Unlocking AGN Variability with Custom ZTF Photometry for High-Fidelity Light Curves and Robust Selection

(Abridged)We explore the potential of optical variability selection methods to identify AGN, including those challenging to detect with conventional techniques. Using the unprecedented combination of depth, sky coverage, and cadence of the ZTF survey, we target even starlight-dominated AGN, known for their redder colours, weaker variability signals, and difficult nuclear photometry due to their resolved hosts. We perform aperture photometry on ZTF reference-subtracted images for 40 million sources across 8,000 deg^2, assemble light curves and classify objects employing an RF algorithm into 14 classes, including 341,938 candidate AGN. We compare variability metrics derived from our photometry to those obtained from ZTF Data Release light curves (DR11-psf), to assess the impact of our analysis. We find that the fraction of low-z quiescent galaxies exhibiting significant variability drops dramatically (from 98% of the sample to 7%) when replacing the DR11-psf light curves with our difference image, aperture photometry (DI-Ap) version. The overall number of variable low-z AGN remains high (99% when using DR11-psf lightcurves, 83% when using DI-Ap), however, implying that our photometry can detect the fainter variability in host dominated AGN. The classifier effectively distinguishes between AGN and other sources, demonstrating high recovery rates even for AGN in resolved nearby galaxies. AGN candidates in eROSITA’s eFEDS field, detected in X-rays and bright enough for ZTF optical observations, were classified as AGN (79%) and non-variable galaxies (20%). These groups show a 2 dex difference in X-ray luminosity but not in X-ray flux. A significant fraction of X-ray AGN are optically too faint for ZTF, and conversely, a quarter of ZTF AGN in the eFEDS area lack X-ray detections, highlighting a wide range of X-ray-to-optical flux ratios in AGN.

💡 Research Summary

**
This paper presents a novel approach to extracting high‑fidelity optical light curves from the Zwicky Transient Facility (ZTF) by performing forced aperture photometry on reference‑subtracted (difference) images. The authors develop a pipeline that (i) selects high‑quality ZTF observations (infobits = 0, maglimit > 20, seeing < 4″), (ii) downloads the corresponding difference images and the catalogues of sources detected on the science frames, (iii) measures fluxes for every source using a fixed 4″ circular aperture on the difference images with SExtractor in dual‑image mode, and (iv) reconstructs a “total” flux for each epoch by adding the measured difference‑image flux to the flux of the same source in the reference image, after converting the reference magnitude to the same zero‑point as the difference image.

A careful per‑epoch, per‑quadrant calibration is then applied. Point‑like sources with 12 < g < 19.5, classtar > 0.7, and magnitude errors < 0.3 are used as calibration anchors (typically 15–several hundred per epoch). The magnitude offset m_tot – m_ref is fitted with a first‑order polynomial; outliers beyond 5σ are iteratively removed. The polynomial is used to correct all measured magnitudes, yielding a calibrated light curve for each source. Because SExtractor’s formal errors severely underestimate uncertainties for bright objects (the Poisson term is lost in the difference image), the authors derive an empirical error model: they compute the RMS of m_diff for the calibration stars in magnitude bins, interpolate this RMS–mag relation, and adopt the larger of this empirical error and the original SExtractor error for each measurement.

Applying this pipeline to ~39.8 million sources over >8,000 deg² (restricted to –29° < Dec < +15° and |b| ≳ 20°) results in a catalogue of >42 million ZTF‑IDs, each with a time‑series of calibrated magnitudes. The authors then compute a suite of variability features (standard deviation, excess variance, structure function parameters, etc.) for each light curve.

To assess the improvement over the standard ZTF Data Release 11 PSF‑photometry light curves (DR11‑psf), they cross‑match their catalogue with the labelled set used in Sánchez‑Sáez et al. (2023). For non‑variable stars the two datasets give comparable results. However, for non‑variable low‑redshift galaxies the classic DR11‑psf light curves spuriously flag ~98 % as variable (due to host‑galaxy contamination and PSF‑photometry systematics), whereas the new DI‑Ap light curves reduce this fraction to ~7 %, demonstrating a dramatic suppression of false variability. For genuine AGN the detection fraction drops modestly from 99 % (DR11‑psf) to 83 % (DI‑Ap), indicating that the new method retains most real variability while being more conservative against noise. Importantly, the DI‑Ap approach recovers weaker variability signals in host‑dominated, low‑luminosity AGN that are often missed or diluted in PSF photometry.

The authors train a Random Forest classifier on 14 astrophysical classes (including stars, variable stars, non‑variable galaxies, low‑z AGN, mid‑z AGN, high‑z AGN, blazars, etc.) using the variability features derived from the DI‑Ap light curves. Cross‑validation yields an overall accuracy > 94 % and especially high completeness and purity for low‑z AGN, even when the nucleus is embedded in a resolved galaxy. Applying the classifier to the full dataset produces 341,938 AGN candidates, split into four redshift‑based subclasses (low‑z, mid‑z, high‑z, blazars).

A key validation is performed against the eROSITA/eFEDS X‑ray catalogue. Within the eFEDS footprint, ZTF‑DI‑Ap AGN candidates that are bright enough for optical monitoring and have X‑ray detections are classified as AGN in 79 % of cases, while 20 % are labeled as non‑variable galaxies. The X‑ray luminosities of the two groups differ by ~2 dex, yet their X‑ray fluxes are comparable, highlighting that optical variability can separate intrinsically luminous X‑ray AGN from lower‑luminosity counterparts. Conversely, a substantial fraction of X‑ray AGN are optically too faint for ZTF, and about a quarter of the ZTF‑selected AGN lack X‑ray detections, underscoring a wide distribution of X‑ray‑to‑optical flux ratios (spanning >2 dex).

In summary, the paper demonstrates that (1) forced aperture photometry on ZTF difference images, combined with robust per‑epoch calibration and empirically derived error estimates, dramatically reduces systematic variability caused by host‑galaxy light and PSF variations; (2) the resulting light curves enable a high‑performance Random Forest classifier that can reliably identify AGN across a broad range of redshifts and host‑galaxy properties, including low‑luminosity, host‑dominated systems that are challenging for traditional colour‑ or X‑ray‑based selections; and (3) cross‑matching with eROSITA confirms that optical variability provides complementary information to X‑ray surveys, revealing a heterogeneous AGN population with diverse X‑ray‑to‑optical ratios.

The methodology is readily scalable to the full ZTF sky and can be directly applied to upcoming large‑scale time‑domain surveys (e.g., LSST) and spectroscopic follow‑up programs such as 4MOST, offering a powerful tool for constructing complete, unbiased AGN samples and probing SMBH growth across cosmic time.