電子封裝熱遷移2007snpb

1、Microstructural evolution and atomic transport by thermomigration in eutectic tin-lead flip chip solder jointsDan Yang, B. Y. Wu, Y. C. Chan, and K. N. TuCitation: J. Appl. Phys. 102, 043502 (2007); doi: 10.1063/1.2769270View online:View Table of Contents:Published by the AIP Publishing LLC.Addition

2、al information on J. Appl. Phys.Journal Homepage: Journal Information: Top downloads: Information for Authors:Downloaded 26 Aug 2013 to 202.38 202.193. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at:JOURNAL OF APPLIED PHYSICS 102, 043502 200

3、7 Microstructural evolution and atomic transport by thermomigration in eutectic tin-lead flip chip solder jointsDan Yang, B. Y. Wu, and Y. C. Chana Department of Electronic Engineering, City University of Tong,K. N. Tu, Tat Chee Avenue,Department of Materials Science and Engineering, UCLA, Los Angel

4、es, California 90095-1595 Received 15 November 2006; accepted 30 June 2007; published online 16 August 2007 The thermomigration behavior of eutectic tin-lead flip chip solder joints at an ambient temperature of 150 °C was investigated in terms of microstructural evolution, atomic transport, and

5、 numerical simulation. Pb accumulation and phase separation were observed in solder joints near a melting temperature after 50 h, which was supported by energy dispersive x-ray and element map analysis. It is believed that Pb atoms migrated from the chip side the hot side to the substrate side the c

6、old side under a temperature gradient. Thermal electrical finite element simulation for the real flip chip test structure showed the existence of a temperature difference between the substrate side and the chip side. In addition, a temperature gradient above 1000 °C/ cm across the adjacent unpo

7、wered solder joints was predicted. This was also verified by temperature measurements with thermocouples. The atomic flux of Pb due to thermomigration was calculated here, which was agreeable with the values originally reported. Also, the driving force of thermomigration was estimated to be 1017 N,

8、even approaching the same order with that of electromigration under a current density of 104 A/ cm2. © 2007 American Institute of Physics. DOI: 10.1063/1.2769270 I. INTRODUCTIONWith the trend toward higher integration and further miniaturization in the microelectronics industry, the cross- sect

9、ional area of conductive lines on the chip has been de-structural observations and marker measurements and their numerical simulation predicted a temperature gradient of 1500 °C/ cm across solder joints. Huang et al. observed a migration of the Sn-rich phase owing to electromigration and thermo

10、migration in tin-lead composite solder joints.6 How- ever, the individual contribution of thermomigration to the failure of solder joints was not specifically investigated. Re- cently, Huang et al. found that Pb atoms moved to the cold side and Sn to the hot side in tin-lead composite solder joints

11、and they suggested that a temperature gradient of 1000 °C/ cm was sufficient to trigger the thermomigration.7 Also, Chuang and Liu reported the thermomigration in eu- tectic tin-lead solder by observing the depletion of the Pb- rich phase at the hot side.8 However, the microstructural evo- luti

12、on due to thermomigration in eutectic tin-lead soldercreased significantly. Due to different electricalsand thermal capacities of individual parts within the flip chip interconnection structure, it is possible that the heat accumu-lated at the chip side is larger than that at the substrate side. Thi

13、s will inevitably lead to a considerable temperature gra- dient across the solder joints, which can provide a driving force for atomic diffusion to trigger thermomigration. In our previous studies, we identified the presence of this tempera- ture gradient by observing a local melting spot in the sol

14、der joints.1 Other research work concerning Joule heating has also predicted this by thermal electrical finite element simulation.2 Recently, hot spots near the entrances of Al traces in the solder joints have been detected by a thermal infrared microscope,3 which provides further support for the ex

15、istence of a temperature gradient across solder joints.Being a complex mass migration process, thermomigra- tion, of which the driving force comes from the energy trans- ported by the moving atoms and the interactions with the usual heat carriers in the lattice,4 has become an important reliability

16、concern for flip chip solder joints. Since it is in- evitable that Joule heat would result from current stressing, thermomigration may occur together with electromigration, if the temperature gradient is sufficiently large. The earliest report regarding the combined effects of electromigration and t

17、hermomigration in the solder joints was given by Basa- ran et al.5 They found that thermomigration in flip chip sol- der joints may assist or counter electromigration with micro-joints has not been studied in. Only the latest researchby Ouyang et al. proposed a redistribution of the Sn andPb-rich ph

18、ase with Pb atoms moving to the cold side in eutectic tin-lead solder joints at 100 °C.9As thermomigration might play an important role in the failure of solder joints, especially for micropower electron- ics, this investigation is intended to study thermomigration phenomena in eutectic tin-lea

19、d solder joints under Joule heating in terms of microstructural analysis and atomic trans- port. In addition, to help understand the mechanism of ther- momigration, a three-dimensional thermal electrical finite el- ement simulation will be conducted to estimate the temperature distribution in the re

20、al flip chip interconnection structure.II. EXPERIMENTThe modules used in this study were flip chip test chips PB08-a Electronic mail: eeycchkits, including dummy chips and .hk0021-8979/2007/102 4 /043502/6/$23.00102, 043502-1© 2007 American Institute of PhysicsDownloaded 26 Aug 20

21、13 to 202.38 202.193. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at:043502-2Yang et al.J. Appl. Phys. 102, 043502 2007 FIG. 2. Color online Sketch of solder joints with four solder joints joints 58 under current stressing.the substrate side

22、 under the current stressing and high ambi- ent temperature was inspected with a M.O.L.E. thermal pro- filer. Two microthermocouples with 0.15 mm diameter, which have a sensitivity of 0.1 °C and an accuracy of±1 °C, were directly attached on top of the chip surface and the copper trac

23、e surface near the powered joints, as shown inFig. 1. High temperatuhesive tape was used for the me-FIG. 1. Color online Flip chip module used in this study The thermo- couples arrowed were used to measure temperature. chanical stabilization of thermocouples on the chip or copper trace.Unpowered sol

24、der joints, especially the most adjacent joints joints 4 and 9 to the stressed joints, were investigated for thermomigration study. After experiments, the samples were ground and polished toward the center of solder joints. These cross sections were then examined with a scanning electron microscope

25、SEM Philips XL40 FEG . The local compositions of solder joints were analyzed using an energy dispersive x-ray EDX system. In addition, an element map- was performed to detect the distribution of each elementof solder joints after thermomigration.250 250 . The chipwas 6 6 mm2 with a thickness of0.635

26、 mm. The Al traces on the chip had a width of 105 mand a thickness of about 2 m. The under bump metalliza- tion layer was composed of a thin film of Al/Ni/Cu and the passivation opening had a diameter of 102 m. There were four rows of 12 solder bumps along each side of the chip with a pitch of 460 m

27、. The bump material was eutectic tin-lead and the bump diameter was about 190 m. The test substrate was a high temperature FR4 board with a thickness of 0.84 mm. The copper pad connected to the solder bumps had a width of about 152 m and a thickness of 35 m, which was plated by a thin 2 m el layer.

28、Both chips and substrates were daisy-chained for electrical continuity.The flip chip was attached to the substrate using flip chip bonding technology by a KarlSuss FCM flip-chip bonder. The bonded samples were then reflowed using a five-zone air convection oven BTU VIP-70 N . In the temperature prof

29、ile, the peak temperature was 230 °C and the time above liqui- dus was about 60 s. Figure 1 shows a typical bonded module prepared for testing.Figure 2 is a sketch of flip chip solder joints used for this study. Two pairs of solder joints joints 5, 6, 7, and 8 were powered with a current of 1.8

30、 A at 150 °C. This correspondsIII. RESULTS AND DISCUSSIONThe electricalof the Al traces, solder joints,and copper conductors were calculated to be approximately201.2, 1.6, and 13.0 m , respectively. These values repre- sent their relative contributions to the Joule heating. It is apparent that

31、the Al trace is the primary heat source becauseof its larger. The local temperatures on top of thechip surface and the copper trace surface near the poweredjoints were measured to be 172.9 and 164.8 °C, respectively, when a current of 1.8 A was applied to the middle four solder joints joints 5,

32、 6, 7, and 8 at the ambient temperature of 150 °C. The temperature difference between the two sur- faces reached about 8 °C. Although the temperature mea- surement was not very accurate because of the contact resis- tance between the thermocouples and the surfaces, it provided the evidence

33、 that the effect of Joule heating was significant for the test structure. Hence, it is believed that ato aage current density, defined by dividing the appliedcurrent by the area of the passivation opening, of about 2.2 104 A/ cm2. A LHT4/30 type oven was utilized to realize a stable and uniform ambi

34、ent temperature of 150 °C.The temperature difference between the chip side andFIG. 3. a SEM micrograph of origi- nal microstructure of solder joints as- reflowed and b local magnified micrograph.Downloaded 26 Aug 2013 to 202.38 202.193. This article is copyrighted as indicated in the abstract.

35、Reuse of AIP content is subject to the terms at:043502-3Yang et al.J. Appl. Phys. 102, 043502 2007 moved to the cold side and Sn to the hot side under a tem- perature gradient of about 1000 °C/ cm at 150 °C.For other unpowered solder joints from joints 1 to 3 and joints 10 to 12 , the ther

36、momigration did not occur instead. As illustrated in Fig. 7, only phase coarsening was observed as compared with the original micrograph. This suggests that phase coarsening at a high homologous temperature rather than thermomigration dominated in this process.An appropriate explanation for the abov

37、e is as follows. For joints 13 and joints 1012, the temperature differences between the chip side and the substrate side were relatively small due to the increasing distances from the powered sol- der joints, i.e., the source of heating. Thus the temperature gradients across them were not high suffi

38、ciently to cause a significant migration. By contrast, there existed a larger tem- perature gradient across the solder joints 4 and 9 that were nearest to the heating source. In general, the flow of atoms is from the hot side to the cold side if there is a temperature gradient.10,11 It is supposed t

39、hat both Pb and Sn atoms would have a tendency to migrate from the hot side to the cold side under this temperature gradient. However, Pb atoms are the dominant species with high diffusivity in the eutectic tin-lead solder above 120 °C.1214 Therefore, Sn atoms migrated slowly and replenished th

40、e vac es due to the depletion of Pb atoms. Macroscopically, the Pb-rich phase migrated to one side and the Sn-rich phase was “pushed” toward the opposite side. This gives a reasonable explanation to the fact that the effect of thermomigration was apparently visible in the most adjacent solder joints

41、.Also, from Fig. 5 it is noticeable that the Pb-rich phase was accumulated at the lower left side, i.e., the colder region of solder joint 4. Likewise, the Pb redistribution in solder joint 9 showed the similar tendency, i.e., Pb migrated to the lower right side or the colder region, as Fig. 4 shows

42、. We recall the lateral thermomigration in composite solder joints.6 Take the solder joint 4 for example, since its right side wasFIG. 4. SEM micrographs of a row of solder joints from joints 1 to 12, with solder joints from joints 5 to 8 under current stressing after 50 h at 150 °C Pb accumula

43、tion in unpowered solder joints 4 and 9 .certain temperature gradient existed in the powered solder joints. Moreover, owing to good thermal conductivity of sili- con chip, this temperature gradient was also formed across the adjacent unpowered solder joints such as joints 4 and 9. Figure 3 illustrat

44、es a typical original microstructure of solder joints before the experiments as-reflowed . Fine scale Pb-rich phase particles light regions were uniformly dis- persed in the Sn-rich matrix dark regions . These two re-gions were -Pb and -Sn phases, respectively.Figure 4 shows the SEM images of the cr

45、oss section of a row of 12 solder joints after 50 h experiment at 150 °C. It is evident that the powered solder joints joints 58 were dam- aged by electron flow plus following partial melting of Sn- rich phase with a low melting point. More significantly, ob- vious Pb thermomigration was detect

46、ed in the most adjacentunpowered joints 4 and 9. Figure 5 shows theed mi-crostructure of joint 4 with a high magnification. According to these micrographs, it is believed that Pb migrated to thesubstrate side the cold side under the temperature gradient across the unpowered solder joints since no cu

47、rrent was ap- plied to them. This was supported by the EDX analysis for local regions. As shown in Fig. 5 b , the average concentra- tion of accumulated Pb at the substrate side was about65.16 at. %, and the concentration of Sn at the chip side approached 86.29 at. %. The width of accumulated Pb-ric

48、h phase band reached approximately 15 m, i.e., a half of the joint standing height. In addition, the separation of Pb and Snmojacent to the heating source, it is possible that atemperature gradient was also established laterally from the right side to the left side. Thus the Pb-rich phase not only m

49、igrated to the substrate side under the vertical temperature gradient, but also moved to the colder region driven by the lateral temperature gradient across this solder joint.It is worth mentioning that large mounts of intermetallilc compounds IMC appeared in unpowered solder joints 4 and 9. This ph

50、enomenon suggests that unpowered solder joints had experienced a very high temperature solid-state aging or even a local melting cycling. The thermomigrationwas apparent according to the element mapof Pb andSn, as shown in Fig. 6. This result agrees well with that ofthermomigration in tin-lead compo

51、site flip chip solder joints reported by Huang et al.7 They observed that Pb atomsFIG. 5. a SEM micrograph of ther- momigration in an unpowered solder joint after 50 h at 150 °C joint 4 and b local magnified micrograph.Downloaded 26 Aug 2013 to 202.38 202.193. This article is copyrighted as ind

52、icated in the abstract. Reuse of AIP content is subject to the terms at:043502-4Yang et al.J. Appl. Phys. 102, 043502 2007 FIG. 6. a Element distribution of Sn on the surface of solder joint 4 and b Pb on the surface of solder joint 4.occurred near the eutectic temperature of solder joints in our fl

53、ip chip test modules. Also, another interesting phenomenon was noticed. As can be seen from Fig. 5 b , the Pb grains were even more uniformly dispersed in the tin-matrix, al- though the bulk of Pb had moved to the substrate side. It means that the lamellar microstructure became much finer after the

54、thermomigration process. This is different from the phase segregation occurred in eutectic tin-lead solder due to electromigration. According to the tin-lead phase diagram,15 since Sn has very small solubility of Pb, there is no issue of precipitation of Pb from Sn. Consequently, it cannot be ex- pl

55、ained well with dual-phase precipitation in a thermal cy- cling process. In a most recent study by Ouyang et al., a similar phenomenon was observed in eutectic tin-lead solder joints. They proposed that the refined microstructure was due to entropy generation during the thermomigration process.9Addi

56、tionally, thermal electrical finite element simulation of the real flip chip test structure was conducted. The param-joint veryto the entrance of electron flow. This is inagreement with the experimental findings that the poweredsolder joints were partially melted because of large Joule heating. Sign

57、ificantly, Fig. 11 shows that there existed a tem- perature difference above 3 °C between the hot side and the cold side in the unpowered joint 9. It means that a tempera-1000 °C/ cm3 °C/ 30 mturegradientabove 1000 °C/ cm was built up across the unpowered joint due to the Joule h

58、eating from the neighboring Al traces. Thissimulation showed the existence of a temperature gradient across solder joints in the test structure, and predicted a tem- perature gradient above 1000 °C/ cm across the most adja- cent unpowered solder joints under such an experimental condition.In or

59、der to understand the mechanism of atomic trans- port, the atomic flux and the driving force during the ther- momigration process were estimated. Taking a central dis- placement x of 7.5 m, the total volume of atomic transport Vtm during the operation time t can be approxi- mately obtained from the product of the displacement and the cross-sectional area