Generator for muons underground

[MoritzNeuberger]

Our usual muon generator (MUSUN) is in some way difficult to work with:

• It is not publicly available. You can get the source code by writing the author.
• The version we currently have access to is only for LNGS, so it requires additional steps for users to adjust to their overburden.
• It does not allow for adjustments for systematic studies, since it would need a rerun of MUSIC (which I also never used before)
• Finally, it's written in Fortran.

I recently stumbled upon mute [link], which is an underground muon flux calculator. Given the primary muon flux, a survival probability tensor as a function of (E_init, E_final, slant depth) given a rock density, and a given overburden (flat or mountain), it can calculate the muon flux as a function of (E, room angle). See details in [pub].

Mute does not have a generator (not yet, at least [link]), but I think it would be great if we could transition to something based on it. This is because it already has the overburden and rock densities for various underground sites implemented, making it relatively easy to adjust for them. In addition, one can easily add custom overburdens for other sites. As far as I understand, shallow overburdens are not yet well reconstructed, but might be in progress in the future [link].

This would ofc require some validation and additional RND work, but including this model could also be very useful for selling remage and for other experiments.

EDIT: I should highlight, for simflow, it is necessary to generate muon input files on the fly and in the long run, I would rather want to have it based on mute than musun because of the systematic studies oppurtunities.

[MoritzNeuberger]

While searching, I also found that the owner of that PR implemented the prepared files to generate muons in their own G4 simulation [link], which they used in this [pub] (it's quite interesting what they did). For our purpose, as well as personally, for the wider low-background community, it would be better to have a more convenient and generalized way to use mute to generate muons. I will contact the author and ask for their plans regarding this.

In the meantime, I played around a bit with the PR that starts implementing some of these features. I implemented a simple sampling module and have already achieved promising results compared to musun. In the most recent mute publication [pub], they also compared their predictions with LVD, on which our musun version is based. Their results seem to reproduce the data quite well, though I've never checked how well our version actually does (would be silly if it didnt).

Here are some early results:

The larger bins in mute at low zenith are due to the slant-depth map.

I'm curious about the differences at low zenith angles. Is it due to differences in overburden modeling? Or is something wrong with my sampling?

Here, interestingly, the sampled mute distribution has a higher mean (293 GeV) than the one you get using their predicted distributions (288 GeV). It might be an artifact of my sampling, or I might not be using mute correctly.

[MoritzNeuberger]

To check whether the difference originates in a difference in overburden modeling, I converted the depth distribution input for musun to the mute format and reran the simulation with that. A difference still is that musun assumes a density of 2.71 g/cm3, and mute 2.72 g/cm3. I am also planning to run mute at a lower density, though I have to generate the survival probability tensor, which takes a while, so for now it's still at 2.72 g/cm3. Below are the same plots as before, with different overburden. It seems to have only a minor effect on the distribution, and the mean energy remains elevated at 293 GeV.

So the two options where the difference might come from are:

• difference in density
• difference in initial muon flux model (<-- my bets are on this one)