Advanced configuration#
In Radiative transfer settings, we saw how to set some of the basic parameters that determine how Hyperion is run, and in this section we discuss some more advanced options.
Monochromatic radiative transfer#
By default, when producing images, the radiative transfer is done over all wavelengths where radiation is emitted. Every emitted photon is propagated, and may be scattered or absorbed and re-emitted (or both) and we follow the propagation until the photon escapes the grid. When the photon is binned into an image or SED (whether making images with the binning or peeling-off algorithms), the photon is binned into a wavelength grid.
This means that producing images at exact wavelengths can be very inefficient, because it may be that 99% of the photons end up at a different wavelength and do not contribute to the images. In order to get around this problem, Hyperion also implements the concept of monochromatic radiative transfer (see section 2.6.4 of the Hyperion paper). In short, the way this algorithm works is the following:
The temperature is first computed as usual in the initial iterations.
Photons are then emitted from the source at the wavelengths of interest and propagated until they are absorbed, at which point they are terminated. In practice, this process can be inefficient in cases where the albedo is low, because photons may not often scatter. Therefore, we instead force the photon to scatter continuously, but decrease its energy by the albedo at each interaction to make sure energy is conserved. Once the energy of the photon has decreased below a certain threshold, it is terminated.
Photons are then emitted from the dust and propagated until they are terminated, as for the source photons described above.
If raytracing is not used, then the photons are peeled off both when the photons are emitted from the source or dust, as well as when they scatter. Otherwise, they are peeled off only when they scatter, and the source and dust components are computed in the raytracing phase.
Note
The algorithm that forces scatterings and lowers the energy of the photon packets is new in Hyperion v0.9.9. Prior to this, the photons were simply propagated until they were absorbed.
To enable monochromatic radiative transfer, before setting the number of
photons, you should call the set_monochromatic()
method. For example, to compute images at 1, 1.2, and 1.4 microns, you would
need to do:
m.set_monochromatic(True, wavelengths=[1., 1.2, 1.4])
where the wavelength
arguments takes a list of wavelengths in microns. By
default, the energy threshold for termination is set to 1.e-10 of the original
photon energy, but this can be changed using the energy_threshold
argument:
m.set_monochromatic(True, wavelengths=[1., 1.2, 1.4],
energy_threshold=1.e-20)
When using the monochromatic mode, it is then necessary to set the number of photons separately for the photons emitted from sources and the photons emitted from dust:
m.set_n_photons(..., imaging_sources=1000, imaging_dust=1000, ...)
This should be used instead of the imaging
option. The number of photons
is the number per wavelength.
Scattered-light images#
In some cases, one might want to compute scattered light images at wavelengths where there is no dust emission. In this case, there is no need to compute the specific energy of the dust, and there is also no need in re-emitting photons when computing images/SEDs. Therefore, one can set:
m.set_n_initial_iterations(0)
m.set_kill_on_absorb(True)
m.set_raytracing(True)
which turns off the specific energy calculation, kills photons as soon as they
are first absorbed, and enables raytracing for the source emission. For the
photon numbers, one can set raytracing_dust=0
to zero, since this is not
needed (there is no dust emission).
Note
This cannot be used for all scattered light images. For example, in a protostar, a K-band image may have a non-negligeable amount of scattered light flux originating from the inner rim of the disk. This technique can only be used when there is no dust emission.
This can be combined with the Monochromatic radiative transfer option described above to avoid wasting photons at wavelengths where they are not needed. When treating only scattering, you will then want to set the following options:
m.set_n_photons(imaging_sources=1000, imaging_dust=0,
raytracing_sources=1000, raytracing_dust=0)
where the values should be adjusted to your model, but the important point is
that initial
is not needed, and imaging_dust
and raytracing_dust
can be set to 0.
For a full example of a model computing scattered light images, see How to efficiently compute pure scattering models.
Miscellaneous Settings#
Set the maximum number of photon interactions:
m.set_max_interactions(100000)
Set the number of output bytes per floating point value for the physical arrays (4 = 32-bit, 8 = 64-bit):
m.set_output_bytes(4)
To set the minimum temperature for dust:
m.set_minimum_temperature(10.)
m.set_minimum_temperature([10., 5., 20.])
If a scalar value is specified, the same value is used for all dust types. If a list is specified, the list should have as many items as dust types, and each item corresponds to the minimum temperature for each dust type.
Similarly, to set the minimum specific energy:
m.set_minimum_specific_energy(1.e-4)
m.set_minimum_specific_energy([1.e-4, 1.e-5, 2.e-5])
By default, photon positions and cells are double-checked every 1 in 1000 cell
crossings. This can be changed
with set_propagation_check_frequency()
:
m.set_propagation_check_frequency(0.01)
Note that values higher than 0.001 (the default) will cause the code to slow down.