© The Authors 2025

1. Introduction

The connection between long gamma-ray bursts (GRBs) and core-collapse supernovae (CCSNe) is well established at present (see Cano et al. 2017 for a review). Numerous associations have been confirmed, with all corresponding to SNe classified as hydrogen- and helium-deficient objects, with broad spectral lines indicative of large kinetic energies (SNe Ic-BL). Although there have been several confirmations published to date, these explosions remain rare among CCSNe. Moreover, the SNe accompanying X-ray flashes (XRFs¹) are even less frequent. The first clear identification of these events was SN 2006aj (Soderberg et al. 2006; Pian et al. 2006), followed a few years later by SN 2010bh (Cano et al. 2011; Olivares et al. 2012). Since then, no further associations were reported until very recently. Given the scarcity of these events, every new object of this type deserves detailed study, as increasing data in this field will improve our understanding of the connections among XRF-SNe, GRB-SNe, and SNe Ic-BL.

A new XRF-SN association was recently reported (Li et al. 2025). The FXT (EP250108a) was discovered by the EP mission on January 8, 2025, and subsequent observations confirmed the existence of an optical counterpart, designated as SN 2025kg and classified as SN Ic-BL (Eyles-Ferris et al. 2025). Since this discovery, several authors have reported their analyses of the photometric and spectroscopic properties of this event Rastinejad et al. (2025) (R25), Li et al. (2025) (L25), Srinivasaragavan et al. (2025) (S25), and Eyles-Ferris et al. (2025) (EF25).

The light curve (LC) of the SN 2025kg was analyzed in R25, S25, and L25. The first two studies modeled the main emission using an analytical radioactive-decay model and derived nickel masses of 0.2–0.6 M_⊙ and $0.{57}_{-0.3}^{+0.6}$ M_⊙, respectively. Moreover, L25 noted that SN 2025kg was brighter than other He-deficient SNe and proposed a magnetar as an additional power source to enhance its luminosity. The XRF properties and early data of SN 2025kg (t < 6 days) were presented in EF25; S25, and L25. Several scenarios were explored for the initial optical emission; however, S25 and EF25 favored models incorporating an extended CSM to explain the early-time emission. Previous works have analyzed the properties of SN 2025kg using analytical prescriptions; here, we model the LC and expansion velocities using numerical simulations, based on public data. Our results are compared with other objects and previous studies, providing a broader context for SN 2025kg.

2. Models and observations

The bolometric LCs of SNe are very sensitive to their progenitor properties and power sources, while their early evolution can be strongly influenced by a nearby CSM. Theoretical LC models are often compared with observations to derive the progenitor and explosion parameters. However, an important degeneracy arises when only photometric data are used. Including the photospheric velocity evolution, as inferred from some spectral lines, can help to break this degeneracy. Fe II velocities have been proposed as an effective tracer of photospheric velocity (Dessart & Hillier 2005). In this work, we use Fe II velocity measurements from S25 for comparison with our models.

Various authors have estimated the pseudo- or bolometric luminosity of SN 2025kg and their results are in remarkable agreement overall (see black symbols in Fig. 1). S25 produced a bolometric LC by applying the color-calibration method provided by Lyman et al. (2014); their data covered the early cooling phase and the main peak from 1 to 35 days. L25 computed a pseudo-bolometric LC through direct flux integration over rest-frame wavelengths of 3000–9000Å. The data provided by R25 was calculated using grizJHK photometry and co-bands from ATLAS. Although the three LCs are similar around the main peak, significant discrepancies appear at early epochs (t < 6 days) during the first peak (cooling phase). It is possible that the non-inclusion of a UV correction accounts for these differences, even though such a correction carries high uncertainties. Accordingly, the earliest data (and the parameters derived on their basis) should be treated with caution.

Fig. 1. Comparison between SN 2025kg and a set of SNe associated with high-energy emission. Top panel: Bolometric LCs. Bottom panel: Photospheric and FeII line velocities. Black symbols show the available data for SN 2025kg (stars from R25, triangles from L25, and circles from S25). Pink, yellow, and cyan circles correspond to XRF-SN 2006aj, GRB-SN 2023pel, and SN 2020bvc, respectively. The black solid line represents our preferred model for SN 2025kg, which includes CSM interaction, a magnetar and some Ni. Black dashed and dash-dotted lines represent the Ni model and the magnetar model (see Sect. 3 for details). Pink and cyan lines correspond to models of SN 2006aj and SN 2020bvc (Román Aguilar & Bersten 2023, and this work), respectively. Error bars have been included when possible. Inset: A model with a different CSM distribution (dotted line) hints at a closer match at early-time velocities.

For modeling purposes, we adopted the bolometric LC of S25 and supplemented it with the measurements of L25 to improve the temporal coverage. In addition, we included data from R25 as a comparison. Although the early data from the latter authors were not used in the modeling, the data point at ∼58 days was included, as it provides additional constraints on the synthesized nickel mass. The times refer to the detection of the XRF at UT 2025-01-08 12:30:28.34, which is taken as the explosion time throughout this paper.

We compared the observational data with the theoretical LC and photospheric velocity evolution calculated using a one-dimensional radiation hydrodynamic code (Bersten et al. 2011). The code simulates the explosion by injecting some energy manually (E) near the core of the progenitor star (M_cut), which we assumed would collapse. This energy is responsible for the formation of a powerful shock wave that propagates inside the star, transforming the thermal and kinetic energy into radiative energy. The code has a crude treatment of radiation transfer, assuming the diffusion approximation for optical photons and gray transfer for the gamma rays produced by the radioactive decay of ⁵⁶Ni. Any nickel distribution is allowed in our code, while the gamma-ray deposition through the SN ejecta is calculated assuming a constant value for the gamma opacities of κ_γ = 0.03 cm² g⁻¹ (Sutherland & Wheeler 1984). However, a detailed treatment is applied for the hydrodynamical variables, including relativistic effects that become important for fast-moving material. The code has been fully described in Bersten et al. (2011) and it has been used to model several SNe of different types, from H-rich to H-free objects (Taddia et al. 2018; Martinez et al. 2022; Bersten et al. 2024), as well as from normal to more extreme SNe (Bersten et al. 2016; Gutiérrez et al. 2021; Orellana & Bersten 2022). Different energy sources can be included in the code in addition to the explosion energy and ⁵⁶Ni decay, such as a magnetar source (Orellana et al. 2018) and/or the presence of some CSM (Englert Urrutia et al. 2020; Ertini et al. 2025).

As an initial input for our simulation, a pre-SN model representing the state of the star prior to the explosion is required. Hydrostatic structures, calculated with stellar evolution codes, are typically employed as pre-SN conditions. However, the outermost regions of these structures are often modified by hand when the effect of a CSM is incorporated, since no evolutionary models consistently reproduce the CSM conditions required to match the observations. In this work, we use the same grid of stellar models recently employed in Roman Aguilar et al. (2025) (hereafter RA25). These models consist of H-free² structures of different masses, specifically stars with main-sequence mass (M_ZAMS) of 13, 15, 18, 20, and 25 M_⊙ which correspond to pre-SN mass of 3.3, 4, 5, 6, and 8 M_⊙. All of these models have a compact structure at the explosion time with R ≲ 5 R_⊙ before modifying its structure, considering the inclusion of CSM.

We visually compared the models and observations using luminosity and line-velocity measurements. For practical purposes, we first focus on reproducing the main peak, which is primarily powered by either ⁵⁶Ni or a magnetar, and subsequently on the early-time emission. We note that this paper only considers the CSM interaction as a possible explanation for the first peak; however, different scenarios have been explored in other works (see e.g., S25; EF25).

3. Results

Figure 1 compares SN 2025kg with other energetic SNe. For instance, SN 2006aj is associated with an XRF (Šimon et al. 2010; Pian et al. 2006), SN 2020bvc is a SN Ic-BL with some evidence of an offset jet (Ho et al. 2020), and SN 2023pel is the most recent GRB-SN (Srinivasaragavan et al. 2024; Hussenot-Desenonges et al. 2024). In the figure, we highlight some aspects: (1) there is a clear diversity in luminosity within the XRF-SN objects; (2) early emission prior to the main peak is present in all the comparison events. Interestingly, this early emission appears to be more frequent in SNe associated with high-energy radiation than in normal SN Ic and it may provide some insights into their progenitor origin; (3) despite the broad range of radiative output (luminosity), the kinetic energy (velocities) seems to exhibit a more homogeneous behavior; and (4) SN 2025kg is remarkably similar to SN 2023pel in terms of both luminosity and the Fe II velocities.

Recently, RA25 presented a detailed hydrodynamic model of SN 2023pel. In that work, we noted that SN 2023pel was brighter than most of GRB-SNe and exhibited relatively low expansion velocities. A magnetar central engine was proposed to account for these intrinsic properties. Given the similarity between both events, a similar explanation is expected to apply to SN 2025kg.

Figure 1 presents some of our preferred models for SN 2025kg; a model powered only by ⁵⁶Ni (the Ni model) and a magnetar, plus some ⁵⁶Ni (the magnetar model, for simplicity). About ∼0.2 M_⊙ of nickel was also included in the latter to improve the agreement with the observations at later times (t ∼ 58 days), and because some ⁵⁶Ni is naturally expected to be synthesized during the explosion. Finally, our selected model consists of a magnetar plus a CSM (Mag + CSM). The parameters of each of these models are listed in Table 1.

All the models presented in Fig. 1 reproduce the data of SN 2025kg reasonably well around the main peak. We note that models including a magnetar contribution require a more massive progenitor (5 M_⊙; see Table 1) than the Ni model. The same was previously noted in our analysis of SN 2023pel. This is because the magnetar model supplies additional energy, which results in a narrower LC. To counterbalance this effect and reproduce the observed LC width, a more massive progenitor is needed to achieve the necessary broadening. The Fe II velocities are also well reproduced, except for the first data point, which none of our models could reproduce. It is possible that the measurement of this line velocity at early epochs is subject to large uncertainties³. Here, we present only our favorite models, although several alternatives were previously explored to select an acceptable solution in each scenario. These calculations involved different progenitor masses and variations in the free model parameters (E, M_Ni, and P and B). However, given the similarity between SN 2025kg and SN 2023pel, the exploration was guided by our previous results.

As mentioned above, both the Ni- and magnetar models provide a good representation of SN 2025kg observations. However, the Ni model requires a large amount of nickel mass (M_Ni= 0.85 M_⊙), especially considering the low ejecta mass of this model (M_ej = 1.9 M_⊙). This high nickel mass is required to account for the high luminosity observed in this object. This was precisely the main argument used by RA25 to favor an additional energy source in the case of SN 2023pel; it was also invoked by Bersten et al. (2016) for SN 2011kl. In a similar way, we believe that this reasoning applies to SN 2025kg and, therefore, we favored the magnetar scenario and adopted it to model the early-time emission.

In this work, we only explore the possibility that CSM interaction powers the early emission. To include the effect of the CSM, we attached some material in the outermost layer of the pre-SN density profile, assuming a stationary wind law (ρ ∝ r⁻²). After exploring various configurations, primarily adjusting the CSM extension and mass, we identified a model that accurately reflects the data. This model, presented in Fig. 1, has a CSM extension and mass of 500 R_⊙ and 0.27 M_⊙, respectively. A wind velocity of 115 km s⁻¹ was assumed in our calculations. Although CSM models reproduce the initial LC well, they yield a poorer match to the first velocity data. In fact, our CSM models produce even lower velocities at those times. However, adopting a different CSM distribution, rather than the steady wind used here, could improve the velocity match without significantly affecting the initial LC behavior (see the inset in Fig. 1).

In Fig. 1, we also present models for the SN 2006aj and SN 2020bvc calculated using the same code and progenitor grid. The parameters of the models are given in Table 1. In these cases, only a nickel power source was explored, due to the relatively normal luminosities. For the early phase, we attached the CSM to the external density profile, as explained above for the case of SN 2025kg. The values found for M_Ni are within the expected range for H-free SNe and are considerably lower than those found for SN 2025kg (Ni model), in agreement with their maximum luminosities. On the other hand, SN 2020bvc has the highest E of the sample, consistent with the behavior of its velocities, whose values remain higher throughout their evolution. SN 2006aj shows the lower M_ZAMS and M_ej, in agreement with showing the narrowest LC. The properties of the CSM are also in concordance with the behavior shown in their LC, as a slower decay in the early phase is associated with higher values of M_CSM and R_CSM⁴; in turn, this produces higher values of the LC minimum. Finally, we note that the model parameters of SN 2023pel are in very good agreement with the value obtained for our preferred model for SN 2025kg (Mag + CSM), which is expected given the similarities between both SNe.

4. Comparison with previous works

Table 1 shows the parameters derived by other authors for the SN 2025kg. As mentioned before, prior studies used analytical models, relying only on photometric data; therefore, differences are to be expected. Despite this, we found a good agreement between our Ni model and that of S25, when the reported uncertainties are considered. All physical properties, and the parameters derived for the CSM, are generally consistent with our results. The agreement is less satisfactory when comparing our Ni model parameters with those derived by R25. However, the authors do not provide an estimate for the explosion energy, which could have a non-negligible impact on the derived mass values. Regarding the CSM properties, the extent and mass values reported by EF25 are considerably larger than ours, particularly the value of R_CSM. This is noteworthy because the data used in their analysis are systematically less luminous than those considered in this work (see the data from R25 in Fig. 1). Therefore, we would have expected some differences in the opposite direction, with lower CSM extension and mass, as found for SN 2006aj and SN 2020bvc (see Table 1). Although the low value for the explosion energy (E ≲1 foe) assumed in EF25 could perhaps explain some of the differences, it is unlikely to be the only reason.

Li et al. (2025) pointed out that SN 2025kg is a highly luminous event and these authors were the first to propose a magnetar as its potential energy source, as we suggest in this work. However, the inferred parameters differ significantly from those found here, with only the M_ej value being broadly consistent with our results. The E value is significantly higher and it was unclear for us which spectroscopic data were used to infer it. Furthermore, with their adopted values of M_ej and E, it appears challenging to explain the evolution of the expansion velocities of SN 2025kg, which are not particularly large. In addition, the values of P and B differ significantly from those used in our modeling. We have tested models using the parameter values presented by L25, but we were unable to reproduce the observations of this SN. The resulting model shows noticeable deviations from the data; however, we were unable to identify the reason for these discrepancies.

5. Conclusions

The luminous SN 2025kg is another example from the small group of SNe accompanied by an XRF. Its LC exhibits two components: an early cooling emission and a main peak. The main peak and the expansion velocities are remarkably similar to those observed in SN 2023pel, which was associated with a GRB.

Our numerical models indicate that a large amount of nickel (M_Ni ∼ 0.85 M_⊙) and a low ejecta mass (M_ej ∼ 1.9 M_⊙) are required to explain the observations when considering a model powered only by ⁵⁶Ni. Alternatively, a model including an additional energy source provided by a magnetar (with P = 2.9 ms and B = 28 × 10¹⁴ G) and a typical nickel mass (∼0.2 M_⊙) could also reproduce the observations. As in the case of SN 2023pel, we also favor the magnetar scenario for SN 2025kg, as we find this model to be physically more plausible (see RA25). On the other hand, the early LC component was modeled by assuming the presence of some CSM located near the progenitor star before the explosion. By adopting an extension of 500 R_⊙ and a CSM mass of 0.27 M_⊙, we were able to reproduce the early-time emission under the assumption of a steady wind profile. These parameters imply an extreme mass-loss episode ∼0.1 yr before the explosion, which may be difficult to justify physically. An alternative is that the progenitor had an extended envelope, as that needed to reproduce early data in SN IIb (Bersten et al. 2012). Such an envelope could naturally arise from binary mass transfer, which could also favor the magnetar formation scenario.

When comparing the properties of SN 2025kg with those of other events associated with high-energy radiation, we find that their luminosities are highly diverse, while their expansion velocities exhibit much smaller variations. Conversely, an early emission component appears to be relatively common among these objects, in contrast to most normal He-deficient SNe. Some of the observed characteristics seem easier to explain within a binary evolution scenario. We therefore speculate that binary progenitors are likely responsible for this class of high-energy transients.

Acknowledgments

M.B. acknowledges support from PIP 112-202001-10034 and PICT 2020-1141 grants and to CSIC-iCOOP exchange program.

References

All Tables

All Figures

Fig. 1. Comparison between SN 2025kg and a set of SNe associated with high-energy emission. Top panel: Bolometric LCs. Bottom panel: Photospheric and FeII line velocities. Black symbols show the available data for SN 2025kg (stars from R25, triangles from L25, and circles from S25). Pink, yellow, and cyan circles correspond to XRF-SN 2006aj, GRB-SN 2023pel, and SN 2020bvc, respectively. The black solid line represents our preferred model for SN 2025kg, which includes CSM interaction, a magnetar and some Ni. Black dashed and dash-dotted lines represent the Ni model and the magnetar model (see Sect. 3 for details). Pink and cyan lines correspond to models of SN 2006aj and SN 2020bvc (Román Aguilar & Bersten 2023, and this work), respectively. Error bars have been included when possible. Inset: A model with a different CSM distribution (dotted line) hints at a closer match at early-time velocities.

In the text