Growth and Selection of Intermetallic Species in Sn-Ag-Cu No-Pb Solder Systems based on Pad Metallurgies and Thermal Histones

L. P. Lehman, R. K. Kinyanjui, L. Zavalij, A. Zribi, E. J. Cotts Binghamton University

Binghamton, NY 13902-6000 Iplehman@binghamton.edu, ecotts@binghamton.edu

Abstract The adoption of Pb-free solders will affect manufacturing

processes and joint reliability for electronics packages. Since SAC solder has a higher melting temperature than eutectic Pb-Sn solder, higher processing temperatures will be required. The higher processing temperatures allow for increased solubility of some elements, such as Ni, within the solder melt. The specific selection of pad metallurgies can influence both the type of intermetallics that form at the pad interfaces, and the age related evolution of those intermetallics. In joints where both Nt and Cu pad metallurgies are present, the fast diffusing Cu species can allow the intermetallic compound, (Cu,Ni)6Sns to form at the opposite, Ni, interface. The absence of a Cu pad can starve the system of Cu and effectively stop the growth of this (Cu,Ni)& intermetallic compound.

The Ag component of the SAC solder can, under proper conditions, result in the growth of AgJSn plates in the solder melt before the Sn phase nucleates. This is controlled by the specific concentration of Ag in the solder, by the coolmg rate of the solder, and by the tendency of the solder joints to undercool by a typical 30°C before the Sn phase nucleates. The critical cooling rate for the onset of this phenomenon is between 3'Cisec. and 0.3"C/sec., within the range of normal reflow practices of the electronics industry.

Introduction Sn is the major constituent for most Pb-free solders of

interest (e.g. eutectic Sn-Cu, Sn-Ag, and Sn-Ag-Cu, where Sn is found at concentrations greater than 90 atomic percent). [l- 101 Relatively small amounts of additional elements such as Cu or Ag are added to Sn to reduce the melting point of the solder alloy, and improve its mechanical and wetting properties. These constituents dramatically affect the mechanical properties of the solder, and possibly the reliability as well. Therefore it is important to understand the evolution of the distribution of these constituents during reflow and long term aging, including changes in the morphology of the intermetallic compounds (IMC's) formed along with Sn by these elements. In Sn-Ag-Cu alloys, Cu&n5 and Ag3Sn form in the equilibrium state, with distinct morphologies, platelets for AglSn and hollow, hexagonal rods for Cu6Sn5. [l-71 We consider the evolution of IMC's during reflow, particularly Ag3Sn, including cooling to the solid state, and during long term aging.

1 Alloys selected for coohg rate study.

b Sn-4.0Ag -0.5Cu,

0 Sn-3.5Ag-O.9Cu,

+ Sn-2.6Ag-0.9Cu, (SAC2)

A Sn-3.5Ag

(SACO)

(SAC1)

Sn

Sn-37Pb

P

i! s U) U)

0 0.5 1.0 Sn Mass% Cu

Figure 1. A portion of a calculated Liquidus Projection Temary phase Diagram based on data by Moon, et.al., [I]. Plotted as symbols on the diagram are 5 of the six alloys used in our cooling rate study.

300 x . A@n ~. ' n . . Liquidus 5- " .

'0 W

g' * 200 2

G . " : , *

100

0 2.5 5 .O

(Sn 0.9Cu) + x wt6 Ag Figure 2. This drawing shows a portion of a quasi-hinary phase diagram that represents a constant copper concentration section of the ternary Sn-Ag-Cu diagram.

Both the composition, and the cooling rate from the melt, of Sn-Ag-Cu alloys profoundly affect their morphologies. Within this context, it is interesting to note that NEMI has selected a relatively Ag rich composition Sn-3.9Ag-O.6Cu,

0-7803-7991 -5/03/$17.00 02003 IEEE 1215 2003 Electronic Components and Technology Conference

while the Japan Electronics and Information Technology Industries Association (JEITIA) has proposed alloys less rich in Ag: Sn-3.0Ag-OSCu and Sn-3.5Ag-0.7Cu. As indicated in Fig. 1, the primary crystallization product is changed from Ag3Sn in the case of the NEMI alloy [Sn-3.9Ag-0.6Cu], to Sn in the case of the JEITIIA alloys: So-3.0Ag-OSCu, and Sn- 3.5Ag-0.7Cu.[1,7] This difference is signifcant because of the large undercooling of the Sn phase observed for these alloys, which allows the Ag3Sn compound to grow to relatively large length scales (tens of microns) in Ag rich SAC alloys (e.g. SACO, Fig. 1) before the nucleation and growth of solid Sn. Some insight into this phenomenon is provided by a sketch of the metastable Ag3Sn liquidus line (Fig. 2).[1,7] This sketch is produced by a simple extrapolation below the eutectic temperature. The suppression of the nucleation and growth of Sn allows the growth of Ag3Sn to extend to a larger fraction of the sample, to larger dimensions, as illustrated in Fig. 3. [1,7] The extent of the growth of Ag3Sn plates depends upon the time below the liquidus before the nucleation of the Sn phase, which depends monotonically on the inverse cooling rate, as reflected in Fig. 3 for the Ag rich SAC alloy, SACO. We examined such phenomena in control samples, isolated Sn-Ag-Cu solder balls of various compositions, prepared under highly controlled conditions. We also studied the evolution of the microstructure of So-Ag-Cu balls in Sn-Ag-Cu solder joints prepared by commercial vendors. We found similar phenomena in both systems, with strong dependences on c o o h g rate from the melt, and upon solder composition.

Figure 3. Scanning electron microscopy back-scattered electron- composition images show SACO, Sn-4.0Ag-OSCn wt%, cooled from 250°C at different cooling rates.

atmosphere. Once melted the ingot is tumed over and melted again. Finally the ingot is sectioned and the microstructure is examined to verify the uniformity of the fmished alloy. The SACO, SnPb and Sn3.5Ag alloy were all obtained from Indium Corporation. From these initial alloys, small samples were prepared for programmed melting and solidification within a Differential Scanning Calorimeter (DSC). Each sample is heated at 0.33"C / s up to the peak temperature of 250% Then the samples were cooled at one of two rates, 0 . 1 T /sand l0C/s.

Alloy Name I Weight Percent I Atomic Percent

Sn SnPb I Sn-37Pb I Sn-25Ph

Table 1. Control samples. solder alloy compositions. The alloy names are arbitrary and provided as a convenience.

The melting event was recorded during the heating phase of these solder alloys and was characterized by the temperature of the onset of melting. The solidification event is recorded during the cooling phase. The temperature at the onset of solidifcation characterizes the solidification event. It is typical to see a larger super cooling of the melt at larger cooling rates.

A commercial vendor provided the second sets of samples. These samples consisted of a silicon chip attached to an organic chip carrier with a near eutectic Sn-Ag-Cu (SAC) solder. The upper pad, on the silicon side, was finished with a Ni layer. The lower pad was fmished with either a copper OSP bump or a copper bump fmished with a Electroless Nickel Immersion Gold (ENIG) fmish. The solder was reflowed at a peak temperature of 235°C and 60 seconds above the liquidus. The nominal reflow profile temperature decreases at approximately 2.4Wsec until the temperature is at least as low as 180°C. After reflow some of these samples were aged at 125°C for up to 1000 hours.

After any heat treatment, samples were sectioned, mounted and polished for metallographic analysis. Metallography included optical and scanning electroo microscopy along with energy dispersive spectroscopy (EDS) and electron probe microanalysis (EPMA). Optical metallography examinations were performed with a Nikon metallograph with digital high-resolution color camera and integrated image processing. Finally the samples were taken for electron microscopy on an SEM. Using composition mode Back Scatter Electron Imaging (Compo BSEI) and spot mode EDS, be identified, ne Same general metallography route was used for the vendor-supplied samples with the addition of an EPMA step to more quantitative[y identify the composition ofsome mcS.

Experimental Procedures A group of solder alloys was Prepared as Control samples

to investigate the effects of different cooling rates on micrOstructure and sdidification temperature. The alloys tested are shown in Table 1. Sn3.5Ag, Sn and SnPb are included in an effort to provide comparative data to the more complex SAC alloys.

One-gram ingots of SAC1 and SAC2 were prepared by arc melting the indicated weight fractions of the pure metal constituents together within an oxygen gettered argon

of the mcs

1216 2003 Electronic Components and Technology Conference

Experimental Results and Discussion Observation of the Microstructure of Control Samples The effect of cooling rate on the microstructure was

examined in control samples of SAC solders (Figures 4 - 7). In addition to varying the cooling rate, the temperahue at the onset of initial melting was measured, and the temperature at the onset of tin phase solidification was also measured (Table 2). Figure 1, above, shows a portion of the Sn-Ag-Cu ternary phase diagram. Five of the alloys investigated in this study are plotted on that diagram. Metallography was performed on many of these samples and the micrographs can be compared with the measured cooling rate and solidification temperatures.

Alloy 1

Rate-

Melt T. Solid T Solid T Under- Under- cool AT cool AT

0.3OC/s I O.l"C/ l.O°C/ 1 O.l"C/s l.O"C/s

g I 221 I 199.1 I 185.7 I 21.9 I 35.3 216 I 185.6 I 182 I 39.4 I 34

theonsetofmelting and the onset of tin phase solidification, dnived fiom DSC runs on the respective alloys.

top. and at ?I.l'C/sec,-at the bottom. Ovals on these bright field optical micrographs indicate primary Ag;Sn plates precipitated in the melt and the rectangles are primary C Q S ~ ~ rods precipitated in the melt.

Sn-based alloys endure significant under cooling before the onset of tin phase nucleation (see Table 2). However, once crystallization ensues, the growth rate of the solid tin phase is very rapid. The growth rate is documented at 20cm/sec at only 10°C of undercooling, [ll]. This means that a typical 30mil BGA ball should be fully solidified in less than 4 milliseconds after the onset of solidification.

Figure 5 . SAC2 cooled at I.O"C/sec, at top, and at O.l"CIsec, at the bottom. No primary Ag;Sn plates are present in either bright field optical micrograph above. Several primary CySul rods precipitated in the melt and are visible in both images.

A consequence of the rapid growth of the tin phase is that all other nucleation and growth is suppressed during and for an interval after the onset of tin phase solidification. This is in part because the tin phase solidification reaction is exothermic. As the tin phase grows, it raises the temperature of the joint, even if only briefly, makiig other nucleation events less likely. Finally, the tin phase grows so rapidly that no significant time period remains for growth of any other nucleated phases.

1217 2003 Electronic Components and Technology Conference

The effect of cooling rate on microstructure was examined for a near eutectic Sn-3.5Ag-0.9Cu wt?? alloy (SACl, see Figure 4). Larger primaryAg3Sn precipitates form at the lower cooling rate of 0.1 C/s. But SACl has a relatively large amount of Cu, so we begin to see a large number of primary C%SnS rods, especially in the O.l0C/sec cooling rate sample. Note that both these samples suffered similar amounts of undercooling, 30.4OC at O.lOC/sec and 34OC at l.O°C/sec, so the time helow the eutectic temperature is primarily responsible for the microstructure differences manifest here.

Figure 6. Sn3.5Ag solder cooled at I.O"C/sec. at top. and at O.lT/sec, at the bottom. The large precipitates visible in these bright field optical micrographs are primary Ag,Sn plates precipitated in the melt.

The SAC2 alloy, Sn-2.6Ag-O.9Cu, is hypoeutectic in silver. and is not expected to produce any primary Ag,Sn plates (Figure 5) . No primary Ag$n plates are present within our cross sections at either cooling rate. This is an even more remarkable observation given that our l.O"C/sec SAC2 sample undercooled 54.5OC. The SAC2 alloy is near eutectic in copper and we find many primary C%SnS rods at both cooling rates. The rods are more numerous, and smaller, at 1 .O°C/sec than at O.l"C/sec.

The Sn3.5Ag solder alloy was used as one of three reference standards for our DSC study, since it is a true

eutectic alloy. The Sn3.5Ag solder provides a rich resource for microstructural studies (Figure 6). As discussed above, primary Ag3Sn plates form as result of entering the metastable AglSn liquidus region of the phase diagram. As expected, larger Ag3Sn plates grow at the lower cooling rate (Figure 6).

field optical image, contrast is derived from the birefringent property of tin. Not more than four crystals are visible in this image.

The extremely rapid growth of the crystalline Sn phase (> 20 cm/s) fiom the undercooled Sn melt [ 111 not only limits the nucleation and growth of other phases, it suppresses nucleation of other Sn crystals. As a result, very few grains can be expected within a typical SAC solder ball (Figure 7). Some solder balls may even be formed as single crystals, at least before they experience any deformation. The paucity of separate Sn crystals is illustrated in Fig. 7, a cross polarizer image of the same sample seen in Figure 4. The cross polarizer image allows the establishment of the locations of the grain boundaries (since tin is birefringent, variations in crystal orientation provide optical contrast when viewed with crossed polarizers). Within Figure 7> only 4 separate grains, or crystal orientations, are revealed. Furthermore, all of these crystals are apparently radiating from a common nucleation site. This structural evidence illustrates the suppression of the formation other nucleation sites by the rapidly growing Sn crystals.

Observation of the Microstructure of Commercially Prepared SAC Solder Joints

The microstructural evolution phenomena that were observed in the control samples were also observed in the vendor samples. Figure 8 shows a group of four images showing SAC joints joining a silicon chip to a substrate. These are all very small joints, processed through a commercial reflow profile, and each shows, in proportion, relatively large primary Ag,Sn plates. The nominal reflow profile has a cooling rate through the melting temperature of about 2.4'C/sec. At this rate one might reasonably expect small or nonexistent primary Ag,Sn plates. But the actual cooling rate of the solder mass can be expected to be slower

1218 2003 Electronic Components and Technology Conference

than the programmed oven profile. The cooling rate of the actual solder mass may be retarded by the shielding effect of the over layers and the thermal mass of the adjoiniig components. The retarding effect will depend on the thermal mass and thermal conductivity of the adjoiniig components and PCB. In any event, as the presence of these plates has been shown by others to affect the mechanical properties of the solder [3-51, thus their presence in these commercial SAC solder joints of nominal high cooling rate may be of interest to those interested in reliability.

Figure 8. SEI micrographs oi near cLlectic SNA& Cu snlller ioints supplied by a vendor. Lower pad is Ni metallur~~nn a silicon-chip. Upper pad is as designated within the images.

Within the vendor-supplied joints, we have SAC solder on either CdOSP or Ni/Au pads. In each case a Ni UBM pad is also present. This difference turns out to be significant; it is emphasized that in one case we have SAC solder with both a Ni metallization, and a CdOSP metallization, while in the other case the SAC solder is only in contact with two Ni metallizations.

Let us consider frst the case when soldering to a CdOSP pad. In the as reflowed condition, Cu& forms at the Cu interface as expected. The morphology is the characterized by a scallop shaped intermetallic layer growing off of the Cu pad surface into the solder (Figure 9). On the Ni interface the IMC morphology is much more complex. The surface is decorated with a fme dispersion of precipitates, among them are found Cu6Sns, Ni,Sn and (Cu,Ni)&,. Many of these precipitates are nearly too small to probe with EPMA. Copper containing intermetallics would not normally be expected at such a Ni interface in a Sn-Pb solder, but as has been previously reported [SI, Cu in the SAC solder can combme with Ni and Sn to form the temary compound. Cu diffuses at very high rates in the Sn matrix, so that the (Cu,Ni)& compound can grow relatively quickly.

Figure 9. SEI micrograph of near eutectic Sn/Ag/Cu solder joint in the as reflowed condition. Cu pad (1). CySns (2), Ag& (3). fine dispersion of CySns. Ni,S& and (CuNi)&nS precipitates (4) and Ni UBM layer(5).

After 1000 hours at 125OC, the microstructure of the sample originally portrayed in Figure 9 has matured and coarsened as seen in Figure 10. The CkSn, on the copper pad has thickened substantially and has lost much of its original scalloped shape. Instead the layer has more uniform thickness, and has several large IMCs growing out of the CySn, layer. The IMCs on the Ni pad have developed a continuous thin layer with many spky precipitates attached to it or protruding through it. All of these precipitates are found to be (Cu,Ni)aSns. No Ni3S% can now be found. The primary Ag3Sn plates are now courser, thicker and less smooth. Again the plates are quite large relative to the solder ball size.

Figure 10. SEI micrograph of near eutectic Sn/Ag/Cu solder joint aged at 1 2 5 T for 1000 hrs. Cu pad (l), CySu, (2), Ag& (3), (Cn,Ni),Su, (4) and Ni UBM layer (5) .

1219 2003 Electronic Components and Technology Conference

In contrast, consider the case when both the pad interfaces in the solder joint are composed of nickel. Figure 11 shows that after reflow, the SAC solder has formed (Cu,Ni)&n~ IMCs at both pad interfaces. The original gold coating of the upper pad has dissolved into the solder and precipitated out throughout the bulk of the solder as f i e (Au,Ni)Sn, intermetallics. The increased solubility of Ni at the higher reflow temperatures needed for SAC alloys helps explain the presence of Ni withii the bulk solder. A very large flat plate of Ag& can be seen which nearly cuts the ball in half.

Figure 11. SEI micrograph of near eutectic Sn/Ag/Cu solder joint in the as reflowed condition. Ni pad (l), (Cu.Ni)& (2), (Au,Ni)S% (3), Ag& (4) and Ni UBM layer (5) .

Figure 12. SEI micrograph of near eutectic Sn/Ag/Cu solder joint aged at 125°C for 1000 Ius. Ni pad ( I ) , (Cu.Ni)& (2). (AuNi)Sn, (3). Ag,Sn (4) and Ni UBM layer (5) .

After 1000 hours at 125"C, the thicknesses of the Cn-Ni- Sn IMCs at both pad interfaces are largely unchanged (Figure 12). They have not grown thicker or courser to any large extent. This too is consistent with our previous observations that the growth of the (Cu>Ni)& abates upon its consumption of all of the Cu in the SAC alloy. In the case of the samples with a CdOSP metallization interface, a constant supply of Cu was available, and the growth of (Cn,Ni)& could continue at the Ni metallization. Without such a source of Cu atoms, the growth is limited. It is also observed that the (Au,Ni)Sn, IMCs have coarsened significantly and many have lost their characteristic as sections of flat plate like structures. These (An,Ni)S& IMCs are observed to have formed structures like pearls on a string.

Figure 13. SEI micrograph of near eutectic SdAgICu solder jomt aged at 150T for 1000 hrs.

Similar observations of the importance of the chemical identity of the pad metallization [CdOSP or Nil to the growth of intermetallic compounds on the chip side (Ni) were made for samples prepared by a different vendor (Figure 13). When Ni is present at both the upper and lower pads of the solder joint, no source of additional copper is available to the solder during aging. We have seen in the pa t , the IMC species that is most kinetically favored tends to he the one that forms fmt. Since (Cu,Ni)&hs tends to grow faster than (Cu, Ni)$n+ it tends to grow fust at the pad interfaces. However, further growth is inhibited due to the f ~ t e supply of copper within the SAC solder joint.

Conclusions We observed the influence of SAC solder composition,

cooling rate 6om the melt, and solder joint pad metallnrgies on the evolution of the microstructure of SAC Pb free solder joints. As the morphology and distribution of intermetallic compounds in SAC solder joints affects the mechanical properties of these joints [3-51, this evolution must be better understood in order to ensure Pb free solder joint reliability.

Acknowledgments Discussions with Peter Borgesen and Donald Henderson,

and support from NSF DM0218129 are gratefully acknowledged.

1220 2003 Electronic Componenis and Technology Conference

References 1. K.-W. Moon, W. J. Boeninger, U. R. Kattner, F. S.

Biancaniello, and C. A. Handwerker “Experimental and Thermodynamic Assessment of Sn-Ag-Cu Solder Alloys” J. Electron. Mater. 29, (2000), pp. 1122.

2. K. N. Tu and K. Zeng, “Reliability Issues of Pb-Free solder Joints in Electronic Packaging Technology” 2002 Electronic Components and Technologv Conference, Seattle, WA.

3 . K. S. Kim, S. H. Huh, and K. Suganuma, “Effects of Cooling Speed on Microstructure and Tensile Properties of Sn-Ag-Cu Alloys” Materials Science and Engineering A333 106 (2002).

4. K. Wu, N. Wade, J. Cui, and K. Miyahara, “Microstructural Effect of the Creep Strength of a Sn- 3.5%Ag Solder Alloy” Journal of Electronic Materials 32,5 (2003).

5 . D. K. Joo and Jm Yu, ‘Zffects of Microstructure on the Creep Properties of the Lead-free Sn-3.5Ag-Cu Solders” 2002 Electronic Components and Technologv Conference, Seattle, WA.

6. A. Schubert, R. Dudek, H. Walter. E. Jung, A. Gollhardt, B. Michel, and H. Reicbl, “Reliability Assessment of Flip- Chip Assemblies with Lead-Free Solder Joints” 2002 Electronic Components and Technologv Conference, Seattle, WA.

7 . Donald W. Henderson, Timothy Gosselm, Amit Sarkhel, Sung K. Kang, Won-Kyoung Choi, Da-Yuan Shih, Charles Goldsmith and Karl J. Puttlitz “Ag3Sn plate formation in the solidification of near temary eutectic Sn- Ag-Cu alloys” Journal of Materials Research 17, 2775 (2002).

8. “The Growth of Intermetallic Compounds at Sn-Ag-Cu SolderKu and Sn-Ag-Cu Soldermi Interfaces and the Associated Evolution of the Solder Microstructure,” A. Zribi, A. Clark, L. Zavalij, P. Borgesen, E. J. Cons, Journal ofElectronic Materials 30, 1157 (2001).

9. “Effects of Reflow Conditions on the Formation of Au-Ni- Sn Compounds at the Interface of Au-Pb-Sn and Au-Sn Solder Joints with Ni Substrate,” Robert K. Kinyanjui, Anis Zribi and Eric J. Cotts, IEEE 2002 Electronic Components and Technology Conference, pp. 161 (2002).

10. “The Kinetics of Formation of Ternary Intermetallic Alloys in Pb-Sn and Cu-Ag-Sn Pb-free Electronic Joints,” A. Zribi, L. Zavalij, P. Borgesen, A. Primavera, G. Westby, and E. J. Cons, IEEE 2001 Electronic Components and Technology Conference, pp.687 (2001).

11. “The Growth Rate of Dendrite in Super Cooled Tin,” A. Rosenberg and W.C. Winegard, Acta Metallurgica, Vol. 2 (1954), pp 342-343

1221 2003 Electronic Components and Technology Conference