To suggest a guiding principle for oxygen uniformity in CZ silicon single crystals, two kinds of experiments from a viewpoint of SiO evaporating from free surface of the melt were examined. First, evaporation rate of SiO from silicon melts with different aspect ratio were measured as a function of A_C/A_S(A_C: contact area between melt and silica crucible, A_S: area of free melt surface) by using thermogravimetric method. The evaporation rate of SiO increased with increasing A_C/A_S, while showing strong temperature dependence. Next, several CZ crystals with varying diameters were grown by the two characteristic growth conditions: the "speed control" and the "temperature control" respectively. These conditions could control the SiO evaporation rate quasi-quantitatively during the crystal growth by changing A_C/A_S It should be noted that axial and radial oxygen distribution in the " speed control" crystal largely varies as a function of A_C/A_S. On the contrary, oxygen was uniformly distributed in the "temperature control" crystal . It was revealed that the marked oxygen heterogeneity in the " speed control" crystal was mainly originated in the variation of SiO evaporation due to high temperature of the melt. Regarding to the influence of boron upon the oxygen distribution in CZ silicon crystals, the oxygen heterogeneity was remarkably emphasized since boron enhanced the evaporation of SiO. It was concluded that fluctuation of SiO evaporation and thus variation of oxygen concentration in silicon melt should be suppressed for oxygen uniformity in silicon single crystals.
The effects of the argon gas flow rate and furnace pressure on the oxygen concentrations in Czochralski (CZ) and transverse magnetic field applied Czochralski (TMCZ) silicon single crystals were examined through experimental crystal growth. The oxygen concentrations in the grown crystals were varied based on the changes of the furnace pressure and the argon gas flow rate . The oxygen concentrations were directly proportional in the CZ crystals and inversely proportional in the TMCZ crystals to the calculated flow velocity of argon gas. To explain the CZ results, newly proposed melt flow pattern model based on the melt flow pattern changes by the argon gas shear stress is applicable. Some numerical simulation works supported by a model experiment with alternative materials confirmed the effects of argon gas shear stress. For the TMCZ results, conventional SiO partial Pressure model or boundary layer thickness model can be applied due to the increased effective viscosity of silicon melt by the magnetic field.
This article discusses the influence of the Marangoni effect on melt flow and oxygen transport in a small silicon Czochralski furnace, based on recent results of global simulation conducted by the authors. Oxygen transfer rate is controlled by the mass-transfer resistances in melt and gas phases. The Marangoni effect exerts a significant change of flow pattern of melt near the gas/liquid interface. The Marangoni effect takes an important roll in modulating the distributions of temperature and oxygen concentration near the interface and consequently the over-all oxygen transport rate. Installing a heat shield provides large shear stress onto the melt surface from the gas flowing along.The Marangoni effect counteracts the stress and minimizes the induced eddy flow and increases oxygen evaporation rate.
This review deals with effects of oxygen partial pressures on the crystal growth of oxides containing transition elements. Rutile (TiO_2) and chromium-doped forsterite (Cr: Mg_2SiO_4) single crystals are successfully grown by the floating zone method. Rutile single crystals are grown under a low oxygen partial pressure of about 10^3 Pa to avoid the formation of low-angle grain boundaries. Zirconium-doping is effective to grow rutile single crystals without low-angle grain boundaries and bubble inclusions at a high growth rate of 10 mm/h under a high oxygen partial pressure of 10^5 Pa. Cr^<4+>-rich Cr: Mg_2SiO_4 single crystals are grown under a high oxygen partial pressure of 1-2 10^5 Pa, which can not be realized in the conventional Czochralski method using an iridium crucible.