THE competitive nature of electronic packaging industry

Transcription

1 IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES 1 Solder Joint Fatigue in a Surface Mount Assembly Subjected to Mechanical Loading Mohd N. Tamin, Yek Ban Liew, Amir N. R. Wagiman, and W. K. Loh Abstract A mechanical approach employing cyclic twisting deformation to a surface mount assembly is examined as an alternative to temperature cycling for evaluating solder joints fatigue performance. This highly accelerated test is aimed at reducing solder joint reliability testing cycle time. In this study, the mechanics of solder joints in a surface mount assembly subjected to cyclic twisting deformation is investigated. For this purpose, a test package with 24 by 24 peripheral-array solder joints is modeled using the finite element method. Unified inelastic strain theory defines the strain-rate-dependent plastic stress-strain response of the 60 Sn 40 Pb solder. Cyclic twisting deformation in the range of 1 per 50 mm length of the assembly board at a rate of 120 s per twist cycle was applied. The calculated stress and strain distributions in the critical solder joint are compared with those predicted for temperature cycling and accelerated temperature cycling tests. Results showed that the accumulated inelastic strain concentrates in a small region of the critical solder joint near the component side for temperature cycling and near the board side for cyclic twisting load cases. The rate of inelastic strain accumulation per fatigue cycle in the solder joint for both temperature cycling (TC1) and mechanical twisting (MT1) tests are similar. Thus, mechanical twisting test imparts similar deformation characteristics to temperature cycling tests in terms of the shear strain range. Low cycle fatigue is dominated by localized shear effect as reflected in the largest shear strain range of the hysteresis loops. The Coffin Manson strain-based model yielded more conservative prediction of fatigue lives of solder joints when compared to energy-based approach. Index Terms Finite element analysis (FEA), low cycle fatigue, mechanical cyclic twisting (MT1), reliability, temperature cycle (TC1). I. INTRODUCTION THE competitive nature of electronic packaging industry calls for a faster method of reliability assessment to meet the fast pace of product development cycle. In this respect, accelerated temperature cycling tests and mechanical cycling tests have been evaluated as an effective acceleration technique to thermal cycling [1] [4]. Temperature cycling of solder assemblies in an environmental chamber is intended to duplicate the characteristics of the actual stress factors experienced by the assembly during processing, reliability testing and service. These include temperature and humidity, shock and vibration, thermal or power cycles. Accelerated temperature cycling profile consists of thermal shock cycles or introduces dwell time period at peak temperature levels. The resulting cyclic stresses and strains induced low cycle fatigue damage in the solder alloy [5] [8]. Mechanical cyclic load test is capable of accelerating the failure by fatigue of solder joint. Fatigue failure occurs by initiation and slow propagation of a crack to a critical length prior to final fracture of the solder joint. Bending tests are useful for evaluating board level assemblies for performance under mechanical stresses. Three- and four-point bend tests have been used to simulate manufacturing, handling, shipping and network conditions of surface mount assemblies and for evaluating mechanical reliability of chip scale packages [2], [4], [9] [12]. Several packages can be tested at multiple load levels with a single three-point bend test. A direct strain measurement method was used for peripheral leaded device where mechanical strain is induced by the application of axial cyclic loading [13]. Torsion test imparts cyclic twisting deformation on an assembled PCB to mimic solder joint fatigue failure mechanisms as observed in temperature cycling tests. Due to twisting load on area array solder joints, diagonal solder joints experience higher strains similar to that observed in a package subjected to thermal load [3]. Finite element analyses (FEAs) have been employed extensively to evaluate fatigue performance and reliability of solder joints in surface mount assemblies [6], [7], [12]. FEA simulations are able to provide insights on low cycle fatigue failure mechanism of solder joint. Its success lies in appropriate representation of the assembly, boundary conditions, applied loads and accurate material model. This study examines solder joint fatigue in a surface-mount test assembly under cyclic twisting deformation. Evolution characteristics of inelastic strains in critical solder joint under temperature cycling (TC1), accelerated temperature cycling by dwell time period (TD1) and mechanical twisting cycles (MT1) are compared in view of validating the accelerated life techniques. Fatigue life prediction models based on strain range [14], [15], and energy density [16] are used to estimate fatigue life of these solder joints. Manuscript received April 20, 2006; revised February 25, This work was supported in part by Intel Technology (Malaysia) under UTM-Intel Contract Research Grant This work was recommended for publication by Associate Editor A. Chandra upon evaluation of the reviewers comments. M. N. Tamin and Y. B. Liew are with the Universiti Teknologi Malaysia, Johor 81300, Malaysia ( taminmn@fkm.utm.my; yekban@yahoo.com). A. N. R. Wagiman and W. K. Loh are with Intel Technology (Malaysia), Penang 11900, Malaysia ( amir.nur.rashid.wagiman@intel.com; wei. keat.loh@intel.com). Color versions of one or more of the figures in this paper are available online at Digital Object Identifier /TCAPT II. FINITE ELEMENT PROCEDURES A surface-mount test assembly with a peripheral array of 92 solder joints, Si-die, FR-4 substrate, Cu Sn intermetallics and layers of copper traces is modeled, as illustrated in Fig. 1(a) for use in temperature cycling cases. In mechanical twisting cases, a full (global) model is subjected to a prescribed twist deformation at the rigid grip ends of the assembly. This full model simulation established the deformation field and identified the location of critical solder joints. Due to symmetry /$ IEEE

2 2 IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES Fig. 1. (a) Exploded view of the quarter model test assembly. (b) Sub-model of the assembly with refined finite element mesh for mechanical twisting. of the observed deformation response, only one quarter of the assembly is considered in a sub-model with refined finite element mesh [Fig. 1(b)]. The global displacement field is imposed on symmetry planes of the quarter sub-model. The diameter, stand-off height and pitch distance of the solder joints is 750 m, 600 m and 1.0 mm respectively. All components are assumed perfectly bonded at contact surfaces. The thin layer of copper traces (0.025 mm) is modeled with one layer of linear incompatible modes elements to accurately reproduce the flexing behavior of thin foil. Previous works have observed that fatigue crack path closely follows pad-solder interface [4], [12]. Consequently, an element layer of 50 m-thick at the solder interface is used to average the plastic strain energy density for use in the life prediction model. Unified inelastic strain theory defines the visco-plastic response of the 60 Sn 40 Pb (wt. %) solder alloys [17] [19]. The constitutive stress-strain equations are represented by Anand model as [17] where is the inelastic strain rate, and are coefficient and strain-rate sensitivity exponent, respectively, is the activation energy for creep, is the gas constant, is temperature in Kelvin scale, is the stress multiplier and is the current stress. The deformation resistance, is given by an evolution equation as The parameters is the hardening/softening constant and represents the corresponding strain rate sensitivity. The saturation value of the resistance, is given by with coefficient, and exponent,. The numerical values of all parameters have been optimized for 60 Sn 40 Pb solder [19]. Other materials are assumed to behave elastically throughout the temperature range of interest. In addition, orthotropic behavior of FR-4 substrate is considered. Temperature dependent (1) (2) (3) properties of these materials are extracted from published literature and summarized in Table I, [5], [17] [21]. Loading parameters for temperature cycling (TC1), accelerated temperature cycling by dwell time period (TD1) and mechanical twisting cycles (MT1) employed in this study are summarized in Table II. The effects of residual stresses and strains due to solder reflow process on subsequent thermal fatigue performance of solder joints are examined for thermal fatigue cases. III. RESULTS AND DISCUSSION The finite element results are presented and discussed in terms of the evolution characteristics of stresses and inelastic strains in the critical solder, the stress-strain hysteresis loops and the predicted fatigue life of the solder using available life models. The critical solder joint that had accumulated the greatest inelastic strain during cyclic twisting deformation is the corner solder. In temperature cycling test, the solder joint located at the symmetry plane parallel with the longer side of the Si-die is the most severely strained solder. A. Stress and Strain Distribution The von Mises stress distribution in the critical solder is compared for both temperature loading and mechanical twisting, as shown in Fig. 2. The stress scale is in MPa. Both temperature cycling and mechanical twisting display similar distribution of stresses in the solder joint for a given geometry of the assembly. Higher stress gradient is predicted for regions near the edge of the solder joint at solder/intermetallic interfaces. Such high stress is favorable to localized yielding of the solder alloy. The corresponding inelastic strain distribution for the cyclic twisting case [Fig. 2(b)] is illustrated in Fig. 3. Inelastic strain concentration is confined to a small region near the solder interface while the bulk of the solder remains elastic. It is worth noting that stress and inelastic strain concentration in the critical solder joint is greatest near the component (top) side for temperature cycling case and near the board (bottom) side for cyclic mechanical twisting. This result is consistent with those observed of the fatigue cracking that preferentially propagate in the solder at the board side during bending test and near the component side in thermal cycle tests [2]. In temperature

3 TAMIN et al.: SOLDER JOINT FATIGUE IN A SURFACE MOUNT ASSEMBLY 3 TABLE I MATERIAL PROPERTIES EMPLOYED IN THE STUDY TABLE II LOAD CASES FOR THERMAL AND MECHANICAL CYCLES EMPLOYED IN THE STUDY cycling tests, stress relaxation occurred with accompanying visco-plastic strain in the solder, thus temperature cycles accelerate inelastic strain accumulation. Similar creep related failure mechanism can also be induced in the solder joint by performing isothermal cyclic mechanical twisting test of the assembly at elevated temperature. The evolution of inelastic strain in the solder joint for the three fatigue loading cases considered is illustrated in Fig. 4. Both thermal fatigue cases (TC1 and TD1) started with an initial residual inelastic strain of 1.37% induced during the solder reflow process. Solder reflow was not modeled for mechanical twisting case. Results show that the rate of inelastic strain accumulation per fatigue cycle in the solder joint for both temperature cycling (TC1) and cyclic mechanical twisting (MT1) tests is similar as depicted by the same slope of the curves. The 10-min dwell time period at both peak temperature levels (TD1) accelerates the accumulated inelastic strain. Stress relaxation that manifests over the hold-time period at peak temperature levels is accompanied by visco-plastic deformation. Cyclic twisting test, if conducted at higher temperature is expected to generate similar creep-fatigue interaction effect. B. Stress-Strain Hysteresis The stress-strain hysteresis response of the solder alloy for the first five mechanical twist cycles of the assembly is shown in Fig. 5. The shear component (23-direction, parallel to the solder/pad interface plane) displays the largest total strain range of 1.2% with the corresponding shear stress range of 13.0 MPa. This large relative displacement within the solder/intermetallic interface attenuates the cyclic interfacial shear stress and strain resulting in crack initiation, as often observed experimentally [2]. The largest stress range of 50.0 MPa is predicted for the normal component (22-direction, parallel to the axis of the solder joint). The tensile part of this stress is favorable to Mode-I crack initiation at the brittle solder/intermetallic interface layer. The ranking of stress and strain ranges by magnitude is similar for both cases of temperature cycling and cyclic mechanical twisting. The fifth hysteresis loops of the shear stress-strain component for all load cases are compared in Fig. 6. The hysteresis loops for temperature cycles, TC1 and TD1 are shifted away from the coordinate origin due to the locked-in residual stress and strain induced during solder re-flow process. Although the shapes of the shear loops are different for thermal and mechanical fatigue, the stress and plastic strain ranges (for TC1 and MT1) are comparable. This plastic strain range is postulated to govern fatigue life of the solder joint. Larger strain range for load case TD1 suggests shorter expected fatigue life (cycles) of the solder joints compared to TC1. C. Solder Fatigue Life Prediction Since low cycle fatigue of solder is governed by temperature, strain rate and stress dependence occurrence of dissipative deformation processes, both inelastic strain and dissipative strain energy or plastic work represent suitable indicators

4 4 IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES Fig. 2. Stress distribution in the critical solder: (a) at 25 C after accumulating four temperature cycles, TC1 and (b) at 0.5 twist during the fifth cycle. (Note the different orientation of each critical solder joint). Fig. 3. Inelastic strain distribution in the critical solder corresponding to the mechanical stress shown in Fig. 2(b). Fig. 5. Hysteresis loops for Cartesian stress-strain components during mechanical twist cycles at 25 C. Indices 11, 22, and 33 refer to the cartesian coordinate axes. Fig. 4. Evolution of inelastic strain in the solder joint for different fatigue load cases considered. Fig. 6. Comparison of shear stress-strain hysteresis loops for mechanical twist cycles (MT1), temperature cycles (TC1) and accelerated temperature cycles (TD1). to evaluate cyclic damage. The calculated plastic shear strain range, for temperature cycling load cases is used with the Coffin Manson relation in the fatigue life prediction model. The coefficient and exponent of the strain-based model have been

5 TAMIN et al.: SOLDER JOINT FATIGUE IN A SURFACE MOUNT ASSEMBLY 5 TABLE III PREDICTED FATIGUE LIVES OF SOLDER JOINTS BY STRAIN-BASE AND ENERGY-BASED MODELS determined for 60Sn 40 Pb solder alloy at as follows [14], [15] 50, 35 and 125 C where is the fatigue life (cycles) of the solder. Darveaux forwarded a fatigue life prediction model for viscoplastic solder based on plastic strain energy density [16], [22], [23] (4) (5) (6) (7) where is the cycles for the crack initiation, is the characteristic crack length (pad diameter), and is the crack propagation rate. The term is the volume averaging of the plastic work density accumulated per cycle for elements at the interface with thickness of 50 m. This volume averaging technique reduces the sensitivity of the calculated strain energy density to mesh size. The calculated strain energy density or plastic work density, and the corresponding fatigue lives of the solder for the three load cases are summarized in Table III. Although the absolute fatigue life prediction is desirable, a life prediction accuracy of 50 pct is generally adequate in view of the complex nature of solder creep behavior and uncertainty in the board level temperature cycling tests [24]. In addition, the different creep models and finite element assumptions only affect the absolute fatigue life, but not the relative trend of fatigue performance [24], [25]. This relative design trend can be evaluated in terms of plastic work density, with reasonable range of fatigue lives. Previous work on different TFBGA packages with eutectic solders indicated that ranges from to MPa [24]. In an experimental validation study, 2512-type resistors with 60Sn 40Pb solders were subjected to temperature ranges of 55 to 125 C (4.5 C/min) and 40 to 120 C (4.0 C/min). The calculated are 7.3 and 6.0 MPa, respectively [26]. The low magnitude of plastic work density calculated in this study ( MPa or psi for load case TC1) is attributed to the fast temperature ramp rate of 33 C/min. This fast temperature ramp rate lowers the relative accumulation of inelastic strains per cycle, thus contributing to longer characteristic life. As expected, it is noted that the accelerated temperature cycling by 10-min dwell time period (TD1) results in shorter life (smaller number of fatigue cycles to failure) of solder joint when compare to temperature cycling (TC1). However, the corresponding reliability testing time is longer. In this respect, thermal shock test is perhaps a favorable alternative for accelerated test. Cyclic mechanical twisting (MT1), for similar rate of inelastic strain accumulation, predicts a moderate number of fatigue cycles to failure but at much shorter testing time. Isothermal mechanical test at elevated temperature could further accelerate the fatigue damage process. Correlations between cyclic mechanical twisting data and temperature cycling test data can be made using Weibull statistics to obtain the acceleration or scale factor for the given geometry and design of the assembly. In addition, isothermal test condition and mechanical twisting load magnitude and frequency are easily achieved in an actual reliability test. In this respect, FEA simulations can be effectively utilized to optimize the accelerated cyclic mechanical test profile. IV. CONCLUSION Solder joint fatigue in a surface mount assembly subjected to cyclic twisting deformation have been analyzed using finite element method. Results show that: the accumulated inelastic strain concentrates in small region near the component side for temperature cycling and near the pad side for the cyclic twisting load cases; similar rate of inelastic strain accumulation per fatigue cycle in the solder joint for temperature cycling (TC1) case can be achieved in mechanical twisting (MT1) test; low cycle fatigue is dominated by localized shear effect as reflected in the largest shear strain range of the hysteresis loops; the Coffin Manson strain-based model yielded more conservative prediction of fatigue lives of solder joints when compared to energy-based approach. REFERENCES [1] J. H. Lau and Y. H. Pao, Solder Joint Reliability of BGA, in CSP, Flip Chip, and Fine Pitch SMT Assemblies. New York: McGraw-Hill, [2] D. T. Rooney, N. T. Castello, M. Cibulsky, D. Abbott, and D. Xie, Materials characterization of the effect of mechanical bending on area array package interconnects, Microelectron. Rel., vol. 44, pp , [3] H. L. J. Pang, K. H. Ang, X. Q. Shi, and Z. P. Wang, Methodology for highly accelerated solder joint reliability test, in Proc. Electron. Packag. Technol. Conf., 2000, pp [4] S. Shetty and T. Reinikainen, Three and four point bend testing for electronic packages, ASME J. Electron. Packag., vol. 125, pp , [5] M. Amagai, Chip scale package (CSP) solder joint reliability and modeling, Microelectron. Rel., vol. 39, pp , [6] H. L. J. Pang, Sensitivity study of temperature and strain rate dependent properties on solder joint fatigue life, in Proc. IEEE Electron. Packag. Technol. Conf., 1998, pp [7] A. Yeo, C. Lee, and H. L. J. Pang, Flip chip solder joint fatigue life model investigation., in Proc. Electron. Packag.Technol. Conf., 2002, pp [8] C. J. Zhai, Sidharth, and R. II. Blish, Board level solder reliability versus ramp rate and dwell time during temperature cycling, IEEE Trans. Devices Mater. Rel., vol. 3, no. 4, pp , Dec [9] J. H. Lau, Solder joint reliability of flip chip & PBGA assemblies under thermal, mechanical and vibrational condition, IEEE Trans. Compon., Packag. Manufact. Technol. B, vol. 19, no. 4, pp , Nov

6 6 IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES [10] E. A. Stout, N. R. Sottos, and A. F. Skipor, Mechanical characterization of plastic ball grid array package flexure using moire interferometry, IEEE Trans. Adv. Packag., vol. 23, no. 4, pp , Nov [11] R. Ghaffarian, Accelerated thermal and mechanical testing of CSP assemblies, NASA Parts Packag. Progr., Nov. 9, [12] R. Dudek, M. Nylen, A. Schubert, B. Michel, and H. Reichl, An efficient approach to predict solder fatigue life and its application to SM and area array components, in Proc. Electron. Comp. Technol. Conf., 1997, pp [13] D. J. Xie, Y. C. Chan, J. K. L. Lai, and I. K. Hui, Fatigue life estimation of surface mount solder joints, IEEE Trans. Compon., Packag., Manufact. Technol. B, vol. 19, no. 3, pp , Aug [14] S. S. Manson, Thermal Stress and Low Cycle Fatigue. New York: McGraw-Hill, [15] H. D. Solomon, Fatigue of 60/40 solder, IEEE Trans. Compon., Hybrids, Manufact. Technol., vol. CHMT-9, no. 4, pp , Dec [16] R. Darveaux and A. Mawer, Thermal and power cycling limits of plastic ball grid array (PBGA) assemblies, in Proc. Surface Mount Technol., 1995, pp [17] L. Anand, Constitutive equations for hot working of metals, J. Plastic., vol. 1, pp , [18] K. Sasaki, K. Ohguchi, and H. Ishikawa, Viscoplastic deformation of 40 Pb/60 Sn solder alloys Experiment and constitutive modeling, ASME J. Electron. Packag., vol. 123, pp , [19] Y. K. Koh, A Unified Constitutive Model for Solder Materials, M. Eng. thesis, Dept. Mech. Eng., Univ. Teknol. Malaysia, Johor, [20] J. Auersperg, Fracture and damage evaluation in chip scale packages and flip chip assemblies by FEA and microdac, in Proc. ASME Symp. Appl. Fract. Mech. Electron. Packag., Dallas, TX, 1997, pp [21] X. Q. Shi, Z. P. Wang, W. Zhou, H. L. J. Pang, and Q. J. Yang, Anew creep constitutive model for eutectic solder alloy, ASME J. Electron. Packag., vol. 124, pp , [22] M. N. Tamin and Y. B. Liew, Numerical modeling of cyclic stressstrain behavior of Sn-Pb solder joint during thermal fatigue, in Proc. Sem. Computat. Exper. Mech., 2005, pp [23] R. Darveaux, Solder joint fatigue life model, in Proc. Surface Mount Technol. Conf., 1997, pp [24] H. S. Ng, T. Y. Tee, K. Y. Goh, J. Luan, T. Reinikainen, E. Hussa, and A. Kujala, Absolute and relative fatigue life prediction methodology for virtual qualification and design enhancement of lead-free BGA, in Proc. Electron. Compon. Technol., 2005, pp [25] B. A. Zahn, Solder joint fatigue life model methodology for 63 Sn37 Pb and 95.5 Sn4Ag05 Cu materials, in Proc. Electron. Comp. Technol., 2003, pp [26] M. Dusek, J. Nottay, C. Hunt, H. Lu, and C. Bailey, An Experimental Validation of Modeling for Pb-Free Solder Joint Reliability, Nat. Phys. Lab, Middlesex, U.K., Tech. Rep., NPL MATC (A) 11, Oct Mohd N. Tamin received the B.Sc. degree in mechanical engineering from Northrop University, Inglewood, CA, in 1984, the M.Sc. degree in mechanical engineering from Washington State University, Pullman, in 1987, and the Ph.D. degree in mechanical engineering and applied mechanics from University of Rhode Island, Kingston, in His primary field of research is computational solid mechanics, fatigue and fracture mechanics. He is currently a Professor with the Department of Applied Mechanics, Faculty of Mechanical Engineering, Universiti Teknologi Malaysia, Skudai, Johor, Malaysia. He is also the Head, Computational Solid Mechanics Laboratory (CSMLab) at the university. His current research on the development of predictive models for reliability of microelectronic components involves micromechanics analysis on fatigue crack initiation and propagation in solder interconnections using cohesive damage concept. Other active research projects include finite element modeling of structural composites for high-strain-rate loading, fatigue and fracture characterization of advanced alloys and welded connections, modeling of materials constitutive behavior and failure investigation of engineering components. Yek Ban Liew received the B.Eng. and M.Eng. degrees in mechanical engineering from Universiti Teknologi Malaysia, Skudai, Johor, Malaysia, in 2003 and 2007, respectively. In March 2007, he became a Process Equipment Engineer at Intel Technology, Kulim, Kedah, Malaysia. Prior to this, he joined Intel Technology as a Graduate Trainee in April, His fields of interest include thermal fatigue, solder joint reliability, product assembly tests, and high volume manufacturing. Amir N. R. Wagiman received the B.Sc degree in mechanical engineering from Tufts University, Medford, MA, in 1990 and the M.Sc. degree in mechanical engineering from Rensselaer Polytechnic Institute, Troy, NY, in Currently, he is the Engineering Manager for the Substrate Materials Group, Intel Technology, Kulim, Kedah, Malaysia. His current job focuses on development of new materials, processes and reliability monitoring metrology tools for the next generation electronic packaging. W. K. Loh, photograph and biography not available at the time of publication.