• Aucun résultat trouvé

An equation of state for fully ionized hydrogen

N/A
N/A
Protected

Academic year: 2021

Partager "An equation of state for fully ionized hydrogen"

Copied!
27
0
0

Texte intégral

(1)

HAL Id: jpa-00212472

https://hal.archives-ouvertes.fr/jpa-00212472

Submitted on 1 Jan 1990

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.

An equation of state for fully ionized hydrogen

G. Chabrier

To cite this version:

(2)

An

equation

of

state

for

fully

ionized

hydrogen

G. Chabrier

Ecole Normale

Supérieure

de

Lyon,

Laboratoire de

Physique (*),

46 Allée d’Italie, 69364

Lyon

Cedex 07, France

Institute for Theoretical

Physics, University

of California, Santa Barbara, California

93106,

U.S.A.

(Reçu

le 2

février

1990,

accepté

le 23 avril

1990)

Résumé. 2014 Nous

développons

un modèle

d’équation

d’état pour des

plasmas complètement

ionisés et nous

appliquons

ce modèle au cas de

l’hydrogène.

Les effets de fort

couplage

entre ions ainsi que les effets à

température

finie pour le gaz d’électrons sont inclus dans le modèle. Nous considérons deux différents modèles dans deux domaines de densité

correspondant

respective-ment aux limites d’interaction faible et forte entre ions et électrons.

L’équation

d’état finale

couvre totalement le domaine de densité et de

température caractéristique

des étoiles de faible masse, des étoiles de la

séquence principale,

des

planètes géantes

et des

expériences

de fusion par confinement inertiel. De

plus

nous donnons une

expression analytique

pour la

règle

des sommes

pour la

compressibilité

du gaz d’électrons et pour la

longueur

d’écran

électronique

à

température

finie, ce

qui

fournit un lien entre les théories

quantique

de Thomas Fermi et

classique

de

Debye

Hückel.

Abstract. 2014 We

developed

a model for the

equation

of state of

fully

ionized

plasmas

which we

apply

to

hydrogen. Strong coupling

effects between ions as well as finite temperature effects for

the electron gas are taken into account. Two different models are considered

corresponding

respectively

to weak and strong limits for the correlation effects between ions and electrons. The final

equation

of state covers the whole

density

and temperature domain characteristic of low-mass stars, main sequence stars,

giant planets

and

inertially

confined

plasmas.

An

analytic

expression

for the electron gas

compressibility

sum rule and for the electronic

screening length

at

finite temperature are derived,

providing

a link between the quantum Thomas Fermi and the classical

Deby

Hückel theories.

Classification

Physics

Abstracts 52.25K - 05.70C

1. Introduction.

The

properties

of dense ionized matter have been

investigated

for

long

not

only

because of the intrinsic interest as a

subject

of statistical

physics

but also because of the

important

applications

to a wide array of

problems

such as inertial confinement fusion

experiments

in

(*) Equipe

associée au CNRS.

(3)

plasma physics,

the

study

of

liquid

metals and other Coulombic

liquids

in condensed matter,

and the

study

of the interior of stars and

giant

planets

in

astrophysics.

Under extreme

conditions,

similar to those encountered in the center of white

dwarfs,

the ions

(nuclei)

form a classical Coulombic fluid embedded in a

rigid background

of

degenerate

electrons. Such a fluid can be modeled

by

the well known one component

plasma (OCP),

whose static and

dynamic

properties

have been studied

extensively [1].

In

general however,

the

rigid

electron gas

approximation

is

rather

crude. In most cases of

interest,

the electron gas

is

polarized by

the ionic

charge

distribution and the

properties

of the

plasma

are modified

by

the response of the electron fluid. Various

investigations

have been made to

analyze

the effect of the electronic

screening

on the

properties

of the

plasma, using

Monte Carlo

experiments

[2,

3]

or variational calculations

[4].

The most extensive

study

of this

problem

has been

performed by

Ichimaru and his collaborators

[5],

who

developed

a

complete theory

of

interparticle

correlation in

dense,

high

temperature

plasmas.

Their

theory,

however,

is restricted to a domain of

relatively

weak

coupling

for the

plasma ( T _ 2)

and therefore is of limited

application

for

astrophysical

purposes. In all these calculations the ion electron interaction is treated within the linear response

theory (LRT).

For

hydrogen plasmas,

attempts

have been made to extend these calculations

beyond

the

LRT,

either

by considering

the electrons as semi-classical

particles [6]

and

using

a

pseudo-potential approximation

for the interactions in the

plasma

[7],

or

by using

sophisticated

theories like the

density

functional

theory (DFT) [8]

or the

quantal hypernetted

chain

theory

[9]

to treat the full non-linear response of the electrons and ions

self-consistently.

Because of

its

approximation,

the first method is confined to a low

density regime,

and its domain of

application

is not

yet

clearly

delimited. This

point

will be addressed in the

present

paper. The other methods

present

a more

complete

treatment of the

problem

but have the

disadvantage

of

being

very cumbersome.

Consequently they

can not be used to

generate

an

equation

of

state

(EOS)

over a wide

density

and

temperature

range.

Moreover,

the pressure and the free

energy must be calculated

by

numerical differentiation or

integration,

which

require

tremendous numerical efforts.

Such an EOS for a

fully

ionized

plasma

is of first

necessity

for

applications

in

plasma physics

and

astrophysics

where the

thermodynamics

of the

plasma

must be known over a

large

domain of

density

and

temperature.

The purpose of the

present

paper is to describe a model

to

generate

such an EOS. Because of its

astrophysical

interest,

we

computed

the EOS for a

pure

hydrogen plasma (Z

=

1 )

and compare it with

published

results. The formalism is

presented

for

arbitrary

ionic

charge

Z and can be

applied

to any

fully

ionized fluid. The

theory developed by

Hubbard and DeWitt

[2]

is accurate

only

in the

high density

regime (rs

1 )

and assumes that the electron gas is

fully degenerate.

Our model takes into

account the

properties

of the electron gas at finite

temperature

and the final EOS table we

computed

covers

essentially

the whole range of

coupling

and

degeneracy

characteristic of a

fully

ionized

hydrogen plasma.

We stress that the purpose of our model is not to

investigate

the onset of localization of bound states in the

hydrogen plasma

but to

generate

an accurate EOS for

fully

ionized

hydrogen.

In

fact,

the EOS we

present

in this paper is

part

of a more

general

calculation for fluid

hydrogen,

in which the presence of atomic and molecular

hydrogen

is considered

[10,11].

For this reason, the determination of the conditions for pressure ionization and

temperature

ionization as

given

in section 2 is

qualitative only

and is meant to

give

approximate

limits to the

temperature

and

density

range of

validity

of the

present

model. The final EOS table and the details of the model we

developed

for

atomic/molecular hydrogen,

(4)

2. Plasma parameters.

We consider a

plasma

of

N;

ions of

charge

Ze and mass M and

Ne

electrons of

charge

- e and mass m in a volume il at a

temperature

T. The concentrations and number densities are defined

respectively by xa -

N a /N

with a = i or e and

n a - N a / Y = xa n

where

n =

NI V

is the total number

density

and N

= N;

+

Ne

is the total number of

particles.

The

condition of

electroneutrality ne

= ni Z is assumed over the whole

plasma.

The

plasma

is characterized

by

the usual

coupling

parameters,

T for the

ions,

Fe

for the

electrons,

and the

density

parameter rs

for the

degenerate

electron gas, defined

respectively

as :

Here a =

(3/4 -ff ni) 1/3

is the

ion-sphere

radius, ae

=

(3/4

7rn,)11’ =

aZ- 1/3

is the mean

interelectronic

distance,

and ao is the Bohr radius.

The

degeneracy

parameter 0

for the electrons is defined as :

where

kTF

=

h2(3

TT’ 2 ne)2/3 /2

me is the Fermi

temperature

for the electron gas. The

temperature

T and the

plasma

mass

density

are related to these dimensionless

parameters

through

the

following

relations :

where A is the nuclear mass number in amu’s. We assume that the atomic nuclei in the

plasma

are all in the

fully

ionized state, which is achieved either

through

the pressure ionization or

through

the

temperature

ionization condition. The pressure ionization condition can be

described

qualitatively by requiring

the attractive Coulomb

potential

per electron

Ze2/ ae

to

be smaller than the Fermi energy - 2

B2/maé i.e.

In the low

density

limit

(log

p m -

4 )

the

temperature

ionization condition is

given

by

the Saha

equation.

In the

p -T

diagram

the 99.9 % ionization

boundary

for a one

component

system

is

given by

where XH is the ionization energy of the considered

species.

At

high density

this condition for a

temperature

ionization is

replaced

by

the

requirement

for the

temperature

to be

greater

than the

binding

energy of an orbital

electron,

i.e. :

(5)

At very

high

density (rs 1 ),

relativistic effects become

non-negligible

in the electron gas. Even

though

such effects could be

easily

included in our model in this

limit,

we will

concentrate on the non-relativistic

domain,

defined

by

the

requirement

that the kinetic energy

be smaller than the electron mass, i.e. :

Quantum

corrections to the

thermodynamic properties

of the ions will be considered in our

calculations. We will define the ionic

quantum

parameter A

as :

In

figure

1 we show the domain of

application

of our

study

in a

p -T

diagram.

The hatched area

in the lower left side of the

diagram

is excluded because of

equation (4)

and the

interpolation

between

equations (5)

and

(6).

The hatched area in the uper left side delimits the

region

of

stability

for the solid

phase.

The fluid-solid coexistence curve is the one obtained from

classical Monte-Carlo calculations

[12],

i.e. T =

178,

corrected for

quantum

effects when the

ionic

plasma frequency

becomes of the order of the

temperature

(ha 2: kT) [13].

Fig.

1. - Phase

diagram

of

fully

ionized

hydrogen.

Solid curves indicate constant values for the

plasma

parameter T and the

degeneracy

parameter 8 defined in the text. Values of the

ion-sphere

radius in units of Bohr radius

(r,)

are also indicated. The hatched

rectangular

area at low temperature and low

density

encloses the

regime

dominated

by

atoms, molecules and

partial

ionization.

Quantum

effects become

important

for the protons above the line

Ai

= 1,

indicating

that the

ion-sphere

radius is smaller than the

thermal de

Broglie wavelength.

The hatched area located at low temperature and

high density

is

essentially

defined

by

r a 178, the value at which the classical OCP model freezes into a

regular

lattice.

Because of

large

quantum effects

(rl > 1 ),

however, this

liquid-solid

transition may not occur in a

hydrogen plasma.

Dashed curves

correspond

to interior models of

Jupiter (J),

a brown dwarf

(BD),

a

(6)

As the

density

decreases

along

an

isotherm,

the electron gas becomes less and less

degenerate,

as seen from

equation (3),

and more and more

strongly

correlated.

Finally,

at low

enough density

we end up with a fluid in which the electrons behave almost as classical

particles.

For this reason, in our

study

of the ion-electron

plasma,

we will be

using

twô

different models in the

high

and low

density regimes,

the

thermodynamic

functions in the

region

of intermediate

coupling being interpolated

between the values calculated in these two

models. The model used in the

high density region

is based on the Two-Fluid

Model

initially

developed by

Ashcroft and Stroud

[14],

solved in our case for a

density-

and

temperature-dependent potential.

The detailed calculations are

presented

in section 3. The models free

energy in the low and intermediate

density regimes

are

given

respectively

in sections 4 and 5

and the final results for the EOS of the H

plasma

are

presented

in section 6.

"

3. The

high density regime.

3.1 SCREENING LENGTH AND EFFECTIVE POTENTIAL. - As

long

as the Coulomb

potential

energy

e2la

is smaller than the Fermi energy of the electrons EF, we can assume that the ion-electron interaction is weak

compared

to the kinetic contribution of the electrons and can be

treated as a

perturbation, retaining

the linear contribution

only.

This condition leads to

equation (4)

and is

supported by

results of the

inherently

non-linear

Density

Functional

Theory

in which bound states appear in the

plasma

for rs

=

2,

indicating

the onset of

strong

non-linear effects

[8].

Under this

condition,

the exact Hamiltonian of the ion-electron

plasma

can be rewritten as

[14] :

Here

He

is the familiar

« jellium »

Hamiltonian

[15]

for the

electrons,

and

Heff

is the Hamiltonian of the screened ionic

fluid,

which takes into account the response of the

polarizable

electron

background

to any variation to the ionic

charge density.

It is

expressed

as :

where K;

is the ionic kinetic

contribution,

p K is the Fourier

component

of the

microscopic

density, and ECK, 0)

is the

screening

function of the electron fluid in the adiabatic

approximation,

to be discussed below. The fluid is

equivalent

to the

superposition

of two

uncoupled

fluids,

i.e. a

jellium

electron gas and a fluid of

pseudo-ions interacting through

the

short range screened

potential :

which is the sum of the external ionic

potential

and the induced

polarizable potential.

We define a dimensionless temperature- and

density-dependent screening

parameter

Qe

as :

Here

KF

=

(3

7T2ne)1/3

is the Fermi wavevector,

Q

=

KKF

and X is

the screened

density

(7)

temperature

static Lindhard function which

yields

to the

following

expression

for the

screening

parameter :

Here

KTF

=

(6

rne

e2/kTp)1/2

is the Thomas-Fermi wavevector, a =

J.L o/kT

and

,

° is the

chemical

potential

of the

noninteracting

electron gas at finite

temperature,

defined

by

the condition :

roc

where

In ( a )

=

Jo

dx xn/ (ex - a

+

1 )

denote the Fermi

integrals.

The chemical

potentials a

0

are calculated with the fit derived

by

Dharma-Wardana and

Taylor [16]

and for the Fermi

integrals

we use

polynomial expressions,

of which values agree to within one

part

in

10- 6

with the numerical estimations

[17].

The

screening

wavevector, or inverse

screening

length, K,,

defined as

Ke

=

KF Qe (Q )

is a

generalization of

the Thomas Fermi and

Debye

Hückel wavevectors recovered

respectively

in the zero

temperature

limit

(0

1 )

and in the

classical limit

(0

>

1 ) at Q

= 0. In the limit of

large wavelengths (6-0), equation (13)

yields :

We note that the

screening

wavevector is

independent

of the

wavelength

in this

limit,

which is

a direct consequence of the

perfect

screening

condition.

Using

the results for the

compressibility

of the

noninteracting

electron gas at finite and zero

temperature,

equations

(12)

and

(15) yield :

where the xoe denote the

respective compressibilities

of the

noninteracting

electron gas at

zero and finite

temperatures.

The

equation (16)

is the

generalization

at

finite

temperature

of

the zero temperature

compressibility

sum

rule for

the electron gas

[15].

We will see in the next

section how to

modify

this relation for an

interacting

electron gas.

In

figure

2,

we show the behavior of the dimensionless

screening

wavevector

aKe

as a function of

Q

for different values of the

density

and

degeneracy

parameters.

A

large

value of

(aKe )

corresponds

to a

relatively

short range effective interaction between the ions

compared

to the bare Coulomb

potential.

At intermediate densities

(rs

=

1 )

we note the

strong

wavelength dependence

of

Ke

and

consequently

the

strong

departure

from the Thomas Fermi model

(aKTF

= 1.56

rs 2)

recovered in the

long wavelength

limit for the

strongly degenerate

case. At short

wavelength,

the

screening

wavevector vanishes and we recover the unscreened

bare Coulomb

potential.

At constant

density, Ke

decreases with

increasing

temperature,

indicating

that the

screening

of the ionic interactions becomes less and less efficient. This is a

consequence of the fact that the electron gas

spreads

out as the

temperature

increases at constant

density.

The same conclusion holds

obviously

as the

density

increases at constant

degeneracy

since the electron gas becomes less

polarizable, becoming ultimately

a

rigid

(8)

Fig.

2. - Electron

screening

wave vector as a function of K for two different densities and two different

temperatures. The dotted line and the short dashed line indicated

respectively

the Thomas Fermi limit and the

Debye

Hückel limit. The dashed lines are RPA calculations whereas the solid line indicates that

a LFC has been included in the calculation of the

polarisability.

The calculation of the distribution functions and then the

thermodynamic

functions of the

system

necessitates the

knowledge

of the screened

potential

Veff (r)

(or screening

function

rheff(r)/(Ze)2)

in the r space. In order to avoid

long

range oscillations in the Fourier

transform of

equation (11),

it is more convenient to calculate the Fourier transform of the non-Coulombic

part

of the effective

potential,

defined

by f/1 (K)

=

Veff (K) -

4 7r

(Ze)2 IK 2

which behaves like

Q - 6 at

large

Q.

The

problem

of the

discontinuity of f/1 (K)

at K = 0 can be circumvented

by rewriting Q

as the sum of two

potentials :

one

being

the difference between a Coulomb and a Yukawa

potential,

of which the Fourier transform is calculated

analytically,

the other one

being

the

complementary

part

f/12(K).

Both have a finite value at K = 0. The final result

reads,

in dimensionless

quantities (x

=

rlae ; q

= ae

K) :

where A o is a

screening length

defined as :

and f/12(X)

is

given by :

with qe(q) = ae Ke(K).

(9)

decreases

rapidly

as the

temperature increases,

at fixed

density

[3].

The function

03C82(x)

represents

the

departure

of the effective

potential

from a Yukawa

potential.

In

figure

3 0393

we compare the effective interionic

potential

Veff(X)

with the Yukawa

potential

e

x

(see

Eq.

(17))

for different values of 0 and rs. For distances smaller than the mean interionic

distance,

the two

potentials

are very similar and become

essentially

the Coulomb

potential

as

r goes to zero. As the

density

decreases

(i.e.

rs

increases)

the

departure

from a

Yukawa

potential

is very

important

in the domain of

strong

degeneracy,

because of the

strong

inhomogeneity

of the electron gas

but,

as the

temperature

increases,

the two

potentials

become very similar

(for

r, = 1 the

highest

discrepancy

is of the order of 10 % at 0

= 0.5)

and become

essentially indistinguishable

at 0 - 1. Below a certain

density and/or

a

certain

temperature,

the effective

potential,

via the function

03C82’

exhibits an

oscillatory

behavior in the

region

of short range order. This

corresponds

to the onset of the well known Friedel oscillations. It defines a transition between a

Thomas-Fermi-type regime,

at

high

density (rs 1 )

and finite

temperature

(0 > 0.1 )

in which the effective

potential

VefT(x)

can

be

approximated by

a Yukawa

potential

of which the

screening length

is

given analytically by

equation (18),

and a

Friedel-type regime

at lower densities

and/or

temperatures.

This

point

has been discussed in detail

by

Gouedard and Deutsch

[18].

It is

important

to stress that

Ao

in

equation (18)

is a

general density-

and

temperature-dependent screening length,

the

density

and the

temperature

dependence being clearly separated (note

that

Ao/qTF

is a

universal function of the

density

parameter

0).

It is a

generalization

of the Thomas Fermi

(Ào

=

qTF)

and

Debye

Hückel

(Ào

1= qDH

=

J"3f:)

screening lengths

recovered

respecti-vely

in the

fully degenerate ( 0 --> 0 high density

(rs 1 )

limit and in the classical

(0

>

1 )

low-density (rs

>

1 )

limit.

The main

approximation

in our model free energy is the linear

approximation

in the

treatment of the electron-ion interaction. The most

powerful, although extremely

cumber-some,

theory

available so far to include non-linear effects in the calculation of this

interaction,

is the

Density

Functional

Theory (DFT) [8].

However,

because of the

complexity

of the

method,

no

equation

of state has been

computed

so far within the framework of the DFT.

Therefore it is not

possible

to

appreciate

the relative contribution of non-linear effects on the

Fig.

3. -

(a)

Effective interionic

potential (Eq. (17) (-)

and Yukawa

potential (---)

for two different densities at 0 = 0.0543.

(b)

Same as

(10)

thermodynamic

functions of the

plasma.

We must resort to a

comparison

of the effective

ion-ion

potentials

of the

respective

models. In

figure

4 we compare the effective interionic

potential (15)

with and without the inclusion of a local field correction

(LFC)

in the dielectric

function,

with the result of

Density

Functional

THeory (DFT)

calculations

[8]

for

rs = 1. As

already pointed

out

by

Perrot

[19]

and Dharma-Wardana et al.

[20],

one of the main consequences of

using

a linear response

theory

is to overestimate the extent of the

screening, leading

to a too

repulsive potential

and an underestimation of the electron

density

at the nucleus. This

stronger

screening

inhibits the formations of bound states. The two

potentials

differ also in both the

amplitude

and the

phase

of the Friedel oscillations

although

this misbehavior is

improved slightly by

the inclusion of a LFC in the RPA dielectric function. However the most

interesting

feature is that the

discrepancy

between the screened

potential

and the DFT

potentials

diminishes very

rapidly

as the

temperature

increases,

indicating

that a

linear

approximation

becomes more accurate in the domain of low and intermediate

degeneracy (

2:

0.2 ),

the electron cloud

becoming

more and more

spread

out. We also note

the

rapid damping

of the Friedel oscillations as the

temperatue

increases,

consequence of the

smearing

of the Fermi

sphere.

As we

already

mentioned,

the

potential

becomes a Yukawa

type

potential

in this

region.

In view of this

comparison, given

the

good

agreement

found for

rs = 1 at 0 - 0.3 between the two

potentials,

we conclude that non-linear effects will

modify

appreciably

an EOS

only

for rs 2:

1 in the

region

of

strong

degeneracy ( B 0.2 ).

In this

region,

however,

bound states are

likely

to appear and a more

complete theory

must be

developed [ 10, 11 ] .

Fig. 4.

- Effective interionic

potential given by equation (17) (-)

or

by

DFT

(---) (Ref. [20])

at

rs = 1 for two different

temperatures. The dotted line indicates the

potential (17)

within the RPA

approximation

for the dielectric function.

3.2 THE THERMODYNAMIC FUNCTIONS. - The free energy related to the Hamiltonian

(9)

is

written

where the trace is taken over the states of the

coupled

electron-ion system. The

superscript

(id)

denotes the

non-interacting

contribution to the free energy

given

either

by

the standard

(11)

where a is the chemical

potential

of the

noninteracting

electron gas defined

by equation (14).

The last two terms on the

right

hand side of

equation (20)

denote

respectively

the non-ideal ionic and electronic

contribution,

to be discussed below.

3.2.1 Screened

ionic fluid.

- Because of the temperature

dependence

of the effective

potential

veff,

the excess internal energy is no

longer equal

to the canonical average of the

potential,

but includes an extra kinetic term. On the other

hand,

the excess pressure includes a term due to

the

density dependence

of the

potential :

where

v ( K _ ) 4’T (Ze)2 2

and

S’ ( K - ) N-1

Px Px * %

is the ionic structure factor of the

K 2

screened ionic fluid.

The Q

and p derivatives are

trivially

rewritten in terms of the dimensionless variables defined in

equations (1)-(3).

The interaction energy eint is

equal

to the

sum of the two first terms on the

right

hand side of

equation (22a).

It is a very well known

feature of the HNC

theory

to allow the calculation of the chemical

potential,

and then of the free energy,

by

direct

integration

of the distribution functions

[21].

The free energy in the HNC scheme is thus obtained with an accuracy

comparable

to those of the internal energy and pressure

contrarily

to methods like Monte Carlo and DFT where numerical

integration

of the interaction energy with

respect

to the

temperature

has to be carried out. This method has been used for

temperature

and

density independent potentials only

but it has been extended

recently

to

temperature

and

dependent

screened

potentials

[22].

More

recently,

DeWitt and

Rogers [23]

derived a more

simple expression

for the same free energy, in the HNC scheme :

where

è(K)

is the dimensionless Fourier transform of the direct correlation function and

h(r)

the

pair

distribution function of the screened ionic fluid.

è,(K)

is defined as

cs (K) - c (K)

+

P 16

V eff (K) / Z2

and cp is

given by :

where

Ao

and

’P2(x)

have been defined in

equations (18), (19).

In order to assess the

validity

of the HNC scheme for the calculation of the

thermodynamics

of a

plasma,

we

compared

out results for the excess pressure

p ex,

internal energy

ue",

interaction energy eint, and free energy

f eX (Eqs. (22), (23))

with

existing

Monte Carlo

(MC)

calculations

(see

Tab.

I).

Such simulations have been carried out for a

rigid

electron

(12)

Table 1.

- Comparison

between MC results

(left

column)

and HNC results

(right

column)

(Eqs.

(22), (23)) for

the ionic excess

thermodynamic functions.

The

quantities

are dimensionless

(

1 Ni kT).

The MC results are taken

respectively from : (a) reference [12] for

rs =

0,

(b)

reference [3] for

calculations with the

finite

temperature Lindhard

function, (c) reference [2a]

and

(d) reference [2b] for

the calculations with the zero-temperature

Lindhard function.

equation (11) [2],

and a finite

temperature

dielectric function

[3].

In most of the cases the

discrepancy

between MC and HNC results in less than 1 % for any value of T from 1 to 160. The

largest disagreement (-

1-2

%)

found for F 2 in the

comparison

with the results of

Totsuji

and Tokami

[3]

can

probably

be attributed to the small number of

particles (64)

involved in the MC calculation in this case, since the

disagreement

diminishes as T increases.

The

discrepancy

on the free energy in the

comparison

with the results of Hubbard and DeWitt

[2]

at small T is due to the poor accuracy of the MC method in this

region,

where the

Debye

radius becomes

larger

than the unit cell

dimension,

and to the

approximated

mean square fit

in this

region [24].

(13)

Dharma-Wardana

[20]

derived two

separate

expressions

for the

exchange

part

and the correlation

part

of the free energy and the chemical

potential.

The correlation

part,

however,

was

approximated by

the

ring

sum

contribution,

i.e. the RPA

expression,

of which range of

validity

is

unknown,

as

pointed

out

by

these authors. It is

known,

moreover, that the RPA scheme

always

underestimates the interaction energy.

Recently

Tanaka,

Mitake and Ichimaru

[5,

26]

calculated the interaction energy of an electron gas at

finite

temperature

using

the

Singwi-Tosi-Land-Sjolander (STLS) approximation [27]

to treat the

strong

coupling

effects between electrons

beyond

the RPA.

They

parametrized

their results as a function of rand 0 and

performed

a

coupling

constant

integration

to obtain a

general

fit for the

exchange

and correlation free energy

Fxx(ne, T). During

this

procedure, they

took into consideration the known violation of the

compressibility

sum rule in the STLS

scheme,

and the

consequent

departure

from the exact Monte Carlo calculations in the classical

[12]

and

fully degenerate

[28]

limits as the

coupling

constant

T or rs

increases,

by anticipating

similar derivations in their

parametrization.

Their free energy and its

temperature

and

density

derivations

reproduce

the

classical results

[12, 29]

and the results for the

ground

state

[28, 30]

with

digressions

of less than 0.6 %. In the

region

of intermediate

degeneracy (0 ~ 1),

their result is in

perfect

agreement

with variational calculations

[31].

Because of the

integration procedure,

the

authors claim a total error of - 5 % on

Fxc.

In any case this result seems to be the most accurate treatment of the

thermodynamic properties

of an

interacting

electron gas at finite

temperature

published

so far.

3.3 LOCAL FIELD CORRECTION. - The dielectric function defined in

equations (10)

and

(11)

is the Lindhard dielectric function calculated within the RPA

approximation.

It is well known that the RPA

approximation

fails as the

density

of the electron gas decreases because it does

not take into account the short range correlations in the motion of the electrons. This

approximation

can be

improved by taking

into account the

polarization

effects in the electron-electron interaction in the

density-density

response function in term of a mean-field

approach :

where

Xo (K)

is the

density-density

response function for the

non-interacting

gas,

v (K)

=

47T2 is

the Hartree

potential

which characterized the

long

range correlations of the

K2

electron

system

and G

(K, w )

is the so-called Local Field Correction

(LFC) describing

all the

exchange

and short range correlation effects

beyond

the RPA. The dielectric function is then written :

where EL

denotes the Lindhard dielectric function

(Eqs. (12), (13)).

In the formulation of a static

theory,

the most

important

effect of the LFC stems from its static evaluation at w =

0,

i.e.

G (K) -

G

(K, w

=

0 ).

Moreover,

in our

calculations,

we will assume the adiabatic

approximation

to be valid and then we will use the static limit

(w

= 0 )

of the dielectric function. The

validity

of this

approximation

has been assessed

(14)

The

long wavelength

behavior of the LFC is

governed by

the

compressibility

sum rule

[ 15].

with

Here

Xo,

=

3 / (2

nEF)

is the

zero-temperature

compressibility

of the

non-interacting

electron

gas and X

xc

=

1

a

2

F xc

T is the

exchange

and correlation contribution.

gas and d -

xxc

1 =

1/2 a n2(

FXC )

T is the

exchange

and correlation contribution.

1/n2

ô2

(

F xc )

T 0

h h d 1 0

Ob 0

n2 an2

/

The small

wavelength

behavior of the LFC is related to the short range behavior of the

electronic

pair

distribution

g (r) through

the exact relation

[33, 34] :

lim

The finite value of

g (0)

is a consequence of

quantal tunneling.

Although equations (27)

and

(28)

have been derived

originally

for the

ground

state

(0 = 0 ), they

are still valid at finite

temperatures

[35].

We can

easily

extend our

previous

result

(16)

for the

compressibility

sum rule for the

non-interacting

gas to the

interacting

electron gas at finite

temperature.

Using equations (16), (26)

and

(27),

we

get :

-lim

where

KD

=

(4

7Tne

f3

e 2)1/2

is the inverse

Debye screening length,

X o =

(ne kT)-1

1 is the

compressibility

of the classical

perfect

gas and

X ( 0 )

denotes now the total

compressibility

of the

interacting

electron gas at finite

temperature

(X ( (o )- 1 =

x - 1 ( o )

+

X xc-1 ( (0 )). The

second

equality

in

equation (29)

can be obtained

by using

the classical version of the Fluctuation

Dissipation

Theorem,

i.e. e(K)-1 =

1 -

K51 K2 Szz(K)

and the

long wavelength

limit of the

charge-charge

structure factor

Szz(K --> 0) =

(K51 K + X o/X )-1

1

[36].

The

equation (29)

provides

a link between the zero

temperature

( 0

=

0)

and

high

temperature

(0

>

1 )

limits of

the

long wavelength

behavior of the dielectric function for an

interacting

electron gas.

A new

screening

length A 0 xc(ne,

T)

associated with the finite

temperature

interacting

electron gas, can then be derived from

equations (26), (27), 29) :

where A F 1 = ae

KF

is the Fermi

screening length

and A o is the

screening length

of the

non-interacting

electron gas

(Eq.

(18)).

The

ratio k F/k 0

is

equal

to the

screening

parameter

Qe(O)

defined in

equation (15).

The second term in the bracket arises from the

exchange

and correlation contribution and can be evaluated

analytically

from the second derivative of the free energy

Fxc

discussed in the

previous

section.

Various

expressions

for G

(K)

have been

proposed

in the

past

by

many

investigators.

In the

zero-temperature case, the most

interesting

solution seems to be the one derived

by

Utsumi and Ichimaru

[37]

hereafter referred to as

GUI.

Not

only

the LFC derived

by

these authors

(15)

compressibility

derives from a

fitting

formula

[30]

in

perfect

agreement

with the

quantum

Monte-Carlo simulation

[28],

and the

expression

for

g (0 )

is the one calculated

through

a

diagrammatic

calculation of the electron-electron ladder

diagrams

[38].

In the case of a

finite

temperature electron

fluid,

the situation is not so

simple

and no

analytical expression

has been derived

yet.

For

large

values of the

degeneracy

parameter 0

the electron fluid becomes

essentially

classical. Thus the

principal

effect that goes into the determination of the LFC is the

interparticle

correlation

brought

about

by

the Coulomb

repulsion,

no

exchange

effects need to be taken into consideration. A self-consistent formulation of such a LFC can be obtained

using

the linear

approximation

and the classical version of the Fluctuation

Dissipation

Theorem,

which

yields :

where

Socp (K)

is the structure factor of the well known One

Component

Plasma

(OCP).

Because of the

long-wavelength

behavior

of Socp (K)

[36],

the

compressibility

sum rule

(27)

is

satisfied,

whereas the condition

(28)

is

automatically

fulfilled since

g (0)

= 0 for a

fully

classical

system.

However the

expression (31)

for the LFC

corresponds

to an electron

OCP,

i.e. a

system

of

point

charges

embedded in a

uniform positive

neutralizing

background,

without interaction with this

background.

This result is

obviously

different from the case of a

lattice of electrons

interacting

with a lattice of

point

ions,

which is the case for the

hydrogen

plasma

in the-classical limit for the electrons. This difference is illustrated in

figure

5 where we

show the electro-electron structure factors obtained in the two situations.

Consequently

we can

expect

the LFC to be different in the two cases.

Fig.

5. - Electron-electron

structure factor calculated within the framework of the HNC

theory

for the OCP model

(-)

and for the TCP model

(---) (Eq. (38))

for r = 1

(and

e = 20 for the

TCP).

Another

approach

for the finite

temperature

LFC has been

proposed recently by

Tanake and Ichimaru

[35],

who extended to finite

degeneracies

the STLS scheme

initially developed

by Singwi et

al. for the

ground

state

[27].

The STLS functional form for G

(K)

satisfies the

(16)

(27).

This

point

has been

partially

corrected

by

Vashista and

Singwi

who

approximated

the LFC

by

the

following

form

[39] :

where

Q

=

KIKF

and the

parameters

A and B are fixed

by

the conditions

(27)

and

(28).

Finally

the last

possible parametrization

scheme for

G ( K)

would be a finite

temperature

extension of the formula first

proposed by

Hubbard

[40],

i.e. :

where A and B are the coefficients used in

(32).

The Hubbard

approximation corresponds

simply

to

using

the Hartree-Fock value of the

pair

correlation function in the STLS scheme. A common feature of the forms for

G (K) given by equations (31)-(33)

is that

they

do not

exhibit a

peak

structure around K = 2

KF,

associated with the

discontinuity

of the electron

density

at the Fermi surface. In the zero

temperature

limit,

the presence of this

peak

in the

LFC,

associated with a

logarithmic singularity

in its

slope

has been demonstrated

rigorously

[41, 42].

Using

the

Density

Functional

Theory (DFT),

Dharma-Wardana and Perrot

[20]

showed

that,

for values of rs of the order of

1,

the

exchange potential

is still dominant

compared

to the correlation

potential

up to 0

of the

order

of 1. Consequently,

since the LFC is

related to the second derivative of the

exchange-correlation potential,

we can assume the

exchange

part

of the

LFC,

and hence the

peak

structure, to be still

important

at intermediate

degeneracy.

For this reason the form we propose for the LFC is a finite

temperature

extension

of the form

given by

Ichimaru and

Utsumi,

that we will refer to as

GUI.

This is achieved

by

using

the finite

temperature

compressibility

in

equation (27b)

instead of the one for the

ground

state. For that we differentiated the

parametrized

form of the

exchange

and

correlation free energy at finite

temperature

mentioned in section 2.2

[5].

This insures

consistency

between the

screening

function

entering

into the calculation of the effective interionic

potential,

and the

thermodynamic properties

of the electron gas at finite

temperature.

The

term

yo in

equation (27b)

is then

given by :

where

Fxc

is the

fitting

formula

given

by equation (3.83)

of reference

[5].

The

zero-temperature

limit and the classical limit

(0 > 0)

of the

compressibility

calculated from this

differentiation

reproduce

the Monte-Carlo results

[28, 12]

within less than 3 %. Because of the lack of

parametrized

form for

g (r

=

0)

at finite

temperature,

we

kept

the zero

temperature

value. It is thus clear that the condition

(28)

is violated. However we found out

that the final

thermodynamic

functions of the screened ionic fluid are

quite

insensitive to any

important change

in

g (0).

This comes from the

fact,

as seen from

figure

2,

that the

polarizability quickly

vanishes

beyond

K ~ 2

KF

and that the effect of the LFC is dominant

only

in the

region

K 2

KF,

so that the short

wavelength

behavior of the LFC does not have

any

quantitative

effect on the

thermodynamics

of the

plasma.

The

advantage

of the form

GUI

is to recover the accurate form in the

zero-temperature

limit and to

give

an accurate

temperature

dependence

in the

long wavelength

limit,

where correlation effects are

dominant.

The behavior of the different form of G

(K)

discussed above is shown in

figure

6 for

0 = 1. We

explored

the

consequence of these different forms for the finite

temperature

LFC

on the

thermodynamic

functions of the H

plasma.

In the

region

of intermediate and low

(17)

region

of intermediate and strong

degeneracy,

the results differ

significantly.

In this

region,

the

quantum

effects between

electrons,

resulting

in a

peak

structure at K - 2

KF,

become

important

and must be taken into account in a proper calculation of an LFC. We conclude

that as

long

as we are interested with the calculations of static

properties,

a

quantum

treatment of the LFC is necessary in the

region

of

strong

and intermediate

degeneracy

(0 1 )

whereas for lower

degeneracies

the

thermodynamics depends essentially

on the

correlation,

i.e. the

long-wavelength (K « KF)

part

of the LFC.

The enhancement of the

screening

due to the inclusion of a LFC

(GÚI)

can be seen in

figures

2 and 3. The effect on the

thermodynamics

of the

plasma

is

quantified

in table II and

shown in

figure

7

by comparison

with RPA results. The difference can reach a few

percents

on

the pressure, the internal energy and the free energy for rs a 1 at

large

T,

and can be more

than 10 % on the

entropy.

Fig. 6.

Fig. 7.

Fig.

6. - Different forms of the LFC at 0 = 1, T = 1. The solid line

(-)

represents

GUI,

the dashed line

(---) G H (Eq. (33)),

the dashed-dotted line

(-.-) Gvs (Eq. (32)),

the dashed-double dotted line

(-..-)

Gocp (Eq. (31)),

and the dotted line

(...) G STLS (Tab.

IV of Ref.

[5]).

Fig.

7. - Relative

importance

of the LFC in percent for the excess ionic free energy

(e, 0),

internal energy

(A, A)

and pressure

(~, ~)

as a function of T for r, = 2

(solid

line

( ))

and rs = 1

(dashed

line

(--- )).

3.4

QUANTUM

CORRECTIONS. - In the low

temperature,

high density region

of the

diagram

shown in

figure

1,

we can

expect quantum

effects on the ions to become

important.

As

long

as

the

parameter

ll

given by equation (8)

is smaller or of the order of

unity,

we can consider

these

quantum

effects to be a small correction to the

thermodynamics

of the ions and we use

the

Wigner expansion

to the lowest order

(-

h 2).

Since the

quantum

corrections occur in the

high

density regime ( rs

«

1 ) (cf.

Eq.

(8)),

we can

approximate

the effective

potential

in this

(18)

Table II. - Excess ionic

thermodynamic

functions

within the RPA

approximation

and with a

Local Field Correction. The

quantities

are dimensionless

(

1 Ni

kT).

For each value

of

rand

rs, the

free

energy on the

first

line has been calculated

from

the

expression

of

Iyetomi

and Ichimaru

(Ref.

[5])

whereas the value on the second line has been calculated

from

equation

(23).

where

Ao

is the

screening length

defined in

equation (18)

or

(30).

Then the

Wigner

expansion

for the

quantum

correction to the ionic free energy

yields :

The first term is the

quantum

correction for the OCP recovered at very

high density,

whereas the second term is the

(negative)

contribution due to the electron

screening.

The correction to

the internal energy

uqu

and pressure

pqu

are

given by

the

temperature

and

density

derivatives. The contribution of the V and T derivatives of the last term in the bracket in

equation

(36)

are

found to be

negligible

so that

ui qu

and

pqu

can be

approximated by :

In table III we show the relative contributions of the different terms of the total excess free

energy

(20)

and the related pressure and internal energy for different values of T and rs. The relative

importance

of the

screening

contribution on the

thermodynamic

functions of

the ionic fluid is shown in

figure

8 as a function of

r,

and then

0,

for different values of the

(19)

Table III.

- Different

contributions to the

excess free

energy, internal energy and pressure

of

the

plasma for

different

densities and temperatures. All the

quantities

are

given

per

particle (IN)

in unit

of

kT. We recall that

u¡qu

= 2

p 9u .

2

f¡qu

(Eq. (36)).

Fig.

8. - Relative

importance

of the

screening

contribution in percent for the excess ionic free energy

(e,

0,

+),

internal energy

(A,

A,

x)

and pressure

(U,

D,

*)

as a function of T

for rs

= 0.1

(solid

line

(-)), rs

= 1

(20)

4. The low

density regime.

In the

low-density regime (rs

>

1 ),

the ion-electron and electron-electron Coulomb energy

become

important

compared

to the Fermi energy of the electrons and the dielectric

formulation

developed

in section 3 can no

longer

be used.

On the other

hand,

in the

low-density,

high

temperature

limit

(0

>

1),

the electron gas is

weakly degenerate

and the

plasma

may be treated as an almost classical

two-component

plasma (TCP).

Quantum

diffraction and

symmetry

effects,

which

prevent

the

collapse

of

purely

classical

systems

of

particles

of

opposite charge,

can be handled

through

pseudopoten-tials,

derived

by expressing

the

quantum-mechanical

Slater sum in a form reminiscent of the classical Boltzmann factor. For the

hydrogen plasma

effective

pair

potentials

have been

derived from exact numerical

computation

of the

two-particle

Slater sum for electron-electron

and

electron-proton pairs

[7],

and

reâd :

where

Àaf3

=

h/(2

71’maf3

kB

T)1/2 and maf3

is the reduced mass of an

a-f3

pair (f3 = 2

denotes

the electron

component).

In

equation (38)

the first term arises from

quantum

diffraction

effects,

while the second takes care of

symmetry.

Because the contribution of bound states to the

electron-proton

Slater sum as well as the

density dependence

of the

potential

are

neglected,

this model Hamiltonian can be used

only

in the low

density, high

temperature

( T >

1

Ry ) region.

Bernu et al.

[44],

considering

the influence of the electron

spin

on the structure and the

thermodynamics

of this

model,

restrained its domain of

reliability

to the

region ’g 2:

5.4,

so

that densities are low

enough

for the

temperature

T to

greatly

exceed the electron

degeneracy

temperature

TF.

We will discuss the

validity

of this limit as well as the

reliability

of the TCP model, The

pair

distribution functions and the

thermodynamics

of the TCP were calculated in

the

regime

of interest within the framework of the HNC

approximation.

On the other

hand,

in the limit of weak

coupling

(r «

1)

and low

degeneracy

(0

>

1),

the

thermodynamics

of the ion-electron

plasma

can be evaluated in term of the

quantum

generalization

of the

ring

expansion [45].

The free energy then reads :

where

(Z2) = L Xa Z;, a

denoting

the ionic and electronic

species.

The first term in

03B1 03B1

equation (39)

is the

leading

term of the classical cluster

expansion,

i.e. the

Debye

term,

whereas the second term in the bracket

represents

the correction for the electron

quantum

effects. The

quantum

diffraction

parameter

ye is

given by :

where Ae

is the electron

equivalent

of the ionic quantum

parameter

defined in

equation (8).

0393

For Coulombic

system

(- 1/r),

the

potential

/3

V - T

decreases as

1/kT

as the

temperature

a

Références

Documents relatifs

2014 The thermal conductivity of a weakly coupled electron gas is obtained within the Wigner function formalism, taking quantum correlations into account

Specially, in the case of the strongly-coupled classical one-component plasma (OCP), the values of the transport coefficients obtained by molecular dynamics [2] and

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des

distribution are: the Pearson correlation coefficient ρ, and the index of regular variation of R (say α) or an auxiliary function of R (say ψ), depending on wether R has a regularly

• The analysis for the lepton pair final state has three improvements: 1) an increased efficiency in the identification of events in which the Higgs boson decays to τ leptons,

Le phénomène de percolation dans les matériaux composites à base de matrice polymérique a été utilisé pour déterminer le seuil de percolation de tels systèmes en élaborant un

We also conclude that despite being at a high alpine site, the vast majority of the air observed was extensively influence by the boundary layer during our campaign (diagnosed from

Atmospheric Pressure Humid Argon DBD Plasma for the Application of Sterilization - Measurement and Simulation of Hydrogen, Oxygen, and Hydrogen Peroxide Formation..