Virtual Thermo-Mechanical Prototyping of Microelectronics Products - Towards Optimized Designing in Reliability
G.Q. Zhang, J. Bisschop, P. Maessen, Philips, P.O. Box 218, 5600 MD Eindhoven, The Netherlands, g.q.zhang@philips.com, jaap.bisschop@philips.com
Abstract
This paper presents the strategy and methodology for virtual thermo-mechanical prototyping of microelectronics products, specified by Philips. The major advantage of using this virtual thermo-mechanical prototyping method is that the design requirements and reliability qualification can be integrated, conducted and optimized at the earlier phase of the product development processes. The results of virtual thermo-mechanical prototyping can be used to predict, evaluate, qualify and eventually optimize the thermal and mechanical behavior of microelectronics products against the actual product requirements prior to major physical prototyping and manufacturing investments. This method is optimized Designing in Reliability based on the method of Physics of Failures.
1. Introduction
Thermo-mechanical (thermal, mechanical and thermo-mechanical) reliability of microelectronics products is one of the major concerns for the electronics industry. Currently, about 65% of all failures in microelectronics products are attributed to thermo-mechanical problems. This is expected to become even more critical in future products due to further miniaturization and function integration, which causes increased power dissipation density, higher interconnection density and higher reliability demands.
Philips has long recognized the vital importance of high quality in electronic components and its crucial effect on the viability and economics of finished equipment. This is especially true for semiconductors, which often perform critical circuit functions, such as handling high frequencies and transmitting high-speed digital data, often in a hostile environment.
Based on the root cause analyses from observed failures of microelectronics products, it is found that most of the thermo-mechanical reliability problems originate from the product and/or process design phase. However, within the electronics industry, the trial-and-error method (designing, building and testing of a multiplicity of physical prototypes) for thermo-mechanical design is still commonly used in product and process development. This method is characterized by several iteration cycles of designing, building and testing physical prototypes, and usually product reliability is only dealt with after physical prototyping. Products’ reliability qualification has been a matter of evaluation with the help of accelerated life tests for many years. Usually these tests were performed as the last step in the qualification of new products, and as a monitor of the performance. Reliability qualification programs used to be based on standard lists of accelerated tests. Most of these tests are updated versions of the MIL Std 883 tests. Reliability qualification programs can be derived from standards like EIA/JEDEC Standard 47, which gives guidelines for reliability qualification.
Due to fierce competition, ever-increasing market drive and strengthening of environmental legislation, the traditional thermo-mechanical design and reliability qualification methods are no longer competitive. These methods cannot guarantee that microelectronics products can be rapidly, economically and environmental-friendly developed and manufactured, and at the same time can satisfy performance and reliability specifications. Therefore, there is an urgent need to develop innovative thermo-mechanical design and reliability qualification methods for the further miniaturized microelectronics products with increasing complexity.
A recently developed methodology for reliability qualification is the Risk Assessment Process from the Semiconductor Quality Management Council (SQMC). This methodology includes a situation appraisal where customer or market requirements are collected, reviewed, and translated to requirements for the construction of the product. The risk is determined by the gap between the requirement and the known performance. Measures depend on identified risks. The risk assessment is an integral part of product development, so that the risk is continuously decreasing in the course of the development process. This methodology enables identification of potential problems, and finding solutions during the development, with a reduced risk in the final stage of the development.
Another development is the so-called Designing in Reliability method. In order to build in reliability in products and technologies a Physics of Failure approach is needed. It is necessary to understand the failure mechanisms and to have Physical models for them. Design rules are derived, based on these models together with experimental data. Many die-related mechanisms have been investigated in the past decades, and most of them are well understood. Some examples are hot carrier degradation, electromigration, threshold shifts due to mobile ions and gate oxide breakdown. These mechanisms are usually evaluated on special structures on wafer level. This allows early detection of process flaws and gives the necessary data for deriving design rules. Another class of failure mechanisms is associated with package-die interactions. These consist of thermo-mechanical mechanisms, caused by differences in expansion coefficients of the different materials and mechanisms induced by moisture and impurities. For many years these mechanisms have been evaluated with standard tests like Temperature Cycling and Temperature Humidity Bias test, which evaluate the ability of a product, but cannot be used for designing in reliability. Especially for the thermo-mechanical mechanisms the basic physics are known. More and more finite element simulations are used to calculate the temperature distribution and the evolution of stress in the device. Combined with the incorporation of physical mechanisms for material fracture and fatigue, these simulations can be used to find safe constructions.
This paper will present our development results for thermo-mechanical design and reliability qualification method for microelectronics products: Virtual thermo-mechanical prototyping. This method is one step further than the Designing in Reliability based on Physics of Failures. It is, in fact, optimized Designing in Reliability.
2. Methodology of virtual thermo-mechanical prototyping
Virtual prototyping is a very promising and emerging technology advance for product and process design and qualification methods. Figure 1 shows Philips’ strategy and methodology for virtual thermo-mechanical prototyping. The results of virtual thermo-mechanical prototyping can be used to predict, evaluate, qualify and eventually optimize the thermal and mechanical behavior of microelectronics products against actual product requirements prior to major physical prototyping and manufacturing investments. It is our believe that the development and application of this virtual thermo-mechanical prototyping method for thermo-mechanical designing and reliability qualification of microelectronics products will contribute to the sustainable business profitability of the electronic industry.
There are several important building blocks for the virtual thermo-mechanical prototyping methods. Among others, ‘reliable and efficient FEM-based thermo-mechanical simulation models’ and ‘Advanced simulation-based optimization methods’ are the two vital ones.
For the building block of “advanced simulation-based optimization methods,” in recent years, much effort [1-4] has been placed on the development and application of methodology and tools. Simulation-based optimization methods combine advanced simulation models that can predict the product and process behavior reliably and efficiently with ‘smart’ optimization methods. They form the most promising technological route towards the realization of product and process optimization. Compared with the physical test-based optimization methods, they are faster, cheaper, more reliable, and capable of carrying out multidisciplinary optimizations. The major advances are the creation of :
- sequential types of Design of Experiment (DOE) methods for simulation-based experiments,
- reliable Response Surface Modeling (RSM) methods for nonlinear responses,
- integrated procedure for simulation-based optimization by linking DOE, simulation, RSM and optimization algorithms. See figure 2 for detail.
The building block of “reliable and efficient FEM-based thermomechanical models” is the basis for virtual thermo-mechanical prototyping. FEM is a well-established technique for predicting the thermal and mechanical behavior of products and processes. This method has made significant progress especially during the last 20 years due to the rapid development of computer hardware and software based on efficient and robust computational algorithms and methodologies. However, the commercially available FEM tools are not specifically developed for application in the microelectronics industry and no attention is paid to the characteristics of TM analysis of micro-electronic products. Thus, some known bottlenecks and challenges are remaining while new ones are emerging, driven by the rapid industrial development and increasing demands on product complexity (further miniaturization and function integration). Examples of the identified bottlenecks are:
- Damage (and damage evolution) modeling (models and criteria) for miniaturized microelectronics products. Known stress or strain distribution alone is not sufficient to predict the thermo-mechanical behavior of microelectronics products.
- Characterization and modeling methodologies for process and geometric dependent material properties and their evolutions.
- Efficient algorithms for nonlinear (material, geometry and boundary) FEM simulations.
- Reliable experimental methods for model verification.
Optimized Designing in Reliability can be realized by:
- developing a reliable FEM-based physics of failure models;
- incorporating the user specifications (loading and environmental conditions) into the FEM- calculations;
- carrying out thermo-mechanical design optimization w.r.t the specified design parameters (geometry, material, process) and the associated design spaces.
3. The demonstrator
Due to the mismatch between the thermo-mechanical properties of different components in electronic packages and both the external and internal constraints, thermal stresses occur. Very often, these thermal stresses cause various types of thermo-mechanical failures during processing, testing, and use.
As the demonstrator, an electronic package consisting of die, solder and heat sink is considered. The vertical die crack, occurring in either the cooling down phase of the die-bonding process or the TCT test, is taken as the critical failure mode. The selected electronic package is used to demonstrate the major procedures and principles of virtual thermo-mechanical prototyping.
A parametric 2D FEM model with axisymmetric elements was developed using MARC. In this model, the solder material is modeled as temperature dependent visco-plastic, the heat sink as temperature dependent ideally plastic, and the die as temperature independent elastic. The maximum tension stresses at the middles of the bottom and top of the die are used as the crack index. If the predicted maximum stress level is close to or higher than the allowable stress level of the die, vertical die crack will occur. Experimental verification shows that this model can reliably predict the thermo-mechanical behavior of the package.
3.1 Package design specification
Three geometric parameters are chosen as the design parameters:
- thickness of the die (Tdie),
- thickness of the heat sink (Tsink), and
- length of the die(Ldie).
All the other parameters (process, material and geometric parameters) are assumed to be constant. Table 1 shows the three design parameters and their ranges of variations.
Using the verified nonlinear FEM model, the maximum tension stresses at the middle of both the bottom (StressBot) and the top (StressTop) of the die are calculated.
The design optimization problem is to choose the geometry parameters of the package such that the maximum of StressBot and StressTop is minimal, within the specified ranges of variations of design parameters. The objective function is defined as MaxStress = max{StressBot, StressTop}.
3.2 Design of computer experiments
A space-filling Latin-Hypercube-Design consisting of 20 design variations is first constructed. FEM simulations are carried out for all the 20 designs, and the StressTop and StressBot are used as the response parameters.
3.3 Developing RSM
For both of StressTop and StressBot, quadratic models with interactions are used for RSM generation. Using the pruning procedure based on cross-validation, the unimportant model terms were deleted; see Figure 3. The regression statistics are listed in Table 3 indicating that the reliability and accuracy requirements are satisfied (error < 5%).
Since these models are accurate enough, there is no need to do extra simulation runs or apply more flexible model types like Kriging models. To really benefit from Kriging in this case, more simulation runs are required. As a rule of thumb it is suggested to take 10-15 times the number of design parameters for Kriging models. In Figure 4 the simulated StressBot versus the predicted ones is plotted, showing the accuracy of the stress prediction using the developed RSM model.
3.4 Prediction and optimization
The optimal design obtained by minimizing the MaxStress target is showed in Table 4. Tension stress reduction of more than 20% can be achieved, compared with the best-simulated package design obtained from Step 2. Figure 5 shows the 3D plot of StressTop as function of die and heat sink thickness.
4. Conclusions
This paper presents part of our research and development results for virtual thermo-mechanical prototyping of electronic packages. The major advantage of using this virtual thermo-mechanical prototyping method is that the design requirement and reliability qualification can be integrated, conducted and optimized at the earlier phase of the product development processes. The results of virtual thermo-mechanical prototyping can be used to predict, evaluate, qualify and eventually optimize the thermal and mechanical behavior of microelectronics products against the actual product requirements prior to major physical prototyping and manufacturing investments. This method is optimized Designing in Reliability based on the method of Physics of Failures, which makes the first time right (die crack free design), shorter-time to market and optimized package design possible.
5. References
- G.Q. Zhang, “The state-of-the-art of simulation-based optimization,” Phips internal report, 1998.
- G.Q. Zhang, J. Janssen, J. Bisschop, Z.N. Liang, F. Kuper, R. Schravendeel, L.J. Ernst, “Virtual thermo-mechanical prototyping of electronic packaging using Philips’ optimization strategy,” IMAPS 2000, USA, 2000
- G.Q. Zhang, A. Tay and L.J. Ernst, “Virtual thermo-mechanical prototyping of electronic packaging - Bottlenecks and solutions of damaging modeling,” 3rd Electronic Packaging Technology Conference (EPTC), Singapore, 2000
- G.Q. Zhang and P. Stehouwer, “Simulation-based optimization in virtual prototyping of electronic packaging,” Proceedings of EuroSimE2000, Kluwer Academic Publisher, 2000
| Table of Contents | Previous Page | Next Page |
