Papers

AVS2004 Session PS1-MoA: Plasma Surface Interactions in Etching

Monday, November 15, 2004 2:00 PM in Room 213A

Monday Afternoon

Start	Invited?	Item
2:00 PM		PS1-MoA-1 Quantitative Plasma Beam Investigation of Polysilicon Sidewall Roughening S.A. Rasgon, Y. Yin, H.H. Sawin (Massachusetts Institute of Technology) For the patterning of sub 100 nm features,a clear understanding of the origin and control of line edge roughness (LER) is extremely desirable, particularly at the gate level where variations in line width can adversely impact the electrical performance of the device. Plasma etching processes often roughen the feature sidewalls, leading to the formation of anisotropic striations. It is this post-etch sidewall roughness which will ultimately affect device performance. Our past research has focused on the observation of sidewall roughness via a novel AFM technique. The resulting images allow the extraction of quantitative information on sidewall roughness and spatial frequency as a function of depth, and vividly highlight the structure of the post-etch sidewall. While these images present a remarkable display of sidewall roughness encountered in common etching processes, a fundamental study of post-etch sidewall roughness remains elusive due to the inherent experimental difficulties encountered. Sidewall roughening during etching depends on the plasma chemistry, ion bombardment energy, and ion incident angle. A true fundamental study requires independent control of all three parameters, impossible to obtain in a conventional plasma etcher. To remedy these difficulties, an inductively-coupled plasma beam source was constructed that allows the exposure of a sample to a realistic ion and neutral flux, of any desired plasma chemistry, while allowing independent control of the ion bombardment energy and incident angle. By rotating the sample to a near-glancing angle, a sidewall can be simulated. This apparatus is used to conduct a fundamental study of sidewall roughness/striation during HBr etching of polysilicon. The resulting AFM images are analyzed for roughness magnitude/spatial frequency using a novel geostatistical technique, and are compared with real sidewalls. Finally, insight into the roughening mechanism is obtained through 3D modeling of the roughening process.
2:20 PM		PS1-MoA-2 3-Dimensional Feature Scale Simulation of Polysilicon Sidewall Roughening H. Kawai, W. Jin, H.H. Sawin (MIT) The line edge roughening has become an important factor as the features shrink. Although there are 2-dimensional simulators that can simulate the artifacts such as faceting and microtrenching, they can not simulate the surface roughness since it is inherently 3-D in nature. Therefore, a 3-dimensional simulator was developed to study the physics of surface and line edge roughening. 2 Â½ -dimensional simulator, that had been developed before, applied Monte Carlo model to model the surface kinetics. Although 3-D simulator also used Monte Carlo model, many changes were made to convert the simulator from 2 Â½ -D to 3-D. These include a new algorithm for the computation of surface normals and fluxes on sidewalls. In addition, since 3-D simulator is more computationally intensive, it is necessary to optimize the simulator by minimizing the computation time while maintaining the accurate results. Simulation domain was discretized into cubic cells with the side of 2.5 nm, and when a particle strikes a surface cell, the local surface conformation was determined. The algorithm was based the fitting of the local region of the surface cells with a polynomial. The cell size of 2.5 nm had been used since the cell size corresponds to the ion induced mixing length. Since the grooved striations formed in line edge roughening have minimum radius of curvatures of about 25 nm, the appropriate degree of polynomial and number of cells to be fitted were selected to allow the representation of surface curvatures of 25 nm or less. This fitted surface is then used to compute the surface normal, scattering angle, and flux on the 3-D surface. The surface normal was used to determine the movement of the surface with material etching or deposition by selecting the appropriate cell for cell addition or collapse.
2:40 PM		PS1-MoA-3 Unraveling the Complex Process Known as 'Plasma Chemistry' M.J. Goeckner (University of Texas at Dallas) 'Plasma chemistry' is perhaps one of the most complex processes known. In general it can be thought of as the interaction between three main scientific subsystem, plasma physics, gas-phase chemistry/physics and surface-phase chemistry/physics. To understand this complexity one simply needs to consider how a given reactive gas-phase specie might interact with a surface. Does it stick to the surface? Does it chemically react with the surface? Does it promote film growth? How does this interaction change the gas composition? How does an altered gas-phase chemistry alter the plasma? Understanding these interactions is key to producing better models of plasmas, allowing the optimization of complete process systems and hence improved product yield. This talk will briefly review how various groups are attacking this complex problem. Then using our fluorocarbon chemistries (CF₄, C₄F₈) studies as an example, we examine how gas and surface chemistries change for different wall conditions (temperature, diameter and material) as well as gas flows and plasma parameters. Based on this knowledge, we will discuss possible interaction mechanisms and how these might affect the process. This will in turn lead to a discussion of possible future studies. This work is supported by a grant from NSF/DOE, CTS-0078669.
3:20 PM		PS1-MoA-5 Stabilization of Radical Composition Drift in Fluorocarbon Plasmas K. Nakamura (Chubu University, Japan); H. Sugai (Nagoya University, Japan); K. Oshima, A. Ando, T. Tatsumi (Sony, Japan) Fluorocarbon discharges have been widely used for etching processes of dielectric thin films for microfabrication. However, these have suffered from various problems, in particular, repeatability of the etching characteristics. The problem becomes recently severe due to narrow process margin for next generation ULSI devices. One of the major origins is plasma-surface interaction on polymer-deposited vessel wall, leading to significant time-variation of radical composition of the plasma. Alternating ion bombardment (AIB) method has been proposed to reduce such interactions by applying a RF bias to the chamber wall¹. This paper reports the effects of the AIB on polymer film deposition onto the chamber walls and the time-variation of radical density in fluorocarbon plasma reactors. 13.56 MHz inductively-coupled plasmas are produced in Ar-diluted C₄F₈ gases in a stainless steel chamber in which two semi-cylindrical electrodes are set. A 400 kHz RF source serves alternating negative bias to the electrodes. The AIB drastically suppressed polymer deposition on the biased wall, and the deposition rate decreases by one order of magnitude with ~100 eV ion bombardment compared to the non-bias case. On the other hand, the AIB also reduces a rise time of densities of the fluorocarbon radicals after the discharge starts, and reached to steady state within ~10 s for CF₂ radicals. ¹K. Nakamura et al: J. Vac. Sci. Techonol. A 18 (2000) 137.
3:40 PM		PS1-MoA-6 Etching of Passivated SiO₂ Film by Fluorocarbon Ions: A Molecular Dynamics Study V. Smirnov, A. Stengatch, V. Pavlovsky (Sarov Labs., Russia); S. Rauf, P. Stout, P.L.G. Ventzek (Freescale Semiconductor) Fluorocarbon plasmas are widely used for etching of dielectric thin films in the microelectronics industry. Fluorocarbon radicals and ions are known to produce a thin passivation layer (~ 2 nm) on the dielectric surface, whereupon energetic ion bombardment leads to dielectric material etching. As the passivation films are extremely thin and in-situ monitoring is difficult during etching, very few experimental studies have been able to probe into the fundamental nature of fluorocarbon based dielectric etching. Computational molecular dynamics (MD) is one technique that has proven useful for such studies. This paper reports about a MD based investigation of fluorocarbon passivated SiO₂ film etching by CF_x (x=1, 2, 3) ions. Our MD code is 3-dimensional and uses the velocity-Verlet method for particle acceleration. Psuodo-potentials for two and three body interactions of Si, O, C, and F have been assembled either using Gaussian based quantum chemistry computations or data available in literature. A variety of fluorocarbon passivation films (with varying thickness and F/C ratio) are prepared by bombarding low to medium energy fluorocarbon ions on SiO₂. Impact of energetic (50-1000 eV) CF_x ions on these passivation films is then investigated, and modeling results are used to determine ion etch yield, nature of sputtered clusters, and their energy and angular distributions. Modeling results clearly demonstrate that presence of a fluorocarbon passivation film enhances etch yield compared to a similar but otherwise unpassivated SiO₂ film. Etch yields peak at an off-normal angle, and SiO_xF_y constitute the bulk of Si containing sputtered clusters.
4:00 PM		PS1-MoA-7 Spontaneous Etching of Silicon with F Atoms and XeF₂: A Unified Model H.F. Winters, D. Humbird, D.B. Graves (UC Berkeley) A recently proposed molecular dynamics simulation of spontaneous etching of undoped silicon with F has been successful in describing a variety of experimental observations.¹ Insights gained from this simulation (denoted HG) have been used to refine a model which explains other spontaneous etch observations, including etching by XeF₂ and various effects of doping on spontaneous etching, among others. The model is based on the assumption that the reaction is proportional to the negative ion concentration on the silicon surface (e.g. at SiF₃^- centers)² using the umbrella-type reaction mechanism observed in the simulation. HG predicts a 5 Å SiF_x layer on silicon exposed to F atoms and it will be demonstrated that XPS data are consistent with this prediction. It will also be shown that XPS data indicate a layer twice as thick (~10 Å) for XeF₂. Experimental data showing that the layer thickness is relatively independent of incident flux and temperature over significant ranges of these parameters as predicted by HG will be presented. Assuming these layer thicknesses are correct allows us to calculate the negative ion concentration on the silicon surface being etched.² These results are correlated with various experiments including reaction probability measurements over the temperature range 200--1000 K and doping experiments with concentrations from ~10¹⁵ to 10²⁰ dopants/cm³. It will be shown that one model can be used to describe the experimental results for the spontaneous etching of Si(111) by both F and XeF₂. A plausible explanation will be presented as to why the doping effect in chlorine is large relative to fluorine even though the opposite trend is observed for spontaneous etch rate. Many other experimental results (etch rate proportional to reactant density, insensitivity of product distribution to doping level, etch product composition, and the doping and reaction probability differences between F and XeF₂) are consistent with this model. Finally, the HG conclusion that etch products are desorbed with significant kinetic energy allows data from modulated beam mass spectrometry calibrated for 300K products to be interpreted properly, leading to determination of the reaction probability from T(surface) = 200 K - 1000 K. The data will be presented and compared with theory. ¹ David Humbird and David B. Graves, J. Appl. Phys, in press, (2004) ² H. F. Winters and D. Haarer, Phys.Rev. B, 36. 6613 (1987); 37 ,10379 (1988)
4:20 PM		PS1-MoA-8 Real-time Spectroscopic Studies of Si Etch Dynamics A.A.E. Stevens, J.J.H. Gielis, M.C.M. van de Sanden, H.C.W. Beijerinck, W.M.M. Kessels (Eindhoven University of Technology, The Netherlands) Nanometer scale control during IC, MEMS/NEMS and photonic device production becomes more and more an issue. Plasma etching and ion beam processing cause the creation of surface roughness and defects, such as dangling and strained bonds. The roughness and defects resulting from the production process end up at critical interfaces in the devices and, thus, influence their performance in a negative way. Therefore, fundamental studies of the creation of roughness and dangling/strained bonds in plasma and ion beam processing of silicon are required. In the past, mass spectrometry studies in well-characterized beam-etching experiments revealed a great deal of information regarding the synergy between ions and etchant during Si etching. At present, more detailed information is desired and can be obtained with spectroscopic surface diagnostics, which have made significant advances over the last years. Hence, spectroscopic ellipsometry (SE) and second harmonic generation (SHG) are employed in situ and real-time during the etching of Si with beams of Ar⁺ [10-2000 eV] and XeF₂. By means of SE the surface roughness is measured. If the etching is dominated by XeF₂ etching, the surface becomes rough (drough ~ 1-10 nm). However, when Ar⁺ ions are driving the etch process, the surface remains relatively smooth (drough < ~1 nm). Complemented with atomic force microscopy measurements, the dynamic roughness scaling theory, expressed in parameters α and β, is applied to find the origin of the roughening processes during the etching. In order to study the creation of dangling and strained bonds SHG studies are being carried out. The role of the ions and etchant in the creation of roughness and defects at etched Si surfaces will be discussed on the basis of the results obtained with these surface-sensitive spectroscopic analysis techniques.
4:40 PM		PS1-MoA-9 Insights into the Ion Energy Dependence of Ion-Assisted Chemical Etch Rates in High-Density Plasmas L. Stafford, J. Margot (Universite de Montreal, Canada); M. Chaker (INRS-Energie, Canada); S.J. Pearton (University of Florida) Over the last few years, important research efforts have been devoted to the development of plasma etching models of various materials in various reactive plasma mixtures. These models generally include a surface kinetic model in which it is usually assumed that the ion-assisted chemical etch rate varies like the square root of the ion energy. This dependence is empirically deduced from the universal energy dependence of physical and ion-assisted chemical etch yields presented by Steinbruchel.¹ In the present work, we show from existing experimental data that the ion-assisted chemical etch rate does not necessarily follow this particular energy dependence. A typical example is provided by the etching of ZnO films in an Ar/Cl2 high-density plasma. To explain this behavior, we propose an analytical model in which the formation rate of the adsorbate is assumed to be proportional to the number of adsorption sites. In the specific case of ZnO, the adsorption sites are generated by thermal desorption of oxygen atoms. This is found to induce a non-linear relation between the etch rate and the square root of the ion energy. For oxide materials with more complicated structures like (Pb,Zr)TiO3 (PZT), ion bombardment is required to generate adsorption sites. In this case, the adsorbate formation depends on ion energy, which results in an etch rate that approximately varies like the square root of the ion energy. The predictions of our model are found to be in excellent agreement with the experimental data reported for several materials, for example ZnO, SiO2, HfO2, PZT, PST, and SBT. In addition, the model includes previous ion-assisted etching models such as that developed by Gottscho and his co-workers.². ¹ C. Steinbruchel, Appl. Phys. Lett. 55, 1960 (1989) ² R.A. Gottscho, C.W. Jurgensen and D.J. Vitkavage, J. Vac. Sci. Technol. B 10, 2133 (1992) .
5:00 PM		PS1-MoA-10 A Model of Multilayer Surface Reactions and Simulation of the Feature Profile Evolution in Etching of Silicon in Chlorine Plasmas Y. Osano, K. Ono (Kyoto University, Japan) A phenomenological model has been made to simulate the feature profile evolution of nanometer-scale etching of Si in Cl₂. The model incorporates an atomistic picture into the model, to analyse the complex surface reactions in the ion-enhanced etching and investigate their effect to the profile evolution, which involves profile anomalies such as bowing, tapering, and microtrenching. To simulate the reaction process at an atomic scale, we employ a feature profile modeling with two-dimensional array of atomic size cell in the entire computational domain. Monte Carlo calculation of the trajectory and stopping of the incoming Cl⁺ ion within the surface layers of Si substrate is then performed, on the basis of kinetics of two-body elastic collision. For surface reactions of Cl neutral reactants, we take into account their adsorption on the very surface layer. The removal of Si atom is assumed to be caused by the reaction on the chlorinated surface in terms of this adsorption process Si_(s)+4Cl_(s) -> SiCl_4(g), where (s) and (g) represent the solid and the gas, respectively. Simulation of the feature profile evolution is performed for etching of sub-100 nm patterns. The effect of neutral-to-ion flux ratio is studied in this calculation. The present model illustrates that changes of the flux ratio have a significant effect on surface anomalies, such as sidewall bowing and tapered feature near the bottom, associated with surface chlorination on the feature surface which varies by the flux ratio and the location within the feature pattern.