Papers

AVS1996 Session MS-FrM: Process & Equipment Modeling

Friday, October 18, 1996 8:20 AM in Room 201A

Friday Morning

Start	Invited?	Item
8:20 AM		MS-FrM-1 RTP Equipment and Their Use in Manufacturing K. Jensen, S. Banerjee, J. Cole, J. Hebb (Massachusetts Institute of Technology) The low thermal budget and single wafer processing characteristics inherent to rapid thermal processing (RTP) have made the technique increasingly used in microelectronics manufacturing. However, challenges remain in the control of thermal uniformity and run-to-run temperature repeatability. We present a systematic approach to addressing these issues based on finite element (FEM) and Monte Carlo (MC) simulations in conjunction with experimental validation. The methodology enables model based design without the expensive and time consuming efforts associated with the traditional "build-and-try-equipment" approach. Effects of reactor configuration, quartz windows, guard rings, and modification of surface radiation properties are delineated through examples. Simulations are also demonstrated to be useful in temperature sensor development by providing quantitative information about the nature and origin of reflected light interference on pyrometer readings. Insight into pattern effects on wafer temperature uniformity are obtained by combining system scale models with multilayer electromagnetic theory for chip and wafer scale radiative properties. These detailed physical models are, however, too computationally intensive to be implemented directly in model based control schemes. Therefore, we also present a systematic way of extracting low order nonlinear models by using the proper orthogonal decomposition method. These resulting low order models show orders of magnitude reduction in computation time while giving excellent agreement with steady state and transient data from the complete model. Applications of this general strategy in control and process simulations of RTP equipment are illustrated. Current and future roles of RTP simulations in manufacturing are also discussed.
9:00 AM		MS-FrM-3 A 3-Dimensional Model of Polysilicon Etch Profiles in High Plasma Density Inductively Coupled Plasma Reactors R. Hoekstra, M. Kushner (University of Illinois, Urbana-Champaign) There is an acknowledged need for integrated process models which self consistently predict etch profiles based on equipment parameters such as bias power, gas pressure and gas mixture. Significant progress in the development of plasma equipment models has now provided the capability to predict reactant fluxes to the wafer, however, there are few examples of where the equipment models have been merged with profile simulators. In this paper we describe an integrated process model which links a plasma equipment model with a profile simulation. The plasma model is the Hybrid Plasma Equipment Model, a general 2-/3-dimensional simulator for plasma tools. The profile modle is a newly developed 3-dimensional Monte Carlo simulation called the Monte Carlo-Feature Profile Model (MC-FPM). The equipment model generates the angular and energy distributions of fluxes of neutrals and ions to the wafer. These fluxes are then used by the MC-FPM to determine the time dependent profile evolution of etch features on the wafer. The MC-FPM enables the user to specify any reaction mechanism (including energy dependence) in terms of fundamental parameters. The etching of 3-d structures in p-Si using chlorine chemistries in Inductively Coupled Plasma reactors have been investigated with the integrated model. The effects of inhibitor fluxes, feature charging and substrate topography will be discussed. /super */Work supported by SRC, NSF, SNLA/Sematech and U of Wisconsin ERC.
9:20 AM		MS-FrM-4 Model for HBr/Cl\sub 2\ Plasma Etching of Polysilicon with Comparisons to Data E. Meeks, J. Shon, R. Larson (Sandia National Laboratories); A. Labun, R. Jones (Digital Equipment Corporation) We describe a chemistry model for etching of polysilicon in mixtures of HBr and Cl\sub 2\. The etch process is modeled using the Sandia well mixed reactor code AURORA, which allows detailed chemical kinetics description via CHEMKIN and SURFACE CHEMKIN interfaces. In the gas phase, electron-impact reaction-rate coefficients are calculated from cross sections of electron-impact collisions with HBr, Cl\sub 2\, H, Br, and Cl, assuming Maxwellian electron energy distribution functions over a range of electron temperatures. We also include neutral reactions with highly reactive radical species, and consider the pressure-dependency of these reactions. At the wafer surface, the silicon etch mechanism is based on experimental measurements of the etch process. The etch mechanism includes adsorption of neutral species, and ion-induced desorption of etch products. The ion reaction rates are based on measured yield functions for ion-enhanced etching. The yield function depends on directed ion energy impacting the wafer surface. In this way, the model can account for rf biasing of the wafer, which serves to increase the ion energy. The ion fluxes are limited by the Bohm criterion, corrected for the presence of negative ions. Data from a response-surface study of etch rate in a Lam 9400 etch tool are compared with simulations, in which pressure, inlet gas mix, rf source power, and rf bias voltage are modified, according to the response surface inputs. The results show good agreement with all the trends observed in the response surface. The simulated etch rates are properly ranked within a range that varies from 1700-2700 Angstroms/minute.
9:40 AM		MS-FrM-5 Stress Distribution on Wafers during CMP Processing D. Wang (Arizona State University); J. Lee (IPEC PLANAR); S. Beaudoin, T. Cale (Arizona State University) The planarization performance of a chemical mechanical polishing (CMP) process is governed by a combination of mechanical and chemical mechanisms. The inherent removal rate of the material is generally considered to be a function of the polishing chemistry. The mechanical aspects of polishing include polishing pressure, pad compressibility and speed of the polishing platen. In particular, the physical properties of the polishing pad and the carrier film play important roles in CMP uniformity. This paper presents results of a project designed to understand the role of pad and carrier film properties on CMP uniformity. Specifically, this paper; 1) describes a wafer scale solid-solid contact model which is solved using I-DEAS (a commercial software package), 2) summarizes the effects of carrier film and pad properties in CMP processes, and 3) proposes that polishing process can be considered as a material yielding process. Results indicate that 1) CMP uniformity improves with increasing polishing pad hardness and carrier film hardness, and 2) the 'Von Mises stress' is the key to understanding polishing, as opposed to the normal stress; i.e., the mechanical part in CMP is a yielding process. The need for low compressibility polishing pads and low compressibility carrier films is demonstrated.
10:00 AM		MS-FrM-6 Profile Simulation of Reactive Ion Etching with Photoresist Erosion J. Park, H. Lee, K. Kim, S. Lee (Samsung Electronics Co., Ltd., Korea) The reactive ion etching of a contact hole with a large aspect ratio needs the more sophisticate control of polymer deposition, which is the major challenge to the SiO2/Si etching with a high selectivity in the Ar/CF4/CHF3 gas system. Especially, the BC(Buried contact) profile of a COB (capacitor-on-bitline) structure must be well adjusted to get an enough process margin and a large contact area. In this paper the profile simulation for RIE(reactive ion etching) process for a contact hole in the Ar/CHF3/CF4 gas system in 1 Giga DRAM technology has been presented. A model has been developed which describes the influence of the photoresist erosion and the polymer-neck generation produced in the real contact etch process. Widening of the photoresist erosion makes CF2 radicals freely move to the inside of the contact hole. The sufficient supply of CF2 radicals to overcome the high energetic ion flux initiates polymer deposition to form a polymer-neck. Based on the empirical photoresist erosion model, the contact etch simulation yields a polymer-neck profile well matched to the real process result.
10:20 AM		MS-FrM-7 Modeling Transisent Enhanced Diffusion: Fundamental Barrier for Simulating Deep Submicron Technologies P. Packan (Intel Corporation) In order to improve MOS device performance, extremely shallow, high concentration dopant profiles are needed. These profiles are necessary for maximizing drive currents and minimizing device channel lengths. In order to form highly doped, shallow dopant profiles, dopant diffusion must be kept to a minimum. For today's deep submicron device technologies, the majority of dopant diffusion is due to transient enhanced diffusion (TED) caused by ion implantation. Surprisingly, lowering the thermal budget of a process does not guarantee a shallower junction. Instead, the complicated interaction of damage annealing kinetics and dopant diffusion determines device junction depths. Point defects, extended defects and dopant interactions all affect dopant diffusion during thermal annealing. In order to accurately simulate advanced deep submicron technologies, these interactions must be well understood and modeled. The impact of TED on current process technologies will be presented. Examples which illustrate fundamental effects which must be modeled will be shown. An overview of our current understanding as well as recent advances in modeling of TED will be presented. The trend toward low energy, high dose implants followed by a high temperature RTA anneal places added importance on accurate modeling of surface and extended defect interactions as well as stringent requirements on the thermal profile of the RTA process. However, fundamental understanding and physically based modeling of these process steps are crucial for designing and simulating future deep submicron technologies.
11:00 AM		MS-FrM-9 Equipment Simulation of SiGe Heteroepitaxy: Model Validation by Ab Initio Calculations of Surface Diffusion Processes M. Hierlemann, C. Werner (Siemens AG, Germany) Incorporation of very little Ge into a Si surface significantly increases the deposition rate during CVD. This is due to the fact that hydrogen/chlorine desorption is faster from SiGe surfaces making available additional sites for adsorption. Two mechanisms are discussed to explain the observed catalytic effect: (i) the diffusion model, where surface diffusion of H/Cl atoms from Si to Ge sites opens an energetically more favourable path for hydrogen/chlorine desorption, and (ii) the collective model, where incorporation of Ge into Si would stimulate an overall change of the electronic structure of the surface leading to higher desorption rates. Ab initio cluster calculations are used in this work to investigate possible diffusion pathways of H and Cl atoms on SiGe surfaces. Binding energies of H and Cl atoms on different surface sites are calculated. The Si-H/Ge-H binding energies do not depend on the neighbouring atoms. This makes the collective model highly unprobable. In a next step the transition states for migration between neighbouring surface sites are located and the relevant activation barriers are calculated. Surface diffusion of H and Cl from Si to Ge is found to be energetically more favourable than desorption of H2, HCl or SiCl2 from Si. Hence, the diffusion model can be considered a valid approach. It is shown in macroscopic reactor simulations that the diffusion model is useful to explain the observed catalytic effect.
11:20 AM		MS-FrM-10 A Molecular Dynamics Simulation of Particle Agglomeration in Reactive Ion Etching Reactors F. Huang (University of Illinois, Urbana); M. Kushner (University of Illinois, Urbana-Champaign) Dust particle contamination continues to be an important issue in microelectronic device manufacturing. Particles collected on surfaces and in exhausts are often agglomerates made up of smaller, monodisperse primary particles. The plasma tool conditions which produce agglomerates are of interest for controlling particle inventory. The shapes of the agglomerates are important with respect to their transport in the plasma. We have previously reported on a particle agglomeration model (PAM) which assesses the propensity to form agglomerates in reactive ion etching (RIE) tools by tracking trajectories of individual monomers and using the energy of colliding primary particles to determine agglomeration rates. In this paper, we discuss results from an improvement to the PAM in which a molecular dynamics techniques are used to resolve the shape and rate of growth agglomerates. We find that larger (> 100s nm) primary particles in high powered plasmas tend to form compact agglomerates. Dendritic filamentary agglomerates are usually formed by small primary particles in low powered plasmas. For these conditions, incoming primary particles can only attach to the edges of the agglomerate where coulomb repulsive forces are lower. Comparisons will be made of the computed shapes to agglomerates which have been collected on wafers and in exhausts.\super *\Work supported by SRC, NSF, SNLA/Sematech and U of Wisconsin ERC.
11:40 AM		MS-FrM-11 Coupling Reactor-scale and Molecular Dynamics Simulations to Feature Profile Evolution Simulations M. Vyvoda, B. Helmer, M. Li, D. Graves (University of California, Berkeley) A major weakness in existing feature profile evolution simulators (FPES) is a heavy reliance on empiricism. It is difficult to unambiguously identify mechanisms responsible for feature shapes because it is usually possible to adjust parameters in the FPES to match observed profiles. By coupling a reactor-scale model of plasma chemistry and a molecular dynamics (MD) simulation of ion- surface interactions to a FPES, we have developed a model that does not rely on arbitrarily chosen parameters. The parameters in the reactor-scale model have been established by comparison to experiment, and the parameters in the MD simulation are interatomic potentials, established in previous work. A major complication in coupling MD simulations with a FPES is a disparity in length and time scales. MD simulations are restricted to length scales of 10\super -8\ m, and time scales of 10\super -12\ s. On the other hand, feature profile dimensions are on the order of 10\super -6\ m and evolve on time scales of minutes, so it is not obvious how to couple the simulations. We have developed a scheme to couple these simulations based on the concept of super-particles from test-particle Monte Carlo, using multiple trajectories sampling different impact parameters in MD. We illustrate the combined reactor-scale model - MD - FPES for the case of poly-Si etching in an Ar/Cl\sub 2\ plasma, highlighting the role of ion reflections from sidewalls and the effects of sputtering on the transport of both Cl and Si within the evolving feature. The mechanisms of aspect ratio dependent etching and microtrenching are identified in the simulation.