- Email: [email protected]
Cu joints
Materials Science & Engineering A 571 (2013) 184–192
Contents lists available at SciVerse ScienceDirect
Materials Science & Engineering A journal homepage: www.elsevier.com/locate/msea
The influence of silver content on structure and properties of Sn–Bi–Ag solder and Cu/solder/Cu joints a ˇ ˇ P. Sebo , P. Svec Sr. b,c,n, D. Janicˇkovicˇ b, E. Illekova´ b, M. Zema´nkova´ a, Yu. Plevachuk d, V. Sidorov e, b ˇ P. Svec a
Institute of Materials and Machine Mechanics, Slovak Academy of Sciences, Racˇianska 75, 831 02 Bratislava 3, Slovakia ´ cesta 9, 845 11 Bratislava 45, Slovakia ´ bravska Institute of Physics, Slovak Academy of Sciences, Du c Faculty of Materials Science and Technology, Slovak University of Technology, J. Bottu 25, 917 24 Trnava, Slovakia d Ivan Franko National University, Department of Metal Physics, 79005 Lviv, Ukraine e Ural State Pedagogical University, Cosmonavtov 26, 620017 Ekaterinburg, Russia b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 30 October 2012 Received in revised form 3 February 2013 Accepted 8 February 2013 Available online 14 February 2013
The effect of silver content on structure and properties of Sn100 xBi10Agx (x ¼ 3–10 at%) lead-free solder and Cu–solder–Cu joints was investigated. The microstructure of the solder in both bulk and rapidly solidified ribbon forms was analyzed by scanning electron microscopy (SEM) and X-ray diffraction. The peculiarities in melting kinetic, studied by differential scanning calorimetry (DSC), and silver influence on it are described and discussed. The wetting of a copper substrate was examined by the sessile drop method in the temperature range of 553–673 K in air and deoxidizing gas (N2 þ 10%H2) at atmospheric pressure. Cu–solder–Cu joints were also prepared in both atmospheres, and their shear strength was measured by the push-off method. The produced solders consisted of tin, bismuth and Ag3Sn phases. The product of the interaction between the solder and the copper substrate consists of two phases: Cu3Sn, which is adjacent to the substrate, and a Cu6Sn5 phase. The wetting angle in air increased slightly as the silver concentration in the solder increased. Wetting of the copper substrate in N2 þ10H2 gas shows the opposite tendency: the wetting angle slightly decreased as the silver content in the solder increased. The shear strength of the joints prepared in air (using flux) tends to decrease with increasing production temperature and increasing silver content in the solder. The equivalent decrease in the shear strength of the joints prepared in N2 þ 10H2 is more apparent. & 2013 Elsevier B.V. All rights reserved.
Keywords: Lead-free solder Cu joints Melting Wetting Shear strength X-ray diffraction
1. Introduction Sn–Pb solders were the most used joining materials in the electronic industry for many years because of their low cost and superior properties. Recently, the regulation of certain hazardous substances (RoHS) and waste electrical and electronic equipment (WEEE) has resulted in extensive research on lead-free solders. Nowadays SnAgCu (SAC) alloys are considered to be the best substitution for lead containing solders with Pb content up to 85 wt% [1,2]. However, for high melting (tl 4 503 K) solders, no proper equivalent is available at the moment. There are many potential candidates. Majority of these systems were studied in the frame of COST MP0602 Action. Many of the studied alloys were based on Sn (Sn–Sb, Sn–Au, Sn–Zn–Ni), ternary alloys e.g. Al–Zn–X (X ¼Ga, Ge, Mg, Ni, Sb and Sn), Al–Sb–Zn, Ag–Sb–Sn and n Corresponding author at: Institute of Physics, Slovak Academy of Sciences, Du´bravska´ cesta 9, 845 11 Bratislava 45, Slovakia. Tel.: þ 421 2 5941 0561; fax: þ 421 2 5477 6085. ˇ E-mail address: [email protected] (P. Svec Sr.).
0921-5093/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.msea.2013.02.013
many others. One potential candidate is also Sn–Bi–Ag system in tin rich corner. The phase diagram for ternary Sn–Bi–Ag alloys was comprehensively investigated both experimentally and numerically in [3–5]. DSC curves for Sn–Bi–Ag solders obtained at heating show endothermal peak in the interval 409–413 K. Some authors associate this with melting of Sn/Bi eutectic, others—with SnþBiþ Ag3Sn-LþAg3Sn reaction [3–5]. As a rule, the fraction of liquid phase at t4413 K is rather low (not exceeding 1% for 8 wt% of Bi content [5]). This phenomenon, however, causes some restrictions in usage of Sn–Bi–Ag alloys in the step soldering technology, in which soldering occurs more than once during the manufacturing process but for soldering of wires and printed circuit boards it is acceptable. The microstructure of Sn–Bi–Ag solder was found to consist of Ag3Sn particles, Sn-rich and Bi-rich phases in different combinations, depending on the cooling rate after reflow [6]. After aging, Bi in SAC–Bi and SnAg–Bi alloys was found as precipitates at grain boundaries and grain interiors [7]. The solid solubility limit of Bi in the Sn–Ag-based alloy is about 4 wt% at room temperature
ˇ P. Sebo et al. / Materials Science & Engineering A 571 (2013) 184–192
[3–5], and with increased Bi addition, the supersaturated Bi would precipitate in the form of pure Bi phase from the Sn matrix. Additions of Bi tend to suppress the formation of large Ag3Sn particles from the solder matrix [8]. No ternary compounds were found in this system [3]. In the Sn–Bi–Ag/Cu solder systems Cu reacts only with Sn, not with Ag or Bi [9]. Because of the formation of the Cu6Sn5 intermetallic compound (IMC), Ag and Bi are expelled from the reaction front into the molten solder between the IMC scallops. The existence of Bi and Ag in the gaps prevents the Cu–Sn grains from joining together and forming a continuous layer. The higher the Bi and Ag concentration in the solder, the more Bi and Ag is formed in the gaps. Bi and Ag additions affect the total thickness of interfacial reaction layers [9]. Several authors have examined wetting and mechanical properties of SnBiAg solders and joints with various concentrations of bismuth. Vianco and Rejent [10] showed that additions of 1–10 wt% Bi into the Sn–3.5 wt% Ag solder improve its wetting performance on Cu. Increasing the Bi content in the solder raised the Cu–solder–Cu joint shear strength to 81 MPa, as determined by the ring-and-plug test [10]. The microhardness reached maximum values of 30 (Knoop, 50 g) and 110 (Knoop, 5 g) for Bi contents greater than 4–5 wt%. He and Acoff [9] investigated the effect of reflow and the thermal aging process on microhardness of Sn–3.7 wt% Ag solder with 0, 1, 2, 3, and 4 wt% Bi. They proved that the microhardness increases with increasing Bi in the solder, regardless of reflow or thermal aging process. Huang and Wang showed that bismuth additions linearly increase the ultimate tensile strength of Sn–Ag lead-free solders [8]. The mechanical properties and the microstructure for the Sn–3 wt% Ag solders with 0, 3, and 6 wt% of Bi, connecting two Cu substrates, were experimentally examined in [9]. It was stated that for the solder between the Cu substrates, the strength increases from 90 to 140 MPa with increasing concentration of Bi from 0 to 6 wt%. The mechanical properties of bulk Pb-free solder alloys containing Bi have been characterized for the as-cast and aged samples in [7,11,12]. Room and elevated-temperature tensile testing showed that the addition of Bi greatly reduced the loss in strength due to aging that occurs in the Sn–Ag–Cu ternary alloys. In SnAg–Bi the tensile strength increased after aging. Hence, it is a common place in all the works that bismuth additions into Sn–Ag or SAC solders improve its mechanical properties and wetting ability on Cu substrate. It is a well known fact that by adding silver to tin one improves its reliability and wetting ability; however, nobody studied precisely the influence of silver content on the structure and properties of Sn–Bi–Ag solder itself and its interaction with substrates. Thus the aim of this work is to determine the effect of silver on (a) melting kinetics, (b) wetting properties and (c) shear strength of Sn100 xBi10Agx alloys and Cu/solder/Cu joints as well as to connect these properties with its microstructure.
2. Experimental A series of lead-free Sn100 xBi10Agx solders with different Ag content (Table 1) was prepared by melting appropriate amounts of the relevant metals, with purity 99.99% and higher, in an induction furnace in an overpressure of argon at approximately 870 K. Their chemical composition was determined using an atomic emission spectrometer Spectrum Flame Modula S. Solders were prepared in two forms: bulk and ribbon. The bulk solder was used for DSC and for wetting experiments. Solders in ribbon form were prepared by rapid quenching of the melt and used for the preparation of copper–solder–copper joints and DSC measurements. The microstructure of the solder in both forms was
185
Table 1 Composition of Sn100 xBi10Agx solders (at%). Solder
Sn
Bi
Ag
B1 B2 B3 B4 B5
86.7 85.3 83.3 81.7 80.0
10.0 10.0 10.0 10.0 10.0
3.3 4.7 6.7 8.3 10.0
investigated by SEM equipped with an energy dispersive analyzer. SEM used was JEOL 7600F with a Schottky FEG emitter and EDX from Oxford Instruments X-max with a 50 mm2 detector. Software used for control of EDX acquisition and semiquantitative analysis was Oxford Instruments INCA version 4.15. The Perkin–Elmer DSC7 differential scanning calorimeter was calibrated for a heating rate of 710 K/min. The same aluminum sample pan was always loosely covered with the lid. A reference pan containing no sample and a flowing (20 ml/min) argon atmosphere were used. The sample mass was 10 mg. Two independent samples were always used; thus, the reproducibility of each effect has been proven and the accuracy of measurement of each quantity has been calculated to be 70.5 K for temperature and 72 J/g for the enthalpy. The instrument was calibrated for the heating regime; therefore, each temperature determined during the cooling should be corrected by the addition of þ0.578 K. Each sample was measured in three subsequent heating and cooling cycles in order to test the reversibility and reproducibility of the effects. Each exothermal or endothermal effect (peak) was characterized by its onset (Tx) and endset (Te) temperature and the transformation enthalpy (DH). Each deduced effect was assigned a call number, which increased following the heating and is written as a subscript of the effect of numerical value, e.g., Txm,1, Tem,1, and DHm1 in the case of the first melting effect and Txs,1, Tes,1 DHs1 in the case of the last cooling effect. The connection between the effects at heating to those at cooling is a natural choice for the cyclic nature of the DSC measurement. The wetting of the copper substrate was studied by the sessile drop method. A cube of solder with the edge length of 4 mm and the substrate were mechanically polished and cleaned in alcohol, followed by etching in 10 vol% sulfuric acid in methanol. The solder was placed on the substrate, which was then introduced into the furnace. Both the substrate and the solder were covered with flux (solution of colophony and diethylammonium dichloride in isopropyl alcohol) prior to insertion into the furnace. An additional wetting of the copper substrate was performed in deoxidizing N2 þ10%H2 gas. The measurements in both atmospheres were done at 553, 573, 623, 653 and 673 K for 1800 s at atmospheric pressure. The drop of the solder was photographed by a digital camera (Olympus 550 with telephoto lens), and the wetting angle was measured using a personal computer. The samples after the wetting experiments were cut (by Buhler diamond saw) to reveal the cross-section. After the metallographic treatment (carborundum paper, diamond paste þemulsion), the microstructure of the interface was studied by scanning electron microscopy using an energy-dispersive X-ray analyzer (EDX) and X-ray diffraction (XRD). For X-ray diffraction analysis, most of the Cu (substrate) and solder was removed from the specimen in order to maximize the interface area. Two X-ray diffractometers were used. One was a conventional horizontal diffractometer HZG-4 with Cu Ka radiation in the Bragg–Brentano configuration with a graphite monochromator in the diffracted beam. The second one was a Bruker D8 Discover Super Speed Solution diffractometer equipped with an 18 kW Cu rotating anode and a TXS generator operating at 12 kW.
ˇ P. Sebo et al. / Materials Science & Engineering A 571 (2013) 184–192
186
This diffractometer was used for spatially resolved phase analysis in the direction perpendicular to the Cu–solder interface [13]. The vertical beam size was restricted by a primary divergence slit yielding a beam width of 0.1 mm at the sample surface. The horizontal size of the beam was limited to 6 mm. The sample holder drive allowed a controlled shift of the sample from the predetermined position of the Cu–solder interface. Consecutive X-ray profiles were taken from the specimen after a shift of 0.1 mm to 1 mm from the copper substrate into the solder. To determine the influence of silver content on intensity of solder interaction with Cu substrate, the thickness of Cu3Sn layer (it appears on the solder–substrate border and is adjacent to the substrate) was measured. As the thickness of this phase is not even, we determined the surface of the phase and its length numerically using SigmaScan Pro 5.0 software from the images obtained by SEM after wetting experiments. The surface and its length were measured from three different places along the interface over a total length of 5 mm. A set of three samples for each measurement were used. The average thickness of Cu3Sn phase was calculated as a ratio of the surface to the length. The copper–solder–copper joints were prepared by inserting solder in ribbon form between two copper plates. One of the plates was a disc with 15 mm diameter and 1.5 mm thickness. The other was a square with an edge length of 15 mm and a thickness of 1.5 mm. Both substrates were covered with flux.
The joints were prepared at 623 K for 1800 s in air. Another set of joints was prepared under the same conditions, with deoxidizing gas atmosphere and without using flux. The preparation and cleaning of the copper substrates were identical to the methods used for the wetting experiments. The shear strength was measured by the push-off method at a loading rate of 1 mm/min.
3. Results and discussion 3.1. Microstructure of the solders, interfaces and soldered joints 3.1.1. Microstructure of the as-prepared solders Fig. 1 shows the microstructure of the as-prepared B1 and B5 solders. Microstructure and phase content for B1 sample obtained by spot analysis are presented in Fig. 2 and Table 2, respectively. The accelerating voltage used was 5 and 15 kV in all cases to eliminate artifacts from the broadening of the spot from X-ray generation volume. In all cases, for the composition used the size of the X-ray generation volume did not exceed 3 mm diameter. All solders consist of Sn matrix (with up to 4 at% of dissolved Bi)þBi particles (the big particles surrounded by the small ones)þ Ag3Sn compound in the form of needles. The relevant X-ray diffraction patterns from the B1, B3 and B5 solders are shown in Fig. 3, confirming the presence of tin, bismuth and Ag3Sn phase. The X-ray diffraction patterns from the other solders are qualitatively identical. The microstructure does not significantly change with the addition of the silver except that the increase of silver content causes the increase of the size of the needles (compare e.g. Fig. 1a and b). Silver atoms are located in Ag3Sn compound only. 3.1.2. Microstructure of the interface It is known that Sn reacts rapidly with copper to form Cu–Sn intermetallic compounds (IMC). These IMC weaken the solder
Sn
Bi
Fig. 2. Microstructure of B1 solder with phase analysis.
Table 2 Concentration of the elements in B1 solder (in at%) shown in Fig. 2 obtained from the EDX analysis.
Ag3Sn
Fig. 1. Microstructure of as-prepared B1 (a) and B5 (b) solders.
Spectrum
Sn
Bi
Ag
Phase
1 2 3 4
24.45 96.72 97.37 –
– 3.28 2.63 100.00
75.55 – – –
Ag3Sn Sn Sn Bi
ˇ P. Sebo et al. / Materials Science & Engineering A 571 (2013) 184–192
intensity [a.u.]
Sn(100-x)Bi10Agx, x = 3.3, 6.7, 10
Ag3Sn Bi Sn
B5
B3 B1
pure Sn
25
30
35
40
50 45 2 theta [deg]
55
60
Fig. 3. X-ray diffraction patterns from B1, B3 and B5 solders.
65
187
joint strength because of their brittleness and weakness. For this reason it is useful to understand and control the kinetics of the interfacial reactions [6]. Fig. 4a shows the interface between the copper substrate and the B1 solder after wetting the substrate at 623 K for 1800 s in air. The existence of two intermetallic compounds is confirmed: the Cu3Sn phase located adjacent to the copper substrate and the Cu6Sn5 phase located at the interface and inside the solder. The X-ray diffraction profiles from the specimen after the wetting of the substrate with B1 solder in air, acquired in the X-scan mode described above, can be seen in Fig. 5. The profile confirms the data collected from the EDX point measurement shown in Fig. 4a. The figure shows five diffraction patterns from the specimen. The X-ray profiles characterize the existence and localization of the phases at the interface and in its vicinity. The first pattern corresponds to the X-ray beam position 0.1 mm below the interface in the copper substrate. The second corresponds to the interface line. Each subsequent pattern was taken after shifting the position of the specimen 0.1 mm further into the solder drop. Ten diffraction patterns were recorded (up to the solder depth of 1 mm), although in Fig. 5 only five of them are presented.
Fig. 4. (a) Interface between the copper substrate (oriented horizontally at the bottom of the image) and solder B1 after wetting the substrate at 623 K for 1800 s in air. A1, A2—Cu3Sn phase; A3–A6—Cu6Sn5 phase; A7, A8—coexistence of Sn- and Bi-phases. (b) EDX copper maps of the interface between the copper substrate and the solders B1 (left) and B5 (right) after wetting the substrate at 623 K for 1800 s in air used for determination of the effective thickness of the Cu3Sn layer (immediately adjacent to the Cu substrate).
ˇ P. Sebo et al. / Materials Science & Engineering A 571 (2013) 184–192
188
Fig. 5. X-ray diffraction pattern from the interface between the substrate and B1 solder after wetting at 623 K for 1800 s in air. Open upside down triangles—Cu3Sn, full upside down triangles—Cu6Sn5. Non-indexed large diffraction peaks correspond to Sn, Bi and Ag3Sn phases as indexed in Fig. 3 and Cu substrate (note the bottom-most diffraction pattern at 0.1 mm).
Table 3 The average thickness of the Cu3Sn phase developed at the interface of B1–B5 solders with copper substrate after wetting at 623 K for 1800 s in N2 þ 10 H2 gas. Solder
At% of Ag in a solder
Thickness of Cu3Sn phase (mm)
B B B B B
3.33 4.67 6.67 8.33 10.00
3.99 3.79 3.99 5.06 4.72
1 2 3 4 5
The presence of the Cu6Sn5 phase was determined from five nonoverlapping peaks (full upside-down triangles in Fig. 5), and the presence of Cu3Sn was determined from two non-overlapping peaks (open upside-down triangles in Fig. 5). The Cu3Sn phase was detected only in a narrow area of the interface adjacent to the Cu substrate while the Cu6Sn5 phase was detected both at the interface area and inside the bulk of the solder drop as well. The entire content of the Cu6Sn5 and Cu3Sn phases is of the order of a few percent of the overall X-ray illuminated area, which is close to the usual detection limit of XRD analysis (typically around 1 vol%), thus explaining their near invisibility in the classical Bragg–Brentano X-ray diffraction analysis. This is especially true for the Cu3Sn phase, which is localized in a very thin layer at the Cu interface. The elemental mapping results indicate the distribution of Cu, Sn, Bi and Ag in Sn–10Bi–10Ag/Cu joint at the interfacial area (Figs. 4b, 6a–g). At the interface between molten Sn–Bi–Ag solder and Cu substrate only Cu reacts with Sn [9] and Ag and Bi are expelled from the reaction front into the molten solder because of the formation of the Cu6Sn5. The change (increase) of Ag concentration in the solder does not have influence on the composition of the substrate–solder interface. In spite of detailed and repeated measurements of the thickness of the Cu3Sn phase for all silver concentrations in the solders, no systematic correlation of the layer thickness with Ag content has been observed (Table 3, Fig. 4b), possibly also due to the scallopy character of the surface of the layer adjacent to the Cu6Sn5 phase.
3.1.3. Microstructure of the soldered joint Fig. 6 shows the microstructure of the joint prepared with B3 solder at 623 K for 1800 s. The EDS maps clearly show the distribution of the individual elements, showing phases rich in copper as well as in tin and bismuth (the brighter the signal, the higher the concentration of the respective element). The measurements of concentration at points in these layers indicate the presence of Cu3Sn phase adjacent to the copper substrate and Cu6Sn5 at the boundary with the solder. More detailed compositional information about the distribution of constituent elements has been obtained using trivariate correlation analysis, TVA, [14] applied to the acquired elemental maps. Such an approach has been used successfully in our previous work to characterize the interface between Cu substrate and Sn–Sb solder [15]. The TVA allows to map the distribution of three elements into clusters with typical compositions which are correlated spatially with the areas of occurrence of such compositions in real images. Fig. 6g shows such mapping of three selected elements present in the image of the solder joint, Fig. 6a, and displayed in the individual elemental maps, Fig. 6c–f. For convenience and based on the information of Ag being in the Ag3Sn phase only (Fig. 6d), the elements Cu, Sn and Bi were mapped only. Fig. 6b shows the phase image constructed on the basis of compositional cluster regions in Fig. 6g remapped into the real space. It can be seen that pure Cu is localized in the joint substrate (left side of the image). The phase consisting of majority of Cu and of Sn (equivalent to Cu3Sn) forms a layer adjacent to the Cu substrate, followed by a phase with nearly equal amount of Cu and Sn (equivalent of Cu6Sn5). The phase equivalent to the solder composition containing majority of Sn with addition of Bi corresponds well to the solder matrix, while nearly pure Bi region is mapped to the particles close to the Cu6Sn5 layer on the solder side. Intermediate Sn–Bi phase is localized mainly at the interface between the Bi particles and the solder matrix. The star-like particle containing Sn and small amount of Bi corresponds to the Ag3Sn phase (the presence of Ag in the particle is confirmed by the elemental map in Fig. 6d). Small black and white regions correspond to clusters with unidentified or empty compositional content. 3.2. Kinetics of melting of SnBiAg solders DSC studies show that under continuous heating, the melting of the B1–B5 bulks is a complex phenomenon that occurs at 413–518 K. The melting process consists of three separate endothermal events: the minor but sharp first effect R1, the massive main effect R2 and the wide final effect R3 (Fig. 7). All three phenomena are reversible. Subsequently, as the melted B1– B5 are cooled, their solidification processes exhibit the same three exothermal events, starting with R3, then R2 and, finally, R1. The DSC curves for all five solders have the same peaks with similar separation, but the temperature and transformation heat values depend on the particular chemical composition of each alloy. The characteristic temperatures and transformation heats of the individual effects are summarized in Table 4. According to the phase analysis, the R1 effect corresponds to the solid/liquid phase transition of the Sn/Bi eutectic. In the case of melting, its temperature is constant and compatible with the corresponding binary phase diagram (412.15 K [8]) for all 5 alloys. Due to the small transformation heat of R1, the content of the eventual liquid phases at 413 K is small, and it decreases from B1 to B5, reflecting the probable decrease of this eutectic phase. The melting of the remaining Sn and Bi from this eutectic starts immediately after R1, forming the main DSC peak, R2, which culminates at approximately the same temperature (Tem2 478 K) for all alloys. The thermal effect R2 can include the effect from the
ˇ P. Sebo et al. / Materials Science & Engineering A 571 (2013) 184–192
189
Fig. 6. (a) Microstructure of the Cu–B3–Cu joint made at 623 K after 1800 s and its phase map (b) constructed using trivariate correlation analysis. EDS maps of individual elements: (c) Cu, (d) Ag, (e) Sn, (f) Bi, (g) trivariate correlation diagram of Cu, Sn and Bi. Cu substrate is oriented vertically in the left part of the SEM image.
2.2 Heat power [ W/g]
Heat power [W/g]
2.2
2.0 heating R1
R2
R3
1.8
2.0 heating
1.8 cooling
cooling 350
400 450 500 550 Temperature [K]
600
350 400 450 500 550 600 650 Temperature [K]
Fig. 7. (a) Heating and cooling DSC curves of the Sn83.33Bi10Ag6.67 solder (B3) in the bulk form. (b) The relation between the heating and cooling DSC curves of B1 (blue dashed line), B3 (black solid line) and B5 (red dot-dashed line) solders in the bulk form. Heating and cooling rates were 7 10 K/min. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Ag3SnþSn-L eutectic reaction. But due to the low content of Ag in the alloys, the eventual heat of melting of this eutectic should be minor and not visible within the complex DSC peak R2. The transformation heat of the complex R2 peak decreases from B1 to B5, along with the content of the first eutectic. The remaining
Ag3Sn and Sn melt in the last broad thermal effect, R3. The end temperature of R3, Tem3, as well as the heat DHm3, dramatically increases in proportion to the increasing content of Ag in the alloy. The Tem3 temperatures are desirable in practice because they determine the completed liquid state; they represent the liquidus
ˇ P. Sebo et al. / Materials Science & Engineering A 571 (2013) 184–192
190
Table 4 Onset temperature Tx and transformation enthalpies DH of three melting (m) and solidification (s) effects (reactions R1, R2 and R3) of relevant solders B1–B5 in the bulk form with various amounts of silver in the as-prepared state.
causes small decrease of DHm1 and DHm2, but significant increase of DHm3 (except for B4). 3.3. Wetting
Solder
B1
B2
B3
B4
B5
Txm1 [K] Txm2 [K] Txm3 [K] Tem3 [K] DHm1 [J g 1] DHm2 [J g 1] DHm3 [J g 1] Txs1 [K] Txs2 [K] Txs3 [K] DHs1 [J g 1] DHs2 [J g 1] DHs3 [J g 1]
413.0 458.7 477.9 518.3 4.7 43.5 2.7 388.8 450.1 470.4 0.6 42.5 1.7
412.5 467.3 478.8 554.3 4.4 43.7 6.4 387.2 446.3 497.4 0.3 43.7 3.9
412.6 462.9 478.4 594.5 4.3 41.2 8.2 388.7 447.2 556.0 0.3 40.6 7.2
412.8 462.3 477.8 604.9 4.3 40.3 15.1 389.0 443.3 570.5 0.3 40.0 9.8
412.1 459.8 476.9 636.2 3.9 39.7 18.4 388.0 448.6 604.2 0.3 35.6 13.9
Fig. 8 shows the time dependence of the wetting angle on the copper substrate for the B1 solder at 553, 573 and 623 K. The wetting angle tends to decrease with increasing temperature and reaches its practically stable values after the first 600 s of exposition. Wetting rate at 623 K for the first 300 s in deoxidizing gas was for all solders around 0.03 deg./s. Wetting angle in air (with flux) was practically constant for the given solder. Fig. 9 shows the time dependence of the wetting angle on the copper substrate for solders at 623 K after 1800 s of wetting in both atmospheres: air (a) and deoxidizing gas N2 þ10% H2 (b). It is worthwhile to mention that these figures demonstrate opposite dependence of the wetting angle on the silver content of solder. When the wetting is performed in air, the wetting angle increases with the content of silver in the solder, but increasing the silver content lowers the wetting angle when wetting is performed in the N2 þ10% H2 gas. This difference may be due to the oxidation of silver, which generally frustrates wetting. These results can be used
Fig. 8. Time dependence of wetting angle of copper substrate by the B1 solder for 553, 573 and 623 K. The accuracy of wetting angle determination is 7 21.
of each solder. Due to the obscure DSC signal of the weak peak R3 in the heating regime (in contrast to the sharp onset of same R3 effect via cooling), both the temperature of liquidus and the enthalpy of the reaction are also determined from the cooling regime by the values Txs3 and DHs3. Both temperatures Tem3 and Txs3 correspond to the liquidus of the alloy; in fact, they differ by 32–62 K (see Table 4). Each difference Txs3 Tem3 is the supercooling necessary for nucleation prior to the start of the growth of the first crystalline phase which has been obtained at the actual cooling rate 10 K/min. Reflecting the rapid quenching, neither the phase composition nor the proportion between the elements in the main binary phases is in equilibrium in the ribbons. This fact is reflected in numerous minor phenomena in their first melting, i.e. several small subpeaks appear. All of these peaks are irreversible. After heating to 723 K, the solidification of the samples resembles that of the bulks and is the same in the next measuring cycle. The transformation (or sometimes the absence) of R1 peak on DSC curves of quenched ribbons allows one to say that precipitation of Bi particles was suppressed. Thus by varying temperature regimes of melt heat treatment before quenching/crystallization it becomes possible to influence the morphology of Bi-rich phase and solubility of bismuth in tin matrix. Summarizing the results from DSC it can be said that the increase of silver content (i) does not influence the values of Txm1, Txm2 and Txm3, but provides the drastic increase of Tem3; and (ii)
Fig. 9. Time dependence of wetting angle of copper substrate for various contents of silver in the solder at the temperature 623 K after 1800 s in atmosphere-air with flux (a) and in N2 þ 10% H2 gas (b). The accuracy of wetting angle determination is as shown in Fig. 8.
ˇ P. Sebo et al. / Materials Science & Engineering A 571 (2013) 184–192
for modifying the regimes of soldering: the utilization of the deoxidizing gas N2 þ10% H2 can shorten the process for the solders with high silver content.
191
Table 5 Shear strength of Cu-–solder–Cu joints with some lead-free solders. Sheer strength [MPa]
Substrate Reference
39 38 39 40 50 51 81
Cu Cu Cu Cu Cu Cu Cu
[10]
350 350 350
42 41 18 16 18
Cu Cu Cu Cu Cu
[10] [10] [17] [17] [17]
1800
350
15.8
Cu
[17]
1800
350
19.5
Cu
Actual paper
1800
350
21
Cu
1800
350
19.5
Cu
Solder
Reflow time [s]
Reflow temp. [1C]
96.5 Sn–3.5 Ag
120 3120 120 3120 120 3120 15
350–400
2
235
1800 1800 1800
3.4. Joint strength The shear strength of the joints was studied in terms of its dependence on two variables: the silver content of the solder and the influence of deoxidizing gas, as compared to the joints prepared in the air atmosphere. Fig. 10 shows the shear strength of the Cu–solder–Cu joints prepared at various temperatures over a period of 1800 s, with all relevant Sn100 xBi10Agx solders, containing x¼3.3–10 at% silver. For the temperatures up to 623 K the joints with B3 solder demonstrate the highest shear strength. For higher temperatures the increase in silver content above 7 at% Ag leads to a small increase in the shear strength of the joints. The shear
95.5 Sn–3.8 Ag–0.7 Cu 93.5 Sn–3.5 Ag–3 Bi 91.84 Sn–3.33 Ag–4.83 Bi 63 Sn–37 Pb 100 Sn 95 Sn–5 Sb 80 Sn–20 Sb 90.8 Sn–7.4 Sb–1.8 Cu 76.1 Sn–20.2 Sb–3.7 Cu 86.7 Sn–10 Bi–3.3 Ag 83.3 Sn–10 Bi–6.7 Ag 80 Sn–10 Bi–10 Ag
235 235 235
[16] [16] [16]
strength of the Cu–solder–Cu joints prepared in deoxidizing gas (Fig. 11) decreases with increasing silver concentration, except for the B3 solder. Shear strength of the joints for various reflow times and reflow temperatures for some alternative solders for comparison with Sn–Pb solder is shown in Table 5. It can be seen that the reflow time does not have a strong effect on the decrease of the shear strength of the joint. Reflow temperature has a more expressive effect on the decrease of the shear strength. Higher reflow temperature as well as reflow time may lead to the growth of the Cu–Sn intermetallic compounds which can have an effect on the strength of the joint. Another reason connected with the increasing of the reflow temperature and reflow time leading to the decrease of the joint strength can be the growth of the Ag3Sn needles. Fig. 10. Shear strength of Cu–solder–Cu joints prepared in atmosphere-air at given temperatures after 1800 s. The accuracy of shear strength determination is 7 0.5 MPa.
3.5. Proposal for optimized solder composition From the obtained results one could suggest the best chemistry for the studied materials i.e. solder of the Sn–10Bi–xAg type (xA(3, 10 wt%). Such solder based on tin could contain 4 wt% Bi and 3 wt% Ag. The solubility of Bi in Sn–Ag alloy is around 4 wt%. Higher Bi content would not be expected to contribute to additional strength through the solid solution mechanism [10]. The lowest wetting angle from all studied alloys with the lowest silver content possesses (B1) at 623 K. The wetting angle of this alloy is low even at 553 K ( 321). Another fact supporting the suggestion of the best solder of Sn–10Bi–xAg type is low content of silver. Vianco and Rejent [12] limited the amount of silver to be below 3.2 wt% to avoid the formation of large Ag3Sn platelets. Large Ag3Sn platelets induce serious degradation of joining mechanical properties. These do not appear in alloys with less than 3.2 wt% Ag content [12]. Shear strength of the Cu–solder–Cu joints is also the highest for B1 solder (3.3 wt% Ag) at 280–300 1C. The relatively low Ag content is also in line with the eventual cost of such solder alloy.
4. Conclusions Fig. 11. Shear strength of Cu–solder–Cu joints prepared in N2 þ 10% H2 gas at given temperatures after 1800 s. The accuracy of shear strength determination is as shown in Fig. 10.
The effects of silver in the Sn100 xBi10Agx (x¼3.3–10 at%) leadfree solder on the transition from the solid to liquid state, on the wetting of the copper substrate and on the shear strength of the
192
ˇ P. Sebo et al. / Materials Science & Engineering A 571 (2013) 184–192
Cu–solder–Cu joints were studied. The influence of the working atmosphere, either air with flux or deoxidizing N2 þ10% H2 gas, the microstructure of the as-prepared solder, and the interaction between the substrate and the solder after wetting at 623 K for 1800 s were also investigated. The obtained results can be expressed as follows:
1. The as-prepared solders consist of tin, bismuth and Ag3Sn phases only. The latter exists in the form of needles. The increase of silver content enhances the length and diameter of needles. 2. The melting and solidification phenomena of the solders consist of three separate events. Their presence confirms the existence of two types of eutectics: the binary Snþ Bi and the ternary eutectic Ag3SnþSn(Bi). While the temperature of the solidus remains constant for all five solders, the temperature of the liquidus dramatically increases with the content of Ag. 3. For wetting in the air, the wetting angle slightly increases with the silver content of the solder, whereas the silver lowers the wetting angle in N2 þ10% H2 gas. The wetting in both cases is very good (wetting angle o 301). 4. For the temperatures up to 623 K the joints with B3 solder demonstrate the highest shear strength. For higher temperatures the increase in silver content above 7 at% Ag brings a small increase in the shear strength of the joints. The shear strength of the joints prepared in the deoxidizing gas at 623 K for 1800 s decreases with increasing silver content. 5. At the interface, there is a layer of the Cu3Sn phase, which is adjacent to the copper substrate. The thickness of this layer grows with the increase of silver in the solder. 6. The methods of x-scan X-ray analysis and trivariate correlation analysis have been conveniently used for the detection of small amounts of individual phases and for their spatial localization at different distances from the Cu–solder joint. Based on the results obtained it can be recommended that silver content should not exceed 6 at% in Sn100 xBi10Agx solders.
Acknowledgments The authors gratefully acknowledge the financial support of the Slovak Grant Agency under the Contracts VEGA 2/6160/26 and 2/0111/11. This work was also supported by the Slovak Research and Development Agency under the Contracts APVV0102-07, APVV-0076-11, APVV-0492-11 and APVV-SK-UA-004309 and by CEX FUN-MAT. V.S. gratefully acknowledges the fellowship of the Slovak Academic Information Agency. This research contributes to the European COST Action MP0602 on Advanced Solder Materials for High Temperature Applications.
References [1] C.M. Miller, I.E. Anderson, J.F. Smith, J. Electron. Mater. 23 (1994) 595–601. [2] N. Sobczak, J. Sobczak, R. Nowak, A. Kudyba, P. Darlak, B. Mikulowski, A. Wojciechowski, J. Mater. Sci. 40 (2005) 2547–2551. [3] S.W. Chen, H.J. Wu, Y.C. Huang, W. Gierlotka, J. Alloys Compd. 497 (2010) 110–117. [4] Z. Li, Z. Cao, S. Knott, A. Mikula, Y. Du, Z. Qiao, CALPHAD: Comput. Coupl. Phase Diagr. Thermochem. 32 (2008) 152–163. [5] L. Zabdyr, G. Garze, CALPHAD: Comput. Coupl. Phase Diagr. Thermochem. 33 (2009) 187–191. [6] M. He, V.I. Acoff, J. Electron. Mater. 35 (2006) 2098–2106. [7] D. Witkin, J. Electron. Mater. 41 (2012) 190–203. [8] M.L. Huang, L. Wang, Metall. Mater. Trans. A 36 (2005) 1439–1446. [9] M. He, V.I. Acoff, J. Electron. Mater. 37 (2008) 288–299. [10] P.T. Vianco, J.A. Rejent, J. Electron. Mater. 28 (1999) 1138–1143. [11] D. Witkin, Mater. Sci. Eng. A 532 (2012) 212–220. [12] P.T. Vianco, J.A. Rejent, J. Electron. Mater. 28 (1999) 1127–1137. ˇ ˇ [13] P. Sebo, Z. Moser, P. Svec, D. Janicˇkovicˇ, E. Dobrocˇka, W. Gasior, J. Pstrus´, J. Alloys Compd. 480 (2009) 409–415. [14] F. Hofer, W. Grogger, G. Kothleitner, P. Warbichler, Ultramicroscopy 67 (1997) 83–103. ˇ ˇ ˇ [15] P. Svec, J. Hoˇsko, P. Svec Sr., I. Maˇtko, P. Sebo, D. Janicˇkovicˇ, in: J. Vajda, M. Weis (Eds.), Proceedings of the 16th International Conference on Applied Physics of Condensed Matter, APCOM, Mala Lucivna, Slovak University of Technology, Bratislava, June 16–18, 2010, pp. 292–297, ISBN 978-80-227-3307-6. [16] S.K. Kang, W.K. Choi, M.J. Yim, D.Y. Shih, J. Electron. Mater. 31 (2002) 1292–1303. ˇ ˇ [17] P. Sebo, P. Svec, D. Janicˇkovicˇ, E. Illekova´, Kovove Mater. 48 (2010) 353–359.
Contents lists available at SciVerse ScienceDirect
Materials Science & Engineering A journal homepage: www.elsevier.com/locate/msea
The influence of silver content on structure and properties of Sn–Bi–Ag solder and Cu/solder/Cu joints a ˇ ˇ P. Sebo , P. Svec Sr. b,c,n, D. Janicˇkovicˇ b, E. Illekova´ b, M. Zema´nkova´ a, Yu. Plevachuk d, V. Sidorov e, b ˇ P. Svec a
Institute of Materials and Machine Mechanics, Slovak Academy of Sciences, Racˇianska 75, 831 02 Bratislava 3, Slovakia ´ cesta 9, 845 11 Bratislava 45, Slovakia ´ bravska Institute of Physics, Slovak Academy of Sciences, Du c Faculty of Materials Science and Technology, Slovak University of Technology, J. Bottu 25, 917 24 Trnava, Slovakia d Ivan Franko National University, Department of Metal Physics, 79005 Lviv, Ukraine e Ural State Pedagogical University, Cosmonavtov 26, 620017 Ekaterinburg, Russia b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 30 October 2012 Received in revised form 3 February 2013 Accepted 8 February 2013 Available online 14 February 2013
The effect of silver content on structure and properties of Sn100 xBi10Agx (x ¼ 3–10 at%) lead-free solder and Cu–solder–Cu joints was investigated. The microstructure of the solder in both bulk and rapidly solidified ribbon forms was analyzed by scanning electron microscopy (SEM) and X-ray diffraction. The peculiarities in melting kinetic, studied by differential scanning calorimetry (DSC), and silver influence on it are described and discussed. The wetting of a copper substrate was examined by the sessile drop method in the temperature range of 553–673 K in air and deoxidizing gas (N2 þ 10%H2) at atmospheric pressure. Cu–solder–Cu joints were also prepared in both atmospheres, and their shear strength was measured by the push-off method. The produced solders consisted of tin, bismuth and Ag3Sn phases. The product of the interaction between the solder and the copper substrate consists of two phases: Cu3Sn, which is adjacent to the substrate, and a Cu6Sn5 phase. The wetting angle in air increased slightly as the silver concentration in the solder increased. Wetting of the copper substrate in N2 þ10H2 gas shows the opposite tendency: the wetting angle slightly decreased as the silver content in the solder increased. The shear strength of the joints prepared in air (using flux) tends to decrease with increasing production temperature and increasing silver content in the solder. The equivalent decrease in the shear strength of the joints prepared in N2 þ 10H2 is more apparent. & 2013 Elsevier B.V. All rights reserved.
Keywords: Lead-free solder Cu joints Melting Wetting Shear strength X-ray diffraction
1. Introduction Sn–Pb solders were the most used joining materials in the electronic industry for many years because of their low cost and superior properties. Recently, the regulation of certain hazardous substances (RoHS) and waste electrical and electronic equipment (WEEE) has resulted in extensive research on lead-free solders. Nowadays SnAgCu (SAC) alloys are considered to be the best substitution for lead containing solders with Pb content up to 85 wt% [1,2]. However, for high melting (tl 4 503 K) solders, no proper equivalent is available at the moment. There are many potential candidates. Majority of these systems were studied in the frame of COST MP0602 Action. Many of the studied alloys were based on Sn (Sn–Sb, Sn–Au, Sn–Zn–Ni), ternary alloys e.g. Al–Zn–X (X ¼Ga, Ge, Mg, Ni, Sb and Sn), Al–Sb–Zn, Ag–Sb–Sn and n Corresponding author at: Institute of Physics, Slovak Academy of Sciences, Du´bravska´ cesta 9, 845 11 Bratislava 45, Slovakia. Tel.: þ 421 2 5941 0561; fax: þ 421 2 5477 6085. ˇ E-mail address: [email protected] (P. Svec Sr.).
0921-5093/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.msea.2013.02.013
many others. One potential candidate is also Sn–Bi–Ag system in tin rich corner. The phase diagram for ternary Sn–Bi–Ag alloys was comprehensively investigated both experimentally and numerically in [3–5]. DSC curves for Sn–Bi–Ag solders obtained at heating show endothermal peak in the interval 409–413 K. Some authors associate this with melting of Sn/Bi eutectic, others—with SnþBiþ Ag3Sn-LþAg3Sn reaction [3–5]. As a rule, the fraction of liquid phase at t4413 K is rather low (not exceeding 1% for 8 wt% of Bi content [5]). This phenomenon, however, causes some restrictions in usage of Sn–Bi–Ag alloys in the step soldering technology, in which soldering occurs more than once during the manufacturing process but for soldering of wires and printed circuit boards it is acceptable. The microstructure of Sn–Bi–Ag solder was found to consist of Ag3Sn particles, Sn-rich and Bi-rich phases in different combinations, depending on the cooling rate after reflow [6]. After aging, Bi in SAC–Bi and SnAg–Bi alloys was found as precipitates at grain boundaries and grain interiors [7]. The solid solubility limit of Bi in the Sn–Ag-based alloy is about 4 wt% at room temperature
ˇ P. Sebo et al. / Materials Science & Engineering A 571 (2013) 184–192
[3–5], and with increased Bi addition, the supersaturated Bi would precipitate in the form of pure Bi phase from the Sn matrix. Additions of Bi tend to suppress the formation of large Ag3Sn particles from the solder matrix [8]. No ternary compounds were found in this system [3]. In the Sn–Bi–Ag/Cu solder systems Cu reacts only with Sn, not with Ag or Bi [9]. Because of the formation of the Cu6Sn5 intermetallic compound (IMC), Ag and Bi are expelled from the reaction front into the molten solder between the IMC scallops. The existence of Bi and Ag in the gaps prevents the Cu–Sn grains from joining together and forming a continuous layer. The higher the Bi and Ag concentration in the solder, the more Bi and Ag is formed in the gaps. Bi and Ag additions affect the total thickness of interfacial reaction layers [9]. Several authors have examined wetting and mechanical properties of SnBiAg solders and joints with various concentrations of bismuth. Vianco and Rejent [10] showed that additions of 1–10 wt% Bi into the Sn–3.5 wt% Ag solder improve its wetting performance on Cu. Increasing the Bi content in the solder raised the Cu–solder–Cu joint shear strength to 81 MPa, as determined by the ring-and-plug test [10]. The microhardness reached maximum values of 30 (Knoop, 50 g) and 110 (Knoop, 5 g) for Bi contents greater than 4–5 wt%. He and Acoff [9] investigated the effect of reflow and the thermal aging process on microhardness of Sn–3.7 wt% Ag solder with 0, 1, 2, 3, and 4 wt% Bi. They proved that the microhardness increases with increasing Bi in the solder, regardless of reflow or thermal aging process. Huang and Wang showed that bismuth additions linearly increase the ultimate tensile strength of Sn–Ag lead-free solders [8]. The mechanical properties and the microstructure for the Sn–3 wt% Ag solders with 0, 3, and 6 wt% of Bi, connecting two Cu substrates, were experimentally examined in [9]. It was stated that for the solder between the Cu substrates, the strength increases from 90 to 140 MPa with increasing concentration of Bi from 0 to 6 wt%. The mechanical properties of bulk Pb-free solder alloys containing Bi have been characterized for the as-cast and aged samples in [7,11,12]. Room and elevated-temperature tensile testing showed that the addition of Bi greatly reduced the loss in strength due to aging that occurs in the Sn–Ag–Cu ternary alloys. In SnAg–Bi the tensile strength increased after aging. Hence, it is a common place in all the works that bismuth additions into Sn–Ag or SAC solders improve its mechanical properties and wetting ability on Cu substrate. It is a well known fact that by adding silver to tin one improves its reliability and wetting ability; however, nobody studied precisely the influence of silver content on the structure and properties of Sn–Bi–Ag solder itself and its interaction with substrates. Thus the aim of this work is to determine the effect of silver on (a) melting kinetics, (b) wetting properties and (c) shear strength of Sn100 xBi10Agx alloys and Cu/solder/Cu joints as well as to connect these properties with its microstructure.
2. Experimental A series of lead-free Sn100 xBi10Agx solders with different Ag content (Table 1) was prepared by melting appropriate amounts of the relevant metals, with purity 99.99% and higher, in an induction furnace in an overpressure of argon at approximately 870 K. Their chemical composition was determined using an atomic emission spectrometer Spectrum Flame Modula S. Solders were prepared in two forms: bulk and ribbon. The bulk solder was used for DSC and for wetting experiments. Solders in ribbon form were prepared by rapid quenching of the melt and used for the preparation of copper–solder–copper joints and DSC measurements. The microstructure of the solder in both forms was
185
Table 1 Composition of Sn100 xBi10Agx solders (at%). Solder
Sn
Bi
Ag
B1 B2 B3 B4 B5
86.7 85.3 83.3 81.7 80.0
10.0 10.0 10.0 10.0 10.0
3.3 4.7 6.7 8.3 10.0
investigated by SEM equipped with an energy dispersive analyzer. SEM used was JEOL 7600F with a Schottky FEG emitter and EDX from Oxford Instruments X-max with a 50 mm2 detector. Software used for control of EDX acquisition and semiquantitative analysis was Oxford Instruments INCA version 4.15. The Perkin–Elmer DSC7 differential scanning calorimeter was calibrated for a heating rate of 710 K/min. The same aluminum sample pan was always loosely covered with the lid. A reference pan containing no sample and a flowing (20 ml/min) argon atmosphere were used. The sample mass was 10 mg. Two independent samples were always used; thus, the reproducibility of each effect has been proven and the accuracy of measurement of each quantity has been calculated to be 70.5 K for temperature and 72 J/g for the enthalpy. The instrument was calibrated for the heating regime; therefore, each temperature determined during the cooling should be corrected by the addition of þ0.578 K. Each sample was measured in three subsequent heating and cooling cycles in order to test the reversibility and reproducibility of the effects. Each exothermal or endothermal effect (peak) was characterized by its onset (Tx) and endset (Te) temperature and the transformation enthalpy (DH). Each deduced effect was assigned a call number, which increased following the heating and is written as a subscript of the effect of numerical value, e.g., Txm,1, Tem,1, and DHm1 in the case of the first melting effect and Txs,1, Tes,1 DHs1 in the case of the last cooling effect. The connection between the effects at heating to those at cooling is a natural choice for the cyclic nature of the DSC measurement. The wetting of the copper substrate was studied by the sessile drop method. A cube of solder with the edge length of 4 mm and the substrate were mechanically polished and cleaned in alcohol, followed by etching in 10 vol% sulfuric acid in methanol. The solder was placed on the substrate, which was then introduced into the furnace. Both the substrate and the solder were covered with flux (solution of colophony and diethylammonium dichloride in isopropyl alcohol) prior to insertion into the furnace. An additional wetting of the copper substrate was performed in deoxidizing N2 þ10%H2 gas. The measurements in both atmospheres were done at 553, 573, 623, 653 and 673 K for 1800 s at atmospheric pressure. The drop of the solder was photographed by a digital camera (Olympus 550 with telephoto lens), and the wetting angle was measured using a personal computer. The samples after the wetting experiments were cut (by Buhler diamond saw) to reveal the cross-section. After the metallographic treatment (carborundum paper, diamond paste þemulsion), the microstructure of the interface was studied by scanning electron microscopy using an energy-dispersive X-ray analyzer (EDX) and X-ray diffraction (XRD). For X-ray diffraction analysis, most of the Cu (substrate) and solder was removed from the specimen in order to maximize the interface area. Two X-ray diffractometers were used. One was a conventional horizontal diffractometer HZG-4 with Cu Ka radiation in the Bragg–Brentano configuration with a graphite monochromator in the diffracted beam. The second one was a Bruker D8 Discover Super Speed Solution diffractometer equipped with an 18 kW Cu rotating anode and a TXS generator operating at 12 kW.
ˇ P. Sebo et al. / Materials Science & Engineering A 571 (2013) 184–192
186
This diffractometer was used for spatially resolved phase analysis in the direction perpendicular to the Cu–solder interface [13]. The vertical beam size was restricted by a primary divergence slit yielding a beam width of 0.1 mm at the sample surface. The horizontal size of the beam was limited to 6 mm. The sample holder drive allowed a controlled shift of the sample from the predetermined position of the Cu–solder interface. Consecutive X-ray profiles were taken from the specimen after a shift of 0.1 mm to 1 mm from the copper substrate into the solder. To determine the influence of silver content on intensity of solder interaction with Cu substrate, the thickness of Cu3Sn layer (it appears on the solder–substrate border and is adjacent to the substrate) was measured. As the thickness of this phase is not even, we determined the surface of the phase and its length numerically using SigmaScan Pro 5.0 software from the images obtained by SEM after wetting experiments. The surface and its length were measured from three different places along the interface over a total length of 5 mm. A set of three samples for each measurement were used. The average thickness of Cu3Sn phase was calculated as a ratio of the surface to the length. The copper–solder–copper joints were prepared by inserting solder in ribbon form between two copper plates. One of the plates was a disc with 15 mm diameter and 1.5 mm thickness. The other was a square with an edge length of 15 mm and a thickness of 1.5 mm. Both substrates were covered with flux.
The joints were prepared at 623 K for 1800 s in air. Another set of joints was prepared under the same conditions, with deoxidizing gas atmosphere and without using flux. The preparation and cleaning of the copper substrates were identical to the methods used for the wetting experiments. The shear strength was measured by the push-off method at a loading rate of 1 mm/min.
3. Results and discussion 3.1. Microstructure of the solders, interfaces and soldered joints 3.1.1. Microstructure of the as-prepared solders Fig. 1 shows the microstructure of the as-prepared B1 and B5 solders. Microstructure and phase content for B1 sample obtained by spot analysis are presented in Fig. 2 and Table 2, respectively. The accelerating voltage used was 5 and 15 kV in all cases to eliminate artifacts from the broadening of the spot from X-ray generation volume. In all cases, for the composition used the size of the X-ray generation volume did not exceed 3 mm diameter. All solders consist of Sn matrix (with up to 4 at% of dissolved Bi)þBi particles (the big particles surrounded by the small ones)þ Ag3Sn compound in the form of needles. The relevant X-ray diffraction patterns from the B1, B3 and B5 solders are shown in Fig. 3, confirming the presence of tin, bismuth and Ag3Sn phase. The X-ray diffraction patterns from the other solders are qualitatively identical. The microstructure does not significantly change with the addition of the silver except that the increase of silver content causes the increase of the size of the needles (compare e.g. Fig. 1a and b). Silver atoms are located in Ag3Sn compound only. 3.1.2. Microstructure of the interface It is known that Sn reacts rapidly with copper to form Cu–Sn intermetallic compounds (IMC). These IMC weaken the solder
Sn
Bi
Fig. 2. Microstructure of B1 solder with phase analysis.
Table 2 Concentration of the elements in B1 solder (in at%) shown in Fig. 2 obtained from the EDX analysis.
Ag3Sn
Fig. 1. Microstructure of as-prepared B1 (a) and B5 (b) solders.
Spectrum
Sn
Bi
Ag
Phase
1 2 3 4
24.45 96.72 97.37 –
– 3.28 2.63 100.00
75.55 – – –
Ag3Sn Sn Sn Bi
ˇ P. Sebo et al. / Materials Science & Engineering A 571 (2013) 184–192
intensity [a.u.]
Sn(100-x)Bi10Agx, x = 3.3, 6.7, 10
Ag3Sn Bi Sn
B5
B3 B1
pure Sn
25
30
35
40
50 45 2 theta [deg]
55
60
Fig. 3. X-ray diffraction patterns from B1, B3 and B5 solders.
65
187
joint strength because of their brittleness and weakness. For this reason it is useful to understand and control the kinetics of the interfacial reactions [6]. Fig. 4a shows the interface between the copper substrate and the B1 solder after wetting the substrate at 623 K for 1800 s in air. The existence of two intermetallic compounds is confirmed: the Cu3Sn phase located adjacent to the copper substrate and the Cu6Sn5 phase located at the interface and inside the solder. The X-ray diffraction profiles from the specimen after the wetting of the substrate with B1 solder in air, acquired in the X-scan mode described above, can be seen in Fig. 5. The profile confirms the data collected from the EDX point measurement shown in Fig. 4a. The figure shows five diffraction patterns from the specimen. The X-ray profiles characterize the existence and localization of the phases at the interface and in its vicinity. The first pattern corresponds to the X-ray beam position 0.1 mm below the interface in the copper substrate. The second corresponds to the interface line. Each subsequent pattern was taken after shifting the position of the specimen 0.1 mm further into the solder drop. Ten diffraction patterns were recorded (up to the solder depth of 1 mm), although in Fig. 5 only five of them are presented.
Fig. 4. (a) Interface between the copper substrate (oriented horizontally at the bottom of the image) and solder B1 after wetting the substrate at 623 K for 1800 s in air. A1, A2—Cu3Sn phase; A3–A6—Cu6Sn5 phase; A7, A8—coexistence of Sn- and Bi-phases. (b) EDX copper maps of the interface between the copper substrate and the solders B1 (left) and B5 (right) after wetting the substrate at 623 K for 1800 s in air used for determination of the effective thickness of the Cu3Sn layer (immediately adjacent to the Cu substrate).
ˇ P. Sebo et al. / Materials Science & Engineering A 571 (2013) 184–192
188
Fig. 5. X-ray diffraction pattern from the interface between the substrate and B1 solder after wetting at 623 K for 1800 s in air. Open upside down triangles—Cu3Sn, full upside down triangles—Cu6Sn5. Non-indexed large diffraction peaks correspond to Sn, Bi and Ag3Sn phases as indexed in Fig. 3 and Cu substrate (note the bottom-most diffraction pattern at 0.1 mm).
Table 3 The average thickness of the Cu3Sn phase developed at the interface of B1–B5 solders with copper substrate after wetting at 623 K for 1800 s in N2 þ 10 H2 gas. Solder
At% of Ag in a solder
Thickness of Cu3Sn phase (mm)
B B B B B
3.33 4.67 6.67 8.33 10.00
3.99 3.79 3.99 5.06 4.72
1 2 3 4 5
The presence of the Cu6Sn5 phase was determined from five nonoverlapping peaks (full upside-down triangles in Fig. 5), and the presence of Cu3Sn was determined from two non-overlapping peaks (open upside-down triangles in Fig. 5). The Cu3Sn phase was detected only in a narrow area of the interface adjacent to the Cu substrate while the Cu6Sn5 phase was detected both at the interface area and inside the bulk of the solder drop as well. The entire content of the Cu6Sn5 and Cu3Sn phases is of the order of a few percent of the overall X-ray illuminated area, which is close to the usual detection limit of XRD analysis (typically around 1 vol%), thus explaining their near invisibility in the classical Bragg–Brentano X-ray diffraction analysis. This is especially true for the Cu3Sn phase, which is localized in a very thin layer at the Cu interface. The elemental mapping results indicate the distribution of Cu, Sn, Bi and Ag in Sn–10Bi–10Ag/Cu joint at the interfacial area (Figs. 4b, 6a–g). At the interface between molten Sn–Bi–Ag solder and Cu substrate only Cu reacts with Sn [9] and Ag and Bi are expelled from the reaction front into the molten solder because of the formation of the Cu6Sn5. The change (increase) of Ag concentration in the solder does not have influence on the composition of the substrate–solder interface. In spite of detailed and repeated measurements of the thickness of the Cu3Sn phase for all silver concentrations in the solders, no systematic correlation of the layer thickness with Ag content has been observed (Table 3, Fig. 4b), possibly also due to the scallopy character of the surface of the layer adjacent to the Cu6Sn5 phase.
3.1.3. Microstructure of the soldered joint Fig. 6 shows the microstructure of the joint prepared with B3 solder at 623 K for 1800 s. The EDS maps clearly show the distribution of the individual elements, showing phases rich in copper as well as in tin and bismuth (the brighter the signal, the higher the concentration of the respective element). The measurements of concentration at points in these layers indicate the presence of Cu3Sn phase adjacent to the copper substrate and Cu6Sn5 at the boundary with the solder. More detailed compositional information about the distribution of constituent elements has been obtained using trivariate correlation analysis, TVA, [14] applied to the acquired elemental maps. Such an approach has been used successfully in our previous work to characterize the interface between Cu substrate and Sn–Sb solder [15]. The TVA allows to map the distribution of three elements into clusters with typical compositions which are correlated spatially with the areas of occurrence of such compositions in real images. Fig. 6g shows such mapping of three selected elements present in the image of the solder joint, Fig. 6a, and displayed in the individual elemental maps, Fig. 6c–f. For convenience and based on the information of Ag being in the Ag3Sn phase only (Fig. 6d), the elements Cu, Sn and Bi were mapped only. Fig. 6b shows the phase image constructed on the basis of compositional cluster regions in Fig. 6g remapped into the real space. It can be seen that pure Cu is localized in the joint substrate (left side of the image). The phase consisting of majority of Cu and of Sn (equivalent to Cu3Sn) forms a layer adjacent to the Cu substrate, followed by a phase with nearly equal amount of Cu and Sn (equivalent of Cu6Sn5). The phase equivalent to the solder composition containing majority of Sn with addition of Bi corresponds well to the solder matrix, while nearly pure Bi region is mapped to the particles close to the Cu6Sn5 layer on the solder side. Intermediate Sn–Bi phase is localized mainly at the interface between the Bi particles and the solder matrix. The star-like particle containing Sn and small amount of Bi corresponds to the Ag3Sn phase (the presence of Ag in the particle is confirmed by the elemental map in Fig. 6d). Small black and white regions correspond to clusters with unidentified or empty compositional content. 3.2. Kinetics of melting of SnBiAg solders DSC studies show that under continuous heating, the melting of the B1–B5 bulks is a complex phenomenon that occurs at 413–518 K. The melting process consists of three separate endothermal events: the minor but sharp first effect R1, the massive main effect R2 and the wide final effect R3 (Fig. 7). All three phenomena are reversible. Subsequently, as the melted B1– B5 are cooled, their solidification processes exhibit the same three exothermal events, starting with R3, then R2 and, finally, R1. The DSC curves for all five solders have the same peaks with similar separation, but the temperature and transformation heat values depend on the particular chemical composition of each alloy. The characteristic temperatures and transformation heats of the individual effects are summarized in Table 4. According to the phase analysis, the R1 effect corresponds to the solid/liquid phase transition of the Sn/Bi eutectic. In the case of melting, its temperature is constant and compatible with the corresponding binary phase diagram (412.15 K [8]) for all 5 alloys. Due to the small transformation heat of R1, the content of the eventual liquid phases at 413 K is small, and it decreases from B1 to B5, reflecting the probable decrease of this eutectic phase. The melting of the remaining Sn and Bi from this eutectic starts immediately after R1, forming the main DSC peak, R2, which culminates at approximately the same temperature (Tem2 478 K) for all alloys. The thermal effect R2 can include the effect from the
ˇ P. Sebo et al. / Materials Science & Engineering A 571 (2013) 184–192
189
Fig. 6. (a) Microstructure of the Cu–B3–Cu joint made at 623 K after 1800 s and its phase map (b) constructed using trivariate correlation analysis. EDS maps of individual elements: (c) Cu, (d) Ag, (e) Sn, (f) Bi, (g) trivariate correlation diagram of Cu, Sn and Bi. Cu substrate is oriented vertically in the left part of the SEM image.
2.2 Heat power [ W/g]
Heat power [W/g]
2.2
2.0 heating R1
R2
R3
1.8
2.0 heating
1.8 cooling
cooling 350
400 450 500 550 Temperature [K]
600
350 400 450 500 550 600 650 Temperature [K]
Fig. 7. (a) Heating and cooling DSC curves of the Sn83.33Bi10Ag6.67 solder (B3) in the bulk form. (b) The relation between the heating and cooling DSC curves of B1 (blue dashed line), B3 (black solid line) and B5 (red dot-dashed line) solders in the bulk form. Heating and cooling rates were 7 10 K/min. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Ag3SnþSn-L eutectic reaction. But due to the low content of Ag in the alloys, the eventual heat of melting of this eutectic should be minor and not visible within the complex DSC peak R2. The transformation heat of the complex R2 peak decreases from B1 to B5, along with the content of the first eutectic. The remaining
Ag3Sn and Sn melt in the last broad thermal effect, R3. The end temperature of R3, Tem3, as well as the heat DHm3, dramatically increases in proportion to the increasing content of Ag in the alloy. The Tem3 temperatures are desirable in practice because they determine the completed liquid state; they represent the liquidus
ˇ P. Sebo et al. / Materials Science & Engineering A 571 (2013) 184–192
190
Table 4 Onset temperature Tx and transformation enthalpies DH of three melting (m) and solidification (s) effects (reactions R1, R2 and R3) of relevant solders B1–B5 in the bulk form with various amounts of silver in the as-prepared state.
causes small decrease of DHm1 and DHm2, but significant increase of DHm3 (except for B4). 3.3. Wetting
Solder
B1
B2
B3
B4
B5
Txm1 [K] Txm2 [K] Txm3 [K] Tem3 [K] DHm1 [J g 1] DHm2 [J g 1] DHm3 [J g 1] Txs1 [K] Txs2 [K] Txs3 [K] DHs1 [J g 1] DHs2 [J g 1] DHs3 [J g 1]
413.0 458.7 477.9 518.3 4.7 43.5 2.7 388.8 450.1 470.4 0.6 42.5 1.7
412.5 467.3 478.8 554.3 4.4 43.7 6.4 387.2 446.3 497.4 0.3 43.7 3.9
412.6 462.9 478.4 594.5 4.3 41.2 8.2 388.7 447.2 556.0 0.3 40.6 7.2
412.8 462.3 477.8 604.9 4.3 40.3 15.1 389.0 443.3 570.5 0.3 40.0 9.8
412.1 459.8 476.9 636.2 3.9 39.7 18.4 388.0 448.6 604.2 0.3 35.6 13.9
Fig. 8 shows the time dependence of the wetting angle on the copper substrate for the B1 solder at 553, 573 and 623 K. The wetting angle tends to decrease with increasing temperature and reaches its practically stable values after the first 600 s of exposition. Wetting rate at 623 K for the first 300 s in deoxidizing gas was for all solders around 0.03 deg./s. Wetting angle in air (with flux) was practically constant for the given solder. Fig. 9 shows the time dependence of the wetting angle on the copper substrate for solders at 623 K after 1800 s of wetting in both atmospheres: air (a) and deoxidizing gas N2 þ10% H2 (b). It is worthwhile to mention that these figures demonstrate opposite dependence of the wetting angle on the silver content of solder. When the wetting is performed in air, the wetting angle increases with the content of silver in the solder, but increasing the silver content lowers the wetting angle when wetting is performed in the N2 þ10% H2 gas. This difference may be due to the oxidation of silver, which generally frustrates wetting. These results can be used
Fig. 8. Time dependence of wetting angle of copper substrate by the B1 solder for 553, 573 and 623 K. The accuracy of wetting angle determination is 7 21.
of each solder. Due to the obscure DSC signal of the weak peak R3 in the heating regime (in contrast to the sharp onset of same R3 effect via cooling), both the temperature of liquidus and the enthalpy of the reaction are also determined from the cooling regime by the values Txs3 and DHs3. Both temperatures Tem3 and Txs3 correspond to the liquidus of the alloy; in fact, they differ by 32–62 K (see Table 4). Each difference Txs3 Tem3 is the supercooling necessary for nucleation prior to the start of the growth of the first crystalline phase which has been obtained at the actual cooling rate 10 K/min. Reflecting the rapid quenching, neither the phase composition nor the proportion between the elements in the main binary phases is in equilibrium in the ribbons. This fact is reflected in numerous minor phenomena in their first melting, i.e. several small subpeaks appear. All of these peaks are irreversible. After heating to 723 K, the solidification of the samples resembles that of the bulks and is the same in the next measuring cycle. The transformation (or sometimes the absence) of R1 peak on DSC curves of quenched ribbons allows one to say that precipitation of Bi particles was suppressed. Thus by varying temperature regimes of melt heat treatment before quenching/crystallization it becomes possible to influence the morphology of Bi-rich phase and solubility of bismuth in tin matrix. Summarizing the results from DSC it can be said that the increase of silver content (i) does not influence the values of Txm1, Txm2 and Txm3, but provides the drastic increase of Tem3; and (ii)
Fig. 9. Time dependence of wetting angle of copper substrate for various contents of silver in the solder at the temperature 623 K after 1800 s in atmosphere-air with flux (a) and in N2 þ 10% H2 gas (b). The accuracy of wetting angle determination is as shown in Fig. 8.
ˇ P. Sebo et al. / Materials Science & Engineering A 571 (2013) 184–192
for modifying the regimes of soldering: the utilization of the deoxidizing gas N2 þ10% H2 can shorten the process for the solders with high silver content.
191
Table 5 Shear strength of Cu-–solder–Cu joints with some lead-free solders. Sheer strength [MPa]
Substrate Reference
39 38 39 40 50 51 81
Cu Cu Cu Cu Cu Cu Cu
[10]
350 350 350
42 41 18 16 18
Cu Cu Cu Cu Cu
[10] [10] [17] [17] [17]
1800
350
15.8
Cu
[17]
1800
350
19.5
Cu
Actual paper
1800
350
21
Cu
1800
350
19.5
Cu
Solder
Reflow time [s]
Reflow temp. [1C]
96.5 Sn–3.5 Ag
120 3120 120 3120 120 3120 15
350–400
2
235
1800 1800 1800
3.4. Joint strength The shear strength of the joints was studied in terms of its dependence on two variables: the silver content of the solder and the influence of deoxidizing gas, as compared to the joints prepared in the air atmosphere. Fig. 10 shows the shear strength of the Cu–solder–Cu joints prepared at various temperatures over a period of 1800 s, with all relevant Sn100 xBi10Agx solders, containing x¼3.3–10 at% silver. For the temperatures up to 623 K the joints with B3 solder demonstrate the highest shear strength. For higher temperatures the increase in silver content above 7 at% Ag leads to a small increase in the shear strength of the joints. The shear
95.5 Sn–3.8 Ag–0.7 Cu 93.5 Sn–3.5 Ag–3 Bi 91.84 Sn–3.33 Ag–4.83 Bi 63 Sn–37 Pb 100 Sn 95 Sn–5 Sb 80 Sn–20 Sb 90.8 Sn–7.4 Sb–1.8 Cu 76.1 Sn–20.2 Sb–3.7 Cu 86.7 Sn–10 Bi–3.3 Ag 83.3 Sn–10 Bi–6.7 Ag 80 Sn–10 Bi–10 Ag
235 235 235
[16] [16] [16]
strength of the Cu–solder–Cu joints prepared in deoxidizing gas (Fig. 11) decreases with increasing silver concentration, except for the B3 solder. Shear strength of the joints for various reflow times and reflow temperatures for some alternative solders for comparison with Sn–Pb solder is shown in Table 5. It can be seen that the reflow time does not have a strong effect on the decrease of the shear strength of the joint. Reflow temperature has a more expressive effect on the decrease of the shear strength. Higher reflow temperature as well as reflow time may lead to the growth of the Cu–Sn intermetallic compounds which can have an effect on the strength of the joint. Another reason connected with the increasing of the reflow temperature and reflow time leading to the decrease of the joint strength can be the growth of the Ag3Sn needles. Fig. 10. Shear strength of Cu–solder–Cu joints prepared in atmosphere-air at given temperatures after 1800 s. The accuracy of shear strength determination is 7 0.5 MPa.
3.5. Proposal for optimized solder composition From the obtained results one could suggest the best chemistry for the studied materials i.e. solder of the Sn–10Bi–xAg type (xA(3, 10 wt%). Such solder based on tin could contain 4 wt% Bi and 3 wt% Ag. The solubility of Bi in Sn–Ag alloy is around 4 wt%. Higher Bi content would not be expected to contribute to additional strength through the solid solution mechanism [10]. The lowest wetting angle from all studied alloys with the lowest silver content possesses (B1) at 623 K. The wetting angle of this alloy is low even at 553 K ( 321). Another fact supporting the suggestion of the best solder of Sn–10Bi–xAg type is low content of silver. Vianco and Rejent [12] limited the amount of silver to be below 3.2 wt% to avoid the formation of large Ag3Sn platelets. Large Ag3Sn platelets induce serious degradation of joining mechanical properties. These do not appear in alloys with less than 3.2 wt% Ag content [12]. Shear strength of the Cu–solder–Cu joints is also the highest for B1 solder (3.3 wt% Ag) at 280–300 1C. The relatively low Ag content is also in line with the eventual cost of such solder alloy.
4. Conclusions Fig. 11. Shear strength of Cu–solder–Cu joints prepared in N2 þ 10% H2 gas at given temperatures after 1800 s. The accuracy of shear strength determination is as shown in Fig. 10.
The effects of silver in the Sn100 xBi10Agx (x¼3.3–10 at%) leadfree solder on the transition from the solid to liquid state, on the wetting of the copper substrate and on the shear strength of the
192
ˇ P. Sebo et al. / Materials Science & Engineering A 571 (2013) 184–192
Cu–solder–Cu joints were studied. The influence of the working atmosphere, either air with flux or deoxidizing N2 þ10% H2 gas, the microstructure of the as-prepared solder, and the interaction between the substrate and the solder after wetting at 623 K for 1800 s were also investigated. The obtained results can be expressed as follows:
1. The as-prepared solders consist of tin, bismuth and Ag3Sn phases only. The latter exists in the form of needles. The increase of silver content enhances the length and diameter of needles. 2. The melting and solidification phenomena of the solders consist of three separate events. Their presence confirms the existence of two types of eutectics: the binary Snþ Bi and the ternary eutectic Ag3SnþSn(Bi). While the temperature of the solidus remains constant for all five solders, the temperature of the liquidus dramatically increases with the content of Ag. 3. For wetting in the air, the wetting angle slightly increases with the silver content of the solder, whereas the silver lowers the wetting angle in N2 þ10% H2 gas. The wetting in both cases is very good (wetting angle o 301). 4. For the temperatures up to 623 K the joints with B3 solder demonstrate the highest shear strength. For higher temperatures the increase in silver content above 7 at% Ag brings a small increase in the shear strength of the joints. The shear strength of the joints prepared in the deoxidizing gas at 623 K for 1800 s decreases with increasing silver content. 5. At the interface, there is a layer of the Cu3Sn phase, which is adjacent to the copper substrate. The thickness of this layer grows with the increase of silver in the solder. 6. The methods of x-scan X-ray analysis and trivariate correlation analysis have been conveniently used for the detection of small amounts of individual phases and for their spatial localization at different distances from the Cu–solder joint. Based on the results obtained it can be recommended that silver content should not exceed 6 at% in Sn100 xBi10Agx solders.
Acknowledgments The authors gratefully acknowledge the financial support of the Slovak Grant Agency under the Contracts VEGA 2/6160/26 and 2/0111/11. This work was also supported by the Slovak Research and Development Agency under the Contracts APVV0102-07, APVV-0076-11, APVV-0492-11 and APVV-SK-UA-004309 and by CEX FUN-MAT. V.S. gratefully acknowledges the fellowship of the Slovak Academic Information Agency. This research contributes to the European COST Action MP0602 on Advanced Solder Materials for High Temperature Applications.
References [1] C.M. Miller, I.E. Anderson, J.F. Smith, J. Electron. Mater. 23 (1994) 595–601. [2] N. Sobczak, J. Sobczak, R. Nowak, A. Kudyba, P. Darlak, B. Mikulowski, A. Wojciechowski, J. Mater. Sci. 40 (2005) 2547–2551. [3] S.W. Chen, H.J. Wu, Y.C. Huang, W. Gierlotka, J. Alloys Compd. 497 (2010) 110–117. [4] Z. Li, Z. Cao, S. Knott, A. Mikula, Y. Du, Z. Qiao, CALPHAD: Comput. Coupl. Phase Diagr. Thermochem. 32 (2008) 152–163. [5] L. Zabdyr, G. Garze, CALPHAD: Comput. Coupl. Phase Diagr. Thermochem. 33 (2009) 187–191. [6] M. He, V.I. Acoff, J. Electron. Mater. 35 (2006) 2098–2106. [7] D. Witkin, J. Electron. Mater. 41 (2012) 190–203. [8] M.L. Huang, L. Wang, Metall. Mater. Trans. A 36 (2005) 1439–1446. [9] M. He, V.I. Acoff, J. Electron. Mater. 37 (2008) 288–299. [10] P.T. Vianco, J.A. Rejent, J. Electron. Mater. 28 (1999) 1138–1143. [11] D. Witkin, Mater. Sci. Eng. A 532 (2012) 212–220. [12] P.T. Vianco, J.A. Rejent, J. Electron. Mater. 28 (1999) 1127–1137. ˇ ˇ [13] P. Sebo, Z. Moser, P. Svec, D. Janicˇkovicˇ, E. Dobrocˇka, W. Gasior, J. Pstrus´, J. Alloys Compd. 480 (2009) 409–415. [14] F. Hofer, W. Grogger, G. Kothleitner, P. Warbichler, Ultramicroscopy 67 (1997) 83–103. ˇ ˇ ˇ [15] P. Svec, J. Hoˇsko, P. Svec Sr., I. Maˇtko, P. Sebo, D. Janicˇkovicˇ, in: J. Vajda, M. Weis (Eds.), Proceedings of the 16th International Conference on Applied Physics of Condensed Matter, APCOM, Mala Lucivna, Slovak University of Technology, Bratislava, June 16–18, 2010, pp. 292–297, ISBN 978-80-227-3307-6. [16] S.K. Kang, W.K. Choi, M.J. Yim, D.Y. Shih, J. Electron. Mater. 31 (2002) 1292–1303. ˇ ˇ [17] P. Sebo, P. Svec, D. Janicˇkovicˇ, E. Illekova´, Kovove Mater. 48 (2010) 353–359.