Aging Studies of PBGA Solder Joints Reflowed at Different Conveyor Speeds

Transcription

1 Aging Studies of PBGA Solder Joints Reflowed at Different Conveyor Speeds S. H.,Fan*, Y. C. Chant, C. W. Tang and J. K. L. La? Department of Electronic Engineering and Department of Physics and Materials Science City University of Hong Kong Tat Chee Avenue, Kowloon, Hong Kong Tel: Fax: shfan0ee.citvu.edu.l-k Abstract PBGA packages are more susceptible to solder joint fatigue problem due to the significant mismatches of the coefficient of thermal expansion (CTE) between bismaleimide triazine substrate (-15ppmPC) and silicon die (-2.5ppmPC). In this paper, the shear cycle fatigue properties of PBGA (Topline DC169) solder joints reflowed with different profiles, and aged at 125OC for 4, 9, 16, 25, and 36 days are studied. The profiles are devised to have the same heating factor, which is defined as the integral of the measured temperature above the liquidus (183OC) with respect to dwell time in the reflow profile, but to have different conveyor speeds. The effects of conveyor speed on the solder joint (non-aged and aged) fatigue lifetimes are investigated. It is found that with increasing the conveyor speed the solder joint shear fatigue lifetime can be improved substantially. Moreover, the shear fatigue lifetimes of aged specimens decrease with extending aging time and the lifetime variation of faster conveyor speed specimens is larger. From our previous work, the heating factor can characterize the reflow profile in the melting section. Using the theory of heat transmission, it can show that solder joint cooling rate during solidification increases with increasing conveyor speed. SEM and optical micrographs show that faster cooling rate causes a rougher interface of the bulk solderlimc and the refinement of the grain size in the solder joints. The thickness of IMC increases with increasing aging time, and the growth rate of samples with faster cooling rate is larger. SEM cross-section views reveal that cracks initiated at the point of the acute angle near the solder pad, then propagated along the interface of the bulk solder/imc layer. The variation in the roughness of the bulk solder/imc interface is likely to be responsible for the fatigue lifetime difference of non-aged solder joints. On the other hand, since the growth of IMC is a diffusion-controlled process, i.e., tin diffuses into copper, and the diffusion activation energy in the grain boundary region is far smaller than that in the lattice region, so diffusion is most likely to be occurred in the coarse grain boundary region. Hence, the IMC growth rate of samples with faster cooling rate is larger and the growth of IMC layer decreases the fatigue lifetimes of aged solder joints. Introduction Recent advances in semiconductor devices have produced small feature sizes, increasing gate count and chip I/O. This trend has put increased emphasis on microelectronic packaging. Ball grid array (BGA) has become the package of choice. The solder balls on BGA packages work as both electrical I/O and mechanical support. Therefore, the mechanical properties and strengths of solder balls become critical to the reliability of BGA packages. On the other hand, BGA packages are more prone to solder joint fatigue problems than other SMT (such as QFP and CGA) packages. This is because the BGA component interconnection to the PCB is only through solder balls, and the standoff height is much smaller than that of other SMT packages. Cu-Sn intermetallic compound (IMC), which serves as the mechanical bonding between the tin-lead solder and the copper pad in a surface mounted solder joint, forms instantaneously when Sn-Pb solder melts on the Cu pads. A thick Cu-Sn IMC layer in surface mount solder joints may not only be created by the long reflow time and high reflow temperature during soldering, but also by prolonged storage and long term operation of the electronic assembly even at room temperature [1,2]. Because of its brittle nature and microstructural mismatch with Sn-Pb solder and copper, too thick an IMC layer causes the interface in the solder joint to be more sensitive to stress [3,4], especially under the action of cyclic loading. The effect is much more severe in small dimension BGA solder joints. This paper investigates the effect of solder alloy microstructure on the kinetics of IMC formation in eutectic soldedcopper PBGA solder joints formed by different reflow profiles. The thickness growth of IMC layer is examined as function of annealing (at 125 C) time. The result of the variation of solder joint fatigue lifetime with annealing time is also reported in this paper. Our research result may contribute to better understanding of diffusion behavior and microstructural evolution of IMC at the eutectic soldedcopper substrate interface in PBGA solder joints, so as to optimize the reflow parameters. Experimental Procedure In this work, PBGA 169T1 S-DC10 components obtained from Topline were soldered to PCBs using various reflow conditions. The details of the experimental procedures are described as follows. 0 Sample Details The 169-pin (solder ball) PBGA components have a peripheral dimension of 23 mm square and its BT substrate thickness is 0.5 mm. The 63wt%Sn/37wt%Pb solder ball has a nominal volume of 0.23 mm3. The solder pad pitch is 1.5. The PCB size is 50 mm x 90 mm, with a thickness of 1.6. The diameter of copper pad on PBGA and PCB side is mm and mm. After reflow, the solder joint has a /00/$ IEEE Electronic Components and Technology Conference

2 maximum diameter of mm in the middle part with the standoff height of mm. Soldering Process A micro-placer machine (MP-2000 SERIES BGA & Fine- Pitch Micro Placer) was used to mount the BGAs on PCB after printing of no-clean flux paste. After mounting, the testing boards were reflowed in a five-zone oven (BTU VIP- 70N) in compressed air environment. The time-resolved temperature reflow profile near the solder bulk and the free gas fluid in the oven chamber were recorded simultaneously by using a wireless profiler (Super M. 0. L. E. E /10). Figure 1 shows a typicaltemperature profile. The PCB and BGA metallizations joined in daisy chains, which can be used to perform electrical continuity tests. Ol aoo 1:OO 2:OO 3:OO 4:OO 5:OO 6:OO TOO &flow time (Mn.) Figure 1. Typical reflow temperature profile measured near the solder (T.,.) and in the ambient gas (Tm). Two inspection procedures are performed on the PBGA package after reflowing. First, an electrical continuity measurement was made to ensure that all of the solder joints are intact. Secondly, an X-ray inspection system (SOFTEX PRO-TEST 125) was used to inspect for solder bridges and any other anomalies in the solder joints. 0 Reflow Parameters According to the general Newton Law of Cooling for convection heat transfer, for a particular time interval At (t,- f2), the system average heat exchange (heating or cooling) rate can be expressed as: t2 q=-=(a~h,(t,(t)-t,(t))dt}/at Q (1) At JI where Q is the system s total heat exchange between a body and a surrounding free fluid from time tl to t2, A is the total heat exchange area, h, is the mean coefficient of heat transfer, and T, and T, are the surface temperature of the object and the temperature of the free fluid respectively. Figure 1 is a typical reflow profile for the PBGA. The 183OC melting temperature of the eutectic solder is defined as the reference temperature line - the liquidus. In our previous work [5], it was found that the integral of the measured I temperature T,. (t) above the liquidus line with respect to time, defined as heating factor, can characterize the reflow profile in the melting section. In this work, the profiles were devised to have the same heating factor, but with different conveyor speeds (IO, 30, 50 idmin for profile #I, #2 and #3 respectively). Our previous work [6] showed that, by varying the conveyor speed, solder joint cooling rate was changed greatly. In the Appendix, we calculate the value of the mean coefficient of heat transfer (AnS. 0 Shear Cycling An INSTRON MINI 44 stress-testing machine was used to test the solder joint shear cycling life. The experimental set-up is schematically shown in Figure 2. In order to reduce the strain on the epoxy-glass substrate when the specimen was cyclically strained, the distance between the steel crosshead and the grip anvil was set to a small value and kept constant among the different test specimens. The test specimen (coupon) was cut carefully from 169-pin assembly in such a way that one specimen contains 3 rows (3 x 13) of solder balls (see Fig.2). The tests were performed at room temperature and 60% relative humidity. The solder joints were cyclically sheared to constant total displacement amplitude. The shear load change with time was recorded, especially the peak loading of each shear cycle. The cycle, which the peak loading drops to 50% of that of the initial cycle, is defined as the fatigue life. The frequency of the tests was 0.43 Hz. The shear displacement amplitudes were set to value of 0.18 mm. The sample size was 15 for each kind of test. Solder Jojnt Grip Anvil Specimen Figure 2. The experimental set-up and the specimen size and geometry for shear cycling. Fatigue Life Distribution In this paper, the lifetime distribution of solder joints is modeled by the Two-Parameter Weibull cumulative distribution function, which has the following form [7, 81: F(x) = 1 - e -(XIQIB (2) In (2), x is the value of the random variable - number of cycles to failure (N) in the present study, B is the shape parameter (Weibull slope), and Q is the scale parameter Electronic Components and Technology Conference

3 (characteristic value). The best fit Weibull parameters (for the median rank) B and Q can be obtained by using the principles of least squares and ranking. The x value corresponding to F (x) = 0.5 is called the number of cycles to failure at 50% failure rate (N50N). In this study, the fatigue lifetime of solder joints is characterized using the parameter Nsosa. Results and Discussion time. The result illustrates that the cooling rate of solder joint solidification has a marked influence on the growth rate of IMC layer. The thickness growth of IMC of the solder joints formed by profile 3# increases from 1.15~ (4-day aging) to 3.55~ (36-day aging), while the IMC thickness growth of the solder joints formed by profile I# increases from 0.97~ to 2.83~. 0 Microstructure Morphology a Square root of aging time days Figure 4. Growth of IMC thickness of samples reflowed by different profiles (#3+, #2x and #lo) vs. square root of aging days. 0 Shear Fatigue Test Result Samples manufactured by different reflow profiles and aged for various days were subject to shear cycling test in order to evaluated the influence of IMC growth on the solder joint fatigue properties. Figure 3. The SEM micrographs of interfacial microstructures for the solders reflowed by (a) Profile #1, (b) profile #2, (c) profile #3. Figure 3 (a-c) are SEM (Philips XLAOG) cross-section micrographs of solder/cu interface of as-solidified solder joints reflowed using profiles #1, #2 and #3. The globular, single layer structure of q-phase Cu6Sn5 interfacial IMC is found in the as-solidified solder joints. Duplex structure, with q-phase Cu6SnS IMC next to the solder and -phase Cu3Sn at the q-phaselcopper interface of interfacial IMC layer was observed in all annealed solder joints. The mean thickness of the IMC layer growth in solder joints annealed for various times were measured with the aid of a powerhl image processing system, OPTIMAS, used in conjunction with a Nikon optical microscope. Since the magnification of optical microscopy is limited, the thickness of the -phase IMC layer, smaller than 1 pn, cannot be resolved independently with acceptable accuracy. So the mean total thickness of the IMC layer was measured. Figure 4 shows the measured result. The IMC thickness increases linearly with the square root of aging n I \ o, = g z Square root of aging days Figure 5. Fatigue lifetime (N50N) of samples reflowed by different profiles (#3 +, #2 x and #lo) vs. aging days. Figure 5 illustrates the test results (NSosa). As shown tkom the figure, the lifetime decreases as the IMC layer in the solder joints thickens and the variation of faster conveyor speed samples is larger Electronic Components and Technology Conference

4 Failure Mechanisms Scanning electron microscopy (SEM) was used to analyze the solder joint failure and fracture by investigating the microstructure, crack initiation and propagation. Figure 6 is the initial crack cross-section SEM image of a solder joint caused by shear cycling test. Figure 7 is a fully developed crack cross-section SEM view of a failed solder joint. It can be seen that the crack initiated at the point of the acute angle near the solder pad, then propagated along the interface of the bulk solder/l;l-phase IMC layer. This result coincides with our previous research results with LCCC solder joints [9]. solder joint. Figure 7. Cross-section SEM view of crack in a failed sold; joint. Solder Microstructures Figures 8 (a-c) show the optical micrographs of bulk solder using profiles # 1, #2 and #3. The dark and light colored areas are the Pb-rich and Sn-rich phases respectively. As cooling rate increases, the grain size of the solder alloy becomes smaller. Fig. 8(a) (profile #1) has a typical lamelldcolony appearance: the two phases are arranged side by side in long range, differently oriented arrays that form colonies. In Fig. 8(b) (profile #2), the colony size becomes smaller and the lamellae become shorter. In Figure 8(c) (profile #3), the lamellae are hardly observable, and the colonies cannot be distinguished. Most areas have been replaced with the equiaxed Pb-rich phase embedded in a Sn-rich matrix. This is because for solder alloy solidified at a faster cooling rate, more nucleus seeds are available and the phase growth is slower. (c) Figure 8. Optical microstructures of solder bulk reflowed by (a) Profile # 1, (b) profile #2, (c) profile #3. 0 Discussion For better understanding the mechanism of intermetallic formation and growth in the solderkopper substrate system, there are several aspects to be focused: First, whether Cu or Sn is the dominant diffusing species? As a general rule for binary diffusion couples, the element with the lower melting point has the larger diffusion constant. Therefore, for Cu-Sn system, tin should diffuse faster. Experiments reported in the literature [ 101 show that Cu6Sn5 and Cu3Sn5 intermetallics form in the copper substrate below any diffision barrier layers (Sn, Ni, CO, Cu6Sn5, etc.) added between the copper and tin. This suggests that the tin diffise into the copper much more rapidly than the copper diffuse into tin. Secondly, Electronic Components and Technology Conference

5 which rate-limiting factor, the diffusion-controlled growth or the interfacial reaction-controlled growth, is the main ratelimiting factor for the intermetallic formation and growth? The results of previous researches [ 1,2,9,11] as well as in this work demonstrate that the relationship between the thickness of IMC layer and aging time is a linearly correlated one, thus the growth is a diffusion-controlled process. So the growth of IMC can be described as [ 121: - d =Jot (3) where d is layer thickness, D is the diffusion coefficient and t is the aging time. The diffusion coefficient is given by the Arrhenius equation: where Do is the diffusion constant, Q is the diffusion activation energy for the growth of interfacial Cu-Sn IMC layer, k is the Boltzmann constant, and Tis the absolute temperature. The third concem should be focused on the factors affecting the tin diffusion progress (or the diffusion activation energy Q). At the beginning of IMC growth, tin is available in the adjacent solder region next to the IMC/solder joint interface. However, along with the aging, the amount of tin available in that region will reduce, so tin diffuses from the inner solder region to the IMC/solder joint interface region. The microstructure and composition of the solder matrix should have influence on the diffusion process. Previous researches [11,13] have shown that, after adding particles of Cu, Cu3Sn, Cu6Sns, Ag, Au, and Ni in eutectic solder alloy, the activation energy for the growth of IMC change dramatically. In this study, by increasing the conveyor speed from 10 to 50 inches per minute during reflow, the solder joint cooling rate during solidification increases 30 times. As the results, the interface of the bulk solder/imc becomes rougher and the grain size of solder alloy becomes smaller. The variation in the roughness (Fig.3) of the bulk solder/imc interface is likely to be responsible for the fatigue lifetime difference of non-aged solder joint [6]. On the other hand, comparing to grain lattice, the structure in grain boundary region is more open, the diffusion activation energy in the grain boundary region is far smaller than that in the lattice region. So for the growth of IMC during aging, the diffusion in the coarse grain boundary region is dominated. Faster cooling rate results in a refined grain size in solder microstructure, and the grain boundary region is more abundant than the slower cooling rate solder alloy microstructure. So the IMC growth rate of the faster cooling rate solder joints is greater than that of slower cooling rate solder joints (see Fig. 4). IMC layer is more brittle than the bulk solder and mismatch in microstructure with Sn-Pb solder and copper, during shear cycling, the shear stress is easily concentrated at the interface of solder bulk / IMC layer, formed higher stress regions, increased susceptibility of the joint to cracking (see Fig. 7) [1,13]. The above results mean that the smaller is the grain size of solder alloy, the more serious is the tin diffusion, the thicker of IMC layer is formed during aging, and the larger the variation of solder joint fatigue lifetimes is (see Fig. 5). Conclusions Based on our investigation of the shear cyclic lifetime of eutectic Sn-Pb solder joints formed using different reflow profiles and aged at 125OC for 4, 9, 16, 25, and 36 days, the following conclusions have been reached: By increasing the conveyor speed in the reflow process, the cooling rate of the PBGA solder solidification can be increased. Heat transmission analysis proves that increase of conveyor speed increases the mean coefficient of the heat transfer during solider solidification, increasing the cooling rate. A high cooling rate during solder joint solidification results in a rougher interface between the solder and the IMC layer, and refines the grain size of the solder bulk. For non-aged solder joints, it is found that increasing the conveyor speed can improve the solder joint shear fatigue lifetime substantially. And the shear fatigue lifetimes of aged specimen decrease with extension of aging time, moreover, the lifetime variation of faster conveyor speed specimens is larger. SEM micrographs of the failed solder joints reveal that cracks initiate at the point of the acute angle near the solder pad, then propagate along the interface of the solder bulk / IMC layer. During aging, the growth of IMC is a diffusion-controlled process, and the diffusion in the coarse grain boundary region is dominant. So the IMC growth rate of faster cooling rate sample solder joints (with smaller grain size) is larger. The growth of IMC layer decreases the shear fatigue lifetimes of aged solder joints. Appendix Calculation of the Mean Coefficient of Heat Transfer h,,, In our previous study to the PBGA solder joints [6], it has demonstrated that the solder bump exchanges heat with the atmosphere in the furnace chamber through forced convection in the reflow process. And due to the BGA structure, only gas flow parallel to the conveyor movement direction can contribute to the heat exchange. The conveyor movement velocity can be taken as the free gas fluid velocity (ujj (Fig. 9). Conveyor Movement Direction --t Flow Direction Figure 9. Schematic of the PBGA assembly on the conveyor during reflow. If we model the solder bumps as cylinders, for an in-line array (Fig. lo), the mean coefficient of heat transfer (II J can be calculated as below [14]: Electronic Components and Technology Conference

6 where the constant C and exponent n are geometric parameters, Re is the Reynolds number, d is the characteristic length, here the diameter of the cylinder, and kf is the thermal conductivity coefficient of the fluid at the film temperature. The film temperature is defined as: The Reynolds number is defined as: where vf is the kinematic viscosity of the fluid, U, is the maximum velocity of the fluid at the region of minimum-flow area of width S,-d shown in Fig. 10, given by: I I I Figure 10. Schematic of geometry of solder joint array. In this work, the solder diameter is taken to be 0.75 mm, and the solder bump pitch (S,,, S ) is 1.5 mm. So SJd and Sdd are both 2. C and n in Eq. (Al) are and respectively. We take Ty and T, as 175 C and 160 C respectively (see Figure l), so the Tf is C. We take standard values for dry air at atmospheric pressure (the values are not strongly pressure-dependent and may be used over a fairly wide range of pressures) of 160 C to calculate the h, (W/(m2. C)). The values of kf and vf are 3.64~ (W/(m.OC)) and 30.09~ (m2/s). The results corresponding to profile #1, 2#, and 3# are 4.2, 8.3 and 11.5 respectively. It can be seen that the mean coefficient of heat of profile #3 is about 3 times of that of profile #l. If we take the value of Ty- T, in Eq. (I) as 10 C (see Figure l), the cooling rate for profile #3 will be 30 times of that of profile #l. References 1. P. L. Tu, Y. C. Chan, and J. K. L. Lai, Effect of intermetallic compounds on the thermal fatigue of surface mount solder joints, IEEE Trans. Comp., Packag.., Manufact, Technol., Part B, Vol. 20, pp , A. C. K. So and Y. C. Chan Aging studies of Cu-Sn intermetallic compounds in annealed surface mount solder joints, Proc. 46 Electron. Comp. & Technol. Conf., Orlando, FL, May 1996, pp J. H. Lau, Solder Joint Reliability. New York: Van Nostrand Reinhold, 1991, pp R. E. Pratt, E. I. Stromswold, and D. J. Quesnel, Effect of solid-state intermetallic growth on the fracture toughness of Cd63 Sn-37Pb solder joints, IEEE Trans. Comp., Packag., Manufact. Technol., vol. 19, pp , Mar P. L. Tu, Y. C. Chan, C. W. Tang, and J. K. L. Lai, Vibration fatigue of pbga solder joint, the 5dh Electronic Components & Technology Conference. 6. S. H. Fan, Y. C. Chan, and J. K. L. Lai, Fatigue Lifetimes of PBGA Solder Joints Reflowed at different Conveyor Speeds, Journal of Electronic Packaging, Submitted for Publication. J. H. Lau, Solder Joint Reliability of BGA, CSP, Flip Chip, and Fine Pitch SMT Assemblies, McGraw-Hill, 1997, pp D. R. Frear, H. S. Morgan, S. N. Burcheit, and J. H. Lau, The Mechanics of Solder Alloy Interconnects. New York: Van Nostrand Reinhold, 1994, pp , Y. C. Chan, P. L. Tu, A. C. K. So, and J. K. L. Lai, Effect of Intermetallic Compounds on the Shear Fatigue of CuI63Sn-37Pb Solder joints, IEEE Trans. Comp., Packag., Manufact. Technol., Part B, vol. 20, No. 4, pp , NOV P. J. Kay and C. A. Mackay, Trans. Inst. Metal Finishing 57, 169 (1979). 11. YUJING WU, JENNIFER A. SEES, CYRUS POURAGHABAGHER, L. ANN FOSTER, JAMES L. MARSHALL, ELIZABETH G. JACOBS, and RUSSELL F. PINIZZOTTO, The Formation and Growth of Intermetallics in Composite Solder, J of Electron. Mater., v01.22, No. 7, pp , R. J. K. Wassink, Soldering in Electronics. London, U.K.: Electro-chemical Publ., 1989, chapter 4, pp Alex C. K. So and Y. C. Chan, Reliability of Studies of Surface Mount Solder Joints-Effect of Cu-Sn Intermetallic Compounds, IEEE Transactions on Component, Packaging, and Manufacturing Technology- Part B, Vol. 19, No. 3, pp , J. P. Holman, Heat Transfer, McGraw-Hill Book Company, Inc. pp , COLOR We alert you to the presence of color in the CD-ROM version of this paper. I 165 i Electronic Components and Technology Conference