Effect of Electrodeposition Conditions on Kirkendall Void Formation ...

<strong>Effect</strong> <strong>of</strong> <strong>Electrodeposition</strong> <strong>Conditions</strong> on Kirkendall Void Formationbetween Electrodeposited Cu Film and Sn-3.5Ag SolderJong Yeon Kim, Jin Yu, and Taek Young Lee*Center for Electronic Packaging MaterialsDepartment <strong>of</strong> Materials Science and Engineering, KAIST373-1 Guseong-dong, Yuseong-gu, Daejeon 305-701, Korea*Department <strong>of</strong> Materials Engineering, Hanbat National University,San 16-1, Dukmyung-dong, Yuseong-gu, Daejeon, 305-719, Koreamax2020@kaist.ac.kr, tel: 82-42-869-4274, fax: 82-42-869-8840AbstractSn-3.5Ag solder on Cu foil and electrodeposited filmswere reflowed at 260°C for 1min. After reflow process, allspecimens showed typical scallop shape <strong>of</strong> Cu 6 Sn 5 at theCu/solder interface. After thermal aging at 150°C, theformation and growth <strong>of</strong> Kirkendall void and two IMCs,Cu 3 Sn and Cu 6 Sn 5 showed quite differently on Cu foil andelectrodeposited Cu films. In the case <strong>of</strong> Kirkendall voids, theelectrodeposited Cu film showed much more at the IMCs thanOFHC Cu foil. The voids in electrodeposited Cu with anadditive were clearly much more than that without additive.And also, most voids <strong>of</strong> electrodeposited Cu with an additivewere distributed at the interface between Cu 3 Sn and Cu, whilethe voids <strong>of</strong> electrodeposited Cu without additive wererandomly distributed in Cu 3 Sn layer. For the growth kinetics<strong>of</strong> Cu 6 Sn 5 and Cu 3 Sn layer, total thickness and relativefractions <strong>of</strong> IMCs were measured after thermal aging at150°C. The ratio <strong>of</strong> Cu 3 Sn layer to total IMC was about 50%except for specimen <strong>of</strong> electrodeposited Cu film withadditive. For the electrodeposited Cu film with additive, ratio<strong>of</strong> Cu 3 Sn to total IMC decreased with the aging time. It isclear that electrodeposition condition <strong>of</strong> Cu film influencesinterfacial reaction and void formation behavior between Cuand solder. The drop impact test showed that electrodepositedCu film with additive degraded drastically with aging time.Fracture occurred at the Cu/Cu 3 Sn interface, where a lot <strong>of</strong>voids existed. In contrast, Cu foil showed much more reliablealthough brittle fracture occurred at the Cu 6 Sn 5 /Cu 3 Sninterface. Therefore, voids occupied at the Cu/Cu 3 Sn interfaceare shown to be degraded seriously drop reliabilities <strong>of</strong> solderjoints.1. IntroductionIn the microelectronic packaging, OSP Cu has beenwidely used as the metallization for the soldering due toshowing good solderablility with solder materials. The issues<strong>of</strong> cost and reliability for electroless Ni in microelectronicsassembly industry triggered the development <strong>of</strong> OSP on Cumetallization. Against the reliable applications <strong>of</strong> OSP Cu tothe microelectronic devices, however, the Kirkendall voidformation between solder and Cu metallization has been a bigconcern because the voids severely degraded mechanicalreliability <strong>of</strong> solder joints [1-4]. During the solid state aging,Cu 3 Sn IMC layer grows between Cu and Cu 6 Sn 5 IMC withreaction time. Kirkendall voids <strong>of</strong>ten observed either at theCu/Cu 3 Sn interface or inside Cu 3 Sn [1-7], which results fromthe difference <strong>of</strong> intrinsic diffusivity <strong>of</strong> two species, Cu andSn. In the case <strong>of</strong> high purity OFHC Cu foil, it is known thatKirkendall voids have been hardly observed at the interface <strong>of</strong>Cu foil and solder [5,6], while the electrodeposited Cu showsa lot <strong>of</strong> voids between solder and Cu layer. The voidformation has been explained with the impurity inelectrodeposited Cu layer. Yang et al. [6] showed that voidswere found in the Cu 3 Sn IMC formed at electrodepositedCu/Sn-3.5Ag interface aged at 190°C for 3 days, while novoids were observed at rolled Cu/Sn-3.5Ag interface afteraging at 190°C for 12 days. They assumed that hydrogenintroduced during electroplating process accelerated voidsformation in the Cu 3 Sn IMC layer. However, theseassumptions were not experimentally established and no clearevidence existed about void formation mechanism.In this study, electrodeposition condition on void formationswas systematically studied. The specimens subjected tothermal aging at 150°C for 960h to confirm IMC growthkinetics and void formation behavior. Then, drop tests wereconducted to establish mean cycle to failure and failure locus.Correlation between voids distribution and the drop reliabilitywere presented and discussed.2. Experimental ProceduresThree types <strong>of</strong> Cu metallization, OFHC Cu foil andelectrodeposted Cu films with and without additive, wereused to react with Sn-3.5Ag solder. In the case <strong>of</strong>electrodeposited Cu, electrolytes and plating conditions weredescribed in Table 1.Table 1. Electroplating condition and bath compositionfor Cu deposition.CuelectrodepositonP N(without additive)P A(with additive)ElectrolyteCuSO 4·5H 2 O,H 2 SO 4 ,DI waterCuSO 4·5H 2 O,H 2 SO 4 ,DI water,commercialadditivesCurrent density 2A/dm 2 2A/dm 2BathtemperatureRoom temperature Room temperatureCopper was electrodeposited on Cu pads with 680µmdiameter defined in printed circuit board (PCB) substrate, anddiameter <strong>of</strong> the solder balls used was 760µm. Thickness <strong>of</strong>1-4244-0985-3/07/$25.00 ©2007 IEEE 1620 2007 Electronic Components and Technology Conference

electrodeposited Cu layer was fixed at about 20µm. Thespecimens were reflowed at 260°C for 1min. After reflow, theaging tests were conducted at 150°C for 240, 480, 720 and960h. Drop tests were conducted to evaluate the mechanicalreliability <strong>of</strong> solder joint. Specimens were assembled usingtwo FR4 PCBs with 16 Sn-3.5Ag solder balls. Solders werereflowed on the top PCB first with either Cu foil orelectrodeposited Cu and then joined with the bottom PCB bythe 2 nd reflow. The pads <strong>of</strong> all bottom PCBs were Cu foilwithout electrodeposited Cu. Thus, top side was reflowedonce more than the bottom side. The shock loading was setaccording to JESD22-B111[8] by adjusting the drop heightand acceleration. The shock pulse from the drop impact wasoptimized to a triangular shape with the peak acceleration <strong>of</strong>1500G for 0.4ms. Two PCBs were daisy-chained, and the insitudetection <strong>of</strong> the threshold resistance <strong>of</strong> 20Ohms was usedas a criterion <strong>of</strong> complete drop failure. The test was stoppedwhen failure did not occur up to 400 drops. Scanning electronmicroscopy (SEM) with energy dispersive x-ray spectroscopy(EDX) was used to investigate IMC thickness, composition,and the failure mode.3. Results and Discussions3.1 Solid state aging testThe microstructure at Cu foil/Sn-3.5Ag solder interfaceafter reflow at 260°C for 1min is shown in Fig. 1. Afterreflow process, Cu 6 Sn 5 IMC formed first at the Cu/Sn-3.5Aginterface. Cu 3 Sn was hardly detected with back scatteredelectron (BSE) image and no voids existed at the Cu/Sn-3.5Ag interface. P A (electrodeposited Cu without additive)and P N (electrodeposited Cu with additive) showed similarmicrostructures to Fig. 1 just after reflow.case <strong>of</strong> P A specimen, more and larger voids were found thanthose in P N . And voids in P A were distributed close to theCu/Cu 3 Sn interface as shown in Fig. 2(b) and (c). Mei et al.[7] reported that voided area at the Cu/Cu 3 Sn interface was30% more than that <strong>of</strong> Chiu’s data [1], although specimensexperienced same aging condition. Chiu et al. [1] indicatedthat increase <strong>of</strong> operating temperature accelerated intensivelyvoids formation. Nevertheless, in the same aging condition, itis sure that voids formation strongly depends on the Cufabrication and deposition condition.(a)(b)Cu 6 Sn 5Cu 3 SnCu 6 Sn 5Cu 3 SnSn-3.5AgCuSn-3.5Ag5µmSn-3.5AgCu5µmCuCu 6 Sn 5(c)Sn-3.5Ag5µmFigure 1. Cross-sectioned BSE micrographs <strong>of</strong> interfacialmicrostructures at Cu foil/Sn-3.5Ag interface after reflowat 260°C for 1min.Figure 2 is BSE images showing Cu/Sn-3.5Ag interfaceafter aging at 150°C for 240h. All specimens showed Cu 3 Sn 5IMC <strong>of</strong> typical planar shape between Cu and Cu 6 Sn 5 .However, Kirkendall voids formation behavior wastremendously different in each specimen. As shown in Fig.2(a), the voids were hardly observed. On the other hand, P Aand P N specimens had clearly Kirkendall voids, even thoughthe amount and distribution <strong>of</strong> voids were not similar. In theCu 6 Sn 5Cu 3 SnCu5µmFigure 2. Cross-sectioned BSE micrographs after thermalaging at 150°C for 240h in (a) Cu foil, (b) P N , and (c) P Aspecimen.Figure 3 shows the microstructure <strong>of</strong> solder joint afteraging at 150°C for 960h. For the specimen with Cu foil,1621 2007 Electronic Components and Technology Conference

egardless <strong>of</strong> increased aging time, voids are not visible insideCu 3 Sn layer as shown in Fig. 3(a). Voids in P N specimendistributed randomly and size was smaller than those <strong>of</strong> P Nspecimen [Fig. 3(b)]. However, in the P A specimen, largervoids located at near the Cu/Cu 3 Sn interface [Fig. 3(c)]. It isinteresting that Kirkendall voids were located at the Cu/Cu 3 Sninterface and suppressed the growth <strong>of</strong> Cu 3 Sn layer byblocking Cu diffusion into Cu 3 Sn. This phenomenon isconsistent with Zeng’s observation. [5] Cu 3 Sn layer wereconverted to Cu 6 Sn 5 layer because voids blocked Cudiffusion. Therefore, Cu 6 Sn 5 layer became thicker than Cu 3 Snlayer as shown in Fig. 3(c).(a)(b)(c)Cu 6 Sn 5Cu 6 Sn 5Cu 3 SnCu 3 SnSn-3.5AgCu 6 Sn 5Cu 3 SnCuSn-3.5AgCuSn-3.5AgCu5µm5µm5µmFigure 3. Cross-sectioned BSE images after thermal agingat 150°C for 960h in (a) Cu foil, (b) P N , and (c) P A specimen.For the growth kinetics <strong>of</strong> Cu 6 Sn 5 and Cu 3 Sn layer, totalIMC thickness and relative fractions <strong>of</strong> the two IMCs weremeasured after thermal aging at 150°C as shown in Fig. 4.Total IMC growth behavior appeared as a function <strong>of</strong> squareroot <strong>of</strong> time about all specimens as given in Fig. 4(a). Sincetwo types <strong>of</strong> IMCs were formed in each specimen, thickness<strong>of</strong> Cu 6 Sn 5 and Cu 3 Sn was measured separately and the ratio <strong>of</strong>Cu 3 Sn layer to total IMC was calculated. Ratio <strong>of</strong> Cu 3 Sn layerto total IMC was about 50% except P A specimen, while ratio<strong>of</strong> Cu 3 Sn to total IMC for P A specimen decreased with theaging time.Total IMC thickness (µm)δ Cu3 Sn /δ total (%)10987654321Cu foilP NP A00 200 400 600 800 1000Aging time (hr)(a)100Cu foilP N80604020P A00 200 400 600 800 1000Aging time (hr)(b)Figure 4. Variation <strong>of</strong> (a) total IMC thickness and (b)ratio <strong>of</strong> Cu 3 Sn to total IMC thickness.Figure 5 showed voided area between Cu and Cu 3 Sn andratio <strong>of</strong> voided length to Cu/Cu 3 Sn interface length measuredby image analyzing program. A comparison between voidedarea <strong>of</strong> P A and P N specimens shows that more voids in P Aspecimen grew between Cu and Cu 3 Sn layer. When voiddensity increased at the Cu/Cu 3 Sn interface, voids coalescedinto disk shape and did not migrate into Cu 3 Sn layer. On theother hand, voids migrate with the growth <strong>of</strong> Cu 3 Sn layer asvoid density between Cu and Cu 3 Sn layer is not high. It is1622 2007 Electronic Components and Technology Conference

clear that randomly distributed small voids in Cu 3 Sn layeraffect much less to the diffusion <strong>of</strong> Cu into Cu 3 Sn as indicatedin Fig. 4.Voided area / total IMC area (%)% voided length10864200 200 400 600 800 1000Aging time (hr)10080604020P NP AP NP A(a)appears that brittle fracture occurred at the interface betweenCu 3 Sn and Cu 6 Sn 5 . Recently, it has been reported that brittlefracture occurred along the interface between Cu 3 Sn andCu 6 Sn 5 layer for Sn-Ag-Cu solder joint with Cu metallization,although no mechanisms were suggested about inter-IMCfailure, yet [9,10]. Therefore, crack propagates much easieralong the interface <strong>of</strong> layered two IMC, Cu 3 Sn and Cu 6 Sn 5 ,than other locations as reported in our previous work [11].N f35030025020015010050Cu foilP A00 200 400 600 800 1000Aging time (hr)Figure 6. Number <strong>of</strong> drops to failure vs. aging time.(a)Sn-3.5Ag00 200 400 600 800 1000Aging time (hr)(b)Figure 5. (a) Ratio <strong>of</strong> voided area to total IMC area and(b) ratio <strong>of</strong> voided length to Cu/Cu 3 Sn interface length3.2 Drop impact testDrop impact test was conducted to confirm effect <strong>of</strong> voidformation at the Cu/Cu 3 Sn interface. Mean cycles to failurefor board level drop tests are presented in Fig. 6 where effect<strong>of</strong> voids occupied at the Cu/Cu 3 Sn interface is clearlydemonstrated. It is very interesting that two opposite trendswere identified for the drop tests. For the tests with Cu foil,drop resistance <strong>of</strong> as-reflowed specimens was poor incomparison to thermally aged specimens. On the contrary, forthe tests with P A , drop resistance decreased drastically withthe thermal aging at 150°C. Cross-sectional BSE images <strong>of</strong>failed specimens reveal the path <strong>of</strong> crack propagation aspresented in Fig. 7 and 8. Brittle failure occurred at theinterface <strong>of</strong> solder joints about all specimens, although dropresistance <strong>of</strong> specimens is quite different. Fig. 7 shows failedinterface <strong>of</strong> solder joint with Cu foil after reflow at 260°C for1min. Focused ion beam(FIB) image as shown in Fig. 7(b)(b)Cu 3 SnCu 6 Sn 5Cu 3 SnCu 6 Sn 5Sn-3.5AgCuCu5µm3µmFigure 7. (a) BSE and (b) FIB images <strong>of</strong> as-reflowedsolder joint with Cu foil after drop test.1623 2007 Electronic Components and Technology Conference

Brittle failure loci for solder joints aged at 150°C for 480hwere presented in Fig. 8. Although brittle failure occurred atthe solder joints, failure site was different and depended onKirkendall void formation. In the case <strong>of</strong> specimens with Cufoil, dominant crack propagation occurred at theCu 3 Sn/Cu 6 Sn 5 interface as shown in Fig. 8(a). From Fig. 7and 8(a), it can be seen that the Cu 3 Sn/Cu 6 Sn 5 interface is notonly weak site but dominant crack path regardless <strong>of</strong> Cu 3 SnIMC thickness. For the specimens with P A , brittle fractureoccurred along the voids formed at the Cu/Cu 3 Sn interface asshown in Fig. 8(b). As shown in reaction study, voidsoccupied completely at the interface between Cu and Cu 3 Snafter aging at 150°C for 480h and the growth <strong>of</strong> Cu 3 Sn IMCwas prohibited by voids. Brittle failure occurred along thevoids formed at the Cu/Cu 3 Sn interface. As shown in Fig.6.drop resistance is very weak when a large amount <strong>of</strong> voidsexist at the Cu/Cu 3 Sn interface. From plane view images <strong>of</strong>specimen with P A after drop test as shown in Fig. 9, it wasclearly demonstrated that fractured interface has an extremelyporous structure, which gives very weak chemical adhesionbetween Cu 3 Sn intermetallics and Cu. Based on the abovedrop impact results, it can be sure that Kirkendall voidsformation <strong>of</strong> solder joint strongly correlated with dropreliability. Additionally, to improve mechanical reliability <strong>of</strong>solder joints with Cu film needs to decrease Kirkendall voidsaccumulated at the interface between Cu 3 Sn and Cu.(a)(b)(a)(b)Cu 3 SnCu 6 Sn 5Cu 3 SnCu 6 Sn 5Sn-3.5AgCu 5µmCuSn-3.5Ag5µmFigure 8. Failure morphology <strong>of</strong> (a) specimen with Cu foiland (b) P A after aging at 150°C for 480h.Figure 9. Plane views <strong>of</strong> solder joint with P A after aging at150°C for 480h (a) x100 and (b) x5000.ConclusionsInterfacial reaction study and drop impact test <strong>of</strong> Sn-3.5Ag solder joint with OHFC Cu foil or electrodeposited Cufilms were performed after thermal aging at 150°C. Additiveinfluenced very critically to the location as well as theformation <strong>of</strong> Kirkendall void. Behavior <strong>of</strong> IMC growth wereinfluenced by Kirkendall voids because Cu diffusion wereblocked by voids. Specimen <strong>of</strong> Cu foil after thermal agingshowed the failure loci through the interface between Cu 3 Snand Cu 6 Sn 5 because <strong>of</strong> brittle properties at the interface.Specimen <strong>of</strong> P A after thermal aging showed the failure locithrough the interface between Cu 3 Sn and Cu 6 Sn 5 because <strong>of</strong>Kirkendall voids.AcknowledgmentsThis work was supported by the Center for ElectronicPackaging Materials (ERC) <strong>of</strong> MOST/KOSEF. (grant # R11-2000-085-05001-0)References1. Chiu, T.-C., Zeng, K, and Stierman, R., Edwards, D., andAno, K., “<strong>Effect</strong> <strong>of</strong> Thermal Aging on Board Level DropReliability for Pb-free BGA Packages”, Proc <strong>of</strong> 54 thElectronic Components and Technology Conf, Las Vegas,NV, June. 2004, pp. 1256-1262.1624 2007 Electronic Components and Technology Conference

2. Zeng, K., Stierman, R., Chiu, T.-C., Edwards, D., Ano K.,and Tu, K. N., “Kirkendall Void Formation in EutecticSn-Pb Solder Joint on Bare Cu and Its <strong>Effect</strong> on JointReliability”, J. Appl. Phy., 97, 024508 (2005).3. Xu, L. and Pang, John H. L., “<strong>Effect</strong> <strong>of</strong> Intermetallic andKirkendall Voids Growth on Board Level Drop Reliabilityfor SnAgCu Lead-free BGA Solder Joint”, Proc <strong>of</strong> 56 thElectronic Components and Technology Conf, San Diego,CA, June. 2006, pp. 275-282.4. Yu, Jin, Sohn, Y. C., Kim, J. Y., Jee, Y. K., and Ko, Y.H., “Impact reliabilities <strong>of</strong> lead-free solder joints withNi(P), Cu and Ni metallizations”, Proc <strong>of</strong> 2006International Conference on Electronics Packaging,Tokyo, April. 2006.5. Laurila, T., Vuorinen, V., and Kivilahti, J. K., “Interfacialreactions between lead-free solders and common basematerials”, Mater. Sci. Eng., R 49 (2005), pp.1-60.6. Yang, W and Messler, R. W., “Microstructure Evolution<strong>of</strong> Eutectic Sn-Ag Solder Joints”, J. Electron. Mater., Vol.23, No. 8 (1994), pp. 765-772.7. Mei, Z., Ahmad, M., Hu, M., and Ramakrishna, G.,“Kirkendall Voids at Cu/Solder Interface and Their<strong>Effect</strong>s on Solder Joint Reliability”, Proc <strong>of</strong> 55 thElectronic Components and Technology Conf, Orlando,FL, June. 2005, pp. 415-420.8. JESD22-B111, “Board Level Drop Test Method <strong>of</strong>Components for Handheld Electronic Products”, JEDECstandard, July. 2003.9. Shin, D. K., Lee, D. Y., Ahn, E. C., Kim, T. H., and Cho,T. J., “Development <strong>of</strong> Multi Stack Package with HighDrop Reliability by Experimental and NumericalMethods”, Proc <strong>of</strong> 56 th Electronic Components andTechnology Conf, San Diego, CA, June. 2006, pp. 377-382.10. Jang, J. W., A.P. De Silva, A. P., Drye, J. E., Post, S. L.,Owens, N. L., Lin, J. K., and Frear, D. R., “Drop Test andFailure Mechanism <strong>of</strong> Sn-3.8Ag-0.7Cu and Sn-37PbSolders in BGA Packages”, Proc <strong>of</strong> 39 th InternationalMicroelectronics and Packaging Society, San Diego, CA,October. 2006, pp. 357-364.11. Kim, J. Y., Sohn, Y. C., and Yu, Jin, “<strong>Effect</strong> <strong>of</strong> Cucontent on the mechanical reliability <strong>of</strong> Ni/Sn–3.5Agsystem”, J. Mater. Res., In press (2007).1625 2007 Electronic Components and Technology Conference