Astron. Astrophys. 323, 271-285 (1997)
6. Stochastic particle acceleration
The interactions between particles and turbulent fluctuations are
characterized by an exchange of energy which can lead to the
acceleration of the particles. This exchange is described by a
diffusion equation of the particles in momentum space.
![[EQUATION]](img162.gif)
The quasi-linear coefficients describe
scattering effects and can be evaluated with the stationary turbulent
spectra. For the Langmuir waves:
![[EQUATION]](img164.gif)
In magnetized plasma the relevant diffusion coefficient is
parallel to the magnetic field direction.
The Alfvén waves contribute essentially to the relaxation of
the anisotropic part of the distribution function and tends to
isotropize the particle distribution function with a pitch-angle
scattering frequency:
![[EQUATION]](img166.gif)
The acceleration by Alfvén waves is of the second order in
with respect to the pitch-angle frequency.
6.1. Derivation of the Fokker-Planck equation
In strong field regime, with an isotropic turbulence, we derive the
diffusion equation in spherical coordinates. Therefore, for Langmuir
turbulence we express the momentum diffusion coefficient
and the pitch-angle scattering frequency
in term of the previous parallel coefficient,
namely for the Langmuir turbulence
![[EQUATION]](img170.gif)
For Alfvén turbulence we introduce an acceleration
coefficient
![[EQUATION]](img171.gif)
The diffusion is balanced by the Inverse Compton losses given by
the Eq. (3), which can be written as ;
describes the losses for a Lorentz factor
![[EQUATION]](img175.gif)
where , and .
we obtain,
![[EQUATION]](img178.gif)
where .
6.2. Characteristic times
The previous stationary turbulent spectra straightly give the
diffusion coefficients and the pitch-angle frequency.
For the Langmuir waves injecting Eq. (37) in Eq. (50) we obtain
![[EQUATION]](img180.gif)
The pitch-angle µ is in the range
. The exact expression of the pitch-angle
frequency is unknown since the core of the
non-linear three waves interaction cannot be derived analytically.
Thus, approximatively is only derived from the
growth rate and the Eq. (51)
![[EQUATION]](img183.gif)
The ratio of the Langmuir acceleration time
to the Alfvén angular diffusion time
is
![[EQUATION]](img186.gif)
For a non relativistic thermal plasma with a kinetic temperature of
order of the Compton temperature , the thermal
speed of the ambient electrons is . The ratio
of the cyclotron to the plasma frequency cannot be greater than
in standard astrophysical conditions. The
Langmuir turbulence is then inefficient at high energy, and all
angular scattering and acceleration processes are dominated by
Alfvén turbulence, even if the Lorentz factor
(or a momentum ) which
balances the two rates tends to increase with an increasing angle.
Anyway, Langmuir turbulence is very efficient to accelerate particles
up to Lorentz factors of order of ten to one hundred for hotter
thermal electrons; which prevents an accumulation of low energy pairs
(the so called "dead-end" problem) and ensures an injection above the
threshold for resonant interaction with Alfvén waves.
As described in Sect. 2, the IC anisotropic emission of a
relativistic particle in a soft photon field is strongly reduced in a
small cone of opening angle . The particle
distribution can then be divided in two different contributions
depending on the pitch-angle of the particle.
The pitch-angle scattering by Alfvén turbulence is the
fastest process inside the cone. Thus, the particle distribution
function is weakly dependent of , and is
obtained by balancing, in the Fokker-Planck equation, the IC cooling
and the acceleration contributions. Note that at low energies, the
distribution function is also angle independent because the angular
diffusion is dominated by Langmuir waves whose pitch-angle frequency
.
Outside the cone the IC cooling is the dominant energetical process
and controls the shape of the distribution function.
6.3. Stationary solutions of the Fokker-Planck equation
6.3.1. Inside the cone
For , the distribution function in the disk
frame can be described by
![[EQUATION]](img196.gif)
where .
At high energies, and for ,
is solution of
![[EQUATION]](img200.gif)
The frequency obtained from Eq. (8) is
![[EQUATION]](img202.gif)
For momenta in the absorption range , and
from Eqs. (53) and (57) scales like
. Here corresponds to
the resonance with the mode of wave number ,
and . The acceleration time scales as p
like the cooling time, leading to a power-law stationary solution with
an index given by
![[EQUATION]](img209.gif)
For , particles resonate with the flatter
transfer spectrum and thus the diffusion coefficient varies as
with . This kind of
spectrum can be associated with an external source of Alfvén
turbulence from the accretion disk, at large scales
( ), and loaded by the MHD jet (see HP). In this
range, the acceleration rate does not follow the variation of the IC
cooling rate and the distribution function drops exponentially with
p.
The maximum momentum can be estimated by
the balance between the IC cooling and the acceleration times.
Moreover, as these times have the same energy dependence this balance
is verified for all momenta , and the spectral
index . Thus, the main effect of the Langmuir
turbulence at low energy and Alfvén turbulence at higher
energies is to built a flat energy distribution function inside the
cone, dominated by high energy pairs with a maximum Lorentz factor
. The high energy particles are submitted to a
strong pitch-angle scattering and are injected in the outer part of
the acceleration cone. The angular diffusion by the Langmuir
turbulence ( ) dominates at small energy the
angular diffusion by Alfvén waves ( ).
The low energy pairs are re-injected in the inner part of the cone
leading to a self-consistent mechanism (see Fig. 2).
![[FIGURE]](img217.gif) |
Fig. 2. Scheme of the acceleration mechanism.
|
6.3.2. Outside the cone
For , the injection process described above
can be treated as sources terms in the Fokker-Planck equation. We have
neglected the width of the acceleration cone, so the injection occurs
at .
The high energy particles are injected outside the cone at
, the cooled low energy pairs are re-injected
in the cone with a momentum . Therefore, we
introduce a source term for the population outside the cone:
![[EQUATION]](img222.gif)
where is the source flux.
The inhomogeneous Fokker-Planck equation (Eq. (56)) can be written
as
![[EQUATION]](img224.gif)
The external cooling frequency is of order
of . Thereafter all the coefficients are
replaced by there average value over µ;
, , and
. Expanding the distribution function in
Legendre polynomials, , we can solve the exact
solution of (64) for ; which gives
![[EQUATION]](img230.gif)
The spectral index is the positive root of
a second degree algebraic equation. Since the acceleration rate is
smaller than the IC cooling rate in this angular region, we get the
following estimate:
![[EQUATION]](img232.gif)
The fundamental component (n = 0) dominates at high energy and thus
we obtain a power-law distribution in . We
therefore obtain a power-law distribution similar to that usually
expected from shocks.
In a more general way, whatever are the momentum dependence of the
diffusion coefficient and the pitch-angle scattering frequency, the
Inverse Compton cooling process imposes the form of
to be in .
In the general case keeping the acceleration coefficient in Eq.
(64), the homogeneous solution is a pile-up distribution function
![[EQUATION]](img234.gif)
For a more general acceleration coefficient
, the solution is
![[EQUATION]](img236.gif)
where . The index
must be lower than 3, in order to have an IC cooling time
( ) dominant in the outer part of the cone.
Moreover we can develop the source term as
![[EQUATION]](img240.gif)
The inhomogeneous distribution function can be cast into the
form
![[EQUATION]](img241.gif)
![[EQUATION]](img242.gif)
where the heaviside function for
and, for
.
For , using the new variable
, the stationary solution is a power-law
function with an index equals to 4. Namely,
![[EQUATION]](img248.gif)
We can write the explicit value of the zero order coefficient
of the previous expansion as
![[EQUATION]](img250.gif)
Therefore, for , the particle energy density
distribution varies as, .
© European Southern Observatory (ESO) 1997
Online publication: June 5, 1998
helpdesk.link@springer.de  |