Discovery of the Optical and Radio Counterpart to the Fast X-Ray Transient EP 240315a

Fast X-ray Transients (FXTs) are extragalactic bursts of soft X-rays first identified ≳10 yr ago. Since then, nearly 40 events have been discovered, although almost all of these have been recovered from archival Chandra and XMM-Newton data. To date, optical sky surveys and follow-up searches have not revealed any multiwavelength counterparts. The Einstein Probe, launched in 2024 January, has started surveying the sky in the soft X-ray regime (0.5–4 keV) and will rapidly increase the sample of FXTs discovered in real time. Here we report the first discovery of both an optical and radio counterpart to a distant FXT, the fourth source publicly released by the Einstein Probe. We discovered a fast-fading optical transient within the 3′ localization radius of EP 240315a with the all-sky optical survey ATLAS, and our follow-up Gemini spectrum provides a redshift, z = 4.859 ± 0.002. Furthermore, we uncovered a radio counterpart in the S band (3.0 GHz) with the MeerKAT radio interferometer. The optical (rest-frame UV) and radio luminosities indicate that the FXT most likely originates from either a long gamma-ray burst or a relativistic tidal disruption event. This may be a fortuitous early mission detection by the Einstein Probe or may signpost a mode of discovery for high-redshift, high-energy transients through soft X-ray surveys, combined with locating multiwavelength counterparts.


INTRODUCTION
In the last decade, a few tens of fast X-ray transients have been discovered with Chandra, XMM-Newton and eROSITA (see e.g., Jonker et al. 2013;Glennie et al. 2015;Bauer et al. 2017;Alp & Larsson 2020;Quirola-Vásquez et al. 2022, 2023).These bursts are soft (0.3 − 10 keV) and exhibit a wide range of timescales, lasting from ∼ 10 1 −10 4 seconds, with a variety of astrophysical interpretations having been invoked to explain their properties.
Events such as CDF-S XT2 (Xue et al. 2019), XRT 210423 (Ai & Zhang 2021;Eappachen et al. 2023) and CDF-S XT1 (Sarin et al. 2021) have been interpreted as resulting from a binary neutron star (BNS) merger.CDF-S XT2 and XRT 210423 both showed a clear plateau in the X-ray lightcurve, followed by a sharp drop, consistent with model predictions for a rapidly spinning magnetar remnant.On the other hand, XRT 000519 showed precursor X-ray emission 4000 and 8000 s before the main flare (Jonker et al. 2013), the timescale of which agrees with the expected orbital timescale of a white dwarf (WD) spiralling towards an intermediate-mass black hole (IMBH) on an eccentric orbit (MacLeod et al. 2016).Glennie et al. (2015) found two FXTs in archival Chandra data, and reported an infrared (IR) Galactic counterpart at a distance of 80 pc for one of them (XRT 120830).They interpret this FXT to be consistent with an M-dwarf super flare, but the other had no detected counterpart.Alp & Larsson (2020) reported 12 FXTs from XMM-Newton and from inference of potential hosts they interpret the FXTs as emission from shock breakout in Wolf-Rayet stars within a dense circumstellar medium or (favoured in two cases) red supergiant progenitors.Eappachen et al. (2024) showed that seven of these have plausible host galaxies with spectroscopic redshifts 0.098 < z < 0.645, with one being a likely Galactic flare star.They proposed one FXT (XRT 110621) is consistent with being a supernova shock breakout (SBO) but the spectroscopic redshifts of the others showed that their peak X-ray luminosities were above that deemed feasible for supernova SBOs.Soderberg et al. (2008) report an X-ray detection which they associate with the SBO from SN 2008D, at a distance of 27 Mpc (see also Chevalier & Fransson 2008;Mazzali et al. 2008;Modjaz et al. 2009).Eappachen et al. (2024) also searched for contemporaneous optical counterparts in the Pan-STARRS and ATLAS wide-field surveys but found none.The detection limits range in depth (from m w ≃ 22 to m o ≃ 18.4; AB mags), and the delay between the bursts and the observations range from 1 − 170 days.The most stringent limit on any contemporaneous optical emission remains the serendipitous observation of the location of CDF-S XT 1 with the Very Large Telescope (VLT) just 80 minutes after the burst (Bauer et al. 2017).With this observation, no associated optical counterpart -or host galaxy -was detected, down to a limiting R-band magnitude, m R > 25.7 AB mag.
The discovery and rapid follow-up of FXTs is expected to accelerate since the launch of the Einstein Probe (EP; Yuan et al. 2022) on January 9 2024.With its instantaneous wide field of view of 3600 square degrees, the mission is designed to survey the available nighttime sky several times per day in the soft X-ray regime (0.5 − 4 keV), and to follow-up detected transients.During its commissioning phase, it has already proven to be a valuable discovery instrument, with four new Xray transient sources reported by mid-March.The first, EPW 20240219aa (Zhang et al. 2024b), has had no multi-wavelength counterpart identified, but an association with a sub-threshold Fermi Gamma-ray Burst Monitor (GBM) detection has been made (Zhang et al. 2024a;Fletcher et al. 2024), suggesting it may be a GRB event.The following two EP transients released are almost certainly Galactic.EPW 20240305aa (Liu et al. 2024b) has been well localised by Swift/XRT (Liu et al. 2024a), and is coincident with a Gaia DR3 star (late A-type or early F-type; Monageng et al. 2024), with radio emission observed by ATCA (An et al. 2024).EP 240309a was detected as a highly variable X-ray source (previously detected by XMM-Newton, Swift and eROSITA; Ling et al. 2024), and has been confirmed as a cataclysmic variable with an orbital period of 3.76 hr (Rodriguez & Kulkarni 2024;Buckley et al. 2024).
The fourth bright transient source publicly released by the Einstein Probe mission, EP 240315a, was detected on 2024 March 15 20:10:44 UTC (T 0 = MJD 60384.84079)by the wide-field X-ray telescope (Zhang et al. 2024c).The EP team reported that the event lasted 1600 seconds, with a peak flux, f X ∼ 3 × 10 −9 erg s −1 cm −2 in the 0.5 − 4 keV band.No previously known X-ray sources were identified in the 3 arcmin localisation radius, making it a candidate extragalactic FXT.
In this Letter, we report the discovery of the optical and radio transient associated with EP 240315a, the first time multi-wavelength counterparts of a 'distant' (D ≳ 100 Mpc) extragalactic FXT have been recorded.Throughout this paper we assume ΛCDM cosmology with a Hubble constant, H 0 = 67.7 km s −1 Mpc −1 , Ω M = 0.309 and Ω Λ = 0.691 (Planck Collaboration et al. 2016).We also assume a line-of-sight Milky Way extinction of E(B − V ) = 0.042 AB mag, which corresponds to A V = 0.130 AB mag (Schlafly & Finkbeiner 2011).scope system, operating a wide-field all-sky survey.AT-LAS continually surveys the sky, typically four times in 24 hr when all four units are operating normally, and we promptly process the data to search for extragalactic transients (Smith et al. 2020).During its normal survey operations, ATLAS observed the localisation region of EP 240315a at MJD 60384.894,1corresponding to T 0 + 1.28 hr (note that the first of the four 30 s exposures was obtained at MJD 60384.88673,or T 0 +1.10 hr).Recall T 0 is the time of the detection from the Einstein Probe (MJD 60384.84079;Zhang et al. 2024c).
Observations were performed by the Sutherland unit in South Africa, with 4 × 30 s exposures obtained using the cyan, or c, filter (analogous to the Pan-STARRS/SDSS g + r filters).During automated image processing (outlined by Smith et al. 2020) the observations were reduced and calibrated photometrically and astrometrically with the reference catalogue RefCat2 (Tonry et al. 2018b), and a reference image was subtracted.We registered the optical transient AT 2024eju (ATLAS24dsx) with skycoordinates of RA = +141.64763,Dec = −9.53401(9 h 26 m 35.43 s , −9 • 32 ′ 02.4 ′′ ), and an observed magnitude, m c = 19.38 ± 0.08 AB mag on the Transient Name Server (Tonry et al. 2024).With no detection of the source in ATLAS images 1.75 d before, no historical variability, and a 0.8 arcmin spatial separation, we reported this as a plausible counterpart to EP 240315a (Srivastav et al. 2024b).
In Figure 1, we present the nightly stacked (4 × 30 s) target and difference images from ATLAS for the detection epoch, and the neighbouring epochs immediately pre-and post-detection.The presence of the transient on MJD 60384.894 is unmistakable, with no evidence for AT 2024eju in the most recent previous observation (indicating no pre-existing transient activity), and no evidence in the subsequent observation (indicating its rapid fade).Figure 1 visually highlights how rapidly AT 2024eju rose and subsequently faded.

Optical photometric follow-up
After the initial discovery with ATLAS, we triggered rapid multi-band follow-up imaging observations with the Pan-STARRS telescopes, the Liverpool Telescope and the Lulin Observatory.All three observatories were triggered and on-source within 24 − 36 hours.
We used the 40-cm SLT located at Lulin Observatory, Taiwan, to obtain r-band images of the field of EP 240315a as part of the Kinder project (Chen et al. 2021).The initial observation with SLT began at MJD 60385.673,or T 0 + 0.832 d.We successfully recovered AT 2024eju in the images, albeit with a marginal detection (Chen et al. 2024a), indicating a fast fade within the first 24 hr of the FXT discovery.Subsequently, we conducted continuous observations of AT 2024eju using both SLT and the Lulin One-meter Telescope (LOT) with i-band imaging.We employed the Kinder pipeline (Yang et al. 2021) to conduct PSF photometry for AT 2024eju without template subtraction.The derived magnitudes and 2σ upper limits were determined by calibrating against Pan-STARRS1 field stars in the AB system.
The 2-m Liverpool Telescope (LT; Steele et al. 2004) was triggered under the program PL24A28 (PI: S. Srivastav).Images were obtained in gri-bands commencing on MJD 60385.848,corresponding to T 0 + 1.007 d.While the observing conditions were poor and the optical counterpart was not detected, our upper limit from non-detections confirmed the rapidly fading nature of AT 2024eju (see Srivastav et al. 2024a).Another set of iz-band images were obtained the next night, at MJD 60386.946, in better conditions.However, given the continuing rapid fade, AT 2024eju was only detected in i-band.The derived magnitudes and 2σ upper limits from the LT images were estimated using the python-based Photometry Sans Frustration (PSF) code2 (Nicholl et al. 2023).
Pan-STARRS observations commenced on MJD 60386.324,or T 0 + 1.483 days.The Pan-STARRS (PS) system is a twin 1.8-m telescope system (Pan-STARRS1 and Pan-STARRS2), both situated atop Haleakala mountain on the Hawaiian island of Maui (Chambers et al. 2016).All observations of AT 2024eju were performed with Pan-STARRS1 (PS1), which has a 1.4 gigapixel camera and 0.26 arcsec pixels.This provides a focal plane with a diameter of 3.0 degrees, and a field-of-view area of 7.06 square degrees, which can be imaged with the grizy filter system (as described by Tonry et al. 2012a).Images were processed with the Image Processing Pipeline (IPP; Magnier et al. 2020a;Waters et al. 2020).The individual exposure frames were astrometrically and photometrically calibrated (Magnier et al. 2020b) and overlapping exposures co-added together with median clipping applied (to produce stacks) on which PSF photometry was performed (Magnier et al. 2020c).We commenced targeted observations on MJD 60386.324(T 0 + 1.483 d).Two epochs of observations were obtained on the first night, with the initial grizy followed ∼ 1.4 hr later by izy imaging.We dropped the gr-bands from all subsequent follow-up due to the non-detections in our first epoch.
Finally, we obtained an epoch of late-time iz-band imaging with the Gemini-North/GMOS-N instrument, under the program ID GN-2024A-Q-221 (PI: M. Huber), at MJD 60403.257(T 0 + 18.42 d).These observations were reduced using the DRAGONS pipeline (Labrie et al. 2023;Labrie et al. 2023), and following standard recipes.AT 2024eju was not detected in these deep stacked images, and the 2σ upper limits were again derived using the PSF code.
The full optical lightcurve information, including our ATLAS, Lulin, LT, Pan-STARRS and Gemini photometry, is presented in Figure 2 and Table 2.

Spectroscopic observation with Gemini and redshift measurement
In addition to the rapidly acquired photometric data, we obtained a spectrum of the optical counterpart to EP 240315a, commencing on MJD 60387.236(T 0 + 2.395 days).Our observation was carried out using the Gemini-North/GMOS-N instrument under program ID GN-2024A-Q-128 (PI: M. Huber) using the R400 grating and 1 ′′ slit width, which provided coverage over the ≈ 4200 − 9100 Å wavelength range.Our observation was split into a number of sub-exposures, with a total on-target exposure time of 8880 s.
We reduced our Gemini observation using the DRAGONS pipeline (Labrie et al. 2023;Labrie et al. 2023) and following standard recipes, with the reduced spectrum calibrated against a standard star.There are a number of narrow absorption lines evident in the spectrum that can be used to estimate the redshift to the system.The reduced spectrum is shown in the top panel of Figure 3.We fit four of these lines as Gaussian absorption components, and estimate the centroids of the features.We find that the four absorption features are centred at ≈ 7259, 7270, 8167 and 8222 Å, which we propose are produced by the N v λλ1238.821,1242.804 and Si iv λλ1393.755,1402.770transitions, respectively.With these line identifications, we estimate the redshift of AT 2024eju to be z = 4.859 ± 0.002.There is evidence for prominent Lyman-α absorption at ≈ 7125 Å, in good agreement with our redshift estimate (see bottom left panel of Figure 3).In the bottom-right panel of Figure 3, we show a composite spectrum of the four absorption features from which we have estimated our redshift.Our derived redshift value is in line with measurements from two GCNs released after the discovery of AT 2024eju (z ≈ 4.859; see Saccardi et al. 2024;Quirola-Vásquez et al. 2024).

Radio observations
We observed the position of EP 240315a with the MeerKAT radio telescope.
The observation was performed as part of program SCI-20230907-JB-01 (PI: J. Bright).MeerKAT is a radio interferometer located in the Karoo desert in South Africa, and a precursor of the Square Kilometre Array (SKA).The instrument consists of 64 × 13.5-m antennas that are currently equipped with UHF, L-band and S-band receivers, covering the 0.5 − 3.5 GHz frequency range.Characterised by a dense core and with a longest baseline of 8 km, the array offers an excellent snap-shot uv -coverage, a large field of view (1.69 square degrees) and ∼ µJy sensitivity (Camilo et al. 2018;Jonas 2018).We observed EP 240315a with MeerKAT starting on MJD 60387.703(T 0 + 2.86 d), for a total on-source time of 42 minutes.We observed at a central frequency of 3.06 GHz (S-band, S3), with a total bandwidth of 875 MHz.PKS J1939-6342 and 3C237 were used as flux and complex gain calibrators, respectively.The data were reduced with the OxKAT pipeline (Heywood 2020), which performs standard flagging, calibration and imaging using tricolour (Hugo et al. 2022), CASA (CASA Team et al. 2022) and WSCLEAN (Offringa et al. 2014), respectively.Specifically, for the imaging part we adopted a Briggs weighting scheme with a −0.3 robust parameter, yielding a 3.5 ′′ × 3.5 ′′ beam and 8 µJy beam −1 rms noise in the target field.We clearly detected a point source at the position of the optical counterpart AT 2024eju (first announced by Carotenuto et al. 2024, see also Figure 1).Fitting for a point source in the image plane, we measure a flux density of 34 ± 5 µJy beam −1 .
Upon the discovery of a radio counterpart with the MeerKAT radio telescope, we obtained a rapid response time request with the enhanced -Multi-Element Radio Linked Interferometer Network (e-MERLIN, DD17004; PI: L. Rhodes).e-MERLIN is a UK-based radio interferometer with a maximum baseline of 217 km, and seven dishes spanning 25 − 76 m in diameter.The facility can observe at L-, C-and K-band.Given the improved phase stability and sensitivity, we requested that our observation be made at C-band with a central frequency of 5.08 GHz and a bandwidth of 0.51 GHz.We obtained two observations; the first commenced on MJD 60389.736(T 0 + 4.90 d) and finished on MJD 60391.083(T 0 + 6.24 d), while the second started on MJD 60396.708(T 0 + 11.87 d) and finished on MJD 60398.042(T 0 + 13.20 d), each with a break in the middle whilst the target was below the horizon.The observations consisted of a series of sixminute scans of the target field, followed by two minutes on the phase calibrator (0933-0819).The target- Optical photometry of AT 2024eju.All non-detections are represented by downward-pointing triangles (and correspond to 2σ upper limits).Error bars correspond to 1σ values.We present the data both in observer and rest frame, and the flux in both magnitude and fν space.The grey band represents the duration of the initial X-ray detection reported by the Einstein Probe (≈ 1600 s; see Zhang et al. 2024c).The epoch of spectroscopic observation with Gemini-North has also been marked (vertical black line).Inset panel: A zoom-in of the first ∼ 1.5 days to emphasise how close our ATLAS c-band detection was to the initial X-ray detection (≲ 80 observer-frame minutes, ≲ 800 rest-frame seconds).Note we do not plot the late-time (T0 + 18.42 d, observer frame) Gemini non-detections.
phase cal loops were book-ended by visits to the flux and bandpass calibrators (J1331+3030 and J1407+2827, respectively).The data were flagged, calibrated and imaged using the e-MERLIN pipeline3 (Moldon 2021).Using a uniform image weighting, we do not find any radio emission at either epoch, with 3σ upper limits of 195 µJy beam −1 and 240 µJy beam −1 , respectively.However, combining the observations reduced the rms noise to 17 µJy beam −1 ,4 enabling us to extract a significant detection of 70 ± 8 µJy beam −1 .

RESULTS AND DISCUSSION
The discovery of the optical counterpart AT 2024eju, with its rapidly fading nature and remarkably high redshift (z = 4.859 ± 0.002), represents the first time an extragalactic FXT has been observed at other wavelengths.The redshift means that our optical observations sampled the emitted ultra-violet flux over the first day of its evolution in the source's rest frame.In Table 1, we present the effective wavelength centroids and widths for the filters with which we performed our observations.In the rest frame of the transient, our initial g-band observation with Pan-STARRS sampled λ rest = 820 +120 −110 Å, and we recovered m g > 23.5 AB mag.With our subsequent redshift estimate extracted from our Gemini spectrum (see Section 2.3), this non-detection is expected, since we sampled wavelengths blueward of the Lyman limit (λ = 911.3Å).Even our reddest filter (y-band) only probed λ rest = 1640 +70 −80 Å, a region still well into the UV.Our dense photometric coverage, with intranight cadence (thanks to our coordinated efforts across multiple observatories, strategically placed at different longitudes) from T 0 + 0.054 d through to T 0 + 5.537 d, corresponds to temporal sampling in the rest frame from T 0 + 13.3 minutes to T 0 + 22.7 hours.Our late-time Gemini observations (T 0 + 18.42 observer-frame days) correspond to a rest-frame phase of T 0 + 3.14 days.
The radio counterpart from the MeerKAT radio telescope, an unresolved point source with a flux density of 34 ± 5 µJy, is also the first radio source to be associated with an extragalactic FXT.The combination of what is almost certainly non-thermal radio emission, exceptional ultra-violet luminosity, and rapid evolution indicates that EP 240315a is most likely related to physical mechanisms that produce highly relativistic jets rather than slower thermal transients (e.g., Granot & van der Horst 2014;Anderson et al. 2017;Alexander et al. 2020).

The early optical and radio fluxes
The optical discovery epoch (the cyan diamond data point in Figure 2) was obtained ∼ 520 s (rest frame) after the EP-WXT stopped detecting the initial X-ray emission.It then faded by ∼ 2 magnitudes within ≈ 0.13 rest-frame days (m c = 19.38 ± 0.08 to m r = 21.34 ± 0.22; AB mags), corresponding to a temporal index of ≈ 0.9.The combination of proximity in time of the initial ATLAS detection and the X-ray counterpart, followed by such a rapid decay could indicate that the earliest optical emission is from the same emitting region and mechanism as the X-ray burst.Similar behaviour has been observed in some long GRBs where large fieldof-view optical facilities have obtained simultaneous optical and gamma-ray detections (e.g., Vestrand et al. 2005;Racusin et al. 2008).It is important to distinguish between the prompt and afterglow emission in the optical data to create the most accurate picture of the early-time emission from this system for future modelling efforts.The afterglow component of the observed emission appears to flatten, or plateau, in the optical (izy) bands.
We combined our e-MERLIN and MeerKAT detections with the published Australian Telescope Compact Array (ATCA) 5.5 and 9 GHz ∼ 100 µJy detections (Leung et al. 2024;Ricci et al. 2024), and found that our observations are consistent with self-absorbed synchrotron emission.Given the spectral regime in which our data sits, we expect the radio counterpart to increase in flux density over the coming months.

What is the nature of EP 240315a/AT 2024eju?
Despite our comprehensive follow-up campaign -the first of its kind for an extragalactic FXT -we cannot conclusively determine the origin of EP 240315a.Our observations are consistent with two classes of extragalactic transients: gamma-ray bursts (GRBs) and jetted tidal disruption events (TDEs).
GRBs are identifiable through their highly variable prompt gamma-ray emission, followed by a smoothly evolving synchrotron afterglow, produced as a highly relativistic jet collides with the circum-burst environment.GRB 240315C, detected by the Neil Gehrels Swift Observatory -Burst Alert Telescope (Swift/BAT) and Konus Wind (KW) instruments (DeLaunay et al. 2024;Svinkin et al. 2024a) was temporally coincident with EP 240315a.The BAT signal began at T 0 + 350 seconds and was detected at 15 − 350 keV.The KW signal was detected from T 0 + 374 seconds at 20 − 1600 keV.The GRB detections lasted ∼ 70 and 47 seconds in the BAT and KW data, respectively.The IPN triangulation of GRB 240315C places EP 240315a just within the annulus of 26.7 • width (Svinkin et al. 2024b).Svinkin et al. (2024a) note that the KW detection had an elevated background due to increased solar flare activity.It is rare, but not completely unprecedented, to have a soft X-ray signal before the GRB itself (see e.g., Murakami et al. 1991;Piro et al. 2005), although in the case of EP 240315a and GRB 240315C the X-ray duration is significantly longer than the two previous cases.It is likely the two signals are related, but further investigation of the high energy data is required.
It is possible that EP 240315a and GRB 240315C fall within the class of ultra-long GRBs.Ultra-long GRBs are events whose prompt emission lasts as long as 10 000 seconds.Like 'regular' long GRBs, ultra-long GRBs have a large range of afterglow luminosities (see e.g., Levan et al. 2014, for multi-wavelength studies).As such, we cannot rule out the possibility that EP 240315a and GRB 240515C correspond to an ultra-long GRB event.
A short GRB (events where the prompt flash of gamma-rays is usually shorter than ∼ two seconds)5 interpretation of EP 240315a requires a BNS merger to have occurred ≈ 1.2 Gyr after the Big Bang (derived from z = 4.859), assuming standard cosmological parameters, as adopted in Section 1. Canonically dominant evolutionary channels involve an initially tight binary of massive OB stars which undergo two CCSNe (e.g., Tauris et al. 2017) and crucial common-envelope Table 1.
Wavelength coverage for the different filters/bands of our observations (both optical and radio).We present the wavelengths/frequencies in both the observer and rest frames, to emphasise the short wavelengths/high frequencies being sampled in the rest frame of the transient.The ATLAS co-band wavelength information is taken from Tonry et al. (2018a), while the Pan-STARRS grizy-band wavelength information is extracted from Tonry et al. (2012b).evolution before coalescence via gravitational inspiral.Binary evolution simulations find a range of peak delay time distributions from ≪ 1 Gyr in rapid population synthesis (Belczynski et al. 2018;Chruslinska et al. 2018;Vigna-Gómez et al. 2018) to as low as 10 Myr in the BPASS detailed stellar evolution models (Eldridge et al. 2019).Additionally, the cosmic star-formation rate (and therefore the merger rate) is suppressed above z ≳ 2 (Madau & Dickinson 2014;Mapelli & Giacobbo 2018), but many models are consistent with significant BNS merger rates at z ≈ 5 (Santoliquido et al. 2021).Consequently, we cannot rule out a BNS origin for EP 240315a on the grounds of stellar evolution and inspiral time alone.However, we strongly disfavour the BNS merger/short GRB scenario through the comparison of radio luminosities and timescales.A sample of radio-detected short GRB afterglows is shown in the left-hand panel of Figure 4. Their luminosities are around two orders of magnitude lower than AT 2024eju, making a short GRB an unlikely origin of EP 240315a/AT 2024eju.Unlike short GRBs, long GRBs (events where the prompt flash of gamma-rays is usually longer than ∼ two seconds), 5 produced by collapsing massive stars, 6 occupy a similar region of transient luminosity parameter space.In all wavebands, their afterglow component is more luminous than their short GRB analogues, likely due (at least in part) to higher kinetic energies in the jets of long GRBs compared to short GRBs (Fong et al. 2015;Aksulu et al. 2022).This is demonstrated best in the left-hand panel of Figure 4.The long GRB radio counterparts are approximately two orders of magnitude more luminous than short GRBs.As such, the long GRB radio population is far more consistent with the position of AT 2024eju in the radio luminosity parameter space.A similar conclusion can be reached regarding the position of AT 2024eju in optical luminosity parameter space.The right-hand panel of Figure 4 shows a sample of optically-detected, high-redshift (z > 2; therefore rest-frame UV) GRB afterglows, alongside our photometric measurements of AT 2024eju.Their luminosities and evolution are extremely consistent. 6We acknowledge the growing evidence for a population of merger-GRB events, including GRB 211211A (Rastinejad et al. 2022;Troja et al. 2022;Yang et al. 2022;Gompertz et al. 2023) and GRB 230307A (Gillanders et al. 2023;Sun et al. 2023;Levan et al. 2024a;Yang et al. 2024) that lie within the long GRB T 90 ≳ 2 s parameter space, where T 90 is the time between the burst emitting 5 and 95 per cent of the detected counts.However, here we are referring to the traditional progenitor system picture, where short GRBs are produced by compact object mergers and long GRB events are produced by massive star core-collapse.
Furthermore, the X-ray decay (as reported by Levan et al. 2024b,c) follows f X ∝ t −2.1 , a decay rate which is consistent with a post-jet-break scenario (which occurs when the bulk Lorentz factor of the jet is less than the inverse of the jet opening angle; Sari et al. 1999;Groh et al. 2013;Wang et al. 2018).However, we note that the spectral slope (Photon index = 1.4 ± 0.5) reported by Chen et al. (2024c) is on the hard side of afterglow spectra, but has large uncertainties (Grupe et al. 2013).
The other main possibility for the origin of EP 240315a is a jetted TDE, also known as a relativistic TDE.There have been two well-studied jetted TDEs, and at least two additional candidate events discovered thus far (Burrows et al. 2011;Levan et al. 2011;Cenko et al. 2012;Brown et al. 2015;Andreoni et al. 2022;Pasham et al. 2023).Whilst not all jetted TDEs have optical counterparts, they all have luminous and highly variable X-ray counterparts.In the right-hand panel of Figure 4, we present the values and limits of rest-frame optical/UV jetted TDE observations (in luminosity space), alongside the GRB rest-frame UV detections.Furthermore, all jetted TDEs so far have luminous, long-lasting radio counterparts, consistent with highly relativistic jets (e.g., Zauderer et al. 2011;Rhodes et al. 2023).
At a redshift of z = 4.859, the isotropic X-ray luminosity of EP 240315a (from the average unabsorbed 0.5 − 4.0 keV flux of 5.3 +1.0 −0.7 × 10 −10 erg s −1 cm −2 , as reported by Zhang et al. 2024c) is L X ≃ 1.3 ± 0.2 × 10 50 erg s −1 , over the rest-frame 3 − 23 keV band.This sits at the top end of the luminosity range for the X-ray flares associated with Swift J1644.The X-ray decay and the photon index measurements are also similar (Burrows et al. 2011;Levan et al. 2024c).The first optical data point of AT 2022cmc from Andreoni et al. (2022) was acquired one day post-burst (rest frame), which is later than the detections of AT 2024eju we report here.However, extrapolation of the AT 2022cmc detections shows that they are consistent with the results we report here for AT 2024eju.
At radio frequencies, both Swift J1644 and AT 2022cmc have reported luminous, slowly evolving counterparts, as shown in the left-hand panel of Figure 4.The radio emission comes from external shocks between the jet and the circum-nuclear environment.The radio detection of AT 2024eju, whilst made earlier than for the other jetted TDEs presented, occupies the same luminosity parameter space.
EP 240315a has characteristics that would allow it to be classified as either a GRB or a jetted TDE. Figure 4 illustrates where our radio and optical discoveries sit compared to other extragalactic transients that have been detected in both the radio and optical bands.Cur-rently, it is not possible to differentiate between the TDE or GRB scenarios.Radio observations of both GRBs and jetted TDEs have found optically thick counterparts at early times (Bright et al. 2023;Rhodes et al. 2023), consistent with our findings here.
In the rest-frame UV, all of our detections are consistent with the low-luminosity end of the GRB afterglow distribution.There are no detections of jetted TDEs at such early times.The earliest UV detection of a jetted TDE recorded is ∼ a few days post-burst (rest frame; Yao et al. 2024), later than our final optical detection.Furthermore, only one jetted TDE has been detected in the UV.The other event only has upper limits (Cenko et al. 2012), an order of magnitude below the detections.This range demonstrates the large possible range of UV parameter space associated with TDEs that is still to be explored, making it very hard to estimate the early UV properties of the jetted TDE family.
We rule out the possibility of AT 2024eju being an FBOT-like transient based on the mismatch between the early evolution of this FXT and that of AT 2018cow, the prototypical FBOT transient.While the peak bolometric luminosity of AT 2018cow roughly matches the early follow-up observations of AT 2024eju (L bol ∼ 10 44 erg s −1 ; Prentice et al. 2018), the rise time is much slower; t rise ≃ 3 rest-frame days, versus ≃ 0.3 rest-frame days for AT 2024eju.Additionally, the early radio lightcurve of AT 2018cow demonstrates a slow rise to maximum radio luminosity of ∼ 100 restframe days (see e.g., Ho et al. 2019), which does not match the early, very luminous radio detection for EP 240315a.This FXT event evolves on a much more rapid timescale than AT 2018cow, both in the optical and radio, leading us to rule out FBOTs as an explanation for this transient.

Archival search for other orphan fast-evolving optical transients
Fast-fading transients are commonly found by ATLAS and the Zwicky Transient Facility (ZTF; Bellm et al. 2019), some of which may be extragalactic counterparts to GRBs or FXTs (Stalder et al. 2017;Andreoni et al. 2021).However, the main issue with identifying such transients is the foreground contamination rate of fast cataclysmic variables (CVs), which also often have no host star in Pan-STARRS or Legacy Survey images.In the ATLAS database, there have been ∼ 400 objects flagged as high-significance transients with no catalogued host (Smith et al. 2020).
We manually checked all of these objects and found that most had decay rates that were too slow to match AT 2024eju (likely supernovae or relatively common CVs), or were characteristic of stellar variability with low signal-to-noise ratios.We found 34 genuine orphans detected on only one night, with signs of rapid fading (evidenced by a non-detection in quick succession to the sole epoch of detection).However, the constraints on their rate of fading still do not allow them to be confidently separated from Galactic CVs.Only by combining with external triggers, such as the Einstein Probe, will we be able to build a better understanding of the optical counterparts to extragalactic FXTs.

SUMMARY AND CONCLUSIONS
In this Letter we presented the discovery of the optical and radio counterparts to the Einstein Probe FXT, EP 240315a.
The optical counterpart, AT 2024eju, was detected as a host-less transient by ATLAS during its routine allsky survey operations just 0.054 days (1.28 hours) after the X-ray signal recorded by the Einstein Probe.We recorded a non-detection 1.7 days prior to discovery, and constrained its rapid fade, with it decaying by ∼ 2 magnitudes in 19 hours post-discovery (see Figure 1).
The radio counterpart to EP 240315a was discovered by the MeerKAT radio telescope 2.86 days after the Xray signal was recorded, and has been shown to originate from optically thick synchrotron emission by follow-up complementary e-MERLIN observations.
Our measured redshift (z = 4.859 ± 0.002) rules out some of the models proposed for FXTs, including supernova shock breakouts, binary neutron star mergers and tidal disruption of white dwarfs.The inferred high luminosity of the X-ray, optical and radio emission implies that this is a relativistic event, and we propose two possible scenarios: a long GRB or a jetted TDE.
To differentiate between the GRB and TDE scenarios, continued monitoring of EP 240315a is needed in both the radio and optical bands.The evolutionary timescale of the afterglow is a clear differentiating feature of GRBs from TDEs, as GRBs evolve much more rapidly -usually decaying on a timescale of ∼ days − weeks.If we consider the optical temporal behaviour of AT 2022cmc to be characteristic of all jetted TDEs, then the TDE lightcurve should plateau.Our Gemini upper limits, obtained at T 0 +18.42 days post-discovery, are deep enough to rule out an AT 2022cmc-like lightcurve and luminosity over the same temporal range.The Gemini upper limits favour a GRB-like lightcurve, but are still consistent with the upper limits obtained for Swift J1644 (Bloom et al. 2011).
In the radio band, we predict that differentiating between the TDE and GRB scenarios will take longer, at least 50 − 100 rest-frame days.The decay rate in-Table 2. Optical and radio photometry of AT 2024eju.Magnitudes have not corrected for the expected foreground extinction of E(B − V ) = 0.042 (Schlafly & Finkbeiner 2011).The errors for the optical photometry are quoted to 1σ, while upper limits are quoted to 2σ significance.The radio upper limits are quoted to 3σ.The final radio column contains the result of concatenating the two e-MERLIN non-detections, resulting in a radio detection.The MeerKAT telescope is operated by the South African Radio Astronomy Observatory, which is a facility of the National Research Foundation, an agency of the Department of Science and Innovation.This work has made use of the "MPIfR S-band receiver system" designed, constructed and maintained by funding of the MPI für Radioastronomy and the Max-Planck-Society.

Optical
e-MERLIN is a National Facility operated by the University of Manchester at Jodrell Bank Observatory on behalf of STFC.• OxKAT (Heywood 2020)

2.Figure 1 .
Figure 1.The location of the counterparts to EP 240315a, marked with cross-hairs (red for the optical location from ATLAS, orange for the radio location from MeerKAT).Top panels: ATLAS stacked (4 × 30 s) target images of the field of AT 2024eju.Left to right: o-band observation taken at T0 − 1.696 d, detection c-band observation taken at T0 + 0.054 d, and the subsequent o-band observation, taken at T0 + 1.304 d.AT 2024eju faded below the detection threshold of ATLAS in ≲ 1.2 days (mo > 21.1 AB mag; 2σ upper limit).Middle panels: Same as the top panels, but here we present the difference images.In the centre panel, the presence of AT 2024eju is unmistakable.Bottom left panel: Pan-STARRS z-band image of the field of AT 2024eju, with the localisation regions from the detections of EP 240315a by the EP-WXT (blue; from Zhang et al. 2024c) and EP-FXT (green; from Chen et al. 2024b) overlaid, to illustrate the spatial coincidence with the optical counterpart AT 2024eju.Bottom right panel: MeerKAT radio image of the field of EP 240315a.There is clear evidence for a bright radio source in the image, coincident with EP 240315a.The beam size for the observations is 3.5 ′′ × 3.5 ′′ , illustrated by the stamp in the lower left corner.Note the much smaller scale in this image compared with the optical images.
Figure 2.Optical photometry of AT 2024eju.All non-detections are represented by downward-pointing triangles (and correspond to 2σ upper limits).Error bars correspond to 1σ values.We present the data both in observer and rest frame, and the flux in both magnitude and fν space.The grey band represents the duration of the initial X-ray detection reported by the Einstein Probe (≈ 1600 s; seeZhang et al. 2024c).The epoch of spectroscopic observation with Gemini-North has also been marked (vertical black line).Inset panel: A zoom-in of the first ∼ 1.5 days to emphasise how close our ATLAS c-band detection was to the initial X-ray detection (≲ 80 observer-frame minutes, ≲ 800 rest-frame seconds).Note we do not plot the late-time (T0 + 18.42 d, observer frame) Gemini non-detections.

Figure 3 .
Figure3.Upper panel: Gemini-North/GMOS-N telluric-corrected spectrum of EP 240315a/AT 2024eju.The Pan-STARRS iz-band observations taken ≈ 0.1 days after the spectral observations are overlaid.The absorption lines from which we estimate the redshift of the system have been marked (vertical red lines).Prominent skylines (from Hanuschik 2003) have been marked with vertical grey lines.Lower left panel: A zoom-in on the region of strong Lyman-α absorption (the width of the red band is representative of our redshift uncertainty).Lower right panel: Composite spectrum of AT 2024eju, constructed from the profiles of the four absorption features we used to measure the redshift to EP 240315a(N v λλ1238.821,1242(N v λλ1238.821,.804and  Si iv λλ1393.755,1402.770).770).Note this composite spectrum has been normalised and transformed to velocity space.The width of the red band is again representative of our estimated redshift uncertainty.

Figure 4 .
Figure 4. Left panel: Radio luminosity of different classes of extragalactic transients adapted fromHo et al. (2020), including data fromLaskar et al. (2023) andRhodes et al. (2023).The yellow star indicates the radio luminosity derived from our MeerKAT detection of EP 240315a.The luminosity is consistent with both GRBs and jetted TDEs.Right panel: AT 2024eju optical (rest-frame UV) detections and Gemini upper limits compared to a sample of rest-frame, UV-detected, long GRB afterglows at redshifts of z > 2(Kann et al. 2010), the UV counterpart of the jetted TDE AT 2022cmc(Yao et al. 2024), and upper limits from the candidate jetted TDE Swift J2058.4+0516(Cenko et al. 2012;Andreoni et al. 2022).