7 Background subtraction techniques

  In order to study the X-ray emission from a particular source it is usually necessary to subtract the X-ray and NXB contributions from within the extraction region. In the case of the LECS (and MECS) it is important to realize that strong sources outside the FOV (and within 2$^\circ$ from the instrument axis) can contribute to the background. A search should be made for the presence of such sources using e.g., the HEASARC's XCOLL catalog.

In the case of the LECS, we present 3 ways of subtracting the background. An estimate of the systematic uncertainty associated with background subtraction may be obtained by comparing results obtained using the different methods.

7.1 Standard background

This is the simplest approach since the standard background includes the contribution from the high (>$\vert 25 \vert ^\circ$) galactic latitude CXB and the NXB. Since the standard background has a longer accumulation time (558.6 ks) than individual BeppoSAX observations, this method does not usually add significant noise. Since the standard background varies with position within the FOV (Fig. 3), the background spectrum should be accumulated using the same extraction region as the source. Note that this technique will underestimate the low-energy background for pointings in directions with bright Galactic emission such as the North Polar Spur (see Sect. 4).

7.2 Background semi-annuli

This technique has the following steps:

1.

Accumulate a spectrum of the two background semi-annuli during the observation of interest. These can be selected in RAWX, RAWY using the data extraction program XSELECT as follows:

-CIRCLE(128.0,128.0,80.0)
-CIRCLE(128.0,128.0,62.85)
-BOX(64.0,192.0,128.0,128.0)
-BOX(192.0,64.0,128.0,128.0).

2.

For source count rates $\mathrel{\hbox{\rlap{\lower.55ex \hbox {$\sim$}} \kern-.3em \raise.4ex \hbox{$\gt$}}}$1 s^-1, subtract the NXB background in the standard extraction region from the total counts in the same region. Then multiply the subtracted counts by the function shown in Fig. 8 to derive the "spill-over'' contribution. This is then subtracted from the extracted semi-annuli spectrum.

3.

Subtract the standard NXB spectrum of the semi-annuli.

4.

Multiply the remaining spectrum by the correction factors shown in Fig. 9. This provides an estimate of the CXB in the source region.

5.

Add the standard NXB spectrum in the source region to this spectrum. This is then the final background to be subtracted before scientific analysis of the source properties.

7.3 Scaled ROSAT PSPC all-sky survey background

ROSAT PSPC all-sky survey diffuse background maps are available in 7 energy bands with a pixel size of $12\hbox{$^\prime$}\times 12$$^\prime$ covering $\sim$98% of the sky. The effects of discrete source counts, non X-ray contamination and X-rays of solar system origin have been eliminated to the greatest possible extent. The contributions of all the sources listed in Table 4 have been excluded. These maps can be used to predict the variation with pointing position of the LECS background. The PSPC count rates closest to the centers of each of the fields included in the LECS standard background were first determined. The mean, LECS exposure weighted, PSPC 0.1-2.5 keV count rate is $1.2 \ 10^{-3}$ s^-1 arcmin^-2.

In order to understand how best to model the PSPC data from different sky positions, 7 channel PSPC spectra were determined for the fields given in Tables 2 and 4. These were then fit with a number of different models including power-laws, thermal bremsstrahlungs, broken power-laws with the hard component $\alpha$ fixed at 1.4 (the value expected above 1 keV), and power-laws with $\alpha = 1.4$,together with thermal components. In all cases low-energy absorption was included. It was found that (1) the power-law and (2) the fixed power-law together with a thermal component give consistently the best fits for almost all the fields. These two models are therefore used in the background analysis except for the Gal cent-3 field, where a broken power-law, with hard index $\alpha$ fixed at 1.4 is preferred.

In order to use this technique to predict the LECS background spectrum at a particular pointing position, the following steps are necessary:

1.

Determine the ROSAT PSPC count rates in the 7 energy bands at the required position (see Sect. 9).

2.

Fit spectral models to the derived spectrum using the appropriate ROSAT all-sky survey response matrix.

3.

Select the model with the lowest reduced $\chi ^2$.

4.

Use the standard LECS response matrix and the parameters of the selected model fit to calculate a predicted background spectrum for the position of interest.

5.

Use the standard LECS response matrix and the CXB fit parameters given below to calculate a predicted spectrum for the standard background field.

6.

Divide the predicted LECS spectrum by that predicted for the standard background to derive a set of PI channel dependent correction factors.

7.

Subtract the standard NXB spectrum (see Sect. 5) measured in the source extraction region from the observed standard background spectrum.

8.

Multiply the subtracted spectrum by the correction factors calculated above.

9.

Add the standard NXB measured in the source extraction region from the standard background spectrum to the scaled spectrum. This is then the final background to be subtracted before scientific analysis of the source properties.

The LECS CXB spectrum is discussed in a companion paper (Parmar et al 1999). In the 0.1-7.0 keV energy range it is well modeled by an absorbed power-law together with two solar abundance thermal bremsstrahlung components (the MEKAL model in XSPEC). When this model is fit to the exposure weighted PSPC spectrum corresponding to the standard background, the best-fit results are a power-law $\alpha$ of 1.48 with normalization of $1.07 \ 10^{-3}$ photon cm^-2 s^-1 keV^-1, a hard bremsstrahlung component temperature (kT) of 0.80 keV with normalization of $1.5 \ 10^{-5}$ photon cm^-5, and a soft bremsstrahlung component kT of 0.159 keV with normalization of $1.9 \ 10^{-4}$ photon cm^-5. For both absorbed models, ${N_{\rm H}}$ equivalent to $3.7 \ 10^{20}$ atom cm^-2 is required. These fit parameters should be used to generate the predicted standard LECS background spectrum.