XMM-Newton Users Handbook

3.3.8 EPIC's sensitivity limits

The EPIC sensitivity limits depend on the sky area, i.e. the true X-ray background and on the `space weather' as described in §§ 3.3.7.1 and 3.3.7.2.

As the sensitivity limits also depend on the angular structure and the spectral characteristics of the source that is observed, it is strongly recommended to use SciSim to get a feeling on the signal to noise which can be achieved with a certain instrument setup and exposure time.

Currently the best statistical results on the EPIC sensitivity limits are based on the Lockman Hole data (Hasinger et al., 2001, A&A 365, L45) and the simulations performed by Watson et al., 2001, A&A 365, L51.

The major uncertainty in predicting the EPIC sensitivity pre-launch was the background levels which would be encountered in orbit. The actual in orbit background levels measured in the first part of the Lockman Hole observation (representative of quiescent background) are $\sim (2.1, 2.9) 10^{-3}$ cts s$^{-1}$ arcmin$^{-2}$ for the pn camera in the soft (0.5 - 2 keV) and hard (2 - 10 keV) bands respectively. Very similar values: $\sim (2.0, 2.6) 10^{-3}$ cts s$^{-1}$ arcmin$^{-2}$ are found for the two MOS cameras combined.

An estimate of the to be expected EPIC background in low background periods can also be derived from blank sky background event files: For each of the blank sky files (based on the instrument-filter-mode classification) count rates have been derived in the standard SSC/PPS (see § B.1) energy bands. These count rates are available from the blank sky count rates page.

Source detections on the Lockman hole data were accepted with likelihood values above 10 (about 4$\sigma$) and inside an off-axis angle of 10 arcmin. The resulting detection statistics are given in Tab. 4 (for further details see Hasinger et al. (2001)).

Table 4: Detection limits (about 4$\sigma$) for different energy bands, based on 100 ksec of data from the Lockman hole (Hasinger et al. (2001), Tab. 2)

Band$^a$	$\Gamma^b$	$ECF^c$	$S_{lim}~^d$
0.2 - 0.5	$2.0\pm0.5$	$7.16\pm1.01$	4.0
0.5 - 2.0	$2.0\pm0.5$	$10.20\pm0.04$	3.1
2 - 10	$2.0\pm0.5$	$1.79\pm0.40$	14
5 - 10	$1.6\pm0.5$	$1.28\pm0.11$	24

Notes to Table 4:
$^a$ Energy band in keV in which the flux is given.
$^b$ Assumed range in photon index.
$^c$ Energy conversion factor in cts s$^{-1}$ per $10^{-11}$ erg cm$^{-2}$ s$^{-1}$.
$^d$ Minimum detected flux in $10^{-16}$ erg cm$^{-2}$ s$^{-1}$.

Watson et al. (2001) used the nominal quiescent background values together with the measured XMM-Newton PSF to compute an EPIC point source sensitivity based on a simple 5$\sigma$ source detection criterion against assumed purely Poissonian background fluctuations, as shown in Fig. 37 ⁴.

Figure 37: EPIC sensitivity (5$\sigma$ minimum detectable flux in erg cm$^{-2}$ s$^{-1}$ in respective bands) as a function of exposure time (from Watson et al., 2001). Sensitivity is computed for an assumed $\alpha = 1.7$ powerlaw spectrum with a column density $N_H = 3\,10^{20}$ cm$^{-2}$. Solid curves are for the nominal background rates. Dashed curves are for background levels enhanced by a factor 3. The EPIC MOS curves correspond to the combination of the two cameras.

Empirical data from analysis of several XMM-Newton fields using the source detection software in the SAS are broadly consistent with these plots. The actual background in an observation depends critically on the fraction of background flares removed, i.e. the trade­off between net background levels and net exposure time. An investigation of a few example fields demonstrates that the effective sensitivity of typical observations is within a factor 2 of the values plotted in Fig. 37. A few observations are affected by enhanced background throughout; here the average background can be several times higher than the nominal values even after the removal of the largest flares.

At very faint fluxes the effective sensitivity is limited by confusion effects. Although a detailed study of source confusion has not yet been carried out, the long XMM-Newton observations of the Lockman Hole (Hasinger et al., 2001) demonstrate that source confusion is not a significant problem in either the soft (0.5-2 keV) or hard (2-10 keV) X-ray bands for an observation duration of about 100 ksec which reaches flux limits $f_X\sim3.1$ and $\sim14\,10^{-16}$ erg cm$^{-2}$ s$^{-1}$ in the soft and hard bands respectively. Recently, Carrera 2007, estimated that the confusion limit in the hard X-ray band (2-10keV) is only reached after 2Ms of observing time and is unreachable in the 5-10 keV band in the foreseeable future.