Atomic Structure and Properties of Epitaxial Thin-film Semiconductor Interfaces

Using high-resolution transmission electron microscopy we have studied atomic structure of interfaces between epitaxial thin films of metals, insulators or semiconductors on semiconductors. For epitaxial cobalt and nickel disilicide we find exceptionally uniform interfaces with a significant dependence of the schottky barrier height on interface structure. For epitaxial alkaline-earth fluorides, the film growth and crystalline structure can be understood from the interface structure. We also report recent results on silicon based semiconductor superlattices. In many cases the electronic properties of technologically important thin films on semiconductors are strongly influenced by atomic structure at their interfaces. Examples are schottky barriers between metals and semiconductors or field-effect transistors involving transport a t an insulator/semiconductor junction. Electron microscopy currently offers sufficient resolution (0.16 0.3 nm) to study the widest interatomic spacings in semiconductors and thus to gain valuable information concerning atomic structure at interfaces. However, when comparing this information with physical measurements e.g. electron transport, which commonly involve sampling macroscopic interface areas > l o i 4 atoms, interpretation is greatly facilitated by uniform interfaces. Most notable amongst these are epitaxial systems, which attain their uniformity through long-range order. We have studied the interface structure of several epitaxial thin-film/semiconductor systems, grown by molecular-beam epitaxy (MBE). These include metals, insulators and semiconductors on Si and other semiconductors. Our studies provide examples of the three classical types of growth. That is commensurate (coherent), in which the overlayer maintains the same lattice parameter as the substrate, incommensurate , in which the interface is totally incoherent and discommensurate , in which commensurate regions are separated by misfit dislocations (discommensurations). Other details of the ;nterface structure can be determined, such as flatness and sharp~:ess and sometimes the local atomic configuration. The results can be related to the physical properties of these interfaces. Experimental Details Films were grown by a variety of vacuum evaporation techniques. Cobalt and nickel -9 disilicides were fabricated in an ultra-high vacuum chamber with base pressure < 10 T by metal deposition on an in-situ cleaned Si surface followed by in-situ annealing. Uniform single-crystals have been grown on both (1 1 I) and (100) silicon substrates using annealing conditions described elsewhere"'. Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphyscol:1985440 C4-370 JOURNAL DE PHYSIQUE Films of alkaline-earth fluorides were deposited in a similar vacuum chamber by effusion of 0 [21 the appropriate fluoride on substrates held at temperatures between 500-700 C . GexSil-x films were grown on (100) Si substrates using molecular beam epitaxy in another UHV chamberI3]. The substrates were held at temperatures between 550-650'~. Samples were thinned for transmission electron microscopy with a combination of mechanical polishing and 4-6kV ~ r + erosion in cross-section geometry along (1 10) zone axes. In this manner, interfaces could be examined edge on for substrates in all the common orientations of semiconductor wafers and with instruments of resolution better than 3 b ; . A JEOL 200CX top-entry electron microscope with spherical aberration coefficient Cs=1.2mm at 200kV accelerating voltage was used. It was operated in axial illumination close to the Scherzer defocus, where the instrument has point-to-point resolution 2.5 A. Where necessary, image simulations were carried out to take account of multiple electron scattering, using the multislice algorithm developed by Cowley and ~ o o d i e ' ~ ' . Epitaxial Schottky Barriers There has been a great technological drive for research on silicide/silicon contacts, both for schottky barrier and ohmic contacts to silicon devices[']. This has arien because of the metallic properties of the silicides and their comparative thermal and electrical stability. The best candidates for epitaxy are the cubic cobalt and nickel disilicides. These are formed as the final phase when reacting a thin film of the metal with a silicon substrate. Typically anneal temperatures of 700-900'~ and times of 30 minutes are requiredr6'. Several groups have studied the interface structure of these silicides with high-resolution microscopy[71 ['I r91. From the point-of-view of the electronic properties of these epitaxial silicides, the investigation of "clean" growth has been a major breakthrough[']. On (1 11) Si, these silicides are,found to be a a mixture of two The two orientations are referred to as A and B: in the former the cubic silicide layer is aligned exactly with the substrate, in the latter it is rotated about (1 11) by 180' ( a "hetero-twin"). Only through insitu deposition and reaction on atomically-clean surfaces can this double-positioning be eliminated. With these methods, single crystals of CoSi2 can be grown which invariably have the B orientationr6'. For Nisi the situation is .less straightforward, but even more 2: intriguing. In this case it is possible through the use of UHV grown templates, to grow single-crystals of either A or B[". Figure 1 shows high-resolution images of each of these thin films in cross-section. Figure (1)a is from an A-type template and (1)b is from a B-type template. The exceptional uniformity and sharpness of these interfaces are clearly visible from these images. These are very similar to the interfaces found in the mixed A+B films studied by Cherns et. al.'']. -using the two models proposed for these interfaces by Cherns, one can model the interface structure in figure 1. The principle of this modelling is the measurement of the relative shift of the lattices across an interface. This is a quantitatively accurate method of studying crystal-crystal interfaces in which the registration is uniform. It has been shown["'-that these measurements can be accurate to better than 0.3 b; under axial bright-field imaging conditions with an instrument of sufficient resolution. We have made such measurements of the Nisi2 interfaces and find the shift to be consistent with the 7-fold model of Cherns, which is shown in figure 2 for both the A and B orientations. This is also consistent with the findings of Cherns et.al.['I, so that it appears that the interface structure is not associated with the considerable differences in growth and crystallography of the higher-quality UHV grown films. Figure 1: HREM images from A (a) and B (b) type Nisi2 templates on (1 11) Si.