Draft version November 16, 2021Typeset using LATEX twocolumn style in AASTeX631

Leptonic non-thermal emission from supernova remnants evolving in the circumstellar magnetic field

Iurii Sushch ,1, 2, 3 Robert Brose ,4, 5 Martin Pohl ,3, 5 Pavlo Plotko ,3 and Samata Das3, 5

1Centre for Space Research, North-West University, 2520 Potcheftroom, South Africa2Astronomical Observatory of Ivan Franko National University of Lviv, Kyryla i Methodia 8, 79005 Lviv, Ukraine

3Deutsches Elektronen-Synchrotron DESY, Platanenallee 6, 15738 Zeuthen, Germany4Dublin Institute for Advanced Studies, Astronomy & Astrophysics Section, 31 Fitzwilliam Place, D02 XF86 Dublin 2, Ireland

5University of Potsdam, Institute of Physics and Astronomy, 14476 Potsdam, Germany

ABSTRACT

The very-high-energy (VHE; E > 100 GeV) gamma-ray emission observed from a number of Supernovaremnants (SNRs) indicates particle acceleration to high energies at the shock of the remnants and a potentiallysignificant contribution to Galactic cosmic rays. It is extremely difficult to determine whether protons (throughhadronic interactions and subsequent pion decay) or electrons (through inverse Compton scattering on ambientphoton fields) are responsible for this emission. For a successful diagnostic, a good understanding of the spatialand energy distribution of the underlying particle population is crucial. Most SNRs are created in core-collapseexplosions and expand into the wind bubble of their progenitor stars. This circumstellar medium features acomplex spatial distribution of gas and magnetic field which naturally strongly affects the resulting particlepopulation. In this work, we conduct a detailed study of the spectro-spatial evolution of the electrons acceleratedat the forward shock of core-collapse SNRs and their non-thermal radiation, using the RATPaC code that isdesigned for the time- and spatially dependent treatment of particle acceleration at SNR shocks. We focus onthe impact of the spatially inhomogeneous magnetic field through the efficiency of diffusion and synchrotroncooling. It is demonstrated that the structure of the circumstellar magnetic field can leave strong signatures inthe spectrum and morphology of the resulting non-thermal emission.

Keywords: Supernova remnants (1667) — Stellar wind bubbles (1635) — Red supergiant stars (1375) — Wolf-Rayet stars (1806) — Gamma-rays (637) — Galactic cosmic rays (567)

1. INTRODUCTION

The very-high-energy gamma-ray emission, that has beenobserved from a handful of Supernova remnants (SNRs),can be explained by both electrons (through inverse Comp-ton (IC) scattering on ambient photon fields) and protons(through hadronic interactions with subsequent decay of neu-tral pions) and it is very difficult to discriminate betweenthese two scenarios. Although acceleration of electrons inSNRs can be directly confirmed through the detection of syn-chrotron emission in the X-ray energy band, similar confir-mation for protons would need a careful analysis of the ob-served gamma-ray emission. One of the key signatures ofthe hadronic radiation process is the so-called pion bump be-tween 100 MeV and a few GeV (Stecker 1973). The exact

Corresponding author: Iurii Sushchiurii.sushch@nwu.ac.za

shape of the pion bump and the energy of its peak is depen-dent on the energy distribution of the parent particle popula-tion (Yang et al. 2018) and its elemental composition (Bhattet al. 2020), but in any case at low energies it is marked by theabrupt decrease of the gamma-ray flux that is not expected inthe spectra of inverse Compton scattering. The first reliabledetection of such a low-energy cut-off associated with a pionbump, and hence direct evidence for proton acceleration, wasfound for two old SNRs, IC 443 and W44, which are inter-acting with molecular clouds (Ackermann 2013). Anotherdiscriminating feature is the shape of the gamma-ray spec-trum above the pion bump energies. The hadronic gamma-ray spectrum above the pion bump is expected to reflect thespectrum of the parent particle population with the spectralindex of photons, α, similar to that of the radiating particles,s. In contrast, the spectral index of the IC gamma-ray emis-sion is with α = (s + 1)/2 generally softer than that of theradiating electrons. Recently, a deep Fermi-LAT survey re-vealed a large population of gamma-ray bright SNRs (Acero

arX

iv:2

111.

0694

6v1

[as

tro-

ph.H

E]

12

Nov

202

1

2 Sushch et al.

et al. 2016), a considerable portion of which were identi-fied as dynamically old SNRs and predominantly hadronicemitters through their spectral shape and potential associa-tion with molecular clouds (Jogler & Funk 2016; Ambrogiet al. 2019; de Ona Wilhelmi et al. 2020)

Typically, the particle spectrum (both for electrons andfor protons) follows a power law with spectral index arounds = 2, as predicted by diffusive shock acceleration (DSA) inthe test-particle approximation (Drury 1983), with an expo-nential cut-off at some maximum energy which is imposed bythe age of the remnant, particle escape, or synchrotron cool-ing (relevant only for electrons). Deviations from the univer-sal value of s = 2 may arise from accounting for e.g. cosmic-ray feedback (Malkov & Drury 2001; Amato & Blasi 2006),secondary acceleration processes (Pohl et al. 2015; Wilhelmet al. 2020), fast motion of downstream turbulence (Haggerty& Caprioli 2020; Caprioli et al. 2020), and inefficient particleconfinement (Malkov et al. 2011). With time the spectrumof confined particles considerably softens to s ≈ 2.7 due tothe escape of high-energy CRs and a rapid reduction of themaximally achievable energy (Brose et al. 2020; Celli et al.2019), as well as, in the case of electrons, continuous syn-chrotron losses (Berezhko & Volk 2004; Diesing & Caprioli2019; Brose et al. 2020).

Substantial variation of this picture can arise from the com-plexity of its environment. The majority of SNRs in theGalaxy are produced in core-collapse events (van den Bergh& Tammann 1991), typically either at the red supergiant(RSG) or at the Wolf-Rayet (WR) stage of stellar evolution(Smartt 2009). In such a case the remnant evolves insidea stellar wind bubble of complicated structure. In particu-lar, the complexity of the circumstellar magnetic field cansubstantially modify the electron distribution through energylosses and changes of the particle confinement, leaving char-acteristic imprints in the spectrum and morphology of the ob-served radiation and their temporal evolution. The resultingIC gamma-ray spectrum would strongly depend on the age ofthe remnant and could considerably differ from that for sim-ple DSA scenarios and in fact be similar to that of hadronicemission, which would add an extra dimension to the dis-crimination problem.

We study the impact of the circumstellar magnetic field onthe distribution of electrons accelerated at the forward shockand their non-thermal emission. For that purpose we usethe RATPaC code (Radiation Acceleration Transport ParallelCode) (Telezhinsky et al. 2012a,b, 2013; Sushch et al. 2018;Brose et al. 2019) to simulate the SNR and the accelerationand transport of energetic particles in it.

Earlier studies of cosmic-ray production in core-collapseSNR (Telezhinsky et al. 2012b, 2013) were based on a rel-atively weak magnetic field in the wind zone, and hence in-significant synchrotron losses, and suffered from an overes-

timation of the effects of interactions of the forward shockwith the contact discontinuity or other shocks due to not idealtemporal resolution. More recently, Gaggero et al. (2018)studied the time evolution of the gamma-ray spectrum fromthe Type II SNR (RSG progenitor) with strong simplifica-tions. At each time step accelerated particles were injectedwith a power-law spectrum up to a certain maximum energythat was subsequently evolved accounting for adiabatic andradiative losses. They noticed some spectral features in theleptonic non-thermal emission, but did not discussed themfurther, as the study was focused on the hadronic emissionand the effectiveness/duration of the ”PeVatron” phase of theSNR.

Below we revisit this problem taking advantage of the finetemporal and spatial resolution offered by the RATPaC codeand solve the transport equation for particles simultaneouslywith the hydrodynamic equations.

2. CIRCUMSTELLAR ENVIRONMENT OFCORE-COLLAPSE SNRS

Core-collapse SNRs are results of explosions of massivestars at the end of their evolution. Throughout their life thesestars drive a stellar wind that shapes a circumstellar bub-ble around the star. The hydrodynamical evolution of thesebubbles has been studied both analytically and numerically(e.g. Weaver et al. 1977; Garcia-Segura & Mac Low 1995;Garcia-Segura et al. 1996a,b; Dwarkadas 2005, 2007; Arthur2007). The general bubble created during a single evolution-ary phase of the star can be roughly divided into four zones.Closest to the star is the free wind (FW). At the terminationshock the wind material is heated and feeds the hot shockedwind (SW). The SW zone ends at a contact discontinuity sep-arating the shocked material of the ISM or an earlier windphase. A bit further out one finds a shock in the outer mediumand eventually reaches the unperturbed ISM. The density inthe FW zone falls off ∝ r−2, and in the SW zone it is ap-proximately constant. This general picture is complicated bychanges in the wind properties on account of stellar evolution(Meyer et al. 2020). The last stage of stellar evolution, RSGor WR, is decisive for the shape of the bubble prior to thesupernova explosion.

The RSG wind is much slower than that of the earliermain-sequence phase and does not provide a simple Weaver-type bubble. However, the deceleration of the main-sequencewind and the eventual backflow of the gas still result in theformation of a shock and a thin shell of RSG and main-sequence wind material (Dwarkadas 2007). At the end ofthe RSG phase, the FW zone of the RSG wind terminatesat the thin shell, beyond which there is the shocked main-sequence wind. In our simulations we ignore the thin shelland approximate the hydrodynamic profile by an abrupt dropin density at the transition point from the free RSG wind to

Supernova remnants in circumstellar magnetic fields 3

the shocked main-sequence wind. The mass-loss rate duringthe RSG phase ranges from 10−7 M�/yr to 10−5 M�/yr ormore (Beasor & Davies 2018)). The wind speed is low, 15 to50 km/s (Mauron & Josselin 2011).

For WR progenitors the environment is shaped by a fastwind that sweeps up the preceding RSG wind. The windspeed ranges from 1000 to 4000 km/s, and the mass-loss rateis similar to that of RSGs, 5× 10−6 M�/yr to 5× 10−5 M�/yr(Crowther 2007). At the termination shock, the gas densityincreases toward the SW.

The changes in the density affect the injection rate intoDSA and modify the shock speed, leading to variations inthe maximum energy to which particles can be accelerated.The temperature of the SW is much higher than that of theFW, and so is the speed of sound. The consequences forthe compression ratio and the spectral index of acceleratedparticles (e.g. Das et al. 2021) will be examined in detail inforthcoming studies. In this paper we focus on the impact ofthe magnetic field.

We construct two generic hydrodynamic profiles describ-ing the circumstellar medium for RSG and WR progenitors.Recent numeric simulations and observations suggest that theX-ray and VHE gamma-ray brightness of leptonic emissionshould strongly decrease after 1000 − 4000 years due to arapidly decreasing maximum electron energy and particle es-cape (H. E. S. S. Collaboration et al. 2018; Brose et al. 2020).Limiting our study to young SNRs, we focus on the FW zoneand the transition to the next zone, which basically consti-tutes the SW of all the previous phases. The location of thetransition point to the SW zone is typically around 5 − 15 pc(e.g. Arthur 2007; Dwarkadas 2007) and here firmly set torTP = 5 pc for both cases, to shorten the simulation time.In the FW the density scales as ∝ r−2 with a mass-loss rateMRSG = 10−5 M�/yr and wind speed VRSG = 20 km/s forthe RSG. For the WR star we set MWR = 10−5 M�/yr andVWR = 2000 km/s. The temperature in the FW is set to104 K, while the pressure follows from the ideal gas law. Forthe RSG progenitor, we assume a sharp decrease of the den-sity by a factor of 100 at rTP, and the flow speed falls to zero.Therefore the circumstellar medium of the RSG can be de-scribed by the following

ρ(r) =

M?

4πr2vwind, r ≤ rTP,

ρ(rTP) × 10−2, r > rTP,(1)

v(r) =

VRSG, r ≤ rTP,

0, r > rTP,(2)

p(r) = ρ(r)RT (r) (3)

T (r) = 104K (4)

For the WR progenitor we have a termination shock withthe usual jump conditions for a strong hydrodynamic shock.

The flow speed downstream of the shock decreases as ∝ r−2

to satisfy the continuity of the mass flux. In total, the circum-stellar medium of a WR progenitor is described by

ρ(r) =

M?

4πr2vwind, r ≤ rTP,

4ρ(rTP), r > rTP,(5)

v(r) =

VWR, r ≤ rTP,

0.25VWR(rTP/r)2, r > rTP,(6)

p(r) =

ρ(r)RT (r), r ≤ rTP,

3/4ρ(rTP)V2WR, r > rTP,

(7)

T (r) =

104K, r ≤ rTP,

p(r)/Rρ(r), r > rTP.(8)

2.1. Magnetic field

Recently, van Marle et al. (2015) performed magneto-hydrodynamical simulations of the wind bubble around themassive star, but ignored the magnetic field in the wind ma-terial. Both RSGs and WR stars exhibit strong surface mag-netic fields which should be transported with the stellar wind.Additionally, a turbulent dynamo might convert a sizablefraction of the bulk flow energy to magnetic field (Marcowithet al. 2018, and references therein).

In the free-wind zone the magnetic field is mostly az-imuthal due to the stellar rotation. Its amplitude is well de-scribed by (Voelk & Forman 1982)

Bwind(r) = B∗R∗r2

√V2

rotr2 + R2∗V2

wind

V2rot + V2

wind

, (9)

with B∗ denoting magnetic field strength at the surface of thestar, R∗ and Vrot are the radius and the rotation velocity of thestar, and Vwind is the wind speed. We ignore the very innerregion with predominantly radial magnetic field, because thatis relevant only during the first days and weeks of the SNR.It is still important to have a rough estimate of the Vwind/Vrot

ratio to correctly scale the magnetic field strength.The initial rotation velocities of massive stars are quite

high, reaching few hundreds km/s (Hunter et al. 2008;Maeder & Meynet 2012). At the RSG phase, the largesize and high mass-loss rate makes the stars slow rotatorswith typical velocities, predicted by all stellar models, of afew km/s (Maeder & Meynet 2012, and references therein).These estimates lead to

Bwind(r) = (0.1 − 0.5)B∗R∗r, for r & R∗Vwind/Vrot (10)

for the FW of a RSG.The rotation velocity of WR stars is very uncertain (see e.g.

Crowther 2007). For some stars the co-rotation of large-scale

4 Sushch et al.

structures indicates values ranging from few tens to few hun-dreds km/s (Chene & St-Louis 2008, 2010; Henault-Brunetet al. 2011), as do strong deviation from spherical symmetry(Harries et al. 1998). On that basis we find

Bwind(r) ∼ (0.01 − 0.1)B∗R∗r

(11)

RSGs exhibit relatively weak surface magnetic field of 1−10 G (Tessore et al. 2017), but a very large size of a fewhundred R�. On the contrary, Wolf-Rayet stars are argued tobe strongly magnetic, up to 1000 G, because the final productof their evolution are highly magnetized pulsars or magnetars(Crowther 2007). There is some evidence that magnetic fieldin some parts of the wind can reach 100 G implying a surfacemagnetic field of 1000 G (de la Chevrotiere et al. 2014). WRstars are also quite compact with a radius of only 1 − 10 R�(Crowther 2007).

At the transition to the SW zone we assume a step-likechange of the magnetic field. For the WR progenitor thefield strength increases by a factor of 4 at the terminationshock and will then linearly increase with radius as the windspeed falls ∝ r−2. In the RSG scenario we have a transition tothe main-sequence wind with highly uncertain magnetic-fieldstrength. In line with the assumption of B2 ∝ ρ we assumea decrease by a factor of 10. Beyond the transition point weassume that the magnetic field stays constant as there is noflow.

Denoting as B0 the initial magnetic field close to the star,the circumstellar magnetic field is expressed as

BCM(r) =

B0

R∗r, r ≤ rTP,

0.1B0R∗rTP

, r > rTP

(12)

for the RSG and

BCM(r) =

B0

R∗r, r ≤ rTP,

4B0R∗rTP

rrTP

, r > rTP

(13)

for the WR star. In the following we consider three configu-rations of the adopted values for B0 and R∗ for both scenarios,RSG and WR (see Table 1). HNA and MNA models corre-spond respectively to a high and a moderate magnetic fieldstrength within the range of allowed values discussed above.The AMP model adopts a field strength lying between thoseof the HNA and MNA models but assumes further amplifica-tion of the field upstream of the shock (see Section 3.2).

3. MODELLING OF THE SNR EVOLUTION

3.1. Hydrodynamics

Table 1. Input parameters describing the circumstellar magneticfield

RSG WRParameter HNAa MNAb AMPc HNA MNA AMPB0 [G] 7 1 3 100 10 50R∗ [R�] 1000 500 1000 10 5 10

a High non-amplified circumstellar magnetic fieldb Moderate non-amplified circumstellar magnetic fielda Circumstellar magnetic field amplified upstream with an amplificationfactor k = 5

The standard gasdynamical equations are solved, meaningwe assume that the magnetic field is dynamically unimpor-tant due to its low strength and that the remnant is not in theradiative phase yet (Petruk et al. 2016). The equations aresolved in 1D for spherical symmetry.

We initialize the ejecta profile by a constant density, ρc, upto the radius rc, followed by a power-law distribution up tothe ejecta-radius Rej with index n = 9 for the core-collapseexplosion,

ρ(r) =

ρc, r < rc,

ρc

(rrc

)−n

, rc ≤ r ≤ Rej,(14)

beyond which the density profile is that of the circumstellarmedium described in Section 2. The velocity of the ejecta isdefined as

vej(r) =r

TSN, (15)

where TSN = 1 yr is the initial time set for hydrodynamicsimulations. Then for the assumed ejecta mass, Mej = 3 M�,and energy, Eej = 1051 erg (for both RSG and WR), anddefining the radius of the ejecta as multiple of rc, Rej = xrc,the initial conditions for simulations can be written as

rc =

103

Eej

Mej

(n − 5n − 3

) 1 − 3n x3−n

1 − 5n x5−n

1/2

TSN, (16)

ρc =Mej

4πr3c

3(n − 3)n

(1 −

3n

x3−n)−1

, (17)

vc =rc

TSN. (18)

The spatial grid used in the hydrodynamic simulations ex-tends to r = 30 pc with 262144 linearly spaced grid points.

3.2. Magnetic field

Upstream of the forward shock of the SNR the ambientmagnetic field is assumed to be amplified by a factor k andto exponentially decline on a length scale ∆l = 0.05Rsh, un-til it reaches the field strength in the far-upstream medium.

Supernova remnants in circumstellar magnetic fields 5

Close to the transition point, rTP, at which the ambient fieldabruptly changes, the thickness of the precursor region is re-defined as ∆l = (rTP − Rsh). After the transition point it ischanged back to 5% of the shock radius.

Assuming a shock compression ratio of 4, the immediatedownstream magnetic field is then given by

Bd =√

11kBwind(Rsh) (19)

The downstream magnetic field is passively transported withthe plasma flow, following the induction equation for idealMHD (Telezhinsky et al. 2013). For HNA and MNA modelswe assume that the magnetic field is not amplified upstreamof the shock (k = 1), while for the AMP model the amplifi-cation factor is k = 5. The field amplification factor, k, has aparticular impact on the rate of synchrotron cooling.

Describing the amplification of the magnetic field only bythe parameter k is of course a crude simplification of the mi-crophysics involved. As far as the resonant streaming insta-bility is concerned, the amplification of turbulence shouldsaturate a level of δB ≈ B0 (Zweibel 1979). An amplifica-tion to δB ≈

√vshUcr/c ≥ B0 in the precursor is possible, if

the non-resonant streaming instability is involved (Lucek &Bell 2000; Bell & Lucek 2001). Cristofari et al. (2021) ar-gue that the saturation level ought to be ∝ v3/2

sh for the Bellscenario and by ∝ vsh for the resonant streaming instability.

The particulars of the nonresonant amplification are clearlybeyond the scope of this paper. The nonresonant instabilitysaturates by a back-reaction of the thermal plasma to CR-streaming or a modification of the bulk flow (Riquelme &Spitkovsky 2009; Niemiec et al. 2010; Kobzar et al. 2017).In both cases, the saturation level is independent of the initiallevel of turbulence or the ambient magnetic field. However,the growth of turbulence is limited by the time that is avail-able for the growth of turbulence, since the turbulence needsto be replenished on timescales of D(p)/v2

sh (Brose et al.2016). For Bell modes at the saturation amplitude

√vshUcr/c

throughout the cosmic-ray precursor one would not even geta single exponential growth cycle (Pohl 2021). This was re-cently shown to be the limiting factor in very young SNRsin dense environments, where the conditions for particle ac-celeration and turbulence growth were deemed ideal (Inoueet al. 2021). Additionally, enhanced damping might suppressfurther turbulence growth when the level of δB/B0 ≈ 1 is sur-passed (Brose et al. 2021). In both cases the amplified fieldcan be approximated by a scaling of the existing, ambientfield.

Behind the shock, the amplified magnetic field may bequickly damped (Pohl et al. 2005), and so the field wouldfall back to the large-scale structure that we describe here.

In Figure 1 we show magnetic-field profiles for the AMPmodel at different SNR ages for the RSG (top panel) and WR(bottom panel) scenarios. Solid lines show the distribution of

the magnetic field at a very early stage of the SNR evolution,dashed lines depict the moment shortly before the transitionto the SW and dashed-dotted lines correspond to the evolu-tion in the SW zone. Gray lines illustrate for reference theprofile for a constant ambient magnetic field of 5 µG. Themagnetic field is generally stronger for a RSG progenitorthan for a WR, and we should expect stronger synchrotroncooling in the SNRs of RSG explosions. Even after the tran-sition to the weaker field in the SW, about 1 µG, the down-stream magnetic field remains at the level of several hundredsµG, as is observed in Cas A (Cowsik & Sarkar 1980; Abey-sekara et al. 2020).

As long as the SNR evolves inside the FW, we see a simi-lar radial variation in the magnetic-field profiles for the RSGand WR scenarios. The circumstellar magnetic field is com-pressed at the shock and further downstream gradually in-creases in strength to peak at the contact discontinuity. Up-stream of the reverse shock the magnetic field is very weak.

When the SNR shock moves into the SW, we see differ-ences between the RSG and WR models, because in the caseof a RSG the circumstellar magnetic field becomes weaker,whereas it turns stronger for a WR progenitor. Moreover, thepost-transition phase of the RSG case model is characterizedby an additional hump closely behind the shock (about 0.5 pcinward after 1000 years, see top panel of Fig. 1). This localboost of the magnetic field is caused by compression at a newcontact discontinuity (CD) between the accelerated forwardshock and a reflected shock. This second CD also coincideswith a pile-up of particles, and both effects combined leadto additional ring structure in the synchrotron morphology ofthe source, that we shall further discuss in Section 4.4.

3.3. Particle acceleration

We simulate the evolution of the differential particle den-sity, N, by solving the transport equation in the form

∂N∂t

= ∇(D∇N − ~vN) −∂

∂p

((N p) −

∇~v3

N p)

+ Q, (20)

where D is the spatial diffusion coefficient,~v is the plasma ve-locity, p represents the energy losses (in our case synchrotronlosses), and Q is the source term.

The diffusion is assumed to scale with the Bohm diffusioncoefficient

D(p) = ηBpc2

3eB, (21)

where ηB ≥ 1 is the Bohm factor and a measure of accel-eration efficiency. A recent systematical study of the accel-eration efficiency suggested a possible time evolution of thediffusion scale factor from ηB ' 10 at several hundred yearsto ηB ≈ 1 after several thousands years (Tsuji et al. 2021).The implied extremely high acceleration efficiency obtainedfor older SNRs is somewhat dubious and fully relies on the

6 Sushch et al.

Figure 1. Radial distribution of the magnetic field both downstreamand upstream of the forward shock of the SNR at different ages,plotted in red. The top panel applies to a RSG progenitor, and thebottom panel is for a WR star. In both cases an amplified modelis assumed and the transition between the FW and the SW zonesis located at r = 5 pc. These profiles are compared to a simplemodel of a constant 5 -µG upstream magnetic field, depicted withgray lines in both panels.

assumption that the maximum energy of electrons is limitedby synchrotron cooling and the cut-off energy of the X-rayspectrum is independent of the magnetic-field strength (Zi-rakashvili & Aharonian 2007). Tsuji et al. (2021) motivatethis assumption by examining the thin filament-like outer X-ray rims which are supposedly cooling limited. It was, how-ever, shown already on several occasions that the thicknessand brightness of X-ray filaments can be equally well ex-plained by the damping of the magnetic field downstream ofthe shock, i.e. spatial distribution of the magnetic field, ratherthan its high strength (Pohl et al. 2005; Rettig & Pohl 2012;Sushch et al. 2018). The inaccuracy of the assumption forolder SNRs is also supported by estimates of ηB < 1 for twoSNRs, namely Vela Jr. and HESS J1731− 347, which shouldbe impossible. Therefore, in this study we adopt the valueηB = 10 for the entire evolution of the SNR up to an age of2000 years.

We use the thermal leakage injection model (Blasi et al.2005),

Q = χnu(Vsh − Vwind)δ(R − Rsh)δ(p − pinj), (22)

where χ is the injection efficiency parameter, nu is the plasmanumber density in the upstream region, Vsh is the shockspeed, Vwind is the wind speed upstream of the shock, Rsh isthe shock radius, and pinj = ξpth is the injection momentum,defined as a multiple of the thermal momentum in the down-stream plasma. The injection efficiency for the compressionratio of 4 is determined as

χ =4√π

ξ3

eξ2 . (23)

Figure 2. Simulated electron spectra for RSG (top) and WR star(bottom) progenitors at different ages of the SNR. Solid coloredlines correspond to different models of the magnetic field in the free-wind zone. The dash-dotted line represents the electron spectrumfor a constant ambient magnetic field of 5 µG.

We solve the transport equation for electrons in sphericalsymmetry using the RATPaC code as described in Telezhin-sky et al. (2012a, 2013); Sushch et al. (2018); Brose et al.(2019), taking into account only a forward shock and ig-noring a reverse shock. Resulting electron spectra at differ-ent moments of time are then used to simulate synchrotronand inverse Compton (scattering on CMB photons) emissionfrom the remnant. In this study we do not consider accel-eration of protons and also assume that cosmic-rays will notdynamically affect the SNR evolution.

4. RESULTS

4.1. Evolution in the free wind

First, we simulate the spectra of electrons and the non-thermal emission at the early stage of the SNR evolution,while it still expands in the free wind. Figure 2 shows theevolution of the spectrum of accelerated electrons for the fourmodels of the ambient magnetic field for RSG (top panel) andWR (bottom) progenitors. The differential density of elec-trons is scaled with R2

shE2 for illustration purposes. For ref-erence, the gray dash-dotted line shows the evolution of theelectron spectrum for the constant ambient magnetic field of5 µG.

For the RSG progenitor, synchrotron cooling can be sub-stantial even if the magnetic field is not amplified upstreamof the shock. The blue line depicts the situation for a strong,but un-amplified circumstellar magnetic field. The electronspectrum features a cooling break at a few tens of GeV af-ter 100 years, that eventually moves to higher energies as thefield strength decreases. Already for the moderate magnetic

Supernova remnants in circumstellar magnetic fields 7

Figure 3. Simulated emission spectra for inverse Compton scat-tering of the CMB, corresponding to the electron spectra shown inFig. 2. The style and labeling is preserved.

Figure 4. Simulated synchrotron spectra corresponding to the elec-tron spectra shown in Fig. 2. The style and labeling is preserved.

field the cooling is inefficient and does not significantly mod-ify the spectrum (green line), but the r−1 distribution of themagnetic field in the free-wind zone can still leave imprintsin the time evolution of the maximum electron energy. For aconstant ambient magnetic field, the maximum energy is agelimited, hence it evolves as Eage

max ∝ V2shtB, where Vsh is the

shock velocity (see e.g. Reynolds 2008). At the free expan-sion stage the shock radius evolves as Rsh ∝ t(n−3)/(n−m) = t6/7

(Truelove & McKee 1999), where the index of the ejecta den-sity profile is n = 9 and the index of the density profile in thewind zone is m = 2. The shock speed declines as Vsh ∝ t−1/7,

yielding Eagemax ∝ t5/7 for the constant magnetic field, whereas

for the realistic scaling, B ∝ r−1, we expect Eagemax ∝ t−1/7.

For the strong circumstellar magnetic field encounteredduring the early stages of SNR evolution, the maximumenergy is limited by synchrotron cooling. The balance ofcooling and acceleration yields Esyn

max ∝ VshB−1/2 (see e.g.Reynolds 2008), and hence the maximum energy only slowlygrows with time, as Esyn

max ∝ t2/7.For the WR progenitor star (Fig. 2, bottom panel), syn-

chrotron cooling is much less efficient even for a very strongmagnetic field. Non-amplified field models result in feature-less spectra roughly similar to those for a constant, weak am-bient field. The amplified field model still shows a coolingbreak at a few hundred GeV early in the evolution, but al-ready 300 years after the Supernova the electrons have a fea-tureless uncooled spectrum. The cut-off energy for the cir-cumstellar magnetic field does not significantly increase withtime, unlike that for the constant ambient field, indicating thatit is still limited by synchrotron cooling.

Figure 3 shows the inverse Compton emission generatedby CMB radiation scattering off the electrons. The gamma-ray spectra of SNRs resulting from the explosion of RSGscan feature a low-energy break and display spectral soft-ening above the break. The spectral break appears belowthe GeV scale at early stages and gradually shifts towardshigher energies, as the ambient magnetic field decreases.The spectral shape visually resembles the one expected forhadronic gamma-ray emission, introducing new difficultiesin discrimination between the leptonic and hadronic scenar-ios, but it is much smoother then the hard transition in thepion-decay spectra, emphasizing the need for reliable databelow 500 GeV.

For the WR progenitor (Fig. 3, bottom panel), we find onlya weak spectral modification, and that only for the amplifiedfield. By and large, there is no significant change of the spec-tral shape as compared to the constant low ambient magneticfield.

Similarly, the simulated synchrotron spectra (Fig. 4) reveala characteristic cooling break for the strong-field configura-tions in the RSG scenario, but nothing of that kind in the WRscenario. The flux normalization for circumstellar field mod-els decreases with time, unlike for the constant ambient field,which can be directly attributed to the decrease of the mag-netic field itself. This would naturally result in the decreaseof the expected radio and X-ray emission, which is furtherdiscussed in Section 4.3.

4.2. Transition to the shocked wind medium

The transition to the SW zone is accompanied by an abruptchange of the physical properties of the medium. To investi-gate the impact of the change of magnetic field on the result-

8 Sushch et al.

Figure 5. Volume-averaged electron spectrum normalized with R3sh

(top panel) and the corresponding gamma-ray (middle panel) andsynchrotron emission spectra (bottom panel) for a RSG progenitor.The spectra are color-coded for SNR age with increments of 100years. We highlight with a dash-dotted line the spectrum close tothe transition from the free RSG wind to the main-sequence wind,when the age is around 800 years.

ing particle and gamma-ray spectrum, we use AMP modelsdescribed in Section 2.

The evolution of the electron, gamma-ray and synchrotronspectra for a RSG progenitor is shown in Fig. 5. Before thetransition to the SW after nearly 800 years, the evolution ofthe electron spectrum is governed by synchrotron cooling and

Figure 6. Same as Fig. 5 but for the WR progenitor. The dash-dotted line gives the spectra close to the transition to the SW afteraround 400 years.

the linear decrease of the magnetic field strength. As theremnant expands, the cooling break shifts to higher energies,while the maximum energy of electrons slowly increases (seeprevious Section). After 800 years, the dash-dotted spec-trum shows a significant shift of the cut-off to lower energieswhich reflects the entrance of the cosmic-ray precursor intothe weak-field region beyond the transition point. The dif-fusion coefficient in the precursor increases, and high-energyparticles can effectively escape. As soon as the shock en-

Supernova remnants in circumstellar magnetic fields 9

ters the dilute shocked medium, the shock velocity increasesconsiderably, and particles are accelerated to high energiesagain, despite the weak magnetic field. Already after 900years a second very-high-energy component emerges in theelectron spectrum, which becomes more significant with timeand eventually starts to dominate (Fig. 5 top panel). In theIC spectrum, a hint of this component could be seen al-ready after 800 years (middle panel). Whereas some of thehigh-energy particles close to the shock can receive a fur-ther boost in energy, low-energy particles get advected down-stream, escape the acceleration process, and are subjected tosynchrotron cooling. This explains the softening of the spec-trum above the cooling break energy at a few hundred GeVshortly after the transition. Later on, the spectrum in this en-ergy range gets harder again due to the rise of the secondcomponent. Reaccelerated particles together with freshly ac-celerated particles constitute this very-high-energy compo-nent of the spectrum. Moreover, cooling becomes ineffectiveclose to the shock due to the much lower magnetic field inthe shocked medium. The resulting concave spectrum can bedescribed as a superposition of spectra of two electron popu-lations: old cooled electrons with a spectral break at around100 GeV and new uncooled electrons with no cooling breakand higher maximum energy. With time, further cooling ofthe first population shifts the break energy to lower energies,and the younger particles become more dominant due to theirincreasing number. This gradually changes the concavity ofthe spectrum until at some point in time the spectrum be-comes featureless again.

The concavity of the electron spectrum is reflected in theradiation spectra (Fig. 5, middle panel for inverse Comptonscattering and bottom panel for synchrotron), strongly chang-ing their shape in the GeV band and the keV band. The ap-pearance of the new high-energy component in the electronspectrum causes a strong temporal variation of the spectralshape in the GeV-to-TeV band. Right after the transition tothe shocked main-sequence wind the very-high-energy spec-trum is soft, but gradually hardens, ending up with a typicalfeatureless IC spectrum when the new high-energy compo-nent takes over completely. This evolution takes around 700years for our setup, suggesting that the different phases ofspectral evolution could potentially be detected. Althoughthe main reason for the concavity of the spectrum is theabrupt change of the magnetic field, the change in densitywould regulate the timescale on which the gamma-ray spec-trum settles to the typical form. The higher the amplitudeof the density drop between the RSG and the main-sequencewind, the slower is the evolution of the spectrum, due to thelower injection rate of particles at the shock.

Similar variations are seen in the synchrotron spectrum, themain difference lying in the duration of the process. Sincethe synchrotron radiation directly depends on the magnetic-

field strength, the new high-energy component takes longerto emerge in the synchrotron spectrum than in the inverseCompton spectrum. Additionally, the normalization of thespectrum decreases as the magnetic-field strength drops.

The situation with the WR progenitor is quite the oppositeand more complicated, but eventually leads to a similar out-come, i.e. a concave spectrum (Fig. 6). The magnetic fieldgenerated by the WR progenitor is weaker than that gener-ated by the RSG progenitor, implying that synchrotron cool-ing is less efficient. Indeed, only at very early times a hint ofa cooling break can be seen, and it disappears quickly. Nev-ertheless, the maximum energy of electrons is still cooling-limited, implying a slow rise of the maximum energy (see theprevious section).

As the forward shock approaches the transition point afternearly 400 years, as in the RSG case the cosmic-ray precur-sor extends to the SW and the stronger magnetic field there,leading to a better confinement. The forward shock instan-taneously decelerates by a factor of 1.5, when it enters thedenser SW, allowing particles of the highest energy to re-turn to the shock and re-enter the acceleration process. Aprominent effect is the sharpened cut-off that is seen after500 years. As the remnant propagates through the SW withthe linearly increasing magnetic field, synchrotron coolingbecomes more efficient and imposes a cooling break and aspectral softening at high energies. Particles located suffi-ciently far downstream still see the lower magnetic field andhence weak cooling (see Fig. 1), but they also have a lowermaximum energy. This again creates a two-componentparticle spectrum and concave inverse Compton and syn-chrotron spectra. With time the strong magnetic field encoun-tered in the SW advects to the far downstream and renderssynchrotron cooling relevant also for the low energy com-ponent. Like for the RSG progenitor, eventually the high-energy component becomes dominant, allowing for consid-erable evolution of the spectral hardness in the GeV-TeV andeV-keV bands. This spectral modification is slower than fora RSG progenitor, opening an even larger window of oppor-tunity for observing it. As expected, the maximum energyof accelerated electrons slowly decreases with time, as theshock slows down and the magnetic field increases linearly.

The interaction with the termination shock can briefly in-crease the compression ratio felt by particles and hence in-duce a spectral bump (Telezhinsky et al. 2013), but the du-ration of this event is too small to give a significant impact(Das et al. 2021), and therefore it is ignored here. More im-portantly, such a shock collision would generate a reflectedshock propagating in a medium of lower density that mighteventually be reflected off the contact discontinuity of theSNR and collide with the forward shock. This process mightrepeat several times. The corresponding boost of the shockspeed results in an instant increase of the maximum elec-

10 Sushch et al.

tron energy, modifying the spectrum accordingly. This wouldresult e.g. in brightening and hardening of the X-ray andTeV emission on very short time scales. The recent detec-tion of rapid variations in X-ray filaments of the Tycho’sSNR (Okuno et al. 2020) may be attributed to the boostingof the forward shock by a trailing shock. A second possi-ble explanation is dynamic magnetic turbulence (Bykov et al.2008). After the boosting event the shock decelerates again,and the maximum energy decreases until the next interactionhappens. The reflected shocks bouncing back and forth be-tween the forward shock and the CD would compress andre-accelerate particles, which can further modify the spec-trum and the source morphology. For the WR progenitor thereflected shock is responsible for somewhat irregular varia-tions of the electron and radiation spectra that are superposedon the main trend determined by the change of the magneticfield.

4.3. Time evolution of the non-thermal emission

In Fig. 7 we show the simulated flux in X-rays, high-energy gamma-rays, and very-high-energy gamma-rays forour AMP models and RSG and WR progenitors. At the earlystages of evolution (in the FW zone) both scenarios, RSG andWR, feature a power-law decrease of the X-ray flux, albeitwith different slope. For the RSG progenitor, the dimming ofthe X-ray emission is determined exclusively by the 1/r scal-ing of the magnetic field strength. The decreasing upstreamdensity, and likewise the injection rate, is fully compensatedby the increasing shock surface, because the synchrotroncut-off does not fall into the energy range in question. Forthe WR progenitor the expected X-ray flux is sensitive tothe change of the cut-off energy of the radiation spectrum,which depends on the maximum electron energy and mag-netic field as Eph

cut ∝ E2maxB (Reynolds 2008; Vink 2012). For

the synchrotron-limited particle acceleration in the FW zonethe cut-off energy evolves as Eph

cut ∝ t−2/7. A consequenceis faster dimming of the X-ray emission than in the RSGcase. When the shock leaves the FW zone, the X-ray fluxabruptly decreases in the case of the RSG and increases forthe WR progenitor, on account of the jump of the magneticfield which besides the synchrotron emissivity also changesthe confinement of high energy particles at the shock. Inthe WR scenario the flux keeps increasing for a few hundredyears, which is partly due to re-acceleration of old particles tohigher energy and partly due to two episodes of shock-shockinteraction. These episodes happen at around 560 years and690 years and are clearly reflected in the VHE light curve(Fig. 7 bottom panel). Later, the forward shock decelerates,causing a gradual decline of the X-ray flux enriched with spo-radic brightenings caused by numerous shock-shock interac-tions.

Unlike synchrotron radiation, the gamma-ray flux gener-ated by inverse Compton scattering simply reflects the num-ber of electrons and does not directly depend on the magneticfield strength. As long as the SNR expands into the FW, theHE and the VHE flux grow, because more and more parti-cles get injected into the acceleration process. Like the X-ray flux, the gamma-ray flux abruptly drops at the transitionto the main-sequence wind in the RSG scenario. This declineis mainly caused by synchrotron cooling of electrons acceler-ated in the FW zone that cannot be replaced on account of theabruptly decreased injection. In line with the energy depen-dence of the cooling time, the TeV-band flux recovers muchfaster than the GeV-scale flux. Later on the flux increases inboth energy bands, reflecting the continuously growing num-ber of accelerated particles.

For the WR progenitor we observe the opposite. At thetransition to the SW, the flux in the entire gamma-ray band isboosted by the higher injection of particles and keeps grow-ing until shock deceleration and synchrotron cooling intro-duce a slow decline, which, naturally, happens earlier in theVHE range than in the HE range. The VHE light curvealso features several bumps, which reflect episodes of shock-shock interaction. The boost of the forward shock instanta-neously increases the maximum energy of accelerated par-ticles and hence the VHE flux. The counterparts of thesebumps in the GeV light curve are much less prominent, if atall visible.

A declining flux of non-thermal X-rays was observed fromCas A (Patnaude et al. 2011) a remnant of the Type-IIb ex-plosion of a RSG (Krause et al. 2008). The 4.2 − 6 keV fluxfrom the whole remnant decreased at a rate of 1.5%/yr over11 years which the authors attributed to a declining maxi-mum energy due to shock deceleration. They argued that thedecrease of the magnetic field alone would not be sufficientto explain such a decline (Patnaude et al. 2011). The requireddeceleration rate of the forward shock considerably exceedsthe observed deceleration with the expansion parameter ofq ' 0.65 for Rsh ∝ tq (Patnaude & Fesen 2009). Note thatthis expansion parameter is compatible both with the Sedov-Taylor stage of an SNR in the FW, q = 2/(5 −m) = 2/3 withm = 2 for a ρ ∝ r−m density profile, and with the free expan-sion of a core-collapse SNR in a medium of constant density,q = (n − 3)/(n − m) = 2/3 for n = 9 and m = 0. The latterreflects exactly the situation that we expect for the RSG sce-nario after the transition to the shocked-main-sequence wind.The expansion of the remnant is still ejecta-dominated as itexpands in a very dilute medium, and the density of the SWis roughly constant. Interestingly, the average dimming rateof the X-ray emission near the transition time in our simu-lations, between 770 and 840 years, is 1.4%/yr, compatiblewith that observed for Cas A. In our simulation the dimmingarises from the escape of the high-energy particles upstream

Supernova remnants in circumstellar magnetic fields 11

Figure 7. Simulated light curves for X-rays at 1 − 10 keV (top),the 1− 100 GeV band (middle), and above a TeV. Blue filled circlesapply to the RSG progenitor, while the orange circles are for theWR scenario.

of the shock, where the magnetic field abruptly decreases.For about 10 years during the transition phase the dimmingcan reach 10%/yr, but as soon as the forward shock enters theSW, the dimming normalizes to 1−1.5%/yr. At this time, theshock expands in a constant-density environment and thusdecelerates again. Note that deviations from spherical sym-metry can smear out this effect over time. It is plausible thatCas A at its age around 350 years recently transitioned fromthe FW to the SW of its progenitor stellar wind bubble, andthe decline of the X-ray emission that we observe reflects thistransition.

The rate of change in the gamma-ray flux is smaller butcan still reach a few %/yr near the transition time. Sucha flux change is potentially detectable with current instru-mentation if we observe the source at the right moment. For

WR progenitors, sporadic brightenings arising from shock-shock interactions are also strong enough to be detected. Anidentification as shock interaction events is likely impossiblethough, because their duration is longer than the time periodavailable for observations. The general brightening in theTeV band in the RSG case, ∼ 0.25%/yr beyond the transi-tion point, and likewise the dimming for the WR scenario,∼ −0.15%/yr, are too slow to be currently detectable.

4.4. Morphology

The two spectral components of electrons, that we sawwhen the SNR shock left the FW, are not only spectrallydistinct but also spatially separated. The pre-transition pop-ulation of electrons, that were accelerated when the shockpropagated through the FW zone, are located further insidethe remnant, whereas the freshly accelerated, more energeticelectrons are found close to the shock. In addition, the speedof the forward shock is modified, changing the separationbetween the forward shock and the CD, which would like-wise be reflected in the morphology. This should be partic-ularly prominent in the radio emission, as low-energy elec-trons preferentially reside close to the CD, and the magneticfield piles up there (Rosenau & Frankenthal 1976).

In Figs. 8 (RSG) and 9 (WR) we show simulated intensitymaps for different ages and different photon energies. Thetop left panels depict the morphology when the SNR shockpropagates through the FW zone. Both X-ray and TeV-scaleemission are bound to the location of the shock where thehigh-energy particles reside. To be noted from the figuresis that the radio shell is located farther at the CD where thelow-energy particles pile up and the magnetic field is high.Time-variable X-ray filaments have been observed in the in-terior of Cas A (Uchiyama & Aharonian 2008) and in a num-ber of other SNRs. They may be caused by the strong field atthe CD, if the site of electron acceleration is very close (Lyu-tikov & Pohl 2004). Our simulation assumes that most of theacceleration happens at the forward shock, and so we do notsee X-ray features associated with the CD.

In the WR scenario the rim of 1-GeV emission is roughlyco-located with the radio shell, reflecting the high densityof low-energy electrons there. In the RSG case, however,the GeV shell is closer to the forward shock, on accountof the strong synchrotron cooling of the TeV-scale electronsthat produce the GeV-scale IC emission, whereas the radio-emitting GeV-band electrons remain unaffected.

In the RSG case, once the remnant transitions to theshocked medium, the forward shock speeds up. The sepa-ration between the shock and the CD increases, and so doesthat between the inner radio ring and the outer X-ray shell.The acceleration of the forward shock also results in the off-set of the 1 GeV ring towards the interior of the remnant,because the production rate of TeV electrons has collapsed.

12 Sushch et al.

Figure 8. Simulated intensity maps for a RSG progenitor at the ages500 years (top left), 900 years (top right), 1500 years (bottom left),and 2000 years (bottom right). Each sectors displays the projectedbrightness at a specific energy, normalized to that at the center. Theleft hemisphere exhibits synchrotron radiation at 1 GHz (top), 1 keV(middle), and 10 keV (bottom), and the right hemisphere shows ICemission at 1 GeV (top), 100 GeV (middle), and 1 TeV (bottom).

Right after the transition, after 900 years (upper right mapof Fig. 8), the highest X-ray and very-high-energy gamma-ray intensity is still seen close to the shock. At later timesa second CD is formed, where particles are piled up and themagnetic field is compressed (Section 3.2). After 1500 years,synchrotron cooling is much less effective, and the GeV-TeVmorphology is similar to that seen during the early phase inthe WR case, i.e. a shell location gradually shifting from theCD for lower energies to the forward shock for higher ener-gies (bottom left map on Fig. 8). The formation of the secondCD also results in the appearance of a second radio ring thatis prominent after 2000 years (bottom right map on Fig. 8).Initially, high X-ray intensity coincides with this second CDon account of the higher magnetic field, but later, after 2000years, a second X-ray shell emerges that is caused by freshlyaccelerated particles at the shock.

For the WR progenitor, the transition to the shockedmedium immediately results in a double-ring structure in theradio and soft X-ray morphology (top left map of Fig. 9).These two rings reflect the locations with enhanced magneticfield: the CD and the abruptly increased field at the shock.Later on, as the magnetic field and particles are advecteddownstream, the synchrotron morphology is mainly deter-mined by the spatial distribution of electrons downstream.Note, that although the radio emission is generated farther inthe interior of the remnant, it no longer coincides with theCD, because we do not find the highest electron density andstrongest magnetic field there. At later stages of evolution

Figure 9. Same as Fig.8, but for the WR progenitor and the ages300 years (top left), 500 years (top right), 1000 years (bottom left),and 2000 years (bottom right).

(bottom maps on Fig. 9), a second ring of GeV-band emis-sion is formed by the electrons near the CD that graduallycool with time. After 1000 years (bottom left map on Fig. 9)the inner ring appears in the 100-GeV map, and after 2000 itis most prominent at 1 GeV.

X-ray and IC very-high-energy gamma-ray emission areproduced by electrons of roughly the same energy. It is there-fore tempting to assume that both emissions should be pro-duced in the same regions and with proportional emissivity,hence the gamma-ray morphology would strongly correlatewith the X-ray morphology. Fukui et al. (2021) made this as-sumption to discriminate between the leptonic and hadronicscenario of the gamma-ray emission, likewise assuming thathadronic gamma-ray radiation should correlate with the gasdistribution. The latter assumption is dubious because wemeasure column densities and not the gas density itself, thatdetermines the hadronic emissivity. Besides, the intensitymaps in Figs. 8 and 9 show clearly that the X-ray and theleptonic gamma-ray morphology do not necessarily corre-late, as the former is also strongly dependent on the localmagnetic field. To illustrate this in more detail we calculateprojected radial profiles of the gamma-ray to X-ray flux ratio(Figs. 10 and 11). Following Fukui et al. (2021), we cal-culate the gamma-ray flux in two energy bands: above 250GeV (solid lines) and above 1 TeV (dashed lines). The X-rayflux is calculated in the energy range of 1 − 5 keV. For bothtypes of SNR we do not see a constant flux ratio. Instead, aclear anti-correlation of the gamma-ray and X-ray morphol-ogy is visible for certain ages of the SNR at length scales of5 − 15% of the shock radius. For larger SNRs, this could bewell resolved even by current generation instruments.

Supernova remnants in circumstellar magnetic fields 13

Figure 10. Intensity ratio of gamma rays and X rays for the RSGscenario, plotted as a function of the projected distance from thecenter and normalized to unity at the center. The X-ray brightnessis integrated between 1 and 5 keV, whereas for the gamma rays weuse two energy bands: above 250 GeV (solid lines) and above 1 TeV(dashed lines). The line color indicates the age of the remnant.

Figure 11. Same as Fig.10, but for the WR progenitor.

4.5. Spatial variation of the GeV spectrum

As already emphasized in the previous subsection, twospatially separated populations of electrons are seen afterthe transition of the forward shock to the shocked medium.We calculated spectral index maps at three different energies,1 GeV, 10 GeV, and 100 GeV, to search for correspondingspatial variations of the spectral index, which may be rel-evant for the discrimination between leptonic and hadronicscenarios. We chose the same ages as for the emission mapsin Fig. 12 for the RSG case and Fig. 13 for the WR progeni-tor. The index plotted is the photon spectral index, α, definedas

Nγ(E) =dNdE∝ E−α, (24)

Figure 12. Photon spectral index maps for 1 GeV (top sector), 10GeV (middle sector), and 100 GeV (bottom sector) calculated forthe RSG scenario at the same ages as in Fig. 8.

Figure 13. Same as Fig. 12, but for the WR case and the ages as inFig. 9.

where Nγ(E) is the differential photon flux. At each locationthe index for particular energy E is calculated as

α(E) = logNγ(E − dE)Nγ(E + dE)

/log

E + dEE − dE

(25)

During the pre-transition stage of the RSG scenario, theGeV-band emission spectrum is flat in E2Nγ(E) representa-tion, characteristic for a cooled electron spectrum (left hemi-sphere of the left map on Fig. 12). The spectral index isroughly constant except at 1 GeV, where we see a signatureof the cooling break. Indeed, electrons close to the shockhave not had the time to cool at the energies relevant for the1-GeV emission. The formation of two distinct electron pop-ulations can be very well seen after 900 years, in the righthemisphere of the left map of Fig. 12, particularly at 1 and10 GeV. The ”old” cooled particles are far behind the accel-erated forward shock, and the ”new” particles close to it ex-

14 Sushch et al.

hibit a much harder spectrum. The difference in spectral in-dex at 10 GeV reaches unity over a distance of 0.2Rsh, whichmay well be detectable with, e.g., Fermi-LAT, although spa-tial resolution can be an issue. This would not be a problemfor CTA measuring at 100 GeV, but there the spectral index isagain spatially constant, suggesting that we already observe”new” particles injected after the transition. Later on, in theright pie of Fig. 12, one can see the ”new” population of par-ticles gradually taking over. The spatial variations disappearexcept for a slight (not-detectable) change in the ring relatedto the second CD.

For the WR case (Fig. 13), the GeV spectrum is not imme-diately affected by the transition. Right before the transition,after 300 years, and right after it, at 500 years, the spectralindex is roughly constant in the GeV band. A spatial varia-tion of the spectral index emerges when the ”old” low-energycomponent is sufficiently cooled for the cooling signature toappear at photon energies around a GeV. After 1000 years(left hemisphere of the right map on Fig. 13), the 10-GeVspectral map exhibits a very similar variation as in the im-mediate post-transition phase of the RSG model. The 1 GeVmap, at the same time, still indicates a featureless uncooledspectrum. The strongest evidence of the formation of twocomponents can be seen at 100 GeV, where the spectral in-dex jumps by ∆α ' 0.3 near 0.85Rsh. After 2000 years thesituation changes again: the particles have had time to cool,and the characteristic signature shifts to lower energies. The1 GeV map at 2000 years resembles that at 10 GeV and 1000years, and likewise for the 10-GeV map and the 100-GeVmap. After 2000 years the 100-GeV map is already domi-nated by the ”new” component.

5. SUMMARY

The remnants of core-collapse Supernovae expand into thecircumstellar medium shaped by the wind of their progenitorstars. The gas and magnetic-field distributions in the windbubbles are complex, which strongly affects not only the evo-lution of the SNR but also the non-thermal emission pro-duced by relativistic particles accelerated at the SNR shock.We studied the impact of the circumstellar medium and inparticular the circumstellar magnetic field on the non-thermalemission of relativistic electrons using the RATPaC software,that is designed for the time- and spatially dependent simula-tions of particle acceleration at SNR shocks. We separatelysimulate the SNRs of RSG and of WR progenitors.

Our simulations demonstrate that during the early evolu-tion of the SNR the electron spectra, and hence those of thenon-thermal radiation, might be effectively modified by se-vere synchrotron losses in the strong magnetic field in theinner wind zone. Synchrotron cooling becomes less efficientwith time as the magnetic field declines as 1/r. Later on, thetransition from the FW to the SW, with the associated abruptchanges in the magnetic-field strength and the gas density,

may strongly modify the electron spectrum, which is sub-sequently reflected in the synchrotron and inverse Comptonspectra. In particular, we see the formation of two compo-nents of particles yielding a concave spectrum, one accel-erated in the free wind and the other one produced in theSW. The concavity of the spectrum changes with time, andso does the spectral hardness in the GeV-TeV range. At sometimes the simulated gamma-ray spectrum produced by elec-trons resemble that expected from hadronic interactions.

The two particle populations formed because of the transi-tion to the shocked medium are also spatially separated. Thiscan result in a complex and energy-dependent morphology ofthe SNR featuring, e.g., multiple bright rings of non-thermalemission. An associated spatial variation of the spectral in-dex can potentially be detected, subject to the angular reso-lution of the instrument. Additionally, the morphology of ICgamma rays and that of synchrotron X rays do not correlate,as the latter is strongly dependent on the distribution of themagnetic field. It is therefore premature to interpret the ab-sence of such a correlation as an indication for the hadronicorigin of the gamma rays.

Variations of the magnetic field influence high-energy elec-trons through the cooling rate and through the efficiency ofconfinement at the shock. Additionally, abrupt changes ofthe gas density, e.g. at the wind-termination shock, lead toacceleration or deceleration of the forward shock, depend-ing on the progenitor type, which again has an impact onthe maximum attainable energy. The interaction of the shockwith the termination shock provides a reflected shock that canreflect off the CD to later catch up with the forward shockagain, triggering an instantaneous acceleration of the latter.Events of this type lead to short episodes of strong bright-ening and/or dimmings in the X-ray and gamma-ray lightcurves. Our simulations suggest that such events may be de-tectable even with current generation instruments.

We used simplified flow profiles for the wind bubbles andin particular concentrate on the first 2000 years of SNR evo-lution. We anticipate that similar, but potentially more com-plex effects can be observed for older SNRs and with refinedmodelling of the wind bubble (e.g. Meyer et al. 2020). Wealso emphasize that the actual time scales at which the effectsdiscussed in this work take place strongly depend on the lo-cation of the transition point which we assumed to be ratherclose to the centre of the star.

The simulations conducted in this study clearly demon-strate that the properties of the accelerated electrons areshaped by the spatial distribution of the ambient medium andparticularly by that of the ambient magnetic field. One con-sequence is a strong temporal variation of the spectra, bright-ness, and morphology of non-thermal emission from core-collapse SNRs. On the one hand, it emphasizes the impor-tance of careful analysis of individual SNRs in pursue of the

Supernova remnants in circumstellar magnetic fields 15

origin of their emission and accordingly the origin of Galac-tic cosmic rays. On the other hand, it offers an opportunity toobtain valuable insights in the properties of SNR from theirnon-thermal emission.

This work is based on the research supported in part bythe National Research Foundation of South Africa (GrantNumber 132276). Robert Brose acknowledges funding froman Irish Research Council Starting Laureate Award (IR-CLA/2017/83).

1

2

3

4

5

REFERENCES

Abeysekara, A. U., Archer, A., Benbow, W., et al. 2020, ApJ, 894,51, doi: 10.3847/1538-4357/ab8310

Acero, F., Ackermann, M., Ajello, M., et al. 2016, ApJS, 224, 8,doi: 10.3847/0067-0049/224/1/8

Ackermann, M., e. a. 2013, Science, 339, 807,doi: 10.1126/science.1231160

Amato, E., & Blasi, P. 2006, MNRAS, 371, 1251,doi: 10.1111/j.1365-2966.2006.10739.x

Ambrogi, L., Zanin, R., Casanova, S., et al. 2019, A&A, 623, A86,doi: 10.1051/0004-6361/201833985

Arthur, S. J. 2007, in Astrophysics and Space Science Proceedings,Vol. 1, Diffuse Matter from Star Forming Regions to ActiveGalaxies - A Volume Honouring John Dyson, 183,doi: 10.1007/978-1-4020-5425-9 10

Beasor, E. R., & Davies, B. 2018, MNRAS, 475, 55,doi: 10.1093/mnras/stx3174

Bell, A. R., & Lucek, S. G. 2001, MNRAS, 321, 433,doi: 10.1046/j.1365-8711.2001.04063.x

Berezhko, E. G., & Volk, H. J. 2004, A&A, 427, 525,doi: 10.1051/0004-6361:20041111

Bhatt, M., Sushch, I., Pohl, M., et al. 2020, Astroparticle Physics,123, 102490, doi: 10.1016/j.astropartphys.2020.102490

Blasi, P., Gabici, S., & Vannoni, G. 2005, MNRAS, 361, 907,doi: 10.1111/j.1365-2966.2005.09227.x

Brose, R., Pohl, M., & Sushch, I. 2021, A&A, 654, A139,doi: 10.1051/0004-6361/202141194

Brose, R., Pohl, M., Sushch, I., Petruk, O., & Kuzyo, T. 2020,A&A, 634, A59, doi: 10.1051/0004-6361/201936567

Brose, R., Sushch, I., Pohl, M., et al. 2019, A&A, 627, A166,doi: 10.1051/0004-6361/201834430

Brose, R., Telezhinsky, I., & Pohl, M. 2016, A&A, 593, A20,doi: 10.1051/0004-6361/201527345

Bykov, A. M., Uvarov, Y. A., & Ellison, D. C. 2008, ApJ, 689,L133, doi: 10.1086/595868

Caprioli, D., Haggerty, C. C., & Blasi, P. 2020, ApJ, 905, 2,doi: 10.3847/1538-4357/abbe05

Celli, S., Morlino, G., Gabici, S., & Aharonian, F. A. 2019,MNRAS, 490, 4317, doi: 10.1093/mnras/stz2897

Chene, A.-N., & St-Louis, N. 2008, in IAU Symposium, Vol. 250,Massive Stars as Cosmic Engines, ed. F. Bresolin, P. A.Crowther, & J. Puls, 139–144,doi: 10.1017/S1743921308020425

Chene, A. N., & St-Louis, N. 2010, ApJ, 716, 929,doi: 10.1088/0004-637X/716/2/929

Cowsik, R., & Sarkar, S. 1980, MNRAS, 191, 855,doi: 10.1093/mnras/191.4.855

Cristofari, P., Blasi, P., & Caprioli, D. 2021, A&A, 650, A62,doi: 10.1051/0004-6361/202140448

Crowther, P. A. 2007, ARA&A, 45, 177,doi: 10.1146/annurev.astro.45.051806.110615

Das, S., Brose, R., M.-A. Meyer, D., et al. 2021, in Proceedings of37th International Cosmic Ray Conference — PoS(ICRC2021),Vol. 395, 988, doi: 10.22323/1.395.0988

de la Chevrotiere, A., St-Louis, N., Moffat, A. F. J., & MiMeSCollaboration. 2014, ApJ, 781, 73,doi: 10.1088/0004-637X/781/2/73

de Ona Wilhelmi, E., Sushch, I., Brose, R., et al. 2020, MNRAS,497, 3581, doi: 10.1093/mnras/staa2045

Diesing, R., & Caprioli, D. 2019, PhRvL, 123, 071101,doi: 10.1103/PhysRevLett.123.071101

Drury, L. 1983, Space Sci. Rev., 36, 57, doi: 10.1007/BF00171901Dwarkadas, V. V. 2005, ApJ, 630, 892, doi: 10.1086/432109—. 2007, ApJ, 667, 226, doi: 10.1086/520670Fukui, Y., Sano, H., Yamane, Y., et al. 2021, ApJ, 915, 84,

doi: 10.3847/1538-4357/abff4aGaggero, D., Zandanel, F., Cristofari, P., & Gabici, S. 2018,

MNRAS, 475, 5237, doi: 10.1093/mnras/sty140Garcia-Segura, G., Langer, N., & Mac Low, M. M. 1996a, A&A,

316, 133Garcia-Segura, G., & Mac Low, M.-M. 1995, ApJ, 455, 145,

doi: 10.1086/176563Garcia-Segura, G., Mac Low, M. M., & Langer, N. 1996b, A&A,

305, 229

16 Sushch et al.

H. E. S. S. Collaboration, Abdalla, H., Abramowski, A., et al.2018, A&A, 612, A3, doi: 10.1051/0004-6361/201732125

Haggerty, C. C., & Caprioli, D. 2020, ApJ, 905, 1,doi: 10.3847/1538-4357/abbe06

Harries, T. J., Hillier, D. J., & Howarth, I. D. 1998, MNRAS, 296,1072, doi: 10.1046/j.1365-8711.1998.01508.x

Henault-Brunet, V., St-Louis, N., Marchenko, S. V., et al. 2011,ApJ, 735, 13, doi: 10.1088/0004-637X/735/1/13

Hunter, I., Lennon, D. J., Dufton, P. L., et al. 2008, A&A, 479,541, doi: 10.1051/0004-6361:20078511

Inoue, T., Marcowith, A., Giacinti, G., van Marle, A. J., & Nishino,S. 2021, arXiv e-prints, arXiv:2108.13433.https://arxiv.org/abs/2108.13433

Jogler, T., & Funk, S. 2016, ApJ, 816, 100,doi: 10.3847/0004-637X/816/2/100

Kobzar, O., Niemiec, J., Pohl, M., & Bohdan, A. 2017, MNRAS,469, 4985, doi: 10.1093/mnras/stx1201

Krause, O., Birkmann, S. M., Usuda, T., et al. 2008, Science, 320,1195, doi: 10.1126/science.1155788

Lucek, S. G., & Bell, A. R. 2000, MNRAS, 314, 65,doi: 10.1046/j.1365-8711.2000.03363.x

Lyutikov, M., & Pohl, M. 2004, ApJ, 609, 785,doi: 10.1086/421290

Maeder, A., & Meynet, G. 2012, Reviews of Modern Physics, 84,25, doi: 10.1103/RevModPhys.84.25

Malkov, M. A., Diamond, P. H., & Sagdeev, R. Z. 2011, NatureCommunications, 2, 194, doi: 10.1038/ncomms1195

Malkov, M. A., & Drury, L. O. 2001, Reports on Progress inPhysics, 64, 429, doi: 10.1088/0034-4885/64/4/201

Marcowith, A., Dwarkadas, V. V., Renaud, M., Tatischeff, V., &Giacinti, G. 2018, MNRAS, 479, 4470,doi: 10.1093/mnras/sty1743

Mauron, N., & Josselin, E. 2011, A&A, 526, A156,doi: 10.1051/0004-6361/201013993

Meyer, D. M. A., Petrov, M., & Pohl, M. 2020, MNRAS, 493,3548, doi: 10.1093/mnras/staa554

Niemiec, J., Pohl, M., Bret, A., & Stroman, T. 2010, ApJ, 709,1148, doi: 10.1088/0004-637X/709/2/1148

Okuno, T., Tanaka, T., Uchida, H., et al. 2020, ApJ, 894, 50,doi: 10.3847/1538-4357/ab837e

Patnaude, D. J., & Fesen, R. A. 2009, ApJ, 697, 535,doi: 10.1088/0004-637X/697/1/535

Patnaude, D. J., Vink, J., Laming, J. M., & Fesen, R. A. 2011, ApJ,729, L28, doi: 10.1088/2041-8205/729/2/L28

Petruk, O., Kuzyo, T., & Beshley, V. 2016, MNRAS, 456, 2343,doi: 10.1093/mnras/stv2746

Pohl, M. 2021, ApJ, 921, 121, doi: 10.3847/1538-4357/ac21cf

Pohl, M., Wilhelm, A., & Telezhinsky, I. 2015, A&A, 574, A43,doi: 10.1051/0004-6361/201425027

Pohl, M., Yan, H., & Lazarian, A. 2005, ApJ, 626, L101,doi: 10.1086/431902

Rettig, R., & Pohl, M. 2012, A&A, 545, A47,doi: 10.1051/0004-6361/201219409

Reynolds, S. P. 2008, ARA&A, 46, 89,doi: 10.1146/annurev.astro.46.060407.145237

Riquelme, M. A., & Spitkovsky, A. 2009, ApJ, 694, 626,doi: 10.1088/0004-637X/694/1/626

Rosenau, P., & Frankenthal, S. 1976, Physics of Fluids, 19, 1889,doi: 10.1063/1.861424

Smartt, S. J. 2009, ARA&A, 47, 63,doi: 10.1146/annurev-astro-082708-101737

Stecker, F. W. 1973, ApJ, 185, 499, doi: 10.1086/152435Sushch, I., Brose, R., & Pohl, M. 2018, A&A, 618, A155,

doi: 10.1051/0004-6361/201832879Telezhinsky, I., Dwarkadas, V. V., & Pohl, M. 2012a, Astroparticle

Physics, 35, 300, doi: 10.1016/j.astropartphys.2011.10.001—. 2012b, A&A, 541, A153, doi: 10.1051/0004-6361/201118639—. 2013, A&A, 552, A102, doi: 10.1051/0004-6361/201220740Tessore, B., Lebre, A., Morin, J., et al. 2017, A&A, 603, A129,

doi: 10.1051/0004-6361/201730473Truelove, J. K., & McKee, C. F. 1999, ApJS, 120, 299,

doi: 10.1086/313176Tsuji, N., Uchiyama, Y., Khangulyan, D., & Aharonian, F. 2021,

ApJ, 907, 117, doi: 10.3847/1538-4357/abce65Uchiyama, Y., & Aharonian, F. A. 2008, ApJ, 677, L105,

doi: 10.1086/588190van den Bergh, S., & Tammann, G. A. 1991, ARA&A, 29, 363,

doi: 10.1146/annurev.aa.29.090191.002051van Marle, A. J., Meliani, Z., & Marcowith, A. 2015, A&A, 584,

A49, doi: 10.1051/0004-6361/201425230Vink, J. 2012, A&ARv, 20, 49, doi: 10.1007/s00159-011-0049-1Voelk, H. J., & Forman, M. 1982, ApJ, 253, 188,

doi: 10.1086/159623Weaver, R., McCray, R., Castor, J., Shapiro, P., & Moore, R. 1977,

ApJ, 218, 377, doi: 10.1086/155692Wilhelm, A., Telezhinsky, I., Dwarkadas, V. V., & Pohl, M. 2020,

A&A, 639, A124, doi: 10.1051/0004-6361/201936079Yang, R.-z., Kafexhiu, E., & Aharonian, F. 2018, A&A, 615, A108,

doi: 10.1051/0004-6361/201730908Zirakashvili, V. N., & Aharonian, F. 2007, A&A, 465, 695,

doi: 10.1051/0004-6361:20066494Zweibel, E. G. 1979, AIP Conf. Proc., 56, 319,

doi: 10.1063/1.32090