Search

Thermal mechanical fatigue (TMF) is an important damage mechanism for solder joints exposed to cyclic temperature environments. Predicting the service reliability of solder joints exposed to such conditions requires two knowledge bases: first, the extent of fatigue damage incurred by the solder microstructure leading up to fatigue crack initiation, must be quantified in both time and space domains. Secondly, fatigue crack initiation and growth must be predicted since this metric determines, explicitly, the loss of solder joint functionality as it pertains to its mechanical fastening as well as electrical continuity roles. This paper will describe recent progress in a research effort to establish a microstructurally-based, constitutive model that predicts TMF deformation to 63Sn-37Pb solder in electronic solder joints up to the crack initiation step. The model is implemented using a finite element setting; therefore, the effects of both global and local thermal expansion mismatch conditions in the joint that would arise from temperature cycling.

The most commonly used solder for electrical interconnects in electronic packages is the near eutectic 60Sn‐40Pb alloy. This alloy has a number of processing advantages(suitable melting point of 183°C and good wetting behaviour). However, under conditions of cyclic strain and temperature (thermomechanical fatigue) the microstructure of this alloy undergoes a heterogeneous coarsening and failure process that makes the prediction of solder joint lifetime complex. A finite element simulation methodology to predict solder joint mechanical behaviour, that includes microstructural evolution, has been developed. The mechanical constitutive behaviour was incorporated into the time‐dependent internal state variable viscoplastic model through experimental creep tests. The microstructural evolution is incorporated through a series of mathematical relations that describe mass flow in a temperature/strain environment. The model has been found to simulate observed thermomechanical fatigue behaviour in solder joints.

We report the first in situ measurements of thermomechanical stresses in a 1000 I/O 250 /spl mu/m pitch piezoresistive flip chip test chip assembled to a 755 I/O 1.0 mm pitch 35 mm Ball Grid Array (BGA). The BGA substrates employed "build-up" dielectric layers containing micro-vias over conventional fiberglass laminate cores. Experimental data, which include in situ stress and die bending measurements, were correlated to closed form and Finite Element Method (FEM) calculations. Cracking and delamination were observed in some of the experimental groups undergoing temperature cycling. Through use of bounding conditions in the FEM simulations, these failures were associated with debonding of the underfill fillet from the die edge that caused stresses to shift to weaker areas of the package.

Solder fatigue tests have been conducted on point focus photovoltaic concentrator cell assemblies to identify a baseline fatigue life and to quantify the fatigue life improvements that result using a copper-molybdenum-copper low-expansion insert between the solar cell and copper heat spreader. The Coffin-Manson and total strain fatigue models for low-cycle fatigue were evaluated for use in fatigue life predictions. Since both of these models require strain calculations, two strain calculation methods were compared: hand-calculated shear strain and a finite element method shear strain. At present, the available theoretical models for low-cycle solder fatigue are limited in their ability to predict failure. >

The thermomechanical fatigue failure of solder joints is increasingly becoming an important reliability issue for electronic packages. The purpose of this Laboratory Directed Research and Development (LDRD) project was to develop computational tools for simulating the behavior of solder joints under strain and temperature cycling, taking into account the microstructural heterogeneities that exist in as-solidified near eutectic Sn-Pb joints, as well as subsequent microstructural evolution. The authors present two computational constitutive models, a two-phase model and a single-phase model, that were developed to predict the behavior of near eutectic Sn-Pb solder joints under fatigue conditions. Unique metallurgical tests provide the fundamental input for the constitutive relations. The two-phase model mathematically predicts the heterogeneous coarsening behavior of near eutectic Sn-Pb solder. The finite element simulations with this model agree qualitatively with experimental thermomechanical fatigue tests. The simulations show that the presence of an initial heterogeneity in the solder microstructure could significantly degrade the fatigue lifetime. The single-phase model was developed to predict solder joint behavior using materials data for constitutive relation constants that could be determined through straightforward metallurgical experiments. Special thermomechanical fatigue tests were developed to give fundamental materials input to the models, and an in situ SEM thermomechanical fatiguemore » test system was developed to characterize microstructural evolution and the mechanical behavior of solder joints during the test. A shear/torsion test sample was developed to impose strain in two different orientations. Materials constants were derived from these tests. The simulation results from the two-phase model showed good fit to the experimental test results.« less

A microstructurally-based computational simulation is presented that predicts the behavior and lifetime of solder interconnects for electronic applications. This finite element simulation is based on an internal state variable constitutive model that captures both creep and plasticity, and accounts for microstructural evolution. The basis of the microstructural evolution is a simple model that captures the grain size and microstructural defects in the solder. The mechanical behavior of the solder is incorporated into the model in the form of time-dependent viscoplastic equations derived from experimental creep tests. The unique aspect of this methodology is that the constants in the constitutive relations of the model are determined from experimental tests. This paper presents the constitutive relations and the experimental means by which the constants in the equations are determined. The fatigue lifetime of the solder interconnects is predicted using a damage parameter (or grain size) that is an output of the computer simulation. This damage parameter methodology is discussed and experimentally validated.

We report the first measurements of in-situ flip-chip assembly mechanical stresses using a CMOS piezoresistive test chip repatterned with a fine pitch full area array. A special printed circuit board substrate was designed at Sandia and fabricated by the Hadco Corp. The flip-chip solder attach (FCA) and underfill was performed by a SEMATECH member company. The measured incremental stresses produced by the underfill are reported and discussed for several underfill materials used in this experiment. A FEM of a one-quarter section of the square assembly has been developed to compare with the measured as-assembled and underfill die surface stresses. The initial model utilized linear elastic constitutive models for the Si, solder, underfill, and PC board components. Detailed comparisons between theory and experiment are presented and discussed.

A new viscoplastic theory for CusilABA and other braze alloys has been developed. Like previous viscoplastic theories,this new theory uses a hyperbolic sine function of effective stress in its kinetic equation for the inelastic strain rate. This new theory has an internal state variable which accounts for isotropic hardening and recovery and a second-order, internal state tensor which accounts for kinematic hardening and recovery. Unlike previous theories, the new theory uses evolution equations for the state variables which describe competing mechanisms of power law hardening and static recovery. The evolution equations used in previous theories describe competing mechanisms of linear hardening, dynamic and static recovery. The new viscoplastic theory was implemented in several finite element codes and used in several metal-to-ceramic brazing simulations. Two approaches for obtaining material parameters for the new viscoplastic theory were developed.

We present a new methodology for predicting the fatigue life of solder joints for electronics applications. This approach involves the integration of experimental and computational techniques. The first stage involves correlating the manufacturing and processing parameters with the starting microstructure of the solder joint. The second stage involves a series of experiments that characterize the evolution of the microstructure during thermal cycling. The third stage consists of a computer modeling and simulation effort that utilizes the starting microstructure and experimental data to produce a reliability prediction of the solder joint. This approach is an improvement over current methodologies because it incorporates the micro-structure and properties of the solder directly into the model and allows these properties to evolve as the microstructure changes during fatigue.

A combined experimental and analytical study of strains developed in encapsulated assemblies during casting, curing and thermal excursions is described. The experimental setup, designed to measure in situ strains, consisted of thin, closed-end, Kovar tubes that were instrumented with strain gages and thermocouples before being over-cast with a polymeric encapsulant. Four bisphenol A (three diethanolamine cured and one anhydride cured) epoxy-based materials and one urethane elastomeric material were studied. After cure of the encapsulant, tube strains were measured over the temperature range of {minus}55{degrees}C to 90{degrees}C. The thermal excursion experiments were then numerically modeled using finite element analyses and the computed strains were compared to the experimental strains. The predicted strains were over estimated (conservative) when a linear, elastic, temperature-dependent material model was assumed for the encapsulant and the stress free temperature T{sub i} was assumed to correspond to the cure temperature {Tc} of the encapsulant. Very good agreement was obtained with linear elastic calculations provided that the stress free temperature corresponded to the onset of the glassy-to-rubbery transition range of the encapsulant. Finally, excellent agreement was obtained in one of the materials (828/DEA) when a viscoelastic material model was utilized and a stress free temperature corresponding to the curemore » temperature was assumed. 13 refs., 20 figs., 3 tabs.« less