A Review on Fundamentals of Grain Refining of Al-Si Cast Alloys

Abstract

Grain refining is considered one of the most important liquid metal processing processes for aluminum alloys. Three different types of grain morphology are possible: columnar, twin columnar and equiaxed. The present work reviews most of the theories that were proposed during the past three decades. These theories were mainly based on thermal analysis and thermodynamics to explain the mechanisms of grain refining of Al-Si based alloys, including the role of the master alloy used i.e., Al-B, Al-Ti, and Al-Ti-B alloys. Other aspects were also examined, mainly the interactions between Si and/or Sr and the grain refining master alloy, superheating of the molten metal as well as holding time prior to casting. This phenomenon is normally termed “poisoning” since it reduces the effectiveness of the added grain refiners. The effects of grain refining on the alloy microstructural characteristics, mechanical properties, machinability, hot tearing etc. have not been addressed in the present article.

1. Introduction

Grain refinement in Al-Si casting alloys is usually assessed by the presence of titanium Ti and boron B. Since the 1980s, thermal analysis has established itself as an important alternative for determining the degree of grain refinement and in predicting the degree of modification of the eutectic silicon. Grain refining is one of the most important liquid metal processing processes for aluminum alloys. Three different types of grain morphology are possible: columnar, twin columnar and equiaxed. It is well known that an equiaxed grain structure provides uniform mechanical properties, reduced hot tear, second phase distribution and microporosity on a fine scale [1, 2, 3, 4, 5].

Grain refining in aluminum alloys aims to increase the number of crystallization sites of the pro-eutectic phase (α-Al phase) and avoid columnar growth. In order to have a fine scale grain size, the most widely practiced way is to present effective nuclei in the liquid metal using the Al-Ti-B grain refiners which usually contain active seeds like TiAl₃, TiB₂, AlB₂ or (Al,Ti)B₂. Thermodynamic studies suggest that these latter particles convert to TiB₂, so that the titanium would diffuse into the (Al,Ti)B₂ particles while the aluminum diffuses out, resulting in the formation of TiB₂ [6, 7, 8].

While the Al-Si alloy system is widely used in industry, constituting around 85–90% of aluminum parts produced, the eutectic silicon in untreated alloys is often very coarse, leading to poor mechanical properties, mainly ductility. These properties are strongly influenced by the morphology of the eutectic silicon. The latter changes from its original raw structure of platelets to a less harmful and finer fibrous structure termed as eutectic Si modification which leads to a significant improvement in the mechanical properties of the products. Modification of the eutectic silicon is usually accomplished by adding certain modifying agents such as strontium Sr. However, over-modification can lead to the formation of porosities and returns the silicon to its original shape, again weakening the characteristics of the alloy. The addition of strontium in Al-Si alloys leads to a considerable increase in the amount of the α-Al dendritic phase and changes the shape of the dendrites. In the presence of a grain refiner like Al-Ti-B, the modifier reduction is considerable since the Sr-Ti interaction alternately changes the volume fraction of the dendritic phase of α-Al and the morphology of the silicon phase [9, 10].

This study aims to establish suitable grain refining mechanisms in Al-Si alloys to study the consequences of the refining-modification interaction in these alloys. Different time and temperature parameters on the thermal analysis curves are analyzed.

1.1 Solidification phenomenon

Grain morphology in as-cast alloys can be categorized as equiaxed or columnar. The grains of an equiaxed structure are nucleated in the liquid pool ahead of the solidification front on particles deliberately added as inoculants or present as impurities. In engineering alloys, the supercooling required to launch a grain onto an inoculant particle is produced by constitutional effects: the splitting of the solute between liquid and solid creates a solute profile ahead of the solidification front, lowering the local liquidus temperature. In the absence of equiaxed grains, growth is typically columnar; columnar grains grow approximately perpendicular to the direction of heat flow and have high elongation. The growth of columnar grains is blocked if there are enough equiaxed grains. Given the importance of grain morphology to the properties of a cast alloy, it is important to predict the conditions that cause the transition from columnar to equiaxed growth [11, 12].

During their growth, equiaxed grains can adopt various morphologies, depending on the rate of cooling and the content of solutes in the liquid metal and the degree to which a grain can grow before impact or collision with other grains. Equiaxed grains initially grow as spheroids, but a smooth solid-liquid interface becomes unstable as the radius of the grain increases. The grain becomes globular and then adopts a strongly branched dendritic structure. This evolution is accelerated by high cooling rates and dissolved solids content.

The morphology of the columnar grains is influenced by cooling carried out constitutionally ahead of the solidification front. In situations where supercooling is greater ahead of this front than at the front itself, the solid-liquid interface is unstable and results in dendritic structures. The greater the constitutional cooling, the more pointed and branched the dendrites.

In the absence of a constitutionally cooled region, the solid-liquid interface remains planar. The observed microstructure can be influenced by magnification of the dendrite arms after solidification is complete; the structure of a solidification front that has been quenched to preserve its structure is often markedly different from one formed without quenching under otherwise identical solidification conditions.

1.2 Importance of thermal analysis

Thermal analysis could provide a reliable method for determining grain size, modifying the eutectic silicon and even quantifying the iron content of the alloy if the latter is equal to or greater than 0.6% by weight [13]. Thermal analysis consists in recording the evolution of the temperature of an alloy as a function of time during solidification. During cooling, heat is released and the temperature rises to a value close to the equilibrium value. This warming process is called recalescence. Supercooling, which represents a thermodynamic force, appears on the solidification curve as a drop in temperature below the equilibrium temperature of the reaction [14].

Grain refinement is a result of two separate processes: nucleation of new liquid aluminum crystals, followed by growth to a limited size. Both of these processes need a driving force which must be provided to the system through supercooling and supersaturation with respect to the equilibrium conditions of the real system. Figure 1a shows that during the entire first period of the solidification process, only those parts of the liquid metal which are in contact with the mold walls are cooled to such a degree that the nucleation of new aluminum grains can take place. Figure 1b shows that nucleation begins above the steady state growth temperature. This means that new crystals can be formed, not only at the first contact of the liquid bath with the cold walls of the mold, but also in the liquid.

Figure 1.

(a) First part of a cooling curve and its derivative obtained from liquid metal close to the wall of the casting mold. (b) Cooling curve and its derivative of a sample to which titanium boride particles have been added. The nucleation temperature is below the liquid metal growth temperature. The recalescence phenomenon shows a very low value of (dT/dt)_max, indicating a sample whose grains are refined.

This phenomenon occurs due to particles with a high nucleation power (titanium borides). The latter becomes active at a supercooling of only 0.1–0.2°C when added to the liquid metal. Seed particles added to liquid metal must be effective substrates for heterogeneous nucleation in order to achieve grain refinement. However, nucleation can occur only if the liquid alloy is sufficiently cooled. In a solidification system, the remaining liquid pool can be cooled only if there is some solute in this liquid pool to limit solid growth or columnar competition with equiaxed solidification. Thus, efficient refining requires heterogeneous nucleation and growth restriction. Al-Ti-B refiners which are generally used for this purpose, consist of particles of TiB₂ with a diameter ranging from 0.1 to 10 μm and Al₃Ti particles with a diameter ranging from 20 to 50 μm, dispersed in an aluminum matrix [15]. Al₃Ti can be a very effective nucleant for aluminum, but this phase dissolves quickly when the refiner is added to commercial purity aluminum, because all the titanium content in the liquid metal is well within the limit of solubility [16, 17]. It is well accepted that some excess titanium (beyond that combined with B in TiB₂) is required for efficient nucleation [14].

Other temperature parameters seen in Figure 1 correspond to:

T_E = The liquidus equilibrium temperature.

T_G = The steady state growth temperature of the molten metal.

T_N = The onset of nucleation temperature. TN is called the nucleation power of the particles present in the liquid metal. This point is most easily recognized by a sudden change in the derivative, as shown in the figure.

T_MIN = The temperature at which the newly nucleated crystals have grown to such an extent that the latent heat released swings out of equilibrium. After this time, the molten metal actually heats up to the steady state growth temperature. The period of time required for this heating is called the recalescence period (t_Rec).

1.3 Nucleation phenomenon

The grain refinement is carried out by the addition of master alloys of the Al-5Ti-1B and/or Al-Ti type in an embossed form. The addition rate of 1 kg/1000 kg gives Ti and B additions of 0.005% and 0.001%, respectively. Such an addition would typically produce grains of equiaxed structure with a grain size ranging from 100 to 150 μm in a small commercial pure aluminum ingot. The phenomenon of grain refining is directly linked to the process of nucleation and growth of aluminum grains. This is based on the nucleation ideas of Volmer and Weber [16].

The theory involves homogeneous and heterogeneous nucleation. In a solidified pure metal, the critical nucleus size for survival is given by:

$$r_{hom\mathit{og}\grave{e}\mathit{ne}}^{\ast }=\frac{-2{\gamma }_{\mathit{sL}}}{\mathrm{\Delta }{G}_v}$$E1

The free energy barrier is given by:

$$\mathrm{\Delta }{G}_{hom\mathit{og}\grave{e}\mathit{ne}}^{\ast }=\frac{16{\mathit{\pi \gamma }}_{\mathit{sL}}^3}{3\mathrm{\Delta }{G}_v^2}\cong \frac{\mathrm{\Delta }{H}_f\mathrm{\Delta }T}{T_m}$$E2

where ${\gamma }_{\mathit{sL}}$ is the surface energy of the interface separating the solid seed from the liquid in J/m².

By substituting $\mathrm{\Delta }{G}_v=\frac{L_v\mathrm{\Delta }T}{\mathit{Ts}}$ in Eq. (2), we obtain the following relation:

$$\mathrm{\Delta }{G}^{\ast }=\left(\frac{16{\mathit{\pi \gamma }}_{\mathit{sL}}^3\mathit{Ts}}{3L_v^2}\right)\left(\frac{1}{\mathrm{\Delta }{T}^2}\right)$$E3

where L_v is the latent heat of solidification per unit volume, ∆T is the supercooling (Ts − T) and Ts is the undercooling temperature. As for heterogeneous nucleation, the critical size of a nucleus is given by:

$$r_{h\acute{e}t\acute{e}\mathit{rog}\grave{e}\mathit{ne}}^{\ast }=\frac{-2{\gamma }_{\mathit{sL}}}{\mathrm{\Delta }{G}_v}$$E4

Eqs. (1) and (4) are identical for both types of homogeneous and heterogeneous nucleation. The potential barrier that the germ must cross to reach its critical size is given by the following equation:

$$\mathrm{\Delta }{G}_{h\acute{e}t\acute{e}\mathit{rog}\grave{e}\mathit{ne}}^{\ast }=\left(\frac{16{\mathit{\pi \gamma }}_{\mathit{sL}}^3}{3\mathrm{\Delta }{G}_v^2}\right)f\left(\theta \right)$$E5

where f(θ) is a function of the contact angle θ on the substrate on which nucleation takes place. Figure 2a shows a nucleated solid on a substrate in a liquid. Figure 2b shows the variation of f(θ) with θ and since f(θ) is always ≤1, the critical free energy for heterogeneous nucleation is always less than or equal to that for homogeneous nucleation. However, it is clear that effective heterogeneous substrates are those with θ close to zero [10]. Undercooling ∆T values are of the order of 0.1–0.2°C for observable nucleation rates in commercial aluminum alloys with grain refiners. Therefore, clearly heterogeneous nucleation takes place. The simplified expression for the heterogeneous nucleation rate per unit volume in m³ s⁻¹ is:

Figure 2.

Schematic representation showing (a) the formation of a spherical wetting of a solid S on a substrate, contact angle and surface tensions, (b) the variation of f (θ) with θ where f (θ) is equal to (2–3 cos θ + cos³θ)/4.

$$I_{h\acute{e}t\acute{e}\mathit{rog}\grave{e}\mathit{ne}}^v={10}^{18}{N}_v^{p}exp\left[\frac{-16{\mathit{\pi \gamma }}_{\mathit{sL}}^{3}f\left(\theta \right)}{3K_B\mathrm{\Delta }{S}^2\mathrm{\Delta }{T}^2}\right]$$

where K_B is the Boltzmann constant J/°C, $N_v^p$ is the number of nucleant/m³ and $I_{h\acute{e}t\acute{e}\mathit{rog}\grave{e}\mathit{ne}}^v$ is the heterogeneous nucleation rate of nucleant/m³ s. Therefore, if the contact angle is near zero, the wetting of the substrate for nucleation is promoted and the rate of nucleation is increased.

When the nucleation sites are homogeneously dispersed in the liquid pool, the result is a fine grainy structure. The important topics for understanding the nucleation phenomena are summarized as follows: (1) the contact angle between the molten metal and the nucleation particles, (2) the interface energy between the molten metal, and (3) the nucleants and the coherence of the lattices of the nucleants and metal liquid. The presence of possible phases at different T in the liquid pool can be evaluated by comparing the free energy ∆G of the reactions.

Based on thermodynamic data, the calculated results are shown in Figure 3. It can be observed that ∆G(TiB₂) is much more negative than ∆G(Al₃Ti) and ∆G(AlB₂) in the range of T from 700 to 1200°K, while ∆G(Al₃Ti) is less negative than ∆G(AlB₂). In other words, the TiB₂ phase is easier to form than the Al₃Ti and AlB₂ phases. With increasing temperature, the changes of free Gibbs enthalpy of TiB₂ and AlB₂ are almost constant while that of Al₃Ti became small. In fact, the theoretical prediction indicates that Al₃Ti particles become unstable when the reaction temperature is increased. From the crystallographic point of view, and to further explain the high stability of the TiB₂ nucleation sites, the hybridization of the 3d orbital of titanium and the 2p orbital of boron is the main reason for the strong bond between these two elements. The bonding behavior between Ti and B layers is a combination of covalent and ionic nature.

Figure 3.

Gibbs free energy of TiB₂, AlB₂, and Al₃Ti as a function of temperature.

1.4 Effect of overheating

As the Ti/B ratio is decisive for better grain refining, the casting temperature also plays a very important role when determining the grain size. Li et al. [18] studied the effect of overheating on pure aluminum before casting. Figure 4 clearly shows the relationship between the superheat temperature and the average grain diameter. It is clear that as the temperature is increased from 725°C to 950°C, the average grain sizes increase linearly. Casting temperature is a significant factor during the performance of grain refining. If the holding temperature is too high after inoculation, some fading or degradation occurs. This can be attributed to the growth and arrangement of TiB₂ particles, leaving a liquid alloy depleted of nucleating particles for efficient grain refining.

Figure 4.

Effect of undercooling temperature on the grain size of pure aluminum.

It is known that the size of the TiB₂ particles that form inside the aluminum depends on the temperature of the liquid metal; at high temperatures, the particles being formed are so large that they can settle to the bottom of the liquid bath by virtue of their greater density.

In general, overheating increases the grain size [11, 19], but in some cases it reduces it [20].

1.5 Master alloys

By adding three master alloys of the type Al-Ti-B, Al-Ti and Al-B with an excess of TiB₂ (Ti/B = 2.22), Lu et al. [21] examined the grain size in an Al-7%Si alloy (composition close to that of the hypoeutectic A356 alloy). The performance of these master alloys is shown in Figure 5. Indeed, the binary Al-Ti alloy is found to be less efficient, while the Al-B alloy is the strongest grain refiner in the Al-Ti alloys, since the grain size changes from 2000 μm to only 200 μm. From a certain level (0.1% by weight), the grain size remains constant even if the amount of master alloy is increased, hence the plateau obtained during supersaturation in master alloy.

Figure 5.

Grain refining of 356 alloy using different master alloys.

In contrast, Al-B alloys show inefficient behavior in pure aluminum [22]. Similar observations were made by Sigworth and Guzowski [6] Cooper et al. [23] proved that the efficiency of residual nucleants of the TiB₂ and Al₃Ti type decreases with the number of recyclings of the Al-Si alloys, which explains the increase in grain size as a function of the number of repetitions of castings [24].

1.5.1 Al-Ti-B

The only controversial point in the ternary system of Al-Ti-B is related to the two borides TiB₂ and AlB₂. Both crystallize in the same crystal structure (hexagonal shape) with similar lattice parameters. The question is to determine if they form a single-phase continuous compound (Al,Ti)B₂ or if they coexist in a two-phase balance AlB₂+TiB₂. A single phase was previously assumed by Hayes et al. [25] considering all experimental data available at that time. Roger et al. [26] investigated a 1000°C isothermal section in the titanium-rich region using quantitative microprobe phase analysis of three molten ternary alloys. They found that TiB₂ is in equilibrium with all Al-rich Al-Ti binary phases. The aluminum-rich corner of the ternary phase diagram was first calculated by Hayes et al. [27]. No ternary parameter was used in this work.

The question whether TiB₂ and AlB₂ exist in two separate phases or as a solid continuous solution has not been resolved. Zupanic et al. [28] investigated arc melted alloys in the triangular composition Al-AlB₂-TiB₂ and found four solid phases: (Al), AlB₁₂, AlB₂ and TiB₂. The AlB₁₂ phase, which is stable at very high temperatures in the Al-B binary system, decomposes during annealing below 900°C. Both borides AlB₂ and TiB₂ were found to coexist even after 1000 h at 800°C. Formation of mixed diboride (Al,Ti)B₂ was not observed [29]. Fjellstedt et al. [30] have produced Al-rich alloys by several different sample fabrication methods. They concluded that only the maximum solubility of aluminum in TiB₂ (up to 0.15 wt% Al) and titanium in AlB₂ (up to 0.2 wt% Ti) exists at 800°C, hence a continuous phase (Al,Ti)B₂ is not stable.

Based on the entity of this data, the AlB₂ and TiB₂ phases were modeled in the work of Gröbner et al. [31] as two separate phases without any solubility. No ternary phase or ternary solubility exists in this system, no ternary parameter is needed for the calculation. The complete phase diagram for the Al-Ti-B system can be calculated from the binary data and the extrapolation. An isothermal section at 500°C is given in Figure 6. It is important to note that the phase equilibria in the Al-rich corner are not changed significantly if a homogeneous ideal range of (Al,Ti)B₂ solution has not been assumed. The reason for this is the much higher thermodynamic stability of TiB₂ compared to that of AlB₂.

Figure 6.

Calculated isothermal section of the Al-Ti-B ternary system at 500°C.

When an Al-Ti-B type master alloy is added to an alloy such as A356 (∼7%Si), several intermetallic phases are created. Among these, there is mention made of intermetallics of the (Al,Si)₃Ti [32, 33] and Ti₆Si₂B [31] type. Ramos et al. [34] carried out detailed research on the ternary phase Ti₆Si₂B which belongs to the Ti-Si-B system. Thanks to X-rays, this phase is characterized by a hexagonal crystalline structure with lattice parameters a = 0.68015 nm and c = 0.33377 nm, and it forms a liquid through the following peritectic reaction: L + TiB + Ti₅Si₃ ↔ Ti₆Si₂B.

As shown in Figure 7, which presents a liquidus projection of the Ti-rich part of the Ti-Si-B system, five regions of primary solidification coexist, namely, Ti_ss, Ti₅Si₃, Ti₆Si₂B, TiB and TiB₂. At 1200°C, the Ti₆Si₂B phase is formed in two-phase fields: Ti_ss, Ti₅Si₃ and TiB. Solidification ends with an invariant ternary eutectic Ti_ss + Ti₆Si₂B + Ti₅Si3, which should correspond to the lowest liquidus temperature in the region shown in the figure. Based on the invariant reaction temperatures in the Ti-rich side of the Ti-Si [35] and Ti-B [36] systems, this ternary eutectic temperature should be less than 1330°C.

Figure 7.

Projection of the liquidus of the Ti-Si-B system in the Ti-rich part. The symbol Δ marks the composition of the Ti₆Si₂B phase.

1.5.2 Al-Ti

One of the effects of the presence of titanium Ti in an aluminum alloy is the reduction in grain size. However, this reduction is no longer achievable in pure aluminum since the number of grains per unit length increases, and consequently, the size of the grains also increases as shown in Figure 8. This size becomes almost constant when the content in titanium is between 0.08% and 0.13% before it begins to increase again with increasing percentage of titanium [36].

Figure 8.

Number of grains per unit length as a function of Ti content in pure aluminum.

Grain refinement in aluminum alloys by the addition of Al-Ti master alloys has been widely applied and studied in recent years. The grain refining mechanism by Al-Ti is not much doubted and can be explained by the action of Al₃Ti particles as heterogeneous nucleating centers [37] as well as by the peritectic theory. The size, morphology and quantity of the nuclei of the different microstructures of the Al-Ti master alloy seem to be important factors in determining the degree of grain refinement. The efficiency of the grain refiner depends on its chemical composition and its processing parameters such as the maintenance at such a temperature, the contact time, the mechanical agitation and the rate of cooling. To show the effect of increasing the percentage of titanium, Simensen [38] investigated a series of Al-7%Si alloys with a cooling rate of the order of 1°C/s. Al-10%Ti rods were added to the liquid metal making alloys with Ti in the range of 0.01–0.18%Ti. The cooling curves showed that at first the grains began to grow at some supercooling. The gradual addition of titanium increased the growth temperature of the alloys according to the following equation: T_growth = 613.2°C + 30.2%Ti (by weight). As for the grain size of the alloys, it was reduced from about 2000 μm to 250 μm when the titanium content increased from 0.01% to 0.12%. The best results were obtained when the Al₃(Ti,Si) phases were nucleated on the TiB₂ particles during cooling, whereas the aluminum grains which form on the Al₃(Ti,Si) intermetallics yield the fine-grained material.

On the other hand, Li et al. [39] investigated the effect of various microstructures on grain refinement by Al-Ti master alloys synthesized at high temperature by mixing aluminum with titanium. The results of their work showed that the variation in experimental parameters, such as the stoichiometric ratio of the initial powders, the particle sizes of the powders, the use of fluxes, etc., led to the formation of various structures of the master alloys, in particular sizes, morphologies and quantities of Al₃Ti particles. The intermetallic particles showed needle-like morphology in the lack of aluminum powder in the initial mixture which was related to the higher reaction temperature (Figure 9a). The same particles had needle-like and blocky forms at the same time (mixed morphology) which resulted from the lower reaction temperature (Figure 9b). Block-like Al₃Ti crystals were formed at the lower reaction temperature when excess aluminum powder was added to the initial mixture (Figure 9c).

Figure 9.

Microstructures Al-Ti master alloys. (a) Master alloy with acicular Al₃Ti particles. (b) Master alloy with a mixture of acicular and blocky Al₃Tiparticles. (c) Master alloy with blocky Al₃Ti particles.

In order to clearly highlight the effect of the increase in titanium on the grain size and on the growth temperature Tc, Tøndel and Arnberg [40] studied the behavior of a binary alloy of the Al-10% Si type, with Ti additions through an Al-6%Ti master alloy. Two series of alloys were prepared: series A cast in a cold mold (0.02–1.5% Ti) and series B cast in a preheated mold (0.02–0.2% Ti). It was essential to compensate for the effect of differences in silicon content on the liquidus temperature before studying the effect of titanium additions on the recorded Tc growth temperature. Temperature data therefore compensated for the true deviation in Si content from a chosen low composition, 9.6%Si, in the claim that Ti additions do not change the slope of the liquidus line of the Al-Si system in small temperature intervals. The calculation was made with a polynomial that describes the liquidus temperature as a function of the Si content in hypoeutectic Al-Si binary alloys: T_liq (°C) = 660–5.59Si–0.14Si.

The phase identification clearly shows that when the Ti content is increased, intermetallic particles will appear in the solidified material. An example of a large Ti (Al,Si)₃ type crystal (∼250 μm) is found in a B-series sample. Microprobe analyses of different samples show that Al₃Ti arises when silicon varies from 12 to 13% in the solution. This may, in fact, be an indication that the Al₃Ti particles of the master alloy have survived in the Al-Si liquid metal which thereby reaches equilibrium composition. The illustration in Figure 10 suggests that the peritectic composition at 0.15% Ti is shifted to lower Ti concentrations when silicon is present because the particle is too large to be a result of exceeding the 0.15% Ti limit by only 0.023% Ti [41, 42, 43].

Figure 10.

Growth temperature Tc and grain size as a function of Ti.

1.5.3 Al-B

The use of Al-B type master alloys (with 1–4%B) to achieve grain refining in Al-Si alloys is very common since this type of refiner is the most powerful in this type of alloys [43, 44]. The main particles that favor germination sites are AlB₂ (more stable in Al-Si) and AlB₁₂. The average AlB₂ particle size varies with the B content in the master alloy as shown in Figure 11a. It is obvious that the small additions in B have a remarkable effect on the solidification process of the alloy. Figure 11b shows the effect of boron additions on the first part of solidification of an Al-9.6%Si alloy. The cooling curve increased by ∼2 to 3°C when 161 ppm B was added. The addition of B eliminates the phenomenon of supercooling and recalescence on the solidification curve.

Figure 11.

(a) Average particle size of AlB₂ as a function of the content of B in Al-B, (b) cooling curve of the Al-9.6%Si alloy as a function of the added B.

Several concepts of the grain refinement mechanism of B on hypoeutectic Al-Si alloys have been adopted: the effect of B grain refinement on the α-Al phase and on the eutectic silicon with different additions of master alloys at 850°C was studied by Wang and Bian [45], where master alloys formed under different temperature conditions were studied to explore the morphologies of the AlB₂ particles; the sample slowly cooled with the addition of grain refiner was made to explore the mechanism of refinement. The master alloy can refine not only the α-Al dendritic phase, but the eutectic silicon. Theoretical analysis indicates that although the AlB₂ particles do not participate directly in the pure Al nucleation process in the presence of silicon, they provide a substrate for the precipitation of a small silicon content on which α-Al will grow without supercooling. As the temperature decreases to the eutectic line, AlB₂ later nucleates the eutectic silicon; AlB₂ particles appear in two different morphologies, namely hexagonal platelet and tetradihedron morphology which depend on temperature processing conditions.

As grain refining is an important process in industrial practice and has been the subject of much study, microstructural characterization of master alloys is useful in monitoring their production to ensure consistency of performance [43, 46]. AlB₂ particles appear in master alloys in two different morphologies, namely, as hexagonal platelets or as a regular tetradihedron. This difference in morphology depends on the state of treatment. The hexagonal platelet morphology is favored by low temperature production, the reaction between Al and KBF₄ is inactive, and the solubility of B in liquid aluminum is rather low. The formation of the AlB₂ phase is affected by the long distance diffusion of B atoms, which makes the obvious growth trend of the crystal. The growth of the AlB₂ particle proceeds by the diffusion of B atoms along <1120> at the edge of the crystal platform, the diffusion rate of B along <0001> is negligible. The schematic representation is shown in Figure 12.

Figure 12.

Schematic representation of AlB₂ morphologies; (a) hexagonal insert, (b) tetra-dihedron.

The tetradihedron morphology is tilted to be formed at high temperature. The chemical reaction of Al and KBF₄ proceeds rapidly at high temperature, therefore, the content of B in the aluminum liquid bath is higher at high temperature than that at low temperature. The diffusion of B atoms has relatively little influence on crystal growth, each plane of AlB₂ grows with almost the same speed. The diffusion velocity of B along <0001> cannot be neglected. When the diffusion velocity of B along <1120> is a little higher than that along <0001>, the tetradihedron morphology will be formed as shown in Figure 12b. No evidence that the morphology has an influence on the efficiency of the refiner was found, however, it seems that the hexagonal plate is always located in the center of two silicon flakes. The tetradihedron morphology is tilted to lie at the center of α-Al [43].

Very small additions of boron to Al-Si alloys lead to the precipitation of aluminum borides. To the Al-Si eutectic liquid, an addition of about 0.01% by weight of B is sufficient for this effect. It is the dominant feature of the calculated Al-B-Si phase diagram section at 0.1%B by weight, as shown in Figure 13. It looks like the binary Al-Si diagram with just AlB₂ as the additional equilibrium phase. The other boride, AlB₁₂, precipitates out of the liquid at high temperature and high in Si. In order to completely dissolve 0.1%B by weight, the liquid must be heated above 775°C. Using the Si-B master alloy, all boride particles should be formed in-situ during solidification, presumably in fine distribution and acting as strong nucleation sites. Only the binary peritectic reaction in the reverse direction, L + AlB₁₂ → AlB₂ is observed upon superheating at T = 972 ± 5°C [41, 42, 43, 44, 45, 46, 47]. If AlB₂ formation is also suppressed by cooling Al-Si-B liquid alloys, the metastable phase diagram should be considered. Such a calculation proves that the relevant phase boundaries in Figure 14 are virtually unchanged, just the saturation phase AlB₂ is replaced by AlB₁₂. This minor difference is also demonstrated by the calculated ternary eutectic:

Figure 13.

Section of the Al-B-Si system calculated at 0.1%B. The composition of the ternary liquid eutectic is 12%Si and only 0.01%B; the temperature is 0.1°C below 577°C of the Al-Si binary eutectic system.

Figure 14.

Evolution of the grain size of the A356 alloy: (a) no addition, (b) Al-10%Ti, (c) Al-5%Ti-1%B, (d) Al-4%B.

$$\mathrm{L}\text{Stable}=\left(\mathrm{Al}\right)+\left(\mathrm{Si}\right)+{\mathrm{AlB}}_2\mathrm{at}576.9°\mathrm{C}\text{and}\mathrm{L}\text{with}12.5\%\mathrm{Si},0.010\%\mathrm{B}\left(\mathrm{wt}.\%\right),$$

$$\mathrm{L}\text{Metastable}=\left(\mathrm{Al}\right)+\left(\mathrm{Si}\right)+{\mathrm{AlB}}_{12}\mathrm{at}576.89°\mathrm{C}\text{and}\mathrm{L}\text{with}12.5\%\mathrm{Si},0.011\%\mathrm{B}\left(\mathrm{wt}.\%\right).$$

Both equilibria are just slightly below the calculated binary eutectic L = (Al) + (Si) at 577°C and L with 12.5% Si by weight. Due to the slow formation of AlB₂ in the liquid metal, it is very likely that the next stable phase, AlB₁₂, will be formed instead. The particles found in the center of the grains are in fact B-rich, but it can be difficult to distinguish between AlB₂ and AlB₁₂ by backscattered electron and X-ray mapping techniques [48, 49]. The work carried out by Gröbner et al. [31] proves that, from the ternary phase diagram and thermodynamics, the two Al borides could be equally well formed. The fact that boron is a very efficient grain refiner in Al-Si alloys, but not in pure aluminum [4], is convincingly explained by the additional presence of dissolved silicon with a high growth restriction factor [50].

Figure 14 gives an overview of the grain size when the A356 alloy is treated by three grain refiners. In the absence of any addition, the grain size amounts to about 1850 μm in the base alloy, see Figure 14a. After an addition of 0.08%Ti using the Al-10%Ti master alloy, the size drops to approximately 800 μm, see Figure 14b. The aluminum grain size continues to decrease when 0.08% Ti is added using Al-5%Ti-1%B, see Figure 14c. A minimum value is obtained thanks to the boron-based refiner (Al-4%B) since the grain size is reduced to 200 μm, see Figure 14d. Adding excess titanium or boron has no effect on reducing the aluminum grain size. On the contrary, an overdose of titanium or boron can lead to deleterious effects on the microstructure of the alloy, and consequently on its mechanical properties.

1.5.4 Effect of master alloys

The literature concerning the influence of the addition of Al-Ti and Al-B master alloys on grain refining is very voluminous. Li et al. [51] investigated the effect of the Ti:B ratio on the solidification structure of a molten aluminum arc which is similar to the welding arc process. Al blocks were prepared, a hole with a diameter of 3.5 mm and another with a diameter of 7.5 mm deep were drilled in the center of each aluminum block in order to hold the two types of powder used (Al-Ti and Al-B). Grain size measurements were made using the linear intercept method, and were conducted at the edge, middle and center positions of each weld. The results are plotted in conventional form in Figure 15. This plot shows a minimum grain size at about 0.07% Ti.

Figure 15.

Effect of of Ti and B additions on the average grain size.

Such a minimum has never been reported previously in welding or casting. However, one must be careful when interpreting such graphs because the Ti:B ratio also affects the performance of grain refiners.

The benchmarks shown in Figure 15 are obtained for different Ti:B ratios. The minimum occurs when the Ti:B ratio approaches an atomic ratio of 2:1, which is the stoichiometric ratio for the formation of TiB₂ and the highest mole number of TiB₂ in the solder. Thus, it is concluded from these data that a stoichiometric TiB₂ is the most effective compound for a good grain refinement of aluminum under these experimental conditions. This disagrees with a few previous reports [52] which point out that excessive Ti is necessary for good grain refining under casting conditions.

1.6 Theories of grain refining

Grain refining is an important technique for improving the properties of aluminum products. Various explanations have been presented in order to provide, a suitable mechanism for grain refining such as nucleating particle theories and phase diagram theories. Both categories of the theories are about the two types of particles present in Al-Ti-B master alloys. Particle theories or boride theory suggest that nucleation occurs on the borides in the master alloy (TiB₂, AlB₂ and (Ti,Al)B₂), while phase diagram theories explain grain refinement by nucleation on the TiAl₃ properitectic phase.

1.6.1 Particle nucleation

Cibula [53] proposed that nucleation is produced on borides or carbides when the latter are present. Borides are added by a master alloy, whereas carbides are formed by a reaction of residual carbon present in the liquid metal with additional titanium which results in a TiC-like form. The nucleation behavior of all borides can be discussed concurrently, since TiB₂ and AlB₂ are known to be isomorphic and hexagonal, with lattice parameters changing only slightly, having a = 0.30311 nm and c = 0.32291 nm, and a = 0.3009 nm and c = 0.3262 nm, respectively. The boron mixed phase is formed by replacing titanium atoms in the lattice with aluminum atoms. The stability of the (Al,Ti)B₂ phase is not known; however, it is thought to convert to TiB₂ after a long hold time. Figure 16 shows the crystal structure of these two locations [54, 55].

Figure 16.

Crystalline structure of AlB₂ and TiB₂.

When the Al-Ti-B master alloy is added, the titanium is present in hypoperitectic amounts (less than 0.15% Ti, Figure 16), where boron particles are often found at the grain centers, with titanium-enriched dendrites growing outside of them. This evidence suggests that borides nucleate in the α-Al phase. However, for other reasons, borides were thought to be less efficient nucleation sites than Al₃Ti. In the master alloys, the borides are pushed or rejected towards the grain boundaries while the aluminides are at the centers of the grains [21]. Recently, Schumacher and Greer [55] confirmed that borides are pushed to grain boundaries and no grain enhancement is observed when there is no dissolved titanium. Additionally, borides are known to need some supercooling to nucleate aluminum, while aluminides need none.

Al₃Ti particles are known to be strong nucleating bodies. If titanium is present at hypo-peritectic concentrations, a dramatic grain enhancement is observed. Al₃Ti are found in the center of the grains at concentrations where they are stable, and multiple orientation ratios have been recognized between Al₃Ti and the aluminum matrix [49]. Obviously, it can be concluded that Al₃Ti is a better nucleant than TiB₂. Phase diagram theories have been developed to explain how Al₃Ti could be an active nucleant with hypo-peritectic compositions.

1.6.2 Theory of phase diagrams

The theories in this category are grouped under this heading because each theory suggests that the grain refinement is caused by a peritectic reaction on the primary Al₃Ti particles. In general, it has been suggested that a peritectic point shifts at low titanium concentrations (e.g. 0.05% Ti) caused by the addition of boron, and that this is the reason for the grain enhancement [54]. Therefore, it has generally been assumed that there is a peritectic Al-Ti-B ternary, and speculations and theories have been based on this assumption. The first attempt to explain the grain refining mechanism dates back to 1951. Crossley and Mondolfo [49] proposed a peritectic theory based on the peritectic reaction in the phase diagram of the Al-Ti system as:

$$\text{Liquid}+{\text{TiAl}}_3=\mathrm{\alpha }-\mathrm{Al}\left(\text{solid solution}\right).$$

It is reasonably clear that the titanium aluminum crystals added by the master alloy are active nucleants and that the observed fading is due to the dissolution of these nucleants with time. Davies et al. [56], and Maxwell and Hellawell [57] observed TiAl₃ particles at the center of the grains of the α-Al dendritic phase. The cooling curves published by Arnberg et al. [58] also support the order of nucleation, i.e., they show no supercooling but a nucleation temperature (T_n) above the melting point (T_f) of the liquid metal. This observation implies that nucleation occurs by a peritectic reaction around the peritectic temperature (665°C) which is higher than the melting point of pure aluminum.

Although the peritectic theory successfully explains the behavior of Al-Ti type master alloys, no consensus has emerged to explain the increased efficiency of commercial grain refiners containing titanium Ti and boron B. The authors of the peritectic theory suggest that the improved performance of boron is due to the shift of the peritectic composition from 0.15% Ti towards the aluminum end of the phase diagram, which ensures the thermodynamic stability of TiAl₃ at low levels of addition of Ti (∼0.02%). Determined phase relationships show that the solubility of TiAl₃ is practically unaffected by the presence of boron B. However, contrary to these thermodynamic predictions, Mondolfo et al. [48] obtained experimental data indicating the effect of boron by shifting the peritectic to the Al-rich end of grain seems to be far from being achieved. Figure 17 presents the aluminum-rich part of the Al-Ti system diagram which clearly shows the domain of existence of the different phases with their formation temperatures.

Figure 17.

Al-Ti binary diagram (Al rich corner).

1.6.3 Theory of peritectic transformation

This theory was very popular in the late 1980s and early 1990s supported by Vader and Noordegraaf [59], and Bäckerud et al. [60, 61]. This theory assumes that TiAl₃ is a stronger nucleant than TiB₂. Therefore, it tries to explain how borides could slow down the dissolution rate of TiAl₃ when an Al-Ti-B master alloy is added to the liquid aluminum bath, so that the more powerful nuclei remain active longer. It suggests that the borides form a shell around the aluminides and therefore slow down the dissolution of the aluminides as the diffusion must proceed through the boride shell. The finally dissolved aluminide leaves a liquid cell inside the boride shell, approximately peritectic in composition. Peritectic reaction takes place to form α-aluminum and growth occurs from there. A schematic representation of the peritectic reaction is shown in Figure 18.

Figure 18.

Schematic presentation of peritectic reaction.

Although this theory appears to fit experimental results, there is strong evidence against it, particularly that described by Johnsson et al. [62]. Borides are very stable in liquid aluminum alloys, compared to TiAl₃ particles, to hypo-peritectic titanium compositions (∼ 0.15% Ti by weight). The peritectic theory suggests that borides are more soluble than TiAl₃, because the borides must dissolve in the liquid bath so that they can reprecipitate onto the more slowly dissolved TiAl₃ particles in the titanium-rich region produced by its dissolution - which does not seem possible. Even with boron in the master alloy, TiAl₃ still dissolves after a few minutes at high temperatures.

Johnsson [63] melted and re-solidified a hypo-peritectic alloy and found that the efficiency of grain refining did not change with the number of cycles. If the peritectic mechanism were to occur, one would expect the efficiency of grain refining to decrease with the number of repetitions, as this would further allow the diffusion of titanium; therefore, the peritectic reaction would cease to occur. Boride shells have been found in the grains of the aluminum, although it is inconclusive that these were the site of nucleation. If they acted as nucleants, this was not the dominant mechanism, as more often boride particles were found in the grain center at hypoperitectic concentrations of titanium. Therefore, the evidence suggests that the peritectic mechanism did not work.

1.6.4 Theory of hyper-nucleation

Jones and Pearson [5] assumed the concept of hyper-nucleation theory at the TiB₂/liquid metal interface. They proved that when titanium is in excess, the titanium atoms segregate at the TiB₂/liquid metal interface, providing a stabilized layer of atoms on the surface of the TiB₂ crystals. This layer, being a full solution of Ti and Al, was predicted to remain stable above the melting point of pure aluminum, i.e., it exists in the liquid metal before casting. During cooling, such a layer will allow the growth of the primary α-Al phase without any supercooling. Although this concept seems the most promising, no experimental evidence is there to support it. Figure 19 presents a model of the hyper-nucleation theory [5, 64, 65, 66]:

Figure 19.

Schematic presentation of the theory of hypernucleation. (a) Excess of Ti (Ti/B > 2.21) in solution, (b) Ti segregation at the interfaces TiB₂-liquid metal, (c) formation of layers of TiAl₃ on TiB₂, (d) nucleation of α-Al through peritectic reaction.

1.6.5 Duplex theory of nucleation

Mohanty et al. [67, 68, 69, 70] suggest that the formation of Al₃Ti is caused by a concentration gradient of titanium towards the boride particles, constituted by an activity gradient towards the borides. Due to the local equilibrium near the borides, the Al₃Ti would be stable and could subsequently nucleate in the α-Al phase, as for alloys whose titanium is found in hyper-peritectic concentrations. Jones [70] supported this titanium gradient theory of segregation, but there is no conclusive evidence that a titanium gradient exists. As early as 1977, Naess and Berg [71] tried to suggest that there would have been a high concentration of titanium around the borides in the liquid pool, but their evidence showed nothing more than the predicted solute profile on the solidification of an Al-Ti alloy.

The duplex theory of nucleation is not totally new. In 1971, Bäckerud [72] claimed to have observed Al₃Ti on boride particles and to have proposed a series of reactions to explain this. It was also mentioned by Cornish [52], who proposed that the role of borides is to facilitate the formation of Al₃Ti at hypo-peritectic concentrations due to a variation in the peritectic composition, which then induces nucleation via the peritectic reaction. They used some of the same arguments as Mohanty et al. [69] about the segregation of titanium to borides. If the TiB₂ particles nucleate the Al₃Ti particles, which, in turn, nucleate the α-Al, then the mechanism is still unexplained, especially since the difference in the expected nucleation temperatures and the non-application of the theory to the alloys of foundries are always a problem. If Al₃Ti particles form on the surface of TiB₂ particles that increase nucleation, then it is the borides that act directly or indirectly as nucleation sites.

1.7 Effect of Si

The alloy composition effect seems to be quite complex. It has been documented for the Al-Si [73, 74, 75, 76, 77, 78] system that the grain size first decreases with increasing the alloying concentration and then, after reaching a minimum, the grain size increases with further additions. The minimum is obtained near the maximum solubility limit and some researchers have therefore reported the minimum in grain size at a maximum range of solidification, and therefore at the maximum time of solidification, suggesting that an alloy with a wide range of solidification grants a longer time for nucleation. However, Bäckerud and Johnsson [79] recently suggested that a cellular structure transition to dendrites with well developed orthogonal branches is responsible for the transition, i.e. a cellular-dendritic transition during the growth of equiaxed crystals. They proposed that the transition occurs at a growth restriction factor of 20. This factor is equal to Σim_iC₀,_i (k_i-1) where m is the slope of the liquidus line, C₀ is the composition of the liquid and k is the equilibrium distribution coefficient for all elements.

In order to identify the effect of increased silicon content on the morphology and grain size in hypoeutectic Al-Si alloys, Lee et al. [80] used six silicon concentrations, 1, 2, 3, 4, 5 and 8 (wt %) combined with five levels of master alloy (grain refiner) of Al-5%Ti-1%B. The different samples were solidified in a preheated cylindrical graphite crucible at a cooling rate of 0.7°C/s. The results indicate that grain size is controlled by a combination of nucleating power and constitutional conditions at the growing crystal interface. Figure 20(a) and (b) illustrate the evolution of the grain size according to the concentrations of silicon Si and titanium Ti.

Figure 20.

(a) Average grain size as a function of Ti content for 6 levels of Si, (b) same as in (a) using Al-5%Ti-1%B master alloy.

The TiSi₂ type phases are the main cause of such poisoning. The micrograph given in Figure 21 proves the presence of these phases. Other researchers have proposed that they are Al₃Ti particles instead of TiB₂, which become coated by certain complex aluminides or silicides formed by the interaction of silicon with Al₃Ti. These intermetallic compounds may be unable to nucleate aluminum during solidification, since they have always been observed at grain boundaries instead of grain centers [81]. However, since boron atoms are too light to be detected by energy dispersive X-ray spectroscopy, the possibility of the existence of borides cannot be ruled out by the absence of boron in the spectrum. Through X-ray energy dispersive spectroscopy, the TiSi₂ phase was identified by the presence of the high intensity peaks relative to titanium and silicon elements. As for Figure 22, it illustrates the distribution of the elements of this phase in a 390 alloy treated with 0.4% Ti and cast after 120 minutes. The spectrum relating to the TiSi₂ phase is given in Figure 23.

Figure 21.

Backscattered electron micrograph identifying the formation of TiSi₂ phases in a 390 alloy (∼17%Si) treated with 200 ppm Sr, 0.4%Ti and cast at 750°C after 120 minutes of dwell time.

Figure 22.

Distribution of aluminum, titanium and silicon in a 390 alloy treated with Al-10%Ti, 200 ppm Sr, showing the TiSi₂ phase and cast after 120 minutes of holding time.

Figure 23.

Dispersive X-ray (EDS) analysis of the TiSi₂ phase proven by high intensity peaks.

1.8 Sr-grain refiner interaction

The addition of strontium as a modifier in Al-Si alloys causes a transition in the morphology of the eutectic silicon from an acicular to a fibrous and fine form. However, strontium can also lead to the formation of a long and columnar α-Al dendritic phase. The Al-5Ti-1B master alloy is often used as a grain refiner to achieve a fine, equiaxed grain structure in aluminum and its alloys. The TiB₂ or/and TiAl₃ particles of the Al-5Ti-1B master alloy are thought to be able to act as nuclei for the primary α-Al phase. However, in Al-Si alloys with a high level of silicon (%Si more than 7% by weight), the grain refining power of the master alloy of Al-5Ti-1B is lower compared to that of Al-3B and Al-3Ti-3B master alloys [81, 82]. It was therefore thought that the silicon poisoned the nucleation nuclei. This is mainly related to the formation of silicon and titanium above the TiAl₃ particles, and Kori et al. [83] specified that this poisoning effect could be neutralized by increasing the level of addition of the main alloy of Al-5Ti-1B. The sequence of the grain refiner and modifier addition has a significant influence on the grain grade of the α-Al dendritic phase when compared to the combined addition as shown in Figure 24.

Figure 24.

Analysis of grain size of 356 alloy treated with Al-1%Ti-3%B master alloy: 1%Ti and 0.02%Sr [83].

When the liquid bath is treated with both a grain refiner and a modifier, the evolution of the Sr concentration in the liquid metal is highly time dependent for higher levels of addition of the Al-Ti-B grain refiners, Ti-B Figure 25 shows this phenomenon. The zero-time concentration is the level of Sr in the liquid aluminum before the addition of either Al-Ti-B grain refiner. After addition, a weakening of Sr is observed for the two liquid alloys; the liquid bath treated with Al-1.5Ti-1.5B master alloy loses its Sr much faster, especially at the initial stage after addition, compared to the liquid bath treated with Al-5Ti-1B. This explains the rapid loss of eutectic modification in the liquid alloy treated with Al-1.5Ti-1.5B, it means that there is insufficient free Sr in the liquid aluminum to modify all the eutectic silicon. The rapid loss of strontium can be explained by external oxidation and vaporization as shown in Figure 25 [84].

Figure 25.

Sr concentrations as a function of time after AlTiB grain refiners were added to achieve 0.15% Ti in the liquid bath. The broken line refers to the case of Sr modification without addition of any grain refiner.

As discussed before, boron B is the most efficient refiner for the A356 alloy. With the addition of strontium and the latter’s reaction with boron B, the overall percentages of strontium and boron capable of acting as a refiner and modifier respectively decrease and therefore would not be as effective as those obtained when each is added individually. In order to demonstrate the resulting reaction, strontium and boron were added to pure aluminum, and the casting was carried out under the same conditions as those existing during individual addition. Microstructural analysis proved that strontium and boron react with each other forming compounds of the SrB₆ type according to the following reaction: Sr + 6B → SrB₆. This last product is confirmed by the results obtained by Li et al. [85] and Nafisi [86]. This type of compound, whose name is strontium hexaboride, is characterized by a very high melting temperature of 2500°C [87] with a weight ratio of Sr:B equivalent to 1.35:1.

This reaction proves that each atom of Sr could react with six atoms of B and consequently form a compound SrB₆. The intermetallics SrB₆ and TiB₂ could act as nucleants but it should be considered that the consumption of boron in the compound is one of the main parameters. It means that AlB₂ consumes less amount of boron in comparison with SrB₆. Considering a constant amount of boron, the density of nucleating particles is much higher in the case of AlB₂ since a lower number of boron atoms is associated with this compound. Therefore, the greater the number of effective nucleants, the greater the probability of having a smaller grain size. Using dispersive X-ray (EDS) analysis, the SrB₆ phase was identified and confirmed by high intensity peaks. The strong affinity between strontium and boron is shown in Figure 26 obtained using the electron microprobe. The size of SrB₆ compounds varies between 5 and 10 μm, and their color is a mixture of dark gray and white [88].

Figure 26.

Mapping produced by the electron microprobe showing the association of strontium and boron in pure aluminum forming SrB₆ type phases [88].

2. Conclusions

Based on the presented survey of literature on the fundamental aspects of grain refining of Al-Si cast alloys, the main points could be summarized as follows:

1. Of primary importance are the grain size, the dimension of the dendritic cells or the interdendritic space, and the shape and distribution of the eutectic mixture, which consists mainly of silicon. A fine, equiaxed grain structure is always desired since a finer grain size promotes improved casting strength by minimizing shrinkage, hot cracking and hydrogen porosity.

2. To achieve grain refining, the most widely practiced way is to introduce effective seed nuclei into the liquid metal using Al-Ti-B grain refiners which usually contain active seeds like TiAl₃, TiB₂, AlB₂ or (Al,Ti)B₂.

3. The introduction of titanium in Al-10%Ti leads to the formation of ultrafine intermetallics of the (Al,Si)₃Ti type. The latter constitute nucleation sites for the α-Al phase. When added to liquid metal, the Al-4%B master alloy shows remarkable potency in comparison to other grain refiners.

4. For master alloys, residual titanium in alloy A356 reacts with boron B to form TiB₂ which subsequently acts as an active seed alongside AlB₂ for the α-Al phase.

5. The addition of strontium and the grain refiner Al-5%Ti-1%B shows a certain affinity between the modifier and boron. This affinity, limited by the outer surface of the TiB₂, partially deactivates the effect of the refiner since the minimum granular size is obtained for a Ti content of 0.2–0.3% by weight, compared to that obtained with the addition of Al-10%Ti and strontium.

6. The introduction of AlB₂ in Al-4%B form in alloys containing traces of titanium leads to the reaction between boron and titanium to form TiB₂. Grain refining is achieved primarily with TiB₂ rather than AlB₂, or both, depending on the titanium content in the given alloy.

7. In the presence of strontium, boron reacts with strontium to form compounds of the SrB₆ type which is supposed to be a very weak refiner.

8. The affinity between titanium and boron is higher than the affinity existing between boron and strontium. Note also that B does not react with Si unlike titanium, however, it leads to better results.

9. The presence of excess silicon in Al-Si alloys leads to a strong interaction between titanium and silicon. This high affinity leads to the formation of (Al,Si)₂Ti-type phases, weakening the nucleation opportunities of the dendritic phase and consequently reducing the degree of grain refinement. Titanium disilicide phase tends to form more when the liquid metal is held for long periods.