Abstract

Fitting AGN mid-infrared spectra efficiently

Peter Scicluna (ASIAA)

Ciska Kemper (ASIAA) Sundar Srinivasan (ASIAA) Catharinus Dijkstra (Almere, Netherlands)

The volume of observational data is increasing at an ever-growing rate, and it is already proving difficult for analysis to keep up with the avalanche of data. For example, the Spitzer archives alone contain thousands of spectra suitable for detailed analysis (Lebouteiller et al., 2015) many of which have yet to be fully exploited, seven years after the end of cryogenic operations. JWST is expected to produce an even larger volume of high-quality spectra, and it will be necessary to adopt new, highly-automated but computationally intensive methods to fully exploit such a large database.

With this problem in mind, we are developing a framework for robustly fitting infrared spectra, incorporating dust radiative transfer modelling and modern techniques that efficiently explore parameter space (e.g. the Pikaia algorithm, Charbonneau 1995, see Dijkstra 2007 and Baier et al., 2010 for an application to dust emission). This framework must be combined with automatic classification schemes to apply the correct class of models to each object. This removes the major part of the human interaction in the analysis process, making the availability of computing time the only bottleneck.

Although the resulting framework is generally applicable, our primary goal is to explore the dust content and mineralogy of AGN tori. AGN disc winds are expected to harbour conditions favourable to dust formation, and may make a significant contribution to the dust budget in the early universe (Elvis et al., 2002). By including a complete treatment of radiative transfer, it is possible to constrain the mineralogy without resorting to simple assumptions such as pure absorption. This will probe the physical conditions in the dust-forming region of a large sample of AGN, and robustly constrain the contribution of AGN-formed dust to galactic dust budgets. So far, the dust mineralogy has been studied in detail for only a handful of quasars, manually (Markwick-Kemper et al. 2007; Köhler & Li 2010; Smith et al. 2010; Xie et al. 2014), while automated fitting techniques typically focus on other aspects of the spectra, and use a standard grain model, such as that of Draine & Li (2007), for the dust. Applying such a tool to a large sample of AGN across the range of luminosities to which JWST-MIRI will be sensitive, will enable a census of dust-forming conditions up to a redshift ~ 1.

Currently, some 2,500 spectra of active galaxies are available in the Spitzer archive (Lebouteiller et al. 2015), which raises expectations of a multiple of that number to be obtained with JWST. If this project is to be completed in less than a million CPU hours, then the time available to explore parameter space, and converge to a solution, is approximately 50 hours per source. In the case of AGN tori, the radiative transfer calculations are potentially very time-consuming (in case of an edge-on view of the torus). Optimization is therefore essential, and we will discuss strategies to reduce the CPU time needed.

Mode of presentation: poster