AVS1996 Session PS-ThM: Etch II

Thursday, October 17, 1996 8:20 AM in Room 201C

Thursday Morning

Start	Invited?	Item
8:20 AM		PS-ThM-1 Negative Ion Assisted Etching of Silicon with Low Energy H. Shindo (Tokai University, Japan); T. Shibayama (Toyo University, Japan); M. Sekine (Toshiba Corporation, Japan); T. Ichiki, Y. Horiike (Toyo University, Japan) Negative ion etching of silicon was studied in halogen gas plasmas of Cl\sub 2\ and SF\sub 6\. The microwave plasma was produced in SUS reactor chamber of 10 cm in diameter and the silicon etching was examined on an electrically biased stage by DC and RF voltages. The etching stage was cooled down to the liquid nitrogen temperature in SF\sub 6\ while in Cl\sub 2\ it was water-cooled. In downstream region, 9 to 15 cm, far from the microwave entrance, the saturation current ratio of positively and negatively biased probe approaches about unity, meaning the large amount of negative ions exist there. In this downstream plasma of both gases, silicon etching was observed at a positive substrate bias with a low threshold voltages, as +5 V in Cl\sub 2\ and +7 V in SF\sub 6\. The threshold is extremely low compared with that in the positive ion, as -45 V in Cl\sub 2\ and -30 V SF\sub 6\. The results strongly suggest that the negative ion etching could be low damage process. A mass and energy analysis by QMS (Hiden EQP-300 ) showed that the dominant negative ion were Cl\super -\ and F\super -\ while in the positive Cl\sub 2\\super +\ and SF\sub 3\\super +\. The ions of Cl\super -\ and F\super -\ are atomically very like the etchants for silicon, and thus this results in a very low etching threshold. Silicon etching in SF\sub 6\ plasma with an admixture of Ar was also studied, in which the non-reactive positive ion of Ar\super +\ was expected to produce as well as the reactive negative ion of F\super -\ and thus only the negative ion etching to occur. The etching with -70 V DC bias was stopped at 30 % Ar, while with +70 V bias the etching was not stopped. This result is very consistent with the above-mentioned fact that the etching threshold of negative ion is very low. The mass analysis of this mixed plasma by QMS showed that SF\sub 3\\super +\ ion completely disappeared at 60 % Ar. This provides a technique of possible negative ion etching with an application of RF bias. An analysis by AFM also showed that the roughness of the etched silicon surface by the negative ion was about one order lower than that by the positive one.
8:40 AM		PS-ThM-2 Analysis of a High Throughput, Reliable, WSi/poly Gate Etch System with Proven 0.25 \mu\m \+-\ 0.025 \mu\m 3-sigma CD Control and High WSi:poly Selectivity K. Olson, L. McCulloch (Tegal Corporation); J. Liu (Sematech/IBM); L. Jerde, J. Almerico (Tegal Corporation) This paper describes the analysis of a system which has demonstrated 0.25 micron WSi/Polysilicon gate etch processing with better than \+-\ 0.025 microns, 3-sigma, critical dimension (CD) control. These results were obtained on 200 mm wafers using a High Density Reflected Electron (HRe\super -\) reactor with Cl\sub 2\ and HBr process gases. High WSi:Poly selectivity offers large advantages for the elimination of profile defects and stringers. It also significantly reduces processing problems associated with thin (40-90 \Ao\) gate oxides. This system has demonstrated the ability to control WSi:Poly selectivity from 1.2:1 to 2.5:1. Typical processes used selectivity's between 1.5-2.0. Factors controlling the WSi:Poly selectivity will be discussed. High selectivity allowed independent optimization of the etch characteristics of each film. WSi and polysilicon etch rates ranged from 2000-3000 \Ao\/minute while within-wafer uniformity was <6%, 3-sigma. Profile angles of 88-90 degrees were demonstrated with 0-1 degrees profile microloading between isolated and dense features. Gate oxide loss of approximately 10 \Ao\ was demonstrated. The system has proven manufacturing performance of >38 wafers/hr and uptime > 98%. Particle performance of <10 (median = 6) adders (\>=\0.3\mu\m) has been demonstrated for multiple runs of >2500 wafers/chamber. In Situ spatially resolved Residual Gas Analysis (RGA) of local gas phase composition was performed. Identification of reactant and product species and their concentration dependencies on process variables such as reactant feed composition and bias power will be discussed.
9:00 AM		PS-ThM-3 Optical Diagnostics of Inductively Coupled Plasmas: Experiment and Simulation D. Economou, R. Wise (University of Houston) We have performed optical emission spectroscopy (OES) and laser induced fluorescence (LIF) measurements of an inductively coupled plasma reactor in chlorine/argon gas mixtures. Radial scans along the 200 mm diameter wafer platen as well as axial scans along the 100 mm reactor height have been performed. Species monitored were Cl atoms by OES and Cl\sub 2\super +\ ions by LIF. The effect of pressure, power, flow rate, and gas composition on the radical and ion density uniformity was examined. Experimental results are compared to two-dimensional plasma simulations using the Modular Plasma Reactor Simulator (MPRES). Possible pitfalls of actinometry as normally applied in practice in high-density low-pressure reactors have been identified, namely: (a) contribution to the argon emission by excitation of metastable states, and (b) gradients in the actinometer gas density due to pressure drop (conductance limitations) and/or due to temperature gradients across the reactor. Interesting multiple steady-states of the plasma have also been observed as the reactor pressure and/or power are varied. Optical diagnostics of the different plasma states will be reported. Work supported by Sandia/SEMATECH and the National Science Foundation.
9:20 AM		PS-ThM-4 Atomic Scale Simulation of Ion-Enhanced Etching: From Picoseconds to Minutes M. Barone (Lam Research Corporation) Predicting the evolution of the topography of an etched feature requires accurate models for the rates of etching and deposition on the surface. These rates are influenced by the chemical nature of species delivered to the surface as well as the efficiency of energy transfer from impinging ions to the substrate. Molecular dynamics (MD) simulation is an ideal tool to probe the brief but intense collision cascade following ion impact, and to determine scattering probabilities of incoming ions (such as ions at grazing incidence, which play an important role in the formation of microtrenches) as a function of incident energy and angle. However, MD simulations alone cannot access all pertinent timescales involved in the evolution of the selvage layer. The etch yields obtained in previous studies of the Ar/Si/Cl etch system, while qualitatively correct as a function of ion energy, were lower than those experimentally obtained in a plasma etch environment and compared more favorably to experimental beam studies \super 1\. The presence of reactive radicals in the plasma which were not accounted for in the simulation partially accounts for this difference. In order to correct for this, and obtain realistic model surface layers, the slower processes occurring between ion impacts must also be investigated. Utilizing Monte Carlo and structural relaxation techniques, the neutral species arriving to the surface between ion impacts and the relaxation of the surface following energy dissipation are taken into account. Probability distributions of etch yield and ion scattering are then determined for a range of incident energies (10-100 eV) and angles (0-85 off normal). The quantitative, physically based mechanisms revealed by atomic scale simulations of ion-assisted etching have the potential to improve the predicative capabilities of topography simulators significantly. \super 1\ M.E.Barone and D.B. Graves, Plasma Sources Sci. Technol. 5 (1996) 1-6
9:40 AM		PS-ThM-5 Optical Investigation of the Etching of Si and Ge in an Inductively Coupled Cl\sub 2\/Ar Plasma J. Choe, I. Herman (Columbia University); V. Donnelly (Bell Laboratories) Several optical probes are used to monitor the etching of Si and Ge in real time within an inductively coupled plasma (ICP) using Cl\sub 2\/Ar mixtures. The ICP reactor has a 2 inch diameter wafer positioned about 2 inches from the quartz window through which RF power at 13.6 MHz is coupled. A Faraday shield is sometimes employed to lessen sputtering from this window. Optical interferometry (6328 \Ao\) is used to monitor the etch rate of polysilicon films as a function of delivered RF power (up to 600 Watts), DC bias of the substrate (+5 to -110 V, applied using RF at 18 MHz), and the fraction of Cl\sub 2\ in the flow mixture at a 20 mTorr total pressure. With maximum RF power and bias, the etch rate is ~2600 \Ao\/min for a 6% chlorine mixture. A sublinear increase in etch rate is observed with increasing chlorine fraction, starting with 100 \Ao\/min in pure Ar, as is a strong dependence of etch rate on DC bias. The intensity of optical emission from Si* (2882 \Ao\) and SiCl* (2823 \Ao\) during etching of c-Si qualitatively tracks the etch rate as process conditions are changed. Optical emission from Ge* (3269 \Ao\) and GeCl* (3393 \Ao\) during the etching of c-Ge tracks process conditions in a similar way; no evidence of Si* or SiCl* emission (attributable to sputtering from the quartz window) is seen. The use of a pulsed XeCl excimer laser for laser-induced thermal desorption (LITD) of adsorbates from the surface followed by laser-induced fluorescence (LIF) of desorbed species, for real-time monitoring of the steady-state surface layer, will also be reported. The work at Columbia was supported by NSF grant DMR-94-11504.
10:00 AM		PS-ThM-6 Spatial Emission Profile and Neutral Density Depletion in an Electron Cyclotron Resonance Plasma P. Lee, S. Lee, H. Chang (Korea Advanced Institute of Science and Technology) In electron cyclotron resonance (ECR) plasmas, spatial neutral ( or radical ) profile and plasma density profile are very important in uniform wafer process. A plasma emission profile could differ from a plasma denisty profile because of depleted neutral density profile. In this paper, we study plasma emission profiles by using an optical probe and the neutral density profiles in a production and a reaction region of an ECR system. In the production region, we observe a plasma density oscillation which is a nonlinear neutral-plasma interaction in the plasma production region, and explains the oscillation as a predator-prey model which describes an oscillating population of fox-rabbit (electron-neutral in plasma ) in ecology. This oscillation might be an evidence of neutral denisty depletion. In the reaction region, we observe the plasma density and the emission profiles. The emission profile differs from the plasma density profile because the neutral density is not uniform. The neutral density profile is strongly affected by the spatial ion energy profile, and which, in turn, is related to magnetic field profile and substrate size. To control the plasma emission profile, we investigate effects of chamber wall heating, substrate size, magnetic field profile, and guard ring. Therefore, we find that the neutral density profile is non-uniform through the ion-neutral collision and could be controllable. We do poly-silicon etch experiments, and study the relation between the etch profile and the emission profile. The actinometric technique is not valid for the relative radical density profile investigation because of non-uniform actinometer density profile.
10:20 AM		PS-ThM-7 Ion and Neutral Temperatures in Inductively Coupled Plasma Etching Reactors S. Rauf (University of Illinois, Urbana); M. Kushner (University of Illinois, Urbana-Champaign) In moderate pressure (100s mTorr) plasma etching reactors, collisional coupling between ions and neutrals is sufficient to equilibrate the bulk ion and neutral gas temperatures. In low pressure (< 10s mTorr) high plasma density reactors, such as inductively coupled plasma (ICP) sources, the specific power deposition is higher and collisional rates are lower. This produces long mean free path effects such as temperature jumps at boundaries and disparities between ion and neutral temperatures. In this paper, results are presented from a computational study of the ion and neutral gas temperature in ICP plasma tools, and comparisons are made to experiments.\super 1\ The model is an improved version of a previously described 2-d hybrid model. \super 2\ The hybrid model combines modules for the electromagnetic fields and electron kinetics with a fluid module for heavy particle transport. Energy equations were added to the fluid module for all neutral and charged species, energy exchange terms were included between all species, and temperature jump conditions were used. Results from the model explain the experimentally observed increase in ion temperature near boundaries as collisional heating in the presheath. Excursions of the gas temperature above ambient observed at pressures of less than 10-20 mTorr correlate with the onset of temperature jump conditions.\super 1\G. Hebner, to be published.\super 2\W. Z. Collison and M. J. Kushner, Appl. Phys. Lett. 68, 903 (1996).\super *\Work supported by NIST, SNLA/Sematech, NSF and the U of Wisconsin ERC.
10:40 AM		PS-ThM-8 The Influence of Different Mechanisms on the Angular Dependence during Dry Etching C. Hedlund, L. Jonsson, I. Katardjiev, S. Berg, H. Blom (Uppsala University, Sweden) Future generations of integrated circuits require dry etching processes that can be controlled to a high accuracy. Simulation of the etch process that can predict topography evolution during etching is needed to avoid time consuming and costly trial and error work. Simulation programs, however, need information about the angular dependence of the etch rates during processing. A general knowledge of the angular dependence of the etch rate for different etch processes is not yet available.We have used a recently developed method for measuring the angular dependence of the etch rate for a variety of etch processes. The method is based on etching materials, deposited on specially patterned silicon wafers. From the data it is possible to extract information about the process and identify its components. We have found that the majority of reactive ion etching (RIE) decomposed to two basic constituents one isotropic and one directional component. It is then possible to plot the dependence of the components versus various parameters , such as pressure, flow etc. Thus for instance etching of poly-silicon with SF\sub6\ is to exhibit a linear dependence between the isotropic etch rate component and the pressure.Several processes, using high density plasmas like inductively coupled plasmas (ICP) and reactive ion etching (RIE), have been studied. The results are used as an input to the topography simulation program DINESE. A number of simulation examples and comparison with experiment are also presented
11:00 AM		PS-ThM-9 Energy and Angle Distributions of Reflected Ions and Sputtered Species within Microfeatures in Plasma Processing B. Helmer, D. Graves (University of California, Berkeley) Previous profile evolution studies have shown that ions reflected from sidewalls can lead to microtrench formation on the bottom of an etched feature. In these studies, the ions were assumed to reflect specularly, retaining all of their incident energy, if the incident angle was above a critical value. In the present study, we describe the distributions of reflected ion energies and angles obtained using molecular dynamics (MD) simulations. Use of these distributions in profile simulations provides a more accurate treatment of reflected ions, and allows assessment of the above assumptions. We also describe the energy and angle distributions of sputtered products, which may deposit or etch other feature surfaces, and play a major role in the evolving profiles. We simulated Ar\super +\, Cl\super +\, and Cl\sub 2\\super +\ impacts onto chlorinated silicon surfaces. The ion incident energy and angle were varied from 20 to 100 eV and 0 to 85 degrees from the surface normal, respectively. We obtained the ion reflection probabilities, sputtering yields, and the energy and angle distributions of reflected ions and sputtered species (Si, Cl, and SiCl\sub x\). For impacts at glancing angles (\>=\ 75 deg.), the distribution of reflected ion energies was quite broad; ions retained < 10% up to > 90% of the incident energy. Ions with the highest reflected energies scattered from the surface with super-specular polar angles (\theta\\sub ref\ > \theta\\sub inc\). Species sputtered by these glancing ions can also receive high energies. For example, Cl atoms with energies up to 40 eV were sputtered by 50 eV Ar\super +\ ions at 75 degrees.
11:20 AM		PS-ThM-10 Effects of NO Addition on the Chemical Downstream Etching of Silicon Nitride and Tungsten M. Blain, R. Jarecki, T. Meisenheimer, J. Stevens, C. Ashby (Sandia National Laboratories) Chemical downstream etching of silicon nitride (Si\sub 3\N\sub 4\) requires discharges containing F, O, and N, e.g. CF\sub 4\/O\sub 2\/N\sub 2\ or NF\sub 3\/O\sub 2\, for efficient etch rates and selectivity to oxide. For example, a 10% addition of N\sub 2\ to a CF\sub 4\/O\sub 2\ discharge causes a factor of six increase in the Si\sub 3\N\sub 4\ etch rate and an 8% drop in the SiO\sub 2\ etch rate, resulting in selectivities approaching 9:1. Measurements of the relative amounts of NO\sub x\ species in the etch chamber suggest that the increase in Si\sub 3\N\sub 4\ etch rate is related to the formation of NO in the discharge. The importance of NO was confirmed by comparing effects of injecting N\sub 2\, NO, NO\sub 2\, and N\sub 2\O into the discharge and into the reaction chamber during a CF\sub 4\/O\sub 2\ discharge. Nitride etch rates increased significantly for NO, relative to the other N\sub x\O\sub y\ species. Even higher etching efficiency was obtained when NO was injected into the chamber during an NF\sub 3\/Ar discharge. Reaction of NO with surface-bound N is proposed as the mechanism for the rate enhancement and comparison of discharge and etch reaction species using mass spectrometry suggests N\sub 2\O as the reaction product. A factor of six increase in the amount of W removed is observed upon injection of NO in the reaction chamber during an NF\sub 3\/Ar discharge. W etch rates are similar both with and without NO injection and initial mass spectrometry data shows no W(NO)\sub x\ or WF\sub x\(NO)\sub y\ species. This indicates that F etching of W is initially accelerated when NO is present and it is proposed that NO reaction with surface WO\sub 3\ may be significant. *Work supported by DOE contract DE-AC04-94AL85000 and by SEMATECH
11:40 AM		PS-ThM-11 Ashing of Photoresist with a Plasma-based Source of Hyperthermal Neutral Atoms M. Goeckner, J. Park, Z. Wang, S. Cohen (Princeton Plasma Physics Lab) A microwave-heated plasma device, employing a planar magnetron configuration for enhancing plasma density, has been built to produce a continuous stream of directed hyperthermal neutrals. These neutrals are generated in the following manner: In a 19-cm diameter 17-cm long cylindrical cavity, a plasma is produced using 2.45 GHz microwaves. The ions are accelerated toward the magnetron surface by a small, ca. -40 Volts, applied bias. The ions are neutralized just prior to impact. For certain surface materials, the accelerated particles are reflected from the surface, retaining about half of their incident energy. A small cathode bias, compared to a standard planar magnetron, is desired to optimize the energy and flux of the reflected neutrals. This is achievable because of the hybridization of the planar magnetron with a microwave cavity. By bouncing the hyperthermal neutrals a second time, the target can be moved out of the line-of-sight to the plasma. This reduces the UV irradiation of the target material. Additionally, the second bounce allows control of the uniformity. For 250 W of coupled microwave power and an Ar pressure of 2 mTorr, n\sub e\ =89 7 x 10\super 10\ cm\super -3\, T\sub e\ =89 5 eV, \phi\\= sub p\ =89 12 V and an ion current of 0.35 A to the cathode have been obtained. Neutral beams produced with Ar/O\sub 2\ plasmas have been used to ash photoresist. Ash rates have been found to vary with bias, total pressure and relative concentrations of Ar and O\sub 2\. Ash rates up to =89 25 nm/min. have been measured. Simple calculations indicate that ash rates on the order of 1 =B5m/min. can be expected in a slightly modified device.