A MODEL FOR THERMAL FATIGUE OF LARGE AREA ADHESIVE JOINTS BETWEEN MATERIALS WITH DISSIMILAR THERMAL EXPANSION zyxw by Are Bjornekletta, Tom Tuhusb and Helge Kristiansenab aSINTEF SI, Oslo, Norway bDept. of Physics, University of Oslo, Norway ABSTRACT zyxwvutsrqp A model describing thermal fatigue of large area adhesive joints such as die bonds, has been developed. It is based on equations for crack growth rate and stress distribution in large area joints. The basic assumption of the model is that cracks grow from the edges of the area towards the center. The thermal resistance of the bond layer was calculated by assuming the cracked part of the layer had infinite thermal resistance. The thermal resistance as a function of the number of thermal cycles was predicted to be different for adhesives with low and high modulu of elasticity. Good agreement with previously reported experiments was obtained. The thermal resistance in silver filled die bond adhesives as a function of the number of thermal cycles was measured in these experiments. INTRODUCTION Large area adhesive bonding of materials with different thermal expansions is common in electronic packages and assemblies. Heat sinks are bonded to various packages for increased cooling efficiency. Silicon chips are attached to chip carriers, hybrid substrates and lead frames. Large area hybrid substrates are attached to heat sinks. The heat transfer characteristics of such bonds are important in electronic systems with high power dissipation. The thermal expansion mismatch creates thermal stress in the bond layer and the adherents. Another aspect of this problem is the cyclic stresses caused by thermal cycling. If the number of cycles is sufficiently high, the bond layer will fatigue. The fatigue process is either crack growth or delamination. The crack increases the thermal resistance of the bond layer. A deteriorating heat transfer capability of the bond is thus caused by the thermal cycles. Electronic systems are often subjected to a thermal cycling test as a part of quality assurance and qualification. MIL-STD- 883C, Method 1010.6 is commonly used. Condition B of this test specifies thermal cycles between zyxwvutsr -55 and 125°C and the number of cycles is typically between 10 and 1000. The temperature range in this test is more severe than in most real applications. However, the number of cycles in real applications may outnumber those of the test considerably. Automotive electronics components may be cycled between 10 and 80°C once or twice every day for ten or more years. Components in low earth orbit spacecrafts may be cycled 14 times per day or 5000 cycles per year caused by consecutive sunlight and shadow. Numbers may be even higher for power on/off cycles. An electrical vehicle in city traffic may start and stop within periods of a few minutes causing temperature cycling of drive electronics components. Temperature control electronics may switch odoff within a period of a few seconds to minutes depending on the thermal time constant of the thermal mass under control. In a concurrent engineering environment1, it will be more desirable to have physical models that describe the effects of thermal fatigue rather than merely observing the outcome of thermal cycling tests. Such tests may even be irrelevant for the operating environment of the system that is tested. With good physics of failure models it is possible to calculate the life time of a system at the design stage. This will be tremendous cost effective compared to building the system and submitting it to a thermal cycling test with a possible redesign if the test is not successfully accomplished. Such models will also enable the designer to tailor the system to various quality levels. We have previously reported experimental results for the increase in thermal resistance in die attach layers during temperature cycling2. These experiments were based on silicon test chips bonded with various silver filled adhesives to substrates with different thermal expansion coefficients. Two separate mechanisms were causing the increase in thermal resistance; (1) crack growth and (2) delamination. In the present work, a model for the fatigue process has been developed. Equations describing the thermal resistance as functions of temperature cycles are developed and they are compared to the previous experimental data. THEORETICAL MODEL Investigation of silver filled die bonds have shown that crack growth is initiated at the edges and in particular at the corners of the bonded area. This is illustrated in Figure 1 where various stages of bond layer fatigue are shown schematically. Figure 2 shows a crossection between two diagonal corners of a bond area. The half diagonal length of the area is L and the length from the center to the crack tip is 1. The starting point for our thermal fatigue model is an equation describing the propagation rate of the crack tip as a function of geometrical, material and environmental parameters. This equation known as the Paris-Erdogan equation3, gives the crack growth rate in materials exposed to cyclic stress, zyxwv 61 zyxwvutsrq = -c(AK)~ In this equation, -dl is the increase in crack length per cycle dN, C and m are material constants and AK is the difference in stress intensity factor between low and high temperature. The 0-7803-1852-&r94/53.00 81994 IEEE 138 Tenth IEEE SEMI-THERW