On joint analysing XMM-NuSTAR spectra of active galactic nuclei

Figure  1.  An example (XMM-Newton obsid 0741330101) of the adopted regions when processing the XMM-Newton EPIC-pn data in SW mode. The image is extracted at 3–10 keV with evselect and plotted in logarithmic scale with ds9. The green circle is the source region, which is optimally determined by eregionanalyse, while the two white circles with a radius of 40 arcsec are the background regions. Note that part of the source region is outside the FOV (often the case for the SW mode), which will be automatically corrected by arfgen. Outside the red circle with a radius of 100 arcsec is the source-free region, used to filter intervals with a flaring background.

Figure  2.  Usable fraction of the GTI after filtering the periods showing flaring background with different thresholds, for the 85 XMM-Newton exposures in SW mode.

Figure  3.  Example NuSTAR FPMA 3–78 keV (ObsID: 60501049002) and XMM-Newton EPIC-pn 3–10 keV (ObsID: 0852210101) light curves (with a time bin of 500 s) of Mrk 1383, to illustrate the merge of NuSTAR and EPIC-pn GTIs. The grey shades represent the dropped time intervals for each instrument, mainly due to the Earth occultation for NuSTAR, and flaring background for EPIC-pn. The blue dots represent the remaining data after merging the GTIs, while the red open circles mark the data dropped due to the merging of GTI. For EPIC-pn (lower panel), we over-plot the background rate curve (grey stars) used to filter the intervals with flaring background, and the corresponding data points filtered out from the source light curve (grey circles). The duration of the exposures and net exposure time (before/after merging GTIs) are labelled, respectively.

Figure  4.  The left panel: photon index $ \varGamma $ of the NuSTAR spectra versus those of the EPIC-pn spectra before (red square) or after (blue circle) correcting the effective area (corresponding to $ \varGamma^{\rm Nu} $, $ \varGamma^{\rm pn} $, and $ \varGamma^{\rm pn-Cor} $ in Table 1, respectively). The colored solid lines show the linear regression results (in comparison with the black 1∶1 line), with the shadows showing the 1$ \sigma $ uncertainty derived through bootstrapping the sample. The middle and right panels show the case for observations before/after 2017-01-01.

Figure  5.  Absorption column density $ N_{\rm H} $ of the NuSTAR spectra versus those of the EPIC-pn spectra before (red square) or after (blue circle) correcting the effective area. The colored solid lines show the linear regression results derived by asurv (in comparison with the black 1∶1 line), with the shadows showing the 1$ \sigma $ uncertainty derived through bootstrapping the sample.

Figure  6.  $ \Gamma $, $ E_{{\rm{cut}}} $ and $ R $ derived through fitting the NuSTAR spectra alone versus joint-fitting with EPIC-pn spectra. For $ E_{{\rm{cut}}} $ and $ R $, we perform linear regression in logarithmic space with asurv^[42] to handle the censored data points (as hollow markers). The colored solid lines show the linear regression results (in comparison with the black 1∶1 line), with the shadow showing the 1$ \sigma $ uncertainty derived through bootstrapping the sample.

Figure  7.  The photon index $ \varGamma $ (y-axis) derived using the effective area corrected XMM-Newton EPIC-pn spectra, after filtering the periods showing flaring background with a threshold of 0.05 counts/s (blue squares) or 0.4 counts/s (brown circles), versus those derived using NuSTAR spectra (x-axis), for the 33 observation pairs with significant flaring background. The colored solid lines show the linear regression results (in comparison with the black 1∶1 line), with the shadow showing the 1$ \sigma $ uncertainty derived through bootstrapping the sample.

Figure  8.  Lost fraction of the NuSTAR net exposure after requiring a perfect simultaneity between NuSTAR and XMM-Newton data.

Figure  9.  The distribution of the relative errors of $ \varGamma $, $ R $ and $ E_{\rm cut} $, with the detection fraction of $ R $ and $ E_{\rm cut} $ provided in the legend. The top panels show the distributions of $ \log_{10}\left( {\rm \dfrac{upper\,error}{value}} \right) $ of the three parameters, while the bottom panels show those of $ \log_{10}\left( {\rm \dfrac{lower\,error}{value}} \right) $. Blue boxes show the results of the joint-fitting of NuSTAR and EPIC-pn, while the orange boxes show the result of fitting the NuSTAR spectra only (from the whole NuSTAR exposure without matching EPIC-pn GTI).