Flow patterns have been made visible in a horizontal water‐cooled epitaxial reactor by injection of TiO2 particles into the gas flow. From these experiments, a stagnant layer model has been developed with which the epitaxial growth of silicon from silane can be described. In the case of a nontilted susceptor, the model predicts an appreciable nonuniformity in thickness along the susceptor, whereas a small angle of tilting of the susceptor should yield a much better uniformity in thickness (2% over a length of 22 cm). Experiments agree very well with the theoretical predictions of the model.

The deposition rate of polycrystalline silicon from a SiH4-H2 mixture is significantly influenced by the addition of ASH3, PH3, and B2H6. At a deposition temperature of 680~ AsH3 causes a decrease by a factor of 7, PH3 causes a decrease by a factor of 2.5, while a two times higher deposition rate is obtained with B2H6 addition. Out of these three dopant hydrides AsH~ and PH3 do not affect the activation energy of the deposition reaction compared to undoped growth (37 kcal/mole). The Arrhenius plot for the deposition of silicon from a B2H6-SiH4 mixture shows two activation energies: 20 kcal/mole at T = 620~176 and 7 kcal/mole below 620~ The experimentally found minimum values of the resistivity of doped polycrystalline silicon can be explained in terms of solid solubility and carrier mobility. At deposition temperatures below 700~ with and without addition of dopants the polycrystalline silicon surface is mirror-like. Significant differences have, however, been observed by electron microscopy. Compared to undoped growth boron was found to lower the etch rate of the polycrystalline silicon film markedly.Increasing interest is being shown in the use of polycrystalline silicon in silicon device technology. The application of polycrystalline silicon films is compatible with silicon device processing where polycrystalline silicon is mainly used as gate material in MOS structures (1) (self-aligned gate).Polycrystalline silicon can be deposited in different ways: by evaporating, by sputtering, and by chemical vapor deposition. Chemical vapor deposition is superior because it permits uniform deposition over oxide steps. To obtain a mirror-like surface, which enables very fine patterns to be etched in it, comparable with those in silicon oxide, the grain size of the polycrystalline silicon film should be as small as possible. DeLuca (2) has found that the grain size of the polycrystalline silicon film decreases with decreasing temperature. However, the deposition rate also decreases with decreasing temperature. At 650~176 an acceptable compromise between growth rate and grain size is realized. In this temperature region most polycrystalline silicon films suitable for high resolution I.C. processing are grown.In the case where polycrystalline silicon is used as the gate material for MOS structures, it may be deposited undoped and subsequently doped by impurity diffusion. In the case where design considerations prohibit high-temperature processing after polycrystalline silicon deposition, doping by codeposition becomes necessary.The present paper reports on the growth of doped polycrystalline silicon films by codeposition of silicon and either arsenic, phosphorus, and boron from the corresponding hydrides using hydrogen as a carrier gas. In particular the temperature dependence of the growth rate was studied. The effects of deposition temperature and dopant concentration on growth rate and resistivity are discussed. ExperimentalThe polycrystaUine silicon films were prepared in an uncooled vertical reactor. The substrates wer...

The extreme values of k plotted against the order of the extrema yield a straight line characterized by a slope dnormals and an intercept δ0/2π . The thickness dnormalm of the deposited layer can be measured with a method described by Eversteyn and van den Heuvel. It turns out that dnormals−dnormalm depends linearly on δ0/2π , the relations being different for N+ and N++, but further independent of substrate resistivity or thickness. After calibration for a particular growth process the thickness can be determined with dnormalm as a standard reference from a conventional infrared multiple interference spectrum.

The thickness of epitaxially grown silicon layers, less than 1 μm thick, can be measured directly with a mercury probe and an, eventually automated, capacitance bridge. The thickness measured in this way with a precision of about 400 Å, is found to be about 10% smaller than the metallurgical thickness. The submicron thickness of polycrystalline layers can be measured similarly and agrees with the metallurgical thickness within the measurement precision of about 200 Å.