Thermomigration in Pb-free SnAg solder joint under alternating current stressing

(1)

Thermomigration in Pb-free SnAg solder joint under alternating current stressing

Hsiang-Yao Hsiao and Chih Chen

Citation: Applied Physics Letters 94, 092107 (2009); doi: 10.1063/1.3089872

View online: http://dx.doi.org/10.1063/1.3089872

View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/94/9?ver=pdfcov Published by the AIP Publishing

Articles you may be interested in

Blech effect in Pb-free flip chip solder joint

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

Failure induced by thermomigration of interstitial Cu in Pb-free flip chip solder joints Appl. Phys. Lett. 93, 122103 (2008); 10.1063/1.2990047

Effect of Sn grain orientation on electromigration degradation mechanism in high Sn-based Pb-free solders Appl. Phys. Lett. 92, 211909 (2008); 10.1063/1.2936996

Study of the thermal stress in a Pb-free half-bump solder joint under current stressing Appl. Phys. Lett. 90, 232112 (2007); 10.1063/1.2747183

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

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 15:27:24

(2)

Thermomigration in Pb-free SnAg solder joint under alternating

current stressing

Hsiang-Yao Hsiao and Chih Chena兲

Department of Materials Science and Engineering, National Chiao Tung University, Hsin-chu 30010, Taiwan

共Received 26 November 2008; accepted 9 February 2009; published online 4 March 2009兲 Lead-free solders have been adopted by the microelectronics industry. However, their thermomigration behaviors are unclear. Thermomigration in eutectic SnAg3.5 solder joints was investigated using an alternating current 共ac兲 of 0.57 A at 100 °C. The ac eliminates the electromigration effect and creates a thermal gradient of 2829 ° C/cm, facilitating the study of thermomigration. Arrays of tiny markers fabricated by a focused ion beam are employed to measure the thermomigration rate. It is found that Sn atoms migrated toward the hot end. The thermomigration flux and molar heat of transport are measured to be 5.0⫻1012 _atoms_/cm2_{and 1.36}

Because of concerns about the environment impact of Pb-containing solders, the electronic packaging industry is in a hurry to replace Pb-containing solders Pb-free solders, most of which have a high Sn content of approximately over 96 wt %.1Therefore, the matrix of Pb-free solder joints con-sists of Sn grains. Additionally, electromigration in flip-chip solder joints has become an important reliability issue in recent years because of the trend toward miniaturization to meet the requirement for improved performance in portable consumer electronics.2,3 With the use of Pb-free solders, electromigration caused by increased current density is a se-rious reliability issue that must be considered in flip-chip packaging.4

Joule heating in the silicon chip generates a thermal gra-dient in a flip-chip solder joint. Since Al traces serve as the major source of heat during accelerated electromigration tests, high current stressing also produces a nonuniform tem-perature distribution, creating a large thermal gradient in a flip-chip solder joint.5This thermal gradient has been dem-onstrated to introduce a component of thermomigration dur-ing accelerated electromigration tests.6,7 In recent years, work has been undertaken to tackle the issue of thermomi-gration in SnPb solder.5–10 The temperature gradient causes Pb atoms in Pb-containing solders to move to the cold end on the substrate end and Sn atoms to migrate to the hot end on the Si side. At test temperatures of over 100 ° C, Pb atoms appear to be the dominant diffusion species, forming voids on the chip end. However, the effect of thermomigration in Pb-free solders is presently unclear. Although the Sn-rich phase segregates to the hot end for the Pb-containing solders,7,8 whether Sn migration is caused by the thermomi-gration of Sn atoms by or the exchanging of positions with the diffusing Pb atoms is unknown. As Pb-free solders are used in the packaging industry, thermomigration in Pb-free solder becomes a critical issue. It may also take place during accelerated electromigration tests.

In this study, tiny markers fabricated by a focused ion beam 共FIB兲 are adopted to monitor the thermomigration of Sn in eutectic SnAg3.5 solder joints. An alternating current

共ac兲 is employed to distinguish electromigration and thermo-migration effects. The thermal gradients that are created by current stressing are measured using infrared microscopy. Therefore, the motion of Sn atoms, the thermomigration flux, and the heat of transport can be studied using this approach. Cross-sectioned solder bumps were employed to facili-tate the detection of the movement of markers and the mea-surement of the thermal gradient inside the solder bumps. A test model, in Fig.1, was constructed. The solder joints com-prise eutectic SnAg3.5 solder bumps with electroplated 5 ␮m Cu/3 ␮m Ni under bump metallurgy 共UBM兲. To ob-serve thermomigration in situ, solder joints were cross sec-tioned into halves, polished laterally until the contact open-ings were exposed, leaving about 50% of the mass of the bump. To prevent electromigration, an ac was adopted for current stressing. The ac generates the same amount of Joule heating but without electromigration. Flip-chip packages were stressed at a temperature of 100 ° C on a hot plate. A current of 0.57 A was passed through the bump, producing a nominal current density of 1.01⫻104 _cm2_.

Before current stressing, the emissivity of the specimen was calibrated at 100 ° C, after which the bump was powered by an ac. The temperature was then measured to map the temperature after the temperature reached a steady state. The

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

[email protected]. FIG. 1.共Color online兲 Schematic of bump used in this study.

APPLIED PHYSICS LETTERS 94, 092107共2009兲

(3)

temperatures in the solder joints were mapped by QFI ther-mal infrared microscopy, with a temperature resolution of 0.1 ° C and a spatial resolution of 2 ␮m. Changes in the maker positions and the surface microstructure were exam-ined by scanning electron microscopy共SEM兲.

A huge thermal gradient can be established using this technique to study thermomigration. Figure2共a兲presents the temperature distribution in the bump before current stressing. The temperature distribution across the solder bump was then quite uniform. When current stressing commences at 0.57 A, the temperature distribution in the joint changes in response to Joule heating, as shown in Fig.2共b兲. The average increase in temperature in the bump due to current stressing was observed to be as high as 40 ° C. In the surrounding underfill, the temperature rises to 60.0 ° C. The thermal gra-dient is defined herein as the difference between the average temperatures in the bump near the substrate共Tsubs兲 and near

the chip 共Tchip兲, divided by the bump height, h:共Tchip

− Tsubs兲/h. Figure 2共c兲 shows the temperature profile along

the dashed line in Fig.2共b兲, where T_chipis higher than T_subs. The thermal gradient was measured to be 2829 ° C/cm.

To measure the heat of transport of Sn and the thermo-migration flux under a specific thermal gradient for eutectic

SnAg3.5 solder joint, inert holes on the solder surface were utilized as markers. Two rows of 0.1 ␮m holes are etched using Ga ions from a FIB on to the cross-sectional surface of the joint; one is close to the chip side: and the other is lo-cated near the substrate end. Figure3共a兲displays the markers on the cross section of the bump before current stressing. The depth of the markers is approximately 100 nm; they each had an area of about 104 _nm2_.

After current stressing, thermomigration causes observ-able mass transport. After 800 h of current stressing, hillocks are present on the chip side, as shown in Fig. 3共b兲. The hillocks comprised Sn using a SEM energy dispersive spec-trum. These hillocks are formed directly by the mass transfer of the solder caused by thermomigration. The fact that Tchip

⬎Tsubsgives rise to a thermal gradient between the chip and

the substrate side, in which the heat is generated in the Al trace. This thermal gradient causes bottom to top mass dif-fusion in the joint, driving the mass transfer of Sn toward the hot chip end. Under current stressing, the markers moved to the substrate end because of thermomigration. The bottom edge of the Si chip was used as a fixed reference frame. Figure 3共b兲 displays an image of the markers after current stressing. The markers moved in the direction opposite the diffusion flux of Sn atoms. This observation suggests that the actual direction of overall diffusion of tin atoms is from the substrate side to the chip side, indicating the influence of the thermomigration under current stressing. This result also re-veals that net mass transport proceeded from the cold end to the hot end. Notably, some voids appeared in the chip end after current stressing, as presented in Fig. 3共b兲. Why the voids form there is unknown. Lu et al.11reported that Cu and Ni atoms diffuse much more quickly along the c-axis than along the a- or b-axis of the Sn crystals, forming a significant number of voids near the Cu or Ni UBM. Further study is required to verify this assertion.

To analyze the thermomigration behavior of Sn atoms, the thermomigration flux was measured by the movement of markers. The positions of the upper-row markers were mea-sured with respect to the reference frame. The

thermomigra-FIG. 2. 共Color online兲 Temperature map of SnAg bump before current stressing, indicating a uniform temperature distribution in solder;共b兲 tem-perature map of bump powered by 0.57 A; 共c兲 temperature profile along dashed line in共b兲; a huge thermal gradient of 2829 °C/cm is generated.

FIG. 3. Cross-sectioned bump with markers before and after thermomigra-tion test at 1.01⫻104 _A_/cm2_{and 100 ° C}_{共a兲 initially; 共b兲 at 800 h. After}

the thermomigration test, the markers moved toward the substrate end.

092107-2 H.-Y. Hsiao and C. Chen Appl. Phys. Lett. 94, 092107共2009兲

(4)

tion flux J in a temperature gradient dT/dx can be written for one-dimensional case:12

JTM=

CDA共Qⴱ/N兲

␬T2 共− dT/dx兲, 共1兲

where C represents concentration; DA= Do exp共−Q/kT兲 is the coefficient of self-diffusion; k is Boltzmann’s constant; N is Avogadro⬘s number; dT/dx is the thermal gradient, and T is the temperature. Qⴱ is a constant, which is the heat of transport.

If the displacement, stressing time, and cross-sectional area of the solder bump are known, the thermomigration flux can be measured. TableIlists the positions of the upper-row markers before and after the current stressing. The markers moved downward to the substrate side. The mean displace-ment of the markers was 3.8 ␮m. The atomic flux in ther-momigration is given by

JTM=

共A⌬X兲␳N

MAt , 共2兲

where␳ is the density of the SnAg3.5 solder共7.42 g/cm3_兲;

M is the molecular weight of SnAg3.5共118.33 g/mole兲, and t is the stressing time. The mean displacement ⌬X can be obtained by measuring the displacements of these markers Thus, the atomic flux associated with thermomigration can be calculated as 5.0⫻1012 _atoms/cm2_.

The experimental atomic flux is substituted into Eq.共1兲 as is the value of DA published by Sun and Ohring.13 Table Isummarizes the measured displacements and local thermal gradients of all the markers. The average molar heat flux Qⴱ is determined to be 1.36 kJ/mole.

The results concerning motion of markers close to the substrate end follow the same trend as those close to near the chip end. The markers also moved toward the substrate side. Yet, the average displacement is only 2.2 ␮m, which is smaller than the value of 3.8 ␮m obtained from the upper-row markers. This difference may be attributed to the fact that the local temperatures at the two locations are different. As shown in Fig.2共c兲, the temperature is 148 and 132 ° C at upper row and at lower row, respectively. As shown in Eq. 共1兲, the atomic flux is a function of the temperature. If the published value of the activation energy for self-diffusion of pure tin共11 700 cal/mol兲 was used,13the calculated atomic flux ratio between the upper row and lower row comes to 1.6, assuming no difference in thermal gradient. This calcu-lated ratio is very close to the measured ratio of the marker movement 3.8/2.2=1.7.

The molar heat flux obtained herein seems quite reason-able. Kuz⬘menko and Golovinskiy14

studied the diffusion of Sn in a thallium-tin liquid alloy at 450 ° C and they obtained a molar heat flux of Sn of approximately 3.58 kJ/mole. In contrast, the testing conditions in this study were ac current stressing of 1.01⫻104 _A/cm2 _{at 100 ° C. However, most}

importantly, in this investigation, the thermal gradient was directly measured by infrared microscopy, while the tem-perature gradient used in the work of Kuz⬘menko and Golov-inskiy was assumed. The Sn atomic flux is one order of magnitude smaller than the thermomigration flux in eutectic SnPb,7–9which is dominated by Pb migration. Therefore, Pb atoms dominate the thermomigration behavior in eutectic SnPb solder.

In summary, thermomigration in eutectic SnAg3.5 solder joints in the flip-chip configuration was observed at 100 ° C under stressing with an ac of 0.57 A. A current of 0.57 A yielded a thermal gradient of 2829 ° C/cm. Thermomigra-tion, which is caused directly by the aforementioned thermal gradient, induces the transport of Sn atoms to the hot end of the bump. As a result, hillocks were formed above the chip side. Based on the movements of the markers fabricated on the cross-sectional surface of the bump by FIB, the molar heat of transport of Sn was calculated as 1.36 kJ/mole.

The authors would like to thank the National Science Council of the Republic of China, Taiwan, for financially supporting this research under Contract No. 95-2221-E-009-088MY3. Ted Knoy is appreciated for his editorial assistance.

1_{K. N. Tu, Solder Joint Technology}_{共Springer, New York, NY, 2007兲.} 2_{K. N. Tu,}_{J. Appl. Phys.} _{94, 5451}_共2003兲.

3_{C. Y. Liu, C. Chen, C. N. Liao, and K. N. Tu,}_{Appl. Phys. Lett.} _{75, 58}

共1999兲.

4_{C. Chen and S. W. Liang,}_{J. Mater. Sci.: Mater. Electron.} _{18, 259}_共2006兲. 5_{S. H. Chiu, T. L. Shao, and C. Chen,}_{Appl. Phys. Lett.} _{88, 022110}_共2006兲. 6_{H. Ye, C. Basaran, and D. C. Hopkins,}_{Appl. Phys. Lett.} _{82, 1045}_共2003兲. 7_{A. T. Huang, A. M. Gusak, K. N. Tu, and Y.-S. Lai,}_{Appl. Phys. Lett.}

88,141911共2006兲.

8_{A. T. Huang, A. M. Gusak, and K. N. Tu,}_{Appl. Phys. Lett.} _{88, 141911}

共2006兲.

9_{Y. C. Chuang and C. Y. Liu,}_{Appl. Phys. Lett.} _{88, 174105}_共2006兲. 10_{H. Y. Hsiao and C. Chen,}_{Appl. Phys. Lett.} _{90, 152105}_共2007兲. 11_{M. Lu, D. Y. Shih, P. Lauro, C. Goldsmith, and D. W. Henderson,}_Appl.

Phys. Lett. 92, 211909共2008兲.

12_{P. G. Shewmon, Thermo- and Electrotransport in Solids}_共TMS,

Warren-dale, PA, 1989兲, Chap. 7.

13_{P. H. Sun and M. Ohring,}_{J. Appl. Phys.} _{47, 478}_共1976兲.

14_{P. P. Kuz}_{⬘menko and L. P. Golovinskiy, Fiz. Met. Metalloved. 30, 1304}

共1970兲.

TABLE I. Measured displacements and thermal gradients at upper-row markers. The reference frame was set at the bottom edge of the Si die.

Marker number 1 2 3 4 5 6 7 8

Marker position before current stressing共␮m兲 ⫺15.1 ⫺15.1 ⫺15.1 ⫺15.1 ⫺15.2 ⫺15.3 ⫺15.3 ⫺15.3

Marker position after current stressing共␮m兲 ⫺19.1 ⫺19.1 ⫺19.1 ⫺19.1 ⫺19.1 ⫺19.1 ⫺19.0 ⫺18.5

Marker displacement共␮m兲 ⫺4.0 ⫺4.0 ⫺4.0 ⫺4.0 ⫺3.9 ⫺3.8 ⫺3.7 ⫺3.2

Thermal gradient共°C/cm兲 2971 2829 2829 2829 2700 2229 2143 1571

Calculated Q共kJ/mole兲 1.17 1.23 1.23 1.23 1.26 1.48 1.50 1.77

092107-3 H.-Y. Hsiao and C. Chen Appl. Phys. Lett. 94, 092107共2009兲