VAR-PZ: Constraining the Photometric Redshifts of Quasars using Variability

1. Problem Statement

The Vera C. Rubin Observatory's Legacy Survey of Space and Time (LSST) is expected to detect over 10 million quasars. While spectroscopic follow-up will provide accurate redshifts for a fraction of these, the vast majority will rely on photometric redshifts (photo-z).

• The Challenge: Traditional photo-z estimation for Active Galactic Nuclei (AGNs) via Spectral Energy Distribution (SED) fitting is notoriously difficult. Luminous Type 1 AGNs possess featureless power-law continua lacking strong spectral lines (except at high redshifts where Lyman break/IGM absorption occurs). This leads to:
  • Severe degeneracies in redshift solutions (multi-peaked probability distributions).
  • High rates of "catastrophic outliers" (large errors).
  • Intrinsic variability of AGNs, which introduces noise into SED modeling if not accounted for.
• The Gap: Existing methods (SED fitting, Machine Learning) struggle with these degeneracies, particularly for luminous quasars where host galaxy features are diluted.

2. Methodology: VAR-PZ

The authors propose VAR-PZ, a novel method that leverages the physical properties of AGN variability to constrain photo-zs. The core hypothesis is that AGN variability follows a Damped Random Walk (DRW) process, where the variability parameters depend on redshift, luminosity, and rest-frame wavelength.

Key Technical Steps:

1. DRW Modeling: AGN light curves are modeled as a stochastic process defined by:

   • $\tau$ (Damping Timescale): The characteristic timescale of variability.
   • $SF_\infty$ (Asymptotic Amplitude): The maximum variability amplitude.
   • The Structure Function (SF) is modeled as: $SF(\Delta t) = SF_\infty(1 - e^{-|\Delta t|/\tau})^{1/2}$.

2. Parametric Relations: The authors adopt a parametric model (based on MacLeod et al. 2010, but re-calibrated) linking DRW parameters to physical properties:

   • $\log_{10} f = A + B \log_{10}(\frac{\lambda_{RF}}{4000 \text{ \AA}}) + C(M_i + 23)$
   • Where $f$ is either $\tau$ or $SF_\infty$, $\lambda_{RF}$ is the rest-frame wavelength, and $M_i$ is the absolute i-band magnitude.
   • Note: The model explicitly excludes Black Hole Mass ($M_{BH}$) dependence to remain applicable to large surveys where $M_{BH}$ is unknown, relying instead on $M_i$ and $\lambda_{RF}$.

3. SED Decomposition & Host Dilution:

   • For every trial redshift, the code performs SED fitting to estimate the intrinsic AGN luminosity ($M_i$) and the AGN flux fraction ($R_{AGN}$) in each band.
   • $R_{AGN}$ is crucial because the host galaxy is non-variable; its presence dilutes the observed variability amplitude ($SF_\infty$).

4. Likelihood Construction (The Prior):

   • Using a Gaussian Process (GP) framework (via the Celerite toolkit) and Maximum Likelihood Estimation (MLE), the method calculates the likelihood that the observed light curve matches the expected DRW parameters for a specific trial redshift.
   • This generates a VAR-PZ Probability Density Function (PDF) for redshift.

5. Hybrid Approach:

   • The VAR-PZ PDF is multiplied with the traditional SED-fitting PDF (e.g., from LePHARE or LRT) to create a posterior distribution.
   • This effectively breaks degeneracies where SED fitting alone fails (e.g., distinguishing between low-z and high-z solutions).

3. Key Contributions

• Novel Variability Prior: Introduction of a physically motivated, redshift-dependent prior based on AGN variability timescales and amplitudes, rather than just spectral features.
• Re-calibration of DRW Parameters: The authors re-estimated the coefficients ($A, B, C$) for the DRW relations using modern SED modeling techniques (Assef et al. 2010) to ensure consistency between the luminosity estimates and the variability model.
• Integration Framework: Demonstrated a robust workflow to combine variability constraints with standard SED fitting, applicable to both existing data (SDSS) and future surveys (LSST).
• Code Availability: The implementation is made publicly available, utilizing efficient GP tools (Celerite) to handle large datasets.

4. Results

Validation on SDSS Stripe 82 (Real Data):

• Sample: 9,210 spectroscopically confirmed quasars ($z \ge 0.3$).
• Performance Improvement:
  • Outlier Fraction: Reduced from 32% (SED only) to 22% when combining with VAR-PZ.
  • Low-Redshift Regime ($z_{SED} < 0.6$): This is where SED fitting fails most. The outlier fraction dropped dramatically from 81% to 33%.
  • Precision (NMAD): Improved from 0.0558 to 0.0579 (excluding outliers) and 0.1051 to 0.0850 (including outliers).
• LePHARE Integration: When applied to the more sophisticated LePHARE SED framework, outliers decreased from 23% to 19% overall, and from 66% to 37% for low-redshift candidates.

Simulations (SDSS vs. LSST):

• Cadence Dependence: The method's efficacy is highly dependent on the temporal baseline. For SDSS (short baselines), the method is limited at $z > 0.6$ because the damping timescale $\tau$ exceeds the available baseline.
• LSST Projections:
  • Simulations of the 10-year LSST survey (Wide-Fast-Deep and Deep Drilling Fields) show massive improvements.
  • Wide-Fast-Deep (WFD): Outlier fraction drops from 32% (SED only) to **15%**.
  • Deep Drilling Fields (DDF): Outlier fraction drops to ~9% due to higher cadence and longer baselines.
  • Faint Objects: Even for faint quasars (3x fainter than SDSS limits), the outlier fraction was reduced from 32% to 17%.

5. Significance

• Enabling LSST Science: As spectroscopic follow-up will be impossible for the millions of quasars detected by LSST, VAR-PZ provides a critical tool to obtain reliable redshifts for cosmological studies, black hole evolution, and AGN feedback research.
• Cost-Effective Accuracy: It utilizes existing photometric time-series data without requiring expensive spectroscopy, significantly reducing the fraction of catastrophic outliers that plague current photo-z catalogs.
• Physical Insight: The method validates the physical connection between AGN variability parameters and redshift/luminosity, reinforcing the DRW model as a useful tool for extragalactic astronomy.
• Future-Proofing: The framework is designed to be adaptable to other variability models and future survey strategies, making it a foundational component for the next generation of large-scale AGN surveys.

In conclusion, VAR-PZ represents a significant step forward in photometric redshift estimation for AGNs, transforming variability from a source of noise into a powerful physical constraint that resolves spectral degeneracies.