7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
1/21
Numerical simulations of dynamics andemission from relativistic astrophysical jets
Petar Mimicawww.uv.es/[email protected]@mimichaninDepartment of Astronomy and AstrophysicsUniversity of Valencia
http://www.twitter.com/mimichaninhttp://www.twitter.com/mimichaninmailto:[email protected]:[email protected]://www.uv.es/mimicahttp://www.uv.es/mimica7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
2/21
Outline
1. Introduction: relativistic jets
2. Special relativistic (magneto)hydrodynamics3. Non-thermal particles in relativistic jets
4. Non-thermal emission processes
5. Overview of applications
6. Conclusions
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
3/21
Relativistic jetsactive galactic nuclei
gamma-ray
burststidal disruption
events
microquasars
appear in a wide variety of astrophysicalscenarios (AGN, TDE, GRB, QSO)
common properties: powered by an accreting compact object collimated relativistic outflows non-thermal emission (synchrotron, IC)
B-fields present, not well constrained
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
4/21
Simulating relativistic jets
1. relativistic hydrodynamics simulationfinite-volumesmethod of linesshock-capturingRiemann solversoptional: coupling to non-thermal particles
2. non-thermal particle evolutionphenomenological shock accelerationradiative and adiabatic losessemi-analytic electron-kinetic eq. solver
spatial advection
3. radiative transfertime-dependent emission and absorptionrelativistic effects (beaming, Doppler)
light-travel timessynchrotron, inverse-Compton scattering
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
5/21
Relativistic hydrodynamics@D
@t+r (Dv) = 0
@S
@t+r (S v + pI) = 0
@
@t+r (SDv) = 0
mass conservation momentum conservation energy conservation
h =5
2
P
c2+
s9
4 P
c22
+ 1 TM approximation to Synge equation of stateMignone et al. Astrophys. J. Supplement 160 (2005) 199de Berredo-Peixoto et al. Modern Phys. Lett. A20 (2005) 2723
W :=1
p1 v2/c2
h := 1 +"
c2+
p
c2
Lorentz factor
specific enthalpy
:= W S := hW2v := hW2c2 p Wc2
relativistic rest-mass density relativistic momentum density relativistic energy density
v
rest-mass density pressure flow velocity
primitive variables must be obtained (recovered) from the conserved ones no analytic solution in general, numerical procedure must be used (and it can fail!)
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
6/21
MRGENESIS
MRGENESIS(Aloyet al. 99 ApJS , Leismann et al.05, A&A, Mimica et al. 07, 09 A&A)
finite volume approach
method of lines: separate semi-discretizationof space and time
time advance: TVD Runge-Kutta methods of2nd and 3rd order
high-resolution shock-capturing scheme
inter-cell reconstruction: up to 3rd order using
PPM algorithm numerical fluxes: Riemann solvers
RMHD: constraint transport to conserve B
orthogonal coordinate systems: Cartesian,cylindrical, spherical
MPI + OpenMP: scales up to 10K cores
HDF5 library for parallel I/O
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
7/21
Non-thermal particles model: relativistic shocks accelerate
electrons to high energies
phenomenological source term:
electron-kinetic equation
Q() = Q0s; min max
:=1 v2/c2
1/2
@n
(, t
)@t + @@ ( n(, t)) = Q()
= ka ks2
adiabaticcompression orexpansion
synchrotronlosses
ka =1
3
D ln
Dtks =
4TB2
3m2ec2
particle energy losses/gains:
special case: pure synchrotron losses (ka = 0): (t) =(0)
1 + ks(0)t
c(t) =1
kstcooling break: maximum for a given time:
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
8/21
Solving electron-kinetic equation
test problem: n(, 0) = K
min
q
@n , t
@t+
@
@( n(, t)) = Q()
coolingbreak
= ka ks2
n(i, t) = n(i(0), t)e2kat 1 + i(0)
ks
kaekat 1
2
i(t) = i(0)ekat
1 + i(0)
ks
ka
ekat 1
1
= e3katN(i(0), i+1(0), t)
moving bins solver
coolingbreak
N(i, i+1, t) :=
Zi+1
i
d n(, t)
Mimica+Mon. Not. R. Astron. Soc. 407 (2010) 2501
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
9/21
Synchrotron radiation
synchrotron emissivity: j() =
p3e3
4mec2B Zd n() R
0
20 =
3e
4mecB
R(x) =1
2
Z
0
d sin2 F
x
sin
F(x) = x
Z1
x
dK5/3()
n() = n(min)
min
q
; min max
j() =
p3e3B
8mec2n(min)
qmin
0
(1q)/2H
02min
, q,max
min
H(x,q, ) =
Zxx/2
d (q3)/2 R()interpolate the function:
advantage: synchrotron computation cost reduced by a factor 50-100 tradeoff: large interpolation table (> 2GB) needs to reside in memory
interpolation table computation: a week on a 16-core machine
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
10/21
Inverse Compton scattering
0: incoming frequency
1: outgoing frequency1: lower electron Lorentz factor cutoff
IC emissivity for monochromatic
incoming emission is analytic problem 1: compute theemissivity for non-monochromaticincoming radiation
problem 2: compute theincoming radiation spectrum
solution: large interpolationtables
Mimica & Aloy,Mon. Not. R. Astron. Soc.421 (2012) 2635
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
11/21
Radiative transfer
dI
ds= j + I
radiative transfer equation:
emitting volume
t1
t2
t3
virtual detector
(observer)
motion (v~c)
towards observer
T1T2T3
s
s0
for a fixed T, equation gives an isochrone (s, t) alongeach line of sights = c(t
T) + s0
j
s
T t
: intensity: emission, absorption
: observer time : jet evolution time
: path towards the detector
synchrotron,inverse-Compton
synchrotronself-absorption
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
12/21
SPectral EVolution code
S. Tabik et al. Computer Physics Communications 183 (2012) 1937
SPEV(Mimica et al., Astrophysical J. 696 (2009) 1142) : non-thermal electron transport and evolution equations time- and frequency-dependent radiative transfer in a dynamically changing background parallelization: MPI (over detector pixels), OpenMP (over particles)
Mimica et al.,Astrophysical J.696 (2009) 1142
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
13/21
SPectral EVolution code
S. Tabik et al. Computer Physics Communications 183 (2012) 1937
SPEV(Mimica et al., Astrophysical J. 696 (2009) 1142) : non-thermal electron transport and evolution equations time- and frequency-dependent radiative transfer in a dynamically changing background parallelization: MPI (over detector pixels), OpenMP (over particles)
Mimica et al.,Astrophysical J.696 (2009) 1142
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
14/21
Simulation building blocksDepending on the application, select one from each row:
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
15/21
Parsec-scale jets
trailing components
trailing components
main component
main component
Mimica et al.,Astrophysical J.696 (2009) 1142
jet perturbation seen as asuperluminal (main) component
radio observations do not directlysee the jet
SPEV, 128 frames, 270 x 18 pixels, 3 frequencies
per run: 100 Kh, 0.5 Tb hydro data, 2x105 Lagrangianparticles, 2x106 line-of-sight segments
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
16/21
syn. peak
Compton peak
Mimicaetal.,Astron.Astrophys.418(
2004)947
Mimica&Aloy,Mon.Not.R.Astron
.Soc.421(2012)2635
Mimic
a&Aloy,Mon.Not.R.Astron
.Soc.401(2010)525
nonmagnetized
weakly magn.
strongly magn.
Blazar flares
blazar light curves: resol. (200x200) ( x T)
900 different shell collisions, 200 Khours
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
17/21
1D simulations: 106 zones, 108 iterations50 - 100 Kh / run 104 snapshots / run
GRB afterglows
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
18/21
Mimica, Giannios & Aloy 2009 , Mizuno et al. 2009
Ejecta-medium interaction
1 1
B
:=B
2
4c2
Sari & Piran 1995
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
19/21
Gamma-ray burst afterglows
references:
Giannios+Astron. Astrophys.478 (2008) 747Mimica+Astron. Astrophys.494 (2009) 879Mimica+Mon. Not. R. Astron. Soc. 407 (2010) 2501
optical flash almost neverobserved: are almost all GRB jets
magnetized?
1
1
optical flash
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
20/21
Required numerical resolutionNeed to resolve o in at least 104 zones for o ~ 10.total computing domaing: Rtot / o ~ 100 - 1000Ntot ~ 106 - 107 Niter ~ 107 - 108(CFL~0.1)
~ 500 - 1000 CPUs/job usedeach job needs ~ 200 (wall clock) hours to finish.
since xo , we would need ~ 2000 h in 500 -1000 CPUs to run realistic modelswitho ~100.
realistic GRB parameters:Rtot / oAG ~ 3x105 Ntot ~ 3x109 Niter ~ 3x1010
6x105 h on 500 CPU 3x104 h on 10000 CPUs
3x105 (wall clock) on 10000 CPUs (o ~100).
0
1
log
-5 -4 -3
log x
27
28
Niter
x10
3
Fig. A.1.Time (upper panel) and the number of iterationsNiter(lower panel) needed to resolve the Riemann problem in planar
coordinates as a function of the spatial discretization x..
10 20 30
0
0
0.5
1
Fig. A.2. Time needed to resolve the Riemann problem in pla-
nar coordinates as a function of the initial Lorentz factor 0 for
x = 1.5625 105.
Mimica+Astron. Astrophys.494 (2009) 879
7/31/2019 Talk by Petar Mimica at the Conference on Computational Physics 2012 in Kobe on October 15th, 2010
21/21
Conclusions MRGENESIS + SPEV is:
a simulation framework that has been successfully applied toAGN, GRB and TDE jets
modular and adaptable to computing at various scales:
small scale computing is used when fast feedback is morevaluable than high spatial and temporal resolution
supercomputing is used when computing time-dependentimages or when performing parameter studies
computation of emission from numerical hydrodynamicsimulations enables direct comparisons with observations
future (MRGENESIS): resistive RMHD, scaling >104 cores
future (SPEV): improved inverse-Compton, polarization,radiative transport in GR