© ESO, 2013

1. Introduction

Be/X-ray binaries (BeXRBs, Reig 2011) are the dominant subclass of high-mass X-ray binaries in the Small Magellanic Cloud (SMC) with more than 100 known systems (including candidates). These systems comprise of a neutron star (NS) that accretes matter from the circumstellar decretion disc of an emission-line star of spectral class B (Be star) leading to, often transient, X-ray emission.

The BeXRB binary SXP 1062 was discovered by Hénault-Brunet et al. (2012) using Chandra and XMM-Newton observations of the star-forming region NGC 602 (Oskinova et al. 2013b). Amongst other BeXRBs in the SMC, SXP 1062 is outstanding owing to: (i) The position in the wing of the SMC, where a younger stellar population is found and less BeXRBs are known compared to the bar of the SMC. (ii) A relatively long spin period of the NS of P_s ~ 1062 s, making it the second slowest known pulsar of the SMC (after SXP 1323, Haberl & Pietsch 2005). (iii) A rapid spin-down of the NS of Ṗ_s = 0.26 s d^-1 (Haberl et al. 2012) that was observed over the interval between 2010 Mar. 25 and Apr. 12 (18 days), whereas in general spin-up is observed during enhanced accretion. (iv) A robust correlation with a supernova remnant (SNR) was found by Hénault-Brunet et al. (2012). Owing to the low density of both BeXRBs and SNRs in the wing of the SMC, this correlation is unlikely by chance. The SNR is clearly seen in optical emission lines (Hα and [O iii]). Additional radio emission and a detailed analysis of the X-ray emission of the SNR can be found in Haberl et al. (2012). The SNR constrains the age of the NS to 10 000−25 000 years, making SXP 1062 a relatively young BeXRB.

These peculiar properties have made SXP 1062 the subject of several theoretical investigations, with the main issue of explaining the long spin period of this young NS. Suggested scenarios include an initially slow spin period (Haberl et al. 2012), an initially and/or presently strong magnetic field (Popov & Turolla 2012; Fu & Li 2012), and the accretion of magnetised matter (Ikhsanov 2012).

Monitoring of Galactic accreting pulsars revealed that individual NSs can switch between spin-up and spin-down phases over short time scales down to a few days and that the NS spins up during intervals of high accretion and spins down over X-ray quiescent periods (Bildsten et al. 1997). This raises the question if the large spin-down of SXP 1062 was observed by chance and if the source continued to spin down with such a high Ṗ_s.

An outburst of SXP 1062 in 2012 was found in the optical with OGLE and in X-rays with the Swift satellite. In this paper, we present our analysis of an XMM-Newton target-of-opportunity observation of SXP 1062 in Oct. 2012, which allows to measure the long-term evolution of the spin period of the NS on a longer time scale of ~900 days since the first XMM-Newton measurement. All uncertainties in this paper are given for 1σ confidence.

2. Observations and data reduction

On 2012 Oct. 9, SXP 1062 was found in X-ray outburst with the Swift satellite. The evolution of the X-ray flux was followed in a monitoring campaign since then (TargetID: 32 580). The XRT spectra were reduced from the cleaned level-3 event files within a circle placed on the source with the ftool¹xselect. Background spectra were created from a nearby point-source-free circular extraction region. The ancillary response files were calculated with xrtmkarf.

The detection of SXP 1062 in X-ray bright state allowed us to ask for an XMM-Newton (Jansen et al. 2001) follow-up observation (ObsID: 0700381801) that was performed on 2012 Oct. 14. SXP 1062 was observed on-axis with all EPIC instruments in full-frame mode using the thin filter for EPIC-pn (Strüder et al. 2001) and the medium filter for both EPIC-MOS (Turner et al. 2001). We processed the data using the XMM-Newton SAS 12.0.1². Because the flaring background was at moderate level (below 8 and 3 counts s^-1 arcmin^-2 for pn and MOS, respectively), no temporal screening of the data was necessary. This yields an exposure of 26.6, 31.1, and 31.2 ks for EPIC-pn, -MOS1, and -MOS2, respectively. Energy spectra and time series of the source and the background were extracted, using the same selection regions as described in Haberl et al. (2012, see their Fig. 1, region A and B) such that emission of the SNR will be subtracted. We used single- and double-pixel events from EPIC-pn and single- to quadruple-pixel events from EPIC-MOS. For energy spectra we rejected events with FLAG ≠ 0. This resulted in 24 810, 8009, and 8908 net counts in the (0.2−12.0) keV band for the three instruments. We binned the EPIC spectra to have a signal-to-noise ratio of ≥5 for each bin. Spectra from the Reflection Grating Spectrometer (RGS, den Herder et al. 2001) with an exposure of 31.6 ks were extracted with rgsproc. We used events within a cross-dispersion point-spread-function width of 90%, which excludes roughly half of the area covered by the SNR. The background events were selected for a point-spread-function width >99%, to exclude emission from the NS and the SNR. The resulting spectra contain 277 (RGS1) and 380 (RGS2) net counts in the (0.35−2.0) keV band. The RGS spectra were binned with 25 net cts bin^-1. Although these spectra have low statistics, they provide the first grating X-ray data of SXP 1062.

Fig. 1 Upper panel: OGLE-IV I-band light curve. Dashed lines indicate the time of optical spectroscopy observations. Middle panel: X-ray fluxes in the (0.2−10.0) keV band from Swift (open squares) and XMM-Newton (open circles) including the 2009 slew-survey data and the 2010 measurements. Lower panel: NS spin period as measured with XMM-Newton.

We used the Southern African Large Telescope (SALT, Buckley et al. 2006a,b; O’Donoghue et al. 2006 to collect long slit spectra of SXP 1062 simultaneous with the XMM-Newton observation. As part of the SALT program 2012-1-RSA_UKSC-003 (PI: Schurch) aimed at studying outbursts from X-ray binary systems, we obtained spectra of SXP 1062 during the evening of 2012 Oct. 13th and morning of the 14th. These data were collected using the Robert Stobie Spectrograph (RSS, Burgh et al. 2003; Kobulnicky et al. 2003). We used the PG2300 grating in two position angles (30.500°, and 48.875°) to achieve coverage from (3835−4924) Å in the blue and (6082−6925) Å in the red, respectively. To achieve the highest resolution we used a 0.6′′ wide slit. Observations were prebinned by a factor 2 producing a spectral resolution of 0.33 Å per binned pixel. Two 400 s observations were taken in the blue on MJD 56 213.99 followed by a 180 s exposure of a ThAr arc lamp. A single 180 s exposure was taken in the red on MJD 56 214.01 followed by a 90 s exposure of an Ar arc lamp. All observations were flat fielded using quartz tungsten halogen (QTH) flats taken immediately after the blue and before the red observations. The image quality during the observations was approaching 3′′. An hour after the observations a spectrophotometric standard star (LTT1020) was observed for both the blue and red setups. The SALT data was first processed using the SALT pipeline, PySALT (Crawford et al. 2010). This performs overscan, gain, cross-talk corrections, and mosaicing of the three 2048 × 4096 CCDs. Subsequent flatfielding, background subtraction, wavelength calibration and extraction of 1D spectra were performed in IRAF v2.16 (Tody 1993). The final spectra were first redshift corrected by 150 km s^-1 and normalised (after being averaged in the case of the two blue 400 s exposures).

An optical monitoring of SXP 1062 is provided by the Optical Gravitational Lensing Experiment (OGLE, Udalski et al. 2008) that covers the source regularly with I-band photometry since the beginning of phase IV (2010 Aug. 6). The data are available in the X-Ray variables OGLE Monitoring (XROM, Udalski 2008) system and presented in the upper panel of Fig. 1.

3. Analyses and results

3.1. X-ray energy spectrum

Spectral analysis was performed with xspec (Arnaud 1996) version 12.7.0u. We used the same model, as described in Haberl et al. (2012), i.e. an absorbed power law where we assume Galactic photoelectric absorption with a column density of N_H,gal = 6 × 10²⁰ cm^-2 with solar abundances according to Wilms et al. (2000), and a free column density, N_H,smc, with abundances set to 0.2 solar accounting for absorption by the interstellar medium of the SMC or within the BeXRB system. The normalisations between the individual instruments were found to be consistent, and we fitted the same model to all five spectra simultaneously. The spectrum and best-fit model are presented in Fig. 2a and the best-fit parameters are listed in Table 1. Formally, the power-law model describes the data well with a reduced ${\mathit{\chi }}_{\mathit{\nu }}^{\mathrm{2}}$ of 1.01. However, in other BeXRBs, a soft excess (Hickox et al. 2004) and iron fluorescent emission are believed to contribute to the X-ray spectrum.

A possible Fe K_α emission line broadened below the instrumental resolution was investigated by fitting an additional Gaussian emission-line profile with central energy of E_c = 6.4 keV, line width of σ = 0 and free normalisation. We obtain a 3.1σ evidence for a line with normalisation of N = (3.1 ± 1.0)    ×    10^-6 photons cm^-2 s^-1, corresponding to an equivalent width of EW = (−40 ± 13) eV. Analogously, we derive for ionised Fe xxv (E_c = 6.7 keV) a 3σ upper limit of N < 2.9 × 10^-6 photons cm^-2 s^-1 and N < 3.1 × 10^-6 photons cm^-2 s^-1 for Fe xxvi (E_c = 7.0 keV).

Fig. 2 a) Energy spectrum of SXP 1062 as observed on 2012 Oct. 14 with EPIC-pn (black), -MOS1 (red), -MOS2 (green), RGS1 (blue), and RGS2 (magenta) together with the best-fit model (solid line) of an absorbed power law. b) Same as a), but fitted with additional contribution of a multi-temperature disc black-body model and Fe Kα emission line, shown by dashed lines. The RGS data are omitted for clarity. c) The residuals in units of σ for the model from a) with a higher binning by a factor of 5. d) Same as c), but for the model from b). e) The residuals for the model of a), binned with only 5 net cts bin-1. Vertical lines mark the energies of prominent emission lines. f) 3σ upper limits for the equivalent width (EW) of a Gaussian absorption line.

To investigate a possible soft excess and to demonstrate systematic uncertainties of the power-law and absorption parameters we fitted additional thermal components. Adding a black-body (BB) emission component to the model results in a slight improvement of the fit with an f-test probability of 2.3 × 10^-7. This formally proves the significance of this component, however see Protassov et al. (2002) for limitations of the f-test. We further fitted a multi-temperature accretion disc model (DiskBB) and a collisionally ionised plasma model with 0.2 solar abundances (APEC). The results are presented in Table 1. The possible contribution of the disc black-body model and Fe K_α line are demonstrated in Fig. 2b, where we omit the RGS spectra for clarity. The residuals of this model, rebinned by a factor of 5, are compared with the residuals of the simple power-law model in Figs. 2d and c. We find that e.g. in the case of a disc black-body a soft excess can only contribute with a luminosity of $\mathrm{7.}{\mathrm{5}}_{-1.1}^{\mathrm{+}\mathrm{0.8}}\mathrm{\times }{\mathrm{10}}^{\mathrm{34}}$ erg s^-1, i.e. $\mathrm{2.}{\mathrm{5}}_{-0.4}^{\mathrm{+}\mathrm{0.3}}\mathrm{\%}$ of the unabsorbed luminosity in the (0.2−10.0) keV band.

To check for spectral features in the RGS grating spectra, we plot the residuals of the power-law fit with a binning of only 5 net cts bin^-1. We find clear indications of emission lines of the helium-like triplets of O vii (0.561−0.574 keV), Ne ix (0.905−0.922 keV), and Mg xi (1.33−1.35 keV) as well as hydrogen-like N vii (0.500 keV), O viii (0.654 keV), and Mg xii (1.47 keV) and neon-like Fe xvii (0.725, 0.826 keV), as marked in Fig. 2e. Due to the low statistics in the spectra, we can only roughly constrain the fluxes of the most convincing lines to $\mathrm{1.}{\mathrm{6}}_{-0.8}^{\mathrm{+}\mathrm{0.9}}\mathrm{\times }{\mathrm{10}}^{-5}$ photons cm^-2 s^-1 (O vii), $\mathrm{4.}{\mathrm{9}}_{-2.5}^{\mathrm{+}\mathrm{3.4}}$ × 10^-6 photons cm^-2 s^-1 (O viii), and $\mathrm{1.}{\mathrm{3}}_{-0.4}^{\mathrm{+}\mathrm{0.5}}\mathrm{\times }{\mathrm{10}}^{-5}$ photons cm^-2 s^-1 (Ne ix). Here we used the unbinned data and C statistics (Cash 1979).

The peculiar properties of SXP 1062 were suggested to be caused by a strong magnetic field (B ~ 10¹⁴ G, Fu & Li 2012), possibly leading to a proton cyclotron feature in the observed energy band. Adding a Gaussian absorption line to the power-law model, we find no significant contribution of this feature. Stepping the absorption line through the complete spectrum, we estimate 3σ upper limits for EW as shown in Fig. 2f, where we assume a line width of σ = 0.2E_c.

Fig. 3 Power-density spectrum of SXP 1062. The best-fit frequency ω = 934 μHz and harmonics are marked by vertical lines.

Fig. 4 Upper figure: pulse profile of the time series, merged from all EPIC instruments in different energy bands for the 2012 observation (black) and an observation in 2010 (red). The individual light curves are background subtracted and normalised to the average count rate (from top to bottom: 0.19, 0.40, 0.44, 0.35, 1.37 cts s-1 in 2012 and 0.020, 0.044, 0.052, 0.037, 0.15 cts s-1 in 2010). Lower figure: hardness ratios as function of pulse phase.

3.2. X-ray pulsations

A fast Fourier transformation of the merged EPIC time series (42 231 cts including background) reveals the pulsation of the NS and various harmonics as presented in Fig. 3. Using a Bayesian detection method as described by Gregory & Loredo (1996) and Haberl et al. (2008), the most probable spin period during the recent XMM-Newton observation is derived to P_s = (1071.01 ± 0.16) s.

The background-subtracted normalised folded light curves are shown in Fig. 4 for the total (0.2−10.0) keV band and the standard sub-bands (0.2−0.5) keV, (0.5−1.0) keV, (1.0−2.0) keV, (2.0−4.5) keV, and (4.5−10.0) keV, where we merged the first two bands to increase the statistics. The HRs are defined by HR_i = (R_i+1 − R_i)/(R_i+1 + R_i) with R_i denoting the background-subtracted count rate in the XMM-Newton standard energy band i (with i from 1 to 4). In Fig. 4, we also show the pulse profile measured in 2010 (ObsID 0602520201), folded with 1062.4 s. The phase offsets are set to have a common maximum in the total-band light curves.

Fig. 5 SALT blue and red (left and right respectively) smoothed spectra of SXP 1062. Clearly visible are the chip gaps between the three CCDs. Dotted lines indicate: Balmer lines (black), He i (green), He ii (light blue), Silicon (red) and other metal lines (dark blue).

3.3. Long-term X-ray light curve

The position of SXP 1062 was covered with the Einstein satellite, but no source is reported in the catalogue of Wang & Wu (1992). Also the ASCA catalogue of Yokogawa et al. (2003) does not contain SXP 1062, which was in the FoV of an observation performed on 1999 May 28 to 29.

We note that both catalogues list a nearby source (No. 67 of Wang & Wu 1992 and No. 105 of Yokogawa et al. 2003). The angular separation of ~4.4′ is too large for a correlation with SXP 1062 with respect to the angular resolution of both satellites. In the recent XMM-Newton observation, as well as in the 2010 observations, we do not find a bright source at this position. Hence, the nearby source might be an X-ray transient. In the case of another BeXRB, the closest possible counterparts are 2dFS 3857 (B0-5V star, 60′′ angular separation), and 2MASS J01281201-7330235 (B3-5III star showing near-infrared variability, 80′′).

Three pointed ROSAT observations of SMC X-1 covered the position of SXP 1062 between 1991 and 1993 (MJD 48 536.2, 48 895.7, and 49 141.0). For these observations, no detection at the position of SXP 1062 is reported in the catalogue of Haberl et al. (2000). Assuming 10 counts as a typical detection limit and the same spectral shape as measured in the recent XMM-Newton observation, we derive upper limits of 1.4, 2.7, and 2.1 × 10^-13 erg cm^-2 s^-1, respectively, which are the lowest limits for the X-ray flux reported so far.

As noted by Hénault-Brunet et al. (2012), the very first X-ray detection of SXP 1062 is listed in the XMM-Newton slew-survey catalogue (Saxton et al. 2008,Release 1.5). The source (XMMSL1 J012746.2-733304) was detected on 2009 Nov. 16 with (4.2 ± 1.5) cts, which according to the spectrum from above, results in a flux of (3.0 ± 1.1) × 10^-12 erg cm^-2 s^-1.

The recent Swift/XRT monitoring allows to follow the evolution of the X-ray outburst in 2012. Since the X-ray spectrum is measured well with XMM-Newton and we find no strong variations of the spectral shape compared to the 2010 observations, we fixed the spectral shape to the PL+DiskBB+Fe model and calculated fluxes for each Swift spectrum, using C statistics. The derived fluxes are listed in Table 2. The X-ray light curve is presented in Fig. 1 and compared to the optical light curve from OGLE and the evolution of the spin period of the NS.

3.4. Optical spectroscopy

Fig. 6 Left: Hα as seen with SALT (normalised) and according modelling with two Gaussians and continuum. The lowest line gives the residuals. The 2dF spectrum is shown on the top for comparison. Middle: same as before, but for Hβ. Right: Hγ as seen with VLT FLAMES (top), 2dF (middle) and SALT (bottom).

Figure 5 shows the fully reduced spectra after being smoothed using a boxcar average of 3 data points. The Hα and Hβ line are shown in more detail in Fig. 6. Neither line could be fitted well using one Gaussian alone. For Hα, broad wings are seen in addition. Hβ clearly exhibits a double-peaked line profile that is also seen in the pre-subtraction images. Fitting both lines with two Gaussians each and using the unsmoothed data, we derive equivalent widths of the Hα and Hβ of EW_Hα = (−26.65 ± 0.09) Å and EW_Hβ = (−2.40 ± 0.29) Å, respectively. Re-analysing previous 2dF spectra (Evans et al. 2004; Hénault-Brunet et al. 2012) with the same method, results in EW_Hα = (−22.02 ± 0.054) Å and EW_Hβ = (−1.58 ± 0.10) Å.

4. Discussion

We analysed an XMM-Newton observation of SXP 1062 and find that the spin period of the NS further increased to 1071 s. Since the last XMM-Newton observation, performed on 2010 Apr. 12, 915 days before the most recent observation, we obtain an average Ṗ_s = (2.27 ± 0.44) s yr^-1. This is factor of 40 less than observed during the 18 day baseline in 2010. By assuming, that this average Ṗ_s is representative for the NS since birth over 25 kyr, one derives an initial P_s of ~20 s. If one assumes that the NS is born with a P_s of a few ms, a high initial magnetic field strength of ~10¹⁴ G (compared to a few 10¹² G as typically found for BeXRBs, e.g. Pottschmidt et al. 2012) and an efficient propeller mechanism are needed (e.g. Fu & Li 2012) to slow the pulsar down within a few kyr to be close to the equilibrium period now.

The more moderate long-term spin-down rate suggests that the source was observed during high spin down phase in 2010 by chance and that the high spin-down rate is not an intrinsic feature of this source. This supports the hypothesis that SXP 1062 is rotating close to its equilibrium period, with spin-up during periods of high accretion rates and spin-down at other times. In this case, we would expect to observe spin-up of the NS occasionally as well, however there is no observational evidence so far. The recent data do not allow to conclude on the detailed evolution as periods of spin-up and -down may have alternated as seen in other BeXRBs. Especially the current Ṗ_s cannot be determined from our single observation and the NS might even have shown spin-up during the X-ray outburst monitored by Swift. A denser X-ray coverage of the source is highly desirable to determine the evolution of P_s and Ṗ_s on a longer time scale and to constrain individual models.

Comparing the pulse profiles of 2010 and 2012, we find indications for a double peaked light curve in 2010, whereas in 2012 the light curve reveals only one clear peak at energies above 1 keV.

The shape of the X-ray spectrum is close to the 2010 measurement, but might contain a soft excess and an Fe K line. If the soft excess is caused by black-body emission, the black-body radius agrees with the radius of an NS and might originate from the NS surface. But also reprocessing of hard X-rays from the NS in nearby material like an accretion disc is possible, as we can only measure a lower limit of the inner disc radius, depending on the inclination (R ~ cos^−1/2Θ). Estimating the inner disc radius according to Hickox et al. (2004) yields R = (L_X/4πσT⁴)^1/2 = 8.5 km.

The high-resolution X-ray spectra from the RGS show indications of emission lines. The emission lines might be explained by the SNR around SXP 1062. However, assuming a homogeneous surface brightness of the SNR and typical SMC abundances of Z = 0.2  Z_⊙, the best-fit plasma model for the SNR of Haberl et al. (2012) accounts only for a few percent of the observed O vii line flux. This points to an additional hot plasma component around the NS as it was e.g. observed in SXP 1323 (Haberl & Pietsch 2005). Deeper X-ray observations of the source in an X-ray bright state are needed here, but we note that the line energies of the helium-like lines point to resonance lines rather than forbidden lines and that the emission of the observed He-like species is stronger than H-like like.

Proton cyclotron absorption lines (e.g. Zane et al. 2001) can be significant features in the X-ray spectra of NSs with high magnetic fields as observed with present-day observatories and might appear in the X-ray spectrum of SXP 1062 (Fu & Li 2012). Neither in the residuals of the simple power-law fit (Fig. 2c) nor by adding an absorption-line component to the model (Sect. 3.1) we can find a convincing contribution of an absorption line. A weak absorption line with equivalent width down to EW > 10 eV can be excluded only in the (0.84−1.17) keV band, whereas a stronger line with EW > 200 eV can be ruled out in the (0.28−7.5) keV band. According to the relation

$\mathit{B}\mathrm{\simeq }\mathrm{1.59}\mathrm{\times }{\mathrm{10}}^{\mathrm{14}}{\mathit{z}}_{\mathrm{G}}^{-1}\left(\frac{{\mathit{E}}_{\mathrm{c}}}{\mathrm{keV}}\right)\mathrm{G}\mathit{,}$where z_G is the gravitational redshift, this corresponds to magnetic fields in the range of (1.3−1.9) × 10¹⁴ G and (0.44−12) × 10¹⁴ G, respectively, for z_G = 1. However, we note that also no convincing detection of a proton cyclotron-absorption feature has been reported for magnetars up to now (Mereghetti 2008).

The long term I-band light curve, as measured with OGLE-IV has been examined by Schmidtke et al. (2012a) and exhibits short term variability of ~9 days, that is likely an alias of sinusoidal modulations of (0.9007 ± 0.0005) d, caused by non-radial pulsations. In addition, two strong outbursts are seen reaching maximum emission around MJD 55 500 (ΔI ~ 0.6 mag) and between MJD 56 151−56 173 (ΔI ≳ 0.15 mag). Schmidtke et al. (2012b) suggested the time-difference between the outbursts to be the likely orbital period P_o = (656    ±    2) d. This is supported by the FRED-like shape (Bird et al. 2012) of the optical outbursts. If so, the X-ray emission in 2010 cannot be caused by a type-I X-ray outburst that can occur during periastron passage of the NS only.

This might be explained by a persistent X-ray state that is also favoured by the fact, that SXP 1062 was always detected above a flux of ~10^-12 erg cm^-2 s^-1 since its discovery. The pulsar must have entered this state after the last ROSAT observation. Comparing the deepest ROSAT upper limit of F_X ≤ 1.4 × 10^-13 erg cm^-2 s^-1 with the average flux of the 2010 XMM-Newton observations reveals an X-ray variability by a factor of 10 and by comparing with the brightest Swift detection, one obtains a maximum X-ray variability by a factor of at least 60. The X-ray flux is also moderately (some 10%) variable on short time scales (days), as measured with Swift and previously with Chandra (cf. Fig. 2 of Oskinova et al. 2013a).

The X-ray outburst around MJD 56 200 may correlate with the second optical outburst. Because our Swift observations did not catch the initial stage of X-ray outburst we cannot judge about the outburst duration and maximum emission. The first Swift observation was performed 58 days after the first measured I-band increase. Regarding the long orbital period, this time interval might still be consistent with a type-I X-ray outburst. In this scenario, the NS accretes matter during periastron passage, leading to X-ray emission, while tidal forces affect the circumstellar disc. A sudden but local perturbation in the disc geometry, such as e.g. warping a part of the disc or the formation of a quasi-steady accretion disc around the NS, might cause the sharp increase in total optical luminosity. Crude estimates show that the observed increase in I-band magnitude can be achieved if the projected disc area increased by about 20%.

The optical counterpart of SXP 1062, the B0 III e star 2dFS 3831, was already observed spectroscopically by Evans et al. (2004) in Sep. 1998 (blue) and Sep. 1999 (red), the later ~4 months after the ASCA observation. On 2010 Oct. 25, 2dFS 3831 was observed spectroscopically with the VLT FLAMES instrument, only 4 days before the first optical outburst. These spectra are discussed in Hénault-Brunet et al. (2012). Our reported SALT spectrum was taken during the decline of the second outburst (dashed lines in Fig. 1).

The post-outburst SALT spectrum (see Fig. 5) is similar to those previously published and show that the Be star disc was at large not affected during the X-ray outburst, as it can be the case in type-II outbursts. The hydrogen Balmer lines have contributing emission filling them in, most likely as a result of a large circumstellar disc. This is also corroborated by a non-split shape of the Hα line, indicating the disc large in size and/or its inclination is small (Hummel & Vrancken 1995). However, the Hβ and Hγ emission lines show split shapes.

Despite overall similarity, there are some differences in spectra obtained at different epochs. The most apparent is the increasing strength of the Hγ line occurring between the 2dF and the VLT measurement. In the SALT spectrum, the Hγ emission line is still seen. The VLT spectrum shows a stronger red-ward wing contribution, while it is more symmetrical in the SALT spectrum pointing to V/R variability (see Fig. 6, right).

Another evidence for an increasing size or density of the disc is given by the increasing equivalent width of the Hα and Hβ emission lines between the 2dF and the SALT measurement by (21.0 ± 0.5)% and (52 ± 21)%, respectively. Thus, the spectroscopic measurements suggest that the decretion disc has been growing over the last ~14 years. This may help to explain the non-detections of SXP 1062 by earlier X-ray missions (Sect. 3.3), because only a small or no decretion disc might have existed in the past (at least before 1999).

The optical and X-ray spectra of SXP 1062 are similar to those observed from other BeXRBs. The unusually slow spin period and its evolution (so far no spin-up was observed), hence, are even more surprising. The young long-period pulsar SXP 1062 continues to challenge the present models of accreting pulsars.

5. Conclusions

We presented the analysis of a recent XMM-Newton observation of SXP 1062 and discuss the results in the context of complementary Swift/XRT and optical data. The main results and conclusions are as follows:

• 1.

  The NS continued to spin down to a spin period ofP_s = (1071.01 ± 0.16) s as measured on 2012 Oct. 14. This implies an average Ṗ_s = (2.27 ± 0.44) s yr^-1 during a baseline of 915 days.

• 2.

  We do not see significant spectral changes in the X-ray spectrum, compared to the 2010 observations, despite of an indication of a soft excess and Fe K_α at higher luminosities (L_X = 3 × 10³⁶ erg s^-1).

• 3.

  We see indications of emission lines in the RGS spectra from prominent thermal emission lines. The SNR can unlikely account completely for this component.

• 4.

  No convincing indication of proton cyclotron-absorption lines are found. If we demand an absorption line with EW > 200 eV, this would exclude a magnetic field of the NS between 4.4 × 10¹³ G and 1.2 × 10¹⁵ G.

• 5.

  The X-ray light curve suggests that SXP 1062 is currently in a persistent X-ray emitting state with a luminosity around L_X = 8 × 10³⁵ erg s^-1. Optical and X-ray outbursts might correlate and be caused by type-I outbursts.

• 6.

  The optical spectra obtained during the X-ray outburst are morphologically similar to the pre-outburst spectra obtained in 1998/99 and 2010. Strong Balmer emission lines, originating from the circumstellar disc of the Be star, are present during all observations. This indicate that the disc is large and not strongly affected by the X-ray outburst.

• 7.

  The evolution of the Balmer emission lines provides evidence for an increase of the circumstellar disc over the last ~14 years. This can explain the observed increase in the persistent X-ray luminosity of SXP 1062 during this time.

Acknowledgments

We thank the XMM-Newton team for scheduling the ToO observation. The XMM-Newton project is supported by the Bundesministerium für Wirtschaft und Technologie/Deutsches Zentrum für Luft- und Raumfahrt (BMWI/DLR, FKZ 50 OX 0001) and the Max-Planck Society. We acknowledge the use of public data from the Swift data archive and thank the Swift team for accepting and scheduling the ToO observation. Some of the observations reported in this paper were obtained with the Southern African Large Telescope (SALT). The OGLE project has received funding from the European Research Council under the European Community’s Seventh Framework Programme (FP7/2007-2013)/ERC grant agreement No. 246678 to AU. MPES is funded through the Claude Leon Foundation Postdoctoral Fellowship program and the National Research Foundation. L.M.O. and RS acknowledge support from the BMWI/DLR grants FKZ 50 OR 1302 and FKZ 50 OR 0907, respectively. LMO acknowledges fruitful discussions within the international team at ISSI (International Space Science Institute) in Bern, especially with M.J. Torrjon and P. Kretschmar. J.S.G. thanks the University of Wisconsin-Madison College of Letters & Science for partial support of this research and the campus for funding the Wisconsin participation in SALT.

References

All Tables

All Figures

	Fig. 1 Upper panel: OGLE-IV I-band light curve. Dashed lines indicate the time of optical spectroscopy observations. Middle panel: X-ray fluxes in the (0.2−10.0) keV band from Swift (open squares) and XMM-Newton (open circles) including the 2009 slew-survey data and the 2010 measurements. Lower panel: NS spin period as measured with XMM-Newton.
In the text

Fig. 2 a) Energy spectrum of SXP 1062 as observed on 2012 Oct. 14 with EPIC-pn (black), -MOS1 (red), -MOS2 (green), RGS1 (blue), and RGS2 (magenta) together with the best-fit model (solid line) of an absorbed power law. b) Same as a), but fitted with additional contribution of a multi-temperature disc black-body model and Fe Kα emission line, shown by dashed lines. The RGS data are omitted for clarity. c) The residuals in units of σ for the model from a) with a higher binning by a factor of 5. d) Same as c), but for the model from b). e) The residuals for the model of a), binned with only 5 net cts bin-1. Vertical lines mark the energies of prominent emission lines. f) 3σ upper limits for the equivalent width (EW) of a Gaussian absorption line.

In the text

	Fig. 3 Power-density spectrum of SXP 1062. The best-fit frequency ω = 934 μHz and harmonics are marked by vertical lines.
In the text

Fig. 4 Upper figure: pulse profile of the time series, merged from all EPIC instruments in different energy bands for the 2012 observation (black) and an observation in 2010 (red). The individual light curves are background subtracted and normalised to the average count rate (from top to bottom: 0.19, 0.40, 0.44, 0.35, 1.37 cts s-1 in 2012 and 0.020, 0.044, 0.052, 0.037, 0.15 cts s-1 in 2010). Lower figure: hardness ratios as function of pulse phase.

In the text

	Fig. 5 SALT blue and red (left and right respectively) smoothed spectra of SXP 1062. Clearly visible are the chip gaps between the three CCDs. Dotted lines indicate: Balmer lines (black), He i (green), He ii (light blue), Silicon (red) and other metal lines (dark blue).
In the text

	Fig. 6 Left: Hα as seen with SALT (normalised) and according modelling with two Gaussians and continuum. The lowest line gives the residuals. The 2dF spectrum is shown on the top for comparison. Middle: same as before, but for Hβ. Right: Hγ as seen with VLT FLAMES (top), 2dF (middle) and SALT (bottom).
In the text