Modeling Blazar Broadband Emission with Machine Learning: Toward a Physical Interpretation of the Blazar Sequence

Blazars, a class of jetted active galactic nuclei, are the most numerous permanent extragalactic gamma-ray sources. Their peculiar double-bumped spectral energy distributions (SEDs) are usually interpreted as non-thermal emission from a relativistic jet of particles closely aligned with the line of sight. Population studies have highlighted a "blazar sequence", i. e. an anticorrelation between the frequency of the low-energy peak and its bolometric luminosity. Its existence and origin are still unclear, despite the influx of new data, including in the TeV band. This work thus aims at contributing to a new sequence that finally includes very high-energy gamma-ray spectra. A number of representative SEDs from a sample of TeV-detected blazars of the "BL Lac" type, binned according to their low-energy peak frequencies, were modeled based on the standard "Synchrotron Self-Compton" scenario: best-fit parameters were compared to search for trends hinting at the mechanisms underlying the sequence. Different techniques, including analytical tools and machine learning, were used to characterize spectral quantities of the selected sources, and their outcomes and performances were discussed.