Weak AGN Selection from Spectroscopic Diagnostics and Time-Domain Variability

This repository contains my research workflow for identifying weak active galactic nucleus (AGN) candidates by combining spectroscopic classification from DESI and optical variability analysis from ZTF light curves.

Project Overview

Weak or obscured AGN can be difficult to identify using optical spectroscopy alone, especially in host-dominated systems. In this project, I explored whether time-domain optical variability can help recover AGN activity that is weak, diluted, or ambiguous in standard spectroscopic diagnostics.

The current workflow combines:

• BPT-based spectroscopic classification of galaxies into Seyfert, LINER, Composite, and Star-forming classes
• ZTF light-curve download and cleaning
• Variability quantification using fractional variability statistics
• Noise-control analysis based on magnitude-dependent baselines
• Quality-control inspection of representative light curves
• Center-matching and follow-up checks motivated by advisor feedback

Scientific Goal

The main goal of this project is to test whether optical variability can help identify weak AGN that may be missed, diluted, or ambiguously classified in emission-line diagnostics.

More specifically, this project aims to answer:

1. Which spectroscopic classes show excess optical variability?
2. Can variability recover AGN-like behavior in weak or host-dominated systems?
3. How much of the apparent variability is real, and how much is driven by photometric noise, systematics, or off-center contamination?

What I Have Done

1. BPT Classification of DESI Sources

I first classified DESI galaxies into standard emission-line categories using BPT diagnostics:

• Seyfert
• LINER
• Composite
• Star-forming

This provided the spectroscopic sample basis for the later time-domain analysis.

2. Download and Cleaning of ZTF Light Curves

I downloaded ZTF light curves for selected subsamples and built a cleaning workflow for the r-band data.

A key issue in the early version was that multiple nearby sources could be mixed into a single light curve. In the current MINFIX version, I improved the pipeline by:

• enforcing central-source uniqueness
• keeping only the closest OID to the target position
• using a single band (r-band) consistently
• applying quality-control filters on photometric measurements

3. Variability Statistics

I computed variability statistics, including fractional variability (Fvar), and compared the variability distributions across different BPT classes.

Initial results showed that LINER, Seyfert, Composite, and Star-forming samples can all contain apparently high-Fvar objects. However, later analysis showed that a simple fixed Fvar > 5% threshold is not reliable across the full magnitude range.

4. Magnitude-Dependent Noise Diagnosis

A major result of the current stage is that Fvar increases systematically toward fainter sources, indicating that photometric noise and systematics dominate at the faint end.

To address this, I introduced a magnitude-dependent noise baseline:

• within each mean-magnitude bin
• use the 95th percentile of Fvar in Star-forming galaxies
• as a reference noise threshold

This provides a more controlled way to evaluate whether Seyfert or LINER samples show excess variability above the expected noise floor.

5. QC on Bright Seyfert Control Sample

To ensure physical reliability before moving to weaker systems, I built a bright, well-sampled Seyfert control sample, using criteria such as:

• Mean_mag < 18
• N > 500

For this control sample, I generated QC booklets with:

• cleaned light curves
• error bars
• outlier flags
• near-limit flags
• metadata summaries

6. High-Variability Case Inspection

I also inspected representative high-variability objects individually, especially:

• high-Fvar LINERs
• Seyfert–LINER morphological comparisons
• Star-forming high-Fvar cases that may be noise-dominated or transient-like

This step is important because high Fvar alone is not sufficient to identify AGN activity.

7. Current Focus

Based on advisor feedback, the project is now focused on:

• checking off-center spectra
• avoiding double counting
• comparing multi-band variability
• examining low-variability Seyferts
• improving variability metrics beyond a fixed Fvar threshold

Repository Structure

weak-agn-selection/
├── docs/        # progress reports, summaries, advisor-related materials
├── figures/     # selected BPT plots, variability figures, and example light curves
├── results/     # summary tables and compact outputs
├── src/         # main analysis scripts