1 $. Heuraux € AE Fusion INRIA Paris
Reflectometry Simulations : Choices to do, and applications to current issues
S. Heuraux$, F. da Silva£, E. Gusakov€, A. Popov€, N. Kosolapova€, K. Syisoeva€, F. Clairet¥, R. Sabot¥
$Institut Jean Lamour-Faculté des Sciences-Nancy-Université UMR 7198 CNRS, BP 70239 F-54506 Vandoeuvre cedex ( [email protected] ) €Ioffé Institute Politekhnicheskaya 26, 194021 St.Petersburg, Russia £Associação EURATOM/IST IPFN, Instituto Superior Técnico, 1046-001 Lisboa Portugal ¥Association Euratom-CEA_Cadarache 13108 St Paul-lez-Durance – France
Schedule: -recall of basic mechanisms involved in turbulence characterisation, -how to choice a relevant set of equations for a given simulation, -current issues: stochastic against Bragg backscattering processes,
beam spreading - impact on the measurements, new concepts of diagnostics, …
conclusions and status of the ERCC 3D-code and Thanks to INRIA for the invitation to this Fusion Summer School
2 $. Heuraux € AE Fusion INRIA Paris
Principle
source
detector
ν(t)
φ(t)
A(t) [φ(t)] ν(t'), φ(t')
[ν1; ν2] (∂ν /∂t) ν [8-155 GHz] λο [mm-cm]
Heuraux et al., Rev. Sci. I (2003) 74, 1501, L. Vermare et al, Plasma Phys Cont Fusion 47, 1895 (2005) T. Gerbaud et al 77, 10E928 (2006), S. Heuraux et al Aims J. Discrete and Cont. Dyn. Sys. - Series S to be published.
Reflectometry : basic principles (1)
δφ
Experimental data from Tore Supra Reflectomter
!(" ) =!WKB (" )+#!(" ,E2 (" ,#n))
A(! ) = AWKB (! )+"A(! ,E2 (! ,"n))
Unperturbed density profile
δn density fluctuations
or !(" )
or A(! )
3 $. Heuraux € AE Fusion INRIA Paris
Reflectometry : basic principles (2)
!"(!n(# ),E 2 (# ,!n),!V,N(# )) = !"source +
!"N +!"!n +!"E2+!" !V +!"!T +!"!
!B +....
In experiments , what should be taken into account ?
!"source
!"N
!"!n
!"E2
!" !V!"!T
!"!!B
+.... Mode conversion, cross-polarisation, multi-reflections,
Hardware noise, thermal phase shift of hardware, plasma emission, … Macroscopic variations of the background plasma parameters, Tranverse gradients effects, Doppler shift if time dependent, Refraction effects, scattering contribution (for- and back-ward)
Damping of the probing wave , resonances (wave trapping), …
Plasma motion inducing Doppler shift => phase shift by n variations, …
Through relativistic corrections on the e- mass, thermal effects …
Magnetic fluctuations -> scattering, magnetic shear effects, …
In addition Numerical errors if a code is used to compute transfer function !
4 $. Heuraux € AE Fusion INRIA Paris
Bragg BackScattering rule
κf
κi κs
Basic mechanisms involved in turbulence characterisation
!"!t
=!F!t
N(F, s)ds+!sed
sc! F !!t
N(F, s)dssed
sc!"#$%&'
fixed probing frequency
!
"F"t
= 0
fast sweep probing frequency
!
"N"t
= 0
(BBS)
! i = 2k(xr )
Measurements based on phase variations
Fluctuations reflectometry or frequency hopping systems Radial and poloidal correlation reflecto.
Density profile reflectometry
Forward scattering Doppler shift (dn motion)
F. Da Silva et al Nuclear Fusion 46, S816 (2006); F. Da Silva et al & A. Popov et al IRW8 reflectometry meeting 2-4 May 2007
Reflection at cut-off layer
NO,X (xc)= 0
Amplitude modulation
Destructive interference
Doppler shift n(t)
5 $. Heuraux € AE Fusion INRIA Paris
!
< "# 2 > = $ k02 LN 2
% f eff
< "n2 >nc
2
Basic mechanisms involved in turbulence characterisation
More efficient
Closer to the cut-off
Gusakov et al. PPCF 44 (2002) 2327 Schubert et al IRW9 Lisbon (2009)
!
< "# 2 > = $ k02 dx G(x)
x edge
x C (k o )
% < "n2(x) >nc
2
Inhomogeneous turbulence case These methods assume E2(r) known
=> New method for the density fluctuation profile reconstruction
Fanack et al PPCF 38 (1996) 1915
6 $. Heuraux € AE Fusion INRIA Paris
• Assumptions versus set of equations describing wave propagation
A. Casiti et al PRL 2009
7 $. Heuraux € AE Fusion INRIA Paris
Reflectometry simulations : How to do it ? (1)
First question : how to choose the relevant equation for the simulation ? need to know exactly the physical system to simulate that is to say what are the relevant assumptions for this system?
A proposal to define it corresponds to a set of equations:
time dependent or not ?
refraction effects only ?
doppler effects there or not?
single polarisation description or not?
Ray-Beam tracing, Helmholtz Wave Eq, Maxwell' Eqs + plasma dynamics
Ray-Beam tracing
Ray-Beam tracing, Helmholtz Wave Eq, Maxwell' Eqs + plas. dyn.
Single description
Coupled equations Coupling to model
8 $. Heuraux € AE Fusion INRIA Paris
From ray tracing to wave equation (1) Ray tracing
Single mode description D(w,k,r,t)=0
Set of coupled Odes to solve
!
"! r "#
= $"D(%,
! k ,! r ,t)
"! k
"! k "#
= "D(%,! k ,! r ,t)
"! r
&
' ( (
) ( (
!
"t"#
="D($,
! k ,! r ,t)
"$"$"#
= %"D($,
! k ,! r ,t)"t
&
' ( (
) ( (
Can be extended to Gaussian beam propagation by one ODE associated to amplitude or using stationnary phase method
Numerical Tools needed for ITER plasma position studies
Quasi-optic description without scattering
RungeKutta 4th order
Hyp. WKB : !!!
!!!dk
dx « k2, !!!!
!!!!d2k
dx2 « !!!
!!!dk
dx k
D. G. Swanson "Plasma Waves", 2nd Ed IoP 2003,ch6.5, ISBN 0 7503 0927 X
G. V. Pereverzev Phys. Plasmas 5, 3529 (1998) R.A. Cairns, V. Fuchs Nucl. Fusion 50 (2010) 095001
A. Richardson, P. Bonoli, J. Wright, PoP 17 (2010) 052107 Or eikonal method with wavepacket for the amplitude description (as quantum phy.)
9 $. Heuraux € AE Fusion INRIA Paris
From ray tracing to wave equation (2)
Helmholtz's equation (full-wave)
Hyp: monochromatic wave, steady state plasma (∆t or lcorr >> 4rc/c)
Single mode description: Computation of the index N(r)
!
"! E + N 2(! r )
! E = 0
Be careful in multi dimensional case, possible cross derivatives more complicated to solve especially for X-mode No Doppler O-mode other method see L-M. Imbert this conférence
Monochromatic and single polarisation probing system
Finite Difference 4th order (Numerov)
C. Fanack, PhD Thesis or et al PPCF 38, 1915 (1996) S. Heuraux, F. da Silva DCDS_S 5, 307 (2012)
10 $. Heuraux € AE Fusion INRIA Paris
Finite Element Method
Actually only few developments on FEM with dispersive media:
In plasma only using equivalent dielectric (Ph Lamalle for ICRH or F. Braun & L. Colas) for ICRH (HFSS or COMSOL multi-Physics) including boundary sheath conditions L. Colas, D. Milanesio, E. Faudot#, M. Goniche, A. Loarte J. N. Mat 390-391 (2009) 959-962, O. Meneghini, S. Shiraiwa, R. Parker PoP, 16 (2009) 090701 Accurate method in vacuum and in complex geometry (commercial software) ALCYON was ICRH code based on functionals, if will be replaced by EVE code developed by R. Dumont (CEA_cadarache) and needs a lot of memory (~10-20 Gbytes) In the case of reflectometry possible ? Yes M. Irzak, et al Nuc Fus 35 1341 (1995)
Monochromatic multi-polarisation probing system
EVE R. Dumont Nuc Fus 49 075033 (2009)
seeTalk R. Dumont friday
Resonances generated by Bragg resonant perturbation with ≠ sources
11 $. Heuraux € AE Fusion INRIA Paris
From ray tracing to wave equation (3)
Shrödinger like's equation (full-wave)
Hyp: quasi-monochromatic wave quasi steady state plasma (∆t or lcorr >> 4rc/c)
Single mode description: Computation of the index N(r)
!
i"t
! E + #
! E + N 2(! r )
! E = 0
Lin et al, Plasma Phys. Cont. Fusion 40 L1 (2001)
!
" >> #t
Restriction on dispersion effects, Quasi-paraxial approximation
Quasi steady state plasma
Parabolic
12 $. Heuraux € AE Fusion INRIA Paris
From ray tracing to wave equation (4)
wave equation (quasi-steady state medium) Hyp: (tf, ∆t or lcorr >> 4rc/c)
!
"t2! E # c 2$
! E +% pe
2 (! r )! E = 0
O-mode or isotrpic plasma
Set of coupled partial differential equations associated to X-mode
Time dependent physical processes or probing system
Hacquin et al, J. of Computational Physics 174, 1 (2001),
Cohen et al, Plas. Phys. Cont Fusion 40, 75 (1998),
V = V/VD where VD=Eo/Bo
and E= E/Eo
Finite Difference + wp
2E rewritting
[!t2 " c2 (!x
2 +!y2 )+! pe
2 ]Ex + c2!x#.!
!E = "! pe
2 vyBo / c
[!t2 " c2 (!x
2 +!y2 )+! pe
2 ]Ey + c2!y#.!
!E =! pe
2 vxBo / c
me!tvx = "e(Ex + vyBo )
me!tvy = "e(Ey " vxBo )
$
%
&&&
'
&&&
13 $. Heuraux € AE Fusion INRIA Paris
From ray tracing to wave equation (5)
wave equation (time-depend medium)
Hyp: single mode polarisation
!t2!E ! c2"
!E +" pe
2 (!r, t)!E = e
#o
!v!tn
!t!v = ! e
me
!E
#
$
%%
&
%%
Just to add ∂tn in the set of coupled partial differential equations associated to X-mode
O-mode or isotrpic plasma
Fast gradient motion, up or down frequency shift amplitude variation
Frequency upshift with ∂tn
Turbulence dynamics, fast events
Finite Difference + wp
2E rewritting + RK45
J. Mendoça, New Journal of Phys.,11(2009) 013029
14 $. Heuraux € AE Fusion INRIA Paris
Cross polarisation simulations
!t2Ez ! c
2!x2Ez +" pe
2 (x, t)Ez =COX Ex,Ey( )!t2Ex +" pe
2 (x, t)Ex = !" pe2 (x, t)vy +CXOx Ez( )
!t2Ey ! c
2!x2Ey +" pe
2 (x, t)Ey =" pe2 (x, t)vx +CXOy Ez( )
!t!v = ! e
me
!E ! e
me
!v "!B
#
$
%%%
&
%%%
1D Case: O-mode and X-mode
O-mode X-mode
dB measurements
Finite difference
N. Katsuragawa, H. Hojo, A. Mase J. Phys. Soc. Jpn. 67 (1998) 2574-2577
©€
15 $. Heuraux € AE Fusion INRIA Paris
Full description: Maxwell's equations
Hyp: linear response of the plasma
!
".! B = 0
".! E = #
$o
" %! E = &'
! B 't
" %! B = µo
! j + 1
c 2'! E 't
(
)
* * *
+
* * *
r total density of charges j current density Associated model fluid or kinetic
Radial direction
60 GHz
Polo
idal
dir
ecti
on
F. da Silva et al , J Plasma Phys. 72 1205 (2006), and Rev. Sci Instr. 79, 10F104 (2008) C. Lechte, IEEE TPS 37 (2009) 1099.
TE and TM modes are usually treated separately
Velocity field mapping, Shear layer detection
Yee's algorithm + J solver
x/lo
50 cm
Eqs Maxwell Eqs + Vlasov or PIC, Required too much computation time for reflectometry simulations
16 $. Heuraux € AE Fusion INRIA Paris
On the role of E2 spatial distribution
17 $. Heuraux € AE Fusion INRIA Paris
The amplification factor of the scattered wave (blue dots or black stars), the squared amplification factor of the probing wave in the cavity (red dots) and the analytical eq. versus the phase F ofδnq
Heuraux IEEE trans Plasmas Sci. 38 (2010) 2150 Gusakov et al. Phys. Cont. Fusion 51 (2009) 065018
Enhanced scattering induced by a resonant cavity
δnf/nc= 3.3·10-5 δnf/nc= 1.6·10-3
δnf
δnq
enhanced scattering factor Ees/Es
amplification factor of the probing wave
Eres/Enp
18 $. Heuraux € AE Fusion INRIA Paris E. Gusakov, S. Heuraux et al Physica Scripta 84 (2011) 04504.
X
Z
Gaussian probing beam
Radial position of the Bragg resonant dn
1
-ao ao 0
Enhanced scattering induced by a resonant cavity(2) In the previous case, evolution of the
scattered field versus dn/nc with and without resonance
Modification of the probing field structure
Role of E2 distribution
19 $. Heuraux € AE Fusion INRIA Paris
Competition between resonant mechanism and Bragg back scattering in turbulent plasma
Density gradient length L= 160 lo, dntur/nc= 5 %
E/Eo
x(pts)
E2 ≠ Cte
20 $. Heuraux € AE Fusion INRIA Paris
Density gradient length L= 160 lo, homogneneous turbulence dntur with cut-off
Competition between resonant mechanism and Bragg back scattering (1) Reflectometry case
dntur/nc= 5 % dntur/nc= 10 %
Average value (over 104 samples) of electromagnetic flux of probing wave
Resonances win and permit to the probing wave to reach the cut-off
Bragg BackScattering wins and reduce the probing depth
Gusakov et al. plasma phys EPS conf dublin 2010 Heuraux et al Cont Plasma Phys in press
The theory is based on a photon diffusion equation only valid at moderate fluctuation level
<E2(r)> vg ~ Cte
<E2(r)> vg ≠ Cte
21 $. Heuraux € AE Fusion INRIA Paris
2000 4000 6000 8000 10000
-4
-2
0
2
4
61
2000 4000 6000 8000 10000
0.2
0.4
0.6
0.8
1
2000 4000 6000 8000 10000
0.2
0.4
0.6
0.8
1
Removing the resonances Taking into the resonances
Simulations with experimental density fluctuation profile
Good news while the path in high turbulence level stays small
22 $. Heuraux € AE Fusion INRIA Paris
Spectrum effects on the probing field structure
0 2000 4000 6000 8000 100000.0
0.5
1.0
1.5
2.0
Ln= 1000 lo for dntur/nc=1% with nmax = 0.8 nc and the same width for the density fluctuations (~75lo) Wavenumbers are resonant from the middle to the top of the density profile
<E2 RM
S/E2
io>
r/lo
0 2000 4000 6000 8000 10000�3
�2
�1
0
1
2
32BBS zone
<E2(r)> vg
T and R are a function of S(kf) and can change at dntur/nc= Cte
23 $. Heuraux € AE Fusion INRIA Paris
Studies on forward scattering
24 $. Heuraux € AE Fusion INRIA Paris
Simulation of Forward scattering
Probing frequency 40 GHz Spectrum without Bragg resonant wavenumber
Spectrum of the backscattered signal frequency shift
backscattered signal simulated
25 $. Heuraux € AE Fusion INRIA Paris
!
"# = 2kpolVpol$ (%Q)2
2 +$ (%Q)2$ &'( & ( ( "# ) kpolVpol
Theory and simulation in agreement more or less for ∆w
2cyl Q= r radius of the beam
2 2
2 28ln 1
2
0.577
cc cx
cxc
n xx l
lc n! " #
$ %%
#
& '( )= * ++ ,- .
/ 01 23
But a factor of the order of two exist for the scattered power and at high level of density fluctuations
Why ?
26 $. Heuraux € AE Fusion INRIA Paris
Reflectometry Computation Requirements
To describe the forward scattering effects (long wavelength contribution)
To recover the theoretical results of the forward scattered power much larger mesh size is required
Usefulness of the testing of the code by using analytical results Be careful on the choice of the turbulence generator: modes summation, burst superposition, …. or coming from turbulence code BUT has intrinsic limitations (few grid points)
2π
0.02
Gives more than a factor of 2 to be verified in 2D but the needs for computations are very high (underway)
Gaussian spectrum
Ln=0.58 m n =47 GHz lc = 7.5 mm
dn/nc=0.001
27 $. Heuraux € AE Fusion INRIA Paris
Direct evaluation of the beam width using averaging in time and samples(8) Cross-Correlation length of E(y)E(y')
a=20°
a=20°
a=20°
Other parameters to characterize forward scattering effects
28 $. Heuraux € AE Fusion INRIA Paris
Power spectrum using sliding FFT Max of the peaks of Sliding FFT
Scattered signal behaviour for pure forward scattering (Gaussian S(k))
Simulation help to know more
29 $. Heuraux € AE Fusion INRIA Paris
Frequency power spectrum using sliding FFT Max of the peaks of Sliding FFT
Scattered signal behaviour with Bragg backscattering Kolmogorov spectrum
With Bragg backscattering
30 $. Heuraux € AE Fusion INRIA Paris
Electric field structure at a given time in purely forward scattering and Kolmogorov cases for Doppler reflectometry on O-mode
In the previous conditions with dn/nc= 10% (RMS)
To see the Doppler shift a long time series is needed ~ 8 105 time steps Left correct simulation but right some scattering contribution are cut Bigger mesh is required for the right field structure (size should be 3000x5000 pts)
31 $. Heuraux € AE Fusion INRIA Paris
2D wave propagation in moving turbulent plasma
wave trapping
local resonance
multi-scattering
beam spreading
How to obtain local resonances ?
turbulence motion inducing Doppler shift
access to Vtur velocity
32 $. Heuraux € AE Fusion INRIA Paris
Conclusions Depending on the Physics you wont to describe:
choose the adapted set of equations For an exact interpretation of the measurements, the E2 should be known especially if resonances exist or make an average to recover <E2>, which looks like a unperturbed case ( photon diffusion equation), valid up to moderate dn/nc Reflectometry Simulation permits to identify the basic mechanisms involved in a given configuration of the reflectometer. Simulation is an essential tool to improve the interpretation of the reflectometry data but also the theoretical studies (role of resonances,…) Theory-simulations are in agreement for each single mechanism (resonance, multi-scattering) However these agreements require to well define the numerical conditions to obtain it, which is not trivial or reachable especially for 2D and 3D cases. So we need help to improve numerical methods to solve PDEs associated to wave propagation
33 $. Heuraux € AE Fusion INRIA Paris
ERCC activities on 3D-Code in EFDA ITM framework
G.D.Conway (IPP), E.Blanco (CIEMAT), S.Hacquin (CEA), S.Heuraux(IJL),
C.Lechte (IPF), A.Sirinelli (CCFE), F.da Silva (IST), S.Soldatov(FZJ)
ERCC European reflectometry code consortium ITM integrated tokamak modeling
EFDA European fusion development agreement
34 $. Heuraux € AE Fusion INRIA Paris
ERCC: 3D full-wave code status!
Source"
Antenna"
Vacuum" Plasma"pol"
rad"tor"
20 - 50cm" 10-15cm"
20-30cm"
Emitting"plane"
3D cartesian grid"
FDTD comp. domain: l/20 grid"
Large physical domain forces use of mixed scheme = source + vac + FDTD "ne(r)"
20cm"
Cutoff"+ turb."
fo = 10 – 180 GHz ® lo ~ 3.0 - 0.17cm"
35 $. Heuraux € AE Fusion INRIA Paris
ERCC: erc3d code module status!
Kernel"
Source"
Profiles"
Turbulence"
Initialize stage!
Antenna CPO"
Edge ne CPO"
Receiver/ IQ"
Equi CPO"
Turb. CPO"
Loop stage! Output stage!
Source step"
Data CPO"
HDF5 files"
Vacuum step"
Turb. step"
Kernel step"
Vac. back-step"
Dectection "& IQ step"
HDF5 files"
Interpolation of profiles &"turb onto siml 3D grid"
36 $. Heuraux € AE Fusion INRIA Paris
ERCC: erc3d code – example results!
Input density profile from “coreprof” CPO"
Cutoff"
Ey field "– xy plane"
Ey field "– xz plane"
x (radial) in grid points"
Launch fo = 70.245 GHz " ∆x = l/20 = 0.2134 mm" Gaussian beam: wo ~ 4 mm" Grid size: 240 x 240 x 240 points"
37 $. Heuraux € AE Fusion INRIA Paris
Critical issues! Large domain: e.g. 32GB RAM ® 6 x108 grid points ® 13 field components
SP= 17cm cube grid! Need lots of memory!
Time: 1 CPU = 6000 hrs for 2048 snap shots (extrapolated from 2D code) ® Need lots of CPUs!
Parallelization: “snap-shot” (easy) but... “domain” (hard but necessary) Need expertise/manpower
Next step
Validation and verification against both 2D code & experimental data
Synthetic turbulence – soon.
Numerical turbulence coming from turbulence code need effort for data exchange
Any help are welcomed, thank you.
ERCC: erc3d code – numerical requirements and near future!