Studies on solder fatigue failure of quad flat nonleaded (QFN) packages subjected to test cycling (e.g., -40/125°C, 1 hour) and field cycling (18/83°C, 7.1 hour) indicated unexpectedly early failures. Daisy chained packages QFN36 and QFN40 dummy components were used on FR-4 test boards. Thermal cycling had been made up to 4,000 test cycles and 17,850 field cycles (15 years). The paper addresses questions as to why this early failure occurs. Microstructural analyses revealed that SAC357 solder degradation in test- and field cycling follow standard processes and cannot explain the deviations. "Secondary effects" are therefore studied. Results on multiple effects are reported based on metallographic analyses, microscopic/SEM analyses, fluorescence microscopy, optical out-of-plane and inplane thermal deformation analyses, and comparison to FEM. It is shown that considering cure induced package shrinkage somewhat affects solder joints loadings, while formation of electronic molding compounds (EMC) surface layers due to thermooxidative cross-linking cannot be proven, even after 15 years of field cycling. Two additional secondary effects were observed, which seriously affect solder fatigue: internal package delamination between leadframe and EMC, and standoff filling (peripheral solder gap) by solder mask and flux residues. A parametric FE study revealed that both delamination and gap filling can reduce solder fatigue life by 1.5X and more. Even more important, by including these effects in FE-based predictions, they come in reasonable agreement with testing results and previous predictions for other components.

The major aim of the study is the validation of FE-simulation based prediction of fatigue life and calculated acceleration factors for selected components, which have been tested in a long-term thermal cycling test with loadings close to service conditions, i.e., 18 degrees C/80 degrees C, 7.1 hours. Theoretical analyses are made and compared to test cycling (TS -40/125 degrees C, 1 hour) results. Some results related to the test duration of up to 13 years (15,650 field cycles) were reported in 1111, in particular for QFP100, FBGA277, and 0201 ceramic chip components. This paper focuses again on FBGA 277 and additionally on QFN packages. More detailed investigations are made on aging of materials, in particular molding compounds. The weakness of the unknown plastic materials properties and neglecting cure shrinkage is overcome by thermal deformation analyses on component level and adjustment by FEA. It is shown that these effects are noticeable for FBGA, but essential for QFN. For example, even the warpage direction at room temperature (RT), calculated without cure shrinkage, arc opposite to the measured ones. Additionally, for the FBGA distribution patterns of cyclic creep strains/ dissipation densities across the balls are compared to failure patterns. Anand creep laws including aging from literature [2] have been investigated. It turned out that creep results-based creep laws with 5 to 60 days aging all agree qualitatively with the failure pattern, only the unaged data set does not.

Reliable lifetime prediction of power and microelectronic devices requires evaluation routines that are reproducible, stable, and comparable across geometries, meshes, and organizations. Conventional path- and area-based post-processing of in-elastic strain often suffers from stress singularities and a priori assumptions about crack paths or regions of interest, introducing mesh and user dependency. This work introduces Automatic Volume Averaging (AVA), a physics-of-failure aligned evaluation routine that ranks elemental inelastic strain within a defined region of interest (ROI) and computes a volume-weighted average over the highest-strained elements up to a user-defined volume threshold. The threshold is selected to match experimental failure definitions, thereby linking simulation metrics to observable degradation. The method is benchmarked against Path Evaluation (PE) and Area Evaluation (AE) using a CR0805 SMT resistor model (ABAQUS, SAC solder with viscoplastic creep) under thermal shock (-40/125 degrees C). Lifetimes are computed with Coffin-Manson/Morrow relations. Experimental cross-sections indicate crack initiation at approximate to 500 cycles, gap failure at approximate to 800 cycles, and full joint failure at approximate to 1700 cycles. AVA yields the closest agreement to all three failure modes; AE tends to underestimate lifetimes, and PE shows the largest deviations due to singularity sensitivity and path placement. AVA reduces mesh/user dependence, scales to 3D assemblies, and provides a practical pathway toward a standardized, comparable damage evaluation routine for reliability assessment.

A long-term temperature cycle test has been running for more than 13 years to answer the most challenging questions about dominating failure modes and degradation processes caused by thermal cycle loadings close to service conditions. Test boards with components QFN40, QFN36,FBGA48, TQFP100, FBGA227 and chip capacitors/resistors size 0201 have been subjected to temperature cycles 18°C/80°C, 7.1 hours. The latter three were analyzed by Finite Element Analysis (FEA) and compared to damage observed from cross sectioning up to damage states at approx. 15.700 cycles (13 years). SAC357 (Sn3.5Ag0.7Cu) was used. Solder fatigue was clearly the dominating failure mechanism for all components except the 0201 ceramic chip capacitors/resistors.SAC solder fatigue is briefly discussed from a microstructural and continuum mechanical perspective. Fatigue life evaluation results based on averaged cyclic creep strain and creep dissipation are compared to testing results. The fatigue life prediction model based on the secondary/primary creep law tends to predict fatigue life reasonably well for test cycling but underestimates it for field cycling. Simulation inherent issues like the use of a combined secondary/primary creep model vs. two Anand’s models from the literature are also addressed.Some failure modes apart from standard ductile solder fatigue are discussed for thermal test-and field-conditions.

Thermal cycling tests showed much higher thermal cycling lifetime for components soldered with a low melting point alloy (LMPA-Q), compared to the standard SnAgCu solder. This is not expected as due to its lower melting point, the LMPA-Q solder is deforming more and having higher creep strains during thermal cycling, in particular if the maximum temperature of the cycling is close to the melting temperature. These higher creep strains are not reflected in the time to fatigue failure.To better understand the physics of failure behaviour of both solders (LMPA-Q vs SnAgCu), the microstructural behaviour of both solders is analysed with scanning acoustic measurements (SEM) and the deformations of the solder joints during thermal cycling are measured using digital image correlation (DIC) method.Finally, the deformation and strain measurements are compared to thermo-mechanical simulations. The simulations also include the grain structures in the solder joints. The outcome reveals the importance of having grains with different orientation into the modelling to explain the outcome of the thermal cycling tests.

The thermo-mechanical reliability of an industrial power module containing IGBTs with soldered die and module attachment and aluminum bonding wires has been investigated. In this kind of power packaging, die attachment, module attachment, and bonding wires attachment can be critical failure modes. It was one aim of the investigation to analyze these failures when the modules are subjected to different kinds of passive and power cycling. Furthermore, prediction of these failures based on finite element (FE) studies are addressed. Coupled (FE-) analyses are reported to understand the thermo-mechanical stresses in the assembly. Verification of evaluation methodologies to predict cyclic damage are in the focus of the paper. Physics-of-failure (PoF) based lifetime prediction is applied for a power stack die attach with soft solder between chip and substrate and between substrate and base-plate. An easy-to-use Coffin-Manson type failure criterion is given based on averaged cyclic creep strain and dissipation density. At the wire bonding interfaces bi-material notch effects cause stronger interface stress singularity, and therefore, damage mechanics criteria already proposed in [1] are updated and applied. Cyclic damage is calculated by the Cohesive Zone Method (CZM) and critical cycle numbers are estimated from that by a modified Paris law. The predictive laws for both thin solder layers and heavy Al bonds are compared to testing results and shown to agree well.

A methodology is presented that allows the evaluation of the thermomechanical reliability of electronic packages using “virtual prototyping.” Here, a virtual flip chip ball grid array (FC-BGA) is examined in comparison to a reference chip scale package (CSP). The comparison is performed using finite element simulation. A combined measurement-simulation technique is used to calibrate the finite element simulations on a reference object. The adjustment is based on the in-plane deformation field obtained by both simulation and optical measurement. For the latter, an optical sensor is used for in-plane deformation and strain field analysis based on the gray-scale correlation method. The findings obtained can be extrapolated to alternative package types with different but similar design to evaluate their suitability for the desired application before physical fabrication.

To fulfil actual and future requirements of xMR based current measurement devices the knowledge of possible deviations during the fabrication process and their impacts is necessary. The digital twin approach for a SO16 sensor device presented in this paper helps to understand the measurement offset change after manufacturing and improving the final product quality. The created FE based virtual prototype model included relevant nonlinear and temperature dependent material descriptions like metal plasticity and viscoelasticity. Beside model uncertainties regarding available data for the specific materials, the properties can fluctuate naturally due to several reasons like different batch qualities, changed ingredients etc. All these issues can have an impact on the resulting intrinsic stresses in the AMR sensor and were considered within the digital twin. An extensive design of experiments (DoE) study with about 40 free input parameters at process and component level was therefore conducted to quantify the uncertainties and their impact on the sensor behavior. With the usage of industrial state-of-the-art meta-modeling algorithms, a massive shrinkage of the initial input parameter space was possible. The parallel generation of a prognostic behavioral model, a so-called compact digital twin, for the resulting stresses in the AMR sensor enables the further usage and a future optimization of the remaining determining parameters.

Exchanging thermo-mechanical simulation models between institutes and companies are useful in research and development projects, in general for electronic components and systems. As different software codes are used at different R&D sites, models need to be shared in a consistent way, with the objective to get the same results independent on the code. This involves in first instance correct transfer of dimensions, boundary conditions and material properties. In this research work, the exchange between different software is applied for two different models. A first model focuses on a combined displacement and temperature cycling load on a single bump, the second model is a full flip chip BGA package soldered to a board.The first model was a learning exercise on how to exchange models. After some iterations, a rather good agreement was achieved. The final benchmarking exercise was done on the full BGA assembly. It was found that there are still variations of about 30% between the different software codes, even after several fine tunings.The outcome of this work could be an initiation for a kind of standard or general guideline on how exchange models.

In this paper, an approach for linking experimental data from active power cycling tests and FE model based results is discussed. An exemplary numerical study of a TO-220 diode as typical discrete power device was conducted with a sliced and an intact specimen. During two electro-thermal DoE analyses several parameters were varied and the model behavior depending on material properties and boundary conditions was investigated. This approach enabled the identification of the most influential parameters and derivation of behavioral models. During a following calibration it was possible to improve the prognostic abilities of the built FE models significantly. Maximum and minimum junction temperatures as well as infrared images were used for a comparison of the thermal results. Additionally thermo-mechanical analyses of the sliced model were compared with microscopic images from the cutting surface. In general, a good match for the temperature and deformation results was determined.

Physics-of-failure (PoF) based lifetime prediction requires appropriate damage metrics dependent on the failure characteristics of the material failing. Power stack die attach with soft solder between chip and substrate or between substrate and base-plate frequently fail by fatigue of the interconnection layer. An easy-to-use failure criterion provided is based on Coffin-Manson type evaluation. Averaging across zones of maximum creep strain/ energy dissipation is developed to calculate mean cycle to failure numbers. However, this criterion does only work for relatively homogeneous plastic/creep strain distributions. For strongly localized plastic zones, which occur at free interface edges of dissimilar materials, application is limited. The (bi-material-) notch effects cause interface stress singularity, and therefore, strong mesh dependence of the maximum stress or cyclic plastic strains. For fatigue in this ductile-brittle transition region, cyclic inelastic strains cannot be applied as a damage metric and fracture and damage mechanics criteria need to be implemented in simulation. Cohesive zone method (CZM) is applied for that reason. It has been implemented for bonded heavy Aluminum wires interconnects.A sequentially coupled electro-thermal-mechanical finite element analysis methodology is applied to get realistic transient temperature loadings as well as mechanical stresses for power cycling loads. For the thermo-mechanical part, process dependent non-linear constitutive models have been developed and introduced for the wire interconnects and die attach. Power cycling tests were run in parallel to the theoretical investigations. Preliminary predictive laws are provided for both thin solder layers and heavy Al bonds.

The thermomechanical reliability of the package and interconnections of assembled flip chip ball grid arrays (FC-BGA) is investigated in comparison to a reference chip scale package (CSP). Comparison is made using finite element (FE-) simulation. A combined measuring-simulation technique is applied to calibrate the finite element simulations on a reference object. Adjustment is made based on the in-plane deformation field evaluated by both simulation and optical measurement. For the latter an optical sensor for in-plane deformation and strain field analysis is used based on grey scale correlation method. A methodology is presented and to extrapolate the knowledge gained to alternative package types of different but similar design in order to evaluate their suitability for the desired application before the physical fabrication (virtual prototyping).

In this paper, the research is focused on lifetime analysis and prediction of silver sintered power devices by experiment and numerical simulation. As researches are emerging on the field of lead-free interconnect technology, silver sintering become an alternative interconnect technology to standard solders. This technology differs from other interconnect technologies because of it’s improved thermal and mechanical properties. Aluminum cladded copper core (CucorAl) wires have been used to prevent early fail of top contacts. Material degradation due to thermal fatigue in the sintered silver still influences the reliability in the interconnect, in particular for high temperature swings upcoming with wide bandgap semiconductors. Plastic deformation is observed in sintered silver interconnect which depends on time and temperature. This research work presents combined experimental and finite element analysis (FEA) results for developing a lifetime model for the sintered silver power modules. Different parameters such as porosities on the sintered silver layer, substrates, plastic and creep constitutive laws, die thickness and sizes, should be considered during the development of lifetime model. In this investigation, Insulated-gate bipolar transistor (IGBT) chips are integrated on different substrates such as printed circuit board (PCB- Help H), insulated metal substrate (IMS), copper lead frame (Cu-lead frame) and direct copper bond (DCB) by silver sintered interconnect. Failure modes such as wire bond lift-off and die bond degradation are observed during this research.

The use of electronics for purpose of autonomous driving requires high performance vehicle computer (HPVC) systems usable in harsh environments. A variety of challenging issues have to be considered from different perspectives. The paper focuses on the perspective of reliability and functional safety. To physically simulate and to analyze the thermal and the thermo-mechanical behaviors of the actual processors in different operation modes, a mock-up system is designed, which has initially been modelled by virtual means. The thermo-mechanical design methodology is essentially based on FE parametric studies, using a full model including 1st and 2nd level interconnects as well as including solder creep and other materials non-linearity.. To gain an overview on overall influences of materials parameters choice, virtual DOE was made at the beginning. Model complexity and size limits virtual DOE, which was performed with reduced geometric details and elastic material modeling. Comparison on material models comprise visco-plastic models for SAC305 and visco-elastic models for polymers. To reach high T-cycle reliability is a challenge for the mock-up system. The effects were analyzed for both first level and second level interconnects. It is shown that first level bumps fatigue or underfill 1 delamination are highest failure risks, which can be basically affected by choice of the assembly technology, i.e. underfill/undermold and soldermask choices. Simulation inherent issued like the use of a combined secondary/primary creep model vs. an Anand's model are compared and good agreement has been observed.

Aluminum based wire-bonds are still representing the most important technology for the chip top side connection in power devices. At the same time, they often limit the reliability and lifetime of the overall power device or module. Hence, knowledge of the wire-bond reliability can be key for predicting and modelling the lifetime of the whole electronics assembly.In this paper, a short overview of load and design based lifetime models for power devices and modules together with their limitations will be given. Based on this, the requirements as well as the necessary process steps for developing a more generic lifetime model for wire-bond degradation using stress and damage parameters are discussed. This includes the development of global-local finite element models as well experimental measurements and test for material characterization, model calibration and validation. Finally, first results of a parametric finite element study evaluating the wire-bond lifetime depending on different geometry parameters and material combinations will be provided.

Silver sintering technology has been introduced as a replacement for the traditional interconnect technology such as soldering on power modules. Solder fatigue and wire bond failures are two major factors that affects the lifetime of power modules. The silver sinter interconnect technology resolves the lifetime issues of conventional solder fatigue. The wire bond degradation is another factor, which limits the lifetime of the power modules. The researches are going on to reduce the curbs of wire bond degradation. The replacement of Aluminium wire bond by CucorAl or Cu-wire bond can be an approach to increase the lifetime of wire bonds in the power modules. In the project SINTER, some aspects of reliability modeling of silver sintered power stacks with different substrate and wire bonding technologies are addressed. In pressure sintered power stacks, the failure mechanism "wire-bond lift-off" is dominating as long as aluminum heavy wire bonds are used. Thermo-mechanical FEA analysis of different wire bond is conducted on PCB and DCB substrates with a passive thermal cyclic load from (TS -40°C/140°C). The major challenge was the extraction of the material characteristics of CucorAl wire bond. The characteristics of the CucorAl wire has been obtained by Nano indentation. Replacement by CucorAl wires can enhance the bonds lifetime without severe changes in metallization and bonding technology. Modeling of standard aluminum wire bonds reliability versus that of CucorAl wire bonds are the subjects of the paper.

In modern microelectronic devices, mechanical robustness and reliability have become more and more challenging, which is mainly due to continuous shrinkage of dimensions. At the same time a wide variety of thin film materials – metals, ceramics and glasses - are stacked and micro-structured for optimized device performance. Perfect adhesion of these thin film layers, even after exposure to thermal treatment or in presence of intrinsic stress, is crucial for device reliability. In particular, interfaces between isolating and metallization layers require careful assessment due to different bonding types of the materials involved.

Reliability is an obvious requirement in particular for the developing use of automotive electronics towards autonomous driving. Micro-electronics must also stay reliable in mounting situations, denoted as “3 rd level reliability”, which has been studied systematically. A methodology combining FEA and high precision optical metrology based measuring techniques was applied to test setups, which achieve "3rd level" component loads either by screwing onto aluminum heatsinks or by superimposed four-point bending loads, as was described in [1]. This paper reports the results of both temperature cycle and temperature shock tests up to 900 cycles -40/125°C, the system level deformation effects at higher cycle loading states as analyzed by the high precision optical measurements, and related FE-analyses, which aimed at adaption of the measuring results from component to system level. From deformation measurements a complicated response of the boards to mounting has been observed, including locally different effective board CTEs as a function of cycle number and an irreversible board stretching. It turned out that standard simulation assumptions like neglecting molding compound initial intrinsic stress, the use of geometrically linear theory, and the assumption of homogeneous temperature fields across the systems do not longer hold for system analyses. Measurements revealed significantly different transient temperature fields at heatsinks and board, dependent on position. The transient temperature gradients across the system were adopted by coupled transient thermo-mechanical simulations. They reflect the transient board deformations different from those simulated by assuming homogeneous temperature changes. Solder joints fatigue loading was considered for these cases and comparison was made to previous results, depending on mounting and w/o or with transient temperature gradients. A significant increase of joint stresses was observed, also with regard to the homogeneous T-distribution assumption. Very early solder fatigue failure was predicted and observed from thermal cycle tests.

Solder fatigue phenomena on lead-free solder joints have been subject of numerous studies. In most cases, reference was made to test cycling conditions. To address the reliability of solder joints subjected to service conditions, long-term studies have been made using two test setups. The first used a standard industrial board with either SAC 305 or Innolot joints and cycling conditions 22°C/95° C, 6 hours, and the other one a test board with SAC 357 joints and cycling conditions 18°C/80°C, 7.1 hours. Both cyclic conditions have been applied for approximately ten years. The tests have reached solder fatigue failure for the most critical components, which are reported in the paper. Acceleration factors are additionally reported for these components related to test cycling results performed in parallel. Two predictive methodologies were applied to evaluate the acceleration prediction by the testing results: analytical models of Norris/Landzberg type and numerical modeling (FEA). It is shown that the estimates of acceleration factors from published Norris/Landzberg formulas result in a wide span of predictive results. The solder damage prediction based on FEA is finally shown to be on the safe side, but its usefulness also for service conditions is obvious.

Recently performed lifetime modelling based on various power cycling tests, including tests with superimposed passive thermal cycles, indicated a strong influence of the mold compound with its characteristic glass transition temperature Tg on the failure mechanism wire-bond lift-off within discrete power device packages. The paper reports the discussion of this influence on the lifetime modelling and details its investigation. It also includes electro-thermo-mechanical finite element simulations as well as the introduction of a new optical in-situ analysis method capable of visualizing the in-plane deformations and strains in the package during power cycling.