Line Radiated Power as a LineShapeModel?

[Mateasek]

Line Radiated Power feature could work in a simmilar way as the TotalRadiatedPower emission model. The advantage would be efficient and precide calculations of individual line intensities.

Some systems need to investigate line ratios, which are experimentally obtained either through system which integrate over a spectral band, as for example photomultipliers with bandpass filters, or through fitting of a spectral line shape.

In case of very narrow lines, the spectral resolution used in Cherab simulations has sometimes have to be very large, to limit numerical errors. The reason is real detectors integrate the received photons, but Cherab with a too large wavelength step can miss a line completely or the following integration can be numerically wrong due to bad sampling. Performing simulations can therefore be sometimes cumbersome, because there can be a number of lines to simulate in a wider wavelength range. The propose feature would remove such problems, but of course would have to be documented properly, to warn users, that the returned spectrum has a no real physical meaning.

From what I learned, the model could be efficiently added as a LineShapeModel, but I'm not completely sure if it should be grouped with other lines, to limit some confusion. It could be added alongside the TotalRadiatedPower into tools.

What do you think @vsnever and @jacklovell ?

[vsnever]

In case of very narrow lines, the spectral resolution used in Cherab simulations has sometimes have to be very large, to limit numerical errors. The reason is real detectors integrate the received photons, but Cherab with a too large wavelength step can miss a line completely or the following integration can be numerically wrong due to bad sampling.

Hi @Mateasek, the add_gaussian_line() function uses Erf to integrate the Gaussian line shape over the spectral bin, so it is impossible to miss a line or calculate the wrong intensity regardless of the spectral resolution. Or did I not understand what you mean?

[Mateasek]

No, you understood correctly, I see now. So the error would not happen, thanks for pointing it out.

Anyway it could be a useful tool for calculations targeted only on line intensities, because you could skip the spectral part.

[vsnever]

The only two PlasmaModels supporting a custom LineShapeModel are ExcitationLine and RecombinationLine. Therefore, you can simply add LineRadiatedPower(PlasmaModel), which doesn't require a line shape. It will use atomic_data.impact_excitation_pec() and atomic_data.recombination_pec() just like TotalRadiatedPower uses line_radiated_power_rate() and continuum_radiated_power_rate().

Alternatively, you can add a DeltaFunctionLineShape which will act like a Dirac delta function.

If you choose the second option, then I think it is better to keep this model along with other line shape models. Perhaps lineshape.pyx needs to be split into multiple files incherab/core/model/lineshape, but that is a separate issue.

[vsnever]

I just remembered that I proposed SOLPSLineEmitter to Cherab-SOLPS, which takes the line shape as a parameter, so the DeltaFunctionLineShape solution will be more universal than the LineRadiatedPower.

[Mateasek]

Yes, this is another option, but then this is what the add_gaussian_line will converge to, with decreasing spectral resolution, so actually we already have the requested functionality. I think I can close this issue. Thanks for helping me to realise this @vsnever.

[vsnever]

Probably, add_gaussian_line() will be slower than add_deltafunction_line(), but if performance is a priority, there are other options like pre-baking the spectral emissivity on a regular grid with RegularGridEmitter, raysect/source#343 and continuous argument set to False.

[Mateasek]

Yes, if there is need for speed, then you should turn to other means as RTMs and the emitters you linked. Thanks again