© The Authors 2024

1. Introduction

Since the 1960s, it has been clear that neutron stars (NSs) and black holes (BHs) in binary systems could be detected through the orbital motion induced on their companion stars (e.g., Guseinov & Zel’dovich 1966). However, these compact stars were first found in different ways: NSs were discovered as radio pulsars (Hewish et al. 1968), while stellar-mass BHs were identified through observations of accretion-powered X-ray binaries (Bolton 1972; Webster & Murdin 1972). The typical steps that led to the identification of most known stellar-mass BHs were the discovery of luminous (transient) X-ray sources, the identification of their optical counterparts, and subsequent spectroscopic observations to determine the binary parameters, and in particular to constrain the mass of the compact object. At present, a few tens of stellar-mass BHs in X-ray binaries have been identified in this way (e.g., Corral-Santana et al. 2016).

The appearance of large stellar surveys in recent years has enabled systematic searches for binary systems with only one visible component and a companion whose mass is compatible with that expected for NSs or BHs. Particularly relevant in this regard is the contribution of the Gaia mission (Gaia Collaboration 2016), which has covered ∼10⁹ Galactic stars for several years. Not only can Gaia identify binary systems through repeated multiband photometry, but it also has superb astrometrical capabilities that provide precise distance information (at least for relatively bright sources). In the case of wide binaries, only astrometry presently allows the detection and study of their dynamics. Although, the primary source of data for these analyses comes from optical observations, multiwavelength follow-up can be useful to confirm the presence of compact objects.

In the present paper, we present X-ray and UV observations obtained with the Neil Gehrels Swift Observatory of a few binaries selected from optical observations as strong candidates for containing a NS. Four targets were chosen because they are the closest among the candidates found in a sample of astrometric binaries (Andrews et al. 2022). They have long orbital periods (≳1 yr), high eccentricities, and are closer than 250 pc. The other two targets are short-period binaries (P < 1 d) for which the masses were determined from photometric and spectroscopic observations (Yuan et al. 2022; Lin et al. 2023). The noncollapsed components in all these binaries are main sequence stars. The main properties of our targets are summarized in Table 1, where we also indicate the short source names adopted in this paper.

2. Observations and data analysis

Our targets were observed with Swift through Target of Opportunity programs as summarised in Table 2, where we give the total exposure times obtained for the X-ray Telescope (XRT) (the observations were in some cases split over different epochs). The XRT was in photon counting mode. The data were processed using the XRTDAS package (version 5.7) distributed with HEASOFT (version 6.31) using the most recent release of the calibration database (CALDB version 10).

Source counts for the spectral analysis of the only source detected with the XRT (1527+3536) were extracted using a circular region of 30 pixels (70.8″), and the background was estimated using a region with the same shape and size and far from the source position. Spectral analysis was performed using the XSPEC package (version 12.13.0c). The upper limits for the nondetected sources were determined from the images in the 0.3–10 keV and 0.3–1 keV energy ranges using the Living Swift-XRT Point Source (LSXPS) catalog online upper limit server (Evans et al. 2023, 2014), which extracts counts from a 12 pixel region centered on the source position, estimates the background from the appropriate background map, and applies the Kraft et al. (1991) Bayesian method to estimate the upper limit.

All the targets were detected with the UV and Optical Telescope (UVOT), which operated with the filters UVW2 (central wavelength λ = 192.8 nm and width δλ = 65.7 nm) for two sources and with UVM2 ( λ = 224.6 nm, δλ = 49.8 nm) for the other four sources (see Table 3). However, we note that the observations executed with the UVW2 filter are affected by contamination from optical photons caused by the red tail of the filter passband. For a K2V type star, it is estimated that the UV portion is less than ∼20% of the photon total flux (Brown et al. 2010), and so any measured flux should be interpreted as an upper bound on the real flux.

3. Results

In all the XRT observations pointed on 2MASS J152748.48+353657.2, a source was detected (S/N > 10) at coordinates right ascension (RA) = 15^h27^m48$\stackrel{\text{s}}{.}$52, declination (Dec) = +35°36′58$\stackrel{″}{.}$8 with an uncertainty of 3.6″ (90% conf. level). This position is only 1.6″ away from the Gaia position of our target and this is the only source in the error box of the ROSAT source 1RXS J152748.8+353658 (positional error 19″, 1σ radius). We are therefore confident that the source detected by the Swift/XRT is the candidate NS binary.

No evidence for variability was found by comparing the data of the three epochs (see Table 4). Therefore, the subsequent analysis was carried out on the summed data, which provided a total of about 180 source counts in ∼18 ks of exposure. We extracted the source spectrum in the 0.3–10 keV energy range and fitted it applying the Cash statistics. As all the models including an absorption component only provided an upper limit of N_H < 10²¹ cm⁻² (90% confidence level), and considering that the source is only 118 pc away, we performed the spectral analysis assuming there is no absorption. Good fits were obtained with either a power-law model with a photon index of Γ = 1.9 ± 0.2, or a thermal plasma model (mekal in XSPEC) with a temperature of $kT=3.5_{-0.8}^{+1.5}$ keV. In both cases, a flux of ∼5 − 6 × 10⁻¹³ erg cm⁻² s⁻¹ was obtained in the 0.3–10 keV range. These values are consistent with the results reported by Mereghetti et al. (2022) based only on the data of the first two epochs (see Table 4). A fit with a black-body model was unacceptable (cstat/d.o.f. = 229.5/121). All the fit parameters are summarized in Table 4, where we also give the results obtained by fitting the individual spectra of the three epochs.

The other five sources were not detected by the XRT. The upper limits of the count rate in two energy ranges are given in Table 3. The corresponding flux limits were computed assuming a power-law spectrum with an index of Γ = 2 and no absorption. Luminosity upper limits were then computed using the source distances reported in Table 3.

We also derived upper limits on the temperature of a hypothetical black body emission under the assumption of a spherical emitting surface with a radius of 10 km, at the distance of the systems. The values are reported in Table 3.

4. Discussion

4.1. X-ray emission from 1527+3536

The nature of the X-ray emission from 1527+3536 was discussed by Lin et al. (2023) on the basis of its possible association with the source 1RXS J152748.8+353658 discovered in the ROSAT All Sky Survey. This association is now confirmed by our more precise localization. The mass estimate of the compact object in this system obtained by Lin et al. (2023), 0.98 ± 0.03 M_⊙, is also compatible with a white dwarf (WD). We also note that Zhang et al. (2024), using high-resolution spectroscopy obtained with the Canada-Hawaii- France 3.6 m Telescope, estimate a mass of 0.69 ± 0.02, which is typical for a WD. However, Lin et al. (2023), based on the X-ray luminosity being lower than that of intermediate polars and on the absence of dwarf nova-like outbursts in the long-term optical data, favor the NS interpretation. In this case, these latter authors claim that despite its binary nature, its X-ray properties resemble those of the so-called X-ray-dim, isolated NSs (XDINS), a small class of nearby, isolated, thermally emitting NSs that might be the descendants of magnetars (see e.g., Turolla 2009).

The Swift/XRT data presented here do not support the XDINS interpretation. In fact, all the isolated NSs of this class have very soft thermal spectra well fit by blackbody models with temperatures of kT < 100 eV. On the contrary, the XRT spectrum of 1527+3536 is rather hard and a BB model gives a poor fit, which might indicate a higher NS age than the range accepted for XDINs. Such a hard X-ray spectrum instead might be consistent with low-level accretion from the companion star. Power-law components have been observed in NS low-mass X-ray binaries in quiescence – that is, when accretion is mostly shut off – and interpreted as possibly due to some residual ongoing accretion (e.g. Degenaar et al. 2012, and references therein). In fact, Lin et al. (2023) interpret the presence of variable Hα emission from this system as evidence for an accretion disk or stream. The observed luminosity of ∼10³⁰ erg s⁻¹ requires an accretion rate of the order of 10¹⁰ g s⁻¹ if the compact object is a NS. In the case of a WD, the inferred accretion rate would lead to dwarf nova phenomena that are, at present, not supported by long time-series data (Lin et al. 2023). Although the late-type companion is not expected to provide such a rate through a strong stellar wind, we note that it might be filling (or close to fill) its Roche lobe, which has a radius of only 0.66 R_⊙.

Another possibility is that the X-ray emission is produced by a rotation-powered NS. In this case, assuming a typical efficiency of ∼10⁻³ for conversion of rotational energy to X-rays yields Ė_rot ∼ 10³³⁻³⁴ erg s⁻¹. In this case, the lack of a radio detection (with a deep upper limit of 5 μJy at 1.2 GHz, Lin et al. 2023) could be explained by an unfavorable beam direction. However, considering the small distance of only 100 pc, the absence of a bright Fermi/LAT γ-ray source at this position (Abdollahi et al. 2020) makes this interpretation unlikely, considering that pulsar beams in γ-rays are much wider than in the radio band (see, e.g., Johnston et al. 2020). Furthermore, no signs of irradiation from an energetic pulsar are visible in the optical data of the companion star.

In summary, if the unseen component of this binary is a NS, we believe that low-level accretion from its companion is the most likely possibility to explain its X-ray emission. The alternative possibility of an accreting white dwarf –which would require an accretion rate that is higher by a factor of ≳500– cannot be ruled out by the current X-ray data. Finally, although the temperature derived with a mekal spectrum is rather high for a K9-M0 star, it cannot be excluded that the noncollapsed component in this system is responsible for the X-ray emission (or at least part of it) if some magnetic coronal activity is present (Preibisch et al. 2005; Drake 2019). Similar conclusions have also been reached for an analogous system, LAMOST J235456.73+335625.9, as shown by Zheng et al. (2023).

4.2. Upper limits

Four of the systems were not detected by the XRT (1220+5841, 1832–0119, 1313+4152, 2128+3316), with very stringent upper limits on the X-ray luminosity; that is, below 10³⁰ erg s⁻¹. All four have orbital periods of longer than many months and thus have very wide orbits. The optical components are low-mass, late-type stars that are clearly not filling their Roche lobes and do not have strong stellar winds. The candidate NSs in these binaries can thus be considered as virtually isolated objects for what concerns their X-ray emission.

Therefore, in addition to the processes discussed above for 1527+3536, we must also consider the possibility of X-ray emission powered by accretion from the interstellar medium (ISM). In a seminal paper, Ostriker et al. (1970), using the Bondi formulation, showed that if a NS accretes from the ISM, the typical expected luminosity L should be in soft X-rays, with

$$\begin{array}{c}\hfill L\sim {10}^{32}{v}_{10}^{-3}n\mathrm{erg}{\mathrm{s}}^{-1},\end{array}$$(1)

where v₁₀ is the velocity of the NS with respect to the medium in units of 10 km s⁻¹, and n is the ISM density in atoms cm⁻³. Treves & Colpi (1991) showed that taking Eq. (1) at face value, a large number of isolated NSs were to appear in the ROSAT survey. Now it is clear that this is not the case, and the problem is rather to understand and interpret this absence. The explanation could be linked to a residual magnetic field, which somehow inhibits accretion, or to velocities that are higher than initially estimated (Popov 2023).

The upper limits on the X-ray luminosity (Table 3) can be reconciled with Eq. (1) assuming a small ISM density, the impediment from the magnetic field, very soft emission peaking below the XRT band, or a moderate velocity of for example ∼100 km s⁻¹ (although such a velocity, typical of isolated NSs, is unlikely for NSs bound in binaries).

The case of 0616+2319 is slightly different, because its short orbital period of only ∼21 hours implies that this is a much more compact binary. However, the parameters derived by fitting its multiwavelength light curves (Yuan et al. 2022) indicate that in this binary as well the noncollapsed component does not fill the Roche lobe. Accretion from the companion’s stellar wind might occur in this system, but the constraints on the mass-accretion rate derived from our upper limit on the X-ray luminosity are not particularly useful, owing to the large distance of this source and the relative weakness of stellar winds from late G-type main sequence stars.

4.3. Search for UV excess

To compare our UV detections and upper limits with the available data for our targets at longer wavelengths, we searched the literature for photometric points using the catalogues available on the VizieR (Ochsenbein et al. 2000) service. For each source, we queried for catalogue entries within 1″ of the Gaia position, and then manually inspected the resulting tables to check for any spurious associations. We then plotted the resulting spectral energy distributions (SED) along with emission curves from a black body with temperature set on the basis of the spectral type of the companion star (as reported in Table 1) or (in the case of 1832–0119) based on the peak emission of the SED.

In the case of 1527+3536, our UVOT flux (Table 3 and Figs. 1, A.1) agrees with the GALEX photometry reported by Lin et al. (2023) and therefore confirms the presence of a sizeable UV excess, which is apparent by comparing with either a black body or the stellar model used by Lin et al. (2023). In our data, there is a possible indication of a 0.1 mag variability on timescales measurable in months (Table 4). The excess could be due to the accretion process of the NS, as proposed by Lin et al. (2023), but one cannot exclude some magnetic coronal activity of the noncollapsed component, as mentioned in the previous section, or emission from the surface of a WD with T_eff ∼ 12 000 K. There is no evidence of UV emission in excess of the flux expected from the companion star for any of the remaining sources.

Fig. 1. Spectral energy distribution for 1527+3536. Red points are literature data from Vizier (GALEX, Gaia, SDSS, 2MASS, PAN-STARRS, WISE data). The green point is Swift UVOT data from our ToO campaign, and the blue line is the spectral fit of the main sequence companion star derived from Lin et al. (2023).

5. Conclusions

We obtained Swift observations of a sample of binary systems selected from optical observations as good candidates for hosting a NS or WD. Only one system, 1527+3536, is detected in X-rays and in the UV. While this emission can be explained by accretion onto a NS from a main sequence companion (almost) filling its Roche lobe, alternatives (such as a white dwarf, or X-ray emission from the main sequence star corona) cannot be completely ruled out.

In all other cases, no X-ray or UV excess emission is detected. As the characteristics of these binaries suggest that accretion from the noncompact object (through Roche lobe or stellar wind) is unlikely, we tried to interpret the data within a XDINS-like scenario. In four cases (1220+5841, 1313+4152, 1832–0119, and 2128+3316), the upper limits we find require that we invoke either a high velocity of the NS, a small ISM density, accretion suppression by the magnetic field, or emission peaking in the very soft X-rays. For 0616+2319, the compactness of the system could allow accretion from the stellar wind of the main sequence star, but because of the larger distance of the system, the constraints we can derive on the accretion rate are not particularly stringent.

Overall, all the systems we observed are compatible with the presence of a NS as the undetected companion, but we did not find any strong evidence of their presence; nor can we exclude their WD nature. Observations in a softer X-ray band (performed with a detector whose sensitivity peaks below the XRT energy band) or in the radio (to check for the presence of a pulsar-like activity) are the next natural steps toward characterizing these systems and fully assessing the viability of using Gaia data to select NS binary systems.

Acknowledgments

This research has made use of the VizieR catalogue access tool, CDS, Strasbourg, France (DOI : https://doi.org/10.26093/cds/vizier). The original description of the VizieR service was published in 2000, A&AS 143, 23. This work made use of data supplied by the UK Swift Science Data Centre at the University of Leicester. AM, FCZ and NR are supported by the H2020 ERC Consolidator Grant “MAGNESIA” under grant agreement No. 817661 (PI: Rea), grant SGR2021-01269 (PI: Rea), and partially supported by the program Unidad de Excelencia María de Maeztu CEX2020-001058-M. FCZ is also supported by a Ramón y Cajal fellowship (grant agreement RYC2021-030888-I). This work has made use of data from the European Space Agency (ESA) mission Gaia (https://www.cosmos.esa.int/gaia), processed by the Gaia Data Processing and Analysis Consortium (DPAC, https://www.cosmos.esa.int/web/gaia/dpac/consortium). Funding for the DPAC has been provided by national institutions, in particular the institutions participating in the Gaia Multilateral Agreement.

References

Appendix A: Spectral energy distribution plots

Fig. A.1. Spectral energy distribution for the systems presented in this study. Red points are literature data from Vizier (GALEX, Gaia, SDSS, 2MASS, PAN-STARRS, WISE data). Green points are Swift UVOT data from our ToO campaign, and blue lines are blackbody emission profiles with temperatures —determined using the stellar type of the main sequence companion star— normalized to the Gaia G flux.

All Tables

All Figures

	Fig. 1. Spectral energy distribution for 1527+3536. Red points are literature data from Vizier (GALEX, Gaia, SDSS, 2MASS, PAN-STARRS, WISE data). The green point is Swift UVOT data from our ToO campaign, and the blue line is the spectral fit of the main sequence companion star derived from Lin et al. (2023).
In the text

	Fig. A.1. Spectral energy distribution for the systems presented in this study. Red points are literature data from Vizier (GALEX, Gaia, SDSS, 2MASS, PAN-STARRS, WISE data). Green points are Swift UVOT data from our ToO campaign, and blue lines are blackbody emission profiles with temperatures —determined using the stellar type of the main sequence companion star— normalized to the Gaia G flux.
In the text