Ripples in the OCEANS: Broad Line Variability of Little Red Dots

Little Red Dots (LRDs) are a unique class of compact, red sources discovered in the JWST extragalactic deep fields. Determining if they are indeed powered by accreting supermassive black holes (SMBHs) is one of the main drivers of the intense study of these objects. Evidence for variability in these objects provides a direct test for the active galactic nucleus (AGN) nature of their central engine. In this study, we present a variability analysis of 6 LRDs observed by the R ∼ 2700 OCEANS survey and leverage archival R ∼ 1000 spectroscopic data from the CEERS and RUBIES surveys. We report marginal detections of Hα broad-line (BL) variability in the LRDs OCEANS-100424/RUBIES-42232 (27% variability at 2.1σ significance) and OCEANS-35829/RUBIES-49140 (GlimmIr/Irony; 50% variability at 1.5σ significance). The other 4 LRDs in our sample do not show evidence for BL variability, with a 1σ upper limit of 4.8% − 30% variability between their epochs of observations. We also find no evidence (< 1σ) for continuum variability in our LRD sample. We compare our results to a sample of SDSS-RM quasars to determine the probability of our broad Hα variability detections. We find that the probability of reproducing 2 variable and 4 nonvariable quasars is 4.71%, corresponding to ∼ 2σ departure from typical quasar variability. The detection of BL Hα variability in 2 LRDs provides some evidence for the AGN nature of these objects as opposed to pure scattering models.

Improvements to AI Systems:

1. Astronomical Time-Series Variability Detector

• Improve AI models to detect and quantify broad-line (BL) variability in low-signal, noisy spectroscopic data (e.g., Hα line flux changes) using Bayesian or likelihood-ratio methods, with explicit uncertainty estimates (σ levels) and upper limits for non-detections.

• The improved system can automatically flag candidate AGNs in JWST-era surveys by comparing multi-epoch spectra, even with sparse or irregular sampling.

2. Cross-Survey Spectral Matching and Epoch Alignment

• Enhance AI to align and co-analyze spectra from different instruments/resolutions (R∼2700 vs R∼1000) by learning spectral continuum and line-profile transformations, reducing systematic errors when comparing variability across surveys.

• The improved system can fuse heterogeneous archival data (e.g., CEERS, RUBIES, OCEANS) to produce consistent variability metrics for faint, compact sources.

3. Quasar Variability Likelihood Model

• Build a generative AI model trained on SDSS-RM quasar variability statistics to predict the probability of observing a given number of variable vs. non-variable sources in a small sample, accounting for redshift, luminosity, and time-lag effects.

• The improved system can quantify the statistical significance (e.g., 2σ departure) of AGN candidacy in small-N extragalactic samples, enabling robust classification of ambiguous objects like Little Red Dots.

4. Continuum vs. Line Variability Disentangler

• Develop an AI that separates continuum variability from broad-line variability in unresolved sources, using spectral decomposition and time-domain modeling.

• The improved system can test alternative explanations (e.g., scattering models vs. accretion disk variability) by predicting expected continuum-to-line variability ratios and comparing them to observations.

5. Automated Upper-Limit Estimator for Non-Detections

• Improve AI to compute rigorous 1σ upper limits on variability (e.g., 4.8%–30%) from noisy multi-epoch spectra, using Monte Carlo simulations or bootstrapping that incorporate flux calibration uncertainties.

• The improved system can provide reliable non-detection constraints for large surveys, helping to rule out AGN activity in compact red sources without requiring manual analysis.

6. Redshift-Epoch Variability Predictor

• Train an AI to predict expected Hα variability amplitude as a function of rest-frame time separation, black hole mass, and Eddington ratio, using quasar ensemble data.

• The improved system can prioritize targets for follow-up variability observations and design optimal observing cadences for JWST or future missions to confirm AGN nature in faint sources.

Little Red Dots (LRDs) are a unique class of compact, red sources discovered in the JWST extragalactic deep fields. Determining if they are indeed powered by accreting supermassive black holes (SMBHs) is one of the main drivers of the intense study of these objects. Evidence for variability in these objects provides a direct test for the active galactic nucleus (AGN) nature of their central engine. In this study, we present a variability analysis of 6 LRDs observed by the R about 2700 OCEANS survey and leverage archival R about 1000 spectroscopic data from the CEERS and RUBIES surveys. We report marginal detections of H alpha broad-line (BL) variability in the LRDs OCEANS-100424/RUBIES-42232 (27% variability at 2.1 sigma significance) and OCEANS-35829/RUBIES-49140 (GlimmIr/Irony; 50% variability at 1.5 sigma significance). The other 4 LRDs in our sample do not show evidence for BL variability, with a 1 sigma upper limit of 4.8 % - 30% variability between their epochs of observations. We also find no evidence (<1 sigma) for continuum variability in our LRD sample. We compare our results to a sample of SDSS-RM quasars to determine the probability of our broad H alpha variability detections. We find that the probability of reproducing 2 variable and 4 nonvariable quasars is 4.71%, corresponding to about 2 sigma departure from typical quasar variability. The detection of BL H alpha variability in 2 LRDs provides some evidence for the AGN nature of these objects as opposed to pure scattering models.