Mei-Ling Wu

Electro-thermal-mechanical modeling of wire bonding failures in IGBT

In power electronic module (PEM), IGBT (Isolated Gate Bipolar Transistor) is a high electrical power device and control wide power range from micro-voltage-ampere to mega-voltage-ampere. In addition, IGBT device usually operates in high-temperature environment, such as engine of automobile. As a result, the thermal problem is one of main concern regardless of the influence from dissipated power or environmental temperature. This paper discusses temperature distribution and mechanical behavior of wire bonding in IGBT based on finite element method. The purpose of this research would focus on the interface of wire bonding and connection substrate. Because the material of wire and substrate are different, it would result in CTE mismatch. Due to thermal damage, it could cause failure in the interface, e.g. wire bond lift off and heel crack. Among these, wire bond lift off is a critical failure and even lead to the degradation of IGBT. In order to simulate electro-thermal-mechanical behavior, two analytical methods are adopted in this research, one is electro-thermal analysis; the other is thermal-mechanical analysis. In this paper, we would use ANSYS software to analyze electro-thermal-mechanical behavior of wire bonding structure, furthermore, the wire bonding structure we build up is symmetrical, so only half of model is analyzed to save more time and reduce computational complexity. In electro-thermal analysis, we would show the temperature distribution in wire bonding structure when electric current is applied to wire. Furthermore, the natural convection and forced convection are taken into consideration in order to correspond to the realistic operational situation. The temperature change is dominated by Joule heating caused by the magnitude of electric current. For the purpose of obtaining the relation between temperature and electric current, nonlinear transient finite element simulation is proposed in this research. The analytical results show that the temperature would change along electric current and there is nearly temperature distribution in wire bonding structure after stopping applying electric current for a period of time. In thermal-mechanical analysis, we would adopt indirect coupling method, that is, use the results obtained from electro-thermal analysis as thermal load to thermal-mechanical analysis. According to analysis, the maximum von-Mises stress occurs in the heel of wire bonding due to CTE mismatch and high Joule heating. Besides, the displacement of wire bonding structure has upward direction, namely, the warpage phenomenon would occur due to temperature change. Under continuing of temperature change, it would bring fatigue failure in the interface of wire bonding and connection substrate. The objective in this research is to improve and increase the mechanical strength of wire bonding by discussing the relation of electro-thermal-mechanical behavior.

Dynamie simulation for microelectronic packaging of Al pad/underlying pad structure during copper wire bonding

For decades, wire bonding technology has been widely used to interconnect IC chip and the substrate in microelectronic package. In recent years, due to the increasing cost of gold, the copper wires started to be employed in microelectronic package instead, lowing the production cost. The additional benefits include superior performance of copper wires, in terms of electrical and thermal property. Thus, it is likely that, the gold wires will be replaced by the copper gradually in the future. However, greater hardness of the copper wire and weak mechanical strength of low-k dielectric layers lead to higher stress in the Al pad and the underlying pad structure during the thermo-sonic bonding process. An additional concerns arises, as the Al pad is squeezed out by copper ball, which may affect the development of fine pitch. Therefore, it is necessary to redesign the bonding pad, as well as adjust the bonding parameters, such as bonding pressure, ultrasonic energy and bonding time. As finite element software enables dealing with the wire bonding process by a transient nonlinear dynamic analysis, a finite element model for copper wire bonding is developed to investigate the mechanical behavior. The simulation results would focus on the dynamic stress response of wire bonding model and the plastic strain of the Al pad. In order to save computational time and reduce modeling complexity, the copper wire bonding is represented by a simplified 2-D finite element model, whereby the complete wire bonding process mechanism is treated as consisting solely of impact and ultrasonic vibration stages. The calculation of nonlinear transient structural behavior is carried out using an explicit time integration scheme. In the present study, the simulation results indicate that the stress wave would rapidly transfer from the bonding interface to the underlying pad structure when copper ball is in contact with the Al pad. With the increasing contact area, the stress wave path would shift from the inside bonding pad to the outside, resulting in the spread of the stress concentration.

Assessing the impact of uncertainty in physics of failure analysis of BGA solder joints fatigue damage

Solder joint fatigue life models will be used for analysis in this paper. Identifying an approach for quantifying the combination of input and uncertainty would enable the determination of more realistic confidence limits on PoF predictions. Variability in the physical parameters can be addressed by existing stochastic methods, which are designed to propagate probability distributions on the parameters through a fixed model structure in order to estimate the statistics of interest on the model response quantities. This paper focuses on uncertainly analysis, that is, how the input data uncertainly affects the output data uncertainly. We are assuming the model is correct and are only dealing with uncertainties due to variations in inputs. Output uncertainty, uncertainly in the correction factor, is determined by Monte Carlo simulation.

Simulation 3D TSV for stress-strain characteristics under mechanical and thermo-mechanical loading

This paper addresses the key stress and strain characteristics issues in three-dimensional integrated circuit (3D IC) packaging, which arise due to mechanical and thermo-mechanical loading. Although 3D IC packaging is known to suffer critical issues due to the reliance on non-mature technologies, it is valued for its high performance and miniaturization, achieved through the short vertical interconnections between individual chips and multi-chips that are stacked together. However, the reliability of through silicon via (TSV) and micro-bumps is still a significant concern and should thus be investigated further, due to the complexity of the architecture and microstructure. In this work, the 3D IC package used in the simulation model is built, after which the model is investigated under thermal cycle loading and mechanical bending cycle loading. In the analyses, both micro-bump and TSV are considered to exhibit bilinear isotropie hardening behaviors. The simulation results indicate that, under mechanical loading, the critical failure occurs on the outer micro-bump, while it is located on the outer TSV under thermo-mechanical loading. We thus posit that these fatigue failure sites could arise from the coefficient of thermal expansion (CTE) mismatch between the silicon chip and the TSV. Based on these findings, a simulation-based optimization methodology is developed with the aim of improving the overall 3D IC reliability. The main objective is to improve the TSV and micro-bump fatigue life when subjected to mechanical and thermo-mechanical loading by optimizing the design factors.

Atomistic simulation analysis for influence of organic PA types on PA/UF adhesion in 2.5D IC package

2.5D IC is an important transition product of 3D IC in the next generation. It has very important commercial value. However, at present 2.5D IC package process has numerous issues that need to be overcome, especially for interface delamination caused by warpage under reflow temperature. Conventionally, the material adopted for the passivation (PA) layer in 2.5D IC package is inorganic metal glass. But for being cost-effective, organic polymer is another choice. Accordingly, the adhesion strength between organic PA and underfill (UF) layers is a critical issue. In this paper, molecular dynamics (MD) simulation is applied to analyze the interface adhesion between organic PA and UF layers affected by five chemical PA types, compensating for essential limitations of experimental and FE simulation analysis. From the pull test curves of MD simulation, the adhesion strength of different PA types to UF layers corresponding to molecular interaction mechanisms such as chemical bond force, non-bond force including Coulomb electrostatic force and van der Waals adsorption, and mechanical interlocking force are examined. The present results reveal that the adhesion strength of the five PA types to UF layers display significant variations, which could be attributed to different strength characteristics in molecular interaction mechanisms due to different PA chemical structures. The ranking order for the maximum adhesion strengths of the pull test curves with five PA chemical types is of acrylic, BCB, epoxy resin of novolak, polyimide, and PBO, types, respectively. The maximum adhesion strengths of the pull test curves with five chemical PA types exhibit the obviously discrepancies, which demonstrates that different PA chemical types play the key role to affect the adhesion strength of PA and UF layers. With regard to the maximum adhesion strength of PA and UF layer affected by the molecular interaction mechanisms, the key influence factors should be Coulomb electrostatic force, van der Waals adsorption, and the mechanical interlocking force because the trend variations are highly related with the value distributions of the maximum adhesion strengths, but due to the page restrictions, the analysis results and interpretations are not presented here. Based on the present study, it could be further extended to establish the design rule for the material development and synthesis of the workability performance promotion of organic PA or UF.

Critical parameter selection for thermal cycle of FBGA fatigue life

This paper will focus on the fast assessment methodology of FBGA fatigue life through simulation and physics of failure (PoF) analysis under thermal cycle. The structure of fine pitch ball grid array (FBGA) that has been investigated, and been modeled by ANSYS to compare with experimental data. There are two temperature cycling will be used, one is used to verify FEA model, and the other one is used to do failure analysis. The aim of this paper is fist discussing the ability of finite element analysis (FEA) in executing the virtual thermal cycling reliability analyzing the reliability of solder joints fatigue life in local modeling. FBGA packages can warp due to local and global mismatch of the coefficients of thermal expansion and the asymmetric package geometry. Temperature cycling condition including dwell time and ramp rate is an important factor that will affect the solder joins reliability. This paper will identify the critical parameters that influence by using a Design of Experiments (DoE) approach using simulation results from ANSYS. Two types of analyses include the physics-based analysis and the statistical-based analysis. The paper presents the physics-based analysis, three steps, uses by DoE tool. The first step we have to select the parameters, such as PCB thickness, PCB Youngs modulus, PCB coefficient of temperature expansion (CTE), solder joints height, die thickness, mold compound thickness, and so on. The second step is to find important factors. And the final step is going to DoE, in this paper used response surface methodology (RSM). The use of DOE and ANOVA to identify the critical parameters and a response surface to generate a functional form will be discussed. In this study, we will concentrate on the approach of deformation information to the critical stress. The critical stress is then fed into a fatigue damage model, which outputs life, or cycles to failure. Throughout the work, DoE and ANOVA techniques will be used to determine the key parameters and help in the development of a fast assessment model of FBGA fatigue life through simulation and physics of failure (PoF) analysis under thermal cycle.

Modeling of pore formation in solid

Quality and microstructure of materials are strongly affected by the shapes of pores in solids, which are encountered in almost all the packaging, micro-electro-mechanical systems (MEMS), and manufacturing fields. This study theoretically analyzes the mechanisms of an entrapped micro-bubble as a pore in the solid. A phase diagram, as introduced previously to be satisfied by interfacial normal stress balance, can be used to control and delineate the growing path and final shape of a pore. The predicted pore shape agrees with experimental results.

Failure modes and FEM analysis of Conductive Anodic Filament resistance during high-frequency electromagnetic of PCB

In the multilayered printed circuit board (PCB), the signal transmission between different layers occurs by way of through hole structure. The geometry of a through hole consists of traces, pads and a perfectly conducting hollow cylinder. Conductive Anodic Filament (CAF) generated easily in PCB when board size is decreased. CAF produced will lead to leakages or short circuit. Many reasons cause the phenomenon of the CAF, one of them is the voltage. In these situations, the signal transmission integrity is concerned. The excessive capacitance is generated near the through hole structure. Such capacitance influences the signal transmission between different layers. A number of factors can cause excessive capacitance, for example: geometry of transmission line, pad of through hole and the interaction of different layers. However, recent application of the multilayered printed circuit board such as cell phones and note books, have decreased the broad size for easy to carry or load more equipment. Decreasing the size of PCB, the excessive capacitance near the through hole structure is inevitable. It is more important to design the complexity of the layout and the geometry of structure. The quick and efficient prediction of the performance is concerned when manufacturing process become more complicated. The simulation model and problem discussion are the subject of this paper. In this study, we demonstrate that the geometry of the transmission line near the through hole may affect the transmission of the signal by excessive capacitance. By using ANSYS 12 to simulate the high frequency electromagnetism, and simulate the transmission line with different shapes near the through hole and extract the Scattering-parameters from the simulation results. Scattering-parameters are used to describe the electrical behavior of linear electrical network. The parameters are helpful for us to design electrical engineering, electronics engineering, and communication systems, and especially for microwave engineering. By comparing Scattering-parameters of different structures, we can understand the signal transmission and reflectivity between two layers. We also analyze the electric field and magnetic field of the structure, and use a method based on the quasi-static approximation to compute the excessive capacitance of the through hole structure. Using the concept of equivalent circuit to convert the original model into the RLC circuit. Equivalent circuit is divided into multiple series parts. Many parameters such as resistance, capacitance, inductance and characteristic impedance will be discussed, then compare these value changes between different shapes of the transmission line near the through hole. The ANSYS model is proposed to simulate the high-frequency electromagnetic of the structure. By these model, we want to predict the signal transmission performance of the through hole structure.

Models for physics of failures analysis during printed circuit board bending

This paper summarizes the force calculation at each pitch position for the structure, where a component (overmold, die, and BT) is attached to a PCB (printed circuit board) through an array of solder joints with an external bending moment applied at the ends of PCB. In this paper, the results from the model proposed by E. Suhir [1] are summarized with corrections in the derivation of the equations. Details of the new model will be described including the methodology, the formulation and the simulated results. Comparisons between the two models will be made with discussions. In this new model, each solder joint between the component and the PCB is replaced by a spring with spring constant. Benefiting from the symmetry of the structure, only half of the structure needs to be considered which simplifies the formulation of the problem and saves the computation time dramatically. The variables (unknowns) to be solved are the forces exerted on each spring due to applied external bending moment. In the subsections to come, detailed derivations will be given followed by numerical results. The setup of the new model and it is essentially the same as the Cantilever-intermediate load problem. The forces exerted on each spring will be solved through a system of linearly independent equations governed by the following conditions. All the listed formulas above are for the case when the number of leads is odd. Basically, similar formulas will be used for the case when the number of leads is even. The goal in this paper will be as follows: (a) Suhirs model deals with continuous case. It is clearly seen from the simulation results of Suhirs model that all the curves are smooth no matter how many number of solder joints are involved. The only parameter that is related to the number of joints as defined by Suhir is the spring constant of the elastic attachment. In the new model, we will solve for the force at each position in a discrete sense. That is why the curves are not that smooth when the number of solder joints is small. As the number of the solder joints becomes large (or in other words, the pitch size is small), the two models yield exactly the same results. (b) Mathematically, the new model is simpler and straightforward. The only math involved is solving the linearly independent system of equations. In Suhirs model, however, the solution process starts from solving the 4th order differential equations. Although analytical solutions can be obtained, special attention has to be paid in imposing proper boundary conditions, which might not be trivial.

Global uncertainty analysis of solder joint fatigue life model in random vibration environment

The global model uncertainty analysis is to estimate the error due to input data approximations rather than sensitivities associated with finite element modeling. In this paper, estimates will be made on the amount of sensitivity in the global modeling which basically involves the sensitivity in the value of the material properties, input loadings, and geometries. Assume the model structure is corrected, how do we estimate a physics of failure (PoF) and predict time to failure or cycle to failure? Identifying an approach for quantifying the combination of input and uncertainty would enable the determination of more realistic confidence limits on Physics of Failure (PoF) predictions. The physics-of-failure (PoF) approach to reliability utilizes knowledge of the life-cycle load (thermal, vibration, mechanical, electrical, photonics, and so on) profile package architecture and material properties to identify potential failure mechanisms and to prevent operational failures through robust design and manufacturing practices. We can decompose the elements of any mathematical model into the physical parameters used in forming equations, such as material properties or geometric dimensions and the model structure, which results from simplifications, assumptions and approximations. Variability in the physical parameters can be addressed by existing stochastic methods, which are designed to propagate probability distributions on the parameters through a fixed model structure in order to estimate the statistics of interest on the model response quantities.