2 Observations and data analysis

GRO J1744-28 was observed with the European Photon Imaging Camera (EPIC), both with the MOS cameras (Turner et al. 2001) on 2001 April 4 at UT 13:14:27 and with the PN camera (Strüder et al. 2001) on the same day starting at UT 14:27:23. The MOS and PN cameras were operating in full window mode with the medium filter. Data reduction was performed using the XMM SAS (Science Analysis Software) and the standard XSPEC packages.

The time histogram of all events above 10 keV showed that the background has only small variations, without any very intense flare, even if the averaged flux is close to the alert level as defined in the standard SAS flare rejection procedure. For this reason we applied no time filtering so that the total duration has been kept to $9199\ {\rm s}$ for the MOS 1&2 cameras and $4699\ {\rm s}$ for the PN camera. A preliminary visual inspection of the MOS 1,2 and PN images showed that a faint source was present at the position of GRO J1744-28 (with no significant flux below 1 keV and above 6 keV). The absence of flux below 1 keV can be explained by the strong absorption towards the Galactic center. Therefore all the further analysis has been made in the 1-6 keV band. Also, it could be noticed that for some unidentified reasons, the background in the MOS2 image appears to be higher than in the MOS1 image.

In this energy band, we generated an image for each of the three EPIC instruments, keeping only the valid X-ray events using the canonical range of pattern specified by the SAS (0-12 for the MOS cameras and 0-4 for the PN camera). We applied to these three images the standard source detection procedure with the following SAS tasks: (i) eexpmap and emask to generate the exposure map and the detection mask for each image; (ii) eboxdetect (in local mode) to generate an input list of source positions; (iii) esplinemap to generate the corresponding background map; (iv) eboxdetect (in map mode) to generate a new list of sources with a better detection sensitivity using the background map; (v) emldetect to perform a maximum likelihood PSF fit to the count distribution of each source. This procedure was also applied to an image made of the sum of all X-ray events of the three instruments in the 1-6 keV band.

 

 

Table 2: Position of GRO J1744-28 as detected by each of the EPIC instruments. The right ascension (RA), declination (DEC), statistical error at $1 \sigma$ level and counts number are the result of the SAS standard source detection procedure. The last line (EPIC) corresponds to the image made of the sum of the PN, MOS1 and MOS2 images.

Inst.	RA	Dec	stat.	counts
			error
PN	$17^{\rm h}44^{\rm m}33\hbox{$.\!\!^{\rm s}$ }1$	$-28^{\circ}44'26''$	$\sim$2.4''	$22.4 \pm 6.1$
MOS1	$17^{\rm h}44^{\rm m}33\hbox{$.\!\!^{\rm s}$ }2$	$-28^{\circ}44'27''$	$\sim$1.9''	$27.4 \pm 6.0$
MOS2	$17^{\rm h}44^{\rm m}33\hbox{$.\!\!^{\rm s}$ }1$	$-28^{\circ}44'16''$	$\sim$6.0''	$18.8 \pm 7.4$
EPIC	$17^{\rm h}44^{\rm m}33\hbox{$.\!\!^{\rm s}$ }2$	$-28^{\circ}44'26''$	$\sim$1.4''	$59.2 \pm 10.0$

The standard SAS source detection procedure that we applied to the data found one weak source within the ROSAT $\sim$10'' error circle of GRO J1744-28. This source was also present in the standard SAS pipeline output. As can be seen in Table 2, it is a detection at the $\sim$4$\sigma$ level for the PN and MOS1 cameras and a detection at the $2.5\sigma$ level only for the MOS2 camera. The number of sources at the $3\sigma$ level in the total field of view of the MOS cameras (radius 15') is about 100 so that the probability to find such a source by chance in an area comparable with the ROSAT error box (radius $\sim$10'') is about 1%. Then we identify this weak source with GRO J1744-28.

The position error obtained with the standard SAS source detection procedure can be dominated by the errors of the absolute astrometry (maximum $\sim$4''). To estimate these systematic errors, we proceeded to identify strong sources in our field of view. Unfortunately only one is strong enough to be used in this way. This bright source which was detected by each instrument has been identified - based on positional coincidence - as the star TYC 6840-38-1 in the Tycho-2 Catalogue (Hog et al. 2000) which contains the position, the proper motion and the two-colour photometric data for the 2.5 million brightest stars in the sky. Its catalogue position is RA $17^{\rm h}43^{\rm m}51\hbox{$.\!\!^{\rm s}$ }3$ and Dec $-28^{\circ }46'38''$ (with a negligible proper motion for our purpose). Table 1 shows the position of this star obtained with each instrument. They are all consistent with the catalogue position, taking into account the statistical error given by the emldetect procedure ($\sim$0.7'') and the quoted accuracy of the XMM-Newton coordinates: the MOS1 and MOS2 positions are distant of $\sim$1'' from each other whereas the PN position is shifted of $\sim$3'' with respect to the MOS2 position. This is in agreement with the negligible offset found in the MOS 1&2 roll angle and the mean offset of $\sim$ $-0.30^{\circ}$ found in the PN roll angle (Tedds & Watson 2001). As the star is located at $\sim$10' of the center of the field of view, the expected shift in the PN position is indeed $\sim$3''. However the same shift at the position of GRO J1744-28 is negligible because the source is at the center of the field of view. The nice agreement bewteen the position of the star obtained with the three EPIC instruments and the catalogue position shows that for our observation, the systematic error in the position is $\sim$1''.

The position of GRO J1744-28 is the same in the three instruments taking into account the statistical error (see Table 2). The best accuracy is obtained when summing the events of the three instruments. The derived position is RA $17^{\rm h}44^{\rm m}33\hbox{$.\!\!^{\rm s}$ }2$ and Dec $-28^{\circ}44'25''$ with an error circle of about 4'' at the $3\sigma$ level (summing in quadrature a statistical error of $\sim$4'' and a systematic error $\sim$1''). The closest other source is located at $\sim$25'' of GRO J1744-28, well outside the ROSAT error circle, and therefore can be easily rejected.

To construct the spectral distribution of the photons, we have decided to sum the two MOS cameras events to increase the statistics, which is possible because the response matrix are similar. We extracted from the PN and MOS 1+2 images the X-ray events within a radius of 15'' of the GRO J1744-28 position. We grouped the photons in 5 energy bands: 1-2 keV, 2-3 keV, 3-4 keV, 4-5 keV and 5-6 keV. Background was estimated using offset regions in the same central CCD. After substracting it, we obtained a detectable flux only in the 1-5 keV band. The resulting spectrum was obtained with XSPEC and is plotted in Fig. 1. The source count rate in the 1-6 keV energy band is $\left(4.75 \pm 1.32\right)\ 10^{-3}\ {\rm counts/s}$ for the PN camera and $\left(3.40\pm 0.76\right)\ 10^{-4}\ {\rm counts/s}$ for the MOS 1+2. Fixing the hydrogen column density to the value of $5.1\times 10^{22}\ {\rm cm^{-2}}$ (Dotani et al. 1996a), we tried to fit the spectrum of GRO J1744-28 either with a power-law or with a blackbody distribution. We found that acceptable fits of the observed spectrum with a power-law could be obtained for a photon index in the range $\sim$2-5. Acceptable fits with a blackbody distribution could be obtained as well for a temperature kT in the range $\sim$0.4-1. keV. The corresponding unabsorbed energy flux in the $1{-}5\ {\rm keV}$ is about $2{-}5\times 10^{-13}\ {\rm erg\,cm^{-2}\,s^{-1}}$ in the power-law case and about $1{-}2\times 10^{-13}\ {\rm erg\,cm^{-2}\,s^{-1}}$ in the blackbody case.

Figure 1: Spectrum of GRO J1744-28: the flux is plotted as a function of energy for the PN camera (top) and the sum of the MOS1 and MOS2 cameras (bottom). The solid line corresponds to the fit of the data by an absorbed blackbody spectrum with a column density $5.1\times 10^{-22}\ {\rm cm^{-2}}$ and a temperature $kT = 0.6\ {\rm keV}$.