Description
Tidal Disruption Events (TDEs) occur when a star is torn apart by the tidal forces of a supermassive black hole, producing a luminous flare that evolves over weeks to months. These transients are valuable for studying astrophysical jet launching, accretion disk formation, and supermassive black hole demographics. The classification of these sources typically requires spectroscopic follow-up, where spectra are used to distinguish TDEs from the much more numerous populations of supernovae and active galactic nucleus (AGN) flares. However, the resources for spectroscopic follow-up are severely limited, making photometric classification methods increasingly important.
To address this, we analyze nuclear transients from the Zwicky Transient Facility (ZTF), focusing on those exhibiting late-time photometric plateaus. We employ a multimodal machine learning approach that integrates both direct light curve analysis using a neural network and extracted photometric features. A key aspect of our approach is the incorporation of temperature evolution, which is particularly useful in distinguishing TDEs from supernovae, where temperature changes are more pronounced. In this poster, I will show how the addition of this neural network to our human-selected features improves classification performance. Tested on ZTF sources, we anticipate that this approach will be applicable to future large-scale surveys, including the upcoming Vera C. Rubin Observatory, where efficient classification frameworks will be crucial for identifying and studying rare nuclear transients.
| Talk category | NOVA Network 3 |
|---|---|
| Preference for a talk or poster | Poster |
| Talk preference for PhD students | 1st year |
