Grain refinement of Al-Si alloys by Nb-B inoculation. Part I: concept development and effect on binary alloys

Abstract The effect of Nb–B inoculation on Al–Si alloys for their grain refinement has been studied through the analysis of binary Al–xSi (where x = 1–10 wt.%) to avoid possible effects of other alloying elements. In Part I of this work the concept development of the Nb–B inoculation is discussed in detail on the basis of the theoretical and fundamental concepts employed (i.e. pro-peritectic particles formation, lattice structures and mismatch as well as analogies between the Al–Ti/Al–Nb or Ti–Si/Nb–Si binary phase diagrams). The systematic study of the addition of different level of Nb–B inoculation to pure Al permitted to determine the best addition rate. From the microstructural and thermal analysis of binary Al–xSi alloys it is found that Nb–B inoculation is highly suitable for Al–Si alloy with Si content greater than 6 wt.%. As results of the Nb–B inoculation the microstructural features of binary Al–xSi alloys (i.e. primary Al α-grains and eutectic phase) are significantly refined. Most importantly, the inoculation of Al–Si cast alloys with Nb–B is not characterised by any visible poisoning effect (i.e. formation of silicides) which is the drawback of using commercial Al–Ti–B master alloys on Al cast alloys. The effect of Nb–B inoculation on commercial Al–Si alloys (which normally include other alloying elements in their chemical composition) is assessed in Part II of this work.


Introduction
Aluminium and its alloys are common engineering materials for structural applications due to the combination of properties they can provide. Specifically: low density (2.7 g/cm 3 ), corrosion resistance in many environments, good specific strength (i.e. strength to density ratio), good thermal conductivity and low electrical resistivity. As for other metals, the primary process employed to fabricate Al alloys is used to differentiate and categorise them as wrought and cast. Cast Al alloys are based on the binary Al-Si phase diagram and are used because of their low melting point, good fluidity, good surface finishing as well as limited solubility for gases (except for hydrogen). Other alloying elements such as magnesium and copper are generally contemplated in the alloy composition to achieve specific properties like improved corrosion resistance or better response to heat-treatments. Depending on the production route employed, Al can be characterised by quite coarse microstructure and, thus, its grain refinement it is a common industrial practise [1]. Usually, the grain refinement is carried out by the addition of commercial master alloy developed and based on the ternary Al-Ti-B system, such as the Al-5Ti-1B master alloy, which is added to the melt prior casting [2][3][4][5]. It is worth mentioning that Ti is the elements with the highest growth restriction factor [6], which plays an important role in the refinement of Al by means heterogeneous nucleation. The mechanism behind the grain refinement of Al has been a topic of debate and various theories were proposed: phase diagram/peritectic theory, peritectic hulk theory, hypernucleation theory, and solute theory [7][8][9]. Summarising, the employment of commercial Al-Ti-B master alloy is based on the scientific fundamental that Ti reacts with B forming TiB 2 particles and with Al forming Al 3 Ti intermetallic particles. When the commercial master alloy is added to the molten Al, TiB 2 particles act as heterogeneous nucleation sites whilst Al 3 Ti intermetallics dissolve into the melt on the base of the peritectic reaction forming a-Al: liquid Al + Al 3 Ti ? a-Al (solid solution) [9,10]. Observation of the interface Al 3 Ti layer on TiB 2 particles suggested that TiB 2 particles in combination with Al 3 Ti contribute to the heterogeneous nucleation of a-Al grains [11]. Commercial Al-Ti-B master alloy are very potent grain refiners for wrought Al alloys, whose Si content is generally lower than 2 wt.%. Nonetheless, in the case of cast Al alloys, where Si content is greater than 4 wt.%, the efficiency of commercial Al-Ti-B master alloys is relatively poor. This is due to the interaction of Ti with Si to form titanium silicides (i.e. TiSi, TiSi 2 and Ti 5 Si 3 ) which depletes the melt of Ti preventing the grain refinement of the alloy. This phenomenon has been the subject of different studies and it is known as poisoning effect [4,[12][13][14]. Many attempts and a lot of effort have been focused in achieving fine equiaxed grain structure in as-cast Al-Si alloys by the addition of small quantities of elements called hardeners (which includes Zr, Nb, V, W, Ta, Ce, etc. [15][16][17][18][19]) but without much success. The latest works focused on the effect of commercial Al-B master alloys (i.e. Al-3B), Al-Ti-C master alloys [20], Al-Ti-B-C master alloys [21] or variants of the Al-Ti-B master alloys [22,23]. The most significant result of research done by Birol [24] is that the Al-3B master alloy performs very well in refining Al-Si alloys grain structure when they are Ti-free, which is not the case for most of the Al cast alloys. When Ti is present as impurity in percentage greater than 0.04 wt.% the Al-3B master alloy refining potency is equal to that of commercial master alloys based on the Al-Ti-B ternary system (i.e. Al-5Ti-1B and Al-3Ti-3B) [24]. Potential heterogeneous nucleation substrates for grain refinement have to be characterised by three main aspects [25]: (1) high melting temperature to prevent their melting when placed in contact with the molten metal to be refined, (2) low lattice mismatch with the nucleating phase and (3) chemically stability (do not interact with the alloying elements). The aim of this work is to report and discuss the grain refinement of Al-Si cast alloys by Nb-B inoculation, chemical composition which was patented [26,27]. In particular, the concept development and the effect of Nb-B inoculation in binary Al-xSi alloys (where x = 1-10 wt.%) are discussed in Part I. Binary Al-Si alloys were chosen because they permitted to quantify the effect of Nb-B inoculation on the microstructural features (primary a-Al dendrites and secondary eutectic phase) without the concern of possible side effects of other alloying elements present in commercial Al-Si cast alloys. The effect of the addition of the Nb-B grain refiner to commercial Al-Si cast alloys solidified over a great range of cooling rates as well as the mechanism governing their grain refinement are assessed in Part II of this work.

Experimental procedure
The work described in this paper is divided into three main sections, namely (1) the study of the addition of Nb to commercial pure Al, (2) the study of the addition of interstitials (B and C) to the Al-Nb system and (3) the study of the addition of the 0.1Nb-0.1B to Al-xSi binary alloys. Therefore, at each specific time different materials and methods were used. Nevertheless, the general path followed is reported in this section and it is applicable for the experiments described unless otherwise specified. The raw materials employed were: Commercial pure Al (Al > 99.5 wt.%, Si = 0.02 wt.%, Fe = 0.08 wt.%, Mn = 0.01 wt.%, Zn = 0.02 wt.%, Ti = 0.06 wt.% and Ga = 0.05 wt.%) supplied by Norton Aluminium: used to test the refining potency of Nb and Nb-B as well as for the production of binary Al-xSi alloys, where x = 1-10 wt.%. Al-50Si master alloy: mixed with pure Al to produce binary Al-xSi alloys, where x = 1-10 wt.%. The Al-xSi alloys were produced in batches of 5 kg and their chemical composition was checked by means of an appositely calibrated FOUNDRY-MASTER Pro equipment (Oxford Instruments). Nb powders (Nb > 99.8%) with big particle size (average particle size 100 lm) and with fine particle size (lower than 45 lm) procured from Alfa Aesar. Initially (Section 3.1.2) the coarse Nb powder was used and it was found that it shows poor dissolution into Al although it permitted to study the sedimentation of Nb particles. Subsequently (from Section 3.1.3 onwards) all the experiments were carried out using the fine Nb powder.
The total amount of Nb added varied during the development of the composition of the novel Nb-B grain refiner ( Fig. 6) but it was then set to 0.1 wt.% (targeted addition) during the study of its performances on Al-xSi alloys.
B powder (B > 98%) with particle size lower than 45 lm: initially used to study the combined effect of Nb and B on pure Al (Section 3.2.2). Lately (from Section 3.2.3 onwards), switched the potassium tetrafluoroborate (KBF 4 ) due to the non-wetting behaviour of the B powder which hinders its dissolution into the melt. Once again, the amount of B added to the melt was changed during the initial experiment ( Fig. 6) but then fixed at 0.1. wt.% (targeted addition) during the study of the performances of Nb-B inoculation. It is worth mentioning that, although 0.1 wt.% was targeted, the actual content of Nb-B, measured by ICP method, is lower due to partial oxidation of the Nb powder during addition to the melt, low B recovery from KBF 4 and non-optimised mixing process. Potassium tetrafluoroborate (KBF 4 > 98%) flux purchased from Alfa Aesar: employed as alternative source of B.
Graphite powder with particle size lower than 20 lm: employed to study the combined effect of Nb and C on pure Al. Commercial Al-5Ti-1B master alloy supplied by London & Scandinavian Metallurgical Co Limited: used to compare the performance and efficiency of Nb-B inoculation. The amount of master alloy added was equal to 0.1 wt.% which is a common percentage employed at industrial level.
Pure Al and binary Al-xSi alloys (where x = 1-10 wt.%) were melted and/or prepared in clay-bonded graphite crucibles and maintained at temperature for, at least, one hour. The melting temperature changed depending on the type of experiment (i.e. production of the binary Al-xSi alloy or grain refinement) and it varied between 720°C and 800°C. Specifically, 800°C was mainly used when the addition of the Nb-B inoculant was considered. In the case of the pouring temperature, this parameter ranges between 660°C and 720°C depending on the nature of the experiments (i.e. study of the influence of the casting temperature). In some cases (e.g. study of the effect of the refiner on the base of the thermal analysis) the melt was left to cool inside a crucible externally lined with a glass wool insulator which permitted a very slow cooling rate (i.e. $0.3°C/s). The cast samples were prepared for metallographic analysis by using the classical route. In the case of macroanalysis, samples were grinded and etched by means of Tuckers' reagent whereas for microstructural analysis the samples were fine polished with OPS and characterised by means of an optical microscope (Carl Zeiss Axioscope A1) and/or SEM (Zeiss Supra 35VP FEG). The TP-1 test (Standard Test Procedure for Aluminum Alloy Grain Refiners) of The Aluminium Association [28] was used to compare the grain refinement efficiency because it permits to maintain a constant cooling rate and to determine the effect of the heterogeneous nucleation induced by the addition of grain refiners (cooling rate is $3.5°C/s). Grain size measurements were carried out on the base of the intercept method as per ASTM: E112. Cooling curves from liquid to solid of the selected composition were measured with K-type thermocouples and recorded by means of dedicated software (NI-VI Logger) collecting 100 data per second.

Results and discussion
3.1. Study of the addition of Nb to commercial pure Al 3.1.1. Analogies/differences between the Al-Ti and Al-Nb binary systems Nb is a promising candidate for the refinement of Al and its alloys because, like Ti, it is characterised by a peritectic reaction with Al. Moreover, Nb forms isostructural phases [29] to titanium aluminide, specifically Al 3 Ti and Al 3 Nb. Fig. 1 shows a sketch of the Al-rich part of Al-X (where X = Ti, Nb, etc.) binary phase diagram to emphasise the peritectic reaction. In the case of the Al-Ti phase diagram, Ti and Al react to form the Al 3 Ti intermetallic at 665°C. This compound has a tetragonal structure based on an ordered cubic close packed structure [30] (Ti content of 0.12-1.15 [29]). Nb has a very similar behaviour because the Al-Nb phase diagram is characterised by a peritectic reaction at approximately 661°C and (Nb content of 0.15-0.22).
From Table 1, Nb is characterised by a body-centred cubic (B.C.C.) structure with lattice parameters similar to those of Al and/or Ti. These data are important to calculate the lattice mismatch, which was done by following Bramfil work [31]: where (hkl) s = a low-index plane of the substrate, [uvw] s = a low-index direction in (hkl) s , (hkl) n = a low-index plane in the nucleating solid, From Eq. (1), a simpler equation can be derived for the plane and direction with the lowest mismatch between the substrate (potential heterogeneous nucleation site) and the nucleating solid which is labelled as lattice parameter mismatch (f). The lattice parameter mismatch (f) reported in Table 1  From the data shown in Table 1, it can be seen that Nb has higher melting point and slightly lower lattice parameter mismatch with Al in comparison to Ti which would favour the solidification of molten Al. Another important point is that the intermetallic Al 3 Nb is characterised by a tetragonal structure with 8 atoms per unit cell exactly as the Al 3 Ti phase and they have equal lattice mismatch with Al.

Effect of Nb addition on pure Al
Initially, a first set of experiments was carried out to study and confirm the promising potential of Nb as grain refiner and for that commercial pure Al was considered. In particular, pure Al was melted at 720°C and different percentages of Nb (0.1-15 wt.%) in the form of powder (average particle size of 100 lm) were added to the melt, left to dissolve for at least 1 h mixing the melt several times and slow cooled (cooling rate $0.3°C/s). The results of the experiments carried out to study the effect of the addition of different level of Nb to Al are summarised in the micrographs shown in Fig. 2.
At it can be seen from the results shown in Fig. 2, even though of its promising characteristics as potential nucleation site, the preliminary experiments with the addition of Nb (0.1-15 wt.%) to pure Al melt were not satisfactory because relatively large primary Al a-grains generally in the order of 1-5 mm were obtained. This unsatisfactory result could be mainly due to two factors: the poor dissolution of Nb in Al and the higher density of Nb with respect to Al.

Effect of solidification temperature combined with Nb inoculation
The dissolution process of solid in liquid metal is commonly described by the Nernst-Shchukarev equation and, consequently, the bulk concentration in the melt changes according to an exponential law (Eq. (3) [35]): where C is the concentration of the dissolved metal in the bulk of the melt measured at the time (t), C S is the saturation concentration, k is the dissolution rate constant, s is the specimen surface area and v is the volume of the melt. The values of the solubility (K) of Nb in Al calculated on the base of Eq. (3) are available in the literature [36] and are reported in Table 2 with the values of the dissolution rate constant (K 1 ) and the coefficient of diffusion (D).
As it can be seen from the data shown in Table 2, the maximum solubility of Nb in Al increases with the processing temperature

Temperature [°C]
Liquid Liquid + α (i.e. Al 3 X)  and, thus, a new set of experiments were programmed. Specifically, Nb content was set to 1 wt.%, the solidification temperature of pure Al melted at 750°C was ranged between 660°and 720°C and the samples were left to solidify inside the crucible (cooling rate $0.3°C/s). A representative example of the results from these set of experiments are summarised in Fig. 3 by means of the macroetched cross-sections of the casts where the sedimented Nb particles were found. From the macrographs shown in Fig. 3 it can be noticed that there is not a significant refinement with the addition of Nb if not in the bottom part of the solidified metal. In order to prove that the grain refinement of the bottom part of the metal left to solidify inside the crucible is actually due to the presence of Nb particles, the cross-section of the casts were cut, polished and microscopically analysed and the results are shown in Fig. 3b).

Comparison of the addition of Al-Ti-B and Nb to commercially pure Al
In order to favour the dissolution of Nb into the molten Al, the processing temperature was increased to 800°C, the particle size of the Nb powder was reduced (<45 lm) and the holding time was prolonged to 2 h. These dissolution conditions were employed for the comparison of the grain refinement efficiency of the Nb addition with that of the commercial Al-5Ti-B master alloy using TP-1 tests (pouring temperature: 680°C). Fig. 4 shows the comparison of the size of the primary a-Al grains without and with the addition of 0.1. wt.% of Nb or commercial Al-5Ti-1B master alloy.
The quantification of the a-Al grain size (Fig. 4) revealed that the solidification of pure Al under the conditions specified in the TP-1 test leads to quite coarse grain size (i.e. 1700 ± 40 lm). The addition of the grain refiners, both the commercial Al-5Ti-1B master alloy and Nb, drastically reduces the grain size of commercially pure Al of about one order of magnitude. Moreover, the commercial grain refiner performs better (i.e. 300 ± 25 lm) in comparison to the addition of Nb (i.e. 1210 ± 15 lm).

3.2.
Study of the addition of interstitials (B and C) to the Al-Nb system 3.2.1. Analogies/differences between the Al-Ti-B/Al-Nb-B and Al-Ti-C/Al-Nb-C ternary systems Keeping in mind that the development of grain refiners was based on the proposed nucleation theories for the Al-Ti-B system [7][8][9] where the combined presence of Ti and B is responsible for the grain refinement, the combined addition of Nb with interstitials such as B and C was considered. The addition of these interstitial elements should lead to the formation of niobium boride (NbB 2 ) and niobium carbide (NbC), respectively. As it can be seen from the data reported in Table 1, TiB 2 and NbB 2 are characterised by the same crystal structure (hexagonal) and have similar lattice parameter mismatch with Al and NbC has a similar structure and lattice parameter mismatch of TiC. Specifically, the lattice mismatch of TiB 2 is slightly higher than that of NbB 2 whilst that of TiC is somewhat lower than that of NbC. The formation and the presence of these hard refractory ceramic particles inside the Al melt should favour the grain refinement of Al by heterogeneous nucleation.

Table 2
Solubility of niobium in aluminium as a function of the processing temperature [36]. inside the crucibles (cooling rate $0.3°C/s). Fig. 5 shows the comparison of the microstructural features between pure Al without and with the addition of Nb and B or C.
As it can be seen in Fig. 5a), commercially pure Al slow cooled inside the crucible is characterised by a very big primary Al a-grains in the order of millimetres. On the one side, the inoculation by the combined addition of Nb and C (Fig. 5b) has very little effect on the grain refinement of commercially pure Al because the size of the primary Al a-grains is of the same order of that of Al without addition of any grain refiner. On the other side, the addition of Nb-B to Al (Fig. 5c) induces a significant reduction of the primary a-Al structure resulting in an average grain size of approximately 350 lm. From this set of experiments it can be evinced that the grain refinement effect of Nb is significantly enhanced by the addition of B but not C.

Optimisation of the B addition
It was found that the addition of the B powder to the Al melt is a limiting factor due to the non-wetting behaviour of B in liquid Al and to the difference in terms of density. Specifically, B floats in the surface of the molten Al because its density (2.08 g/cm 3 ) is lower than that of Al (2.7 g/cm 3 ). A common industrial practise is the use potassium fluorides to introduce the alloying elements needed to fabricate the grain refinement. An example is the employment of potassium tetrafluoroborate (KBF 4 ) and potassium hexafluorotitanate (K 2 TiF 6 ) fluxes to obtain the commercial Al-5Ti-1B master alloy generally used as grain refiner for wrought

µm]
Al Nb Al -5Ti -1B Al 3 Ti þ AlB 2 ! TiB 2 þ 4Al ð6Þ Similarly, in the case of addition of pure Nb powder and potassium tetrafluoroborate to molten Al, the individual reactions can be described by means of Eqs. (5), (8) and (9) and the global reaction by Eq. (10): Nb þ 3Al ! Al 3 Nb ð8Þ The employment of potassium tetrafluoroborate flux is not only advantageous from the point of view of introducing B into the molten Al preventing the floating of the powder but it also contributes to the dissolution of the Nb powder particles. This is because the reaction taking place between Al and the potassium tetrafluoroborate salt is exothermic and, therefore, locally the temperature inside the melt is much higher. The local increment of the temperature increases the solubility as well as diffusion rate of Nb in Al (see Table 2) and, thus, it will boost up the dissolution of the Nb powder particles added to the melt. Once identified the best source for adding B and confirmed that the procedure employed guarantee the complete dissolution of Nb, the attention was focused on the optimisation of the amount of B to be added for an effective grain refinement. For this set of experiments, pure Al was melt at 800°C, Nb powder addition was varied between 0.01 wt.% and 5 wt.% whilst the total amount of B, in the form of KBF 4 , added was ranged between 0.05 wt% and 1 wt.%. After one hour of homogenisation, TP-1 tests were done at a pouring temperature of 680°C. The results of the optimisation of the Nb/B addition are presented in Fig. 6.
Form Fig. 6, it can be noticed that the addition of Nb or B alone has also a very little effect on the final grain size of Al. When considering the Nb-B inoculation of Al (Fig. 6), it can be seen that the Nb-B mixture is very effective for grain refinement with the exception of the products where the lowest amount of B considered (i.e. 0.05 wt.%) is employed whose grain size is comparable to that of the reference material and/or the addition of exclusively Nb.

Comparison of the addition of Al-Ti-B and Nb-B to commercially pure Al
On the base of the results of the variation of the grain size as a function of the combined effect of Nb and B (Fig. 6), the amount of these two elements was set to 0.1 wt.% (0.1Nb-0.1B) because this is the lowest percentage which guarantees the grain refinement without significantly change the chemical composition of the material. In these set of experiments the addition of the commercial Al-5Ti-1B master alloy to pure Al was compared with that of the 0.1Nb-0.1B powders addition (TP-1 test). Representative micrographs of the results of the experiments done are shown in Fig. 7a-c) whilst the variation of the primary Al a-grain size is graphed in Fig. 7d).
As it can be seen from the results presented in Fig. 7, there is not any significant difference in terms of grain size between the commercially pure Al without and with the addition of any grain refiner when considering a solidification temperature of 660°C because the average grain size is approximately 200 lm. Nonetheless, it can be noticed that the mean primary Al a-grain size significantly increases with the increment of the pouring temperature reaching almost 3300 lm without the addition of any grain refiners whereas the increment of the grain size is far less limited after addition. Moreover, it can be seen that the difference in grain size between the reference material and the grain refined materials also increases with the solidification temperature reaching one order of magnitude (i.e. 300 lm vs. 3000 lm) at 720°C. When considering the effect of the each grain refiners, it can be seen that the commercial Al-5Ti-1B master alloy performs slightly better than the 0.1Nb-0.1B powders because there is a constant difference of approximately 200 lm in between them independently of the pouring temperature studied.

3.3.
Study of the addition of the 0.1Nb-0.1B grain refiner to Al-xSi binary alloys 3.3.1. Analogies/differences between the Ti-Si and Nb-Si binary systems As reported in the introduction, the grain refinement of Al-Si alloys cannot be performed efficiently by Al-Ti-B master alloys because when the Si percentage is greater than 4 wt.%, Ti and Si react leading to the formation of silicides (poisoning). A better understanding of the compounds that form between Ti, Si and Ti Si Nb Fig. 8. The Ti-Nb-Si ternary system (adapted from [37]) showing that Ti-Si compounds form and are stable at lower temperatures in comparison to Nb-Si compounds. Nb can be achieved by studying the Ti-Nb-Si ternary phase diagram (Fig. 8).
From the Ti-Si binary phase diagram (Fig. 8) five titanium silicides (Ti 3 Si, Ti 5 Si 3 , Ti 5 Si 4 , TiSi and TiSi 2 ) can be formed. In the case of the binary Nb-Si phase diagram (Fig. 8) it can be noticed that Nb has a similar behaviour to Ti and three intermetallic compounds, precisely Nb 5 Si 3 , Nb 3 Si and NbSi 2 , can form. Nevertheless, there is a very important difference among the two binary phase diagrams and it is that Ti 3 Si is stable at much lower temperature (<1170°C) as compared to Nb 3 Si which is stable between $1700°C and 1980°C (Fig. 8) [37]. It is worth mentioning that, in general, niobium silicides are high temperature intermetallics whose melting temperature is higher with respect to that of titanium silicides. As it was found experimentally, Si has very little solubility in both AlNb 3 and AlNb 2 intermetallics, and only $2 at.% in Al 3 Nb [38]. This suggests that an Al melt at 800°C there will be extremely less probability to form niobium silicides than titanium silicides and, therefore, the Nb-based phases would act as heterogeneous nucleation sites for Al without poisoning.

Comparison of the addition of Al-Ti-B and Nb-B to hypoeutectic binary Al-xSi alloys
In order to clearly identify the effect of the addition of the Nb-B refiner, binary Al-xSi alloys were studied. It is worth mentioning that, commonly, commercial Al-Si cast alloys used in industry have a Si content greater than 4 wt.%. Nonetheless, in this study Al-1Si and Al-2Si alloys were also produced and studied to better understand and characterise the Nb-B inoculation behaviour. The results of the characterisation of the TP-1 tests done from a pouring temperature of 700°C are summarised in Fig. 9. In particular, a comparison of the microstructural features by means of micrographs of selected binary Al-7Si and of the variation of the primary Al a-grain size with the Si content is presented. From Fig. 9 it can be seen that the refinement of the binary Al-7Si alloy (Fig. 9a) through the addition of commercial Al-5Ti-1B master alloys (Fig. 9b) is not effective because both materials are characterised by a very coarse dendritic microstructure. Conversely, the grain refinement with the 0.1Nb-0.1B refiner (Fig. 9c) leads to a much finer microstructural features. From a Si content of about 4 wt.% the primary Al a-grain size continuously increases and the binary Al-10Si alloy has similar grain size (2150 lm) to that of pure Al. In the case of the commercial Al-5Ti-1B master alloy addition, the primary Al a-grain size increase continuously starting from a Si content following the same trend of the binary Al-xSi alloys and reaching a maximum value of 1850 lm in the binary Al-10Si alloy. Considering the addition of the 0.1Nb-0.1B refiner in the form of powders, it can be noticed that this is highly efficient in refining commercially pure Al, because the grain size is reduced of one order of magnitude. Nonetheless, the most important point is that Nb-B inoculation leads to the grain refinement of binary Al-xSi alloys throughout the whole spectrum of Si content analysed (i.e. 1-10 wt.%).

Effect of the Nb-B inoculation in binary Al-xSi alloys on the base of thermal analysis
Thermal analysis has been used to assess the performance of grain refiners and it is an efficient tool to determine the refining efficiency of a grain refiner [8]. Fig. 10 shows the cooling curves of the binary Al-5Si alloy without and with Nb-B inoculation studied to measure the undercooling and the nucleation temperature of the primary Al a-grains as well as to determine whether heterogeneous nucleation is the main mechanism of solidification.
Concerning the data of the thermal analysis ( Fig. 10 and Table 3), the nucleation temperature (T N ) is identified as the point where the cooling curve starts to deviate from linearity. This is the temperature when the first Al crystal nucleates and starts to grow in the 20 mm 20 mm   [39].
From the analysis of the data concerning the primary a-Al nucleation ( Fig. 10a and b and Table 3), it can be seen that Nb-B inoculation leads to a somewhat higher nucleation temperature but the most important point is that the undercooling is lower in comparison to the untreated alloy (Al-5Si). A perfect heterogeneous nucleating site would reduce the activation energy for nucleation to zero. Thus, the extent of undercooling is indirectly proportional to the efficiency of a grain refiner in providing potent nucleating sites [8]. From the literature available for the grain refinement of Al-Si alloys, it is suggested that a good grain refinement is achieved when the undercooling is below 0.3°C [40]. In the case of Nb-B inoculation, the undercooling is lowered in comparison to the untreated materials indicating that it is actually a potent nucleation agent which induces heterogeneous nucleation. From the analysis of the macroetched surface of the binary Al-5Si alloy it was found that the Nb-B inoculation is highly effective even at very slow cooling rate. Specifically, the mean primary a-Al grain size of the untreated Al-5Si alloy (Fig. 10c) is approximately 2500 ± 130 lm whereas Nb-B inoculation leads to a grain size of about 380 lm.
The cooling curve of the binary Al-10Si alloy is presented in Fig. 11 where it can be seen that Nb-B inoculation lowers the eutectic nucleation temperature and leads also to a greater recalescence. Specifically, the untreated alloy shows low recalescence prior to eutectic growth ($0.3°C) whereas after Nb-B inoculation the material has a DT of 1.2°C.
It is well known that the larger the magnitude of the recalescence, the higher the level of modification [41]. The term ''modification'' identifies the addition of chemicals to induce the change of the morphology of the Al-Si eutectic phase normally composed of large plates and some lamellar structure (faceted phase) to a much finer and fibrous eutectic flake-like phase. The most common elements used in industry for the modification of the coarse plate-like eutectic phase of Al-Si cast alloys are strontium and sodium [39]. The results of the eutectic nucleation after Nb-B inoculation indicates that this composition is not only highly effective for the refining of the primary a-Al grain structure of Al-Si cast alloys but, simultaneously, has some effect (i.e. refine but not modify) on the Al-Si eutectic phase of the binary Al-xSi alloys. The analysis of the micrographs of the untreated Al-10Si alloy (Fig. 11c) shows that the microstructure is composed of dendritic coarse primary a-Al grains (around 2000 lm), quite large and unevenly distributed eutectic needles (50-80 lm) and quite large amount of primary Si particles of approximately 20 lm. As expected on the base of the data plotted in Fig. 9d), Nb-B inoculation in the binary Al-10Si alloy significantly reduces the primary a-Al grain size (Fig. 11d). Nevertheless, the microstructural analysis confirms also that the addition of Nb-B to Al-xSi alloys is also able to refine the eutectic phase which switches to an even distribution of fine intermetallic particles of about 10-30 lm. Table 3 Primary a-Al or eutectic nucleation temperature (T N ), minimum temperature (T min ), growth temperature (T G ) and undercooling for primary a-Al nucleation or recalescence for eutectic nucleation (DT) as measured from the cooling curves shown in Fig. 10.

Conclusions
From the study of the addition of Nb both alone and in combination with B to Al and Al-Si alloys it can be concluded that: The Al-Nb binary diagram is characterised by a peritectic reaction which permits the formation of Nb aluminide particles which are able to refine the microstructure of commercially pure Al. Nevertheless, Nb is not as efficient as Ti or the commercial Al-Ti-B master alloy due to its higher lower growth restriction factor. The combined addition of B and Nb to commercially pure Al leads to finer microstructure in comparison to the untreated material due to the reduction of the constitutional undercooling. The final grain size is comparable, slightly coarser, to that obtained by adding commercial Al-Ti-B master alloy. The developed Nb-B grain refiner can efficiently refine binary Al-xSi alloys due to the formation of niobium borides which are more stable than titanium borides, therefore, significantly limiting the so called poisoning effect. The best performance of the novel Nb-B grain refiner is for Al-Si alloy with Si content greater than 6 wt.%. The addition of the novel Nb-B grain refiner does not only leads to finer primary Al a-grains but also to finer Al-Si eutectic phase due to the more homogeneous distribution of the alloying elements in the solidification front and eutectic pools.