Learning about the Torus from AGN demographics

  • 1 Results from X-ray population studies
  • 2 What can modellers learn about the torus from this?
  • 3 What should observers take away from this?

    See the video explanation, or read below

    1 Results from X-ray population studies

    Buchner+15 worked out the intrinsic number of AGN as a function of accretion luminosity, obscuring screen density and redshift. This is possible in X-ray because the selection function is well understood and independent of the host galaxy.

    Important results are:

    1. 3/4 AGN in the Universe are seen obscured. 3/4 of sight lines from the central SMBH are obscured. More technically: The fraction of obscured AGN ( $N_{H}>10^{22}\text{cm}^{-2}$) averaged over cosmic time is $77_{-5}^{+4}\%$.
    2. 1/3 AGN in the Universe are seen Compton-thick. 1/3 of sight lines from the central SMBH are Compton-thick. More technically: The fraction of AGN with $N_{H}>10^{24}\text{cm}^{-2}$ averaged over cosmic time is $38_{-7}^{+8}\%$. (Buchner+15: $38_{-7}^{+8}\%$, supported by ultra-hard X-ray selection finding $30\%$ Ricci in prep., local infrared selection by Annuar+14 finding $\sim30\%$, and another X-ray survey study by Aird+15 finding $\sim20\%$).
    3. The fraction of Compton-thick AGN does not seem to change with luminosity.
    4. The fraction of obscured AGN decreases with increasing luminosity.

      Obscured fraction of AGN, X-ray column density

      1. $f_{\text{obsc}}$ is $75\%$ at $L_{X}=10^{42-43}\text{erg/s}$ and $40\%$ at $L_{X}=10^{45-46}\text{erg/s}$, when not considering the constant Compton-thick population. See figure, z=0.5-1:
    5. It is not generally true that the fraction of obscured AGN increases with redshift. It increases for $L_{X}=10^{44-45}\text{erg/s}$ and decreases for $L_{X}=10^{42-43}\text{erg/s}$. Why?

      1. The luminosity-dependence has a critical turn-over luminosity $L_{\text{crit}}\approx10^{43}\text{erg/s}$ where it shows a peak (Burlon+11 in the local Universe, Buchner+15)
      2. The entire shape shifts to higher luminosities with redshift! (Buchner+15). See figure. The star is always at the same position. Compare between redshift panels.

    2 What can modellers learn about the torus from this?

    Compton-thick and Compton-thin covering fractions illustrated as a torus.
    1. Aim for opening angles of only $30\text{\textdegree}$ or a covering fraction of $75\%$. Everything else is $N_{H}>10^{22}\text{cm}^{-2}$. (see illustration)
    2. For Compton-thick densities, aim for opening angles of $60\text{\textdegree}$ or a covering fraction of $35\%$. (see illustration)
    3. Not all torii must look the same. But if you model lower opening angles, keep in mind that another part of the population must have higher obscuration. Think about them too.
    4. The obscuration-luminosity anti-correlation means that the obscuration is tightly linked to the accretion process, and therefore nuclear and should be explained by torus models. Try to make a model that increases the covering with luminosities, and then decreases it again with high luminosities. This could be explicitly through a causation $L_{X}\rightarrow f_{\text{obsc}}$ (feedback) or $f_{\text{obsc}}\rightarrow L_{X}$ (feeding), or implicitly due to a correlation.
    5. Redshift-evolution: If you have a critical luminosity $L_{\text{crit}}$ where this effect sets on, $L_{\text{crit}}$ has to depend on something else that could evolve with redshift. For example, if it is dependent on Eddington ratio, the mass can evolve over redshift (downsizing in black hole mass).

      1. Note that just putting more gas at high redshift is ruled out: This would increase $f_{\text{obsc}}$ at all luminosities (relation goes up), not shift the peak luminosity.
      2. Buchner+15 systematically discusses a few broad classes of models and why they are ruled out.

    3 What should observers take away from this?

    1. If you select AGN at some specific redshifts, and compare to a lower redshift bin at the same luminosity:

      1. You are not looking at the same population. Scale down the luminosities you compare to.
    2. When you create an obscured AGN sample and an unobscured AGN sample:

      1. The obscured sample will have a lower average luminosity, therefore either lower Eddington rate or higher mass.
      2. Use intrinsic luminosity matching.

    4 References

    (Web|ADS|PDF) Buchner et al. (2014) -- We investigated the opening angle and geometry of the torus through a new Bayesian Spectral analysis method applied to Chandra Deep Field South data.

    (Web|ADS|PDF) Buchner et al. (2015) -- With a large sample spreading X-ray luminosity, redshift and column density, we plot these quantities against each other (incorporating selection effects). More technically, the space density as a function of L,z,NH is investigated, as well as the fraction of Compton-thick AGN, the fraction of obscured AGN. We also discuss torus models that could reproduce the luminosity-dependent obscuration and its evolution.

    (PDF|image) Poster for the TORUS2015 conference. Contains a comparison to the model of Wada (2012), which roughly reproduces the expected column density distribution.

    AGN catalogues for the CDFS and AEGIS-X fields are released at the MPE X-ray surveys website. This includes detection and redshift catalogues as well as derived spectral parameters, such as luminosities, column densities and probability of the object being obscured/Compton-thick.

    Contact me for questions, any additional plots or the computed space densities. The space densities are available as a table in txt and fits format, with the columns described in the CDS catalogue.

  • Impressum: Johannes Buchner. Previously a PhD student at MPE/Germany under Antonis Georgakakis and Kirpal Nandra, since April 2015 Post-doctoral fellow at PUC/Chile with Franz Bauer. Contact: johannes [dot] buchner [dot] acad [ät] gmx.com.