3 Results and discussion

3.1 Source position

Figure 2 shows the XMM-Newton error circle we have derived in the last section. It is entirely included in the intersection of the ROSAT error circle (Augusteijn et al. 1997) and the IPN error ellipse (Hurley et al. 2000). This strengthens the identification of the source observed by XMM-Newton with GRO J1744-28, despite the fact that this source is too weak to allow the search for pulsations (such pulsations were also impossible to study with the ROSAT observation). The association of the proposed optical/near-infrared counterpart (Augusteijn et al. 1997; Cole et al. 1997) with GRO J1744-28 is rejected without any ambiguity with the more precise XMM-Newton error circle we have obtained in the previous section. It is then important to carry out deeper searches for the real IR counterpart of GRO J1744-28 using the better determined XMM-Newton position.

Figure 2: An $\sim$ $1' \times 1'$ square region containing the approximate $3\sigma$ IPN error ellipse, the $\sim$10'' radius ROSAT error circle and the $3\sigma$ XMM error circle ($\sim$4'' radius). The two positions of the proposed optical/near-infrared counterpart (Augusteijn et al. 1997; Cole et al. 1997) are respectively marked with a star and a cross.

3.2 Source luminosity in quiescence

Despite a very low statistics, we tried in the last section to estimate the spectrum of GRO J1744-28 in quiescence. This spectrum appears to decrease very fast at high energy, as it is indicated by the photon index found when fitting with a power-law. The unabsorbed luminosity in the 1-5 keV band is

$$L \simeq \left( 1 {-} 6 \right) 10^{33}\ d_{10}^{-2}\ {\rm erg/s}$$ (1)

where d₁₀ is the distance of the source in unit of 10 kpc. The quiescent luminosity is then about five orders of magnitude smaller than the outburst luminosity

$$L \simeq 10^{38}\ d_{10}^{-2}\ {\rm erg/s}\$$ (2)

and is compatible with the typical quiescent luminosity of a neutron star transient (Asai et al. 1998). On the other hand, if we assume a blackbody distribution, the range in temperature we obtain (0.4-1. keV) is slightly outside the typical range of neutron star transients in quiescence ( $kT \sim 0.2{-}0.3$ keV; Asai et al. 1998). The temperature we find corresponds to an emitting area

$$A = \frac{L}{\sigma T^{4}} \simeq \left( 10^{9} {-} 10^{11} \right)\ d_{10}^{-2}\ {\rm cm^{2}} .$$ (3)

Such an area corresponds to a sphere of radius $10^{4}{-}10^{5}\ {\rm cm}$ which is well below the size of a neutron star. Then the observed luminosity cannot be associated with the thermal emission of the total surface of a non accreting neutron star. Many other origins can be proposed, such as the emission due to the accretion on the magnetic poles or the emission from the accretion onto the neutron star magnetosphere (Campana et al. 1998). However the large uncertainty we get on the spectral distribution of the X-ray photons makes very difficult any detailed comparison with a particular model.

Note added in proof: After submission of this paper, an independent work on GRO J1744-28 in quiescence based on Chandra observations was submitted to ApJ Letters by R. Wijnands and Q.D. Wang. Their results are broadly consistent with ours.

Acknowledgements

F.D. acknowledges financial support from a postdoctoral fellowship from the French Spatial Agency (CNES).