The Extreme-ultraviolet Variability Experiment, or EVE, was never designed to measure Doppler velocities. Its mission is to track how the Sun’s extreme ultraviolet output varies over time, information that matters because that radiation is the primary energy input to Earth’s ionosphere and thermosphere. EVE observes the full solar disk from 6 to 106 nanometers with a spectral resolution of just 0.1 nanometers, using its Multiple EUV Grating Spectrographs, known as MEGS. At that modest resolution, detecting the tiny wavelength shifts produced by moving plasma would seem impossible. Yet two fortunate circumstances changed the calculus: the SDO spacecraft holds its solar pointing with exceptional stability, and its geosynchronous orbit provides an unusually stable thermal environment, keeping the instrument’s wavelength scale steady enough to capture genuine solar motions.

The new data product fits Gaussian profiles to 70 carefully selected emission features, each originating from a distinct layer of the solar atmosphere. Roughly half of the lines come from the transition region, a thin boundary layer where temperatures climb from about 25,000 kelvin to 0.6 million kelvin, and half come from the corona above, where temperatures exceed 0.6 million kelvin and can surpass 6 million kelvin during flares. The fitting algorithm models each feature of interest with a Gaussian plus two additional Gaussians for blended neighboring lines and a linear background, yielding fitted values for intensity, wavelength center, and line width. When the GOES X-Ray Sensor detects a flare, the algorithm automatically subtracts a pre-flare spectrum to isolate the flare’s own spectral signature, flagging those events in the data product.

Converting wavelength shifts to velocities is straightforward in principle: the shift between the flare spectrum and the pre-flare spectrum, multiplied by the speed of light and divided by the reference wavelength, gives the line-of-sight velocity. Positive values are red shifts, indicating plasma receding from the observer, which for flares near disk center means downflow. Negative values are blue shifts, indicating approaching plasma, or upflow. The team emphasizes that measured pre-flare wavelengths, rather than theoretical values from the CHIANTI spectral database, must serve as the reference, because small systematic offsets in the instrument’s wavelength scale would otherwise contaminate the result.

Not every spectral line is trustworthy for this work. Many EUV features are blends of multiple emissions at MEGS resolution, and different lines dominate at different plasma temperatures, so a blended feature can shift spuriously as solar activity changes. The team used CHIANTI model spectra for quiet Sun, active region, and flare conditions to estimate this blend uncertainty for each of the 70 features, expressing it as a velocity error. Forty-two lines, nine from MEGS-A and thirty-three from MEGS-B, came in below the 30 kilometers-per-second threshold and are flagged as the best choices for studying flare dynamics. The rest carry blend errors large enough to swamp any genuine solar signal.

Perhaps the most technically demanding part of the study involved disentangling an optical artifact from real solar physics. The MEGS-B instrument uses two Rowland-circle spectrographs in tandem, and because its CCD sensor is flat rather than curved along the Rowland circle, most wavelengths are slightly out of focus. That defocus makes the measured wavelength scale sensitive to where an active region sits on the solar disk. Raytrace modeling of the original optical design confirmed that active regions near the east limb produce wavelength shifts in one direction, west-limb regions in the opposite direction, with the sign flipping between central and outer wavelengths. Earlier reports of surprisingly fast prograde-rotation velocities of about 50 kilometers per second in coronal lines, published by Hudson and colleagues in 2022, turn out to be an artifact of this optical behavior rather than a genuine solar flow.

To validate and calibrate the correction, the team exploited a rare window in April 2019 near the minimum of solar cycle 24, when a single active region, NOAA 12738, crossed the disk alone over six days. By subtracting a spectrum from a spotless day to remove the full-disk contribution, and then fitting the remaining active-region spectrum, the researchers measured wavelength shifts that matched a tuned raytrace model remarkably well. The tuning revealed that the Sun’s center sits about 2.5 arc-minutes east of MEGS-B’s optical center, a misalignment inherited from the compromise positioning of EVE’s three channels and the jolts of launch. The resulting correction equations, expressed as parabolic functions of wavelength scaled by flare position, are now available for users of the data product, though the corrections are not applied automatically because the processing pipeline does not know each flare’s location.

The payoff comes in the flare statistics. For the X2.2 flare of 15 February 2011, the first X-class event of the SDO mission, all chromospheric, transition region, and cool coronal features below 1 million kelvin showed maximum red shifts averaging 75 plus or minus 24 kilometers per second, while hotter coronal lines showed blue shifts averaging minus 114 plus or minus 75 kilometers per second, with the Fe XIV line at 21.14 nanometers reaching an extraordinary minus 213 kilometers per second. The X9.0 flare of 3 October 2024 told a similar story, with downflows averaging 37 plus or minus 16 kilometers per second and upflows averaging minus 94 plus or minus 58 kilometers per second. In both cases, and across a broader sample of 15 disk-center X-class flares, the velocity maxima almost always occurred during the impulsive phase, the brief interval when magnetic reconnection releases energy most furiously.

These patterns match the standard picture of flare physics, often called the CSHKP model, in which reconnection drives chromospheric evaporation upward into hot coronal loops while cooler material drains downward. The transition from red shifts to blue shifts between 1 and 2 million kelvin, first noted by Milligan and Dennis in 2009 using Hinode data, appears clearly in the EVE results. The findings also align with earlier EVE-based analyses, including Hudson and colleagues’ 2011 report of a 50 kilometers-per-second red shift in He II and a 100 kilometers-per-second blue shift in Fe XXIV, and Otsu and Asai’s 2024 detection of a dramatic minus 400 kilometers-per-second blue shift during a filament eruption. For limb flares, the team found that even after location corrections, coronal lines still blue-shift by about 150 kilometers per second relative to the gradual phase, hinting at non-radial outflows that may accompany coronal mass ejections.

The practical lesson for solar physicists is that the EVE Level 4 Lines product, despite the instrument’s modest resolution and full-disk field of view, opens a routine path to measuring flare plasma dynamics across an entire solar cycle. Users must apply the optical wavelength-shift correction based on flare location for anything away from disk center, or risk mistaking instrumental artifacts for solar flows of 50 to 200 kilometers per second. But with 42 X-class flares already analyzed for MEGS-A and 103 for MEGS-B, and countless smaller events awaiting study, the new data product promises to deepen understanding of the magnetic explosions that drive space weather, disturb satellite orbits, and occasionally paint auroras across skies far from the poles.

Subject of Research: Measurement of solar flare Doppler velocities using the SDO EVE Level 4 Lines data product

Article Title: Solar Doppler Velocity Results from the SDO EVE Level 4 Lines Data Product

Article References: Solar Doppler Velocity Results from the SDO EVE Level 4 Lines Data Product. (n.d.). https://doi.org/10.1007/s11207-026-02727-w

Image Credits: AI Generated

DOI: 10.1007/s11207-026-02727-w

Keywords: solar flares, Doppler velocity, SDO, EVE, extreme ultraviolet, corona, transition region, space weather, magnetic reconnection, spectroscopy, Solar Dynamics Observatory, MEGS

The Solar Dynamics Observatory has transformed solar science since its launch, delivering continuous, multi-wavelength monitoring of the Sun and generating roughly 1.5 terabytes of data every day. AIA captures the full solar disk in multiple extreme ultraviolet, ultraviolet, and visible channels, each sensitive to plasma at different temperatures and heights in the solar atmosphere, producing 57,600 images daily. HMI complements this view by measuring the photospheric magnetic field through spectropolarimetric observations, providing magnetograms, Dopplergrams, and continuum images. Together these instruments underpin real-time monitoring of solar flares and coronal mass ejections that can threaten power grids, GPS, and satellite communications. Yet despite this data abundance, analysis methods in solar physics have remained largely task-specific, relying on handcrafted features, curated event catalogs, or models trained independently for each objective, a fragmentation that motivated the search for shared, reusable representations.

Foundation models offer a way out of this fragmentation. Analogous to large language models or vision models trained on massive datasets, a foundation model learns broadly useful structure from unlabeled data and can then be adapted to many downstream tasks with lightweight task-specific heads or fine-tuning. SDO’s archive is uniquely suited to this paradigm because of its long-term continuity, consistent calibration, and multi-channel coverage, which allow self-supervised training without extensive manual labeling. SDOFMv2 builds on earlier efforts including SDOFMv1, the first masked-autoencoder-based foundation model for SDO imagery, as well as Solaris, a forecasting-centered transformer model, and Surya, a large-scale spatiotemporal model trained on native 4096 by 4096 resolution observations with roughly 366 million parameters.

At the heart of SDOFMv2 is a masked autoencoder built on a Vision Transformer backbone. During pretraining, a high fraction of image patches is hidden and the model must reconstruct the missing regions, forcing it to learn global spatial structure rather than local pixel interpolation. The team introduced several solar-specific innovations. A region-aware masking and reconstruction strategy explicitly distinguishes the inner solar disk, informative off-limb structures such as prominences, and nearly empty background regions, which can occupy roughly 20 percent of a square full-disk image. Background areas are strongly down-weighted in the reconstruction loss so the model concentrates its capacity on physically meaningful solar morphology rather than trivial black corners.

The loss function itself was redesigned around the physics of solar imagery. A weighted, region-aware objective balances reconstruction of the disk, prominences, and background, while a Huber loss provides robustness to the extreme statistical imbalance between the quiet Sun and intense, sparse active regions. Because magnetograms contain sparse, high-frequency magnetic structures with weaker spatial correlations than smooth EUV emission, the researchers added a denoising fine-tuning stage in which Gaussian noise is injected into inputs and the model learns to recover the clean signal. This stage, combined with a reduced masking ratio of 10 percent for magnetic data, substantially improved HMI reconstruction quality. A signum-logarithmic transformation was also applied to compress the enormous dynamic range of solar intensities and magnetic fields while preserving sign information.

The framework includes three configurations: an AIA-only model with approximately 114 million parameters, an HMI-only model with about 111 million, and a joint AIA+HMI model with 114 million. Notably, the models were trained on workstation-scale hardware, four compute nodes with a total of eight NVIDIA RTX 5090 GPUs, requiring roughly three weeks and an estimated 670 GPU-hours per model. The team found that data engineering mattered as much as raw compute: converting the dataset to the cloud-optimized Zarr format reduced image loading times from 20 to 40 seconds to under 100 milliseconds, and a RAID 0 array of NVMe SSDs achieving about 21 gigabytes per second kept the GPUs continuously fed. An interleaved monthly splitting strategy ensured that training, validation, and test sets each sampled the full range of solar activity across the cycle.

Qualitative and quantitative evaluations showed substantial improvements in reconstruction fidelity over SDOFMv1, with sharper intensity gradients, better preserved coronal loops, and faithful recovery of off-limb prominence emission that the earlier model largely omitted. Analyses of the learned latent space using principal component analysis revealed that the dominant modes of variation consistently highlight active regions, magnetic concentrations, and prominences, while attention maps from the transformer encoder showed that different attention heads specialize in complementary solar structures and adapt their focus as solar conditions change. Ablation studies indicated that modality-specific models remain strongest for pure reconstruction, and that among joint models the ViT-Base configuration offered the best balance of capacity and performance, with the larger ViT-L variant not surpassing it under equal training budgets.

The true test of a foundation model is transfer, and here SDOFMv2 was evaluated on three downstream tasks against SDOFMv1, Solaris, and Surya in one of the most comprehensive benchmarking studies of heliophysics foundation models to date. In regression of the F10.7 solar radio flux, a widely used ground-based proxy for solar EUV emission, Surya achieved the best overall performance with a coefficient of determination of 0.918, consistent with its high-resolution, forecasting-oriented pretraining, while SDOFMv2’s AIA-based variants remained competitive and outperformed Solaris on mean squared error and R-squared. In classification of solar wind phenomena into streamer belt, sector reversal, coronal hole, and ejecta categories using Parker Solar Probe in situ labels, fine-tuned SDOFMv2-AIA delivered the highest accuracy, precision, and F1-score, and all SDOFMv2 variants outperformed the corresponding SDOFMv1-based results. In missing-channel reconstruction, where an entire AIA wavelength must be inferred from the others, the AIA-only model achieved a mean R-squared of 0.96, improving on SDOFMv1’s 0.94 and exceeding both Solaris and Surya on agreement with ground truth.

The results carry a broader message about how machine learning is reshaping heliophysics. Modality-specific and joint foundation models exhibit complementary strengths, and accurate pixel-level reconstruction turns out not to be a strict prerequisite for useful downstream representations, as the HMI model achieved competitive solar wind classification despite lower reconstruction fidelity. The authors acknowledge limitations, including the reduced spatial resolution of the SDOML dataset, where each pixel corresponds to roughly 1,353 kilometers on the solar surface, and the absence of Dopplergrams and continuum intensity products, and they point toward temporal forecasting objectives and cross-modal alignment as future directions. By releasing model weights, training code, and a benchmark evaluation framework openly, the team has established strong, reproducible baselines that could accelerate space-weather research and lower the barrier to entry for smaller groups hoping to build on solar foundation models.

Subject of Research: A multi-instrument AI foundation model for the Solar Dynamics Observatory with transferable applications in solar physics and space weather

Article Title: SDOFMv2: A Multi-Instrument Foundation Model for the Solar Dynamics Observatory with Transferable Downstream Applications

Article References: Hong, J., Martin, D., & Gallego, J. (2026). SDOFMv2: A Multi-Instrument Foundation Model for the Solar Dynamics Observatory with Transferable Downstream Applications. Solar Physics, 301(9), Article 137. https://doi.org/10.1007/s11207-026-02740-z

Image Credits: AI Generated

DOI: 10.1007/s11207-026-02740-z

Keywords: Solar Dynamics Observatory, foundation models, machine learning, solar physics, masked autoencoder, space weather, heliophysics, solar magnetic fields, solar wind, self-supervised learning, AIA, HMI

The research, led by Divya Kumari of Jai Prakash University and Pawan Kumar of Patna University, together with colleagues including Sanjay Kumar, Sadashiv, and Alok Ranjan Tiwary, is published in the journal Astrophysics and Space Science. Rather than relying on observations alone, the team combined multi-wavelength imaging from NASA’s Solar Dynamics Observatory with a sophisticated mathematical technique known as Non-Force-Free-Field extrapolation, which takes measured magnetic maps of the Sun’s surface and reconstructs the three-dimensional magnetic field permeating the corona above them. Because the solar corona is far too hot for spacecraft to sample directly, such extrapolations are among the most powerful tools available for peering into the invisible structures where flare energy is stored and released.

The observations immediately revealed something distinctive. Instead of the familiar pairs of parallel flare ribbons that typically accompany eruptions from simple magnetic arcades, this flare lit up in a circular pattern, a ring of bright emission surrounding a central kernel, while a second, unrelated-looking brightening appeared far away from the main event. During the impulsive phase, these initial brightenings matured into a circular ribbon and a remote ribbon, a combination that solar physicists recognize as the fingerprint of a so-called fan-spine magnetic configuration, in which field lines sweep downward from a coronal null point like the ribs of an umbrella converging on a central pole.

The extrapolated magnetic field confirmed the suspicion with striking precision. Embedded in the corona above the active region, the reconstructed field contained a genuine three-dimensional magnetic null, a point where the magnetic field strength drops to zero and the topology fundamentally reorganizes. From this null, a fan surface of field lines spread outward and downward, enclosing a dome whose footprint on the solar surface corresponded closely with the observed circular ribbon. Beneath the dome, tucked inside the flaring kernel, sat a pre-existing arc-shaped filament, visible in H-alpha data from the Global Oscillation Network Group. Intriguingly, this filament remained stable throughout the flare itself, only showing signs of eruption shortly after the main event had subsided, suggesting the filament may have been a consequence of, rather than the sole trigger for, the explosive energy release.

To understand why energy accumulated at this location in the first place, the researchers turned to measurements of the photospheric magnetic flux beneath the fan structure. They found clear evidence of flux cancellation during the flare, a process in which opposing magnetic polarities collide and mutually annihilate at the solar surface. Flux cancellation is widely regarded as a key mechanism for concentrating magnetic free energy in the corona, and in this case the team proposes that the cancellation progressively transformed a sheared arcade of field lines beneath the fan dome into a twisted magnetic flux rope, the canonical precursor of both flares and coronal mass ejections. The gradual buildup of free energy implied by this evolution is exactly the kind of slow-loading process that precedes the sudden, catastrophic release of an X-class event.

The topology analysis added a second crucial ingredient. Surrounding the fan foot-points, the extrapolation revealed a circular quasi-separatrix layer, or QSL, a volume of space where magnetic field lines connect in an extremely sensitive, steeply gradient manner. Unlike true separatrices, QSLs are not sharp boundaries but thin volumetric shells across which the connectivity of field lines changes abruptly. When magnetic stresses build up across such a layer, field lines can slip rapidly through the plasma in a process called slipping reconnection, transferring energy and threading foot-points along extended, often ring-shaped tracks on the surface. The close spatial match between the circular QSL footprint and the observed circular ribbon is a textbook validation of this picture.

Quantitatively, the team found that the ratio of current density to magnetic field strength, a proxy for how efficiently thin current sheets can form and dissipate magnetic energy, was markedly enhanced in the vicinity of the null point. This enhancement signals that the region was primed for the spontaneous development of current sheets, the thin, intense sheets of electric current where magnetic reconnection actually converts stored magnetic energy into heat, accelerated particles, and radiation. In other words, the active region contained not just the right geometry, but the right thermodynamic conditions, for reconnection to ignite and cascade.

Synthesizing these strands of evidence, the authors propose a two-stage trigger mechanism for the X1.5 flare. First, slipping magnetic reconnection within the circular QSL switched on, producing the initial circular brightening as energy was redistributed along the fan surface. Then, reconnection intensified at the three-dimensional null itself, amplifying the circular ribbon and simultaneously driving reconnection along the spine, which mapped down to the distant, remote ribbon. This sequential picture elegantly explains the observed timing and morphology of the event without requiring the filament to erupt first, and it reinforces a growing consensus that null-QSL systems are among the most efficient energy-release engines on the Sun.

Why does this matter beyond the elegant physics? X-class flares are the most powerful explosions in the solar system, capable of disrupting satellite operations, degrading GPS accuracy, and, in extreme cases, threatening power grids and astronauts. The flare of 10 May 2022 was geoeffective enough to draw multiple independent studies, and understanding its trigger mechanism refines the diagnostics that space-weather forecasters rely upon. Circular ribbon flares, in particular, may provide early warning signatures, because the fan-spine topologies that produce them can often be identified in magnetograms hours before the energy release begins. As solar observatories deliver ever-richer magnetic measurements, and as extrapolation techniques like the Non-Force-Free-Field method used here grow more accurate, the ability to recognize a loaded null-QSL system in a restless active region could shift space-weather prediction from reactive observation toward genuine anticipation. For now, NOAA Active Region 13006 stands as a vivid reminder that the Sun’s most violent outbursts are governed by invisible geometry, and that decoding that geometry is the surest path to foreseeing the next storm.

Subject of Research: Magnetic topology and trigger mechanism of an X1.5-class solar flare in NOAA active region 13006

Article Title: Investigation of an X1.5 class solar flare associated with a 3D null – QSL system in NOAA active region 13006

Article References: Kumari, D., Kumar, P., Kumar, S., Sadashiv, & Tiwary, A. R. (2026). Investigation of an X1.5 class solar flare associated with a 3D null – QSL system in NOAA active region 13006. Astrophysics and Space Science, 371(9), Article 99. https://doi.org/10.1007/s10509-026-04631-y

Image Credits: AI Generated

DOI: 10.1007/s10509-026-04631-y

Keywords: solar flare, magnetic reconnection, 3D null point, quasi-separatrix layer, circular ribbon, fan-spine topology, flux cancellation, flux rope, active region 13006, magnetic field extrapolation, space weather, Solar Dynamics Observatory