By Guotao Wang and Dr. Paul Ho, University of Texas (Austin); Drs. Terry Hayden and Eumi Pyun, 3M Electronic Products Division, Austin; and Dr. Douglas Gundel, 3M Fiber Optics and Electronics Technology Center, Austin Since the early days of IBM's controlled collapse chip connection, (C4)(1), solder joint reliability/thermal fatigue in solder array assemblies has been a critical factor because of the CTE mismatch between the silicon die and the carrier. The following general formula approximates the maximum possible shear strain in the flip-chip-on-board construction(2):
CTEi = coefficient of thermal expansion for the die or board ΔT = temperature range of the stress cycle DNP = distance from the neutral point of the solder joints H = die-to-board standoff height In reality, the resistance of the solder joints to deformation reduces their strain below this level with larger solder joints offering more resistance. Because modern devices are migrating toward fine-pitch (which limits solder joint size) and toward larger sizes (which increases DNP), this equation indicates limitations for the FC-on-board construction.
One solution to reduce the solder joint stress and increase reliability in FC solder joints is to underfill the solder joints(2,3) (Figure 1), but this is not desirable for many solder array assemblies, because it requires additional process steps and limits rework. BGA packages, whether based on conventional rigid substrates or higher-routing-density flex circuits, may also suffer from similar board-level reliability issues, but a package offers a significant opportunity for reduction of these stresses through an increase in the package CTE.
Package geometry and materials selection can dramatically affect these solder ball strains, and hence reliability. For example, a CUEBGA (Figure 2) differs from a conventional flex-based package, such as Texas Instruments' µStar or Amkor's fleXBGA4 by incorporating a copper stiffener that is laminated to the flex circuit prior to assembly (Table 1)(5,6). It also differs from other package constructions that provide a relatively thick compliant layer between the die and the solder balls (such as a thicker die attach adhesive(7) or elastomer as in µBGA™), to reduce the solder ball strain, or others that alter other geometrical factors such as the solder ball pad size(7). Improved Handling and Reliability This addition improves handling during first-level assembly manufacturing(6), enhances heat dissipation(5) and results in an increase in the overall package CTE to better match that of the PWB(5,6). End-user reliability testing has recently demonstrated that the 12 mm CUEBGA package construction with stiffener meets tough telecom equipment attachment reliability requirements (two-parameter Weibull characteristic life of 6300 cycles for 0 to 100°C stressing(7)). In previous CUEBGA studies, moiré interferometry of this package by itself (not mounted to the board) proved useful for experimentally determining that its CTE was close to the PWB(5,6). In the present study, cross-sectional moiré interferometry was employed to determine stresses and strains throughout the package in board-mounted devices. The measured solder ball stresses were then compared to three dimensional package FEA. The results of this investigation are given below. They suggest that there is an opportunity to decrease solder ball stresses further through the judicious selection of adhesive materials in the construction. A low modulus adhesive8 selected for use in this package has the required high-temperature durability for package fabrication and service, as well as the requisite properties for low solder ball stresses.
Solder Ball Stress Analysis Moiré interferometry is a whole-field optical interference technique with high resolution and high sensitivity for measuring the in-plane displacement and strain fields9. Recently, this method has been used to measure the thermal-mechanical deformations of electronic packages with the objective of studying electronic package reliability(10). Figure 3 shows the optical system of moiré interferometry.
Standard moiré interferometry employs a grating frequency of 1200 lines/mm, which yields a spacing of the interference fringe corresponding to 417 nm of in-plane displacement. While the sensitivity is adequate for measuring the overall thermal deformation of electronic packages, it is not sufficient for measuring thermal deformation in high-density electronic packages-particularly for small features such as solder balls. For this purpose, high-resolution moiré interferometry was developed based on the phase shifting technique. In this study, high-resolution moiré interferometry was carried out, in addition to the standard interferometry, using an IBM PEMI (Portable Engineering Moiré Interferometer) which was modified for phase-shifting operation. The detection sensitivity for phase-shift moiré interferometry can reach 26 nm per fringe(11). Package Analysis Moiré analysis was carried out for the 12 mm CUEBGA package with 0.8 mm pitch that had been reflowed to a two-layer PWB and cross-sectioned. A schematic of the 12 mm CUEBGA package with the two sections that were analyzed is shown in Figure 4. (Section 07 and section 45 correspond to the seventh (middle) solder ball line and the diagonal solder ball line respectively.)
Each specimen was cut and polished to the mid-section of the solder balls. A low-viscosity, brittle adhesive was selected to adhere a 1200 lines/mm grating on the cross-section of the specimen at 102°C. This adhesive was chosen to reduce the adhesive thickness (and hence the shear-lag effect). The deformation at this temperature was taken as a reference (zero) deformation state. The experiment was performed at a room temperature of 22°C, thereby providing a temperature change of 80 degrees centigrade (ΔT= -80°C). Finite Element Analysis Three-dimensional FEA of the 12 mm CUEBGA package was performed using commercial FEA software(12). FEA was first used to duplicate the same sectioned sample as the moiré results and then used to analyze the entire package. Due to the symmetric structure and deformation, only a quarter of the package was modeled. Temperature-dependent elastic-plastic properties for eutectic solder were input in the model. Initially, all other materials but the solder were considered to be isotropic linearly elastic. Without considering the plastic property for the die-attach adhesive, the FEA result did not agree with the moiré result. For this reason, a micro-tensile system was employed separately to measure the stress-strain relationships for the die-attach, polyimide, and the proprietary adhesive. In this system, two computer-controlled encoder stepper motors are used to generate static sample displacements of less than 1 micron, and a load cell provides data for determining the applied load. The stress-strain curves generated by this system for the three adhesives were then input into the FEA model.
Results An example of a moiré fringe pattern is shown in Figure 5, which also includes a superimposed outline of the interfaces. Results from both section 07 and section 45 were similar. A number of qualitative observations regarding the package deformation can be made from the patterns:
Critical Findings The critical findings relative to the solder joint reliability are the solder ball strains. Figure 6 (a) plots the average normal strains in solder balls. εx is approximately constant in all the solder balls and it is about -0.1 percent. The εy displays a somewhat varying value. The outermost solder ball has the highest εy and it is equal to approximately -0.4 percent. The shear strain is given in Figure 6 (b). The two components of the shear strain are plotted separately in addition to the shear strain, εxy.
|
|||||||||||||||||||||||||||||||||||||||||||||||||
(1)
Copyright © 2001