Searching for the Donor Stars of ULX Pulsars

TL;DR

This study searches for optical counterparts of two ultraluminous X-ray pulsars using HST data, aiming to identify their donor stars and understand their nature, with mixed results and some candidate identifications.

Contribution

First detailed optical counterpart search for ULX pulsars M82 X-2 and NGC 5907 X-1, providing candidate donor stars and insights into their properties.

Findings

Detected potential donor star candidates for M82 X-2.

Did not find the expected red giant donor for NGC 5907 X-1.

Identified possible AGB star counterparts for NGC 5907 ULX-2.

Abstract

We report on our search for the optical counterparts of two ultraluminous X-ray pulsars with known orbital periods, M82 X-2 and NGC 5907 X-1, in new and archival HST observations, in an effort to characterize the donor stars in these systems. We detect five near-infrared sources consistent with the position of M82 X-2 that are too bright to be single stars. We also detect seven sources in the WFC3/UVIS F336W image whose photometry matches that of 10-15 M$_\odot$ stars turning off the main sequence. Such stars have densities consistent with the properties of the donor star of M82 X-2 as inferred from X-ray timing analysis, although it is also possible that the donor is a lower mass star below our detection limit or that there is a significant contribution from the accretion disc to the optical emission. We detect three candidate counterparts to NGC 5907 X-1 in the near-infrared. All of these are too bright to be the donor star of the ULX, which based on its orbital period is a red giant. The high background at the location of NGC 5907 X-1 precludes us from detecting this expected donor star. The recently discovered NGC 5907 ULX-2 also falls within the field of view of the near-infrared imaging; we detect four sources in the error circle, with photometry that matches AGB stars. The star suggested to be the counterpart of NGC 5907 ULX-2 by Pintore et al. (2018) falls outside our 2-$\sigma$ error circle.

Figures

Tables

Table 1. Table 1: List of HST observations

Target	Prop ID	Obs date	Inst.	Filter	Exp. time (s)
NGC 5907	15074	2017-12-13	WFC3/IR	F160W	5612
	6092	1996-03-31	WFPC2	F450W	780
	6092	1996-03-31	WFPC2	F814W	480
M82	11360	2010-01-01	WFC3/UVIS	F225W	1665
	11360	2010-01-01	WFC3/UVIS	F336W	1620
	11360	2010-01-01	WFC3/IR	F110W	1195
	11360	2010-01-01	WFC3/IR	F160W	2395

Table 2. Table 2: List of Chandra observations

Target	Obs ID	Obs start date	Inst.	Exp. time (ks)
NGC 5907	12987	2012-02-11	ACIS-S	16.0
	20830	2017-11-07	ACIS-S	51.3
M82	10542	2009-06-24	ACIS-S	118.6
	10543	2009-07-01	ACIS-S	118.5

Table 3. Table 3: Limiting magnitudes of HST images at the position of the ULXs

Source	HST filter	Limiting magnitude
M82 X-2	F225W	25.1
	F336W	26.4
	F110W	18.8
	F160W	18.8
NGC 5907 X-1	F450W	25.5
	F814W	23.6
	F160W	22.0
NGC 5907 ULX-2	F450W	25.8
	F814W	25.0
	F160W	23.7

Table 4. Table 4: Dolphot magnitudes of sources labeled in Figure 1 . Listed magnitudes are not corrected for extinction

Source	$m_{\mathrm{F160W}}$	$m_{\mathrm{F814W}}$	$m_{\mathrm{F450W}}$
NGC 5907 X-1
1	$21.61\pm 0.01$	$23.53\pm 0.14$	$25.0\pm 0.2$
2	$21.75\pm 0.01$	$>23.6$	$>25.5$
3	$21.91\pm 0.02$	$>23.6$	$>25.5$
NGC 5907 ULX-2
1	$22.68\pm 0.02$	$>25.0$	$24.7\pm 0.3$
2	$22.73\pm 0.02$	$>25.0$	$>25.8$
3	$23.21\pm 0.03$	$>25.0$	$>25.8$
4	$23.38\pm 0.04$	$>25.0$	$>25.8$

Table 5. Table 5: Dolphot magnitudes of sources labeled in Figure 2 . Listed magnitudes are not corrected for extinction

Source	$m_{\mathrm{F160W}}$	$m_{110W}$	$m_{\mathrm{F336W}}$	$m_{\mathrm{F225W}}$
M82 X-2
1	$16.719\pm 0.001$	$18.411\pm 0.002$	$24.9\pm 0.1$	$>25.1$
2	$16.006\pm 0.001$	$17.386\pm 0.001$	$>26.4$	$>25.1$
3	$16.153\pm 0.001$	$17.537\pm 0.001$	$25.3\pm 0.2$	$>25.1$
4	$16.822\pm 0.002$	$18.442\pm 0.002$	$>26.4$	$>25.1$
5	$16.046\pm 0.001$	$17.436\pm 0.001$	$>26.4$	$>25.1$
6	$>18.8$	$>18.8$	$24.39\pm 0.09$	$>25.1$
7	$>18.8$	$>18.8$	$25.4\pm 0.2$	$>25.1$
8	$>18.8$	$>18.8$	$25.3\pm 0.2$	$>25.1$
9	$>18.8$	$>18.8$	$25.6\pm 0.2$	$>25.1$
10	$>18.8$	$>18.8$	$25.6\pm 0.2$	$>25.1$
11	$>18.8$	$>18.8$	$25.9\pm 0.3$	$>25.1$
12	$>18.8$	$>18.8$	$25.7\pm 0.3$	$>25.1$

Full text

stars: neutron — X-rays: individual(M82 X-2, NGC 5907 X-1, NGC 5907 ULX-2) — infrared: stars

††facilities: HST (WFPC2 and WFC3), Chandra (ACIS), Gaia††software: astropy (Astropy Collaboration et al., 2013), DOLPHOT (Dolphin, 2000), MESA (Paxton et al., 2011, 2013, 2015), CIAO (Fruscione et al., 2006)

\turnoffedit

1 Introduction

Ultraluminous X-ray sources (ULXs) are off-nuclear point-like sources with $L_{\rm X}\geq 10^{39}$ erg s*-1*, exceeding the Eddington limit for a 10 $M_{\odot}$ black hole (e.g., Fabbiano, 1989). Although early papers on ULXs suggested intermediate mass black holes (IMBHs; $10^{2}\lesssim M_{\rm{BH}}\lesssim 10^{5}\,M_{\odot}$) as the accretors in these systems (e.g. Colbert & Miller, 2005), a growing body of evidence — mainly from broadband X-ray spectroscopy — favors ‘stellar mass’ objects exceeding the Eddington limit for the majority of sources (Gladstone et al., 2009; Sutton et al., 2013b; Walton et al., 2018). A combination of super-Eddington accretion and geometric collimation would then explain the large luminosities. This interpretation was proven to be correct by the discovery of X-ray pulsations from M82 X-2 (Bachetti et al., 2014) and, more recently, from three other ULXs , showing that in these objects the accretor is in fact a neutron star. A fifth neutron star ULX (M51 ULX8), without X-ray pulses but identified through a cyclotron resonance feature, was reported by Brightman et al. (2018). For a recent review on ULXs we refer the reader to Kaaret et al. (2017).

Accreting magnetized neutron stars can, in principle, reach these super-Eddington luminosities through a number of mechanisms. For example, magnetic fields $\gtrsim 10^{12}$ G will collimate the accretion flow, allowing material to accrete onto the polar regions while radiation escapes from the sides of the column. Strong surface magnetic fields ($\gtrsim 10^{13.5}$ G) reduce the scattering cross section for electrons, reducing the radiation pressure and increasing the effective Eddington luminosity (Basko & Sunyaev, 1976; Herold, 1979; Mushtukov et al., 2015). However, configurations with low magnetic fields have also been proposed (e.g. Kluzniak & Lasota, 2015). All accretion theories trying to explain the seemingly super-Eddington luminosities of ULXs — with either neutron star or stellar mass black hole accretors — include a highly super-Eddington mass transfer rate from the donor star, which is difficult to reach through wind accretion. The favored scenario is therefore mass transfer through Roche lobe overflow from a massive donor star.

Very few ULX mass donor stars have been identified. The large distances to ULXs in combination with their often bright accretion discs make their donor stars hard to detect, and even more difficult to confirm spectroscopically. The only systems where spectroscopic signatures of donor stars are detected are M101 ULX1, with a Wolf-Rayet donor star (Liu et al., 2013); three ULXs (NGC 253 J004722.4-252051, NGC 925 J022721+333500 and NGC 4136 J120922+295559) with M-type supergiant donors (Heida et al., 2015, 2016); and the ULX pulsar (ULXP) NGC 7793 P13, which has a blue supergiant (B9Ia) donor (Motch et al., 2011, 2014). The recently discovered ULXP SN2010da in NGC 300 likely has a supergiant (sg) B[e] or yellow sg donor star (Lau et al., 2016; Villar et al., 2016). These are all massive stars, confirming the idea that many ULXs are an extreme class of high-mass X-ray binaries and possible progenitors of gravitational wave sources . However, there is a significant observational bias in favor of massive companions because they are much brighter than low-mass stars. Wiktorowicz et al. (2017) predict that the majority of neutron star ULXs have low mass ($\leq 1.5$ $M_{\odot}$), red giant donors.

For ULXs with a black hole accretor, phase-resolved spectroscopic observations of the donor star are the only direct way to obtain limits on the mass of the compact object. In ULXPs we can turn this around: since neutron stars have been observed to have a very narrow range of masses ($1-2$ $M_{\odot}$, see Lattimer, 2012, for a review), the mass of the compact object is known , and — unless the system is viewed at a very low inclination — we can obtain orbital parameters of the system by observing the modulation of the pulse period due to the Doppler effect. This gives us the mass function and therefore limits on the mass and average density of the donor star. In combination with direct observations of the donor it is then possible to investigate if these systems are indeed fed through Roche lobe overflow, as was recently shown for NGC 7793 P13 (Fürst et al., 2018). It is also a potential way to identify ULXs with low-mass donor stars.

In this paper we analyze new and archival Hubble Space Telescope (HST) observations of M82 X-2 and NGC 5907 X-1. We combine the photometry with orbital parameters of these systems derived from X-ray timing analyses by Bachetti et al. (2014) and Israel et al. (2017a) to put constraints on their respective donor stars. All magnitudes in this paper are Vega magnitudes.

2 Data reduction and analysis

2.1 HST observations

We obtained a deep HST WFC3/IR F160W image of NGC 5907 (proposal ID 15074) and downloaded archival WFPC2 images of NGC 5907 and WFC3 images of M82 (Lim et al. 2013, e.g.; see Table 1). There are additional deep WFPC2/F606W images of NGC 5907, but NGC 5907 X-1 is just outside the field of view.

We use DOLPHOT version 2.0 (Dolphin, 2000) for the photometric analysis. Following the DOLPHOT manuals for the WFPC2 and WFC3 modules, we photometer the flt (WFC3) and c0m (WFPC2) images separately, using the drizzled images produced by the HST pipeline as the reference image. The ULXs are located in crowded regions with high and variable backgrounds, especially NGC 5907 X-1. We find that setting fitsky = 3, with other parameter values as recommended in the manuals, yields the best results.

2.2 ULX positions

We determined the positions of the ULXs and other X-ray sources in archival deep Chandra observations of the two galaxies (see Table 2) using the CIAO tool wavdetect. There are two very deep Chandra observations covering M82 X-2, but none of the X-ray sources have unique optical counterparts in the HST images. Instead we register both Chandra observations and the drizzled HST images using Gaia DR2 sources in the field of view. We use the CIAO tool reproject_aspect to register the Chandra observations, only considering X-ray sources with S/N $>5$ (194 sources in obsid 10542 and 209 sources in obsid 10543, excluding M82 X-2 itself). There are 4 and 3 sources with matches in the Gaia catalog for obsid 10542 and 10543, respectively. Given the low number of sources we only calculate the x- and y-shift, leaving the rotation and scale fixed. We adopt the average residual distance between the Gaia and Chandra positions, $0.3^{\prime\prime}$, as a measure of the astrometric uncertainty. The positional uncertainty due to localizing the ULX in the Chandra image is negligible ($<0.05^{\prime\prime}$). The position of M82 X-2 in the two Chandra observations is consistent to within $<0.1^{\prime\prime}$ — we adopt the position of the ULX in obsid 10543, RA = 09:55:51.21, Dec = +69:40:44.1 (J2000). To register the drizzled HST images we use the Starlink/GAIA ‘fit to star positions’ tool, with 21 and 19 matched sources for the IR (F160W) and UVIS (F336W) images, respectively. We adopt the rms of the fit as the final astrometric uncertainty: this is $0.12^{\prime\prime}$ for the IR image and $0.04^{\prime\prime}$ for the UVIS image. The uncertainty of the ULX position on the HST images is dominated by the uncertainty in the registration of the Chandra observation and is $0.3^{\prime\prime}$ for both the IR and UVIS images.

For NGC 5907 we retrieved two Chandra observations from the archive. The shorter one (obsid 12987) contains NGC 5907 X-1 but no X-ray sources that can be used for cross-matching with the HST image or other optical catalogs. The longer observation (obsid 20830) is a recent DDT observation obtained while NGC 5907 X-1 was in an off-state (Pintore et al. 2018; a previous off-state of this source was observed by Walton et al. 2015). This observation does contain two X-ray sources with unique counterparts in our WFC3/IR image. Using reproject_aspect, we first register the short Chandra observation to the longer one using 8 common X-ray sources with S/N $>4$. The average residual after the shift is $0.23^{\prime\prime}$. We then use the two overlapping sources between the longer Chandra observation and our WFC3/IR image to register the Chandra observations to the HST image. With only two sources we cannot calculate the uncertainty in this step; we conservatively assume an error of $0.3^{\prime\prime}$. Quadratically adding the two uncertainties gives a total astrometric uncertainty of $0.4^{\prime\prime}$ (again, the error in the localization of the ULX in the Chandra image is negligible). The position of the ULXP on the WFC3 image is RA = 15:15:58.66, Dec = +56:18:10.3 (J2000).

The new ULX reported by Pintore et al. (2018, NGC 5907 ULX-2) is visible in Chandra observation 20830. Its position in our WFC3 F160W image is RA = 15:16:01.13, Dec. = +56:17:51.5 (J2000), with an astrometric uncertainty of $0.3^{\prime\prime}$ that is mainly due to the registration of Chandra observation 20830 to our HST image.

2.3 Limiting magnitudes

3 Results

3.1 NGC 5907

We detect three sources brighter than the limiting magnitude inside the 2-$\sigma$ error circle of NGC 5907 X-1 in our F160W image (see center right panel in Figure 1 and Table 4). The errors listed are the statistical errors as calculated by Dolphot. The brightest source is also detected in the F814W images, at $m_{\mathrm{F814W}}$ $\approx 23.5$ and $m_{\mathrm{F450W}}$ $\approx 25.0$. No point sources are detected in the F450W images.

We detect sources brighter than the limiting magnitude in the 2-$\sigma$ error circle of NGC 5907 ULX-2 in our F160W image (see right-hand panel in Figure 1 and Table 4). The Pintore et al. (2018) preferred counterpart is visible in our F160W and F814W images (indicated with a magenta circle in Figure 1), but its position is inconsistent (at $>3$-$\sigma$) with our localization of the ULX. This is due to our smaller localization uncertainty: our $1-\sigma$ uncertainty is $0.3^{\prime\prime}$ versus $0.42^{\prime\prime}$ for Pintore et al. (2018), as they used a different method to register the Chandra and HST images.

3.2 M82 X-2

We detect five sources in the 2-$\sigma$ error circle of M82 X-2 in the NIR images, with $17.4\leq m_{\mathit{F110W}}\leq 18.4$ and $16.0\leq m_{\mathit{F160W}}\leq 16.8$ (see right-hand panel in Figure 2 and Table 5). No point sources are significantly detected in the F225W image.

4 Discussion

We searched for counterparts to two ULXPs in new and archival HST observations. Both sources are located in crowded regions with high extinction — M82 X-2 in the center of M82, and NGC 5907 X-1 in the dust lane of the edge-on spiral galaxy NGC 5907. We detect multiple potential counterparts to both ULXPs as well as to a second, recently discovered ULX in NGC 5907. Given the distance modulus to NGC 5907 ($m-M=31.17\pm 0.09$, Tully et al. 2013), the absolute magnitudes of the counterparts to NGC 5907 X-1 in the F160W filter are $-9.5\leq M_{\mathit{F160W}}\leq-9.3$.

4.1 Reddening

NGC 5907 ULX-2 is located further away from the central dust lane; Pintore et al. (2018) found no local extinction on top of the foreground Galactic absorption from their analysis of the X-ray spectrum of this source.

The foreground Galactic extinction in the direction of M82 is $A_{V}=0.485$ (Schlegel et al., 1998). On top of that there is significant extinction due to gas and dust ($A_{V}\geq 1.55$) in the core of the galaxy, where M82 X-2 is located (Hutton et al., 2014).

4.2 Limits on donor star properties from X-ray timing

Assuming circular orbits and accretion via Roche lobe overflow, the density of the donor star can be determined directly given the orbital period of the binary system and the mass of the accretor, using the equation for the Roche radius (Eggleton, 1983) and Kepler’s laws. For M82 X-2, the orbital period was reported by Bachetti et al. (2014) as $2.53260\pm 0.00005$ days, with an eccentricity $\epsilon<0.003$ and a projected semi-major axis ($a\sin(i)$) of $22.225\pm 0.004$ ls.

For NGC 5907 X-1 these values are not as well constrained. Israel et al. (2017a) performed a likelihood analysis to obtain the orbital parameters of the system (their Figure 2). They found a most-probable period $P_{\mathrm{orb}}=5.3^{+2.0}_{-0.9}$ days with $a\sin(i)=2.5^{+4.3}_{-0.8}$ ls.

For M82 X-2, the lower limit on the donor star mass is $\sim 5$ $M_{\odot}$, assuming a neutron star mass of 1.4 $M_{\odot}$ and a system inclination $\leq 60^{\circ}$ (Bachetti et al., 2014). For NGC 5907 X-1, the mass function for the most likely parameters reported by Israel et al. (2017a) is $6\times 10^{-4}$ $M_{\odot}$, implying a lower limit for the donor star of $\sim 0.1$ $M_{\odot}$. The measurements of Israel et al. (2017a) also allow for a longer orbital period (up to $\sim 21$ days) with a larger projected semi-major axis, that would imply a higher minimum mass for the donor star (e.g. 29 $M_{\odot}$ for an orbital period of 18 days and $a\sin(i)=200$ ls).

We plot the relation between allowed masses and radii for M82 X-2 and NGC 5907 X-1 in Figure 3. For comparison, we also show the relation for NGC 7793 P13, a ULXP with a blue supergiant companion with a mass of 18 – 23 $M_{\odot}$ and a radius of 96 – 125 $R_{\odot}$ (Motch et al., 2014; Fürst et al., 2018). In the background we show evolution tracks of single, non-rotating stars at solar metallicity from the MESA Isochrones and Stellar Tracks (MIST) project (Choi et al., 2016; Dotter, 2016).

From Figure 3 we can see that the donor star of M82 X-2 is most likely just turning off the main sequence (as was also shown by Fragos et al. 2015), while the donor of NGC 5907 X-1 .

The limiting magnitude of our WFC3/F160W image is too bright to detect the donor star, due to the high background in this region. The F160W magnitudes and F450W and F814W limits of the sources detected in the error circle of NGC 5907 ULX-2 are consistent with asymptotic giant branch (AGB) stars. Without information on the orbital period and accretor mass in this system, we cannot say anything about the likelihood that one of these sources is the donor star to the ULX.

The near-IR counterparts to M82 X-2 are even brighter, and no single star in the MIST simulations matches these F110W and F160W absolute magnitudes. These sources

5 Conclusions

We obtained new WFC3/NIR images of NGC 5907 and used those in conjunction with deep archival WFPC2 images to search for the donor star of NGC 5907 X-1. The sources we detect are too bright to be the donor star of the ULXP, which based on its orbital period should be a red giant (see Figure 3). The high background at the location of NGC 5907 X-1 precludes us from detecting the expected donor star, . The recently discovered NGC 5907 ULX-2 also falls within the field of view; we detect sources in the error circle, with photometry that matches AGB stars. The star suggested to be the counterpart by Pintore et al. (2018) falls outside our 2-$\sigma$ error circle.

We also retrieved deep WFC3/NIR and WFC3/UVIS images of M82 X-2 from the HST archive to search for the donor star of this ULXP. The sources detected in the NIR images are too bright to be single stars, but we also detect in the UVIS F336W image whose photometry matches that of $M_{\odot}$ stars turning off the main sequence. Such stars have densities consistent with the donor star of M82 X-2, although it is also possible that the donor is a lower mass star fainter than our detection limit .

Characterizing the donor stars of ULXPs is important for testing models of binary evolution. To date, the only system with a spectroscopically characterized donor star is NGC 7793 P13, which hosts a blue supergiant star (Motch et al., 2011, 2014). However, the majority of ULXPs are expected to have lower mass red giant donor stars (Wiktorowicz et al., 2017). Shorter period systems such as M82 X-2 and NGC 5907 X-1 may be examples of such lower mass systems, but the crowded environments and the large distances to these sources make direct detection and especially spectroscopic characterization unfeasible with currently available instruments. For the nearest ULXs this may change when the first 30-40 m class telescopes become operational.

We would like to thank the anonymous referee whose comments helped to significantly improve the paper. Based on observations made with the NASA/ESA Hubble Space Telescope, obtained from the Data Archive at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5-26555. These observations are associated with program 15074. Support for Program number 15074 was provided by NASA through a grant from the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Incorporated, under NASA contract NAS 5-26555. The scientific results reported in this article are based in part on data obtained from the Chandra Data Archive. 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. The work of DS was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with NASA.