-like silicon, both of which remain unmodified in these samples, are observed to grow ahead of the modified eutectic interface as shown in Fig. 3(b), therefore they are forming from a liquid which has a relatively low strontium concentration compared to that immediately at the interface. The avoidance of the high strontium region enables them to maintain an unmodified form. However, at higher strontium levels, e.g. 0.029 wt%, the entire sample area assumes a modified form, suggesting that the bulk liquid concentration is high enough to suppress unmodified silicon formation. Banding is also not observed in partially modified samples solidified at R > 15 pm/s since, as for R < 5 pm/s, growth requirements are not the same for the flake and the fibrous mode (Fig. 8). In this case the structural development is similar to that of partially modified castings, with flake colonies rejecting strontium ahead of the advancing interface. This rejected strontium concentrates in regions adjacent to the flake colony which grow in a modified form. Evidence of the difference in growth requirements between the two structural types can readily be seen in these samples since the modified interface lags that of the unmodified by about 3°C. Finally, the results of the present study suggest that growth rate does influence impurity modification behaviour (Fig. 2), however this is probably related to a segregation effect. The silicon growth mode is highly sensitive to the amount of strontium immediately ahead of the interface, and the results of the present study suggest that the fibrous form is favoured when the strontium concentration exceeds a critical level. Increasing the solidification rate should lead directly to higher interface strontium concentrations [12], i.e. at higher growth velocities this critical interface concentration may be achieved in alloys containing lower bulk strontium concentrations. This and other [l-3] studies on the partially modified AI-Si eutectic have provided useful and interesting information regarding the segregation behaviour of the strontium modifier during solidification. The results also indicate that the flake-to-fibrous transition is associated with a critical level of strontium at the solid/liquid interface, and that the boundary between the two types of silicon is discrete, with no gradual structural change apparent. This evidence strongly supports the hypothesis that modification is a result of impurity element adsorption at the silicon/surface. The most dominant facets in unmodified primary and eutectic silicon are associated with {111) planes [9, 141. This suggests that the {11l} planes have the lowest surface energy and therefore grow the most slowly, eventually bounding the crystal [15]. An impurity which is positively adsorbed at an interface should lower the





surface tension of that interface according to Gibb’s adsorption isotherm, i.e.

where e is the surface tension, r, is the adsorption of impurity i at the interface, and pi is the chemical potential of i. In the case of the silicon phase in AI-5 alloys, positive adsorption of an impurity such as strontium should reduce the surface tension (and therefore surface energy) of other planes relative to the {11 I}-type, allowing these other planes to “compete” as bounding planes on the external surface. Fredriksson et al. [9] showed this to be the case for primary silicon; in the presence of sodium, facets additional to the {111l-type were observed on silicon particles growing freely from the melt. Similarly, recent studies [2, 16. 171have indicated that impurity modified fibrous eutectic silicon is bounded by numerous microfacets, in contrast to unmodified flake silicon which has only one significant faceted surface corresponding to a {ill}-type plane. The adsorption argument is also consistent with the observation that a critical level of strontium is necessary in order for a modified structure to form. Progressively higher levels of strontium should gradually lower the surface energies of other silicon planes. These will not be observed to be bounding faces, however, until their surface energies are comparable to that of {Ill}-type planes; this would be expected to occur at a critical strontium concentration. 5. CONCLUSION

The present study has shown that in unidirectionally solidified samples, as in castings, the partially modified AI-Si microstructure consists of discrete regions of unmodified and modified silicon. The distribution of the two different types of structure is dependent on the growth rate and has been described as being either of a mixed, banded or colony form. The presence of these three types of structure at different growth rates has been rationalised by considering the growth requirements for the flake and fibrous silicon, in combination with information regarding the segregation behaviour of strontium in these alloys. The results suggest that a critical strontium level is required for modification to occur; this is consistent with the hypothesis that modification is a result of impurity adsorption onto the faceting phase during growth. TM. D. Hanna and A. Hellawell, MRS Symp. Proc. 19,411 (1983). $B. Clossct, H. Dugas, M. Pekguleryuz and J. E. Gruzleski, MetaN. Trans. 17A, 1250 (1986).





Acknowledgements-The authors would like to thank Dr M. A. Savas for his help and useful discussions. L. Clapham has been supported by an Ontario Graduate Scholarship during the course of this work.

REFERENCES and R. W. Smith, J. Crystal Growth 79,866 (1986). L. Clapham, Ph.D. thesis, Queen’s Univ. Kingston, Canada (1987). L. Clapham and R. W. Smith, J. Crvsral Growth. To be published. 0. A. Atasoy. F. Yilmaz and R. Elliott, J. Crystal Growth 66, 137 (1984). 0. A. Atasoy, klumi~um 59, E530 (1983). D. C. Jenkinson and L. M. Hogan, J. Crystal Growth 28, 171 (1975). A. Hellawell, Prog. Mater. Sci. 15, 1 (1973). K. Kobayashi, P. H. Shigu and R. Ozaki, Trans. Japan Inst. Metals 21, 417 (1980). H. Fredriksson, M. Hillert and N. Lange, J. Inst. Metals 101, 285 (1973). S. M. D. Glenister and R. Elliott, Metal Sci. 15, 181 (1981). W. M. Rumball, Metallurgia 78, 141 (1968). W. A. Tiller, J. W. Rutter, K. A. Jackson and B. Chalmers, Acta metall. 1, 428 (1953). S. C. Flood and J. D. Hunt, Mefal Sci. 15, 287 (198 I). M. G. Day and A. Hellawell, Proc. R. Sot. 305A, 473 1968. G. Wulff, Z. Krist. 34, 449 (1901). Shu-Zu Lu and A. Hellawell, J. Cryst. Growth 73, 316 (1985). Shu-Zu Lu and A. Hellawell, Metall. Trans. 18A, 1721 (1987).

1. L. Clapham 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17.

APPENDIX Delermination of Approximule Strontium Concentration to Produce Cells in a Uni-Directionally Solid$ving AILSi Eufectic Sample From Tiller et al. [12]. for cell formation: c,>c’D fi

x-- k, (1 -k,)

where C,, and k, refer respectively to the bulk concentration and the overall equilibrium distribution coefficient of strontium in an Al-Si eutectic alloy, m is the slope of the liquidus line in the phase diagram, and D is the diffusion coefficient of strontium in the liquid alloy. For colony formation: G, = 230”C/cm R = 0.00278 cm/s D _ 5 x lo-’ cmZ/s m N 1.7”C/wt%


The value of k, for strontium in AI-Si is not known, however it can be assumed that its lowest value (giving the lowest value of C, for cell formation) would be k, = 0.01 since this is the value for strontium in aluminum:. Using the above values for G,, D, R, m and k,, the lowest bulk concentration of strontium that could produce cells in AI-Si would be C, _ 0.025 wt% Sr.