Abstract
We aim to construct an exceptionally deep ($V \lesssim 27$) catalog of variable objects in selected Galactic and extragalactic fields visited multiple times by the Hubble Space Telescope (HST). While HST observations of some of these fields were searched for specific types of variables before, we attempt a systematic study of the population of variable objects of all types at a magnitude range not easily accessible with ground-based telescopes. The variability timescales that can be probed range from hours to years depending on how often a particular field has been visited. For source extraction and cross-matching, we rely on the Hubble Source Catalog (HSC), which includes $10^7$ objects detected with WFPC2, ACS, and WFC3 instruments. Lightcurves are corrected for systematic effects and screened for bad measurements. For each lightcurve, we compute variability indices sensitive to a broad range of variability types, characterizing overall scatter and smoothness. Candidate variables are selected based on index values significantly higher than expected for objects of similar brightness. The Hubble Catalog of Variables (HCV) provides a deep, multi-instrument look at the changing sky.
Introduction: Expanding the Time-Domain Frontier
Most variability surveys, such as the Zwicky Transient Facility or the upcoming Vera C. Rubin Observatory, are ground-based. While they offer wide-sky coverage, they are limited by atmospheric blurring, making it difficult to resolve individual variable stars in "crowded" fields like globular cluster cores or neighboring galaxies. Furthermore, previous Hubble-based variability searches were often "targeted," focusing on specific objects like Cepheid variables for distance measurements.
The Hubble Catalog of Variables (HCV) represents the first systematic effort to mine the Hubble Source Catalog (HSC) for variable objects of all types, reaching unprecedented depths of $V \approx 27$. By leveraging decades of archival data, the HCV turns the telescope’s high-resolution imagery into a precision tool for time-domain astronomy, identifying everything from flickering stars in our galaxy to erupting sources in distant universes.
Methodology and Variability Indices
The HCV pipeline processes millions of lightcurves using a suite of robust statistical indices. Because archival data often contains inhomogeneous sampling and poorly defined measurement errors, the team prioritized "non-parametric" indices that do not rely on reported error bars, making detection more reliable in the face of "messy" archival data.
Two primary metrics used are:
- Interquartile Range (IQR): A robust measure of the amplitude of brightness changes, resistant to outliers or single bad data points.
- Inverse von Neumann ratio ($1/\eta$): A measure of "smoothness" that distinguishes gradual, real variability from random noise.
As shown in Figure 2, the efficiency of these indices (measured by the $F_1$-score) improves significantly as the number of observations ($N$) increases, with IQR and $\sigma$ performing particularly well even with relatively few data points. To handle the massive computational load, the pipeline was implemented in Java and parallelized using Apache Spark.
Data Processing and Validation
The project utilizes the Hubble Source Catalog version 2, encompassing approximately 10 million objects detected by the WFPC2, ACS, and WFC3 instruments. Key technical steps include:
- Data Cleaning: Excluding measurements near frame edges or those from outdated pipelines.
- Local Zero-Point Corrections: Calculating "local" corrections for each star based on the average behavior of its neighbors to compensate for residual sensitivity variations across the detector.
The authors validated these methods using specific test fields, including the M31 (Andromeda) halo and the M4 globular cluster.
Key Results and Insights
The HCV highlights the power of HST’s sharp Point Spread Function (PSF). This resolution allows for the detection of variables in regions where ground-based data would show only a single, blurred source. In the HSC v2, the team identified 89 fields with at least 5 visits (1.7 million sources) and 7 fields with 25 or more visits (290,000 sources).
Figure 6 illustrates the spatial distribution of sources in a typical field and the resulting lightcurves, demonstrating the pipeline's ability to track brightness changes over long baselines. The catalog reaches a magnitude of $V \approx 27$, far deeper than any wide-field ground-based variability survey.
Limitations and Future Outlook
The primary challenge for the HCV is the irregular "cadence" of Hubble observations. Because HST is a pointed observatory, temporal sampling is often sparse, making it difficult to classify the specific type of variable (e.g., pulsating star vs. binary system) in fields with very few visits.
Future work involves integrating Machine Learning and Principal Component Analysis (PCA) to better combine information from the 20+ variability indices calculated. The first version of the Hubble Catalog of Variables has since been released and is available via the Mikulski Archive for Space Telescopes (MAST), providing a vital resource for studying the deep, variable universe.
Summary updated