TEMPERATURE-DEPENDENCE OF THERMAL-EXPANSION OF CERAMICS AND METALS FOR ELECTRONIC PACKAGES

We have measured the x-y (in-plane) expansivity over the range of −40–140°C of several of the ceramics used in modern hybrids, single-chip packages, and multichip modules. We compare these thermal expansion data to those of silicon and the materials used for leads, lids, heat spreaders, and sinks. High performance packages, usually made of ceramic dielectrics, may be required to survive under extreme temperatures, such as −50–150°C. Thermally induced strains constitute a major limiting factor to their reliability under these conditions. Material selection based on the coefficient of linear thermal expansion (which we call expansivity) can, therefore, be critical. It is not always appreciated how temperature-dependent expansivity can be. The expansivity of pure polycrystalline alumina varies from 4.1 to 7.1 ppm/K as the temperature is varied from −40 to 140°C. Thus it may be inadequate to estimate thermal strains from a single number reported for a given material. This is especially true for ceramics, where suppliers are mostly interested in dynamics of firing and cooling over many hundreds of degrees, so they use and report data averaged over a temperature range including temperatures many hundreds of degrees above room temperature. For circuit reliability, we are mostly interested in the temperature range to which the circuit will be subjected. This may include temperatures as low as −65°C, but seldom involves temperatures above 150°C. True expansivity is defined as (1/L)dL/dT where L is a length, and T is the temperature. This is not the same as an average over a temperature range, as usually reported. Our measurements use strain gages which are calibrated against the expansivity of silicon as a standard. The expansivity measurements were reproducible within a standard deviation of 0.26 ppm/K. We measured the expansion in the x and y directions on both sides of a sample of the substrate. That way, if there is appreciable bowing of the material, that is also measured. We have measured the expansivity of aluminas of different purities (92–99.5%), low temperature co-fired ceramics, aluminum nitride, and beryllia. With aluminas, the expansivity increases slightly with increasing purity, especially at the higher temperatures. Low temperature co-fired ceramics can be formulated to have a dielectric constant (κ) anywhere between 4.8 and 8.0. We measured a κ = 4.8 sample which was expansivity matched to Si, and a κ = 8.0 sample matched to alumina. We found no significant effect of metallization of the multilayers on the expansivity. Aluminum nitride is well matched to silicon, whereas beryllia is well matched to alumina. In addition to the ceramic measurements, we measured Kovar, Invar, Alloy 42, molybdenum, and two composites: copper-molybdenum-copper and copper-Invar-copper. Kovar is well matched to alumina ceramic at room temperature; below and above room temperature, however, the expansivity of Kovar diverges from that of alumina. All of these low expansivity “nickel steels” have low expansivity over a specific temperature range. Above and below that range, the expansivity is considerably greater. For Kovar, the minimum expansivity occurs above the range of these measurements, so as the temperature is lowered, the expansivity of Kovar increases, whereas the expansivity of all ceramics investigated decreases. At −40°C, the expansivity of Kovar is almost twice that of alumina. A similar temperature-dependence is observed for Alloy 42. For Invar, the minimum expansivity occurs near room temperature, so over most of the range of these measurements, its expansivity increases with increasing temperature. In the composite structures, interesting results are obtained depending on the elastic limits of the copper, which can be exceeded, causing large hysteresis effects, especially in the case of copper-Invar-copper. Based on these measurements, the package designer can make appropriate materials choices to avoid the problems caused by large thermal stresses. © 1990 IEEE