HAL Id: jpa-00210638
https://hal.archives-ouvertes.fr/jpa-00210638
Submitted on 1 Jan 1987
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of
sci-entific research documents, whether they are
pub-lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Low-frequency electric microfield calculations by
iterative methods
B. Held, P. Pignolet
To cite this version:
B. Held, P. Pignolet. Low-frequency electric microfield calculations by iterative methods. Journal de
Physique, 1987, 48 (11), pp.1951-1961. �10.1051/jphys:0198700480110195100�. �jpa-00210638�
Low-frequency
electric microfield calculations
by
iterative methods
B. Held and P.
Pignolet
Laboratoire
d’Electronique
des Gaz et desPlasmas, I.U.R.S.,
Université dePau,
avenue del’Université,
64000 Pau, France
(Reçu
le 15 décembre 1986, révisé le 29 mai1987, accept6
le 1erjuillet 1987)
Résumé. - Dans cet
article,
une méthode itérative utilisant larègle
du second moment estproposée
pour lecalcul de la composante basse
fréquence
dumicrochamp
en unpoint chargé.
Les résultats, en bon accord avec ceux de travaux récents, sont déduits d’une simulationnumérique particulièrement simple.
Cette méthodepermet
d’entreprendre
des calculs pour desmélanges
d’ions avec des temps de calcul nonprohibitifs.
Abstract. - An iterative method
using
the second moment rule isproposed
for the calculation of thelow-frequency
component of the microfield at acharged point.
Theresults,
ingood
agreement with other recentworks,
are deduced from a useful andsimple
numerical simulation. This methodpermits
extensive calculations for ionic mixtures withoutprohibitive
calculation times.Classification
Physics
Abstracts 52.701. Introduction.
During
the last few years, because of the Starkbroadering diagnostic
ofstripped
ions immersed in ionicperturbers,
many papers have been devoted tothe microfield calculations
[1]
(and
referencesgiven
there), [2-15].
The
density
andtemperature
effects are ingeneral
introduced into themicrofield,
using
the correlationfunction
9 (y).
In thepast,
9 (y)
was deduced from theDebye-Huckel
potential
with or withoutlineari-zation
[1].
Recent studiesimprove
thismethod,
introducing
a more accurate function9 (y) [6]
orpropose an all-order resummation of the
Baranger-Mozer series
[9-15].
In a
previous
paper[16]
hereafter referred to asI,
weproposed
ageneral
semi-empirical
expression
forthe correlation function
9 (y)
in theone-component
and the two-ioniccomponent
plasma
cases.The purpose of this paper is to
present
an iterative method for the microfieldcalculations,
taking
advan-tage
of correlation functions deduced from I.The calculations are
performed
at acharged point
(Za e )
for the lowfrequency
component
of the microfieldH({3), using
the second moment rule as aconstraint for the resummation of the
Baranger-Mozer series from the second order term to
infinity.
This numerical simulation
permits
extensivecalcu-lations for ionic mixtures without
prohibitive
calcula-tion time.In section
2,
the formalism ispresented
for the microfield and the second moment rule in the case oftwo-component
plasma.
Section 3 is devoted to
applications
of theformal-ism where useful
expressions
aregiven
for theparameters,
the correlationfunctions,
the low fre-quencycomponent
of the microfield and the secondmoment rule used as the
convergent
criterion for thecomputational
method.The results are
presented
in section 4 andcom-pared
with other theories andfinally,
section 5gives
general
conclusions.2. Formalism.
2.1 GENERAL FORMALISM. - We consider
a
plasma
with two ionic
components
(a, b).
If N is the totalnumber of ions
(Na,
Nb )
and E the total microfieldat the
origin
0,
it is convenient to consider theFourier transform of the
probability
distributionW(E)
[1] :
Introducing
theprobability
of occurrence of agiven
configuration
of the Nparticles,
it ispossible,
by
a clusterexpansion
method[17-18],
to expressF (k)
with adevelopment
ofintegral
functions ofcorrelation functions
g (r) [1].
Finally,
theprobability
distributionW(E)
is ob-tainedby
inverting
the Fourier transform:The main difficulties which appear with this method are the
knowledge
of the correlationfunc-tion g
(r)
and the resummation over all the orders forthe
development
ofF (k) :
thesepoints
areparticu-larly
important
forincreasing
values of theplasma
parameter,
i.e. forincreasing
values of electronicdensity
orhighly stripped
ions(fusion applications).
In
I,
we introduced ageneral semi-empirical
expression
for the correlation function. We takeadvantage
of theseresults,
introducing
accurateg (r)
functions in thegeneral
formalism. The firstorder terms of
F (k )
are calculatedclassically
[1]
and we introduce a corrective function to serve as asubstitute for the resummation over all the other
orders.
Assuming
that this correction isproportional
to the first order terms, the coefficient of
proportion-ality
is determinedby
an iterative methodusing
the second moment rule(Sect. 3).
2.2 SECOND MOMENT RULE. - The second
moment
rule is a useful constraint for the microfield
calcula-tions
[19-22].
If V is the totalpotential
energy of thesystem
andZa e
thecharge
of the centralion,
thesecond moment can be written in the form
[19] :
where
Vo
is thegradient
withrespect
to theposition
of the central ionic
charge.
Using
the non-linearized Poisson-Boltzmannequation,
it ispossible
to introduce the correlationfunction in the second moment rule
expression :
where
ne, na,
andnb
are the mean values for theelectronic and ionic
densities,
uaa (r)
anduab (r)
being
thetwo-body
potential
functions.In first
approximation,
we assume that thetwo-body
potential
functionsuaa (r )
anduab (r )
aregiven
by
theDebye-Hfckel
screenedpotential,
thisnon-linear formulation
(4) taking
care of morecoupling
effects.
Finally,
the second moment rule can beexplicitly expressed using
theappropriate
correlationfunction
g (r ).
3. General results.
3.1 PARAMETERS. -
Assuming
that theplasma
iscomposed
of two ioniccomponents
(a, b)
and anelectronic one, the
description
of thisplasma
can begiven
by
fourparameters
[1] :
- Correlation :
where re
is the mean inter-electroniclength
andAj)
theDebye
length
for the electrons(Appen-dix A) ;
-
Proportion :
where
C a (C b)
is the concentration ofions ;
- Number
of charges
on ions a and b :Za,
Zb.
Under these
conditions,
the two-ionicplasma
parameter
r[16]
can be written(Appendix A) :
where
Z
is the average ioniccharge,
Z
the root meansquare ionic
charge
(Appendix A)
andF,
theelec-tronic
plasma
parameter :
3.2 CORRELATION FUNCTION. - In
I,
thecorrela-tion funccorrela-tion
g (r)
is deduced from asemi-empirical
method,
using
physical
and numerical constraints.These
results,
ingood
agreement
with Monte-Carlosimulations,
are introduced in thegeneral
formalism,
the
general
expressions
for a two-ioniccomponent
plasma
being given
by
[16] :
The
screening
potentials Haa (y )
andHab (y )
arededuced from the
screening potentials
at theorigin
[16] :
where y is the reduced
length
(Appendix A)
andf(y)
theoscillating
function :with
the
screening potentials
at theorigin
Hij(o),
theplasma
screening
length
d and theparameters
A’(T), B’(T), C’(T) being
defined in I.3.3 MICROFIELD. - General
expressions
for themicrofield are deduced from the
general
formalismintroducing :
- the reduced microfield :
with
-
a new variable for the
expression
ofF (k) :
with
With this
notation,
ifH(13 )
is theprobability
distribution of the scalar-microfield
/3,
it ispossible
to write :
where
F (u )
isgiven
by :
with
1/’ fa
and1/’ fb
being
given
by [1] :
where
jo
is thespherical
Bessel function of orderzero and
x
being
a reducedlength
defined inAppendix
A.The
asymptotic
behaviour ofH(Q )
can begiven
with an
analytical expression [1] :
Finally,
using
parameters
introduced in section3.1,
we find(Appendix A) :
and the
asymptotic
form(for
high
values of(3) :
where ya and Yb are
given
inAppendix
A as :3.4 SECOND MOMENT RULE. - The second moment
rule can be
expressed
using
the definition of thereduced microfield and the
parameters
(Appen-dix
A).
For atwo-component
plasma :
the
one-component
plasma
resultbeing
obtained in the limit p = 0.Taking,
for thetwo-body
potential
functionu(r)
the screenedDebye-Huckel potential:
it is easy to see that :
with
It is of interest to note
that,
for aone-component
plasma
without correlation(Ae --+ 0),
the limit isgiven by [19] :
3.5 ITERATIVE METHOD FOR THE MICROFIELD CAL-CULATIONS. - The microfield calculation
is per-formed
using
thegeneral
formalism results for thefirst order terms in the
development
ofF (u )
(Eqs. (33)
and(34)).
Thehigher
order terms arecalculated with the second moment rule
assuming
that
(22) :
where
k1/l’1
1 is the corrective term.In
practice,
the microfield is calculated in foursteps :
- calculation of the correlation function
g (y )
with asemi-empirical
method[16] :
the first orderterms
41,
1(23)
to(25)
and the second moment rule«(3 . 2)
r(38)
are deduced fromg (y)
- for k =
0,
determination of the microfielddistribution
H (/3 )
(Appendix A)
- verification of the conditions of normalization :
- calculation of the second moment :
and
comparison
with the second moment rule. After an incrementationon k,
the last threesteps
are stillperformed
and so on, until the convergenceis reached.
Finally,
the finalko
value is determinedby
a linearinterpolation
using
the last two values of4.
Applications.
In tables I and
II,
the results for the test ofconverg-ence are
presented
for aone-component
and atwo-component
plasma
for various values of theplasma
parameter
r(7) ;
the second momentrule J3 2)
r
(38)
iscompared
with the secondmoment (/3
2) c
(N )
calculated from the microfield
distribution,
Nbeing
the number of iteration on k.
Table I. -
Comparison
between the second momentrule
/3
2) r
and the secondmoment (
f3 2) c
(N )
calculatedfor
several valuesof
k(N) (one
component
plasma).
Table II. -
Comparison
between the second momentrule f3
2) r and
the second momentf3 2) c
(N )
calculated
for
several valuesof
k (N ) (two-component plasma).
Fig.
1. - Microfield distributions ofH(¡3)
for aone-component
plasma (test
of convergence for various values ofk)
with v = 0.173, p = 1,Za
=Zb
= 1 and r = 0.01.Fig.
2. - Same asfigure
1 with v =0.2, p
= 1,Za
=Fig.
3. - Same asfigure
1 with v =0.4, p
= 1,Za
=Zb
= 35 and r = 20.03.Fig.
4. -Microfield distributions of
H(13 )
for atwo-component
plasma
(test
of convergence for various values ofk)
with v =0.2,
p =0.5, Za
=9,
Zb
= 1 and r = 0.32.The
corresponding
microfield distributionsH(13 )
are
plotted
infigures
1 to 3 for aone-component
plasma
and infigures
4 to 6 for atwo-component
plasma.
For
increasing
valuesof k,
the microfieldH(13 )
isshifted towards smaller values of the reduced mic-rofield
13
with apinch
effect on the distribution andon enhancement of the maximum.
Comparing
(13 2) rand (13 2) c (N)
(Tab.
I andII),
it is easy to determine theconvergent
ko
valuesFig.
5. -Same as
figure
4 with v =0.8, p
=0.5, Za
= 9,Zb = 1 and r = 5.13.
Fig.
6. - Sameas
figure
4 with v =0.6, p
=0.5, Za
= 17,Zb
= 1 and r = 8.39.(Appendix A)
and to deduce the final microfields distributionH (f3).
These results(Fig.
7 to12)
arecompared
with the more recent resultsgiven
by
other theoriesusing
theDebye-Huckel
screenedpotential approximation
[2-8] :
next-nearest-neigh-bour
(NNN)
approximation, Hooper
(H)
orTighe-Hooper
(TH)
calculations,
adjustable
parameter
exponential (APEX) approximation,
Monte Carlo(MC)
simulations.Fig.
7. -Definitive microfield distributions of
H(J3)
fora one-component
plasma
compared
with the results ofIglesias
et al.[2-8]
withko =
0.138,
v =0.173,
p = 1 9
Za = Zb = 1
and r = 0.01.Fig.
8. - Same asfigure 7
withko = 0.415,
v = 0.2,p = 1, Za=Zb=9
and F=0.52.with H and APEX. In the other cases, the
discrepan-cy is in
general
small,
our resultsbeing
inside thedispersion given
by
NNN, TH,
APEX or MC.5. Conclusion.
In this paper, a
general
iterative method for themicrofield
calculations has beenpresented
in view ofFig.
9. - Sameas
figure
7 withko
=0.456,
v =0.4,
P = 1, Za=Zb=35
and F = 20.03.Fig.
10. -Definitive microfield distributions of
H(,B )
fora two-component
plasma
compared
with the results ofIglesias
et al.[2-8]
withko =
0.304, v =0.2,
p =
0.5,
Za = 9, Zb =1
and r=0.32.applications
to the Starkdiagnostic
ofhighly stripped
ions.
The results are in
good
agreement
with APEX and the calculation times are very shortcompared
withMonte-Carlo simulations or more usual theories
[1].
With this iterative
method,
it ispossible
tocon-sider extensive
applications,
taking
care of theFig.
11. - Same asfigure
10 withko
=0.171,
v =0.8,
p
= 0.5, Za = 9, Zb = 1
and r = 5.13.Fig.
12. - Same asfigure
10 withko
=0.423,
v =0.6,
p=0.5, Za=17, Zb=l
and T=8.39.Zb =1,
i.e.v =1.732 )
and the Holtsmark limit(v = 0).
Finally,
it is of interest to note that the accuracy ofthese numerical results could
certainly
beimproved
by
introducing
the electronicdegeneracy
in the electronicscreening
function[23-26]
orusing
effec-tive
two-body potential
function solution of thenon-linearized Poisson-Boltzmann
equation.
Other cases were studied for different values of the
parameters
v, p,Za, Zb
and these results areavailable on
request.
Appendix
A. General results.1. PARAMETERS. - We consider a two-ionic
com-ponent
plasma
with aneutralizing
background.
The electronic
plasma
parameter
Te
is definedby :
where re
is the inter-electroniclength :
Introducing
the average ioniccharge
Z
[16] :
the root mean square ionic
charge
Z :
and the electronic
Debye length :
the two-ionic
plasma
parameter
r can be written[16] :
In order to express the microfield and the second
moment rule
explicitly,
it isinteresting
to introducethe relation for the concentrations and the conditions
of
neutrality :
In these
conditions,
we obtain the rates[1] :
Finally,
for the correlation functionapplications,
it is useful to define two reduced
lengths
x and y :where r is the
position
of thecharge particle,
andwith
the relation between x and y
being given by :
2. MICROFIELD. - The
asymptotic
behaviour of the microfield at acharged
point
(ion a)
isgiven by
[1] :
1with
and the notation
For the ionic
part
of themicrofield,
p and/3
areconnected
by equations (31)
and(32).
Using
the definitions for x and y, it is easy to show thatFor
high
values off3,
theexponential integral
termvanishes and the
asymptotic expression
becomesfairly
simple :
where
with k = a or b.
3. SECOND MOMENT RULE. - It is
possible
togive
explicit expressions
for the second momentrule,
using
the reduced microfield and theparameters.
For a
one-component
plasma :
it is easy to show that :
and we find :
Noticing
that :with
the second moment rule
becomes,
using
the reducedlength x :
For a
two-component
plasma :
4. CONVERGENT METHOD. - For the microfield
calculations,
we consider threeregions.
For smallvalues of the reduced microfield
f3,
H(13)
iscom-puted starting
from thegeneral
formalism(21), (24),
(25), (45)
andusing
theKrylov-Skoblya
approxima-tions for the Fourier transform[27].
Thesecalcula-tions are
performed
until numerical noise appears,criterium on the
slope
stops
the numerical pro-cedure. Forlarge
values off3,
the microfield iseasily
deduced from the
asymptotic development (28)
to(30)
and(35).
In the intermediateregion,
we use anappropriate
function for theinterpolation
ofH(f3 ).
This function is introduced from thegeneral
asymp-toticdevelopment, assuming
that :The four
parameters
are deduced from constraints on the value ofH(i3 )
and the first derivative of thisterm for the last
computed point (Hl
=H({3I),
and the first
asymptotic point
The numerical values of
H({3)
can beadjusted
after verification of the conditions of normalization :
Noticing
that(45) :
the microfield distribution calculation is
performed
for every iterative value of k.
N
being
the iterationnumber,
(f3 2) r the
secondmoment rule
(38),
(f3 2) c (N)
the calculation of thesecond moment
(47)
from the N-thH (/3 )
microfielddistribution,
thediscrepancy
A (N )
can be written :Assuming
that:the definitive
ko
value is calculated when thecon-vergence is reached.
k(N ) being
the N-th value ofk,
ko
isgiven
by :
where
(N - 1 )
and N are the last two iterationsbefore and after the convergence.
References
[1]
HELD, B., J.Physique
45(1984)
1731.[2]
IGLESIAS, C. A., HOOPER C. F. Jr., DE WITT, H. E.,Phys.
Rev. A 28(1983)
361.[3]
IGLESIAS,C. A.,
LEBOWITZ, J. L.,Phys.
Rev. A 30(1984)
2001.[4]
ALASTUEY, A., IGLESIAS,C. A.,
LEBOWITZ, J. L., LEVESQUE, D.,Phys.
Rev. A 30(1984)
2537.[5]
HELD, B., PIGNOLET, P.,Rapport
GRECO ILM(1984)
151.[6]
DE LUZE, O., Thèse de 3ecycle,
Orsay
[7]
DUFTY,
J. W., BOERCKER, D. B., IGLESIAS, C. A.,Phys.
Rev. A 31(1985)
1681.[8]
IGLESIAS,
C. A., DE WITT, H. E., LEBOWITZ, J. L.,MCGOWAN,
D., HUBBARD, W. B.,Phys.
Rev. A 31(1985)
1698.[9]
PERROT,
F., DHARMA-WARDANA, M. W. C.,Physi-ca 134A
(1985)
231.[10]
HELD,
B., PIGNOLET, P.,Rapport
GRECO ILM(1985).
[11]
HELD, B., DEUTSCH, C., GOMBERT, M. M.,Phys.
Rev. A 31
(1985)
921.[12]
HELD, B.,Phys.
Rev. A 31(1985)
1939.[13]
ANGELIE, A.,GILLES,
D., Ann.Phys.
Fr.Colloq.
11(1986)
C3-157.[14]
HELD, B., PIGNOLET, P., Ann.Phys.
Fr.Colloq.
11(1986)
C3-155.[15]
DHARMA-WARDANA, M. W. C., PERROT, F.,Phys.
Rev. A 33