Efficient Monte Carlo Efficient Monte Carlo continuum radiative transfer continuum radiative transfer

with SKIRTwith SKIRT

Maarten Baes

2nd East-Asia Numerical Astrophysics Meeting, Daejeon, Korea

3 November 2006

Brussels

Why continuum radiative transfer…

• the ISM is extremely dusty

Detailed continuum radiative transfer simulations are

necessary to investigate the effect of dust on observable

properties of all dusy systems…

• dust strongly affects the radiation field at all wavelengths

- X-ray: scattering - UV and optical: extinction - IR and submm: emission

Radiative transfer equation

• we take into account the effects of- extinction

• condition of thermal equilibrium:

Radiative transfer equation

• we take into account the effects of- extinction- multiple anisotropic scattering

• condition of thermal equilibrium:

Radiative transfer equation

• we take into account the effects of- extinction- multiple anisotropic scattering- thermal dust re-emission, assuming thermal

equilibrium

• condition of thermal equilibrium:

Radiative transfer equation

• we take into account the effects of- extinction- multiple anisotropic scattering- thermal dust re-emission, assuming thermal

equilibrium- multiple dust grain populations

• condition of thermal equilibrium:

Monte Carlo radiative transfer

• probabilisitic technique >< deterministic technique

• RT simulations in which a large number of photons are followed individually through the dusty medium

• the trajectory of each photon is determined by (pseudo) random numbers

• the radiation field is reconstructed by classifying the photons by position, propagation direction, wavelength…

ADVANTAGES

• conceptually simple, natural treatment of emission, absorption and scattering

• all geometries possible (3D simulations)

• rather economic in memory → large grids are possible

• very flexible: multiple anisotropic scattering, polarization, kinematics, dust clumping…DISADVANTAGES

• Poisson noise

- error analysis is difficult

- accuracy goes as N-1/2 → efficiency !?

Monte Carlo radiative transfer

SKIRT

• Stellar Kinematics Including Radiative Transfer

• allows all geometries for sources and sinks: dust cells

• several dust cell geometries: spherical, cylindrical, cuboidal,…

Steinacker et al. 2003

SKIRT

• strongly optimized through the use of deterministic elements

- forced (first) scattering Witt 1977

- peeling-off technique Yusef-Zadeh et al. 1984

- continuous absorption Lucy 1999

- partly polychromatic photon packages Baes 2006, MNRAS, submitted

• computing power: dedicated cluster with 16 x 2 Gb memory

• two major modes:

- LTE → modelling the dust temperature distribution and the SED of dusty systems

- KINE → modelling the observed kinematics of dusty galaxies

SKIRT in LTE mode

• LTE radiative transfer:

- radiative equilibrium: energy absorbed = energy emitted

- the absorbed energy determines the dust temperature• frequency distribution adjustment technique

Bjorkman & Wood 2001 Baes et al. 2005, NewA, 10, 523

– no iteration is necessary

– immediate re-emission: guaranteed flux conservation

– works with all optical depths• polychromatic photon packages: very efficient

1D benchmark tests

Ivezić et al. (1997) benchmark tests

- star + spherical envelope

- V-band optical depths 1-1000

Polychromatic photon packages

(re-)emissioneach photon package initially contains photons of all wavelengths

exitif it leaves the galaxy: contribution to the SED at all wavelengths

scatteringloss of polychromatism

minimal computational overheadsignificant gain in efficiency

Baes 2006, MNRAS, submitted

2D benchmark tests

Pascucci et al. (2004) benchmark tests

- star + axisymmetric envelope

- V-band optical depths 0.1-100

SKIRT 2D

benchmark

SKIRT 3D vs benchmark

Efficiency of Monte Carlo RT

• “common wisdom” about Monte Carlo RT: numerically demanding

• comparison between SKIRT and other codes used in Pascucci et al.

SKIRT 2D: 2.5 MBySKIRT 3D: 58 MBy

MC RT codes can be very efficient when modern optimization techniques are used.

Limited memory usage is extra advantage when moving to 3D

Baes 2006, MNRAS, submitted

Application 1: Circumstellar discs

• large homogeneous survey of post-AGB stars- they all seem to be binary systems- they have a MIR excess due to dust starting at the sublimation temperature- MIR-submm SED and VLTI data suggest

circumbinary discs

• question: how do the temperature distribution and the emerging radiation field depend on the structure of the circumstellar medium ?

Application 1: Circumstellar discs

We can see some systematic effects, but in general the structure of the dust temperature distribution is rather insensitive to the structure of the ISM.

density temperature

Application 2: spiral galaxy atlas

• simulation of a large set of spiral galaxy models

- images at various inclinations and passbands - global and spatially resolved spectral energy distributions- attenuation maps- dust temperature distributions

• scientific goals

- investigate the systematic effects of physical parameters on the observables (luminosity, dust content, bulge-to-disc ratio, inclination…)

- construct an optimized galaxy dust mass estimator for IRAS, Spitzer, Akari,… data

- provide a database for statistical / cosmological applications

Optical depth

Bu

lge lu

min

osity

Optical depth B

ulg

e lu

min

osity

Application 2: spiral galaxy atlas

R-band images Spitzer MIPS 160 μm

images

Conclusions

• 2 modes: LTE and KINE

• SKIRT = efficient 3D Monte Carlo radiative transfer code

• uses efficient optimization techniques

• reproduces the 1D and 2D benchmark test easily

• ready to go….

- models for circumbinary discs around post-AGB stars

- atlas of dusty spiral galaxy models- simulations of accretion discs in the centre of

AGNs- kinematics of dusty galaxies and galactic nuclei

- your radiative transfer problem ???

Thank you…

EANAM 2004Japan

EANAM 2006Korea

EANAM 2008China

EANAM 2010Iran ?

EANAM 2012Belgium!

See you there !