Search for glitches of gamma-ray pulsars with deep learning

TL;DR

This paper introduces a novel deep learning-based method to detect pulsar glitches in gamma-ray data, leveraging the Fermi-LAT telescope observations and convolutional neural networks to improve detection accuracy.

Contribution

The paper presents a new approach combining weighted H-test statistics and CNNs for glitch detection in sparse gamma-ray pulsar data, a novel application in this field.

Findings

High accuracy demonstrated on Monte Carlo simulations

Method ready for application to real gamma-ray data

Potential to discover new pulsar glitches in gamma-ray observations

Abstract

The pulsar glitches are generally assumed to be an apparent manifestation of the superfluid interior of the neutron stars. Most of them were discovered and extensively studied by continuous monitoring in the radio wavelengths. The Fermi-LAT space telescope has made a revolution uncovering a large population of gamma-ray pulsars. In this paper we suggest to employ these observations for the searches of new glitches. We develop the method capable of detecting step-like frequency change associated with glitches in a sparse gamma-ray data. It is based on the calculations of the weighted H-test statistics and glitch identification by a convolutional neural network. The method demonstrates high accuracy on the Monte Carlo set and will be applied for searches of the pulsar glitches in the real gamma-ray data in the future works.