Sn-0 . 5 Cu (-x ) Al Solder Alloys : Microstructure-Related Aspects and Tensile Properties Responses

In this study, experiments were conducted to analyze the effect of 0.05 and 0.1 wt.% Al additions during the unsteady-state growth of the Sn-0.5wt.%Cu solder alloy. Various as-solidified specimens of each alloy were selected so that tensile tests could also be performed. Microstructural aspects such as the dimensions of primary, λ1, and secondary, λ2, dendritic arrays, and intermetallic compounds (IMCs) morphologies were comparatively assessed for the three tested compositions, that is, Sn-0.5wt.%Cu, Sn-0.5wt.%Cu-0.05wt.%Al, and Sn-0.5wt.%Cu-0.1wt.%Al alloys. Al addition affected neither the primary dendritic spacing nor the types of morphologies identified for the Cu6Sn5 IMC, which was found to be either globular or fibrous regardless of the alloy considered. Secondary dendrite arm spacing was found to be enlarged and the eutectic fraction was reduced with an increase in the Al-content. Tensile properties remained unaffected with the addition of Al, except for the improvement in ductility of up to 40% when compared to the Sn-0.5wt.%Cu alloy without Al trace. A smaller λ2 in size was demonstrated to be the prime microstructure parameter associated with the beneficial effect on the strength of the Sn-0.5wt.%Cu(-x)Al alloys.


Introduction
Solder alloys play a crucial role in the reliability of electronic systems.Therefore, these alloys must not only provide a set of properties to fulfill both the mechanical and electrical demands, but should also not increase the cost-effective production of electronic assemblies.Traditional Sn-Pb solder alloys have been widely used as they best meet the aforementioned requirements [1].However, due to the inherent toxicity of lead (Pb), these alloys have been increasingly banned worldwide [2].
Lead free solder alloys have been investigated with the view to replacing Pb containing solders.Among the several candidate alloy systems that have emerged as substitutes, the Sn-Cu system has received significant attention due to its low melting point and low cost [3][4][5].The demands for the improvement of the mechanical performance of the solder joints has generated the need to optimize the composition and manufacturing processing without compromising other soldering properties such as wettability, solderability, electrical and thermal conductivity.In this context, minor additions of In order to produce them, the alloys were directionally solidified (DS) by using a water-cooled setup.The utilized system allowed for unsteady-state solidification conditions to be attained, which means that the temperature varies with time and position across the casting.The solidification system was built in such a way that the alloy could be remelted in situ, being heated by radial electrical wiring around a cylindrical stainless steel crucible.The heating stage may be suppressed when the melt temperature achieves 10% above the alloy liquidus temperature.At the same time, the water inlet at the bottom of the crucible may be initiated.After, the variation of temperature is monitored and registered by a set of thermocouples laterally inserted in strategic positions across the length of the casting.Fine type J thermocouples (0.2 mm diameter wire) were used inside the molten material in different positions from the water-cooled surface of the casting.The calculation of the cooling rates ( Ṫ), growth rates (V), and thermal gradients (G) was possible thanks to the thermal data associated with the solidification stage.The ternary Sn-0.5wt.%Cu-0.1wt.%Alalloy was also solidified in a stepped copper mold using centrifugal casting under an Ar atmosphere.This alloy was melted in situ and the heating system was power controlled to achieve a desired melting temperature at which the heating system was turned off and the mold rotated, permitting the molten alloy to rapidly solidify by the mold surface.
The microstructures were examined by means of an Olympus inverted metallurgical microscope model 41GX (Olympus Co., Tokyo, Japan) and a scanning electron microscope (EDS/SEM) from Zeiss (Oberkochen, Germany, Auriga 40 model).
The revealed dendritic microstructures allowed for measurements of the primary dendritic spacing (λ 1 ) and the secondary dendrite arm spacing (λ 2 ) by virtue of the triangle and intercept methods, respectively [25] (Figure 1).The spacing between the Cu 6 Sn 5 globules and between the Cu 6 Sn 5 fibers were also experimentally characterized by identifying regions with a predominance of either globular (round) or fibrous shapes.
The instrument used to determine the Cu and Al contents on the longitudinal section of the DS casting was a Rigaku Rix 3100 x-ray fluorescence (XRF) spectrometer.The x-ray diffraction (XRD) patterns for indexing the formed phases were acquired by a Siemens D5000 diffractometer with a 2-theta range from 20 • to 90 • , CuKα radiation, and a wavelength, λ, of 0.15406 nm.Both devices are from Shimadzu (Kyoto, Japan).
The tensile strengths and elongations of the Sn-Cu-Al solder alloys were investigated using a MTS 810 machine according to the specifications of the ASTM Standard E 8M/04 at a strain rate of 2.5 × 10 −3 s −1 .The tensile samples were machined in a dog-bone shape with a gage length of 30 mm, a width of 4 mm, and thickness of 2 mm.showing the methods used to measure dendritic and eutectic spacing.The intercept method was applied for λ2, λf, and λ g, whereas the triangle method was applied for λ1.'L, Lf and Lg' are the length of lines and 'n' is the number of intercepted phases.

Solidification Thermal Parameters and Macrosegregation Profiles
The cooling curves acquired during directional solidification of the Sn-0.5wt.%Cu and Sn-0.5wt.%Cu-xwt.%Al(x = 0.05 and 0.1) alloys are shown in Figure 2, which were obtained from a set of thermocouples placed along the centerline of the casting along its length.Thermocouples may have the positions of their tips slightly changed during the assembling procedure into the mold.In this sense, the positions of the thermocouples took their reference from the metal/mold interface obtained from the post mortem longitudinal sections of the solidified ingot containing the thermocouple tip.For positions closer to the cooled bottom of the castings, a sharper drop in the temperature was observed.For other positions, the rate of cooling decreased progressively due to the increase in the thermal resistance associated with the increasing thickness of the solid layer from the heat-extracting base and due to the increase in the metal/mold interface thermal resistance in addition to the metal contraction dynamics along the solidification.showing the methods used to measure dendritic and eutectic spacing.The intercept method was applied for λ 2 , λ f , and λ g , whereas the triangle method was applied for λ 1 .'L, L f and L g ' are the length of lines and 'n' is the number of intercepted phases.

Solidification Thermal Parameters and Macrosegregation Profiles
The cooling curves acquired during directional solidification of the Sn-0.5wt.%Cu and Sn-0.5wt.%Cu-xwt.%Al(x = 0.05 and 0.1) alloys are shown in Figure 2, which were obtained from a set of thermocouples placed along the centerline of the casting along its length.Thermocouples may have the positions of their tips slightly changed during the assembling procedure into the mold.In this sense, the positions of the thermocouples took their reference from the metal/mold interface obtained from the post mortem longitudinal sections of the solidified ingot containing the thermocouple tip.For positions closer to the cooled bottom of the castings, a sharper drop in the temperature was observed.For other positions, the rate of cooling decreased progressively due to the increase in the thermal resistance associated with the increasing thickness of the solid layer from the heat-extracting base and due to the increase in the metal/mold interface thermal resistance in addition to the metal contraction dynamics along the solidification.Figure 3a shows the plot of position (P) from the metal/mold interface and the corresponding time (t) of the eutectic/liquidus fronts passing by each thermocouple, concerning the eutectic alloy Figure 3a shows the plot of position (P) from the metal/mold interface and the corresponding time (t) of the eutectic/liquidus fronts passing by each thermocouple, concerning the eutectic alloy and the ternary alloys, respectively.A function of type P = at b , with 'a' and 'b' being constants, was used to fit these experimental points.The derivative of this function with respect to time yielded another Metals 2019, 9, 241 6 of 19 function for the tip growth rate (V) (Figure 3b).It can be realized that the addition of 0.1 wt.%Al into the Sn-0.5wt.%Cualloy did not affect the growth rate, since the binary alloy and this ternary alloy presented the same (V) profile.On the other hand, the addition of 0.05wt.%Algenerated a lower (V) profile than the aforementioned one.The cooling rate ( Ṫ), as shown in Figure 3c, was determined by computing the time-derivative (dT/dt) of each cooling curve after the passage of the liquidus and eutectic isotherms.In contrast to the (V) profiles, ( Ṫ) ones did not follow the same behavior.This can be explained by the definition of the cooling rate, which is the product between the thermal gradient and growth rate, i.e., ( Ṫ) = G.V. Although they presented the same (V) profiles, the Sn-0.5wt.%Cu and Sn-0.5wt.%Cu-0.1wt.%Alalloys differed from each other by having distinct thermal gradients (G) (Figure 3d).and the ternary alloys, respectively.A function of type P = at b , with 'a' and 'b' being constants, was used to fit these experimental points.The derivative of this function with respect to time yielded another function for the tip growth rate (V) (Figure 3b).It can be realized that the addition of 0.1 wt.%Al into the Sn-0.5wt.%Cualloy did not affect the growth rate, since the binary alloy and this ternary alloy presented the same (V) profile.On the other hand, the addition of 0.05wt.%Algenerated a lower (V) profile than the aforementioned one.The cooling rate (Ṫ), as shown in Figure 3c, was determined by computing the time-derivative (dT/dt) of each cooling curve after the passage of the liquidus and eutectic isotherms.In contrast to the (V) profiles, (Ṫ) ones did not follow the same behavior.This can be explained by the definition of the cooling rate, which is the product between the thermal gradient and growth rate, i.e., (Ṫ) = G.V. Although they presented the same (V) profiles, the Sn-0.5wt.%Cu and Sn-0.5wt.%Cu-0.1wt.%Alalloys differed from each other by having distinct thermal gradients (G) (Figure 3d).extraction.No evidence of equiaxed grains was observed.Figure 5 shows the Cu and Al contents along the entire DS casting length, obtained by x-ray fluorescence (XRF).The average contents of Cu and Al were close to the nominal composition of the alloys.Although higher fluctuations of the Al content were observed in the Sn-0.5wt.%Cu-0.1wt.%Alalloy, it could not be considered a segregation since there was no tendency of increase or decrease in the Al profile.It is worth noting that Al will be mainly constricted in the Sn-rich primary phase since a very small solidification freezing range and complex dendritic arrangement may avoid this light element being transferred to the open liquid.As such, a prime factor appears to contribute to the observed fluctuations in the Al content demonstrated throughout the macrosegregation profiles.This is the thermal instability induced by the unsteady-state regime of heat flow extraction during the growth of the Sn-rich phase.The Al profiles are mainly a consequence of this unsteadiness, which is inherent in the process.Figure 4a-c show that columnar macrostructures prevailed along the entire casting length for all alloy compositions examined.The grains' growth direction was predominantly parallel to the heat extraction.No evidence of equiaxed grains was observed.Figure 5 shows the Cu and Al contents along the entire DS casting length, obtained by x-ray fluorescence (XRF).The average contents of Cu and Al were close to the nominal composition of the alloys.Although higher fluctuations of the Al content were observed in the Sn-0.5wt.%Cu-0.1wt.%Alalloy, it could not be considered a segregation since there was no tendency of increase or decrease in the Al profile.It is worth noting that Al will be mainly constricted in the Sn-rich primary phase since a very small solidification freezing range and complex dendritic arrangement may avoid this light element being transferred to the open liquid.As such, a prime factor appears to contribute to the observed fluctuations in the Al content demonstrated throughout the macrosegregation profiles.This is the thermal instability induced by the unsteadystate regime of heat flow extraction during the growth of the Sn-rich phase.The Al profiles are mainly a consequence of this unsteadiness, which is inherent in the process.

Dendritic Arrays and Their Evolutions along the Length of the Alloys Castings
The microstructures from optical microscopy, as demonstrated in Figures 6-8, were those associated with the samples collected along the length of the castings from the cooled bottom, referring to three different cooling rates for the analyzed alloys (1.0 °C/s, 0.45 °C/s, and 0.15 °C/s).

Dendritic Arrays and Their Evolutions along the Length of the Alloys Castings
The microstructures from optical microscopy, as demonstrated in Figures 6-8, were those associated with the samples collected along the length of the castings from the cooled bottom, referring to three different cooling rates for the analyzed alloys (1.0 • C/s, 0.45 • C/s, and 0.15 • C/s).These samples were selected to comparatively visualize the effect of cooling rates on the resulting microstructure.Dendritic growth could be observed for all Sn-Cu alloys composed of β-Sn rich dendrites with the eutectic mixture (β-Sn + Cu 6 Sn 5 ) forming the interdendritic regions [18].These phases were identified by the XRD analysis as shown in Figure 9a,b.CuAl IMCs could not be detected.This was in agreement with the results found by Li et al. [26], which demonstrated that such phases could only be clearly detected for Al-contents higher than 1000 ppm.
Regions closer to the cooled surface of the castings were submitted to higher cooling rates (represented by 1 • C/s samples in Figures 6-8).Thus, more refined microstructures could be formed, which become gradually coarser for the samples toward the top of the castings at a cooling rate of 0.15 • C/s (Figures 6-8).Comparing the microstructures characterizing the three examined alloys, it was noted that the primary dendritic spacing, λ 1 , for the same range of cooling rate, remained very close for all alloys, indicating that the addition of Al did not affect the growth of the primary dendritic stems.On the other hand, the addition of Al seemed to affect the stability at the solid/liquid interface from the primary branches and later increased the secondary dendritic spacing.Hence, the λ 2 tended to increase with the increase in Al content.
Table 2 shows the influence of the Al addition on the eutectic area fraction.The presence of Al appears to have inhibited the eutectic reaction, or, that is to say, to favor the growth of the β-Sn matrix.Such benefits in the growth of β-Sn are mainly translated through the coarsening of the secondary dendritic branches, as clearly seen in Figures 6-8.A study performed by Uddin and co-authors [27] described the effect of Al addition (1, 2, and 4 wt.%) on the microstructure of a Sn-8wt.%Zn-3wt.%alloy.The main microstructural change reported by the authors consisted of a decrease in the α-Zn volume fraction with the increasing addition of Al due to high reactivity between the Zn and Al elements.
It is widely accepted that the eutectic Cu 6 Sn 5 IMC size and fraction can be strongly affected by variations in β-Sn undercooling during alloy solidification [28].It appears that the presence of Al is able to decrease the eutectic area fraction (as shown in Table 2) since it induces a decrease in the β-Sn undercooling at the solidification front.As such, the faster growing β-Sn surrounds the tip of the solidifying eutectic Cu 6 Sn 5 , inhibiting its development.
Although there is a good susceptibility of the formation of CuAl x intermetallic compounds (IMCs) in Al-containing Sn-Cu alloys as mentioned in the literature [26], these IMCs can only be significantly detected at higher levels of Al (1000 ppm).The influence (or not) of Al on the formation of CuSn based IMCs will be discussed next.
undercooling at the solidification front.As such, the faster growing β-Sn surrounds the tip of the solidifying eutectic Cu6Sn5, inhibiting its development.
Although there is a good susceptibility of the formation of CuAlx intermetallic compounds (IMCs) in Al-containing Sn-Cu alloys as mentioned in the literature [26], these IMCs can only be significantly detected at higher levels of Al (1000 ppm).The influence (or not) of Al on the formation of CuSn based IMCs will be discussed next.X-ray diffraction (XRD) patterns of the (a) Sn-0.5wt.%Cu-0.05wt.%Aland (b) Sn-0.5wt.%Cu-0.1wt.%Alsolder alloys for specific positions along the length of the directional castings.P is the position from the metal/mold interface.

Cu 6 Sn 5 Morphology
The Sn-rich side of the Sn-Cu alloy system showed a eutectic reaction at 0.7 wt.%Cu and at 227 • C.This eutectic microstructure occurred between the Cu 6 Sn5 IMC and the Sn phase.Most studies have stated that rod-like Cu 6 Sn 5 particles must form [29]. Minor additions of Al in the Sn-Cu alloys are not expected to suppress or affect such reactions as those seen in Figure 10 by means of detailed SEM microstructures.The demonstrated microstructures refer to the interdendritic regions observed over high magnification.
Modification of the eutectic morphology is widely practiced in other metallic systems whose eutectic reaction predominates.This is the case of the hypoeutectic Al-Si alloys [30], which may result in the leading Si phase transitioning from a plate-like to a fibrous morphology.The control of the Si size and morphology in Al-Si alloys is well known to be essential to meet the progressively increasing demands for automotive and aerospace applications utilizing these alloys [31,32].
Trace concentrations of rare earth elements (that is, Ce and La) have been demonstrated to significantly alter the eutectic microstructure of the Sn-0.7wt.%Cualloy [33], providing a better distribution of finer particles within the as-solidified alloy.This suggests that microstructural control is possible by utilizing third elements.However, very little attention has been given to the modification of the Cu 6 Sn 5 morphology.
The present results showed that the Cu 6 Sn 5 phase developed mainly as a mixture of fibers and globules.These features could be observed even for the non-modified alloy.Some of the fibers appeared to be broken, as if they were perforated irregularly at different distances along their lengths.Such results bring new debate around the morphology of the Cu 6 Sn 5 phase.This means that a given morphology of Cu 6 Sn 5 does not depend on Al (at least for the tested Al-contents).It is associated, hence, with the solidification progress.The employed unsteady-state solidification conditions, typified in this study by cooling rates altering from 0.05 • C/s to 2.8 • C/s, induce thermal-driven perturbations, which may stimulate an irregular front of solidification [34].
If comparing the SEM microstructures on the left (i.e., fast cooling rate) with those on the right (i.e., slow cooling rate) in Figure 10, larger particle spacing as well as bigger particles can be observed to the right.This is largely due to the reduction of 1 order of magnitude in the cooling rate during the formation of the eutectic microstructure from the left to right in Figure 10.
EDS/SEM analysis of the alloys revealed that Al was concentrated in the Sn-rich phase, probably in solid solution as can be seen in Figure 11.The final distribution of the elements can be seen within the phases.Sn can be seen with high intensity within the Sn-rich matrix whereas Cu predominates, constituting the Cu 6 Sn 5 IMC.These results confirmed those previously demonstrated by utilizing XRD analysis where none of the Al-containing intermetallic was detected.EDS/SEM analysis of the alloys revealed that Al was concentrated in the Sn-rich phase, probably in solid solution as can be seen in Figure 11.The final distribution of the elements can be seen within the phases.Sn can be seen with high intensity within the Sn-rich matrix whereas Cu predominates, constituting the Cu6Sn5 IMC.These results confirmed those previously demonstrated by utilizing XRD analysis where none of the Al-containing intermetallic was detected.

Mechanical Properties vs. Secondary Dendrite Arm Spacing
The size of λ2 was shown to change significantly, as demonstrated in Figures 6-8.The increase in λ2 was shown to have a deleterious effect on the tensile strength as can be seen in Figure 12, despite having no considerable effect in the elongation of the three tested Sn-Cu alloys.The elongation values remained almost unaltered with λ2.Modified Hall-Petch-type correlations were adopted to represent some of the experimental scatters, which were the σu-ultimate tensile strength, and σy = 0.2-yield strength experimental variations with λ2.
Yang et al. [7,35] reported that the Al-modified Sn-0.7wt.%Cualloys may exhibit lower tensile strength and higher ductility when compared with the binary Sn-0.7wt.%Cualloy.Due to the utilization of a massive non-water-cooled mold, σu values changing from 10 MPa to 17 MPa were obtained in the cited study [7].In addition, the microstructures of all tested specimens were not standardized as accomplished at this point in the present investigation.Hence, a direct comparison was much more sustainable here, since specimens with similar microstructural parameters (i.e., λ2) were tested as plotted in Figure 12.In order to assess the finer secondary dendritic spacing in the

Mechanical Properties vs. Secondary Dendrite Arm Spacing
The size of λ 2 was shown to change significantly, as demonstrated in Figures 6-8.The increase in λ 2 was shown to have a deleterious effect on the tensile strength as can be seen in Figure 12, despite having no considerable effect in the elongation of the three tested Sn-Cu alloys.The elongation values remained almost unaltered with λ 2 .Modified Hall-Petch-type correlations were adopted to represent some of the experimental scatters, which were the σ u -ultimate tensile strength, and σ y = 0.2-yield strength experimental variations with λ 2 .
Yang et al. [7,35] reported that the Al-modified Sn-0.7wt.%Cualloys may exhibit lower tensile strength and higher ductility when compared with the binary Sn-0.7wt.%Cualloy.Due to the utilization of a massive non-water-cooled mold, σ u values changing from 10 MPa to 17 MPa were obtained in the cited study [7].In addition, the microstructures of all tested specimens were not standardized as accomplished at this point in the present investigation.Hence, a direct comparison was much more sustainable here, since specimens with similar microstructural parameters (i.e., λ 2 ) were tested as plotted in Figure 12.In order to assess the finer secondary dendritic spacing in the specimens for tensile tests, extra rapidly solidified (RS) castings were fabricated and their results plotted in Figure 12.
failure.A slightly larger elongation was observed in the Sn-0.5wt.%Cu-0.1wt.%Alsolder alloy, especially if restricted to λ2 −1/2 values lower than 0.220.The larger elongation of this alloy can probably be explained by the lower fraction of the brittle Cu6Sn5 IMC composition of the microstructure, compared to that related to the non-modified alloy.This is also reliable for samples related with higher cooling rates (i.e., λ2 −1/2 ~0.34)where an average δ value of about 44% was associated with the Sn-0.5wt.%Cu-0.1wt.%Alalloy, in comparison to the 37% obtained for the Sn-0.5wt.%Cualloy.Up to now, the methods utilized to examine the formation of microstructures in Al containing Sn-Cu alloys have not permitted the calculation and comparison of cooling rates as well the examination of equivalent samples in terms of λ2 in size.This means that the effects of cooling rate joined with the variations of alloy Al content remain largely unknown in past literature.In the present research, a methodology permitting the microstructure scaling laws of solders to be determined over a wide range of cooling rates has become available.The Al-containing alloys were characterized by having a lower eutectic fraction than the Sn-0.5wt.%Cualloy.It is possible to infer that this characteristic may be balanced with the solid solution strengthening induced by the Al.As a result, similar experimental evolutions were attained

Figure 1 .
Figure 1.Graphic representation of (a,c) transverse and (b) longitudinal sections of the DS castingshowing the methods used to measure dendritic and eutectic spacing.The intercept method was applied for λ2, λf, and λ g, whereas the triangle method was applied for λ1.'L, Lf and Lg' are the length of lines and 'n' is the number of intercepted phases.

Figure 1 .
Figure 1.Graphic representation of (a,c) transverse and (b) longitudinal sections of the DS castingshowing the methods used to measure dendritic and eutectic spacing.The intercept method was applied for λ 2 , λ f , and λ g , whereas the triangle method was applied for λ 1 .'L, L f and L g ' are the length of lines and 'n' is the number of intercepted phases.

Figure 3 .Figure 3 .
Figure 3. Experimental plots for the Sn-Cu-Al solder alloys corresponding with (a) position vs. time at which either the liquidus or the eutectic front reaches each thermocouple; (b) growth rate (V) as a Figure 3. Experimental plots for the Sn-Cu-Al solder alloys corresponding with (a) position vs. time at which either the liquidus or the eutectic front reaches each thermocouple; (b) growth rate (V) as a function of position; (c) cooling rate ( Ṫ) as a function of position (P); and (d) thermal gradient (G) as a function of position (P).R 2 is the coefficient of determination.

Figure
Figure4a-cshow that columnar macrostructures prevailed along the entire casting length for all alloy compositions examined.The grains' growth direction was predominantly parallel to the heat Metals 2019, 9, x FOR PEER REVIEW 7 of 19 function of position; (c) cooling rate (Ṫ) as a function of position (P); and (d) thermal gradient (G) as a function of position (P).R 2 is the coefficient of determination.

Figure 6 .
Figure 6.Representative optical microstructures for selected positions along the length of the Sn-0.5wt%Cusolder alloy casting.Additionally, these include the related solidification thermal parameters (Ṫ and v) and primary, λ1, secondary, λ2, and dendrite arm spacing.

Figure 6 .
Figure 6.Representative optical microstructures for selected positions along the length of the Sn-0.5wt%Cusolder alloy casting.Additionally, these include the related solidification thermal parameters ( Ṫ and v) and primary, λ 1 , secondary, λ 2 , and dendrite arm spacing.

Table 1 .
Chemical composition (wt.%) of the metals used to prepare the alloys.

Table 2 .
Eutectic area fractions measured along the length of the DS Sn-Cu-xAl alloy castings.

Table 2 .
Eutectic area fractions measured along the length of the DS Sn-Cu-xAl alloy castings.