IEEE TRANSACTIONS ON COMPONENTS, PACKAGING, AND MANUFACTURING TECHNOLOGY PART B, VOL. 20, NO. 1, FEBRUARY

IEEE TRANSACTIONS ON COMPONENTS, PACKAGING, AND MANUFACTURING TECHNOLOGY PART B, VOL. 20, NO. 1, FEBRUARY 1997 87 Effect of Intermetallic Compounds on the Thermal Fatigue of Surface Mount Solder Joints P. L. Tu, Yan C. Chan, Senior Member, IEEE, and J. K. L. Lai Abstract The effect of Cu-Sn intermetallic compounds (IMC) on the fatigue failure of solder joint during thermal cycling has been studied. The samples consist of components [leadless ceramic chip carrier (LCCC)] soldered onto FR-4 printed circuit board (PCB), and are prepared by conventional reflow soldering using a 63Sn-37Pb solder paste. The specimens are subjected to thermal cycling between 035 C and 125 C with a frequency of two cycles per hour until failure. The results indicate that the fatigue lifetime of the solder joints depends on the thickness of IMC s layer between Cu-pad and bulk solder, and the relation of the lifetime to the thickness can be described as a monotonically decreasing curve. The lifetime is very sensitive to the thickness of the IMC when the thickness is less than 1.4 m. During thermal cycling, the thickness of the IMC layer increases and then the interface between IMC and solder becomes gradually flatter. The results of X-ray diffraction and scanning electron microscope (SEM) analysis show that cracks propagate along the interface between the IMC layer and the solder joint. The Cu 3Sn ("-phase) is also found to form between the Cu-pad and -phase during thermal cycling. On the basis of the above results, the thick and flattened IMC layer is shown to responsible for the fatigue failure of solder joint during thermal cycling. The results of this paper can be used to optimize the reflow soldering process for the fabrication of robust solder joints. Index Terms Fatigue failure, intermetallic, solder joint, surface mount technology (SMT), thermal cycles. I. INTRODUCTION THE reliability of solder joints is one of the critical issues in surface mount technology (SMT) printed circuit board (PCB) assemblies. Due to the mismatch of the coefficient of thermal expansion (CTE) between surface mount components and the PCB, thermal fatigue is a very common failure mode of solder joint in electronic components [4] [6], [8] [11], [14]. Therefore, there needs to be better understanding of thermal fatigue behavior of solder joints. In this research field, the intermetallic layer between the solder and the Cu-pad has drawn the most attention. When molten Sn-Pb solder contacts with the Cu-pad surface, the intermetallic compounds (IMC) are formed between the solder and the pad. The IMC layer is entirely made up of Cu-Sn intermetallic species ( -phase, -phase et al.) [1] [3], [7], [11], [13] and serves as a bonding Manuscript received May 6, 1995; revised November 1, 1996. This paper was supported by the Universities Grants Council of Hong Kong (CERG Project 904109). The authors are with the Department of Electronic Engineering, City University of Hong Kong, Kowloon, Hong Kong (emails: 94871340@plink.cityu.edu.hk; eeycchan@cityu.edu.hk; apjoelai@cityu.edu. hk). Publisher Item Identifier S 1070-9894(97)01438-2. material for the solder joints. The thickness of the IMC layer depends on the reflow time and temperature [3] [5], [9]. Too thick IMC may result in mechanical failure in the operating environment, such as thermal or power cycling. Thus the electrical performance and lifetime of the joint are degraded as a result of the intermetallic growth [11]. The results of most researches indicate that the growth of IMC layer plays a degrading role in the mechanical strength of solder joints. It is found that the mechanical strength decreases with the increase in the amount of IMC ( -phase) at the interface [3], [7], [11]. However, different results have been reported. No failure in the intermetallic interface of solder joints in electronic packages was observed at a slow deformation rate [5], [10]. In another situation, when the thickness of the IMC ( -phase) layer was increased from 0.7 1.3 m, the shear force, required to fracture the solder joints, increased by about 20% and reached a maximum value [14]. Obviously, the effect of intermetallic compounds on the property of solder joint is very complicated, especially in the case of thermal fatigue. The aim of this paper is to study the effects of Cu-Sn IMC on the solder joint failure during thermal cycling. A thermal cycling method is employed to test the lifetime of solder joint prepared by a usual SMT process. Metallographic, scanning electron microscope (SEM) and X-ray diffraction methods are used to examine the microstructures of the solder joints at different stages of thermal cycling. The results of this paper can contribute to achieving the optimal technique to control the IMC growth in a production process, and hence to devising a good method for maximizing the operating life of surface mount solder joints. II. EXPERIMENTAL PROCEDURE A schematic configuration of the samples is shown in Fig. 1. Surface mount passive components [leadless ceramic chip carrier (LCCC): 1206, Resistor: 10 ] were assembled on FR-4 PCB using a standard infrared reflow. The solder paste used is a 63Sn/37Pb eutectic alloy (MULTICORE SN63 ABS90). Care was taken to keep the quantity of solder paste printed on each copper pad fairly constant so that the effect of solder thickness on the formation and growth of the Cu- Sn intermetallic layer could be minimized. The thickness of the paste was about 150 m, as measured by laser section microscopy. The assemblies were preheated to 100 C for 100 s and then reflowed inside a three-zone infrared oven (PRECISOLD 1070 9894/97$10.00 1997 IEEE

88 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING, AND MANUFACTURING TECHNOLOGY PART B, VOL. 20, NO. 1, FEBRUARY 1997 TABLE I THERMAL SHOCK CHARACTERISTICS FOR SAMPLES WITH DIFFERENT IMC THICKNESS Fig. 1. Schematic of the surface mount solder joint. Fig. 2. Thermal shock conditions. PS-3000). The parameters of the solder technique is shown in Table I. In order to investigate the effect of IMC on the fatigue failure, samples with various IMC thickness were obtained by controlling the soldering time or temperature, and were then subjected to thermal shock, which was performed in an air to air condition in a TABAI TSA-70L thermal shock chamber. The temperature range of thermal cycling was from 35 C to 125 C and samples were held at each temperature for 15 min, as shown in Fig. 2. The electrical resistance of the solder joint was measured continuously by a computer via analog-todigital/digital-to-analog (AD/DA) cards to capture the fatigue failure of the solder joints [4], [5], [9], [12]. For the sake of comparing the electrical properties of samples with different thickness, hundreds of solder joints were monitored under the same experimental conditions. Failures were determined by a persistent detected peak in electrical resistance during thermal shock, which is indicative of the development of cracks in shear deformation. The mean thickness of the interface intermetallic layer was measured using a powerful image processing system (OPTIMAS) and a Nikon optical microscope. Apart from the infrared reflow, the thermal shock can also affect the IMC thickness. A SEM and an optical microscope were used to observe the micro-structural details of the solder joints in order to study the crack morphology formed during thermal shock. The phase and crystal structure of different samples were identified by means of X-ray diffraction using a Siemens D-500 diffractometer with a Cu target. III. RESULTS AND DISCUSSION The IMC thickness was measured using an image analyzer. The results are shown in Fig. 3. It can be seen that the IMC thickness increases with the reflow time, according to exponential growth law. That is different to previously reported results [15]. The samples with various IMC thickness were subjected to temperature shock in order to evaluate the influence of IMC on solder joint failure. The fatigue lifetime of different sets of samples is summarized in Table II. The reliability of the solder joints is modeled by using Weibull s probability density function and analyzed by Johnson s statistical method with ranking [9]. The two-parameter Weibull cumulative distribution function has the following form where is a random variable (i.e., number of cycles to failure in the present study), is a shape parameter (the Weibull slope), and is a scale parameter (a characteristic value). Applying the principles of least squares method and ranking to the temperature cycling test results (Table II), the best fit Weibull parameters ( and ) are calculated, and summarized in Table I. Using these values and (1), the life distribution of solder joints, with different IMC thickness under these test conditions, are shown in Fig. 4. It is found that the thinner the IMC layer, the greater the number of cycles to failure as the Weibull distribution curve moves to higher. The lifetime distribution can be (1)

TU et al.: SURFACE MOUNT SOLDER JOINTS 89 Fig. 3. Growth of IMC thickness with reflow time. Fig. 5. IMC thickness versus number of cycles to failure at 50% and 1% failure. Fig. 4. Weibull plot of the LCCC solder joint lifetime with different IMC thicknesses(k). TABLE II THERMAL CYCLING TEST RESULTS FOR STATISTICAL ANALYSIS CYCLES TO FAILURE (1 CYCLE = 31 min) characterized by the scale parameter ( ). Table I shows the - value for different IMC thickness. This parameter is equivalent to the number of cycles to failure when 50% of samples have failed ( ). The early failure 1% level ( ) and the value of as a function of IMC thickness is shown in Fig. 5. The relation of the (similar to ) to the IMC thickness can be described as a monotonously decreasing curve. For various, the lifetime decreases rapidly until 1.4 m and then the lifetime decreases more slowly. The, there are of more practical interest than 50% failure times, possesses same character with. When the IMC thickness increases to 1.4 m from 0.95 m, the decreases to 30%. The results reveal that there is a tradeoff between the IMC s thickness and the solder joint quality. From Fig. 4, the optimal reflow time is around be 20 seconds and the value of is 0.95 m. It should be noted that the solder joints cannot be formed successfully when the reflow time is less than 20 s. Microstructure of the failed solders has been observed by metallographic, SEM, and X-ray to clarify the failure mechanism. The cross section morphology of the solder joint is shown in Fig. 6. In Fig. 6(a), it can be observed that the Cu-Sn intermetallic ( -phase) layer is very thin, in this case, the precipitate don t constitute a major feature microstructure nor do it appear to be involved in thermal fatigue of solder joints [5], so the joint lifetime is long. For much increased reflow time, the IMC layer becomes thicker and continuous [Fig. 6(b)]. As a result, the solder joint lifetime decreases rapidly. In this case, the bulk solder and IMC interface is very uneven. The longer the reflow time, the rougher the interface. The thick IMC layer enlarges the probability of the failure of solder joints (a negative effect), whereas the uneven interface gives a positive effect on the joint performance. This explain why the reduction rate of the joint lifetime decreases as the initial IMC thickness increases. When the IMC thickness is more than 1.4 m, the positive effect caused by the uneven factor increases rapidly. Hence, the lifetime decreases less rapidly with IMC thickness. The changes of the IMC in terms of thickness and shape can also contribute to solder joint failure. The increase of the intermetallic layer thickness can be monitored by metallographic examination during thermal shock. The IMC thickness increases linearly with the square root of the cycle number, and does so more rapidly for the samples reflowed for longer

90 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING, AND MANUFACTURING TECHNOLOGY PART B, VOL. 20, NO. 1, FEBRUARY 1997 (a) (b) Fig. 6. SEM micrographs of solder joints showing the Cu-Sn IMC layersafter (a) reflowing for 20 s and (b) reflowing for 600 s. times. For specimens reflowed for 100 s, the function of the IMC thickness as the number of thermal cycles is shown in Fig. 7. This function can be described by the following [9] where is the layer thickness and is hold time at 125 C, proportional to. is a diffusion coefficient, given by (2) (3) here is a diffusion constant ( m /s) [13], is an activation energy (1.09 ev) [13], is Boltzmann constant, and is the absolute temperature. The test condition is described elsewhere [13]. To compare the experimental result with (2), the calculated value of the equation is also plotted in Fig. 7. The difference between both curves can be attributed to the error in measuring the temperature and to the difference of energy. The inner temperature of the solder is difficult to measure, and the temperature of the chamber is substituted for the temperature of the solder joints. As the test temperature is exchanged, the IMC interface is held at 125 C for a time shorter than 15 min. The activation energy may also be large in the condition of exchange temperature. The phenomenon is useful to correct the IMC thickness calculation and the fatigue failure prediction, in the case of the thermal variation. The shape change of the IMC during thermal shock is shown in Fig. 8. The IMC morphology is different from the one formed during reflow. One of the differences is the shape of the IMC. The interface between the IMC and the bulk solder in the thermal cycled samples is smooth. Another is its chemical components. The Cu Sn ( -phase) is formed by thermal shocks between the Cu and the -phase, assuming that there is no - phase and only Cu Sn -phase [3] is present in the solder before thermal shock. The phase and crystal structure of the Cu/solder interface was identified by X-ray analysis and the results are shown in Fig. 9. The smooth IMC layer can degrade the solder joint lifetime more than the uneven layer, due to its poor ability to resist shearing. The growth of the IMC, during thermal shock, affected the lifetime of the solder joints. This is a dominant Fig. 7. IMC thickness after a different number of thermal cycles. Fig. 8. Cross section morphology of the solder joint (Reflowing: 230 C 2100 s and number of thermal shock cycles: 1000). reason for solder joint failure. Compared to the shape of the IMC layer, the -phase formed by consuming the -phase have only a small effect on the lifetime because the crack does initiate and grow in the boundary of -phase. In the samples reflowed for a long time, the IMC layer is thick and uneven, which allows the IMC to grow rapidly and to reduce the solder joint lifetime. The thick and flattened IMC layer is responsible for initiating the crack, as shown in Fig. 10.

TU et al.: SURFACE MOUNT SOLDER JOINTS 91 Fig. 9. The X-ray diffraction pattern of the solder cross section (Reflowing: 230 C2100 s and number of thermal shock cycles: 1000). Fig. 10. Micrographic section of the bottom fracture surface. The failure mechanism can be revealed by X-ray diffraction and SEM analysis of the fracture surface of the failed solder joint. Figs. 10 and 11 show that the cracks do initiate and propagate along the IMC/bulk solder interface. Normally, there exists a difference of CTE between the PCB and the components (e.g., ppm/ C, ppm/ C). The CTE mismatch will result in a greater shear stress in the solder joints during the temperature change or power cycling [9], thus thermal fatigue failure of the solder occurs. It is reported by D. Frear [10] and S. M. Le [5] that all failures occur in the bulk solder of the 60Sn-40Pb joint. However, the presence of an intermetallic compound may cause cracks, which propagate along the intermetallic/cu interface during the pull-off test, according to Bi-Shiou et al. [11]. Moreover, our result shows that the cracks also can initiate and propagate along the IMC/bulk solder interface (see Fig. 10). After the components are removed, the bottom fracture surface of solder joint can be analyzed by X-ray diffraction and the results are shown in Fig. 11. On the bottom of fracture surface, there are intermetallics of -phase and -phase. A small amount of Sn and Pb-rich phases also exists on fracture surface, which does not mean the crack occurs in bulk solder because the phases only adhere at the interface. Therefore, the crack is mainly in the IMC/solder interface. There is also another type of intermetallic phase between the component metallization and the solder, and the effect of this intermetallic phase is not yet certain. Another possible mechanism for failure is related to the mode of deformation. The super-plastic eutectic Sn-Pb alloy

92 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING, AND MANUFACTURING TECHNOLOGY PART B, VOL. 20, NO. 1, FEBRUARY 1997 Fig. 11. X-ray diffraction pattern of the bottom fracture surface of the surface mounted assemble. deforms by grain rotation and grain boundary sliding [10]. During thermal shock, when the PCB is displaced relative to the component (LCCC), the strain would be mainly due to the bulk solder. The IMC layer is harder than the bulk solder (Hv -,Hv ) [9], and so the shear stress will concentrate in the IMC/solder interface and lead to crack nucleation and growth. According to the above results and discussions, the intermetallic compounds can greatly affect the fatigue lifetime of solder joints. Thus it is important to control and monitor the IMC micro-morphogenesis to improve the reliability of SMT assemble. IV. CONCLUSION The effects of Cu-Sn IMC on solder joint failure during thermal shock are summarized as follows: 1) The Cu Sn IMC is formed between the Cu-pad and the solder joint, and no Cu Sn phase exists. The longer the reflow time, the thicker the IMC layer. The IMC/bulk solder interface is found to be uneven. The longer the reflowing time, the rougher the IMC/solder boundary. 2) During thermal cycling, the IMC thickness increases linearly with the square root of cycle number. The IMC layer flattens gradually at the interface between the IMC and the solder, and the -phase is formed between Cupad and -phase. The flat IMC/solder boundary degrades the solder joint performance and the fatigue lifetime. 3) The lifetime of solder joints goes down rapidly with the increase of IMC thickness during reflow. However, the lifetime decreases slowly once thickness of the IMC exceeds about 1.4 m. Our experimental results show that the best reflow condition are 230 C for 20 s. 4) During thermal cycling, the fatigue failure occurs mainly in the IMC/solder interface. The Cu Sn -phase, formed during the thermal cycling, has little influence on the lifetime. REFERENCES [1] J. K. Wassin, Soldering in Electronics. New York: Electrochemical Publications, 1990. [2] H. H. Manko, Solders and Soldering: Materials, Design, Production and Analysis for Reliable Bonding. New York: McGraw Hill, 1964, pp. 135 139. [3] A. C. K. So and Y. C. Chan, Reliability studies of surface mount solder joints -Effect of Cu-Sn intermetallic compounds, in Proc. 45th Electron. Comp. Conf., 1995, pp. 1073 1080. [4] D. R. Frear, Thermomechanical fatigue of solder joint: A new comprehensive test method, IEEE Trans. Comp., Hybrid, Manufact. Technol., vol. 12, pp. 492 501, Apr. 1989. [5] S. M. Le et al., Grain boundary sliding in surface mount solders during thermal cycling, IEEE Trans. Comp., Hybrid, Manufact. Technol., vol. 14, pp. 628 633, Mar. 1991. [6] V. K. Garj, Reliability on surface mount solder joint, in Proc. 4th. Int. Microelectronics Systems 95 Conf., June 19 23, 1995. [7] L. E. Felton, Copper-Tin Intermetallics Joints. Troy, NY: Rensselaer Polytechnic Institute, 1991, pp. 88 120. [8] Y.-H. Pao et al., An experimental and finite element study of thermal fatigue fracture of PbSn solder joints, ASME Trans. J. Electron. Packag., vol. 115, pp. 1 8, Mar. 1993.

TU et al.: SURFACE MOUNT SOLDER JOINTS 93 [9] D. R. Fear, H. S. Morgan, S. N. Burcheit, and J. H. Lau, The Mechanics of Solider Alloy Interconnects. New York: Van Nostrand Reinhold, 1994, pp. 42 86, 361 378. [10] D. Frear, D. Grivas, and J. W. Morris, A microstructural study of the thermal fatigue failures of 60Sn-40Pb solder joints, J. Electron. Mater., vol. 17, no. 2, pp. 171 180, 1988. [11] B. S. Chiou, J. H. Change, and J. G. Du, Metallurgical reactions at the interface of Sn-Pb solder and electroless copper-plated AIN substrate, IEEE Trans Comp.,, Packag., Manufact. Technol., vol. 18, pp. 537 542, Aug. 1995. [12] J. H. Constable and C. Lizzul, An investigation of solder joint fatigue using electrical resistance spectroscopy, IEEE Trans. Comp., Packag., Manufact. Technol., vol. 18, pp. 142 152, 1995. [13] A. C. K. So and Y. C. Chan, Aging studies of Cu-Sn intermetallic compounds in annealed surface mount solder joints, in Proc. 46th. Electron. Comp. Conf., Orlando, FL, May 1996. [14] S. F. Dirnfeld and J. J. Ramon, Microstructure investigation of coppertin intermetallics and the influence of layer thickness on shear strength, Welding J. Suppl., pp. 373-s 377-s, Oct. 1990. [15] A. C. K. So and Y. C. Chan, Reliability studies of surface mount solder joints Effects of Cu-Sn intermetallic compounds, IEEE Trans. Comp., Packag., Manufact. Technol., vol. 19, pp. 661 668, 1996. P. L. Tu received the B.Eng. degree from Beijing University of Aerospace and Astronautics, China in 1982, and the M.Eng. degree from Nanjing University of Aerospace and Astronautics, China, in 1988. From 1989 to 1995, he was employed as a Lecturer at the Department of Materials Science at the Nanjing University of Aerospace and Astronautics, China. Since 1995, he has worked as a Research Assistant in the Department of Electronic Engineering, City University of Hong Kong. His current research focuses on fatigue life analysis, solder joint defect studies, and reflow soldering. Yan C. Chan (M 85 SM 95) received the B.S. degree in electrical engineering, the M.S. degree in materials science, and the Ph.D. degree in electrical engineering from the Imperial College of Science and Technology, University of London, U.K., in 1977, 1978, and 1983, respectively. He joined the Advanced Technology Department of Fairchild Semiconductor, CA, as a Senior Engineer. In 1985, he was appointed to a Lectureship in Electronics at the Chinese University of Hong Kong. From 1987 to 1991, he worked in various senior operations and engineering management functions in electronics manufacturing including SAE Magnetic (HK) Ltd. and Seagate Technology. He set up the Failure Analysis and Reliability Engineering Laboratory for SMT PCBA at Seagate Technology, Singapore, prior to joining the City University of Hong Kong as a Senior Lecturer in Electronic Engineering in 1991, and was promoted to University Senior Lecturer in 1993. His current technical interests include surface mount technology, electronic materials, and component reliability. Dr. Chan is Chairman of the IEEE Hong Kong Centre. J. K. L. Lai graduated with first class honors from Keble College, Oxford University, U.K., in 1971. From 1974 to 1985, he was employed as a Research Officer at the Central Electricity Research Laboratories, Leadherhead, Surrey, U.K. In 1984, he was appointed Project Leader of the Remaining Life Study Group and a member of the Remnant Life Task Force of the Central Electricity Generating Board, U.K. He is now a Professor in the Department of Physics and Materials Science and Head of the Materials Research Centre, City University of Hong Kong. He has performed over forty cases of constancy in the area of failure analysis of metallic components in Hong Kong. He holds one patent and has published over 50 technical papers in refereed international journals and conference proceedings. He is also the author of over 20 industrial reports on failure analysis and materials for engineering design, issued by the Central Electricity Generating Board, U.K. Dr. Lai is a member of the International Institute of Weld Working Group on Creep, the Pressure Equipment Advisory Committee of the Labour Department, the Plastics Training Board of the Vocational Training Council, and the Consumer Council of Hong Kong.