4 Source classification and parameter determination in the hardness ratio plane

In Paper I the technique of X-ray spectral fitting has been applied to a sample of 26 background AGN in the field of the LMC. These AGN were taken from the ROSAT PSPC catalog of LMC X-ray sources of HP99 and candidate AGN selected here.

A different approach to constrain the spectral parameters is to use the hardness ratios $H\!R1$ and $H\!R2$. These hardness ratios are commonly available for ROSAT PSPC X-ray sources published e.g. in catalogs. There are two techniques to calculate these hardness ratios. The first is to fit the ROSAT PSPC point-spread function to data binned spatially in the standard energy bands and to determine the background from a spline-fit applied to these binned data. Such a procedure can be applied to a large sample of sources detected in the field of view of a PSPC observation. The second technique is to determine the source and the background counts in the standard energy bands from observational data which have been corrected for the ROSAT PSPC instrument functions. The background counts have been derived from a spatial region close to the source.

In order to constrain the spectral parameters of a source the observed hardness ratios $H\!R1$ and $H\!R2$ have been compared with the hardness ratios determined from simulations. Such a comparison was performed with a $H\!R1$ - $H\!R2$ grid derived from such simulations. Free parameters to be varied in the simulations were the powerlaw photon index $\Gamma$ of the adopted photon spectrum, the hydrogen absorbing column density $N_{\rm H}$, and the metallicity X. The redshift z has not been varied in the simulations. It has always been set to z=0.

Figure 2: Enlarged $H\!R1$ - $H\!R2$ grid calculated for background X-ray sources (AGN) and X-ray binaries, showing the effect of foreground absorption. The full lines give $\Gamma$ isolines, the dashed lines represent $N_{\rm H}$ isolines. The band for AGN type spectra is shown which is defined by $-\Gamma$ = (2.0:2.5). The $N_{\rm H}$ isolines have been labelled at the intersection with the $-\Gamma = 2.0$ line with $N_{\rm H}^{\rm LMC}$ in units of 1020 cm-2. The leftmost (long dashed) line is for zero LMC gas. The calculations include foreground galactic gas with $N_{\rm H}^{\rm gal} = 3\times10^{20}~{\rm cm}^{-2}$ (upper panel) and $5\times10^{20}~{\rm cm}^{-2}$ (lower panel), while for the LMC gas a mean metallicity of -0.3 dex is assumed.

In the case of the LMC background AGN, it is expected that the parameters can be confined to well defined ranges. For example in recent work it has been found that AGN have canonical powerlaw indices which can be confined to a narrow range $-\Gamma = 2.0$ to 2.5 for the ROSAT PSPC (cf. Brinkmann et al. 2000). These canonical values may still somewhat depend on the chosen energy range (instrument). Also the mean metallicity of the LMC gas is quite well constrained from observational work (e.g. Dopita & Russell 1992). These facts, in principle, allow the hydrogen absorbing column density towards an AGN from simulations to be determined. The redshift of the AGN has only a minor effect on the simulated spectra for the expected redshift range covered by the AGN sample (cf. Comastri et al. 1995). For the hydrogen column density $N_{\rm H}$ the model must account for the galactic contribution as well as the LMC contribution. These two components are assumed to have different metallicities and these models will be termed hybrid models. AGN can also show intrinsic absorption (cf. Comastri et al. 1995). But most of the intrinsically absorbed AGN will not be detected in the ROSAT band as the value of the absorbing column is large and the fluxes are low.

In Fig. 2 I give the $H\!R1$ - $H\!R2$ grid for a hybrid $N_{\rm H}$ model. For the first component with galactic metallicity a value of $N_{\rm H}^{\rm gal}$ of 3 and $5\times 10^{20}\ {\rm cm}^{-2}$ has been used, for the second component absorbing columns with mean LMC metallicities (X = -0.3 dex with respect to galactic interstellar absorption abundances, Morrison & McCammon 1983) which range from $10^{20}\ {\rm cm}^{-2}$ to $10^{22}\ {\rm cm}^{-2}$ in steps of $10^{20}\ {\rm cm}^{-2}$. In this classification scheme X-ray binaries cover $-\Gamma$ values of $\sim$(0.5-1.6) and AGN cover $-\Gamma$ values of $\sim$(1.8-3.0).

The range chosen for the powerlaw photon index $\Gamma$ is the range which is presently considered to be the most reliable in the ROSAT PSPC band and is considered to be the canonical band. This range of $\Gamma$ values agrees with the range of $\Gamma$ values required in Paper I to classify AGN from the result of X-ray spectral fitting.

I have derived the values for the hardness ratios and the errors in the hardness ratios making use of the same source and background regions as chosen for X-ray spectral fitting in Paper I. The spectral data have been binned in the standard energy bands used in the hardness ratio definition and the data have been corrected using the EXSAS correction package (Zimmermann et al. 1994).

In Table 1 the catalog of the reanalyzed spectrally hard X-ray sources as taken from the catalog of HP99 is given. One additional source, RX J0536.9-6913, is contained in the catalog which is not contained in the catalog of HP99 but which has been investigated in Paper I and found to be consistent with an absorbed AGN. In the table first the sources classified as X-ray binaries are given, then the AGN, and at the end of the table a few sources classified as SNR or foreground stars are given. For the last 11 sources in the table no classification is given. In Col. 1 of Table 1 the ROSAT name is given, in Cols. 2 and 3 the source index from the same catalog and the catalog of Sasaki et al. (2000), in Cols. 4 and 5 the count rate of the broad (0.1-2.4 keV) and hard (0.5-2.0 keV) band, in Cols. 6 and 7 the hardness ratios $H\!R1$ and $H\!R2$ including $1\sigma$ errors, in Cols. 8 and 9 the column density of the galactic and LMC H  I derived from 21-cm Parkes data (Brüns et al. 2001), in Col. 10 the LMC column density derived from the hardness ratio analysis from this work, in Col. 11 the source classification, and in Col. 12 references and notes to individual sources.

4.1 Source classification in the HR1-HR2 plane

Figure 3: Sources from Table 1 classified as X-ray binaries (XRB, upper two panels) and background AGN (lower two panels). First and third panel: sources with $\delta H\!R1 \le 0.2$ and $\delta H\!R2\ \le 0.2$ are given. Second and fourth panel: sources with $\delta H\!R1 \le 0.85$ and $\delta H\!R2\ \le 0.85$ are given. Also shown are the simulated (powerlaw slope -1.0, -2.0, -2.5, and -3.0) tracks for hydrogen columns in excess of galactic hydrogen columns ( $N_{\rm H}^{\rm LMC} > 0.0$) and assuming a value for the galactic hydrogen column density of $N_{\rm H}^{\rm gal} = 3\times 10^{20}\ {\rm cm^{-2}}$. For sources with $H\!R1 > 1.0$ and $H\!R2 > 1.0$ a value of 1.0 is given and the error bar towards the minimum allowed value.

I made simulations in which I varied the powerlaw photon index $\Gamma$ of the source spectrum and the LMC hydrogen column density $N_{\rm H}^{\rm LMC}$ assuming reduced metallicities which are expressed with the logarithmic decrement X. Different values have been assumed for the powerlaw photon index $\Gamma$ for X-ray binaries and AGN ( $-\Gamma = 0.5$ to 1.6 for X-ray binaries and $-\Gamma = 1.8$ to 3.0 for AGN).

Figure 4: Probability distribution of $H\!R1$ and $H\!R2$ values in the $H\!R1$ - $H\!R2$ plane. The 68%, 95.4% and 99.7% confidence contours are given for sources with $\delta H\!R1 \le 0.2$ and $\delta H\!R2\ \le 0.2$ from Table 1 classified as X-ray binaries (XRB, upper panel) and as AGN (lower panel). Also given are the tracks with constant powerlaw photon index $-\Gamma = 1.0,2.0,2.5$ and 3.0.

From the location of a source in the $H\!R1$ - $H\!R2$ plane a tentative source classification has been made. Sources which have hardness ratios which coincide with the range of $\Gamma$-tracks for X-ray binaries or AGN have been classified accordingly. In addition the galactic and LMC column at the location of a source has been used for a source classification. AGN are supposed to be seen through the galactic and total LMC absorbing column and X-ray binaries are seen through at least the galactic and at most the total LMC absorbing column. Of course there may be sources which have spectral properties which deviate from the standard values and the classification may not be unique. Especially the similarity between the spectral properties of the low-mass X-ray binary LMC X-2 and AGN is striking (LMC X-2 is located in the regime of AGN type spectra). A few of the sources which have not been classified as candidate X-ray binaries by HP99 and which show time variability in X-rays are located in the AGN regime and could also be time variable AGN. The source RX J0532.7-6926 (with number 914 in the catalog of HP99) has been classified as a LMXB by Haberl & Pietsch (1999b) from a time variability analysis. From X-ray spectral fitting follows that this source has a very steep spectrum ( $-\Gamma\sim3.0$) and could also be a time variable AGN. We did not include this source in the further analysis in the class of X-ray binaries.

In the color - color diagram ($H\!R1$ - $H\!R2$ plane), Fig. 3, I show the location of the sources classified as X-ray binaries and as background AGN as given in Table 1. It is obvious that X-ray binaries and background AGN cover in general different areas in this diagram as expected due to the different steepness of their spectral slopes. Background AGN have steeper slopes and are found in regimes of lower values for $H\!R2$ than X-ray binaries (cf. Fig. 4) although there is some region of overlap (some X-ray binaries have as steep X-ray spectra as AGN). Due to absorption by galactic gas with column densities in the range $\sim$ $(3{-}7)\times 10^{20}\ {\rm cm}^{-2}$ the value of $H\!R1$ does not extend to values $H\!R1\la0$.

Figure 5: Sources from Table 1 classified as background AGN or X-ray binaries and in the LMC $N_{\rm H}$ - $H\!R1$ and $H\!R2$ plane. Tracks for three different powerlaw photon indices $-\Gamma = 0.5$, 1.8 and 2.5 (upper to lower curves) are given. The sample of 49 AGN (filled circle) and X-ray binaries (open circle) is shown. Sources with accurate hardness ratio values $\delta H\!R1 \le 0.20$ and $\delta H\!R2\ \le 0.20$ have been used. The LMC $N_{\rm H}$ has been derived local to the X-ray source from 21-cm Parkes data and making use of the galactic foreground $N_{\rm H}$ derived from 21-cm Parkes data. Two AGN with $N_{\rm H}^{\rm LMC} < 10^{20}\ {\rm cm}^{-2}$ are not shown in the figure.

I can obtain information about the total hydrogen column density due to LMC gas from the source shape. Due to the dependence of the ROSAT PSPC instrument point-spread function on the energy (the point-spread-function becomes narrower with increasing energy) AGN seen through high LMC columns appear sharp and pointlike while AGN seen through low LMC columns have broader images. This fact helps to confirm large LMC columns derived in the direction of background AGN. But some classified AGN may be X-ray galaxies and intrinsically extended so this argument is not a perfect one.

A new source has been detected in the merged ROSAT PSPC pointings (see Table 2) which is not contained in the catalog of HP99.

A second source is given which is not included in the catalog of HP99, the heavily absorbed background source RX J0532.0-6919 in the 30 Dor complex which coincides in position with the radio source MDM 65 of Marx et al. (1997). In addition two sources are given for which significantly improved positions (compared to those given in HP99) were derived.

Column 1 of the catalog of Table 2 gives the catalog index, Col. 2 the ROSAT source name, Cols. 3 and 4 the source position, the right ascension (RA) and declination (Dec) for the epoch J2000 with the 90% confidence positional uncertainty (Col. 5), the likelihood of existence $L_{\rm exi} = -\ln(P)$ (Col. 6), with P the probability that the detected source is due to excess counts measured above a mean local background. For the first source, RX J0529.4-6713, in addition to the coordinates the values of the hardness ratios $H\!R1$ and $H\!R2$ are given. The source parameters have been determined by applying the maximum likelihood source detection task to the merged data in the field of the corresponding source.

 

 

Table 2: New X-ray sources detected in merged ROSAT PSPC observations and X-ray sources from the HP 99 catalog for which the positions have been improved.

(1)	(2)	(3)	(4)	(5)	(6)
Source	Source	RA	Dec	$P_{\rm e}$	$L_{\rm exi}$
No.	Name	(J2000)	(J2000)
	RX J	h m s	$^\circ$$\arcmin$ $\arcsec$	($\arcsec$)
1	0529.4-6713	05 29 25.8	-67 13 24	11	91
2	0529.7-6713	05 29 47.0	-67 13 50	11	10
3	0536.9-6913	05 36 57.9	-69 13 29	17	125
4	0553.2-7144	05 53 13.4	-71 44 03	25	11

Notes on sources: Source 1: $H\!R1 = -0.849\pm0.142$, $H\!R2 = 0.474\pm0.020$, close to source 2;
Source 2: HP 494, close to source 1; Source 3: 180 ksec exposure, MDM65, see also Table 1;
Source 4: HP 1303, 17$\arcmin$ off-axis.

In Fig. 5 I show the location of 49 X-ray sources from Table 1 with accurate values for the hardness ratios $\delta H\!R1 \le 0.20$ and $\delta H\!R2\ \le 0.20$ and which have been classified either as background AGN or as X-ray binaries in the LMC $N_{\rm H}$ - $H\!R1$ and $H\!R2$ plane respectively. The two AGN HP 37 and HP 352 which have $N_{\rm H}^{\rm LMC} < 10^{20}\ {\rm cm}^{-2}$ are not shown in this figure. The source HP 414 has not been included in the sample as it may be a foreground object and also the source HP 914 has not been included as it may either be a LMXB or an AGN. I also give tracks for powerlaw photon indices $-\Gamma = 0.5$, 1.8 and 2.5. Sources classified as XRB are preferrentially found in the $-\Gamma = 0.5$ to 1.8 band while sources classified as AGN are preferentially found in the $-\Gamma = 1.8$ to 2.5 band. There are a few exceptions, e.g. the XRB LMC X-4 has a steep powerlaw photon index and is outside the $-\Gamma = 0.5$ to 1.8 band. The source classification has been made using the LMC $N_{\rm H}$ - $H\!R2$ diagram (in the LMC $N_{\rm H}$ - $H\!R1$ diagram there is more scatter as the value for $H\!R1$ is less accurately determined than for $H\!R2$).