Influence of Solder Reaction Across Solder Joints Kejun Zeng FC BGA Packaging Development Semiconductor Packaging Development Texas Instruments, Inc. 6 th TRC Oct. 27-28, 2003 Austin, TX 1
Outline Introduction Solder reactions of metals Dissolution of metals in molten solders IMC formation during reflow IMC formation during solid state aging Solder reactions across joints Cu 6 Sn 5 appears on Au/Ni(P) pad of substrate AuSn 4 forms throughout solder joint Influence of metals on reliability across joints Au enhances consumption of Ni(V)/Cu UBM Ni plating is consumed by Cu 6 Sn 5 Summary 6 th TRC Oct. 27-28, 2003 Austin, TX 2
Introduction With the electronic devices being continuously scaled down, solder reaction is becoming one of the major concerns for packaging reliability. Due to the Pb-free requirement, new surface finishes have appeared in the market and more are being studied. Persistence of the Black Pad problem results in the application of OSP- Cu or bare Cu. Local effects of solder reaction on joint reliability has been extensively studied, but its effects on the other side of the joint is relatively new to the industry. Theories for the study of solder reactions: Local equilibrium Reaction path Dominant diffuser 6 th TRC Oct. 27-28, 2003 Austin, TX 3
Dissolution of metals in molten solders Stable phase diagram Stable Metastable Metastable phase diagram W Actual concentration of Cu in molten solder at the interface can be higher than what the stable phase diagram indicates, depending on the surface condition of Cu and the time above the eutectic T. 6 th TRC Oct. 27-28, 2003 Austin, TX 4
1 1 Solubility (Mole fraction) 0.1 0.01 0.001 Au Ag Pd Cu Ni Solubility (Mole fraction) 0.1 0.01 0.001 Au Cu Pd Ni 0.0001 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 1000/T(K) Metal solubility in Sn-Pb 0.0001 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1 1000/T(K) Metal solubility in Sn-Ag Thermodynamically calculated equilibrium saturation solubility of metals in molten solders. (1) Ni has the lowest solubility in solders. (2) Solubility of metals in Sn-Ag, SnCu, and Sn-Ag- Cu solders is higher than in Sn-Pb solder. 6 th TRC Oct. 27-28, 2003 Austin, TX 5
Metal Solder Schematic concentration profile of metal in moleten solder. Higher solubility greater gradient higher diffusion rate higher dissolution rate. Dissolution rates of various metals in the 60Sn-40Pb solder as a function of T [Bader, 1969]. Note that the latest experimental results showed that the solder reaction of Pd was very fast [Wang & Tu, 1995]. 6 th TRC Oct. 27-28, 2003 Austin, TX 6
IMC formation during reflow 350 C 400 C Reaction path (arrows exaggerated for readability). After saturation solubility is reached, Cu 6 Sn 5 forms at interface. Because Cu 6 Sn 5 is not in equilibrium with Cu, Cu 3 Sn appears between Cu 6 Sn 5 and Cu. Following the same procedure, it is predicted that in high Pb solder joint Cu 3 Sn is the first IMC to form. This is in agreement with experimental results by Grivas et al., 1986. 6 th TRC Oct. 27-28, 2003 Austin, TX 7
Cu 6 Sn 5 220 C Cu 3 Sn Cu Eutectic SnPb/Cu joint after 1 reflow. SEM image (Texas Instruments). Cu 6 Sn 5 Cu 3 Sn Ni(V) After reflow process, Cu 6 Sn 5 is usually observed in the eutectic solder joint. A thin layer of Cu 3 Sn forms but is not easy to see. Al Al SiO 2 Si 1µm Eutectic SnPb/Cu/Ni(V)/Al joint after 3 reflows. Voids are marked by red arrows. TEM image. (Courtesy of K. N. Tu, UCLA) 6 th TRC Oct. 27-28, 2003 Austin, TX 8
IMC formation during solid state aging 175 C Solder Cu 6 Sn 5 Cu 3 Sn Cu During solid state aging, both Cu 6 Sn 5 and Cu 3 Sn grow thicker. SnPb/Cu joint after 40 days at 150 C. Since Cu is the dominant diffuser in Cu 6 Sn 5, Kirkendall voids form in the Cu 3 Sn layer. SIM image (Texas Instruments). 6 th TRC Oct. 27-28, 2003 Austin, TX 9
Ni can be dissolved into AuSn 4 by replacing gold atoms. This is because the dissolution of Ni can lower Gibbs energy of AuSn 4. Thermodynamic calculation shows that the max. solubility is 10 at.% or 4.9 wt.% and the max. energy change is 3 kj/mole of atoms. SnPb/Au/Ni/Cu BGA joint after 500 hours at 160 C. During reflow, Au plating was totally dissolved away. After aging, Au diffused back to the interface and form (Au,Ni)Sn 4. SEM image. (Courtesy of C. R. Kao, National Central University, Taiwan). 6 th TRC Oct. 27-28, 2003 Austin, TX 10
Cu 6 Sn 5 appeared on other side Cu Solder Ni Cu 6 Sn 5 (Cu,Ni) 6 Sn 5 Ni Cu Ni 3 Sn 4 Solder a) After 3 reflows b) After 150 C/1000 hour baking SnPb flip chip bumps on Ni/Au pad. Cu in the interfacial compounds came from Cu UBM on the die side. (Cu,Ni) 6 Sn 5 composition (at.%): 46.7Cu, 8.2Ni, 45.1Sn. (Texas Instruments) This is an indication that Ag and Pd can also reach the other side of a joint after reflow. 6 th TRC Oct. 27-28, 2003 Austin, TX 11
AuSn 4 formed throughout joint Au Eutectic Sn-Pb solder cap on Au at 200 C. a) after 5 seconds, b) after 60 seconds. AuSn 4 compound has extended all the way to the surface of the cap. With a diffusivity of 10-5 cm 2 /s, Au atoms can diffuse a distance of 100 µm in 5 seconds to the cap top. (Kim & Tu, 1996). 6 th TRC Oct. 27-28, 2003 Austin, TX 12
Au enhances consumption of Ni(V)/Cu UBM FR4 Ni(P): 10 µm Au: 0.125 µm Cu pad Cu: 300 nm Solder Ni(V): 400 nm Al: 400 nm Passivation layer Si SiO 2 Joint structure Solder/Cu/Ni(V)/Cu/Al Ni(P)/Au/solder/Cu/Ni(V)/Al Solder Sn-37Pb Sn-3.5Ag-1.0Cu Sn-37Pb Sn-3.5Ag-1.0Cu Fraction of Ni(V) dissolved by molten solder No dissolution after 20 reflows 60% after 10 reflows 30% after 10 reflows; 100% after 20 reflows 50% after 1 reflow; 100% after 3 reflows (See Zeng & Tu, Mater. Sci. Eng. Reports, 2002) 6 th TRC Oct. 27-28, 2003 Austin, TX 13
Au/Ni(P) pad Au/Ni(P) pad SnAgCu bump SnAgCu bump (a) Cu/Ni(V)/Al UBM (b) Cu/Ni(V)/Al UBM SEM images of cross-section of a eutectic SnAgCu flip chip joint. (a) As-bonded. (b) After 10 reflows, IMC crystals have been spalled into bulk solder. Ni(V) layer has been completely dissolved. The presence of Au has enhanced the dissolution of Ni(V) layer. (Courtesy of M. Li, Institute of Materials Research and Engineering, Singapore) 6th TRC Oct. 27-28, 2003 Austin, TX I N N O V A T E. C R E A T E. M A K E T H E D I F F E R E N C E. 14 TM
Cu 6 Sn 5 compound Cr Si 5 µm 1µm (a) SnPb on Au/Cu/Cr UBM. Cu 6 Sn 5 scallops were decorated with small particles. [Liu&Tu, 1996] (b) SnPb on Cu/Ti UBM. Surface of Cu 6 Sn 5 scallops were smooth. [Kim&Tu, 1996] Cu 6 Sn 5 r Cu 6 Sn 5 R Ni(V) Ni(V) When Au is present, Cu 6 Sn 5 crystals become spherical, exposing Ni(V) to the solder. 6 th TRC Oct. 27-28, 2003 Austin, TX 15
Ni plating is consumed by Cu 6 Sn 5 Ni (Cu,Ni) 6 Sn 5 Solder If one side of a joint has Ni plating and the other side has bare Cu or OSP-Cu, the Cu may decrease the reliability of the Ni plating. Here is a such an example. After 1000 hours of baking at 150 C, Cu 6 Sn 5 had dissolved a great amount of Ni from the Ni plating so that the Ni plating became porous.(texas Instruments) 6 th TRC Oct. 27-28, 2003 Austin, TX 16
Thermodynamically calculated phase diagram showed that the maximum solubility of Ni in Cu 6 Sn 5 is 6Cu:4Ni, corresponding to 21.8 at.% Ni. Experimentally, the Ni content in the interfacial (Cu,Ni) 6 Sn 5 layer in page 16 was measured by EDX to be 21 at.%. The driving force for the dissolution of Ni into Cu 6 Sn 5 is the Gibbs energy change. 6 th TRC Oct. 27-28, 2003 Austin, TX 17
Summary Ranking of dissolution rate of metals corresponds to that of saturation solubility of the metals, with Au being the fastest and Ni the slowest. Dissolution rate of metals in the eutectic SnAg solder is higher than in the eutectic SnPb solder. Dissolved metals may diffuse across the solder joint during reflow process, altering the solder reactions on the other side of the joint and thus influence the joint reliability. When assessing solder joint reliability, influence from the other side of the joint should also be considered if the surface coatings are different on the two sides. Interfacial reaction is controlled by thermodynamics (Gibbs energy change) and diffusion kinetics. 6 th TRC Oct. 27-28, 2003 Austin, TX 18
References W. G. Bader, Weld. J. Res. Suppl., 1969, 28, 551s-557s. D. Grivas, D. Frear, L. Quan, and J. Morris, J. Electr. Mater., 1986, 15, 355-359. Y. Wang, H. K. Kim, H. K. Liou, and K. N. Tu, Scr. Metall. Mater., 1995, 32, 2087-2092. P. G. Kim and K. N. Tu, J. Appl. Phys., 1996, 80, 3822-3827. H. K. Kim, K. N. Tu, and P. A. Totta, Appl. Phys. Lett., 1996, 68, 2204-2206. A. A. Liu, H. K. Kim, K. N. Tu, and P. A. Totta, J. Appl. Phys., 1996, 80, 2774-2780. C. Y. Liu, K. N. Tu, T. T. Sheng, C. H. Tung, D. R. Frear, and P. Elenius, J. Appl. Phys., 2000, 87, 750-754. C. E. Ho, W. T. Chen, and C. R. Kao, J. Electr. Mater., 2001, 30, 379-385. K. Zeng and J. K. Kivilahti, J. Electr. Mater., 2001, 30, 35-44. K. N. Tu and K. Zeng, Mater. Sci. Eng. Reports, 2001, 34, 1-58. K. Zeng and K. N. Tu, Mater. Sci. Eng. Reports, 2002, 38, 55-105. 6 th TRC Oct. 27-28, 2003 Austin, TX 19
Acknowledgements The thoughts on this topic and the approaches taken in the study were initiated and developed when the author was with Lab of Electronics Production Technology, Helsinki University of Technology, Finland Electronic Thin Film Lab, University of California at Los Angeles, USA The author would like to thank Prof. J. Kivilahti and Prof. K. N. Tu for inspiring discussions. Support from Texas Instruments is acknowledged. 6 th TRC Oct. 27-28, 2003 Austin, TX 20