Numerical Procedure to Extract Physical Properties from Raman Scattering Data in a Flow Reactor

Abstract

A parameter estimation procedure was established to extract physical parameters related to chemical vapor deposition ͑CVD͒ processes, which consists of in situ experimental data collection and computational analysis. Specifically, mass-transport behavior in an up-flow cold-wall CVD reactor was monitored using in situ Raman spectroscopy. A two-dimensional axisymmetric model of the reactor was developed and combined with genetic and simplex algorithms for property estimation. The numerical procedure was unambiguously able to extract binary mass diffusivities and Raman cross sections from the same data set for the methane/ nitrogen and ammonia/nitrogen cases. The procedure developed in this study is general and expected to be particularly useful in extracting diffusivities and cross sections for reaction intermediate species and kinetic parameters for complex reacting systems.Metallorganic chemical vapor deposition ͑MOCVD͒ is widely used to deposit thin semiconductor films with good crystalline quality and electrical properties. 1 However, MOCVD processes are influenced by several factors such as the choice of carrier gas, 2 substrate material and its surface preparation, 3 and the gas dynamics inside the reactor including the thermal boundary conditions. 4 Obtaining a realistic reactor model based on a fundamental understanding of the process would be useful, e.g., for optimizing an existing process or designing a new reactor.Accurate reaction kinetic information about homogeneous and heterogeneous chemical reactions in the reactor and the thermophysical properties of the gas and solid phases are required in developing a realistic model. Previously reported kinetic data on the decomposition of metallorganic sources have most often been obtained from simple flow cell or flow tube experiments. 5-7 In those analyses, homogeneous conditions are usually assumed to be established in the cell, which simplifies the subsequent data analysis, although conditions in a real MOCVD process are not truly homogeneous. The kinetic data inferred from flow cell or flow tube experiments could be confounded with other process-specific influences such as flow instability in the reactor and memory effects. 8 As an example, the pyrolysis of trimethylindium, first studied by Jacko and Price, 9 has since been examined in a flow cell experiment by several investigators. 10-13 The reported activation energy for the first-order decomposition reaction ranges from 35.9 to 54.0 kcal/mol, while frequency factors range from 10 12.0 to 10 17.9 s −1 . This wide range of values reflects the uncertainties associated with the determination of decomposition rate constants of metallorganic sources, even though some of the variation could be interpreted by the chemical effects of carrier gas on the decomposition reactions. Accurate transport properties of the gas mixture in an MOCVD reactor are also required for developing a realistic process model, but they are not well reported and usually have been estimated using the Chapman-Enskog equation. 14,15 A we...

Homogeneous Decomposition Mechanisms of Diethylzinc by Raman Spectroscopy and Quantum Chemical Calculations

“…This figure also shows the temperature profile calculated using a detailed 2-D asymmetrical reactor model. This model is described in greater detail elsewhere, including its validation. , Briefly, the simulation uses the finite element method (FEM) to simultaneously solve the steady-state momentum, heat, and mass balance equations in a reacting system. It is noted that a ∼160 °C difference exists between the set-point temperature and the gas-phase temperature extrapolated to the heater surface.…”

“…Atomic Zn was not expected to be detected by Raman spectroscopy because the most probable emission radiates from the excited state ( 1 P 1 o ) to the ground state ( 1 S 0 ) with just 1.6 ns lifetime, which is too short to be detected. 25 Therefore, Raman scattering experiments were used to search for the possible decomposition fragments • ZnC 2 H 5 , HZnC 2 H 5 , and ZnH 2 .…”

“…It is noted that the high temperature reduced the signalto-noise level, and transverse and axial diffusion diluted the precursor as well as reaction products. The existence of a validated reactor model 25 and rate constants from the computational studies allowed comparison of the experimental profiles to those predicted by the reaction simulation.…”

Model and Parameter Uncertainty in Distributed Systems

“…7 Recently, problem inversion for reaction mechanisms in metallorganic vapor decomposition processes microelectronic applications was studied. 8 Prior research performed data assimilation for air pollution models. [9][10][11] Inverse problems have been used to prevent excessive aerodynamic heating of space vehicles during re-entry into the earth's atmosphere.…”

Discovery of transport and reaction properties in distributed systems

“…Recently, problem inversion for reaction mechanisms in metallorganic vapor decomposition processes microelectronic applications was studied. 1 Prior research performed data assimilation for air pollution and weather models. [2][3][4][5] Inverse problems aim at preventing excessive aerodynamic heating of space vehicles during reentry into the earth's atmosphere.…”