arXiv subjects

Benjamin J. Ricketts

Publications and source records attributed to Benjamin J. Ricketts.

3 recordsLinked to original sources

arXiv2026

Emulation of non-linear 1D spectral models: relativistic X-ray reflection

The use of machine learning techniques to approximate computationally expensive models has become increasingly prevalent in a wide variety of fields within astronomy. We discuss the implementation of emulators for 1-dimensional models in the context of the astrophysical numerical model reltrans, a black hole X-ray spectral model that models the effects of relativistically smeared emission from an accretion disk. We argue that the decision of whether and how to emulate should follow from a systematic characterisation of the target model, and we demonstrate a diagnostic workflow: examining how the spectrum varies with individual parameters. We adopt a modular strategy, emulating only the relativistically convolved reflection spectrum (1-10% of the total flux) rather than the full model. Using an operator-learning architecture with Fourier feature embeddings and FiLM conditioning, we reproduce the reflection spectrum to O(0.1)% precision across 0.1-100 keV with a 4-10x speed-up that scales considerably better under vectorised evaluation. This emulator, RTFAST2, recovers the true parameters of simulated observations without the systematic posterior biases of our previous work. We conclude that no architecture is universally transferable and bespoke emulators motivated by a model's specific structure are required. The modular approach taken in this work presents a promising strategy for future emulators of numerical models.

astro-ph.IM↗

arXiv2026

Pushing the Limit of Asteroseismic Detection for Cool Dwarfs using TESS and Deep Learning

Asteroseismology provides a powerful probe of stellar interiors by detecting stellar oscillations, including solar-like oscillations, which are stochastically excited by near-surface convection. While thousands of solar-like oscillators have been identified in evolved stars, only a limited number of main-sequence cool dwarfs have confirmed oscillations due to the low amplitudes of their signals. In this work, we train a convolutional autoencoder on TESS two-minute light curves to automatically identify solar-like oscillation features in cool dwarf main sequence and sub-giant stars. Using catalogs of confirmed oscillators for training and validation, our network achieves a classification accuracy of 99.8% on the test set, along with Precision of 0.945, Recall of 0.998, and F1 Score of 0.971. From the Asteroseismic Target List, our model identifies 3463 potential solar-like oscillators (probability greater than 0.5). After further analysis, we find a list of 24 candidate stars that have the potential to exhibit solar-like oscillations. Notably, several of these candidates occupy regions of the color-magnitude diagram that are accessible only through more resource-intensive radial velocity observations, thereby has the potential of extending the detection frontier of TESS-based asteroseismology. Our candidate catalog provides a valuable foundation for follow-up efforts aimed at expanding the sample of cool-dwarf solar-like oscillators. This will ultimately improve our understanding of stellar structure and evolution across the lower main sequence and strengthen the evidence for using deep learning techniques to study stellar light curves.

astro-ph.SR↗

arXiv2023

Mapping the X-ray variability of GRS1915+105 with machine learning

Black hole X-ray binary systems (BHBs) contain a close companion star accreting onto a stellar-mass black hole. A typical BHB undergoes transient outbursts during which it exhibits a sequence of long-lived spectral states, each of which is relatively stable. GRS 1915+105 is a unique BHB that exhibits an unequaled number and variety of distinct variability patterns in X-rays. Many of these patterns contain unusual behaviour not seen in other sources. These variability patterns have been sorted into different classes based on count rate and color characteristics by Belloni et al (2000). In order to remove human decision-making from the pattern-recognition process, we employ an unsupervised machine learning algorithm called an auto-encoder to learn what classifications are naturally distinct by allowing the algorithm to cluster observations. We focus on observations taken by the Rossi X-ray Timing Explorer's Proportional Counter Array. We find that the auto-encoder closely groups observations together that are classified as similar under the Belloni et al (2000) system, but that there is reasonable grounds for defining each class as made up of components from 3 groups of distinct behaviour.

astro-ph.HE↗