Numerical Schemes for Streamer Discharges at Atmospheric Pressure Jean PAILLOL*, Delphine BESSIERES - University of Pau Anne BOURDON – CNRS EM2C Centrale Paris Pierre SEGUR – CNRS CPAT University of Toulouse Armelle MICHAU, Kahlid HASSOUNI - CNRS LIMHP Paris XIII Emmanuel MARODE – CNRS LPGP Paris XI STREAMER GROUP The Multiscale Nature of Spark Precursors and High Altitude Lightning Workshop May 9-13 – Leiden University - Nederland Outline • • • • • • • • Plasma equations Integration – Finite Volume Method Advection by second order schemes Limiters – TVD – Universal Limiter Higher order schemes – 3 and 5 – Quickest Numerical tests – advection Numerical tests – positive streamer Conclusion Equations in one spatial dimension 2D schemes for discharge simulation real 2D schemes 2D = 1D + 1D (splitting) Coupled continuity equations N e ( N eWe ) N e D S ( ) N e We N e N p 1 t x x x N p ( N pW p ) S N e We N e N p 1 N n N p 2 t x N n ( N nWn ) N e We N n N p 2 t x Poisson equation e div E ( N p N e N n ) 0 Advection equation – 1D N e ( N eWe ) N e D S ( ) N e We N e N p t x x x N e ( N eWe ) S' t x N e ( N eWe ) 0 t x S’ can be calculated apart (RK) N ( wN ) 0 t x N ( x, t ) ( f ( x, t )) 0 and t x f ( x, t ) w( x, t ) N ( x, t ) Outline • • • • • • • • Plasma equations Integration – Finite Volume Method Advection by second order schemes Limiters – TVD – Universal Limiter Higher order schemes – 3 and 5 – Quickest Numerical tests – advection Numerical tests – positive streamer Conclusion Finite Volume Discretization Computational cells t n+1 UPWIND n n-1 i-2 i-1 i-3/2 i i-1/2 i+1 i+1/2 Control Volume i+2 i+3/2 x Integration N ( x, t ) ( f ( x, t )) 0 t x and f ( x, t ) w( x, t ) N ( x, t ) Integration over the control volume : d xi1 / 2 N ( x, t )dx f ( xi 1/ 2 , t ) f ( xi 1/ 2 , t ) 0 dt xi1 / 2 Introducing a cell average of N(x,t): 1 N i (t ) xi 1/ 2 xi 1/ 2 xi 1 / 2 xi 1 / 2 N ( x, t )dx then : N i (t n 1 1 ) N i (t ) xi 1/ 2 xi 1/ 2 n t f ( x , t )dt t f ( x , t )dt i 1 / 2 i 1 / 2 t n t n n1 n1 Integration N ( x, t ) ( f ( x, t )) 0 t x and f ( x, t ) w( x, t ) N ( x, t ) Integration over the control volume : d xi1 / 2 N ( x, t )dx f ( xi 1/ 2 , t ) f ( xi 1/ 2 , t ) 0 dt xi1 / 2 Introducing a cell average of N(x,t): 1 N i (t ) xi 1/ 2 xi 1/ 2 xi 1 / 2 xi 1 / 2 N ( x, t )dx then : N i (t n 1 1 ) N i (t ) xi 1/ 2 xi 1/ 2 n t f ( x , t )dt t f ( x , t )dt i 1 / 2 i 1 / 2 t n t n n1 n1 Integration N ( x, t ) ( f ( x, t )) 0 t x and f ( x, t ) w( x, t ) N ( x, t ) Integration over the control volume : d xi1 / 2 N ( x, t )dx f ( xi 1/ 2 , t ) f ( xi 1/ 2 , t ) 0 dt xi1 / 2 Introducing a cell average of N(x,t): 1 N i (t ) xi 1/ 2 xi 1/ 2 xi 1 / 2 xi 1 / 2 N ( x, t )dx then : N i (t n 1 1 ) N i (t ) xi 1/ 2 xi 1/ 2 n t f ( x , t )dt t f ( x , t )dt i 1 / 2 i 1 / 2 t n t n n1 n1 Flux approximation How to compute t n1 t n f ( xi 1/ 2 , t )dt ? f ( xi 1/ 2 , t ) w( xi 1/ 2 , t ) N ( xi 1/ 2 , t ) ~ N ( x , t ) N i ( x, t ) Assuming that : w( xi 1/ 2 , t ) wi over ~ f ( xi 1/ 2 , t ) wi Ni ( xi 1/ 2 , t ) xi 1/ 2 , xi 1/ 2 (t n , t n 1 ) Flux approximation How to choose the approximated value 0th order 1st order xi-3/2 ~ Ni ( xi 1/ 2 , t ) ~ n Ni ( x, t n ) Ni ~ Ni ( x, t n ) Nin ( x xi ) in xi-1 xi-1/2 xi xi+1/2 Control Volume ? Linear approximation xi+1 xi+3/2 x Advect exactly tn+1 wi tn xi-3/2 t n1 tn xi-1 xi-1/2 xi t n1 f ( xi 1/ 2 , t )dt wi n t t n1 wi n t 1st order t n1 tn xi+1/2 xi+1 xi+3/2 x ~ N i ( xi 1/ 2 , t )dt ~n N i ( xi 1/ 2 wi (t t n )) dt xi 1 / 2 xi 1 / 2 wi dt ~n N i ( x)dx 2 dt f ( xi 1/ 2 , t )dt wi N in dt ( xi 1/ 2 xi ) in wi dt wi2 in 2 Update averages [LeVeque] 1 n 1 n N i (t ) N i (t ) dxi t n1 tn 1 t n1 f ( xi 1/ 2 , t )dt f ( xi 1/ 2 , t )dt n t dxi 1st order t n1 tn t n1 tn 2 dt f ( xi 1/ 2 , t )dt wi N in dt ( xi 1/ 2 xi ) in wi dt wi2 in 2 2 dt f ( xi 1/ 2 , t )dt wi 1 N in1dt ( xi 1/ 2 xi 1 ) in1wi 1dt wi21 in1 2 Note that : if dxi dx and wi w 2 2 n 1 dx w dt n 1 n n n n ( i i 1 ) N i N i wdt ( N i N i 1 ) w dt dx 2 2 Update averages [LeVeque] 1 n 1 n N i (t ) N i (t ) dxi t n1 tn 1 t n1 f ( xi 1/ 2 , t )dt f ( xi 1/ 2 , t )dt n t dxi 1st order t n1 tn t n1 tn 2 dt f ( xi 1/ 2 , t )dt wi N in dt ( xi 1/ 2 xi ) in wi dt wi2 in 2 2 dt f ( xi 1/ 2 , t )dt wi 1 N in1dt ( xi 1/ 2 xi 1 ) in1wi 1dt wi21 in1 2 Note that : if dxi dx and wi w 2 2 n 1 dx w dt n 1 n n n n ( i i 1 ) N i N i wdt ( N i N i 1 ) w dt dx 2 2 UPWIND scheme Update averages [LeVeque] 1 n 1 n N i (t ) N i (t ) dxi t n1 tn 1 t n1 f ( xi 1/ 2 , t )dt f ( xi 1/ 2 , t )dt n t dxi 1st order t n1 tn t n1 tn 2 dt f ( xi 1/ 2 , t )dt wi N in dt ( xi 1/ 2 xi ) in wi dt wi2 in 2 2 dt f ( xi 1/ 2 , t )dt wi 1 N in1dt ( xi 1/ 2 xi 1 ) in1wi 1dt wi21 in1 2 Note that : if dxi dx and wi w 2 2 n 1 dx w dt n 1 n n n n ( i i 1 ) N i N i wdt ( N i N i 1 ) w dt dx 2 2 UPWIND scheme Approximated slopes ~ Ni ( x, t n ) Nin ( x xi ) in 0 n i Upwind * n n N N i Lax-Wendroff ** in i 1 xi 1 xi ** Second order accurate * First order accurate xi-3/2 xi-1 xi-1/2 N in N in1 Beam-Warming ** xi xi 1 N in1 N in1 Fromm ** n i xi 1 xi 1 n i xi xi+1/2 xi+1 xi+3/2 x Numerical experiments [Toro] ntotal = 401 w dt 0.4 dx w Periodic boundary conditions After one advective period Upwind Beam-Warming Lax-Wendroff Fromm Outline • • • • • • • • Plasma equations Integration – Finite Volume Method Advection by second order schemes Limiters – TVD – Universal Limiter Higher order schemes – 3 and 5 – Quickest Numerical tests – advection Numerical tests – positive streamer Conclusion Slope Limiters 2 2 n 1 dx w dt n 1 n n n n ( i i 1 ) N i N i wdt ( N i N i 1 ) w dt dx 2 2 N n 1 i wdt wdt wdt n n n n N ( N i N i 1 ) 1 ( i i 1 ) dx 2 dx n i in ( in1/ 2 )( N in1 N in ) in1/ 2 N in N in1 n n N i 1 N i : correction factor Smoothness indicator near the right interface of the cell How to find limiters ? TVD Methods ● Motivation First order schemes poor resolution, entropy satisfying and non oscillatory solutions. Higher order schemes oscillatory solutions at discontinuities. ● Good criterion to design “high order” oscillation free schemes is based on the Total Variation of the solution. ● Total Variation of the discrete solution : TV ( N n ) N in1 N in i ● Total Variation of the exact solution is non-increasing TVD schemes TV ( N n1 ) TV ( N n ) Total Variation Diminishing Schemes TVD Methods ● Godunov’s theorem : No second or higher order accurate constant coefficient (linear) scheme can be TVD higher order TVD schemes must be nonlinear. ● Harten’s theorem : 0 ( ) min mod( 2,2 ) 0 ( ) 0 upwind ( ) 1 Lax Wendroff ( ) Beam Warmin g 1 ( ) Fromm 2 TVD region TVD Methods ● Sweby’s suggestion : 2nd order Avoid excessive compression of solutions 2nd order Second order TVD schemes minmod ( ) min mod(1, ) ( ) max( 0, min( 1,2 ), min( 2, )) superbee 1 ( ) max( 0, min( ,2,2 )) Woodward 2 ( ) 1 Van Leer After one advective period minmod Van Leer Woodward superbee Universal Limiter [Leonard] N n 1 i wi 1/ 2 dt H n wi 1/ 2 dt H n N N i 1/ 2 N i 1/ 2 dxi dxi 1 n i N iH1n/ 2 High order solution to be limited tn Ni+1 Ni+1/2 Ni NF Ni-1 ND NC NU xi-3/2 xi-1 xi-1/2 xi xi+1/2 Control Volume xi+1 xi+3/2 x After one advective period Fromm method associated with the universal limiter Outline • • • • • • • • Plasma equations Integration – Finite Volume Method Advection by second order schemes Limiters – TVD – Universal Limiter Higher order schemes – 3 and 5 – Quickest Numerical tests – advection Numerical tests – positive streamer Conclusion Advect exactly Finite Volume Discretization 1 N i (t n 1 ) N i (t n ) dxi t n1 t n t f ( x , t )dt t f ( x , t )dt i 1 / 2 i 1 / 2 t n t n n 1 n 1 t n1 f ( xi 1/ 2 , t )dt wi n N ( xi 1/ 2 , t )dt t xi 1 / 2 xi 1 / 2 wi dt N n ( x)dx tn+1 wi tn xi-3/2 xi-1 xi-1/2 xi xi+1/2 xi+1 xi+3/2x Integration [Leonard] d ( x ) N ( x) dx Assuming that is known : 1 dxi t n1 tn 1 f ( xi 1/ 2 , t )dt dxi N i (t n 1 t n1 tn 1 N i ( xi 1/ 2 , t )dt dxi i x i 1 / 2 wi 1 / 2 dt dxi ) N i (t ) n d ( x) xi1/ 2 wi1/ 2dt dx dx xi 1 / 2 i i * dxi i i * i 1 i 1 * dxi dxi 1 High order approximation of * Y function is determined at the boundaries of the control cell by numerical integration Yi+1 Yi Yi-1 tn Y i* Yi-2 dt.wi xi-2 xi+3/2 xi-3/2 xi-1 xi-1/2 xi xi+1/2 xi+1 Control Volume Polynomial interpolation of (x) Y i* x High order approximation of * * is determined by polynomial interpolation Polynomial order Interpolation points Numerical scheme 1 i-1 i UPWIND 2 i-1 i i+1 3 i-2 i-1 i i+1 5 i-3 i-2 i-1 i i+1 i+2 …… …… Lax-Wendroff 2nd order QUICKEST 3 (Leonard) 3rd order QUICKEST 5 (Leonard) 5th order …… Universal Limiter applied to * [Leonard] (x) is a continuously increasing function (monotone) Yi+1 dt.wi Y i* xi-3/2 Yi Yi-1 Yi-2 xi-2 xi+3/2 tn xi-1 xi-1/2 xi xi+1/2 xi+1 x Outline • • • • • • • • Plasma equations Integration – Finite Volume Method Advection by second order schemes Limiters – TVD – Universal Limiter Higher order schemes – 3 and 5 – Quickest Numerical tests – advection Numerical tests – positive streamer Conclusion Numerical advection tests ● Ncell = 401, after 5 periods ● Ncell = 401, after 500 periods MUSCL superbee QUICKEST 3 MUSCL Woodward QUICKEST 5 Ncell = 1601, after 500 periods MUSCL superbee MUSCL Woodward QUICKEST 3 QUICKEST 5 Celerity depending on the x axis ~ N ( x , t ) N i ( x, t ) Celerity w( xi 1/ 2 , t ) wi over xi 1/ 2 , xi 1/ 2 (t n , t n 1 ) x Celerity depending on the x axis ~ N ( x , t ) N i ( x, t ) Celerity w( xi 1/ 2 , t ) wi over xi 1/ 2 , xi 1/ 2 (t n , t n 1 ) x Celerity depending on the x axis ~ N ( x , t ) N i ( x, t ) Celerity w( xi 1/ 2 , t ) wi over x xi 1/ 2 , xi 1/ 2 (t n , t n 1 ) Quickest 5 Quickest 3 After 500 periods Woodward Initial profile x Outline • • • • • • • • Plasma equations Integration – Finite Volume Method Advection by second order schemes Limiters – TVD – Universal Limiter Higher order schemes – 3 and 5 – Quickest Numerical tests – advection Numerical tests – positive streamer Conclusion Positive streamer propagation Plan to plan electrode system [Dahli and Williams] streamer Cathode x=0 Initial electron density Anode x=1cm 1014cm-3 108cm-3 x=0 x=1cm x=0.9cm E=52kV/cm radius = 200µm ncell=1200 Positive streamer propagation Charge density (C) 2ns x=0 Zoom UPWIND x=1cm Positive streamer propagation Charge density (C) 2ns x=0 Zoom UPWIND x=1cm Charge density (C) 4ns Woodward Quickest minmod superbee Zoom Conclusion Is it worth working on accurate scheme for streamer modelling ? YES ! especially in 2D numerical simulations Advection tests Quickest 5 Quickest 3 TVD minmod Error (%) 0.78 3.8 3.41 26.5 22.77 Number of cells 1601 401 1601 201 1601 High order schemes may be useful
© Copyright 2026 Paperzz