Effects of Ni and Cu Additive on Electromigration in Sn Solder Lines
S. C. Hsu and C. Y. Liu
Dept. of Chemical Engineering and Materials Engineering,
National Central University, Chung-Li, Taiwan, R. O. C.
Abstract:
The effect of Cu content in Sn(Cu) alloys on the interfacial reaction between Ni thin film and Sn(Cu) alloys has investigated. We have found that the variation of Cu content has a strong influence on the spalling of the Ni thin film. With small Cu additives in the Sn, spalling was deferred to longer reflowing time. When the Cu content increased to about 1.0wt.%, a layer of Cu-Sn compound formed on the Ni thin film and no spalling was observed after 20 minutes reflowing. The possible mechanism of spalling deferring is proposed. A Cu flux from the solder to the interface compensated the ripening flux of the semi-spherical compound grains, therefore, spalling was retarded. The driving force of the Cu flux was attributed to the reduction of Cu solubility due to the presence of Ni at the interface of the Ni thin film. The Cu flux from solder to the interface is calculated to be in the same order with the ripening flux of Cu6Sn5 compound grains, which confirms the proposed mechanism of spalling deferring. For the Sn(Cu) alloys having Cu content over 1.0wt.%, the Cu-Sn compound layer grew so fast that the surface of the interfacial compound layer was free of Ni. There was no Cu flux to compensate the ripening flux, therefore, the ripening flux dominated and spalling occurred after short reflowing time.
Key words: Pb-free solder, Thin film UBM, interfacial reaction, Cu diffusivity.
The corresponding author's e-mail address is .
Introduction:
In the advanced IC (Integrated Circuits) packaging, the thin-film metallization is commonly used as UBM (Under Bump Metallization) on the chip side to bond small C4 solders balls. [1] Yet, the fast consumption of the limited metal thin layer due to the reaction with Sn-bearing solders is always a concern for the reliability of solder joints.[2,3,4] Tri-layer Cu/Ni(V)/Al UBM has successfully used as flip-chip UBM for many years. No spalling was observed after 40 minutes reflowing, when it reacted with the eutectic SnPb solder.[5] However, when the eutectic SnPb solder was replaced with Sn-rich Pb-free solders, spalling in tri-layer Cu/Ni(V)/Al UBM was found after several reflows.[6] Therefore, it is a challenging task to develop a thin film UBM without spalling for flip-chip Pb-free solder bumps.
In previous study by Wang, the spalling of Ni thin film was found to be prevented, if a Cu reservoir was introduced into the structure of C4 (Controlled Collapse Chip Connections) solder joints during the soldering reaction.[7] The Cu reservoir could be the incorporated Cu particles inside the solder or the Cu pad in the package side. The major role of Cu reservoir was a source to provide Cu atoms to form a layer Cu-Sn compound on Ni thin film, which was believed to be the key for the prevention of spalling.
Some reports have demonstrated that the small amount of Cu additives can improve the Pb-free solders’ mechanical properties, such as, fatigue, creep, and tensile strength.[8,9,10] So, it seems inviting us to understand the role of Cu additives in the interfacial reaction between Cu-bearing Pb-free solders and Ni thin film metallization. It will also be interesting to study that will the limited Cu additives in Cu-bearing Pb-free solders have the same effect as the Cu reservoir. In this study, nine dilute Sn(Cu) were used to study the effect of the variation of Cu content on the interfacial reaction between Sn(Cu) alloys and the Ni thin film.
Experimental Procedures:
The Ni/Ti metallization was deposited on an oxidized Si wafer by E-beam evaporation process. The Ti layer is for the adhesion purpose and the Ni layer is simulated to the wettable metal bond pad. The thickness of Ti and Ni are 500 Å and 2000 Å, respectively. 99.99% purity of Sn and Cu purchased from Alfa AESAR Inc. were used to prepare nine SnCu alloys, which range 0, 0.2, 0.6, 1, 1.4, 1.8, 2.2, 2.6, and 3.0. To produce accurate composition of SnCu alloys, weight Sn and Cu in desired weight percentage by a digital balance (SCALTEC SBC31), then, place them into evacuated and sealed quartz tubes. Heat quartz tubes to a constant temperature of 900°C in a furnace for 150 hours. During the heat treatment, swirled quartz tubes to evenly mix the metal elements. After the heat treatment, quartz tubes were quickly quenched by water cooling. Cut 2mg of each of nine alloys and melt them under the flux ambient to produce a spherical solder ball.
The reflowing process steps are in following. Right amount of flux was poured into a stainless beaker. Put the beaker on a hot plate and maintain the flux at a constant temperature of 250°C. The Si wafer coated with Ni/Ti metallization was cut into 2x2 cm square pieces. Those Si pieces were immersed into the heated flux for 30 seconds to clean the Ni surface. Then, 2mg solder balls were placed on the top of Si pieces. As solder ball touched Ni surface, the molten solder started spreading and reacting with Ni/Ti metallization. Five different reaction times were carried out for each composition, which are 30sec, 1min, 5min, 10min, and 20min.
After certain reflowing time, Si chips were removed from flux and cooling in the air for one minute, then they were cleaned by alcohol in an ultrasonic machine. To perform the SEM examination, mount reflowed-samples by the epoxy resin. First, the epoxy mounted samples were cut by a diamond saw and abraded by coarse sandpapers. Then, samples were finished by 0.3μm Al2O3 powder polishing. Before SEM observation, samples were slightly etched by light acid (10%H2SO4) for 10 seconds to delineate the morphology of the compound layer.
Results:
Fig.1 shows the cross-sectional SEM images of interfacial reactions between low Cu-content Sn(Cu) alloys and Ni thin film. As seen in Fig. 1(a) and (b), the original Ni thin film could not be observed at the interface after 30 seconds reflowing. We believe that the Ni thin film was completely consumed by the formation of the Ni-Sn compound layer and, then, spalled into the molten solder. It is striking that, as shown in Fig.1(c), a layer of interfacial compound was found to reside at the interface, as Ni thin film reacted with Sn0.6Cu for 30 seconds. It implied that 0.6wt.% of Cu content could defer the on-set time of spalling to longer reaction time. Using EDX analysis, the compound layer was determined to be the Ni3Sn4 compound phase dissolved with Cu. The Cu in (Ni,Cu)3Sn4 intermetallic compound layer should attribute to the Cu content in the molten solder. Yet, we found that spalling occurred after 5 minutes reflowing, as shown in Fig.1(d). The big separated Ni-Sn grains on the Ti surface indicated the occurrence of the ripening process.
As the Cu content in Sn(Cu) alloy increased to 1.0 wt.%, remarkably, we found that spalling could not be observed even after 20 minutes reflowing! Fig.2 is the SEM cross-sectional images of interfacial reaction between Sn1.0Cu and Ni thin film for different reflowing times. The EDX analysis shows that a scallop-like Cu6Sn5 compound layer covered the original Ni thin film after 30 seconds reflowing, as seen in Fig.2 (a). The substantial portion of the initial Ni thin film remained un-reacted. With prolonged reflowing, the growth of the interfacial compound layer was very sluggish. After 20 minute reflowing, EDX results show that a thin Ni thin film layer still can be observed and covered by a (Cu, Ni)6Sn5 intermetallic compound layer. Interestingly, the morphology and size of the scallop-like compound grains show no big change during the reflowing process, as seen in Fig.2. It indicates that ripening did not occur in compound grains.
Fig.3 is the cross-sectional SEM images of interfaces between Sn1.8Cu/ Ni thin film for different reflowing times. Spalling did not occur in the initial 5 minutes reflowing. As the reflowing was prolonged to 10 minutes, the ripening process occurred and the resultant big compound grains started departing from the Ti surface, as seen in Fig.3 (c). Fig.4 is the SEM cross-sectional images of the interfacial reaction between Sn3.0Cu and Ni thin film. Unlike the scallop-type compound layer of other Sn(Cu) alloys after 30 seconds reflowing, the morphology of interfacial layer of Sn3.0Cu is relatively flat. In addition, the primary Cu6Sn5 compound phase was observed in the solder, as indicated by black arrows in Fig.4 (a) and (b). After one minute reflowing, the primary Cu6Sn5 compound precipitates inside the solder were greatly reduced. We believe that the primary Cu6Sn5 compound particles in the solder were quickly expensed by the growth of the compound layer. Coalescence or ripening between primary Cu-Sn compound particles and interfacial compound layer were the two possible processes for the consumption of the primary Cu6Sn5 particles. After 5 minutes reflowing, big chunk of compound grains were found at the interface and about departing from the surface of Ti layer, as seen in Fig.4 (c). The average thickness of compound layer is estimated to be about 4.2μm. In the Fig.4 (d), clearly, we can observe that the chunky Cu6Sn5 compound grains are floating upward in the solder after 20 minutes of reflowing.
Table one is a summary for the on-set time of spalling versus different Sn(Cu) alloys. The on-set time of spalling strongly depends on the Cu concentration in the Sn(Cu) alloys. The initial small amount of Cu additives deferred the on-set time of spalling to longer reflowing time. Amazingly, no spalling can be found for the Sn1.0Cu alloy. As Cu content in the Sn(Cu) alloys are over 1.0 wt. %, the spalling backed to occur. Note that in region of high Cu-content alloys, the more Cu additives, the earlier spalling occurred.
Discussions:
(1) Review of spalling phenomenon
The typical picture of the spalling process in the Cu thin film is illustrated in Fig.5. First, the Cu thin film was completely converted to a scallop-appearing compound layer after a short soldering reaction, as shown in Fig.5 (a). Then, the ripening process occurred among scallop-like compound grains; the bigger scallop-like compound grains grew at the expense of smaller scallop-like compound grains. [2] Due to the mass conservation, the incremental of the bigger grain in radius, dr1, was smaller than the decrease of smaller compound grain in radius, dr2. Thus, a small portion of the Ti surface was exposed to the molten solder, as shown in Fig.5 (b). As a result, unstable triple junctions of three interfacial energies, i.e., γTi/Solder, γTi/compound, and γSolder/compound, occurred, as indicated by three arrows in Fig.5 (c). The unbalanced interfacial energies at the triple junctions led to the shape transformation from semi-spherical grains to spherical grains, as seen in the dash circle in Fig.5 (c). The spherical compound grains had very little contact with the Ti surface, so, they tend to float into the molten solder due to the gravity effect. The resultant high interfacial energy between the Ti surface and the molten solder caused the dewetting of the molten solder on the Ti surface, as shown in Fig. 5 (d).
From the review above, we recognize that the ripening process, which caused the exposure of the Ti surface to the molten solder, is the key step for the occurrence of spalling. In the present study, we found that the Cu additives in Sn could defer or prevent the spalling of the Ni thin film during the soldering reaction. So, we believe that the Cu additives in solders should have a strong influence on the repining process in the interfacial compound layer. In the following, we will first discuss the influence of Cu additives on the interfacial reaction between Sn(Cu) alloys and Ni thin film and, then, lead to the possible mechanism of the spalling deferring.
(2) Formation of interfacial compounds on Ni thin film
SEM cross-sectional results have shown that a layer of Cu-Sn compound formed on the Ni thin film, as the Ni thin film reacted with Sn(Cu) alloys having Cu content over 0.6wt.%. A similar finding on the bulk Ni substrate was reported by Kao.[11,12] It is plausible that Kao explained this finding well by using the diffusion paths in the Sn-Cu-Ni isotherm. Here, we found that it also can be interpreted by the concept of the Cu solubility limit in Sn. Fig.6 is the enlarged Sn corner of the Sn-Cu-Ni ternary phase diagram calculated by Zeng et. al. [13,14]. According to this diagram, the solubility limit of Cu in Sn is estimated to be about 1.1 wt. %, which is close to the experimental value, 1.5 wt. %, reported from Steen.[15] The slight discrepancy between experimental and calculation values could be due to the difference in temperature. It is of interesting that with little additional Ni in Sn, as indicated by a black dot in Fig.6, the Cu solubility limit is greatly reduced to about 0.6 wt%.
In the very early stage of the soldering reaction, Ni would dissolve into the local region of the molten solder near the interface of Ni thin film before the formation of the Ni-Sn compound. As pointed out earlier, the presence of little Ni dissolution in the liquid Sn would cause the reduction of the Cu solubility limit to be around 0.6 wt.% at 250 °C. Therefore, for Sn(Cu) alloys having the Cu content over 0.6 wt.%, the Cu concentration near the Ni surface became over-saturated. The access Cu would instantaneously precipitate out as a Cu-Sn compound layer on the Ni interface to meet the Cu solubility limit under the presence of Ni. Consequently, an instant Cu concentration difference would be established between the Ni interface and the bulk of molten Sn(Cu) alloys and generate a Cu flux from solder to Ni interface. As long as sufficient Ni dissolution could be maintained near the interface of the Ni side, the Cu atomic flux would exist and the Cu-Sn compound layer would grow. Yet, during the formation of Cu-Sn compound layer on Ni side, the Ni dissolution could be incorporated into the formation of the Cu-Sn compound and formed a Cu-Ni-Sn ternary compound. The decreasing Ni dissolution at the Ni surface could diminish the Cu atomic flux toward the Ni interface. The Cu flux would resume, if the Ni dissolution could be replenished by the diffusion of Ni through the Cu-Sn compound layer from Ni thin film. So, we believe that Cu flux from solder to Ni interface could likely to be an on-and-off process rather than a continuous process. It really depended on the Ni diffusion rate in the Cu-Sn compound and the thickness of Cu-Sn compound on the Ni side.