Metal-THINGS: On the Metallicity and Ionization of ULX Sources in NGC 925

The Metal-THINGS survey is a large program focused on obtaining IFU spectroscopy for a unique sample of nearby galaxies using the 2.7 m Harlan Schmidt telescope at the McDonald Observatory. It employs the George Mitchel spectrograph (GMS, formerly known as VIRUS-P, Hill et al. 2008), as well as the Multi Unit Spectroscopic Explorer (MUSE; Bacon et al. 2010)—on the Very Large Telescope (VLT). The selected galaxies are based on the THINGS survey (Walter et al. 2008), which observed 34 large nearby galaxies with the Very Large Array (VLA), obtaining HI data at high spatial and spectral resolution.

The GMS is a square array of 100 × 102 arcsecs, with a spatial sampling of 4.2 arcsecs, and a 0.3 filling factor. The IFU consists of 246 fibers arranged in a fixed pattern. NGC 925 was observed as part of the Metal-THINGS survey (Lara-Lopez et al. 2020, in preparation) with a red setup covering wavelengths from 4400 to 6800 Å, using the low resolution grating VP1, with a resolution of 5.3 Å. This wavelength range allows us to measure the following strongest emission lines: Hβ, [OIII]λλ4959, 5007, [OI]λ6300, Hα, [NII]λ6584, and [SII]λλ6716, 6731.

Every pointing is observed with three dither positions to ensure a 90% surface coverage. Due to the extended nature of our galaxies, sky exposures are taken off-source for sky subtraction purposes during the reduction process. Our observing procedure consists of taking a 15 minute exposure per dither, followed by a sky exposure, and the repetition of this process until 45 minutes are reached per dither, per pointing. A calibration star was observed every night using six dither positions to ensure a 99% flux coverage. Calibration lamps (Neon + Argon for the red setup) were observed at the beginning and end of every night to wavelength calibrate the spectra. The average seeing during the observations was 1.5 arcsecs.

NGC 925 was observed in 2017 December, 2018 January, and 2019 January using GMS. A total of 13 pointings were observed (Figure 1), which provide a total of 9,594 individual spectra for this galaxy. From our spectroscopic data, we estimate that NGC 925 is at a redshift of 0.00185, and adopted a cepheids-based distance of 9.2 Mpc (Saha et al. 2006), which gives a scale of 0.044 kpc/'', meaning each 4.2'' fiber in the IFU corresponds to ∼184 pc.

Zoom In Zoom Out Reset image size

The basic data reduction, including bias subtraction, flat frame correction, and wavelength calibration, was performed using P3D. ¹⁹ Sky subtraction and dither combination was performed using our own routines in Python. Flux calibration was performed following Cairós et al. (2012) using six dither positions, ensuring a 97% coverage. We used IRAF (Tody 1986) to create a calibration function using the packages standard and sensfunc.

The stellar continuum of all flux-calibrated spectra was fitted using STARLIGHT (Cid Fernandes et al. 2005; Mateus et al. 2006; Asari et al. 2007). Briefly, we used 45 simple stellar populations (SSP) models from the evolutionary synthesis models of Bruzual & Charlot (2003) with ages from 1 Myr up to 13 Gyr and metallicities Z = 0.005, 0.02, and 0.05. The stellar continuum was subtracted from the spectra, and the emission lines were measured using Gaussian line-profile fittings. For a more detailed description, see Zinchenko et al. (2016, 2019a).

The measured emission line fluxes were corrected for interstellar reddening using the theoretical Hα/Hβ ratio and the reddening function from Cardelli et al. (1989) for R_V  = 3.1.

Figure 2 shows the extinction corrected Hα surface brightness map for NGC 925. For visualization purposes, and only for this figure, an interpolation of the 13 pointings is displayed. In the same figure, the three identified ULXs are marked with crosses. All further figures that use GMS data (Figures 4 to 6) show the data of individual fibers.

Zoom In Zoom Out Reset image size

To measure the size of the nebulae associated with the ULX sources, we use high spectral and spatial resolution data from the Fabry–Perot interferometer PUMA (Rosado et al. 1995). Observations of NGC 925 were done on 2018 December 2 at the 2.1 m telescope of the Observatorio Astronómico Nacional in San Pedro Mártir, Baja California, México (OAN-SPM). We used a 2048 × 2048 spectral CCD detector, with a binning factor of 4, resulting in a field of view of 10' in 512 × 512 pixels with a spatial sampling of 129 pixel⁻¹. Its Finesse (∼24) leads to a sampling spectral resolution of 0.41 Å (19.0 km s⁻¹). The spatial resolution of PUMA is 57.5 pc per pixel.

To isolate the redshifted Hα emission of the galaxy, we used a filter centered at Hα (λ6570 Å) with an FWHM of 20 Å. We scanned the Fabry–Perot interferometer through 48 channels with an integration time of 120 s per channel, obtaining an object data cube of dimensions 512 × 512 × 48. The calibration data cube was obtained under the same conditions as the object data cube, just after its completion, and it has the same dimensions as the data cube. We used a hydrogen lamp whose 6562.78 Å line was close to the redshifted Hα from the galaxy due to its low recession velocity. The wavelength calibration of the PUMA object cube was carried out using the Fabry–Perot specialized software ADHOCw developed by J. Boulesteix (Amram et al. 1989), which uses the calibration cube to compute the phase map, a 2D map producing, pixel per pixel, the channel at which the Hα line is at rest. Once wavelength was calibrated, we obtained a lambda cube with 48 channels of different consecutive velocities. We also constructed a spatial smoothing (x,y) with an FWHM of (2,2) pixels. It is worth nothing that the data are not flux calibrated.

The radial velocity field was obtained by computing the velocity value of the barycenter, the "center of mass" of the velocity profile, of each of the 512 × 512 line profiles as well as the gas velocity dispersion of the line profiles. These quantities were also obtained with the ADHOCw software.

Lastly, the velocity dispersion, σ, for each pixel was computed from the Hα velocity profiles after correction from instrumental and thermal widths, assuming that all profiles (instrumental, thermal) are described by Gaussian functions. The velocity dispersion was estimated using σ = (${\sigma }_{\mathrm{obs}}^{2}$ - ${\sigma }_{\mathrm{inst}}^{2}$ - ${\sigma }_{\mathrm{th}}^{2}$)^1/2, where σ_inst = FSR/(${ \mathcal F }$ × $2\sqrt{2\mathrm{ln}2}$) = 16.4 km s⁻¹ and σ_th = 9.1 km s⁻¹. FSR = 912 km ⁻¹ is the free spectral range of PUMA and ${ \mathcal F }$ is its Finesse. The thermal broadening was computed according to σ_th = (kT_e /m_H)^1/2 assuming an electronic temperature of T_e = 10⁴ K (Spitzer & Arny 1978).

Figure 3 depicts the main kinematic properties of the nebulae possibly associated with the ULX sources. The left column shows the radial velocity (top) and velocity dispersion (bottom) fields of NGC 925 obtained from the PUMA lambda cube. The right panels show close-ups of the monochromatic Hα image (not calibrated in flux) and the radial velocity field, centered at the locations of the three ULXs.

Zoom In Zoom Out Reset image size

We find that ULX-1 and ULX-2 are associated with large diameter nebulae. The dimensions of the nebulae are listed in Table 1. The diameters were measured from our monochromatic images taking into account the PUMA plate scale of 129 pixel⁻¹ and assuming a distance to NGC 925 of 9.2 Mpc. The diameters of the nebulae are similar to others found for ULXs located inside nebulae (e.g., Pakull & Mirioni 2002; Abolmasov et al. 2007).

We integrated the radial velocity profiles of the pixels corresponding to each of the nebulae in order to get a global velocity profile of each nebula. We do not detect internal differences in the line profiles, at the limit of PUMA's velocity sampling resolution (∼20 km s⁻¹). The systemic velocities of the nebulae follow the rotation of the galaxy and are also listed in Table 1. For ULX-3, we did not detect a nebulae, but two faint intensity blobs around the E-W direction separated by ∼350 pc. The size and distance to each blob is listed in Table 1.

The range of velocity dispersion values in the nebulae associated with ULX-1 and ULX-2 are shown in Table 1, with values from 70–100 km s⁻¹ and 50–107 km s⁻¹, respectively. The velocity dispersion of the blobs associated with ULX-3 range from 26–104 km s⁻¹ for E_blob and 47–100 km s⁻¹ for W_blob. The high velocity dispersion of the nebulae and blobs associated with the ULXs suggest the presence of shocks ionizing the gas. However, we cannot resolve any expansion signature, presumably driven by an expanding bubble (ULX bubble) or jets from the ULX source. On the other hand, ULX bubbles can be distinguished from supernova remnants (SNRs) and HII regions due to their larger size (diameters of ∼100–300 pc, e.g., Pakull & Grisé 2008), comparable to the ones found for ULX-1 and ULX-2 in this work.

The first method we use is the S-calibration (Pilyugin & Grebel 2016), which relies on the following strong line ratios: S₂ = [S ii] λ6717 + λ6731/Hβ, R₃ = [O iii] λ4959 + λ5007/Hβ, and N₂ = [N ii] λ6548 + λ6584/Hβ. For a detailed description of the method, see Pilyugin & Grebel (2016). The S-calibration takes into account dependencies on the ionization parameter, is compatible with the metallicity scale of H ii regions, and shows a smaller scatter compared to auroral line metallicities.

A signal-to-noise (S/N) cut of 3 is applied in each of the emission lines used in the method. Additionally, we select only SF regions using the standard [N ii]-BPT diagram (Baldwin et al. 1981), adopting the discrimination of K03, as shown in Figure 4. Our final SF sample consists of 1692 spaxels. The metallicity map for NGC 925 is shown in Figure 5 (left panel).

Zoom In Zoom Out Reset image size

In order to further characterize the ULX environment, we also estimate the ionization parameter U, defined as the ratio of ionizing photon density to hydrogen density. The ionization parameter has proven to be a useful diagnostic tool, since any force that clears ionized gas from the interior of an HII region will have the effect of reducing the ionization parameter (Yeh & Matzner 2012).

We estimate the ionization parameter following the prescription of Dors et al. (2017):

$$\begin{eqnarray}&&\mathrm{log}{(U)}_{D}=c\times S2+d\end{eqnarray} \tag{ 1 }$$

where S2 = log([S ii]λλ6717,31/Hα), c = −0.26 × (Z/Z_⊙) − 1.54, and $d=-3.69\,\times \,{(Z/{Z}_{\odot })}^{2}$ + 5.11 × (Z/Z_⊙) −5.26. We used the oxygen abundance of the Sun as Z_⊙ = 8.69 (Amarsi et al. 2018). The resulting ionization map is shown in Figure 5 (right panel).

To estimate the metallicity and ionization gradients, we use the isophotal radius R₂₅ = 321.6 arcsec from Schmidt et al. (2016). We estimate the inclination i = 687 and position of the major axis (PA) = 2933 through the rotational curve using the approach of Pilyugin et al. (2019). Linear fits are performed in "R" with the package "HYPER-FIT" see Robotham & Obreschkow (2015) for details.

The metallicity gradient, using galactrocentric distances R_g = R/R₂₅, is shown in Figure 6. The fitted parameters yield:

$$\begin{eqnarray}&&12+\mathrm{log}{({\rm{O}}/{\rm{H}})}_{{\rm{S}}}=-0.0948(\pm 0.0056){{\rm{R}}}_{{\rm{g}}}+8.3124(\pm 0.0038)\end{eqnarray} \tag{ 2 }$$

with a root mean square error (RMSE) of the residuals of 0.07 dex. We define the RMSE as $\sqrt{{\sum }_{i=1}^{n}{({\hat{y}}_{i}-{y}_{i})}^{2}/n}$, where ${\hat{y}}_{i}$ are the predicted values, and y_i the observed values. The metallicity gradient we find is typical for galaxies of this mass (e.g., Kreckel et al. 2019; Zinchenko et al. 2019b).

Similarly, we analyze the ionization gradient, obtaining the following:

$$\begin{eqnarray}&&\mathrm{log}{({\rm{U}})}_{D}=0.3813(\pm 0.0248){{\rm{R}}}_{g}-3.267(\pm 0.016),\end{eqnarray} \tag{ 3 }$$

with an RMSE = 0.2427 (see Figure 6). The values of metallicity and ionization for ULX-1 and ULX-3 are listed in Table 2.

Interestingly, Figure 6 indicates that the data from the two fibers overlapping with ULX-1 have a very low gas metallicity for their galactocentric distance. Furthermore, ULX-3 exhibits a very high ionization parameter. This is discussed further in Section 5.

To confirm the previous result using a second independent method, we use the Bayesian spectral analysis tool beagle (Chevallard & Charlot 2016), which incorporates models for the emission from stars and photoionized gas. Specifically, the stellar models (latest version of the Bruzual & Charlot 2003, stellar evolutionary code) are computed using a standard Chabrier (2003) IMF with an upper mass cutoff of 100 M_⊙. These models have been combined with CLOUDY (v13.03 Ferland et al. 2013) to compute the corresponding nebular emission from the ionized gas in HII regions (Gutkin et al. 2016). For our analysis, we adopt a constant star formation history and the model of Charlot & Fall (2000) to describe dust attenuation. One interesting feature of BEAGLE is that it takes into account multiple parameters that can impact the line ratios simultaneously and thus the reported error bars appropriately reflect the parameter degeneracies. We remeasured the flux-calibrated spectra using the software Pyspeclines ²⁰ by fitting Gaussians to the spectra prior the continuum subtraction and correction for stellar absorption. This is because BEAGLE templates include both continuum and stellar absorption features. The emission lines fed into BEAGLE are Hα, Hβ, [OIII]λ5007, [NII]λ6584, [SII]λ6716, 6731. We use BEAGLE to fit the line fluxes of the three ULX spectra, as well as the 399 spaxels at the galactocentric distances 0.4 < R/R₂₅ < 0.6 indicated in Figure 6.

Zoom In Zoom Out Reset image size

Our results are illustrated in Figure 7, which shows the probability distributions of the nebular physical properties: oxygen abundance 12+log(O/H)_B, interstellar metallicity (Z_ISM)_B, the dust-to-metal mass ratio (ξ_d)_B, and the ionization parameter at the edge of the Strömgren sphere log(U_S)_B. We note that the two ionization parameters inferred with the BEAGLE and Dors et al. (2017) methods are not directly comparable; the first is the ionization parameter at the edge of the Strömgren sphere, while the second at the inner edge of the gas cloud. A summary of the median values for 12+log(O/H)_B, log(U_S)_B and (Z_ISM)_B is given in Table 2. We note that the gas metallicities given by BEAGLE depend on a grid with values of (Z_ISM)_B and (ξ_d)_B (Gutkin et al. 2016), which would add to the gas metallicities an intrinsic error of ∼0.05 dex based on the average spacing of the grid.

Zoom In Zoom Out Reset image size

We find that the difference in gas metallicities between the S-calibration and BEAGLE ranges from ∼0.1 to ∼0.4 dex. This is in agreement with previous work comparing photoionization-based metallicities with the direct-Te method (e.g., Vale Asari et al. 2016). Clearly, the gas metallicity 12+log(O/H) in the environment of ULX-1 is lower in comparison to ULX-3. On the other hand, the ionization parameter of ULX-3 is just slightly higher than the ULX-1 spectra.

In order to highlight the low-metallicity values of the ULX-1 region using the two different metallicity methods, we show a histogram of the 399 spaxels around ULX-1 in Figure 8, in the interval 0.4 < R/R₂₅ < 0.6. The median metallicity using the S-calibration in this interval is 8.26, with a 1σ scatter of 0.075. The gas metallicities of ULX-1(∗) and ULX-1(•) are 0.16 and 0.22 dex below the median value, respectively. Since comparisons between different metallicity methods reveal large discrepancies (e.g., Kewley & Ellison 2008; López-Sánchez et al. 2012), we also show in Figure 8 the gas metallicity distribution estimated by BEAGLE for the same spaxels. The median metallicity is 8.42, with a 1σ scatter of 0.14, where the gas metallicities of ULX-1(∗) and ULX-1(•) are 0.12 and 0.16 dex below the median value, respectively. Both the S-calibration and BEAGLE estimations confirm that the gas metallicities of ULX-1(∗) and ULX-1(•) are very low even compared to other HII regions in the galaxy at the same galactocentric radius.

Zoom In Zoom Out Reset image size

We find that ULX-1 and ULX-2 are associated with nebulae of the order of (major/minor axis) 285/285 and 285/232 pc, respectively. The dimensions of these nebulae suggest they are ULX bubbles, which can be distinguished from SNRs and HII regions due to their larger size (∼100–300 pc; Pakull & Grisé 2008). ULX-3 is associated with two nearby Hα blobs that are separated by ∼350 pc. Furthermore, the high velocity dispersion of the nebulae and blobs suggest the presence of shocks, although we cannot resolve any expansion signature with the PUMA data. On the other hand, the analyzed BPT diagrams indicate the ionization is consistent with HII regions. We explore as well the [SII]/Hα ratio, that can distinguish between shock-ionized and photoionized regions, where shock-ionized regions have [SII]/Hα > 0.4 (Dodorico et al. 1980). For our data, these ratios are 0.43, 0.29, and 0.2 for ULX-1(•), ULX-1(∗) and ULX-3, respectively. Therefore, while it indicates that ULX-1(•) is possibly consistent with a shock-ionized region, the result is inconclusive as ULX-1(∗) is below this threshold.

To further explore the possibility of shocks in the ULX-1 spaxels and their impact on the gas metallicity, we use the MANGA sample of Pilyugin et al. (2020). From this sample, we examine the gas metallicity, using the S-calibration, of SF-type regions with gas velocity dispersions of ∼100 km s⁻¹, and spaxels with sizes of ∼300 pc. We find that the gas metallicity for these regions follows the metallicity gradient of the host galaxy. Therefore we conclude that if shocks were present, they would have a minimum impact on the gas metallicities.

We note that the combination of different resolutions and depths make the maps from PUMA and the GMS not directly comparable. For instance, the high resolution data of PUMA does detect Hα emission in the vicinity of ULX-2 (see Figure 3). However, the same region is not as bright in the Hα surface brightness map from the GMS (Figure 2). This can be explained due to the different resolution, plus the extra requirement of the Hβ line in emission to correct for the extinction in the GMS data. Additionally, data from PUMA show two bright blobs around the location of ULX-3. It is likely that the lower spatial resolution of GMS is mixing the information of these blobs, and that a deeper image from PUMA could potentially unveil more gas around the ULX-3 position.

The ULX emission line ratios are located in the star-forming region of the BPT diagram, thus allowing us to use conventional prescriptions for gas metallicity and ionization. We note, however, that photoionization models predict that AGN-ionized regions with low metallicity can populate the left side of the [NII]-BPT diagram. For instance, if the emission lines of any of the ULXs are powered by accretion around a black hole, we would not identify it from the BPT diagram if it is in a regime with Z < ∼1/3 Z_⊙ (e.g., Groves et al. 2006; Feltre et al. 2016), which is the case for ULX-1.

From the GMS observations, we find that the two fibers closest to ULX-1 have a very low metallicity with respect to their galactocentric distance. This result is supported by metallicities estimated with the S-calibration and BEAGLE (see Figures 6 & 8). On the other hand, their ionization parameter seems to be within the dispersion.

Overall, the results for ULX-1 agree with Heida et al. (2016), who suggest a bright nebulae in the vicinity of ULX-1 with very low gas metallicity. However, we are uncertain of the origin of the emission lines, as pointed out earlier in this section.

We emphasize that IFU analyses, with surveys such as Metal-THINGS, enable the study of the entire region surrounding ULXs, and not just the potentially contaminated regions very close to ULXs, as is the case for single fiber or multiobject spectroscopy (MOS) of point-like sources.

Results in the literature show that ULXs are frequently found in low-metallicity environments (e.g., Linden et al. 2010; Mapelli et al. 2010; Prestwich et al. 2013; Basu-Zych et al. 2016; López et al. 2019; Schaerer et al. 2019). Among the possible explanations, Linden et al. (2010) show that the majority of the ULX population can be explained by the tail of the high-mass X-ray binary (HMXB) population, where the low-metallicity environment favors active Roche lobe overflow (RLO-HMXBs). In this scenario, the lower metallicity binaries will have, on average, tighter orbits that will result in an increased number of RLO-HMXBs that can drive much higher accretion rates. Therefore, lower metallicity HMXB populations are expected to be more luminous (e.g., Linden et al. 2010; Fragos et al. 2013).

A different ULX origin is proposed by Kim et al. (2017), who conclude that a ULX in NGC 5252 is most likely the nucleus of a small, low-mass galaxy accreted by NGC 5252. To explore this possibility, we report a histogram of the metallicity of the 27 spaxels within a radius of ∼400 pc centered on ULX-1, and find a very similar distribution as shown in Figure 8 using the S-calibration, with a median metallicity of 8.24, and a 1σ scatter of 0.076. Since the low-metallicity gas is not spatially extended, we consider this possibility highly unlikely for ULX-1.

In contrast to ULX-1, the gas metallicity and ionization parameter of ULX-3 are both high (Figure 6). This is more consistent with the presence of a neutron star within an evolved stellar population, as postulated by Earnshaw et al. (2020). This scenario is not uncommon, since several neutron star ULXs have been discovered in recent years (e.g., Bachetti et al. 2014; Fürst et al. 2016; Israel et al. 2016; Carpano et al. 2018; Chandra et al. 2020).