Couple questions on custom energy injection and more

[ferromatteo]

Hello, I'm currently working on the modeling of a very rich Long GRB lightcurve and I came across this wonderful software :D

Since I'm not really an expert regarding AG modeling nor related codes I would like to understand if this can be feasible or not:
Basically I would like to:

1. Fit multiple episodes of energy injection (not just single magnetar spin-down, but maybe N gaussian-like generic episodes)
   2 (and 3). Fit host galaxy extinction as a free parameter (rather than fixing it), or at least understand if there is an easy way to sum this term in the likelihood (as well as possible N fixed fluxes related to a possible underlying host, since my lightcurve reaches late times where I see a chromatic softening of the powerlaws)

While I discovered the Ejecta class allows custom E_dot profiles, integrating this with the Fitter class for MCMC parameter estimation seems non-trivial to me, while for the magnetar I just add it to the Setup and define the three parameter priors.

For the other two: let's say a possible AV/E(B-V) parameter of the host given an extinction law and the N-magnitudes of the possible host I would like just to know if I can easily access the log-likelihood before running the fitter, after the rest of the configuration is set, in order to just sum the two related terms.

As a side note, I found that when I run with more complex configuration selected (like specifically enabling klein-nishina corrections, ssc cooling) the kernel of the notebook where I'm currently running my fit dies very quickly, dunno If this could be related to my dataset or I'm missing something else.

Thank you very much in advance and for the great effort of this amazing package (and sorry if my requests are silly or unfeasible, I just started playing with the code, which btw is extremely user friendly =D).

Matteo

[YihanWangAstro]

Hi Matteo,

No worries at all, we’re always here to help. Any feedback is extremely welcome!

As you may have noticed, VegasAfterglow has two components:
(1) a black-box MCMC module designed for beginners, and
(2) a more flexible light-curve calculation module where you can customize the jet, medium, and other physics.

Yes, using the Ejecta class is the correct way to implement multiple energy injections. For this reason, instead of relying on the black-box MCMC module, it’s better to write your own likelihood function (simply using VegasAfterglow as the backend light curve generator and computing the χ² yourself) and then call whichever MCMC sampler you prefer. This approach gives you full control and lets you take full advantage of the flexibility built into VegasAfterglow. You can incorporate extinction in this framework as well. (In fact, there are successful examples of including extinction; see Figs. 8 and 9 of this paper: https://arxiv.org/pdf/2509.22784.)

If you run into any difficulties while setting up your MCMC, feel free to reach out.

Regarding the crash, could you please provide more details so we can debug it?
1. VegasAfterglow version
2. Platform / OS
3. A minimal code snippet that reproduces the crash

The older version is quite greedy in memory usage when SSC is enabled, so the crash could simply be due to memory exhaustion. In the latest version, we optimized the memory footprint (at the cost of a bit of performance), which should prevent this issue. However, if you’re already using the newest version, there might be another underlying problem.

If there is anything we can improve or if you encounter any inconvenience when using the code, please let us know. And it’s great to hear that VegasAfterglow has been helpful to you!

Best,
Yihan

[ferromatteo]

Hello Yihan,

thank you very much for reaching out so quickly, I will try to do as suggested and in case I will reach out :D

Regarding the crash, you are right, my current version (installed via pip) was 1.0.2, I've updated now to 1.0.3, but I see that there should be already the 1.0.4 (although not available via pip, correct?). Anyway I guess that the problem was probably caused by that (I have a Mac M2 with OS Tahoe 26.0.1).

I will set up an external MCMC with my own likelihood and I will let you know if I get the same errors with this version :)

kudos! (And thanks for the patience)

Matteo