Simulation of feature profile evolution for thin film processes involving simultaneous deposition and etching Nathan Marchack, Calvin Pham, John Hoang Prof. Jane P. Chang, University of California, Los Angeles 2010 Main Objectives Essential Elements Motivation Back End of Line (BEOL) Cu ionized PVD Via etch: top view Gas flow rate Profile evolution Plasma 3. Simulation of Feature Profile Evolution – Use experimental beam systems to measure the pertinent kinetics parameters (such as sticking and recombination coefficients and formulate reaction mechanisms) as well as reactor scale models (to determine plasma parameters, such as wafer flux and ionization fractions) to be incorporated into a Monte Carlo simulator to account for surface evolution, especially with competing etching/deposition processes Expand kinetic database for profile evolution; validate using model systems: Cu IPVD & High-k etch in Cl2 containing chemistries Henrik Schumacher, CMOS Chip Structure. www.wikimedia.org Singer, Peter, Semi. Int. 2008 Front end of Line (FEOL) Current Generation MOSFET Structure Selective High k Gate Etch TEM Micrography of Current Generation High k Devices Si SiO2 Poly High k After Poly Etch After High k Etch High k NiSi, NiSi(Pt) Porous Low k Metal Adapted from http://finepolymers.com/feol-cleans Braun, A.E. Semi. Int. (2000) Reactor scale model and PIC/analytical model provide fluxes, IED and IAD Surface normal determines the direction of surface advancement Translated mixed layer (TML) kinetics extracted from MD and beam data Auth, C. et al. Intel Tech. Journal (2008) Stringent control of feature shape and geometry as dimensions decrease Lower process and development cost by predicting profile evolution Method of Approach Cu IPVD: Overview Cu IPVD: Effect of Fluxes Reactor scale flux ratios with varying power SEM Courtesy of Novellus Systems Total Ion Flux (Ar+ and Cu+) vs. Power Vyas, V., Kushner, M. JVST A 24(5) 2006 Total Cu flux and ionization fraction vs. Power HCM IPVD Reactor Simulation Domain Physics Models Kinetic Models Surface normal is calculated using polynomial fitting + q Ion Scattering Ar+ on Cu (111) 45° off normal scattering Liu, X.-Y. et al. Thin Film Solids 422, 2002 Ion Etching Cu+ Sputtering of Cu (111) Angular Dependence Coronell et al. APL 73(26), 1998 Neutral Adsorption Ion Adsorption Physical Sputtering Ion Induced Production Spontaneous Removal by incoming C flux Large bottom to top coverage ratio indicates high Cu ionization fraction Singer, Peter. Semiconductor International 2002 10 kW Power 30 kW Power 50 kW Power Sample simulations with flux ratios from above High Neutral to Ion ratio results in very low bottom deposition 21% Cu+, 16% Cu 63% Ar, Eion = 25, sE = 10 11% Cu+, 33% Cu 56% Ar , Eion = 25, sE = 10 7% Cu+, 42% Cu 51% Ar , Eion = 25, sE = 10 Simulations with higher ionization ratios 75% Cu+, 20% Cu, 5% Ar+ Eion = 50, sE = 10 85% Cu+, 20% Cu, 5% Ar+, Eion = 50, sE = 10 Higher ionization ratio and energy yields higher bottom coverage and faceting Top View Vyas, V., Kushner, M. JVST A 24(5) 2006 The TML model allows for individual reactions to be modeled and provides information about the surface composition. Robust model can easily incorporate complex kinetics and composition Reactor scale model can be used to generate normalized flux ratios and ionization fractions that can be implemented to feature scale model Cu IPVD: Effect of Ion Angle/Energy Application of TML Model Cu IPVD: Effect of TaN Barrier SEM Courtesy of Novellus Systems Cu IPVD Data Sticking on sidewalls No Sticking on sidewalls Ar+ on Cu(111) Cu+ on Cu(111) Angular Dependencies of Ion etch yield Kress, J. D. et al. JVST A 17(5), 1999 Angular dependency Angular dependent etching Eion M B Cl O Adsorption Flux Removal Flux + + Γion + + + + Γneutral Movement Flux Faceting observed due to angle dependant etching Perform experiments with selected materials and chemistry to obtain data. Test parameter accuracy by comparing to surface composition data, etc. Final verification involves using TML parameters in feature scale model and comparing profile evolution. Increasing RF Power Energy dependency Increasing Energy (Flux ratio constant 75% Cu+, 20% Cu, 5% Ar+ E=15 eV, sE = 10 25 eV, sE = 10 35 eV, sE = 10 45 eV, sE = 10 Overhang from sputter SEM Courtesy of Novellus Systems Hoang et al. JVST B 26(8) 2008 SEM images show low sidewall deposition on TaN diffusion barrier layer, possibly as a result of lacking surface interactions with ions Modeled profiles with and without sidewall sticking The TML model is built from experimentally obtained data and allows for a realistic physical description of a process to be constructed. Sticking coefficient and etch yield is a function of impact angle and energy Compare Models in Etching Complex Oxide Compare Models in Etching Complex Oxide Future Goals Continue to refine surface evolution algorithms in feature scale model Apply more complicated Translated Mixing Layer kinetics in order to simulate feature profile evolution of high-k dielectric material Special Acknowledgements Herbert Sawin at MIT Ron Kinder at Novellus Goal 1 Goal 2 Eion M B Cl O Adsorption Flux Removal Flux Eion M B Cl O Adsorption Flux Removal Flux + + + + Γion + + + + Γion + + + + Γneutral Γneutral Movement Flux Movement Flux Balance Surface Sites Translating Mixed Layer Balance Surface Sites Translating Mixed Layer Funded by For a Cl2 process dominated by etching, there is good agreement between the models. For BCl3 chemistry, both etching and deposition regimes can be modeled.