Failure induced by thermomigration of interstitial Cu in Pb-free flip chip solder joints

(1)

Failure induced by thermomigration of interstitial Cu in Pb-free flip chip solder joints

Hsiao-Yun Chen, Chih Chen, and King-Ning Tu

Citation: Applied Physics Letters 93, 122103 (2008); doi: 10.1063/1.2990047

View online: http://dx.doi.org/10.1063/1.2990047

View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/93/12?ver=pdfcov Published by the AIP Publishing

Articles you may be interested in

Modeling of electromigration on void propagation at the interface between under bump metallization and intermetallic compound in flip-chip ball grid array solder joints

J. Appl. Phys. 107, 093526 (2010); 10.1063/1.3369442 Blech effect in Pb-free flip chip solder joint

Appl. Phys. Lett. 94, 011912 (2009); 10.1063/1.3067863

The time-dependent melting failure in flip chip lead-free solder interconnects under current stressing Appl. Phys. Lett. 93, 041907 (2008); 10.1063/1.2963473

Electromigration in Pb-free flip chip solder joints on flexible substrates J. Appl. Phys. 99, 023520 (2006); 10.1063/1.2163982

Polarity effect of electromigration on kinetics of intermetallic compound formation in Pb-free solder V-groove samples

J. Appl. Phys. 97, 063514 (2005); 10.1063/1.1861151

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 140.113.38.11 On: Wed, 30 Apr 2014 22:53:33

(2)

Failure induced by thermomigration of interstitial Cu in Pb-free flip chip

solder joints

Hsiao-Yun Chen,1Chih Chen,1,a兲 and King-Ning Tu2

1_{Department of Materials Science and Engineering, National Chiao Tung University, Hsin-chu 30010,} Taiwan, Republic of China

2_{Department of Materials Science and Engineering, UCLA, Los Angeles, California 90095-1595, USA}

共Received 24 May 2008; accepted 5 September 2008; published online 23 September 2008兲 Failure induced by thermomigration in Pb-free SnAg flip chip solder joints has been investigated by electromigration tests under 9.7⫻103 A/cm2at 150 ° C. The fast interstitial diffusion of Cu atoms from underbump metallization into Sn matrix caused void formation at the passivation opening on the chip side. The Cu diffusion was driven by a large thermal gradient and led to void formation even in the neighboring unpowered bumps. When the thermal gradient is above 400 ° C/cm, theoretical calculation indicates that the thermomigration force is greater than the electromigration force at 9.7⫻103 _A_/cm2_{stressing. © 2008 American Institute of Physics.}

关DOI:10.1063/1.2990047兴

Electromigration in flip chip solder joints has become a major reliability issue in recent years because of the minia-turization trend to meet the demand of higher performance in portable consumer electronics.1Much research has been con-ducted to provide scientific understanding of this reliability issue.2–8Serious current crowding effect has played a crucial role in the failure mechanism in electromigration,3,9causing void formation near the thin-film underbump metallizations10 共UBMs兲 or enhancing dissolution of metallization of the thick-film UBM in flip chip solder joints.11,12

Moreover, thermomigration accompanies

electromigration.4,5,13 The Joule heating of the Al intercon-nect on the chip side builds up a large temperature gradient across the solder joints. For Pb-containing solder, Pb atoms move to the cold end on the substrate side and the net effect is void formation on the chip side. However, failure induced by thermomigration in Pb-free flip chip solder joints has not been reported yet.

In this letter, electromigration and thermomigration in eutectic SnAg solder joints with Cu UBMs are reported. It is found that thermomigration of Cu dominates the failure mechanism. The fast Cu interstitial diffusion14 from Cu UBM into Sn-based solder matrix left supersaturated vacan-cies to form voids below the UBM on the chip side. The large thermal gradient has caused the migration of Cu atoms and the formation of Cu6Sn5 intermetallic compounds

共IMCs兲 to the cold end, and the Cu atoms also served as the dominant diffusion species in electromigration. An indepen-dent study on thermomigration was performed to verify the results. Theoretical calculation was conducted to support the observed phenomena.

To study electromigration and thermomigration simulta-neously, a special layout of four solder joints was adopted, as illustrated in Fig.1共a兲. On the chip side, an Al trace connects the four bumps together. Its dimensions are 2550 ␮m long, 100 ␮m wide, and 1.5 ␮m thick. The four bumps were de-noted as bump 1 through bump 4. The solder joints consists of eutectic SnAg solder bumps with electroplated 5 ␮m

thick Cu as UBM on the chip side. The bump dimensions are 130 ␮m wide and 70 ␮m high with a UBM opening of 120 ␮m in diameter. On the substrate side, six Cu nodes were fabricated and they are labeled as nodes N1–N6. The dimensions of the Cu lines on the FR5 substrate are 25 ␮m thick and 100 ␮m wide.

To facilitate the observation of electromigration and thermomigration, the joints were polished laterally and ap-proximately to their centers, as shown in Fig.1共b兲. After the polishing, the widths of the Al traces and the Cu lines also decreased accordingly. Electric currents were applied only through nodes N3 and N4, as shown in Fig.1共b兲. Therefore, bumps 1 and 4 did not have current passing through, so did the Al line connecting bumps 1 and 2 and that connecting bumps 3 and 4. Yet, the four bumps have experienced almost the same Joule heating because the Al line and the Si die possess excellent heat conduction. The applied current was 0.55 A at 150 ° C on a hot plate, which corresponds to an average current density of 9.7⫻103 _A_/cm2 _{in the UBM}

opening. Kelvin probes were employed to monitor the resis-tance changes in both bumps 2 and 3;15 thus the degree of damage as a function of time can be measured for these two bumps.

The temperature distribution in the solder bump during the current stressing was measured by a QFI infrared 共IR兲 microscope, which has 0.1 ° C temperature resolution and 2 ␮m spatial resolution. The thermal gradient is 1143, 1643, 1857, and 1143 ° C/cm for bumps 1, 2, 3, and 4, respec-tively, under stressing at 0.55 A and 150 ° C.

We found that voids formed in the chip end of the solder joint for all the four bumps after the current stressing for 76 h. The current was terminated when the resistance of bump 3 reached three times its original value. In bump 3, the void formed in the cathode end 共the chip side兲 because of elec-tromigration. However, we observed that serious damage due to void formation also occurred in the anode end 共the chip side兲 of bump 2. Its resistance also increased up to two times the original value. According to the thermomigration results on Pb-containing solder,4,5,13Pb atoms moved to the cold end on the substrate side; thus Kirkendall voids formed on the chip side. It was also found that Sn atoms migrated to the hot

a兲_{Author to whom correspondence should be addressed. Electronic mail:}

[email protected].

APPLIED PHYSICS LETTERS 93, 122103共2008兲

(3)

end on the chip side.5For the Pb-free solder in this study, it might be expected that Sn atoms would migrate to the hot end on the chip side since Sn has a positive value of heat transport number.16 Thus we expect void formation at the cold end on the substrate. Thus, the voids in the chip side for bump 2 cannot be accredited to the thermomigration of Sn in the solder.

Instead, we propose that thermomigration of Cu atom dominates the failure mechanism in this study. The scanning electron microscopy 共SEM兲 images for bumps 1–4 support this assumption, as illustrated in Figs. 2共a兲–2共d兲. We note that only thermomigration has taken place in bumps 1–4. Because the thermal gradients were slightly smaller than those in bumps 2 and 3, smaller voids occurred in them. Furthermore, the voids replaced the original position of the Cu6Sn5IMCs or that between the Cu UBM and the IMC. In

contrast, in the SnPb solder, the thermomigration voids ac-cumulate below the IMC.4,5,13Therefore we propose that the damage was due to the Cu migration from the hot end to the cold end.

The Cu interstitial diffusion in Sn matrix is extremely fast14especially driven by a large thermal gradient. Based on Huntington’s17results, Cu atom has the tendency to move to the cold end in thermomigration. When Cu dissolves inter-stitially in Sn, the original substitutional site of the Cu atom would be replaced by a vacancy. When the vacancies become supersaturated, the nucleation and growth of void can occur. Although there should be a net vacancy flux form the chip side to the substrate side since the vacancy concentration should be greater at the chip side, the diffusion rate of va-cancies is much less than that of the interstitials. In turn the Sn flux in exchange with the vacancy flux is less, too. Thus vacancy supersaturation and void formation tend to occur below the Cu UBM.

Furthermore, a higher Cu solubility in Pb-free solders may also facilitate the thermomigration of Cu and the forma-tion of Cu–Sn IMCs. It was reported that the solubility of Cu in eutectic SnPb at 220 ° C is 0.18 wt. %, whereas it reaches 1.54 wt. % in Pb-free solders at 260 ° C.18

Another evidence that lends support to Cu diffusion can be illustrated in Figs.3共a兲and3共b兲. Figure3共a兲represented the cross-sectional view of bump 1 before current stressing, and Fig. 3共b兲 was the view after 60 h stressing when the resistance of bump 3 reached to 2.3 times its original value. While bump 1 was unpowered, the damage is clear near the upper left-hand corner, as indicated by the left arrow in Fig. 3共b兲. In addition, from the backscattered SEM image shown in Fig. 3共b兲, the IMC has formed toward the substrate side and the Cu UBM has dissolved almost completely into sol-der. This can be accredited to the thermomigration of Cu atoms by fast interstitial diffusion and reaction with Sn at-oms to form IMC inside the solder bump.

Since the electromigration force in bump 2 also pushed the Cu atoms to the chip side, the void formation in the chip side infers that the thermomigration force overwhelmed the electromigration force. To verify this point, theoretic calcu-lations are performed below. The electromigration force is represented as19

FIG. 1. Schematic diagram of the experimental setup used in this study.共a兲 The layout before the bumps was polished. The distribution of current den-sity was also shown in the figure.共b兲 Schematic drawing of the samples after polishing.

FIG. 2. Cross-sectional SEM images showing the microstructure changes of unpowered bump 1 and bump 4 after current stressing at 9⫻104 _A_/cm2_at

150 ° C for 76 h.共a兲 Bump 1. 共b兲 Bump 2. 共c兲 Bump 3. 共d兲 Bump 4. Voids formed in all the four bumps.

FIG. 3. Cross-sectional SEM images representing the microstructure for the un-powered bump 1 in Cu UBM system before and after current stressing in bumps 2 and 3 at 0.55 A at 150 ° C for 60 h共a兲 before and 共b兲 after the current stressing.

122103-2 Chen, Chen, and Tu Appl. Phys. Lett. 93, 122103共2008兲

(4)

F = ZⴱeE = Zⴱe␳j, 共1兲 where Zⴱ is the effective charge number of ions, ␳ is the resistivity, j is the electric current density, and E is the elec-trical field. If we assume that the driving force is independent of the mechanism of diffusion, we take the effective charge number of Cu to be 6.4,20 resistivity as 1.67⫻10−8 _{⍀ m,}

and the current density in the current crowding region as 9.7⫻103 _A/cm2_{. The force is 1.66⫻10}−18 _{N and the work}

done by the force in an atomic jump distance of 3 ⫻10−8 _{cm is 4.98⫻10}−28 _{J. On the other hand, in}

thermo-migration, the driving force is21

F = −Q

ⴱ

T

冉

⳵T

⳵x

冊

, 共2兲

where Qⴱ is heat of transport. With a thermal gradient tem-perature difference across an atomic jump as ⌬T/⌬x, the thermal energy change would be 3k⌬T. When the thermomi-gration force balances the electromithermomi-gration force, we have 3k⌬T=4.98⫻10−28 _J, _which _yields _{⌬T=1.2⫻10}−5 _K

across an atomic jump. Therefore, the critical thermal gradi-ent is 400 ° C/cm. In other words, when the thermal gradient is greater than 400 ° C/cm, thermomigration force on Cu atoms would be bigger than that of electromigration force. Our IR thermal gradient measurement agrees with this cal-culation. So voids form in the chip/anode side for bump 2.

Additionally, when we performed isothermal annealing of the flip chip solder joints at 165 ° C for 90 h, by taking into account the higher temperature due to Joule heating, we found no void formation, but the Cu6Sn5IMCs grew thicker

below the UBM, yet they did not migrate to the substrate side. Also when we have changed the Cu UBM to Ni UBM, no obvious change was observed after 160 h under the same current stressing condition. So what is shown in Figs. 2and 3 is unique to the Cu UBM in thermomigration. Regarding thermal migration of Ag₃Sn IMCs, it is not clear so far and more studies are needed to verify whether they move under a thermal gradient.

In summary, thermomigration in Pb-free SnAg solder joints with Cu UBM has been investigated during electromi-gration tests under 9⫻103 A/cm2 at 150 ° C. Void forma-tion was observed in the powered bumps, regardless the di-rection of the current flow, as well as in the unpowered bumps. We propose that the fast interstitial diffusion of Cu

from the Cu UBM into Sn solder left behind supersaturated vacancies for void formation. The driving force of the Cu diffusion is the large thermal gradient accredited to Joule heating across the solder bumps. A critical thermal gradient of approximately 400 ° C/cm was found, above which ther-momigration force would be bigger than the electromigration force tested in the present study.

The authors would like to thank the National Science Council of R.O.C. for the financial support through Grant No. 95-2221-E-009-088MY3. Also Professor K.N.T. acknowledged the support from Seoul Technopark, South Korea.

1_{K. N. Tu,}_{J. Appl. Phys.} _{94, 5451}_共2003兲.

2_{C. Y. Liu, C. Chen, C. N. Liao, and K. N. Tu,}_{Appl. Phys. Lett.} _{75, 58}

共1999兲.

3_{E. C. C. Yeh, W. J. Choi, and K. N. Tu,}_{Appl. Phys. Lett.} _{80, 580}_共2002兲. 4_{H. Ye, C. Basaran, and D. C. Hopkins,}_{Appl. Phys. Lett.} _{82, 1045}_共2003兲. 5_{A. T. Huang, A. M. Gusak, K. N. Tu, and Y.-S. Lai,}_{Appl. Phys. Lett.} _88,

141911共2006兲.

6_{S. H. Chiu, T. L. Shao, C. Chen, D. J. Yao, and C. Y. Hsu,}_{Appl. Phys.}

Lett. 88, 022110共2006兲.

7_{S. W. Liang, Y. W. Chang, and C. Chen,} _{Appl. Phys. Lett.} _{88, 172108}

共2006兲.

8_{C. Y. Liu, J. T. Chen, Y. C. Chuang, K. Lin, and S. J. Wang,}_{Appl. Phys.}

Lett. 90, 112114共2007兲.

9_{T. L. Shao, S.-W. Liang, T. C. Lin, and C. Chen,} _{J. Appl. Phys.} _98,

044509共2005兲.

10_{L. Y. Zhang, S. Q. Ou, J. Huang, K. N. Tu, S. Gee, and L. Nguyen,}_Appl.

Phys. Lett. 88, 012106共2006兲.

11_{Y. H. Lin, Y. C. Hu, C. M. Tsai, C. R. Kao, and K. N. Tu, Acta Mater. 53,}

2029共2005兲.

12_{J. W. Nah, K. W. Paik, J. O. Suh, and K. N. Tu,}_{J. Appl. Phys.} _{94, 7560}

共2003兲.

13_{H. Y. Hsiao and C. Chen,}_{Appl. Phys. Lett.} _{90, 152105}_共2007兲. 14_{B. F. Dvson, T. R. Anthony, and D. Turnbull, J. Appl. Phys. 38, 3408}

共1967兲.

15_{Y. W. Chang, S. W. Liang, and C. Chen,} _{Appl. Phys. Lett.} _{89, 032103}

共2006兲.

16_{W. Jost, Diffusion in Solids, Liquids and Gases, 3rd ed.}_{共Academic, New}

York, 1960兲, Chap. 12.

17_{H. B. Huntington, Diffusion}_{共American Society for Metals, Metals Park,}

OH, 1973兲, Chap. 6.

18_{K. Zeng, and K. N. Tu,}_{Mater. Sci. Eng., R.} _{38, 55}_共2002兲. 19_{I. A. Blech,}_{J. Appl. Phys.} _{47, 1203}_共1976兲.

20_{B. Chao, S. H. Chae, X. Zhang, K. H. Lu, J. Im, and P. S. Ho,}_{Acta Mater.} 55, 2805共2007兲.

21_{D. V. Ragone, Thermodynamics of Materials}_{共Wiley, New York, 1995兲,}

Vol. 2, Chap. 8.

122103-3 Chen, Chen, and Tu Appl. Phys. Lett. 93, 122103共2008兲