A&A 504, 185–197 (2009) Astronomy

DOI: 10.1051/0004-6361/200912180 &


c ESO 2009 Astrophysics

A search for thermally emitting isolated neutron stars

in the 2XMMp catalogue
A. M. Pires1,2 , C. Motch2 , and E. Janot-Pacheco1 ,

1
Instituto de Astronomia, Geofísica e Ciências Atmosféricas, Universidade de São Paulo, R. do Matão 1226, 05508-090 São Paulo,
Brazil
e-mail: [email protected]
2
CNRS, Université de Strasbourg, Observatoire Astronomique, 11 rue de l’Université, 67000 Strasbourg, France
Received 25 March 2009 / Accepted 4 June 2009

ABSTRACT

The relatively large number of nearby radio-quiet and thermally emitting isolated neutron stars (INSs) discovered in the ROSAT All-
Sky Survey, dubbed the “Magnificent Seven”, suggests that they belong to a formerly neglected major component of the overall INS
population. So far, attempts to discover similar INSs beyond the solar vicinity failed to confirm any reliable candidate. The good
positional accuracy and soft X-ray sensitivity of the EPIC cameras onboard the XMM-Newton satellite allow us to efficiently search
for new thermally emitting INSs. We used the 2XMMp catalogue to select sources with no catalogued candidate counterparts and
with X-ray spectra similar to those of the Magnificent Seven, but seen at greater distances and thus undergoing higher interstellar
absorptions. Identifications in more than 170 astronomical catalogues and visual screening allowed us to select fewer than 30 good
INS candidates. In order to rule out alternative identifications, we obtained deep ESO-VLT and SOAR optical imaging for the X-ray
brightest candidates. We report here on the optical follow-up results of our search and discuss the possible nature of 8 of our candidates.
A high X-ray-to-optical flux ratio together with a stable flux and soft X-ray spectrum make the brightest source of our sample, 2XMM
J104608.7-594306, a newly discovered thermally emitting INS. The X-ray source 2XMM J010642.3+005032 has no evident optical
counterpart and should be further investigated. The remaining X-ray sources are most probably identified with cataclysmic variables
and active galactic nuclei, as inferred from the colours and flux ratios of their likely optical counterparts. Beyond the finding of
new thermally emitting INSs, our study aims at constraining the space density of this Galactic population at great distances and at
determining whether their apparently high density is a local anomaly or not.
Key words. stars: neutron – X-rays: individuals: 2XMM J104608.7-594306 – catalogs

1. Introduction magnitudes of mB ∼ 25–28 were detected for several sources,

implying high logarithmic X-ray-to-optical flux ratios of ∼ 4–5.
Around ten years ago, seven X-ray bright thermally emitting and Relative to radio pulsars, the neutron star spin periods are longer
radio-quiet isolated neutron stars (INSs) sharing similar proper- and distributed in a much narrower range, P ∼ 3–10 s. Six of
ties were identified in the ROSAT All-Sky Survey data (Voges the sources show sinusoidal X-ray pulsations with pulsed frac-
et al. 1999). This group is now commonly referred to as the tions between ∼1% and 18%. The M 7 are believed to be nearby
“Magnificent Seven” (or M 7, for simplicity; see Haberl 2007, (d <
for a review) for the reason that they clearly stand apart from the ∼ 500 pc), as inferred from the distribution of the interstel-
lar medium in the line-of-sight and the equivalent hydrogen col-
population of rotation-powered radio pulsars. In spite of many umn densities measured in their X-ray spectra (Posselt et al.
searches for similar objects in the ROSAT data (e.g. Rutledge 2007). Furthermore, HST parallaxes are available for the two
et al. 2003; Chieregato et al. 2005; Agüeros et al. 2006), no X-ray brightest members (Kaplan et al. 2002, 2007) with re-
new candidate has been identified since the discovery of the last sults largely consistent with those estimated by Posselt et al.
member, RX J2143.0+0654 (Zampieri et al. 2001). (2007). The seven sources do not show either persistent or tran-
A soft blackbody spectrum undergoing low interstellar ab- sient radio emission to a rather sensitive limiting flux of ∼10 μJy
sorption (kT ∼ 40–100 eV and NH ∼ few 1020 cm−2 ) is common (Kondratiev et al. 2008) and are not associated with supernova
to the seven sources, as well as the absence of a non-thermal remnants. However, unconfirmed claims of detection at long
component extending towards higher energies. We note, how- wavelengths exist (Malofeev et al. 2007).
ever, the detection in the optical/UV of a possible non-thermal
power-law component in the spectrum of RX J0720.4-3125 Proper motion measurements in the optical (for the three
(Kaplan et al. 2003). Very faint optical counterparts with blue INSs with bright enough counterparts, Neuhäuser 2001; Kaplan
 et al. 2002; Motch et al. 2003; Kaplan et al. 2007; Motch et al.
Based on the public data archive of XMM-Newton an ESA sci-
ence mission with instruments and contributions directly funded by
2005; Zane et al. 2006) and in X-rays (for RX J1308.6+2127,
the ESA Member States and the USA (NASA) and a Chandra Legacy Motch et al. 2008, 2009) have established these objects as cool-
programme. Optical observations were performed at the European ing, middle-aged (∼105–106 yr) neutron stars, probably originat-
Southern Observatory, Paranal, Chile, under programme ID 079.D- ing in the nearby OB associations of the Gould belt. Intense mag-
0633(A), and at the Southern Astrophysical Research Telescope, Cerro netic fields B ∼ 1013 –1014 G (see Haberl 2007; van Kerkwijk
Pachón, Chile. & Kaplan 2007, for reviews) are inferred from the presence of

Article published by EDP Sciences

186 A. M. Pires et al.: A search for INSs in the 2XMMp catalogue

one or more broad absorption lines in the spectra of the sources population properties and evolutionary links with other groups
and from the spin-down rates detected in three cases by tim- of Galactic INSs.
ing studies (Kaplan & van Kerkwijk 2005a,b; van Kerkwijk & Despite the relatively small field-of-view and low sky cover-
Kaplan 2008), where magnetic dipole braking is assumed. For age of the XMM-Newton Observatory, its large effective area and
these cases, characteristic ages τch = P/2Ṗ ∼ 1.5–4 Myr over- good positional accuracy at soft X-ray energies make it ideal to
predict by a factor of ∼2–10 the kinematic travel times inferred look for faint INS candidates beyond the Gould Belt. In Pires
by tracing back the trajectories from their likely birth places (e.g. & Motch (2008) we reported on the preliminary results of a
Motch et al. 2008). The spectral absorption lines are commonly programme aimed at identifying new thermally emitting INSs
interpreted as due to atomic transitions in a partially ionised hy- in the 2XMMp catalogue. In the present paper we discuss the
drogen atmosphere (e.g. Lai 2001) or as proton cyclotron fea- nature of a handful of the X-ray brightest of our INS candi-
tures (e.g. Zane et al. 2001). In spite of the high B, their low dates, selected from more than 7.2 × 104 EPIC pn sources with
magnetospheric activity allows for a direct view of the surface count rates above 0.01 s−1 . The paper is structured as follows:
of the neutron star. Atmosphere models and the surface emissiv- in Sect. 2 we describe the methodology which was applied to
ity under extreme physical conditions can thus be tested in order select the 2XMMp sources and restrict the number of INS can-
to derive radii (e.g. Ho et al. 2007) and eventually constrain the didates; in Sect. 3 the optical data obtained on a subsample of our
equation of state of the neutron star interior, provided that dis- candidates are described together with their analysis and results;
tances are known. Unfortunately, the current lack of understand- Sect. 4 is devoted to the analysis of the X-ray emission of these
ing of the surface composition, magnetic field and temperature candidates. Finally, the discussion, conclusions and summary are
distributions have limited any definite conclusion. given in Sects. 5 and 6. The discovery of the new thermally emit-
ting INS 2XMM J104608.7-594306, the X-ray brightest among
The locus occupied by the M 7 in the P – Ṗ diagram, some-
our sample, and the discussion of its nature as, in particular, a
what intermediate between radio pulsars and magnetars, sug-
cooling or an accreting neutron star have been presented in Pires
gests that they could be linked to other populations of INSs. In
et al. (2009).
particular, the similar spin periods and intense magnetic fields
raised the possibility that some of the M 7 could have evolved
from the younger and more energetic magnetar objects (Heyl
& Kulkarni 1998). Magnetic field decay provides an additional 2. Selection of candidates
source of heating of the neutron star crust and dramatically We searched for new thermally emitting INSs similar to the
changes the cooling history of neutron stars with B > 13
∼ 10 G. M 7 in the pre-release of the XMM-Newton serendipitous source
Aguilera et al. (2008) have shown that the range of observed catalogue, 2XMMp1 . This version of the catalogue, released in
temperatures of the M 7 is consistent with them being born as July 2006, contains source detections drawn from 2400 individ-
magnetars with B ∼ 1014 –1015 G, provided that their current ual XMM-Newton EPIC observations made between 2000 and
ages are around 106 yr (which is in agreement with the kine- 2006. It contains more than 120 000 unique X-ray sources with
matic ages). On the other hand, if the neutron star magnetic field a median flux of 2.4 × 10−14 erg s−1 cm−2 (in the total energy
has not changed substantially over its lifetime, an evolutionary band 0.2–12 keV); around 20% of the catalogued sources have
link with the high B radio pulsars (HBPSRs) can be suggested fluxes below 1 × 10−14 erg s−1 cm−2 . Detection positional accu-
(e.g. Zane et al. 2002). Relatively to the M 7, these objects are racy is generally <2 (68% confidence radius) and the total
generally located at much greater distances and show consider- sky area covered is ∼285 deg2 . Despite the small percentage of
ably higher spin-down energy (Ė). However, if the HBPSRs are sky coverage (0.7%), the EPIC instruments provide much better
a factor of ∼100 younger, then the difference in Ė can be ex- positional accuracy at soft energies and faint fluxes relative to
plained within the standard scenario for pulsar evolution assum- ROSAT, making the XMM-Newton catalogue a powerful tool to
ing the usual magnetic dipole braking model (Kaplan 2008). The look for unidentified INSs. In the near future, the ROSAT succes-
M 7 could then be long-period radio pulsars for which the narrow sor eROSITA2 mission should perform the first imaging all-sky
emission beam simply does not sweep over the earth. Finally, the survey up to 10 keV with unprecedented spectral and angular
only rotating radio transient (RRAT) discovered so far at X-ray resolution, and can potentially increase the number of cooling
energies (McLaughlin et al. 2007) also shares strikingly similar INSs up to one order of magnitude (Posselt et al. 2008).
properties with the M 7 – namely the position on the P – Ṗ dia-
In order to select candidates, we only considered detec-
gram and the presence of a broad absorption line in its thermal
tions in the EPIC pn camera at a limiting count rate of 0.01 s−1
spectrum. The source is however hotter (kT ∼ 140 eV), fainter
(0.2–12 keV), corresponding to a total of ∼7.2 × 104 X-ray
and expected to be more remote, at ∼3.6 kpc.
sources, and then required these to be well detected and point-
Considering that within 1 kpc the M 7 appear in comparable like. Translating into the more precise catalogue parameters,
numbers to young (< ∼few Myr) radio and γ-ray pulsars (Popov we selected sources with a maximum likelihood of detection
et al. 2003), they may represent the only identified members of a PN_DET_ML > 8 and required a low probability for extension
large, yet undetected, elusive population of radio-quiet and ther- (EP_EXTENT_ML < 4), both properties computed by the cata-
mally emitting INSs in the Galaxy. It should be noted, however, logue pipeline with the SAS3 task emldetect. None of the source
that the proximity of OB associations of the Gould Belt to the quality or variability flags provided by the catalogue was used in
solar vicinity is an important factor that might explain this ap- the selection and screening process.
parent local overdensity of INSs with similar temperatures, ages For the remaining ∼4.6 × 104 entries, we then required as
and magnetic field intensities. Since the neutron star cooling is main selection criteria (i) no correlations (within 3 σ) with the
strongly dependent on mass and, to a lesser extent, on magnetic
field, a scenario where these stars evolved from a common pro- 1
http://xmmssc-www.star.le.ac.uk/Catalogue/
genitor population of massive stars could be considered. The xcat_public_2XMMp.html
2
discovery of similar sources at greater distance is then manda- http://www.mpe.mpg.de/projects.html#erosita
3
tory in order to make any progress towards understanding their http://xmm.esac.esa.int/sas
 A. M. Pires et al.: A search for INSs in the 2XMMp catalogue 187

1.0 1.0
I: kT ≤ 100 eV, NH ≤ 1 × 1021 cm−2
II: kT ≤ 120 eV, NH = (1 − 5) × 1021 cm−2
III: kT = 120 − 200 eV, NH = (1 − 5) × 1021 cm−2
0.5 IV: kT ≤ 200 eV, NH = (5 − 10) × 1021 cm−2 0.5
HR2

HR2
0.0 0.0

IV
−0.5 −0.5
III

II
−1.0
I
−1.0 *
* *** **
−1.0 −0.5 0.0 0.5 1.0 −1.0 −0.5 0.0 0.5 1.0

HR1 HR1

Fig. 1. Left: boundaries of the 4 different selection regions considered in the HR1 × HR2 diagram. Right: positions of the final list of INS candidates
in the same diagram (filled circles with error bars). The known ROSAT-discovered INSs (stars) occupy the less absorbed and cooler portion of the
diagram. Dotted lines denote soft absorbed blackbodies of different temperatures (50 eV, 80 eV, 100 eV, 120 eV, 150 eV and 200 eV, from bottom
to top) with hydrogen column densities in the range 1019 to 1022 cm−2 . Contours show the hardness ratio distribution of quasars from SDSS DR3
(see text). The candidates selected for follow-up optical observations are highlighted by open circles.

USNO-A2.0 optical catalogue and (ii) a soft energy distribu- pipeline (Watson et al. 2009) provides lists of archival entries in
tion. Applying the first criterion led to the rejection of more than over 170 archival catalogues having a position consistent within
2.1 × 104 X-ray sources having a positive match with USNO-A2 3 σ with that of the 2XMMp source.
entries. For requirement (ii), instrumental response and auxiliary So as to more easily classify the 561 remaining sources
files were used to simulate the count rates and hardness ratios which fulfilled the above selection criteria, we considered 4 dif-
(HR) of absorbed blackbodies (in a grid of temperature and col- ferent regions in the HR1 × HR2 diagram, based on temperature
umn density), taking into account the different optical blocking and column density (see left panel of Fig. 1). We note that,
filters. Although the effective area of the three (thin, medium and although this diagram represents our main means of selecting
thick) filters at energies below 1 keV can differ by a factor of up INS candidates, we required soft emission in the HR3 × HR4
to 3, we verified that their influence in the HR diagrams is minor. diagram as well so as to avoid X-ray sources with significant
The HR is defined as emission above 2 keV. For each source in these regions, we vi-
Ci+1 − Ci sually checked the X-ray and optical images4 and searched for
HRi = i = 1, . . . , 4 (1) possible identifications in astronomical catalogues and databases
Ci+1 + Ci like NED and Simbad. Source selection and screening pro-
where Ci and Ci+1 are source counts in two contiguous en- cesses used the facilities provided by the XCat-DB (Motch et al.
ergy bands of the five pre-defined of the 2XMMp catalogue, 2007), a database hosting the 2XMM catalogue and its associ-
0.2–0.5 keV, 0.5–1 keV, 1–2 keV, 2–4.5 keV and 4.5–12 keV. We ated pipeline products, including the archival cross-correlations.
thus selected the catalogue sources with an X-ray energy dis- Many of the sources turned out to be false detections in ex-
tribution compatible (within 1σ) with that of template black- tended diffuse emission mostly of supernova remnants (SNRs)
bodies of kT ≤ 200 eV undergoing absorptions in the range or due to out-of-time events. The high number of spurious de-
NH = 1019 –1022 cm−2 , by defining their positions in HR dia- tections among our selected sources reflects the fact that we did
grams. not use the source quality flag information in the early steps of
The X-ray column densities of the catalogue sources were the selection process. X-ray sources located in the direction of
then verified to be less than or equal to the total Galactic value nearby galaxies (sometimes with an optical or infrared coun-
in order to discard intrinsically absorbed objects. Sources with terpart visible in the finding charts) were classified as “extra-
very large positional uncertainties (r90 > 4 , where r90 is the galactic” and thus were no longer regarded as potential INS can-
source 90% confidence level error circle) or which were too far didates. Finally, because the spectral properties of some faint
from the optical axis (at off-axis angles θ > 11 ) were also X-ray sources are not well determined in short exposures, the
discarded. After applying these selections, we ended up with last step in our procedure was to check if any source with no
∼1000 sources. Of these, a total additional number of 461 X-ray evident classification (and thus a potential INS candidate) was
sources with matches in one or more of the optical/IR catalogues spectrally harder in other (longer) XMM-Newton exposures.
USNO-B1.0, SDSS, GSC-2, APM and 2MASS was discarded as In Table 1 we list the results of the final screening of sources,
well. We note that our main sources of information on the pos- sorted by selection regions. The numbers refer to detections that
sible astronomical content of the X-ray source error circles were might have entered more than once or refer to more selection
the cross-correlations computed by the Astronomical Catalogue
4
Data Subsystem. This module of the XMM-Newton reduction Finding charts from SDSS or created from digitized optical plates.
188 A. M. Pires et al.: A search for INSs in the 2XMMp catalogue

Table 1. Results of visual screening of the selected 2XMMp sources

Knots in SNRs
INS/INS cand
1.0

spurious/gap
Extragalactic

Other
discriminated by selection region.

I II III IV
Known INSs 16 0 1 1 0.5
INS candidates 1 5 24 2
Solar system objects 2 0 0 0
Extragalactic X-ray sources 12 31 39 6
Bright knots in SNRs 9 59 18 51 0.0
Diffuse emission in clusters of galaxies 2 6 9 0 1.0 I
Spurious out-of-time events / edge 10 70 58 8
Spurious in wing of bright object 2 15 18 6 II
Sources in CCD gap 3 13 46 5 III
Other 2 3 8 0 0.5

Relative frequency
Total 59 202 221 79 IV

0.0
5
regions and not to unique sources . It is worth noting that the 1.0
known M 7 or, more precisely, the pn observations of the M 7 in-
cluded in the compilation of the 2XMMp, were among the final
list of INS candidates, providing a test of the efficiency of our 0.5
procedure in selecting thermally emitting neutron stars.
Overall, we found 59 very soft and low-absorbed sources
with X-ray emission consistent with blackbodies of temperatures
kT ≤ 100 eV and column densities NH ≤ 1 × 1021 cm−2 (region I 0.0
in Fig. 1). In this region we retrieved the totality (16) of pn ob- 1.0
servations of the M 7 carried out until 2006. Two observations
of Saturn and a number of bright knots in the extended diffuse
emission of Galactic SNRs (of the supernova SN 1006 and of 0.5
the relatively old and nearby Cygnus Loop) also are included in
this region. Interestingly, we noted that there were some cases
(included as “other” in Table 1) in which the X-ray source had
no optical candidate but was nonetheless related to (optically 0.0
bright) high proper motion (HPM) stars. For these cases, we Type of object
checked that the position of the X-ray source was compatible
with the position of the HPM star at the time of the X-ray obser- Fig. 2. Histograms showing the relative frequency of the 2XMMp
vation. Region I provided 1 INS candidate. sources that fulfilled the selection criteria applied in order to find new
INS candidates. The four plots show the same histograms in the differ-
There were 202 more absorbed and slightly hotter sources, ent selection regions (see Fig. 1).
consistent with blackbodies of kT ≤ 120 eV and NH = (1–
5) × 1021 cm−2 (region II). This selection region provided 4 INS
candidates (in addition to the INS candidate that had also been as “other” in Table 1 while the pulsar was included as “known
selected in region I). Most of the sources consisted of bright INSs”.
knots in SNRs – the Galactic remnants of Puppis, RCW 86 and We found 79 most absorbed sources, with X-ray emission
SN 1006 as well as SNRs in galaxies M 33, M 31, M 51 and the consistent with kT ≤ 200 eV and NH = (5–10) × 1021 cm−2
Small Magellanic Cloud – and of spurious detections (mostly of (region IV). The detections mainly consisted of knots in the
bright objects on the edge of the CCDs or sources located in the Galactic SNRs Puppis, RCW 86, RCW 89 and Tycho as well as
CCD gaps) and of out-of-time events. in some extragalactic SNRs in the Small and Large Magellanic
A total of 221 sources were found in region III (kT = 120– Clouds. This region provided no new INS candidate – the objects
200 eV and NH = (1–5) × 1021 cm−2 ). Relative to region II, a in Table 1 correspond to INS candidates that already had been
smaller fraction of the total number of selected sources con- selected in region II and to the pulsar found in region III. The
sisted of spurious detections in SNRs, which in this region were results sorted by selection region can be seen in the histograms
mostly extragalactic. Also relative to region I and II, a larger of Fig. 2.
number of extragalactic X-ray sources were found. Region III Thus, out of 7.2 × 104 serendipitous EPIC pn sources above
also provided the largest number of INS candidates (22, in 0.01 s−1 , fewer than 30 candidates met all the selection crite-
addition to 2 repeated entries of selection regions I and II). ria; these are thus intrinsically soft sources not associated with
Interestingly, we detected in this region the X-ray emission of any catalogued optical or infrared object and not likely to be
pulsar PSR J1722-3712 and the quiescent neutron stars XMMU spurious. In the right panel of Fig. 1 we show the positions of
J164143.8+362758 and CXOU J132619.7-472910 in the glob- the final list of unique INS candidates in the hardness ratio dia-
ular clusters M13 and ω Cen. These two sources were included gram HR1 × HR2 . The lowest left part of the diagram is occupied
5
In other words, a soft source which was observed several times by by the soft, low-absorbed M 7 while our candidates, undergoing
XMM-Newton, e.g. one of the M 7, can enter the same selection region higher photoelectric absorptions, move upwards, along with the
more than once, provided that its spectral properties are compatible. blackbody lines of hotter temperatures. Contour lines in the same
Similarly, one INS candidate can show HR that are consistent with more diagram show the hardness ratio distribution of quasars from
than one selection region, and in this case it is counted accordingly. the SDSS DR3 (Schneider et al. 2005) that have counterparts
 A. M. Pires et al.: A search for INSs in the 2XMMp catalogue 189

Table 2. Properties of the INS candidates selected for optical follow-up.

Source identification RA Dec ba NHGal b r90 c pn CRd Selection Optical

(J2000) (J2000) (degrees) (cm−2 ) (arcsec) (s−1 ) region datae
2XMM J104608.7-594306 10 46 08.7 −59 43 06.1 −0.60 1.35 × 1022 1.33 0.060(4) II, III, IV , ♦
2XMM J121017.0-464609 12 10 17.1 −46 46 11.2 +15.52 8.70 × 1020 2.60 0.027(6) III 
2XMM J010642.3+005032 01 06 42.4 +00 50 31.3 −61.79 3.18 × 1020 3.80 0.020(5) III 
2XMM J043553.2-102649 04 35 53.2 −10 26 50.0 −34.80 5.80 × 1020 2.70 0.019(3) III ‡
2XMM J031459.9-291816 03 14 59.9 −29 18 15.5 −58.42 1.32 × 1020 2.41 0.018(4) III ‡
2XMM J214026.1-233222 21 40 26.2 −23 32 22.3 −46.95 3.51 × 1020 1.90 0.0181(20) III †
2XMM J125904.5-040503 12 59 04.6 −04 05 02.3 +58.73 1.81 × 1020 1.90 0.0129(21) III †
2XMM J125045.7-233349 12 50 45.7 −23 33 47.7 +39.31 7.37 × 1020 3.30 0.0122(21) III †
a
Galactic latitude.
b
Total Galactic extinction (Dickey & Lockman 1990).
c
Positional error (90% confidence level).
d
EPIC pn count rate in total energy band (0.2–12 keV).
e
Obtained during different observing periods:  ESO-VLT P79 (pre-imaging), † SOAR 2007A, ‡ SOAR 2007B, ♦ SOAR 2008A.

Table 3. Description of the optical observations.

Target Night Telescope Exposures Exp. Time Total Exp. Time FWHM Airmass
(s) (s) (arcsec)
XMM J1046 22-02-2007 ESO-VLT 2 × B, 2 × R 150, 150 300, 300 0.81, 0.68 1.27, 1.26
XMM J1210 25-02-2007 ESO-VLT 2 × B, 2 × R 150, 150 300, 300 0.83, 0.70 1.16, 1.15
XMM J1250 09-05-2007 SOAR 14 × U, 7 × B, 3 × R 500, 200, 200 7000, 1400, 600 0.90, 0.74, 0.80 1.10, 1.03, 1.01
XMM J1259 10-07-2007 SOAR 4 × U, 4 × B, 4 × R 500, 200, 200 2000, 800, 800 0.75, 0.68, 0.62 1.02, 1.01, 1.01
XMM J2140 11-07-2007 SOAR 4 × U, 4 × B, 3 × R 500, 200, 200 2000, 800, 600 1.11, 0.95, 0.80 1.19, 1.23, 1.25
XMM J0106 21-07-2007 ESO-VLT 2 × B, 2 × R 150, 150 300, 300 0.62, 0.58 1.16, 1.15
XMM J0314 29-01-2008 SOAR 4 × U, 5 × B, 4 × R 500, 200, 200 2000, 1000, 800 0.85, 0.82, 0.72 1.18, 1.22, 1.24
XMM J0435 30-01-2008 SOAR 4 × U, 4 × B, 4 × R 500, 200, 200 2000, 800, 800 1.14, 1.16, 1.10 1.10, 1.11, 1.12
XMM J1046 08-02-2008 SOAR 6×V 1120 6720 0.94 1.23

in the 2XMMi catalogue with a maximum likelihood of detec- The 8 INS candidates in Table 2 were observed using the
tion greater than 20 (total of 796 sources; the list of correlated 8.2 m ESO Very Large Telescope (ESO-VLT) and the 4.1 m
sources was extracted using the XCat-DB). Their location is Southern Astrophysical Research Telescope (SOAR) facilities in
above our selection region III. We can see already that several of Chile during 2007 and 2008. Table 3 shows the log of all the
our INS candidates have hardness ratios compatible with those observations that we have obtained and analysed so far. We note
of the SDSS quasars. that the exposure times, seeing and airmasses listed in Table 3 are
We list in Table 2 the equatorial coordinates and count rates averaged per filter. Hereafter, we adopt the convention “XMM
of a subsample of the candidates, which were the object of Jhhmm” to designate the 8 X-ray sources discussed in this work.
follow-up observations in the optical during 2007 and 2008. The
choice of sources for follow-up was made by selecting the X-ray
3.1. Observations and data reduction
brightest INS candidates that were visible from the southern
hemisphere. Positions are derived by the catalogue pipeline with We obtained deep imaging of the fields of our candidates un-

the SAS task emldetect. The 90% confidence level error circle der photometric sky conditions (see the log of observations
on the position is given by r90 = 2.15 σ2 + σ2syst , where σ is the in Table 3). The FOcal Reducer low/dispersion Spectrograph
nominal error as given by emldetect. The systematic error σsyst (FORS2; Appenzeller et al. 1998) and SOAR Optical Imager
on the detection position is provided by the 2XMMp catalogue (SOI; Schwarz et al. 2004) were used. The FORS2 detector con-
and varies with the success or failure of boresight correction of sists of a mosaic of two 2 k × 4 k MIT CCDs (15 μm) and is op-
the EPIC sources with optical matches in the USNO-B1.0 cata- timized for the red band. It provides imaging at a pixel scale of
logue. 0.25 pixel−1 using standard read-out mode and a 2 × 2 binning
(6.8 × 6.8 field-of-view). With this configuration, the gain is
1.25 e− ADU−1 and the read-out noise 2.7 e− . The SOI instru-
3. Optical follow-up ment uses a mosaic of two E2V 2 k × 4 k CCDs (15 μm) and it
is optimized for the blue and UV bands. It has a field-of-view of
We are presently conducting an optical campaign with the pur- 5.3 × 5.3 and a pixel scale of 0.15 pixel−1 using a 2 × 2 bin-
pose of investigating the nature of the X-ray brightest INS candi- ning. It was used in slow read-out mode, minimizing the read-out
dates of our sample. The immediate goal is to find a possible al- noise to 3.1 e− with a gain of 0.4 e− ADU−1 .
ternative identification – essentially faint polar-type cataclysmic Only pre-imaging data6 were obtained with the ESO-VLT
variables (CVs), late-type stars or active galactic nuclei (AGN) – while ∼90% of the proposed observing time was executed by
for the selected sources, through the analysis of the spectra and
the colour indexes of the optical objects that may be present in 6
The pre-imaging had the purpose of selecting potential optical can-
the X-ray error circles. didates for spectroscopy or deep imaging.
190 A. M. Pires et al.: A search for INSs in the 2XMMp catalogue

SOAR in three different observing periods. Dithering patterns 3.2.2. Photometry

with offsets of 2 and 5 were chosen for the SOAR and the
ESO-VLT observations, respectively. Seeing during the several Two different approaches were adopted to measure magnitudes:
nights of observation varied between 0.58 and 1.2 FWHM, standard PSF fitting, as implemented in the DAOPHOT (Stetson
with mean values of 0.70 and 0.88 for the ESO-VLT and 1987) package for the IRAF environment and, for the fields
SOAR nights. of candidates located at high Galactic latitudes (|b| > ◦
∼ 20 , see
Table 2), variable elliptic apertures and background maps using
We used the already bias and flat field corrected frames pro- SExtractor 2.5.010 .
vided by the ESO-VLT pipeline in the analysis. Images in each PSF fitting using a two-dimensional Gaussian function was
filter were combined using an algorithm to remove cosmic rays adopted whenever there were a fair number of isolated and bright
and bad pixels based on CCD statistics (ccdclip in IRAF7 task non-saturated stars available, allowing a good determination of
imcombine). SOAR data were reduced using standard proce- the modeled PSF to apply to all detected sources on the image
dures within IRAF V2.14 (Tody 1986). The observations were (including faint and overlapping stars). We note, however, that
bias and overscan subtracted and combined skyflat and dome none of the fields is severely crowded, not even the one located
flat images in each filter were used to correct the pixel-to-pixel at b ∼ 0◦ . When this was not the case, we adopted SExtractor
response variations across the CCDs. The science frames in each instead. This software is most commonly used to reduce galaxy
filter were also combined to eliminate cosmic rays and increase survey data but it also performs well in moderately crowded stel-
the signal-to-noise ratio of the data. lar fields. As many extragalactic objects were likely to be present
The observation of photometric standard star fields (Landolt in our high Galactic latitude fields, its usage is convenient since
1992) was arranged to be executed during the same night as our it is possible to derive, for instance, information on how elon-
SOAR targets. Fields RU 149, Mark A and PG1323-085 were gated an object is; additionally, making use of a neural network,
selected to calibrate the observations of sources XMM J0435, SExtrator classifies any given object as stellar or non-stellar.
XMM J0314, XMM J2140, XMM J1259 and XMM J1250. The However, this classification is less reliable for faint fluxes. Its
deeper observations of XMM J1046 were calibrated with the main advantage is that flux measurements of extended objects
standards PG1047+003 and SA 104. For each field (as appropri- tend to be more accurate when compared to the ones obtained
ate depending on the requested observations of the science tar- using fixed aperture photometry, since elliptical apertures with
get) we obtained 10 s (U or V) and 5 s (B or R) single exposures. variable sizes (based on the object intensity) are used instead.
The airmasses at which the observations of standard stars were Another advantage is that a global background map is created,
carried out were similar to the ones of the science targets. As which better accounts for local spatial variations of brightness
only one exposure was taken in each filter (at a given airmass), it due e.g. to nebular emission or caused by scattered light from
was not possible to determine the extinction coefficients; mean bright objects.
values for the periods were adopted instead. On the other hand, Using SExtractor, we created catalogues gathering position,
the zero-point magnitudes and colour coefficients were deter- intensity and shape information on every optical object present
mined from our observations of the standard stars in each night. in our high Galactic latitude frames. Source detection was car-
The fit to the standard system was performed relative to the U −B ried out by convolving a Gaussian filter having the mean image
(for the U magnitudes) and B − R (for the B and R magnitudes) FWHM. The resulting list of sources in each field was then cor-
colours. The V SOAR instrumental magnitudes (period 2008A) related to keep the optical objects which were detected in all
were transformed adopting a null colour coefficient and the mean filters. In the following subsections we use the information ex-
value of the extinction coefficient for the period. The instrumen- tracted from these optical catalogues to discuss the nature of our
tal magnitudes of the ESO-VLT images were transformed using sample of INS candidates and field objects.
directly the extinction coefficients and colour terms provided by
the ESO-VLT calibration web pages8 . We compared the R mag-
3.2.3. X-ray/optical associations
nitudes of non-saturated GSC-2 stars present in the fields with
our results and concluded that the agreement is good with a mean We found at least one optical object inside the X-ray error circles
dispersion of ∼0.03 mag. (90% confidence level) of all INS candidates but one, the X-ray
brightest source XMM J1046 (see Fig. 3). In this case no optical
object brighter than the limiting magnitude of our present data
3.2. Data analysis and results is present within ∼4.3 (>∼5 σ) from the position of the X-ray
source (see discussion in 3.2.5).
3.2.1. Astrometry Table 4 lists the R magnitudes and B − R and U − B11
colours of the possible optical counterparts of the remaining
The astrometric calibration was performed using the USNO- seven sources, as well as their positional offsets relative to the
B1.0, 2MASS and GSC-2 catalogues and the GAIA 4.2-1 soft- X-ray coordinates and the implied logarithmic X-ray-to-optical
ware9 . The pixel coordinates of the non-saturated catalogued flux ratios, uncorrected for absorption. The candidate counter-
stars in each combined science frame were determined by fitting parts were well detected in all (U BR or BR) filters, with the ex-
a two-dimensional Gaussian function to their intensity profiles. ception of the two optical objects possibly associated with XMM
An astrometrical solution was then computed by fitting the pixel J0106, only detected in the R image. Among the possible coun-
to the celestial coordinates. Overall, our astrometric errors are of terparts, these were also the faintest ones (R ∼ 24.5).
∼0.15 or better. To estimate the reliability of each X-ray/optical association,
we defined a likelihood ratio (LR) based on the probabilities that,
given the positions, an optical object is the true counterpart of
7
http://iraf.noao.edu
8 10
http://www.eso.org/observing/dfo/quality http://terapix.iap.fr/rubrique.php?id_rubrique=91
9 11
http://star-www.dur.ac.uk/∼pdraper/gaia/gaia.html Only for SOAR data.
 A. M. Pires et al.: A search for INSs in the 2XMMp catalogue 191

Fig. 3. R images of the fields of the observed INS candidates. Candidates XMM J1046, XMM J1210 and XMM J0106 were observed with the
ESO-VLT in period P79 while candidates XMM J0435, XMM J0314, XMM J2140, XMM J1259 and XMM J1250 were observed using the SOAR
telescope. In these images, north and east point upwards and to the left, respectively. Circles show the 90% confidence level error on the positions.

Table 4. Results of follow-up investigations.

Candidate r R U−B B−R LR Pchance e log( fX / fR )a

(arcsec) (%)
XMM J1210 1.55 20.13(3) ? 1.35(3) 8.4 2.8 0.037 0.18
XMM J0106 2.75 24.51(9) ? > 2.0
∼ 1.2 22.4 0.178 1.93
1.68 24.58(9) ? >
∼ 1.9 1.2 22.4 0.162 1.96
XMM J0435 1.20 22.15(4) −0.51(8) 0.93(7) 15.2 2.3 0.068 0.77
XMM J0314 0.67 21.37(3) −0.13(7) 1.43(4) 32.1 1.7 0.022 0.41
XMM J2140 1.26 23.78(9) −1.08(16) 1.06(15) 6.0 6.4 0.156 1.40
XMM J1259 1.34 21.64(4) −1.08(4) 0.76(5) 25.5 1.0 0.098 0.61
XMM J1250 1.80 22.19(4) −0.71(4) 0.74(4) 8.4 4.1 0.029 0.57
a
The X-ray flux is computed in the 0.15–3 keV energy band assuming an absorbed power-law model.

the X-ray source or is just a random object that happens to lie distribution of spurious associations and therefore estimate the
angularly close to its position. Following de Ruiter et al. (1977), chance probability of X-ray/optical associations in our data, we
LR can be defined as: performed Monte Carlo simulations randomly changing the po-
⎡  ⎤⎥ sition of the X-ray source accross the optical fields. For a given
1 ⎢⎢⎢ x2 ⎥
LR = exp ⎢⎣ ⎢ 2λ − 1 ⎥⎥⎥⎦ (2) simulated X-ray source, we then computed the LR quantity given
2λ 2 by Eq. (2) for each optical object (if any) lying inside r90 . The
overall distribution of positive matches (among 1000 simula-
where λ is defined as λ = πσ2 ρ, σ is the error on the X-ray tions) can be seen in the histograms of Fig. 4. As expected, the
and optical positions and ρ is the surface density of brighter or distribution of LR is highly asymmetrical, strongly peaking at
equally bright objects on the image. The dimensionless variable low values and then monotonically extending towards high LR.
x = r/σ is the significance of the position offset (r). Hashed bars in Fig. 4 highlight the number of simulated random
We list in Table 4 the LR quantities for every possible X-ray/optical associations with LRi ≥ LRX , where LRi and LRX
X-ray/optical association, which ranges from ∼1 to 32. Although are the likelihood ratios of simulated and the actual X-ray/optical
objects with large LR are more likely to constitute a true asso- association of the given field.
ciation, the computation of actual probabilities of identification
requires a calibration of LR taking into account the a priori prob- In general, the simulations show that the probability of a
ability that any X-ray source has a counterpart in the optical sam- chance association is low for the optical objects found inside r90
ple considered (see e.g. Pineau et al. 2008). on the real data (Table 4). Out of 1000 simulations, we usually
LR strongly depends on the local density of optical objects; found that less than 3% of the cases had an X-ray/optical asso-
the chance of a spurious association rises with the number of ob- ciation with a likelihood ratio greater than or equal to the real
jects of similar or brighter magnitudes. In order to derive the LR one. The two exceptions are the associations found for source
192 A. M. Pires et al.: A search for INSs in the 2XMMp catalogue

Fig. 4. Histograms showing the LR distribution of X-ray/optical associations in the fields of candidates for which possible optical counterparts were
found. Each histogram is computed from 1000 Monte Carlo simulations. The number of matchings is given in the right upper corners. Hashed
bars show simulated X-ray/optical associations which have a LR greater than or equal to the one found for the X-ray source.

XMM J0106, with Pchance ∼ 22% and, to a lesser degree, for and CVs. Among the five sources, the proposed counterpart of
source XMM J2140, with Pchance ∼ 6%. These are the two X-ray XMM J0314 is the only one that is somewhat more isolated from
sources with the faintest optical candidates and the only ones the bulk of these two populations of X-ray emitters, although
with log( fX / fR ) > 1. However, while the two candidate coun- some quasars at still low redshifts from Richards et al. (2001)
terparts for XMM J0106 were detected only in the R band, the do show similar colours. None of the optical sources exhibits
one for XMM J2140 was well detected in the U, B and R filters, colours typical of late-type stars. The strong blue/UV excess,
and it shows very blue colours. We argue in the following that, with u − g < 0, shown by the proposed counterparts of sources
based on the results obtained here and together with the analy- XMM J1259 and XMM J2140 is remarkable. The colours of the
sis of their colours and X-ray-to-optical flux ratios (Sect. 3.2.4), optical objects in the error circles of XMM J1250 and XMM
the optical objects found inside the X-ray error circles of all INS J0435 are more intermediate but still consistent with those of
candidates are very likely to constitute their true optical counter- AGN and CVs.
parts, with, again, the exception of source XMM J0106.
Interestingly, when compared to other uncatalogued field ob-
jects present in our SOAR frames and detected in all filters, the
3.2.4. Identification with other classes of X-ray emitters colours and even the UV excess exhibited by our sample of opti-
cal candidates do not stand out from the overall field population
As mentioned before, some populations of astrophysical objects (see right pannel of Fig. 5). For further comparison, we also plot
are expected to pollute our sample of INS candidates. In order as density contours SDSS objects present in five regions of the
to identify these among the sources having optical candidates, sky located at similar column densities as the SOAR fields. More
we compared their positions in the colour-colour g − r × u − b precisely, we gathered SDSS entries in a SOAR field-of-view
diagram with those of the major classes of X-ray emitters (left centered on five different directions, requiring r magnitudes be-
pannel of Fig. 5). The magnitudes in the Sloan photometric sys- tween 21 and 24, which is the range covered by our SOAR field
tem were computed using the Johnson-Morgan-Cousins trans- objects, and u , g and r errors smaller than 0.5. Overall, the
formation equations in Fukugita et al. (1996). In this diagram plot shows that the SOAR and SDSS objects constitute the same
are plotted, as contours, the spectroscopically identified popula- population of “blue field objects” located at high Galactic lati-
tion of quasars (Richards et al. 2001) and cataclysmic variables tudes, with colours not particularly distinguishable from those of
(Dillon et al. 2008, and references therein) from the SDSS, as the sample of soft X-ray counterparts. Apart from these blue ob-
well as late-type stars (Bilir et al. 2005). Only the five INS can- jects, a smaller fraction of the SOAR field population occupies
didates observed with SOAR are shown, since we do not have U the same location in the colour diagram as the one of quasars
magnitudes for the two candidates observed with the ESO-VLT identified at more remote redshifts (z > 3; see open triangles
(sources XMM J1210 and XMM J0106). We note, however, that in Fig. 5). Most (> ∼99%) of the SOAR objects do not have a
the B − R colours of the optical candidates of these sources are correlation with the 2XMMi and none with the FIRST (radio)
among the “reddest” of our sample (Table 4). catalogues. According to Mateos et al. (2008), the expected den-
The u − g and g − r colours of the optical candidates sity of soft (0.5–2 keV) X-ray sources at high Galactic latitudes
of the SOAR targets are clearly consistent with those of AGN |b| > 20◦ , with fluxes greater than fX > 10−14 erg s−1 cm−2 is
 A. M. Pires et al.: A search for INSs in the 2XMMp catalogue 193

1.5 1.5

*
1.0 * 1.0
*
J0435 *
* * J0314
* **
** * ** * **
*
g’−r’

g’−r’
0.5 J2140* * *** * * * *
* 0.5
* *
* ** * * ** *** * *** * * * * *
* * ** * * ** * * **
J1259************************************* * ****** * *** *
*********************
********************************* * *** * * *
* * ************************************ * * *
************************* *** *
***************************************************************** * *** * * * *
*
**
* *** * * * * * ** *
* * * * *
* ******************************* * **** * * ** *
* * ** *** *** ** * *
0.0 * 0.0
*************************************** ** **
*** ************* * *
* ********* * *
*** ***** * *
*
* * *
J1250
−0.5 −0.5
*
−0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 −0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0

u’−g’ u’−g’

Fig. 5. Left: colour–colour g − r × u − g diagram comparing the locations of different astrophysical populations of X-ray emitters. Red stars (and
contours) show the position of identified QSOs at redshifts lower than 3. Hashed blue and solid green contours are CVs and late-type stars from
the SDSS, respectively (see text). The five INS candidates observed with SOAR are represented by plain black circles with labels. Right: same
diagram showing the colours of field objects detected in the three (U, B and R) filters of our SOAR observations (plain red circles). The population
of SDSS objects located at similar column densities is shown for comparison as a density contour. Open triangles are identified quasars located at
redshifts greater than 3. Optical candidates of the soft X-ray sources are shown as large blue circles.

∼100 deg−2, corresponding to ∼0.75 sources in a typical SOAR Table 5. Upper limits on the B, V and R magnitudes for the optical
field-of-view – which is roughly in agreement with our results. counterpart of XMM J1046.
In summary, the SOAR field objects do not seem to be active
in X-rays or radio wavelengths, despite their blue colours being Filter msyn σdet mdet log(FX /Fλ )
similar to those of the INS candidates. B 26.0 2.4 26.2(5) >
∼ 2.2
V 25.5 2.5 25.9(4) >
∼ 3.1
We analysed the extent distribution of field objects and opti- >
R 24.2 2.4 25.5(5) ∼ 3.2
cal counterparts using the ellipticity derived by SExtractor. The
goal was to verify the possibility of separating stars/AGN from
galaxies by measuring their point spread functions. Defining el-
lipticity as usual (e = 1 − b/a, where a and b are the semi-major To test the detection limit specifically in the region covered
and semi-minor axis lengths, respectively) we found that most by the error circle of the X-ray source, small12 sections of the B,
(>
∼60%) of the SOAR field objects are not elongated (e < ∼ 0.2). V and R images were analysed so as to minimize the contribu-
The counterparts of the INS candidates are also rather point- tions of nearby bright stars and of nebular emission. The frames
like, with mean ellipticity ∼0.08 – the most elongated one being were binned by a factor of 2 in order to increase the signal-
source XMM J2140 and the two objects possibly associated with to-noise ratio. Images of synthetic stars (created using the PSF
XMM J0106 (see values in Table 4). model derived from the data) of progressively fainter magnitudes
were added at the position of the X-ray source. The composite
images were then subjected to automatic detection and to mag-
3.2.5. Optical upper limits for XMM J1046 nitude measurement using DAOPHOT. In this process, the syn-
thetic star was no longer detected or was rejected (while trying to
The only INS candidate of our sample that remains without an fit the PSF model) when the signal-to-noise ratio was worse than
optical candidate is source XMM J1046. The short ESO-VLT ex- S /N ∼ 2.5. We then defined the limiting magnitude as the mag-
posure obtained in 2007 revealed no counterparts brighter than nitude of the faintest synthetic star still succesfully measured at
R ∼ 25. The location of XMM J1046 towards the Carina star this confidence level. We give in Table 5 the magnitudes of the
forming region explains the presence of many bright stars and faintest simulated star (msyn ) detected at a signal-to-noise ratio
of relatively intense diffuse emission in its optical field. The R σdet , along with the measured magnitude (mdet ). The discrepancy
filter, which includes the strong Hα line, is thus particularly af- between the simulated and detected values of the R magnitude
fected by background intensity variations, impacting the quality is due to a somewhat poorly defined PSF model in the wave-
of the photometry. In the light of this, we obtained additional length range most affected by background intensity variations.
deep SOAR exposures in the V band in 2008. The V band ex- Taking the limiting magnitudes as upper limits on the brightness
cludes some of the strongest lines found in H ii regions (Balmer of the optical counterpart of XMM J1046, Table 5 also lists the
emission as well as [N ii] and [S ii] emission lines) and is thus implied X-ray-to-optical flux ratios corrected for photoelectric
best suited for a deep upper limit on the optical emission of
XMM J1046. However, the V filter is still somewhat contami- 12
54 × 46 , 128 × 100 and 70 × 70 respectively for the R, B and
nated by the nebular [O iii] emission. V images.
194 A. M. Pires et al.: A search for INSs in the 2XMMp catalogue

Table 6. X-ray spectral analysis of the INS candidates (excluding XMM J1046).

Candidate OBSID texp a d.o.f. kT fX,14 b C Goodnessc Γ fX,14 C Goodness

(ks) (eV) (erg s−1 cm−2 ) (%) (erg s−1 cm−2 ) (%)
XMM J1210 204710101 2.1 22 253+53
−40
+0.8
3.6−0.7 14 4.6 2.0(4) +1.0
4.6−0.9 17 9.9
XMM J0106 150870201 1.7 5 247+116
−60
+0.9
2.6−0.7 5 18.4 1.5(8) +1.4
4.6−1.3 5 19.8
XMM J0435 307001301 7.3 24 200+29
−24
+0.3
2.00−0.29 22 26.8 2.4(4) 2.8(4) 21 22.7
XMM J0314 201750901 11.6 17 229+58
−44 1.7(3) 24 76.7 2.1(4) 2.5(5) 21 60.0
XMM J2140 008830101 10.5 15 144+14
−12 2.00 +0.22
−0.24 17 58.5 +0.24
2.94−0.23 +0.3
2.59−0.25 9 6.8
XMM J1259 203020101 9.8 26 176+21
−18
+0.26
2.15−0.25 44 96.7 +0.23
2.43−0.22 3.1(3) 35 79.4
+34 +0.26
XMM J1250 303561001 7.8 11 186−27 1.30−0.24 6 6.7 2.6(4) 1.7(3) 6 7.0
a
Exposure times are filtered for background flares.
b
Observed flux refers to range 0.15–3 keV in units of 10−14 erg s−1 cm−2 .
c
The “goodness-of-fit” is derived from a number of 1000 Monte Carlo simulated spectra.

absorption and interstellar extinction. In order to compute the (Smith 2006). The distance to XMM J1046 is likely to be com-
unabsorbed fluxes we adopted the best blackbody fit parame- parable. Together with the optical data, the high lower limit on
ters of the source (see Sect. 4) and derived AV = 1.96 using the the X-ray-to-optical flux ratio, log(FX /FV ) > +0.3
∼ 3.1−0.1 , pratically
Predehl & Schmitt (1995) relation between the X-ray absorp- excludes any other possibility than an INS. The timing analysis
tion and optical extinction. For the other bands, we adopted the applying a Rayleigh Zn2 test shows no pulsations up to a rather
Aλ /AV extinction relations of Cardelli et al. (1989). non-constraining 30% upper limit (3σ), in the 0.073–100 s pe-
riod range.
4. X-ray analysis
4.2. Other INS candidates
XMM J1046 is very likely a new thermally emitting INS with
properties that, at first glance, seem similar to those of the M 7. Each one of the other INS candidates was serendipitously ob-
A detailed X-ray and optical analysis of this object has been re- served only once by the XMM-Newton detectors. The event files
ported in Pires et al. (2009). We recall below the main results were reprocessed using SAS 8.0.0, applying standard procedures
presented in that paper and refer to it for details on the reduction as for the source XMM J1046. For the spectral analysis, the
and analysis of the X-ray data of this and the other seven INS low signal-to-noise ratio spectra were only fitted assuming an
candidates. absorbed blackbody or power-law using XSPEC 12.4. Being
roughly one order of magnitude fainter than XMM J1046, the
low number of counts prevents well constrained spectral fits; in
4.1. XMM J1046 particular, it does not permit a strong constraint on the value of
XMM J1046 was detected in 16 XMM-Newton observations of the column density. Blackbody fits show rather high tempera-
the well studied binary system Eta Carinae and of the Wolf- tures, typically kT >∼ 200 eV, and column densities ranging from
Rayet star HD 93162 (WR 25), as well as in one Chandra obser- 0 to 4.2 × 1021 cm−2 , at a 68% confidence level. Although appar-
vation of the Carina Nebula. Although the observing conditions ently in disagreement with our selection criteria (kT ≤ 200 eV),
were far from optimal – in many observations, XMM J1046 was the significant errors in HR explain the inclusion of these X-ray
located near the edge or in CCD gaps and often the effective ex- sources among the selected sources. We thus decided to hold NH
posure times were short – we have analysed the available X-ray fixed at the Galactic value (Dickey & Lockman 1990, Table 2)
archival data on this source, which span more than six years. in order to obtain better constrained spectral fits. The results of
The event files were reprocessed using SAS 7.1.0 and the blackbody and power-law fits for these sources can be seen
CIAO 4.0.113, applying standard procedures. Using XSPEC 12.4 in Table 6 (errors are 1σ). The quality of each fit (“goodness”)
(Arnaud 1996, 2004), we tested different models (blackbody, corresponds to the fraction of simulations yielding a better fit
power-law, bremsstrahlung, Raymond-Smith, . . . ). All param- statistic than the actual data, with high values implying bad fits.
eters were allowed to vary freely. Due to the low number of Whereas the relatively large number of counts collected for
counts, we applied the C-statistic (Cash 1979) in order to derive XMM J1046 safely allows one to exclude a power-law shape for
the best fit parameters and their uncertainties. Whenever possi- its X-ray spectrum, the much lower signal-to-noise spectra of the
ble for a given observation, data from all EPIC cameras were seven sources studied here are equally well described by either
analysed simultaneously to better constrain the spectral parame- a hot blackbody or by a soft power-law energy distribution, with
ters. spectral indexes usually larger than 2.
We have shown that the X-ray emission of XMM J1046 is
best described by a soft blackbody, with mean kT ∼ 117 ± 14 eV 5. Discussion
and NH ∼ (3.5 ± 1.1) × 1021 cm−2 and a stable 0.15–3 keV ob-
served flux of fX ∼ (1.03 ± 0.06) × 10−13 erg s−1 cm−2 (errors are At energies below 2 keV, the X-ray spectra of quasars and
3σ). There is no evidence for a hard non-thermal component. In Seyfert I galaxies usually show an excess relative to the extrap-
this work, we adopt the notation Fλ to denote the unabsorbed olation of the high energy (2–10 keV) power-law. It is generally
fluxes while fλ is used for the observed flux. The NH is a fac- believed that this “soft excess” is produced by the scattering of
tor of ∼10 higher than those typical of the M 7 and is consistent thermal optical/UV photons from the accretion disk surrounding
with the one towards Eta Carinae, which is located at ∼2.3 kpc the central black hole to soft X-ray energies, by a population of
ambient hot electrons (e.g. Atlee & Mathur 2009, and references
13
http://cxc.harvard.edu/ciao4.0/index.html therein). Over a limited energy range – in particular, that covered
 A. M. Pires et al.: A search for INSs in the 2XMMp catalogue 195

by the sensitivity of the ROSAT instruments, 0.1–2.4 keV – the dubious of the X-ray/optical associations. They have been de-
energy distribution of the soft excess is roughly described by a tected only in the R filter, preventing colour estimation, whereas
power-law of spectral index > ∼2 (e.g. Boller et al. 1996; Grupe the optical candidate of source XMM J2140 has been well de-
et al. 1998; Page et al. 2004). This is in agreement with the tected in the U, B and R filters. XMM J0106 was detected in a
results obtained in Sect. 4, for the INS candidates with optical very short exposure and therefore its HR errors are more poorly
counterparts (Table 6). Moreover, X-ray emitting quasars cluster determined; however, the two possible optical counterparts de-
in a region of the HR1 × HR2 diagram that is above selection re- fine an already very high log( fX / fR ) ∼ 2, characteristic of the
gion III (contours in Fig. 1), the location where we expect to find most extreme classes of X-ray emitters as AM Her systems and
sources with emission compatible with the hottest blackbodies BL Lac objects (Schwope et al. 1999), and it is worthy of further
that entered our selection. At faint fluxes, the 2XMMp sources investigation.
show rather large HR errors, reflecting the fact that the low num-
Our best INS candidate is XMM J1046, for which no optical
ber of detected photons prevents a precise determination of the
source spectral energy distribution. As it is evident from Fig. 1, counterpart is detected down to the limiting magnitude of our
several of the INS candidates share hardness ratios common to present data. On the basis of the analysis reported in Pires et al.
(2009), the identification of XMM J1046 with an INS is very
both (galactic and extragalactic) populations. For these reasons,
our search procedure is sensitive to optically faint AGN, espe- likely. This could be the first example of a presumably radio-
cially those showing soft X-ray excess. quiet and X-ray dim INS, located at a significantly greater dis-
tance than the M 7. Its spatial location in the Galactic plane and
Magnetic CVs exhibit a low temperature component de-
the derived value of the column density suggest that it may be
tectable in soft X-rays or at UV wavelengths, depending on
physically associated with the Carina Nebula, a giant H ii region
whether the source is in a high or low accretion state (e.g.
harbouring a large number of massive stars and with ongoing
Ramsay & Cropper 2004; Ramsay et al. 2004). This emission
active star formation (see e.g. Smith & Brooks 2007, and refer-
is due to the reprocessing of hard X-ray photons (of several tens
ences therein).
of keV) produced in a shock of the accretion flow into the pho-
tosphere of the white dwarf. In principle, faint polars showing a Several attempts to identify new thermally emitting INSs
particularly large soft X-ray excess and with uncatalogued opti- have been carried out in the years that followed the discovery
cal counterparts can be present in our sample of INS candidates of the seven sources, usually by cross-correlating ROSAT data
as well. with optical, radio and IR catalogues (e.g. Rutledge et al. 2003;
Active coronae of late-type stars manifest themselves as soft Agüeros et al. 2006; Chieregato et al. 2005; Treves et al. 2007).
X-ray emission with luminosities generally below 1031 erg s−1 . However, these searches are hampered by the large ROSAT posi-
This population dominates soft X-ray samples at low Galactic tional errors at faint fluxes, especially in the populated regions of
latitudes (Motch et al. 1997). It is well known, however, that the Galactic plane. For these regions, many spurious (low signif-
the high latitude X-ray sky at flux levels of 10−14 erg s−1 cm−2 is icance) candidate optical/IR counterparts enter the X-ray error
dominated by AGN (e.g. Maccacaro et al. 1982; Barcons et al. circle of a given ROSAT source. This makes the probability of
2007; Mateos et al. 2008); the stellar content, although more im- erroneously assigning an identification significant, thus exclud-
portant in soft X-ray-selected samples such as ours, should be ing the source as a potential INS candidate. The final effect is
less than 10%. In the optical, none of the INS candidates show that the Galactic plane is largely “avoided” by the usual cross-
colours typical of M, G, K stars (Fig. 5). It is expected that late- correlation algorithms (see e.g. Rutledge et al. 2003, Fig. 1), and
type stars have log( fX / fR ) < −1 (e.g. Maccacaro et al. 1982; proposed INS candidates are usually located at high Galactic lat-
Stocke et al. 1991) which is clearly not the case for any of our itudes.
INS candidates, all with log( fX / fR ) > 0. Therefore, we do not
A long-term project that investigates INS candidates selected
expect any of the investigated X-ray sources to be identified with
from the ROSAT Bright Source Catalogue (Voges et al. 1999),
late-type stars.
making use of follow-up investigations with the Swift satel-
Based on the X-ray/optical associations discussed in
lite, is presently being conducted by Rutledge et al. (2008) and
Sect. 3.2.3, the objects found inside the error circles of the INS Letcavage et al. (2009). Follow-up X-ray and optical observa-
candidates are very likely to constitute their true optical coun-
tions of one candidate with no evident counterpart led to the dis-
terparts, all with a very low (< ∼5%) probability of chance asso- covery of Calvera, a likely INS (whose exact nature is, however,
ciation, with the exception of XMM J0106. The strong UV ex- still unclear) with a very large FX /FV >
cess exhibited by several of the sources (XMM J1259, XMM ∼ 8700 (Rutledge et al.
2008). Not an exception, Calvera is located at high b, implying
J2140, and XMM J1250) is clearly consistent with those of
a remote distance of d = 8.4 kpc and a vertical velocity in ex-
AGN and CVs (Fig. 5). However, optical colours alone would cess of ∼5100 km s−1 to explain its current position well above
not be sufficient to distinguish these X-ray sources from the
the plane, considering a standard cooling time and thermal emis-
population of (X-ray and radio-quiet) blue objects located at
sion similar to the M 7. Instead, it is more likely that Calvera is
high Galactic latitudes. In spite of the more average colours of a nearby radio pulsar.
the optical candidates for sources XMM J1210, XMM J0435
and XMM J0314, their high X-ray-to-optical flux ratios leave According to the detailed population synthesis calculations
no doubt that they are likely of extragalactic origin. According of Posselt et al. (2008), it is expected that, in general, new
to Barcons et al. (2007), the fraction of X-ray sources with unidentified cooling INSs at faint soft X-ray fluxes are to be
log( fX / fR ) > 1, where obscured AGN are expected, although found in the Galactic plane, at small angular distances from their
larger in hard X-ray-selected samples, is of just a few percent birth star forming regions – in particular, of the rich OB associ-
in soft ones. As mentioned in Sect. 3.2.3, two out of seven of ations located beyond the Gould Belt, such as Carina, Vela, and
our candidates show log( fX / fR ) > 1, XMM J0106 and XMM Cygnus-Cepheus. At the flux limit applied in our search, the ex-
J2140. These are the sources with the most elongated optical pected number of cooling neutron stars having the same prop-
candidates as well. In particular, the two optical objects found erties as those exhibited by XMM J1046 – i.e. a slightly hotter
in the error circle of source XMM J0106 correspond to the most source located at a greater distance relative to the M 7 – is of
196 A. M. Pires et al.: A search for INSs in the 2XMMp catalogue

only ∼50–80 in the whole sky14 , which translates into less than are currently developping a model for this Galactic population
one source present in the 2XMMp catalogue. The fact that we which aims at making the least number of uncertain assumptions
found one source exhibiting these unique characteristics corrob- – even at the expense of having a less sophisticated description
orates the Posselt et al. (2008) conclusions that a search for new of the real population – while, based on Monte Carlo simulations
cooling INSs should not be blind, i.e. one has to look preferen- and observations, in particular with the XMM-Newton, would be
tially in the most promising regions of the sky – i.e. near more statistically robust in order to constrain important properties of
remote OB associations. this population of unique Galactic remnants, especially beyond
Population synthesis constitutes a fundamental tool in order the Solar vicinity.
to compare observational results from e.g. surveys with theoret-
ical expectations for a given population of astrophysical objects.
If a large number of objects are known, as for the case of radio 6. Summary and conclusions
pulsars, it allows the testing of many of the population physi- Our search for new thermally emitting INSs in the 2XMMp cat-
cal (and theoretically unknown) properties. Recent examples are alogue has revealed a number of interesting and previously un-
the work of Faucher-Giguère & Kaspi (2006), conducted for the known soft X-ray sources. Deep optical imaging revealed likely
population of radio pulsars, and the study carried out by Story optical counterparts for six of them which, based on blue opti-
et al. (2007) with the population of disk millisecond radio and cal colours and X-ray-to-optical flux ratios around 1–10, identify
γ-ray pulsars. them as, most likely, AGN and CVs. Source XMM J0106 has no
In the particular case of thermally emitting INSs, population evident optical candidate and should be further investigated.
syntheses conducted in the early 90s predicted that a large num- Source XMM J1046 is a newly discovered thermally emit-
ber of old neutron stars accreting from the ISM would be de- ting INS, possibly radio-quiet and similar to the M 7. As is ex-
tected by ROSAT (Treves & Colpi 1991; Blaes & Madau 1993; pected for cooling neutron stars, its X-ray emission is described
Madau & Blaes 1994). It soon became clear that the unrealistic by an intrinsically soft and thermal energy distribution, stable
overprediction was due to several simplfying assumptions used on long time scales and with no evidence for magnetospheric
in the model. Several mechanisms act in order to inhibit accre- activity. The present lower limit on the X-ray-to-optical flux ra-
tion, such as the presence of the magnetic field of the neutron tio already excludes standard classes of X-ray emitters. Its likely
star (Toropina et al. 2003; Ikhsanov 2007) and that of an in- location in the Carina Nebula is in agreement with the most up-
falling material which is weakly magnetized (Perna et al. 2003) to-date expectations of population synthesis models.
or heated by the emergent X-rays (e.g. Blaes et al. 1995). It is This work confirms the use of the XMM-Newton catalogue
believed that accretion should proceed at a rate well below the of sources as an efficient tool to identify new thermally emitting
Bondi-Hoyle one, thus affecting the actual number of sources INSs and other particular classes of soft X-ray sources. A simi-
that can be observed by current X-ray missions. The population lar search using the Chandra data with its excellent astrometric
of cooling nearby INSs from the Gould Belt has been extensively accuracy would be instrumental in selecting sources, especially
modeled by Popov et al. (2000, 2003, 2005). In Popov et al. those located in the Galactic plane, where it is more likely to find
(2006), the expected number of sources at a given flux was used new INS candidates.
to constrain cooling curves of INSs. Posselt et al. (2008) im-
proved many of the features of their population synthesis model,
mainly by providing a new description of the ISM distribution Acknowledgements. We thank N. Grosso for discussions about Chandra data
and source XMM J1046. The work of A. M. P. is supported by FAPESP (grant
and that of the massive progenitors. Interestingly, they used the 04/04950-4), CAPES (grant BEX7812/05-7), Brazil, and the Observatory of
spatial distribution of the synthetic population in order to define Strasbourg (CNRS), France.
search strategies to identify new cooling INSs.
The expected number of such sources to be detected by References
current X-ray missions is strongly model-dependent. It comes
Agüeros, M. A., Anderson, S. F., Margon, B., et al. 2006, AJ, 131, 1740
from the parametrization of properties inherited at birth (kick Aguilera, D. N., Pons, J. A., & Miralles, J. A. 2008, ApJ, 673, L167
velocity and mass distribution of the compact remnant) to that Appenzeller, I., Fricke, K., Fürtig, W., et al. 1998, The Messenger, 94, 1
of the cooling history and emission mechanism of the neutron Arnaud, K. A. 1996, in Astronomical Data Analysis Software and Systems V,
star. Together with realistic descriptions of the Galactic gravita- ed. G. H. Jacoby, & J. Barnes, ASP Conf. Ser., 101, 17
Arnaud, K. 2004, BAAS, 36, 934
tional potential, the ISM distribution and the detector character- Atlee, D. W., & Mathur, S. 2009, ApJ, submitted,
istics, these ingredients determine the neutron star thermal and [ArXiv:0901.3566]
dynamic evolution and detectability. Similarily, from the obser- Barcons, X., Carrera, F. J., Ceballos, M. T., et al. 2007, A&A, 476, 1191
vational point of view, selection biases and the area effectively Bilir, S., Karaali, S., & Tunçel, S. 2005, Astron. Nachr., 326, 321
covered by the survey (considering the constraints applied while Blaes, O., & Madau, P. 1993, ApJ, 403, 690
Blaes, O., Warren, O., & Madau, P. 1995, ApJ, 454, 370
correlating X-ray sources with other wavelengths) have to be Boller, T., Brandt, W. N., & Fink, H. 1996, A&A, 305, 53
taken into account in order to derive correct observational up- Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245
per limits on the number of sources at a given flux. Overall, even Cash, W. 1979, ApJ, 228, 939
the simplest models have a level of complexity that is evident Chieregato, M., Campana, S., Treves, A., et al. 2005, A&A, 444, 69
de Ruiter, H. R., Arp, H. C., & Willis, A. G. 1977, A&AS, 28, 211
from the number of assumed hypotheses – many of them involv- Dickey, J. M., & Lockman, F. J. 1990, ARA&A, 28, 215
ing significant theoretical and observational uncertainties. We Dillon, M., Gänsicke, B. T., Aungwerojwit, A., et al. 2008, MNRAS, 386, 1568
Faucher-Giguère, C.-A., & Kaspi, V. M. 2006, ApJ, 643, 332
14
The range shown here for the expected number of sources is Fukugita, M., Ichikawa, T., Gunn, J. E., et al. 1996, AJ, 111, 1748
Grupe, D., Beuermann, K., Thomas, H.-C., Mannheim, K., & Fink, H. H. 1998,
derived from Posselt et al. (2008) adopting the dotted and hashed A&A, 330, 25
curves of their Fig. 5 as lower and upper limits, respectively. See Haberl, F. 2007, Ap&SS, 308, 181
Posselt et al. 2008 for details on the assumed parameters. EPIC pn Heyl, J. S., & Kulkarni, S. R. 1998, ApJ, 506, L61
counts were converted to ROSAT PSPC counts using WebPIMMS Ho, W. C. G., Kaplan, D. L., Chang, P., van Adelsberg, M., & Potekhin, A. Y.
(http://heasarc.gsfc.nasa.gov/Tools/w3pimms.html). 2007, Ap&SS, 308, 279
 A. M. Pires et al.: A search for INSs in the 2XMMp catalogue 197

Ikhsanov, N. R. 2007, Ap&SS, 308, 137 Pires, A. M., Motch, C., Turolla, R., Treves, A., & Popov, S. B. 2009, A&A,
Kaplan, D. L. 2008, in 40 Years of Pulsars: Millisecond Pulsars, Magnetars and 498, 233
More, ed. C. Bassa, Z. Wang, A. Cumming, & V. M. Kaspi, AIP Conf. Ser., Popov, S. B., Colpi, M., Prokhorov, M. E., Treves, A., & Turolla, R. 2000, ApJ,
983, 331 544, L53
Kaplan, D. L., & van Kerkwijk, M. H. 2005a, ApJ, 628, L45 Popov, S. B., Colpi, M., Prokhorov, M. E., Treves, A., & Turolla, R. 2003, A&A,
Kaplan, D. L., & van Kerkwijk, M. H. 2005b, ApJ, 635, L65 406, 111
Kaplan, D. L., van Kerkwijk, M. H., & Anderson, J. 2002, ApJ, 571, 447 Popov, S. B., Turolla, R., Prokhorov, M. E., Colpi, M., & Treves, A. 2005,
Kaplan, D. L., van Kerkwijk, M. H., Marshall, H. L., et al. 2003, ApJ, 590, 1008 Ap&SS, 299, 117
Kaplan, D. L., van Kerkwijk, M. H., & Anderson, J. 2007, ApJ, 660, 1428 Popov, S. B., Turolla, R., & Possenti, A. 2006, MNRAS, 369, L23
Kondratiev, V. I., Burgay, M., Possenti, A., et al. 2008, in 40 Years of Posselt, B., Popov, S. B., Haberl, F., et al. 2007, Ap&SS, 308, 171
Pulsars: Millisecond Pulsars, Magnetars and More, ed. C. Bassa, Z. Wang, Posselt, B., Popov, S. B., Haberl, F., et al. 2008, A&A, 482, 617
A. Cumming, & V. M. Kaspi, AIP Conf. Ser., 983, 348 Predehl, P., & Schmitt, J. H. M. M. 1995, A&A, 293, 889
Lai, D. 2001, Rev. Modern Phys., 73, 629 Ramsay, G., & Cropper, M. 2004, MNRAS, 347, 497
Landolt, A. U. 1992, AJ, 104, 340 Ramsay, G., Cropper, M., Wu, K., et al. 2004, MNRAS, 350, 1373
Letcavage, R., Rutledge, R. E., Fox, D. B., & Shevchuk, A. H. 2009, in American Richards, G. T., Fan, X., Schneider, D. P., et al. 2001, AJ, 121, 2308
Astronomical Society Meeting Abstracts, 213, 470.03 Rutledge, R. E., Fox, D. W., Bogosavljevic, M., & Mahabal, A. 2003, ApJ, 598,
Maccacaro, T., Gioia, I. M., Zamorani, G., et al. 1982, ApJ, 253, 504 458
Madau, P., & Blaes, O. 1994, ApJ, 423, 748 Rutledge, R. E., Fox, D. B., & Shevchuk, A. H. 2008, ApJ, 672, 1137
Malofeev, V. M., Malov, O. I., & Teplykh, D. A. 2007, Ap&SS, 308, 211 Schneider, D. P., Hall, P. B., Richards, G. T., et al. 2005, AJ, 130, 367
Mateos, S., Warwick, R. S., Carrera, F. J., et al. 2008, A&A, 492, 51 Schwarz, H. E., Ashe, M. C., Boccas, M., et al. 2004, in SPIE Conf. Ser. 5492,
McLaughlin, M. A., Rea, N., Gaensler, B. M., et al. 2007, ApJ, 670, 1307 ed. A. F. M. Moorwood, & M. Iye, 564
Motch, C., Guillout, P., Haberl, F., et al. 1997, A&A, 318, 111 Schwope, A. D., Hasinger, G., Schwarz, R., Haberl, F., & Schmidt, M. 1999,
Motch, C., Zavlin, V. E., & Haberl, F. 2003, A&A, 408, 323 A&A, 341, L51
Motch, C., Sekiguchi, K., Haberl, F., et al. 2005, A&A, 429, 257 Smith, N. 2006, ApJ, 644, 1151
Motch, C., Michel, L., & Pineau, F. X. 2007, in Astronomical Data Analysis Smith, N., & Brooks, K. J. 2007, MNRAS, 379, 1279
Software and Systems XVI, ed. R. A. Shaw, F. Hill, & D. J. Bell, ASP Conf. Stetson, P. B. 1987, PASP, 99, 191
Ser., 376, 699 Stocke, J. T., Morris, S. L., Gioia, I. M., et al. 1991, ApJS, 76, 813
Motch, C., Pires, A. M., Haberl, F., Schwope, A., & Zavlin, V. E. 2008, in 40 Story, S. A., Gonthier, P. L., & Harding, A. K. 2007, ApJ, 671, 713
Years of Pulsars: Millisecond Pulsars, Magnetars and More, ed. C. Bassa, Tody, D. 1986, in SPIE Conf. Ser. 627, ed. D. L. Crawford, 733
Z. Wang, A. Cumming, & V. M. Kaspi, AIP Conf. Ser., 983, 354 Toropina, O. D., Romanova, M. M., Toropin, Y. M., & Lovelace, R. V. E. 2003,
Motch, C., Pires, A. M., Haberl, F., Schwope, A., & Zavlin, V. E. 2009, A&A, ApJ, 593, 472
497, 423 Treves, A., & Colpi, M. 1991, A&A, 241, 107
Neuhäuser, R. 2001, Astron. Nachr., 322, 3 Treves, A., Campana, S., Chieregato, M., et al. 2007, Ap&SS, 308, 167
Page, K. L., Schartel, N., Turner, M. J. L., & O’Brien, P. T. 2004, MNRAS, 352, van Kerkwijk, M. H., & Kaplan, D. L. 2007, Ap&SS, 308, 191
523 van Kerkwijk, M. H., & Kaplan, D. L. 2008, ApJ, 673, L163
Perna, R., Narayan, R., Rybicki, G., Stella, L., & Treves, A. 2003, ApJ, 594, 936 Voges, W., Aschenbach, B., Boller, T., et al. 1999, A&A, 349, 389
Pineau, F.-X., Derriere, S., Michel, L., & Motch, C. 2008, in AIP Conf. Ser., Watson, M. G., Schröder, A. C., Fyfe, D., et al. 2009, A&A, 493, 339
1082, 15 Zampieri, L., Campana, S., Turolla, R., et al. 2001, A&A, 378, L5
Pires, A. M., & Motch, C. 2008, in 40 Years of Pulsars: Millisecond Pulsars, Zane, S., Turolla, R., Stella, L., & Treves, A. 2001, ApJ, 560, 384
Magnetars and More, ed. C. Bassa, Z. Wang, A. Cumming, & V. M. Kaspi, Zane, S., Haberl, F., Cropper, M., et al. 2002, MNRAS, 334, 345
AIP Conf. Ser., 983, 363 Zane, S., de Luca, A., Mignani, R. P., & Turolla, R. 2006, A&A, 457, 619