Online material

Appendix A: Observations and data reduction

Appendix A.1: XMM-Newton

EPIC

The European Photon Imaging Camera (EPIC – Strüder et al. 2001; Turner et al. 2001) data of NGC 5548 were first processed using the Science Analysis System (SAS v13.0). The EPIC instruments were operating in the Small-Window mode with the thin-filter applied. Periods of high flaring background were filtered out by applying the #XMMEA_EP and #XMMEA_EM filterings for pn and MOS, respectively. This is done by extracting single event, high energy lightcurves, in order to create a set of good-time-intervals (GTIs) for use in conjunction with the internal GTI tables, to exclude intervals of flaring particle background exceeding 0.4 count/s for pn and 0.35 count/s for MOS. The EPIC spectra and lightcurves were extracted from a circular region centred on the source with a radius of 40′′. For pn the background was extracted from a nearby source-free region of radius 50′′ on the same CCD as the source, whereas for MOS a background radius of 100′′ was used from a CCD other than that of the source. For archival (2000, 2001) EPIC data, there was weak pile-up (few % for pn); however, for the 2013 observations due to the suppressed X-ray fluxes no pile-up was detected. The single and double events were selected for both the pn (PATTERN <= 4) and MOS (PATTERN <= 12). The EPIC lightcurves were background-subtracted and corrected (using the epiclccorr SAS task) for various effects on the detection efficiency. Response matrices were generated for the spectrum of each observation using the rmfgen and arfgen tasks. With EPIC-pn operating in the Small-Window mode, the final cleaned EPIC-pn exposure for each XMM-Newton observation was on average about 37 ks, which is about 67% of the duration of an observation (i.e. about 55 ks as given in Table 1).

The 2013–2014 EPIC-pn data of NGC 5548 were found to be affected by larger than expected long-term degradation/evolution of the charge transfer inefficiency (CTI). This creates a shift of about 50–60 eV at 6 keV. The CTI problem had previously been found in the Mrk 509 campaign EPIC-pn data (Ponti et al. 2013), and so the long-term CTI was corrected and implemented in new SAS CCFs. For the new EPIC-pn data of NGC 5548, further CTI corrections had to be applied and the use of the new CCFs were not sufficient to correct for all remaining CTI. We thus used an ad-hoc correcting energy shift as described in more detail in Cappi et al. (in prep.), in which the full data analysis and modelling of the EPIC data is reported. This energy shift results in a poor fit near the energy of the gold M-edge of the telescope mirror, therefore, the 2.0–2.4 keV region was omitted from our spectral modelling of the 2013 EPIC-pn data.

RGS

The RGS (den Herder et al. 2001) instruments were operated in the standard Spectro+Q mode for the XMM-Newton observations of NGC 5548. The processing of the RGS data has been performed at a more advanced level than the standard SAS pipelines. The details of this enhanced RGS data reduction technique are reported in Kaastra et al. (2011). The procedure takes advantage of the multi-pointing mode of XMM-Newton and utilises accurate relative calibration for the effective area of the RGS. It also incorporates an accurate absolute wavelength calibration to improve the method for stacking time-dependent RGS spectra and enhancing the efficiency of the spectral fitting. The Kaastra et al. (2011) procedure was used to produce a combined RGS-1/RGS-2 spectrum and response matrix (in SPEX format) for each XMM-Newton observation of NGC 5548, including archival 2000 and 2001 observations. Also using this procedure a single stacked RGS spectrum and response matrix from the 12 XMM-Newton observations of summer 2013 was created for spectral fitting.

OM

The Optical Monitor (OM; Mason et al. 2001) photometric filters were operated in the Science User Defined image/fast mode for the 2013–2014 observations and in the Image mode for the archival observations of NGC 5548. In each 2013–2014 observation, OM images were taken with the V,B,U,UVW1,UVM2, and UVW2 filters, with an exposure time of 4.4 ks for each image. The OM images of NGC 5548 were processed with SAS v13.0 omichain pipeline with the standard parameters, to apply the necessary corrections including the removal of Modulo-8 fixed pattern noise, and produce images ready for aperture photometry. The aperture photometry on each image was done using the omsource program. For those archival datasets in which exposures in some of the six OM filters have not been taken, we used the relations derived in Mehdipour et al. (in prep.) between different photometric filters from extensive Swift monitoring of NGC 5548, to calculate the missing fluxes in order to carry out a uniform and consistent spectral fitting of the data. Source and background regions were selected to extract count rates, with the necessary corrections applied, including those for the point spread function (PSF), coincidence losses and time-dependent sensitivity. The OM count rates were extracted from a circle of 5.0′′ centred on the source nucleus. The background was extracted from a source-free region of the same size. The OM count rates were then transformed into the standard OGIP FITS format (Arnaud et al. 1992) to be used together with their corresponding response files in X-ray spectral fitting packages. Lastly, the files were converted into the SPEX format using the auxiliary trafo program of the SPEX package.

In addition to the OM photometric filters, spectra with the Optical and UV OM grisms were also obtained in each 2013 XMM-Newton observation in the Full Low mode. The exposure time of each grism spectrum is 5.0 ks. The grism data were first processed with the SAS v13.0 omgchain pipeline. The necessary corrections, including that for Modulo-8 fixed pattern noise and removal of scattered light features, were applied to obtain undistorted and rotated grism images. We then used the omgsource program to interactively identify zero and first order dispersion spectra of our source, properly define the source and background extraction regions and extract the calibrated spectrum from the grism images of each observation. For the Optical grism the extraction and calibration was extended beyond the standard results to just below 7000 Å as described in Mehdipour et al. (2011) in order to capture the Hαλ6563 line. Similar to the OM broadband filter data, the grism spectra were also converted into SPEX format for simultaneous spectral modelling with other data.

Appendix A.2: Swift

XRT

For the Swift X-ray Telescope (XRT; Burrows et al. 2005) data reduction, we used the procedure detailed in Evans et al. (2009), which is an enhanced version of the standard Swift processing including several modifications. This tool is made available online on the UK Swift Science Data Centre (UKSSDC⁴). The XRT instrument operated in the photon counting (PC) mode for nearly all of the NGC 5548 observations. The XRT data have been corrected for bad pixels and effects of vignetting and PSF to produce cleaned event files using the xrtpipeline script. The data were not significantly affected by pile-up at the observed low count rates of the source. Using the event lists, exposure maps were made, which together with the source spectrum, were used to create a corresponding ancillary response file (ARF) for each interval. The corrections have been applied to both spectra and lightcurves. The optimum extraction radius for data products depends on the PC-mode count rate as reported in Evans et al. (2009), which for NGC 5548 translated most of the time into an extraction radius of 25 pixels (59.0′′). The XRT lightcurves at different energies were constructed from each Swift snapshot observation of NGC 5548. The default grades of 0–12 in the PC mode were used for event selections. The corresponding response matrix files (RMF) for the Swift observations were obtained from the HEASARC Calibration Database (CALDB).

UVOT

The Swift UV/Optical Telescope (UVOT – Roming et al. 2005) data of NGC 5548 from Image-mode operations were taken with the six primary photometric filters of V,B,U,UVW1,UVM2 and UVW2. The uvotsource tool was used to perform aperture photometry using a circular aperture radius of 5.0′′. The standard instrumental corrections and calibrations according to Poole et al. (2008) were applied. Any data from when Swift had bad tracking were removed. The continuous monitoring of NGC 5548 indicated occasional exposures in UVOT filters in which the flux suddenly dropped by about 20% or more compared to the adjacent mean; these exposures were found to be clustered at particular regions of the detector and their cause and fix is currently being investigated by the UVOT calibration team. These occasional observations have been filtered out of our analysis (about 2% of all UVOT observations). For the purpose of spectral fitting with SPEX, the filter count rates and the corresponding response matrices in each filter were created.

In addition to the photometric exposures, there was an extensive monitoring of NGC 5548 with the UV grism of UVOT between 1 April and 12 September 2013. The grism observations of NGC 5548, which were taken about every two days, were performed in the Clocked mode, which is the optimum mode for low background as well as no presence of zeroth orders in the upper left half of the detector. The Swift UVOT Grism Calibration and Software UVOTPY⁵ v1 Kuin et al. (in prep.) was used to produce the UV grism spectra of NGC 5548. This software is more enhanced than the standard tools available within HEASOFT. The uvotgraspcorr program was used to aspect correct the UVOT grism images. The uvotgetspec module of UVOTPY was used to extract first-order spectra of NGC 5548 and also produce the corresponding RMF response matrices. The grism spectra were then converted into SPEX format for spectral modelling.

Appendix A.3: NuSTAR

The NuSTAR observations were reduced using the NuSTAR Data Analysis Software (NUSTARDAS) v1.3.1, which was utilised as part of HEASOFT v6.14 distribution. For instrumental calibration of our NuSTAR data, CALDB version v20131007 was used. The data were processed with the standard pipeline script nupipeline to produce level 1 calibrated and level 2 cleaned event files. The data from the South Atlantic Anomaly passages have been filtered out and event files were cleaned with the standard depth correction, which significantly reduces the internal background at high energies. The source was extracted from a circular region (radius ~110′′), with the background extracted from a source-free area of equal size on the same detector. Then the nuproducts script was run to create level 3 products (spectra, lightcurves, ARF and RMF response files) for each of the two hard X-ray telescope modules (FPMA and FPMB) onboard NuSTAR. The spectra and corresponding response files of the two telescopes were combined for spectral modelling using the mathpha, addrmf, and addarf tools of the HEASOFT package.

Appendix A.4: INTEGRAL

The INTEGRAL IBIS and ISGRI data were reduced with the Off-line Scientific Analysis (OSA) v10.0 software. The spectra were extracted using the standard spectral extraction script ibis_science_analysis. The total INTEGRAL ISGRI spectrum for each observation, and the associated RMF response and ARF ancillary files, were then generated using the OSA spe_pick tool.

Appendix A.5: Chandra LETGS

The three Chandra observations were taken with LETGS using the HRC-S camera. The LETGS data were first reduced using the Chandra Interactive Analysis of Observations (CIAO v4.5) software to create level 1.5 event files. The rest of the processing, until the creation of final level 2 event files, was carried out using an in-house software first described by Kaastra et al. (2002). This procedure follows the same steps as in the standard pipelines of CIAO, but results in improved wavelength and effective area calibration. The outcome of this procedure is the production of spectral files and their corresponding response matrices in the SPEX format (Kaastra et al. 1996).

Appendix A.6: HST COS

The six HST COS observations were taken concurrently with XMM-Newton Obs. 1, 4, 8, 11, 12 and 14. Each COS observation consists of an HST two-orbit visit. The observations through the Primary Science Aperture were taken with gratings G130M (1770 s per visit) and G160M (2215 s per visit), covering the far-UV spectral range from 1132 to 1801 Å. In addition to the routine processing of data with the calibration pipelines of STScI, further enhanced calibrations (as described in Kriss et al. 2011 for Mrk 509) were applied to produce the best-quality COS spectra possible. They include refined flux calibrations that take into account up-to-date adjustments to the time-dependent sensitivity of COS, improved flat-field corrections and wavelength calibration and an optimal method for combining exposures comprising a single visit. More details on the COS calibration of NGC 5548 data are given in Kaastra et al. (2014), and in Kriss et al. (in prep.), in which the full analysis of the HST COS spectra are also reported. For the purpose of modelling the UV continuum as part of our broadband continuum modelling, we extracted the COS flux from six narrow spectral bandpasses, which are free of emission and absorption lines: 1155–1165, 1364.5–1369.5, 1470–1490, 1730–1740, 1730–1770 and 1785–1800 Å. In order to use these measurements in our simultaneous spectral fitting with other optical/UV and X-ray data, we converted the COS fluxes into count rate units used in the OGIP spectral file format. To do this we used the time-dependent sensitivity curves as a function of wavelength, obtained for the COS data in Kriss et al. (in prep.), which are in units of (count s^-1 pixel^-1)/(erg cm^-2 s^-1 Å^-1), to calculate the count rate and the effective area over the six narrow bandpasses. For the G130M and G160M modes, the pixel size is about 0.010 Å and 0.012 Å, respectively.

Appendix A.7: OCA

The Observatorio Cerro Armazones (OCA) monitoring of NGC 5548 was carried out between 20 May and 25 July 2013. For our observations the 25 cm Berlin Exoplanet Search Telescope II (BEST-II) was utilised, which is equipped with a Peltier-cooled 4096 × 4096 pixel Finger Lakes Imager CCD KAF-16801. It has a field-of-view of 1.7 × 1.7 degrees with a pixel size of 9 μm. Photometric measurements were taken in the optical Johnson B,V and R filters. For each filter, 6 dithered images were acquired, with 150 s exposure time for each individual image. There were observations on 27 days, from which 24 had

exposures in B, 19 in R, and 7 in the V filter. Data reduction and processing of the OCA data was performed in the same manner as reported in Pozo Nuñez et al. (2013).

Appendix A.8: Wise observatory

The Wise Observatory (WO) monitoring of NGC 5548 was done with the Centurion 46 cm telescope, and images were obtained with the standard Bessel B,V,R, and I filters. They cover the time from beginning of June to end of September 2013. There were also further monitoring from December 2013 to July 2014. The observations were done with STL-6303 CCD which gives a field of view of 70 × 50 arcmin and each exposure time was 300 s. Standard data reduction was carried out with the Image Reduction and Analysis Facility (IRAF) packages using PSF photometry. Each night a few exposures in each filter were obtained and after confirming that there were no intra-night variations within the uncertainties, they were averaged to produce one measurement for each night. Calibration to absolute magnitudes was done in each filter using several stars in the field which their absolute magnitudes were taken from the SIMBAD database.

© ESO, 2015