Copyright © 2013 by Annual Reviews. All rights reserved
|
| Annu. Rev. Astron. Astrophys. 2013. 51:63-104
Copyright © 2013 by Annual Reviews. All rights reserved |
3D dust RT is a rich and diverse field, with applications across a broad range of astrophysical topics from dust near stars to entire galaxies. Correctly modeling the effects of dust on the transfer of radiation is critical to studying many astrophysical objects, including the dust itself. Recent years have seen an impressive improvement in observational capabilities across the electromagnetic spectrum, and this has shown that the dust distribution in many regions is strongly 3D. This requires methods to compute the dust RT that can handle 3D structures and return solutions in a reasonable amount of time. The most common 3D dust RT solver is based on MC techniques, with RayT features in its modern accelerated form. A few applications have used pure RayT solvers. Both methods face the challenges of grid discretization, determination of uncertainties in solutions, and accurate comparison between observations and the model calculations. Almost 30 codes are currently able to deal with the full 3D dust RT, with code variations arising from the prime field of application. There is no 3D dust RT benchmark; currently, code comparisons are done using 2D benchmarks.
Several trends indicate that the future of 3D dust RT is bright. The number of people actively involved in 3D dust RT is growing, and the number of new published codes has increased significantly in recent years. A 3D approach to modeling complex distributions is becoming common in many fields requiring 3D dust distributions. The continuing increase in available computing power and storage will support this trend, allowing a full transition from 2D to 3D dust RT for all objects showing 3D signatures. A prominent example of this trend is circumstellar disks with (proto)planets, where the MHD simulations have been 3D for years, dust RT modeling often was 2D, and observations are now reaching the resolution necessary to identify the 3D signatures of disk deformation due to a planet. In addition, modern online tools are expected to support the access to the codes by users through sophisticated interfaces.
8.3. Future Benchmarks
For progress in 3D dust RT to continue, 3D dust RT benchmarks need to be
established. Given the complexity of the codes, increasing number of
acceleration algorithms, and large number of specific applications, it
is critical to provide a quantitative comparison between
codes. Experience with existing dust RT benchmarks and similar efforts
in other areas indicate a suite of 3D dust RT benchmarks is
needed. Ideally, each benchmark would focus on a particular part of the
RT solution (e.g., scattering, polarization, equilibrium dust emission,
or nonequilibrium dust emission) in a 3D geometry. This would provide a
clear test of different aspects of 3D dust RT and support the
participation of all codes in at least part of the suite.
Given the impressive flow of new data from ground- and space-based
observatories now and projected for the coming years, it is clear that
the demand to accurately model 3D dusty structures will rise
strongly. Interfaces that can simulate observations with different
telescope properties will become necessary to perform modeling. We
expect a rise of 3D dust RT modeling efforts that rely on automated
fitting processes rather than by-hand explorations of the model
parameter space. Because the number of sources of multiwavelength data
will rise, collaborations between observers and modelers will become
more frequent. The ultimate goal of 3D dust calculations is to model
multiwavelength images and derive quantitative and statistically sound
information about 3D structures, embedded sources, and the dust itself.
8.5. Future Connections to Nondust Radiative Transfer
Codes
Another future direction is the coupling of 3D dust RT codes with codes
describing other physical effects in astrophysical objects. This trend
is already happening with 2D dust RT codes, and the extension to 3D dust
RT codes is clearly the next step. A variant of this type of connection
is already happening where 3D dust RT is used to calculate the radiation
field in a dust distribution generated with an MHD code; furthermore,
MHD codes that make use of simple dust RT could be tested or the simple
algorithms improved by comparison with full 3D dust RT
solutions. Chemical network calculations could be based on a more
realistic estimate of the incoming radiation calculated from 3D dust
codes. Finally, a combined calculation for 3D line and dust RT would
enable line and continuum data to be simultaneously investigated using
the same underlying physical model.
Conferences and keyword-related publication searches have often been
used in the past to improve the unfortunately rare communication of new
numerical algorithms from applied mathematics to astrophysics. The basic
issue is the sheer flow of new findings and the different languages of
the two communities. Recent MC improvements have been developed mainly
by coders working in the field, and additional efforts should be made to
enable community-crossing exchange on algorithms and error control. As a
result of communications between coders preferring different solvers, we
expect hybrid solvers making use of the advantages of the various
approaches to appear more frequently. Given the increase in complexity
in the modeled objects, we expect future activities to establish grid
generation algorithms that are optimized for 3D dust RT; besides the
octree or AMR-style grids that are now routinely implemented in 3D dust
RT codes, unstructured grids as used in line RT
(Paardekooper, Kruip & Icke 2010),
and MHD codes
(Springel 2010)
are an interesting alternative. The inclusion of time dependence in the 3D
RT problem, which could be important in the context of star formation or
episodic accretion, will also need to be tackled with new algorithms
(see, e.g.,
Harries 2011).
The increasing availability of massively parallel machines will support
algorithms that are optimized to run on many processors.
8.7. Input Physics Improvements
The improvement of the solvers is not restricted to developing
algorithms that provide accelerated solutions. The interaction of
radiation with cosmic dust is still not fully understood, and the
variation of the dust properties with environment is an area of active
research. The various continuum radiation sources such as stars, PDRs,
AGN accretion discs, and the interstellar radiation field are areas of
vigorous investigation. For example, efforts based on existing and
upcoming large-scale surveys are being made to update the 3D structure
of the stars in the Milky Way. Consequently, we expect to achieve a
better understanding of the observed radiation from future research on
the optical properties of dust and improved data on the stellar and
nonstellar sources that enter the 3D dust RT equation.
A major challenge in 3D dust RT that this review highlights is how to
account for and mitigate systematic uncertainties in the dust RT
solution. They arise from under-resolving grids, not propagating
rays/photons to important cells, and/or uncertainties in the underlying
dust grain models. As under-resolving of the dust and radiation field
grid is often a result of constraints on computer memory and speed,
improvements in algorithms to implicitly handle optimal grids are
needed. The preprocessing steps necessary for the RayT solver address
some of these issues, but need further automating. The issue of not
propagating enough rays/photons into particular cells has been solved
for both RayT (placement of rays) and MC (biased emission), but both
currently require hand-tuning. An algorithm to automatically add
additional rays/photons similar to that used for AMR would clearly be
useful. Finally, uncertainties in the assumed dust grain model provide a
systematic uncertainty in the dust RT modeling that is difficult to
quantify. Different dust grain models can be used to provide an estimate
of this uncertainty, but the best way to reduce this uncertainty is to
support the improvement of dust grain models through the use of improved
laboratory and observational data.
DISCLOSURE STATEMENT
The authors are not aware of any affiliations, memberships, funding, or
financial holdings that might be perceived as affecting the objectivity
of this review.
The authors acknowledge the support of the Ghent University for two
excellent week-long meetings in Ghent, Belgium, where a large portion of
the work on this review was done. We thank Simon Bruderer, Jacopo Fritz,
Gianfranco Gentile, Michiel Min, Kirill Tchernyshyov, Ewine van
Dishoeck, and Adolf Witt for providing comments on this review that
significantly improved it.
ACKNOWLEDGEMENTS