Online material

Appendix A: Outliers

Based on Fig. 1, we identified 11 main outliers in the distribution of intrinsic column densities (N_H(z)). If this is a true cosmological effect no outlier should be present. Therefore we carefully study these outliers in order to find an explanation for their low intrinsic column densities. We approached the X-ray spectra of these sources and tried to verify if the evaluation of the column density is completely satisfactory from a statistical point of view. In this respect we disregard AGN with an intrinsic column density log N_H< 19 and z< 0.3 because of a too small intervening length scale and, more importantly, because of the AGN influence in cleaning a substantial fraction of the surrounding environment. Five AGN were discarded in this way. For the other six AGN, we provide a detailed description.

Appendix A.1: QSO B0909+5312 at z = 1.377

QSO B0909+5312 was observed by XMM-Newton in May 2003. Data were retrieved from the XMM-Newton archive and reprocessed using SAS 13.5.0. Useful XMM-Newton data were taken from the entire 13 ks and 17 ks exposures of pn and MOS detectors, respectively. Data were extracted following common practice and spectrally binned to 50 counts per energy channel; arf and rmf files were generated using SAS tasks. Fitted energy ranges are 0.2−10 keV and 0.3−10 keV for the pn and MOS instruments, respectively. Eitan & Behar (2013) fitted the same data with an absorbed power-law model. They adopted a Galactic column density of cm^-2 (Kalberla et al. 2005) plus and intrinsic column density at the AGN redshift. Based on their spectrum they derive an upper limit of N_H(z) < 7 × 10¹⁹ cm^-2. We fit our data with a higher binning to increase statistics. We adopted the TBABS model within XSPEC to account for X-ray absorption (Wilms et al. 2000) and with the wilm solar abundance pattern (this is very similar to Asplund et al. (2009) and in the case of oxygen, the main driver for the absorption, is exactly the same) and the vern photoelectric absorption cross sections. We obtain similar results (N_H(z) < 5 × 10¹⁹ cm^-2, 90% confidence level for one parameter of interest, i.e. Δχ² = 2.71) but with a reduced (for 187 degrees of freedom, d.o.f.). Fitting the same data with a broken power law we obtain instead (185 d.o.f.). An F-test indicates that the addition of the spectral break is significant at the > 8σ confidence level. The intrinsic column density in this case is N_H(z) < 5 × 10²⁰ cm^-2, perfectly consistent with our lower envelope.

Appendix A.2: Two AGNs in the Lockman hole at z =1.204 and z = 1.113

J105239.6+572432 (J1052 at z = 1.113) and J105316.9+573552 (J1053 at z = 1.204) are two Type I AGN in the Lockman hole with a very low absorption (N_H(z) < 6 × 10¹⁹ cm^-2 and N_H(z) < 9 × 10¹⁹ cm^-2, respectively (Corral et al. 2011). We took the longest XMM-Newton observation of the Lockman hole not affected by soft proton flare covering 100 ks. Data were extracted as above and binned to 25 counts per spectral bin.

J1052 can be fit with a simple power law, resulting in a (288 d.o.f.) and N_H(z) < 3 × 10²⁰ cm^-2 ( cm^-2). A better fit is obtained with a broken power

law, with (286 d.o.f.) and N_H(z) < 6 × 10²⁰ cm^-2. An F-test indicates that the addition of the spectral break is significant at > 8σ.

J1053 falls on the dead CCD in MOS1 and across dead columns on the pn detector. The fit with a single power law is good (, 129 d.o.f.) and returns N_H(z) < 2 × 10²⁰ cm^-2.

Both values are consistent with our lower envelope.

Appendix A.3: QSO B1345+584 at z = 0.646

QSO B1345+584 is a Type I AGN. Corral et al. (2011) derived an upper limit on the intrinsic column density of N_H(z) < 1 × 10¹⁹ cm^-2 for this source. It was observed by XMM-Newton in Jun 2001 for 40 ks, even if it was not on-axis. Because of a strong proton flare activity at the end of the observation only the first 25 ks and 29 ks were retained for the pn and MOS2 instruments, respectively (the source falls on a CCD gap in MOS1). Data were extracted as above and binned to 25 counts per spectral bin. The fit with a power law (assuming cm^-2) provides a (308 d.o.f.) with a clear under prediction of the high energy spectral part. In this case the intrinsic column density is negligible (N_H(z) < 2 × 10¹⁹ cm^-2). Fitting the data with a broken power law ,we obtain (306 d.o.f.). An F-test indicates that the addition of the spectral break is significant at > 8σ. The intrinsic column density in this case is N_H(z) < 7 × 10¹⁹ cm^-2, in line with similar upper limits at this redshift.

Appendix A.4: QSO B1157–1942 at z = 0.450

QSO B1157–1942 is the lowest redshift quasar (z = 0.45) in the AGN-E13 sample. They quote a very low intrinsic column density of N_H(z) < 1 × 10¹⁹ cm^-2. XMM-Newton observed the quasar on Jun 2008. The exposure time is 25 ks and 24 ks for the MOS and pn detectors, respectively. Data were extracted as above and binned to 50 counts per spectral bin. Fitting the data with an absorbed power -law model ( cm^-2) , we derive N_H(z) < 6 × 10¹⁸ cm^-2 (in line with previous results) and (791 d.o.f.). The residuals show clear signs of an underlying, redshifted iron line. Adding a free Gaussian component at the quasar redshift, the fit improves to (788 d.o.f., F-test probability > 8σ). The derived limit on the intrinsic column density N_H(z) < 1 × 10²⁰ cm^-2 is in line with our expectations.

Appendix A.5: 2E 1852 at z = 0.312

This Type I AGN is reported to have an intrinsic column density N_H(z) < 1 × 10¹⁹ cm^-2 (Corral et al. 2011). The object was observed by XMM-Newton on Apr. 2000 in small window with the two MOS and in full window with the pn. Some flaring activity was present and filtered out reducing the exposure time to 54 ks and 22 ks for the MOS and pn, respectively. MOS data were fitted in the 0.6−10 keV range to encompass calibration uncertainties in small window mode. All data were binned to have at least 25 counts per spectral bin. We fit the spectra with an absorbed power-law model ( cm^-2). We derive a good fit with (1316 d.o.f.) and N_H(z) = (4.3 ± 0.3) × 10²⁰ cm^-2, much larger than the reported value.

© ESO, 2015