Elemental Composition Evolution during the 2024 September 30 Solar Eruption: A Comparison of Hot and Cool Plasma Components with Solar Orbiter/SPICE, Hinode/EIS, and Chandrayaan-2/XSM

The paper presents novel observations from Solar Orbiter SPICE and EUI, Hinode/EIS, and Chandrayaan-2 XSM instruments, finding rapid abundance changes in the emitting plasma, on timescales of minutes, during the eruptive M7.6-class solar flare observed on 2024 Sept 30. These instruments have wide temperature coverage and find contrasting abundance-evolution patterns between the hotter and cooler plasma components. 3D reconstruction of the active region and additional observations from the Solar Orbiter STIX X-ray telescope show how the hot and cool plasma components, emitting in different spectral regions and observed with the various instruments, sample the plasma composition evolution in distinct locations within the observed flaring plasma. The bright post-flare loop tops observed by SPICE show coronal FIP bias, while the hot plasma observed with XSM exhibits FIP-bias decreasing from coronal to photospheric during the impulsive phase. The authors interpret these observations as evidence of the X-ray diagnostics seeing hot coronal reconnection outflows mixing with chromospheric plasma as flare loops sequentially energize and relax, explaining why the FIP bias decreases from coronal to a hybrid; and the cool loop tops seen with SPICE show coronal abundances due to coronal material deposited near the looptops.

The paper describes the observations and reduction of data from each instrument. SPICE observed a range of spectral lines with temperature coverage in the range of 10 5–10 6 K, suitable for differential emission measure (DEM) analysis. The SPICE data suffer from a non-uniform point-spread function (PSF) in the x − y − λ space, requiring further reduction steps to alleviate this issue and make the data usable for further analysis. The authors performed a deskewing of the apparent effect of the PSF in the x − y − λ space, removing the artifacts of the PSF by estimating it through the data themselves. After applying the reduction, they found a shift of [–5, +1] pixels/Å, making the rightmost two columns of the raster unusable.

Hinode/EIS observed during the flare and produced a raster scan over the AR, with the spectral scan taking place during 2024 Sept 30 22:52–23:58 UT, with the slit scanning from west to east. The line ratio of S X 264.3 Å and Si X 258.4 Å is used as a FIP bias proxy derived from the Hinode data.

Chandrayaan-2/XSM observes the Sun-as-a-star and measures the solar X-ray spectrum in the energy range of 1–15 keV with an energy resolution of 0.175 keV (FWHM) at 5.9 keV and time cadence of 1 s. XSM observed the Sun in the 1–15 keV energy range during the impulsive phase of the flare until 23:58 UT, after which an attenuator was inserted to reduce the photon flux, so only spectra from 2 keV onward were available. The authors analyzed XSM spectra only during the impulsive phase of the flare, fitting the 1-minute cadence XSM spectra with a two-temperature component model, considering the temperature and emission measure of each component as free parameters, while the elemental abundances of Mg, Si, S, Ar, Ca, and Fe were treated as free parameters.

STIX is a hard X-ray imaging instrument on Solar Orbiter with an energy range of 4–150 keV. The authors used the data from the lowest energy bands (6–10 keV) observed with STIX to localize the sources of the Sun-as-a-star signals detected with XSM.

For the DEM and abundance analysis with SPICE, the authors used the DEM algorithm of J. Plowman & A. Caspi (2020) via the EMToolkit interface to reconstruct the DEM from the SPICE lines. The temperature contribution functions are precomputed for the SPICE lines using CHIANTI v8.0.7 assuming electron density of 10 8 cm-3. The coronal plasma abundance based on the SPICE spectral observations is derived from a DEM inversion approach, where the abundance is left as a free parameter. In this approach, the abundances of all elements vary together between the values of a photospheric (E. Caffau et al. 2008) to coronal abundance models (A. Fludra & J. T. Schmelz 1999). The inversion technique varies a single free parameter that linearly interpolates between the two abundance models, and that parameter effectively determines all of the elemental abundances.

The authors also performed a separate analysis of the evolution of the FIP-sensitive line ratios S V 786 Å to N IV 765 Å and Mg IX 706 Å to Ne VIII 770 Å, to investigate the plasma composition evolution during the flare progression. The resulting FIP biases show clearly the abundance evolution during the flare progression, with the looptops getting filled with increased FIP-bias plasma, corroborating the findings from the DEM+abundance inversion.

To interpret the observed Doppler shifts in the SPICE data in terms of downflows and upflows, the authors used the CROBAR framework to construct a model of the AR based on magnetograms and EUV observations. CROBAR reconstructs the coronal magnetic field through a linear force-free field (LFFF) extrapolation of the observed AR from a photospheric magnetogram, using the available SDO/HMI line-of-sight magnetic field strength. The resulting reconstruction provides a 3D rendering of the AR, where each coronal loop has a known set of three-dimensional coordinates associated with it, as well as thermodynamic and magnetic plasma properties along the loop length. The CROBAR results point to the fact that footpoints of the post-flare loops closer to disk center seen by Solar Orbiter are oriented along the line-of-sight, and the ones closer to the limb are pointing away from Solar Orbiter. Hence, the authors can interpret the data of the blueshifts at the loop footpoints as chromospheric evaporation in the closer footpoints, while the redshifts observed above the looptops correspond to downflows.

The Hinode/EIS observations show an increase in intensity, which corresponds to the bright spot in the left side of the figure highlighted by the red arrow, at times after 23:30 UT, when the flare enters its impulsive phase. The Si X 258.4 Å to S X 264.3 Å ratio clearly shows the enhanced FIP bias in the flaring region, which somewhat agrees with the SPICE conclusions.

The XSM observations show a decrease in the abundances of all low-FIP elements, such as Mg, Si, Ca, and Fe, during the impulsive phase of the flare, which is consistent with previous XSM observations. Near the peak of the flare, Si, S, and Fe all show a decreasing abundance. Sulfur shows a more modest decrease compared to the other low-FIP elements. Argon, which is the only high-FIP element in the XSM analysis, does not show a statistically significant change during the impulsive phase of the flare. The increase of the hot component EM in the model is consistent with chromospheric evaporation material being heated, which has photospheric abundance. This newly heated material emitting in X-ray produces the drop in the FIP bias observed with XSM.

The STIX data show the source of the X-ray emission is located in a persistent location at solar helioprojective coordinates of about [+3000′′, –980′′] with a slight drift to the west, seen from Solar Orbiter. The X-ray radiation detected by XSM is emitted from a source distinct from the bright EUV features. The 6–10 keV energy band is dominated by thermal plasma emission, likely coming from the plasma upflows associated with chromospheric evaporation, where photospheric FIP bias plasma is being ablated and elevated into the loops.

The authors conclude that the discrepancy between the EUV and X-ray inferred abundance evolution is due to the different solar regions emitting the two types of solar radiation. In particular, XSM shows the dropping FIP bias consistent with chromospheric evaporation in the hottest loops, whereas SPICE is looking in the older, cooling, post-flare loops, which experience mass loading from above, as suggested from the detected SPICE Doppler shifts.

The paper's conclusions state: "We have described a set of observations of the M7.6-class 2024 Sept 30 flare with Solar Orbiter and a few other observing assets in the vicinity of Earth. The SO observations include SPICE data with a variety of spectral lines, suitable for analysis of the plasma composition during the flare evolution. After appropriate calibration of different datasets, we applied a combined approach to infer the DEM and composition simultaneously of the flaring plasma from the SPICE data. The approach, inverts at the same time for both DEM distribution and elemental abundance. The combined DEM+abundance inversion has the plasma abundance as a free parameter, linearly varying the composition of all elements between photospheric and coronal abundance model values. We found that the bright looptops seen by SPICE exhibit coronal FIP bias compared to the footpoints. When we computed the FIP bias for the Mg IX 706 Å to Ne VIII 770 Å, and S V 786 Å to N IV 765 Å lines using two-line ratios, we found similar results, corroborating the previous analysis. This distinct computation supports the robustness of the previous result based on the combined DEM+abundance inversion approach. Furthermore, using CROBAR we were able to reconstruct the geometry of the AR and associate the Doppler velocity observed in the SPICE data with the evolution of the plasma composition. The observed redshifts above the looptops correspond to downflows based on the CROBAR reconstruction, and are compatible with reconnection-driven downflows, which would transport plasma with coronal (enhanced) FIP bias. When these cooling looptops become visible in the SPICE spectral lines, they exhibit enhanced looptop FIP-bias, possibly due to the coronal material transport from the reconnection outflows. The presence of the plasma downflows is inferred from the redshifts present in the SPICE Doppler shifts, interpreted through the CROBAR AR modeling."

"Comparison of the SPICE observations with other EUV and X-ray observations provides a complementary picture of the physical evolution of the plasma with different temperatures. The Hinode/EIS EUV observations, which only caught a very small part of the impulsive phase of the flare, also show an increase in the FIP bias in the flaring AR apparent in the Si X 258.4 Å to S X 264.3 Å ratio map in Figure 9. This result agrees with the finding from SPICE, that the brightest parts of the flaring regions exhibit increased FIP bias, despite measuring different temperatures. When comparing the flaring plasma composition detected with the Chandrayaan-2/XSM instrument, however, we find a significant discrepancy between the abundance evolution detected in the soft X-ray data with the SPICE EUV data. In particular, we find a decrease in the abundance of the low-FIP elements, and no change in the high-FIP element Ar abundance, as shown in Figure 11. To locate the X-ray sources of the XSM emission, we use SO/STIX imaging data. The location of the soft X-rays is displaced from the EUV loop tops as shown in Figure 12. Our results support the interpretation that the apparent EUV/X-ray discrepancy can arise because the diagnostics sample distinct thermal and spatial plasma components, possibly influenced by different processes. In particular, the X-ray data samples chromospheric material (with photospheric abundance) being evaporated into the corona and being heated, as suggested by the modeling of the XSM X-ray data presented in Figure 11. On the other hand, the SPICE EUV results show that the cooling post flare loops exhibit enhanced looptop FIP-bias. The observed downflows towards the looptops (as interpreted with the CROBAR modeling) are compatible with a reconnection downflow transporting coronal material (elevated FIP-bias) to these looptops. These old loops were previously seen in the XSM data as the hot loops with photospheric FIP bias, which now have cooled, and exhibit photospheric abundances everywhere, except at their looptops."

"In conclusion, we were able to infer the spatially resolved SPICE measurements of rapid FIP-bias evolution in post-flare loop plasma. The results show the complexity of the phenomenon. The comparison between SO/SPICE and XSM shows the clear discrepancy of the plasma composition evolution. In particular, we suggest that the contrasting behavior arises from differing regions emitting the different radiation detected by the two instruments. We suggest this by using the lowest energy channel of the STIX instrument, which observed the event from the SO vantage. This shows the potential of SPICE for observing composition changes during AR flares and provides spatiotemporal constraints of the FIP fractionation process."

Improvements to AI Systems:

1. Multi-Instrument Data Fusion for Solar Flare Analysis

• Improvement: Develop an AI model that integrates heterogeneous datasets (EUV spectra, X-ray spectra, imaging, magnetograms) with varying spatial, temporal, and spectral resolutions, accounting for instrument-specific artifacts (e.g., non-uniform PSF, attenuator insertion, raster timing).

• Capability: Automatically cross-calibrate and co-align observations from Solar Orbiter, Hinode, and Chandrayaan-2 to produce a unified, time-resolved 3D map of plasma composition, temperature, and dynamics during flares.

2. Real-Time DEM and Abundance Inversion with Uncertainty Quantification

• Improvement: Enhance the DEM+abundance inversion algorithm (e.g., Plowman & Caspi 2020) with a neural network that learns the mapping from multi-line intensities to DEM and FIP-bias, including error propagation and non-LTE effects.

• Capability: Instantaneously infer plasma composition and temperature distributions from new observations, with confidence intervals, enabling rapid classification of flare phases and detection of abundance anomalies.

3. PSF Deconvolution and Artifact Correction via Self-Supervised Learning

• Improvement: Train a convolutional neural network (CNN) to estimate and remove the non-uniform PSF in x-y-λ space directly from the data, without requiring external calibration targets, by exploiting redundancy across spectral lines and spatial scans.

• Capability: Automatically clean SPICE-like spectral rasters, recover lost edge pixels, and improve signal-to-noise for faint lines, enabling more reliable abundance measurements in low-emission regions.

4. Automated 3D Coronal Loop Reconstruction and Doppler Interpretation

• Improvement: Integrate CROBAR’s linear force-free field extrapolation with a transformer-based model that predicts line-of-sight Doppler shifts from reconstructed loop geometry, magnetic field topology, and flare timing.

• Capability: Automatically classify upflows/downflows in flaring loops, identify evaporation vs. reconnection outflows, and predict where FIP-bias enhancements will appear in EUV vs. X-ray diagnostics.

5. Cross-Wavelength Abundance Discrepancy Resolution

• Improvement: Build a physics-informed neural network that learns the relationship between EUV (SPICE/EIS) and soft X-ray (XSM) abundance measurements, incorporating thermal structure, spatial source location (from STIX), and plasma evolution models.

• Capability: Predict when and where EUV and X-ray abundance signals will diverge, and automatically flag regions where chromospheric evaporation vs. coronal downflows dominate, providing a unified picture of flare composition dynamics.

6. Rapid Flare Phase Classification and Forecasting

• Improvement: Train a temporal convolutional network on multi-instrument time series (e.g., XSM EM, SPICE FIP-bias, STIX source drift) to identify flare phases (pre-impulsive, impulsive, gradual) and predict upcoming abundance changes.

• Capability: Provide early warnings of chromospheric evaporation onset and FIP-bias drops, enabling targeted follow-up observations and real-time alerts for space weather models.

7. Automated FIP-Bias Proxy Selection and Validation

• Improvement: Use a reinforcement learning agent to select optimal spectral line pairs (e.g., S V/N IV, Mg IX/Ne VIII, Si X/S X) based on temperature sensitivity, blending, and signal-to-noise, dynamically adapting to instrument state.

• Capability: Automatically generate robust FIP-bias maps with minimal human intervention, even during flares with rapidly changing plasma conditions.

8. Simulation-to-Real Transfer for Flare Plasma Modeling

• Improvement: Train a generative model on synthetic flare simulations (e.g., from CROBAR + radiative transfer) to predict SPICE/XSM observations, then fine-tune on real data via domain adaptation.

• Capability: Enable AI to infer hidden plasma parameters (e.g., reconnection rates, mass loading) from observed abundance patterns, and test alternative flare models against multi-wavelength data.

Solar plasma composition differs between the photosphere and corona over a range of timescales, with preferential enhancement of elements with low first ionization potential (FIP). However, the physical origin of the FIP fractionation remains incompletely understood. Furthermore, during flares, the FIP bias also exhibits rapid changes, associated with fast transport of material with different FIP biases. We present novel observations from Solar Orbiter SPICE and EUI, Hinode/EIS, and Chandrayaan-2 XSM instruments, finding rapid abundance changes in the emitting plasma, on timescales of minutes, during the eruptive M7.6-class solar flare observed on 2024 Sept 30. These instruments have wide temperature coverage and find contrasting abundance-evolution patterns between the hotter and cooler plasma components. 3D reconstruction of the active region and additional observations from the Solar Orbiter STIX X-ray telescope show how the hot and cool plasma components, emitting in different spectral regions and observed with the various instruments, sample the plasma composition evolution in distinct locations within the observed flaring plasma. The bright post-flare loop tops observed by SPICE show coronal FIP bias, while the hot plasma observed with XSM exhibits FIP-bias decreasing from coronal to photospheric during the impulsive phase. We interpret these observations as evidence of the X-ray diagnostics seeing hot coronal reconnection outflows mixing with chromospheric plasma as flare loops sequentially energize and relax, explaining why the FIP bias decreases from coronal to a hybrid; and the cool loop tops seen with SPICE show coronal abundances due to coronal material deposited near the looptops.