The Optical Variability of Active Galactic Nuclei: A Probe of Black Hole Physics
Abstract
Active Galactic Nuclei (AGN) exhibit strong, rapid optical luminosity fluctuations that are often described as stochastic (exhibiting intrinsic randomness). The physical mechanisms that produce stochastic fluctuations may be studied by characterizing the correlation structure of such data in both the time domain and the frequency domain. We find evidence that AGN do not exhibit a universal optical PSD; the PSD shape (or light curve correlation structure) may be a function of fundamental AGN properties. In this work, we develop an application of continuous-time autoregressive moving average models (CARMA) to AGN optical light curves, and we discuss their relationships to more traditional techniques such as the structure function (SF) and the power spectral density (PSD).We present a study of AGN variability using the Damped Harmonic Oscillator (DHO) or CARMA(2,1) model. We estimate DHO timescales and optical variability amplitudes for 7492 objects with both SDSS (Stripe 82) r-band and CRTS V-band light curves. We employ clustering methods (t-distributed stochastic neighbor embedding) in this work to test the non-universality of AGN variability (deviations from the damped random walk; DRW or CARMA(1,0)). This approach groups together light curves with similar correlation structure as described by DHO timescales.We identify three classes of AGN variability. We find a ``long memory" variability class characterized by milder variability amplitudes (∼12\%) and steeper SFs than DRWs. We also detect a variability class characterized by an excess of short timescale variability (<100 days). The onset of the excess takes the form of a secondary break in the SF/PSD at short timescales. The break timescale (CRTS light curves) is anti-correlated with bolometric luminosity (ext{log }au_{perturb} = (-0.26± 0.03) ext{ log } L_{Bol} +13.2). Finally, we identify failed DHO model-fits with characteristic timescales that are greater than the timescale detection limit of a single 8-10 year survey. We estimate that for 40\% of AGN we cannot meaningfully estimate timescales without combining decades-long surveys. We find a strong discrepancy between SDSS and CRTS light curves for the same AGNs at timescales relevant to the secondary SF break. The discrepancy calls into question the impact of sparse cadence (SDSS) and large photometric uncertainties (CRTS) on similar methods such as the SF and PSD that are used to characterize the correlation structure of stochastic data.We also explore a Kepler/K2 dataset of well-sampled (∼30 minute cadence) light curves for ∼ 4000 spectroscopic AGN and AGN candidates. We detect instrumental systematics in K2 data that require careful error mitigation. Our investigations provide characterizations of the noise dependence on CCD channel, pixel row/column, and magnitude. We provide recommendations for continued development of processing software to rehabilitate K2 data for the study of AGN PSD shapes at short timescale as a function of physical properties (luminosity, black hole mass and Eddington ratio). The results of this investigation suggest that the optical variability of AGN specifically, at short timescales (<100 days), may contain the richest insights for AGN science.
- Publication:
-
Ph.D. Thesis
- Pub Date:
- 2019
- Bibcode:
- 2019PhDT.......151M
- Keywords:
-
- Astrophysics