MNRAS 000, 1–11 (2023) Preprint 5 January 2023 Compiled using MNRAS LATEX style file v3.

Multiwavelength Scrutiny of X-ray Sources in Dwarf Galaxies:

ULXs versus AGN

Erica Thygesen,1★ Richard M. Plotkin,2,3 † Roberto Soria,4,5,6 Amy E. Reines,7 Jenny E. Greene,
Gemma E. Anderson,9 Vivienne F. Baldassare,10 Milo G. Owens,2 Ryan T. Urquhart,1 Elena Gal
James C. A. Miller-Jones, 9 Jeremiah D. Paul,2 and Alexandar P. Rollings2
1 Center for Data Intensive and Time Domain Astronomy, Department of Physics and Astronomy, Michigan State University, East Lansing, MI 48824, USA
2 Department of Physics, University of Nevada, Reno, NV 89557, USA
arXiv:2301.01317v1 [astro-ph.HE] 3 Jan 2023

3 Nevada Center for Astrophysics, University of Nevada, Las Vegas, NV 89154, USA
4 College of Astronomy and Space Sciences, University of the Chinese Academy of Sciences, Beijing 100049, China
5 INAF - Osservatorio Astrofisico di Torino, Strada Osservatorio 20, 10025, Pino Torinese, Italy
6 Sydney Institute for Astronomy, School of Physics A28, The University of Sydney, Sydney, NSW 2006, Australia
7 eXtreme Gravity Institute, Montana State University, Bozeman, MT, USA
8 Department of Astrophysical Sciences, Princeton University, Princeton, NJ 08544, USA
9 International Centre for Radio Astronomy Research, Curtin University, GPO Box U1987, Perth, WA 6845, Australia
10 Department of Physics and Astronomy, Washington State University, Pullman, WA 99163, USA
11 Department of Astronomy, University of Michigan, 1085 S University, Ann Arbor, MI 48109, USA

Accepted XXX. Received YYY; in original form ZZZ

ABSTRACT
Owing to their quiet evolutionary histories, nearby dwarf galaxies (stellar masses 𝑀★ .
3 × 109 𝑀  ) have the potential to teach us about the mechanism(s) that ‘seeded’ the growth of
supermassive black holes, and also how the first stellar mass black holes formed and interacted
with their environments. Here, we present high spatial-resolution observations of three dwarf
galaxies in the X-ray (Chandra), the optical/near-infrared (Hubble Space Telescope), and the
radio (Karl G. Jansky Very Large Array). These three galaxies were previously identified as
hosting candidate active galactic nuclei on the basis of lower resolution X-ray imaging. With our
new observations, we find that X-ray sources in two galaxies (SDSS J121326.01+543631.6
and SDSS J122111.29+173819.1) are off nuclear and lack corresponding radio emission,
implying they are likely luminous X-ray binaries. The third galaxy (Mrk 1434) contains two
X-ray sources (each with 𝐿 X ≈ 1040 erg s−1 ) separated by 2.008, has a low-metallicity (12
+ log (O/H) = 7.8), and emits nebular He ii 𝜆4686 line emission. The northern source has
spatially coincident point-like radio emission at 9.0 GHz and extended radio emission at 5.5
GHz. We discuss X-ray binary interpretations (where an ultraluminous X-ray source blows a
‘radio bubble’) and active galactic nucleus interpretations (where a ≈ 4 × 105 𝑀  black hole
launches a jet). In either case, we find that the He ii emission cannot be photoionised by the
X-ray source, unless the source was ≈30—90 times more luminous several hundred years ago.
Key words: galaxies: dwarf — stars: black holes — radio continuum: galaxies — X-rays:
galaxies

1 INTRODUCTION known to host nuclear black holes (e.g., Filippenko & Ho 2003;
Barth et al. 2004; Reines et al. 2011, 2013; Schramm et al. 2013;
There is abundant evidence that supermassive black holes (SMBHs;
Moran et al. 2014; Sartori et al. 2015; Mezcua et al. 2016, 2018;
106 . 𝑀BH . 109 𝑀  ) ubiquitously exist at the centres of large
Pardo et al. 2016; Ho & Kim 2016; Chen et al. 2017; Chilingarian
galaxies (e.g., Kormendy & Ho 2013), some of which accrete and
et al. 2018; Nguyen et al. 2019; Baldassare et al. 2020; Martínez-
shine as active galactic nuclei (AGNs). Some lower-mass dwarf
Palomera et al. 2020; Cann et al. 2021; Schutte & Reines 2022), with
galaxies (which we define by stellar masses 𝑀★ . 3 × 109 𝑀  ) are
some mass estimates as low as 𝑀BH ≈104 𝑀  (e.g., Baldassare et al.
2015; Woo et al. 2019). These black holes represent the lower-mass
★ E-mail: thygesen@msu.edu end of the SMBH population, and we refer to them here as ‘massive
† E-mail: rplotkin@unr.edu

© 2023 The Authors

2 E. Thygesen et al.
black holes’ (mBHs; 104 . 𝑀BH . 106 𝑀  ). An actively accreting Table 1. Properties of the three dwarf galaxies in our sample. Column
mBH can affect how dwarf galaxies provide feedback to their larger 1: galaxy names. The full designations of the second and third galax-
scale environments (e.g., Dashyan et al. 2018; Trebitsch et al. 2018; ies are SDSS J121326.01+543631.6 and J122111.29+173819.1. Column
Mezcua et al. 2019), and more generally, mBHs represent a phase 2: distances to each galaxy, assuming 𝐻0 = 73 km s−1 Mpc−1 for Mrk
that nuclear black holes must pass through as they grow to SMBH 1434 and SDSS J1213, and using the Tully-Fisher relation for SDSS
sizes over cosmological time scales (e.g., Volonteri 2010). Given J1221(Kashibadze et al. 2020). Column 3: stellar masses, following the
that dwarf galaxies have had relatively quiet evolutionary histories, methodology of Reines & Volonteri (2015). Column 4: logarithm of star
formation rates, based on far-ultraviolet and infrared luminosities (Hao et al.
constraining the fraction of dwarf galaxies hosting mBHs in the
2011; Kennicutt & Evans 2012). Column 5: metallicities when available in
local Universe, along with the mBH mass distribution, may provide
the literature (taken from Shirazi & Brinchmann 2012 for Mrk 1434 and
constraints on the mechanism(s) that formed the first black holes in Zhao et al. 2013 for SDSS J1221).
the Universe (e.g., Ricarte & Natarajan 2018; Inayoshi et al. 2020;
Volonteri et al. 2021). The fraction of dwarf galaxies hosting an Name D log 𝑀★ log SFR 12+log (𝑂/𝐻 )
mBH is still relatively unknown, with current empirical constraints (Mpc) (𝑀 ) (𝑀 yr−1 )
implying & 30−50% occupation (Miller et al. 2015; Gallo & Sesana (1) (2) (3) (4) (5)
2019; Greene et al. 2020).
Mrk 1434 30.7 6.6 −0.9 7.8
Stellar mass black holes (𝑀BH ≈10 𝑀  ) and neutron stars are
SDSS J1213 32.7 7.3 −2.2 ...
also observed within some dwarf galaxies in the form of X-ray bi-
SDSS J1221 16.1 8.0 −1.5 8.3
naries (XRBs). XRBs serve as probes of stellar populations within
galaxies, with the number and/or luminosity of XRBs expected to
scale with the star formation rate, stellar mass, and metallicity of
density at frequency 𝜈, and the radio spectral index 𝛼 = 0 for a flat
the host galaxy (e.g., Grimm et al. 2003; Gilfanov 2004; Linden
spectrum; Ho 2008). Such unresolved, flat spectrum radio emission
et al. 2010; Mineo et al. 2014; Lehmer et al. 2021). Intriguingly,
is usually interpreted as a partially self-absorbed synchrotron jet
lower-metallicity galaxies appear to contain an excess of luminous
(Blandford & Königl 1979). Meanwhile, rapidly accreting XRBs
XRBs compared to Solar-metallicity galaxies (Prestwich et al. 2013;
do not launch jets that would be detectable beyond distances of a
Brorby et al. 2014; Douna et al. 2015; Ponnada et al. 2020; Lehmer
few Mpc (Fender et al. 2004). Thus, the presence of unresolved
et al. 2021), which may be a consequence of lower-metallicity pro-
radio emission has the potential to exclude hard X-ray sources as
genitor stars having weaker stellar winds, and therefore producing
rapidly accreting XRBs.
black hole remnants that are more numerous and/or more massive
In this paper, we present high spatial-resolution X-ray (Chan-
(e.g., Heger et al. 2003; Mapelli et al. 2010). Besides tracing stel-
dra), optical/near-infrared (Hubble Space Telescope; HST), and ra-
lar populations, the energy output from XRBs can also provide
dio observations (Karl G. Jansky Very Large Array; VLA) of three
feedback to their host galaxies. For example, line emission from the
nearby dwarf galaxies that each host at least one hard X-ray point
high-ionisation He ii 𝜆4686 line (𝜒ion = 54.4 eV) has been observed
source. These three targets were initially identified as AGN candi-
from some ultraluminous X-ray sources (ULXs),1 which is often
dates by Lemons et al. (2015), but with the caveat that the positions
interpreted as an X-ray photoionised nebula (Pakull & Angebault
of their X-ray sources were poorly determined. From the multiwave-
1986; Moon et al. 2011). Extrapolating such ULX phenomenology
length data presented here, we better locate the positions of the X-ray
in the local Universe to higher redshifts, XRBs could have con-
sources within these three galaxies, and we attempt to constrain the
tributed to the X-ray heating of the intergalactic medium during the
nature of each source (i.e., XRB or mBH). In Section 2 we detail
Epoch of Reionisation and Cosmic Dawn (e.g., Mirabel et al. 2011;
our sample selection and data reduction. We outline our results in
Ponnada et al. 2020). Thus, characterising both the XRB and mBH
Section 3, followed by a discussion in Section 4. Our conclusions
populations in nearby dwarf galaxies, particularly as a function of
are presented in Section 5. Unless stated otherwise, uncertainties
host galaxy metallicity, is important for understanding the formation
are reported at the 68% confidence level.
of the first black holes and galaxies in the Universe.
X-ray observations are commonly used to identify accreting
compact objects, since hard X-ray emission (&1–2 keV) is a uni-
versal signature of accretion. However, in several cases, it is very 2 OBSERVATIONS AND DATA REDUCTION
challenging to determine the mass of an accreting object via X-
ray observations alone. In particular, both a rapidly-accreting XRB 2.1 Sample
and a weakly-accreting mBH/SMBH can have comparable X-ray
Our three targets were selected from the survey by Lemons et al.
luminosities in the 1039 − 1041 erg s−1 range, and they can also
(2015), who cross matched ∼44,000 nearby dwarf galaxies (𝑧 <
display similar X-ray spectral shapes (below ≈50 keV). Folding
0.055) from the NASA-Sloan Atlas2 to the Chandra Source Catalog
in multiwavelength information is therefore essential for differ-
(CSC Release 1.1; Evans et al. 2010). They identified 19 galaxies
entiating between rapidly accreting XRBs and weakly-accreting
with hard X-ray point sources (2–7 keV), of which 10 contained an
mBHs/SMBHs. It is well established that weakly accreting SMBHs
X-ray source positionally consistent with the galaxy optical centre
(i.e., low-luminosity AGNs with 𝐿 bol . 0.01𝐿 Edd , where 𝐿 bol is the
(given positional uncertainties, we note that not every galaxy has
bolometric luminosity and 𝐿 Edd = 1.3×1038 [𝑀BH /𝑀  ] erg s−1 is
a well defined nucleus). They presented these 10 galaxies as AGN
the Eddington luminosity) emit compact, usually unresolved, radio
candidates.3
emission with a flat spectrum ( 𝑓 𝜈 ∝ 𝜈 −𝛼 , where 𝑓 𝜈 is the radio flux

2 http://www.nsatlas.org/
1 We define ULXs as having X-ray luminosities 𝐿𝑋 > 1039 s−1 .
erg ULXs 3 Since publication of Lemons et al. (2015), there is new theoretical evi-
are most commonly interpreted as super-Eddington neutron star or black dence that mBHs do not need to reside in the nucleus (e.g., Bellovary et al.
hole XRBs (see, e.g., Feng & Soria 2011; Kaaret et al. 2017). 2019).

MNRAS 000, 1–11 (2023)

 X-ray Sources in Dwarf Galaxies 3
Chandra provides exquisite spatial resolution (≈0.00 4) for tar- SDSS catalog using wcs_match. We found only two common X-
gets located at the telescope’s aimpoint, but the resolution degrades ray/optical sources for Mrk 1434, zero common sources for SDSS
for sources located farther away. Of the 10 AGN candidates in J1213, and one common source for SDSS J1221. Thus, we applied a
Lemons et al. (2015), they found that four galaxies contain X- translational astrometric correction for Mrk 1434 (Δ𝑥 = 0.97, Δ𝑦 =
ray sources that are far enough from the aimpoint to have large 1.32 pixels) and for SDSS J1221 (Δ𝑥 = 0.01, Δ𝑦 = 0.96 pixels)
positional uncertainties (>5 00 , which is comparable to the pro- using wcs_update. No astrometric correction was applied to SDSS
jected size of the entire dwarf galaxy). Of these four galaxies, J1213.
three contained X-ray sources with hard X-ray luminosities >3𝜎 We next re-ran wavdetect on the aligned event files (filtered
(>1.2 dex) larger than expected from the galaxy-wide contribu- from 0.5-7 keV, now including each target dwarf galaxy) to de-
tion from X-ray binaries, given the stellar mass and star forma- termine positions in the aligned reference frame of X-ray sources
tion rate of each galaxy (see Sections 4.3 and 4.4 of Lemons et al. hosted by each dwarf galaxy. We used the same wavdetect param-
2015). These three galaxies include: Mrk 1434 (𝑧 = 0.00747), SDSS eters as above, except we used ecf=0.3 when generating the point
J121326.01+543631.6 (𝑧 = 0.00797; hereafter SDSS J1213), and spread function map to allow the detection of fainter point sources.
SDSS J122111.29+173819.1 (𝑧 = 0.00699; hereafter SDSS J1221; wavdetect identified two X-ray sources in Mrk 1434, one source in
see Table 1). Of particular interest, Mrk 1434 is a metal-poor galaxy SDSS J1213, and one source in SDSS J1221. The positions of each
(12 + log (𝑂/𝐻) = 7.8; Shirazi & Brinchmann 2012) and its optical X-ray source are listed in Table 2. We estimated 95% uncertainties
spectrum from the Sloan Digital Sky Survey (SDSS; York et al. of each X-ray position based on the distance from the telescope aim-
2000) shows nebular He ii line emission (Shirazi & Brinchmann point and the number of counts detected by wavdetect, following
2012). Equation 5 in Hong et al. (2005). Note, this 95% positional uncer-
To better constrain the locations of the X-ray sources relative tainty represents the statistical error on each source. For SDSS J1213
to their host galaxies, we obtained new Chandra X-ray and HST in particular, where we could not perform an astrometric alignment
optical/near-infrared observations for these three galaxies. We also of the Chandra image, there is an additional systematic uncertainty
obtained new VLA radio observations for one target, SDSS J1213, that could be as large as 2 00 (although 0.00 8 is more typical).4
while archival VLA data were already available for the other two We then measured the number of counts from each X-ray
sources. We adopt distances for each galaxy based on their redshifts, source using srcflux. We adopted circular apertures centred at
using 𝐻0 = 73 km s−1 Mpc−1 , except for SDSS J1221, which is a each wavdetect position with radii of 5 pixels, except for Mrk
member of the Virgo cluster (VCC 459). For this galaxy, we use a 1434, which contains two X-ray sources, where we adopted radii of
distance of 16.1 Mpc based on the Tully-Fisher relation (Kashibadze 2.5 pixels to avoid the regions from each X-ray source from over-
et al. 2020). For all three galaxies, we adopt star formation rates lapping. The number of background counts per pixel was estimated
from Lemons et al. (2015), which are based on (dust-corrected) far- from nearby source-free regions of each image. These measure-
ultraviolet and infrared luminosities and the relationships from Hao ments were performed in both broad (0.5-7.0 keV) and hard (2.0-7.0
et al. (2011) and Kennicutt & Evans (2012). For SDSS J1221, we keV) images, and we detected 19–73 counts from each source in the
scale the star formation rate from Lemons et al. (2015) to 16.1 Mpc. broad band and 8–23 counts in the hard band. All X-ray detections
For stellar mass estimates, following Reines & Volonteri (2015), (in all bands) are significant at the >99% level according to the
we use the colour-dependent mass-to-light ratios from Zibetti et al. confidence tables in Kraft et al. (1991).
(2009). Finally, spectra were extracted for each X-ray source us-
ing specextract and fit using an absorbed powerlaw model
(tbabs*powerlaw) in the Interactive Spectral Interpretation Sys-
2.2 Chandra tem v1.6.2 (ISIS; Houck & Denicola 2000), adopting Cash statis-
We obtained new Chandra observations (Cycle 17; PI Plotkin) with tics (Cash 1979) given the relatively low number of counts per
each galaxy centred at the aimpoint of the S3 chip of the Advanced source. We initially left the column density as a free parameter.
CCD Imaging Spectrometer (ACIS; Garmire et al. 2003). Data However, for three X-ray sources 𝑁 𝐻 converged to zero, in which
were telemetered in VFAINT mode. Chandra data reduction was case we froze the value to the Galactic column density and refit the
carried out using the Chandra Interactive Analysis of Observations spectrum. Model fluxes were calculated using the cflux convolu-
(ciao) software version 4.13 (Fruscione et al. 2006) and caldb tion model. Spectral parameters and model fluxes are reported in
v4.9.5. The Chandra data were reprocessed using chandra_repro Table 3.
to create new level 2 event files and bad pixel files, and to apply the
latest calibration files. We then searched for background flares using
the deflare script, and we did not find any periods with elevated 2.3 Hubble Space Telescope
background levels. We observed each galaxy with the Wide Field Camera 3 (WFC3)
Next we aligned the event file astrometry to the SDSS reference aboard HST for one orbit per galaxy (PI Plotkin; program 14356).
frame. We first excluded areas on each X-ray image occupied by For Mrk 1434 and SDSS J1221 we observed in both the F110W
the dwarf galaxy, so that our astrometric corrections would not be and F606W filters (with the IR and UVIS channels, respectively),
influenced by sources within each target galaxy. We then filtered and for SDSS J1213, which is a fainter galaxy, we took observations
each Chandra image to 0.5-7.0 keV and ran wavdetect to identify only in the F110W filter. Observations in each filter were taken over
X-ray point sources, adopting wavelet scales of 1,2,4,8, and 16, four dither positions, and we used the IRSUB512 subarray for Mrk
setting sigthresh to 10−6 (i.e., approximately one false positive 1434 and SDSS J1221. Total exposure times in each filter are listed
per chip), and using a point spread function map (at 2.3 keV) with in Table 4. Data were downloaded from the Mikulski Archive for
an enclosed count fraction (ecf) of 0.9. The relatively large ecf Space Telescopes (MAST), and individual exposures were aligned
was chosen to help filter out weak X-ray sources, which would not
have sufficient positional accuracy for astrometric alignment. We
then cross-matched X-ray sources identified by wavdetect to the 4 https://cxc.harvard.edu/cal/ASPECT/celmon/

MNRAS 000, 1–11 (2023)

4 E. Thygesen et al.

Table 2. Details of Chandra observations. Column 1: name of X-ray source. Column 2: Chandra obsID. Column 3: date of observation. Column 4: exposure
time. Columns 5 & 6: right ascension and declination of each X-ray source. Column 7: radius of the 95% positional uncertainty of each Chandra source, based
on Equation 5 of Hong et al. (2005). Column 8: aperture corrected net count rate (in counts per ks) in the broad X-ray band (0.5-7.0 keV). Aperture corrections
of 0.90, 0.95, and 0.96 were used for Mrk 1434, SDSS J1213, and SDSS J1221, respectively. Column 9: aperture corrected net count rate (in counts per ks) in
the hard band (2.0-7.0 keV). Aperture corrections of 0.87, 0.93, and 0.93 were used for Mrk 1434, SDSS J1213, and SDSS J1221, respectively.

Source obsID Date Exp Time Right Ascension Declination 𝑝err Net Rate (0.5-7.0 keV) Net Rate (2.0-7.0 keV)
(ks) (J2000) (J2000) (00 ) (ks−1 ) (ks−1 )
(1) (2) (3) (4) (5) (6) (7) (8) (9)

Mrk 1434 X-N 18059 2016 Jan 26 5.0 10:34:10.19 +58:03:49.0 0.35 8.00 ± 2.22 2.98 ± 1.40
Mrk 1434 X-S 18059 2016 Jan 26 5.0 10:34:10.11 +58:03:46.3 0.36 6.83 ± 2.04 2.06+1.36
−0.94
SDSS J1213 18060 2016 Aug 04 7.0 12:13:26.12 +54:36:34.1 0.38 2.78 ± 1.10 1.15+0.89
−0.60
SDSS J1221 18061 2016 Feb 13 7.0 12:21:11.00 +17:38:18.0 0.33 10.82 ± 2.11 3.46 ± 1.24

Table 3. Chandra spectral parameters, fluxes, and luminosities. Column 1: name of X-ray source. Column 2: column density. Column 3: best-fit photon index.
Column 4: best-fit Cash statistic/degrees of freedom. Columns 5 & 6: logarithms of the unabsorbed model X-ray flux and luminosity from 0.5-10 keV, estimated
using the cflux convolution model. Columns 7 & 8: logarithms of the unabsorbed model X-ray flux and luminosity from 2-10 keV, estimated using the cflux
convolution model.

Broad (0.5-10.0 keV) Hard (2.0-10.0 keV)

Source 𝑁𝐻 Γ C-stat/d.o.f. log Flux log Luminosity log Flux log Luminosity
(1020 cm−2 ) (erg s−1 cm−2 ) (erg s−1 ) (erg s−1 cm−2 ) (erg s−1 )
(1) (2) (3) (4) (5) (6) (7) (8)

Mrk 1434 X-N <56.9𝑎 1.3 ± 0.4 16.0/13 −12.8 ± 0.1 40.2 ± 0.1 −13.0 ± 0.2 40.1 ± 0.2
Mrk 1434 X-S 0.6𝑏 1.7 ± 0.4 10.3/13 −13.1 ± 0.1 40.0 ± 0.1 −13.3 ± 0.2 39.8 ± 0.2
SDSS J1213 1.4𝑏 1.3 ± 0.5 5.8/12 −13.3 ± 0.2 39.8 ± 0.2 −13.5 ± 0.2 39.6 ± 0.2
SDSS J1221 2.7𝑏 1.6 ± 0.3 27.1/32 −12.8 ± 0.1 39.6 ± 0.1 −13.0 ± 0.1 39.5 ± 0.1
𝑎 Best-fit column density 𝑁 = 8.0 × 1020 cm−2 , reported as an upper limit (95% confidence level) because the uncertainty on
𝐻
the best-fit value extends down to the Galactic value of 0.6 × 1020 cm−2 .
𝑏 Column density frozen to the Galactic value during fitting, taken from Dickey & Lockman (1990).

and combined using AstroDrizzle in the DrizzlePac software Table 4. Summary of HST observations. Column 1: galaxy name. Column
(Hack et al. 2013).5 The F110W drizzled images were created with 2: date of observations. Column 3: filters used for observations. Column 4:
plate scales 0.00 06 pix−1 for Mrk 1434 and SDSS J1221, and 0.00 09 exposure times in the F110W/F606W filters, respectively, when both filters
pix−1 for SDSS J1213. All F606W images have plate scales 0.00 03 were used. All observations were taken through HST Proposal ID 14356.
pix−1 .
We aligned the HST astrometry to the Gaia Data Release Source Date Filter Exp. Time
2 (Gaia Collaboration et al. 2018) reference frame using the (min)
tweakreg task within AstroDrizzle (after excluding sources (1) (2) (3) (4)
falling within each galaxy).6 For Mrk 1434, the corrections re- Mrk 1434 2016 Apr 16 F110W/F606W 8.6/30.9
sulted in astrometric shifts by (Δ𝑥 = 1.8, Δ𝑦 = 0.0) pixels (from SDSS J1213 2016 Apr 16 F110W 43.7
two common sources) and (Δ𝑥 = 1.9, Δ𝑦 = 2.2) pixels (from nine SDSS J1221 2016 Apr 9 F110W/F606W 8.6/26.9
common sources) in the F110W and F606W filters, respectively. For
SDSS J1213, we shifted the F110W filter by (Δ𝑥 = 0.6, Δ𝑦 = 2.8)
pixels (five common sources). Finally, for SDSS J1221 we could not aligned the F110W filter to the F606W filter (via three common
identify enough common sources between the HST image and the sources between the two HST filters).
Gaia catalog in the F110W filter (which has a smaller field of view).
So, we only aligned the F606W filter to the Gaia frame, shifting by
(Δ𝑥 = 0.2, Δ𝑦 = 5.3) pixels (four common sources), and we then 2.4 Very Large Array
Mrk 1434 and SDSS J1221 both had archival datasets (PI Satyapal,
14A-358) from the VLA, while new data were obtained for SDSS
5 https://hst-docs.stsci.edu/drizzpac
6
J1213 for this study (PI Plotkin, SH0563). All three galaxies were
We note that we aligned HST images to the Gaia frame and the Chandra observed in the most extended A configuration. Both Mrk 1434 and
X-ray images to the SDSS frame, because we generally found a larger num-
SDSS J1221 observations were from 4.5-6.5 GHz (C band) and
ber of common HST/Gaia sources vs. common HST/SDSS sources (and
vice-versa for Chandra). Compared to the statistical uncertainty on each
8-10 GHz (X band), while SDSS J1213 was observed only from
Chandra position (0.00 3–0.00 4), we do not expect a meaningful offset between 8-12 GHz.
the absolute astrometry of SDSS vs. Gaia, such that systematic uncertainties The Common Astronomy Software Applications (CASA; CASA
in our astrometric alignments are dominated by the small number of sources Team et al. 2022) software package version 5.1 was used to carry
used to apply the corrections. out standard data reduction. We used 3C 286 to perform delay

MNRAS 000, 1–11 (2023)

 X-ray Sources in Dwarf Galaxies 5
and bandpass calibrations, and to set the flux density scale. Nearby extended emission is not simply resolved out at the higher radio fre-
phase calibrators (see Table 5) were observed to solve for the time- quency, since the smallest baselines of the VLA in A configuration
dependent complex gain solutions. Imaging was performed using are sensitive to structures up to ≈5 00 at 9.0 GHz, which is larger
the task tclean, using two Taylor terms (nterms=2) to account for than the ≈1 00 angular size of the 5.5 GHz emission.
the wide fractional bandwidth and natural weighting to maximise At 9.0 GHz, the chance of a random alignment of a background
sensitivity. We achieved root-mean-square (rms) sensitivities rang- radio point source falling within the Chandra error circle is very
ing from 3.7 to 8.7 𝜇Jy bm−1 in each observing band (see Table 5). small. Integrating the differential source counts tabulated by de
The only X-ray source for which we found coincident radio Zotti et al. (2010) at 8.4 GHz, and assuming a flat radio spectrum
emission is Mrk 1434 X-N, where we found radio detections at both (as expected if the 9.0 GHz emission is from a compact jet; see
5.5 and 9.0 GHz within the X-ray error circle. We used imfit to fit Section 4.1.2), we expect only ≈3 × 10−5 sources with 𝑓peak >
two-dimensional Gaussians in the image plane (at each frequency) 0.036 mJy within the X-ray error circle. The chance of a statistical
to calculate the size of the radio structure, and to measure peak fluctuation as large as 0.036 mJy (i.e., 4𝜎rms ) within the X-ray error
and integrated flux densities. As discussed further in Section 3.1, circle (which contains ≈240 pixels in the radio map) is also very
the 5.5 GHz emission is slightly extended (with integrated flux small (𝑝 = 3 × 10−5 ). Thus, we believe the 9.0 GHz emission is
density 𝑓int = 0.191 ± 0.036 mJy) while the 9.0 GHz is point-like indeed physically associated with the galaxy. However, we note that
( 𝑓peak = 0.036 ± 0.009 mJy). The centroids of the radio emission the radio source lies toward the edge of the X-ray error circle. Thus,
at each frequency are offset by 0.00 20 ± 0.00 07. For the other two even though the radio source formally falls within the Chandra
galaxies, we place 3𝜎rms limits on their radio flux densities. We positional uncertainty, its association specifically with Mrk 1434
note that SDSS J1221 displays radio emission aligned with a likely X-N is less clear, particularly after considering that the Chandra X-
H ii region toward the eastern outskirts of the galaxy that is not ray astrometry of Mrk 1434 was aligned to the optical frame using
associated with X-ray emission, so we do not discuss that radio only two common X-ray/SDSS sources.
emission in this paper. Finally, we note that towards the southwest of the 0.00 35 Chan-
dra X-ray error circle of Mrk 1434 X-N, there is an optical/near-
infrared source that appears red in the HST composite image (see the
3 RESULTS zoom-in of Figure 1a). If that source is a background quasar it may
also be responsible for the X-ray and/or radio emission. However,
In the following subsections we present the multiwavelength results the random alignment of such a background quasar is very unlikely,
for each galaxy, deferring discussions regarding the possible nature as described below. The AB magnitude of the HST source in the
of each X-ray source to Section 4. Composite HST images are shown F606W filter is 18.8, which we convert to SDSS i≈18.7 assuming
for each galaxy in Figure 1, including the locations of X-ray sources. a typical quasar spectrum (Vanden Berk et al. 2001). We then con-
sider SDSS Type 1 quasar counts from 0.3 < 𝑧 < 3.5 (Richards
et al. 2006; Ross et al. 2013), and we find only a negligible number
3.1 Mrk 1434 of background quasars (≈ 6 × 10−7 ) are likely to fall within the
Mrk 1434 hosts two X-ray sources separated by 2.00 8 (see Figure 1a), Chandra X-ray circle by random chance (note, the random align-
both of which are classified as ULXs: the northern source (Mrk ment of a radio-loud or a Type 2 quasar would be even rarer). That
1434 X-N), which is located toward the galactic nucleus, has an source is likely intrinsic to the galaxy.
unabsorbed hard X-ray luminosity 𝐿 2−10 keV = (1.2 ± 0.6) × 1040
erg s−1 , and the southern source (Mrk 1434 X-S) has 𝐿 2−10 keV =
3.2 SDSS J1213 and SDSS J1221
(5.8 ± 0.2) × 1039 erg s−1 . The X-ray spectra of each source are
fit by powerlaw models with photon indices of Γ = 1.3 ± 0.4 for SDSS J1213 and SDSS J1221 each contain a single hard X-ray
Mrk 1434 X-N and Γ = 1.7 ± 0.4 for Mrk 1434 X-S. Neither source point source near the outskirts of each galaxy (Figure 1b-c). The
shows evidence for significant intrinsic absorption. hard (2-10 keV) X-ray luminosities of the sources are 𝐿 2−10 keV =
It is unlikely that either hard X-ray source is a superposed fore- (4.3 ± 2.4)×1039 and (2.9 ± 0.8)×1039 erg s−1 , respectively (Table
ground/background object. Given the density and flux distribution 2), such that both sources are classified as ULXs. The chance of a
of hard X-ray sources in the cosmic X-ray background (see, e.g., superposed foreground/background object is negligible (we expect
Equation 2 of Moretti et al. 2003), we expect to only find 0.001 and only 0.005 hard X-ray background sources for SDSS J1213 and
0.003 hard X-ray sources with 2-10 keV fluxes similar (or brighter) 0.001 sources for SDSS J1221; Moretti et al. 2003). Neither galaxy
than Mrk 1434 X-N and Mrk 1434 X-S, respectively, within the contains radio emission within the Chandra X-ray circles to 3𝜎rms
projected size of the galaxy (which we conservatively approximate upper limits of < 1.4 × 1035 erg s−1 at 10.0 GHz for SDSS J1213,
as a circle with a 20 00 radius). and to limits of < 3.0 × 1034 erg s−1 and < 4.9 × 1034 erg s−1 at
Radio emission is detected only from the northern source, Mrk 5.5 and 9.0 GHz, respectively, for SDSS J1221.
1434 X-N. At 5.5 GHz, the emission is extended with major and
minor axis full width half maxima of 1.00 1×0.00 6 (160 pc × 90 pc),
respectively, covering ≈3.5 synthesised beams. The centroid of the
4 DISCUSSION
5.5 GHz emission is 0.00 16 from the X-ray position (for reference,
the 95% Chandra error circle is 0.00 35), and the integrated luminos- In the following subsections we discuss possible interpretations for
ity is 𝐿 5.5,int = (1.2 ± 0.2) × 1036 erg s−1 . At 9.0 GHz we detect the X-ray sources in our sample of three dwarf galaxies. We focus
a point source located 0.00 32 from the X-ray position, with a peak primarily on Mrk 1434 since it exhibits the most complex phe-
luminosity 𝐿 9.0,peak = (3.7 ± 0.8) × 1035 erg s−1 . We do not detect nomenology (i.e., two X-ray sources, one of which is coincident
any extended radio structures at 9.0 GHz, thereby indicating that the with radio emission). We provide arguments for/against XRB in-
emission seen at 5.5 GHz has a steep radio spectrum (our 5.5 and terpretations in Section 4.1.1 and for/against AGN interpretations
9.0 GHz radio maps have similar sensitivities; see Table 5). Note, in Section 4.1.2. In Section 4.1.3 we discuss whether the observed

MNRAS 000, 1–11 (2023)

6 E. Thygesen et al.

Table 5. Summary of VLA observations. Column 1: galaxy name. Column 2: VLA Program ID. Column 3: date of observation. Column 4: name of phase
calibrator. Column 5: central frequency of each observation. Column 6: bandwidth of each observation. Column 7: the time spent integrating on each galaxy.
Column 8: the size of the (elliptical) synthesised beam along the major and minor axes. Column 9: rms noise of each image.

Source Program Date Phase Calibrator 𝜈 Δ𝜈 𝜏 𝜃bm 𝜎rms

(J2000) (GHz) (GHz) (min) (00 × 00 ) (𝜇Jy bm−1 )
(1) (2) (3) (4) (5) (6) (7) (8) (9)

Mrk 1434𝑎 14A-358 2014 Feb 24 1035+564 5.5 2.0 8.5 0.45×0.38 8.7
Mrk 1434𝑏 14A-358 2014 Feb 24 1035+564 9.0 2.0 8.5 0.27×0.24 8.6
SDSS J1213 SH0563 2016 Sep 30 1219+482 10.0 4.0 39.5 0.28×0.23 3.7
SDSS J1221 14A-358 2014 Feb 26 1158+248 5.5 2.0 25.8 0.42×0.38 5.9
SDSS J1221 14A-358 2014 Feb 26 1158+248 9.0 2.0 26.0 0.25×0.23 5.8

int = 0.191 ± 0.036 mJy and 𝑓peak =

𝑎 Extended radio emission detected near Mrk 1434 X-N at 5.5 GHz, with 𝑓

0.054 ± 0.009 mJy bm−1 . The centroid of emission is located at RA=10h 34m 10.1867s ± 0.0042𝑠 , Dec=58◦ 030 49.00 1481
± 0.00 0763.
𝑏 Point-like radio emission detected near Mrk 1434 X-N at 9.0 GHz, with 𝑓 −1
peak = 0.036±0.009 mJy bm . The emission
is located at RA=10h 34m 10.2045s ± 0.0039s , Dec=58◦ 030 49.00 2883 ± 0.00 0460.

(a) Mrk 1434 3"

0.5"

0.5 arcsec

3 arcsec

(b) SDSS J1213 (c) SDSS J1221

5"
5"

5 arcsec 5 arcsec

Figure 1. (a) Composite HST image of Mrk 1434 in the F606W (blue/green) and the F110W (red) filters. The locations of the two X-ray point sources are
shown as red cross hairs, with the dashed red circles illustrating the sizes of the 95% positional errors from Chandra. The Zoom-in of the centre of the galaxy
shows the location of Mrk 1434 X-N relative to the radio emission, where yellow contours show the extended 5.5 GHz radio emission (1.00 1 × 0.00 6; contours
drawn at 3, 4, 5×𝜎rms ) and the magenta contours show the unresolved emission at 9.0 GHz (contours drawn at 3, 4×𝜎rms ). The sizes of the VLA synthesised
beams are 0.00 45 × 0.00 38 (5.5 GHz) and 0.00 27 × 0.00 24 (9.0 GHz), respectively. Note, the SDSS spectroscopic fibre, from which the nebular He ii emission is
detected, has a diameter of 300 and is placed at the centre of the galaxy. (b) HST image of SDSS J1213 in the F110W filter, with the location of the X-ray source
marked by the red cross hair and dashed circle. (c) HST composite image of SDSS J1221 in the F606W (blue/green) and the F110W (red) filters, with the
location of the X-ray source marked by the red cross hair and dashed circle. In all images, north is up and east is to the left.

MNRAS 000, 1–11 (2023)

 X-ray Sources in Dwarf Galaxies 7
X-ray flux is sufficient to explain He ii line emission observed in most likely be synchrotron radiation with a steep spectrum from
the SDSS spectrum of Mrk 1434. A discussion on the nature of the supernova remnants. Note, our data exclude free-free radio emission
X-ray sources in the other two galaxies is presented in Section 4.2. from an H ii region, which would produce a flat spectrum that would
be detectable at 9.0 GHz.

4.1 Mrk 1434

4.1.2 AGN Interpretations
4.1.1 XRB Interpretations
AGN can also produce extended radio emission, which is another
As shown in Section 3.1, both X-ray sources in Mrk 1434 are viable explanation for the 5.5 GHz radio structure. However, in light
physically associated with the galaxy and luminous enough to be of the discussion in the previous subsection that a super-Eddington
classified as ULXs. The observed X-ray luminosity, however, is XRB is also capable of producing the observed extended emission
higher than expected from the luminous tail of the galaxy’s XRB at 5.5 GHz, the resolved radio complex does not provide useful
population. The luminosities of both X-ray sources are above the diagnostics for attempting to discriminate between XRB vs. AGN.
cutoff of the low-mass XRB luminosity function (e.g., Gilfanov Since the X-ray spectra of Mrk 1434 X-N and Mrk 1434 X-S (Γ =
2004), so in the following we only consider high-mass XRBs using 1.3 ± 0.4 and Γ = 1.7 ± 0.4, respectively) are consistent with low-
the metallicity-dependent luminosity function from Lehmer et al. luminosity AGNs (Younes et al. 2011; Yang et al. 2015), we focus
(2021). For Mrk 1434, with 12 + log (𝑂/𝐻) = 7.8 and SFR=0.12 the following discussion on AGN scenarios with Eddington ratios
𝑀  yr−1 , Lehmer et al. (2021) predict a total 0.5-8.0 keV X-ray 𝐿 bol /𝐿 Edd . 0.01. For such weakly accreting AGN, we expect to
luminosity (i.e., from all X-ray point sources) of 𝐿 0.5−8.0 keV = observe unresolved radio emission from a partially self-absorbed
(1.7 ± 0.15) × 1039 erg s−1 (where the error bar represents the 68% compact jet (Ho 2008). By combining X-ray and radio luminosities,
confidence interval provided by Lehmer et al. 2021). They also we can then make crude estimates on black hole masses by appealing
predict only 0.03+0.04
−0.02
ULXs with 𝐿 0.5−8.0 > 1040 erg s−1 . For to the fundamental plane of black hole activity (Merloni et al. 2003;
reference, the unabsorbed 0.5–8.0 keV model luminosities of Mrk Falcke et al. 2004). For Mrk 1434 X-N, we then interpret the the
1434 X-N and Mrk 1434 X-S are (1.3 ± 0.4)×1040 and (0.8 ± 0.2)× unresolved 9.0 GHz radio emission as arising from a compact jet,
1040 erg s−1 , respectively. Thus, the combined X-ray luminosity of and we utilise the fundamental plane regression by Gültekin et al.
both ULXs is ≈10 times higher than expected relative to the Lehmer (2019),
et al. (2021) luminosity function, which is significant even after  
considering uncertainties and intrinsic scatter. log 𝑀BH /108 𝑀  = (0.55 ± 0.22) +
Even though the above suggests that it is statistically unlikely  
for both sources to be XRBs, small number statistics could influence (1.09 ± 0.10) log 𝐿 5 GHz /1038 erg s−1 −
the above arguments, and it is worth exploring XRB interpretations.  
In particular, the extended 5.5 GHz radio emission from Mrk 1434 (0.59 ± 0.16) log 𝐿 2−10 keV /1040 erg s−1 ,
X-N could represent a ‘ULX bubble’, as similar types of extended (1)
radio structures have been observed from other ULXs, making the
radio emission a signature of a ULX outflow shocking the nearby which has an intrinsic scatter ≈1 dex. We estimate that Mrk 1434
interstellar environment (e.g., Pakull et al. 2010; Soria et al. 2010, X-N would have 𝑀BH ≈ 4×105 𝑀  if powered by an mBH (see Ta-
2021; Cseh et al. 2012; Urquhart et al. 2019). If the 5.5 GHz radio ble 6). Note, we assume a flat radio spectrum to convert the observed
emission is indeed a ULX bubble, then with 𝐿 5.5,int = (1.2 ± 0.2) × radio luminosity at 9.0 GHz to 5.0 GHz for use in the fundamen-
1036 erg s−1 it would represent the most luminous ULX bubble yet tal plane (we cannot use our 5.5 GHz radio map to estimate the 5
observed by a factor of ≈6 (Pakull et al. 2010; Soria et al. 2010, GHz luminosity because we do not have enough signal-to-noise to
2021). Meanwhile, the projected size of ≈160 pc × 90 pc (1.00 1×0.00 6) attempt to decompose a point source embedded within the extended
in diameter is fairly typical compared to other ULX bubbles, where radio emission observed at 5.5 GHz). Similarly, the lack of radio
diameters range from ≈25–350 pc (Soria et al. 2021; also see Table 1 emission from Mrk 1434 X-S implies 𝑀BH . 4 × 105 𝑀  (where
of Berghea et al. 2020 and references therein). Taking the peak we adopt a 3𝜎rms upper limit, based on the observed 𝜎rms near
flux density of the 5.5 GHz structure, and extrapolating to 1 GHz Mrk 1434 X-S in our 5.5 GHz image). These mass estimates imply
assuming a spectral index 𝛼 = 0.7, the intensity of the radio bubble Eddington ratios (𝐿 2−10 keV /𝐿 Edd ) of ≈ 2 × 10−4 and & 1 × 10−4 for
in Mrk 1434 X-N would be 𝐼1 GHz ≈ 6 × 10−16 erg s−1 cm−2 Hz−1 Mrk 1434 X-N and Mrk 1434 X-S, respectively, which, assuming
sr−1 , which is relatively large but reasonable compared to other bolometric corrections of ≈10, are consistent with Eddington ratios
ULX radio bubbles with similar physical sizes (see Figure 5 of for which the fundamental plane can be applied (see, e.g., Plotkin
Berghea et al. 2020). et al. 2012).
Although a ULX bubble is one interpretation of the 5.5 GHz
emission, we stress that it is not a unique (or necessary) explanation. 4.1.3 On the Origin of Nebular He ii Emission
Adopting SFR = 0.12 𝑀  yr−1 for Mrk 1434 and the relation
between star formation rate and the 1.4 GHz specific luminosity In the following we determine whether the X-ray emission from Mrk
from Kennicutt & Evans (2012), we expect 𝐿 5.5,SF ≈ 3.9 × 1036 erg 1434 is a strong enough source of photoionisation to explain the
s−1 (we convert from 1.4 GHz to 5.5 GHz assuming a spectral index strength of the He ii emission in the SDSS spectrum of Mrk 1434.
𝛼 = 0.7). Considering that the intrinsic scatter on the conversion The observed He ii line flux is 𝐹4686,obs = (7.5 ± 0.1) × 10−16
between SFR and radio luminosity is on the order of ±0.3 dex erg s−1 cm−2 , which translates to a photon flux of 𝑁4686,obs =
(Murphy et al. 2011), the observed extended structure at 5.5 GHz (1.8 ± 0.1) × 10−4 photons s−1 cm−2 . Every photon emitted in the
could be produced entirely by star formation processes. Since the He ii line requires 5.2 ionizing photons incident on singly ionised
extended radio structure at 5.5 GHz is not detected at 9.0 GHz, helium (Pakull & Angebault 1986). Given the ionisation potential
the dominant radio emission mechanism in such a scenario would of singly ionised helium (𝜒ion = 54.4 eV), and considering that

MNRAS 000, 1–11 (2023)

8 E. Thygesen et al.

Table 6. mBH mass estimates and limits. Column 1: galaxy name. Column times more luminous several hundred years ago, which appears to
2: logarithm of the hard X-ray luminosity. Column 3: logarithm of the radio be on the only viable way for the He ii emission to be powered by
luminosity at 5 GHz, assuming a flat radio spectrum. For Mrk 1434 X-N, X-ray photoionisation.
this luminosity is based on the unresolved emission detected at 9 GHz. For If the extended radio emission is produced by an outflow shock-
all other X-ray sources, limits are placed as 3𝜎rms . Column 4: logarithm of ing the interstellar medium, then one must also consider the pos-
the black hole mass (or limit) if X-ray sources are weakly accreting mBHs, sibility of the He ii emission being produced by ionisation from a
based on the fundamental plane of black hole activity (Gültekin et al. 2019). radiative shock (e.g., Dopita & Sutherland 1995). According to the
Uncertainties on log 𝑀BH are ≈1 dex.
MAPPINGS III libraries of line ratios for radiative shocks (Allen
et al. 2008), assuming a shock velocity of 300 km s−1 , we expect the
Source log 𝐿2−10 keV log 𝐿5 GHz log 𝑀BH
luminosity of the He ii 𝜆4686 emission line 𝐿 4686 ≈ 4 × 10−4 𝐿 rad ,
(erg s−1 ) (erg s−1 ) (𝑀 )
where 𝐿 rad is the total radiative luminosity of the shock.7 Assuming
(1) (2) (3) (4)
that the kinetic power required to inflate a bubble 𝑃kin ≈ 77/27𝐿 rad
Mrk 1434 X-N 40.1 ± 0.4 35.3 ± 0.1 5.6 (Weaver et al. 1977), then explaining the observed He ii line via
Mrk 1434 X-S 39.8 ± 0.3 <35.2 <5.6 shock ionisation requires an outflow with 𝑃kin ≈ 6 × 1041 erg s−1 .
SDSS J1213 39.6 ± 0.4 <34.9 <5.3 We do not have a reliable method to independently estimate
SDSS J1221 39.5 ± 0.2 <34.4 <5.0 𝑃kin (especially considering that other emission lines in the SDSS
spectrum are dominated by star formation). However, for an order of
magnitude estimate, we calculate the minimum synchrotron energy
−3 dependence on
the photoionisation cross section has a steep 𝐸 ph of the 5.5 GHz radio emission, which is 𝑊min ≈ 2 × 1052 erg
photon energy, 𝐸 ph , then producing the observed SDSS He ii line (Longair 1994).8 A 300 km s−1 shock would take ≈ 3 × 105 yr to
flux requires a photon flux in the extreme ultraviolet (54–300 eV) inflate a bubble with a 160 pc diameter, such that the average power
of 𝑁54−300 eV = 5.2𝑁4686,obs = (9.1 ± 0.1) × 10−4 photons s−1 stored in internal energies of the synchrotron emitting structure is
cm−2 . Note, this photon flux is an underestimate because we have 𝑃¯min ≈ 2 × 1039 erg s−1 (i.e., the average power in particles and
not corrected the observed SDSS line flux for extinction. in the magnetic field). Thus, an outflow would need to carry &102
The 3 00 SDSS spectroscopic fibre is centred near Mrk 1434 times more power in order for a shock to be the sole ionisation
X-N, such that if the He ii emission arises from photoionisation by source of the observed He ii emission line. Of course, 𝑃¯min is a
the X-ray source, we expect the emission to be dominated by Mrk minimum energy estimate, and the power in bubbles/cavities carved
1434 X-N. We do not have direct measurements on the extreme out by kinetic outflows have sometimes been observed to be larger,
ultraviolet flux from 54-300 eV, so we extrapolate the Chandra sometimes by factors of several hundreds (e.g., Ito et al. 2008), such
X-ray spectrum into the extreme ultraviolet. Our best-fit powerlaw that the above does not exclude the possibility of shock ionisation.
+2.5 × 10−4 photons s−1 cm−2
model predicts a photon flux of 0.3−0.2 For comparison, the ULX NGC 6946 MF16 (Roberts & Col-
(note the large range in uncertainty because we are extrapolating bert 2003) has a luminous and compact radio bubble (Berghea et al.
the model to energies lower than the Chandra X-ray band). Thus, 2020), which suggests a relatively powerful outflow. Adopting the
while high-energy radiation from Mrk 1434 X-N may contribute to NGC 6946 MF16 bubble line flux in the [Fe ii] 𝜆16440 emission
some of the He ii photoionisation, the observed X-ray source is too line (4.2 × 10−15 erg s−1 cm−2 ) and a distance of 7.8 Mpc (Long
faint, by a factor of ≈30, to supply all of the photoionising photons. et al. 2020), the MAPPINGS III libraries for a 300 km s−1 shock
If we assume a thermal X-ray emission model (tbabs*diskbb), it (with Solar abundances) suggest a kinetic power of 𝑃kin ≈ 7 × 1040
becomes even more difficult for the X-ray source to explain the He ii erg s−1 . Thus, the kinetic power of NGC 6946 MF16 (i.e., one of the
photionisation, as the extrapolated 54-300 eV extreme ultraviolet most powerful known ULX radio bubbles) is an order of magnitude
flux becomes ≈90 times too faint. Adding a contribution of photons lower than the power required for shock ionisation to be responsible
form Mrk 1434 X-S would only increase the above photon flux by for the observed strength of the He ii emission line near Mrk 1434
a factor of ≈2, for either spectral model. X-N. Thus, if the He ii line is powered by shock ionisation, then it
There is currently no evidence for significant X-ray variabil- would represent one of the most powerful bubbles carved by a ULX
ity from Mrk 1434 over the past 1–2 decades. Coincidentally, the outflow yet observed.
SDSS spectrum and the archival Chandra observation from Lemons Intriguingly, Mrk 1434 is one member of a population of 182
et al. (2015, Chandra obsID 3347) were both taken in May 2002 star forming galaxies with nebular He ii emission that were identi-
(separated by ≈2 weeks). The archival data from 2002 show nearly fied by Shirazi & Brinchmann (2012). The ratios of He ii/H𝛽 relative
identical X-ray luminosities (log 𝐿 2−10 keV = 40.1 and 39.9 erg s−1 to [N ii] 𝜆6584/H𝛼 are inconsistent with AGN for these galaxies.
for Mrk 1434 X-N and Mrk 1434 X-S, respectively; see Table 2 of Typically, when an AGN is absent, Wolf-Rayet stars are considered
Lemons et al. 2015) compared to the Chandra observations pre- the primary stellar population capable of producing enough extreme
sented here, which were taken nearly 14 years later (see Table 2 of ultraviolet flux above the 54 eV He ii ionisation edge. However, Shi-
this paper). There are also two X-ray detections of Mrk 1434 in the razi & Brinchmann (2012) inspected the SDSS spectra for broad
third XMM-Newton serendipitous source catalog (3XMM; Rosen emission features indicative of Wolf-Rayet stars, and they found no
et al. 2016) in 2007 and 2008. Both X-ray sources are blended to- Wolf-Rayet signatures in the spectrum of Mrk 1434. Thus, without
gether due to XMM-Newton’s poorer spatial resolution. Comparing
the XMM-Newton fluxes to the combined fluxes of both sources in 7 Given the low metallicity of Mrk 1434, we adopt the MAPPINGS III
the Chandra observations, X-ray variability is smaller than a factor
model grid with Small Magellanic Cloud abundances. We also assume an
of ≈2 over the four observations. However, considering the light interstellar medium density of 1 cm−3 and equipartition of magnetic and
travel time between the X-ray source and the ionised medium, it is thermal pressures.
feasible that Mrk 1434 X-N was more active in the past. The pro- 8 We adopt 𝐿 36 erg s−1 , a bubble diameter of ≈160 pc, and an
5.5 ≈ 10
jected radius of the SDSS spectroscopic fibre is 730 light years, and ion to electron energy ratio of 𝜂 = 40. We note that 𝑊min ∝ 𝜂 4/7 , and the
we cannot exclude the possibility that Mrk 1434 X-N was ≈30–90 proper value of 𝜂 is not well constrained.

MNRAS 000, 1–11 (2023)

 X-ray Sources in Dwarf Galaxies 9
concrete evidence that Mrk 1434 X-N was indeed brighter several extended A configuration. The largest angular scale to which the
hundred years ago to power the He ii emission via photoionisation, VLA is sensitive to radio emission at our observing frequencies
and/or lacking a reliable estimate of the kinetic power of an outflow (X-band) and configuration (A) is 5.00 3, such that our VLA obser-
for shock ionisation, the source of extreme ultraviolet photons in vations would not detect flux from extended jets larger than ≈850
Mrk 1434 remains a mystery. Another plausible explanation could and ≈410 pc for SDSS J1213 and SDSS J1221, respectively. On the
be photoionisation from extreme ultraviolet photons emitted by ex- other hand, the radio cores of weakly accreting AGN (bolometric
otic stellar populations (like rapidly rotating stars) in metal-poor luminosities 𝐿 bol < 0.01𝐿 Edd ) have flat radio spectra and are com-
environments (see the discussion in Section 6 of Shirazi & Brinch- pact enough that their radio emission should not be ‘resolved out’
mann 2012). It is very plausible that several of the above scenarios at VLA resolutions (see, e.g., Orienti & Prieto 2010). Thus, if only
contribute toward producing the He ii line, and Shirazi & Brinch- considering mBHs in the weak accretion regime, we can use our ra-
mann (2012) recovered a heterogeneous population (multiple mech- dio upper limits in conjunction with the fundamental plane to place
anisms may even contribute to producing the He ii emission within mass limits of 𝑀BH < 2 × 105 and <1 × 105 𝑀  for SDSS J1213
a single galaxy). For example, Senchyna et al. (2020) conclude and SDSS J1221, respectively. Requiring 𝐿 bol < 0.01𝐿 Edd , and as-
that X-ray photoionisation cannot explain nebular He ii emission suming X-ray bolometric corrections of 10, then places lower limits
across a sample of nearly a dozen metal-poor galaxies. Meanwhile, on black hole masses of & 3 × 104 (SDSS J1213) and & 2 × 104 𝑀 
there are several well-established examples of X-ray sources that (SDSS J1221). Thus, there is a relatively narrow range of mass
are indeed sufficient to power nebular He ii emission (e.g., Pakull & where our VLA observations could ‘miss’ the compact radio jet
Angebault 1986; Moon et al. 2011; Schaerer et al. 2019; Simmonds from a weakly accreting mBH. Note, our radio limits do not place
et al. 2021). Further observational constraints, ideally via system- useful constraints on the possibility of a more rapidly accreting
atic X-ray surveys of metal-poor dwarf galaxies under high spatial mBH with 𝐿 bol > 0.01𝐿 Edd , which would correspond to a mass
resolution, are required to understand the level to which ULXs con- 𝑀BH . 104 𝑀  for both sources. Nevertheless, even though our
tribute extreme ultraviolet radiation in metal-poor galaxies, which data do not exclude the possibility of mBHs, Occam’s razor proba-
has implications for understanding sources of ionisation and heating bly suggests that the simplest and most likely scenario is that these
of the intergalactic medium in the early Universe. are luminous XRBs.

4.2 SDSS J1213 and SDSS J1221 4.3 An Update to Lemons et al. (2015)
Our new Chandra observations confirm the conclusion of Lemons After considering the above multiwavelength observations, all 10 of
et al. (2015) that both X-ray sources are more luminous than ex- the dwarf galaxy AGN candidates identified by Lemons et al. (2015)
pected from the XRB populations in each galaxy, as described be- (via hard X-ray emission) now have sufficient spatial resolution to
low. Unlike for Mrk 1434, the luminosities of both X-ray sources determine if the X-ray sources indeed reside in galactic nuclei. Our
in SDSS J1213 and SDSS J1221 are low enough that we should study reduces their number of AGN candidates to 7–8 (adopting an
consider both high-mass and low-mass XRBs. Following Lemons AGN definition that requires nuclear sources). It is very unlikely that
et al. (2015), we therefore adopt the relation from Lehmer et al. any of these 7–8 nuclear sources are chance alignments with fore-
(2010), which predicts the hard X-ray luminosity from low-mass ground/background X-ray emitting objects. Adopting the hard (2-10
and high-mass XRBs as a function of stellar mass and star for- keV) X-ray fluxes and X-ray position error circles of the nuclear can-
XRB /erg s−1 = (9.05 ± 0.37) × 1028 (𝑀 /𝑀 ) +
mation rate: 𝐿 2−10 didates from Table 2 of Lemons et al. (2015), and replacing the X-ray
★ 
  flux and positional uncertainty of Mrk 1434 X-N with the values
(1.62 ± 0.22) × 10 39 −1
𝑆𝐹 𝑅/𝑀  yr , with an intrinsic scatter of presented here, the Moretti et al. (2003) cosmic X-ray background
±0.34 dex. The Lehmer et al. (2010) relation predicts 𝐿 2−10 XRB = predicts only 0.003 sources to fall within the nuclei of the eight
37 37 −1
1.2 × 10 and 5.6 × 10 erg s for SDSS J1213 and SDSS J1221, possible nuclear mBH candidates. Obtaining 7–8 viable AGN can-
respectively. The predicted luminosities are ≈3 times higher if we didates is a significant result, considering that (a) the Lemons et al.
instead adopt the calibrations in Lehmer et al. (2019). Thus, the ob- (2015) dwarf galaxy survey was archival and therefore serendipitous
served X-ray luminosities are ≈120–360 and ≈17–50 times higher in nature, and (b) the three dwarf galaxies with follow-up presented
than expected, for SDSS J1213 and SDSS J1221, respectively.9 here represent three of their most unlikely AGN candidates (given
In light of recent theoretical motivation for ‘wandering’ mBHs the poor spatial resolution of their archival Chandra data). Lemons
(Bellovary et al. 2019, 2021, also see, e.g., Mezcua & Domínguez et al. (2015) found X-ray sources in 19 galaxies total (i.e., the re-
Sánchez 2020; Reines et al. 2020; Greene et al. 2021; Sargent maining 11–12 galaxies host off-nuclear X-ray sources, most likely
et al. 2022 for observational searches), an X-ray source being ‘off- XRBs). Thus, if a luminous X-ray source is detected in a dwarf
nucleus’ does not on its own preclude the possibility of an accreting galaxy, our study (very roughly) implies a 30–40% chance10 that
mBH. It is possible that these sources are mBHs launching jets it could be a nuclear mBH, which supports the viability of using
that are either (a) beneath our radio detection limit or (b) that are X-ray surveys to identify mBHs in low-mass galaxies, as long as
very extended and ‘resolved out’ by the VLA when it is in its most the survey is performed with sufficient sensitivity and spatial res-
olution. We stress the importance of high spatial-resolution X-ray
observations. For example, Mrk 1434 was previously identified as
9 The Lehmer et al. (2010) relation is calibrated to galaxies with approxi- an AGN from an XMM-Newton survey (Birchall et al. 2020), while
mately Solar metallicities. The metallicity of SDSS J1213 is unknown, and our higher spatial-resolution Chandra observation clearly resolves
the metallicity of SDSS J1221 is log (𝑂/𝐻 ) + 12 = 8.3 (Zhao et al. 2013).
If we adopt the metallicity-dependent Lehmer et al. (2021) relation for high-
mass XRBs, the X-ray luminosity of the X-ray source in SDSS J1221 is still 10 This number is an upper limit, and it neglects biases inherent to an
≈20 times higher than expected for a galaxy with its star formation rate and archival/serendipitous survey, which is out of the scope of this paper to
metallicity. quantify.

MNRAS 000, 1–11 (2023)

10 E. Thygesen et al.
the ‘nuclear’ X-ray source into two distinct sources (and even then, ment. This research made use of Astropy,11 a community-developed
it remains unclear if either source is indeed an accreting mBH). core Python package for Astronomy (Astropy Collaboration et al.
2013, 2018).

5 SUMMARY AND CONCLUSIONS DATA AVAILABILITY

We have presented a multiwavelength study of three nearby dwarf The data underlying this article are available in the Chandra Data
galaxies that host ULXs. Two galaxies in our sample, SDSS J1213 Archve under ObsIDs 18059, 18060, and 18061 (https://cda.
and SDSS J1221, each contain single off-nuclear X-ray sources harvard.edu/chaser/), in the Barbara A. Mikulski Archive
that we suspect are luminous XRBs. The third galaxy, Mrk 1434 for Space Telescopes under program ID 14356 (dx.doi.org/
hosts two X-ray sources separated by 2.00 8. The northern source 10.17909/3bxp-zt07), and in the National Radio Astronomy
(Mrk 1434 X-N) also displays extended radio emission at 5.5 GHz Observatory Data Archive under programs 14-358 and SH0563
and point-like radio emission at 9.0 GHz. It remains unclear if (data.nrao.edu).
the X-ray sources in Mrk 1434 are XRBs or AGNs (especially
Mrk 1434 X-N), although either scenario is intriguing. If XRBs,
then the combined X-ray luminosity of both sources is larger than
REFERENCES
expected for a galaxy with Mrk 1434’s star formation rate and
(low) metallicity. Futhermore, the extended radio emission at 5.5 Allen M. G., Groves B. A., Dopita M. A., Sutherland R. S., Kewley L. J.,
GHz could then represent the most luminous ‘ULX bubble’ ever 2008, ApJS, 178, 20
observed in the radio, although we stress that the 5.5 GHz radio Astropy Collaboration et al., 2013, A&A, 558, A33
emission can also be attributed entirely to star formation within the Astropy Collaboration et al., 2018, AJ, 156, 123
galaxy, or to an AGN jet. Regardless of the correct scenario, we Baldassare V. F., Reines A. E., Gallo E., Greene J. E., 2015, ApJ, 809, L14
Baldassare V. F., Geha M., Greene J., 2020, ApJ, 896, 10
find that the line emission from He ii in Mrk 1434 is inconsistent
Barth A. J., Ho L. C., Rutledge R. E., Sargent W. L. W., 2004, ApJ, 607, 90
with a nebula being powered by the central X-ray source, unless the
Bellovary J. M., Cleary C. E., Munshi F., Tremmel M., Christensen C. R.,
central source underwent a period of higher activity several hundred Brooks A., Quinn T. R., 2019, MNRAS, 482, 2913
years ago, or if the the nebula is shock ionised by an outflow that Bellovary J. M., et al., 2021, MNRAS, 505, 5129
is an order of magnitude more powerful than yet observed from Berghea C. T., Johnson M. C., Secrest N. J., Dudik R. P., Hennessy G. S.,
a ULX. If Mrk 1434 X-N is an AGN, then the 9.0 GHz radio El-khatib A., 2020, ApJ, 896, 117
emission may represent a compact synchrotron jet from a low- Birchall K. L., Watson M. G., Aird J., 2020, MNRAS, 492, 2268
luminosity AGN power by an mBH with 𝑀BH ≈ 4 × 105 𝑀  . Blandford R. D., Königl A., 1979, ApJ, 232, 34
We conclude by stressing the importance of high spatial-resolution Brorby M., Kaaret P., Prestwich A., 2014, MNRAS, 441, 2346
observations when performing multiwavelength searches for mBHs CASA Team et al., 2022, arXiv e-prints, p. arXiv:2210.02276
Cann J. M., et al., 2021, ApJ, 912, L2
in dwarf galaxies.
Cash W., 1979, ApJ, 228, 939
Chen C.-T. J., et al., 2017, ApJ, 837, 48
Chilingarian I. V., Katkov I. Y., Zolotukhin I. Y., Grishin K. A., Beletsky
Y., Boutsia K., Osip D. J., 2018, ApJ, 863, 1
ACKNOWLEDGEMENTS Cseh D., et al., 2012, ApJ, 749, 17
Dashyan G., Silk J., Mamon G. A., Dubois Y., Hartwig T., 2018, MNRAS,
We thank the anonymous referee for helpful comments that im- 473, 5698
proved this manuscript. Support for this work was provided by the Dickey J. M., Lockman F. J., 1990, ARA&A, 28, 215
National Aeronautics and Space Administration through Chandra Dopita M. A., Sutherland R. S., 1995, ApJ, 455, 468
Award Number GO6-17079X issued by the Chandra X-ray Center, Douna V. M., Pellizza L. J., Mirabel I. F., Pedrosa S. E., 2015, A&A, 579,
A44
which is operated by the Smithsonian Astrophysical Observatory
Evans I. N., et al., 2010, ApJS, 189, 37
for and on behalf of the National Aeronautics Space Administration Falcke H., Körding E., Markoff S., 2004, A&A, 414, 895
under contract NAS8-03060. This research is based on observa- Fender R. P., Belloni T. M., Gallo E., 2004, MNRAS, 355, 1105
tions made with the NASA/ESA Hubble Space Telescope obtained Feng H., Soria R., 2011, New Astron. Rev., 55, 166
from the Space Telescope Science Institute, which is operated by Filippenko A. V., Ho L. C., 2003, ApJ, 588, L13
the Association of Universities for Research in Astronomy, Inc., Fruscione A., et al., 2006, in Society of Photo-Optical Instrumentation En-
under NASA contract NAS 5–26555. These observations are as- gineers (SPIE) Conference Series. p. 62701V, doi:10.1117/12.671760
sociated with program HST-GO-14356. Support for Program No. Gaia Collaboration et al., 2018, A&A, 616, A1
HST-GO-14356 was provided by NASA through a grant from the Gallo E., Sesana A., 2019, ApJ, 883, L18
Space Telescope Science Institute, which is operated by the As- Garmire G. P., Bautz M. W., Ford P. G., Nousek J. A., Ricker Jr. G. R.,
2003, in Truemper J. E., Tananbaum H. D., eds, Proc. SPIEVol. 4851,
sociation of Universities for Research in Astronomy, Incorporated,
X-Ray and Gamma-Ray Telescopes and Instruments for Astronomy.. pp
under NASA contract NAS5-26555. RMP and JDP acknowledge 28–44, doi:10.1117/12.461599
support from the National Science Foundation under grant No. Gilfanov M., 2004, MNRAS, 349, 146
2206123. RS acknowledges support from grant number 12073029 Greene J. E., Strader J., Ho L. C., 2020, ARA&A, 58, 257
from the National Natural Science Foundation of China (NSFC). Greene J. E., et al., 2021, ApJ, 917, 17
AER acknowledges support provided by NASA through EPSCoR Grimm H. J., Gilfanov M., Sunyaev R., 2003, MNRAS, 339, 793
grant number 80NSSC20M0231. GEA is the recipient of an Aus-
tralian Research Council Discovery Early Career Researcher Award
(project number DE180100346) funded by the Australian Govern- 11 http://www.astropy.org

MNRAS 000, 1–11 (2023)

 X-ray Sources in Dwarf Galaxies 11
Gültekin K., King A. L., Cackett E. M., Nyland K., Miller J. M., Di Matteo Reines A. E., Sivakoff G. R., Johnson K. E., Brogan C. L., 2011, Nature,
T., Markoff S., Rupen M. P., 2019, ApJ, 871, 80 470, 66
Hack W. J., Dencheva N., Fruchter A. S., 2013, in Friedel D. N., ed., Astro- Reines A. E., Greene J. E., Geha M., 2013, ApJ, 775, 116
nomical Society of the Pacific Conference Series Vol. 475, Astronomical Reines A. E., Condon J. J., Darling J., Greene J. E., 2020, ApJ, 888, 36
Data Analysis Software and Systems XXII. p. 49 Ricarte A., Natarajan P., 2018, MNRAS, 481, 3278
Hao C.-N., Kennicutt R. C., Johnson B. D., Calzetti D., Dale D. A., Mous- Richards G. T., et al., 2006, AJ, 131, 2766
takas J., 2011, ApJ, 741, 124 Roberts T. P., Colbert E. J. M., 2003, MNRAS, 341, L49
Heger A., Fryer C. L., Woosley S. E., Langer N., Hartmann D. H., 2003, Rosen S. R., et al., 2016, A&A, 590, A1
ApJ, 591, 288 Ross N. P., et al., 2013, ApJ, 773, 14
Ho L. C., 2008, ARA&A, 46, 475 Sargent A. J., Johnson M. C., Reines A. E., Secrest N. J., van der Horst A. J.,
Ho L. C., Kim M., 2016, ApJ, 821, 48 Cigan P. J., Darling J., Greene J. E., 2022, ApJ, 933, 160
Hong J., van den Berg M., Schlegel E. M., Grindlay J. E., Koenig X., Laycock Sartori L. F., Schawinski K., Treister E., Trakhtenbrot B., Koss M., Shirazi
S., Zhao P., 2005, ApJ, 635, 907 M., Oh K., 2015, MNRAS, 454, 3722
Houck J. C., Denicola L. A., 2000, in Manset N., Veillet C., Crabtree D., Schaerer D., Fragos T., Izotov Y. I., 2019, A&A, 622, L10
eds, Astronomical Society of the Pacific Conference Series Vol. 216, Schramm M., et al., 2013, ApJ, 773, 150
Astronomical Data Analysis Software and Systems IX. p. 591 Schutte Z., Reines A. E., 2022, Nature, 601, 329
Inayoshi K., Visbal E., Haiman Z., 2020, ARA&A, 58, 27 Senchyna P., Stark D. P., Mirocha J., Reines A. E., Charlot S., Jones T.,
Ito H., Kino M., Kawakatu N., Isobe N., Yamada S., 2008, ApJ, 685, 828 Mulchaey J. S., 2020, MNRAS, 494, 941
Kaaret P., Feng H., Roberts T. P., 2017, ARA&A, 55, 303 Shirazi M., Brinchmann J., 2012, MNRAS, 421, 1043
Kashibadze O. G., Karachentsev I. D., Karachentseva V. E., 2020, A&A, Simmonds C., Schaerer D., Verhamme A., 2021, A&A, 656, A127
635, A135 Soria R., Pakull M. W., Broderick J. W., Corbel S., Motch C., 2010, MNRAS,
Kennicutt R. C., Evans N. J., 2012, ARA&A, 50, 531 409, 541
Kormendy J., Ho L. C., 2013, ARA&A, 51, 511 Soria R., Pakull M. W., Motch C., Miller-Jones J. C. A., Schwope A. D.,
Kraft R. P., Burrows D. N., Nousek J. A., 1991, ApJ, 374, 344 Urquhart R. T., Ryan M. S., 2021, MNRAS, 501, 1644
Lehmer B. D., Alexander D. M., Bauer F. E., Brandt W. N., Goulding A. D., Trebitsch M., Volonteri M., Dubois Y., Madau P., 2018, MNRAS, 478, 5607
Jenkins L. P., Ptak A., Roberts T. P., 2010, ApJ, 724, 559 Urquhart R., et al., 2019, MNRAS, 482, 2389
Lehmer B. D., et al., 2019, ApJS, 243, 3 Vanden Berk D. E., et al., 2001, AJ, 122, 549
Lehmer B. D., et al., 2021, ApJ, 907, 17 Volonteri M., 2010, A&ARv, 18, 279
Lemons S. M., Reines A. E., Plotkin R. M., Gallo E., Greene J. E., 2015, Volonteri M., Habouzit M., Colpi M., 2021, Nature Reviews Physics, 3, 732
ApJ, 805, 12 Weaver R., McCray R., Castor J., Shapiro P., Moore R., 1977, ApJ, 218, 377
Linden T., Kalogera V., Sepinsky J. F., Prestwich A., Zezas A., Gallagher Woo J.-H., Cho H., Gallo E., Hodges-Kluck E., Le H. A. N., Shin J., Son
J. S., 2010, ApJ, 725, 1984 D., Horst J. C., 2019, Nature Astronomy, 3, 755
Long K. S., Blair W. P., Winkler P. F., Lacey C. K., 2020, ApJ, 899, 14 Yang Q.-X., Xie F.-G., Yuan F., Zdziarski A. A., Gierliński M., Ho L. C.,
Longair M. S., 1994, High energy astrophysics. Vol.2: Stars, the galaxy and Yu Z., 2015, MNRAS, 447, 1692
the interstellar medium. Vol. 2 York D. G., et al., 2000, AJ, 120, 1579
Mapelli M., Ripamonti E., Zampieri L., Colpi M., Bressan A., 2010, MN- Younes G., Porquet D., Sabra B., Reeves J. N., 2011, A&A, 530, A149
RAS, 408, 234 Zhao Y., Gao Y., Gu Q., 2013, ApJ, 764, 44
Martínez-Palomera J., Lira P., Bhalla-Ladd I., Förster F., Plotkin R. M., Zibetti S., Charlot S., Rix H.-W., 2009, MNRAS, 400, 1181
2020, ApJ, 889, 113 de Zotti G., Massardi M., Negrello M., Wall J., 2010, A&ARv, 18, 1
Merloni A., Heinz S., di Matteo T., 2003, MNRAS, 345, 1057
Mezcua M., Domínguez Sánchez H., 2020, ApJ, 898, L30 This paper has been typeset from a TEX/LATEX file prepared by the author.
Mezcua M., Civano F., Fabbiano G., Miyaji T., Marchesi S., 2016, ApJ, 817,
20
Mezcua M., Civano F., Marchesi S., Suh H., Fabbiano G., Volonteri M.,
2018, MNRAS, 478, 2576
Mezcua M., Suh H., Civano F., 2019, MNRAS, 488, 685
Miller B. P., Gallo E., Greene J. E., Kelly B. C., Treu T., Woo J.-H., Baldas-
sare V., 2015, ApJ, 799, 98
Mineo S., Gilfanov M., Lehmer B. D., Morrison G. E., Sunyaev R., 2014,
MNRAS, 437, 1698
Mirabel I. F., Dijkstra M., Laurent P., Loeb A., Pritchard J. R., 2011, A&A,
528, A149
Moon D.-S., Harrison F. A., Cenko S. B., Shariff J. A., 2011, ApJ, 731, L32
Moran E. C., Shahinyan K., Sugarman H. R., Vélez D. O., Eracleous M.,
2014, AJ, 148, 136
Moretti A., Campana S., Lazzati D., Tagliaferri G., 2003, ApJ, 588, 696
Murphy E. J., et al., 2011, ApJ, 737, 67
Nguyen D. D., et al., 2019, ApJ, 872, 104
Orienti M., Prieto M. A., 2010, MNRAS, 401, 2599
Pakull M. W., Angebault L. P., 1986, Nature, 322, 511
Pakull M. W., Soria R., Motch C., 2010, Nature, 466, 209
Pardo K., et al., 2016, ApJ, 831, 203
Plotkin R. M., Markoff S., Kelly B. C., Körding E., Anderson S. F., 2012,
MNRAS, 419, 267
Ponnada S., Brorby M., Kaaret P., 2020, MNRAS, 491, 3606
Prestwich A. H., Tsantaki M., Zezas A., Jackson F., Roberts T. P., Foltz R.,
Linden T., Kalogera V., 2013, ApJ, 769, 92
Reines A. E., Volonteri M., 2015, ApJ, 813, 82

MNRAS 000, 1–11 (2023)