Specular reflection measurements offer the possibility of providing useful information on both silicon and sapphire surfaces. The infrared surface analysis of sapphire (NBS Spec. Publs. 400-19, pp. 22-24, and 400-25, pp. 20-23) [54] is a useful means of detecting lattice-related damage, but is much less sensitive to surface scratch lines which may remain after the associated lattice damage is removed by substrate annealing. This is not surprising since the dimensions of these imperfections are much smaller than the infrared wavelength used, and scattering losses are relatively small. Thus, reflectance measurements at ultraviolet wavelengths comparable in magnitude with the dimensions of surface defects should be more effective in detecting surface irregularities. Similar considerations apply to epitaxial silicon films which, in turn, exhibit a surface texture related to the substrate surface condition and deposition parameters. Epitaxial silicon films are generally accepted or rejected on the product line in a subjective manner by visual inspection for light scattering effects or "haze." Films which exhibit haze appear to have inferior electrical properties; a fast, nondestructive, quantitative test is needed to place the screening on an objective basis. flected component can be ignored and eq (7) becomes Initial experiments were performed on four bulk silicon substrates which had been variously polished in order to introduce different degrees of surface roughness. The reflectance for each surface was measured as a function of wavelength and the ratio R/R at 240 and 280 nm was determined for each specimen; an aluminum mirror was used to determine Ro. The logarithm of this ratio is plotted against (on a logarithmic scale) the size of the polishing grit in figure 15. The values of σ to be used in eq (8) are not known but are related to the size of the polishing grit. Scanning electron micrographs show that surface roughness is related to the polishing grit size. It appears from these preliminary (7) 0.10 0 0.1 where R is the reflectance of a nonideal surface, R is the reflectance of an ideal surface (without roughness), A is the wavelength of the incident radiation, and ▲ is an instrumental acceptance angle. The second term takes into account diffusely reflected light. If the reflectance measurement is made at wavelengths much longer than the root mean square surface roughness, the diffusely re T Figure 15. Specular reflection data for four bulk silicon substrates polished with different polishing grits and for six silicon-onsapphire composites. (a: bulk silicon at 240 nm; : bulk silicon at 280 nm; o: siliconon-sapphire at 280 nm.) a. Acceptable quality. Magnification: ~600x. b. Rejected specimen. Magnification: ~600x. Figure 16. Scanning electron micrographs of silicon-on-sapphire surfaces. results that this plot may be used as a reference curve for determining silicon surface roughness on this scale and at these wavelengths. On this basis, reflection data were obtained at 280 nm on six silicon-on-sapphire composites. These results are also plotted in figure 15; in this case the point for each specimen is placed on the 280-nm line at the measured value of reflectance. The horizontal position, then, should give some indication of the surface roughness. Scanning electron micrographs of two typical specimens are presented in figure 16. Specimen 1, which is equivalent to a very small grit size, had an acceptable surface smoothness which compares favorably with a highly polished bulk silicon surface, while specimen 4, which is equivalent to a significantly larger grit size, had been rejected from the product line because of haze. (M. T. Duffy*. P. J. Zanzucchi*, and G. W. Cullen*) 4.2. Impurities in Sapphire Substrates Besides surface roughness (see sec. 4.1.), the presence of impurities introduced unintentionally in the epitaxial silicon film can also affect the performance of silicon-onsapphire devices. One potential source of these undesired impurities is the polished sapphire substrate. In order to determine the typical impurity levels in commercially available polished sapphire substrates, three substrates from different sources were char acterized by neutron activation analysis.+ Two specimens cut from half of each of the circular polished sapphire slices, one near the center and one near the edge, were weighed, sealed in polyethylene, and irradiated in a neutron flux of 6 x 1013 cm-2.s-1 for 30 min in the NBS reactor. After a cooling period of about 30 min, the specimens were cleaned with a 50-percent hot nitric acid solution to remove possible external contamination. The specimens were then counted using a Ge (Li) detector-multichannel analyzer system. Calcium was determined 30 min after irradiation and manganese was determined 120 min after irradiation. Standards were aqueous solutions of the two elements. Subsequently, the specimens were again irradiated, this time for 10 h. After a cooling period of about 10 days to eliminate 24 Na matrix activity, the specimens were counted to determine the amounts of molybdenum, chromium, antimony, scandium, iron, cobalt, and zinc which were present. This counting was also Work performed at RCA Laboratories under NBS Contract No. 5-35915. NBS contact for additional technical information: K. F. Galloway. The substrates were supplied by Rockwell International Corporation which had obtained them from various sources for evaluation under Air Force Contract F19628-75-C-0108. 4.3. Electron Spectroscopy Techniques Effect of Ion Flux Distribution on Depth Resolution in Auger/Ion Beam Profiling of OxideSilicon Structures As a result of an electrical breakdown and attempts to repair the ion gun in the Auger spectrometer, the depth resolution was seriously deteriorated. This was manifested by an apparent widening of the oxide-silicon interface in a previously measured specimen, and also by a large scattering in the data. In order to clarify the source of the deterioration, the flux distribution in the primary ion beam was measured for typical operating conditions. Consider the experimental conditions outlined in figure 17. The specimen surface, S, is positioned at the focal point, F, of the cylindrical mirror analyzer, CMA, with its normal at an angle 30 deg with respect to the primary electron beam, EB. The ion beam, IB, is incident on the surface at an angle ʼn and is carefully aligned so that it strikes the specimen at the focal point, as depicted in the figure. The ion current (flux) distribution at the specimen surface is also shown in the figure; the electron beam is considered infinitesimally narrow since it is much smaller than the ion beam. During the alignment procedure there is a finite possibility of displacing the specimen along the z (CMA) axis. The broken lines and curves show how the ion and electron beams are displaced. Notice that in this case the electron beam strikes the wall of the ion sputtered crater in the specimen. A similar situation arises when the ion beam is not perfectly aligned and hits off the focal point of the CMA, or if a shift occurs in the ion beam alignment during sputtering. The latter effect may result from thermal drift in the ion optics, particularly when the total sputtering time is long (hours). It can be shown that the fractional change in apparent interface width is equal to the fractional change in ion flux density provided that fluctuations in sputtering yield and specimen density can be neglected [57]. To demonstrate the importance of reducing flux inhomogeneities to a minimum, a 100-nm thick oxide thermally grown on a (100) silicon surface was profiled using two ion guns, NBS Activation Analysis Section, Analytical Chemistry Division. Imm MATERIALS CHARACTERIZATION BY PHYSICAL ANALYSIS METHODS S z (CMA AXIS) EB Figure 17. Schematic of experimental conditions for ion sputter etching and Auger electron spectroscopy. (Note that the angles e and n are not in the same plane.) one normal and one with a defect in the optics. With the former the observed width of the oxide-silicon interface was 4.3±0.3 nm. With the defective gun there was a large scatter in the observed width but the average of the minimum width was about 6 nm. The spatial distribution of the ion beam current in the two guns was measured by moving a square Faraday cup 250 μm on a side, along These results show that in Auger/ion depth Figure 18. Spatial distributions of ion beam current. (⚫: stationary beam directed at CMA focal point (z=0); ▲: rastered beam with ion current at z=0 reduced to one-half of its stationary value.) total depth milled. that this is not the niques are used. MATERIALS CHARACTERIZATION BY PHYSICAL ANALYSIS METHODS The present work shows case if proper tech- W. E. Spicer, and Y. E. Strausser*) Interface Broadening by Electron Stimulated Desorption KLL - Another important parameter in Auger analysis and in Auger/ion depth profiling is the primary electron fluence. To demonstrate the adverse effect of a high primary electron current density on depth reslution, the interface profile of the oxide film described above was measured with a 5-μA primary electron beam. In one case the beam was held stationary; in the other case it was rastered over an area 250 um by 250 μm. The defective ion gun was used to obtain both profiles. The two 502-eV 0, profiles are shown in figure 20. The horizontal axis is based on the sputter rate associated with the ion beam and the sputtering time. The interface appears to be significantly wider when measured with the stationary electron beam. The early reduction in the measured oxygen density occurs because the sputtering rate in the area where the specimen is exposed to both ions and electrons is greatly enhanced [59] as illustrated in the inset of the figure. As a result the electron beam reaches the interface ahead of the main portion of 0.8 04 |