# Robust detection of quasi-periodic variability: A HAWKI mini survey of late T dwarfs

## TL;DR

This study uses HAWK-I J-band photometry and Bayesian analysis to investigate quasi-periodic variability in late T dwarfs, highlighting the importance of robust methods to avoid false detections due to red noise.

## Contribution

It introduces a Bayesian approach combined with false alarm probability to reliably detect quasi-periodic variability in brown dwarf lightcurves affected by red noise.

## Key findings

- ULAS J2321 shows tentative 1.64-hour variability but is not statistically significant.
- Traditional false alarm probability methods may overestimate variable objects.
- A hybrid Bayesian approach improves robustness in variability detection.

## Abstract

We present HAWK-I J-band light curves of five late-type T dwarfs (T6.5-T7.5) with a typical duration of four hours, and investigate the evidence for quasi-periodic photometric variability on intra-night timescales. Our photometry reaches precisions in the range 7-20 mmag, after removing instrumental systematics that correlate with sky background, seeing and airmass. Based upon a Lomb-Scargle periodogram analysis, the latest object in the sample - ULAS J2321 (T7.5) - appears to show quasi-periodic variability with a period of 1.64 hours and an amplitude of 3 mmag. Given the low amplitude of variability and presence of systematics in our lightcurves, we discuss a Bayesian approach to robustly determine if quasi-periodic variability is present in a lightcurve affected by red noise. Using this approach, we conclude that the evidence for quasi-periodic variability in ULAS J2321 is not significant. As a result, we suggest that studies which identify quasi-periodic variables using the false alarm probability from a Lomb-Scargle periodogram are likely to over-estimate the number of variable objects, even if field stars are used to set a higher false alarm probability threshold. Instead we argue that a hybrid approach combining a false alarm probability cut, followed by Bayesian model selection, is necessary for robust identification of quasi-periodic variability in lightcurves with red noise.