11institutetext: European Space Agency (ESA), European Space Research and Technology Centre (ESTEC), Keplerlaan 1, 2201 AZ Noordwijk, the Netherlands
11email: ashley.chrimes@esa.int
22institutetext: Department of Astrophysics/IMAPP, Radboud University, PO Box 9010, 6500 GL Nijmegen, The Netherlands 33institutetext: Department of Physics, University of Warwick, Gibbet Hill Road, CV4 7AL Coventry, United Kingdom 44institutetext: Inter-University Institute for Data Intensive Astronomy, Department of Astronomy, University of Cape Town, Private Bag X3, Rondebosch 7701, South Africa 55institutetext: South African Astronomical Observatory, P.O. Box 9, 7935 Observatory, South Africa 66institutetext: Oxford Astrophysics, Denys Wilkinson Building, Keble Road, Oxford, OX1 3RH, United Kingdom

Multi-wavelength observations of the Luminous Fast Blue Optical Transient AT 2023fhn

Up to similar-to\sim200 days post-explosion
A. A. Chrimes, ESA Research Fellow1122    D. L. Coppejans 33    P. G. Jonker 22    A. J. Levan, 2233    P. J. Groot,, 224455    A. Mummery 66    E. R. Stanway 33
(Received September 15, 1996; accepted March 16, 1997)
Abstract

Context. Luminous Fast Blue Optical Transients (LFBOTs) are a class of extragalactic transients notable for their rapid rise and fade times, blue colour and accompanying luminous X-ray and radio emission. Only a handful have been studied in detail since the prototypical example AT 2018cow. Their origins are currently unknown, but ongoing observations of previous and new events are placing ever stronger constraints on their progenitors.

Aims. We aim to put further constraints on the LFBOT AT 2023fhn, and LFBOTs as a class, using information from the multi-wavelength transient light-curve, its host galaxy and local environment.

Methods. Our primary results are obtained by fitting galaxy models to the spectral energy distribution of AT 2023fhn’s host and local environment, and by modelling the radio light-curve of AT 2023fhn as due to synchrotron self-absorbed emission from an expanding blast-wave in the circumstellar medium.

Results. We find that the neither the host galaxy nor circumstellar environment of AT 2023fhn are unusual compared with previous LFBOTs, but that AT 2023fhn has a much lower X-ray to ultraviolet luminosity ratio than previous events.

Conclusions. We argue that the variety in ultraviolet-optical to X-ray luminosity ratios among LFBOTs is likely due to viewing angle differences, and that the diffuse, yet young local environment of AT 2023fhn - combined with a similar circumstellar medium to previous events - favours a progenitor system containing a massive star with strong winds. Plausible progenitor models in this interpretation therefore include black hole/Wolf-Rayet mergers or failed supernovae.

Key Words.:
Supernovae: individual: AT 2023fhn – Supernovae: general – Stars: black holes – Black hole physics – Stars: winds, outflows – Stars: circumstellar matter

1 Introduction

Luminous Fast Blue Optical Transients (LFBOTs) are a rare class of rapidly-evolving, hot, multi-wavelength extragalactic transients. The prototypical example, AT 2018cow (‘the Cow’, Prentice et al., 2018), is the nearest and best-studied event of this class so far. Its characteristic early-time features include a peak optical absolute magnitude of similar-to\sim–20 with a rapid rise and decay timescale of similar-to\sim5 days, constraining the 56Ni mass in the ejecta to <<<0.0040.0040.0040.004 M, and ruling out standard core-collapse or thermonuclear supernova models (Perley et al., 2019; Margutti et al., 2019). The optical-ultraviolet (UV) emission is well fit by a hot black-body and power-law component, where the black-body temperature was 3×104similar-toabsent3superscript104\sim 3\times 10^{4}∼ 3 × 10 start_POSTSUPERSCRIPT 4 end_POSTSUPERSCRIPT K initially, falling to 1.5×104similar-toabsent1.5superscript104\sim 1.5\times 10^{4}∼ 1.5 × 10 start_POSTSUPERSCRIPT 4 end_POSTSUPERSCRIPT K over two weeks (Prentice et al., 2018; Perley et al., 2019). The optical spectra were largely featureless, with broad hydrogen absorption features (indicative of a high outflow velocity) appearing and disappearing between 2–8 days, and narrow He lines appearing after similar-to\sim20 days. At other wavelengths, AT 2018cow was X-ray and radio bright (Rivera Sandoval et al., 2018; Margutti et al., 2019; Ho et al., 2019; Nayana & Chandra, 2021). The X-ray emission was well in excess of power-law extrapolations from the radio (e.g. Ho et al., 2019) and was also highly variable after a break in the light-curve, which declined as Lt2proportional-to𝐿superscript𝑡2L\propto t^{-2}italic_L ∝ italic_t start_POSTSUPERSCRIPT - 2 end_POSTSUPERSCRIPT after similar-to\sim20 days (Migliori et al., 2024). The broadband X-ray spectrum, and X-ray variability, cannot be explained by an external shock origin, and the synchrotron self-absorbed radio emission - consisting of a slow rise and rapid decay - was instead smoothly evolving, indicating a distinct physical origin from the highly variable X-rays. An interpretation is that a central engine powers the X-ray emission, while an expanding blast wave produces the radio emission (e.g. Ho et al., 2019; Margutti et al., 2019). The slow radio variability timescale set the size of the emission region at 5–6 days at <3×1015absent3superscript1015<3\times 10^{15}< 3 × 10 start_POSTSUPERSCRIPT 15 end_POSTSUPERSCRIPT cm, while the X-ray variability gave a length scale similar-to\simfive times smaller (Ho et al., 2019). Therefore, the X-rays appear to originate from a central engine or internal shock, while the radio emission is generated externally. A claim of quasi-periodic oscillations in the X-rays can interpreted as evidence for a <850Mabsent850subscript𝑀direct-product<850M_{\odot}< 850 italic_M start_POSTSUBSCRIPT ⊙ end_POSTSUBSCRIPT central engine (Pasham et al., 2021), while a separate claim of similar-to\sim250s quasi-periodicity instead implies an intermediate mass (103-105 M) black hole (Zhang et al., 2022). Synchrotron modelling of the sub-millimetre and radio data revealed a mildly-relativistic expansion velocity (similar-to\sim0.1c𝑐citalic_c) into a wind-like extended circumstellar medium (CSM) with a high density of similar-to\sim105 cm-3 (Margutti et al., 2019; Ho et al., 2019). For Wolf-Rayet-like wind speeds of similar-to\sim1000 km s-1, this implies a mass-loss rate M˙=104˙𝑀superscript104\dot{M}=10^{-4}over˙ start_ARG italic_M end_ARG = 10 start_POSTSUPERSCRIPT - 4 end_POSTSUPERSCRIPT-103superscript10310^{-3}10 start_POSTSUPERSCRIPT - 3 end_POSTSUPERSCRIPTMsubscript𝑀direct-productM_{\odot}italic_M start_POSTSUBSCRIPT ⊙ end_POSTSUBSCRIPT yr-1 (Margutti et al., 2019).

Since AT 2018cow, several more LFBOTs have been discovered. Confirmed events include AT 2018lug/ZTF 18abvkwla (‘the Koala’, Ho et al., 2020), CSS161010 (Coppejans et al., 2020), AT 2020xnd/ZTF 20 acigmel (‘the Camel’, Perley et al., 2021; Bright et al., 2022; Ho et al., 2022), AT 2020mrf (Yao et al., 2022), AT 2022tsd (‘the Tasmanian Devil’, Matthews et al., 2023) and AT 2023fhn (‘the Finch’, Chrimes et al., 2024). Despite variety (e.g. in peak luminosity), they share the same key features of hot, largely featureless spectra at early times, optical luminosities rivalling gamma-ray burst (GRB) afterglows and superluminous supernovae, plus bright X-ray and radio emission. They are estimated to occur at <<<0.1% of the local core-collapse supernova rate (Ho et al., 2023d).

Recent developments have provided further insight into the origin of LFBOTs. Polarimetry of AT 2018cow demonstrated the emission region to be highly aspherical, indicative of an accretion disc (Maund et al., 2023). Unexpectedly, AT 2018cow was found to be UV (Sun et al., 2022; Chen et al., 2023; Sun et al., 2023; Inkenhaag et al., 2023) and X-ray (Migliori et al., 2024) bright at late times, several years post-explosion. This emission has been interpreted as from a black hole accretion disc. Estimates for the black hole mass range from similar-to\sim10–100 M (super-Eddington accretion) to similar-to\sim103–104 M (sub-Eddington, from X-ray observations, Migliori et al., 2024) and similar-to\sim1000 M (UV observations, Inkenhaag et al., 2023). Magnetar central engine models struggle to produce both the early and late UV emission (Chen et al., 2023). Further evidence for a black hole accretion scenario comes from minute-long optical flares, up to several months post-explosion, from AT 2022tsd (Ho et al., 2023c). An interpretation is that the central engine is undergoing highly variable, short-lived bursts of accretion.

Several models have been put forward to explain LFBOTs. Tidal disruptions of compact, hydrogen-poor stars (such as white dwarfs) around intermediate mass black holes (IMBHs) can plausibly explain the optical rise and fall timescale, spectral features and X-ray variability timescale (Perley et al., 2019; Kuin et al., 2019). However, the dense CSM inferred from radio observations is hard to explain in such a scenario (e.g. Margutti et al., 2019). Other possibilities include failed supernovae, in which a black hole is formed and the emission is powered by accretion onto the natal black hole rather than radioactive decay in the ejecta (Perley et al., 2019; Quataert et al., 2019), choked jets (e.g. Gottlieb et al., 2022; Soker, 2022), highly aspherical supernovae (‘ellipsars’, DuPont et al., 2022), and the mergers of compact objects and/or massive stars (Lyutikov & Toonen, 2019; Uno & Maeda, 2020; Schrøder et al., 2020), such as black holes and Wolf-Rayet stars (Metzger, 2022). A dense outflow from the progenitor may result in dust echoes (Metzger & Perley, 2023). CSM shock interaction models have also been put forward (e.g. Fox & Smith, 2019; Xiang et al., 2021; Pellegrino et al., 2022; Khatami & Kasen, 2023), but the X-ray variability, broadband spectral evolution, late-time UV/X-ray emission from AT 2018cow and giant optical flares from AT 2022tsd all indicate the presence of a central engine.

In this paper, we present multi-wavelength radio, optical, UV, and X-ray observations of the LFBOT AT 2023fhn up to similar-to\sim200 days post-explosion. We place AT 2023fhn in the context of other LFBOTs so far, in terms of its host galaxy, optical/UV/X-ray light-curve, and radio emission, with the event energetics and blast wave properties inferred from synchrotron modelling of the radio observations. Throughout, we use a flat ΛΛ\Lambdaroman_ΛCDM cosmology with ΩmsubscriptΩm\Omega_{\rm m}roman_Ω start_POSTSUBSCRIPT roman_m end_POSTSUBSCRIPT=0.3 and H0subscript𝐻0H_{0}italic_H start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT=70 kms-1Mpc-1. All magnitudes are reported in the AB system (Oke & Gunn, 1982).

2 Observations and data reduction

2.1 X-ray

We obtained four epochs of Chandra X-ray Observatory (CXO) ACIS-S observations of AT 2023fhn up to similar-to\sim200 days post-explosion. The epochs consist of 1, 2, 6 and 14 observations, respectively (full details are provided in Table 1). The data are reduced, and transient fluxes measured, with standard CIAO (v4.13, caldb v4.9.3, Fruscione et al., 2006) procedures. The images are reprocessed and filtered to the energy range 0.5-7.0 keV. wavdetect is used to find point sources, and srcflux used to measure the flux (or upper limits) at the location of AT 2023fhn. We merged the datasets in each of the four epochs (with merge__\__obs) to increase the signal-to-noise ratio. The mean (mid-point, exposure-time weighted) observation times of these epochs are 15.0, 28.9, 64.5 and 210.9 days (since JD–2460045, or 12:00 UT on 10-Apr-2023). The total exposure times per epoch are similar-to\sim30, 60, 83 and 193 ks respectively. Finally, the fluxes are de-absorbed by assuming a photon index Γ=2Γ2\Gamma=2roman_Γ = 2 (e.g. Rivera Sandoval et al., 2018), and a Galactic neutral hydrogen column density of NH=2.78×1020subscript𝑁H2.78superscript1020N_{\rm H}=2.78\times 10^{20}italic_N start_POSTSUBSCRIPT roman_H end_POSTSUBSCRIPT = 2.78 × 10 start_POSTSUPERSCRIPT 20 end_POSTSUPERSCRIPT cm-2 (Dickey & Lockman, 1990).

Table 1: All CXO observations of AT 2023fhn from programme 24500143 (PI: Chrimes). ObsID, exposure start times (since JD-2460045) and data mode are listed. All observations are made with ACIS-S. The fluxes FX are unabsorbed and measured in the energy range 0.5-7.0 keV. Individual observations in each of the four epochs are merged as indicated. Uncertainties are given at 1σ𝜎\sigmaitalic_σ, upper limits at 2σ𝜎\sigmaitalic_σ.
ObsID Start date texp Data mode FX
JD-2460045 ks erg s-1 cm-2
26624 14.78957 29.68 FAINT (7.61.8+2.2subscriptsuperscriptabsent2.21.8{}^{+2.2}_{-1.8}start_FLOATSUPERSCRIPT + 2.2 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 1.8 end_POSTSUBSCRIPT)×1015absentsuperscript1015\times 10^{-15}× 10 start_POSTSUPERSCRIPT - 15 end_POSTSUPERSCRIPT
26625 27.98310 29.68 FAINT }}\left.\begin{array}[]{l}\\ \end{array}\right\}\hskip 2.84544pt}(4.52.9+4.7subscriptsuperscriptabsent4.72.9{}^{+4.7}_{-2.9}start_FLOATSUPERSCRIPT + 4.7 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 2.9 end_POSTSUBSCRIPT)×1016absentsuperscript1016\times 10^{-16}× 10 start_POSTSUPERSCRIPT - 16 end_POSTSUPERSCRIPT
27833 29.47145 29.67 FAINT
26626 61.80356 16.88 VFAINT }<8.2×1016casesmissing-subexpressionmissing-subexpressionmissing-subexpressionmissing-subexpression8.2superscript1016\left.\begin{array}[]{l}\\ \\ \\ \\ \\ \end{array}\right\}\hskip 2.84544pt<8.2\times 10^{-16}start_ARRAY start_ROW start_CELL end_CELL end_ROW start_ROW start_CELL end_CELL end_ROW start_ROW start_CELL end_CELL end_ROW start_ROW start_CELL end_CELL end_ROW end_ARRAY } < 8.2 × 10 start_POSTSUPERSCRIPT - 16 end_POSTSUPERSCRIPT
27895 62.33516 10.94 VFAINT
27835 65.12251 13.89 FAINT
27905 65.45429 13.89 FAINT
27906 65.79704 13.89 FAINT
27907 66.13969 13.89 FAINT
26627 198.66317 10.74 VFAINT }<3.5×1016casesmissing-subexpressionmissing-subexpressionmissing-subexpressionmissing-subexpressionmissing-subexpressionmissing-subexpressionmissing-subexpressionmissing-subexpressionmissing-subexpressionmissing-subexpressionmissing-subexpressionmissing-subexpression3.5superscript1016\left.\begin{array}[]{l}\\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \end{array}\right\}\hskip 2.84544pt<3.5\times 10^{-16}start_ARRAY start_ROW start_CELL end_CELL end_ROW start_ROW start_CELL end_CELL end_ROW start_ROW start_CELL end_CELL end_ROW start_ROW start_CELL end_CELL end_ROW start_ROW start_CELL end_CELL end_ROW start_ROW start_CELL end_CELL end_ROW start_ROW start_CELL end_CELL end_ROW start_ROW start_CELL end_CELL end_ROW start_ROW start_CELL end_CELL end_ROW start_ROW start_CELL end_CELL end_ROW start_ROW start_CELL end_CELL end_ROW start_ROW start_CELL end_CELL end_ROW end_ARRAY } < 3.5 × 10 start_POSTSUPERSCRIPT - 16 end_POSTSUPERSCRIPT
28997 198.96634 10.74 VFAINT
28998 199.27127 11.12 VFAINT
27837 205.89744 13.4 VFAINT
29031 206.23509 13.3 VFAINT
29032 206.57170 13.5 VFAINT
29034 207.96147 10.93 VFAINT
29033 208.26639 14.39 VFAINT
27838 215.80840 16.85 VFAINT
29054 216.19957 17.84 VFAINT
29056 216.60163 14.39 VFAINT
28991 218.76822 9.94 VFAINT
28999 219.12318 18.69 VFAINT
29055 219.54352 16.85 VFAINT

2.2 UV-optical

A second epoch of HST imaging was obtained on 23/24 October 2023 (the first was on 17 May 2023, Chrimes et al., 2024), using the WCF3 instrument and six filters (F225W,F336W,F555W,F763M,F814W,F845M𝐹225𝑊𝐹336𝑊𝐹555𝑊𝐹763𝑀𝐹814𝑊𝐹845𝑀F225W,F336W,F555W,F763M,F814W,F845Mitalic_F 225 italic_W , italic_F 336 italic_W , italic_F 555 italic_W , italic_F 763 italic_M , italic_F 814 italic_W , italic_F 845 italic_M). Full details are given in Table 2. The data are reduced with drizzlepac (Hoffmann et al., 2021), re-drizzling the charge-transfer-efficiency-corrected __\__flc input images with North oriented up and a final pixel scale of 0.025 arcsec pixel-1 (pixfrac=0.8). Image stamps around the location of AT 2023fhn in epochs 1 and 2 are shown in Figure 1. Visible in the bottom left is the presumed satellite of the larger spiral to the south (see Figure 3). Both galaxies lie at a common redshift of similar-to\sim0.24 (Ho et al., 2023b; Chrimes et al., 2024).

Table 2: All HST data for AT 2023fhn, from program 17238 (PI: Chrimes). Filter, exposure start times (JD-, where is 12:00 UT on 10-APR-2023) and exposure durations texp are given. All observations are with WFC3 in the UVIS channel.
Filter Start date texp
JD–2460045 s
F555W 36.87666 1092
F814W 36.89272 1092
F555W 196.42824 990
F814W 196.44313 1092
F225W 196.49431 1068
F336W 196.51027 1068
F845M 196.56034 990
F763M 196.57515 1068

2.3 Radio

We obtained radio observations with the Karl G. Jansky Very Large Array (VLA) between 22 Apr 2023 and 16 December 2024 (programme SC240143, PI: Chrimes). Details of the observations are listed in Table 3. The observations were taken in standard phase-referencing mode using 3C286 as a flux density and bandpass calibrator, with ICRF J101447.0+230116, FIRST J101644.3+203747, FIRST J101353.4+244916 and ICRF J095649.8+251516 as complex gain calibrators. The observations were calibrated using the VLA Calibration Pipeline versions 2023.1.0.124 and 2022.2.0.64 in CASA 6.5.4 and 6.4.1 respectively, with additional manual flagging. The images were created using the tclean task in CASA with Briggs weighting with a robust parameter of 1. In the observations where the source was not detected we quote the upper limit on the flux density as three times the local RMS. The one exception to this is during the last epoch (see Table 3) where the synthesized beam (resolution element) was large and included other sources. In this case we quoted the upper limit as the flux density at the source location. For the observations where we detected the target, we fitted the flux density using the imfit task within CASA and constrained the fit to the synthesized beam.

The observations up to similar-to\sim12 days post JD-2460045 are already published (Chrimes et al., 2024) and all produced non-detections. In the similar-to\sim87–95 day and 138similar-toabsent138\sim 138∼ 138 day epochs we have sufficient data points for fitting a synchrotron self-absorbed spectrum. The KU band (15 GHz) data point at 138 days has sufficient signal-to-noise to split into 3 (centred on 13, 15 and 17 GHz), as listed in Table 3, increasing the points at similar-to\sim138 days to 7 (with 6 detections). We fit a self-absorbed synchrotron model to the similar-to\sim87-95 and similar-to\sim138 day epochs in Section 4.4.

Table 3: AT 2023fhn flux densities from our VLA programme (SC240143, PI: Chrimes). Observation start times are listed with respect to JD–2460045 (12:00 on 10-Apr-2023). The quoted uncertainties do not include the systematic uncertainty of 5% on the absolute flux calibration at these frequencies. Upper limits are given as 3 times the local RMS. aThe resolution of the last three observations was lower (as the VLA was in D configuration at the time), so we could not remove contaminating sources at the target location and have consequently listed the upper limit as the flux density at the source location.
Start date Freq. Bandwidth texp Flux Density
JD–2460045 GHz GHz min. μJy/beam𝜇Jybeam\mu\rm{Jy/beam}italic_μ roman_Jy / roman_beam
11.80740 1.52 1.024 38 <<<130
11.78309 3.00 2.048 32 <<<35
11.76507 6.00 4.096 23 <<<18
11.74688 10.00 4.096 23 <<<18
11.72090 15.08 6.144 35 <<<11
11.69229 22.00 8.192 35 <<<17
11.66552 33.00 8.192 33 <<<25
87.59185 1.52 1.024 39 <<<45
87.56657 3.00 2.048 33 110±plus-or-minus\pm±8
87.54257 6.00 4.096 32 128±plus-or-minus\pm±5
95.58690 10.00 4.096 31 105±plus-or-minus\pm±7
95.56438 15.00 6.144 30 71±plus-or-minus\pm±7
95.52738 22.00 8.192 47 60±plus-or-minus\pm±10
137.17440 1.52 1.024 39 <<<58
137.14178 3.00 2.048 44 110±plus-or-minus\pm±10
137.10567 6.00 4.096 50 221±plus-or-minus\pm±5
138.16972 10.00 4.096 42 197±plus-or-minus\pm±6
138.13664 13.00 2.050 42 180±plus-or-minus\pm±10
138.13664 15.00 2.050 42 160±plus-or-minus\pm±10
138.13664 16.96 2.050 42 140±plus-or-minus\pm±20
249.95139 6.00 4.096 35 <<<311a
249.90997 10.00 4.096 57 <<<181a
249.86861 15.08 6.144 57 <<<125a
Refer to caption
Figure 1: F555W and F814W HST/WFC3 image stamps at the location of AT 2023fhn in the first epoch (May 2023, left) and the second epoch (October 2023, centre). The right-hand panels show difference images (epoch 1 - epoch 2). North is up, East is left, and the stamps are 2.5 arcsec on each side.

3 Environmental analysis

3.1 Local environment

The second epoch of HST imaging presented in this paper allows us to examine the environment directly underlying the transient after it has faded. As noted by Chrimes et al. (2024), there is diffuse emission in the vicinity of the transient. To characterise this faint underlying population, we place 0.2 arcsec (and 0.4 arcsec) apertures at the location of AT 2023fhn in all six epoch 2 images. The images are aligned with x𝑥xitalic_x-y𝑦yitalic_y shifts using 5 common point sources in every image, with respect to the location of AT 2023fhn in the epoch 1 F555W image. The rms of these relative astrometric alignments is similar-to\sim5–10 mas, better than the absolute astrometry of the images (which have been aligned with the Gaia DR3 reference frame), and much smaller than the aperture size. We perform photometry with photutils, estimating the background with either the median image background (with medianbackground) or an annulus (1.5 to 4 times the aperture radius, with pixels values clipped at 3σ𝜎\sigmaitalic_σ). The appropriate encircled energy corrections for each filter and aperture are applied. Magnitudes are then calculated using the photplam and photflam header keywords111https://hst-docs.stsci.edu/wfc3dhb/chapter-9-wfc3-data-analysis/9-1-photometry, and are listed in Table 4. The only detections are in F555W and F814W. To investigate the nature of these detections, we place eight 0.4 arcsec apertures at equal spacing around the location of AT 2023fhn in a circle of radius 20 pixels (0.5 arcsec). With the F555W filter and median background subtraction, we have significant detections in 5/8 apertures, with a mean magnitude of 25.9±plus-or-minus\pm±0.6 in these apertures - consistent with the measurement at the precise location of AT 2023fhn. This demonstrates that the emission in this area is from an extended, diffuse background, rather than any significant contribution by residual light from AT 2023fhn. This can also be seen in Table 4, where the magnitudes calculated with annulus background subtraction are fainter, since the local background is elevated. Larger apertures also give brighter magnitudes, despite encircled energy correction (unlike point sources in the field). We similarly disfavour any significant contribution from a compact cluster at this specific location, which would appear as a point source in the image given the physical scale at this redshift of similar-to\sim100 pc pixel-1. However, the presence of a globular cluster (which would favour an IMBH interpretation, e.g., Lützgendorf et al., 2013) cannot be ruled out, as even the brightest globular clusters would be far below detection limits at this distance and limiting magnitude (Chrimes et al., 2024). Shifting the circle of apertures 5 arcsec to the north, well away from the galaxies, we find non-detections in all eight apertures with a 3σ𝜎\sigmaitalic_σ upper limit of 26.7. We therefore conclude that there is extended, diffuse emission from an underlying stellar population at the location of AT 2023fhn.

We now estimate the age and dust extinction of this underlying population. First, we correct for the (low) Galactic extinction of E(BV)𝐸𝐵𝑉E(B-V)italic_E ( italic_B - italic_V )=0.0254 (Schlafly & Finkbeiner, 2011)222https://irsa.ipac.caltech.edu/applications/DUST/ using the filter effective wavelengths (Rodrigo et al., 2012; Rodrigo & Solano, 2020) and the Python extinction package (Barbary, 2016) with a Fitzpatrick (1999) extinction law and RV=3.1subscript𝑅V3.1R_{\rm V}=3.1italic_R start_POSTSUBSCRIPT roman_V end_POSTSUBSCRIPT = 3.1. To estimate the age and local (intrinsic) extinction, we fit the Galactic-extinction corrected F225W,F336W,F555W𝐹225𝑊𝐹336𝑊𝐹555𝑊F225W,F336W,F555Witalic_F 225 italic_W , italic_F 336 italic_W , italic_F 555 italic_W and F814W𝐹814𝑊F814Witalic_F 814 italic_W photometry to BPASS (Binary Population and Spectral Synthesis v2.1, Eldridge et al., 2017; Stanway & Eldridge, 2018) single-age spectral templates. These are constructed by assuming that a stellar population of 106M is formed instantaneously, and left to evolve with no further star formation. We use these simple stellar populations since the limited data available to model solely the local environment of AT 2023fhn precludes a more complex procedure, including, for example, the star-formation history (however, see the next section). A fixed metallicity of half-Solar is adopted (Z=0.01𝑍0.01Z=0.01italic_Z = 0.01 by mass fraction). We therefore simply fit for the age of the population, the luminosity (i.e. mass) of the stellar population is then allowed to freely vary to minimise χ2superscript𝜒2\chi^{2}italic_χ start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT. Four data points are used (F225W,F336W,F555W𝐹225𝑊𝐹336𝑊𝐹555𝑊F225W,F336W,F555Witalic_F 225 italic_W , italic_F 336 italic_W , italic_F 555 italic_W and F814W𝐹814𝑊F814Witalic_F 814 italic_W) where the upper limits are treated as data points with zero flux and an uncertainty equal to the flux of the 1σ𝜎\sigmaitalic_σ upper limit. We therefore have 2 fit parameters and 4 data points for 2 degrees of freedom. Fitting is performed by multiplying the (de-redshifted) filter response curves (Rodrigo et al., 2012; Rodrigo & Solano, 2020) with the BPASS spectra to extract fluxes and hence magnitudes from the spectra. These are compared with the absolute magnitudes in each filter, after correction for a range of intrinsic extinction values from AVsubscript𝐴VA_{\rm V}italic_A start_POSTSUBSCRIPT roman_V end_POSTSUBSCRIPT=0.0 to 1.0. The intrinsic extinction correction uses the rest-frame effective wavelength of each filter. The F763M𝐹763𝑀F763Mitalic_F 763 italic_M and F845M𝐹845𝑀F845Mitalic_F 845 italic_M filters are not used in this fit since the upper limits are shallower than the F555W𝐹555𝑊F555Witalic_F 555 italic_W and F814W𝐹814𝑊F814Witalic_F 814 italic_W detections, and so provide no additional constraints.

The results are shown in Figure 2. The top panel shows the best-fit single-age BPASS spectrum. The lower panel shows log(χ2)10{}_{10}(\chi^{2})start_FLOATSUBSCRIPT 10 end_FLOATSUBSCRIPT ( italic_χ start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ) across the parameter space (indicated by the shading). Each pixel represents a unique combination of AVsubscript𝐴VA_{\rm V}italic_A start_POSTSUBSCRIPT roman_V end_POSTSUBSCRIPT and a BPASS simple stellar population at a given age. The 68% and 90% confidence intervals are indicated by white contours (where the Δχ2Δsuperscript𝜒2\Delta\chi^{2}roman_Δ italic_χ start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT intervals are from Avni, 1976).

We also measure the local surface brightness in epoch 2 (in a 0.5 arcsec radius around AT 2023fhn’s position), giving 25.1 mag arcsec-2 in F555W𝐹555𝑊F555Witalic_F 555 italic_W and 24.65 mag arcsec-2 in F814W𝐹814𝑊F814Witalic_F 814 italic_W. This compares well with the 25.2 mag arcsec-2 and 24.6 mag arcsec-2 values from the transient-subtracted images in Epoch 1 (see Chrimes et al., 2024). The F336W𝐹336𝑊F336Witalic_F 336 italic_W surface brightness is 25.76 mag arcsec-2, which after Galactic extinction correction is 25.27 mag arcsec-2. The rest-frame central wavelength of F336W𝐹336𝑊F336Witalic_F 336 italic_W is similar-to\sim2700Å. This allows for a better comparison with the UV (usuperscript𝑢u^{\prime}italic_u start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT) surface brightness distribution for supernova environments, as reported by Kelly & Kirshner (2012) than made by Chrimes et al. (2024) with F555W𝐹555𝑊F555Witalic_F 555 italic_W. The Galactic extinction-corrected F336W𝐹336𝑊F336Witalic_F 336 italic_W surface brightness is in the faintest similar-to\sim10% for local supernova values; this is therefore faint but not unprecedented. We note that supernovae type IIb are the most likely supernova sub-class to explode in young, but low surface brightness environments, and are also found at the highest host-normalised offsets on average (Kelly & Kirshner, 2012).

Table 4: HST magnitudes m𝑚mitalic_m, and their uncertainties δm𝛿𝑚\delta mitalic_δ italic_m, for the second epoch of AT 2023fhn imaging at 200similar-toabsent200\sim 200∼ 200 days (Table 2). In all six filters, two photometry methods are listed - aperture photometry with median background estimation, and aperture photometry with annulus background estimation. Two aperture sizes (and hence enclosed energy corrections) are given in each case.
Filter Method Bkg. Aper. m δ𝛿\deltaitalic_δm
F225W photutils Median 0.2′′ ¿26.1 -
F225W photutils Annulus 0.2′′ ¿26.1 -
F225W photutils Median 0.4′′ ¿25.4 -
F225W photutils Annulus 0.4′′ ¿25.5 -
F336W photutils Median 0.2′′ ¿26.6 -
F336W photutils Annulus 0.2′′ ¿26.6 -
F336W photutils Median 0.4′′ ¿25.9 -
F336W photutils Annulus 0.4′′ ¿25.9 -
F555W photutils Median 0.2′′ 26.9 0.2
F555W photutils Annulus 0.2′′ 27.1 0.3
F555W photutils Median 0.4′′ 25.8 0.2
F555W photutils Annulus 0.4′′ 25.6 0.1
F763M photutils Median 0.2′′ ¿26.0 -
F763M photutils Annulus 0.2′′ ¿26.0 -
F763M photutils Median 0.4′′ ¿25.3 -
F763M photutils Annulus 0.4′′ 25.0 0.3
F814W photutils Median 0.2′′ 26.4 0.2
F814W photutils Annulus 0.2′′ 26.5 0.3
F814W photutils Median 0.4′′ 25.3 0.2
F814W photutils Annulus 0.4′′ 25.2 0.2
F845M photutils Median 0.2′′ ¿25.6 -
F845M photutils Annulus 0.2′′ ¿25.6 -
F845M photutils Median 0.4′′ ¿24.9 -
F845M photutils Annulus 0.4′′ ¿24.9 -
Refer to caption
Refer to caption
Figure 2: Epoch 2 HST photometry at the location of AT 2023fhn, corrected for Galactic extinction, and fit to BPASS single age spectral models while allowing the (intrinsic) extinction to vary. Upper panel: the best fit single-age BPASS spectrum, with local extinction applied. Wavelengths are observer frame. Lower panel: the colourmap corresponds to the fit log10(χ2superscript𝜒2\chi^{2}italic_χ start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT) as a function of extinction and age. We find a best fitting combination of age=166+4subscriptsuperscript164616^{+4}_{-6}16 start_POSTSUPERSCRIPT + 4 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT - 6 end_POSTSUBSCRIPT Myr and A=V0.10.1+0.6{}_{\rm V}=0.1^{+0.6}_{-0.1}start_FLOATSUBSCRIPT roman_V end_FLOATSUBSCRIPT = 0.1 start_POSTSUPERSCRIPT + 0.6 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT - 0.1 end_POSTSUBSCRIPT (uncertainties at 68% confidence, Avni, 1976). The 68% and 90% confidence regions are indicated by white contour lines.

3.2 Global host properties

We next consider how the overall host galaxy properties compare with the local environment of AT 2023fhn, and how they compare with the hosts of other LFBOTs. To do this, we perform spectral energy distribution (SED) fitting of the integrated light of the host. By host, we refer to the spiral and satellite galaxy together, since their proximity likely results in interactions (e.g. tidal) and therefore the two galaxies can be considered as one interacting system. Furthermore, the two galaxies are not spatially resolved in ground-based imaging (e.g. PanSTARRS), which we used to add photometric points to the SED.

We attempt to collect as close to 100% of the galaxy light from HST photometry as possible. We measure the Petrosian radius Rpetrosubscript𝑅petroR_{\rm petro}italic_R start_POSTSUBSCRIPT roman_petro end_POSTSUBSCRIPT (Petrosian, 1976) of the spiral galaxy with the statmorph package (Rodriguez-Gomez et al., 2019, η=0.2𝜂0.2\eta=0.2italic_η = 0.2) and adopt 1.5Rpetrosubscript𝑅petroR_{\rm petro}italic_R start_POSTSUBSCRIPT roman_petro end_POSTSUBSCRIPT as a radius that encloses similar-to\sim100% of the flux (e.g. Conselice, 2003). We account for the projected ellipticity and orientation of the galaxy using the ellip and theta outputs. A pixel mask is produced using these parameters as measured from the F555W𝐹555𝑊F555Witalic_F 555 italic_W image, and applied to the other HST images, as shown in Figure 3. The flux within the mask is summed, and background subtraction (as for the local environment measurements above) uses the sigma-clipped median background, scaled for the number of pixels in the mask. Repeating the procedure for the satellite galaxy produces a 1.5Rpetrosubscript𝑅petroR_{\rm petro}italic_R start_POSTSUBSCRIPT roman_petro end_POSTSUBSCRIPT pixel mask that lies entirely within the spiral’s mask. We therefore use spiral pixel mask alone as it captures similar-to\sim100% of the flux from both galaxies.

To supplement the HST data we add host photometry from archival catalogues. For additional optical points we use PanSTARRS data release 2 (Chambers et al., 2016). We use the catalogued Kron magnitudes (Kron, 1980, in g𝑔gitalic_g, r𝑟ritalic_r, i𝑖iitalic_i, z𝑧zitalic_z and y𝑦yitalic_y), which capture similar-to\sim90% of the light of extended sources, and increase the fluxes by a further 10% to approximate the similar-to\sim100% flux value 333https://outerspace.stsci.edu/display/PANSTARRS/PS1+Kron+photometry+of+extended+sources. The Kron radii for the spiral (4.41, 4.62, 4.36, 3.33 and 2.89 arcsec in g,r,i,z,y𝑔𝑟𝑖𝑧𝑦g,r,i,z,yitalic_g , italic_r , italic_i , italic_z , italic_y respectively) extend past the position of the satellite in g,r,i𝑔𝑟𝑖g,r,iitalic_g , italic_r , italic_i, so the system can be considered blended. Effective wavelengths for these filters are from Tonry et al. (2012). We also add far-UV and near-UV photometry from GALEX (Martin et al., 2003), plus W1, W2 and W3 detections from WISE. The spiral and satellite cannot be separated at the spatial resolution of these surveys, and neither galaxy is detected in 2MASS. The full list of photometry used to performed SED fitting is provided in Table 5.

To perform SED fitting we use prospector (Leja et al., 2017; Johnson et al., 2021), which makes use of FSPS (Flexible Stellar Population Synthesis Conroy et al., 2009; Conroy & Gunn, 2010) and Python-FSPS (Johnson et al., 2023). For the Markov Chain Monte Carlo (MCMC) implementation we use emcee (Foreman-Mackey et al., 2013). We again use BPASS (Binary Population and Spectral Synthesis v2.1, Eldridge et al., 2017; Stanway & Eldridge, 2018) for the spectral models. Before being passed to prospector, the input photometry is corrected for Galactic extinction (as described in Section 3.1). We fit four parameters: the stellar mass Msubscript𝑀M_{\star}italic_M start_POSTSUBSCRIPT ⋆ end_POSTSUBSCRIPT, intrinsic extinction AVsubscript𝐴VA_{\rm V}italic_A start_POSTSUBSCRIPT roman_V end_POSTSUBSCRIPT, population age tagesubscript𝑡aget_{\rm age}italic_t start_POSTSUBSCRIPT roman_age end_POSTSUBSCRIPT and the timescale for an exponentially declining star-formation history τ𝜏\tauitalic_τ. The redshift is fixed at z=0.238𝑧0.238z=0.238italic_z = 0.238, and the luminosity distance at DL=1192subscript𝐷L1192D_{\rm L}=1192italic_D start_POSTSUBSCRIPT roman_L end_POSTSUBSCRIPT = 1192 Mpc.

We run the MCMC with 128 walkers and 512 iterations; the full list of MCMC set-up parameters and joint posterior distributions (in the form of a corner plot) are provided in Appendix A. The maximum a posterior (MAP) spectrum is shown in Figure 4, with the associated properties from the posterior distribution listed in Table 6. Thus far, the metallicity Z𝑍Zitalic_Z has been fixed at half-Solar, based on the approximate mass of 1010superscript101010^{10}10 start_POSTSUPERSCRIPT 10 end_POSTSUPERSCRIPT M and the mass-metallicity relation (Tremonti et al., 2004; Gallazzi et al., 2005). A similar table containing the results when metallicity is allowed to vary is also provided in Appendix A. In this case, the mass and SFR are similar, such that fixing Z𝑍Zitalic_Z at a more realistic value does not change our results in a qualitative sense. In the delayed-τ𝜏\tauitalic_τ model, the current star-formation rate (SFR) is proportional to (t/τ)e(t/τ)𝑡𝜏superscript𝑒𝑡𝜏(t/\tau)e^{(-t/\tau)}( italic_t / italic_τ ) italic_e start_POSTSUPERSCRIPT ( - italic_t / italic_τ ) end_POSTSUPERSCRIPT. The absolute value is obtained by normalisation with respect to the mass formed, yielding a SFR of similar-to\sim4 M yr-1. The galaxy pair is therefore dominated by a fairly typical star-forming spiral, but is perhaps notable for the likely presence of tidal interactions between the spiral and its satellite. In Figure 5 we plot its mass versus SFR, comparing with the host galaxies of previous LFBOTs. The galaxy has a high SFR and mass for LFBOT hosts, lying slightly above average in terms of specific star formation rate (sSFR), but well below the sSFR of the host of ZTF 18abvkwla.

Refer to caption
Figure 3: HST imaging of the galaxy hosting AT 2023fhn in the six epoch 2 filters. Pixels within 1.5 Rpetro of the spiral galaxy centroid are selected as associated with the host, and given an orange-purple colourmap (see text for details). This region fully encompasses the satellite galaxy. The location of AT 2023fhn is marked with a +++ sign. The image cutouts are 13 arcsec on each side. North is up and east is left.
Table 5: Host galaxy photometry used for SED fitting. All magnitudes are in the AB system, and before Galactic extinction correction. The filter effective wavelengths and Galactic extinction at that wavelength - assuming E(B-V)=0.0254, R=V3.1{}_{\rm V}=3.1start_FLOATSUBSCRIPT roman_V end_FLOATSUBSCRIPT = 3.1 and a Fitzpatrick (1999) extinction law - are also listed. We increase the PanSTARRS fluxes by 10% over the values below, as described in the text.
Filter Source m err λeffsubscript𝜆eff\lambda_{\rm eff}italic_λ start_POSTSUBSCRIPT roman_eff end_POSTSUBSCRIPT [Å] A(λ𝜆\lambdaitalic_λ)
FUV GALEX 20.93 0.31 1548.85 0.20
NUV GALEX 20.74 0.25 2303.37 0.22
F225W HST 20.60 0.05 2358.70 0.20
F336W HST 20.40 0.03 3359.11 0.13
g𝑔gitalic_g PS 19.70 0.01 4810.00 0.09
F555W HST 19.34 0.01 5235.33 0.08
r𝑟ritalic_r PS 19.17 0.01 6170.00 0.07
i𝑖iitalic_i PS 18.93 0.01 7520.00 0.05
F763M HST 18.93 0.01 7602.85 0.05
F814W HST 18.84 0.01 7954.84 0.04
F845M HST 18.68 0.01 8430.20 0.04
z𝑧zitalic_z PS 19.02 0.02 8660.00 0.04
y𝑦yitalic_y PS 19.00 0.04 9620.00 0.03
W1 WISE 18.91 0.07 33526.00 0.00
W2 WISE 18.82 0.13 46028.00 0.00
W3 WISE 16.92 0.38 115608.00 0.00
Refer to caption
Figure 4: Host galaxy photometry and best-fit spectrum from prospector. The model spectrum is red-shifted into the observer frame. A light blue shaded region encloses the 90% confidence interval on the posterior flux distribution at each wavelength. The photometry is from GALEX, PanSTARRS, WISE and HST/WFC3 as listed in Table 5, and is corrected for Galactic extinction with the Python module extinction at the filter effective wavelengths. The corresponding galaxy properties are listed in Table 6.
Table 6: Host galaxy properties derived from prospector SED fitting. The median values from the marginalised posterior distributions are quoted, with uncertainties bounding the 68% confidence interval on each parameter.
Host property Value
Msubscript𝑀M_{\star}italic_M start_POSTSUBSCRIPT ⋆ end_POSTSUBSCRIPT / M (0.93±0.04)×1010plus-or-minus0.930.04superscript1010(0.93\pm 0.04)\times 10^{10}( 0.93 ± 0.04 ) × 10 start_POSTSUPERSCRIPT 10 end_POSTSUPERSCRIPT
SFR / M yr-1 4.00.7+0.9subscriptsuperscriptabsent0.90.7{}^{+0.9}_{-0.7}start_FLOATSUPERSCRIPT + 0.9 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.7 end_POSTSUBSCRIPT
AVsubscript𝐴VA_{{\rm V}}italic_A start_POSTSUBSCRIPT roman_V end_POSTSUBSCRIPT 0.15±0.07plus-or-minus0.07\pm 0.07± 0.07
tagesubscript𝑡aget_{{\rm age}}italic_t start_POSTSUBSCRIPT roman_age end_POSTSUBSCRIPT/Gyr 0.950.09+0.11subscriptsuperscriptabsent0.110.09{}^{+0.11}_{-0.09}start_FLOATSUPERSCRIPT + 0.11 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.09 end_POSTSUBSCRIPT
τ𝜏\tauitalic_τ/Gyr 0.260.04+0.06subscriptsuperscriptabsent0.060.04{}^{+0.06}_{-0.04}start_FLOATSUPERSCRIPT + 0.06 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.04 end_POSTSUBSCRIPT
Refer to caption
Figure 5: Stellar mass versus SFR for LFBOT host galaxies, including AT 2023fhn. Other LFBOT host data are from Perley et al. (2019, AT 2018cow), Ho et al. (2020, ZTF 18abvkwla), Coppejans et al. (2020, CSS161010) and Perley et al. (2021, ZTF 20acigmel). Lines of constant specific star formation rate (sSFR/yr-1) are drawn in grey. The core-collapse supernova host galaxy sample of Schulze et al. (2021) is plotted as transparent cyan points.

4 Transient emission

4.1 UV-optical

We now compare the UV-optical constraints on AT 2023fhn’s light-curve with previous LFBOTs. All times used in Section 4 are in the rest-frames of the LFBOTs. Comparison data are corrected for Galactic extinction of E(BV)𝐸𝐵𝑉E(B-V)italic_E ( italic_B - italic_V )=0.08 (AT 2018cow, Prentice et al., 2018) and E(BV)𝐸𝐵𝑉E(B-V)italic_E ( italic_B - italic_V )=0.07 (ZTF 20acigmel, Perley et al., 2021), their UV light-curves (in absolute magnitude) are compared with AT 2023fhn in Figure 6. We fit the light-curve of AT 2018cow in 2 phases, early (<200absent200<200< 200 d) and late-time, with a fit of the form M=alog(t)b+cM=a\log(t)^{b}+citalic_M = italic_a roman_log ( italic_t ) start_POSTSUPERSCRIPT italic_b end_POSTSUPERSCRIPT + italic_c. For the fit to AT 2018cow, we assume that the late-time UV is dominated by residual transient emission (Sun et al., 2022, 2023; Chen et al., 2023; Inkenhaag et al., 2023). We shift the AT 2018cow best-fit up in absolute magnitude such that it lies between the early-time ATLAS c𝑐citalic_c-band and FORS2 u𝑢uitalic_u-band AT 2023fhn detections (Ho et al., 2023b). The extrapolated curve passes below the late-time HST F225W and F336W upper limits reported in this work. Another LFBOT with good UV photometric coverage is ZTF 20acigmel, but here we consider only the early, pre-break phase due to a lack of late-time constraints. ZTF 20acigmel starts brighter than AT 2018cow and fades faster, whereas AT 2023fhn is the most luminous LFBOT yet at UV-optical wavelengths. A final addition to Figure 6 are bands of constant UV absolute magnitude, corresponding to late-time emission from black holes of different masses in the tidal disruption event model of Mummery et al. (2024). This model yielded a black hole mass of similar-to\sim103 M for AT 2018cow. Assuming similar evolution, the HST F336W point source upper limit for AT 2023fhn tentatively constrains the accreting black hole mass in a TDE interpretation to 105less-than-or-similar-toabsentsuperscript105\lesssim 10^{5}≲ 10 start_POSTSUPERSCRIPT 5 end_POSTSUPERSCRIPT M.

Refer to caption
Figure 6: The UV data points for AT 2023fhn, compared with the UV light-curves of AT 2018cow (Early u𝑢uitalic_u-band and late-time F222W and F336W, Prentice et al., 2018; Perley et al., 2019; Inkenhaag et al., 2023) and ZTF 20acigmel (Perley et al., 2021). A light-curve fit to the AT 2018cow data is increased in luminosity to intercept the sole early-time AT 2023fhn UV point, the subsequent F336W limit at similar-to\sim112 rest-frame days lies just above the expected UV magnitude at this epoch, assuming identical evolution to the Cow. A similar fit is made for the early-time ZTF 20acigmel points. Cyan horizontal bands show the expected UV absolute magnitudes at late times for accretion discs around intermediate mass black holes of different masses, following a tidal disruption event (Mummery et al., 2024).

4.2 X-ray

Figure 7 shows our X-ray observations of AT 2023fhn, and the X-ray light-curves of other LFBOTs. The AT 2018cow broken power-law and late-time plateau fit of Migliori et al. (2024) is also shown. AT 2023fhn is the faintest LFBOT in X-rays at early times. Assuming a shallow decay initially, similar to AT 2022tsd, ZTF 20acigmel and AT 2018cow, the break time can be - at the latest - similar to AT 2018cow and ZTF 20acigmel. There appears to be a correlation between break time and X-ray luminosity, with brighter LFBOTs transitioning to a steeper decay at later times. Assuming instead that epochs 1 and 2 are on the same phase of the light-curve, the decay index n=2.10.9+0.7𝑛subscriptsuperscript2.10.70.9n=2.1^{+0.7}_{-0.9}italic_n = 2.1 start_POSTSUPERSCRIPT + 0.7 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT - 0.9 end_POSTSUBSCRIPT (where Ltnproportional-to𝐿superscript𝑡𝑛L\propto t^{-n}italic_L ∝ italic_t start_POSTSUPERSCRIPT - italic_n end_POSTSUPERSCRIPT). Expectations for the X-ray decay rate are t1superscript𝑡1t^{-1}italic_t start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT (shock power), t2superscript𝑡2t^{-2}italic_t start_POSTSUPERSCRIPT - 2 end_POSTSUPERSCRIPT (magnetar central engine) and t5/3superscript𝑡53t^{-5/3}italic_t start_POSTSUPERSCRIPT - 5 / 3 end_POSTSUPERSCRIPT (fallback, i.e. a TDE). Overall, the detections and upper-limits are consistent with AT 2023fhn behaving like a fainter version of previous LFBOTs in the X-ray band, and demonstrates that they can exhibit several orders of magnitude of variety in their X-ray luminosity.

Refer to caption
Figure 7: The X-ray light-curve of AT 2023fhn compared with other LFBOTs. All data are in the similar-to\sim0.5-10 keV range, are unabsorbed and from Rivera Sandoval et al. (2018, AT 2018cow); Kuin et al. (2019, AT 2018cow); Migliori et al. (2024, AT 2018cow), Coppejans et al. (2020, CSS161010), Bright et al. (2022, ZTF 20acigmel); Ho et al. (2022, ZTF 20acigmel), Yao et al. (2022, AT 2020mrf) and Matthews et al. (2023, AT 2022tsd). Given the marginal nature of the second AT 2023fhn measurement, we also plot the 2σ𝜎\sigmaitalic_σ upper limit at this epoch. A broken-power law and late-time plateau interpretation of AT 2018cow’s light-curve is shown by the solid black line (Migliori et al., 2024).

4.3 UV/X-ray ratio

Motivated by the fact that AT 2023fhn appears to be the brightest LFBOT yet at UV-optical wavelengths, and the faintest in terms of X-ray luminosity, in Figure 8 we show the ratio of X-ray to UV luminosity for the 3 LFBOTs with such constraints. The data points for AT 2023fhn take the X-ray detections at 12 and 23 rest-frame days, and the corresponding point on the shifted AT 2018cow light-curve in Figure 6. The uncertainties shown are exclusively from the X-ray observations. For AT 2018cow, we take the ratio of the X-ray fit of Migliori et al. (2024) in Figure 7, and our fit to the UV light-curve fit in Figure 6. Finally, for ZTF 20acigmel we take the ratio of the X-ray luminosity with the UV light-curve fit at the same time. LFBOTs therefore exhibit at least 3similar-toabsent3\sim 3∼ 3 orders of magnitude in their X-ray/UV luminosity ratio, even at similar times in their evolution. This is plausibly a viewing angle effect. A qualitative prediction of tidal disruption models is a trade-off between UV-optical and X-ray luminosity as a function of viewing angle, where on-axis angles (which may also be aligned with a beamed outflow) would see a higher X-ray luminosity (Dai et al., 2018; Hayasaki & Jonker, 2021). Differences in LX/LUV are also expected for different black hole masses and spins, due to varying accretion disc formation rates (which in turn affects the delay betweeen peak X-ray and UV/optical emission, Jonker et al., 2020). However, a scenario in which the peak X-ray emission is delayed due to a delay in forming the inner accretion disc is hard to reconcile with the energetics and (variability) timescales of LFBOT emission, which demands energy input from a central engine and therefore active accretion (e.g. Ho et al., 2019; Margutti et al., 2019). Alternatively, the range of Lxsubscript𝐿xL_{\rm x}italic_L start_POSTSUBSCRIPT roman_x end_POSTSUBSCRIPT/LUVsubscript𝐿UVL_{\rm UV}italic_L start_POSTSUBSCRIPT roman_UV end_POSTSUBSCRIPT could reflect differences in the circumstellar media, which we investigate in the following Section.

Refer to caption
Figure 8: The ratio of X-ray to UV luminosity for LFBOTs AT 2023fhn, AT 2018cow and ZTF 20acigmel. The AT 2023fhn points use the dashed orange line in Figure 6 and the two Chandra X-ray detections (errorbars reflect the X-ray uncertainties only). The black curve is the ratio of the Migliori et al. (2024) X-ray light-curve fit (see the solid black line, Figure 7) and a broken power-law fit to the UV observations (the solid black line in Figure 6). The small k𝑘kitalic_k-corrections are neglected in this comparison. The evolution past 200 days (drawn as a dashed line) is highly uncertain due to the sole X-ray detection. The ZTFacigmel data are from Ho et al. (2022); Bright et al. (2022), where we have taken the ratio of X-ray points and the power-law fit to the ZTFacigmel UV-light-curve in Figure 6. The uncertainties on these points again solely reflect the X-ray measurement uncertainties.

4.4 Radio

Assuming that the radio emission is synchrotron-dominated with self-absorption - as we will see, the radio SED of AT 2023fhn is consistent with this - and that the peak of the SED occurs at the synchtrotron self-absorption (SSA) frequency, we can estimate several shock parameters, and properties of the circumstellar medium. We follow the synchrotron self-absorption model of Chevalier (1998) (see also Soderberg et al., 2005). Adopting this framework for AT 2023fhn is reasonable since this best fits other LFBOTs studied so far (based on the brightness temperature, which precludes thermal emission, and the spectral shape, e.g. Margutti et al., 2019; Coppejans et al., 2020; Ho et al., 2020; Nayana & Chandra, 2021; Ho et al., 2022; Yao et al., 2022; Bright et al., 2022).

We fit the radio spectrum at similar-to\sim90 and similar-to\sim138 days (similar-to\sim70 and similar-to\sim110 rest-frame days) following Chevalier (1998); Granot & Sari (2002); Chevalier & Fransson (2006). At a given time t𝑡titalic_t the radio SED has the form,

F(ν)=Fpk[(ννpk)sβ1+(ννpk)sβ2]1s,𝐹𝜈subscript𝐹pksuperscriptdelimited-[]superscript𝜈subscript𝜈pk𝑠subscript𝛽1superscript𝜈subscript𝜈pk𝑠subscript𝛽21𝑠F(\nu)=F_{\rm pk}\left[\left(\frac{\nu}{\nu_{\rm pk}}\right)^{-s\beta_{1}}+% \left(\frac{\nu}{\nu_{\rm pk}}\right)^{-s\beta_{2}}\right]^{-\frac{1}{s}},italic_F ( italic_ν ) = italic_F start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT [ ( divide start_ARG italic_ν end_ARG start_ARG italic_ν start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT end_ARG ) start_POSTSUPERSCRIPT - italic_s italic_β start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT end_POSTSUPERSCRIPT + ( divide start_ARG italic_ν end_ARG start_ARG italic_ν start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT end_ARG ) start_POSTSUPERSCRIPT - italic_s italic_β start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT end_POSTSUPERSCRIPT ] start_POSTSUPERSCRIPT - divide start_ARG 1 end_ARG start_ARG italic_s end_ARG end_POSTSUPERSCRIPT , (1)

where F(ν)𝐹𝜈F(\nu)italic_F ( italic_ν ) is the flux density, Fpksubscript𝐹pkF_{\rm pk}italic_F start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT is the flux at the peak (break) frequency νpksubscript𝜈pk\nu_{\rm pk}italic_ν start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT where the optically thick and thin power laws intersect, s𝑠sitalic_s is a smoothing factor and β1subscript𝛽1\beta_{1}italic_β start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT and β2subscript𝛽2\beta_{2}italic_β start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT are spectral indices in the optically thick and thin regimes, respectively. In our case the cooling frequency νcsubscript𝜈c\nu_{\rm c}italic_ν start_POSTSUBSCRIPT roman_c end_POSTSUBSCRIPT lies at higher frequencies than probed by our observations (similar-to\sim400-800 GHz), where νcsubscript𝜈c\nu_{\rm c}italic_ν start_POSTSUBSCRIPT roman_c end_POSTSUBSCRIPT is given by 18πmece)/(t2σt2B3)18\pi m_{e}ce)/(t^{2}\sigma_{t}^{2}B^{3})18 italic_π italic_m start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT italic_c italic_e ) / ( italic_t start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT italic_σ start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT italic_B start_POSTSUPERSCRIPT 3 end_POSTSUPERSCRIPT ) (DeMarchi et al., 2022). We therefore expect F(ν)ν(p1)/2proportional-to𝐹𝜈superscript𝜈𝑝12F(\nu)\propto\nu^{-(p-1)/2}italic_F ( italic_ν ) ∝ italic_ν start_POSTSUPERSCRIPT - ( italic_p - 1 ) / 2 end_POSTSUPERSCRIPT in the optically thin regime, where p𝑝pitalic_p is the power law index of the electron energy distribution in the shock (i.e. the number N𝑁Nitalic_N of electrons with Lorentz factor γesubscript𝛾e\gamma_{\rm e}italic_γ start_POSTSUBSCRIPT roman_e end_POSTSUBSCRIPT goes as N(γe)γepproportional-to𝑁subscript𝛾esuperscriptsubscript𝛾e𝑝N(\gamma_{\rm e})\propto\gamma_{\rm e}^{-p}italic_N ( italic_γ start_POSTSUBSCRIPT roman_e end_POSTSUBSCRIPT ) ∝ italic_γ start_POSTSUBSCRIPT roman_e end_POSTSUBSCRIPT start_POSTSUPERSCRIPT - italic_p end_POSTSUPERSCRIPT).

Using the scipy curve_fit function, and working with rest-frame times and central frequencies throughout this section, we fit equation 1 to the similar-to\sim90 day and similar-to\sim138 day (observer frame) data. At 90 days we have 6 data points (5 detections, 1 upper limit, we combine the 87 and 95 day data for this epoch), and at 138 days we have 7 data points (6 detections, 1 upper limit). There are 5 parameters to fit: Fpksubscript𝐹pkF_{\rm pk}italic_F start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT, νpksubscript𝜈pk\nu_{\rm pk}italic_ν start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT, β1subscript𝛽1\beta_{1}italic_β start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT, β2subscript𝛽2\beta_{2}italic_β start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT and s𝑠sitalic_s. The best-fit values for these parameters and their uncertainties are listed in Table 7. The optically-thin spectral index of -0.59 (138 days) yields an electron energy spectral index of similar-to\sim2.2, which is relatively shallow - 2.5 is expected from theory, while values closer to similar-to\sim3 are often measured in gamma-ray bursts, tidal disruption events and supernovae (e.g. Chevalier & Fransson, 2006; Cendes et al., 2023). Values from other LFBOTs are also in the range similar-to\sim2–3 (Margutti et al., 2019; Ho et al., 2020; Coppejans et al., 2020; Yao et al., 2022; Bright et al., 2022).

The peak flux Fpk and (rest-frame) frequency at the peak flux νpksubscript𝜈pk\nu_{\rm pk}italic_ν start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT (at the intersection of the power-laws, rather than the fitted peak) allow us to estimate the radius of the shock, circumstellar density at that radius, and the CSM surface density parameter AM˙/vwproportional-tosubscript𝐴˙𝑀subscript𝑣𝑤A_{\star}\propto\dot{M}/v_{w}italic_A start_POSTSUBSCRIPT ⋆ end_POSTSUBSCRIPT ∝ over˙ start_ARG italic_M end_ARG / italic_v start_POSTSUBSCRIPT italic_w end_POSTSUBSCRIPT (see DeMarchi et al., 2022, for a detailed description of the modelling assumptions). Following the formulism of Chevalier (1998) (see also Chevalier & Fransson 2006; DeMarchi et al. 2022; Bright et al. 2022), we first have the shock radius Rpsubscript𝑅pR_{\rm p}italic_R start_POSTSUBSCRIPT roman_p end_POSTSUBSCRIPT, given by,

Rp=4×1014(ϵeϵB)119(f0.5)119(Fpk(1+z)mJy)919(DθMpc)1819(νpk5GHz)1cm,subscript𝑅p4superscript1014superscriptsubscriptitalic-ϵesubscriptitalic-ϵB119superscript𝑓0.5119superscriptsubscript𝐹pk1zmJy919superscriptsubscript𝐷𝜃Mpc1819superscriptsubscript𝜈pk5GHz1cmR_{\rm p}=4\times 10^{14}\left(\frac{\epsilon_{\rm e}}{\epsilon_{\rm B}}\right% )^{\frac{-1}{19}}\left(\frac{f}{0.5}\right)^{\frac{-1}{19}}\left(\frac{F_{\rm pk% }}{{\rm(1+z)\,mJy}}\right)^{\frac{9}{19}}\left(\frac{D_{\theta}}{{\rm Mpc}}% \right)^{\frac{18}{19}}\left(\frac{\nu_{\rm pk}}{5\leavevmode\nobreak\ {\rm GHz% }}\right)^{-1}{\rm cm},italic_R start_POSTSUBSCRIPT roman_p end_POSTSUBSCRIPT = 4 × 10 start_POSTSUPERSCRIPT 14 end_POSTSUPERSCRIPT ( divide start_ARG italic_ϵ start_POSTSUBSCRIPT roman_e end_POSTSUBSCRIPT end_ARG start_ARG italic_ϵ start_POSTSUBSCRIPT roman_B end_POSTSUBSCRIPT end_ARG ) start_POSTSUPERSCRIPT divide start_ARG - 1 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_f end_ARG start_ARG 0.5 end_ARG ) start_POSTSUPERSCRIPT divide start_ARG - 1 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_F start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT end_ARG start_ARG ( 1 + roman_z ) roman_mJy end_ARG ) start_POSTSUPERSCRIPT divide start_ARG 9 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_D start_POSTSUBSCRIPT italic_θ end_POSTSUBSCRIPT end_ARG start_ARG roman_Mpc end_ARG ) start_POSTSUPERSCRIPT divide start_ARG 18 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_ν start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT end_ARG start_ARG 5 roman_GHz end_ARG ) start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT roman_cm , (2)

where Dθ is the angular diameter distance, and ϵesubscriptitalic-ϵ𝑒\epsilon_{e}italic_ϵ start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT and ϵBsubscriptitalic-ϵ𝐵\epsilon_{B}italic_ϵ start_POSTSUBSCRIPT italic_B end_POSTSUBSCRIPT are the fraction of the shock energy in electrons and in the magnetic field, respectively. The average shock velocity can then be calculated as Rp/tobs=Γβcsubscript𝑅psubscript𝑡obsΓ𝛽𝑐R_{\rm p}/t_{\rm obs}=\Gamma\beta citalic_R start_POSTSUBSCRIPT roman_p end_POSTSUBSCRIPT / italic_t start_POSTSUBSCRIPT roman_obs end_POSTSUBSCRIPT = roman_Γ italic_β italic_c, where β=v/c𝛽𝑣𝑐\beta=v/citalic_β = italic_v / italic_c, ΓΓ\Gammaroman_Γ is the Lorentz factor and tobssubscript𝑡obst_{\rm obs}italic_t start_POSTSUBSCRIPT roman_obs end_POSTSUBSCRIPT is the rest-frame observation time. Next we have, for the internal magnetic field B𝐵Bitalic_B,

B=1.1(ϵeϵB)419(f0.5)419(Fpk(1+z)mJy)219(DθMpc)419(νpk5GHz)G𝐵1.1superscriptsubscriptitalic-ϵesubscriptitalic-ϵB419superscript𝑓0.5419superscriptsubscript𝐹pk1zmJy219superscriptsubscript𝐷𝜃Mpc419subscript𝜈pk5GHzGB=1.1\left(\frac{\epsilon_{\rm e}}{\epsilon_{\rm B}}\right)^{\frac{-4}{19}}% \left(\frac{f}{0.5}\right)^{\frac{-4}{19}}\left(\frac{F_{\rm pk}}{{\rm(1+z)\,% mJy}}\right)^{\frac{-2}{19}}\left(\frac{D_{\theta}}{{\rm Mpc}}\right)^{\frac{-% 4}{19}}\left(\frac{\nu_{\rm pk}}{5\leavevmode\nobreak\ {\rm GHz}}\right){\rm G}italic_B = 1.1 ( divide start_ARG italic_ϵ start_POSTSUBSCRIPT roman_e end_POSTSUBSCRIPT end_ARG start_ARG italic_ϵ start_POSTSUBSCRIPT roman_B end_POSTSUBSCRIPT end_ARG ) start_POSTSUPERSCRIPT divide start_ARG - 4 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_f end_ARG start_ARG 0.5 end_ARG ) start_POSTSUPERSCRIPT divide start_ARG - 4 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_F start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT end_ARG start_ARG ( 1 + roman_z ) roman_mJy end_ARG ) start_POSTSUPERSCRIPT divide start_ARG - 2 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_D start_POSTSUBSCRIPT italic_θ end_POSTSUBSCRIPT end_ARG start_ARG roman_Mpc end_ARG ) start_POSTSUPERSCRIPT divide start_ARG - 4 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_ν start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT end_ARG start_ARG 5 roman_GHz end_ARG ) roman_G (3)

and for the wind density (the mass loss rate M˙˙𝑀\dot{M}over˙ start_ARG italic_M end_ARG over the wind velocity),

M˙vw(1000kms1104Myr1)=2.5×105(1ϵB)(ϵeϵB)819(f0.5)819×(Fpk(1+z)Jy)419(DθMpc)819(νpk5GHz)2(tpkdays)2˙𝑀subscript𝑣w1000superscriptkms1superscript104subscript𝑀direct-product𝑦superscript𝑟12.5superscript1051subscriptitalic-ϵBsuperscriptsubscriptitalic-ϵesubscriptitalic-ϵB819superscript𝑓0.5819superscriptsubscript𝐹pk1zJy419superscriptsubscript𝐷𝜃Mpc819superscriptsubscript𝜈pk5GHz2superscriptsubscript𝑡pkdays2\frac{\dot{M}}{v_{\rm w}}\left(\frac{1000\leavevmode\nobreak\ {\rm kms^{-1}}}{% 10^{-4}\leavevmode\nobreak\ M_{\odot}yr^{-1}}\right)=2.5\times 10^{-5}\left(% \frac{1}{\epsilon_{\rm B}}\right)\left(\frac{\epsilon_{\rm e}}{\epsilon_{\rm B% }}\right)^{\frac{-8}{19}}\left(\frac{f}{0.5}\right)^{\frac{-8}{19}}\times\\ \left(\frac{F_{\rm pk}}{{\rm(1+z)\,Jy}}\right)^{\frac{-4}{19}}\left(\frac{D_{% \theta}}{{\rm Mpc}}\right)^{\frac{-8}{19}}\left(\frac{\nu_{\rm pk}}{5% \leavevmode\nobreak\ {\rm GHz}}\right)^{2}\left(\frac{t_{\rm pk}}{{\rm days}}% \right)^{2}start_ROW start_CELL divide start_ARG over˙ start_ARG italic_M end_ARG end_ARG start_ARG italic_v start_POSTSUBSCRIPT roman_w end_POSTSUBSCRIPT end_ARG ( divide start_ARG 1000 roman_kms start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT end_ARG start_ARG 10 start_POSTSUPERSCRIPT - 4 end_POSTSUPERSCRIPT italic_M start_POSTSUBSCRIPT ⊙ end_POSTSUBSCRIPT italic_y italic_r start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT end_ARG ) = 2.5 × 10 start_POSTSUPERSCRIPT - 5 end_POSTSUPERSCRIPT ( divide start_ARG 1 end_ARG start_ARG italic_ϵ start_POSTSUBSCRIPT roman_B end_POSTSUBSCRIPT end_ARG ) ( divide start_ARG italic_ϵ start_POSTSUBSCRIPT roman_e end_POSTSUBSCRIPT end_ARG start_ARG italic_ϵ start_POSTSUBSCRIPT roman_B end_POSTSUBSCRIPT end_ARG ) start_POSTSUPERSCRIPT divide start_ARG - 8 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_f end_ARG start_ARG 0.5 end_ARG ) start_POSTSUPERSCRIPT divide start_ARG - 8 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT × end_CELL end_ROW start_ROW start_CELL ( divide start_ARG italic_F start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT end_ARG start_ARG ( 1 + roman_z ) roman_Jy end_ARG ) start_POSTSUPERSCRIPT divide start_ARG - 4 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_D start_POSTSUBSCRIPT italic_θ end_POSTSUBSCRIPT end_ARG start_ARG roman_Mpc end_ARG ) start_POSTSUPERSCRIPT divide start_ARG - 8 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_ν start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT end_ARG start_ARG 5 roman_GHz end_ARG ) start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ( divide start_ARG italic_t start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT end_ARG start_ARG roman_days end_ARG ) start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT end_CELL end_ROW (4)

Under the assumption that the CSM is dominated by fully ionised hydrogen, the electron number density can be related to M˙/vw˙𝑀subscript𝑣𝑤\dot{M}/v_{w}over˙ start_ARG italic_M end_ARG / italic_v start_POSTSUBSCRIPT italic_w end_POSTSUBSCRIPT by ne=M˙/(4πmpr2vw)subscript𝑛𝑒˙𝑀4𝜋subscript𝑚𝑝superscript𝑟2subscript𝑣𝑤n_{e}=\dot{M}/(4\pi m_{p}r^{2}v_{w})italic_n start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT = over˙ start_ARG italic_M end_ARG / ( 4 italic_π italic_m start_POSTSUBSCRIPT italic_p end_POSTSUBSCRIPT italic_r start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT italic_v start_POSTSUBSCRIPT italic_w end_POSTSUBSCRIPT ) - where mpsubscript𝑚𝑝m_{p}italic_m start_POSTSUBSCRIPT italic_p end_POSTSUBSCRIPT is the proton mass - so that,

ne=1.02418(1ϵB)(ϵeϵB)619(f0.5)619×(Fpk(1+z)Jy)2219(DθMpc)4419(νpk5GHz)4(tpkdays)2cm3subscript𝑛e1.024181subscriptitalic-ϵBsuperscriptsubscriptitalic-ϵesubscriptitalic-ϵB619superscript𝑓0.5619superscriptsubscript𝐹pk1zJy2219superscriptsubscript𝐷𝜃Mpc4419superscriptsubscript𝜈pk5GHz4superscriptsubscript𝑡pkdays2superscriptcm3n_{\rm e}=1.02418\left(\frac{1}{\epsilon_{\rm B}}\right)\left(\frac{\epsilon_{% \rm e}}{\epsilon_{\rm B}}\right)^{\frac{-6}{19}}\left(\frac{f}{0.5}\right)^{% \frac{-6}{19}}\times\\ \left(\frac{F_{\rm pk}}{{\rm(1+z)\,Jy}}\right)^{\frac{-22}{19}}\left(\frac{D_{% \theta}}{{\rm Mpc}}\right)^{\frac{-44}{19}}\left(\frac{\nu_{\rm pk}}{5% \leavevmode\nobreak\ {\rm GHz}}\right)^{4}\left(\frac{t_{\rm pk}}{{\rm days}}% \right)^{2}{\rm cm}^{-3}start_ROW start_CELL italic_n start_POSTSUBSCRIPT roman_e end_POSTSUBSCRIPT = 1.02418 ( divide start_ARG 1 end_ARG start_ARG italic_ϵ start_POSTSUBSCRIPT roman_B end_POSTSUBSCRIPT end_ARG ) ( divide start_ARG italic_ϵ start_POSTSUBSCRIPT roman_e end_POSTSUBSCRIPT end_ARG start_ARG italic_ϵ start_POSTSUBSCRIPT roman_B end_POSTSUBSCRIPT end_ARG ) start_POSTSUPERSCRIPT divide start_ARG - 6 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_f end_ARG start_ARG 0.5 end_ARG ) start_POSTSUPERSCRIPT divide start_ARG - 6 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT × end_CELL end_ROW start_ROW start_CELL ( divide start_ARG italic_F start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT end_ARG start_ARG ( 1 + roman_z ) roman_Jy end_ARG ) start_POSTSUPERSCRIPT divide start_ARG - 22 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_D start_POSTSUBSCRIPT italic_θ end_POSTSUBSCRIPT end_ARG start_ARG roman_Mpc end_ARG ) start_POSTSUPERSCRIPT divide start_ARG - 44 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_ν start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT end_ARG start_ARG 5 roman_GHz end_ARG ) start_POSTSUPERSCRIPT 4 end_POSTSUPERSCRIPT ( divide start_ARG italic_t start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT end_ARG start_ARG roman_days end_ARG ) start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT roman_cm start_POSTSUPERSCRIPT - 3 end_POSTSUPERSCRIPT end_CELL end_ROW (5)

Additionally we have, for the internal shock energy U=UB/ϵB𝑈subscript𝑈Bsubscriptitalic-ϵBU=U_{\rm B}/\epsilon_{\rm B}italic_U = italic_U start_POSTSUBSCRIPT roman_B end_POSTSUBSCRIPT / italic_ϵ start_POSTSUBSCRIPT roman_B end_POSTSUBSCRIPT,

U=1.859×1046(1ϵB)(ϵeϵB)1119(f0.5)819×(Fpk(1+z)Jy)2319(DθMpc)4619(νpk5GHz)1erg𝑈1.859superscript10461subscriptitalic-ϵBsuperscriptsubscriptitalic-ϵesubscriptitalic-ϵB1119superscript𝑓0.5819superscriptsubscript𝐹pk1zJy2319superscriptsubscript𝐷𝜃Mpc4619superscriptsubscript𝜈pk5GHz1ergU=1.859\times 10^{46}\left(\frac{1}{\epsilon_{\rm B}}\right)\left(\frac{% \epsilon_{\rm e}}{\epsilon_{\rm B}}\right)^{\frac{-11}{19}}\left(\frac{f}{0.5}% \right)^{\frac{8}{19}}\times\\ \left(\frac{F_{\rm pk}}{{\rm(1+z)\,Jy}}\right)^{\frac{23}{19}}\left(\frac{D_{% \theta}}{{\rm Mpc}}\right)^{\frac{46}{19}}\left(\frac{\nu_{\rm pk}}{5% \leavevmode\nobreak\ {\rm GHz}}\right)^{-1}{\rm erg}start_ROW start_CELL italic_U = 1.859 × 10 start_POSTSUPERSCRIPT 46 end_POSTSUPERSCRIPT ( divide start_ARG 1 end_ARG start_ARG italic_ϵ start_POSTSUBSCRIPT roman_B end_POSTSUBSCRIPT end_ARG ) ( divide start_ARG italic_ϵ start_POSTSUBSCRIPT roman_e end_POSTSUBSCRIPT end_ARG start_ARG italic_ϵ start_POSTSUBSCRIPT roman_B end_POSTSUBSCRIPT end_ARG ) start_POSTSUPERSCRIPT divide start_ARG - 11 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_f end_ARG start_ARG 0.5 end_ARG ) start_POSTSUPERSCRIPT divide start_ARG 8 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT × end_CELL end_ROW start_ROW start_CELL ( divide start_ARG italic_F start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT end_ARG start_ARG ( 1 + roman_z ) roman_Jy end_ARG ) start_POSTSUPERSCRIPT divide start_ARG 23 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_D start_POSTSUBSCRIPT italic_θ end_POSTSUBSCRIPT end_ARG start_ARG roman_Mpc end_ARG ) start_POSTSUPERSCRIPT divide start_ARG 46 end_ARG start_ARG 19 end_ARG end_POSTSUPERSCRIPT ( divide start_ARG italic_ν start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT end_ARG start_ARG 5 roman_GHz end_ARG ) start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT roman_erg end_CELL end_ROW (6)

We assume equipartition (ϵesubscriptitalic-ϵ𝑒\epsilon_{e}italic_ϵ start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT=ϵBsubscriptitalic-ϵ𝐵\epsilon_{B}italic_ϵ start_POSTSUBSCRIPT italic_B end_POSTSUBSCRIPT=1/3), where the magnetic energy density, the energy density in electrons and the energy density in protons contribute equally as destinations for the converted kinetic energy in the shock. We further assume f=0.5𝑓0.5f=0.5italic_f = 0.5 for the filling factor. If the emission region is modelled as a disc of radius R𝑅Ritalic_R and thickness S𝑆Sitalic_S on the sky, whose volume is πR2S𝜋superscript𝑅2𝑆\pi R^{2}Sitalic_π italic_R start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT italic_S, an equivalent spherical volume can be given by 4/3πR343𝜋superscript𝑅34/3\pi R^{3}4 / 3 italic_π italic_R start_POSTSUPERSCRIPT 3 end_POSTSUPERSCRIPT. The filling factor is the fraction of this equivalent spherical volume producing emission (Chevalier, 1998).

We list the inferred properties of AT 2023fhn’s blast-wave in Table 7. Results for the fiducial parameters of ϵe=0.1subscriptitalic-ϵ𝑒0.1\epsilon_{e}=0.1italic_ϵ start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT = 0.1 and ϵB=0.01subscriptitalic-ϵ𝐵0.01\epsilon_{B}=0.01italic_ϵ start_POSTSUBSCRIPT italic_B end_POSTSUBSCRIPT = 0.01 are also listed. These properties are compared with other LFBOTs in Figures 9, 10 and 11. In LFBOTs the expanding blast-wave typically shows a SSA spectrum that decreases in peak flux and frequency over time. However we note that AT2023fhn shows an increase in peak flux between 90similar-toabsent90\sim 90∼ 90 and 138similar-toabsent138\sim 138∼ 138 days post explosion. A similar increase was seen in CSS161010 between 69 and 99 days post explosion (Coppejans et al., 2020). This could potentially be caused by an increase in density, or inhomogeneities in the CSM, but we are not able to test this scenario given our weak constraints on the SSA peak at 90similar-toabsent90\sim 90∼ 90 days post explosion.

Finally, we calculate a dimensionless normalisation of the wind density parameter Aner2M˙/vwproportional-tosubscript𝐴subscript𝑛𝑒superscript𝑟2proportional-to˙𝑀subscript𝑣𝑤A_{\star}\propto n_{e}r^{2}\propto\dot{M}/v_{w}italic_A start_POSTSUBSCRIPT ⋆ end_POSTSUBSCRIPT ∝ italic_n start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT italic_r start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ∝ over˙ start_ARG italic_M end_ARG / italic_v start_POSTSUBSCRIPT italic_w end_POSTSUBSCRIPT (Chevalier & Li, 2000),

A=M˙(5×1011gcm1)×4πvwsubscript𝐴˙𝑀5superscript1011gsuperscriptcm14𝜋subscript𝑣𝑤A_{\star}=\frac{\dot{M}}{(5\times 10^{11}\leavevmode\nobreak\ {\rm g% \leavevmode\nobreak\ cm^{-1}})\times 4\pi v_{w}}italic_A start_POSTSUBSCRIPT ⋆ end_POSTSUBSCRIPT = divide start_ARG over˙ start_ARG italic_M end_ARG end_ARG start_ARG ( 5 × 10 start_POSTSUPERSCRIPT 11 end_POSTSUPERSCRIPT roman_g roman_cm start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT ) × 4 italic_π italic_v start_POSTSUBSCRIPT italic_w end_POSTSUBSCRIPT end_ARG (7)

where A=1subscript𝐴1A_{\star}=1italic_A start_POSTSUBSCRIPT ⋆ end_POSTSUBSCRIPT = 1 for a Wolf-Rayet-like wind with M˙=105˙𝑀superscript105\dot{M}=10^{-5}over˙ start_ARG italic_M end_ARG = 10 start_POSTSUPERSCRIPT - 5 end_POSTSUPERSCRIPTMyr-1 and vw=1000subscript𝑣𝑤1000v_{w}=1000italic_v start_POSTSUBSCRIPT italic_w end_POSTSUBSCRIPT = 1000km s-1. From our best fits to the radio data, we derive that AT 2023fhn at similar-to\sim70–110 rest-frame days has A1similar-tosubscript𝐴1A_{\star}\sim 1italic_A start_POSTSUBSCRIPT ⋆ end_POSTSUBSCRIPT ∼ 1 (0.1×104similar-toabsent0.1superscript104\sim 0.1\times 10^{-4}∼ 0.1 × 10 start_POSTSUPERSCRIPT - 4 end_POSTSUPERSCRIPT M yr-1 for vwsubscript𝑣wv_{\rm w}italic_v start_POSTSUBSCRIPT roman_w end_POSTSUBSCRIPT = 1000 km s-1). This mass loss rate is consistent with that of Wolf-Rayet stars. As shown in Figure 11, this density is also consistent with that of the other LFBOTs. The constraints on the synchrotron self-absorption peak at 90similar-toabsent90\sim 90∼ 90 days post explosion were unfortunately insufficient to constrain the density profile of the CSM around AT2023fhn.

Refer to caption
Refer to caption
Figure 9: Radio observations of AT 2023fhn (see Table 3) placed in the context of other LFBOTs. Upper limits from NOEMA and the VLA as reported by Ho et al. (2023a) and Ho (2023) are also shown. Left: radio SED for AT 2023fhn, with a broken-power law fit to the t=138𝑡138t=138italic_t = 138 day (grey line) and t=90𝑡90t=90italic_t = 90 day (green line) data. Data point rest-frame times are indicated by the colourbar. The 90% confidence regions on the fits are shown by light grey/green shading. Right: radio light-curve for AT 2023fhn (central frequencies indicated by the colourbar) and other LFBOTs. The 10 GHz AT 2023fhn detections are connected by a dashed orange line, to aid the eye in comparing with other LFBOTs. Data for the other LFBOTs, all at (10±plus-or-minus\pm±2) GHz, are from Ho et al. (2019, AT 2018cow); Margutti et al. (2019, AT 2018cow), Ho et al. (2020, ZTF 18abvkwla), Coppejans et al. (2020, CSS161010), Bright et al. (2022), Ho et al. (2022, ZTF 20 acigmel) and Yao et al. (2022, AT 2020mrf).
Refer to caption
Refer to caption
Figure 10: Left: the product of rest-time ΔtΔ𝑡\Delta troman_Δ italic_t and the rest-frame peak frequency νpksubscript𝜈pk{\nu}_{\rm pk}italic_ν start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT at that time, versus the peak radio spectral luminosity Lθν,pk=Lν,pk/(1+z)3subscript𝐿𝜃𝜈pksubscript𝐿𝜈pksuperscript1𝑧3L_{\rm\theta\nu,pk}=L_{\rm\nu,pk}/(1+z)^{3}italic_L start_POSTSUBSCRIPT italic_θ italic_ν , roman_pk end_POSTSUBSCRIPT = italic_L start_POSTSUBSCRIPT italic_ν , roman_pk end_POSTSUBSCRIPT / ( 1 + italic_z ) start_POSTSUPERSCRIPT 3 end_POSTSUPERSCRIPT, with lines of constant M˙/Vw˙𝑀subscript𝑉𝑤\dot{M}/V_{w}over˙ start_ARG italic_M end_ARG / italic_V start_POSTSUBSCRIPT italic_w end_POSTSUBSCRIPT (in units of 10-4 M yr-1 / 1000 km s-1) and lower limits on the blast-wave velocity shown. Both AT 2023fhn epochs are shown, we note that constraints on the later point (138 days) are stronger and the parameters at this epoch better constrained. All comparison data points adopt ϵe=ϵB=1/3subscriptitalic-ϵ𝑒subscriptitalic-ϵ𝐵13\epsilon_{e}=\epsilon_{B}=1/3italic_ϵ start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT = italic_ϵ start_POSTSUBSCRIPT italic_B end_POSTSUBSCRIPT = 1 / 3. Right: lower limits on the average blast-wave velocity (blast-wave radius over the rest-frame time) in units of c𝑐citalic_c versus the internal energy of the shock U=UB/ϵB𝑈subscript𝑈𝐵subscriptitalic-ϵ𝐵U=U_{B}/\epsilon_{B}italic_U = italic_U start_POSTSUBSCRIPT italic_B end_POSTSUBSCRIPT / italic_ϵ start_POSTSUBSCRIPT italic_B end_POSTSUBSCRIPT. Other LFBOT data are taken from Ho et al. (2019, AT 2018cow); Margutti et al. (2019, AT 2018cow); Nayana & Chandra (2021, AT 2018cow), Ho et al. (2020, ZTF 18abvkwla), Coppejans et al. (2020, CSS161010), Bright et al. (2022);Ho et al. (2022, ZTF 20 acigmel) and Yao et al. (2022, AT 2020mrf). The mildly-relativistic regime is shaded. We have scaled the Bright et al. (2022) ZTF 20 acigmel point following equations 2 and 6 to align with our assumption of equipartition.
Refer to caption
Figure 11: Circumstellar density ne at the radius of the shock Rp for AT 2023fhn and previous LFBOTs. The densities for ZTF 20 acigmel (Bright et al., 2022, Table 6) have been reduced by a factor similar-to\sim16 to align them with the equipartition ϵe=ϵB=1/3subscriptitalic-ϵ𝑒subscriptitalic-ϵ𝐵13\epsilon_{e}=\epsilon_{B}=1/3italic_ϵ start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT = italic_ϵ start_POSTSUBSCRIPT italic_B end_POSTSUBSCRIPT = 1 / 3 assumption used for all other measurements (for reference, see how equation 5 scales with ϵesubscriptitalic-ϵ𝑒\epsilon_{e}italic_ϵ start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT and ϵBsubscriptitalic-ϵ𝐵\epsilon_{B}italic_ϵ start_POSTSUBSCRIPT italic_B end_POSTSUBSCRIPT). As in Figure 10, both AT 2023fhn epochs are shown, where the best constraints are from the epoch at 138 days. Other LBFOT results are from Ho et al. (2019, AT 2018cow); Nayana & Chandra (2021, AT 2018cow), Ho et al. (2020, ZTF 18abvkwla), Coppejans et al. (2020, CSS161010), Bright et al. (2022, ZTF 20 acigmel, their fit to ne(r) is adopted, shown as a red solid/dash line either side of a possible break) and Yao et al. (2022, AT 2020mrf). Lines of constant M˙˙𝑀\dot{M}over˙ start_ARG italic_M end_ARG are shown for vw=1000subscript𝑣𝑤1000v_{w}=1000italic_v start_POSTSUBSCRIPT italic_w end_POSTSUBSCRIPT = 1000 km s-1. Note that due to different assumptions in the modelling, the densities and/or mass loss rates derived between different authors and objects can differ by up to a factor 5 (DeMarchi et al., 2022).
Table 7: Summary of radio SED fit results using the similar-to\sim90 day data, left, and similar-to\sim138 day data, right, with equation 1. Above the single solid lines we list the broken power-law fit parameters. Below, we list the inferred event properties at each epoch under the synchrotron blastwave model as described in Section 4.4, assuming equipartition (ϵesubscriptitalic-ϵ𝑒\epsilon_{e}italic_ϵ start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT=ϵBsubscriptitalic-ϵ𝐵\epsilon_{B}italic_ϵ start_POSTSUBSCRIPT italic_B end_POSTSUBSCRIPT=1/3). The uncertainties on the event properties are statistical only, and are underestimated due to the presence of systematic errors arising from fixed values of f𝑓fitalic_f, ϵesubscriptitalic-ϵe\epsilon_{\rm e}italic_ϵ start_POSTSUBSCRIPT roman_e end_POSTSUBSCRIPT and ϵBsubscriptitalic-ϵB\epsilon_{\rm B}italic_ϵ start_POSTSUBSCRIPT roman_B end_POSTSUBSCRIPT. We also allowed s𝑠sitalic_s to vary between 0 and 1, with s=1𝑠1s=1italic_s = 1 providing the best-fit in each case. The (rest-frame) fit parameters νpksubscript𝜈pk\nu_{\rm pk}italic_ν start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT and Lν,pk are defined at the intersection point of the two power laws (Chevalier, 1998). Below the double lines we give parameter values calculated with ϵe=0.1subscriptitalic-ϵ𝑒0.1\epsilon_{e}=0.1italic_ϵ start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT = 0.1 and ϵB=0.01subscriptitalic-ϵ𝐵0.01\epsilon_{B}=0.01italic_ϵ start_POSTSUBSCRIPT italic_B end_POSTSUBSCRIPT = 0.01 (note that only the results with ϵesubscriptitalic-ϵ𝑒\epsilon_{e}italic_ϵ start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT=ϵBsubscriptitalic-ϵ𝐵\epsilon_{B}italic_ϵ start_POSTSUBSCRIPT italic_B end_POSTSUBSCRIPT=1/3 are plotted in the relevant figures).
t90similar-to𝑡90t\sim 90italic_t ∼ 90 days
Parameter Unit Value
νpksubscript𝜈pk\nu_{\rm pk}italic_ν start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT GHz 4±plus-or-minus\pm±4
Lν,pksubscript𝐿𝜈pkL_{\rm\nu,pk}italic_L start_POSTSUBSCRIPT italic_ν , roman_pk end_POSTSUBSCRIPT erg s-1 Hz-1 (44+5subscriptsuperscriptabsent54{}^{+5}_{-4}start_FLOATSUPERSCRIPT + 5 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 4 end_POSTSUBSCRIPT)×1029absentsuperscript1029\times 10^{29}× 10 start_POSTSUPERSCRIPT 29 end_POSTSUPERSCRIPT
β1subscript𝛽1\beta_{1}italic_β start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT 3.0±plus-or-minus\pm±0.7
β2subscript𝛽2\beta_{2}italic_β start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT -0.66±plus-or-minus\pm±0.04
Rpsubscript𝑅pR_{\rm p}italic_R start_POSTSUBSCRIPT roman_p end_POSTSUBSCRIPT 1017superscript101710^{17}10 start_POSTSUPERSCRIPT 17 end_POSTSUPERSCRIPT cm 1.3±plus-or-minus\pm±0.7
v/c=Γβ𝑣𝑐Γ𝛽v/c=\Gamma\betaitalic_v / italic_c = roman_Γ italic_β 0.7±plus-or-minus\pm±0.4
M˙˙𝑀\dot{M}over˙ start_ARG italic_M end_ARG/vw 10-4 M yr-1 / 1000 km s-1 0.08±0.01plus-or-minus0.01\pm 0.01± 0.01
nesubscript𝑛𝑒n_{e}italic_n start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT cm-3 2114+33subscriptsuperscriptabsent3314{}^{+33}_{-14}start_FLOATSUPERSCRIPT + 33 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 14 end_POSTSUBSCRIPT
B𝐵Bitalic_B G 0.24±0.02plus-or-minus0.02\pm 0.02± 0.02
U𝑈Uitalic_U 1049superscript104910^{49}10 start_POSTSUPERSCRIPT 49 end_POSTSUPERSCRIPT erg 2.21.4+4.3subscriptsuperscriptabsent4.31.4{}^{+4.3}_{-1.4}start_FLOATSUPERSCRIPT + 4.3 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 1.4 end_POSTSUBSCRIPT
Rpsubscript𝑅pR_{\rm p}italic_R start_POSTSUBSCRIPT roman_p end_POSTSUBSCRIPT 1017superscript101710^{17}10 start_POSTSUPERSCRIPT 17 end_POSTSUPERSCRIPT cm 1.20.4+0.6subscriptsuperscriptabsent0.60.4{}^{+0.6}_{-0.4}start_FLOATSUPERSCRIPT + 0.6 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.4 end_POSTSUBSCRIPT
v/c=Γβ𝑣𝑐Γ𝛽v/c=\Gamma\betaitalic_v / italic_c = roman_Γ italic_β 0.60.2+0.3subscriptsuperscriptabsent0.30.2{}^{+0.3}_{-0.2}start_FLOATSUPERSCRIPT + 0.3 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.2 end_POSTSUBSCRIPT
M˙˙𝑀\dot{M}over˙ start_ARG italic_M end_ARG/vw 10-4 M yr-1 / 1000 km s-1 1.0±0.2plus-or-minus0.2\pm 0.2± 0.2
nesubscript𝑛𝑒n_{e}italic_n start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT cm-3 (3.42.2+5.3subscriptsuperscriptabsent5.32.2{}^{+5.3}_{-2.2}start_FLOATSUPERSCRIPT + 5.3 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 2.2 end_POSTSUBSCRIPT)×102absentsuperscript102\times 10^{2}× 10 start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT
B𝐵Bitalic_B G 0.15±0.01plus-or-minus0.01\pm 0.01± 0.01
U𝑈Uitalic_U 1049superscript104910^{49}10 start_POSTSUPERSCRIPT 49 end_POSTSUPERSCRIPT erg 1912+37subscriptsuperscriptabsent3712{}^{+37}_{-12}start_FLOATSUPERSCRIPT + 37 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 12 end_POSTSUBSCRIPT
t138similar-to𝑡138t\sim 138italic_t ∼ 138 days
Parameter Unit Value
νpksubscript𝜈pk\nu_{\rm pk}italic_ν start_POSTSUBSCRIPT roman_pk end_POSTSUBSCRIPT GHz 5±plus-or-minus\pm±3
Lν,pksubscript𝐿𝜈pkL_{\rm\nu,pk}italic_L start_POSTSUBSCRIPT italic_ν , roman_pk end_POSTSUBSCRIPT erg s-1 Hz-1 (6±plus-or-minus\pm±4)×1029absentsuperscript1029\times 10^{29}× 10 start_POSTSUPERSCRIPT 29 end_POSTSUPERSCRIPT
β1subscript𝛽1\beta_{1}italic_β start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT 2.81±plus-or-minus\pm±0.02
β2subscript𝛽2\beta_{2}italic_β start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT -0.6±plus-or-minus\pm±0.2
Rpsubscript𝑅pR_{\rm p}italic_R start_POSTSUBSCRIPT roman_p end_POSTSUBSCRIPT 1017superscript101710^{17}10 start_POSTSUPERSCRIPT 17 end_POSTSUPERSCRIPT cm 1.20.5+0.3subscriptsuperscriptabsent0.30.5{}^{+0.3}_{-0.5}start_FLOATSUPERSCRIPT + 0.3 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.5 end_POSTSUBSCRIPT
v/c=Γβ𝑣𝑐Γ𝛽v/c=\Gamma\betaitalic_v / italic_c = roman_Γ italic_β 0.40.2+0.1subscriptsuperscriptabsent0.10.2{}^{+0.1}_{-0.2}start_FLOATSUPERSCRIPT + 0.1 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.2 end_POSTSUBSCRIPT
M˙˙𝑀\dot{M}over˙ start_ARG italic_M end_ARG/vw 10-4 M yr-1 / 1000 km s-1 0.320.03+0.10subscriptsuperscriptabsent0.100.03{}^{+0.10}_{-0.03}start_FLOATSUPERSCRIPT + 0.10 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.03 end_POSTSUBSCRIPT
nesubscript𝑛𝑒n_{e}italic_n start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT cm-3 (1.00.5+3.2subscriptsuperscriptabsent3.20.5{}^{+3.2}_{-0.5}start_FLOATSUPERSCRIPT + 3.2 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.5 end_POSTSUBSCRIPT)×102absentsuperscript102\times 10^{2}× 10 start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT
B𝐵Bitalic_B G 0.310.02+0.04subscriptsuperscriptabsent0.040.02{}^{+0.04}_{-0.02}start_FLOATSUPERSCRIPT + 0.04 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.02 end_POSTSUBSCRIPT
U𝑈Uitalic_U 1049superscript104910^{49}10 start_POSTSUPERSCRIPT 49 end_POSTSUPERSCRIPT erg 2.82.2+2.6subscriptsuperscriptabsent2.62.2{}^{+2.6}_{-2.2}start_FLOATSUPERSCRIPT + 2.6 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 2.2 end_POSTSUBSCRIPT
Rpsubscript𝑅pR_{\rm p}italic_R start_POSTSUBSCRIPT roman_p end_POSTSUBSCRIPT 1017superscript101710^{17}10 start_POSTSUPERSCRIPT 17 end_POSTSUPERSCRIPT cm 1.10.5+0.3subscriptsuperscriptabsent0.30.5{}^{+0.3}_{-0.5}start_FLOATSUPERSCRIPT + 0.3 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.5 end_POSTSUBSCRIPT
v/c=Γβ𝑣𝑐Γ𝛽v/c=\Gamma\betaitalic_v / italic_c = roman_Γ italic_β 0.40.2+0.1subscriptsuperscriptabsent0.10.2{}^{+0.1}_{-0.2}start_FLOATSUPERSCRIPT + 0.1 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.2 end_POSTSUBSCRIPT
M˙˙𝑀\dot{M}over˙ start_ARG italic_M end_ARG/vw 10-4 M yr-1 / 1000 km s-1 4.00.4+1.2subscriptsuperscriptabsent1.20.4{}^{+1.2}_{-0.4}start_FLOATSUPERSCRIPT + 1.2 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.4 end_POSTSUBSCRIPT
nesubscript𝑛𝑒n_{e}italic_n start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT cm-3 (1.60.7+5.2subscriptsuperscriptabsent5.20.7{}^{+5.2}_{-0.7}start_FLOATSUPERSCRIPT + 5.2 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.7 end_POSTSUBSCRIPT)×103absentsuperscript103\times 10^{3}× 10 start_POSTSUPERSCRIPT 3 end_POSTSUPERSCRIPT
B𝐵Bitalic_B G 0.190.01+0.03subscriptsuperscriptabsent0.030.01{}^{+0.03}_{-0.01}start_FLOATSUPERSCRIPT + 0.03 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.01 end_POSTSUBSCRIPT
U𝑈Uitalic_U 1049superscript104910^{49}10 start_POSTSUPERSCRIPT 49 end_POSTSUPERSCRIPT erg 2520+23subscriptsuperscriptabsent2320{}^{+23}_{-20}start_FLOATSUPERSCRIPT + 23 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 20 end_POSTSUBSCRIPT

5 Discussion

In this Section, we discuss the host galaxy and derived properties of AT 2023fhn in the context of the other LFBOTs discovered thus far.

The host of AT 2023fhn, taking the spiral and satellite as one interacting system, is broadly consistent with the host galaxies of core-collapse supernovae, and slightly above four other LFBOT hosts in terms of specific star formation rate (although below the host of ZTF 18abvkwla). The host offset and faint, diffuse emission at the transient location (see Section 3.1) are consistent with the tail of the core-collapse supernova distribution (see also Chrimes et al., 2024). While the local and broader environment is consistent with a core-collapse origin, it is interesting to consider whether the high sSFR is related to tidal interaction between the spiral and satellite galaxy. Such interactions may be associated with an elevated tidal disrupted event (TDE) rate, which show a bias towards occurrence in post-starburst galaxies and galaxies undergoing interactions/mergers (French et al., 2016). The non-nuclear location of AT 2023fhn - at high offset from both the spiral and satellite - is difficult to explain in such a scenario (Chrimes et al., 2024). On the other hand, the measured optical magnitudes at the location of AT 2023fhn (see Table 4) allow a contribution from a point source with absolute magnitude as bright as similar-to\sim-14. A significant contribution from a point source at the location of AT 2023fhn is disfavoured (as the precise location has similar brightness to its immediate surroundings, as explained in Section 3.1), but the presence of a globular cluster or ultra-compact dwarf galaxy - which may host massive black holes (e.g. Seth et al., 2014) - cannot be ruled out. The presence of such an undetected cluster or ultra compact dwarf galaxy would be consistent with the upper limit on the black hole mass inferred from our late-time UV observations of 105less-than-or-similar-toabsentsuperscript105\lesssim 10^{5}≲ 10 start_POSTSUPERSCRIPT 5 end_POSTSUPERSCRIPT M (Mummery et al., 2024), given black hole - host galaxy/cluster mass relations (Kormendy & Ho, 2013; Lützgendorf et al., 2013).

The UV-optical, X-ray and radio evolution of AT 2023fhn is broadly similar to other LFBOTs. Notably, however, the X-ray to UV luminosity ratio of AT 2023fhn is an order of magnitude lower than AT 2018cow at similar times, and up to 3 orders of magnitude lower than ZTF 20acigmel. As we show in Section 4.4, it is difficult to attribute this variety to differences in the circumstellar medium density or blast-wave propagation, as AT 2023fhn has a blast-wave velocity, energy and CSM comparable with other LFBOTs. This is consistent with the evidence from other LFBOTs thus far that the X-ray emission arises from a distinct mechanism, namely central engine activity. AT 2023fhn is only the third LFBOT with a mildly relativistic outflow (v=0.40.2+0.1c𝑣subscriptsuperscript0.40.10.2𝑐v=0.4^{+0.1}_{-0.2}citalic_v = 0.4 start_POSTSUPERSCRIPT + 0.1 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT - 0.2 end_POSTSUBSCRIPT italic_c), in common with CSS161010 and ZTF 18abvkwla, demonstrating that the blast-wave is engine-driven. As can be seen in Figure 11, all LFBOTs with sufficient constraints from radio observations thus far have a wind-like (in the sense that density decreases with distance), albeit not r2superscript𝑟2r^{-2}italic_r start_POSTSUPERSCRIPT - 2 end_POSTSUPERSCRIPT, circumstellar density profile. This suggests the CSM was produced by the progenitor system (i.e. through winds), rather than the explosion occurring in a pre-existing dense ISM, which would produce a flat density profile.

An alternative explanation for the variety in UV/X-ray ratios is the viewing angle, where the asymmetric outflow and accretion disc are being viewed from different angles. In this interpretation, the viewing angle to AT 2018cow was closer to perpendicular to the plane of the accretion disc (although not exactly perpendicular, Margutti et al., 2019). This conclusion was also reached by Maund et al. (2023) based on the high polarization (and possibly for AT 2022tsd, given the observation of late-time optical flares Ho et al., 2023c). AT 2023fhn, meanwhile, would have been seen close to edge-on, well off-axis from any asymmetric outflow (e.g. from a choked jet) and with the inner accretion disc obscured (where choked jets and/or the inner disc dominates the X-ray luminosity, Jonker et al., 2020). The effect of viewing angle as an explanation for different LFBOT X-ray luminosities has also been posited by Metzger (2022) and Migliori et al. (2024).

With the fundamental LFBOT requirement of a low 56Ni ejecta mass, and magnetar central engines struggling to explain all aspects of LFBOT phenomenology (e.g. the late-time emission in AT2018cow, Chen et al., 2023; Li et al., 2024), constraints on the possible progenitor models are tightening. IMBH TDE models remain plausible, if a dense wind-like CSM can be produced (e.g. by the ejection of stripped mass during the disruption event). However, the star-forming nature of the host galaxy population, and the locations of LFBOTs within them, likely disfavour such an interpretation. Other plausible models include the delayed merger of black holes and Wolf-Rayet stars (Metzger, 2022), and failed supernovae (Quataert et al., 2019). In AT 2023fhn, the mass-loss wind parameter A1{}_{\star}\sim 1start_FLOATSUBSCRIPT ⋆ end_FLOATSUBSCRIPT ∼ 1 - higher than many radio loud supernovae (Chevalier & Fransson, 2006) and collapsar GRBs (Gompertz et al., 2018; Chrimes et al., 2022). Such a dense circumstellar environment likely require a short-lived evolutionary stage with enhanced mass loss, for example pre-explosion winds from a blue supergiant or Wolf-Rayet star (Margutti et al., 2019).

6 Conclusions

We list here our conclusions about the nature of AT 2023fhn and its place in the context of other LFBOTs and extragalactic transients more generally,

  1. 1.

    Although relatively isolated compared with other LFBOTs and indeed most core-collapse supernovae, AT 2023fhn is otherwise consistent with a core-collapse event, associated with a typical star-forming galaxy and located in a young (albeit diffuse) stellar population.

  2. 2.

    The low X-ray to UV luminosity ratio demonstrates orders of magnitude of variety in this parameter among LFBOTs, which may be indicative of differences in viewing angle. In this interpretation, the relatively low X-ray luminosity of AT 2023fhn is due to an edge-on viewing angle, such that the inner accretion disc is obscured and we are well off-axis from any choked jet/asymmetric outflow.

  3. 3.

    The CSM properties are similar to previous LFBOTs, and are indicative of a dense surrounding medium. Given the wind-like ne density profiles of other LFBOTs, and our ne measurements of AT 2023fhn which continue this trend, it is likely that the dense CSM was produced by wind-like mass-loss from the progenitor system itself (rather than the progenitor exploding in a pre-existing dense ISM).

  4. 4.

    An IMBH TDE interpretation remains possible, only if there exists a pre-existing dense CSM or if the early stages of the tidal disruption produce such an environment. The host galaxy is likely undergoing tidal interactions, which may elevate the TDE rate. While the non-nuclear location and host properties rate favour a core-collapse origin, the explosion of AT 2023fhn in an undetected globular cluster or ultra-compact dwarf galaxy cannot be ruled out.

Despite mounting evidence, the origin of LFBOTs is still ambiguous. Two approaches will elucidate which of the proposed scenarios contribute to the LFBOT population. The first is to grow the sample, enabling statistically robust comparisons of offsets and host galaxy properties to be made with other classes of transient. This will be possible with the advent of new wide-field, deep sky surveys such as those performed by the Vera Rubin observatory (Ivezić et al., 2019). The second is to perform detailed studies of future local events - like AT2018cow - across the electromagnetic spectrum. Such events offer the best opportunity to search for underlying clusters, monitor the long-term evolution, understand the detailed emission physics, and ultimately, determine their progenitors. Although there is much progress still to be made, based on AT 2023fhn and the growing population of LFBOTs, we deem a massive star progenitor with strong winds but low ejecta mass the most likely scenario. This favours models such as black hole/Wolf-Rayet mergers or failed supernovae.

Acknowledgements.
A.A.C. acknowledges support from the European Space Agency (ESA) as an ESA Research Fellow. P.G.J. has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant agreement No. 101095973). P.J.G. is partly supported by NRF SARChI Grant 111692.
Observations analysed in this work were taken by the NASA/ESA Hubble Space Telescope under program 17238. This research has made use of software provided by the Chandra X-ray Center (CXC) in the application of the CIAO package (Fruscione et al., 2006). The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc.
Computing facilities were provided by the Scientific Computing Research Technology Platform of the University of Warwick. This research has made use of the Spanish Virtual Observatory (https://svo.cab.inta-csic.es) project funded by MCIN/AEI/10.13039/501100011033/ through grant PID2020-112949GB-I00. The Pan-STARRS1 Surveys (PS1) and the PS1 public science archive have been made possible through contributions by the Institute for Astronomy, the University of Hawaii, the Pan-STARRS Project Office, the Max-Planck Society and its participating institutes, the Max Planck Institute for Astronomy, Heidelberg and the Max Planck Institute for Extraterrestrial Physics, Garching, The Johns Hopkins University, Durham University, the University of Edinburgh, the Queen’s University Belfast, the Harvard-Smithsonian Center for Astrophysics, the Las Cumbres Observatory Global Telescope Network Incorporated, the National Central University of Taiwan, the Space Telescope Science Institute, the National Aeronautics and Space Administration under Grant No. NNX08AR22G issued through the Planetary Science Division of the NASA Science Mission Directorate, the National Science Foundation Grant No. AST-1238877, the University of Maryland, Eotvos Lorand University (ELTE), the Los Alamos National Laboratory, and the Gordon and Betty Moore Foundation. This publication makes use of data products from the Wide-field Infrared Survey Explorer, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology, funded by the National Aeronautics and Space Administration.

References

  • Avni (1976) Avni, Y. 1976, ApJ, 210, 642
  • Barbary (2016) Barbary, K. 2016, extinction v0.3.0
  • Bright et al. (2022) Bright, J. S., Margutti, R., Matthews, D., et al. 2022, ApJ, 926, 112
  • Cendes et al. (2023) Cendes, Y., Berger, E., Alexander, K. D., et al. 2023, arXiv e-prints, arXiv:2308.13595
  • Chambers et al. (2016) Chambers, K. C., Magnier, E. A., Metcalfe, N., et al. 2016, arXiv e-prints, arXiv:1612.05560
  • Chen et al. (2023) Chen, Y., Drout, M. R., Piro, A. L., et al. 2023, ApJ, 955, 43
  • Chevalier (1998) Chevalier, R. A. 1998, ApJ, 499, 810
  • Chevalier & Fransson (2006) Chevalier, R. A. & Fransson, C. 2006, ApJ, 651, 381
  • Chevalier & Li (2000) Chevalier, R. A. & Li, Z.-Y. 2000, ApJ, 536, 195
  • Chrimes et al. (2022) Chrimes, A. A., Gompertz, B. P., Kann, D. A., et al. 2022, MNRAS, 515, 2591
  • Chrimes et al. (2024) Chrimes, A. A., Jonker, P. G., Levan, A. J., et al. 2024, MNRAS, 527, L47
  • Conroy & Gunn (2010) Conroy, C. & Gunn, J. E. 2010, ApJ, 712, 833
  • Conroy et al. (2009) Conroy, C., Gunn, J. E., & White, M. 2009, ApJ, 699, 486
  • Conselice (2003) Conselice, C. J. 2003, ApJS, 147, 1
  • Coppejans et al. (2020) Coppejans, D. L., Margutti, R., Terreran, G., et al. 2020, ApJ, 895, L23
  • Dai et al. (2018) Dai, L., McKinney, J. C., Roth, N., Ramirez-Ruiz, E., & Miller, M. C. 2018, ApJ, 859, L20
  • DeMarchi et al. (2022) DeMarchi, L., Margutti, R., Dittman, J., et al. 2022, ApJ, 938, 84
  • Dickey & Lockman (1990) Dickey, J. M. & Lockman, F. J. 1990, ARA&A, 28, 215
  • DuPont et al. (2022) DuPont, M., MacFadyen, A., & Zrake, J. 2022, ApJ, 931, L16
  • Eldridge et al. (2017) Eldridge, J. J., Stanway, E. R., Xiao, L., et al. 2017, PASA, 34, e058
  • Fitzpatrick (1999) Fitzpatrick, E. L. 1999, PASP, 111, 63
  • Foreman-Mackey (2016) Foreman-Mackey, D. 2016, The Journal of Open Source Software, 1, 24
  • Foreman-Mackey et al. (2013) Foreman-Mackey, D., Hogg, D. W., Lang, D., & Goodman, J. 2013, PASP, 125, 306
  • Fox & Smith (2019) Fox, O. D. & Smith, N. 2019, MNRAS, 488, 3772
  • French et al. (2016) French, K. D., Arcavi, I., & Zabludoff, A. 2016, ApJ, 818, L21
  • Fruscione et al. (2006) Fruscione, A., McDowell, J. C., Allen, G. E., et al. 2006, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 6270, Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, ed. D. R. Silva & R. E. Doxsey, 62701V
  • Gallazzi et al. (2005) Gallazzi, A., Charlot, S., Brinchmann, J., White, S. D. M., & Tremonti, C. A. 2005, MNRAS, 362, 41
  • Gompertz et al. (2018) Gompertz, B. P., Fruchter, A. S., & Pe’er, A. 2018, ApJ, 866, 162
  • Gottlieb et al. (2022) Gottlieb, O., Tchekhovskoy, A., & Margutti, R. 2022, MNRAS, 513, 3810
  • Granot & Sari (2002) Granot, J. & Sari, R. 2002, ApJ, 568, 820
  • Hayasaki & Jonker (2021) Hayasaki, K. & Jonker, P. G. 2021, ApJ, 921, 20
  • Ho (2023) Ho, A. Y. Q. 2023, Transient Name Server AstroNote, 174, 1
  • Ho et al. (2023a) Ho, A. Y. Q., Bremer, M., Schulze, S., & Perley, D. 2023a, Transient Name Server AstroNote, 100, 1
  • Ho et al. (2023b) Ho, A. Y. Q., Liu, C., Andreoni, I., et al. 2023b, Transient Name Server AstroNote, 93, 1
  • Ho et al. (2022) Ho, A. Y. Q., Margalit, B., Bremer, M., et al. 2022, ApJ, 932, 116
  • Ho et al. (2023c) Ho, A. Y. Q., Perley, D. A., Chen, P., et al. 2023c, Nature, 623, 927
  • Ho et al. (2023d) Ho, A. Y. Q., Perley, D. A., Gal-Yam, A., et al. 2023d, ApJ, 949, 120
  • Ho et al. (2020) Ho, A. Y. Q., Perley, D. A., Kulkarni, S. R., et al. 2020, ApJ, 895, 49
  • Ho et al. (2019) Ho, A. Y. Q., Phinney, E. S., Ravi, V., et al. 2019, ApJ, 871, 73
  • Hoffmann et al. (2021) Hoffmann, S. L., Mack, J., Avila, R., et al. 2021, in American Astronomical Society Meeting Abstracts, Vol. 53, American Astronomical Society Meeting Abstracts, 216.02
  • Inkenhaag et al. (2023) Inkenhaag, A., Jonker, P. G., Levan, A. J., et al. 2023, MNRAS, 525, 4042
  • Ivezić et al. (2019) Ivezić, Ž., Kahn, S. M., Tyson, J. A., et al. 2019, ApJ, 873, 111
  • Johnson et al. (2023) Johnson, B., Foreman-Mackey, D., Sick, J., et al. 2023, dfm/python-fsps: v0.4.6
  • Johnson et al. (2021) Johnson, B. D., Leja, J., Conroy, C., & Speagle, J. S. 2021, ApJS, 254, 22
  • Jonker et al. (2020) Jonker, P. G., Stone, N. C., Generozov, A., van Velzen, S., & Metzger, B. 2020, ApJ, 889, 166
  • Kelly & Kirshner (2012) Kelly, P. L. & Kirshner, R. P. 2012, ApJ, 759, 107
  • Khatami & Kasen (2023) Khatami, D. & Kasen, D. 2023, arXiv e-prints, arXiv:2304.03360
  • Kormendy & Ho (2013) Kormendy, J. & Ho, L. C. 2013, ARA&A, 51, 511
  • Kron (1980) Kron, R. G. 1980, ApJS, 43, 305
  • Kuin et al. (2019) Kuin, N. P. M., Wu, K., Oates, S., et al. 2019, MNRAS, 487, 2505
  • Leja et al. (2017) Leja, J., Johnson, B. D., Conroy, C., van Dokkum, P. G., & Byler, N. 2017, ApJ, 837, 170
  • Li et al. (2024) Li, L., Zhong, S.-Q., Xiao, D., et al. 2024, ApJ, 963, L13
  • Lützgendorf et al. (2013) Lützgendorf, N., Kissler-Patig, M., Neumayer, N., et al. 2013, A&A, 555, A26
  • Lyutikov & Toonen (2019) Lyutikov, M. & Toonen, S. 2019, MNRAS, 487, 5618
  • Margutti et al. (2019) Margutti, R., Metzger, B. D., Chornock, R., et al. 2019, ApJ, 872, 18
  • Martin et al. (2003) Martin, C., Barlow, T., Barnhart, W., et al. 2003, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 4854, Future EUV/UV and Visible Space Astrophysics Missions and Instrumentation., ed. J. C. Blades & O. H. W. Siegmund, 336–350
  • Matthews et al. (2023) Matthews, D., Margutti, R., Metzger, B. D., et al. 2023, Research Notes of the American Astronomical Society, 7, 126
  • Maund et al. (2023) Maund, J. R., Höflich, P. A., Steele, I. A., et al. 2023, MNRAS, 521, 3323
  • Metzger (2022) Metzger, B. D. 2022, ApJ, 932, 84
  • Metzger & Perley (2023) Metzger, B. D. & Perley, D. A. 2023, ApJ, 944, 74
  • Migliori et al. (2024) Migliori, G., Margutti, R., Metzger, B. D., et al. 2024, ApJ, 963, L24
  • Mummery et al. (2024) Mummery, A., van Velzen, S., Nathan, E., et al. 2024, MNRAS, 527, 2452
  • Nayana & Chandra (2021) Nayana, A. J. & Chandra, P. 2021, ApJ, 912, L9
  • Oke & Gunn (1982) Oke, J. B. & Gunn, J. E. 1982, PASP, 94, 586
  • Pasham et al. (2021) Pasham, D. R., Ho, W. C. G., Alston, W., et al. 2021, Nature Astronomy, 6, 249
  • Pellegrino et al. (2022) Pellegrino, C., Howell, D. A., Vinkó, J., et al. 2022, ApJ, 926, 125
  • Perley et al. (2021) Perley, D. A., Ho, A. Y. Q., Yao, Y., et al. 2021, MNRAS, 508, 5138
  • Perley et al. (2019) Perley, D. A., Mazzali, P. A., Yan, L., et al. 2019, MNRAS, 484, 1031
  • Petrosian (1976) Petrosian, V. 1976, ApJ, 210, L53
  • Prentice et al. (2018) Prentice, S. J., Maguire, K., Smartt, S. J., et al. 2018, ApJ, 865, L3
  • Quataert et al. (2019) Quataert, E., Lecoanet, D., & Coughlin, E. R. 2019, MNRAS, 485, L83
  • Rivera Sandoval et al. (2018) Rivera Sandoval, L. E., Maccarone, T. J., Corsi, A., et al. 2018, MNRAS, 480, L146
  • Rodrigo & Solano (2020) Rodrigo, C. & Solano, E. 2020, in XIV.0 Scientific Meeting (virtual) of the Spanish Astronomical Society, 182
  • Rodrigo et al. (2012) Rodrigo, C., Solano, E., & Bayo, A. 2012, SVO Filter Profile Service Version 1.0, IVOA Working Draft 15 October 2012
  • Rodriguez-Gomez et al. (2019) Rodriguez-Gomez, V., Snyder, G. F., Lotz, J. M., et al. 2019, MNRAS, 483, 4140
  • Schlafly & Finkbeiner (2011) Schlafly, E. F. & Finkbeiner, D. P. 2011, ApJ, 737, 103
  • Schrøder et al. (2020) Schrøder, S. L., MacLeod, M., Loeb, A., Vigna-Gómez, A., & Mandel, I. 2020, ApJ, 892, 13
  • Schulze et al. (2021) Schulze, S., Yaron, O., Sollerman, J., et al. 2021, ApJS, 255, 29
  • Seth et al. (2014) Seth, A. C., van den Bosch, R., Mieske, S., et al. 2014, Nature, 513, 398
  • Soderberg et al. (2005) Soderberg, A. M., Kulkarni, S. R., Berger, E., et al. 2005, ApJ, 621, 908
  • Soker (2022) Soker, N. 2022, Research in Astronomy and Astrophysics, 22, 055010
  • Stanway & Eldridge (2018) Stanway, E. R. & Eldridge, J. J. 2018, MNRAS, 479, 75
  • Sun et al. (2022) Sun, N.-C., Maund, J. R., Crowther, P. A., & Liu, L.-D. 2022, MNRAS, 512, L66
  • Sun et al. (2023) Sun, N.-C., Maund, J. R., Shao, Y., & Janiak, I. A. 2023, MNRAS, 519, 3785
  • Tonry et al. (2012) Tonry, J. L., Stubbs, C. W., Lykke, K. R., et al. 2012, ApJ, 750, 99
  • Tremonti et al. (2004) Tremonti, C. A., Heckman, T. M., Kauffmann, G., et al. 2004, ApJ, 613, 898
  • Uno & Maeda (2020) Uno, K. & Maeda, K. 2020, ApJ, 897, 156
  • Xiang et al. (2021) Xiang, D., Wang, X., Lin, W., et al. 2021, ApJ, 910, 42
  • Yao et al. (2022) Yao, Y., Ho, A. Y. Q., Medvedev, P., et al. 2022, ApJ, 934, 104
  • Zhang et al. (2022) Zhang, W., Shu, X., Chen, J.-H., et al. 2022, Research in Astronomy and Astrophysics, 22, 125016

Appendix A SED fitting MCMC results

In this appendix we provide the joint posterior parameter distributions for the host galaxy of AT 2023fhn, in the form of a corner plot (Figure 12) including stellar mass, metallicity, extinction, population age and timescale for an exponentially declining star formation history. These are provided as outputs from emcee SED-fitting using prospector. For the MCMC initial values (and flat priors) we use A=V0.05{}_{\rm V}=0.05start_FLOATSUBSCRIPT roman_V end_FLOATSUBSCRIPT = 0.05 (0¡AV¡2), t=age1{}_{\rm age}=1start_FLOATSUBSCRIPT roman_age end_FLOATSUBSCRIPT = 1 Gyr and M=1010absentsuperscript1010=10^{10}= 10 start_POSTSUPERSCRIPT 10 end_POSTSUPERSCRIPT M (106superscript10610^{6}10 start_POSTSUPERSCRIPT 6 end_POSTSUPERSCRIPT¡M/M¡1012superscript101210^{12}10 start_POSTSUPERSCRIPT 12 end_POSTSUPERSCRIPT), with a flat prior on τ𝜏\tauitalic_τ of 0.1¡τ𝜏\tauitalic_τ/Gyr¡100. For our fiducial run, the redshift is fixed at z=0.238𝑧0.238z=0.238italic_z = 0.238 and the metallicity at Z=0.5Z𝑍0.5subscript𝑍direct-productZ=0.5Z_{\odot}italic_Z = 0.5 italic_Z start_POSTSUBSCRIPT ⊙ end_POSTSUBSCRIPT. Also provided in Table 8 are results when Z𝑍Zitalic_Z is allowed to vary a free parameter.

Refer to caption
Figure 12: emcee output from prospector for the host of AT 2023fhn. Produced using the corner.py code (Foreman-Mackey, 2016) via prospector. The metallicity is fixed at 0.50.50.50.5Z.
Table 8: Host galaxy properties derived from prospector SED fitting, as in Table 6, but allowing the metallicity Z𝑍Zitalic_Z to vary.
Host property Value
M / M (1.210.05+0.07)×1010subscriptsuperscript1.210.070.05superscript1010(1.21^{+0.07}_{-0.05})\times 10^{10}( 1.21 start_POSTSUPERSCRIPT + 0.07 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT - 0.05 end_POSTSUBSCRIPT ) × 10 start_POSTSUPERSCRIPT 10 end_POSTSUPERSCRIPT
SFR / M yr-1 7.31.0+1.2subscriptsuperscriptabsent1.21.0{}^{+1.2}_{-1.0}start_FLOATSUPERSCRIPT + 1.2 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 1.0 end_POSTSUBSCRIPT
Z/Z 0.080.02+0.02subscriptsuperscriptabsent0.020.02{}^{+0.02}_{-0.02}start_FLOATSUPERSCRIPT + 0.02 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.02 end_POSTSUBSCRIPT
AV 0.400.06+0.06subscriptsuperscriptabsent0.060.06{}^{+0.06}_{-0.06}start_FLOATSUPERSCRIPT + 0.06 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.06 end_POSTSUBSCRIPT
tage/Gyr 1.80.4+0.5subscriptsuperscriptabsent0.50.4{}^{+0.5}_{-0.4}start_FLOATSUPERSCRIPT + 0.5 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.4 end_POSTSUBSCRIPT
τ𝜏\tauitalic_τ/Gyr 1.10.6+1.2subscriptsuperscriptabsent1.20.6{}^{+1.2}_{-0.6}start_FLOATSUPERSCRIPT + 1.2 end_FLOATSUPERSCRIPT start_POSTSUBSCRIPT - 0.6 end_POSTSUBSCRIPT