Microstructure-property relations in as-atomized and as-extruded Sn-Cu (-Ag) solder alloys

Rapidly solidi ﬁ ed Sn-based solder alloys can provide metallurgical features such as low segregation and ﬁ ne intermetallic compounds (IMCs). These features can be obtained in a controlled way by Impulse Atomization that provides powders of various sizes corresponding to a variety of cooling rates and undercoolings to which diverse microstructures are associated. In the present investigation, rapid so-lidi ﬁ cation of Sn-0.7 wt%Cu and Sn-0.7 wt%Cu-3.0 wt%Ag alloys have been examined through the production of a wide size range of impulse atomized powders. The microstructures and hardness resulting from the generated powders have been compared with those of directionally solidi ﬁ ed (DS) specimens. Regular cells > dendrites and reverse dendrites > cells transitions were identi ﬁ ed, and high-speed eutectic cells were found to prevail for the examined Sn-0.7 wt%Cu powders of size smaller than 300 m m. It is shown that the Vickers microhardness of the ternary Sn-3.0 wt%Ag-0.7 wt%Cu alloy is directly in ﬂ uenced by both the presence of tertiary dendrite arms ( l 3 ) and the cooling rate/powder size dependent eutectic fraction. Also, compaction and extrusion of the atomized powders were carried out in order to consolidate the samples so that tensile tests could be carried out. Tensile strength and ductility of samples corresponding to different powder sizes and compositions were thus measured and the re-sults are found to be consistent with their microstructures.


Introduction
Atomization techniques provide experimental access to very high undercooling providing opportunity to investigate the fundamentals of nucleation and growth in highly driven metallic systems. A wide range of cooling rates (Ṫ L ) can be obtained due to the typically produced wide range of droplet sizes [1]. Impulse atomization (IA) is a single-fluid atomization technique where molten metal is pushed through an orifice to form a ligament that is transferred into a chamber with a stagnant inert atmosphere and break up into spherical droplets before rapidly losing heat to the gas. The atomized droplets are accelerated by gravity while rapidly solidifying into powders that are collected at the bottom of the chamber. Due to its controlled gas-droplet heat transfer, IA offers conditions to generate reliable microstructural-spacing vs. coolingrate relationship for metallic alloys [2].
According to Plookphol et al. [3] new processing routes may be required for the production of the SAC305 lead-free solder powders in order to improve their properties and consequently make them industrially more attractive. Centrifugal atomization is stressed as one of the alternatives by Plookphol et al. [3]. The Sn-3 wt%Ag-0.7 wt%Cu (SAC307) and Sn-0.7 wt%Cu solder alloys are among the Sn-based solder family which has been used instead of conventional Sn-Pb solder alloy. These lead-free alloys provide compatible or even superior mechanical properties than those of the traditional Pb-Sn [4]. Recent legislations and regulations in many countries (RoSH e Restriction of Hazardous Substances Directive, WEEE e Waste Electrical and Electronic Equipment) have led to an increased use of the above mentioned alloys [5].
A number of studies devoted to derive microstructure-cooling rate type inter-relations during atomization can be noted in literature. Most of those investigations have been dedicated for Albased and Fe-based alloys [1,2]. However, the effects of cooling rate on the as-atomized microstructures and sizes in Sn-Cu (-Ag) alloys remain undetermined.
Recent literature [5e7] indicates that one way of improving the properties of the solder is the judicious selection and innovation of the processing techniques. Impulse atomization, a rapid solidification technique, is one of the alternative techniques to produce high performance materials with refined microstructures. Indeed, it is well known that rapid solidification of metallic alloys results in refined microstructures with reduced microsegregation and improved mechanical properties of the final products as compared to traditional as-cast components [8,9]. Typical cooling rates during reflow procedures for solders in industrial practice remain in the range of 3e10 K/s [10].
Rapidly-solidified lead-free solders are assumed to propitiate lower melting point, narrower melting interval, better wettability and improved mechanical performance because of the fine, uniform and low segregation microstructure compared with the ascast lead-free solders [11]. Features such as high melting rate, low soldering temperature and short time are highly desirable concerning real processing conditions, even though rapidly-solidified microstructure will be affected during re-solidification step within the solder joint.
The cooling rate during liquid-to-solid transformation is considered a key factor in the formation of b-Sn dendrites and Ag 3 Sn plates in Sn-Ag-Cu alloys. If the cooling rate is high, the Ag 3 Sn plates do not have sufficient time to achieve significant growth even if they had nucleated. Large plates may not be commonly seen in as-received solder balls, or in rapidly cooled solder joints (at a rate of 1 K/s or higher) [12]. For example, in the case of the directionally solidified Sn-2.0 wt%Ag alloy [13], it has been stated that the range of cooling rates from 0.15 to 1.15 K/s is associated with microstructures having a mixture of spheroid and fiber-like IMCs while at higher cooling rates (>8.0 K/s) only Ag 3 Sn spheroids develop. The as-solidified microstructures and their morphologies are well-known to be governed by the temperature gradient (G L ), the growth rate (V L ) and the cooling rate ( _ T ¼ G L Â V L ). Several studies have been carried out during recent years [14e18] with a view to determine experimental interrelations of microstructure-cooling rate for lead-free Sn-based alloys. Some of these studies regarding some metallic systems of interest have established that planar front > cells > dendrites transitions occur with increasing V L [14,16]. Further increase in V L changes the dendritic front back to cellular and planar front, the latter associated with the limit of absolute stability [19,20].
Experimental studies on this reverse transition from dendrites to cells are rare in the literature for metallic systems. Trivedi et al. [21] examined the conditions under which the high-velocity morphological transition occurs during the steady-state growth of carbon-tetrabromide. Fu et al. [22] investigated the effect of cooling rate in as-cast samples of AISI 304 stainless steel, and reported the transition from dendritic austenite to cellular austenite at high cooling rates.
The aims of this study is primarily to determine the evolution of microstructure (morphology and scale of the b-Sn matrix and fraction of intermetallics) during impulse atomization of the Sn-0.7 wt%Cu and Sn-3.0 wt%Ag-0.7 wt%Cu alloys considering a wide range of powder sizes. Secondly, mechanical strength, ductility, porosity and wetting angle of extruded powders will be examined in order to compare not only the effect of alloy composition but also these experimental data with available data obtained from slow cooling conditions. The effects of powder size and extrusion temperature on the final microstructures will also be assessed.

Impulse atomization
The Sn-0.7 wt%Cu and Sn-3.0 wt%Ag-0.7 wt%Cu alloys were  prepared at the Advance Materials Processing Laboratory (AMPL) of the University of Alberta. In both cases commercially pure elements (Sn, Ag and Cu shots) were mixed in a clay-graphite crucible and heated to 1200 C by an induction furnace. The molten alloy was stabilized for 30 min, which allowed a complete mixing of the elements. After that, the melt was cooled to 500 C before being rapidly solidified (RS) by IA, into spherical powders. These powders were then collected in an oil filled beaker before being washed, dried and sieved into different sizes. For these IA experiments, 50 mm nozzles size were used and the sizes of powders obtained range from 125 mm to 1400 mm. For this paper, the following size ranges were examined: i. 125e150 mm, 180e212 mm and 250e300 mm for the Sn-0.7 wt%Cu and ii. 106e180 mm, 212e250 mm, 250e300 mm and 425e500 mm for the Sn-3.0 wt%Ag-0.7 wt%Cu alloy. Vickers hardness tests were performed on the powders by using a test load of 50 g and a dwell time of 15 s. The Vickers hardness of each powder size associated with each alloy was the average of at least 10 measurements. Also, for each powder size a counting of the eutectic area fraction was performed using an image processing software (Image J). The triangle and intercept methods proposed by Gündüz and Çadirli [23] were used in order to quantify either cell spacing/primary (l 1 ) dendritic spacing (l c ) -triangle, or secondary (l 2 ) dendritic spacing e intercept. The images were acquired and analyzed in an optical image processing system Olympus, GX51 (Olympus Co., Japan). The average microstructural spacing associated with each droplet size was determined through at least 20 measurements.
Details concerning the experimental data obtained by directional solidification of both Sn-0.7 wt%Cu and Sn-3.0 wt%Ag-0.7 wt %Cu can be found in previous studies [24,25].

Compaction and extrusion of the powders
Single punch compaction of the droplets was performed at room temperature under normal atmosphere. Powders (~6.5 g) of known size range were introduced in a hollow cylindrical geometry and sandwiched by two T-shape cylinders. The assembly was then placed under 2000 Kg load. The compacted samples were extruded at room temperature as well as at 150 C. Two size ranges were chosen for each alloy, which were: 125e150 mm/250e300 mm for the Sn-0.7 wt%Cu alloy and 106e180 mm/250e300 mm for the Sn-3.0 wt%Ag-0.7 wt%Cu alloy. Hence, a direct comparison between the results of the two examined alloys is possible.
Extruded samples from Sn-0.7 wt%Cu and Sn-3.0 wt%Ag-0.7 wt% Cu powder alloys were fabricated with an extrusion rate of 5 mm/ min through a die. A rod of around 150 mm long and of 2.5 mm diameter was obtained due to a reduction ratio of 10:1 in the cross section of the pressed preform. At every processing stage, specimens from both alloys were polished and etched (solution of 92% (vol) CH 3 OH, 5% (vol) HNO 3 and 3% (vol) HCl) for metallography. Microstructural characterization was performed using a Field Emission Gun-Scanning Electron Microscope (SEM) Philips (XL30 FEG).
Tensile testing on the Sn-Cu and Sn-Ag-Cu as-extruded samples was performed according to specifications of ASTM Standard E 8M/ 04 and tested in an Instron 5500R machine at a strain rate of about 3 Â 10 À4 s À1 . In order to compare wetting behavior, cylinder bars (2.5 mm-height Â 2.5 mm-diameter) were obtained from the DS and the as-extruded Sn-3.0 wt%Ag-0.7 wt%Cu alloy samples. The samples were cleaned, dried and finally coated with adequate flux for testing. Cu (copper) substrates with the same level of surface roughness were used for the wetting tests. The measurements of the contact angles (q) were carried out in a Goniometer Krüss DSHAT HTM Reetz GmbH model from the average of q R and q L (Rright and L-left) values provided by a computational method (tangent-2). A desired purging gas atmosphere was maintained by passing argon through the furnace. Three specimens of the Sn-Ag-Cu alloy were placed on the tester, heated and melted individually. A standard thermal cycle was imposed and a contact angle (q a ) was determined as the average of continuously measured contact angles as the form of the molten droplet was changing till an equilibrium regime was reached. l ¼ AṪ Àn established between the microstructural spacing and the solidification cooling rate (Ṫ) for the DS samples [24,25]. A and n are the alloy dependent parameters, as defined in the mentioned references [24,25].  SAC307 alloy generally presents a dendritic Sn-rich matrix surrounded by a eutectic mixture, where Ag 3 Sn and Cu 6 Sn 5 particles arrange a ternary eutectic mixture. On the other hand, the cellular/ dendritic array in the Sn-0.7 wt%Cu alloy is constituted by a solid solution of Cu in Sn, with a surrounding eutectic mixture formed by Cu 6 Sn 5 intermetallic compound (IMC) randomly distributed in the Sn-rich phase. By comparing the resultant extrapolated cooling rate values as can be seen in Fig. 1a and b by the open star points in the graphs, ranges from 76 to 248 K/s and from 25 K/s to 123 K/s for the Sn-3.0 wt%Ag-0.7 wt%Cu and Sn-0.7 wt%Cu alloys were attained, respectively. The presence of silver (Ag) in the ternary alloy composition seems to cause higher cooling rates to be obtained during IA since Cu chemistry in both alloys is exactly the same. Also, atomization conditions (nozzle size, gas and melt overheating level) were maintained the same in both cases. Silver (Ag) content seems to be able to change intrinsic thermal properties of the alloy leading to increasing on heat transfer efficiency during solidification of the atomized powders.

Results and discussions
The measured l 1 , l c (Sn-0.7 wt%Cu) and l 2 (Sn-3.0 wt%Ag-0.7 wt %Cu) for all examined powder size ranges were inserted (open stars). The related cooling rate values were obtained through the continuation of the experimental cooling curves obtained for the aforementioned DS experiments [24,25]. A regular transition from dendrites to cells was observed in a previous article during directional solidification of the Sn-0.7 wt%Cu alloy [24]. Considering the morphology and scale of the rapidly solidified b-Sn matrix, cells with lower cell spacing than those found in DS specimens can be seen. According to Kurz and Fisher [26] the solidification interface morphologies can vary from planar at very low growth rates to cells and to dendrites which become finer and finer until they give rise to cellular-type structures when close to the limit of absolute stability. The cell growth observed in this study is a typical rapid solidification microstructure. Due to the higher cooling rates experienced by the Sn-0.7 wt%Cu powders during IA cooling, high-cooling eutectic cells seems to prevail. A typical micrograph showing this morphology can be seen in Fig. 2a. Also, a representative microstructure of a coarser Sn-Cu powder with size higher than 1000 mm was inserted in Fig. 1a. In this case, b-Sn cells and dendrites (black arrows) were found to grow concurrently, which characterizes a microstructural transition with no prevalence of a single morphology. These powder particles may be associated with lower cooling rate values during solidification, which may be considered closer to the highest values obtained for DS. It seems that a reverse dendrites > cells transition in Sn-0.7 wt%Cu alloy may start to happen for cooling rates higher than 12.0 K/s. Previous studies regarding the determination of these features and of the reverse transition from dendrites > cells for Sn-Cu alloys have not been found in literature. Thus, it is the first study which emphasizes such aspects in Sn-Cu alloys. A recent investigation by Brito et al. [27] for the directionally solidified Al-3.0 wt%Mg-1.0 wt%Si alloy found that a high-cooling rate cellular region is related to cooling rate higher than 2.0 K/s, followed by a dendritic region associated with cooling rate lower than 0.8 K/s. Figs. 1b and 2b show that all examined conditions (DS and IA) allowed a fully dendritic pattern to be developed in the SAC307 alloy. Typical micrographs associated with the different levels of cooling rates (fast and slow cooling conditions) have been inserted in the graph of Fig. 1b with a view to   Fig. 3 shows the variation of hardness with microstructural length scales. As can be seen, Vickers hardness increases with increasing microstructural spacing of the as-atomized Sn-0.7 wt% Cu-3.0 wt%Ag alloys. However, the values of hardness remain practically unaltered for the Sn-0.7 wt%Cu alloy. The experimental fittings in Fig. 3 represent the hardness evolutions obtained for distinct powder size ranges, with each size range corresponding to a mean microstructural spacing. In a previous study emphasizing the characterization of the directionally solidified SAC307 alloy [25,28], hardness is shown to be positively affected by the presence of tertiary dendritic branches. Hardness was found to vary from 13.1 to 14.2 HV, being these values associated with l 2 range from 52.0 to 9.0 mm. If considered the rapidly solidified IA powder, the l 2 range was 3.7e5.5 mm with hardness experimental range from 14.8 to 16.7HV. Higher HV values of the rapidly-solidified samples were due to the decrease on l 2 value. As can be noted in Figs. 1b and 2b, the dendritic arrangement characterizing the coarser SAC307 powder (see Fig. 2b: Ṫ L ¼ 76 K/s and size of 425e500 mm) depicts more extensive presence of tertiary branches than that shown in Fig. 2b (Ṫ L ¼ 172 K/s and size of 212e250 mm). Despite the presence of coarser l 2 for the coarser powder size, the growth of very fine tertiary arms seems to be able to highly distribute the Ag 3 Sn and Cu 6 Sn 5 intermetallic particles, which are well known as reinforcement particles to improve strength. As a consequence, higher hardness may be obtained with increasing powder size.

Microstructural length scales and associated hardness
Under equilibrium conditions, the solid solubility of Cu and Ag in Sn is very limited, with values of 0.04 wt%Ag and 0.0063 wt% for Cu at the eutectic temperature. Nevertheless, fast cooling during impulse atomization may promote the extension of solid solubility of these elements for higher values as stated by Snugovsky and collaborators [29] for non-equilibrium, supersaturated solid solutions. As a consequence, a decrease in the eutectic fraction is expected to occur with the increase on cooling rate. So, in order to check the consequences of supersaturation in the present SAC307 powder samples, the average area fraction occupied by the eutectic mixture was measured and the results obtained were: 20.4%, 24% and 29.4% for the following size ranges: 106e180 mm, 250e300 mm and 425e500 mm, respectively. The higher proportion of eutectic is probably another factor affecting hardness since such constituent is formed by a mixture of the hard Ag 3 Sn and Cu 6 Sn 5 intermetallic particles and the ductile b-Sn phase. The eutectic is located in the interdendritic regions. The same features were reported by Anderson et al. [30], which examined the phases forming the microstructures of the Sn-3.5Ag-0.95Cu (wt.%). According to the literature and Thermo-Calc computations [17], the eutectic volumetric fraction for the SAC307 alloy is 61% under equilibrium conditions.
In the case of the Sn0.7 wt%Cu, hardness of the IA powders may be not affected by the increase in the length scale of the microstructure. The same behavior as a function of the microstructure was reported for tensile strength of this alloy, which was maintained essentially constant, even for the case of regions with more refined structural arrangements [24].

Strength and ductility in relation with microstructures
Typical stress-strain curves are plotted considering tests performed at high temperature (150 C) and at room temperature for two distinct powder size ranges used for extrusion of both examined alloys (Fig. 4). In both examined alloys, the cold samples exhibited low ductility values with around 7% corresponding to the total elongation of the cold extruded Sn-0.7 wt%Cu powder (125e150 mm) and around 5% in the case of both cold extruded samples with the SAC307 alloy (both 106e180 mm and 250e300 mm). This is probably due to the presence of higher level of porosity and fissures which were observed along the transverse sections ( Fig. 5a and b), which anticipated fracture. Porosity has been counted based on optical images using image processing software (Image J). Some typical transverse optical images for the SAC307 alloy can be seen in Fig. 5. Large cracks and aligned pores refer to the cold extruded samples either for fine or coarse powders. On the other hand, a small number of round pores can be seen in the hot extruded samples (Fig. 5c and d). The measured area fractions of irregular porosity are about 1.90 ± 0.6% and 1.10 ± 0.4% for the cold extruded powders with 106e180 mm and 250e300 mm, respectively. In contrast, the determination of spherical-like porosity fractions resulted in 0.22 ± 0.06% and 0.52 ± 0.18% for the hot extruded powders with 106e180 mm and 250e300 mm, respectively.
As discussed in a previous article [31], finer grains seem to be responsible for the higher strength observed for the extruded Sn-Cu samples at room temperature as can be seen in Fig. 6a. The same characteristic can be observed in the microstructures associated with the cold extruded SAC307 alloy powders, which present finer grains than those typically found for the hot extruded alloy samples. Microstructural details can be observed through the SEM images in Fig. 6. A more balanced set of tensile parameters (ultimate tensile strength e s u and elongation to fracture e d) for the Sn-0.7 wt%Cu alloy can be seen by the solid lines in Fig. 4a. These curves refer to the samples subjected to extrusion under high temperature.
Larger and more homogeneously-distributed grains can be seen in hot extruded SAC307 alloy samples which allowed higher ductility to be achieved (see Fig. 6).
If the coarser powder size range (250e300 mm) of the SAC307 alloy is considered, increase up to 48% and 125% of the s u and d values were obtained when compared with those values for the equivalent Sn-0.7 wt%Cu extruded sample. In general, the plastic deformation amounts were highly increased in the case of the hotextruded samples.
White and black arrows in Fig. 7 reveal the presence and characteristics of Ag 3 Sn and Cu 6 Sn 5 IMCs, respectively, within the SAC307 solder alloy. A general precipitation of very refined and spheroidal-like Ag 3 Sn IMCs can be observed. It seems that a large amount of Ag 3 Sn spheroids has been developed for the samples processed at 150 C with coarser powder size. It is well known that spheroids of reduced size may be more homogeneously distributed along the microstructure and as a consequence the resulting mechanical strength may be improved. The combined presence of very fine Ag 3 Sn spheroids, low porosity and homogeneous distribution of b-Sn grains seems to be the reason why the hot extruded SAC307 alloy powders with 250e300 mm in size achieved sounder properties.
The Sn-0.7 wt%Cu-3.0 wt%Ag solder alloy tested in a previous study [17] through unsteady-state directional solidification exhibited s u values between 30 and 35 MPa and d values from 12 to 14%. Therefore, the production of powders followed by pressing and extrusion allowed more reliable tensile properties to be developed especially considering the gains in ductility. Fig. 8 shows the experimental evolutions of the contact angles for the SAC307 alloy against copper (Cu) substrate. For any of the examined conditions, i.e., as-extruded and as-cast, higher contact angles can be noted for the first measurements, representing the very initial contact conditions. Same experimental fluctuations can also be observed. The higher contact angles associated with the first stage of measurements are due to the fact that the melt was not able to spread. Fluctuations are due to convection currents within the alloy drop. A final stage indicates a period of certain stability for times higher than 300 s, which means that the angle reached a steady-state regime, resulting in an energetic equilibrium with the surface of the substrate.

Wettability of the investigated samples
The as-extruded SAC307 samples presented slightly lower average contact angles than those found for the directionally solidified alloy sample, achieving relatively close q a values between each other. According to the literature [32] the uniform chemical composition of the rapidly-solidified Sn-0.7 wt%Cu alloy may improve its wettability if compared with as-cast alloy (slow cooling conditions). Furthermore, the freezing range of the rapidlysolidified Sn-0.7 wt%Cu alloy tends to be narrow, which would help to uniform melt. As a consequence, more effective spreading is expected to occur when liquidus temperature is reached.

Conclusions
From the results obtained in the present investigation, the following conclusions can be drawn: -The as-atomized microstructures of the Sn-0.7 wt%Cu and Sn-3.0 wt%Ag-0.7 wt%cu alloys are basically arranged by highvelocity cells and Sn-rich dendrites, respectively. In the case of the Sn-0.7 wt%Cu alloy, a reverse microstructural transition from dendrites-to-cells has been noticed taking into consideration the prevalence of dendrites growing under slow cooling conditions (DS specimens), the cellular growth for rapidlysolidified powders and the presence of both cells and dendrites under intermediate cooling conditions for atomized powders higher than 1000 mm in size; -The extrapolations of the experimental cellular and dendritic growth laws allowed estimating the ranges from 76 to 248 K/s and from 25 K/s to 123 K/s for the rapidly-solidified Sn-3.0 wt% Ag-0.7 wt%Cu and Sn-0.7 wt%Cu alloys, receptively. Silver (Ag) addition resulted in increase of cooling rate during impulse atomization (IA). In the case of the Sn0.7 wt%Cu, hardness of the IA powders may be not affected by the increase in the length scale of the microstructure. However, hardness has increased with increasing powder size in the Sn-3.0 wt%Ag-0.7 wt%Cu alloy. Arrangement of an extensive distribution of hard Ag 3 Sn and Cu 6 Sn 5 intermetallic particles due to the growth of very fine dendrite tertiary arms and higher fraction of eutectic in coarser powders may justify the hardness evolution in this alloy; -The combined presence of very fine Ag 3 Sn spheroids, low porosity and homogeneous distribution of b-Sn grains seems to be the reason why the hot extruded SAC307 alloy powders with 250e300 mm in size achieved sounder properties like mechanical strength and ductility. Other than, high level of porosity and cracks within the cold extruded samples of both examined alloys caused low elongation-to-fracture values to be attained; -A more optimized balance of properties has been observed for the hot extruded powders with 250e300 mm (150 C). The following values were obtained: strength of 40 MPa, 62% of elongation-to-fracture, porosity proportion of 0.52% and average contact angle of 29.1 .