SciPost Submission Page
Dark matter or millisecond pulsars? A deep learning-based analysis of the Fermi Galactic Centre Excess
by Florian List, Nicholas L. Rodd, Geraint F. Lewis
This Submission thread is now published as
Submission summary
| Authors (as registered SciPost users): | Florian List |
| Submission information | |
|---|---|
| Preprint Link: | scipost_202209_00003v1 (pdf) |
| Code repository: | https://github.com/FloList/GCE_NN |
| Date accepted: | 2022-11-28 |
| Date submitted: | 2022-09-01 22:16 |
| Submitted by: | List, Florian |
| Submitted to: | SciPost Physics Proceedings |
| Proceedings issue: | 14th International Conference on Identification of Dark Matter (IDM2022) |
| Ontological classification | |
|---|---|
| Academic field: | Physics |
| Specialties: |
|
| Approaches: | Computational, Phenomenological |
Abstract
The $\gamma$-ray Galactic Centre Excess (GCE) remains one of the few observed high-energy signals for which a dark matter (DM) origin is a plausible explanation. We present a deep learning-based analysis of the $\gamma$-ray sky in the Galactic Centre region, carefully accounting for the mathematical degeneracy between faint point-sources (PSs) such as millisecond pulsars (MSPs) and DM-like Poisson emission. Using recent models for the Galactic foregrounds, we find that relatively few bright PSs just below \textit{Fermi}'s detection threshold seem unlikely to explain the GCE, although we continue to find evidence for PSs. Looking ahead, further improvements in the modelling of the $\gamma$-ray sky will be crucial for distinguishing between a DM-like and point-like morphology of the signal.
Published as SciPost Phys. Proc. 12, 034 (2023)
Reports on this Submission
Anonymous Report 1 on 2022-10-21 (Invited Report)
- Cite as: Anonymous, Report on arXiv:scipost_202209_00003v1, delivered 2022-10-21, doi: 10.21468/SciPost.Report.5947
Strengths
1. This study tries to study the puzzling origin of Galactic center gamma excess with convolutional neural networks. This is a novel method and definitely will lead to some interesting understandings of the problem.
2. The manuscript provides a nice summary of their method, as well as potential extensions to improve it in the future.
Weaknesses
While the work is mostly focusing on the method, it seems that the modeling uncertainties are the crucial factor that doesn't allow for a conclusive answer at this moment.
Report
This manuscript clearly demonstrates the methodology that uses convolutional neural networks to analyze the Fermi-Lat data. While still haunted by modeling uncertainties, the method seems promising in the near future. It thus meets the criteria of SciPost Physics Proceedings, and should be published here.
Requested changes
None.