Reactor-scale Simulations Research Articles

How Predictive Are Reactor-Scale Simulations of Wbg Semiconductor Epitaxy? a Sensibility Analysis of SiC and Nitride Epitaxy Models

Process modeling can be the perfect complement to process development: in addition to offering a cost-effective way of exploring the parameter space of a given process or new reactor configurations, models are able to provide information regarding details of surface and gas phase chemistry that is not readily available experimentally. This information can be extremely useful in helping us rationalize trends in yield, epitaxial quality and defect concentration, or why a particular set of experimental conditions leads to the highest quality materials. In the case of physical vapor and chemical vapor epitaxy, the fundamental aspects of reactor-scale simulations are well established. However, it is our knowledge of homogeneous and heterogeneous processes what limits the predictive nature of our models. Every reactor rate and transport coefficient comes with an associated error that propagates into the model solution. In addition to this, the complexity of the overall kinetics of the growth process complicates our ability to identify rate limiting processes exclusively based on experimental trends of growth rates as a function of experimental parameters. Consequently, knowing how sensitive our model is to the different assumptions will help us both better understand the margin of error of model predictions and identify the key steps in the growth process. In this talk I will summarize the sensitivity analysis studies that we have carried out on the heterogeneous and homogeneous processes during SiC, GaN, and AlN epitaxy. Using datasets compiled from different models available in the literature, we will compare the predicted results in both 0.5D and simplified 1D and 2D models that will allow both local and non-local sensitivity studies. The advantage of 0.5D models is that they can be used as a first order approximation to simulate epitaxy on planetary systems where wafer to wafer variability is lower than in-wafer variability. We will use these models to identify both the key steps and the main sources of error as a function of the experimental conditions. Finally, we will study the impact that reduction on complexity has on model predictions.

Improving atomic layer deposition process through reactor scale simulation

Abstract In order to modify atomic layer deposition (ALD) characteristics of Al2O3, three-dimensional gas transports and film depositions are investigated through reactor scale simulations inside two different viscous flow reactors. In the top-inlet reactor (TIR), the gaseous species are directly injected into the substrates from the upper surface of the reactor while in the bottom-inlet reactor (BIR), the inlet is at the bottom of the reactor and next to the substrate. The numerical procedure to simulate the ALD process is thoroughly explained by using the multi-species and multi-reaction chemistry phenomena. The reactants are trimethylaluminum (TMA) and ozone, and the simulations are performed in an operating pressure of 10 Torr (1330 Pa) and two substrate temperatures of 250 °C and 300 °C. Due to the chemistry mechanism used in this study, a long ozone exposure is a crucial parameter to deliver a sufficiently oxidized substrate. For a specific reactor type, deposition rates are higher on the hotter substrate due to both a larger surface reaction rate constant and greater concentrations of the oxygen atoms on the substrate. At a fixed substrate temperature, higher deposition rates are obtained by using the TIR. The same deposition rate distributions are obtained among all cycles for each ALD process that result in the dependency of the film thickness only on the numbers of ALD cycles. For the substrate at 250 °C, the growth rates are equal to 3.78 A/cycle and 4.43 A/cycle in the BIR and the TIR, respectively, and for the substrate at 300 °C, the growth rates are equal to 4.52 A/cycle and 6.49 A/cycle in the BIR and the TIR, respectively.

Thermodynamics and Kinetics of SiC CVD Epitaxy

The ultimate goal of process modeling in the context of thin film growth is not only to predict the growth rate at a reactor scale, but also to provide details of other aspects that are crucial in process development, such as the microstructure and the epitaxial quality of the films. With the fundamental aspects of reactor-scale CFD simulations already well established, chemistry becomes the key issue that needs to be addressed in order to achieve predictive reactor-scale simulations. The problem is not just the complexity of SiC CVD epitaxy in terms of the number of reactions and/or species, but the fact that the kinetics of some of these processes, especially those involving surface reactions, is not known beyond the most common precursor combinations. This lack of knowledge limits the predicting abilities of simulations in two ways: 1) they introduce an uncertainty in the model results that needs to be accounted for, especially when unknown coefficients are determined by fits to experimental data, and 2) they prevent us from understanding the impact of chemistry on film microstructure and epitaxial quality, where heterogeneous processes play a central role.In this talk, I will present our research on non-halogenated and halogenated chemistries for SiC CVD. As part of a greater effort aimed at developing multiscale simulations of ALD and CVD for predictive scale up,[1] we are carrying out both thermodynamic and kinetic analyses of the gas phase and surface kinetics in SiC CVD epitaxy in order to identify the main sources of uncertainty. In addition to using standard sets of reactions previously compiled in the literature for SiC CVD,[2,3] we are exploring the combustion chemistry, etching, and Si CVD and PECVD literature to understand the consistency and uniqueness of the different rate coefficients, and its impact in the main predicted species in the gas phase. The goal of this work is to quantify the level of uncertainty and ultimately to improve the predictive abilities of our models as our knowledge progresses, using for instance Bayesian analysis. Likewise, we seek to identify gaps in the experimental data that will help us drive experimental research towards those reactions/pathways that have a greater weight on the synthesis processs. Finally, we will address the impact of surface chemistry on film microstructure. Besides reviewing the underlying assumptions of previous models in the literature,[2,3,4] we will explore the impact of kinetics on the roughness and the stability of step-flow growth. These could be used to establish limits on unknown heterogeneous kinetic processes based on the experimental characterization of the film microstructure and the stability of the step-flow growth as a function of the miscut angle.[1] Yanguas-Gil and Elam, Simple ALD Multiscale Simulation (SALMS), ANL-SF-13-073, Argonne National Laboratory (2013).[2] Danielsson, Henry and Janzen, J. Crystal Growth 243, 170 (2002).[3] Nichizawa and Pons, Chem. Vap. Deposition 12, 516 (2006).[4] Camarda et al, J. Crystal Growth 310, 971 (2008).

Multiscale Simulations of ALD in Cross Flow Reactors

Two of the biggest challenges in reactor-scale simulations of ALD processes are our lack of knowledge on the mechanistic details of surface chemistry and the need to integrate reactor scale and feature scales in the simulations. In order to assist our process development and scale up we have developed a multiscale simulation code that allows us to study the impact of surface chemistry on the coating of large area substrates with high surface area/high aspect-ratio features. Our code, based on open-source libraries, takes advantage of the ALD surface chemistry to achieve an extremely efficient two-way coupling between reactor and feature length scales, and it can provide simulated quartz crystal microbalance and mass spectrometry data at any point of the reactor. By combining experimental surface characterization with simple analysis of growth profiles in a tubular cross flow reactor, we are able to extract a minimal set of reactions to effectively model the surface chemistry, including the presence of spurious CVD. These reactions are then used as an input for the large-scale 3D simulations. In this work, we will focus on two aspects of the problem: the impact of non-ideal surface chemistry and the role of reactor geometry in the coating of both planar and high surface area substrates. Finally, we will discuss the impact of moving substrates characteristic of roll-to-roll processes.

BIOTC: An open-source CFD code for simulating biomass fast pyrolysis

The BIOTC code is a computer program that combines a multi-fluid model for multiphase hydrodynamics and global chemical kinetics for chemical reactions to simulate fast pyrolysis of biomass at reactor scale. The object-oriented characteristic of BIOTC makes it easy for researchers to insert their own sub-models, while the user-friendly interface provides users a friendly environment as in commercial software. A laboratory-scale bubbling fluidized bed reactor for biomass fast pyrolysis was simulated using BIOTC to demonstrate its capability. Program summaryProgram title: BIOTC-2.1.xCatalogue identifier: AESJ_v1_0Program summary URL:http://cpc.cs.qub.ac.uk/summaries/AESJ_v1_0.htmlProgram obtainable from: CPC Program Library, Queen’s University, Belfast, N. IrelandLicensing provisions: Standard CPC licence, http://cpc.cs.qub.ac.uk/licence/licence.htmlNo. of lines in distributed program, including test data, etc.: 82295No. of bytes in distributed program, including test data, etc.: 266197Distribution format: tar.gzProgramming language: C++.Computer: All capable of running Linux.Operating system: Linux.Has the code been vectorized or parallelized?: Parallelized with MPI.Classification: 22.External routines: MPI, OpenFOAM (http://www.openfoam.org)Nature of problem:Computational fluid dynamics (CFD) simulation of biomass fast pyrolysis at reactor scale can help reveal the details of the process and develop an understanding of the underlying mechanisms for reactor operation and optimization. However, the existing CFD codes, commercial or open-source, still pose difficulties for users to carry out an efficient simulation. Therefore, an open-source CFD code that integrates the merits of mainstream commercial and open-source codes, i.e., user-friendly interface and convenient sub-model implementation, is by all means necessary for reactor-scale simulation of biomass fast pyrolysis.Solution method:A multi-fluid model (MFM) is used to solve the multiphase fluid dynamics, while a global reaction mechanism is employed to solve the chemical reactions in the fluidized-bed reactors. Partial differential equation solvers and ordinary differential equation solvers provided by OpenFOAM are used to solve the MFM conservation equations and the chemical reaction equations, respectively. The coupling of MFM and chemical reactions is realized by a time-split numerical scheme.Running time:It depends on the dimension of the reactor and the complexity of the in-reactor process. Typically, for the tutorial case provided, the run time ranges from several hours to tens of hours.The small test case provided takes approximately 30 min on a serial machine.

The effective convectivity model for simulation of melt pool heat transfer in a light water reactor pressure vessel lower head. Part II: Model assessment and application

The paper reports detailed assessments and representative application of the effective convectivity model (ECM) developed and described in the companion paper ( Tran and Dinh, submitted for publication). The ECM capability to accurately predict energy splitting and heat flux profiles in volumetrically heated liquid pools of different geometries over a range of conditions related to accident progression is examined and benchmarked against both experimental data and CFD results. Augmented with models for phase changes in binary mixture, the resulting PECM (phase-change ECM) is validated against a non-eutectic heat transfer experiment. The PECM tool is then applied to predict thermal loads imposed on the reactor vessel wall and Control Rod Guide Tubes (CRGTs) during core debris heatup and melting in the BWR lower plenum. The reactor-scale simulations demonstrate the PECM's high computational performance, particularly needed to analyze processes during long transients of severe accidents. The analysis provides additional arguments to support an outstanding potential of using the CRGT cooling as a severe accident management measure to delay the vessel failure and increase the likelihood of in-vessel core melt retention in the BWR.

Reactor-scale profile of group-V composition of InGaAsP studied by fluid dynamics simulation and in situ analysis of surface kinetics

Reactor-scale simulation of metalorganic vapor phase epitaxy (MOVPE) by computational fluid dynamics is now widely attempted, but it is still difficult to predict group V composition precisely for InGaAsP. In this study, desorption of As and P from the surface of GaAs, InAs, GaP, InP, InGaAs and InGaP was measured around growth temperatures using in situ reflectance anisotropy (RA) spectroscopy. Time transient curves of RA were measured when group V supply was turned on/off and the rate constants of the desorption of As and P were measured assuming first-order kinetics. These data were incorporated into the simulation to obtain the composition of InGaAsP. The simulated distribution of group V composition showed better agreement with experimental data than the results using previous kinetic data.

Transient ALD simulations for a multi-wafer reactor with trenched wafers

For a large multi-wafer vertical batch ALD reactor, transient, 3-dimensional, multi-scale simulations have been performed for the TEMAH pulse step during a HfO 2 ALD process. A bi-directional, multi-scale coupling has been established between continuum reactor scale simulations and molecular trench scale simulations. For describing the molecular deposition process inside narrow deep trenches, two different models have been developed and implemented as boundary condition for the flow simulation: an analytical model based on the model proposed by Gordon, and a numerical Knudsen diffusion model. Both trench models have been validated against DSMC simulation results for a single trench. Based on the 3D reactor simulation results, important timescales in the complete process are identified. It is found that several timescales are of the same order of magnitude, rendering predictions based on engineering rules more difficult. Timescales for trenched wafers are found to be much larger than for flat wafers, leading to much longer required cycle times.

Modeling of Silicon Nanodots Nucleation and Growth Deposited by LPCVD on SiO2 : From Molecule/surface Interactions to Reactor Scale Simulations

ABSTRACTWe present first results combining models at continuum and atomistic (DFT) levels to improve understanding of key mechanisms involved in silicon nanodots (NDs) synthesis on SiO2 by Low Pressure Chemical Vapor Deposition (LPCVD) from silane SiH4. In particular, by simulating an industrial LPCVD reactor using the CFD code Fluent, we find that the deposition time could be increased and then the reproducibility and uniformity of NDs deposition could be improved when highly diluting silane in a carrier gas. A consequence of this high dilution seems to be that the contribution to deposition of unsaturated species such as silylene SiH2 highly increases. This result is important since our first DFT calculations have shown that silicon chemisorption on silanol Si-OH or siloxane Si-O-Si bonds present on SiO2 substrates could only proceed from silylene (and probably from other unsaturated species). The silane saturated molecule could only contribute to NDs growth, i.e. silicon chemisorption on already deposited silicon bonds. Increasing silylene contribution to deposition in highly diluting silane could then also exalt silicon nucleation on SiO2 substrates and then increase NDs density.

A computational framework for modelling grain-structure evolution in three dimensions

We describe a simulation framework designed to track individual grains in a material during simulations of formation, processing and usage. The framework, which we call parallel level-set environment for nanoscale topography evolution, is designed to fill the clear and present need to account for grain structure in understanding and predicting the performance of structures in some products, such as metal lines in integrated circuits. It is the realization of ‘grain continuum’ models of films by Cale et al. and can be used to complement discrete atomistic simulations and to link them to continuum simulations. We demonstrate the use of multiple level-set methods to track islands nucleated on substrates, during growth and impingement to form a polycrystalline film. Individual grains in the film are defined in the finite-element data structure. We briefly discuss how this simulation tool might be used in an integrated multiscale process simulation environment previously described by Bloomfield et al. to establish a link from atomistic simulations upwards to feature-, pattern- and reactor-scale simulations.

Gas phase and surface kinetic processes in polycrystalline silicon hot-wire chemical vapor deposition

Experiments and numerical simulations have been conducted to determine critical parameters for growth of polycrystalline silicon via hot-wire chemical vapor deposition. Reactor-scale simulations performed using the Direct Simulation Monte Carlo (DSMC) method have revealed a number of important phenomena such as a sharp drop of 1700 K in the gas temperature from the wire to substrate. The gas-phase reaction of silicon atoms produced on the wire with ambient silane molecules has been studied using ab initio quantum chemical calculations. Results reveal that collisional stabilization of the adduct (H3SiSiH) is unlikely under typical growth pressures, but an energetically favorable, low-pressure pathway has been found that leads to the formation of Si2H2 and H2. Threshold ionization mass spectrometry measurements of radicals have revealed that at the pressure characteristic of growth (2-200 mTorr of 1% SiH4 in He), the radical SiH2 is predominant. Finally, film growth studies reveal that hot-wire-produced atomic hydrogen may preferentially etch amorphous silicon and suppress the formation of small nuclei.

Mechanistic feature-scale profile simulation of SiO2 low-pressure chemical vapor deposition by tetraethoxysilane pyrolysis

Simulation of chemical vapor deposition in submicron features typical of semiconductor devices has been facilitated by extending the EVOLVE [T. S. Cale, T. H. Gandy, and G. B. Raupp, J. Vac. Sci. Technol. A 9, 524 (1991)] thin film etch and deposition simulation code to use thermal reaction mechanisms expressed in the Chemkin format. This allows consistent coupling between EVOLVE and reactor simulation codes that use Chemkin. In an application of a reactor-scale simulation code providing surface fluxes to a feature-scale simulation code, a proposed reaction mechanism for tetraethoxysilane [Si(OC2H5)4] pyrolysis to deposit SiO2, which had been applied successfully to reactor-scale simulation, does not correctly predict the low step coverage over trenches observed under short reactor residence time conditions. One apparent discrepancy between the mechanism and profile-evolution observations is a reduced degree of sensitivity of the deposition rate to the presence of reaction products, i.e., the by-product inhibition effect is underpredicted. The cause of the proposed mechanism’s insensitivity to by-product inhibition is investigated with the combined reactor and topography simulators. This is done first by manipulating the surface-to-volume ratio of a simulated reactor and second by adjusting parameters in the proposed mechanism such as the calculated free energies of proposed surface species. The conclusion is that simply calibrating mechanism parameters to enhance the by-product inhibition can improve the fit to profile evolution data; however, the agreement between with reactor-scale data and simulations decreases. Additional surface reaction channels seem to be required to simultaneously reproduce experimental reactor-scale growth rates and feature-scale step coverages.

Benefits and Limitations of Radiatively Heated Suszeptors

ABSTRACTIn radiatively heated single wafer reactors, suszeptors are used to control the wafer temperature. This temperature, however, might be significantly different than the suszeptor temperature. The reason is that the conductive coupling of the suszeptor to the wafer is often small, even when the wafer is placed directly on the suszeptor, and the temperature difference is then determined by the optical properties and the spectral distribution of the illumination.In this paper these temperature differences are calculated with reactor scale simulation for typical geometries. Furthermore the pattern effect is investigated whose magnitude depends on the pattern and the configuration of wafer and suszeptor.The radiation is modeled with Monte Carlo and the optical properties of the surfaces are calculated from the electromagnetic theory of thin films.A commercial system with a thick suszeptor for Si-CVD at reduced pressure serves as an example for the size of the discussed effect. The comparison of the simulation with measurements provides a model validation. Then, a typical RTP system equipped with a thin suszeptor is investigated. The configuration is identified under which the thin suszeptor leads to a reduced temperature variation in case of a changing wafer coating and in case of a pattern.

Strategies for the Reduction of Pattern Effects

AbstractThe pattern effects in hot processes arise from a spatial variation of the radiative heat flux imbalance between emission and absorption. There are two scenarios to reduce the pattern effect: a very reflective chamber facing the patterned side with an illumination from the backside or a controlled double sided illumination on the other hand. The paper demonstrates the mechanism of both scenarios with a simple mathematical model and discusses the limitations of the approaches. The results will be exemplified with the help of reactor scale simulations involving a detailed Monte Carlo radiation model featuring continuous spectral dependence.