XMM-Newton Users Handbook

3.4.4.6 The RGS Background

The RGS instrumental background has different components, each with its own characteristics and time dependency. The main components are listed below. The background characterisation has been performed with in-flight data by analysing clear sky images (where there is no source emission) and fitting a model.

• Particle background (minimum ionising particles, protons and ions). These particles deposit most of their energy outside the 0.35 to 2.5 keV energy band and can thus mostly be rejected by the on-board software. There is an additional on-board rejection based on the spatial shape of the events. Except for the period when the spacecraft is passing through the radiation belts (i.e. for spacecraft elevation below 46000 km) and occasional solar storms, the in-flight measured particle background is quite stable. Typically it produces 2.7$\pm$1 events cm$^{-2}$ s$^{-1}$, which is a factor 2 higher than expected pre-launch. This number applies for event rates before any on-board selection and across the whole field of view. The quoted variation is larger than the statistical one and is due to variations in solar background.

• Low energy electrons which enter through the telescope. Electrons with energies up to 20 keV are rejected by the electron deflector at the exit plane of the mirrors. Higher energy electrons create secondary radiation in the spacecraft and detector shielding.

• Fluorescence lines in the detector housing due to interactions with the electron and other minimum ionising particles. The expected strong fluence from the detector housing material, Al, is strongly suppressed by a Au coating. One still expects to see Al K$\alpha$ and Au-M emission in the RGS energy band. These lines are indeed observed at a very low intensity ($\le$ 10$^{-3}$ counts cm$^{-2}$ s$^{-1}$).

• Calibration sources. They can easily be modelled. Nevertheless, the fraction spilling into the standard event selection regions is very small, since these are offset in the cross dispersion direction and additionally placed at locations where their energy signal is outside the order selection regions.

• Readout noise. Although strictly speaking this is not a background component, the noise characteristics of each CCD has a tail which cannot be distinguished from proper X rays. It is roughly constant in dispersion and cross dispersion directions (with the whole detector, i.e. before any spectral extraction), and can be modelled by an exponential pulse height dependency. The typical detector noise count rate is 10$^{-3}$ counts s$^{-1}$ Å$^{-1}$, for the background in the extracted spectrum.

• Soft protons entering through the mirrors. It has been found to be consistent with distributions coming from two different components, one distribution originating from protons being reflected off the grating array that results in a beam centred on the zero-order position and the other distribution having a much larger angular spread consistent with a diffuse non-scattered component. While the widths of these two distributions appear to be fixed, the overall count rate and shape are clearly variable with time. Overall particle count rates variations by a factor 20 have been measured for similar length observations. The changes in the background shape are related, according to the models, to changes in the input particle energy spectrum and on the fraction of protons that are reflected by the gratings. It is found that this fraction is correlated with the overall count rate. Quiescent background periods seem to be dominated by the diffuse component. The count rate in quiet periods is 1-5x10$^{-3}$ counts s$^{-1}$ Å$^{-1}$. It should also be noted that background fluence is generally higher on the CCDs that are closer to the primary focus (CCDs 8 and 9).

The average background spectrum has been analysed for quiet-time periods. This has been performed using clear sky images (or “blank” images) and selecting low background periods (after inspection of the light curves). The average first and second order RGS1 and RGS2 background spectra are shown in Fig. 87. The total exposure time is about 170 ks. The spectra have been extracted in a standard way for the energy, and with the full field-of-view in the cross-dispersion direction. The background count rate in the quiet period corresponds to 1-5x10$^{-5}$ counts s$^{-1}$ per RGS resolution element for a point source.

The fraction of the total exposure time which corresponds to quiet background periods is unpredictable, as it is basically linked to the solar activity itself. During the first years of XMM-Newton operations there have been both “active” and “quiet” background epochs, during which about 50% and nearly 100% (respectively) of most exposures where taken with quiet background. A significant fraction of the data collected so far has much higher background than the one shown in Fig. 87. This should be taken into account for exposure time estimates, specially if detection of weak features is the main objective of the observations.

Figure 87: The average quiet background spectra from first (top) and second (bottom) order. RGS1 is shown in black and RGS2 in red. An enhancement of the count rate below 7 Å in each RGS is due to a change in the width of the pulse height filter at that wavelength. There is a bump around 32 Å in the RGS1 spectrum. The origin has not been fully understood but the most likely explanation is a somewhat higher dark current for CCD2 in this RGS. The lower background in the first order spectrum of RGS2 in this range seems to be related to the use of “single node readout” mode in this RGS.

Each of the background components described above has a very different impact on the RGS sensitivity. The first four are either rejected on board or have low intensity. The major constituents are therefore a tail on the CCD response due to readout noise, and the soft proton radiation.