EC comparisons with JetSet: DT

[cosimoNigro]

Hello,

let us discuss in this thread the comparisons for EC with jetset that were suggested by Justin. I started to work on them in the jetset_ec_test branch. Here let us comment on the DT.
Here the link to the jetset documentation on External Compton. I think there is a 1:1 correspondence with the parameters we use, see the test_jetset_ec_dt.py script in the branch.

I used the same distances of the crosscheck:

• close to the DT we have a similar shape and a factor 4 between jetset and agnpy note in this case the transformation feature used for the BLR (transforming the electron distribution to the disk frame) makes the computation of the SED fail.

• outside the photon field we have the usual order-of-magnitudes disagreement

Let me know what do you think.

[cosimoNigro]

Performing some of the tests @jsitarek requested in another thread:

before giving up with jetset, could you make two simple test:

1. compare the two with the case of non-relativistic blob (e.g. Gamma = 1.001)

Here goes the non-relativistic case:

• inside the DT

• outside the DT

2. compare the two forcing the observation angle to be on axis (theta << 1/ Gamma).

Here goes the on-axis case:

• inside the DT

• outside the DT

Scripts to perform these tests with jetset are now available in the master, see the jetset_tests repository.

[jsitarek]

follow up, concerning the e-mail discussion. With the fix in jetset the agreement look good at small distances, but gets progressively worse at high distances. I think this is consistent with the fact that different geometry is used for DT (sphere in jetset, ring in agnpy). I think it is worth to mention this different geometry, and no need to include a comparison for DT case in the paper.

[cosimoNigro]

After some work by @andreatramacere in jetset the comparison improved a lot:

the only controversial feature I see is the spectrum computed with @agnpy at large distances cutting off at lower frequencies compared to jetset. @andreatramacere any suggestion?

[andreatramacere]

@cosimoNigro at larger distances the cut-off I think that is better estimated by agnpy, even though we are at a flux level that is barely detectable, at least for distance sources. On the contrary, it would be interesting to check, to which extent, for close radio galaxies, this might be tested. This could be an interesting discussion

[jsitarek]

@andreatramacere, I'm not sure how exactly you do the integration in jetset, but it might be that simply increasing considerably the angular binning in the integrals improves the accuracy of this integral.

what is R_DT in this plot? if it is expected that for large R/R_DT jetset has difficulties, and as Andrea mentioned this is not really a physical case since the emission is really tiny, maybe a more reasonable case of R/R_DT equal to a few or 10 at most can be tested instead ?

[andreatramacere]

@jsitarek within the BLR, jetset and agnpy should agree, as long as the seed photon field is the same. Outside, no, because I am not doing all the extra angular integration that are done in agnpy, to avoid too long computational time. In any case, I found a way to scale the flux level in such a way that it is not too different from the exact angular integration. My feeling is that the comparison between jetset and agnpy is meaningful, in terms of validation, within the BLR/DT, and if the fields are the very same. For the BLR, seed fields are not identical between jetset and agnpy. So this kind of clarification and discussion should be addressed in the text of the paper. We can test R/R_DT at lower as you suggest. But, I repeat, outside the BLR/DT we can't call it actual validation, I would rather say we are testing the difference between IC kernel with angular dependence, and IC kernel based on assuming isotropic seed photons. As I discussed yesterday with @cosimoNigro, an interesting possibility would be to build a matrix embedding the correction to apply to the isotropic case, via interpolation. In this case, we would get a fast and accurate EC computation