A targeted search for FRB counterparts with Konus-Wind

We present results of the search for hard X-ray/soft $\gamma$-ray emission in coincidence with publicly reported (via Transient Name Server, TNS; http://www.wis-tns.org/) fast radio bursts (FRBs). The search was carried out using continuous Konus-Wind data with 2.944 s time resolution. We perform a targeted search for each individual burst from 581 FRBs, along with a stacking analysis of the bursts from 8 repeating sources in our sample and a separate stacking analysis of the bursts from the non-repeating FRBs. We find no significant associations in either case. We report upper bounds on the hard X-ray (20 - 1500 keV) flux assuming four spectral models, which generally describe spectra of short and long GRBs, magnetar giant flares, and the short burst, coincident with FRB 200428 from a Galactic magnetar. Depending on the spectral model, our upper bounds are in the range of $(0.1 - 2) \times10^{-6}$ erg cm$^{-2}$. For 18 FRBs with known distances we present upper bounds on the isotropic equivalent energy release and peak luminosity. For the nearest FRB 200120E, we derive the most stringent upper bounds of $E_{\text{iso}}\leq$2.0 $\times 10^{44}$ erg and $L_{\text{iso}}\leq$1.2 $\times 10^{44}$ erg s$^{-1}$. Furthermore, we report lower bounds on radio-to-gamma-ray fluence ratio $E_{\text{radio}}/E_{\text{iso}} \geq 10^{-11}-10^{-9}$ and compare our results with previously reported searches and theoretical predictions for high-energy counterparts to FRBs.


INTRODUCTION
Fast radio bursts (FRBs) are exceptionally bright (∼Jy), shortduration (∼ms) radio transients, discovered serendipitously in 2007 (Lorimer et al. 2007).The dispersion measures (DM) of observed FRBs are well in excess of the expected Milky Way contribution, which implies they are originating from extragalactic distances (see, e.g.Cordes & Chatterjee (2019); Petroff et al. (2022), for a review).Over 600 unique sources have been reported thus far by different radio telescopes (see Table 1), including 492 sources detected by the Canadian Hydrogen Intensity Mapping Experiment Fast Radio Burst (CHIME/FRB) Project (CHIME/FRB Collaboration et al. 2021).Among them, only 18 (Chatterjee et al. 2017;Ravi et al. 2019;Bhandari et al. 2020Bhandari et al. , 2022) ) FRBs have been localized with enough (sub-arcsecond to arcsecond) precision to identify their host galaxies and redshifts, which confirms extragalactic origins and reveals a wide range of galaxy types and local environments surrounding the FRBs (Heintz et al. 2020).More than half of these localizations have been provided by the Australian Square Kilometre Array Pathfinder (ASKAP; Macquart et al. 2010).While most FRBs are only seen once ("one-offs"), a small fraction (∼ 4 %) of them have been found to produce multiple bursts ("repeaters") (Spitler et al. 2016;CHIME/FRB Collaboration et al. 2021;Fonseca et al. 2020;Andersen et al. 2023).
It remains an open question whether all FRBs repeat, and whether multiple progenitor populations of FRBs exist.
Until now, no clear physical picture of either the central-engine that produce a FRB or the mechanism by which the emission is gen-★ E-mail: ridnaia@mail.ioffe.ru† Transient Name Server (TNS), http://www.wis-tns.org/erated has emerged.A wide range of models have been proposed, none of which is able to explain alone the variety of observed events (see Platts et al. 2019 for a review).The most debated progenitor models include magnetars as their central-engines, with the FRB emission originating inside or outside of the magnetosphere (Popov & Postnov 2010;Zhang 2020;Kumar et al. 2017;Katz 2014;Lyubarsky 2014;Metzger et al. 2019;Beloborodov 2017Beloborodov , 2020)).The recent discovery of a FRB-like event from the Galactic magnetar SGR 1935+2154 (FRB 200428;Bochenek et al. 2020;CHIME/FRB Collaboration et al. 2020) strongly suggests that at least some fraction of FRBs may originate from magnetars.The bright radio burst FRB 200428 was accompanied by the simultaneous emission of hard X-rays with properties similar to those of the short bursts typical of Galactic magnetars (Mereghetti et al. 2020;Ridnaia et al. 2021;Li et al. 2021;Tavani et al. 2021), except for the peculiarly hard energy spectrum (Ridnaia et al. 2021).A couple more coincident radio and high energy events were detected from the same source (Dong & Chime/Frb Collaboration 2022;Wang et al. 2022;Frederiks et al. 2022;Maan et al. 2022;Huang et al. 2022;Li et al. 2022), characterized by much fainter radio emission and longer duration than FRB 200428, and softer X-ray spectra typical of magnetar bursts.
To date, there is no other confirmed multi-wavelength or multimessenger transient being associated with any FRB.The presence or absence of a simultaneous or delayed emission corresponding to FRBs in different wavebands would be essential to constrain the emission mechanisms and to identify the FRB progenitor(s).In the last years, many multi-wavelength searches for FRB counterparts have been carried out at all wavelengths with no confirmed results (see, e.g., Nicastro et al. 2021 for a review).In high-energy domain, a number of systematic searches has been made by using archival data of Fermi (GBM, Martone et al. 2019;LAT, Principe et al. 2023), INTE-GRAL (IBIS-ISGRI, Mereghetti et al. 2021), AstroSat (CZTI, Anumarlapudi et al. 2020), Insight-HXMT (HE, Guidorzi et al. 2020), AGILE (MCAL, GRID; Verrecchia et al. 2021), and data of multiwavelength campaigns involving multiple instruments (Cunningham et al. 2019;Trudu et al. 2023).However, most of these studies were based on small FRB samples (less than 50 sources) or only focused on certain objects.
In this work, taking the advantages of the huge increase in the number of detected FRBs and continuous full-sky observations covering the entire current era of FRBs, performed by the Konus-Wind -ray spectrometer (KW), we carry out a targeted search for possible hard X-ray/soft -ray counterparts to over 700 publicly reported bursts from repeating and non-repeating FRBs in KW archival data.The structure of this paper is the following.In Section 2 we provide the FRB sample used in the search and briefly describe our search methodology and upper bound calculations.In Section 3 we present our results, to then discuss it and provide our summary and future prospects in Section 4.

FRB sample
For our analysis we extract all publicly reported FRBs from TNS (799 events, accessed on 2022 April 27).Six events had to be discarded due to incomplete event information, such as FRB coordinates or burst time arrival, and 25 events due to gaps in the KW data at the time of interest.In addition, we decided to exclude 14 repeating sources, which have less than six bursts per source and have no accurate localization.Thus, the FRB sample used in our analysis consists of 721 events detected with 14 radio telescope facilities (see Table 1) between 2001 January 25 and 2022 January 5.This includes 573 thus far one-off FRBs and 148 bursts from eight repeating sources: FRB 121102A, FRB 180814A, FRB 180916B, FRB 181030A, FRB 190303A, FRB 190711A, FRB 200120E, FRB 201124A.Full list of FRB events considered in the analysis and their measured parameters are available at the webpage 1 .Figure 1 shows the dispersion measure distribution of the selected FRBs.To derive upper bounds on the radio-to-high-energy fluence ratio, we use fluence measurements from the first CHIME/FRB catalog (CHIME/FRB Collaboration et al. 2021).

Konus-Wind analysis
Konus-Wind is a gamma-ray spectrometer which has been successfully operating since November 1994 (Aptekar et al. 1995).KW orbit is far from the Earth magnetosphere (since 2004 at distance of ∼ 5 light seconds) that enables nearly uninterrupted observations of entire sky under very stable background.The continuous KW waiting-mode data consist of count rates in ∼ 20 − 80 keV (G1), ∼ 80 − 320 keV (G2), and ∼ 320 − 1300 keV (G3) bands with temporal resolution of 2.944 s.These data are a valuable resource for various studies on hard X-ray/soft gamma-ray transients (Kozlova et al. 2019;Ridnaia et al. 2020).
To search for FRB counterparts, we first estimate the burst arrival time  0 at the KW position for each FRB.For this, we make two time corrections: a frequency-dependent time delay due to dispersion of the radio frequency with respect to soft -rays (infinite frequency) FRBs: 573 "one-off" events and 148 bursts from 8 repeating sources.18 FRB sources that have been associated with a host galaxies, have a luminosity distances range from 3.6 Mpc to 4 Gpc.and a propagation time delay between KW and the telescope site.The combined corrections range from few milliseconds to hundreds of seconds, with a mean (median) value of 9.2 (4.9) s.We then search for significant (> 5) excess over background during the 400 s time interval centered on  0 .While the search interval length of 400 s is chosen arbitrarily, we were motivated by the discovery of the Galactic FRB 200428 accompanied by the simultaneous emission of hard X-rays and by theoretical predictions of very weak highenergy emission on time scales of (at most) minutes after the radio signal (Lu et al. 2020;Metzger et al. 2019).The search is performed in six energy channel combinations (G1, G2, G3, G1+G2, G2+G3 and G1+G2+G3), on temporal scales from 2.944 s to 100 s, similar to Svinkin et al. (2019).The linear background approximation is estimated using two time intervals, before ( 0 -1000 s,  0 -250 s) and after ( 0 +250 s,  0 + 1000 s) the search interval.

Upper bound on the peak flux and fluence
In the case of non-detection of a significant counterpart in the KW data, we estimate upper bounds on its peak energy flux and energy fluence using four template spectral models, which represent typical short and long GRBs (Svinkin et al. 2016;Tsvetkova et al. 2017), huge initial pulses of magnetar giant flares (MGFs; Svinkin et al. 2021), and the Galactic SGR/FRB 200428 event (Ridnaia et al. 2021).These models are characterized by the Band function (Band et al. 1993) or an exponentially cut off power law (CPL), with the parameters listed in Table 2.
In this work we use upper bound on the gamma-ray flux defined as the upper edge of a (frequentist) confidence interval for the flux of the source, according to Kashyap et al. (2010).To estimate an upper bound  ub on the source counts in a particular KW light curve, measured in the energy band Δ, we use the bin with the maximum count rate, for which  max is the observed number of counts,  bg is the estimated number of background counts, and  bg is the error of the background estimation.We define  ub (corresponding to the confidence level CL, hereafter CL=0.9) so that the probability to observe  >  max , assuming that the counts have Gaussian distribution with  =  2 = ( ub +  bg +  2 bg ), equals CL (see Figure 2).We find  (red) we estimate the background count rate  bg (the dashed line).We define upper bound on the source counts  ub (corresponding to the 90 % confidence level) so that the probability to observe  >  max , assuming that the counts have Gaussian distribution with  =  2 = ( bg + ub ), equals 0.9.that the last term  2 bg contribute less than a percent to the  2 , and therefore can be omitted from our calculations.
The upper bound on the source counts then can be converted into a fluence (peak flux) upper bound in the standard energy range (20 -1500 keV) by using the count-to-energy conversion factor  dependent on Δ, the template spectrum, the FRB sky location (the angle of incidence), and the corresponding KW detector response.The maximum value of  ub or  ub /2.944 s is adopted as the upper bound on the corresponding short (< 2.944 s) event energy fluence or the long event peak energy flux, respectively.

Stacking analysis
Current theories make widely varying predictions about FRB highenergy counterparts, with expected emission being faint (below the threshold sensitivity of the present telescopes) in most of the models (see e.g.Chen et al. 2020).Assuming that parameters determining the hard X-ray/soft -ray emission have the same values for all the FRBs, we can employ the stacking analysis.The stacking analysis is a powerful technique that makes it possible to detect sources below the detection threshold.It brings down the statistical noise by combining the signal of many individually undetected sources.
We perform a stacking analysis of the KW data by summing up the background subtracted count rates of the individual event light curves, centered on  0 and then devided by the number of the summed events.To calculate upper bounds for the resulted light curve we use a similar procedure as described in 2.2.1, except estimating an upper bound  ub on rates instead of  ub on counts.Two sets of upper bounds were computed: one based on the bin with the maximum count rate (assuming that all FRB events have the same large (> 3 s) time delay between FRB and its high-energy counterpart) and the other, on the bin comprising  0 (the non-delayed case).
We carry out a stacking analysis of the bursts from each repeating source in our sample and a separate stacking analysis of the bursts from the non-repeating FRBs.

Candidate transients
Our search resulted in two candidate transient events, coincident in time with FRB 160206A and FRB 171019A (see Figure 3).The first one turns out to be a GRB 160206B, which was also detected and localized by Fermi-GBM (trigger 476446756/bn160206430).The GRB localization is inconsistent with the FRB position, which lies far outside the 3  GBM localization region.
In the second case, a KW ecliptic latitude response (Svinkin et al. 2022) for the KW-detected transient is inconsistent with the position of FRB 171019A.Moreover, the FRB position is outside Earthocculted part of the sky for Swift and the source is located right at the edge of the BAT coded field of view, and so a FRB-related GRB Table 3. Upper bounds on the 20-1500 keV fluence (peak flux).

Spectral template
Upper bounds range (10 −7 erg cm −2 ) Typical long GRB * 1 -4 Typical short GRB 5 -10 MGF, GRB 200415A 9 -20 SGR/FRB 200428 1 -7 * For the long burst template, we provide upper bounds on the peak flux in units of 10 −7 erg cm −2 s −1 .might be captured by BAT as a count rate excess.We examine BAT data around the time of FRB 171019A and the KW transient 2 and find no significant count rate increase at 5  level.This, together with the shape of the KW light curve and its detectors' response hints towards this transient being an accidentally coincident GRB.
Thus, we conclude that both candidate transients found in our search are physically unrelated to the FRBs.Based on the continuous KW observations between November 1994 and August 2017 (Kozlova et al. 2019) we estimate an average KW GRB detection rate to ∼ 0.8 GRB per day.Assuming this rate, an expected number of GRBs detected by KW during the total exposure time of our search (∼ 3.25 days) is 2 +3 −2 (95% conf.), which is consistent with the number of the observed events.

Upper Bounds
Our search did not reveal any significant hard X-ray/soft -ray emission associated with the 721 FRB events through the TNS and detected between 2001 January 25 and 2022 January 5. Following the procedure of Section 2 we have set upper bounds, that are presented in Table 3.The stacked data analysis allows us to set a factor of 20 (25 in the case of upper bounds based on the bin comprising  0 ) on average more stringent than individual upper bounds.Figure 4 summarises the results.
For the FRBs with measured redshifts we estimate upper bounds on the total isotropic-equivalent energy release  iso and peak luminosity  iso (see Table 4).While we calculate these upper bounds for each of the four spectral templates, the bounds listed in Table 4 are given on  iso for short GRBs template and on  iso for long GRBs.On average, upper bounds for MGF spectral template results in a factor of two less stringent values and in a factor of two and a half more stringent values for SGR/FRB 200428 template.The upper bounds derived from the stacked data analysis are reported for the repeating FRBs (the upper bounds computed using  0 bin are given in parenthesis).
Using the derived fluence/peak flux upper bounds and the available radio fluences from the first CHIME/FRB catalog, we calculate the lower bounds on radio-to-gamma-ray fluence ratio  FRB .The provided radio fluences are lower bounds due to the telescope sensitivity at the centre of the field of view is assumed (CHIME/FRB Collaboration et al. 2021).We show the distribution of these ratios in Figure 5 for the repeating and non-repeating FRBs from the joint TNS and CHIME/FRB sample.

DISCUSSION AND CONCLUSIONS
Our results, derived with one of the largest FRB sample used so far, are consistent with that found from previous studies.Cunningham et al. (2019) searched for high-energy counterparts to 23 FRBs in GBM, LAT, and BAT data and found  FRB ≥ 10 5 − 10 7 Jy ms erg −1 cm 2 , which is comparable with the derived in this paper.A search for long-duration (1 to 200 s) -ray emission coincident with FRBs was carried out by Martone et al. (2019) in cumulative GBM light curves.They obtained a deep upper limit  FRB > 10 8 Jy ms erg −1 cm 2 .Both primary classes of FRB models (magnetospheric and maser-shock models) predict prompt high-energy counterparts and specify the ratio between the energy emitted by the counterpart and by the FRB itself (Metzger et al. 2019;Cooper & Wĳers 2021;Yang & Zhang 2021).In order to compare our results with theoretical predictions, we have set limits on the radio-to-gamma-ray fluence ratio in dimensionless units.Assuming radio fluence and frequency bandwidth values from the literature (see Table 5) for 12 non-repeating FRBs with known distances we found  radio / iso ≥ 10 −11 − 10 −9 .Although the timescales and energy ranges are not identical to our analysis, this is consistent with the ratios obtained over different FRB samples with different instruments (10 −10 − 10 −7 , Nicastro et al. ( 2021)) and only approaches the ratios expected from theory (10 −6 Lyubarsky (2014) to 10 −5 Metzger et al. ( 2019); Yang & Zhang (2021)).However, one should keep in mind that intrinsic fluence ratios may be significantly different from the observed ones due to beaming effects and that we need statistical limits on fluence ratio of lots of FRBs to constrain the models.Unfortunately, the extragalactic distances and the expected faintness of FRB counterparts put them below the detection thresholds of currently available telescopes, observing at frequencies above the radio band.The nearest and brightest FRBs are the most promising candidates for multi-wavelength observations that could strongly constrain FRB emission models as the models become more quan- titative.At present, only two extragalactic FRBs are located in a relative proximity to us, i.e., FRB 181030A from a star-forming spiral galaxy NGC 3252 at a distance of 20 Mpc (Bhardwaj et al. 2021) and FRB 200120E from a globular cluster in M81 at 3.6 Mpc (Kirsten et al. 2022).From the stacked data analysis of nine bursts from FRB 181030A and six bursts from FRB 200120E in our sample, we derive the most stringent upper bounds on  iso ≤ 6.5×10 45 erg and  iso ≤2.0 ×10 44 erg for short bursts from FRB 181030A and FRB 200120E, respectively.Based on the bounds obtained from our observations, we can exclude GRBs with  iso ≥ 7 × 10 50 erg, that are the majority of the observed by KW population (∼ 97%; Tsvetkova et al. 2017Tsvetkova et al. , 2021)), as counterpart candidates of localized FRBs from our sample.A magnetar flare origin of FRBs is consistent with the derived bounds in terms of either gamma-ray energetics, or radio-to-gamma-ray fluence ratios.For almost all FRBs considered, we can not rule out the occurrence of an extragalactic MGF, with isotropic energy similar to or smaller than that of GRB 200415A ( iso ∼ 1.3 × 10 46 erg) or GRB 051103 ( iso ∼ 5.3 × 10 46 erg) (Svinkin et al. 2021).MGFs with radio-to-gamma-ray fluence ratio similar to that of the 2004 giant flare from the Galactic magnetar SGR 1806-20 ( GF < 10 7 Jy ms erg −1 cm 2 ; Tendulkar et al. 2016) are partly consistent with  FRB ≥ 10 6 − 10 8 Jy ms erg −1 cm 2 derived in this paper.The SGR/FRB 200428 event having a radio-to-gamma-ray fluence ratio of ∼ 7 × 10 11 Jy ms erg −1 cm 2 (∼ 10 −5 in dimensionless units) (CHIME/FRB Collaboration et al. 2020;Ridnaia et al. 2021) is far above our lower bounds on radio-to-gamma-ray fluence ratio.The most stringent KW bounds are placed using the stacked data analisys of bursts from two nearest FRB repeaters, these bounds rule out MGFs, but do not rule out short magnetar bursts with the typical emitted energies below 10 42 erg.
Both detections and non-detections of FRB counterparts in multiwavelength and multi-messenger search campaigns are of great importance.As in the case of other transient phenomena, for example, GRBs, collecting observational data at as wide as possible energy band are crucial for progress towards our understanding of these enigmatic events.The search of high-energy FRB counterparts with KW is a work in progress.With the rapid growth of FRB population we will soon be able to study several more close-by sources, and hence, significantly tighten bounds reported here.

Figure 1 .
Figure1.Dispersion measure (DM) distribution of our selected sample of 581 FRBs: 573 "one-off" events and 148 bursts from 8 repeating sources.18 FRB sources that have been associated with a host galaxies, have a luminosity distances range from 3.6 Mpc to 4 Gpc.

Figure 3 .Figure 4 .
Figure 3. Candidate transient sources found for FRB 160206A (12 significance, left) and FRB 171019A (5.3 significance, right).Arrival time of FRBs corrected for delays due to dispersion and propagation marked by red line.

Figure 5 .
Figure 5. Lower bounds on the radio-to-gamma-ray fluence ratio distribution of FRBs with radio fluences measured by CHIME/FRB.The bounds are in units of Jy ms erg −1 cm 2 .

Table 2 .
The four source spectrum models used in upper bound calculations.Upper bound calculation.For the bin with the maximum observed count rate  max

Table 4 .
Upper bounds on the total isotropic equivalent energy release (short GRBs template) and peak luminosity (long GRBs template) for FRBs with measured redshifts.The upper bounds derived from the stacked data analysis are reported for the repeating FRBs.
a FRB source at a distance of 3.6 Mpc with a formally negative redshift b Bhandari et al. (2022) c Prochaska et al. (2019) d Law et al. (2020) e Bhardwaj et al. (2021) f Kirsten et al. (2022)

Table A1 .
Properties of the bursts from FRB 20200120E