The influence of NbB inoculation on dendritic spacing and grain size of an aluminum 2017 alloy at different cooling rates

The Al-4Cu-Mg 2017 alloy is a significant Al extrusion alloy. As a result, knowing its behavior in terms of grain refining during casting stages becomes critical for the next processing steps. In the face of the shortcomings exhibited by TiB inoculants, the addition of NbB could be tested for this alloy as an alternative. The present article investigates the effects of NbB addition on grain size (GS), on secondary dendrite arm spacing (λ2), and on hardness of centrifugally cast (CC) and directionally solidified (DS) 2017 alloy samples, encompassing a broad range of cooling rates and solidification conditions. The λ2 did not change much as a result of the NbB inoculation, either for DS or CC samples, decreasing from 7.2 to 7.0 μm in the case of the 4-mm-thick CC samples and increasing from 17.0 to 22.5 μm in the case of the DS samples. The average GS, on the other hand, reduced from 293 to 136 μm for the NbB-containing CC samples and from 756 to 313 μm for the DS samples associated with cooling rates of 10 K/s. Both processes can achieve nearly the same GS reduction ratio, namely 2.2 for the CC and 2.5 for the DS. This implies that GS reduction ratio was barely affected if similar cooling rates were compared while taking the experimental range for both processes into account. Moreover, Vickers hardness results revealed that GS had little influence on this property for both alloys studied. The experimentally obtained data translated a variety of solidification condition outlines for the 2017 alloy. Through this data, the understanding of an alternative inoculant in a significant Al alloy series advanced significantly.


Introduction
The Al-4 wt% Cu alloy (2xxx Series) is one of the most important commercial Al alloys. It is frequently used in the fabrication of different components for several purposes, such as structural parts in the automotive and aircraft industries, due to their high mechanical properties, which may be adjusted through further processing [1,2]. In particular, the 2017 alloy is considered an extrusion alloy of relevance. In recent decades, Al extrusion has become increasingly popular in manufacturing, with a global market size of 28.08 million metric tons in 2020 [3]. This market is being driven by a surge in the development of ecologically friendly, energyefficient, green buildings, as well as a developing automotive industry, particularly electric vehicles.
Processability of components during production is highly correlated with the as-cast resulting microstructure length scales; therefore, controlling the grain size (GS) during the casting stage becomes a critical factor. Grain refinement of these alloys is extremely important, as the formation of fine α-Al equiaxed grains during cast production may promote increased formability and homogeneity required by these alloys for the next stages of manufacturing.
Furthermore, the composition of the 2017 alloy is not significantly different from that of the 204.0 and 206.0 Al-Cu cast alloys [4]. As a result, understanding the inoculation of the 2017 alloy with an NbB refiner broadens knowledge of typical alloys applied in foundry processes employing sand and permanent molds. These alloys (204.0 and 206.0) contain 4 to 6% Cu and 0.25 to 0.35% Mg, with highly restrictive impurity (iron and silicon) limits, and in some cases also contain 0.25 to 0.35% Mn or Cr. The 2xxx alloys also have the highest mechanical strengths and hardnesses of all casting alloys at elevated temperatures (up to 300 °C), and this factor accounts for their use in some applications. Inoculation, dendritic spacing, and grain size control are also critical for these alloys.
Nucleation theory can be used to assess the basic requirements for an efficient nucleant [5]. To propitiate crystal formation on a nucleant, the interface between the nucleant and the liquid should have a higher energy than the interface between the nucleant and the solid crystal. Providing a nucleant crystal relationship associated with a good crystallographic fit between the respective crystal lattices is one way to maximize this condition [6][7][8]. Liu [5] divided the Al refining theories into four categories, and among the theories related to the peritectic reaction, the duplex nucleation theory in which TiB 2 is covered by a thin layer of Al 3 Ti acts as a nucleation site for α-Al; and the peritectic reaction theory through which α-Al nucleates from liquid and Al 3 Ti, seem to be well accepted as Al-Ti-B refiners. Despite such knowledge, this type of refiner has some limitations.
According to Ding et al. [9], the traditional Al-Ti-B refiners have some shortcomings when facing refining 2xxx alloys. They can be poisoned by Si, Zr, V, Cr, etc. Because most 2xxx alloys contain these elements, the Al-Ti-B refiner may not be the best choice. Regarding the grain growth controlling in Al alloys, it is worth noting the effectiveness of the NbB inoculant in other alloys of industrial importance [10][11][12], especially in conditions of intermediate/high solidification rates, compatible with various industrial casting processes.
Al-Nb-B master alloys offer an effective method for refining Al-Si casting alloys, as discussed recently by Li et al. [13]. Fundamental information about the grain refining mechanism of refiners based on Nb-B in Al-Si cast alloys was discussed, revealing that NbB 2 provides the main nucleation site for α-Al, together with a metastable Al3Nb phase that forms in its basal plane, in accordance with the duplex theory.
The nucleating sequence proposed by Li et al. [13] point out that the Al 3 Nb intermetallics dissolves at temperatures above the α-Al melting point and then re-precipitates on the NbB 2 basal plane at a certain orientation relationship, reducing the misfit between NbB 2 /α-Al from 8.4 to nearly 1%. In the meantime, the exposed Nb atoms on some NbB 2 particle surfaces can be substituted by Al to form (Al,Nb) B 2 intermediate layers where the SiB 6 phase particles precipitate. This phase also reduces the misfit of NbB 2 /α-Al and contributes to the heterogeneous nucleation. The role of Nb-B refiners on Al-Cu alloys has not been discussed, but the grain refinement of an Al-3wt% Cu alloy by Al-5Ti-B was discussed recently by Wu et al. [14] using the same techniques as Li et al. [13]. It was possible to make some parallels between the two refiners, especially because of their crystallographic similarities and the mechanisms proposed, which are quite similar. Therefore, given the importance of the 2xxx alloys, the possibility of using this inoculant, which has not yet been widely verified, becomes attractive. Correct inoculant addition and process control can provide favorable GS and secondary dendritic arm spacing (λ 2 ) arrangement for a given alloy.
GS and λ 2 are considered key microstructural features in Al alloys [15]. Aside from their effects on the alloy properties, the relationship between them may be used to tailor grain morphology, ranging from dendritic grains when the GS/ λ 2 ratio is high to cellular grains when the coarsening of λ 2 is so great that the ratio between them is small. λ 2 , on the other hand, is less susceptible to nucleation and is more connected with heat extraction during growth [16].
The purpose of this study is to evaluate the effects and efficiency of refining resulting from the addition of the NbB inoculant in the 2017 alloy, because no scientific studies in this regard could be found in the literature. Furthermore, quantitative relationships between GS, λ 2 , and solidification cooling rate will be established for both alloys considering a large range of cooling rates through centrifugal and directional solidification techniques. The changes generated by the addition of NbB on GS and λ 2 will be outlined. Vickers hardness tests were also performed on the samples with and without NbB to assess the impact of the changes on this property for a variety of microstructure length scales and conditions of solidification.

Materials and methods
Thermodynamic calculations using the Thermo-Calc® software, version 2021, together with a TCAL 7 database (v7.1) through the CALPHAD method were performed for both the master alloy and the Al-Cu alloys studied.
The 2017 alloy was first produced in an induction furnace, Inductotherm, VIP Power-Trak model, with a power source of 50 kW and a frequency of 3.2 kHz, from commercial-grade raw materials with charge balance. The addition of the NbB inoculant was carried out through an additional melting operation so that the Al-5.1 wt% Nb-1 wt% B master alloy could be introduced, resulting in the modified 2017 alloy having 0.5 wt% Nb and 0.1 wt% B. The inoculation has been verified through grain size reduction occurrence as well as qualitative chemical proof of concentrations of Nb present in some phases performed by chemical mappings by the EDS/SEM technique. To prevent gases trapped inside the material during solidification, both alloys were degassed for 2 min before being poured. A thin stainless steel tube covered with silicoaluminous refractory paint was inserted into the alloy in the liquid state before the metal pouring process. Argon gas was inserted through this channel, forming small gas bubbles that pass through the liquid metal between the end of the tube and reach the surface.
The 2017 base alloy composition can be seen in Table 1. These alloys were cut, cleaned, and inductively melted in a hermetic chamber set perpendicular to a rotating axis in the centrifugal casting equipment (Linn High Therm, Titancast 700 VAC model). Afterwards, the force generated by the chamber's rotation drove the molten alloy into the Cu mold cavity (400 rpm). Samples with 3-mm and 4-mm thick plate geometry of both alloys were generated at relatively high solidification rates. Two thicknesses were adopted to cause variation in the solidification cooling rate.
For the directional solidification, the molten alloy was poured into a mold. When the melt temperature reached 5% above the liquidus temperature, the electric heaters were turned off, and the water flow at the bottom of the container was initiated. For this reason, the cooling system was started as soon as the melt reached the correct temperature. The watercooled base promoted the onset of solidification and kept the system running until the casting was fully formed. The mold was cylindrical and longitudinally split, made of AISI 310 stainless steel, with a height of 160 mm, an internal diameter of 60 mm, and a wall thickness of 5 mm. The liquid-contact surface of the carbon-steel base plate had a 1200 mesh finish. The evolution of temperatures along the casting length was monitored using fine K-type thermocouples, whose tips were distributed longitudinally along the length of the casting.
The solidification cooling rates were calculated concerning the movement of the liquidus α-Al isotherm throughout the casting length. After each thermocouple had passed the liquidus isotherm, the cooling rate (Ṫ) was calculated by taking the local time (t) derivative of each cooling curve (dT/dt).
Several CC and DS samples were ground and etched with a Keller's solution (1% HF + 1.5% HCl + 2.5% HNO 3 in water) for 30 s to reveal the grains. The same samples were ground, polished, and etched with 3% HF in water for 15 s to reveal the dendritic array. After capturing images with an optical microscope, the primary, λ 1 , and secondary, λ 2 , dendritic spacings were assessed using the triangle and intercept methodologies, respectively [17], using an optical microscope, whereas the grain size (GS) was calculated using the Heyn linear intercept method [18] after images were taken using a stereomicroscope.
The microstructures of the Al-5.1Nb-1B master alloy and of the 2017 and 2017-NbB alloy samples were analyzed by a scanning electron microscope (SEM) equipped with an energy-dispersive spectrometer (EDS).
Vickers hardness tests were carried out with a test load of 2 kgf and a dwell duration of 15 s. The average of at least 10 measurements was used to determine the representative hardness value for each sample. Figure 1 shows the Scheil curve describing the master alloy solidification, calculated by the CALPHAD method. As discussed by Li et al. [14], the NbB 2 phase is the main heterogeneous nucleation site in NbB-based refiners and although its presence has not been verified in Fig. 1, this phase is commonly observed in the Al-Nb-B master alloy [10,11,19]. In fact, ref. [19] describes this phase as arising from the (Al 3 Nb + →AlB 2 NbB 2 + 4Al) reaction, whose preceding phases are indicated on the Scheil curve in Fig. 1. Figure 2 shows a representative SEM image (BSE) of the master alloy and a specific region where spot semiquantitative chemical analyses were performed by EDS points and EDS mapping. Figure 2a reveals a distribution of intermetallic particles of varying sizes, dispersed in the Al matrix. According to the EDS mapping shown in Fig. 2b, these particles are possibly Nb-based intermetallics. Although the presence of B element cannot be confirmed with precision by EDS analysis, due to its low atomic weight, chemical mapping shows this element in association with Nb, in the ring-shaped structure that surrounds the Al-and Nb-enriched phase. This may corroborate to the NbB 2 description as a result of a dissociation of the Al 3 Nb phase [19]. As a matter of fact, other authors [10,11,19] that studied the NbB master alloys identified the presence of niobium aluminides (Al 3 Nb) and niobium borides (NbB 2 ) through SEM and XRD in alloys with similar composition. Besides, Bolzoni and Babu [19]

Directional solidification (DS)
The experimental cooling curves for the eight thermocouples inserted into the melt during solidification of the 2017 and 2017-NbB alloys are shown in Fig. 3a and b. The liquidus temperatures were progressively reached from the bottom to the top of the casting, revealing a strong unidirectional heat extraction. These data were processed so that it was possible to calculate the solidification cooling rates and velocities (see Fig. 3c and d, respectively) along the length of the castings and compare them to the microstructure features and parameters. The details on the methods employed for determining the solidification thermal variables can be seen elsewhere [20,21].
Through the knowledge of experimental growth laws, the variation in solidification velocity and cooling rate have been mapped along the casting in directional solidification studies with a transient heat extraction regime and connected to the resulting microstructure [22][23][24][25]. Figure 3c and d, respectively, depict the behaviors of the studied castings' cooling rate and solidification velocity. Similar patterns in the Solidification velocity (mm/s) solidification kinetics of the two alloys can be seen, allowing the microstructural length scale differences between the two alloys to primarily be attributed to the presence of the inoculant if considering a specific cooling rate or solidification velocity during the subsequent microstructural analysis. The solidification velocity consists of the speed with which the solid-liquid interface travels through the volume of the material during the solidification process. This parameter is obtained by measuring the solidification temperature of the matrix phase, Al in this case, at different points in the volume during the solidification process, establishing the corresponding times of passage. This factor also indicates the position of this interface along the volume of metal during the solidification process. The experimental interrelations between position and time result in the solidification velocity during solidification. The 0.5 wt% Nb-0.1 wt% B addition does not appear to have had an impact on the solidification kinetics, which mainly depend on the alloy wettability, thermal properties, and solidification range [26][27][28]. As a result, it is possible to treat the changes in the cooling rate and solidification velocity along the castings as being nearly identical.
The grains that form as both alloys solidify are shown in Fig. 4. Images corresponding to two distinct cooling rates can be examined. The GS is significantly reduced by NbB, and each alloy also seems to be sensitive to the cooling rate. The 2017-NbB alloy has a fine-grained, highly homogeneous microstructure, as that typically found in industrial processes. Figure 5 shows two examples of optical microstructures for each DS alloy. One is related to slow cooling down and another to higher cooling process. Interestingly when both alloys are compared for the same cooling rate of approximately 1 K/s, both samples demonstrate similar dendritic scales.
Both alloys are distinguished by the exclusive presence of dendrites with well-defined ramifications. This structure is typical for the as-cast state and is formed by α-Al dendrites surrounded by intermetallic phases that mainly crystallize as eutectics and are dispersed throughout the interdendritic areas. The addition of NbB is not in favor of this feature's refinement under such conditions. A reduction trend in the secondary dendritic arm spacing happens as the solidification velocity increases as can be seen in Fig. 6. However, the inoculation with NbB was not effective in decreasing λ 2 .
As the solidification velocity increases, there is a decreasing tendency in the secondary dendritic arm spacing, as seen in Fig. 6 for either non-inoculated or inoculated alloys. This means that the solidification velocity may impact such microstructure parameter. In contrast, NbB inoculation showed no significant impact on λ 2 in micrographs comparing the same velocity for both alloys.
Experimental relationships between λ 1 , λ 2 , and GS during solidification could be developed with the solidification thermal parameters using the optical images in Figs. 4, 5, and 6, their data, and additional points for the microstructural parameters, as shown in Fig. 7.
The main interest in GS is to correlate this parameter with the cooling rate, providing an understanding of the microstructural uniformity for different types of casting processing [10][11][12]. Moreover, λ 2 value fundamentally depends on the total time that the secondary dendritic branch is in contact with the liquid, which is linked to the solid/liquid interface displacement during solidification, that is, to the solidification velocity [29]. The dependence of λ 1 on thermal gradient, G, and solidification velocity, v, was established by numerous researchers [30,31]. The cooling rate, which is equal to G × v, can also be used to translate this dependence, as can be seen in Fig. 7. In Fig. 7, points are experimental results and lines represent an empirical fit to the experimental points, expressed as power functions. Figure 7a is elaborated considering the whole microstructural variation of several samples taken transversally to the direction of heat flux and the association of the primary dendritic spacing with the cooling rate for both alloys. The primary dendritic spacing in the 2017 alloy varied between 55 and 212 µm while in the 2017-NbB alloys ranged from 57 to 349 μm. The balance between the whole data taking into account small changes in the cooling rate and λ 1 permitted establishing a function of the cooling rate according to the experimental law 1 = 220Ṫ −0.55 . Microstructure coarsening in both alloys is due to both the increase in thermal resistance imposed by the progressive increase in the solidified layer and the rise in metal/mold interfacial thermal resistance as solidification progresses. It can be seen that the experimental variation of λ 1 with cooling rate follows a − 0.55 power law, as also observed by Rocha et al. for binary Al-Cu chemistries [32]. The observations of Bouchard and Kirkaldy [30,31] that an exponential relationship with − 0.5 exponent best generates their experimental results in unsteady-state solidification are in good agreement with this finding. While Rocha et al. [32] found that the Cu content has little effect on λ 1 for binary Al-Cu alloys with a single power law representing the variation of spacings with the cooling rate for Al-5Cu, Al-8Cu, and Al-15Cu alloys, agreeing with Sharp and Hellawell [33], who demonstrated that λ 1 is almost independent of composition, the current findings indicate that the low Si, Fe, Mn, and Mg alloying contents may play a role in slightly lowering λ 1 .
For the 2017 alloy, the secondary dendritic spacing varied from 17.0 to 38.5 µm as a function of the solidification velocity, according to the experimental law 2 = 24.5v −2∕3 , while in the modified alloy, the variation was from 22.5 to 47 µm, under the experimental law 2 = 28.2v −0.6 . It is observed that the grain size is sensitive to the cooling rate, but the NbB-modified alloy is less sensitive and governed by the experimental law GS = 475Ṫ −1∕5 . A similar cooling rate sensitivity through experimental growth equations deriving a − 1/5 exponent was recently reported for the GS in a 6201 alloy [20]. It is clear that the addition of NbB resulted in a more refined and uniform grain distribution.

Centrifugal casting (CC)
Rapidly solidified samples have been generated to verify the microstructures and changes induced by the addition of NbB. Recent studies in the literature suggest that the cooling rates experienced should fall between 100 and 300 °C/s. For instance, Sousa et al. [12] calculated that the centrifuged cast 6201 alloy sample cooling rates were approximately 240 K/s. Additionally, calculations performed by Gouveia et al. [34] for centrifugally cast Al-15 wt% Cu-7 wt% Si and Al-22 wt% Cu-7 wt% Si alloy samples ranged from 200 to 300 K/s. Figures 8 and 9 show some examples of grain morphologies formed in the CC Increase in the nucleation rate of the inoculated alloy during solidification caused the change from columnar to equiaxed grains. The proportion of equiaxed crystals generated as the solidification front advances determines the growth of equiaxed grains, which causes an increase in the number of nucleation sites to promote the blockage/ inhibition of the columnar structure [35]. The presence of the inoculant induced the formation of a greater number of nucleation sites, thus promoting the development of a microstructure formed by equiaxed grains, especially generated in the case of high cooling rates, such as those typified in solidification by centrifugation.
The average GS values associated with the 2017 alloy were 290 μm and 293 μm for the samples having 3 mm and 4 mm, while for the 2017-NbB alloy were 127 μm and 136 μm considering the same thicknesses. Even being processed at relatively high cooling rates, the inoculant proved to be efficient in reducing GS by at least 2 times. Moreover, different thicknesses were not effective in changing GS. This suggests that GS was unaffected by variations in cooling rate during centrifugal casting. (c) Table 2 shows the whole data obtained for the CC samples.
The presence of primary aligned dendrites can be seen in the left image of Fig. 10, in which optical images are shown. However, because their fraction was significantly less than the number of observed randomly developed secondary branches, only λ 2 was measured. The λ 2 did not undergo a significant variation as a function of the NbB inoculation, varying, for example, from 7.2 to 7.0 μm due to inoculation in the case of the 4-mm-thick samples. In contrast, a variation from 8.6 to 7.2 μm was observed for the 2017-NbB alloy due to the increase in cooling rate (i.e., decreasing thickness). Another interesting topic is the control of micromorphologies formed in Al alloys, which depends on the relationships between GS and λ 2 [16]. In this study, GS/ λ 2 ratios were approximately 32 and 15 for the 2017 and 2017-NbB alloys, respectively, considering the 3-mm-thick samples. Easton et al. [16] noticed that cells could be formed only for GS/ λ 2 lower than 3 in 2014 alloy samples inoculated with TiB 2 . The present results suggest that the fully dendritic morphology in Fig. 10 may be justified based on the high GS/ λ 2 ratios found here.
The results of thermodynamics computations of the 2017 and 2017-NbB alloys, in Figs. 11 and 12, provided the two main pieces of information. First, the Nb and B elements constituting the Al 3 Nb (AL3TI_D022) and AlB 2 (ALB2_C32) intermetallics, respectively, are formed before the crystallization of the α-Al phase, providing the possibility of acting as nucleation agents. The second information is that the fraction of intermetallics does not change between the analyzed alloys, disregarding the small fraction of intermetallics formed by Nb and B. Therefore, this calculus confirms that possible variations in the alloy's mechanical behavior can be attributed to the particle morphology and size variations, instead of the fraction of intermetallics. Moreover, the inoculant elements are present in the added master alloy, as discussed before, but this simulation is important to evaluate the chemical stability of the inoculant particles. It is possible to verify that both Nb and B demonstrate chemical stability, as they do not react with the other elements to form phases, a positive aspect for refiners. Figures 13 and 14 show SEM images of the CC 2017 and 2017-NbB samples, respectively. Figure 13 shows a typical microstructure formed in the as-cast state of the alloy of interest produced through centrifugal solidification. It is formed by α-Al solid solution dendrites and precipitates of intermetallic phases that mainly crystallize as eutectics and are dispersed throughout the interdendritic areas [28]. Cu, Mg, Mn, and Si, the four primary alloying elements, formed intermetallic phases with Al or between themselves and Fe. EDS mapping confirmed the distribution of the main elements throughout the microstructure, and the main formed phases were α-Al, Al 2 Cu, Mg 2 Si, and AlCuFe phases, as identified in the inset image at left in Fig. 13. The formation of the Al15Si2M4 phase, as predicted by the CALPHAD method, may be indicated by the presence of Mn and Si traces in the mapping of elemental spectra. These phases are comparable with those predicted by the thermodynamic calculations.
A combination of SEM observations regarding the morphology and contrast of the phases and EDS mappings was To complement the phase analysis and also to compare the results of centrifugal casting with those of directional solidification, regarding the phases forming the microstructure, Fig. 15 shows SEM images emphasizing the formed intermetallics under two different conditions, one to more rapid cooling rates of approximately 10 K/s, and the other to slow cooling at 1 K/s. Regardless of the cooling rate taken into account, for the 2017 and 2017-Nb alloys, the morphology of the interdendritic phases show the presence of lamellar Al 2 Cu (brighter phase) and black Mg 2 Si forming a ternary eutectic mixture with the α-Al phase (see inset images in Fig. 15). Similar phases and morphologies for the interdendritic zones of the 2017 alloy were found by Grazyna et al. [28]. This kind of eutectic has not been clearly seen in the CC samples. Moreover, the DS samples showed a higher fraction of globular structures filled with the eutectic mixture in the form of pockets, in contrast with more elongated phases formed in the microstructure of the CC samples.

Hardness Vickers
As can be seen in Table 2, the following mean hardness values were obtained for the CC samples: 74 HV for the 3-mm and 4-mm-thick 2017 alloy samples and 83 HV for both 2017-NbB alloy samples. From these hardness results, it was possible to observe that the samples having NbB had higher values of hardness, most likely due to the inoculating effect, which caused a reduction in GS. Another typical factor of importance in the study of mechanical properties of cast samples is λ 2 , which does not appear to have any influence on the variation of hardness for the alloys solidified under CC conditions. In Fig. 16, the results for the tested alloy samples' Vickers hardness as a function of GS are displayed. The behavior that the results show can give rise to two main arguments. The first is a reference to the ineffectiveness of raising hardness while lowering GS. Hardness did in fact slightly rise as GS increased. This suggests that the effects of GS on hardening are minimal for these alloys produced under such conditions. The refining by NbB addition was clearly not enough to improve the hardness of the 2017 alloy. While grain size may not be the most crucial factor for hardness, it can have an impact on a variety of other properties of these alloys, including their electrical conductivity, toughness, ductility, corrosion resistance, and wear resistance; all of which require further investigation.
The second argument explains why the hardness of both alloys rises as GS rises. It appears that further and strong effects of hardening are occurring. One can assume that the distribution, nature, and size of the reinforcing phases forming the microstructure may explain the hardness steadiness as a function of grain size with the tendency of slightly increasing.   This is because CALPHAD computations showed no effective change in the intermetallics fractions due to NbB addition. It appears that the larger intermetallics may increase overall hardness even for samples with larger spacings.
The evaluation of the SEM results indicates that in the case of the CC samples, the well-defined eutectics were not formed, and the distribution of intermetallics is concentrated in very thin interdendritic films. In the case of the DS samples, however, several areas composed of the ternary eutectic and globules of the same eutectic embedded in the α-Al matrix were observed. It seems that this last configuration is more effective as an obstacle making dislocations difficult to move and thus acting as a hardening factor during loading along the hardness test.
Additionally, the Al15Si2M4 phase, which is shown with an arrow in the inlet detail in Fig. 15, has been most frequently seen in the DS samples, particularly in those that solidified at lower cooling rates (i.e., higher GS), and it appears to have a significant hardening effect as well.
Further research is needed, according to Quested [36], to completely understand the specifics of grain refinement in different Al alloys, including Al-Cu-based ones. Quested [36] claimed that Al-Ti-B refiner may not always be successful in decreasing grain size, as also shown by Ding et al. [9] after examining 2xxx alloys. The presence of a certain solute content can interact with TiB2 particles, rendering  ineffectual in the grain-initiation process, which is not necessarily advantageous to grain refining. The current research seems to offer a different alternative to improve a high solute (Al-4Cu-Mg 2017) alloy, resulting in an effective GS reduction, by adding Al-Nb-B.

Conclusions
• The variations in solidification velocity (v L ) and cooling rate (Ṫ) have been mapped along the length of the directionally solidified (DS) 2017 and 2017-NbB alloys castings and connected with the resulting microstructure. The 0.5 wt% Nb-0.1 wt% B addition was shown not to have an impact on the solidification kinetics, i.e., the changes in v L and Ṫ along the length of the castings were nearly identical. • Both alloys were shown to be characterized by the exclusive presence of α-Al dendrites surrounded by intermetallic phases that mainly crystallize as eutectics and are dispersed throughout the interdendritic areas. The grain size (GS) was shown to be significantly reduced by the addition of NbB, and each alloy was also shown to be sensitive to Ṫ. Experimental relationships relating λ 1 and GS as a function of Ṫ and λ 2 as a function of v L have been developed. A reduction trend in λ 1 and λ 2 happens as Ṫ and v L, respectively, increase. However, the inoculation with NbB was not effective in decreasing λ 2 . • The average GS was shown to be reduced from 293 to 136 μm for the NbB-containing centrifuged cast (CC) samples and from 756 to 313 μm for the DS samples associated with Ṫ of 10 K/s. Both processes can achieve nearly the same GS reduction ratio due to the NbB, namely 2.2 for the CC and 2.5 for the DS. However, Vickers hardness tests have shown that GS has little effect on such property for any alloy examined. Funding The authors acknowledge FAPESP (grants 2019/23673-7 and 2020/02697-2) and CNPq. This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior-Brasil (CAPES)-Finance Code 001.

Competing interests
The authors declare no competing interests.