Magnetic Energy:
Transport and Dissipation
Åke Nordlund Niels Bohr Institute
University of Copenhagen
Spineto Summer School, June 2012 SWIFF WP3: Coupling at the Sun
The Magnetic Energy Equation
To discuss and understand the transport and
dissipation of magnetic energy, it is fundamentally useful to consider the equation that controls the change of magnetic energy with time:
The three quantities on the right have names after
historic persons: The Poynting flux, the Lorentz work, and the Jould dissipation.
So, let’s deriv it!
Induction eqn:
Magnetic energy:
Vector identity:
Electic current:
Electric field:
All together:
Collect terms:
or
where
Time evolution of magnetic energy in sub-surface magnetoconvection
starting with zero field
Poynting flux
Blue = up, red = down
Evolution of Poynting flux
Energy equation parts, early
Time evolution
Time evolution detail
Surface region
Magnetic dissipation in
a horizontal slice just below the surface, with
24 km grid size
Ditto, with 12 km grid size
Late time, high res balance
Zoom into small ROI
Work, dissipation, sum of the two:
Ditto, smoothed
Box car averaged over 5x5x5 cells
Magnetic energy density evolution
Dependence on numerical resolutoion
Very long time evolution
The same, as averages
Magnetic and kinetic energy density
Transition to 2x numerical resolution
Surface detail, 24 – 12 km transition
Small ROIs, in dissipation
Now up into the
Chromosphere & Corona!
Solar Chromosphere & Corona Conditions:
Which of the MHD approximations are valid?
Enough charged particles to make a plasma?
Certainly! Typical Debye lengths are a few mm!
Slow enough motions relative to the speed of light?
Well ”yes, most of the time!”
Frequent enough collisions for thermal behavior?
No!! Mean-free-paths similar to loop lengths
Coulomb collisions dominate
Have interesting run-away behavior!
The one most important macro-scale kinetic effect:
Particle acceleration!
Might be a recursive accerelation (ping-pong) effect
But more likely it is a direct acceleration effect!
Mean free path comparable to structure scales
Coulomb cross sections decrease with energy!
Minority population can have run-away
Need particle-in-cell codes to investigate!
Computational Costs
MHD codes, cost per cell
On Intel Nehalem, IBM Power-6, and similar CPUs
Costs of the order 2-10 microseconds per cell-update
Depending on equation-of-state, radiative transfer, …
Particle-in-cell (PIC) codes
On similar CPUs
Costs of the order 1 microsecond per particle-update
Need 10-50 particles per cell
Naïve bottom line: only about 5-10 times slower
However...; physical scales!!
Masson et al. 2009
Flare ribbon
FROM OBSERVATIONS TO PIC
MHD
simulation
?
PIC
simulation
06:27 UT
TRACE and RHESSI observations
Image: H. Reid
12-25 keV (>50%) 25-100 keV (>50%)
A dense electrons beam is accelerated by the E- field in the reconnection region and travels along the fan magnetic field lines. The impact on the photosphere creates X- rays and UV emission.
INITIAL EXTRAPOLATED BFIELD
Potential extrapolation, side view
Potential extrapolation, top view
3D nullpoint reconnection
nullpoi nt
magnetic field lines
Sequence of modeling
SOHO cut-out, potential field extrapolation, MHD model, PIC-cutout, particles accelerated
APPLIED DRIVER
The elliptical driver applies a sub-Alfvénic velocity of about 20 km/s to the lower boundary after about 150 sec of relaxation time.
Flux emergence on 16th Nov ‘02
SPINE/FAN MOTION AND J//B
About 5h 30min. real time
RESISTIVE MHD SIMULATION
Code: Stagger MHD code (6th order space, 3th order in time)
Computational box size: 62 x 175 x
100 [Mm]
Stretched mesh, max. resolution: 80 [km]
Boundaries: closed
SPINE/FAN MOTION AND J//B
About 5h 30min. real time
Initial average vertical structure
Overview of MHD experiments
Stretched mesh, grid size = f(index)
Magnetic field line connectivity
Electric currents in the current sheet
Relative displacement of the spines
Current density in the fan plane
Perpendicular and parallel electric fields
Evolution with time of
magnetic energy and magnetic dissipation
RESULTS MHD
Current density in horizontal slice Current parallel B
Magnetic dissipation (blue surface) and magnetic field lines near the null
Rate of change of magnetic energy, and rate of magnetic dissipation
Why no flare?
1. Numerical resolution is not the reason, since
Runs with different resolution are very similar
All runs behave as expected from scaling arguments
2. Kinetic effects (non-MHD behavior) would not helpt – there is simply not enough free energy available for a C-class flare
Free energy is proportional to the angle difference across the current sheet, which we show scale with the drive speed, indendent of resistivity …
… and the solar boundary motions were much slower!
MHD experiment passing over to PIC (Particle-in-Cell) experiment
MHD snapshot provides
A cut-out of the ”dome” with uniform mesh
Initial conditions (including electric currents)
Boundary conditions (vector B-field at boundaries)
PIC setup
Box size covers the ”dome” structure
Stops about 1.3 Mm above the solar surface
Mesh size 70, 35, and 17.5 (!) km
The 17.5 resolution case has more mesh points (~ 3 billion) than any MHD model so far
Particle-in-cell simulations
Field equations (Maxwell):
Equations of motion (Newton):
Large scale PIC-simulations
Copenhagen PhotonPlasma Code
MPI-parallelized PIC code
Includes particle-interactions
Most relevant in the corona: Coulomb scattering
Modular: also Compton scattering, pair creation & annihilation, ...
Could also do collisional, time dependent ionization this way!
Scales to thousands of CPUs / cores
Pleiades (NASA/Ames); ~47.000 cores, tested up to 4096 cores
Uses about 1 core-microseconds per particle-update
(95% efficiency from 8 to 4096 cores)
PIC Scaling
Gyroradius:
Debye length:
PIC Scaling
L_Debye ~ 0.5 dx
L_skin ~ 6 - 8 dx > L_Debye
v_drift < v_thermal_ion ~ v_sound < v_Alfven <
v_thermal_electron < c
100 Mm
175 Mm
PIC runs at JUGENE (Jülich)
The largest run, 5L
Grid: 2518x1438x923 (3.3 billion cells)
Particles: 135 billion
Ran for 36h on 262.144 (2^18) cores on JUGENE
about 10 million core-hours
Scaling on JUGENE
Both the MHD code and the PIC code scale to the full size of the machine (218 cores)
Initial PIC state: 44x25x16 Mm box
Projection planes
VERTICAL
PROJECTION
HORIZONTAL
PROJECTION
CURRENT DENSITY IN THE NULL PLANES
Impact area of electrons
Comparison with TRACE and RHESSI observations
Image: H.Reid
12-25 keV (>50%) 25-100 keV (>50%)
A dense electrons beam is accelerated by E//B in the reconnection region and travels down the
inner spine and along the fan magnetic field lines.
The impact on the
photosphere creates X- rays and UV emission.
J//B maximum
Energy distribution
Different simulations
Energy histogram, sampled from Ntot=1.35 1011
The dN/dE power-law index of -1.75 Implies that the electric
current is mainly carried by the low energy electrons, while energy is mainly carried by the high energy ones.
The histograms here show dN/dlnE = E dN/dE, so the dN/dE
distribution index is steeper by one unit; i.e. has slope approximately -1.75
PARTICLE POSITIONS
IN THE POWER-LAW TAIL
PARTICLE POSITIONS
IN THE POWER-LAW TAIL
Location of accelerated particles
Inner spine Outer spine
Fan plane
Trace of accelerated particles
Losers, in low energy part of tail
X - Z Plane
X - Y Plane
Tracing particles in the power-law tail...
MOVIE WITH EVOLVING BACKGROUND
add movie of accelerated particles from lower part of power tail, their trajectories, edotv, and energies
add screenshot of the failed ones.( their trajectories, edotv, and energies)
Lower energy power law particles occur in essentially the same region
Total Current Density
Total electric
current density in slices through the null point
neighborhood
Electric current contributions
Parallel (accelerating) electric field
Note the scale! stabilizes
in a few seconds
Summary
Need to consider kinetic effects in context!
micro-scales; modified dissipation details
meso-scales; dissipation hierarchy
macro-scales; particle acceleration!
New tool: Particle-In-Cell (PIC) simulations
can now handle resolutions ~ MHD a few years ago
careful scalings need to be applied
radiation can be included (Monte Carlo RT)
time dependent ionization is possible
synthetic observations over a broad range (gamma ray, X- ray, visual, radio, ...) possible ..
Take home: myths to kill!
Ideal MHD
School book magnetic fields
Sweet-Parker current sheets
Scalings cannot be done
No collision -> no magnetic feld aligned E