A Phenomenological Study of the Accretion Disk in the Super-Eddington AGN I Zw 1

TL;DR

This study uses three years of high-cadence optical monitoring of the super-Eddington AGN I Zw 1, supplemented by UV/optical data, to perform continuum reverberation mapping and probe the accretion-disk structure. Through cross-correlation and PyROA analyses, it finds a lag-wavelength spectrum with disk-dominated high-frequency signals and evidence for a secondary, BLR-like reprocessor at larger radii that lengthens lags at low frequencies. The high-frequency lags are compatible with a compact disk size near the fiducial thin-disk value, while the low-frequency behavior requires a diffuse-continuum reprocessor, likely associated with the BLR, and the u-band excess suggests Balmer-jump effects. The optical-to-mid-IR size compilation indicates the disk remains largely separate from the BLR, challenging models that require the disk to extend into the BLR, and demonstrating the usefulness of Fourier lag analyses in disentangling multiple reverberation components in AGN interiors.

Abstract

The structure of the accretion disk in AGN is still an unsolved question, especially how it may change with Eddington ratio. Here we examine the accretion disk in the super-Eddington AGN I Zw 1 using reverberation mapping of the optical continuum. We use three years of optical monitoring with Las Cumbres Observatory at sub-day cadence in $uBgVriz_s$. The lag-wavelength spectrum, calculated using the cross correlation method and PyROA, shows a $u$-band excess. PyROA lags are equally well fitted with a thin and slim disk profile. The UV/optical AGN spectral energy distribution is consistent with a thin disk. The disk size at 4495 Å for a thin disk model is $4.23\pm0.24\:\mathrm{ld}$ and for a slim disk model is $1.71\pm0.09\:\mathrm{ld}$, larger by a factor of $2-4$ than the fiducial disk size of $1.07\pm0.15\:\mathrm{ld}$ as determined using the Eddington ratio. We find evidence of different size scales probed with different variability timescales. Lags evaluated at longer variability timescales increase as do frequency-resolved lags at low frequencies, which we interpret as an additional secondary reprocessor at large radii consistent with the broad-line region (BLR) in I Zw 1. The high frequency lags, predicted well with just a disk, are fit with a thin disk profile and a size of $0.61\pm0.37\:\mathrm{ld}$. This indicates that the actual disk size may be on the order of the fiducial size. We also collate the most extensive set of directly measured internal sizes of an AGN, from optical to mid-infrared with reverberation mapping and optical interferometry. Assuming that the disk is indeed the fiducial size, these show little evidence that the accretion disk extends into the BLR significantly, tentatively disfavouring the failed radiatively accelerated dust driven outflow BLR formation model.

Figures (14)

• Figure 1: The LCO light curves in all bands for Years 2--4, the PyROA model (solid line), and its $68$% confidence interval shown in grey, including the reference light curve $X(t)$ (top panel). The right panel shows the marginalised posterior distributions for the inter-band lags as calculated by PyROA, with its mean (solid line) and $68$% confidence interval denoted by the dotted lines. Lags are measured with respect to the $g$-band.
• Figure 2: Lag spectrum for Years 2--4 as calculated using ICCF, using the centroid lag $\tau_\mathrm{cent}$ and with reference to the $g$-band in the AGN rest frame. Year 2 is denoted by the circles, Year 3 by the triangles, and Year 4 by the squares. Lags plotted here are presented in Table \ref{['tab:lags']}. Lags increase with wavelength for all years.
• Figure 3: Lag spectrum as calculated using PyROA simultaneously for all three years in the AGN rest frame, with most of the errorbars too small to be visible. As in Fig. \ref{['fig:iccf_lags']}, lags increase with wavelength. A thin disk profile with $\tau \propto \lambda^{4/3}$ and a slim disk profile with $\tau \propto\lambda^{2}$ is fitted to these data (Table \ref{['tab:fits']}). These are shown with the solid lines and their error regions are shaded. As a comparison, the fiducial thin disk profile for I Zw 1 with its mass and bolometric luminosity is illustrated with the dashed line, with $\tau_0 = 1.07\:\textnormal{days}$ as calculated in Section \ref{['disc:rm_signals_source']}.
• Figure 4: Lags calculated using PyROA while varying the variability stiffness parameter $\Delta$ with $\Delta = 3, 5, 10, 20\:\mathrm{days}$ in the AGN rest frame. $\Delta = 3$ is denoted by the circles, $\Delta = 5$ by the squares, $\Delta=10$ by the triangles, and $\Delta = 20$ by the stars. As $\Delta$ increases, the reference light curve stiffens and longer variability timescales are probed. The plot shows that as that variability timescale increases, the magnitudes of the lags increase.
• Figure 5: The lag-frequency and coherence spectra for all bands derived from Fourier analysis. The studied frequency range is $0.013-2.9\:\mathrm{days}^{-1}$ ($1.5\times10^{-7}-3\times10^{-5}\:\mathrm{Hz}$), however data at frequencies above $0.5\:\mathrm{days}^{-1}$ (vertical dotted lines) is uninformative as the ROA washes out the variations (and correlates adjacents points, also increasing its coherence). The data points are plotted with circles, the simple thin disk is denoted with the dashed line, and the disk+secondary reprocessor model is denoted by the solid line. The model's uncertainty envelope is shown by the shaded region. The median delay of the secondary reprocessor in this model is $\tau_M = 20\:\mathrm{days}$.