The workshop consisted in the cross-collaboration of the
participants on the 11 projects listed below. The subjects range from numerical
methods (6) to properties of real and simulated turbulence (5,8,9), with,
in between, turbulence generation mechanisms, large scales (1,2) and small
scales (11), and the relaxation of collisionless/dispersive turbulence
in (3,7,10). Some subjects attracted more collaborators than others. However,
everyone took profit of knowing in some detail what the others were working
on.
1. List of projects
(1) MHD within shear flows - F. Lignieres
and J. Leorat
(2) Forced helical turbulence in spherical
geometry - Wolfgang Dobler
(3) Kinetic effects in the expanding
solar wind: Petr Hellinger, Roland Grappin, Pavel Travnicek, Andre Mangeney
(4) Parametric instability of Alfven
waves in the expanding wind.: L.Primavera and R.Grappin
(5) Origin of the magnetic excess in
3D incompressible MHD turbulence (continuation of last CIAS workshop):
W. Mueller and R. Grappin
(6) High Order Schemes for turbulence:
Tony Arber and Wolfgang Dobler
(7) Filamentation of dispersive Alfven
waves: Andre Mangeney, Petr Hellinger, Pavel Travnicek
(8) Finite size scaling of the magnetic
field magnitude from the simulated MHD turbulence: Bogdan Hnat and Tony
Arber.
(9) The Lyapunov spectrum of WIND--MFI
data: B. Hnat and L. Nocera
(10) Turbulence in the DNLS-Burgers
equation: L. Nocera
(11) Electron acceleration in a turbulent
electrostatic wave field: Filippo Pantellini and Petr Hellinger
=> Top of page
2. Reports on projects
(1) MHD within shear flows
Collaborators: F. Lignieres and J. Leorat
Our objective is to study the non linear dynamo action
by shear turbulence in stably stratified atmosphere like stellar radiative
zone. Such shear
flows occur within a flow at larger (stellar) scale,
not represented in the simulations. Some modelisation of the effects of
the large scale flow,
considered as given, on the simulated scales of the shear
is thus needed. This may be done for example, by applying a uniform rotation
or
by imposing the mean flow to keep a given uniform gradient.
During this week, we considered two types of shear flows
in slab geometry and started to experiment numerically with a code solving
the non linear
MHD equations.
a) In the first configuration a shear layer anticyclonic
with respect to the local rotation may be unstable at a rather low Reynolds
number (cf Lezius & Johnston, 1996, JFM, 77,153-175). We have checked
the occurence of the linear instability (the resulting flow is 2D). We
have now to verify if the flow in non linearly saturated regime is an efficient
kinematic dynamo (for a related example, see P.C. Matthews' work)
b) In the second configuration , turbulence is maintained
by an uniform shear (see for example Schumacher JFM,441,109-118, 2001).
The
numerical study has just started, and stratification
effects will be examined next.
=> Top of page
(2) Forced helical turbulence in spherical
geometry
Collaborators: Wolfgang Dobler
Magnetic helicity conservation is considered one of the
big problems for cosmic turbulent dynamos, which are operating at very
large magnetic
Reynolds numbers. The magnetic helicity can only evolve
on the diffusive time scale, which is determined by the molecular magnetic
diffusivity, not a turbulent
diffusivity, and the corresponding timescales are extremely
long for stars or galaxies.
The project is dedicated to investigating the helicity
balance (production + flux) for spherical turbulent dynamos. Currently
the dynamo is realised
by imposing, within the sphere, a stochastic helical
forcing at a given small wavenumber. The velocity field quickly evolves
into a helical state
with positive sign of helicity in one hemisphere and
negative in the other (just as the helicity of the imposed forcing).
The magnetic field generated by this flow is quite asymmetrical
and has prominent small structures even after the (statistically) steady
state is
achieved.
A comparison with a simulation where the forcing has
the same sign of helicity in the whole sphere shows that in this latter
case (after an
almost identical phase of linear growth) the field grows
faster and attains a considerably higher steady-state value than in the
case
discussed before. This difference is most probably due
to two effects, the different efficiency of magnetic field generation (which
manifests itself in the
different critical magnetic Reynolds numbers for the
corresponding mean-field dynamo models with symmetric or antisymmetric
alpha-effect),
and the exchange of magnetic flux and magnetic helicity
through the equatorial plane. The next step is to disentangle these two
effects, which requires us to adapt
tools from slab simulations for calculating magnetic
helicity and its flux.
In the long perspective,, the addition of gravity, rotation
and an energy equation with central heating and surface cooling will allow
us to drop the
forcing altogether and to generate helical turbulence
and the accompanying magnetic fields in a much more realistic fashion.
=> Top of page
(3) Kinetic effects in the expanding
solar wind.
Collaborators: Petr Hellinger, Roland Grappin, Pavel
Travnicek and Andre Mangeney
We consider hybrid simulation taking into account solar
wind expansion. For different plasma parameters we consider how the ideal
plasma system
could stop adiabatic perpendicular cooling that results
from a radial expansion. In the case of cold plasma there is no way to
stop the
cooling while for a hot plasma fire hose instabilities
are able to transfer energy from parallel to perpendicular direction and
to stop
perpendicular cooling. The situation is strongly modified
when we add to the plasma a population of alpha particles with a drift
velocity
with respect to the protons. In this case the ratio of
the drift velocity to the local Alfven velocity increases during the expansion.
When the ratio exceeds a certain treshhold (between 1
and 2) various instabilities appear that slow down the alphas and
heat alphas and protons. The generated waves are able
to stop the adiabatic cooling of protons even for the
cold plasma. Such instabilities may explain the observation
that show that solar wind protons are not adiabatic.
Another process we considered during the workshop is
bringing the energy from large to small resonant scales: we considered
a shear (parallel
to the mean field). We currently are considering two
cases:
a) almost all energy is within the shear
b) the shear and the wave have comparable energies.
The first case corresponds to waves within large-scale
stream shear; the second to parallel Alfven waves interacting with perpendicular
(non propagating) Alfven waves, which might lead to regimes
resembling turbulence.
=> Top of page
(4) Parametric instability of Alfven
waves in the expanding wind.
Collaborators: L.Primavera and R.Grappin
A steady rise of z-/z+ (the ratio of reflected over outward
propagating Alfven wave energy) with increasing heliocentric distance is
observed
by Ulysses, with a saturation at about 2.2 AU. A comparable
evolution has also been obtained via simulations starting with
circularly polarized Alfven waves propagating in a homogeneous
medium, subject to parametric instability. The question is whether this
mechanism
still works within the expanding solar wind. This question
is related to the theme of the relative importance of the compressible
and incompressible
components in this turbulent medium. It is clear that
even a circularly polarized Alfven wave propagating within an expanding
medium leads to
compressive fluctuations. However, at the same time,
expansion slows down significantly nonlinear coupling, as has been shown
in the early work by
Grappin Mangeney Velli 1993 and Grappin & Velli 1994.
We used the expanding box model (coordinates are comobile
with a given uniform radial wind) to study first an all-radial case (mean
field, as well as
wavevector are radial). For reasonable parameter values
(beta=0.1, expansion parameter 0.1, corresponding to 6 hours waveperiods),
the
numerical simulations lead to a growth rate (measured
in wave frequency) comparable to the unexpanding case. This is not unexpected
in the radial
case, as both the relative wave amplitude and plasma
beta increase with distance almost at the same rate.
We plan to study soon the more realistic non-radial case
where both the wavevector and mean field rotate (in opposite directions)
when distance
increases, so that the wave no longer remains a circularly
polarized wave.
=> Top of page
(5) Origin of the magnetic excess in
3D incompressible MHD turbulence (continuation of last CIAS workshop)
Collaborators: W. Mueller and R. Grappin
Simulations by Biskamp and Mueller (2000) found that decaying
turbulence starting with equipartition between magnetic and kinetic energies
and
with no mean magnetic field give rise to a Kolmogorov
slope for the total (magnetic+kinetic) energy spectrum. Our aim is to use
a simple approach to obtain a phenomenology of the process leading to the
magnetic excess. We find that the case with no magnetic helicity is well
explained by a balance between the Alfven relaxation (linear propagation
within the local mean field, that is basically a non-local coupling in
Fourier space) and the dynamo effect, estimated as a local coupling.
A problem remains for the case with large-scale magnetic
helicity. In that case, the magnetic excess is much enhanced, and not only
in the large scales where the magnetic helicity is confined, but also at
medium scales. Hence, classical (EDQNM closure) estimations of the effects
of helical terms cannot explain the large magnetic excess at medium scales.
We tested an hypothesis of non WKB propagation in which the large scale
helical structures are fixed, but a wave propagating in this structure
shows no more magnetic excess than without helicity. We are thus left with
the conclusion that a large-scale magnetic excess leads to a cascade with
reduced Alfven effect, even though the (isotropic) Alfven time is significantly
smaller than the nonlinear time. The only explanation is that the average
Alfven time is larger than the
isotropic average, due to anisotropies in spectral space,
most structures having their wavevectors perpendicular to the local mean
field. This anisotropy would be larger when large scales have a large magnetic
helicity. This idea has to be tested in detail.
=> Top of page
(6) High Order Schemes for turbulence
Collaborators: Tony Arber and Wolfgang Dobler
High order finite difference schemes have the advantage
over spectral codes of being easy to code and parallelise. For compressible
simulations in which shocks may form spectral techniques are also unsuitable.
The aim was to evaluate the effectiveness of a 6th order in space, third
order in time finite difference code for strongly driven turbulence. This
has high accuracy for smooth regions but is capable of employing localised
shock viscosity where needed. Results were compared directly with those
from a 2nd order shock code based on Lagrangian remap techniques. By testing
the finite difference code on simple tests such as the Brio-Wu shock we
determined that upwind biasing, shock resistivity and some thermal conduction
were required to remove all false oscillations. With these parameters set
the code was tested on a volume forced turbulence simulation where the
forcing was through delta-correlated, large amplitude Beltrami waves. The
code does work for this problem. Hence, the same high order scheme can
be used for problems ranging from linear waves, weakly non-linear interactions
and strongly driven turbulence. However, for the strongly driven turbulence
test the results were no more accurate than those from the shock capturing
code which suggests that the optiminal strategy for strongly non-linear,
compressible turbulence is probably a PPM or ENO based scheme. The high
order finite difference scheme is freely availably to all participants
at the workshop for future work.
=> Top of page
(7) Filamentation of dispersive Alfven
waves
Collaborators: Andre Mangeney, Petr Hellinger and Pavel
Travnicek
We consider the filamentation of a dispersive Alfven wave
in low beta plasma. This work is motivated by Cluster II observations behind
a low Mach number quasi-perpendicular shock. This study is done in two
steps. First, we consider a nearly perpendicular shock with a relatively
low beta and we study the formation of the Alfven wave behind the shock
with a two-dimensional (2-D) hybrid code. Second, we have started a detailed
study of the filamentation process using 2-D (and in the future 3-D) hybrid
code: we start with a homogeneous plasma with a single, left handed, circularly
polarised Alfven wave. The Alfven wave is unstable in the transverse direction:
The filamentation process leads to magnetosonic waves perpendicular to
the Alfven wave propagation.
=> Top of page
(8) Finite size scaling of the magnetic
field magnitude from the simulated MHD turbulence.
Collaborators: Bogdan Hnat and Tony Arber.
Statistical properties of the interplanetary magnetic
field fluctuations can provide an important insight into the solar wind
turbulent cascade. Analysis of the Probability Density Functions (PDF)
of the velocity and magnetic field fluctuations has shown that these exhibit
non-Gaussian properties on small time scales while large scale features
appear to be uncorrelated. Recently we have applied the finite size scaling
technique to explore the scaling of the magnetic field energy density fluctuations
as seen
by WIND. We have found a single scaling sufficient to
collapse the curves over the entire investigated range. The rescaled PDF
followed a non Gaussian distribution with asymptotic behavior well described
by the Gamma distribution arising from a finite range Levy walk. Such mono
scaling suggests that a Fokker-Planck approach can be applied to study
the PDF dynamics. This finite size scaling technique has never been applied
to the simulated MHD turbulence before. During the workshop we applied
this method to the magnetic field magnitude time series and compared it
with the results obtained from the real observations. Interestingly, we
identified a dual scaling (in time) which, in general, resembles that found
in the observed data from the real driven systems such as the magnetosphere.
The PDFs for small scales are consistent with the Levy distribution with
alpha index of about 1.32 while the large scale PDFs are subdiffusive with
the scaling index of about 0.3. The time scale where this break occurs
can be related to the turnover time of the smallest driven structure in
the simulation. This result will be investigated further by studying different
driving methods in the simulation. Finally, we will attempt to apply the
results of this analysis to the physical turbulent systems where the dual
scaling has been identified.
=> Top of page
(9) The Lyapunov spectrum of WIND--MFI
data
Collaborators: B. Hnat and L. Nocera
This project is aimed at applying modern concepts and
techniques of signal data analysis to the time series provided by the observations
of magnetic field fluctuations in the solar wind as performed
by the MFI probe on WIND; it was conceived entirely during the last week
of
the workshop and it is part and complement to the project
on the computation of the PDF of the same data.
Specifically, our task is to compute the whole spectrum
of Lyapunov exponents of the time series using the technique of phase space
reconstruction and to estimate the information dimension
of the series: this latter point is a contribution to the debate on whether
solar wind fluctuations are to be regarded as deterministic
chaos (finite fractal dimension) or rather as infinite dimensional stochastic
noise.
The large amount of records in the series (N>10^6, sampled
at an interval of 92 sec for more than three years) provides an excellent
opportunity also to test several aspects of the phase
space reconstruction technique in situations of large embedding dimensions
D, viz Eckman's 'N>10^D' conjecture, Kaplan and Yorke's
information dimension formula and Hardy's and linear reconstruction algorithms.
Preliminary results with embedding dimensions as large
as 7 show that all but one of the Lyapunov exponents are positive and that
the information dimension of the time series should be larger than 6.
=> Top of page
(10) Turbulence in the DNLS-Burgers
equation
Collaborators: L. Nocera
This project is motivated by the numerical simulations
of quasi-parallel propagating Alfven waves in the solar wind which can
be described by the Derivative Nonlinear Schrodinger Burgers (DNLS-Burgers)
Equation. Specifically, during these simulations, a turbulent state appears
to converge to a state where the magnetic field is mostly concentrated
into isolated peaks. These peaks, or 'structures', look strikingly similar
to the solitons of the DNLS equation with zero dissipation, though of course,
these solitons cannot solve the complete DNLS-Burgers equation.
To explain this state of affairs, we conjecture that
those structures appear due to a relaxation process much in the same way
as it happens for two dimensional turbulence, according to Taylor's hypothesis.
We verified that the properties hold true for the DNLS-Burgers equation:
1. existence of nonlinear invariants; 2. existence of
selective dissipation of invariants; 3. existence of an inverse cascade.
In particular we verified that hypothesis 2 is actually
met by applying the general perturbation theory of solitons to the DNLS-Burgers
equation; we also verified numerically that this property holds true by
actually computing the invariants for the specific case
in which an initial non vanishing oblique algebraic soliton state was evolved
according to the DNLS-Burgers equation. Though proving hypothesis
3 was not possible on general grounds, we took space spectra out of the
simulations of the DNLSB equation and found that the inverse cascade process
indeed exists in the specific situation of turbulence evolving out of the
same algebraic soliton state mentioned above.
Last, using techniques of variational calculus, we found
that the distribution of the magnetic field is given by the 'hyperbolic
vanishing soliton', which was finally observed in our direct numerical
simulations of the DNLS-Burgers equation.In conclusion, we proved that
turbulence in the DNLS-Burgers equation can evolve into self-organized
states.
=> Top of page
(11) Electron acceleration in a turbulent
electrostatic wave field
Collaborators: Filippo Pantellini and Petr Hellinger
The plasma of the upper solar chromosphere is characterized
by the fact that the collisional mean free path is large compared to the
typical wave length of the electrostatic plasma waves. Under such conditions
the waves may stochastically accelerate (Fermi mechanism) a fraction of
the electrons to high energy enabling them to escape upward in the solar
atmosphere. If the acceleration mechanism is efficient enough to overcome
the thermalizing effect of collisions a non thermal electron population
is generated. The presence of non thermal electron populations in the solar
atmosphere can substantially affect the transport properties of the atmospheric
plasma. For example, transport properties are of fundamental importance
in the problem of the heating of the solar corona.
During this workshop we could define what essential ingredients
a numerical code must have in order to simulate the acceleration. For example,
what local characteristics and what kind of height dependence shall one
assume for the spectrum. A working fortran version of the code has been
written during the workshop. The code still needs some testing, but the
first, and most important steps, have been successfully accomplished during
the workshop.
=> Top of page