UHV-TEM

Ultra high vacuum transmission electron microscopy

Silicides Silicides are important components of high density integrated circuits. By depositing metals on Si using in situ evaporators, we can examine the structures and phase transitions which occur in these complex and interesting materials. TiSi2: the effect of oxygen on the C49 to C54 phase transformation TiSi2 is used in low resistance contacts where it is formed by annealing a layer of Ti deposited on Si. Annealing at 600-700°C first forms a high resistivity metastable phase of TiSi2 known as C49, and this must be converted into the lower resistivity, stable C54 phase by a rapid thermal anneal. Unfortunately, the conversion of C49 TiSi2 to C54 TiSi2 is retarded in small areas, only occurring at temperatures above ~900°C. It is important to understand why the phase transition is so difficult and how it can be encouraged to occur at lower temperatures. Because Ti is such an efficient gettering agent, sputtered films of Ti contain substantial amounts of oxygen and even relatively clean films oxidise further between growth and annealing. We therefore used the capabilities of the UHV-TEM to examine the effect of oxygen impurities on the TiSi2 phase transition. This was done by comparing the phase transformation in clean, in situ evaporated Ti with the same reaction in ex situ sputtered films which contained appreciable amounts of oxygen. The phases formed during heating of the UHV evaporated film appear indistinguishable from those for the sputtered film. However the kinetics of the C49-C54 phase transformation are quite different in the two cases. The transition occurs smoothly in the UHV film, with large C54 grains steadily consuming the smaller C49 grains. The reaction velocity is independent of the orientation of the C49 grains. In contrast, the sputtered film undergoes a jerky transformation, pinning at grain boundaries then moving rapidly across C49 grains. We attribute the differences in reaction kinetics to oxygen segregation to the grain boundaries. Since C49 and C54 have the same stoichiometry, the reaction in the UHV film can occur without long range diffusion. However, the reduction in grain boundary density which occurs during the transformation requires diffusion of any grain boundary segregants to other sites, a process which appears to reduce the reaction rate in the sputtered film. The differences in reaction kinetics are striking and show that during normal processing the propagation of the phase boundary can be rate-limited by oxygen and other impurities at grain boundaries in the C49 phase. This suggests that changes in processing conditions which reduce the oxygen content might accelerate the phase transition and thereby improve contact properties. Figure 1: Phases formed during the heating of a 10nm thick UHV-deposited Ti film. As-deposited, the grain size is ~10nm (top image left; note that the dark lines are due to bending of the specimen) and annealing at low temperatures results in the formation of a small grained TiSi phase (not shown). The TiSi2 C49 phase appears at 700°C (top image right, with faulted structure and 100nm grain size) while C54 forms at 850°C (image right, with stacking faults, pinholes and micron-sized grains). Figure 2: Video frames showing the transformation from the C49 to the C54 phase. The numbers indicate the time elapsed in seconds since the first frame. (left) The phase transition in a 10nm UHV deposited film, recorded during heating at 825°C. The reaction front moves from left to right smoothly with no pinning. The bright feature is a pinhole. (right) Phase transition at the same temperature in a 10nm ex situ sputtered film containing ~ 30% oxygen as measured by Auger electron spectroscopy. A single C54 grain is in a strong diffracting condition and appears dark. The C49 phase has a similar grain structure to that shown in the first sequence although the grain boundaries are less visible due to a greater substrate thickness. The arrow marks a reference point on the specimen. The reaction occurs in a jerky fashion with pinning events at the grain boundaries (frames 1-3) followed by rapid jumps across several grains (frames 4 and 5). CoSi2: the formation of islands during reactive epitaxy on Si(111) Unlike TiSi2, CoSi2 has a good lattice match with Si and grows epitaxially on Si. If Co is deposited at high temperature, CoSi2 forms directly without any intermediate phases. However, as we saw with Ge on Si, strain due to lattice mismatch (0.9% at 900°C) leads to the formation of small islands rather than a flat film. We observed the growth of these islands in real time by evaporating Co onto Si at high temperature in situ. Nucleation and growth were visible, but unlike the case of Ge on Si, coarsening was not significant during growth. It is difficult for islands to grow freely on the surface since, due to the twin relationship between the silicide and Si lattices, growth down a surface step necessitates formation of a partial dislocation. By analogy with Ge on Si, as islands grow above a critical diameter, usually about 500nm, dislocations form at the CoSi2/Si interface. Real time observations show that these dislocations are introduced at the edges and move towards the center, in a manner similar to that observed in Ge islands on Si(111). It is possible to calculate dislocation configurations using a program which accounts for interactions between pairs of dislocations and between dislocations and surfaces. We can predict the structures seen very well, suggesting interesting possibilities for the creation of nanoscale patterns. Figure 3: Dislocation configurations in CoSi2 islands on Si(111). The images were taken in a 220 dark field condition for the Si substrate, showing dislocations as lines of strong bright or dark contrast. (top) A partly relaxed island, recorded at the growth temperature of 850°C. (middle) After holding the specimen at this temperature for 30 minutes the same island shows a more symmetrical and more developed dislocation array, although the center remains unrelaxed. (bottom) A smaller island, this time fully dislocated, recorded after cooling the specimen. Each image is compared with a calculation, using the Paranoid algorithm, of the expected dislocation configuration.