IEEE TRANSACTIONS ON DEVICE AND MATERIALS RELIABILITY, VOL. 8, NO. 3, SEPTEMBER 2008 471 Temperature-Level Effect on Solder Lifetime During Thermal Cycling of Power Modules Mounira Bouarroudj, Zoubir Khatir, Jean-Pierre Ousten, and Stéphane Lefebvre Abstract—In this paper, we show that, during thermal cycling, the solder lifetime of power modules is not only dependent on temperature variation, but we also highlight the inﬂuence of some other key parameters such as upper and lower dwell temperature levels. In particular, we show the inﬂuence of these parameters on the solder crack initiation and propagation in the solder layer between the direct copper bonding and base plate of high-power insulated gate bipolar transistor modules. For this purpose, both experimental and numerical investigations have been carried out. Concerning thermal cycling tests, three temperature proﬁles have been done: −40 ◦ C/120 ◦ C, 40 ◦ C/120 ◦ C, and −40 ◦ C/40 ◦ C. Results have shown that stress values in the solder are monitored by the low temperature level and that the strain is monitored by the high-level one. We observed that the relative magnitude of strain variations is larger than that of stress variation. In or- der to understand experimental results, ﬁnite-element simulations with various high and low temperatures have been performed. Results have pointed out that the solder exhibits two different mechanical behaviors, depending on whether the upper dwell temperature (Tmax) exceeds or not a homologous temperature of approximately 0.74 T m . When Tmax is below this value, shear strain variations remain in relatively small range values, and shear stress variations have a linear dependence with the temperature variation. In these conditions, only energy-based models should be used for solder lifetime estimation. On the contrary, when Tmax is above 0.74 T m , shear stress variations reach a saturation value while inelastic shear strains increase signiﬁcantly. Therefore, in these conditions, either strain- or energy-based models could be used for solder lifetime estimation. Finally, the thermal cycling behaviors of a lead-free solder (SnAg3Cu0.5) and a lead-based one (SnPb37) are numerically compared. Index Terms—Finite-element analysis (FEA), insulated gate bipolar transistors (IGBTs), packaging, power electronic modules, thermal cycling tests. I. I NTRODUCTION T HERMAL cycling is often responsible for thermomechan- ical damages of power electronics devices [1], [2]. Such constraints lead to crack initiation followed by crack propaga- tion inside solder attach materials. In the case of insulated gate bipolar transistor (IGBT) power modules, the weakest attach Manuscript received February 28, 2008; revised April 11, 2008. First published August 12, 2008; current version published October 16, 2008. M. Bouarroudj is with the French National Institute for Transport and Safety Research, 94114 Arcueil, France (e-mail: bouarrou@inrets.fr). Z. Khatir and J.-P. Ousten are with the Laboratory of New Technology, French National Institute for Transport and Safety Research, 94114 Arcueil, France (e-mail: khatir@inrets.fr; ousten@inrets.fr). S. Lefebvre is with the Conservatoire National des Arts et Métiers and SATIE Laboratory, ENS de Cachan, 94235 Cachan, France (e-mail: lefebvre@satie.ens-cachan.fr). Color versions of one or more of the ﬁgures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identiﬁer 10.1109/TDMR.2008.2002354 layers are the solders between direct copper bonding (DCB) and base plates due to the mismatch of the coefﬁcients of thermal expansion and the large areas of these solders [2]–[5]. For example, in the case of automotive applications, wished failure rates are very low in spite of very severe operating conditions. Therefore, the lifetime assessment of a given power device packaging technology is an important stake. This can be achieved by using models that require both experimental and simulation results. In particular, accelerated tests may be used but it must be possible to extrapolate the lifetime results from accelerated conditions to normal operating conditions [6]. Generally, the acceleration factor (AF) between two test conditions characterized by temperature variations ΔT 1 and ΔT 2 for solder lifetime in thermal cycling conditions is com- puted by using a simple Cofﬁn–Manson relation [6]: AF = (ΔT 1 /ΔT 2 ) −m . This model of extrapolation is only related to the temperature variation and is probably too simple to be used in all operating conditions. Obviously, the highly nonlinear behavior of solder joints makes the lifetime prognosis in real conditions difﬁcult to estimate. Therefore, it is important to understand the role played by each parameter of thermal-cycle proﬁles in the solder joint lifetime. Some answers have been already reported concerning the effects of dwell time durations and temperature ramp rate in a solder joint lifetime [7]–[10]. Some papers [10]–[12] have already reported test results on the effects of Tmax and Tmin on fatigue life during thermal cycling of ﬂip chip assemblies with bumps, but such results are fewer concerning large solder joints. In [11], Pang et al. have proposed scale factors for correlating ﬂip chip solder joint fatigue life from accelerated conditions with large ΔT to smaller temperature ranges. In this paper, focusing only on large solder joints, we show that a solder lifetime is not only dependent on temperature vari- ation (ΔT ) but we also emphasize the inﬂuence of some other key parameters, such as dwell temperature levels. In particu- lar, we highlight how the upper and lower temperature levels (Tmax, Tmin) affect the shear stress and inelastic strain varia- tions during thermal cycles and ﬁnally play a signiﬁcant role in the AF. In a ﬁrst step, thermal cycling tests were performed on integrated six-pack IGBT power modules (600 V–200 A) where the studied layers are solder joints between alumina DCBs and copper base plates. Several test conditions with two ΔT ’s and different upper and lower temperature levels have been realized on which solder crack initiations have been detected and crack propagation rates have been measured. In a second step, these experimental tests are completed by ﬁnite-element analyses (FEAs) to show the mechanical behavior dependence of the solder behavior with temperature levels. Finally, the thermal 1530-4388/$25.00 © 2008 IEEE