betaFeedback

WorksPhysics and Astronomy

Rapid Variability of Radiative Signatures from Driven Relativistic Reconnection

Abstract

Rapid variability in energetic flares is a distinctive feature of high-energy emission from many magnetically dominated astrophysical sources. We investigate the radiative variability of driven relativistic magnetic reconnection using Particle-in-Cell simulations of strongly magnetized flux tubes pushed to merge at driving speed vpushv_{\rm push}. This setup captures the onset of reconnection and provides a controlled model for single reconnection-driven flares near compact objects, where magnetic structures can be twisted, sheared, and forced to merge by turbulent accretion flows or relativistic frame dragging. We find that the merger efficiently accelerates particles into nonthermal distributions, producing synchrotron emission whose variability tracks the reconnection phases. We identify two regimes: a quasi-steady, weakly driven regime for vpush≤0.1 cv_{push}\leq0.1\,c, and strongly driven regime for vpush>0.1 cv_{\rm push}>0.1\,c. The transition occurs when the driving exceeds the typical inflow-to-outflow velocity ratio of ∼0.1\sim0.1 in steady collisionless relativistic reconnection, i.e., the average rate at which reconnection can process incoming flux, potentially providing an additional source of variability. The lightcurves rise on 11–33 light-crossing times of the reconnecting region, and the particle spectra develop power laws with index p≃1.8p\simeq1.8–2.42.4. These timescales are compatible with observed variability in high-energy sources, including the very-high-energy flares of M87∗^* and the γγ-ray flares of the Crab Nebula, making this model a viable framework for rapid flaring in active galactic nuclei and pulsar wind nebulae.

Introduction

Observations of rapidly variable high-energy sources, in particular by γ\gamma-ray telescopes such as AGILE and Fermi-LAT ([8, 59]), strongly suggest the existence of compact regions of high electromagnetic energy density, where this energy is efficiently dissipated via radiative processes and charged particles are accelerated beyond their radiation-reaction limit on timescales comparable to the light-crossing time of the emitting region. A prominent example is the powerful γ\gamma-ray flaring activity of the Crab Nebula (>100>100 MeV), with events typically lasting from a few days to weeks. During these flares, the γ\gamma-ray flux increases many-folds, while no corresponding enhancement is observed in other energy bands; the brightest flares even exhibit resolvable intra-day variability ([1, 10, 58, 60]). Rapid variability is also routinely detected in active galactic nuclei (AGNs) ([6, 2, 49, 7] (L. Gu et al. 2025)). For example, GeV flares on hourly timescales are observed in flat-spectrum radio quasars such as 3C 273 ([46]) and 3C 279 ([39]), while several BL Lac objects, e.g., PKS 2155−304 ([4]) and Markarian 501 ([5]), flare in the TeV band with variability timescales as short as a few minutes.

In most of these astrophysical sources, the observed nonthermal emission is the combined output of two radiative channels operated by populations of relativistic leptons (e±e^{\pm}): synchrotron emission in the local magnetic field, and inverse-Compton (IC) scattering off low-

Corresponding author: Email: [email protected], [email protected] energy photon fields, whether externally supplied (e.g., accretion-disk or dusty-torus radiation in AGNs) or self-generated by the same synchrotron-emitting particles (synchrotron self-Compton, SSC). [e.g. 15, 53] Because both radiative channels carry a direct imprint of the particle acceleration process itself, there is a direct link between the shape and variability of the high-energy emission and the physical mechanism responsible for energizing the radiating particles in the first place.

This general class of phenomena of large-scale, extremely rapid conversion of electromagnetic energy into a nonthermal, high-energy particle population has recently been termed magnetoluminescence [13, 12, 67, 14]. In this picture, macroscopic instabilities in a magnetically dominated outflow give rise to localized regions of unscreened electric field, in which a small population of particles is accelerated efficiently, while processing a large plasma volume on a short timescale; additional ingredients, such as bulk relativistic motion, may be required to fully explain observations.

Magnetic reconnection is widely regarded as one of the most promising physical realizations of this process, and has repeatedly been proposed as the source of high-energy, rapidly variable radiation in extreme astrophysical environments, including black-hole accretion disks and their coronae [29, 11, 62, 50, 9, 34, 26], pulsar wind nebulae [19, 21, 65, 63], gamma-ray bursts [41, 28, 45], and neutron star and magnetar magnetospheres (A. Y. Chen & A. M. Beloborodov 2014; B. Cerutti et al. 2016a,b; H. Hakobyan et al. 2019; K. M. Schoeffler et al. 2019, 2023). Particle-in-Cell (PIC) simulations of radiative reconnection in the strong-cooling regime have further demonstrated particle acceleration beyond the synchrotron burnoff limit (>160> 160 MeV; [20, 21, 23]), as observed, for instance, in Crab Nebula flares.

In this paper, we present a relativistic, reconnection-driven flare model based on the rapid merger of two flux tubes of pair plasma [42, 52, 51, 31, 61, 30]. Unlike standard setups built around a single, pre-formed Harris sheet, our configuration self-consistently generates the current layer from the compression and collision of two force-free magnetic structures, allowing us to follow reconnection from its onset through to the nonlinear, dynamical merger of the two flux ropes. This offers a controlled framework for modeling a single flaring event, triggered by the driving speed vpushv_{\mathrm{push}} at which the flux tubes are pushed together, while letting the underlying reconnection dynamics develop self-consistently. This setup also intrinsically accounts for multi-scale coupling between the current sheet and the global system, crucial to a complete understanding of energy partition and particle acceleration in reconnecting systems. Physically, this model is motivated by environments around compact objects, where e.g. magnetic structures threading accretion flows can be twisted, sheared, and driven into coalescence by turbulent motions or relativistic frame-dragging effects [29], providing a natural trigger for reconnection-powered flares. By systematically varying vpushv_{\mathrm{push}}, we characterize how the resulting synchrotron emission variability and its timescales depend on the strength of the external driving, and assess to what extent this model is compatible with the rapid variability observed in high-energy sources.

The paper is organized as follows. In § 2 we describe the numerical setup, the simulation parameters, and the methods used for the emission analysis. In § 3 we present our results on the temporal and spatial variability of the system’s synchrotron emissivity, the nonthermal particle spectra, and their connection to the underlying reconnection stages and dynamics. In § 4 we discuss the astrophysical implications of our model and summarize our conclusions.

Methodology

Numerical setup

We perform 2.5D numerical simulations of electron–positron plasma with the PIC code TRISTAN-MP v2 [38]. The setup is taken from our previous work on reconnection onset-to-coalescence dynamics [31] and is briefly summarized as follows. We initialize two identical cylindrical Lundquist-type force-free flux tubes in a square domain of size Lx=Ly=4RL_x = L_y = 4R and 14336 × 14336 cells (R/de=896R/d_e = 896), where RR is the radius of the tubes centered in (x1,y1)=(2R,R)(x_1,y_1) = (2R,R); (x2,y2)=(2R,3R)(x_2,y_2) = (2R,3R) and de=c/4πnee2/med_e = c/\sqrt{4\pi n_e e^2/m_e} is the electron skin depth, with nen_e electron number density. The in-plane (x−y)(x-y) and out-of-plane (z)(z) magnetic-field profiles are given by [42, 52, 51]

Bϕ(r)=B0J1(αr/R),(1)B_\phi(r) = B_0 J_1(\alpha r/R), \tag*{(1)}
Bz(r)=B0J02(αr/R)+C,(2)B_z(r) = B_0 \sqrt{J_0^2(\alpha r/R) + C}, \tag*{(2)}

where rr is the radial distance from the tube axis, J0J_0 and J1J_1 are Bessel functions of the first kind, α≈3.8317\alpha\approx3.8317 is the first zero of J1J_1, B0B_0 sets the characteristic magnetic-field strength, and the constant CC controls the relative guide-field contribution in the vicinity of the current-sheet midplane where reconnection occurs. To impose a force-free initial state, the guide field has a nontrivial profile: it peaks at the centers of the tubes, decreases toward the edges, and exhibits a slight bump near the region of contact between the tubes. At the boundaries, electromagnetic fields are damped to zero, and particles are allowed to periodically re-enter the simulation box. We set an initial total magnetization σ0=B02/(4πn0mec2)=60\sigma_0 = B_0^2/(4\pi n_0 m_e c^2) = 60 (with n0n_0 the total plasma number density) and the in-plane magnetization from the reconnecting magnetic field in the (x,y)(x,y)-plane is σin=0.16 σ0∼10\sigma_{\mathrm{in}} = 0.16\,\sigma_0 \sim10.

To induce the onset of reconnection, the tubes are pushed toward each other with equal and opposite velocity vpush=±vpushy^\mathbf{v}_{\mathrm{push}} = \pm v_{\mathrm{push}}\hat{\mathbf{y}}, corresponding to a driving timescale tdrive=R/vpusht_{\mathrm{drive}} = R/v_{\mathrm{push}} and inducing a motional electric field E=−vpush×B/c\mathbf{E} = -\mathbf{v}_{\mathrm{push}} \times\mathbf{B}/c. As discussed in C. Granier et al. [31], we distinguish three stages in the evolution of this model. Phase I: formation of the current sheet; Phase II: slow start of reconnection up to ∼0.1%\sim0.1\% dissipation of the in-plane magnetic energy and plasmoid formation; Phase III: rapid coalescence of the tubes between tmergt_{\mathrm{merg}} and trelaxt_{\mathrm{relax}} (30%30\% of energy dissipated), Δtmerg≡trelax−tmerg\Delta t_{\mathrm{merg}} \equiv t_{\mathrm{relax}} - t_{\mathrm{merg}}. The external driving significantly influences the duration of the current sheet formation, as well as the linear tearing phase (Phases I–II). In the subsequent nonlinear merging phase (Phase III), the evolution becomes effectively independent of the initially imposed velocity.

We consider eight 2D simulations with a range push velocities vpush/c={0.01,0.02,0.05,0.1,0.2,0.3,0.4,0.5}v_{\mathrm{push}}/c = \{0.01, 0.02, 0.05, 0.1, 0.2, 0.3, 0.4, 0.5\}. We analyze how the emission evolves through stages of reconnection as a function of the driving timescale. Such driving speeds are in line with Alfvénic-like motions in high-σ\sigma relativistic plasma ([44, 43]) and observations of ultrafast winds from AGNs ([66] (L. Gu et al. 2025)).

Synchrotron emission

The radiation power spectrum of a single relativistic electron of mass mem_e and charge ee is given by ([15, 53])

dPsyncdν=3e3B~⊥mec2ννc∫ν/νc+∞K5/3(ω) dω(3)\frac{\mathrm{d}P_{\mathrm{sync}}}{\mathrm{d}\nu} = \frac{\sqrt{3}e^3\widetilde{B}_{\perp}}{m_e c^2}\frac{\nu}{\nu_c}\int_{\nu/\nu_c}^{+\infty} K_{5/3}(\omega)\,\mathrm{d}\omega \tag*{(3)}

where ν\nu is the radiation frequency, K5/3K_{5/3} is the modified Bessel function of order 5/35/3,

νc=3eB~⊥γ24πmec(4)\nu_c = \frac{3e\widetilde{B}_{\perp}\gamma^2}{4\pi m_e c} \tag*{(4)}

is the critical frequency (i.e., the frequency at which a relativistic particle emits most of its synchrotron radiation), and

B~⊥=(E+β×B)2−(β⋅E)2(5)\widetilde{B}_{\perp} = \sqrt{(\mathbf{E}+\boldsymbol{\beta}\times\mathbf{B})^2-(\boldsymbol{\beta}\cdot\mathbf{E})^2} \tag*{(5)}

is the effective (relativistically invariant) magnetic field strength perpendicular to the particle motion, as found in the L. D. Landau & E. M. Lifshitz [40] formulation of the synchro-curvature radiation force. E\mathbf{E} and B\mathbf{B} are the background electromagnetic fields and β=v/c\boldsymbol{\beta} = \mathbf{v}/c is the particle 3-velocity.

To model the emission physics from our setup, we construct synthetic synchrotron radiation. We store the values of E\mathbf{E} and B\mathbf{B} interpolated from the grid to the particle position at every time step, compute the emitted radiation power spectrum per single particle from Eq. (3), and then sum over the total particle distribution to obtain the full synchrotron spectra νFν\nu F_\nu. Furthermore, by integrating Eq. (3) over the whole frequency range, we obtain the total energy emitted via synchrotron radiation over time, i.e., lightcurves L(t)=dEsync/dtL(t) = \mathrm{d}E_{\mathrm{sync}}/\mathrm{d}t, presented in the next section. Extensive applications of the method can also be found in (B. Reville & J. G. Kirk 2010; B. Cerutti et al. 2016a,b; H. Hakobyan et al. 2019, 2023; B. Cerutti et al. 2025). Note that particle cooling is not included in our simulations; radiative losses do not feed back on the particle motion.

Results

In Figure 1 and Figure 2 we show the evolution of two representative runs for vpush=0.02cv_{\mathrm{push}} = 0.02c (weak drive) and vpush=0.3cv_{\mathrm{push}} = 0.3c (strong drive). In the first row, we display the evolution of the lightcurve; in the second row, the spatial distribution of the out-of-plane current density JzJ_z (with selected magnetic field lines); and in the bottom row, the spatial distribution of the total synchrotron power Psync=(4/3)σT(γ2B~⊥2/(8π))cP_{\mathrm{sync}} = (4/3)\sigma_T(\gamma^2 \widetilde{B}_{\perp}^2/(8\pi))c.

Evolution of a single run at different stages for $v_{\rm push}=0.02c$

Figure 1. Evolution of a single run at different stages for vpush=0.02cv_{\rm push}=0.02c: lightcurve temporal evolution (top row); spatial distribution of the out-of-plane current density JzJ_z with selected magnetic field lines (second row); spatial distribution of the total synchrotron power (third row); zoom in on the current-sheet area for the total synchrotron power spatial distribution of the highest energy particle with Lorentz factor γ>σin\gamma> \sigma_{\rm in} (bottom row).

Evolution of a single run at different stages for $v_{\mathrm{push}} = 0.3c$

Figure 2. Evolution of a single run at different stages for vpush=0.3cv_{\mathrm{push}} = 0.3c: lightcurve temporal evolution (first row); spatial distribution of the out-of-plane current density JzJ_z with selected magnetic field lines (second row); spatial distribution of the total synchrotron power (third row); zoom in on the current-sheet area for the total synchrotron power spatial distribution of the highest energy particle with Lorentz factor γ>σin\gamma> \sigma_{\mathrm{in}} (last row).

In the weakly driven case (Figure 1), the system appears to evolve gradually, driven by large-scale electromagnetic stresses: as a result of the initial reconnection phase, field lines that initially enclose the separate flux tubes merge and create an envelope of tension force that rapidly drives the cores together ([42]). In the strongly driven case (Figure 2), a large current sheet of size L∼2RL \sim2R forms extremely rapidly, and the initial compression of the magnetic field triggers the formation of a first-generation plasmoids (t=2.43 R/ct = 2.43\,R/c) that is almost immediately ejected, forming a two-arm structure of turbulent outflow (t=3.68 R/ct=3.68\ R/c). This corresponds to a first increase in emission of roughly two orders of magnitude above the baseline; only then does the actual merger start, and the system evolves similarly to the vpush=0.02 cv_{\rm push}=0.02\ c case (t=5.44t=5.44–8.93 R/c8.93\ R/c).

Lightcurve variability

In Figure 3 we characterize the system variability for different initial drives vpushv_{\rm push} by comparing the timescales and evolution of: (i) the total synchrotron emission L(t)=dEsync/dtL(t)=dE_{\rm sync}/dt (top panel); (ii) the in-plane magnetic-energy variation [EB(t)−EB(0)]/EB(0)[E_B(t)-E_B(0)]/E_B(0) (middle panel); (iii) the normalized reconnection rate (bottom panel), defined as

Figure 3. Time evolution of: (top panel) lightcurves; (middle panel) the in-plane magnetic energy variation; (bottom panel) the normalized reconnection rate.

R(t)=1vABx,upddt[max⁡(Az)−min⁡(Az)],(6)R(t)=\frac{1}{v_A B_{x,\mathrm{up}}}\frac{\mathrm{d}}{\mathrm{d}t}\left[\max(A_z)-\min(A_z)\right], \tag*{(6)}

where AzA_z is the out-of-plane component of the magnetic vector potential along the current sheet, Bx,upB_{x,\mathrm{up}} is the upstream reconnecting field, and vA=cσx,up/(1+σx,up)v_A=c\sqrt{\sigma_{x,\mathrm{up}}/(1+\sigma_{x,\mathrm{up}})} is the corresponding Alfvén speed, with σx,up=Bx,up2/(4πn0mec2)\sigma_{x,\mathrm{up}}=B_{x,\mathrm{up}}^2/(4\pi n_0m_ec^2). The synchrotron lightcurves for all eight simulations (top panel of Figure 3) share a similar qualitative structure: a sigmoid-shaped emission trend with a sharp rise of several orders of magnitude at the onset of reconnection, before reaching a similar saturation level. The onset time of the rise is delayed systematically with the push strength, occurring earlier for faster pushes and later for slower ones, although presenting clear saturation for vpush={0.2cv_{\mathrm{push}} = \{0.2c–0.5c}0.5c\}, whose dynamics tend to partially overlap from the beginning. The shape of the rise is, however, not universal: we can distinguish two dynamical regimes for vpushv_{\mathrm{push}} above or below ≈0.1c\approx0.1c (representative examples are described in Figure 1 and Figure 2).

The vpush={0.01cv_{\mathrm{push}} = \{0.01c–0.1c}0.1c\} lightcurves rise smoothly and monotonically until saturation. Simultaneously with the rise in emission, we observe progressive dis-

sipation of the in-plane magnetic energy (middle panel) and one broad peak of order 0.15–0.3 in the reconnection rate R(t)R(t) (bottom panel). The relevant dynamics and most of the emission are then associated with one rapid, coherent reconnection event corresponding to Phase III, the dynamical merger.

By contrast, the {0.2c\{0.2c–0.5c}vpush0.5c\}v_{\mathrm{push}} lightcurves show a distinctly multi-step rise: an initial jump, a partial plateau/bump at an intermediate value, and one more subsequent increase interval before reaching saturation (Figure 3, top). In the in-plane magnetic energy, we observe a rapid increase (between 1–3 R/cR/c) as a result of the initial compression of the tubes under the strong push. The reconnection rate shows the same two-stage pattern (Figure 3, bottom panel), which, for the stronger-push runs, is markedly bursty, exhibiting two distinct narrow peaks (order of ≲0.15\lesssim0.15) rather than a single dominant event. The first peak occurs during the compression phase, and the second peak during the merger. Thus, with increasing vpushv_{\mathrm{push}}, the system develops noticeable pre-merger dynamics, i.e., a compression phase for t≲3 R/ct \lesssim3\,R/c (dotted gray line in Figure 3), governed by the initial drive and influencing the evolution of both reconnection and emission.

The transition between the two regimes occurs around vpush=0.1cv_{\mathrm{push}} = 0.1c. Interestingly, this is close to vin/vout∼0.1v_{\mathrm{in}}/v_{\mathrm{out}} \sim0.1 the standard value for the asymptotic inflow-to-outflow ratio (i.e., reconnection rate) in steady-state, collisionless relativistic reconnection, when vout∼vA∼cv_{\mathrm{out}} \sim v_A \sim c (L. Sironi & A. Spitkovsky [56]; L. Sironi et al. [57]). Our system does not reach a steady state, but the resulting regimes in our setup could be related to whether the external driving exceeds or remains below the average rate at which collisionless reconnection can steadily process incoming flux. For vpush≥0.1cv_{\mathrm{push}} \ge0.1c, the flux tubes are driven together faster than the reconnection layer can dissipate the incoming magnetic flux; the system instantly accumulates large flux and stress, which is released almost immediately in a discrete burst (first two snapshots in the bottom row of Figure 2). For vpush≤0.1cv_{\mathrm{push}} \le0.1c, the system remains closer to marginally stable reconnection, producing smooth, single event energy release and emission.

Rescaling all lightcurves by a time factor t0t_0 (Figure 4) highlights self-similarity in the weaker-push family (vpush≤0.1cv_{\mathrm{push}} \le0.1c) as they collapse remarkably well onto a single curve described by a logistic function in a log–log space

Figure 4. Time-rescaled lightcurves in logarithmic space, with a fit to the logistic model in Eq. (7).

log⁡10L(t)=Lmin⁡+Lmax⁡−Lmin⁡1+exp⁡(−k(log⁡10t−log⁡10t0))(7)\log_{10}L(t) = L_{\min} + \frac{L_{\max} - L_{\min}}{1 + \exp\left(-k\left(\log_{10}t - \log_{10}t_0\right)\right)} \tag*{(7)}

where Lmin⁡L_{\min} and Lmax⁡L_{\max} are the lower and upper asymptotes of log⁡10L\log_{10} L, corresponding to the baseline and saturation fluxes 10Lmin⁡10^{L_{\min}} and 10Lmax⁡10^{L_{\max}}, kk is the growth rate or steepness of the curve, and t0t_0 is the inflection time (in physical units of R/cR/c). For the black dotted logistic curve, shown as reference in Figure 4, we have Lmin⁡≈2L_{\min} \approx2, Lmax⁡≈7L_{\max} \approx7, and k≈40k \approx40.

Given the analytic model in Eq. (7), the estimated rise time τriselog⁡L\tau_{\mathrm{rise}}^{\log L} for our simulated lightcurves to grow from 10% to 90% of the total variation is

τriselog⁡L=t0(10log⁡9/k−10−log⁡9/k)≈3 R/c(8)\tau_{\mathrm{rise}}^{\log L} = t_0\left(10^{\log9/k} - 10^{-\log9/k}\right) \approx3\,R/c \tag*{(8)}

for the vpush={0.01cv_{\mathrm{push}} = \{0.01c–0.1c}0.1c\} cases.

We note that τriselog⁡L\tau_{\mathrm{rise}}^{\log L} measures the time needed for log⁡10L\log_{10} L to go from 10% to 90% of its total variation, roughly corresponding to an increase from ≃0.003%\simeq0.003\% to ≃32%\simeq32\% of the maximum linear flux. As we are interested in identifying when and how the most marked signatures of the emission appear, we characterize the rise via the logarithmic variations, whereas on a linear scale the variations would be dominated by the area near the maximum and the onset of the rise would be compressed onto the baseline.

For the stronger push cases vpush={0.2cv_{\mathrm{push}} = \{0.2c–0.5c}0.5c\}, it is more challenging to identify a unique model of the timescale evolution with one defined rise interval. Considering the fragmented character of the lightcurves (Figure 4), a single fitting function cannot describe all curves well. We therefore estimate the variability timescale qualitatively, directly from the curves shown in the top panel of Figure 3 (see also Appendix § 6 for a zoom-in). During the compression phase, there is a very sharp rise in τrise∼1 R/c\tau_{\mathrm{rise}} \sim1\,R/c followed by a small decrease of similar duration, while the merger phase shows a characteristic τrise∼2 R/c\tau_{\mathrm{rise}} \sim2\,R/c. Overall, the simulated emission shows a variability timescale of 11–3 R/c3\,R/c, in good agreement with different high-energy astrophysical sources where synchrotron losses are the dominant radiative mechanism (A. A. Abdo et al. [1]; M. Tavani et al. [60]; A. Abramowski et al. [2]; J. C. Algaba et al. [7]). We discuss qualitative comparisons with observational cases more elaborately in § 4.

Particle energy spectra

High-energy synchrotron radiation is produced by accelerated nonthermal particles (G. B. Rybicki & A. P. Lightman [53]). We discuss, in the following, the formation and evolution of the nonthermal component of the particle energy distribution and the corresponding synchrotron spectra νFν\nu F_{\nu} (defined in § 2.2 above).

In Figure 5 we show the particle energy spectra (top panel) and the synchrotron spectra (bottom panel) for the two reference cases discussed above, vpush=0.02cv_{\mathrm{push}} = 0.02c (purple) and vpush=0.3cv_{\mathrm{push}} = 0.3c (red). For each case, we plot the distributions at the start (tmergt_{\mathrm{merg}}, dotted lines) and end (trelaxt_{\mathrm{relax}}, solid lines) of the dynamical merger (Phase III).

Figure 5. Electron energy distributions (top panel) and corresponding synchrotron spectra for the two reference cases vpush=0.02cv_{\mathrm{push}} = 0.02c (purple) and vpush=0.3cv_{\mathrm{push}} = 0.3c (red). We plot the distributions at the start (tmergt_{\mathrm{merg}}, dotted lines) and end (trelaxt_{\mathrm{relax}}, solid lines) of the dynamical merger (Phase III).

As discussed above, the dynamics during the merger event are governed by self-consistent, large-scale electromagnetic stresses, largely independently of the initial driving speed. Consistently, Figure 5 shows that the post-merger particle and synchrotron distributions of the two runs nearly overlap (similarly for all other cases, see Appendix § 6, Figure 7). The two cases differ primarily in the pre-merger phase, where injection physics retains a dependence on the initial drive: the stronger-push run injects a larger population of particles into the nonthermal pool during the pre-merger (compression) phase, but with a somewhat softer power-law index, as reflected in the modest difference between the energy and synchrotron spectra at tmergt_{\mathrm{merg}} and trelaxt_{\mathrm{relax}}. The weak-push case instead starts from a smaller nonthermal population, which is subsequently accelerated much more efficiently during the merger itself, producing the pronounced jump between the pre- and post-merger distributions visible in Figure 5 (top panel).

Figure 7. Particle energy distributions (left) and particle synchrotron spectra (right) for all the runs at post-merger time trelaxt_{\rm relax}.

We find a nonthermal power-law index pp in the range 1.8–2.4 (including also all other vpushv_{\mathrm{push}} cases, see Appendix § 6, Figure 7), indicating efficient particle acceleration. The corresponding synchrotron spectral index is in agreement with the theoretical expectation for a power-law particle energy distribution and isotropic velocity distribution with respect to the local magnetic field, νFν∝νs\nu F_{\nu} \propto\nu^{s} with s=(3−p)/2s = (3-p)/2 (G. B. Rybicki & A. P. Lightman [53]; L. Comisso et al. [25]; L. Comisso & B. Jiang [24]). These results are consistent with the findings of C. Granier et al. [30] in a similar setup across a range of driving strengths and guide-field amplitudes. There, the high-energy cutoff converged to a common asymptotic value, (γcut/σin)max⁡≃50(\gamma_{\mathrm{cut}}/\sigma_{\mathrm{in}})_{\max} \simeq50, only weakly dependent on drive strength, guide field, or dimensionality (as in our case, with (γcut/σin)max⁡≃100(\gamma_{\mathrm{cut}}/\sigma_{\mathrm{in}})_{\max} \simeq100; see Appendix § 6). This is attributed to electric-field-limited acceleration, where the maximum energy is set by the reconnection electric field integrated over the fast-merging phase. The nonthermal spectra were similarly robust, with p≃1.6p \simeq1.6–2.0, comparable to our range. Particle tracking in C. Granier et al. [30] further showed that the reconnection layer acts as a largely injection-energy-agnostic accelerator, with the energy gain per crossing controlled mainly by the residence time in the current sheet and the local electric field.

As synchrotron cooling is not included dynamically in the particle motion in our runs (i.e., from a physical standpoint, cooling can be considered dynamically weak), the plasmoids carry most of the high-energy particles, as can be observed in the two single-case evolutions of PsyncP_{\mathrm{sync}} in Figure 1–Figure 2 (bottom row). These particles lose energy on timescales much longer than the characteristic injection timescale from the current sheet, and the plasmoids are rapidly replenished with freshly accelerated plasma. In this regime, most of the high-energy photons are therefore produced inside the largest plasmoids (H. Hakobyan et al. 2023) and the synchrotron peaks are set by the particles energy cutoff, νpeak∼γcut∼100σin\nu_{\mathrm{peak}} \sim\gamma_{\mathrm{cut}} \sim100\sigma_{\mathrm{in}}, as visible in Figure 5 and Figure 7. Moreover, plasmoids may provide a secondary energization channel: as they grow, their interiors compress and amplify the internal magnetic field, roughly linearly in time, so that the trapped particles gain energy through conservation of their magnetic moment (H. Hakobyan et al. [35]).

Conclusions

We have presented a relativistic, reconnection-driven flare model based on 2D PIC simulations of two merging flux tubes of pair plasma, spanning a range of external driving speeds vpush/c=v_{\mathrm{push}}/c = {0.01,0.02,0.05,0.1,0.2,0.3,0.4,0.5}\{0.01, 0.02, 0.05, 0.1, 0.2, 0.3, 0.4, 0.5\}. Our main findings can be summarized as follows.

Despite showing distinct transient features in the pre-merger phase and during the rise of the emission, all simulations converge to a quite similar post-merger state, with similar particle energy distributions, synchrotron spectra, and lightcurve saturation levels. This confirms that the final outcome is largely insensitive to the details of the initial driving.

Magnetic reconnection triggered by the merger of the two flux tubes drives fast variability in synchrotron emission, on timescales of the order of the light-crossing time of the reconnecting region (τrise∼1\tau_{\mathrm{rise}} \sim1–3 R/c3\,R/c). This variability is directly tied to the underlying reconnection dynamics: we identify two distinct regimes, separated at vpush≈0.1 cv_{\mathrm{push}} \approx0.1\,c, corresponding respectively to a quasi-steady, single-event energy release (weak push) and a strongly driven, multi-step, bursty release (strong push), with the transition governed by whether the external driving exceeds the characteristic rate at which collisionless relativistic reconnection can steadily process incoming flux. The resulting rise timescales are compatible with observed variability in several classes of high-energy sources, providing a plausible physical basis for interpreting rapid flaring episodes through a reconnection-driven scenario.

Taking for example the case of a low-luminosity AGN such as M87* at the center of the homonymous radio galaxy, the inferred size for the emitting region during a rapid energetic flare is of a few RSR_{\mathrm{S}} ([3, 48] (H. Hakobyan et al. 2023)), where RS=2GMBH/c2∼1.8×1015 cmR_{\mathrm{S}} = 2GM_{\mathrm{BH}}/c^2 \sim1.8 \times10^{15}\,\mathrm{cm} is the Schwarzschild radius ([27]). Our scaling would then give to a τrise≲3 RS/c∼2.08\tau_{\mathrm{rise}} \lesssim3\,R_{\mathrm{S}}/c \sim2.08 days, compatible, for instance, with typical timescales from major observational events like the April 2010 VHE-flare ([2]), showing a γ\gamma-ray flux (above 350 GeV) rise of τrise=(2.44±0.40)\tau_{\mathrm{rise}} = (2.44 \pm0.40) days and the VHE rapid flare observed in April 2018, during the EHT multi-wavelength campaign, with τrise=(1.88±0.27)\tau_{\mathrm{rise}} = (1.88 \pm0.27) days ([7]). We note that our simulated rise fits correspond to an approximate power-law growth in linear time L(t)∼tkL(t) \sim t^k around the inflection point t0t_0 rather than the exponential growth used in the observational fits from [2, 7]. Nonetheless, when focusing only on the timescale variability of the rise, they appear as comparably rapid episodes.

We can apply a similar qualitative reasoning to the case of the Crab nebula: the flare rise times (observations vary from a few days to two weeks; [1, 60]) favor a compact emission region, close to the pulsar, of size Rp≲1016 cmR_{\mathrm{p}} \lesssim10^{16}\,\mathrm{cm}. This gives a rise time for our setup of τrise≲3 Rp/c∼11\tau_{\mathrm{rise}} \lesssim3\,R_{\mathrm{p}}/c \sim11 days, again aligned with the recorded observational window.

This study was not designed to map our results directly onto specific observations, but rather to assess the viability of the model and identify the range of its applicability. We focused on variability timescales, without attempting to relate simulation units to observed synchrotron fluxes. A next step in this direction would be to start from a specific astrophysical source, extrapolate its physical parameters, and use them to construct a matched simulation setup. The sources discussed above (VHE flares from BL Lac objects, M87*, and pulsar wind nebulae) span different radiative-cooling regimes, and in some cases it may be necessary to include the radiation-reaction (drag) force in the particle equations of motion to properly capture the effect of cooling on the dynamics. Whether and how this would affect the resulting emission timescales remains an open question, and represents an interesting direction for future work.

A further important ingredient to explore would be the composition of the current sheet ([23]). While most studies of relativistic reconnection, including this one, focus on pair plasmas, many astrophysical environments of interest are expected to host a mixture of ions (primarily protons) in addition to leptons [33, 64, 9]. Since for ions radiative losses are often much weaker over the timescales of interest, they can in principle be accelerated more efficiently by reconnection than radiating leptons. This question has not been systematically addressed and constitutes a promising avenue for future work in systems (like ours) with driven relativistic reconnection and where the turbulent region outside the current sheet is taken into account.

Another extension of this work is to compute the emission from a population of interacting flux tubes undergoing successive mergers, as in island-merger models of decaying magnetic turbulence [68] (Y. Cai et al. 2026). Computing the radiative output in this setup would test whether the lightcurve signatures identified here for a single merger remain visible when many reconnection events overlap, or whether their superposition produces qualitatively different effects.

Acknowledgments

CM acknowledges support from the FWO PhD Fellowship fundamental research 1105426N granted by the Research Foundation – Flanders (FWO). FB acknowledges support from the FED-tWIN programme (profile Prf-2020-004, project “ENERGY”) issued by BELSPO, and from the FWO Junior Research Project G020224N granted by the Research Foundation – Flanders (FWO).

Software: TRISTAN-MP V2 [38]

Supplementary diagnostics

Strongly driven lightcurves

As discussed in § 3.1, for vpush≥0.2cv_{\mathrm{push}} \ge0.2c the lightcurves cannot be described by an analytic profile; thus we estimate their rise times directly from the curves. The left panel of Figure 6 shows a zoom on the first 12 R/cR/c of the vpush={0.2cv_{\mathrm{push}} = \{0.2c–0.5c}0.5c\} lightcurves, where the multi-step structure is clearly visible. All runs exhibit a sharp jump of about two orders of magnitude within τrise∼1R/c\tau_{\mathrm{rise}} \sim1 R/c at t≈2t \approx2–2.5R/c2.5 R/c, occurring slightly earlier for faster pushes. This first burst coincides with the compression phase, and it is followed by a by a second, more gradual rise over τrise∼2R/c\tau_{\mathrm{rise}} \sim2 R/c associated with the dynamical merger (Phase III), before the emission saturates at L∼107L \sim10^7 (a.u.), the same level reached by the weakly driven runs.

Figure 6. Left: zoom in on strong-push (0.2c-0.5c) lightcurves to qualitatively estimate the timescale variability; Right: high-energy cutoff Lorentz factor of electrons in time for all runs.

Evolution of the high-energy cutoff

The right panel of Figure 6 shows the time evolution of the cutoff Lorentz factor γc\gamma_{\mathrm{c}} of the particle distribution, normalized to the in-plane magnetization, γcut/σin\gamma_{\mathrm{cut}}/\sigma_{\mathrm{in}}, for all eight runs. In the strongly driven runs, γcut\gamma_{\mathrm{cut}} grows already during the compression phase, reaching γcut/σin∼10\gamma_{\mathrm{cut}}/\sigma_{\mathrm{in}} \sim10–5050 within a few R/cR/c, consistent with the larger nonthermal population injected before the merger (§ 3.2). In the weakly driven runs, γcut\gamma_{\rm cut} stays at ≲σin\lesssim\sigma_{\rm in} until the merger starts and then rises by about two orders of magnitude in a few R/cR/c, showing that particles in these runs are energized almost entirely during Phase III. Regardless of the drive, all runs saturate at a common value (γcut/σin)max⁡≃100(\gamma_{\rm cut}/\sigma_{\rm in})_{\max} \simeq100.

Post-merger spectra

Figure 7 extends Figure 5 to all runs and shows the particle energy distributions (left panel) and synchrotron spectra (right panel) at the end of the dynamical merger, trelaxt_{\rm relax}, from the weakest (orange) to the strongest (purple) drive. The distributions nearly overlap over most of the energy range, supporting analyses of § 3.2 and conclusions that the post-merger state is largely insensitive to the details of the initial driving. All runs develop a nonthermal tail between γ−1∼1\gamma- 1 \sim1 and ∼102\sim10^{2}, cut off at γ∼103\gamma\sim10^{3}. The residual dependence on vpushv_{\rm push} is a systematic softening of the tail with increasing drive, from (γ−1) dN/dγ∝γ−0.7(\gamma- 1)\,\mathrm{d}N/\mathrm{d}\gamma\propto\gamma^{-0.7} to ∝γ−1.4\propto\gamma^{-1.4}.

The synchrotron spectra show the same behavior. Below the peak, they follow the single-particle low-frequency asymptote νFν∝ν4/3\nu F_\nu\propto\nu^{4/3}. In the nonthermal range, the slopes range from νFν∝ν0.6\nu F_\nu\propto\nu^{0.6} (weak push) to ∝ν0.3\propto\nu^{0.3} (strong push), in agreement with the expected s=(3−p)/2s=(3-p)/2 for p≃1.8p\simeq1.8 and 2.4, respectively. The weakest-drive runs extend to slightly higher frequencies, consistent with their marginally harder particle spectra. Consistent with a weak-cooling regime behaviour, the synchrotron peaks track the particle energy cutoff, νpeak∼γcut∼100 σin\nu_{\rm peak}\sim\gamma_{\rm cut}\sim100\,\sigma_{\rm in}.

References

  1. [1]Abdo, A. A., Ackermann, M., Ajello, M., et al. 2011, Science, 331, 739, doi: 10.1126/science.1199705DOI
  2. [2]Abramowski, A., Acero, F., Aharonian, F., et al. 2012, ApJ, 746, 151, doi: 10.1088/0004-637X/746/2/151DOI
  3. [3]Aharonian, F., Akhperjanian, A. G., Bazer-Bachi, A. R., et al. 2006, Science, 314, 1424, doi: 10.1126/science.1134408DOI
  4. [4]Aharonian, F., Akhperjanian, A. G., Bazer-Bachi, A. R., et al. 2007, ApJL, 664, L71, doi: 10.1086/520635DOI
  5. [5]Albert, J., Aliu, E., Anderhub, H., et al. 2007, ApJ, 669, 862, doi: 10.1086/521382DOI
  6. [6]Albert, J., Aliu, E., Anderhub, H., et al. 2008, ApJL, 685, L23, doi: 10.1086/592348DOI
  7. [7]Algaba, J. C., Baloković, M., Chandra, S., et al. 2024, A&A, 692, A140, doi: 10.1051/0004-6361/202454097
  8. [8]Atwood, W. B., Abdo, A. A., Ackermann, M., et al. 2009, ApJ, 697, 1071, doi: 10.1088/0004-637X/697/2/1071DOI
  9. [9]Bacchini, F., Werner, G. R., Granier, C., & Vos, J. 2025, ApJL, 991, L9, doi: 10.3847/2041-8213/ae0197DOI
  10. [10]Balbo, M., Walter, R., Ferrigno, C., & Bordas, P. 2011, A&A, 527, L4, doi: 10.1051/0004-6361/201015980DOI
  11. [11]Beloborodov, A. M. 2017, ApJ, 850, 141, doi: 10.3847/1538-4357/aa8f4fDOI
  12. [12]Blandford, R., East, W., Nalewajko, K., Yuan, Y., & Zrake, J. 2015, arXiv e-prints, arXiv:1511.07515, doi: 10.48550/arXiv.1511.07515DOI
  13. [13]Blandford, R., Simeon, P., & Yuan, Y. 2014, Nuclear Physics B Proceedings Supplements, 256, 9, doi: 10.1016/j.nuclphysbps.2014.10.002DOI
  14. [14]Blandford, R., Yuan, Y., Hoshino, M., & Sironi, L. 2017, SSRv, 207, 291, doi: 10.1007/s11214-017-0376-2DOI
  15. [15]Blumenthal, G. R., & Gould, R. J. 1970, Rev. Mod. Phys., 42, 237, doi: 10.1103/RevModPhys.42.237DOI
  16. [19]Cerutti, B., Uzdensky, D. A., & Begelman, M. C. 2012, ApJ, 746, 148, doi: 10.1088/0004-637X/746/2/148DOI
  17. [20]Cerutti, B., Werner, G. R., Uzdensky, D. A., & Begelman, M. C. 2013, ApJ, 770, 147, doi: 10.1088/0004-637X/770/2/147DOI
  18. [21]Cerutti, B., Werner, G. R., Uzdensky, D. A., & Begelman, M. C. 2014, ApJ, 782, 104, doi: 10.1088/0004-637X/782/2/104DOI
  19. [23]Chernoglazov, A., Hakobyan, H., & Philippov, A. 2023, ApJ, 959, 122, doi: 10.3847/1538-4357/acffc6DOI
  20. [24]Comisso, L., & Jiang, B. 2023, ApJ, 959, 137, doi: 10.3847/1538-4357/ad1241DOI
  21. [25]Comisso, L., Sobacchi, E., & Sironi, L. 2020, ApJL, 895, L40, doi: 10.3847/2041-8213/ab93dcDOI
  22. [26]Deka, P. J., Bacchini, F., Zhou, M., & Granier, C. 2026, arXiv e-prints, arXiv:2605.20318, doi: 10.48550/arXiv.2605.20318DOI
  23. [27]Gebhardt, K., Adams, J., Richtstone, D., et al. 2011, ApJ, 729, 119, doi: 10.1088/0004-637X/729/2/119DOI
  24. [28]Giannios, D. 2008, A&A, 480, 305, doi: 10.1051/0004-6361:20070985
  25. [29]Goodman, J., & Uzdensky, D. 2008, ApJ, 688, 555, doi: 10.1086/592345DOI
  26. [30]Granier, C., Bacchini, F., Groselj, D., & Sironi, L. 2026, 3D Kinetic Simulations of Driven Reconnection in Merging Flux Tubes, https://arxiv.org/abs/2604.10431
  27. [31]Granier, C., Gröselj, D., Comisso, L., & Bacchini, F. 2025, ApJ, 992, 193, doi: 10.3847/1538-4357/ae0738DOI
  28. [33]Guépin, C., Cerutti, B., & Kotera, K. 2020, A&A, 635, A138, doi: 10.1051/0004-6361/201936816DOI
  29. [34]Hakobyan, H., Levinson, A., Sironi, L., Philippov, A., & Ripperda, B. 2025, ApJL, 995, L73, doi: 10.3847/2041-8213/ae286aDOI
  30. [35]Hakobyan, H., Petropoulou, M., Spitkovsky, A., & Sironi, L. 2021, ApJ, 912, 48, doi: 10.3847/1538-4357/abedacDOI
  31. [38]Hakobyan, H., Spitkovsky, A., Chernoglazov, A., et al. 2023, TRISTAN-MP v2.6, Princeton University, doi: 10.5281/zenodo.7566725
  32. [39]Hayashida, M., Nalewajko, K., Madejski, G. M., et al. 2015, ApJ, 807, 79, doi: 10.1088/0004-637X/807/1/79DOI
  33. [40]Landau, L. D., & Lifshitz, E. M. 1975, Course of Theoretical Physics, Vol. 2, The Classical Theory of Fields (Oxford: Pergamon Press)
  34. [41]Lyutikov, M. 2006, New Journal of Physics, 8, 119, doi: 10.1088/1367-2630/8/7/119DOI
  35. [42]Lyutikov, M., Sironi, L., Komissarov, S. S., & Porth, O. 2017, Journal of Plasma Physics, 83, 635830602, doi: 10.1017/S0022377817000071X
  36. [43]Mahlmann, J. F., Philippov, A. A., Mewes, V., et al. 2023, ApJL, 947, L34, doi: 10.3847/2041-8213/acacda
  37. [44]Parfrey, K., Beloborodov, A. M., & Hui, L. 2013, ApJ, 774, 92, doi: 10.1088/0004-637X/774/2/92DOI
  38. [45]Peng, C.-F., Lu, R.-J., Liang, W.-Q., & Chen, Z.-g. 2024, ApJ, 973, 62, doi: 10.3847/1538-4357/ad65d6DOI
  39. [46]Rani, B., Lott, B., Krichbaum, T. P., Fuhrmann, L., & Zensus, J. A. 2013, A&A, 557, A71, doi: 10.1051/0004-6361/201321440DOI
  40. [48]Rieger, F. M. 2019, Galaxies, 7, 28, doi: 10.3390/galaxies7010028DOI
  41. [49]Rieger, F. M., & Levinson, A. 2018, Galaxies, 6, 116, doi: 10.3390/galaxies6040116DOI
  42. [50]Ripperda, B., Bacchini, F., & Philippov, A. A. 2020, ApJ, 900, 100, doi: 10.3847/1538-4357/abababDOI
  43. [51]Ripperda, B., Porth, O., Sironi, L., & Keppens, R. 2019, MNRAS, 485, 299, doi: 10.1093/mnras/stz387DOI
  44. [52]Ripperda, B., Porth, O., Xia, C., & Keppens, R. 2017, MNRAS, 467, 3279, doi: 10.1093/mnras/stx379DOI
  45. [53]Rybicki, G. B., & Lightman, A. P. 1985, Radiative Processes in Astrophysics (New York: Wiley-VCH)DOI
  46. [56]Sironi, L., & Spitkovsky, A. 2014, ApJL, 783, L21, doi: 10.1088/2041-8205/783/1/L21DOI
  47. [57]Sironi, L., Uzdensky, D. A., & Giannios, D. 2025, ARA&A, 63, 127, doi: 10.1146/annurev-astro-020325-115713DOI
  48. [58]Striani, E., Tavani, M., Piano, G., et al. 2011, ApJL, 741, L5, doi: 10.1088/2041-8205/741/1/L5DOI
  49. [59]Tavani, M., Barbiellini, G., Argan, A., et al. 2009, A&A, 502, 995, doi: 10.1051/0004-6361/200810527DOI
  50. [60]Tavani, M., Bulgarelli, A., Vittorini, V., et al. 2011, Science, 331, 736, doi: 10.1126/science.1200083DOI
  51. [61]Velberg, A., Stanier, A., Li, X., et al. 2025, arXiv e-prints, arXiv:2511.15988, doi: 10.48550/arXiv.2511.15988DOI
  52. [62]Werner, G. R., Philippov, A. A., & Uzdensky, D. A. 2019, MNRAS, 482, L60, doi: 10.1093/mnrasl/sly157DOI
  53. [63]Werner, G. R., & Uzdensky, D. A. 2017, ApJL, 843, L27, doi: 10.3847/2041-8213/aa7892DOI
  54. [64]Werner, G. R., & Uzdensky, D. A. 2024, ApJL, 964, L21, doi: 10.3847/2041-8213/ad2fa5DOI
  55. [65]Werner, G. R., Uzdensky, D. A., Cerutti, B., Nalewajko, K., & Begelman, M. C. 2016, ApJL, 816, L8, doi: 10.3847/2041-8205/816/1/L8DOI
  56. [66]Xrism Collaboration, Audard, M., Awaki, H., et al. 2025, Nature, 641, 1132, doi: 10.1038/s41586-025-08968-2DOI
  57. [67]Yuan, Y., Nalewajko, K., Zrake, J., East, W. E., & Blandford, R. D. 2016, ApJ, 828, 92, doi: 10.3847/0004-637X/828/2/92DOI
  58. [68]Zhou, M., Loureiro, N. F., & Uzdensky, D. A. 2020, Journal of Plasma Physics, 86, 535860401, doi: 10.1017/S0022377820000641DOI

Paper details

vlnc:W_9eg9zkde
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents