Supernovae, by Chandra and XMM-Newton

Abstract

X-ray emission from supernovae can arise from multiple interactions during their evolution. The immediate explosion is sufficiently energetic to generate X-rays; so, too, is the impact of the shock as it runs into circumstellar matter from earlier mass loss phases. A considerable range of physics is on display during the evolution of such X-ray emission. This paper reviews some of the results of observing supernovae obtained by XMM-Newton and Chandra over the past 25 years. Each satellite has contributed significantly to the collection of observations and to our increased understanding of supernovae.

1. Introduction

Humanity recently passed the 25th anniversaries of the launches of the X-ray satellites Chandra and XMM-Newton (hereafter, XMM). Both satellites were built to study a portion of the high-energy universe, roughly X-rays between 0.2 and ∼8 (Chandra) or ∼12 (XMM) keV.

This paper reviews the progress made in understanding supernovae using these two satellites over the past twenty-five years. The study of supernovae in the X-ray band has definitively progressed: a review in 1995 covered ∼5 objects (Schlegel 1995 [1]) while a review in 2025 covered ∼115 objects (Dwarkadas 2025 [2]). Our understanding of the complexity and evolution of the X-ray emissions of supernovae over the past twenty-five years has also deepened as a direct result of Chandra and XMM. A more general review of the contributions to X-ray astronomy of Chandra and XMM may be found in Wilkes et al. (2022 [3]).

Both observatories have been invaluable for making progress in our understanding of the X-ray Universe. XMM has the largest effective area for an imaging instrument ever flown1 [4]. Chandra has the sharpest point spread function ever flown: the on-axis 30% and 90% encircled energy radii for any energy lie below 0.5 and 2 arc seconds, respectively [5]2.

That the above two instrumental properties are on separate satellites means that this review will emphasize different results depending upon the satellite. For XMM, its larger effective area allows us to follow supernovae that are farther away and to follow those that are nearby to fainter stages than typical of previous instruments. For Chandra, its higher spatial resolution allows us to isolate a supernova in a crowded field, as well as to study the surroundings, which can, in principle, lead to mass estimates of progenitors. That the two instruments have differing properties does not imply that they must observe unique objects. Quite a few supernovae have been observed using both instruments.

This review will not be comprehensive, in that no attempt is made to summarize every supernova observed by XMM and Chandra during the past twenty-five years. Instead, representative objects have been chosen to provide a pathway for comparing and contrasting the results from XMM and Chandra observations, as well as the progress delivered by both when observing a single supernova. Supernovae that occurred before mid-2024 were included in the initial overview.

Furthermore, there have been several very good reviews over the past five to ten years covering aspects of X-ray emission connected to supernovae. Chandra (2017 [6]) summarized the radio and X-ray emission of supernovae in dense environments. Frisch & Dwarkadas (2017 [7]) described the effects of supernovae on the local interstellar medium.

Some abbreviations used in this review include CSM—CircumStellar Medium; GMRT—Giant Meter Radio Telescope (India); GRB—Gamma-Ray Burst; SLSN—SuperLuminous SuperNova; VLA—Very Large Array; W-R—Wolf–Rayet (star); XRT—X-ray Telescope (Swift).

2. Advancements from XMM and Chandra

2.1. XMM: Type Ic SN2002ap in NGC 628

The Type Ic SN2002ap in NGC 628 (=M74) was a nearby supernova, as NGC 628 lies ∼7.3 Mpc distant. Optical observations indicated broad spectral lines (e.g., Mazzali et al. [8]; Kinugasa [9]) that have been cited as indicating a very energetic explosion (‘hypernova’). Mazzali et al. [8] presented a model requiring a kinetic energy of 4–10 $\times {10}^{51}$ ergs and a 20–25 M_⊙ C+O progenitor. An interacting binary would provide an ‘easier’ explanation for the lack of H and He that defines the SN Ic sub-class. Indeed, a recent review explicitly stated that Ic’s originate ‘either from core-collapses of very massive stars or from less massive stars in binary systems’ (Solar et al. (2024) [10]). But then there is that word ‘hypernova’.

The use of ‘hypernova’ describes an apparent connection between supernovae and gamma-ray bursts (hereafter, GRBs), first detected with the Ic SN1998bw and the GRB 980425, and defined increasingly well over the past ∼20 years (e.g., Mazzali et al. (2008) [11]; Della Valle 2022 [12]. GRB 030329 showed significant similarity to SN1998bw in optical monitoring. After about one week, the optical spectra exhibited strong similarity to the optical spectra of SN1998bw (e.g., Garnavich et al. [13]), strongly implying that some GRBs originate in core-collapse supernovae. However, the supernova associated with GRB 030329 was not followed in the X-ray band at all, hence the interest in Ic supernovae that are followed, such as SN2002ap.

Sutaria et al. (2003 [14]) used India’s GMRT in combination with XMM to study the Type Ic SN2002ap. The supernova was ∼280” from the nucleus (Nakano 2002 [15]), so XMM’s use was possible. The supernova was faint in the X-ray band, with a flux of ∼${10}^{-14}$ erg s⁻¹ cm⁻². The fitted spectrum matched an absorbed power law of index −2.6 or a thermal bremsstrahlung spectrum with a temperature of 0.8 keV, a seemingly low temperature (see below). The fitted column density was ∼5 $\times {10}^{20}$ cm⁻².

Forward progress in our understanding from observations of SN2002ap has been the result of a simultaneous combination of GMRT and XMM. Two different interpretations of the data exist: are there two different populations of electrons generating the radio and X-ray emission or is there one location with a different emission mechanism?

First, that of Sutaria et al. (2003 [14]): Extrapolating the GMRT upper limits on the radio spectrum to the X-ray band shows that direct synchrotron emission produces an insignificant fraction of the observed X-rays. SN2002ap’s prompt emission must then arise from different populations of electrons to describe the radio and X-ray emission. Direct synchrotron emission as a mechanism for early emission has been eliminated for SN2002ap. The remaining questions are as follows: Does it play a role in other supernovae? If so, what properties dictate when it is present and when it is absent?

Sutaria et al. (2003 [14]) then described two possible origins for SN2002ap. First, the progenitor could have been a Wolf–Rayet (W-R) star. W-R stars have empirically measured $\dot{M}\sim$1.5–3 $\times {10}^{-5}$ M_⊙ yr⁻¹, which would be sufficient to explain the X-ray measurements (e.g., Nugis & Lamers (2000) [16]). Second, interacting binaries can easily provide the necessary mass transfer rate. A limitation on ‘interacting binaries’ as an explanation is the supernova class: Ic. Efforts over the past ∼25 years have shown that Ic supernovae shed most of their H and He layers before the supernova. Consequently, to invoke an interacting binary model, one must acknowledge, or have the data to show, that such a model remains viable. The expected mass transfer rate for the interacting binary model is $\sim {10}^{-4}$ M_⊙ yr⁻¹ for ∼3000 years, with a wind speed of ∼100 km s⁻¹. Given the measurements obtained, both the W-R star and the interacting binary models remain viable.

At this point, the reader may wonder why the discussion of SN2002ap seems to have two separate branches: the Ic behavior, and the seemingly off-topic GRB discussion. The branches are, in fact, separated here for a very specific reason: no GRB was associated with SN2002ap. That likely was disappointing to the SN-GRB crowd, but Totani (2003 [17]) noted that the jet could simply be pointed away from us. The SN-GRB connection has been discussed for some time, yet some Ib/Ic supernovae are associated with GRBs and some are not. However, Totani [17] noted that a GRB jet should have swept up sufficient interstellar matter within a few to ten years that shocks would exist and be detectable in the radio—a clear test for the jet hypothesis. To date, no radio observations appear to have been carried out (or, at least, published) for SN2002ap. Similarly, Ramirez-Ruiz & Madau (2004 [18]) described the expected Compton scattering that should occur for a misaligned jet, again directly testing the overall SN-GRB connection.

A second interpretation of the SN2002ap data was published by Bjornsson & Fransson (2004 [19]). They focused on the radio and X-ray emission, arguing that the emission arises from the same region (in contrast to Sutaria et al. (2003 [14]) as a result of inverse Compton cooling. They demonstrated that inverse Compton cooling can explain not only the steep radio spectrum, but also the ‘low temperature’ X-ray emission. The authors further showed that radio/X-ray emission yields a measure of the ratio of the energy densities in the magnetic field and relativistic electrons of approximately unity.

Normally at this point there would be a comparison or a contrasting of the two models. However, a recent pre-print by Fiore, Menegazzi, & Stratta [20] reviewed the GRB-SN connection in considerably more detail than is possible here and argued that a ‘conclusive answer remains elusive’. The interested reader should definitely consult that paper. And it remains the case that some fraction of long-duration GRBs do not have accompanying supernovae and some Ic supernovae do not exhibit GRBs. Is this physics or a lack of necessary data? At the moment, the answer eludes us.

The situation may be even more extreme than the above paragraphs imply: Soker (2024, [21]) explicitly compared the neutrino mechanism for generating potentially all core-collapse supernovae with the ‘jittering jets’ mechanism. The comparison hinges on the apparent inability of the neutrino mechanism to produce the observed energies, a potentially fatal limitation for the model. One potential path forward is the unification of both explosion mechanisms: the jittering jets start the explosion in the core, which then triggers neutrino heating, which then boosts the jet energies sufficiently to disrupt the star.

The community has reached this point at least partially because of the improved resolution of X-ray detectors. Nailing down the detailed physics will require the next generation of X-ray observatories, particularly a combination of very high spatial and spectral resolution. This portion of the review has been extra long precisely because SN2002ap provides a good case to illustrate the physics problems facing the community.

2.2. Chandra: Type IIn SN2017hcc

Discovered on 2017 October 2.38, SN2017hcc lies at a distance of ∼75 Mpc and was found within a day or so of its explosion. It was quickly classified as a IIn supernova and reached its peak brightness around day 40–45 (Prieto et al. (2017 [22]). The first X-ray observations were obtained at days 29, 47, and 70 using Swift XRT and UVOT. All of the X-ray observations yielded upper limits of a few $\times {10}^{40}$ erg s⁻¹.

Chandra et al. (2022 [23]) used Chandra to detect SN2017hcc at day 727; previous observations using the Swift XRT led to upper limits. The Chandra observation yielded a flux of ∼$2.9 \times {10}^{-14}$ erg cm⁻² s⁻¹, about an order of magnitude fainter than the XRT upper limits. That flux yielded an unabsorbed luminosity of ∼$2 \times {10}^{40}$ erg s⁻¹. SN1998S was of comparable brightness at that age; the X-ray luminosities of other IIns are at least several factors larger.

Chandra et al. (2022 [23]) concluded that 2017hcc was X-ray and radio faint, while also reporting strong IR emission. That is a puzzle: a ‘dense’ CSM implies strong dust formation, hence strong IR emission. However, most strongly emitting SNe in the IR are also strong X-ray emitters, consistent with a dense stellar wind. SN2017hcc does not follow that behavior, suggesting an asymmetry in the CSM.

2.3. Chandra: Type SLSN I SN2017egm

SN2017egm was a super-luminous, He-rich supernova. Superluminous supernovae (=SLSNe) are roughly 2–3 mag more luminous than normal supernovae (Gal-Yam 2019 [24]). The first recognized SLSN was SN2005ap (not to be confused with SN2002ap described above) at a redshift of ∼0.28 (Quimby et al., 2007 [25]). Because of the distance, no X-ray observations were obtained. In the years since SN2005ap, about 100 SLSNe have been detected. With one exception, none of the SLSNe have been detected in the X-ray band with upper limits for nearby events of ∼5 $\times {10}^{41}$ erg s⁻¹ [24]. The single exception is the SLSN SCP06F6 at z = 1.189, for which XMM yielded ${\mathrm{L}}_X\sim {10}^{45}$ erg s⁻¹ in the 0.2–2 keV band—the most luminous X-ray detection of a supernova ever.

Enter SN2017egm. Most of the SLSNe detected to date have been found in low-mass, metal-deficient galaxies (Gal-Yam 2019 [24]). SN2017egm was H-poor (typical of SLSN), nearby (z = 0.03, a rare occurrence to date), and exploded in a massive, metal-rich galaxy (a unique object?). Strong He I emission at $\lambda$10830 Å was present in spectra obtained at day 105 (Zhu et al., 2023 [26]).

Four observations were obtained with Chandra, at days 4, 85, 136, and 323 after the optical peak. None of the exposures yielded a detection. The data were merged, yielding a total exposure of 96 ks, which still did not lead to a detection. The 3$\sigma$ upper limit, assuming a powerlaw spectrum with a $\mathsf{\Gamma }=2$ index and the column density toward the source of $1.08 \times {10}^{20}$ cm⁻², was $1.9 \times {10}^{39}$ erg s⁻¹ in the 0.5–10 keV band and with an X-ray-to-optical luminosity ratio $<{10}^{-3}$. These are very strict upper limits. Based on a review by Margutti et al. (2018 [27]) (see Section 5 below), the vast majority of SLSN upper limits are above a few $\times {10}^{42}$ erg s⁻¹, with an average upper limit closer to 10⁴³ erg s⁻¹. The first thought, of course, is that the adopted column density is incorrect by orders of magnitude. However, for SN2017egm, adopting a column of 10²⁴ cm⁻² leads to an upper limit of 10⁴¹ erg s⁻¹, still about an order of magnitude below the others.

As explored by Zhu et al. (2023 [26]), all of the simple models invoked to explain SN2017egm’s behavior failed to do so (e.g., enhanced ${}^{56}\mathit{Ni}$ production, energy from rapidly rotating magnetic stars, pair instability), leading to more complex models. Multiple circumstellar matter shells could explain the multiple light curve peaks seen in SN2017egm. But the X-ray-to-optical luminosity ratio appears to be a stumbling block.

2.4. Chandra: Type Ib SN 2014C

A campaign that started around day 300 post-explosion using Chandra, NuSTAR, and the VLA in coordination followed SN2014C until day ∼2300 (Brethauer et al., 2022 [28]). They used the data to investigate the mass loss and circumstellar distribution of matter. The result was a significant advance in understanding mass loss history and shock interactions post-core collapse. Margutti et al. (2017 [29]) had analyzed the X-ray data prior to day ∼300; Thomas et al. (2022 [30]) had also analyzed the joint Chandra and NuSTAR datasets, but had not removed from the NuSTAR data the emission from nearby sources3.

Brethauer et al.’s joint data analyses between Chandra and NuSTAR provided a much-improved spectral fit. From the combined analyses, they found a linear decline in the column density from ≈$3 \times {10}^{22}$ cm⁻² by a factor of ten over their ∼2000 day study. The fitted bremsstrahlung temperature reached a peak of 23 keV around day 500–600. Interestingly, their subsequent detected decline was t^−0.5, instead of the adiabatic expansion value of t⁻².

Brethauer et al. (2022 [28]) demonstrated that the radio and X-ray emissions arise from two different emitting locations, given the different inferred shock velocities: they inferred that the radio emission arises from the fastest-moving material, while the X-ray emission arises from the densest material. They also noted that VLBI the observations are consistent with a spherical shell, while the optical observations revealed a broad H$\alpha$ emission that emerged around day 127. Since spectra obtained within days of the explosion indicated H-poor material, the appearance of the broad H$\alpha$ then implied a shock running into a circumstellar medium. The only apparent path for no interactions prior to that time is a quasi-spherical circumstellar shell impacted by the forward shock and generating radio emission, plus a H-rich circumstellar disk impacted by the shock that also slowed the shock and generated the X-rays and broad H$\alpha$ emission. Other than SN1987A, this is perhaps one of at most a handful of supernovae that have been ‘disassembled’ to such a degree.

2.5. XMM and Chandra: Type IIn SN2010jl

SN2010jl generated significant interest, as it was detected in the X-ray band with Chandra by Chandra et al. (2012 [31]) and Ofek et al. (2013 [32]). The fitted X-ray spectrum at one month post-explosion implied a very high column density (10²⁴ cm⁻²) that dropped by a factor of approximately three over the next year, suggesting a very dense circumstellar medium. But the assumed model used to fit the spectra was not the best choice given the conditions. Ofek et al. (2014 [33]) re-fit the data and adopted a high, but more reasonable, column of $2 \times {10}^{23}$ cm⁻². However, their fits showed a drop of ∼3 from approximately one month to one year post-explosion, consistent with the results of Chandra et al. (2012 [31]). Ofek et al. [33] also used NuSTAR to observe a supernova for the first time, demonstrating that the shock temperatures generated by typical shock velocities (∼${10}^4$ km s⁻¹) would be >10 keV, as was the case for SN2010jl. From the X-ray observations, Ofek et al. [33] inferred a circumstellar mass larger than 10 M_⊙, hence a very dense circumstellar medium. Chandra et al. (2015 [34]) questioned that result, noting that the presence of an Fe K$\alpha$ supports a high value of column density. Clearly, we have more to learn from the circumstellar matter surrounding a supernova.

3. Older Supernovae

While XMM and Chandra are ‘old’, they were launched on a specific date. That automatically allowed both satellites to look at supernovae that occurred before their launch. As mentioned earlier, XMM’s larger effective area allowed the study of objects that lay beyond the reach of Chandra or earlier missions. Chandra’s sharper PSF allowed the study of the environments of earlier supernovae to increase the reliability of earlier observations.

3.1. Chandra and XMM: Type II SN1987A

The first ‘old’ supernova to be discussed is SN1987A, as both satellites were used to study its emission via their dispersion gratings. A brief summary is in order.

SN1987A was first detected in the Large Cloud of Magellan on 1987 Feb 23 and was the first non-Solar object detected with neutrinos (McCray 2017 [35]; original papers: Hirata et al. (Kamiokande-II) 1987, 1988 [36,37]; Bionta et al., 1987 [38]; Bratton et al. (IMB) 1988 [39]). The light curve was unusual because, while the spectrum showed SN1987A was a Type II supernova, the light curve continued to brighten for several months after the explosion [35]. This behavior becomes understandable if the progenitor was a blue, rather than a red, supergiant. The progenitor was, in fact, identified from pre-supernova images of the LMC as a blue supergiant (Sonneborn 1988 [40]).

Observations about five years after the explosion showed three circumstellar rings that have been estimated to have been ejected about 20 Kyr before the supernova occurred (Crotts, Kunkel, & Heathcote 1995 [41]). An image obtained in 2010 with the Hubble Space Telescope was presented in McCray (2017 [35]). Further, the symmetry of the circumstellar rings suggests that the progenitor was a binary star. However, as yet no evidence has been identified as signaling the existence of a companion [35]. That has led to speculation that the putative binary was a binary, but that common envelope evolution occurred before the explosion (Chevalier & Soker 1989 [42]; Podsiadlowski, Joss, & Rappaport (1990) [43]). Such a scenario could explain why no companion has been detected, as well as why SN1987A’s progenitor was ‘blue when it blew’4.

Technically, SN1987A is now a supernova remnant, as the luminosity is dominated by shocked emission from the blast wave (McCray 2017 [35]). The energetics of the explosion coupled with the neutrino temperature (∼4 MeV) and the short duration (∼10 s) both imply that a neutron star should be present [35]. To date, no neutron star has been found. However, evidence is accumulating that a neutron star is almost certainly present: JWST detected emission lines of ionized argon around the core, which require high-temperature ionizing radiation for the lines to be produced (Fransson et al. (2024) [44]). Regardless, it is interesting to watch one supernova evolve into a remnant. Historically, supernovae went off, disappeared, and perhaps emerged as a remnant at some point. Consider, briefly, SN1885A in the Andromeda Galaxy, the second-nearest supernova. Observations could not be obtained until the late 20th century, by which time the supernova was a remnant (Fesen et al., 2017, Fesen et al., 1999, Fesen et al., 1989; [45,46,47]). It very likely has also been recovered with Chandra (Prucker et al. (2025) [48]).

Chandra and XMM both carry dispersion gratings. The luminosity of SN1987A is sufficient to allow both dispersion gratings to be used. The emission at this time mostly resides in the 0.5–2 keV band (6–20 Å). In that band, there are H- and He-like emission lines for Ne, Mg, and Si, as well as partial coverage of the O VII and O VIII emission lines. There are also emission lines from the L-shell of Fe XVII and Fe XX.

In general, a single-temperature model does not fit all the lines nor any single line completely—not surprising, given the range of densities and temperatures present. Using emission measure distributions, one can detect a bi-modal distribution of temperatures with kT ≈ 0.5 and ≈2 keV. The line widths typically have FWHM ≈ 500 km s⁻¹. A typical shock temperature–velocity relation translates the 2 keV temperature to a velocity of 1000 km/s, which is not observed. The 2 keV gas has been produced by the reflected shock, which slows the gas while simultaneously raising its temperature.

What about longer-term monitoring? Four such studies are readily apparent. Dewey et al. (2008; [49]) reported a ‘significant decrease’ between 2004 and 2007 of the highest-temperature component of the shocked plasma from ∼2.7 to ∼1.9 keV. While that may not seem like a large change, it represents a good bit of cooling in about three years.

Second, Sturm et al. [50] used the RGS on XMM in 2003 and once-a-year from 2007 to 2009. A simple display of the four spectra (Figure 1 in their paper) shows a clear evolution of the line emission from 2003 to 2007 to 2008/2009, with an increase in the O VII and Fe XVII lines in the 0.7–0.8 keV band.

Third, Ravi et al. [51] described the most recent (2018) deep exposure Chandra grating observation of SN1987A in comparison with grating observations obtained in 2004, 2007, and 20115. They found that the line emission and broadband fits of two-component shock models yielded consistent results for shock temperatures. A comparison with grating spectra obtained in 2007 and 2011 showed that the shocks are encountering less dense circumstellar material, suggesting that the shocks may be exiting the immediate environment before encountering the material deposited by the supergiant wind well before the supernova explosion. The elemental abundances support that interpretation, as the authors have not yet detected significant evolution.

Finally, Bray et al. [52] adopted a different approach: they took shorter Chandra HETG exposures but at a higher cadence: instead of 200+-ks exposures (2011, 2018), they obtained 50–70 ks HETG exposures. Necessarily, the signal-to-noise ratio was lower, so the statistical uncertainties were larger. Nevertheless, evolution was detected in the emission measures (outgoing and reverse shocks), whereas the abundances generally did not show evolution.

In summary, the dispersive gratings on both XMM and Chandra have provided high-resolution X-ray spectra of SN1987A. That we are so close to this supernova also means we have been able to witness the evolution of the supernova to a supernova remnant. That witnessing gets directly to the evolution, the physics, of an explosion with time.

3.2. Chandra and XMM: Type II SN1986J in NGC 891

A supernova was identified in the edge-on galaxy NGC 891 from VLA observations. After their launch, Chandra and XMM observed NGC 891 in 2000 and 2002, respectively, as described by Temple, Raychaudhury, & Stevens (2005) [53]. SN1986J had been observed previously with ROSAT and ASCA. From the ROSAT observations, Houck et al. (1998) [54] had determined that the X-ray light curve declined as $t^{-2}$—a fast decline rate. The 51-ks Chandra observation permitted correcting contamination in the ASCA data by emission from a nearby source. By the time XMM observed SN1986J (for ∼15 ks (MOS), ∼10 ks (EPIC)), it had faded: Temple, Raychaudhury, & Stevens (2005) [53] reported having difficulty determining the flux and had to simultaneously fit the spectra from all three data frames (EPIC + MOS1 + MOS2) to constrain the model parameters sufficiently well. The result was an even steeper decline rate: $\sim t^{-3}$. Alternatively, as the authors noted, it is possible that the ROSAT fluxes included non-SN point sources, thereby steepening the decline. However, one can also argue that the decline from Chandra to XMM, a period of about two years, was in fact fast given that the larger effective area for XMM over Chandra should have compensated for the shorter exposure times obtained by XMM. Clearly, we have more to learn about late-time emission behavior as the shocks propagate through the CSM.

There are additional older supernovae that could be covered here, e.g., SN1993J (Chandra et al., 2009 [55]) or SN1978K both observed by Chandra and XMM, but in the interests of keeping this review relatively short, these objects will be skipped.

4. On-Going Supernova: Type IIL SN2023ixf

Given the increasingly dense monitoring of sky variability, the number of supernovae discovered per day has increased considerably over the lifetimes of XMM and Chandra. As a result, while this review is limited to a particular range of time, there are supernovae that are under active monitoring. One warrants attention: SN2023ixf in M101, the closest supernova since SN1993J.

This supernova occurred well out in an arm of M101, so both Chandra and XMM could obtain spectra with little concern regarding contamination from the nucleus. Chandra obtained ACIS observations around day 13 and day 85; XMM observed on days 9, 30, and 58. SN2023ixf was also the first time that NuSTAR observed a supernova—its harder bandpass permitted a measurement of the shock temperature of ∼30–35 keV, well beyond Chandra or XMM’s capabilities for constraining the temperature.

For Chandra, the column density dropped roughly a factor of eight over the ∼75 day gap (Chandra et al., 2024; Nayana et al. (2024) [56,57]). Folding in an earlier NuSTAR observation shows the column density dropping by a factor of ten to the time of the first Chandra observation. From additional observations, the authors inferred the existence of a complex structure of circumstellar matter.

5. Additional Results of X-Ray Observations

Finally, now that there are so many X-ray detections, the community is starting to see comparison studies. Two stand out to the author at the time of this review.

5.1. XMM: Supernova Density

Sun et al. (2022) [58] described using the results of a search through the XMM archive to study the luminosity function and event rate density of supernova shock breakouts. The study was based on the discovery of twelve fast X-ray transients from a search of the entire XMM archive (Alp & Larsson 2020 [59]). The transients lasted from 30 to 10⁴ s and suggested temperatures in the 0.1–1 keV range based on blackbody fits to the spectra. Shock breakouts provide one possible explanation for the temperature range and duration. Redshifts ranged from 0.13 to 1.17 for nine objects based on the redshifts of the nearest, assumed host, galaxies6.

Sun et al. (2022) [58] then used the data to examine the luminosity function and event rate density. The luminosity function can be fit with a single or broken powerlaw, with somewhat similar slopes but differing local event rate densities (range from 0.03 to ∼50,000 per Gpc⁻³ yr⁻¹ for minimum luminosities in the $5 \times {10}^{42}$ erg s⁻¹ (SN shock breakout) to $7 \times {10}^{49}$ erg s⁻¹ (short GRB). For a significant range in peak flux (∼${10}^{-6}$ to ∼${10}^{-13}$ erg s⁻¹ cm⁻²), the slope of the log N-log S relation is −3/2 as expected for Euclidean geometry.

While these candidate shock breakouts most likely cannot be further investigated, they were detected by XMM because of its larger effective area. Because of their detection, a future X-ray instrument could be built to pursue them past their discovery burst.

5.2. XMM and Chandra: H-Poor Superluminous Supernovae

Margutti et al. (2018) [27] published a systematic survey of the X-ray emission of H-poor superluminous supernovae (SLSNe), objects that emit X-rays above 10⁴³ erg s⁻¹, and recognized as a subgroup only in 2009 (Chomiuk et al., 2011 [60], Quimby et al., 2011 [61]). They examined 26 nearby SLSNe using Chandra and XMM (and Swift) from a few days to ∼2000 days post-explosion. The authors divided the 26 into three sub-groups: a ‘gold’ sample with very good observations; a ‘bronze’ sample with good X-ray but sparse optical or IR coverage, and an ‘iron’ sample with sparse coverage.

From a comparison across the data, the authors argued that the data do not support undetected but extra X-ray emission, and do not support a massive and ‘distant’ CSM. The lack of a massive CSM argues for a central engine that powers the source, as well as significant mass loss during, at most, the final ∼10 years prior to the explosion. That inference then moves the discussion toward compact objects and away from extended envelopes of red supergiants, for example.

Once compact objects are involved, jets or relativistic outflows and magnetic fields take center stage. Margutti et al. (2018) [27] constrained the presence of jets: if jets are present, they are either pointing far off-axis (>30 degrees) or SLSNe host failed jets that do not break through the stellar envelope. The authors also constrained the presence of a magnetic field, arguing that the data require either a very strong field (B $>2 \times {10}^{14}$ G) or a ≈5 times weaker field but a large ejected mass (>7 M_⊙).

6. Discussion, Conclusions, and ‘What’s Next?’

That XMM-Newton and Chandra have made significant contributions to the study of the X-ray emission of supernovae is not in doubt. Both missions have been critical to detecting circumstellar interactions months or years post-explosion. Both instruments are operationally limited in their ability to respond quickly to new events, hence the need for Swift. Both instruments are operationally limited in bandpass, yet are used to study objects where our understanding would be enhanced with a broader bandpass (e.g., 0.2 to 40 or 50 keV).

The answer to this section’s question should be relatively clear to anyone following the evolution of supernovae. We need three things: (i) a much larger ‘Chandra’ that clearly separates the X-ray emission of supernovae from the other X-ray-emitting sources in and around galaxies, so that the community can follow supernovae for a longer time; (ii) a much larger ‘Swift’ to follow the early development of supernovae; and (iii) the use of much higher resolution spectroscopic instruments, e.g., Resolve on XRISM. A spectrometer with the resolution of Resolve (few eV!) or better would permit following emission lines from highly ionized elements. That would address some of the physics problems raised in this review by yielding details of the temperature, density, absorption, and velocities of the prior phases of mass loss.

For a much larger ‘Swift’, SN2008D is an example (Mazzali et al., 2008; [11]): Swift detected an X-ray flash that the authors attributed to the development of a relativistic jet as a black hole formed during the core collapse of the progenitor. However, no subsequent follow-up observations appear to have been carried out.

With a little thought, one might identify a need for a more dedicated X-ray instrument to follow most supernovae7. Such an instrument would be useful; however, the current road block to following bright supernovae with higher cadence is largely the exposure time. Assume we have an instrument with a 10x larger effective area—that immediately leads to a 8–10× decrease in the exposure times needed. And that means it is less ‘expensive’ to follow a supernova with additional observations. The benefit: mapping mass loss with more detail than is possible at present. That would then reveal the degree to which our understanding of massive stars needs improvement.