Extraction of Kinetic Parameters for the Chemical Vapor Deposition of Polycrystalline Silicon at Medium and Low Pressures

The deposition of silicon (St) from silane (Sill4) was studied in the silane pressure range from 0.5 to 100 Pa (0.005 to 1 mbar) and total pressure range from 10 to 1000 Pa using N2 or He as carrier gases. The two reaction paths, namely, heterogeneous and homogeneous decomposition could be separated by varying the amount of wafer area per unit volume (wafer-distance variation) and the Sill4 partial pressure as well as the total pressure. Rate constants were derived by fitting the experimental results. The heterogeneous reaction path could be described by only the adsorption rate constants of reactive species and the desorption rate constant of hydrogen using a Langmuir-Hinshelwood mechanism. Hydrogen and phosphine were found to suppress the deposition rate at low silane pressures. At high silane pressures or high total pressures the unimolecular decomposition of silane dominates. The unimolecular rate constant was found to be one to two orders larger than literature values based on RRKM analyses of high pressure rate data. The relative efficiency of SiH4-N2 and SiH4-He collisions compared with SiH4-SiH4 collisions in the unimolecular gas-phase decomposition of Sill4 has been investigated. Helium was found to be a weak eollider compared to silane and nitrogen. Polycrystalline silicon plays an important role in the semiconductor industry, namely, in integrated circuits (IC), thin film transistors (TFT), solar cells, and sensors. An extensive review of many scientific aspects and applications of polysilicon is given in Ref. 1. In general polysilicon is deposited in hot-wall tube reactors where loads of 50 to 100 wafers are processed in one run. The depositions are performed using 100% Sill4 or 23 % Sill4 diluted in an inert gas at Sill4 partial pressures of between 10 and 30 Pa (0.1 to 0.3 mbar) at 625~ m Under these conditions the growth rate is dominated by a Langmuir-Hinshelwood type of heterogeneous Sill4 decomposition with H2 release from Si-H bonds as the rate determining step. 4'~ The unimoleeular gas-phase decomposition becomes important at higher Sill4 or higher total pressures. The homogeneous path is, in general, less desirable, because large polymers may be formed that coalesce to particles. 6 Thickness inhomogeneity at the wafer edge, due to local variation of the area per unit volume, is another problem related to highly reactive product formation in gas-phase reactions. 6-9 However, in the case of in situ phosphorus or arsenic doping, the surface sites become blocked by PH3or AsH3related surface species, and the growth rate is strongly reduced.7'10'" In order to achieve reasonable growth rates the homogeneous path has then to be promoted thereby creating molecules and radicals that are more reactive than Sill4 and are able to compete with PH:~ or AsH3 for surface sites. The inhomogeneity at the wafer edge related to the gasphase path and the discontinuity of area per unit volume at the wafer edge, the so called "bull's eye" effect, is caused by the high reactivity of the gas-phase products and can be solved by adjusting the quartz carrier design or by using cages. ~2'~:~ In this way more area becomes available at the wafer edge to scavenge excess reactive gas-phase products. The gas-phase decomposition regime may also become important in single-wafer reactors where high throughputs are desired. If the total pressure is allowed to rise in singlewafer systems, a 2 % nonpyrophoric dilution of SiH~ related to its spontaneous flammability.~4 The relative contributions of the different paths have been modeled in the past for both batch and single wafer reactors (see Ref. 15-20, and many other earlier references cited therein). In this paper we report on a study of the decomposition of Sill4 at low and medium pressures as well as the influence of the carrier and dopant gases. Special attention is paid to the unimolecular gas-phase decomposition of Sill4 into Sill2, for which, in general, values were used, which up to now were derived by means of RRKM analyses 21-2~ of experiments 24-2~ at much higherpressure and/or lower temperature than generally applied in low pressure chemical vapor deposition (LPCVD) of polysilicon. In this paper values of the unimolecular rate constant and of the collision efficiency of the carrier gases He and N2 are derived by making use of the effect that, due to the high reactivity of gas-phase reaction products, the growth rate becomes dependent on wafer spacing. We used this phenomenon for the separation of homogeneous and heterogeneous contributions to the total growth rate. Experimental Deposition.-The depositions were carried out on oxidized 3 in. wafers in a three-zone hot-wall horizontal tube reactor with 12 cm inner diameter. The length of the heating element is 90 cm with a flat temperature zone of 30 cm. The quartz wafer-carrier with slots at 5 mm spacing was positioned radially symmetrically in the reactor. The first and last five slot positions are always loaded with 3 in. wafers. In the experiments we varied the wafer spacing, gas flow rate, gas composition, and pressure. For the experiments presented in Fig. 2, 3, 4, 9, 10, and 11, the wafer spacing was varied alternatingly at 5 and 15 mm with the same run. The experiments were performed in the temperature range from 525 to 700~ this paper deals mainly with the experiments at 625~ The temperature at the wafer-carrier position was calibrated by in situ thermocouple measurements and by using solid phase epitaxy of an amorphous layer, created by implantation of Si in monocrystalline silicon wafers, as a temperature monitor. 27 All experiments were performed with a so-called flat temperature profile at the carrier position. The wafer temperature was found to be independent of the pressure. With experiments at 500 Pa (5 mbar) we found that the temperature at the wafer position was no longer constant when the flow was larger than 1 slm. Therefore, the total flow was kept below i slm in the 500 and the 1000 Pa experiments. All gases except Sill4 had contamination levels below 1 ppm. The purity of Sill4 was 99.995%. J. Electrochem. Soc., Vol. 140, No. 7, July 1993 9 The Electrochemical Society, Inc. 2089 Downloaded 23 Jun 2009 to 130.89.112.86. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp 2090 J. Electrochem. Soc., Vol. 140, No. 7, July 1993 9 The Electrochemical Society, Inc. Growth-rate measurements.--Growth rates were measured by weighing and by measuring the height of an etched step. The weighing method yields global growth rates whereas step-height measurements give information about ~'~ the local growth rate and uniformity within the wafer. As a ~ standard procedure the growth rates were measured at different positions in the wafer carrier, with various flow rates E and various interwafer spacings. In a horizontal hot-wall reactor, where Si wafers are placed in series with respect to the gas flow, the growth rate 9 +~ will decrease in the axial direction due to depletion of the (~ reactant. The undepleted growth rate, ro for high P~clet ~values can be determined by plotting the growth rate, rx, at _E position x against the ratio A J M F according to Eq. 1 28 r = ro[1 v{1 + (e 1)(psi~4o/Ptot)}AxroMF -1] [1] O~_ 9 where Ax is the total hot surface area (wafers, wall, cantilever, carrier) upstream the position where r is measured, MF is the mass flow of SiH~, v is the overall order of the reaction, PSIH4o is the inlet partial pressure of Sill4, Ptot is the total pressure, and 9 is the number of gaseous molecules formed by the decomposition of one molecule of reactant. Other authors ignored the hot surface area upstream the flat zone. 29'3~ Although this area does not play a significant role in low depletion cases it cannot be ignored in most of the practical cases, and it is, in our opinion, the source of the lower activation energy of about 150 kJ mol 1 generally reported for LPCVD hot-wall silicon-deposition systems. 2'~6'3~-33 We found an apparent activation energy of 180 kJ 9 mo1-1 in the temperature range 550-650~ which is in agreement with values reported for single-wafer systems and molecular-beam reactors and in hot-wall reactors under conditions of low depletion. 5'3~ We determined the effective wall area upstream the flat zone by measuring the growth rate on silicon wafers placed close to the wall. The effective surface area of the entrance region was found to be 1200 cm 2. The growth rates were also measured by varying the wafer spacing within the same run. When a surface reaction is the rate-determining step the growth rate is independent of the wafer spacing. However, when the gasphase reaction leading to highly reactive components becomes important the growth rate will become a function of the density of surface area per unit volume. Gas-phase reactions have been reported for the deposition of silicon from Sill4. ~,24,34,3~,39-4~. These gas-phase reactions with Sill2, Si2H6, Si3Hs, and Si2H4 as major constituents contribute to "~ the total deposition rate at higher pressures. ~ When gas-phase reactions become important, extrapolation according to Eq. 1 is no longer valid because the loss in the entrance region is now no longer determined by the effective wall area alone but also by the effective entrance volume. It is more convenient then to measure the loss in the entrance region for each run. The loss can be calculated +~ by measuring the growth rate on silicon wafers placed horizontally close to the wall. The partial pressure of Sill4 at the first wafer position can be calculated by equating the total flux, i.e., convective and diffusion flux, at the first wafer position to the inlet flux minus the measured loss in the entrance region. (_~ Results and Discussion Growth rate.--The results of growth-rate measurements at various positions, flow rates, and wafer spacings are presented in Fig. 1 through 4. In Fig. 1 the wafer spacing was kept constant at 10 mm and the total pressure remained at 133 Pa using N2 as a carrier gas. The undepleted growth rates, ro, were obtained by extrapolation to the zero axis. In Fig. 2, 3, and 4 the wafer spacing was varied in order to separate surface and gas-phase decomposition of Sill4. The undepleted total growth rates for 10 mm wafer spacing are presented in Fig. 5, along with results with the addition of PH3. Also included in Fig. 5 are some data points (open circular markers) obtained previously in another reactor with N2 and H2 as carrier gases. 42 In Fig. 6 the uniformity 250