AVS1997 Session PS1-WeM: Plasma Reaction Mechanisms

Wednesday, October 22, 1997 8:20 AM in Room A7/8

Wednesday Morning

Start	Invited?	Item
8:20 AM	Invited	PS1-WeM-1 Plasma Chemistry Mechanisms for High-Density-Plasma Aluminum Etching E. Meeks, P. Ho, R.J. Buss, A. Ting (Sandia National Laboratories) A methodology for developing and testing plasma chemistry mechanisms is presented, including description of the Chemkin software and its capabilities in facilitating plasma-etch and plasma-deposition modeling. Models incorporating the plasma Chemkin software range from 0-D to 2-D axisymmetric plasma-flow models, which are used to demonstrate the modeling approach. An example application for the modeling techniques is given for metal-etch processes in high plasma density reactors. Having derived and assembled reaction kinetic mechanisms for BCl₃/Cl₂/Ar etching of aluminum, we test these mechanisms using detailed kinetic descriptions of the gas and surface chemistry in the two models. The first model is that of a well mixed reactor, while the second model solves 2-D axisymmetric plasma-flow equations including electromagnetic fields for inductive power coupling and full momentum equations for all charged and neutral species. Experimental data taken in the Sandia GEC-ICP cell¹ for mixtures of BCl₃/Cl₂/Ar provide validation of the gas-phase kinetics. Validation is shown through comparison of the models with measurements of electron density, Cl^- density, electron temperature, plasma potential, and relative BCl concentrations. We explore the sensitivity of model predictions to uncertain reaction parameters and report dominant and rate-limiting reaction paths, using formal sensitivity analysis techniques. Validation of the surface etch mechanism for blanket-aluminum wafers uses etch-rate data available in the literature covering a wide regime of Cl₂ partial-pressures. Comparisons of the well mixed reactor code and 2-D model results show 2-D transport effects on the bulk plasma behavior that are especially apparent in the GEC cell geometry. Finally, the modeling explores wafer-uniformity dependencies on reactor parameters with the 2-D plasma model. ¹G. Hebner and C. Fleddermann, submitted to Journal of Vacuum Science and Technology A.
9:00 AM		PS1-WeM-3 Effect of CHF₃ Addition On Cl₂/BCl₃ Chemistry Based Stacked TiN/AlCu/Ti/TiN Metal Etching M. Araki, K. Tokashiki, H. Miyamoto (NEC Corporation, Japan) The useful etching process for stacked TiN/AlCu/Ti/TiN wiring with less than 0.6µm pitch has been developed and the etching characteristics have been studied. To fabricate devices of less than 0.25µm dimensions, a DUV resist process is applied. This application requires high etching selectivity technology because of thinner resist thickness and lower anti-plasma erosion than the conventional i-line resist. Furthermore the ratio of TiN thickness in the stacked TiN/AlCu/Ti/TiN wiring will become larger. Improvement of etching selectivity to the photo-resist during TiN etching as well as AlCu etching is one of most important subject for the metal etching. In our experiment, it was found that CHF₃ addition to Cl₂/BCl₃ gas chemistry strongly enhanced TiN film etch rate from 250 nm/min to 450 nm/min. Optimized additional ratio was 15 to 25 % of CHF₃. As a result, this chemistry effectively improved the overall etching selectivity. In this experiment, the inductive coupled plasma (ICP) etching tool was used. Plasma optical emission spectroscopy (OES) was performed to investigate which species is dominant for the enhancement of TiN etch rate. OES revealed that CN spectrum (386,387 nm of wave length) was dominant during TiN etching under CHF₃ additive plasma. Proposed model of the TiN rate enhancement is as follows. A Nitrogen atom in the TiN film reacts with a C atom from the added CHF₃ gas and volatile cyan compounds such as CN are formed on the etched TiN surface. Thus, the TiN etch rate strongly increaces. In conclusion, CHF₃ addition to Cl₂/BCl₃ gas chemistry was found to strongly enhance TiN film etch rate, improving the overall etching selectivity due to the formation of volatile CN. This etching technology has an advantage for stacked TiN/AlCu/Ti/TiN wiring with DUV resist mask.
9:20 AM		PS1-WeM-4 Simple Method for Measuring Etching Rate Constants, Etchant Production and Predicting Uniformity D.L. Flamm, J.P. Verboncoeur (University of California, Berkeley) This paper presents means to solve two problems: 1) measuring first order gas-surface reaction rate constants without the benefit of sophisticated diagnostic equipment and 2) predicting the uniformity of chemical plasma etching as a function of processing parameters in certain reactor geometries. Both are closely related. We show that the reaction rate constant for etching may be derived from a single measurement of an etching uniformity profile. This technique relies on the fact that, under suitable conditions, the profile depends on the importance of diffusive transport limitation relative to chemical reaction. As a consequence, the ratio of the etching rate constant k(T), to the diffusivity, D(T,p), can be extracted from profile shape. Once the rate has been uncovered, an inverse problem, determining the etching rate uniformity as a function of substrate-wall spacing and pressure, can be solved analytically. Experimental profile uniformity data at several temperatures can be used to find the rate as a function of temperature. It is also shown that previous relationships of the loading effect do not apply when etching is nonuniform across a wafer. However the form of the usual loading effect equation can be restored by defining a profile average substrate area, A_eff, which can be determined from the rate constant analysis. Once this is done,loading effect data can be used as a means to find the absolute flux of radicals from various plasma sources as a function of source operating conditions.
9:40 AM		PS1-WeM-5 Etching of Photoresist through Polymer Layer, Resist Selectivity and Resist Faceting in Oxide Etch H. Shan, C.H. Bjorkman, K. Doan, M. Welch (Applied Materials, Inc.) Precision etching for sub-0.25 µm ULSI fabrication era still is nothing but micro-machining through pattern transfer, therefore, resist integrity during the etch is vital to both device yield and performance. Resist etch rates in flat areas and feature edges---resist faceting---are two key process indicators. In oxide etch, their importance is acute because oxide films are thick, feature sizes small, aspect ratios high, and overetch long. In this paper, we first report observation of etching of resist through a thick layer of simultaneously deposited polymer, then discuss its relation to resist faceting and oxide etch mechanisms. A recipe with very high resist selectivity was used to etch oxide in both ICP and MERIE etchers. After etch, severe resist faceting was seen and there remained a thick layer of polymer on top of the resist, but not on etched oxide surface. When etch duration was increased, the polymer thickness remained essentially constant, but the thickness of the resist underneath the polymer decreased with the etch time. Further, the resist loss rate was the highest initially. When a comparison recipe with lower resist selectivity was used, no resist faceting was observed though the resist etch rate at flat areas was higher. Therefore, resist was etched through the polymer layer, and the etch could be either diffusion or surface reaction limited, depending on the polymer layer thickness and polymer density. Resist faceting was resulted from presences of a thinner polymer at the resist edge and extra resist-etching radicals due to diffusion limitation through the resist. Finally, the model is used to elucidate generic oxide etch mechanisms, and methods for increasing selectivity and decreasing resist faceting are discussed.
10:00 AM		PS1-WeM-6 A Spatially-averaged Model for Plasma Etch Processes: Comparison of Different Approaches to Electron Kinetics. P. Ahlrichs, U. Riedel, J. Warnatz (Universität Heidelberg, Germany) Plasma etching of semiconductor materials is a standard manufacturing technique used in integrated circuit fabrication facilities. Due to increasing demands on feature size and high anisotropic etch rates, there is the need to operate such plasma-etch reactors at low pressure (1 - 10 mtorr). Under these highly diffusive conditions the time-scale of diffusion is fast compared to the chemical time-scale, leading to a reaction controlled regime. Therefore, the chemical kinetics in the gas phase and the gas surface interaction plays an important role for the understanding and the optimization of plasma-etch reactors. A well stirred reactor model is used to compute spatially averaged species composition by solving species, mass, and electron-energy balance equations. The reactor is characterized by the chamber volume, surface area, mass inflow, inflow composition, pressure, power deposition, and energy loss to its surroundings. This approach reduces the computational resources needed for detailed chemical kinetics calculations and allows parametric studies with respect to reaction schemes, reactor conditions, and modeling assumptions. Application of the model to chlorine chemistry provides the dependence of the plasma composition on pressure and power input to the plasma. The computation of the rates of electron impact reactions requires the determination of the electron energy distribution function (EEDF). In the calculations the EEDF, discretized on the energy-axis, is solved fully coupled to the species equations. The results obtained are compared to the often used simplification of assuming a Maxwellian distribution function or precomputing the EEDF off-line for typical compositions and temperatures¹. A comparison with experimental data reported in the literature ²^,³ is shown. Sensitivity analysis and reaction flow analysis are used to identify rate limiting reactions and to understand the dependence of the computed solution on model parameters. ¹E. Meeks, J. W. Shon, IEEE Transactions on Plasma Science, 23, 4, 539 (1995). ²E. Meeks, J. W. Shon, Y. Ra, R. Jones. J. Vac. Sci. Technol. A 13,2884 (1995). ³K. Ono, M. Tuda, K. Nishikawa, T. Oomori, K. Namba. Jpn. J. Appl. Phys. 33, 4424 (1994).
10:20 AM		PS1-WeM-7 Investigation of Plasma Uniformity for Various Feedstock Gases in an Inductively-Coupled Plasma Source W.Z. Collison, T.Q. Ni, M.S. Barnes (Lam Research Corp.) Various feedstock gases are applied for different applications ( poly-silicon etch, metal etch, oxide etch, etc.) in an inductively-coupled plasma etch reactor. In this work, plasma modeling and probe measurement are used to study plasma uniformity for several of the most commonly used gases including N₂, Cl₂, BCl₃, CF₄ etc.. It is shown that plasma density distributions for various gases are significantly different. Ions and electrons are more diffusive in CF₄ plasma and N₂ plasma , producing plasma density profiles which are more uniformly distributed in the reactor. Cl₂ and BCl₃ plasmas are less diffusive and the plasma density profiles tend to peak near the source region. The mechanism underlying this observed behavior is studied and the implication on etch reactor design (aspect ratio, coil configuration, gas injection, etc.) is discussed.
10:40 AM		PS1-WeM-8 Measurement of Neutral Kinetic Energies and Source Gas Cracking in Ar, N₂, and Cl₂ Electron Cyclotron Resonance Plasmas R.S. Goodman, N. Materer, S.R. Leone (University of Colorado, Boulder and NIST) Neutral mean kinetic energies, ion intensities and neutral source gas cracking from Ar, N₂ and Cl₂ electron cyclotron resonance plasmas are measured by modulated beam time-of-flight mass spectrometry. The mean kinetic energies of neutral species excited in the plasma are found to range between 0.04 eV and 0.45 eV, depending on plasma conditions and species measured. Mean kinetic energies increase at a nearly constant rate with decreasing pressure from 8.0 x 10^-2 Pa to 2.5 x 10^-2 Pa with constant applied microwave power. At pressures below 2.5 x 10^-2 Pa, the neutral mean kinetic energies sharply increase. This increase in neutral mean kinetic energy is attributed to an abrupt increase in the ion flux out of the source. This effect is much stronger for atomic neutrals than for molecular neutrals where internal degrees of freedom can accept energy in momentum transfer collisions. The increase in kinetic energy can be separated into two contributions, i) thermal at higher pressures and ii) non-thermal at lower pressures. The time-of-flight distributions are characterized by a two-component form consisting of a slow component and a fast Gaussian component which accounts for non-thermal species produced in the source and an effusive Maxwell-Boltzmann distribution. Cracking of N₂ and Cl₂ is also examined as a function of source pressure at constant microwave power. The N:N₂ flux ratio of particles emanating from the electron cyclotron resonance plasma varies between 0.2 and 1.4. The Cl:Cl₂ flux ratio varies from 10 to 16, indicating a very high degree of dissociation in the plasma. Both flux ratios decrease with increasing source pressure. The N atom flux peaks at ~ 1.3 x10^-2 Pa and decreases on either side of this pressure, while the total flux of Cl increases with increasing source pressure over the entire range.
11:00 AM		PS1-WeM-9 Atomic-Scale Mechanisms of Halogen Etching of Cu Surfaces C.Y. Nakakura, G. Zheng, E.I. Altman (Yale University) While metal etching reactions are used to pattern gates and interconnects, the lack of a low-temperature dry etching process for Cu has delayed the transition to Cu interconnects in integrated circuits. To address this issue, we have studied halogen etching of Cu using temperature programmed desorption (TPD), low energy electron diffraction (LEED), and scanning tunneling microscopy (STM) with the goal of developing a kinetic model of the process based on atomic-scale microscopic observations. For Cu(100), Br₂ and Cl₂ adsorb dissociatively, forming the c(2x2) structure. Near saturation of this layer, steps rotate to align in the [010] and [001] directions, leading to a zig-zag pattern with a high kink density. Further exposure results in etching, which proceeds with halide formation at step edges and, thus, can be pictured as the reverse of step flow growth. The reaction of Br₂ on Cu(100) occurs anisotropically, resulting in the observation of long, narrow peninsulas and channels. For Cl₂ on Cu(100), the etching anisotropy is not as pronounced. Steps are etched until multi-atom high steps form low index facets. The Cu halide species formed at the steps are mobile at room temperature, aggregating into clusters and forming ring structures related to stress relief. Above 500 K the Cu halide sublimes, completing the etching process. Annealing the chemisorbed layer prior to reaction increases the reactivity for Br₂ on Cu(100), but not for Cl₂. Evidence of variations in adsorbate structure along step edges will be presented, demonstrating structures which have increased reactivity, accounting for the anisotropy. In addition, STM movies will be shown to demonstrate the dynamics of the etching process.
11:20 AM		PS1-WeM-10 The Effects of Plasma Processing Parameters on the Surface Reactivity of OH(X²Π) in Tetraethoxysilane/O₂ Plasmas During Deposition of SiO₂ E.R. Fisher, K.H.A. Bogart (Colorado State University); J.P. Cushing (Stanford Research Systems) The OH(X²Π) radical in a 20:80 tetraethoxysilane (TEOS)/O₂ plasmas has been characterized during deposition of SiO₂ using the Imaging of Radicals Interacting with Surfaces (IRIS) method. This method combines spatially resolved laser-induced fluorescence and molecular beam techniques to examine the plasma-surface interface. The reactivity of OH at the surface of a growing SiO₂ film has been determined as a function of the applied rf plasma power (P), and the substrate temperature (T₂). The reactivity (R) of OH during deposition of silicon dioxide on a 300 K Si substrate is 0.41±0.04. R decreases as substrate temperature increases, but is unaffected by increasing rf power. Translational and rotational temperatures (Θ_T and Θ_R, respectively) of the OH radical are also determined. For a 20:80 TEOS/O2 plasma (P = 85 W), Θ_T = 912 ± 20 K and Θ_R = 450 ± 20 K. Θ_T is significantly higher than Θ_R and increases with increasing rf power. Using isotopically labeled ¹⁸O₂ as a precursor, the source of the oxygen in OH is identified as the O₂ gas, not oxygen from the ethoxy groups on TEOS. The implications of our work on the mechanism for OH formation in the plasma, the surface reactivity of OH radicals, and the possible roles of OH in the deposition of SiO₂ from TEOS/O2 plasmas will be discussed.
11:40 AM		PS1-WeM-11 Modeling of Silicon Dioxide Deposition in High Density Plasma Reactors and Comparisons of Model Predictions with Experimental Measurements R.S. Larson, E. Meeks, P. Ho, C. Apblett (Sandia National Laboratories); S.M. Han, E.A. Edelberg, E.S. Aydil (University of California, Santa Barbara) High density plasma deposition of SiO₂ is an important process in integrated circuit manufacturing. A list of gas-phase and surface reactions has been compiled for modeling plasma-enhanced chemical vapor deposition of SiO₂ from SiH₄, O₂ and Ar gas mixtures in high density plasma reactors. The gas-phase reactions include electron impact, molecule-molecule, ion-ion, and ion-molecule reactions. The surface reaction mechanism is based on the insight gained from attenuated total reflection Fourier transform infrared spectroscopy experiments and includes adsorption of radical species on to the SiO₂ surface, ion-enhanced desorption from the surface layer, radical abstractions, as well as direct ion- energy-dependent sputtering of the oxide film. A well-mixed reactor model that consists of conservation equations averaged across the reactor volume was used to model three different kinds of high-density plasma deposition chambers. Experimental measurements of total ion densities, relative radical densities, and net deposition-rate, as functions of plasma operating conditions, have been compared to the model predictions. The results show good quantitative agreement between model predictions and experimental measurements. The compiled reaction set and surface reaction network description was thus validated and can be employed in more sophisticated two or three dimensional plasma simulations.