Lifetime Prediction of an IGBT Power Electronics Module ... - IEEE Xplore

Lifetime Prediction of an IGBT Power Electronics Module under Cyclic Temperature Loading Conditions Hua Lu*, Chris Bailey University of Greenwich 30 Park Row, London SE10 9LS * Email: [email protected] Abstract The lifetime of an IGBT power electronics module under cyclic temperature loading conditions has been analyzed using Finite Element Analysis method. The failure mechanisms that have been taken into account are the fatigue of the chip-mount-down solder joint, the substrate attach solder joint, the busbar solder joint and the Aluminum wirebond. The results show that the lifetime of the module is about 1000 cycles under the -40 to 125C cyclic temperature loading condition. The critical failure location has been found to be the busbar solder joint and the lack of compliance of the busbar design is the cause of the problem. The objective of this paper is to demonstrate the methodology of using physics of failure approach for the reliability analysis of power electronics modules and highlights the important design parameters that affect the thermal-mechanical fatigue failures of these components. Power Module Failure Mechanisms Power electronic modules (PEM) are self-contained power electronics components that are widely used in aerospace, automotive and alternative energy generation and distribution applications. They play an important role in the conversion, control and delivery of electrical power [1]. PEMs have highly inhomogeneous structures. They are made of semiconductor, ceramic, copper, aluminum, polymer, and sometimes composite materials. These materials are assembled together in the packaging manufacturing process using soldering, direct bond copper (DBC), wirebond, and pressure contact interconnect techniques. In the service and accelerated qualification test conditions, the temperature of the power module changes with time. This gives rise to fluctuating stress and strain due to the mismatch of coefficient of thermal expansion in PEMs. The stress and strain will lead to the degradation and ultimately failure of the interconnection and the modules and this is one of the most important PEM failure mechanisms. As the current trends in the power electronics applications leads to ever greater currents passing through the components and ever smaller spatial profile, this failure is bound to become more and more critical to the PEM reliability. In order to predict the reliability of PEMs effectively, it is crucial that the failure mechanisms are well understood and the lifetime are predicted accurately. In this paper, the methods of using Finite Element computer simulation combined with experimental results to predict the lifetime of PEMs (Figure 1) is described. This approach to reliability analysis is called the Physics of Failure (PoF) method. PoF method takes into account the root causes of failures and the lifetimes are predicted using computer simulations

appropriate for the underlying physical processes that have caused the failure.

Figure 1. An IGBT power electronics module design. Under cyclic electric or passive thermal-mechanical loading, the failures of PEMs may be caused by physical phenomena such as solder joints fatigue, wirebond, cracking fatigue, isolation substrate delamination [2-6]. In this work, four failure mechanisms are included in the analysis: the chip mount-down solder joint, the substrate mount-down solder joint, the busbar solder joint and the aluminum wire bond. In this paper, the design parameters that affect these failures are discussed and the lifetime of an ad hoc PEM design has been predicted. Lifetime Prediction Methods Busbar Solder joints PEM busbars are made of copper. They serve as electric terminals and they are connected to the substrate using soldering technique. In Figure 2, a PEM with four busbars is shown.

Figure 2. PEM with four busbars. Busbars may fail because of the solder joint fatigue or delimination at the interfaces around the joint. For the solder joint fatigue mechanism, the lifetime can be predicted using similar prediction method that has been used in

2009 International Conference on Electronic Packaging Technology & High Density Packaging (ICEPT-HDP) 978-1-4244-4659-9/09/$25.00 ©2009 IEEE

274

microelectronics solder joint reliability analysis [7]. In this type of analysis, the response of the solder joint under a prescribed cyclic loading condition is obtained using Finite Element Analysis method, and the damage indicators such as the accumulated plastic/creep strain per cycle Δε p or plastic work density per cycle

ΔW p in solder joints can be

calculated in the post-processing stage of the analysis. In general, damage indicator distribution is not uniform and as the crack grows, the geometry of the structure changes and this results in changes stress and damage indicator distribution. As an approximation, however, these are not taken into account and the maximum values or the average values in the solder volume where crack propagate are used for reliability prediction of busbars. The mechanical properties of solder is highly nonlinear. Even at room temperature, creep is an important deformation mechanism. In this work it is assumed that the solder joints are made of Sn3.5Ag and the visco-plastic/creep constitutive equation for SnAg alloy is,

ε

cr

= A × sinh n (ασ e ) exp(

−Q ) RT

(1)

where R is the gas constant, T is the temperature in Kelvin, σe is the von Mises equivalent stress, A, n, α, Q are material constants and their values are listed in Table I [8]. Table I: Creep parameters for solder materials.

SnAg

A(s)

n

α(1/MPa)

Q/R

9.00E+05

5.5

0.06527

8690

To calculate the Sn3.5Ag solder joint lifetime after the damage indicator has been obtained, Equation 2 is used.

L / N = 0.00562(Δε p )

1.023

(2)

where L is the crack length and N is the number of load cycles that have caused this crack. The unit of L is in mm. The crack length is predefined by failure criteria. In this analysis, it is defined as Large solder joints.

In PEMs, chip-mount and substrate solder joints are thin and much larger than microelectronics solder joints. For example, for the power module shown in Figure1, the width of the substrate is 58 mm and the solder thickness is 0.1mm and the aspect ratio is 580. In contrast, BGA and flip chip solder joints have ball shapes and the aspect ratio is in the order of magnitude of 1. In order to take into account of the effect of the stress distribution and changes caused by crack propagation, a method that is based on the Miner's linear damage accumulation is used in this paper for the lifetime prediction of these solder joints [9]. In illustrate how this method works, let's assume the 2D solder joint structure shown in Figure 3 is subject to cyclic loading and crack is expected to propagate along the soldersubstrate interface where strain has the highest value. The crack path can be conceptually divided into a number of segments. Each segment can be defined as one or a number of Finite Element mesh elements in the solder joint are. These segments can be regarded as small solder joints and the damage indicator is a constant within each solder segment. For these small "solder joints", Equation 2 can be used to calculate the lifetime in terms of number of cycles to failure [8]. Damage indicators are not uniform in the solder joint. This means that the solder segments will not fail at the same time. At the edge the solder joint, the solder joint segment fails first and then the one next to it will fail and so on so forth, resulting in a crack propagation process that starts from the edge towards the centre of the solder joint. The lifetime of the first segment, i.e. the crack tip, can be calculated using Equation 2. The lifetime of the next segment has to be calculated differently because when the first segment is not cracked, the load on the second segment is not as severe as the load on the first segment, but after the failure of the first section, this second segment becomes the crack tip and load increases. In effect, this second segment will have experienced two load levels before it fails. Similarly, the third sections experiences three levels of load before its failure. As it shown in [9], the shape of the damage indicator distribution remains very much the same and this means that the damage indicator value is a function of the distance from the crack tip only. By exploiting this characteristics of the large solder joint and by using the Miner's linear damage accumulation rule, the lifetime of each solder segment, NFi, can be calculated using Equation 3.

0 ⎛ α1 ⎜ ⎜ α 2 α1 ⎜ ... ... ⎜ ⎝ α k α k −1 where

Figure 3. A 2D substrate solder joint model.

0 ⎞⎛ NF1 ⎞ ⎛1⎞ ⎟⎜ ⎟ ⎜ ⎟ 0 ⎟⎜ NF2 ⎟ 2 = N1 ⎜ ⎟ ⎜ ... ⎟ ... ... ⎟⎜ ... ⎟ ⎟⎜ ⎟ ⎜ ⎟ ... α1 ⎠⎝ NFk ⎠ ⎝k⎠

... ...

(3)

α i = N1 / N i . N i are the number of cycles to failure

under the ith load level and they are calculated using Equation 2.

2009 International Conference on Electronic Packaging Technology & High Density Packaging (ICEPT-HDP)

275

Wirebond lifetime In PEMs, wirebonds carry strong electric currents and they are often made from heavy Aluminum wires. In thermal mechanical cyclic test environment, the difference between the thermal expansion coefficients of the wire and the substrate material causes mechanical fatigue at the wiresubstrate interface (bond lifting) or at the bond heel (heel cracking) as shown in Figure 4. For wirebonds with no globtop, the dominant failure will be bond-lifting under thermalmechanical loading conditions. Under power cycling condition, however, this may not be the case because wire flexing will cause more damage to the heels. In this work, thermal-mechanical loading is applied and only the fatigue at the bond interface is assumed to be the failure mechanism.

Figure 5. Finite Element model of a busbar. The compliance can also be enhanced by reducing the thickness of the copper. As shown in Figure 6, the plastic work density in busbar solder joint decreases as copper thickness is reduced from 2.5 to 1 mm. In this analysis, the top- part of the busbar is constrained in the vertical direction. 0.24 3

dW(MPa)

2

y = 0.0041x - 0.0312x + 0.0927x + 0.1292

0.23 0.22 0.21 0.2 0.19

Figure 4. Possible wirebond failure sites.

1

In this work, the accumulated plastic strain per cycle is used as the damage indicator for the analysis of the wirebond reliability. For the failure at the bond interface, lifetime can be calculated from the following lifetime model.

N = 1.18ε p−3.492

(4)

The failure criteria on which this lifetime model is based is a 90% reduction in shear strength. Effects of Design Parameters on the Reliability of Power Modules Busbars The reliability of the busbars is to a great extent determined by their mechanical compliance. The compliance can be controlled by changing the thickness and the shape of the busbars. Figure 5 shows a computer model of typical busbar design. The bends in the busbar are used to increase the compliance.

1.5

2

2.5

Cu thickness (mm)

Figure 6. Plastic work density in busbar solder joint as a function of copper thickness. Other parameters such as the copper's Young's Modulus and the yield stress have also been found to have little impact on the busbar solder joint. Large solder joints Table II shows the accumulated plastic strain values for 15 substrate designs. It can be seen that the reliability of large solder joint is mainly affected by its thickness. The size of the substrate will also has impact on substrate solder joint reliability because it takes longer for the cracks to propagate in a larger solder joint than in a small one. In contrast, in a non-underfilled flip chip or ball grid array, chip size greatly affect the reliability because the solder joints are not strong enough to stop the relative movement of the materials on the two sides of the joints and the stress in solder joints increase as chip size increases.


276

Table II: Effect of design parameters on reliability of substrate solder joint. Solder Substrate Baseplate ∆εp thickness(mm) thickness(mm) thickness(mm) 0.1 60 9 0.0158 0.1 60 5 0.0144 0.1 54 9 0.0158 0.1 54 5 0.0144 0.1 57 7 0.0154 0.2 60 7 0.0092 0.2 54 7 0.0092 0.2 57 9 0.0094 0.2 57 5 0.0087 0.2 57 7 0.0092 0.3 60 9 0.0070 0.3 60 5 0.0066 0.3 54 9 0.0070 0.3 54 5 0.0066 0.3 57 7 0.0069 0.1 60 9 0.0158 Wirebonds The reliability of wirebonds has been found to be affected by wire diameter, loop height, purity and manufacturing process. In Figure 7 and Figure 8 the maximum accumulated plastic strains at the end of three thermal-mechanical cycles are shown as functions of wire diameter and the loop height. In this case, the wire diameter has more important effect on the reliability than the loop height. However, by reducing the wire diameter, the current carrying capability is reduced and therefore may not always be an design option.

Figuire 8. Plastic strain strain as a function of the loop height. Lifetime Prediction of The IGBT Module In order to predict the lifetime of the IGBT module shown in Figure 1, three models have been used. In the first model, there are the three busbars mounted on the substrate using SnAg solder (Figure 9). There are two different busbar solder joint sizes in this PEM. They are 4x2x0.1mm and 4x2.5x 0.1mm respectively. The busbar in the middle has the larger solder joints.

Figure 9. Busbar Model of the IGBT model.

Figure 10. Deformation of the PEM under a fixed thermal load of 100°C.

Figure 7. Plastic strain as a function of the wire diameter in Al wire.

As busbars, baseplate and much of the substrate are made of copper, CTE mismatch is not the most important factor in causing damage in busbar solder joint. Instead, the boundary constraints may have great impact on the solder joint reliability. PEMs are usually filled with silicone gel and epoxy resin and as temperature changes they expand or contract to push the top of the casing up and down. This movement causes cyclic deformation the damage in the busbar solder joints. To include this effect, the expansion and contraction of the casing has been modeled and the the time


277

dependent displacement values are used as mechanical loading conditions for the busbar simulation. Figure 10 shows the deformed shape of the PEM casing. The chip mount-down and substrate solder joints are contained in one single model. Figure 11 shows half of the cross section of the model. The solder joints are both 100 microns in thickness. Figure 12 shows the stress and strain in the structure. While the stress is the highest in the silicon the strain concentration is in the solder joints. Figure 14. The accumulated plastic strain distribution after three thermal cycles.

Figure 11. Substrate and chip mount-down solder joint model

Cyclic temperature loading conditions are applied to all models. The ramp time and dwell time are 15 minutes and the minimum and maximum temperatures are -40°C and 125°C respectively. In order to define the lifetime, it is necessary to define the failure criteria. For the busbar, failure is defined as having 50% of area delamination while for the chip and the substrate solder failure is defined as the 20% area delamination. The failure criteria for the wirebonds is 90% reduction in shear strength. The lifetimes of the components in the PEM are listed in Table III. For this particular design, the busbar solder joint is the first to fail and its lifetime is the lifetime of the whole PEM. The results also show that chip solder would fail before substrate solder although the damage indicator Δε p has a lower maximum value in the chip solder. This is because the size of the chip solder is much smaller than the substrate solder.

Figure 12. Stress (top) and strain (bottom) distribution in the solder joints Part of the wirebond model is shown in Figure 13. The Al wire has a diameter of 375µm, the length of the bond foot is 1 mm, the loop length is 2.85 mm and the loop height is about 1.2mm. Figure 14 shows that plastic strain concentrates at the bod interface.

Figure 13. The wirebond model.

Table III: Lifetime predictions Component Lifetime (number of cycles) busbar 946 wirebond 10489 Chip solder 14889 Substrate solder 30000 Conclusions Physics of Failure methods have been used to analyze the reliability of power electronics module designs. The lifetime of a power module has been predicted. Because of a lack of compliance, a busbar has been found to be the first to fail. Acknowledgments The authors wish to acknowledge the support of the Innovative electronics Manufacturing Research Centre (IeMRC) and the United Kingdom Technology Strategy Board for the project ‘Modelling of Power Modules for Lifetime, Accelerated Testing, Reliability and Risk’. The authors would like to thank Semelab Ltd and Dynex Semiconductor Ltd. for their great contribution to the work and unreserved support. The author would also like to acknowledge the support from Goodrich Engine Control, SR Drives Ltd., Areva T&D Ltd and Rolls Royce Plc.


278

References 1. Sheng, W.W. and Colino, R.P., Power Electronic Modules, CRC Press (2005) 2. Shammas, N.Y.A., “Present problems of power module packaging technology” , Microelectronics Reliability, Vol. 43, Issue 4 (2003), pp. 519-527 3. Pooch, M.-H., Dittmer, K.J., Gabisch, D., “Investigations on the damage mechanism of aluminum wire bonds used for high power applications”, Proc. EUPAC 96, (1996) pp. 128-131 4. Günther, M., Wolter, K, Rittner, M, Nüthter, “Failure Mechanisms of Direct Copper Bonding Substrates”, Proceedings of Electronics Systemintegration Technology Conference (ESTC), Dresden, Germany, 2006, pp.714718 5. Dupont, L., Khatir, Z., Lefebvre, S., and Bontemps, S., “Effects of metallization thickness of ceramic substrates on the reliability of power assemblies under high temperature cycling”, Microelectronics Reliability, vol. 46 (2006) pp.1766–1771 6. Yoshiyuki Nagatomo and Toshiyuki Nagase, “The study of the power Modules with High Reliability for EV Use”, Proceedings of the 17th International Electric Vehicle Symposium (2000). 7. Syed, A., "Accumulated creep strain and energy density based thermal fatigue life prediction models for SnAgCu solder joints", Proceedings of the 54th Electronic Components and Technology Conference, pp.737-746, (2004) 8. Lau, J.H. (editor), Ball Grid Array Technology, McGrawHill (1995), p.396 9 Lu, H. Tilford, T., Bailey, C. and Newcombe D.R., "Lifetime Prediction for Power Electronics Module Substrate Mount-down Solder Interconnect", Proceedings of The 2007 International Symposium on High Density Packaging and Microsystem Integration, 2007, pp.40-45


279