|
Title |
Diagnostics for compact objects type for Extra-galactic X-ray Binaries |
Alternative Title |
Διαγνωστικά για το είδος των συμπαγών αντικειμένων σε εξωγαλαξιακά διπλά συστήματα ακτίνων Χ |
Author
|
Βάσιλας, Νικόλας
|
Thesis advisor
|
Ζέζας, Ανδρέας
|
Abstract |
X-ray binaries are systems consisting of a compact object accreting material from a donor star. They are objects
of great importance for the study of the late stages of stellar evolution and the mechanism of supernova explosions.
While they are crucial for understanding compact objects it is usualy very difficult to characterize their nature.
X-ray spectra information can be used in order to infer the nature of the compact objects. In this work, we develop
a diagnostic method for the characterization of compact objects based on the X-ray colors or hardness ratio of X-ray
binaries. For that reason, we used a large sample of X-ray spectra of Galactic X-ray binaries for which we know
the nature of the primary component (black hole 15716, Atoll neutron star 2688, and pulsar 657) in order to train
machine learning models capable of predicting accurately the label of the objects.
The training of such a model requires good-quality spectra. Our approach was to simulate based on those data
spectra observed with different detectors. From those simulated spectra we calculated the non-parametric metric of
colors and total intensity normalized at 10 kpc which we used for the training of our models. We considered two
X-ray observatories, NuSTAR and the proposed HET and LET detectors on the HEXP mission.
By training the diagnostic models with different band and feature combinations we were able to investigate the
optimal combinations of those for each instrument. We found that models trained with bands reaching 12 keV
and features containing only one color and the total intensity behave well but are not as accurate as combinations
including higher energy bands and more colors. Finally, by comparing the diagnostics for NuSTAR and HET we
notice that the HET instrument needs spectra with one order of magnitude lower number exposure time (which
translates to one order of magnitude less counts) in order to make an accurate classification of the type of each
object.
|
Language |
English |
Issue date |
2024-07-11 |
Collection
|
School/Department--School of Sciences and Engineering--Department of Physics--Graduate theses
|
|
Type of Work--Graduate theses
|
Permanent Link |
https://elocus.lib.uoc.gr//dlib/7/0/4/metadata-dlib-1719306273-34690-28445.tkl
|
Views |
524 |