The global market for semiconductor chips is $270 billion. Growth in this industry is driven by the need to enhance performance by packing more transistors into the same chip area, with a relationship between feature size and time that approximately follows Moore's Law. As the feature size shrinks below 45 nm, the limits in performance of conventional materials are being reached. The industry is moving towards a metal gate/high-k dielectric combination in which fully silicided (FUSI) metal gates are receiving a lot of attention for the metal layer. Advanced submicrometer CMOS technology requires the development of new materials, centered on the utilization of high dielectric constant materials for the gate oxide. Along with the new oxide materials, new gate metallizations are being used. Because of their desirable work function and thermal properties, metal silicides are receiving a lot of attention. An important example is the Ni-Si intermetallic system, which is made by thermally annealing separately deposited Ni and Si. The particular intermetallic phase or phases that form depend on the ratio of Ni to Si and on the annealing process. Further, the thermal annealing process must be compatible with other processes that are used in making the semiconductor devices on the wafer. This favors rapid and ultra-rapid processes. Conventional calorimetry has served as a key tool for the development of phase diagrams, but is limited to heating rates much slower than required for characterizing these fast annealing processes.
Because silicide formation is an irreversible process, study of the effect of heating rate on silicide formation requires a separate sample for each measurement. Using a micromachined differential scanning calorimeter (DSC) developed at NIST, studies of nickel/silicon multi-layers has shown the surprising result that ultra-rapid heating enhances the mixing between Ni and Si phases. The NIST DSC is able to make measurements at ramp rates that are much faster than are possible with conventional DSC instruments, and are relevant to rapid thermal annealing processes. An advantage of these chip calorimeters is that they can be formed as arrays, so that one deposition process can produce many samples for scanning at different rates. Six-element arrays were used to produce both melting point samples for calibration (which are reversible) and Ni/Si samples. A element is shown to the right. For each device a series of identical scans were collected, with the Ni/Si reaction usually complete by the second scan. The scans were analyzed to produce enthalpies and reaction temperatures as a function of scan rate. After the DSC measurements are completed, samples consisting of a superlattice of Ni/Si layers are prepared for TEM examination using an elaborate FIB process involving the cutting out of a square of the DSC device, thinning it, and welding it to a TEM grid. TEM observations show the unexpected result that there is enhanced interdiffusion of Ni and Si at high ramp rates compared to slow ramp rates. The results point towards the likelihood of metastable melting. This idea is supported by calculations which show that if certain NiSi phases are suppressed there can be a dramatic lowering of the melting temperature. The enhanced mixing at lower temperature is desirable for the formation of silicides.
Work in the coming year will focus on investigating the utility of micromachined calorimeters to characterize processes in liquids. In drug development, differential scanning calorimetry has been shown to be a useful tool for characterizing formulation stability. Nanocalorimetry offers the potential to measure smaller samples with higher throughput and higher scan rates than is possible with conventional instruments.