Papers

AVS1997 Session MS+VM-WeM: Modeling of Processes and Equipment

Wednesday, October 22, 1997 8:20 AM in Room J1/4

Wednesday Morning

Start	Invited?	Item
8:20 AM		MS+VM-WeM-1 Computational Modeling as a Design Tool in Microelectronics Manufacturing M. Meyyappan (NASA Ames Research Center) Plans to introduce pilot lines or fabs for 300 mm processing are in progress. The IC technology is simultaneously moving towards 0.25/0.18 micron. The convergence of these two trends places unprecedented stringent demands on processes and equipments. More than ever, computational modeling is called upon to play a complementary role in equipment and process design. The pace in hardware/process development needs a matching pace in software development: an aggressive move towards developing " virtual reactors " is desirable and essential to reduce design cycle and costs. This goal has three elements: reactor scale model, feature level model, and database of physical/chemical properties. With these elements coupled, the complete model should function as a design aid in a CAD environment. This talk would aim at the description of various elements. At the reactor level, continuum, DSMC(or paricle) and hybrid models will be discussed and compared using examples of plasma and thermal process simulations. In microtopography evolution, approaches such as level set methods compete with conventional geometric models. Regardless of the approach, the reliance on empricism is to be eliminated through coupling to reactor model and computational surface science. This coupling poses challenging issues of orders of magnitude variation in length and time scales. Finally, database development has fallen behind; current situation is rapidly aggrevated by the ever newer chemistries emerging to meet process metrics. The virtual reactor would be a useless concept without an accompanying reliable database that consists of: thermal reaction pathways and rate constants, electron-molecule cross sections, thermochemical properties, transport properties, and finally, surface data on the interaction of radicals, atoms and ions with various surfaces. Large scale computational chemistry efforts are critical as experiments alone cannot meet database needs due to the difficulties associated with such controlled experiments and costs.
9:00 AM		MS+VM-WeM-3 A Self-Consistent Kinetic (Plasma-DSMC) Model for Chemically Reacting Low Pressure Plasma Reactors T.J. Bartel, J.E. Johannes, M.L. Hudson (Sandia National Laboratories); D.J. Economou (University of Houston) The Direct Simulation Monte Carlo (DSMC) technique, a kinetic simulation of the Boltzmann equation, has been extended in a self-consistent manner to model the chemically reacting plasmas in low pressure systems. Our 2D DSMC code has been optimized for the parallel computing environment to overcome the inherent CPU constraints of the method. We use a bulk plasma approximation for modelling the electrons and electro-static fields for these reactor systems. We assume local charge neutrality in the bulk and a sheath model to compute the potential jump from the surface into the bulk. We simulate the neutral and ion species as computational particles. We include charge exchange, recombination and electron impact gas phase reactions. The sheath thickness is still very small at these pressures and the Debye length is much smaller than a reactor dimension. Therefore, we do not transport electrons; the local electron density is determined from summing the local ion densities. The Inductively Coupled Power (ICP) deposition is computed using the ORMAX code. The Maxwellian electron temperature is determined from a local energy balance and the electro-static space charge field is computed assuming the electrons are in a Boltzmann equilibrium. We will compare this model with experimental data: the GEC reference cell for both a Cl2 and a Cl2/BCl3 discharge (no wafer) at 10 to 20 mtorr and a High Density Plasma commercial etch tool with C2F6 chemistry at 5-10 mtorr. We will also compare the DSMC results to results from continuum based models for these problems.
9:20 AM		MS+VM-WeM-4 Three-Dimensional Fluid Simulation of Inductively Coupled Plasma Reactors T. Panagopoulos, D.J. Economou (University of Houston) A three-dimensional fluid simulation of inductively coupled plasma reactors has been developed. A modular approach was used to couple in a self-consistent manner the disparate time scales of plasma and neutral species transport. Complex chemical reactions involving electrons, ions and neutrals as well as surface chemistry were included in the simulation. The Poisson equation was not solved in the bulk plasma. Instead, the electron density in the bulk was calculated from the electroneutrality constraint. The power deposited into the plasma was an input to an electron energy module where the electron temperature and the rate coefficients of electron-impact reactions are calculated. These were in turn used as source terms in separate neutral and charged species transport modules. By iterating among the modules, a self-consistent solution was obtained. Results have been obtained for both argon and chlorine chemistries. As expected, the gas density was maximum near the inlet and minimum near the pumping port. Further, the gas density was higher near the walls of the reactor compared to the center due to ion pumping. It was also observed that the plasma density was locally higher near the exit port where the surface-to-volume ratio for ion losses to the walls is smaller. The global maximum of the plasma density for the argon discharge was located at the center of the reactor, while for the chlorine discharge off-center peaks were observed. Several spatial patterns obtained using different reactor and inflow/outflow configurations showed a non-uniform dependence of all essential variables on the azimuthal direction. The location of both gas inlet(s) and pump outlet played a key role.
9:40 AM		MS+VM-WeM-5 Two and Three Dimensional Modeling of SiO₂ Chemical Mechanical Planarization Based on Pad Deformation T.K. Kim, Y.H Kim, H.J. Lee, J.T. Kong, S.H. Lee (Samsung Electronics Co. Ltd., Korea) Chemical mechanical polishing(CMP) has been used as a fundamental planarization tool to satisfy the design rule of multiple levels of wiring with decreasing pitch and photolithographic depth of focus. This paper describes the modeling of SiO₂ CMP process based on the nonlinear pseudo-plastic deformation from the contact between a pad and a wafer. Using ABAQUS, a program to analyze solid mechanics by the finite element method, the elastic stress of the SiO₂ surface is calculated and nonlinearly converted to the normalized evolution rate of the target surface. A time dependent CMP profile is then generated by the marker presentation to develop the new points of the surface. In order to verify the model, a CMP profile simulator is applied to the test patterns with the trench depth of 0.5µm. Two and three dimensional transient profiles for the island patterns with the pattern width of 0.25, 0.50 and 1.00 µm, agrees well with the experimental data and explain the nonlinear relationship of the elastic stress of the patterns independent o The CMP profile simulator is also applied to find rules to minimize the nonplanarity(i.e, to analyze the dummy pattern effect and height difference effect prior to steps in 256Mbit DRAM technology). Furthermore, this simulator is used to understand the mechanical contact properties between the wafer and the polishing pad according to the equipment parameters, such as the pad hardness and the down pressure.
10:00 AM		MS+VM-WeM-6 A Physically based Analytic Collimated PVD Model for Control and Prediction. M.M. IslamRaja (Texas Instruments Inc.) Collimated Physical Vapor Deposition (PVD) is the most commonly used technique to deposit titanium in contacts and vias used in multilevel metalization for VLSI technology. One of the drawbacks of Collimated PVD is the decrease in the deposition rate with usage due to deposition on the collimator walls. This requires frequent adjustments of the deposition time to achieve the required deposition thickness. The adjustment for the deposition time is made by using test wafers to calculate the deposition rate. The use of test wafers is not desirable as it increases the cost and cycle time for this process. Empirical models for the change in deposition rate as a function of usage are not accurate enough to eliminate the usage of test wafers. A physically based analytic model has been developed to calculate the transport of the sputtered species through the collimator. At any instant, the model predicts the fraction and directionality of the sputtered species passing through the collimator as well as the deposition on the collimator walls. The simple, analytic closed form solution of the model can be used to control the deposition process by continuously adjusting the deposition time required to give the desired thickness. This eliminates the use of test wafers resulting in better cycle time and lower cost for Collimated PVD processes. The model results are in very good agreement with experimental deposition rate data over the entire life-time of the collimator. In addition to the deposition rate, the model also predicts the across wafer non-uniformity and the directionality of the depositing species. Therefore, it can also be used to optimize the collimator configuration. As the model is physically based it is not dependent on the type of target or the collimator aspect ratio. Another advantage of the model is that it is much simpler and several orders of magnitude faster than the Monte Carlo techniques currently being used to simulate the species transport through the collimator.
10:20 AM		MS+VM-WeM-7 Composition Control by Current Modulation in DC-Reactive Sputtering T. Nyberg, S. Berg, C. Nender (Uppsala University, Sweden) It is well known that the reactive sputtering process generally shows a hysteresis effect when the reactive gas supply is used as the control parameter. The key to control the composition of the deposited film is primarily to find some way of controlling the partial pressure of the reactive gas. It has earlier been suggested that a partial pressure variation also can be accomplished by other means than varying the reactive gas supply. In this study we want to report about a detailed investigation concerning current control as an alternative method to tune the partial pressure level. We will show that if the reactive sputtering process is carried out in this mode of operation both the deposition rate and the partial pressure vs the ion current curves exhibit hysteresis effects. Varying the target current will directly influence the gettering rate of the reactive gas thereby varying the residual partial pressure. By varying the power supply level from high to low current the process may avalanche from metallic to compound sputtering mode. By periodically shifting between proper high to low current it is possible to tune the average partial pressure to the desired value. Results from experiments and computer simulations will clearify the relationship between the involved key parameters for this mode of operation.
10:40 AM		MS+VM-WeM-8 Topography Simulation for Redeposition Effects on Etch Pattern Resolution of Platinum Cell Capacitor Electrodes of DRAM in Ar/Cl₂/O₂ Sputter Etching J.S. Park, H.W. Kim, H.J. Lee, W.J. Yoo, J.T. Kong, S.H. Lee (Samsung Electronics Co. Ltd., Korea) In this paper, three dimensional plasma modeling and topography simulation for the effects of redeposition on the etch resolution of the platinum film which has been used as an electrode material in a DRAM cell capacitor, is presented. Although sputtering is known to be a proper method for platinum etching, redeposition of non-volatile by-products like Pt or PtCl_x during the etching on the sidewall of the mask pattern makes it difficult to define the subquarter micron etch patterns. To predict the etch resolution accurately, an etch profile simulator has been developed considering the selectivity of the oxide mask against the platinum film and sticking coefficients of etch by-products as calibration parameters. Simulation results show that the etch pattern resolution depends on the oxygen content in the Cl₂/O₂/Ar gas, as well as geometry effects which are a thickness ratio between the platinum film and the oxide mask, and the distance between the nearest neighboring cell electrodes. The sidewall slopes of the cell electrodes measured by SEM agree well with the simulation results.
11:00 AM		MS+VM-WeM-9 Effects of Wall Conditioning on Ion and Neutral Densities and Etching Rate in High Density Plasma Reactors Using BCl₃/Cl₂/Ar Gas Mixtures* S.J. Choi (Sandia National Laboratories); R. Veerasingam (The Pennsylvania State Unversity) The predictive capability of numerical plasma reactor models depends sensitively on the accuracy of the plasma chemistry mechanisms, the database, and the surface chemistry that are included in the models. For low pressure (few mTorr to 10s mTorr) etching reactors used in the semiconductor industry, the boundary conditions for the model are very crucial since the gas phase chemistry does not dominate the chemical processes because of the relatively large mean free paths. Rather, it is extremely difficult to set the correct boundary conditions for the surface chemistry since the chamber wall condition is determined by the wall temperature, surface type (wall material and the covered chemical species), and the process history (wall coverage) of the reactor. In this paper, numerical simulations are performed using two-dimensional plasma reactor models (MPRES[1] and HPEM[2]) to investigate the spatial distributions of neutral and ion densities and fluxes to the wafer surface for different wall conditions. The sensitivity of the surface condition is demonstrated with a aluminum etching chemistry (BCl₃/Cl₂/Ar) in an ICP GEC reactor geometry. [1] D.P. Lymberopoulos and D.J. Economou, J. Vac. Sci. Technol. A, 12, 1229 (1994). [2] P.L.G. Ventzek, R.J. Hoekstra, and M.J. Kushner, J. Vac. Sci. Technol. B, 12, 461 (1994). *This work performed at Sandia National Laboratories supported by the U.S. Department of Energy under contract DE-AC04-94AL85000.
11:20 AM		MS+VM-WeM-10 Investigation of the Cl₂/BCl₃/Ar Chemical Mechanisms and the Effects on Ion and Neutral Densities and Etching Rate in High Density Plasma Reactors* R. Veerasingam (The Pennsylvania State Unversity); S.J. Choi (Sandia National Laboratories) Modeling the complex chemistry in low pressure etching reactors that are used widely in the semiconductor wafer industry is an important step in the development of newer tools with lower process pressure in the trend towards smaller critical dimensions. In order to make predictive statements using numerical plasma reactor models, it is vital to include the important chemistry mechanisms and the wall boundary conditions correctly in the models since the model calculations depend sensitively on them. Understanding the effect of the chemistry mechanism and the uncertainties in the database on the modeling calculations will provide the modeler and the tool developers greater insight and guidance for more eficient tool designs. In this paper, we focus on the plasma chemistry using the latest available database for BCl₃ in a Cl₂/BCl₃/Ar gas mixture used typically for aluminum etching. Numerical simulations are performed for an ICP GEC reactor geometry using two-dimensional plasma reactor models (MPRES[1] and HPEM[2]) to investigate the spatial distributions of neutral and ion densities and fluxes to the wafer surface. [1] D.P. Lymberopoulos and D.J. Economou, J. Vac. Sci. Technol. A, 12, 1229 (1994). [2] P.L.G. Ventzek, R.J. Hoekstra, and M.J. Kushner, J. Vac. Sci. Technol. B, 12, 461 (1994). *This work performed at Sandia National Laboratories supported by the U.S. Department of Energy under contract DE-AC04-94AL85000.
11:40 AM		MS+VM-WeM-11 Trends in Wafer-Edge Aluminum Etch Rate Uniformity in a Commercial ICP Plasma Etch System D.F. Beale, S.C. Siu, R. Patrick (Lam Research Corporation) Effects of process, power and chamber geometry on the wafer-edge etch rate uniformity of Al etched in Cl₂ were both measured and simulated in the showerhead-injected Lam TCP 9600 etch reactor. Parameters used in the experimental matrix included pressures of 6 to 24 mT, flows from 25 to 200 sccm, mixture fractions of 0 to 50% N₂, plasma powers of 0-350 Watts and chamber heights of 6-12 cm. Distinctive features of this study are the large number of input parameters studied in a commercial reactor, the emergence of utilization fraction as a factor determining etch profile as new tools push to ever-lower residence times, and the quantitative connection between Peclet number and edge uniformity over a large range of process conditions. Despite the number of parameters varied, the average etch rate was found to significantly depend only on flow rate and, to a lesser extent, residence time. This result simplified the data analysis and simulation that followed. After trends in etch uniformity with plasma power were determined experimentally, the study was further simplified by conducting the remainder of the measurements and all simulations at zero plasma power. A previously validated¹ fluid simulation was used in the explanation of uniformity trends. These trends were explained in terms of three factors: the wafer-edge Peclet number, the pressure and the utilization fraction. The importance of each factor was confirmed in etch rate measurements and simulations. Further simulations show how the reported trends differ when side injection, a collimator (edge ring) or a novel distributed injection scheme is used. ¹D.F. Beale, P.L. Jones, S.C. Siu, R. Patrick, 43rd AVS National Symposium, MS+PS-WeM5, Philadelphia (1996).