Solidification behaviour of AZ91D alloy under intensive forced convection in the RDC process

Transcription

1 Acta Materialia 53 (5) Solidification behaviour of AZ91D alloy under intensive forced convection in the RDC process Z. Fan *, G. Liu BCAST (Brunel Centre for Advanced Solidification Technology), Brunel University, Uxbridge, Middlesex UB8 3PH, UK Received 17 April 5; received in revised form 24 May 5; accepted 25 May 5 Available online 14 July 5 Abstract Rheo-diecasting (RDC) is a new semisolid processing technology for production of near net shape components. In this work, the solidification behaviour of AZ91D alloy under intensive forced convection in the RDC process was investigated experimentally to understand the effects of the intensity of forced convection, shearing time and shearing temperature on the nucleation and growth behaviour. It was found that under intensive forced convection, heterogeneous nucleation occurred continuously throughout the entire volume of the solidifying melt. All the nuclei could survive due to the uniform temperature and composition fields created by the forced convection. This has been named as continuous effective nucleation. It is also found that the nuclei grow spherically with an extremely fast growth rate. This makes the primary solidification essentially a coarsening process, in which Ostwald ripening takes place by dissolution of the smaller particles. Secondary solidification of the intensively sheared semisolid slurry takes place also through effective nucleation, but with dendritic growth. Increasing the intensity of forced convection enhances nucleation and promotes the formation of the primary phase during the secondary solidification in the shot sleeve. The final solidification microstructure is strongly dependent on the presence of turbulence rather than the shear rate. Ó 5 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Solidification; Nucleation; Growth; Forced convection; Magnesium 1. Introduction Semisolid metal (SSM) processing opens a new dimension for solidification control through externally applied forced convection [1 3]. It pushes the boundary of solidification science from the conventional solidification under static conditions, to solidification under controlled dynamic environment [3]. Technologically, it promises to deliver new casting processes, which are capable of producing cast components with high integrity, fine and uniform microstructure, and therefore enhanced performance. Scientifically, it provides the scientific community with new approaches to study * Corresponding author. Tel.: ; fax: address: Zhongyun.Fan@brunel.ac.uk (Z. Fan). nucleation and crystal growth under controllable dynamic conditions. Historically, research on SSM processing has been mainly concentrated on technological development of casting techniques, research on nucleation and growth under forced convection has been very limited [1 3]. Limited understanding of nucleation rate, growth morphology and the mechanisms for the formation of the globular structures has been achieved [4]. The conventional belief [1] is that under forced convection the initial dendrites would fragment through either the bending of dendrite arms followed by liquid penetration of the high angle grain boundaries [5], or through remelting at the root of dendrite arms due to solute enrichment and thermal-solutal convection [6]. The detached dendrite arms then undergo a coarsening process to provide the observed /$. Ó 5 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:.16/j.actamat

2 4346 Z. Fan, G. Liu / Acta Materialia 53 (5) globular particles. Although this dendrite fragmentation theory provides a mechanism for grain multiplication under forced convection, it is difficult to explain why the detached dendrite arms would grow into globules rather than dendrites. More importantly, the dendrite fragmentation theory lacks of convincing experimental evidence. The early theoretical investigation on the subject by Vogel and Cantor [7] used a stability analysis and concluded that laminar melt flow destabilises the solid liquid interface and promotes dendritic growth with an enhanced growth rate. The effects of laminar flow on the solidification morphology have also been studied theoretically by an analytical approach with Oseen viscous flow approximation [8], by the cellular automaton model [9], and by phase-field models [ 13]. More recently, the belief has been that the globular structure is more likely to be a result of spherical growth under forced convection, rather than a consequence of dendrite arm detachment [3]. Molenaar et al. [14,15] proposed that the growth is cellular, based on their experimental observations, and that for alloys with a low Prandtl number the thermal boundary layer is hardly affected by stirring, while the hydrodynamic boundary layer is significantly reduced and mass transport is dominated by convection. Ji and Fan [16] experimentally investigated the effect of turbulent flow on growth morphology and concluded that the globular structure is a direct result of spherical growth under intensive forced convection, and no dendrite or dendrite fragments were ever observed. Based on their experimental results, they proposed that the growth morphology changes from dendrite to sphere via rosette with the increasing shear rate and degree of turbulence. This is in good agreement with the theoretical analysis by Qin and Fan [17] using stability analysis and boundary element method, by Das et al. [18] using a Monte Carlo simulation technique and by Qin and Walleck [19] using phase field modelling. However, the nucleation behaviour and coarsening mechanisms during isothermal shearing are still not clear. It is necessary to conduct a systematic investigation on the solidification behaviour under intensive forced convection. This work aims to understand the solidification behaviour of AZ91D alloy under intensive forced convection provided by the twin-screw melt shearing device, used in the rheo-diecasting (RDC) process developed recently by BCAST at Brunel University []. Experiments were carried out to investigate the effects of the intensity of forced convection, shearing time, and shearing temperature on the nucleation and growth behaviour of AZ91D alloy in the RDC process. Experimental results will be discussed in the context of the current understanding of the subject from both theoretical analysis and experimental investigations in the literature. 2. Experimental The RDC process is an innovative one-step SSM processing technique for manufacturing near net shape components of high integrity directly from liquid alloys []. The process innovatively adapts the well-established high shear dispersive mixing action of the twin-screw mechanism to the task of in situ creation of SSM slurry, with fine and spherical solid particles, followed by direct shaping of the SSM slurry into a near-net shape component using the existing cold chamber high pressure die casting (HPDC) process. A detailed description of the RDC process can be found elsewhere [21,22]. The RDC equipment consists of two basic functional units, a twin-screw slurry maker and a standard cold chamber HPDC machine. The twin-screw slurry maker has a pair of screws rotating inside a barrel. The screws have specially designed profiles to achieve co-rotating, fully intermeshing and self-wiping. The basic function of the twin-screw slurry maker is to convert the liquid alloy into high quality semisolid slurry through solidification of the liquid alloy under high shear rate and high intensity of turbulence. During the slurry making process, there is an enormous amount of ever-changing interfacial area between the solidifying alloy and the slurry maker, providing an enhanced heat transfer. To avoid oxidation, a protective gas mixture of N 2 containing.4vol.%sf6 was used in the melting furnace and the twin-screw slurry maker. The fluid flow inside the slurry maker is characterised by high shear rate, high intensity of turbulence, and a cyclic variation of shear rate. It is, therefore, rather complex and difficult to quantify. For pure laminar flow, the shearing intensity can be described by shear rate. However, the shear rate in the twin-screw slurry maker is continuously changing; an elemental volume of alloy melt will experience a cyclic variation of shear rate, with the shear rate being highest at the intermeshing region, and lowest between the screw root and the barrel s inner surface. The degree of turbulence is even more difficult to quantify. However, it is important to point out the following facts: Both shear rate and degree of turbulence is proportional to the screw rotation speed; the higher the screw rotation speed, the higher the shear rate and the degree of turbulence. The flow is turbulent even at low screw rotation speed; Reynolds number cannot be used to predict the transition from laminar flow to turbulent flow. Both shear rate and the degree of turbulence are functions of the screw profile design and screw arrangement. Therefore, for simplicity, screw rotation speed is used in this investigation as a measure of the intensity of forced convection.

3 Z. Fan, G. Liu / Acta Materialia 53 (5) During the RDC process, a predetermined dose of liquid alloy from the melting furnace is fed into the slurry maker. The liquid alloy is continuously cooled to the SSM processing temperature while being mechanically sheared by a pair of closely intermeshing screws, converting the liquid into semisolid slurry. The fraction of solid of the semisolid slurry is controlled by setting the barrel temperature. The semisolid slurry is then transferred to the shot chamber of the HPDC machine for component shaping. AZ91D alloy used in the present study was provided by Magnesium Elektron Ltd (Manchester, UK) and the alloy composition is listed in Table 1. The Mg alloy ingots were melted at 675 C and fed into the slurry maker at 6 C. The slurry maker was operated at a temperature range between 585 and 6 C. The rotation speed of the twin-screw was varied between and 9 rpm, and the shearing time between 3 and s. The AZ91D alloy melt was sheared in the slurry maker for a predetermined period of time and then transferred into a 28-ton cold chamber HPDC machine (LK Machinery, Hong Kong), to produce standard tensile test samples. The die used for casting has six cavities, of which four are tensile test samples and two are Charpy test samples. The dimensions of the tensile test samples are 6mm in gauge diameter, 6 mm in gauge length and 1 mm in total length. For all the experiments in this investigation, the die temperature was kept at 2 C. Specimens for microstructural characterisation were cut from the middle section of the tensile test samples. The microstructure of the alloy was examined by optical microscopy (OM), with quantitative metallography, and scanning electron microscopy (SEM). The specimens for OM and SEM were prepared by the standard technique of grinding with SiC abrasive paper and polishing with an Al 3 O 2 suspension solution, followed by etching in an aqueous solution of 6 vol.% ethylene glycol, vol.% acetic acid, 1 vol.% concentrated HNO 3. A Zeiss optical microscope was utilised for the OM observations and the quantitative measurements, while a Jeol JXA-84A scanning electron microscope, equipped with an energy dispersive spectroscopy (EDS) facility, was used to perform the SEM examinations with an accelerating voltage of kv. The shape factor (F) was calculated using F =4pA/P 2, where A and P are the total area and the peripheral length of the primary particles, respectively. Therefore, F is 1 for perfectly spherical particles. 3. Results 3.1. General microstructure of the RDC AZ91D alloy Solidification takes place in the RDC process in two distinct stages, as illustrated schematically in Fig. 1. Solidification inside the twin-screw slurry maker under intensive forced convection to produce semisolid slurry is referred to as primary solidification, while the solidification of the remaining liquid inside the shot sleeve and the die cavity without shearing is referred to as secondary solidification. The primary solidification can also be divided into two sub-stages, namely, continuous cooling from the pouring temperature to the SSM processing temperature, and isothermal shearing at the SSM processing temperature. The secondary solidification process includes solidification occurring during the slurry transfer to the shot sleeve, mould filling and solidification in the die cavity. Fig. 2 shows the typical microstructure of the AZ91 D alloy produced by the RDC process. The relatively large and spherical particles are primary a-mg particles (a 1 ) produced inside the twinscrew slurry maker under high shear rate and high intensity of turbulence. Secondary solidification began when the semisolid slurry was transferred to the shot sleeve. Due to the relatively low temperature of the shot sleeve T Primary solidification Continuous cooling Isothermal shearing Slurry transfer Secondary solidification Mould filling Solidification in die cavity Time Solid state cooling Solid state cooling Fig. 1. Schematic illustration of the solidification process in the RDC process. Table 1 Chemical composition of the AZ91D alloy ingot used in this work (in wt.%) Zn Al Si Cu Mn Fe Ni Be Others each < <.1

4 4348 Z. Fan, G. Liu / Acta Materialia 53 (5) Fig. 3. SEM micrograph of the RDC AZ91D alloy showing the fine a-mg grains (dark region, and denoted as a 3 ) formed inside the die cavity during the secondary solidification. The white phase is the eutectic b-phase. Fig. 2. The typical microstructure of AZ91D alloy produced by the RDC process. The large globular primary solid particles (denoted as a 1 ) are formed inside the twin-screw slurry maker during the primary solidification, and the dendrite fragments (denoted as a 2 ) are formed inside the shot sleeve during the secondary solidification. (a) Low magnification; (b) higher magnification. (usually below 4 C), further volume fraction of the primary a-mg phase was produced in the remaining liquid. Such particles usually have a dendritic morphology and are denoted as a 2. The semisolid slurry is then delivered to the die cavity with a controlled velocity (usually a few m/s) and pressure. The dendrites formed in the shot sleeve are fragmented when they pass through the narrow gate, resulting in the dendrite fragments observed in the final microstructure (see Fig. 2(b)). The remaining liquid in the semisolid slurry then solidified in the die cavity under high cooling rate, typically in the order of 3 K/s, provided by the metallic die block. The resulting microstructure is fine a-mg particles (a 3 ) delineated by the eutectic b-phase (see Fig. 3) Effects of shearing time Experiments were carried out at a shearing temperature of 593 C to investigate the effects of shearing time on the formation of the primary phase inside the twinscrew slurry maker. The measured volume fraction of the primary a-phase is plotted in Fig. 4 as a function of shearing time and screw rotation speed. Fig. 4 indicates that solidification inside the twin-screw slurry maker occurred in two stages. The first stage is a continuous cooling process, where the initially superheated melt was cooled continuously to the semisolid processing temperature, producing the desired volume fraction Volume Fraction of α 1 (%) Continuous cooling Isothermal shearing 1 Shearing Time (s) 8rpm rpm Fig. 4. Volume fraction of the primary particles (a 1 ) formed in the twin-screw slurry maker as a function of shearing time and screw rotation speed. The shearing temperature was 593 C. The solid lines represent the best fit to the experimental data. Also marked in the figure is the boundary between continuous cooling and isothermal shearing during the primary solidification.

5 Z. Fan, G. Liu / Acta Materialia 53 (5) Particle Size of α 1 (µm) Continuous cooling 1 Shearing Time (s) Isothermal shearing 8rpm rpm Power (rpm) Power (8rpm Fig. 5. Size of the primary particles (a 1 ) formed in the twin-screw slurry maker as a function of shearing time and screw rotation speed. The shearing temperature was 593 C. The solid lines represent the best fit to the experimental data. Also marked in the figure is the boundary between continuous cooling and isothermal shearing during the primary solidification. Shape Factor of α Continuous cooling 1 Shearing Time (s) Isothermal shearing 8rpm rpm Fig. 6. Shape factor of the primary particles (a 1 ) formed in the twinscrew slurry maker as a function of shearing time and screw rotation speed. The shearing temperature was 593 C. The solid line represents the best fit to the experimental data. Also marked in the figure is the boundary between continuous cooling and isothermal shearing during the primary solidification. Particle Density of α 1 (mm -2 ) Continuous cooling 1 Shearing Time (s) Isothermal shearing 8rpm rpm Fig. 7. Density of the primary particles (a 1 ) formed in the twin-screw slurry maker as a function of shearing time and screw rotation speed. The shearing temperature was 593 C. The solid lines represent the best fit to the experimental data. Also marked in the figure is the boundary between continuous cooling and isothermal shearing during the primary solidification. of the solid particles. This stage lasted about 15 s, and solid volume fraction increased with the increase of shearing time. Once the melt reaches the semisolid temperature, it experiences an isothermal shearing process, where solid volume fraction was fairly constant. At the continuous cooling stage, screw rotation speed had little effect on the solid volume fraction, but it did affect the final fraction of solid before the isothermal shearing started. Higher screw rotation speed led to a smaller fraction of solid, suggesting that intensive shearing suppresses partially the formation of the primary phase. Figs. 5 and 6 present the particle size and shape factor of the primary phase as a function of shearing time under two different screw rotation speeds, respectively. At the early stages of shearing, for instance the first 3 s, the primary particles were fairly spherical, as indicated by the shape factor in Fig. 6. Further increase in shearing time and screw rotation speed only improved the shape factor slightly. However, the primary particle size increases with the increase in shearing time, and the particle growth rate appeared to be higher at a lower screw rotation speed (see Fig. 5). The solid lines in Fig. 5 represent the power law fit to the experimental data. The measured density of the primary particles formed in the twin-screw slurry maker is plotted in Fig. 7, asa function of shearing time and screw rotation speed. At the continuous cooling stage, particle density increased with the increase of shearing time, while at the isothermal shearing stage it decreased with further increase in shearing time. There was a maxima in the density time curve, which coincided with the transition between the

6 43 Z. Fan, G. Liu / Acta Materialia 53 (5) continuous cooling and isothermal shearing. The increase in particle density implied that there was a continuous nucleation during the continuous cooling stage, while the decrease in particle density suggested that Ostwald ripening took place through dissolution of smaller particles during the isothermal shearing stage. Shearing in the twin-screw slurry maker also had an effect on the formation of the primary phase in the shot sleeve. The measured volume fraction of the dendrite fragments formed during the secondary solidification is given in Fig. 8, as a function of shearing time and screw rotation speed. Generally, at the continuous cooling stage, volume fraction of a 2 decreased with the increase of shearing time, while it is almost constant with prolonged shearing time at the isothermal shearing stage. Fig. 8 also revealed that high screw rotation speed promoted the formation of the a 2 particles, suggesting that increasing the intensity of forced convection enhances secondary solidification in the shot sleeve. Fig. 9 shows the measured total volume fraction of a 1 and a 2 as a function of shearing time and the screw rotation speed. The total volume fraction of a 1 and a 2 increased with the increase in shearing time at the continuous cooling stage, while it became almost constant at the isothermal shearing stage. A comparison between Figs. 8 and 9 indicated that the formation of the primary phase in the shot sleeve is a temperature-controlled process. At a given shearing temperature, if the shearing in the twin-screw slurry maker produced less solid phase (smaller volume fraction of a 1 ), the volume fraction of Volume Fraction of α (%) Continuous cooling 8 rpm rpm 1 Shearing Time (s) Isothermal shearing Fig. 8. Volume fraction of the primary phase (a 2 ) formed in the shot sleeve as a function of shearing time and screw rotation speed. The shearing temperature was 593 C. The solid lines represent the best fit to the experimental data. Also marked in the figure is the boundary between continuous cooling and isothermal shearing during the primary solidification. Volume Fraction of α 1 +α 2 (%) 4 Continuous cooling 1 a 2 would increase to keep the total fraction of solid fairly constant. Therefore, it can be concluded that both increase in the shearing intensity and prolonged shearing time, promote the formation of the primary phase in the shot sleeve Effect of shearing intensity Isothermal shearing Shearing Time (s) 8rpm rpm Fig. 9. Total volume fraction of the primary particles (a 1 + a 2 )asa function of shearing time and screw rotation speed. The shearing temperature was 593 C. The solid lines represent the best fit to the experimental data. Also marked in the figure is the boundary between continuous cooling and isothermal shearing during the primary solidification. The results presented in the previous section showed that the intensity of forced convection measured by the screw rotation speed had a strong effect on the formation of the primary phase. Shearing experiments were conducted at a shearing temperature of 593 C and a shearing time of 35 s to establish such effects. The measured particle size and shape factor of the primary particles formed during the primary solidification are presented in Fig.. The primary particle size decreased slightly with the increase in screw rotation speed, while the shape factor of the primary particles increased slightly with the increase in screw rotation speed. The measured volume fraction and particle density of the primary phase formed in the twin-screw slurry maker are plotted in Fig. 11, as a function of the screw rotation speed. Fig. 11 revealed that fraction of solid decreased with the increase in screw rotation speed, indicating that intensive shearing suppressed, to some degree, the formation of the primary phase in the twin-screw slurry maker. However, intensive shearing did not promote nucleation in the twin-screw slurry

7 Z. Fan, G. Liu / Acta Materialia 53 (5) Particle Size of α 1 (µm) Particle Size of α1 Shape Factor of α Shape Factor of α 1 Volume Fraction of α1+α2 (%) Vol. Fraction of α1+α2 Vol. Fraction of α Volume fraction of α2 (%) Rotation Speed (rpm) Fig.. Size and shape factor of the primary particles (a 1 ) formed in the twin-screw slurry maker as a function of screw rotation speed. The shearing temperature was 593 C and shearing time was 35 s. The solid lines represent the best fit to the experimental data Rotation Speed (rpm) Fig. 12. Volume fraction of the primary particles formed in the shot sleeve (a 2 ) and the total volume fraction of a 1 + a 2 as a function of screw rotation speed. The shearing temperature was 593 C and shearing time was 35 s. The solid lines represent the best fit to the experimental data. Volume Fraction of α 1 (%) Vol. Fraction of α1 Density of α1 2 1 Particle Density of α 1 (mm -2 ) the total volume fraction of the primary phase (a 1 + a 2 ). This is consistent with the experimental results presented in the previous section. Therefore, it can be concluded that intensive shearing partially suppresses the formation of primary phase during primary solidification, promotes the formation of primary phase in the shot sleeve, but does not affect the total volume fraction of the primary phase. The effects of shearing intensity on the formation of a 2 are presented in Fig. 13, in which the density of a 2 and total density of a 1 and a 2, are plotted against the Rotation Speed (rpm) Fig. 11. Volume fraction and particle density of the primary particles (a 1 ) formed in the twin-screw slurry maker as a function of screw rotation speed. The shearing temperature was 593 C and shearing time was 35 s. The solid lines represent the best fit to the experimental data. maker, as indicated by the constant particle density in Fig. 11. It is necessary to point out that all the samples has been sheared for 35 s; some particles have already dissolved due to Ostwald ripening. In contrast to the results presented in Figs. and 11 for the primary solidification, intensive shearing promoted the formation of a 2, as indicated by the increase in a 2 volume fraction with screw rotation speed in Fig. 12. However, intensive shearing did not change Particle Density of α 1 + α 2 (mm -2 ) 6 4 Density of α1+ α2 Density of α Rotation Speed (rpm) Fig. 13. Density of primary particles formed in the shot sleeve (a 2 ) and the total particle density as a function of screw rotation speed. The shearing temperature was 593 C and shearing time was 35 s. The solid lines represent the best fit to the experimental data. 7 Density of α2 (mm -2 )

8 4352 Z. Fan, G. Liu / Acta Materialia 53 (5) screw rotation speed. It is interesting to note that the particle density of a 2 and the total density of a 1 and a 2 increased with the increase in screw rotation speed. It is necessary to point out that particles of a 2 refer to the fragments formed during the mould filling, from the original dendrites formed in the shot sleeve, and therefore the density of a 2 is considerably higher than the density of dendrites. However, since the die casting conditions have been kept constant for all the experiments, the increase in density of dendrite fragments corresponds to the increase in the density of dendrites. Hence, it is concluded that intensive shearing promotes nucleation of the primary particles in the shot sleeve at the secondary solidification stage. The increase in the total particle density in Fig. 13 was mostly due to the increase of a 2 density, since intensive shearing had very little effect on the density of a 1 (see Fig. 11) Effect of shearing temperature All the experiments in this section were conducted with a screw rotation speed of rpm and 35 s shearing time, the shearing temperature was varied between 585 and 597 C. The measured volume fraction of the primary phase is presented in Fig. 14, as a function of shearing temperature, in comparison with the equilibrium fraction of solid for the same alloy predicted by the CALPHAD approach [23]. Fraction of solid decreased with the increase in shearing temperature, and the experimental data was close to the thermodynamic predictions. In addition, it was also found that.5 shearing temperature did not have a strong effect on particle size and morphology of the primary phase formed in the twin-screw slurry maker, as shown by the experimental results presented in Fig. 15. This is in good agreement with the theoretical prediction by Wan and Sahm [24,25], that fraction of solid only has a very small effect on the coarsening process if the fraction of solid is low Effect of intensive shearing on the solidification inside the die cavity The effect of shearing temperature on the formation of a 3 is presented in Fig. 16, which revealed that shearing temperature had very little effect on the size of a 3 formed inside the die cavity. In the temperature range investigated in this work, the particle size of a 3 remained at about 6 lm, irrespective of the shearing temperature. In addition, further experiments were conducted to investigate the effects of shearing intensity and shearing time on the formation of a 3. The results were found to be very similar to those presented in Fig. 16. The a 3 particles were fine in size (around 6 lm), equiaxed in morphology and uniform in distribution, as shown in Fig. 3. It appeared that the formation of a 3 was controlled by the cooling rate rather than by the shearing conditions in the twin-screw slurry maker. However, it should be pointed out that intensive forced convection in the RDC process produced a fine and uniform microstructure inside the die cavity, while the HPDC process with no shearing produced a coarser and non-uniform microstructure [26], indicating the importance of uniformity of both temperature and composition fields in Solid fraction Particle Size of α1 (µm) Shape Factor of α T, o C Fig. 14. Volume fraction of the primary particles (a 1 ) formed in the twin-screw slurry maker as a function of shearing temperature (triangles) in comparison with the thermodynamic predictions by the CALPHAD approach (the solid line). The screw rotation speed was rpm and shearing time was 35 s. Particle size of α1 Shape Factor of α Shearing Temperature ( o C).65 Fig. 15. Size and shape factor of the primary particles (a 1 ) formed in the twin-screw slurry maker as a function of shearing temperature. The screw rotation speed was rpm and shearing time was 35 s. The solid lines represent the best fit to the experimental data..6

9 Z. Fan, G. Liu / Acta Materialia 53 (5) Particle Size of α3 (µm) creating a fine and uniform microstructure. Turbulent flow is crucial to achieve such uniformity. However, once the temperature and composition are uniform, further increase of the intensity of forced convection does not improve the microstructure. 4. Discussion Temperature ( C) Fig. 16. Size of the a-mg particles (a 3 ) formed in the die cavity as a function of shearing temperature. The screw rotation speed was rpm and shearing time was 35 s. The solid line represents the best fit to the experimental data Nucleation under intensive forced convection effective continuous nucleation Classic nucleation theory [27,28] predicts that homogeneous nucleation rate remains very small until undercooling (DT) reaches a critical value, when it increases very fast with the increase in DT. It also predicts that heterogeneous nucleation rate follows the same trend, except that a much reduced critical undercooling is required compared with homogeneous nucleation. This explosive manner of nucleation has been termed Big Bang nucleation by Chalmers [29]. However, a large nucleation rate is a necessary condition for grain refinement, but not a sufficient one on its own. The final effect of grain refinement is also dependant on the survival rate of the nuclei produced by the Big Bang nucleation. In the conventional casting processes, overheated liquid metal is poured into the relatively cold mould. Heterogeneous nucleation takes place immediately in the undercooled liquid close to the mould wall. The majority of the nuclei are transferred by the convection caused by mould filling to the overheated liquid region and dissolved; only a small proportion of the nuclei survive and contribute to the final microstructure, resulting in a coarse and non-uniform microstructure. It is, therefore, clear that an important step towards microstructural refinement is to make sure that every single nucleus formed during nucleation can survive and contribute to the final microstructure. % nucleus survival rate can be achieved by creating the following conditions: (1) uniform temperature and chemical composition throughout the entire volume of the liquid alloy during the continuous cooling process; (2) welldispersed heterogeneous nucleation agents; (3) rapid extraction of latent heat to prevent recalescence. Under such conditions, nucleation will occur throughout the entire volume of the liquid and each nucleus will survive and contribute to the final solidified microstructure, producing a fine and uniform microstructure. The authors refer to nucleation under such conditions as effective nucleation. As has been discussed previously, the melt flow inside the twin-screw slurry maker is characterised by high shear rate, high degree of turbulence and cyclic variation of shear rate. Such characteristics make the twin-screw mechanism very powerful for dispersive mixing. Consequently, the temperature and composition fields inside the slurry maker are extremely uniform throughout the entire alloy melt during both the continuous cooling stage and the isothermal shearing stage. It is also likely that the dispersive mixing power of the twin-screw mechanism can disperse any potential agglomerates of nucleation agents, and hence increases their potency for heterogeneous nucleation. In addition, the twin-screw mechanism is extremely efficient for heat transfer, avoiding any chance for recalescence. Therefore, the twin-screw mechanism is an important approach for achieving effective nucleation. This argument is well supported by the experimental results presented in the previous section (e.g., Figs. 2 and 5). During the continuous cooling stage of the primary solidification, volume fraction of a 1 increases with shearing time (Fig. 4), as expected according to the thermodynamic prediction. However, at the same time the particle density of a 1 also increases continuously (Fig. 7). This increase in particle density can not be explained by the dendrite fragmentation mechanisms [1], either by dendrite arm bending [5] or by remelting at the secondary arm root [6], due to the following three reasons: (1) no dendrite fragments were observed at any time; (2) particle growth showed the same kinetics during both continuous cooling and isothermal shearing stages (Fig. 5); (3) the shape factor of a 1 particles reaches a high value (.8) at the very beginning, and increases very slowly afterwards (Fig. 6). If the fragmentation mechanisms were operative, the particle size would decrease initially with time before the coarsening process, as

10 4354 Z. Fan, G. Liu / Acta Materialia 53 (5) has been confirmed by experiments in the literature [5,], and initially, the shape factor would not be high. The increased particle density during the continuous cooling can be explained by a continuous nucleation mechanism. The continuous nucleation can be understood in terms of the competition between nucleation and growth under small undercooling. The average cooling rate during the continuous cooling stage is about 4 K/s, which is small compared with those achieved by most of the die casting processes. Low cooling rate in combination with intensive melt stirring, makes a small undercooling more likely during the primary solidification, leading to a relatively small nucleation rate. However, under the intensive forced convection the nuclei grow spherically with an extremely fast growth rate, as will be discussed in the following section. Nevertheless, once the spherical particles reach a critical size corresponding to the fraction of solid for the melt temperature, the driving force for fast growth will diminish because spherical particles are energetically stable [24]. In competition with the growth of the existing primary particles, further cooling favours the nucleation of new ones. Therefore, when the temperature of the melt is reduced below the liquidus of the alloy, heterogeneous nucleation takes place, and all the nuclei will grow very rapidly to a volume fraction corresponding to the melt temperature. Further cooling of the melt under intensive forced convection will repeat the above cooling-nucleation-growth-cooling process, until the melt reaches the semisolid processing temperature. In addition, continuous nucleation is also favoured by the rapid heat transfer and uniform temperature field, which prevent any recalescence. Based on the above discussion, it can be concluded that nucleation under intensive forced convection is an effective and continuous process. Although continuous nucleation has been suggested by Hallewell [6] as a possible mechanism for rheocasting, the current experimental results presented in Fig. 7 are believed to be the first experimental confirmation of the existence of continuous nucleation. In addition, the experimental results in Fig. 11 suggest that increasing the intensity of forced convection does not increase nucleation rate. This implies that although high shear rate and high degree of turbulence is necessary to achieve effective nucleation, by providing uniform temperature and composition fields, further increase in the intensity of forced convection does not increase nucleation rate once both temperature an composition fields are already uniform. A further implication of the results in Fig. 11 is that intensive forced convection does not enhance the absolute nucleation rate. Grain refinement under intensive forced convection is achieved by effective nucleation, through increasing the nucleus survival rate Growth under intensive forced convection fast spherical growth The experimental results in Fig. 5 imply that growth of the primary particles after the nucleation inside the twin-screw slurry maker is very fast. This is evident by the measured particle size at very early stages of solidification, which is already around 35 lm after only 3 s of shearing. In fact, particles with a size less than 35 lm have never been observed in AZ91D alloy during the continuous cooling, suggesting that the growth stage between nuclei and 35 lm particles is probably shorter than 1 s. This is also supported by the plot of particle size against shearing time in Fig. 5. Particle growth shows the same kinetics during both the continuous cooling and isothermal shearing stages. This is contrary to the normal cases, where at first growth occurs until the fraction of solid reaches an equilibrium value before coarsening commences. This would show a distinct change in kinetics from growth stage to coarsening stage. The other fact established by the present experimental results is that solidification under intensive forced convection produces spherical particles during the entire primary solidification process. There is no evidence of dendrites or dendrite fragments observed at this stage. This suggests that particle growth inside the twin-screw slurry maker is spherical growth, rather than dendrite fragmentation followed by spheroidisation, as suggested by the dendrite fragmentation theory [1]. Therefore, it can be concluded that particle growth under high shear rate and high degree of turbulence is spherical growth with a very fast growth rate. This conclusion is well supported by the theoretical predictions by various techniques. Vogel and Cantor [7] investigated theoretically the effect of laminar flow on the stability of the solid liquid interface based on the stagnant boundary layer approach, and concluded that laminar flow enhances growth rate. More recent theoretical work has introduced turbulent flow into the solidification process. These include the stability analysis using a boundary element method by Qin and Fan [17], Monte Carlo simulation by Das et al. [18] and phase field approach by Qin and Wallach [19]. All these studies concluded that intensive forced convection, in particular turbulent flow, during solidification promotes spherical growth with enhanced growth rate. A general agreement is that laminar flow reduces the stagnant boundary layer and causes particle rotation, resulting in a rosette type of morphology, while turbulent flow can penetrate the inter-dendritic arm space, and therefore destabilise dendritic growth and promote spherical growth [7,17 19]. In addition, the experimental results in Figs. 4 and show that increasing the intensity of forced convection decreases the fraction of solid produced by primary solidification. There are two possible mechanisms

11 Z. Fan, G. Liu / Acta Materialia 53 (5) responsible for such results. One possible mechanism is friction heating caused by intensive forced convection in the semisolid slurry, resulting in an increased slurry temperature and therefore a decreased fraction of solid. The other possibility is that intensive shearing in the semisolid state, particularly with a high degree of turbulence, may have changed the energy level of the remaining liquid, or even changed the status of atomic clusters in the remaining liquid, in such a way that the liquidus of the alloy is displaced to a lower temperature, leading to a decreased fraction of solid at the primary solidification stage Particle coarsening under intensive forced convection The experimental results in Fig. 5 reveal that solidification inside the twin-screw slurry maker is mainly characterised by particle coarsening, since the growth rate is extremely fast. In addition, the solid particles have a fairly spherical morphology (with the shape factor being.8) in the very early stages of solidification, and prolonged shearing only slightly increases the shape factor (see Fig. 6). The major change occurring at the isothermal shearing stage is the decrease in particle density (see Fig. 7). Thus, it can be concluded that particle coarsening, under intensive forced convection, is achieved by Ostwald ripening through the dissolution of the smaller particles, rather than through diffusion of solid matters from areas with high curvature to areas with low curvature. Ostwald ripening is described by the classical LSM theory [31,32], which predicts that particle size (d) increases with time (t) according to the following equation: d n ðd Þ n ¼ kt; ð1þ where d is the initial particle size, k is the coarsening rate constant, and n is the coarsening exponent. It is generally believed that n is 3 for volume diffusion-controlled coarsening, n is 4 for grain boundary diffusioncontrolled coarsening, and n is 2 for interfacial reaction-controlled coarsening. Power law was used to fit the experimental data in Fig. 5 to extract n and k for different screw rotation speeds. The results are presented in Table 2. It is interesting to note that the coarsening exponent n is much larger under intensive forced convection than that under static condition. The values for n are 8.2 and 12.7 for Table 2 Summary of coarsening rate coefficient (k) and coarsening exponent (n) as a function of screw rotation speed Screw rotation speed (rpm) n k screw ration speeds of and 8 rpm, respectively. The values of n are much larger than 7/3 predicted theoretically by Wan and Sahm [24,25] for the case of laminar melt flow. In addition, the results in Table 2 suggest that both n and k are strongly dependent on the intensity of forced convection, and that they increase with the increase in the intensity of forced convection. Therefore, the coarsening process under intensive forced convection in the twin-screw slurry maker can be described by the following equations: d 8.2 ¼ t at rpm; ð2þ d 12.7 ¼ t at 8 rpm. ð3þ Eqs. (2) and (3) suggest that Ostwald ripening is extremely slow under intensive forced convection in comparison with coarsening under static conditions. This slow coarsening rate can be attributed to the unique solidification behaviour under intensive forced convection. As discussed previously, under intensive forced convection, crystals grow in a spherical manner with an extremely fast rate. Consequently, when the primary phase reaches the predetermined fraction of solid, the solid particles have morphology close to spherical (see Fig. 6) and a particle size with a narrow distribution around the mean diameter [33]. Under such conditions, the driving force for Ostwald ripening is substantially reduced, resulting in a very slow coarsening rate. Increasing the intensity of forced convection by increasing the screw rotation speed will improve particle morphology and narrow the distribution of particle size, leading to a further reduction of coarsening rate (see Fig. 5) Solidification behaviour of AZ91D alloy in the RDC process In this section, the solidification behaviour in the RDC process is summarised based on the experimental results presented in Section 3 and discussions made in Section 4. As illustrated in Fig. 1, the primary solidification in the RDC process can also be divided into two substages, continuous cooling and isothermal shearing. During continuous cooling, which lasts about 15 s, heterogeneous nucleation occurs continuously and uniformly throughout the entire volume of the melt, with a small undercooling. Due to the extremely uniform temperature and composition in the melt, every single nucleus survives, achieving % nuclei survival rate. The growth rate at this stage is extremely fast, with the maximum particle size being capped by the melt temperature. As a result of intensive forced convection, the growth of the nuclei takes place in a spherical manner. The fast and spherical growth leads to the formation of fine and spherical particles with a narrow size distribution. The isothermal shearing stage is basically a

12 4356 Z. Fan, G. Liu / Acta Materialia 53 (5) coarsening process. This is achieved through dissolution of the smaller particles. At this stage, coarsening due to mass transport from high curvature region to the low curvature region in the same particle is not significant, because of the large shape factor achieved by spherical growth at the continuous cooling stage. The secondary solidification starts when the semisolid slurry is delivered to the shot sleeve. Due to the low heat capacity and relatively low temperature of the shot sleeve, heterogeneous nucleation occurs in the intensively sheared liquid, and nearly all the nuclei survive. However, the nuclei grow dendritically in the absence of shearing. The resultant dendrites are fragmented when they pass through the narrow gate during the mould filling. Inside the die cavity, the remaining liquid still has largely uniform temperature and composition fields, due to the intensive shearing in the slurry maker and through the gate. Under the large cooling rate (about 3 K/s) provided by the metallic die block, nucleation is expected to take place throughout the entire remaining liquid, with a high nucleation rate. Numerous nuclei compete to grow, and the solidification finishes even before the instability occurs, producing fine and spherical a 3 particles. It is clear from the above discussion that, the uniformity of temperature and composition in the remaining liquid is crucial to the formation of a fine and uniform microstructure. Therefore, once the uniformity of temperature and composition is achieved, further increase in shearing intensity and shearing time will not lead to further structural refinement. (5) Particle coarsening occurs at the primary solidification stage through Ostwald ripening by consumption of the smaller particles. (6) The coarsening rate is extremely slow due to the spherical particle morphology and narrow particle size distribution. (7) Increasing the intensity of forced convection decreases both volume fraction and particle size, but does not enhance nucleation rate. (8) Increasing both the intensity of forced convection and shearing time increases the volume fraction and the nucleation rate of the primary phase formed in the shot sleeve. (9) Intensive forced convection promotes effective nucleation throughout the entire remaining liquid inside the die cavity and produces a fine and uniform microstructure. However, this effect is independent of the intensity of forced convection beyond a critical level. () Turbulent flow is more effective than laminar flow for creating fine and spherical solid particles during semisolid processing. Acknowledgements The authors thank Dr. A. Das in BCAST at Brunel University for useful discussions on the experimental results. The financial support from EPSRC (UK), Ford Motor Co. and Magnesium Elektron Ltd. (UK) is also acknowledged with gratitude. 5. Conclusions (1) The fluid flow inside the twin-screw slurry maker is characterised by a high shear rate, a high degree of turbulence and a cyclic variation of shear rate. Both shear rate and turbulence, although difficult to quantify, can be represented by the screw rotation speed. The higher the screw rotation speed, the higher the shear rate and the degree of turbulence experienced. (2) Both melt temperature and composition fields inside the slurry maker are extremely uniform during the primary solidification, owing to the strong dispersive mixing power of the twin-screw mechanism. (3) Heterogeneous nucleation occurs continuously throughout the entire volume of the melt during the continuous cooling. All the nuclei can survive and contribute to the final microstructure. This is called effective continuous nucleation. (4) The nuclei grow spherically with a very fast growth rate under intensive forced convection. References [1] Flemings MC. Metall Trans A 1991;22:957. [2] Kirkwood DH. Int Mater Rev 1994;39:173. [3] Fan Z. Int Mater Rev 2;47:49. [4] Boettinger WJ, Coriell SR, Greer AL, Karma A, Kurz W, Rappaz M, et al. Acta Mater ;48:43. [5] Doherty RD, Lee H-I, Feest EA. Mater Sci Eng 1984;65:181. [6] Hellawell A. Proceedings of semisolid processing of alloys and composites conference, Sheffield; p. 6. [7] Vogel A, Cantor B. J Cryst Growth 1977;37:9. [8] Miyata Y. ISIJ Int 1995;35:6. [9] Mullis AM. Acta Mater 1999;47:1783. [] Tong X, Beckermann C, Karma A, Li Q. Phys Rev E 1;64: [11] Beckermann C, Dirpers HJ, Steibach I, Karma A, Tong X. J Comp Phys 1999;154:468. [12] Tonhardt R, Amberg G. Phys Rev E ;62:828. [13] Jeong JH, Goldenfield N, Dantzig JA. Phys Rev E 1;64: [14] Molenaar JMM, Salemans FWHC, Katgerman L. J Mater Sci 1985;:4335. [15] Molenaar JMM, Katgerman L, Kool WH, Smeulders RJ. J Mater Sci 1986;21:389. [16] Ji S, Fan Z. Metall Mater Trans A 2;33:3511.