The ZTF Source Classification Project: II. Periodicity and variability processing metrics

TL;DR

This paper introduces the ztfperiodic codebase and computational metrics for classifying variable sources in ZTF data, emphasizing GPU-optimized algorithms and scalability for large-scale surveys.

Contribution

It presents a new GPU-optimized period-finding algorithm suite and metrics for large-scale variable source cataloging in ZTF data releases.

Findings

GPU acceleration improves processing speed

Metrics enable scalable catalog generation

Framework applicable to future surveys like LSST

Abstract

The current generation of all-sky surveys is rapidly expanding our ability to study variable and transient sources. These surveys, with a variety of sensitivities, cadences, and fields of view, probe many ranges of timescale and magnitude. Data from the Zwicky Transient Facility (ZTF) yields an opportunity to find variables on timescales from minutes to months. In this paper, we present the codebase, ztfperiodic, and the computational metrics employed for the catalogue based on ZTF's Second Data Release. We describe the publicly available, graphical-process-unit optimized period-finding algorithms employed, and highlight the benefit of existing and future graphical-process-unit clusters. We show how generating metrics as input to catalogues of this scale is possible for future ZTF data releases. Further work will be needed for future data from the Vera C. Rubin Observatory's Legacy Survey of Space and Time.