Context. Sky surveys produce enormous quantities of data on extensive regions of the sky. The easiest way to access this information is through catalogues of standardised data products. XMM-Newton has been surveying the sky in the X-ray, ultra-violet, and optical bands for 20 years. Aims. The XMM-Newton Survey Science Centre has been producing standardised data products and catalogues to facilitate access to the serendipitous X-ray sky. Methods. Using improved calibration and enhanced software, we re-reduced all of the 14 041 XMM-Newton X-ray observations, of which 11 204 observations contained data with at least one detection and with these we created a new, high quality version of the XMM-Newton serendipitous source catalogue, 4XMM-DR9. Results. 4XMM-DR9 contains 810 795 detections down to a detection significance of 3σ, of which 550 124 are unique sources, which cover 1152 degrees2 (2.85%) of the sky. Filtering 4XMM-DR9 to retain only the cleanest sources with at least a 5σ detection significance leaves 433 612 detections. Of these detections, 99.6% have no pileup. Furthermore, 336 columns of information on each detection are provided, along with images. The quality of the source detection is shown to have improved significantly with respect to previous versions of the catalogues. Spectra and lightcurves are also made available for more than 288 000 of the brightest sources (36% of all detections).

The inner radius of the accretion disc around a black hole in the low/hard state can be measured in one of two ways. Firstly, via the extent of broadening of the iron emission line, and secondly, from the luminosity and temperature of the weak soft component seen in this state, assuming it is the disc. We use both of these methods on all the low/hard state spectra taken in timing mode of XMM-Newton 's EPIC-pn. We find that the two methods are not consistent with each other, and the difference is not always in a single direction. The two methods are neither model independent, nor are they independent of current calibration issues. We find that the remaining small residuals in the EPIC-pn timing mode response at the ≤ 3% level can have a dramatic effect on the fit parameters for the reflected spectrum. There is also a mismatch in cross-calibration with RXTE, which makes it difficult to use simultaneous data to extend the bandpass of the spectral fits. Nonetheless, it is clear from the data that the iron line is noticeably broader and stronger at higher L/L Edd , which is consistent with the truncated disc models. We also show that it is likely that the soft component changes character, from a stable component consistent with a truncated disc at high L/L Edd , to a variable one with much smaller radius at low L/L Edd . This adds to growing evidence for a complex soft component in the low/hard state, possibly resulting from clumps torn from the edge of the truncated disc.

We extract all the XMM–Newton European Photon Imaging Camera (EPIC) pn burst mode spectra of GX 339−4, together with simultaneous/contemporaneous RXTE data. These include three disc‐dominated and two soft intermediate spectra, and the combination of broad bandpass/moderate spectral resolution gives some of the best data on these bright soft states in black hole binaries. The disc‐dominated spectra span a factor of 3 in luminosity, and all show that the disc emission is broader than the simplest multicolour disc model. This is consistent with the expected relativistic smearing and changing colour temperature correction produced by atomic features in the newest disc models. However, these models do not match the data at the 5 per cent level as the predicted atomic features are not present in the data, perhaps indicating that irradiation is important even when the high‐energy tail is weak. Whatever the reason, this means that the data have smaller errors than the best physical disc models, forcing use of more phenomenological models for the disc emission. We use these for the soft intermediate state data, where previous analysis using a simple disc continuum found an extremely broad residual, identified as the red wing of the iron line from reflection around a highly spinning black hole. However, the iron line energy is close to where the disc and tail have equal fluxes, so using a broader disc continuum changes the residual ‘iron line’ profile dramatically. With a broader disc continuum model, the inferred line is formed outside of 30Rg, so it cannot constrain black hole spin. We caution that a robust determination of black hole spin from the iron line profile is very difficult where the disc makes a significant contribution at the iron line energy, i.e. in most bright black hole states.