Transcript A部門大会OHP
XIV International Conference on Gas Discharges and their Applications (GD2002) in Liverpool The University of Liverpool, 1-6 September 2002 Developing an Efficient PIC/MC Simulator for RF Glow Discharge using a Higher Order Sampling Technique K.Satoh1, M.Kozaki1, H.Date2, H.Itoh1 and H.Tagashira1 1 2 Department of Electrical and Electronic Engineering, Muroran Institute of Technology, Muroran 050-8585, Japan College of Medical Technology, Hokkaido University, Sapporo 060-0812, Japan [email protected] Agenda 1. Motivation & Objective 2. Methodology 3. Simulation model & conditions 4. Results 5. Conclusions MURORAN INSTITUTE OF TECHNOLOGY Introduction Particle-In-Cell/Monte Carlo (PIC/MC) simulator is one of the most reliable and useful simulators for RF glow discharges. The behaviour of charged particles both in equilibrium and in non-equilibrium regions is calculated accurately and the effects of boundaries are treated properly. The velocity distributions of the charged particles, which are important and useful information to understand the characteristics of RF glow discharges, are obtained. However, statistical fluctuations are included in the PIC/MC essentially. Traced MC particles have weight factor irrespective to the position in a gap, so that number of the MC particles in the sheaths is smaller than that in the bulk plasma. Since the sheath regions are important to sustain the discharge, the statistical fluctuations due to the shortage of MC particles may cause the error of the PIC/MC simulation. In order to remove the instability of the PIC/MC simulation, the statistical fluctuations must be reduced. MURORAN INSTITUTE OF TECHNOLOGY (Ardehali et al., IEEE Trans. Plasma Sci., Vol.25, pp.1081-5, 1997) Sheath regions are split and MC particles are added in the regions. Fluctuations are reduced with a small amount (< 5%) of increase in computation time. Legendre Polynomial Weighted Sampling (LPWS) (Ventzek et al., J.Appl.Phys., vol,75, pp.3785-88, 1994) The higher order derivatives of the bin of energy distribution are sampled, and then the density distribution in the bin is described using Legendre polynomials. The distributions are multiplied by the weight of B-spline, and then superposed to eliminate the discontinuity between successive bins. sheath sheath bulk Number of MC particles SPLIT-PIC/MC simulator Ion density Approaches to reduce fluctuations in MC simulation This technique has a wider applicability to MC sampling. In this work we developed a PIC/MC simulator coupled with LPWS to reduce the statistical fluctuations in the density distributions of charged particles. The developed simulator is applied to the simulation of a RF glow discharge in nitrogen. MURORAN INSTITUTE OF TECHNOLOGY METHODOLOGY - Conventional sampling & principle of LPWS •Conventional method (simple counting) n(z) e Ni=Ni+1 〇 bin •Legendre Polynomial Weighted Sampling n(z) bin e 〇 n1() n0() N0 z0 N1 z1 ni() Ni Ni+1 Ni+2 z2 zi zi+1 Dz z Number of electrons in a bin is added. If (zi<z≦zi+1) then Ni Ni 1 To obtain detailed distribution n(z) ・fine mesh(bin) is needed. ・statistical fluctuations increase. ・a number of electron must be sampled. ・considerable computing time z0 z1 z2 zi ni+2() ni+1() zi+1 Dz z 1 χ 1 Density distribution ni(x) in a bin is described using Legendre polynomial. ni ( ) P0 ( ) f 0 P1 ( ) f1 P2 ( ) f 2 ・・・ 1 1 f 0 1 P0 ( )ni ( )d 1 2 2 P0 1, P1 , P2 2 (3 1) 3 1 f1 P1 ( )ni ( )d 2 1 2 z ( zi 1 zi ) / 2 5 1 zi 1 zi f P2 ( )ni ( )d 2 2 1 To obtain detailed distribution n(z) ・increasing terms of the polynomial. ・accurate n(x) is independent on bin size. MURORAN INSTITUTE OF TECHNOLOGY METHODOLOGY - LPWS : overlap sampling & superposition Problem ni zi ni+1 Solutions zi+2 zi+1 ・To sample enough number of particles. ・Overlap sampling & superposition of ni and ni+1 using B-spline (LPWS). ① ni(z) { n(z) At boundary, ni and ni+1 are not necessarily continuous. Superposition (smoothing) by the primary B-Spline 1 χ 1 S(z) z Si(z) z ② { n(z) Si+1(z) S(z) z Si(z) Si+1(z) Generally, m 1 n( z ) ni ( z ) Si ( x) , z i 1 x z zi zi z zi 1 zi 1 zi m : the order of B-spline, Si : B-spline n(z) n(z) ② z zi n( z ) (1 ) ni ni 1 zi 1 zi ni+1(z) 1 χ 1 ① 0 χ 1 zi-1 zi zi+1 zi-2 z MURORAN INSTITUTE OF TECHNOLOGY METHODOLOGY - PIC/MC simulation coupled with LPWS ① conventional Flowchart Calculating flight, collision & scattering of MC particles in Dt Charge assignment n(z) between electrodes is obtained by LPWS, and then charge is assigned to grid points. e q Dz 1 Dz ② ① Charge is assigned to grid points as a function of its position between grid points. n(z) 1q Dz q z ② using LPWS n(z) Calculating electric field using poisson’s eq. Dz z Mesh width PIC/MC simulator PIC/MC simulator with LPWS MURORAN INSTITUTE OF TECHNOLOGY Simulation Conditions Electron-N2 collision cross sections x=0 (cm) (Y.Ohmori et al., J. Phys. D: vol.21, pp.724-9, 1987) -14 10 Cg=2.5 (pF) qm -15 10 qi 2 cross section (cm ) Cb=2.0 (pF) ・N2 ・0.5 Torr ・20 ℃ ・n0=1.0×107 (cm-3) ・MC partical : 30,000 x=3 (cm) -16 10 qv(10) qex(20) -17 10 -18 10 -19 10 V=300sinwt w/2p =13.56MHz LPWS for ne(z) and np(z) Number of terms (Legendre Polynomial) : 7 terms ○ B-spline : primary B-Spline ○ 1 10 electron energy (eV) 2 3 10 10 (Phelps, J.Phys.Chem. Ref. Data, vol.20, pp.557-73, 1991) 10 -13 Qm 10 -14 Q(Vib) 2 number for field calculation (Dx≦lD) 60 meshes (Dx=0.05cm), 300 meshes (Dx=0.01cm) ○ Simulation Step (Dt≦1/fe ) Dt(=0.123ns)=1/600cycle 0 10 N2+-N2 collision cross sections cross sections [cm ] ○ Mesh -1 10 10 -15 QCT Q(N+) 10 10 -16 -17 Q(N 3+) Q(391) Q(300-500) 10 -18 0.1 1 10 100 1000 electron energy [eV] MURORAN INSTITUTE OF TECHNOLOGY Spatiotemporal variations of electron & ion densities 1499-1500 cycle 2.5x10 8 2.0x10 8 1.5x10 8 1.0x10 8 60 grid points for charge assignment Ion density 2.5x10 8 2.0x10 8 1.5x10 8 1.0x10 8 5.0x10 7 5.0x10 -3 7 0.0 0.0 P 0.5 1.0 1.5 positio n (cm) 2.0 2.5 1500.0 1499.8 1499.6 ) 1499.4 cle 1499.2 cy e( 1499.0 tim 3.0 0.0 0.0 P G 2.5x10 8 2.0x10 8 1.5x10 8 1.0x10 8 2.5x10 8 2.0x10 8 1.5x10 8 1.0x10 8 5.0x10 7 0.5 1.0 1.5 positio n (cm) 2.0 2.5 3.0 1500.0 1499.8 1499.6 ) e 1499.4 cl cy 1499.2 e( tim 1499.0 G 5.0x10 7 0.0 0.0 P 0.5 1.0 1.5 positio n (cm) 2.0 2.5 1500.0 1499.8 1499.6 ) 1499.4 cle 1499.2 cy e( 1499.0 m i t 3.0 -3 PIC/MC with LPWS electron density (cm ) -3 electron density (cm ) Standard PIC/MC electron density (cm ) -3 electron density (cm ) Electron density 0.0 0.0 P 0.5 1.0 1.5 positio n (cm) G Statistical fluctuations are reduced by LPWS. 2.0 2.5 3.0 1500.0 1499.8 1499.6 ) 1499.4 cle cy 1499.2 e( tim 1499.0 G MURORAN INSTITUTE OF TECHNOLOGY Spatiotemporal variations of electric field 1499-1500 cycle 60 grid points for charge assignment Standard PIC/MC PIC/MC with LPWS 20000 0 10000 0 electric field (V/m) 10000 electric field (V/m) 20000 -10000 -10000 1500.0 1499.8 1499.6 time 1499.4 (cyc le) 1499.2 1499.0 0.0 P 0.5 1.0 1.5 2.0 ) n (cm pisitio 2.5 -20000 3.0 G 1500.0 1499.8 1499.6 time 1499.4 (cyc le) 1499.2 1499.0 0.0 P 0.5 1.0 1.5 2.0 ) n (cm pisitio 2.5 -20000 3.0 G The fluctuations in the charged particle densities do not make large influence on the electric field in this case. MURORAN INSTITUTE OF TECHNOLOGY Spatial variations of densities & electric field electron density 300 grid points for charge assignment ion density 6 6 300x10 300x10 250 Standard PIC PIC with LPWS 250 100 Standerd PIC PIC with LPWS -3 ion density (cm ) -3 electron density (cm ) 2999.3 cycle 200 150 100 Electric field 200 150 100 50 50 0 0 0.5 1.0 1.5 2.0 2.5 3.0 0.5 Position [cm] 1.0 1.5 2.0 Position [cm] Net charge density 6 -3 net charge density (cm ) 150x10 2.5 3.0 electric field (V/cm) 50 0 100 -50 50 0 Standerd PIC PIC with LPWS -50 Standerd PIC PIC with LPWS -100 0.5 1.0 1.5 Position [cm] 2.0 2.5 3.0 -100 0.5 1.0 1.5 2.0 2.5 3.0 Position [cm] 60 meshes → 300 meshes Fluctuations in the densities by the standard PIC/MC increase. (both in sheath and in bulk plasma) Fluctuation in the densities by PIC/MC with LPWS do not increase significantly. Spatial variation of electric fields by the PIC/MC with agrees well with that by the standard PIC/MC. However, in the bulk small fluctuations are found. MURORAN INSTITUTE OF TECHNOLOGY Spatiotemporal variations of electron density 100 cycle 6 At 100 and 200 cycle, the electron density profiles by these methods do not agree, however, the profiles have a tendency to agree each other at 500 cycle. -3 electron density (cm ) 250x10 200 150 100 50 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 position (cm) 200 cycle -3 electron density (cm ) It is likely that the simulation using the PIC/MC with LPWS tends to reach convergence faster than the standard PIC/MC simulation. 6 250x10 200 150 100 50 0 0.0 It is found that the density profile by the PIC/MC with LPWS increases faster than that of the standard PIC/MC. 0.5 1.0 1.5 2.0 2.5 3.0 Similar tendency is reported by Ardehali et al. for the comparison between the SPLIT-PIC/MC and the standard PIC/MC simulation. (IEEE Trans. Plasma Sic., vol.25, pp.1081-5, 1997) position (cm) 500 cycle 6 electron density (cm -3 ) 300x10 250 200 Standard PIC/MC PIC/MC with LPWS DSMC (Date et al, 1992) Propagator 150 100 50 0 0.0 0.5 1.0 1.5 position (cm) 2.0 2.5 3.0 MURORAN INSTITUTE OF TECHNOLOGY Comparison with measured data Generation rate of excited molecules (C3u) density Experiment PIC/MC with LPWS 6.0x10 2 4.0x10 2 3 density of C u (a.u.) Spatiotemporal variation of excited molecule (C3u) density K 2.0x10 0.0 0 K 6.0x10 14 4.0x10 14 2.0x10 14 2 A A 0 A A 37 Ti m Gap ) (mm ) s) (n 74 0 s (n K 30 e e m Ti 37 excitation rate of C3u p=0.5Torr, P=60W, f=13.56MHz 0.0 3 K 2 74 1 0 tio pos i n (cm ) H.Itoh et al., Trans. of IEE Japan, Vol.121-A, pp.465-70, 2001 Large peaks are seen near the instantaneous cathode (K), and small peaks are also seen near the instantaneous anode (A). MURORAN INSTITUTE OF TECHNOLOGY Computation time method computation time (for 100 cycle) accuracy PIC/MC simulator 14.35 hours 〇 PIC/MC with LPWS simulator 14.70 hours ◎ DEC Alpha station 533MHz The increase of computation time ・・・ 2.4% It is entirely fair to say that accuracy of PIC/MC simulation is improved by LPWS without adding MC particles. MURORAN INSTITUTE OF TECHNOLOGY Conclusions A PIC/MC simulator coupled with LPWS is developed in this work. It is found that statistical fluctuations are reduced substantially by this simulator with the slight increase (2.4%) of computation time. It is also found that accuracy of particle simulation can be improved without adding MC particles by the developed simulator. MURORAN INSTITUTE OF TECHNOLOGY Direct Simulation Monte Carlo (DSMC) (Date et al, J.Phys.D, vol.25, pp.442-53, 1992) before after E Calculating electric field using Poisson’s eq. z n(z) n(z) z E slab slab z E e e e e e Powered electrode 4.5 4 f (vz,vr) Grounded electrode Powered electrode 4.5 3.5 4 3 3.5 2.5 3 2 2.5 1.5 f (vz,vr) 1 0.5 0 Grounded electrode 2 1.5 1 0.5 0 Free flight, collision and scattering in Dt are calculated by MC method. MURORAN INSTITUTE OF TECHNOLOGY Spatial variations of densities & electric field 60 grid points for charge assignment Net charge density electron density 1499.7 cycle 6 6 150x10 250x10 net charge density (cm ) 100 -3 -3 electron density (cm ) Standard PIC PIC with LPWS 200 150 100 50 0 0.5 1.0 1.5 2.0 2.5 3.0 50 0 -50 Standerd PIC PIC with LPWS Position [cm] -100 0.5 ion density 1.0 Standerd PIC PIC with LPWS 2.0 2.5 3.0 Position [cm] Electric field 6 250x10 1.5 100 -3 ion density (cm ) 200 electric field (V/cm) 150 100 50 0 0.5 1.0 1.5 2.0 2.5 3.0 50 0 -50 Position [cm] Standerd PIC PIC with LPWS -100 0.5 1.0 1.5 2.0 2.5 3.0 Position [cm] 1499.7 cycle MURORAN INSTITUTE OF TECHNOLOGY Comparison with other method (propagator) Propagator method p=0.5Torr, f=13.56MHz PIC/MC with LPWS 300 400 150 200 A A 100 75 0 0 -100 -75 A A 3 -200 3000.00 K 3000.25 3000.50 3000.75 time (cy 3001.00 0 cle) 2 ) (cm 1 n o i t i pos 0.0 18.5 37.0 time (n s) 3 K 55.5 electric field (V/cm) electric field (V/cm) 225 K K 300 -150 2 1 74.0 0 m) n (c o i t i pos Double layer is formed in the nitrogen rf glow discharge. Spatiotemporal variation of electric field by PIC/MC with LPWS agrees well with that by propagator method. MURORAN INSTITUTE OF TECHNOLOGY