Testing CaII K chromospheric flare precursors with spectropolarimetric inversions and data-constrained LARE2D simulations

1 Introduction

Solar flares have been recognized since the first white-light observations by Carrington and Hodgson in 1859 as among the most energetic explosive phenomena in the solar atmosphere (Carrington, 1859; Hodgson, 1859). They release approximately 1 0 29–1 0 32 erg over minutes to hours, primarily through the rapid conversion of stored magnetic energy in stressed coronal fields, a picture developed through early magnetic flare theories and reconnection models (Giovanelli, 1946; Sweet, 1958; Parker, 1957; Petschek, 1964; Carmichael, 1964; Sturrock, 1966; Hirayama, 1974; Kopp and Pneuman, 1976) and reviewed in modern treatments (Fletcher et al., 2011; Shibata and Magara, 2011; Benz, 2017). Recent studies have continued to extend this picture through both observational and modelling approaches (Babu et al., 2025; Babu et al., 2026).

The energy liberated drives intense heating of the chromosphere and corona, non-thermal particle acceleration, and large-scale mass motions, with the reconnected field lines retracting to form a post-flare arcade of hot loops above the polarity-inversion line (PIL; Moore et al., 1987; Kopp and Pneuman, 1976; Priest and Forbes, 2002). The chromospheric footprints of this process, such as the flare ribbons, are among the most well-studied observational signatures, tracing where energetic electrons and conduction fronts deposit their energy into the denser lower atmosphere (Fletcher et al., 2011; Milligan, 2015; Druett et al., 2017). The practical consequences of large flares extend to solar energetic particle (SEP) events, geomagnetic storms, and radio blackouts, all of which motivate a sustained effort to identify reliable pre-flare indicators (Toriumi and Wang, 2019; Georgoulis et al., 2021).

The search for early warning signatures has proceeded along two complementary lines: statistical characterization of pre-eruptive magnetic configurations and direct detection of chromospheric precursor emission. On the magnetic side, large-scale photospheric vector-magnetogram surveys using SDO/HMI have demonstrated that active regions with high magnetic flux, strong polarity-inversion-line gradients, and elevated non-potentiality metrics are more likely to flare (Leka and Barnes, 2003; Leka and Barnes, 2007; Bobra and Couvidat, 2015; Florios et al., 2018). Machine-learning classifiers trained on these descriptors now achieve moderate operational skill for ≥ M-class events, but face a reproducible ceiling when photospheric data alone are used (Nishizuka et al., 2017; Deshmukh et al., 2023): the photosphere encodes the boundary condition, not the free energy available in the corona above it. On the chromospheric side, a growing body of evidence shows that compact brightenings, plasma jets, and velocity transients near PILs often precede the impulsive phase by several to tens of minutes (Chifor et al., 2007). The Interface Region Imaging Spectrograph (IRIS; De Pontieu et al., 2014) observations of the Mg II and Si IV lines have further demonstrated that spectral profiles at flare-ribbon sites separate from their quiescent counterparts well before any detectable soft X-ray rise, providing a chromospheric diagnostic with genuine predictive content (Panos et al., 2018; Panos and Kleint, 2020; Woods et al., 2021; Panos et al., 2023; Zbinden et al., 2024). More recently, Kumar et al. (2026), using eight flare sequences from the Rapid Oscillations in the Solar Atmosphere (ROSA; Jess et al., 2010) instrument at the Dunn Solar Telescope (DST) showed that compact CaIIK brightenings adjacent to eventual flare kernels peak 10–45 min before the primary-kernel brightening and the associated GOES 1–8 Å. This operationally simple, ground-based diagnostic requires no spectrograph and provides a practical short-lead precursor with demonstrated reproducibility across multiple event classes. Whether such chromospheric precursors are active participants in flare initiation or passive early signatures of energy release remains an open question. In the passive interpretation, a precursor brightening simply traces low-level heating from the earliest stages of coronal reconnection, and it carries predictive value without any causal role. The statistical and machine-learning studies of IRIS spectra fall largely in this category. They demonstrate that pre-flare spectral profiles separate measurably from quiescent ones (Panos et al., 2018; Panos and Kleint, 2020; Woods et al., 2021; Panos et al., 2023; Zbinden et al., 2024), but they do not establish whether the emitting plasma influences the eruption itself. The active interpretation instead treats localized low-atmosphere disturbances as triggers that push an already stressed field past its stability threshold.

Theoretical support for this view may comes from the tether-cutting scenario, in which small-scale reconnection low above the PIL initiates the main explosion (Moore et al., 2001), from the trigger-field model of Kusano et al. (2012), in which compact bipolar intrusions of specific orientation at the PIL destabilize the overlying arcade, and from flux-emergence simulations in which newly arriving flux perturbs the pre-eruptive equilibrium (Archontis and Hood, 2012). Chifor et al. (2007) found X-ray precursors co-spatial with the eventual eruption onset site and interpreted them as small-scale reconnection associated with filament destabilization. Bamba et al. (2013) and Bamba et al. (2017) identified compact chromospheric brightenings at exactly the trigger regions predicted by the Kusano model in X-class flares. These studies established a spatial and temporal association between compact low-atmosphere activity and flare onset. What has rarely been tested is energetic sufficiency, that is, whether a perturbation with the observed amplitude, depth, and duration can in fact destabilize the specific stressed configuration above it. We use the Lagrangian Remap code LARE2D (Arber et al., 2001), which solves the resistive magnetohydrodynamic equations on a two-dimensional Eulerian grid with a staggered mesh and operator-split Lagrangian remap, and is well suited to modelling eruption onset in stratified atmospheres (Archontis and Hood, 2012; Leake et al., 2013). Previous studies with LARE2D and similar resistive MHD frameworks have shown that a pre-stressed coronal arcade near the torus-instability threshold (Kliem and Török, 2006) can be pushed into eruption by low-altitude perturbations. If the thermal energy deposited in the chromosphere by the CaII K precursor is sufficient to perturb the overlying, already-stressed arcade beyond that threshold, the resulting reconnection would proceed from the bottom up, consistent with the observed lead times. This is the approach adopted in this work.

Full-Stokes spectropolarimetry, combined with inversion of the polarised radiative transfer equation, offers a direct route to this stratified picture. By recording the Stokes ( I , Q , U , V ) profiles of multiple spectral lines simultaneously, one recovers depth-stratified profiles of the magnetic field vector, temperature, and LOS velocity from the deep photosphere into the upper chromosphere in a single observational scan (Ruiz Cobo and del Toro Iniesta, 1992; Del Toro Iniesta and Ruiz Cobo, 1996). Ground-based instrumentation at the Dunn Solar Telescope (DST), in particular the Spectropolarimeter for Infrared and Optical Regions (SPINOR; Socas-Navarro et al., 2006), facilitates the simultaneous observation of the photospheric FeI 5,250 Å and chromospheric CaII 8,542 Å lines, thereby providing a vertically extended sampling of the solar atmosphere at each slit position. Previous work using comparable spectropolarimetric observations has detected chromospheric magnetic-field changes during flares (Kleint, 2017) and used full-Stokes CaII 8,542 Å inversions to investigate the chromospheric magnetic field in an M-class flare (Kuridze et al., 2018), but systematic pre-versus-post-flare comparisons at multiple ribbon footpoints spanning both magnetic polarities have not been published.

In this work, we test whether the compact CaIIK precursor observed before the 2025 June 13 GOES C3.1 flare represents a physically meaningful low-atmosphere trigger rather than only an early brightening. The event was accompanied by chromospheric jets and a precursor brightening previously identified by Kumar et al. (2026). We use simultaneous SPINOR FeI 5,250 Å and CaII 8,542 Å St profiles to infer the depth-dependent magnetic field strength | B |, line-of-sight velocity v LOS, and temperature T at five flare-ribbon footpoints spanning both magnetic polarities. By comparing matched pre-flare and post-flare epochs, we examine how the photosphere–chromosphere system responds at individual footpoints and whether those changes are consistent with reconnection-driven back-reaction, chromospheric flows, and localized heating. We then use the LTE-derived chromospheric temperature enhancement at the precursor site, Δ T ∼ 2000 K, to constrain a LARE2D MHD experiment that tests whether such a localized perturbation can drive reconnection in a pre-stressed arcade. The subsequent NLTE analysis gives a representative enhancement of approximately Δ T ∼ 1000 K; we therefore test amplitudes spanning Δ T = 500–2000 K to determine whether the trigger depends on the inversion treatment or the precise thermal amplitude. The paper is organised as follows. Section 2 describes the observations, inversion methodology, and LARE2D simulation setup. Section 3 presents the atmospheric stratifications, footpoint-dependent changes, and simulation results. Sections 4 and 5 summarize the main conclusions of the work.

2 Observations and methods

Figure 1 summarises the observational context for the 2025 June 13 GOES C3.1 flare. The event was observed with ROSA at the DST and was previously identified by Kumar et al. (2026) as one of the events showing a compact CaII K precursor before the main flare kernel. Here we use that prior CaII K timing result as context and focus on the new spectropolarimetric and numerical analysis of this event. Panel (A) shows a ROSA CaII K context image at 13:31 UT, approximately 27 min before the impulsive flare peak. The marked regions show the eventual flare-ribbon footpoints, the jet-associated trigger region, and the compact precursor/flare-trigger footpoints. Panel (B) compares the GOES 1–8 Å soft X-ray flux with the detrended CaII K light curve from the jet-trigger site. The CaII K brightening peaks approximately 20 min before a clear GOES rise and about 27 min before the impulsive flare peak at 13:57 UT.

The magnetic context is also taken from the Stokes-V-anchored flux-proxy connectivity analysis of Kumar et al. (2026). That analysis used signed point sources at the identified footpoint centroids to trace approximate source-to-sink streamlines in the plane of the sky. The method was used only as a geometric connectivity proxy. The pre-flare proxy field showed compact closed connections linking the precursor, jet, and flare-ribbon footpoints, while the post-flare proxy field showed rerouting toward more open or extended trajectories. In the present paper, this prior CaII K and connectivity information is used only to define the event context and the footpoint locations. The new analysis focuses on the depth-dependent SPINOR/HAZEL inversions and on a data-constrained LARE2D trigger experiment for the same event.

2.1 SPINOR spectropolarimetric observations

Full-Stokes spectropolarimetric observations were obtained with SPINOR (Socas-Navarro et al., 2006) at the Dunn Solar Telescope. Two spectral channels were recorded simultaneously: the primarily photospheric Fe I 5,250 Å line ( g eff = 3.0 ) and the chromospheric CaII 8,542 Å line ( g eff = 1.1 ). The active region was located at approximately ( 75 0 ″ , − 22 0 ″ ), corresponding to μ = 0.52. Three slit-scan epochs are available: a pre-flare scan from 13:33:00–13:53:20 UT with 300 slit steps (Δ t step ≈ 5.3 s), an impulsive-phase scan from 13:57:43–14:00:05 UT with 26 steps and a limited field of view, and a post-flare scan from 14:05:43–14:27:09 UT with 300 slit steps. The data were reduced with the standard SPINOR pipeline, including dark- and flat-field correction, polarimetric calibration, and continuum normalization to the local limb-darkening-corrected quiet-Sun intensity I c (Sellers et al., 2025). The limb-darkening correction was computed using the tabulated values of Allen (1973).

Figure 2 shows the two-dimensional SPINOR Stokes maps for the 2025 June 13 event in Fe I 5,250.2 Å. Stokes I provides the intensity context and separates the dark umbra from the surrounding penumbra. Stokes Q and U show the linear polarization signal, which is strongest in the penumbral region and is consistent with a more inclined magnetic field. Stokes V traces the line-of-sight magnetic polarity, with red and blue regions marking opposite-polarity structures within and around the sunspot. We selected five footpoints (FPs) from the enhanced Stokes-V signal associated with the flare-ribbon locations. FP 1 and FP 2 lie in the negative-polarity umbra/penumbra, while FP 3–FP 5 lie in the positive-polarity region. A timing caveat is important for interpreting the pre-flare scan. SPINOR builds the map sequentially as the slit scans across the field of view, so footpoints at different solar-X positions are not observed at the same time. The elapsed time between FP 1 and FP 5 in the pre-flare scan is ≈ 11.4 min. As a result, FP 3–FP 5 were observed only 5–13 min before flare onset and may already include precursor or early-phase signatures. The observation times for each footpoint are listed in Table 1.

TABLE 1. Footpoint positions and observation times. Coordinates are in the common (ROSA-anchored) helioprojective frame of Figures 1, 2, accurate to ∼ 2 ″. Δ t values are relative to the flare onset time of 13:57:43 UT.

Possible flare precursor conditions.

2.2 St inversion with HAZEL

Physical parameters were retrieved from the observed Stokes profiles using the HAnle and ZEeman Light inversion code (HAZEL, Asensio Ramos et al., 2008). The inversion framework uses polarized radiative transfer calculations to iteratively adjust a depth-stratified model atmosphere until the synthetic Stokes profiles reproduce the observed Stokes I, Q, U, and V spectra. For the photospheric diagnostic, we use the SIR-based inversion approach (Ruiz Cobo and del Toro Iniesta, 1992), in which the atmosphere is described on a log ⁡ τ scale and the depth variation of each atmospheric quantity is controlled by a small number of nodes. The full atmospheric profile is then obtained by interpolation between nodes. This approach regularizes the inversion while retaining sensitivity to vertical gradients in temperature, magnetic field, and line-of-sight velocity.

Before inversion, the Stokes profiles at each footpoint were spatially averaged over a 5 × 5 pixel box centered on the selected footpoint. This averaging improves the signal-to-noise ratio of the weak linear polarization signals, Stokes Q and U, while preserving the local footpoint structure. At μ = 0.52, the box corresponds to a projected area of roughly 360 × 360 km 2 on the solar surface. The noise level was estimated from the standard deviation of Stokes I, Q, U, and V in the continuum windows over a 50 × 50 pixel quiet-Sun region away from the active region. Typical 1 σ noise levels were ≈ 10.4 % I c for Fe I 5,250 and ≈ 6.6 % I c for CaII 8,542 in Stokes I, and 0.5–1.5 % I c in Stokes Q, U, and V. The same noise estimate was used as the weighting vector in the inversion for each footpoint.

Two spectral diagnostics were inverted independently for each footpoint. The photospheric diagnostic used the FeI 5,250.2 Å ( g eff = 3.0 ) and FeI 5,250.6 Å ( g eff = 1.5 ) line blend, fitted over the wavelength range 5,249.7–5,251.2 Å. These lines form mainly over log ⁡ τ ≈ 0 to − 2, with the strongest sensitivity near log ⁡ τ ≈ − 1 (Ruiz Cobo and del Toro Iniesta, 1992). The Harvard-Smithsonian Reference Atmosphere (HSRA; Gingerich et al., 1971) was used as the initial photospheric model. The continuum intensity I c was measured in the clean 5,252.25–5,252.75 Å window, corrected for limb darkening at μ = 0.52 using the tabulated values of Allen (1973), and used to normalise all four St parameters. The main line properties, fitting windows, approximate sensitivity ranges, and initial model atmospheres are summarised in Table 2.

TABLE 2. Summary of spectral lines and inversion setup.

The chromospheric diagnostic used the Caii 8,542.09 Å line ( g eff = 1.1 ) and was fitted over 8,540.0–8,544.4 Å. The wings of this line sample the photosphere. Its core forms in the chromosphere (log ⁡ τ ≈ − 4 to − 6; de la Cruz Rodríguez et al., 2012), where densities are too low for collisions to maintain local thermodynamic equilibrium (Leenaarts et al., 2009). Each footpoint profile was therefore inverted twice with the SIR engine inside Hazel2, once in LTE mode and once in NLTE mode. In the LTE mode, the CaII level populations follow Saha–Boltzmann statistics of the local temperature. This ties the line source function to the local Planck function at all depths. In the NLTE mode, a neural-network emulator predicts the departure coefficients of the CaII levels for the current model atmosphere at every iteration of the inversion (see details in Vicente Arévalo et al., 2022). The two inversions are otherwise identical. They use the same 8,540.0–8,544.4 Å window (photospheric blends at 8,540.76 and 8,543.79 Å masked), the same node schedule and six-cycle strategy, the same initial FAL-C chromospheric model (Fontenla et al., 1993), and the same continuum normalization. Any difference between the retrieved atmospheres is therefore attributable to the line-formation physics alone.

Fit quality alone cannot arbitrate between the two treatments: χ 2 depends on the noise model, and the NLTE forward problem is the more flexible of the two. We therefore evaluate the comparison in three complementary ways. First, the fits are inspected profile by profile at all five footpoints and both epochs (Figure 3). Particular attention is paid to whether each treatment reproduces the central emission reversals that develop at the flaring footpoints. Second, the retrieved temperature stratifications are correlated against the LTE-valid Fei 5,250 inversions over the photospheric overlap ( log ⁡ τ ∈ [ − 2,0 ] ). The two spectral diagnostics must agree there regardless of the chromospheric treatment, so this serves as an independent anchor. Third, the retrieved temperatures and line-of-sight velocities are compared at fixed reference depths, log ⁡ τ = 0 and − 3, across all footpoints and epochs (Table 3). The outcome of this comparison is presented in Section 3, together with its one quantitative consequence for the simulation input.

TABLE 3. Temperatures and line-of-sight velocities (LTE/NLTE) at two reference depths. Negative velocities are blueshifts. Field components are bounded at | B x , y , z | ≤ 500 G.

2.3 Simulation setup using LARE2D

We use the LARE2D code (Arber et al., 2001) to test whether the chromospheric energy release inferred from the HAZEL inversions is sufficient to trigger the eruption of a pre-stressed coronal arcade. LARE2D solves the two-dimensional compressible resistive MHD equations on a Lagrangian-remap grid; the governing equations, normalization, and numerical scheme are described in full by Arber et al. (2001), and we refer the reader there for details. Our domain spans x ∈ [ − 30,30 ] Mm across the PIL and y ∈ [ 0,40 ] Mm from the photospheric base to the low corona, discretized on a uniform 1024 × 1024 mesh. The normalizations are set by the observed active-region properties: B n o r m = 100 G from the HAZEL-inferred upper-chromospheric field strengths at the ribbon footpoints (Section 3), ρ n o r m = 1.67 × 1 0 − 8 kg m − 3 from the FAL-C quiet-Sun chromosphere at log ⁡ τ ≈ − 4 (Fontenla et al., 1993), and L n o r m = 1 Mm (Arber et al., 2001). These give an Alfvén speed v A ≈ 69 km s − 1 and crossing time t n o r m ≈ 14.5 s. The initial magnetic configuration is a force-free sheared arcade above the PIL, with an amplitude B 0 = 200 G, half-width L a = 15 Mm, and shear parameter α s = 0.90, placing the arcade just below the torus-instability threshold (Kliem and Török, 2006) and consistent with the HAZEL-inferred active-region field geometry. The photospheric field strengths retrieved at log ⁡ τ = − 1 span approximately 0.4–0.87 kG and describe compact flux concentrations a few hundred kilometers across, whereas B 0 represents the large-scale arcade field at the chromospheric and coronal heights included in the model. The normalization B n o r m = 100 G is therefore adopted as a representative model scaling rather than being taken directly from the inverted chromospheric field strengths, which carry substantial cycle-to-cycle scatter and are additionally limited by the bound imposed on the field components (Section 3). The arcade amplitudeB 0 = 2 B n o r m = 200 G is lower than the compact photospheric field strengths, as expected when photospheric flux expands into the larger-scale arcade. A systematic sweep in B 0, which would alter the ratio of perturbation energy to magnetic energy through the plasma-β, is left for future work.

The CaII K precursor brightening is introduced as a localized perturbation to the specific internal energy at the PIL,

δ ε = Δ T γ − 1 T n o r m exp − x 2 σ x 2 − y − y 0 2 σ y 2 , (1)

centered at y 0 = 3.0 Mm (upper chromosphere) with σ x = 8.0 Mm and σ y = 1.5 Mm. The amplitude Δ T = 2 000 K is taken directly from the HAZEL inversion of the pre-flare brightening (Section 3), making equation (Equation 1) an observationally constrained input rather than a free parameter.

The CaII,K light curve at the active-region PIL shows three temporally distinct phases: a burst of jet activity peaking ≈ 20 min before the flare, a waning of that activity as the arcade stores sufficient free energy to erupt, and a near-background floor after the GOES peak. We reproduce this temporal structure with a converging footpoint shear and co-spatial thermal injection applied directly to the chromospheric interior cells. The shear profile is given by Equation 2 as

δ v x = − v e f f t sin π x x m a x G y y d t , (2)

is antisymmetric about the PIL, pushing field lines from both sides toward the forming current sheet without violating the line-tied photospheric boundary.

Here v e f f has units of v A t n o r m − 1 (an acceleration amplitude, not a velocity): at each timestep the code increments v x by v e f f × d t, so the equilibrium footpoint shear speed reached by the plasma is sub-Alfvénic, set by the balance between the driver and the magnetic tension restoring force. Both the shear and the co-spatial thermal source are spatially confined by Gaussian envelopes centred at y 0 = 3.0 Mm (σ x = 8.0 Mm, σ y = 1.5 Mm). To avoid a nonphysical step-function onset the amplitude is ramped as v e f f ( t ) = v p h a s e ( t ) tanh ( t / t r a m p ) with t r a m p = 10 t n o r m ≈ 145 s. The three-phase target acceleration is:

v p h a s e t = 2.0 v A t n o r m − 1 ≈ 9.5 k m s − 2 t ≤ 83 t n o r m j e t p h a s e , 0 – 20 m i n , 1.0 v A t n o r m − 1 ≈ 4.8 k m s − 2 83 < t ≤ 124 t n o r m w a n i n g , 20 – 30 m i n , 0.2 v A t n o r m − 1 ≈ 0.95 k m s − 2 t > 124 t n o r m g r a d u a l , > 30 m i n . (3)

The Phase 1 boundary at 20 min is set by the observed CaII K lead time (Kumar et al., 2026); the amplitude v 1 = 2.0 v A ensures that the PIL current repeatedly exceeds the reconnection threshold, reproducing the bursty jet reconnection seen in the CaII K context movies. Phase 2 begins when the imposed driver is reduced by 50 % (20 min), reflecting the observed drop in CaII K intensity, and ends at the observed flare onset (30 min); this tests whether the eruption is self-sustaining once triggered. The Phase 3 floor ( 0.2 v A ) represents residual chromospheric footpoint motions during the gradual phase. The co-spatial thermal source follows the same three-phase envelope with Phase 1 amplitude E 1 = 0.040 (normalized), consistent with Δ T ≈ 2 000 K above background. Furthermore, Phase 1 is the interval of strongest chromospheric forcing, corresponding to the observed period of enhanced CaII K jet and precursor activity before the main soft X-ray flare rise; Phase 2 tests whether the magnetic system continues to reconnect after the strongest forcing has weakened; and Phase 3 is the gradual post-flare stage, in which only a weak residual driver remains and the arcade is allowed to relax.

Reconnection is enabled by a current-threshold anomalous resistivity, η = 5 × 1 0 − 3 for | J | > j m a x = 16 and zero otherwise (Priest and Forbes, 2002). The threshold is calibrated so that the unperturbed arcade current (| J z | ≈ 13–14) remains in the ideal regime; only the PIL-concentrated current driven by the shear velocity genuinely exceeds j m a x, giving a Lundquist number S ≈ 200 in the reconnection layer, consistent with Petschek-type fast reconnection in chromospheric current sheets (Petschek, 1964; Yokoyama and Shibata, 1995). The simulation runs to t e n d = 300 t n o r m ≈ 72.5 min, producing 101 snapshots at 3 t n o r m ≈ 43 s intervals. Key parameters are collected in Table 4.

TABLE 4. LARE2D simulation parameters.

Based on the NLTE analysis described in (Section 2.2), additional revisions to both parameters of (Equation 1) are made: the precursor enhancement is Δ T ≈ 1000 K (range ≈500–1400 K) rather than 2000 K, and it is measured at log ⁡ τ ≈ − 3 to − 4, which in the FAL-C stratification corresponds to y ≈ 0.8–1.5 Mm (mid-chromosphere) rather than the upper-chromospheric y 0 = 3 Mm adopted above. We therefore repeated the experiment with the energy injection centered at y 0 = 1.25 Mm and restricted to σ y = 0.5 Mm, ensuring that the deposited energy remains confined within the empirically inferred formation layer and does not penetrate into the photosphere, for temperature increments of Δ T = 500,1000,2000 K, while holding all other parameters constant. Physically, this is a more demanding test of the trigger scenario: because the thermal source and the converging shear share the same vertical envelope (Equations 1–3), the entire driver now acts on substantially denser plasma, where a given energy and velocity input is harder to convert into magnetic stress above the arcade. The deeper, narrower injection is nearly energy-matched to the original (the higher ambient density is offset by the smaller injection volume; the total deposited thermal energy ratio is 1.14 for Δ T = 1000 K), so any change in outcome is expected to isolate the effect of placement in the stratification rather than that of injected energy.

3 Results

3.1 Pre-flare and post-flare atmospheric stratification

Figures 4, 5 show the depth-stratified temperature T, total magnetic field strength | B |, and LOS velocity v LOS for all five footpoints in the Fe I 5250 Å (photosphere, top rows) and CaII 8542 Å (chromosphere, bottom rows) lines at the pre- and post-flare epochs, respectively. Quoted values are extracted at reference depths log ⁡ τ = − 1 (Fe I) and log ⁡ τ = − 3 (CaII); FPs 3–5 were observed only 5–13 min before flare onset, and their pre-flare values may already reflect precursor conditions.

In the pre-flare photosphere (Fe I 5250, log ⁡ τ = − 1), temperatures cluster in the range 4950–5350 K across all five footpoints, with no significant polarity-dependent offset. The photospheric field strengths span approximately 0.4–0.87 kG, with FP,1 and FP,4 returning the strongest fields, FP,2 an intermediate field strength, and FP,3 and FP, 5 the weakest. The distribution, therefore, shows no simple ordering by magnetic polarity. Photospheric line-of-sight velocities at this depth are small at every footpoint (| v | ≤ 2.2 km, s − 1), consistent with quiescent granular motions rather than a coherent pre-flare photospheric flow (Figure 4).

In the chromosphere (CaII 8542, log ⁡ τ = − 3) the footpoints are hotter, spanning 5440–8190 K. The positive-polarity ribbon footpoints FP 5 and FP 3 are the warmest, at 8190 and 7088 K, consistent with early ribbon-site heating already underway 5–13 min before onset (Bamba et al., 2017). With the field components bounded at 500 G, the retrieved chromospheric field strengths lie between 510 and 865 G at all five footpoints. The line-of-sight velocities are small. At log ⁡ τ = − 1, where the line is best constrained and the cycle-to-cycle scatter is smallest, FP 1 returns − 3.4 km s − 1 while the remaining footpoints lie between − 0.5 and − 0.1 km s − 1. FP 1, the jet-precursor site, is therefore the only footpoint carrying a measurable pre-flare velocity signal, and it is a blueshift. The same blueshift is recovered at log ⁡ τ = − 3 (− 4.2 km s − 1) and persists into the post-flare map (− 3.5 km s − 1), so it is stable in both depth and time. Its amplitude is comparable to the cycle-to-cycle scatter, so we treat it as suggestive rather than conclusive. Because μ = 0.52 and the field geometry at each footpoint is unknown, we quote line-of-sight velocities throughout and do not convert them into vertical flows.

The chromospheric stratification at FP 5 differs from the other footpoints, reflecting its location as much as its timing. FP 5 lies closest to the eventual flare ribbon, where precursor energy deposition concentrates as non-thermal particles or conduction fronts arrive along newly reconnecting field lines. It was also scanned only 5.4 min before onset, the latest of the five footpoints. Two signatures point to a single interpretation: its chromospheric temperature is the highest in the sample, near 8190 K at log ⁡ τ = − 3, and its CaII 8542 Å core shows a central emission reversal. Together, these indicate that FP 5 provides the earliest direct spectropolarimetric snapshot of ribbon-site heating in this event, already underway before any detectable GOES rise. The line-of-sight velocity does not contribute to this picture. It is a small redshift at both epochs, + 4.1 and + 2.0 km s − 1 at log ⁡ τ = − 3.

Figure 5 shows the corresponding post-flare stratifications. At log ⁡ τ = − 1, the signed total-field changes are approximately + 3, + 24, + 1, 0, and − 14 G at FP 1–FP 5, respectively. Thus, the photospheric field changes do not exceed approximately 24 G: FP 1–FP 3 strengthen slightly, FP 4 is unchanged, and only FP 5 weakens. The chromospheric response is larger and is not uniform across the sample. The trigger footpoints FP 3 and FP 4 heat by + 260 and + 190 K at log ⁡ τ = − 3, while FP 1, FP 2, and FP 5 cool by − 318, − 168, and − 626 K respectively. With the exception of FP 2, none of these changes exceeds its cycle-to-cycle uncertainty, so we do not interpret them individually or infer a polarity-dependent heating pattern from them. The chromospheric field strength increases by + 325 G at FP 3 and decreases by − 319 G at FP 5, with changes below 50 G at the remaining three footpoints. We note that the retrieved B z at FP 3, FP 4, and FP 5 lies at or close to the ± 500 G bound imposed on the field components, so the chromospheric field changes at these footpoints should be regarded as lower limits. The blueshift at the jet-precursor site FP 1 persists into the post-flare map, decreasing from − 3.4 to − 2.3 km s − 1 at log ⁡ τ = − 1, while the other four footpoints stay within ± 1.1 km s − 1 at both epochs.

Figure 6 summarizes the signed changes Δ ( p o s t − p r e ) across all five footpoints. At log ⁡ τ = − 3, the two treatments agree in velocity sign at FPs 1–4 but differ substantially in amplitude. At FP 3 and FP 4, the LTE inversion returns + 6.4 and + 11.6 km s − 1, compared with + 1.4 and+ 2.5 km s − 1 in NLTE. At FP 5, the treatments differ in sign, with LTE returning − 11.3 km s − 1 and NLTE returning+ 4.1 km s − 1. Because FP 5 has the most complex profile and the poorest fit in the sample, its chromospheric velocity is the least well determined, and we do not interpret its sign.

3.2 LTE versus NLTE

Under the common 500 G bound on the field components, both LTE and NLTE treatments reproduce the observed profiles well. The best-cycle χ 2 values lie between 0.5 and 1.7 at FPs 1–4 for both treatments, increasing at FP 5 to 5.0 in LTE and 5.5 in NLTE. Fit quality, therefore, does not discriminate between the treatments. The photospheric anchor test also shows no consistent preference: correlations with the LTE-valid FeI 5250 Å inversions over log ⁡ τ ∈ [ − 2,0 ] give mean Pearson coefficients of r = 0.71 (LTE) and 0.76 (NLTE) before the flare, and 0.58 and 0.39, respectively, after the flare. Thus, NLTE gives the higher coefficient before the flare, whereas LTE gives the higher coefficient afterward. The treatments differ more strongly in the retrieved chromospheric stratifications. At the absorption-control footpoint FP 2, the pre-flare LTE–NLTE temperature difference at log ⁡ τ = − 3 is only − 26 ± 111 K. The corresponding LTE/NLTE line-of-sight velocities at log ⁡ τ = 0 are approximately − 17.0 km s − 1 before the flare and + 2.0 km s − 1 afterward. At log ⁡ τ = − 1, the velocity magnitudes from both treatments are below 1 km s − 1. At the remaining footpoints, the absolute LTE–NLTE temperature differences at log ⁡ τ = − 3 span 823–2214 K before the flare and 765–1893 K afterward; at log ⁡ τ = 0, the corresponding ranges are 94–241 K and 37–564 K, respectively. The largest differences, therefore, occur at chromospheric depths and coincide with profiles whose line cores depart from simple absorption. For the quantitative comparison below, we use the NLTE solutions as the reference, following the line-formation treatment described in Section 2.2. Relative to FP 2 at log ⁡ τ = − 3, the pre-flare NLTE temperature excess is + 732 ± 316 K at the precursor footpoint FP 1 and + 1650 ± 604 and + 1260 ± 549 K at the trigger footpoints FP 3 and FP 4, respectively. These excesses correspond to approximately 2.3–2.7 σ. Section 3.3 reports the sensitivity of the simulation results to perturbation amplitudes spanning Δ T = 500–2000 K.

3.3 Chromospheric jets, energy transport, and sensitivity to the NLTE-Revised perturbation

Figure 7 shows the evolution of the peak normalised current density | J z | m a x at the PIL. In the reference run, it first exceeds the anomalous-resistivity threshold j m a x = 16 att = 3.6 min, and panel (a) resolves this onset region. Reconnection then proceeds as a series of discrete bursts rather than as a single sustained episode. The bursts continue throughout Phase 1, reaching a peak of 28.6 att ≈ 20 min, and persist into Phase 2, with a secondary peak of ≈ 21.9 at t ≈ 26 min. After the transition to Phase 3, | J z | m a x decreases steadily, falls below j m a x at t = 36.2 min, and reaches 1.82 by t = 47.8 min. In panel (b), all three corrected-height runs cross the reconnection threshold within one snapshot cadence of 43 s and are delayed by only ≈ 0.7 min relative to the reference run. The Δ T = 500 K run shows no detectable onset delay relative to the Δ T = 2000 K run. The corrected-height runs produce 7–8 discrete episodes during Phases 1–2, compared with 5 in the reference run. Active reconnection ends 7–10 min earlier in the corrected-height runs (t = 26.8–29.7 min, compared with 36.2 min), while their late-time current level is approximately five times higher (⟨ | J z | m a x ⟩ ≈ 6, compared with 1.2 for t > 60 min). Nevertheless, all three corrected-height runs form a PIL current sheet and initiate sustained, bursty reconnection. Outcome metrics for all four runs are collected in Table 5.

TABLE 5. LARE2D simulation results. (a) Trigger outcome for the corrected-height amplitude sweep (y 0 = 1.25 Mm, σ y = 0.5 Mm) compared with the reference run. A burst is a contiguous excursion of | J z | m a x above j m a x = 16 at the 43-s snapshot cadence, counted on the rising edge; onset is the first crossing sustained over at least two consecutive snapshots; cessation is the last snapshot above j m a x. Phase 1 counts for the corrected-height runs include a single-snapshot excursion at t = 2.2 min associated with the initial relaxation transient common to all three runs. (b) Additional diagnostics for the 1024 × 1024 reference simulation. Normalised quantities use B n o r m = 100 G, ρ n o r m = 1.67 × 1 0 − 8 kg m − 3, L n o r m = 1 Mm, v A = 69 km s − 1.

To separate the effects of the thermal source and the converging shear, we repeated the reference experiment with the thermal source removed and the shear retained. Reconnection still occurs: | J z | m a x first exceeds j m a x at t = 4.4 min, compared with t = 3.6 min in the reference run, and the subsequent evolution is indistinguishable. During Phase 1, compact enhancements in | J z | repeatedly form near x = 0 and y ≈ 1–2 Mm, while the corresponding V y + maps show narrow upward-velocity columns extending from the chromosphere into the low corona. Figure 8 a shows the temporal coupling between | J z | m a x and the upward velocity at y = 10 Mm. Their normalized Phase-1 cross-correlation peaks at a lag of 5.8 min in the reference run (Figure 8c), while the corresponding jet launch speeds are 28–55 km s − 1 (Table 5). Jet bursts are detected at y = 3, 6, and 10 Mm (Figure 8b), with peak speeds of ≈ 40 km s − 1 near the transition region and up to ≈ 85 km s − 1 at y = 10 Mm. Figure 8b also shows that the activity continues through Phase 2. The dashed curves in Figures 8a–c show that the NLTE-revised run likewise produces outflows that reach the low corona, with amplitudes comparable to those of the reference run at all three heights. Its Phase-1 cross-correlation peaks at 8.0 min, compared with 5.8 min for the reference run (Figure 8c). The temperature height–time diagram in Figure 9a shows that heating is initially strongest below and around the transition region, and that, from t ≈ 5 min, the enhanced temperature extends above y = 3 Mm. At y ≈ 5–10 Mm, the coronal temperature rises by Δ T ≈ 0.5 MK above background. The corresponding velocity diagram in Figure 9b shows that the upward velocity declines at all heights after t ≈ 30 min. The two-dimensional maps in Figure 10 show the accompanying restructuring. At t = 25 min (Figure 10b), the main reconnection region has shifted upward to y ≈ 6–8 Mm, with newly reconnected field lines forming a compact hot-loop system at 1–3 MK and a field geometry that is less strongly sheared than during Phase 1. At t = 45 min (Figure 10c), the compact PIL current layer has dispersed, no localized region with | J z | > j m a x remains, the hot loops have cooled, and the streamlines have relaxed toward a less-sheared configuration.

4 Discussion

The results presented here support a connected physical picture in which the compact CaII K precursor, the stratified footpoint response, and the MHD evolution are different stages of the same flare-triggering sequence. Photospheric magnetic descriptors such as strong-field gradients, unsigned flux, shear, and non-potentiality near PILs have proven useful for identifying flare-productive active regions (Leka and Barnes, 2003; 2007; Bobra and Couvidat, 2015; Florios et al., 2018; Toriumi and Wang, 2019; Georgoulis et al., 2021). However, these quantities characterize the stored energy budget rather than the perturbation that determines when that energy is released. Chromospheric precursors fill this gap because they sample the layer where localised heating, flows, and small-scale reconnection can occur minutes before the coronal energy release (Chifor et al., 2007; Bamba et al., 2013; 2017). The CaII K precursor studied here is notable in this context. It is not merely an early intensity enhancement but is associated with a measurable chromospheric temperature excess (Δ T ≈ 1000 K from the NLTE inversion), a coherent but low-significance pre-flare blueshift at the jet-precursor footpoint FP 1, and a magnetic pathway connecting the precursor site to the eventual flare ribbons through the flux-proxy connectivity model (see Figure 3 of Kumar et al. (2026). This goes beyond what IRIS spectral classification can offer for individual events (Panos et al., 2018; Panos and Kleint, 2020; Woods et al., 2021; Panos et al., 2023), because the spectropolarimetric inversions provide the actual atmospheric stratification at each footpoint rather than a statistical profile-shape discriminant.

At log ⁡ τ = − 1, the pre-flare photospheric field strengths span approximately 0.4–0.87 kG, with no simple polarity-dependent ordering (Figure 4). Photospheric line-of-sight velocities are small at every footpoint, with | v | ≤ 2.2 km s − 1. The chromospheric velocities are somewhat larger but remain modest: at log ⁡ τ = − 3, their magnitudes remain below 9 km s − 1 and show no polarity-dependent ordering that is robust to the inversion treatment. The post-flare stratifications are shown in Figure 5, and the signed changes are summarized in Figure 6. At log ⁡ τ = − 1, the retrieved total field-strength changes are approximately + 3, + 24, + 1, 0, and − 14 G at FP 1–FP 5, respectively. At log ⁡ τ = − 3, the corresponding NLTE changes are + 48, − 11, + 325, + 17, and − 319 G. The photospheric changes are therefore small, whereas the chromospheric response is larger at FP 3 and FP 5. The two negative-polarity footpoints strengthen slightly at log ⁡ τ = − 1 but change in opposite directions at log ⁡ τ = − 3, while the positive-polarity footpoints also show mixed responses. Thus, the changes do not define a coherent pattern organized by magnetic polarity, and we find no clear evidence for a polarity-asymmetric Lorentz-force back-reaction (Fisher et al., 2012). Observational support for such stepwise changes comes from Sudol and Harvey (2005) and Petrie and Sudol (2010), who documented abrupt, permanent | Δ B L O S | changes of up to ∼ 300 G (median ∼ 90 G) coincident with the impulsive phase using GONG magnetograms, and from Castellanos Durán et al. (2018), who found permanent B L O S changes in 59 of 75 flares observed with SDO/HMI, with weaker flares less likely to show detectable changes. The retrieved B z retains its sign at every footpoint, so we also find no evidence for a reorientation of the field vector. Two caveats apply. First, the chromospheric B z values at FP 3, FP 4, and FP 5 lie at or close to the ± 500 G bound imposed on the field components, so the retrieved field strength changes at these footpoints should be regarded as lower limits. Second, flare heating can shift the response function of the Ca II 8542 Å line to deeper layers (Kuridze et al., 2018; Yadav et al., 2021); an apparent change in chromospheric field strength may therefore reflect a change in formation height rather than a true intensification. Our measurements do not distinguish between these possibilities.

Second, the velocity changes are small at most footpoints (Figure 6). If non-thermal electrons or thermal conduction fronts deposit their energy above the temperature-minimum region, the velocity response is expected to be dominated by the chromosphere (Allred et al., 2015; Druett et al., 2017). Impulsive heating drives upward chromospheric evaporation together with downward-propagating chromospheric condensation (Fisher and Hawley, 1990; Druett et al., 2017). Line-of-sight condensation velocities of 10–40 km s − 1 have been measured in Hα and CaII 8542 Å during C- and M-class flares (Kuridze et al., 2018; Kleint, 2017) and reproduced in 1D radiative-hydrodynamic simulations using RADYN (Allred et al., 2015). We do not resolve a comparable signature in these data. At log ⁡ τ = − 1, the photospheric Δ v values remain below approximately 1 km s − 1. At log ⁡ τ = − 3, the NLTE changes range from − 7.9 to + 2.5 km s − 1 across the five footpoints. The largest change occurs at FP2, while FP5 changes from + 4.1 to + 2.0 km s − 1. These changes show no coherent footpoint-dependent condensation pattern.

Third, the temperature changes do not exceed their propagated cycle-to-cycle uncertainties at four of the five footpoints (see Figure 6). At log ⁡ τ = − 3, the chromospheric changes are + 260 ± 929 K at FP3 and + 190 ± 813 K at FP4, with cooling of − 318 ± 379, − 168 ± 86, and − 626 ± 1299 K at FP1, FP2, and FP5, respectively. Only the change at FP2 exceeds its cycle-to-cycle uncertainty, so we do not infer a polarity-dependent heating pattern. This is consistent with flare simulations in which the energy-flux spectrum and column depth vary from footpoint to footpoint depending on the local magnetic mirror ratio and loop length (Allred et al., 2015). Given the large uncertainties, however, these measurements do not distinguish between uniform energy deposition and genuine footpoint-to-footpoint variations.

The LARE2D simulation connects these observations to a physical trigger test. The perturbation is not arbitrary. The original reference run uses the earlier LTE-derived amplitude of Δ T = 2000 K, whereas the NLTE-corrected run adopts Δ T = 1000 K from the revised HAZEL/SIR inversion of the precursor site. The driver phasing follows the observed CaII K light curve (Kumar et al., 2026), and the arcade shear ( α s = 0.90 ) represents the pre-stressed magnetic configuration. Under these data-constrained conditions, the chromospheric forcing produces a current sheet at the PIL, launches bursty reconnection during Phase 1, and allows the arcade to continue reconnecting after the driving amplitude drops in Phase 2. This result is consistent with a bottom-up scenario in which a compact chromospheric perturbation does not supply the full flare energy but pushes an already stressed arcade past the reconnection threshold (Moore et al., 2001; Kusano et al., 2012). Similar ideas have been explored in the context of flux emergence (Archontis and Hood, 2012), small-scale tether-cutting (Moore et al., 2001), and trigger-field reconnection near the PIL (Kusano et al., 2012; Bamba et al., 2017). In our case, the advance is that the perturbation amplitude is tied directly to the chromospheric inversion output and the observed precursor timing (Kumar et al., 2026), rather than being a free parameter. The experiments reported in Section 3.3 sharpen what this test establishes. Because temperature is the least reliably recovered quantity from CaII 8542 Å inversions (de la Cruz Rodríguez et al., 2012; Leenaarts et al., 2009), the trigger was examined across a factor of four in Δ T and at two injection heights (Figure 7), rather than by adopting a single value. The trigger outcome is robust across all tested runs, although the subsequent evolution depends on the injection height. Thus, the central conclusion does not depend on the precise amplitude inferred from the inversions: the tested amplitudes span and extend beyond the NLTE-supported interval, and all produce a PIL current sheet that crosses the reconnection threshold. A control experiment in which the thermal source is removed while the converging shear is retained shows that, within the model, the shear determines whether the reconnection threshold is crossed, whereas the thermal perturbation advances the onset by approximately one snapshot interval. This division of roles is consistent with the observed precursor being a jet-associated brightening rather than a static thermal enhancement (Kumar et al., 2026). Moving the injection into the denser mid-chromosphere shortens the interval over which reconnection remains active and leaves a higher residual current at late times, as summarized in Table 5, indicating that the coupling between the driver and the overlying arcade depends on where in the stratification the energy is deposited. Taken together, these results support interpreting the simulation as a test of sufficiency rather than a reconstruction: a chromospheric disturbance with the observationally constrained location, timing, and thermal-amplitude range can push a near-critical arcade past the reconnection threshold, but the observations do not establish the relative contributions of converging motion and heating in the actual event.

The combination of three normally separate approaches is the principal contribution of this study: CaII K imaging identifies the precursor and its timing, spectropolarimetric inversions provide height-dependent diagnostics at individual footpoints, and the MHD experiment tests whether a perturbation with the observed properties can influence a flare-ready arcade. Each component alone would leave an ambiguity. The CaII K brightening alone could be a passive early emission feature. The inversions alone would reveal a structured atmosphere but not test whether that structure can trigger reconnection. The simulation alone would be idealized without observational constraints. Together, they support the interpretation that the precursor is physically meaningful.

Several limitations must be noted. The SPINOR scan is sequential, so the five footpoints are not co-temporal. FPs 3–5 are sampled only 5–13 min before onset and may already include precursor contamination. The CaII 8542 Å inversions carry larger uncertainties beyond log ⁡ τ ≈ − 4 where the line response weakens, so only changes within the reliable formation range should be interpreted (de la Cruz Rodríguez et al., 2012; Kuridze et al., 2018). The LARE2D model is two-dimensional with an idealized arcade geometry, whereas the observed active region has a fully three-dimensional magnetic topology. The simulation is best understood as a data-constrained trigger experiment, not as a unique reconstruction of the event. Test runs at lower grid resolution produced broader current layers that did not exceed the anomalous-resistivity threshold, indicating that the trigger outcome depends on resolving sufficiently thin current sheets at the PIL. Three-dimensional, higher-resolution models will be needed to address this convergence issue. Despite these caveats, the combined evidence favors an interpretation of compact CaII K precursors as more than passive brightenings in flare-ready regions. In this event, the precursor is connected to localised chromospheric heating, strong pre-flare velocity structure, pre-existing magnetic connectivity to the eventual ribbon sites, and a perturbation amplitude capable of driving reconnection in a data-constrained simulation. This does not imply that every CaII K brightening is a flare trigger. Rather, it suggests that such brightenings may become physically important when they occur at magnetically connected footpoints in an already stressed arcade. A larger sample of events with matched spectropolarimetric coverage, including both precursor-positive and precursor-negative cases, will be needed to determine whether these brightenings are necessary, sufficient, or conditional indicators of flare onset.

5 Concluding remarks

We have investigated whether compact CaII K brightenings observed before solar flares can be interpreted as physically meaningful chromospheric triggers, rather than only as empirical early-warning signatures. For the 2025 June 13 GOES C3.1 flare, we combined ROSA CaII,K imaging (Kumar et al., 2026), SPINOR full-Stokes spectropolarimetry in Fe I 5250 Å and CaII 8542 Å, HAZEL/SIR inversions at five ribbon footpoints spanning both magnetic polarities, flux-proxy connectivity modeling, and data-constrained LARE2D simulations.

• The pre-flare atmosphere is already non-uniform across the five ribbon footpoints. At log ⁡ τ = − 3, the chromospheric line-of-sight velocities remain below 9 km s − 1. FP 1 shows a persistent but low-significance blueshift, whereas the sign of the velocity at FP5, whose profile contains a central emission reversal, depends on the inversion treatment and is not interpreted as a flow direction. The NLTE inversion also gives a localised precursor temperature enhancement of Δ T ≈ 1 000 K. Together, these signatures show that the lower atmosphere is already structured and dynamically active before any detectable GOES soft-X-ray rise.
• The post-flare inversions show small photospheric field changes and larger but non-uniform chromospheric changes (Figure 6). Neither the magnetic nor thermal response exhibits a coherent polarity dependence, B z retains its sign at every footpoint, and only the temperature change at FP2 exceeds its propagated cycle-to-cycle uncertainty. We therefore find no clear evidence for either a polarity-asymmetric Lorentz-force back-reaction (Fisher et al., 2012) or a reorientation of the field vector. The chromospheric field changes at FP3–FP5 remain limited by the ± 500 G bound imposed on the field components and by the possible shift of the CaII 8542 Å response function to deeper layers during flare heating (Kuridze et al., 2018; Yadav et al., 2021).
• The flux-proxy connectivity model confirms that the precursor site and eventual ribbon footpoints are magnetically linked before the flare, providing a geometric pathway through which a localized chromospheric perturbation can influence the overlying stressed field. Pre-flare connectivity is dominated by compact closed linkages (FP3→FP2, FP4→FP1), while post-flare connectivity reorganizes toward predominantly open or extended trajectories.
• The LARE2D simulations provide a data-constrained test of whether the observed chromospheric disturbance is sufficient to initiate reconnection in a pre-stressed arcade. A thermal perturbation constrained by the inversions, combined with a shear driver whose three-phase timing follows the observed CaII K lightcurve, forms a PIL current sheet and crosses the reconnection threshold. This trigger outcome is robust across Δ T = 500–2000 K and at both injection heights (Figure 7; Table 5), although the cessation time and late-time residual current depend on the injection height. The no-thermal control further shows that the converging shear is sufficient to trigger reconnection, while the thermal perturbation advances the onset by approximately one snapshot interval. During Phase 1, repeated bursts launch outflows with characteristic speeds of 28–55 km s − 1, reaching approximately 85 km s − 1 at y = 10 Mm; the coronal response follows the PIL-current enhancements after a lag of 5.8–8.0 min (Figure 8). Reconnection continues through Phase 2 after the forcing is halved, showing that continuous strong driving is not required. These results support interpreting the simulation as a sufficiency test: a compact chromospheric disturbance can push an already stressed arcade past the reconnection threshold without supplying the full flare energy.

Correction note

This article has been corrected with minor changes. These changes do not impact the scientific content of the article.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. I have used gen-aI for language editing and formatting.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: José Juan González-Avilés, National Autonomous University of Mexico, Mexico

Reviewed by: João M. da Silva Santos, National Solar Observatory, United States

Pradeep Kayshap, VIT Bhopal University, India