Energy Evolution from the Chromosphere to the Heliosphere in the 2021 October 28 Solar Eruption Facilities: GOES (EXIS, SUVI), OVRO:SA, SDO (HMI, AIA), SOHO (LASCO), SolO, STEREO (EUVI, Cor), WIND, Hinode (XRT), Fermi (GBM, LAT) Katharine K. Reeves, Daniel B. Seaton, Cynthia Cattell, Bin Chen (陈彬), Liam David, Federico Fraschetti, Joe Giacalone, Phillip Hess, Andryi Koval, Dana W. Longcope, Surajit Mondal, Christopher S. Moore, Sophie Musset, Tatiana Niembro, Daniel Pacheco, Yeimy J. Rivera, Soumya Roy, Xudong Sun (孙旭东), Durgesh Tripathi, Domenico Trotta, Matthew J. West, Sijie Yu (余思捷), Chunming Zhu Abstract We perform a detailed study of the energetics for a well-observed solar eruption and flare that occurred on 28 October 2021. This event included a GOES class X1.0 flare, a global EUV wave, and a coronal mass ejection that reached speeds of > > 2000 km s -1 . The event was observed from a variety of spacecraft in NASA’s Heliophysics System Observatory, including multiple missions near Earth, STEREO-A off the Sun-Earth line, and Solar Orbiter, near the Sun-Earth line at about 0.8 au. Using remote sensing, in situ observations, and in some cases scaling laws based on previous observations, we characterize the following quantities: free magnetic energy, energy in non-thermal electrons, energy in non-thermal ions, bolometric energy, energy deposited in the chromosphere, thermal energy radiated in the flare loops, energy dissipated by the EUV wave, CME kinetic and gravitational potential energy, CME energy flux in the heliosphere, and the energy partition in the CME shock. We find that the total energy released during the event is consistent with estimates of the pre-event stored magnetic energy, and the CME kinetic + potential energy dominates the energy partition. \turnoffedit I Introduction Solar eruptions serve as a pathway to convert the Sun’s stored magnetic energy into other forms, including thermal energy, kinetic energy, and gravitational potential energy. The release of stored energy manifests in a variety of ways, including heating in the chromosphere, heated flare loops in the corona, accelerated electrons and ions, EUV waves, and accelerated ejecta in coronal mass ejections. Several previous attempts have been made to characterize and understand the energy partitioning in solar eruptive events. Emslie et al. 2004 and Emslie et al. 2005 examined two well-observed solar eruptions associated with X-class solar flares that occurred on 21 April 2002 and 23 July 2002. For these events, the studies found that the CME contains the largest fraction of the released energy, though the uncertainties in the energy budget calculations are large. A more comprehensive study by Emslie et al. 2012 evaluated 38 solar eruptive events, mostly associated with M- and X-class flares, and similarly concluded that most of the energy released during solar eruptions goes into the kinetic energy of the CME. On the other hand, in a study of 399 M and X flares, Aschwanden et al. 2017 found that only approximately 7% of the released magnetic energy is partitioned into the launch of the CME. The process of accounting for the energy partitioned into different manifestations in a solar eruption is complex because the energy released can transform through multiple states (see Benz 2017 , and references therein) . For example, electrons that are accelerated in the corona by the released magnetic energy may encounter the chromosphere and deposit energy there, generating hard X-rays and heating the chromosphere. Flare loops, which appear subsequently, are also a result of this energy deposited in the chromosphere via a heat induced pressure differential that drives hot plasma up into the corona. Similarly, magnetic energy that is transformed into the kinetic energy of the CME can then be further converted into energetic particles through s shock formed as the CME propagates into space. Thus a thorough and systematic accounting of all of this energy, spanning multiple observational regimes, is required to completely address the energy partition during these eruptive events. The eruption that we study in this paper occurred on 28 October 2021, and it was associated with an X1.0 flare in AR 12887, located in the southern hemisphere near the central meridian of the Sun. A CME and an EUV wave were associated with this event, and the CME caused the first ground level enhancement (GLE) of Solar Cycle 25 ( Klein et al. 2022 ; Papaioannou et al. 2022 ) , as well as disturbances to the ionosphere ( Habarulema et al. 2022 ; Fagundes et al. 2024 ) when it passed the Earth. A unique feature of this event is that it was observed by Solar Orbiter in a part of its orbit that was roughly aligned with the Earth-Sun line. Solar Orbiter was at about 0.8 au, allowing for a thorough diagnosis of the evolution of the in situ energy fluxes as a function of time. In this work, we perform a thorough accounting of the energy partitioning of the 28 October 2021 event from the corona to the heliosphere to derive an energy budget for this eruption. In § II we present the observations of this event. In § III we catalog the energetics of the eruption, and § IV presents the discussion and conclusions. II Observations The event we study is an eruption that occurred on 2021 October 28, originating near the central meridian of the Sun in the southern hemisphere. The eruption produced an X-class flare (SOL2021-10-28T15:17), an EUV wave, and a fast CME. Below we summarize the observations of this event used in our exploration of the energetics. We combine the remote sensing and in situ measurements of multiple instruments from ground- and space-based observatories. In Figure 1 we show, in the Earth Ecliptic (HEE) coordinate system, the locations of the spacecraft: BepiColombo ( Benkhoff et al. 2021 ) , Parker Solar Probe ( Fox et al. 2016 , PSP , ) , Solar TErrestrial RElations Observatory Ahead ( Kaiser 2005 , STEREO-A , ) , Solar Orbiter ( Müller et al. 2020 , SolO , ) , Solar Dynamics Observatory ( Pesnell et al. 2012 , SDO , ) , SOlar and Heliospheric Observatory ( Domingo et al. 1995 , SoHO , ) , the Wind spacecraft ( Ogilvie & Desch 1997 ) and the Geostationary Operational Environmental Satellite ( Krimchansky et al. 2004 , GOES , ) . On Earth, we use the radio observations from the Expanded Owens Valley Solar Array ( Gary et al. 2018 , EOVSA, ) . BepiColombo and PSP were located conveniently for multi-spacecraft analysis, however, datasets contain large data gaps during the event preventing the eruption identification and were excluded within this investigation. Figure 1: Locations of BepiColombo , Parker Solar Probe , STEREO-A , Solar Orbiter and Earth ( SDO , SoHO , GOES and Wind spacecraft and EOVSA observatory) at the eruption onset time of 2021 October 28 15:00 UT. In Table 1 , we summarize the radial distance ( R ), longitude ( lon ) and latitude ( lat ) of the multiple observatories. The locations are given for the approximate eruption onset time (2021 October 28 15:00 UT) as well as when the in situ signatures are measured by Solar Orbiter (2021 October 30 22:00 UT) and Wind (2021 November 01 09:21 UT) spacecraft. Earth locates the SDO , SoHO , GOES and Wind spacecraft, as well as the EOVSA observatory. Table 1: Locations (R- radial distance, lat - latitude and lon- longitude) in HEE coordinates of BepiColombo , Parker Solar Probe , STEREO-A , Solar Orbiter and Earth ( SDO , SoHO , GOES and Wind spacecraft and EOVSA observatory) during the eruption onset time (2021 October 28 15:00 UT) and when the in situ signatures are measured by Solar Orbiter (2021 October 30 22:00 UT) and Wind (2021 November 01 09:21 UT) spacecraft. Eruption location R [au] lat [ ∘ ] lon [ ∘ ] BepiColombo at onset 0.411 89.843 6.414 at 0.81 au 0.432 96.012 6.665 at 0.99 au 0.447 99.542 6.746 Parker Solar Probe at onset 0.623 -53.419 -3.380 at 0.81 au 0.594 -53.646 -3.388 at 0.99 au 0.573 -53.678 -3.391 STEREO-A at onset 0.958 -37.465 0.075 at 0.81 au 0.958 -37.343 0.07 at 0.99 au 0.958 -37.267 0.067 Solar Orbiter at onset 0.803 -2.569 -2.78 at 0.81 au 0.819 -1.817 -2.543 at 0.99 au 0.828 -1.395 -2.392 Earth (Wind) at onset 0.993 0.0 0.0 at 0.81 au 0.992 0.0 0.0 at 0.99 au 0.992 0.0 0.0 II.1 Pre-eruptive Active Region The source region for the event, AR 12887, was already well formed when rotated onto the disk, showing a quadrupolar configuration (Figure 2 , top) as observed by the Helioseismic and Magnetic Imager ( Scherrer et al. 2012 , HMI;) on SDO . The subsequent evolution featured significant flux cancellation between the south-eastern, negative sunspot and the surrounding positive magnetic flux, which led to the disintegration of the spot (Figure 2 , bottom). Part of its negative flux moved westward and collided with a small patch of positive flux at the center (circle, see also the accompanying animation). There was strong shearing motion and apparent flux cancellation between the newly formed bipole. Figure 2: HMI normalized continuum I_{c} (left) and vertical magnetic field B_{z} for 2021 October 25 15:00 UT (top) and 2021 October 28 15:00 UT (right), just before the flare. Circles in the bottom row highlight a small bipole that undergoes intense shearing motion. The animation shows the continuous evolution of the continuum and magnetic field observations of the region from October 24 23:59 UT to October 29 22:59 UT. II.2 Flare Observations The flare SOL2021-10-28T15:17 was an X1.0 class flare as observed by the X-Ray Sensor (XRS) on GOES , shown in Figure 3 . It started at 15:17 UT and reached its peak at 15:35 UT. The flare was a relatively long duration event – even by two hours after the event peak, the GOES flux had still not returned to pre-flare values. Figure 3: GOES XRS observations of the flare. The peak time is marked with a vertical blue line, and the flare duration is indicated by the blue shading. Figure 4: AIA images of the corona in several EUV passbands before the onset of the flare (15:00 UT) and during the flare (15:28 UT). The flare was observed in the EUV by the Atmospheric Imaging Assembly ( Lemen et al. 2012 , AIA, ) on SDO and by the Solar Ultraviolet Imager ( Darnel et al. 2022 , SUVI;) on GOES . AIA provides images in seven EUV passbands (94, 131, 171, 193, 211, 304, and 335 Å) as well as several UV channels (1550, 1600, 1700 Å). Figure 4 shows AIA images of the flaring region before and during the flare in several EUV passbands. SUVI imaged the flare and associated eruption with similar EUV channels to AIA (94, 131, 171, 195, 284, 304 Å). Hinode /X-Ray Telescope ( Golub et al. 2007 , XRT;) data were saturated during the flare peak, but there was some usable data during the decay phase in the Be-thick and Be-med filters. This flare was also observed by the Expanded Owens Valley Solar Array (EOVSA), which is a 13 element microwave interferometer that makes quasi-simultaneous observations over the 1–18 GHz band. The event occurred when the Sun was just rising at EOVSA and only 9 of the 13 antennas were tracking the Sun. Hence the data quality and spatial resolution are somewhat degraded. The EOVSA observations are shown in Figures 5 . In panel a), we show the dynamic spectrum from EOVSA and panel b) shows the microwave lightcurve between 3–15 GHz with the soft X-ray lightcurve from GOES overlaid. We have saturated the colorscale to 1000 SFU. In panel c), we overlay an example image at 13 GHz over a nearby AIA 1600 Å  image. While it is evident that the microwave source location matches that of the flaring region, the microwave source is practically unresolved. Despite the limitations of the data, we are able to obtain meaningful constraints on the nonthermal energetics of the flare as is described in Section III.2.2 . Figure 5: Panel a): EOVSA dynamic spectrum. The colorbar has been saturated at 1000 SFU. Panel b): The red line shows the smoothed EOVSA lightcurve between 3–15 GHz. The soft X-ray lightcurve obtained from GOES is also shown in the same plot in blue. Panel c): EOVSA image at 13 GHz at 15:28:05 UT, shown with cyan contours over an AIA 1600 Å  image taken close in time. The flare was observed by the Spectrometer/Telescope for Imaging X-rays ( Krucker et al. 2020 , STIX,) instrument on Solar  Orbiter . In the upper panel of Figure 6 the STIX soft X-ray (4–15 keV, solid blue) and hard X-ray (15–25 keV, red dashed) contours are over-plotted on the AIA 1600 Å observation. The AIA 1600 Å observation is in negative colormap, i.e. the brighter features appear darker. The 15–25 hard X-ray contours are cospatial with the AIA 1600 Å ribbons. In the lower panel of Figure 6 we plot the light curve from various STIX channels in comparison to GOES 1–8 Å light curve. The attenuator on STIX kicks in before the soft X-ray peak of the flare to avoid saturation. The attenuator affects the softer X-ray flux considerably more than hard X-ray flux. We use the STIX spectra to obtain the constraints on the non-thermal energy deposited at the foot-points by the accelerated electrons in Section III.2.1 . The Fermi Gamma-ray Burst Monitor ( Meegan et al. 2009 , GBM,) observed the flare in similar energy bands as STIX during the decay phase, from 15:50 - 16:14 UT. Figure 6: STIX soft and hard X-ray contours overplotted on an AIA 1600 Å image with an inverted colorscale. STIX light curves from various channels are shown in the bottom panel in comparison to the GOES soft X-ray flux in 1–8 Å. The STIX light curves are made with 4 s time bins and the light travel time is taken into account. The contours in the top panel are marked with same colors as the light curves in bottom channel. II.3 EUV Wave One of the most prominent features of the eruption was a large, circular global-scale EUV Wave ( Long et al. 2017 , sometimes referred to an “EIT Wave”;) . Although such waves can usually be seen over a wide range of EUV passbands, corresponding to a wide range of temperatures, they are often most easily seen in warm coronal bands, such as the SUVI 195 Å band (which features strong contributions from Fe xi and xii at about 1.5 MK; see Darnel et al. 2022 , Table 4). Because these waves are diffuse and highly dynamic, they are generally observed most clearly in running-difference images. Figure 7 shows an overview of the wave’s evolution as seen by SUVI and the STEREO-A Extreme Ultraviolet Imager ( Howard et al. 2008 , EUVI, ) . Early in the event (upper left panel, in normal imaging) the wave is essentially circular in appearance, while later (subsequent panels) part of the wave that travels off of the solar disk retains its circular appearance and travels at essentially constant velocity, while the part of the wave observed on-disk begins to interact with the local magnetic structures and both slows and dissipates. Figure 7: 195 Å images of the EUV wave from SUVI and EUVI-A. The upper left image shows a normal view of the wave, enhanced with a radial filter ( Seaton et al. 2023 , see ) and temporal unsharp masking technique, while the remaining panels are running difference. The two rightmost panels enclosed in the dashed line show the wave from both GOES and STEREO-A perspectives, separated by about 37.5^{\circ} in longitude and 39 s in time. II.4 Coronal Mass Ejection Figure 8 shows the CME’s early evolution observed by SUVI and SoHO Large Angle and Spectrometric Coronagraph ( Brueckner et al. 1995 , LASCO) C2 instrument. The CME is a strong halo event, directed southward, with the dense, bright prominence core trailing the main CME bubble (visible especially in the second and third panels of the figure). Figure 8: Composite view of the CME using SUVI and C2 data. Arrows indicate the location of the expanding CME as it passes through the LASCO field of view. The CME in the northwest began before the onset of the halo event we study here. Note that the overlaid timestamps correspond to the SUVI image at the center of the frame, which is sampled at higher cadence than the LASCO data. The animation shows the complete observation of the eruption from 15:00U̇T to 16:59 UT on October 28. II.5 Interplanetary Shocks: multi-point in-situ measurement An interplanetary shock (IPs) associated with this eruption crossed Solar Orbiter located at an heliocentric distance \sim 0.8 au on 2021 October 30 at 22:02:07 UT and subsequently Wind on 2021 October 31 at 9:33:16 UT. Table 2 lists the shock parameters used in the present analysis. The IPs crossed both s/c in an oblique configuration ( \theta_{Bn}\sim 45-60^{\circ} at both s/c). The two spacecraft were close to radial alignment (see Fig. 1 ) at the time of the eruption. Thus, this event offers one of the first opportunities to measure the evolution with heliocentric distance of the energy partition at IPs, including not only the ram, thermal, and magnetic energies but also the contribution from energized particles (both ions and electrons). SolO in-situ measurements revealed no energetic ( \gtrsim 100 keV) ions at the shock ( Trotta et al. 2023 ; Dimmock et al. 2023 ) , which are generally expected at such a Mach number shock (see Table 2 ). In the present paper we make use of multi-spacecraft in-situ measurements to provide insight on the low and high energy cutoff for energetic particles probing the shock at a different spatial locations. Table 2: Parameters for the 2021 October 30 and 31 shock at both SolO and Wind , determined with Szabo 1994 and Viñas & Scudder 1986 technique for SolO and Viñas & Scudder 1986 technique for Wind . Parameter SolO Wind Passage Time (UT) 22:02:07 Oct 30 09:33:16 Oct 31 Speed [km s -1 ] 413.1\pm 2.8 437.7\pm 2.2 Normal (RTN) [0.743, -0.003 , [0.801, 0.299, -0.669 ] -0.518 ] Normal 1\sigma error [0.009, 0.017, [0.009, 0.015, 0.010] 0.014] \theta_{Bn} [ ∘ ] 43.7\pm 8.8 58.7\pm 8.5 M_{fms} 3.6\pm 0.3 2.59\pm 0.09 II.5.1 SolO and Wind Magnetic Field and Plasma Data For the plasma environment at SolO (see Fig. 9 ), we use the magnetic field measured with a resolution of 64 vectors / s by MAG ( Horbury et al. 2020 ) . Ion bulk flow, density and temperature are the plasma moments measured by the Proton Alpha Sensor of the Solar Wind Analyser suite ( Owen et al. 2020 , SWA-PAS,) , with a 3 second time resolution. For Wind in-situ measurements (see Fig. 10 ), we use the Magnetic Field Instrument ( Lepping et al. 1995 , MFI,) 11 vectors / s magnetic field measurements in RTN coordinates ( Koval & Szabo 2023 ) and 92-second proton density, bulk velocity and temperature data 1 1 1 These data are determined with non-linear fitting of the current distribution function from the Solar Wind Experiment ( Ogilvie et al. 1995 ; Ogilvie et al. 2021 , SWE,) suite. . These quantities are also plotted in Figures 9 and 10 . Shock parameters at SolO are computed using a least-squares fit of the full set of Rankine-Hugoniot conservation equations to the magnetic field and thermal plasma measurements ( Szabo 1994 ; Viñas & Scudder 1986 ) , assuming thermal equilibrium of electrons and protons. For the shock parameters at Wind , we use the publicly available solution from the CfA shock database 2 2 2 https://lweb.cfa.harvard.edu/shocks . Among different methods for shock parameter determination included in the CfA database, for this shock the selected method is RH08, which is a least-squares fit of the reduced set of Rankine-Hugoniot conservation equations devoid of temperature terms 3 3 3 Since the shock normal direction in the CfA database is in GSE coordinates, we convert it to RTN coordinates. . Figure 9: SolO time-series of flow velocity components in the shock frame (not in the spacecraft frame), magnetic field components in RTN coordinates, and proton density and temperature for the 2021 October 30 shock. The vertical green line in each panel denotes the time of the shock passage and the four vertical lines indicate the start and end times of the upstream and downstream averaging intervals for the IPs analysis in Section III.10 . Figure 10: Wind time-series of flow velocity components in the shock frame (not in the spacecraft frame), magnetic field components in RTN coordinates, and proton density and temperature for the 2021 October 31 shock. The vertical green line in each panel denotes the time shock of shock passage and the four vertical lines indicate the upstream and downstream averaging intervals for the IPs analysis in Section III.10 . II.5.2 SolO and Wind Energetic Particle Data For SolO , we use data from the SupraThermal Electron Proton (STEP) instrument, the Electron Proton Telescope (EPT), and the High-Energy Telescope (HET). The energy ranges for all instruments are plotted in the top panel of Figure 11 . We average the signal across all 15 SSDs present in each channel. Details on the instruments can be found in Rodríguez-Pacheco et al. 2020 and Wimmer-Schweingruber et al. 2021 . From STEP, we use the data product 4 4 4 The STEP electron differential intensities for each energy channel were calculated from the available data products as Electron_Flux_01_Mult \times (Integral_01_Flux - Magnet_Flux_01). containing proton energies 5.73-68.0\,\text{keV} in 24 bins (STEP-p) and 4.08-62.7\,\text{keV} electrons in 32 bins (STEP-e) both with a one-second cadence. Among the 4 apertures, the flux is not isotropic: the EPT partial pressures, integrated over the entire EPT energy range (see David et al. 2022 , and Sect. 30 below for details) after the shock, normalized to the Sun-facing aperture, are 1 , 0.40 , 0.55 , and 0.65 in the Sun , Asun , North , and South apertures, respectively. The greater particle pressures (or fluxes) of the two apertures Sun (facing the Sun) and South (facing southward) can be explained with the source location with respect to SolO at the time of shock passage: Sun indicates a source (shock) located inside SolO orbit and South a source mostly coming from southern heliolatitudes; consistently with this geometry, remote observations show a CME directed southward. Thus, we discard fluxes from the Asun and North apertures, for both EPT and HET, and sum the fluxes from the Sun and South apertures only, over a 2\pi solid angle. We use the data product containing 49.5\,\text{keV}-6.13\,\text{MeV} protons (EPT-p) and 31.2-471\,\text{keV} electrons both with a 5-second cadence. We use the HET data product containing 7.05-105\,\text{MeV} protons (HET-p) and 453\,\text{keV}-18.8\,\text{MeV} electrons (HET-e) both with a 5-second cadence. As with the EPT data, for HET we use only the fluxes from the Sun and South apertures. There exists overlap in the energy ranges of STEP-p and EPT-p which we removed by excluding the upper two and lower three bins, respectively. This splitting was chosen to minimize the remaining overlap of 470\,\text{eV} ( 57.8\,\text{keV}-58.3\,\text{keV} ) which is small enough to not affect the energetic proton contribution to the shock. Data are unavailable between the EPT-p and HET-p ranges from 6.13-7.05\,\text{MeV} , but due to the small fluxes and thus energy contributions of particles at such energies, this gap is likely to have no effect on our analysis. In addition, the overlap between STEP-e and EPT-e was removed by excluding the upper eight bins of STEP-e. For the Wind spacecraft, the omnidirectional flux of energetic protons (9 channels in the range 70\,\text{keV} to 6.8\,\text{MeV} ) is retrieved from Three-Dimensional Plasma and Energetic Particle Investigation (3DP) instrument (SOSP-p data product) with a 12.5-second cadence. Wind omnidirectional flux of energetic electrons data from 3DP is retrieved from three data products: ELSP-e ( 5.18\,\text{eV}-1.11\,\text{keV} ), EHSP-e ( 136\,\text{eV}-27.6\,\text{keV} ) and SFSP-e ( 27.0\,\text{keV}-517\,\text{keV} ) with 100-second, 100-second, and 12.5-second cadences, respectively ( Lin et al. 1995 ) . The overlap between ELSP-e and EHSP-e was eliminated by removing the first six EHSP-e bins, leaving a 226\,\text{eV} gap ( 1.11\,\text{keV}-1.34\,\text{keV} ) in the flux data. We did not compensate for the small overlap ( 27.0\,\text{keV}-27.7\,\text{keV} ) between SFSP-e and EHSP-e since removing either the highest EHSP-e or lowest SFSP-e energy bin would leave a large energy range ( 18.9\,\text{keV}-27.0\,\text{keV} or 27.7\,\text{keV}-40.1\,\text{keV} , respectively) unaccounted for. The energy ranges used and removed are summarized in the bottom panel of Figure 11 . All data were downloaded from the NASA CDAWeb database 5 5 5 https://cdaweb.gsfc.nasa.gov/ . Figure 11: Top panel: Energy ranges for the SolO /STEP, EPT, and HET instruments used in this study, where solid lines correspond to protons and dashed lines to electrons. The black lines denote the energy ranges that were removed due to overlap.Bottom panel: Energy ranges for the Wind data products used in this study, where solid lines correspond to protons and dashed lines to electrons. The black line over EHSP-e denotes the energy range that was removed due to overlap with ELSP-e. The black line over ELSP-e indicates the energies that correspond to the thermal electron spectrum and were also removed. III Results: Energetics Figure 12: Energy partitioning and flow in the 28 October 2021 event. Underlying CME cartoon modified from Reeves & Forbes 2005 . In this section, we undertake the ambitious task of cataloging the energy build-up and release for the well-observed eruption that occurred on 28 October 2021, originating from AR 12887. Figure 12 summarizes the energy flow over the course of the event. Stored magnetic energy is initially converted into the kinetic energy of the CME, direct heating in the supra-arcade region, and accelerated particles due to reconnection. These quantities then cascade into other forms of energy as the event progresses, such as the energy deposited in the chromosphere that eventually drives heated plasma into flare loops, and accelerated particles across the CME shock. We do not discuss the accelerated particles escaping anti-sunward due to reconnection in this paper. These particles represent a small fraction of the total event energy; a detailed discussion of these particles is given by Klein et al. 2022 . Table 3: Energy partition into flare- and CME-associated strucures. Type Energy a a Energies are in units of 10 31 ergs. Method Primary Magnetic free 8.2 \pm 2.6 NLFFF Flare Nonthermal Energy electrons (footpoints) 1.6 \pm 0.4 STIX electrons (loops) 0.05 {}^{+0.01}_{-0.02} EOVSA electrons (decay) 0.01–0.02 Fermi/GBM ions 0.6 \pm 0.8 Scaling Law Bolometric Energy total 1.6–3.0 Scaling Law Deposited in Chrom. prompt 1.2 \pm 0.4 UFC tail 0.5 \pm 0.15 UFC Thermal Loop Energy peak (§ III.6.1 ) 0.6 GOES Flux peak (§ III.6.2 ) 0.49^{+0.07}_{-0.06} DEMs tail 0.3 GOES Flux CME Kinetic @ 14.87 R ⊙ 4.12 \pm 2.77 COR2 data Potential @ 14.87 R ⊙ 4.73 \pm 2.33 COR2 data EUV Wave (total) 0.9 \pm 0.4 Patsourakos & Vourlidas (2012) Total Energy a a Energies are in units of 10 31 ergs. Fraction Magnetic free 8.2 \pm 2.6 100% Flare 2.25 \pm 0.9 13–56% CME 9.75 \pm 3.6 > 56% Using the available observations coupled with scaling laws derived from from previous studies of ensemble events ( Kretzschmar 2011 ; Emslie et al. 2012 , i.e.) , we make estimates of the following quantities, organized by section: § III.1 : Magnetic energy § III.2 : Energy in non-thermal electrons § 8 : Energy in the non-thermal ions § III.4 : Bolometric energy § III.5 : Energy deposited in the chromosphere § III.6 : Thermal energy radiated in the flare loops § III.7 : Energy in the EUV wave § III.8 : CME kinetic & gravitational potential energy § 4 : CME energy fluxes in the heliosphere § 30 : Energy partition at CME shock These calculations give us a very detailed diagnosis of the energy partition in this event, tracking energy flow all the way from storage, prior to the onset of the eruptive flare, through many key phenomena during the event, and beyond, to the energy flux in the CME as it passes through the heliosphere. In several cases, we use multiple methods or estimates of related parameters to validate our energy calculations and find consistent values, which provides confidence that the estimates we report here are reasonable. Table 3 summarizes the full accounting of the energy partition into the primary forms of energy present in flare- and CME-associated structures for this event. It also presents total energies for the flare and CME components. The total flare energy is derived from the energy in the nonthermal electrons (§ III.2 ) and ions (§ 8 ). The total energy partitioned into the CME is given by the kinetic and potential energy (§ III.8 ) and the energy dissipated from the EUV wave (§ III.7 ). The range in the Fraction column is derived based on the extrema of the values, given the errors. Direct heating of plasma in the supra-arcade region is also a possible primary result of the eruption reconnection (see Figure 12 ), but we cannot measure this quantity directly due to the superposition of supra-arcade structures and flare loops, which are a secondary structure formed due to chromospheric evaporation. However, studies examining limb flares have shown that the thermal energy release due to the direct heating component is small compared to the thermal energy found in the flare loops ( Roy et al. in review ) . We note that the total energy expended during the event is not simply the sum of the items in Table 3 , as many lines essentially count the same energy in different ways. For example, much of the flare-associated heating is driven by the energetic particles, and thus adding these two lines would double count much of this energy. Details of the calculations of each of the energy terms are given in the following sections. III.1 Magnetic Energy Figure 13: Magnetic properties of AR 12887. From top to bottom: Stonyhurst longitude of the AR centroid, unsigned magnetic flux, net electric current, mean torsional parameter \alpha , magnetic energy from extrapolations, and ratio between the magnetic free energy ( E_{f} ) and the potential energy ( E_{p} ). Blue, pink, and orange symbols show the results for NLFFF, potential field, and magnetic free energy, respectively. The error bands for the second to fourth panels are 3\sigma values from the SHARP database. The six magnetic field model times are 1 , 2 , 3 , 3.25 , 3.5 , and 3.63 days after Oct 25 00 UT. The error bars for the last two panels show the minimum and maximum of all the extrapolation models that pass the solenoidality test at each time. No error bar is provided for the last time step because only one model passes the test. The vertical dotted line indicates the GOES flare start time. Figure 14: Magnetic energy flux of AR 12887 based on the PDFI method. Top: instantaneous energy flux. Bottom: accumulated energy since 2021 October 24 00:00 UT. The values are from the PDFI database ( Hoeksema et al. 2020 ) . Red curves show 2\sigma formal uncertainty estimate based on a Monte-Carlo method; the uncertainty is small for the accumulated energy. Flux is taken to be zero during HMI data gaps. The vertical dotted line indicates the GOES flare start time. We evaluated the photospheric magnetic properties of AR 12887 using the Space-weather HMI Active Region Patches ( Bobra et al. 2014 , SHARP;) over five days, shown in Figure 13 . As a consequence of magnetic flux cancellation, the extensive properties (e.g., totals) generally decreased over time. For example, the unsigned magnetic flux and the net electric current decreased by about 34\% and 66\% , respectively. On the other hand, the intensive properties (e.g., means), such as the mean torsional parameter (measure of magnetic twist), exhibited an increase starting late on 2021 October 27, when the central small bipole started to form. As apparent flux cancellation commenced in the small bipole on 2021 October 28, mean \alpha started to decrease as well, dropping by about 69\% over the next 1.4 days. To assess the coronal magnetic field energy, we performed nonlinear force-free field (NLFFF) and potential field extrapolations ( Wiegelmann et al. 2012 ) at selected times leading to the flare. For each time step, four different models were calculated using different sets of free model parameters following Thalmann et al. 2020 . The differences between these four models allow for an estimate of the systematic error. Further, we calculated the energy for the non-solenoidal component of the modeled coronal field ( Valori et al. 2013 ) , and rejected those with large errors. The spread is generally small for the NLFFF ( E_{n} ) and PF ( E_{p} ) energy: the maximal difference between models is 2\% – 8\% of the mean. However, the spread is much larger for the magnetic free energy E_{f} , defined as the difference between the NLFFF and potential field energy ( E_{n}-E_{p} ). The spread can be as large as 70\% of the mean: significant systematic uncertainty is expected. Additionally, uncertainty due to spectropolarimetric noise can be estimated from pseudo Monte Carlo method. For HMI magnetograms, different noise realizations will lead to a typical relative error of 10\% - 20\% for E_{f} ( Sun et al. 2012 , e.g.,) . We note that the systematic uncertainty here is likely a lower bound. The widely used potential field extrapolation method adopts a Neumann boundary condition, that is, the vertical field B_{z} on the lower boundary is required to match the observation exactly. This approach overfits the magnetogram in the presence of noise, and the consequence is not entirely clear. In an illuminating example, Welsch & Fisher 2016 combine the Neumann and the Dirchlet (in the form of \nabla_{h}\cdot\bm{B_{h}} ) boundary condition weighted by their relative noise level. The method seeks a solution that best fits the observed magnetic field vector in a statistical sense, and the extrapolated B_{z} is statistically similar to the observation in terms of noise. Surprisingly, the “hybrid” potential field solution has a much lower magnetic energy than the fiducial method (by 15\% — 38\% ), which suggests that our free energy may be significantly underestimated. Over the 3.6 days leading to the flare, we found that E_{n} decreases from 1.1\times 10^{33} to 6.8\times 10^{32} erg, by about 4.6\times 10^{32} erg, or 40\% (Figure 13 ). This decrease is closely related to the apparent flux cancellation. The magnetic free energy at 2021 October 28 15:12 UT, right before the X-class flare, is about 8.0\times 10^{31} erg, about 13% of the potential field energy. We additionally assessed the change of coronal magnetic energy by integrating the estimated magnetic energy flux through the photosphere. To this end, photospheric electric field \bm{E} was to be inferred using the “PTD-Doppler-FLCT Ideal” ( Kazachenko et al. 2014 , PDFI;) technique. The vertical component of the Poynting flux vector \bm{E}\times\bm{B} was then integrated over the AR and in time. We found significant negative energy flux from 2021 October 25–27, and on 2021 October 28 (Figure 14 ), consistent with the drop in photospheric proxies (Figure 13 ) and is likely caused by the magnetic flux cancellation. The integrated energy flux from 2021 October 25 00:06 UT to 2021 October 28 15:06 UT is estimated to be 1.5\times 10^{33} erg, about 3.2 times the change of energy from the coronal extrapolation model E_{n} . We have performed an additional Monte Carlo test to estimate the effect of random noise on the estimate ( Avallone & Sun 2020 ) . The effect is rather small on the accumulated energy; we expect the system uncertainty that is not well understood to dominate. III.2 Energy in Non-Thermal Electrons III.2.1 Non-thermal Energy from X-rays The energy in the non-thermal electrons is estimated from the STIX observations. Figure 6 shows that the majority of the emission from hard X-rays comes from the two footpoints in this event. Following Emslie et al. 2012 , we perform a spectral fit to the STIX observations, to constrain the injected electron spectra F_{0}(E_{0}) . We have used the thermal model ‘ f_vth ’ and cold thick target model ‘ thick2 ’ publicly available in the OSPEX package to fit the STIX spectrum as a function of time. Throughout the fitting process, only the high-energy cutoff was kept fixed at \mathrm{3.2~\times~10^{4}~keV} . The typical low energy cutoff is within the 10–15 keV range. ( Roy et al. in review , For further details regarding the fitting procedure, please refer to) . The constrained electron energy distribution spectra is used to estimate the instantaneous energy deposited by the electrons in the foot points as: U_{e}(t=t^{\prime})~=~A~\Delta t\int_{E_{min}}^{E_{max}}~E_{0}F_{0}(E_{0})(t=t^{\prime})dE_{0}, (1) where the electron energy distribution in constrained by fitting the count spectra for the time interval t^{\prime}-\Delta t/2\leq t\leq t^{\prime}+\Delta t/2. (2) The cumulative non-thermal energy deposited at the foot points as a function of time is estimated by adding the instantaneous non-thermal energy from every time bin: U^{cumulative}_{e}(t)~=~\sum_{t^{\prime}=t_{0}}^{t}U_{e}(t^{\prime}). (3) The cumulative non-thermal energy as a function of time, as inferred from STIX observations is shown in Figure 15 (solid line with yellow shading) in comparison to cumulative non-thermal energy constrained from EOVSA microwave observations (discussed in § III.2.2 ) shown with blue triangles. The energy deposited via the non-thermal electrons at the foot-points levels off at \sim~(1.6~\pm~0.4)\times 10^{31} ergs. The uncertainty in the energy deposited is calculated by propagating the fit parameter uncertainties in Equation 1 . Figure 15: Comparison between nonthermal energies calculated from STIX data and EOVSA. Yellow shading on the STIX data indicates the error bounds. Note that the EOVSA energies are multiplied by a factor of 10. During the decay phase starting at 15:50 UT, we use Fermi GBM to calculate the cumulative energy for non-thermal energy in electrons above 15 keV. We use two different cutoff energies fixed in the spectral fitting, one at 15 keV and one at 20 keV. These values represent the spread obtained when the cutoff energy is a free parameter in the fitting process. Both values lead to equally satisfying \chi^{2} values. We find that the maximum value of the cumulative energy from non-thermal electrons in this phase is about 0.01-0.02\times 10^{31} ergs. Unfortunately energetic particle hits on the STIX detector after 15:50 UT make that data unreliable, so it is not possible to directly compare the two measurements. III.2.2 Microwave spectral diagnostics The microwave spectra obtained from EOVSA enable us to constrain the properties of the nonthermal electron distribution during the flare based on the nonthermal gyrosynchrotron emission model. The gyrosynchrotron emission spectrum depends on several parameters: the vector magnetic field, the nonthermal electron distribution, the density and temperature of the thermal plasma, the area of the emission source, and the depth along the line of sight (LOS). Following Chen et al. 2020 , we perform a Markov chain Monte Carlo (MCMC) analysis to explore this parameter space and fit the observed spectra. For the spectral modeling, we use the fast gyrosynchrotron modeling codes developed in Fleishman & Kuznetsov 2010 and Kuznetsov & Fleishman 2021 . We assume that the nonthermal electron distribution is isotropic and and has a power-law form f(E)=dn_{\rm nth}/dE\propto E^{-\delta^{\prime}} , where n_{\rm nth}=\int_{E_{\rm min}}^{E_{\rm max}}f(E)dE is the total nonthermal electron density, with {E_{\rm min}} and {E_{\rm max}} being the low- and high-energy cutoff, respectively, and \delta^{\prime} is the power-law index. E_{\rm max} are fixed at 10 MeV respectively. We denote the magnetic field as B , density of thermal electrons as n_{th} , angle between the magnetic field and LOS by \theta . We assume that the temperature of the gyrosynchrotron emitting plasma is 7 MK and its depth along the LOS is 7.2 Mm. Due to poor spatial resolution of the data, we have modeled the integrated flux density of the flaring source. While the flare source is inhomogeneous in general, we assume a homogeneous distribution of plasma parameters within the source to limit the degrees of freedom during the spectral modeling procedure to a manageable number. This approximation holds reasonably well in case of well-resolved sources, but the poor spatial resolution of the EOVSA data in this case was an issue. In particular, assuming the entire flare source has the same source area over 3--15 GHz, over the emission height is an over-simplification. We have devised a simple method to take into account the possible variation in source area during spectral modeling. We assume that the fractional error incurred due to our assumption of equal area over the entire frequency range will be same as the fractional change of the total area of the source over the same frequency range. The total source area was obtained from the images produced using standard techniques 6 6 6 http://www.ovsa.njit.edu/wiki/index.php/Calibration_Overview , https://github.com/suncasa/suncasa-src/blob/master/examples/eovsa_flare_slfcal_example.py . In Figure 16 , we show the variation of the microwave spectra with time 7 7 7 The flux scale of EOVSA is determined regularly by matching the observed EOVSA flux with that reported by the Radio Solar Telescope Network (RSTN). The RSTN flux products do not include formal statistical uncertainties. Inter-comparisons between RSTN sites and with independent instruments indicate that the reported flux densities are typically accurate to within \sim 5–10% ( Giersch & Kennewell 2022 ) . Therefore, we have taken a fiducial 10% systematic uncertainty on the absolute flux, shown as the error bars in the figure. . We find that around 15:27 UT, the spectral peak is at frequency \lesssim 3 GHz, indicating small quantities of nonthermal electrons prior to 15:27 UT, which is also during the start of the flare. At 15:29 UT, the peak of the spectrum is about 5 GHz, indicating enhancement in the number density of nonthermal electrons. Around 15:31 UT, the spectrum becomes flat, a feature that persists at 15:33 UT. This flattening indicates that by this time, not many nonthermal electrons are present in the system, which is also consistent with the STIX results (see Section III.2.1 ). We have used the MCMC chains to constrain the evolution of the energy flux of nonthermal electrons, with energy above 20 keV. The results are shown in Figure 15 , compared with the energy derived from STIX. The error bars correspond to the 67.5 % confidence interval. We note that substantial uncertainties of the total energy flux of the microwave-emitting electrons may also arise from the assumed homogeneous source and single-power law distribution of the source electrons. We find that while the total energy estimated from the microwave data is much smaller than that estimated from STIX data, likely due to the fact that X-rays and microwaves probe different regions of the flare. While the microwave observations primarily probe nonthermal electrons trapped inside the flaring loop, at the loop top region, X-ray observations probe nonthermal electrons which are deposited at the chromosphere. Nevertheless, interestingly, we find that the temporal evolution of the nonthermal energy estimated from both X-rays and microwave data is qualitatively similar in that they both start to flatten around 15:30 UT. Figure 16: The temporal evolution of the microwave spectrum from EOVSA. The flattening of the spectra as time goes on indicates a decreasing number of nonthermal electrons in the flare loop system. III.3 Energy in Non-Thermal Ions The best way to measure flare-accelerated ions is using the 2.223 MeV neutron-capture gamma-ray line, for example as observed by RHESSI ( Shih et al. 2009 , e.g.) . Unfortunately this data does not exist for the 28 October 2021 flare, so we use a scaling law found by ( Emslie et al. 2012 ) for an ensemble of fourteen events with suitable RHESSI data: E_{nti}/E_{nte}=0.34\pm 0.5 (4) where E_{nte} is the energy in the nonthermal electrons and E_{nti} is the energy in the nonthermal ions. This relation was also used by Aschwanden et al. 2017 in their estimates of nonthermal ion energy. Using Equation 4 , we find that the energy in the nonthermal ions is on the order of 0.6\pm 0.8\times 10^{31} ergs. Together with the accelerated electron energies found from EOVSA and STIX, this estimate gives a total of \sim 2.25\times 10^{31} ergs of energy in the accelerated particles in this event. It is worth noting that this event was detected with Fermi Large Area Telescope ( Atwood et al. 2009 , LAT,) showing the presence of ions 8 8 8 See https://hesperia.gsfc.nasa.gov/fermi/lat/qlook/ max_likelihood/2021/10/lat_maxlike_20211027_20211030.png with energies of \geq 100 MeV. Klein et al. 2022 speculate that the long duration gamma-ray emission in this event could be due to the partial trapping and mirroring of relativistic protons in large-scale coronal magnetic fields. III.4 Bolometric Energy Figure 17: Bolometric energy vs. GOES Flux. Events used in Emslie et al. 2012 are shown as asterisks, and events used in Kretzschmar 2011 are shows as diamonds. The fits to all events and to the Kretzschmar events alone are shown as solid and dashed lines, respectively. The estimate of the bolometric energy for the 28 October 2021 event from the Kretzschmar fit ( 2.1\times 10^{31} ergs) is shown as a square and that from the Emslie+Kretzschmar fit ( 3.9\times 10^{31} ergs) is shown as a triangle. The GOES flux of the 28 October flare has been rescaled to be consistent with the older observations. Most of the energy radiated from the flare should come from the bolometric energy, which is the integrated irradiance over the entire electromagnetic spectrum. Thus the bolometric energy provides a useful check on the energies of the non-thermal particles, since these particles hitting the lower atmosphere produces the radiation ( Warmuth & Mann 2020 , e.g) . Previous accounts of energy partitioning in solar eruptions have found relationships between the bolometric energy and the peak GOES flux. Emslie et al. 2012 uses measurements for a few flares from the TIM instrument on SOURCE combined with estimates from an empirical model for the rest of their events. Aschwanden et al. 2017 uses the relationship found by Kretzschmar 2011 , who calculated the total solar irradiance of set of 2100 superimposed flares. We use these previous observations to determine the bolometric energy for the 28 October 2021 event. Figure 17 shows the 28 October flare compared to previous estimates of the relationship between X-ray irradiance and bolometric energy by ( Kretzschmar 2011 ) and Emslie et al. 2012 . Following Aschwanden et al. 2017 we derive scaling-law relationships between X-ray flux and bolometric energy for both the Kretzschmar flare distribution and total population of flares reported by both Emslie and Kretzschmar. These relationships show that the bolometric energy for an X1.0 flare is between 2.1-3.9 \times 10^{31} ergs. Note that reanalysis of the relationship between operationally reported flare class and calibrated irradiance indicates a correction factor of 1.43 should be applied to X-ray flux measurements obtained by GOES-15 and earlier ( Woods et al. 2024 , see the discussion in the introduction of) , while more recent data, including the irradiance reported for the October 28 flare, do not require this correction. To maintain the consistency of the previously data in the figure and its the source papers, we did not apply this factor to Figure 17 . Instead, we reduce the flare class for the October 28 flare by a factor of 1.43 (to 7\times 10^{-5}\,\mathrm{W/m^{2}} ) to account for the underestimate of irradiance in historical data; thus the bolometric intensities for the October 28 flare are 3.0\times 10^{31} ergs for the Emslie+Kretzschmar fit and 1.6\times 10^{31} ergs for the Kretzschmar-only fit. Both of these values fall well within the uncertainty of the overall distribution. III.5 Energy Deposition in the Chromosphere While there is not appropriate irradiance data to gauge the total bolometric energy in this event, we can infer the energy deposition in the chromosphere from the AIA 1600 Å images of flare ribbons using the Ultraviolet Footpoint Calorimeter method ( Qiu et al. 2012 ; Liu et al. 2013 , UFC,) . Recent work by Qiu 2021 has shown that this method gives energies that are comparable to the bolometric energies shown in Figure 17 . Thus the UFC method serves as a useful check on the estimate of the bolometric energy provided in III.4 . For the 28 October X1.0 flare, there are two roughly parallel flare ribbons traced in magenta and cyan over the HMI magnetogram in Figure 18 a. The north (south) ribbon lies in positive (negative) flux. Ribbon pixels are identified by their brief, impulsive brightening in AIA 1600 Å, as illustrated in Figure 19 a. We find 14,760 different pixels fitting this criteria. Following the classical pattern of a two-ribbon flare, successive brightening sweeps outward, as seen in Figure 18 b. The ribbons move very rapidly from 15:25 – 15:30 UT to cover the innermost regions (i.e. the blue pixels) and have mostly ended by 15:40 UT (yellow). (a) (b) Figure 18: The chromospheric flare ribbons. (a) is the HMI line-of-sight magnetogram between \pm 300 G in greyscale. Ribbons from 15:31 UT traced shown as magenta and cyan curves. Regions enclosing most of the ribbons are outlined in red and yellow. (b) A map of all ribbon pixels, showing the time of the peak emission as color, according to a scale shown in a color bar. The ribbon locations from 15:31 UT traced shown as magenta and black curves, matching those in panel (a). (a) (b) (c) Figure 19: The UFC procedure for inferring power from light curves. (a) shows the ribbon in AIA 1600 Å on a logarithmic gray scale. Three pixels are identified by colored diamonds. The light curves from these pixels are plotted in an inset, although without accounting for solar rotation. (b) shows how the light curve of a given ribbon pixel, in this case the cyan pixel from (a), is translated into an energy flux. The light curve is plotted in cyan in the bottom panel, after accounting for co-rotation. Its conversion to energy flux is shown as a black curve in the upper panel. The prompt and tail components are over-plotted in magenta and orange respectively. (c) shows the full set of GOES, AIA, and XRT light curves (black) along with the version synthesized using UFC. Each of the 14,760 ribbon pixels is interpreted as the footpoint of a flare loop which has been impulsively energized. The light curve of each pixel (DN/s), computed after accounting for solar rotation, is converted to energy flux ( {\rm erg\,cm^{-2}\,s^{-1}} ) as shown in Figure 19 b. This calculation is done using two empirical parameters ( Qiu et al. 2012 ) . The resulting heating profile is fed as input to the highly efficient, zero-dimensional Enthalpy-Based Thermal Evolution of a Loop model ( Klimchuk et al. 2008 , EBTEL,) to derive the time evolution of the half-loop. The time profiles of temperature and density are used to synthesize that half-loop’s contribution to each of the AIA passbands, XRT filters, and both GOES X-ray channels. This procedure is repeated for each of the 14,760 ribbon pixels, to build up the full-flare light-curve of each passband, as shown in Figure 19 c. The empirical parameters are then adjusted to optimize the agreement between observed and synthesized light curves. After a number of iterations we obtain the reasonable agreement shown in Figure 19 c, and take the energy profile (i.e. Figure 19 b) as the actual flare energy flux incident on the chromosphere. Owing to the nature of EBTEL, this flux may take the form of either thermal conduction or non-thermal particles; it is simply the flux of flare energy into that pixel. The light curve of a ribbon pixel typically consists of a short, intense peak, followed by a slower tail (see the bottom panel of Figure 19 b). These two components have been interpreted as the results of prompt energy flux from magnetic reconnection, followed by a long tail of lower energy input ( Qiu & Longcope 2016 ; Zhu et al. 2018 ) . The energetic contributions of these two components are distinguished as magenta and green curves in the upper panel of Figure 19 b. Combing the UFC energy fluxes from all 14,760 ribbon pixels gives the power plotted in Figure 20 b, with the separate contributions of the prompt and tail components plotted in magenta and green respectively. Figure 20: Time history of the energy inferred from UFC. (a) shows the accumulated energy of the total (black) and the prompt components (magenta). (b) is the power of the total input (black) prompt component (magenta) and tail component (green). (c) shows the rate of magnetic reconnection in the positive (red) and negative (blue) ribbon footpoints. The ribbons reveal the progress of magnetic reconnection which is assumed to power the flare. The reconnection rate is computed by mapping the HMI magnetogram to each ribbon pixel, and attributing that pixel’s flux to the time the AIA 1600 Å light curve first crosses half maximum ( Forbes & Priest 1984 ; Poletto & Kopp 1986 ) . The resulting reconnection rate is shown separately for the positive (red) and negative (blue) ribbon in Figure 20 c. The two versions agree to a reasonable extent, although the stronger, more compact nature of the negative flux makes that slightly more reliable. In the end, \Delta\Phi\simeq 3.3\times 10^{21} Mx of flux is reconnected — about 20% of all flux in the active region. The peak rate, \dot{\Phi}\simeq 10^{19} Mx s -1 , is at the higher end of rates observed in other large flares ( Jing et al. 2005 ; Zhu et al. 2020 ) , confirming our assumption of an unusually impulsive flare. The reconnection occurs primarily during the flare’s impulsive phase (15:25 – 15:35 UT), and has largely subsided by 15:40 UT, when the radiative cooling phase begins. The energy flux inferred from UFC, plotted in Figure 20 b, reflects the reconnection history shown in Figure 20 c. It is not surprising that the prompt component, plotted in magenta, resembles the reconnection curves, and has largely subsided by 15:40 UT. Its integral, plotted in the same color on Figure 20 a, levels off at \Delta E\simeq 1.2\times 10^{31} ergs. The tail contribution, on the other hand, builds up over the evaporative phase (15:30 – 15:40 UT), peaking at P_{\rm tail}=0.3\times 10^{28}\,{\rm erg/s} at 15:40 UT. Its integrated contribution accounts for the difference between the black and magenta curves in Fig. 20 a. This difference reaches \Delta E_{\rm tail}=0.5\times 10^{31} ergs by 16:40. The combination of the peak and tail energies calculated in this manner is 1.7 \times 10^{31} ergs, very similar to the lower value of 1.6 \times 10^{31} ergs estimated for the bolometric energy in § III.4 . III.6 Thermal Energy We use two different methods to approximate the thermal energy of the plasma heated during the eruption. In the first method (section III.6.1 ), we use the flux from the GOES XRS and a single-loop approximation to estimate the physical quantities necessary to derive the thermal energy. In section III.6.2 , we employ differential emission measure calculations using AIA and SUVI imaging data to estimate these quantities. III.6.1 Single loop approximation The thermal energy of the flare loops can be obtained by using the two GOES X-ray channels to derive temperature, T , and volume emission measure, EM , ( Thomas et al. 1985 ; Garcia 1994 ) shown in Figure 21 a, after assuming coronal abundances. A phase portrait of T vs. EM , shown in Figure 21 b, reveals this flare’s evolution (clockwise as indicated by the red arc) to be unusually similar to that of a single impulsively-heated loop ( Jakimiec et al. 1992 ) . Adopting this interpretation provides an estimate of the flare’s energetics without recourse to imaging. Figure 21: Global characteristics of the flare from GOES . (a) shows the evolution of the temperature (green) and emission measure (violet) derived from the two X-ray bands. The bottom axis is in minutes from temperature peak, while the top is UT on 2021 October 28. Evolution under mechanical equilibrium is shown by dashed curves of corresponding colors following the triangle. (b) is a phase portrait formed by plotting T vs. EM in green. The red arc indicates the clockwise sense of evolution. Dashed lines show the ideal characteristics of evaporation (black) and radiative cooling (magenta). The beginning of each phase is marked by a symbol, which appears in (a) to show the times of these phases. The phase portrait shows that evaporation ends at approximately 15:41 UT, with T_{B}=10.9 MK as indicated by the triangles in Figure 21 . After this the loop remains in mechanical equilibrium, with no gradient in the plasma pressure p ( \partial p/\partial\ell\simeq 0 ), cooling radiatively to reduce pressure p and peak temperature T_{\rm pk} . While traditional equilibrium treatments assume a uniform volumetric heating, h , time evolution under mechanical equilibrium differs through the addition of -\partial p/\partial t , which must be uniform, like the pressure itself. Thus, even as it remains in mechanical equilibrium, the loop cools according to {d\,T_{\rm pk}\over dt}={2\,T_{\infty}^{3/2}\over\tau_{c,B}\,T_{B}^{1/2}}\left[\left({T_{\infty}\over T_{\rm pk}}\right)^{2}-\left({T_{\rm pk}\over T_{\infty}}\right)^{3/2}\right]~~, (5) where T_{\infty}~=~0.69\left({hL^{2}\over\kappa_{0}}\right)^{2/7}~~, (6) is the asymptotic temperature (i.e. when \partial p/\partial t=0 ) and \tau_{c,B} is the conductive time at the beginning of the radiative cooling phase. The green dashed curve in Figure 21 a, is the solution with \tau_{c,B}=92 min and an asymptotic temperature T_{\infty}=8.0 MK, found by fitting the curve. Demanding that the cooling leg remains in equilibrium (details are relegated to an appendix) yields an estimate of the full loop length L=102 Mm, and requires V/L^{2}\simeq 3.0 Mm, from which we obtain the flare volume, V=3.1\times 10^{28}\,{\rm cm}^{3} , without using an image. The analysis above and in the appendix can be used to compute the energy evolution of the flare. The net thermal energy, U_{\rm th}~={3\over 2}\,p\,V~=~3\,k_{\rm b}\,T\,{\sqrt{EM\,V}}~~, (7) plotted as a red curve on Figure 22 a, peaks at U_{\rm th}\simeq 0.6\times 10^{31}\,{\rm ergs} at 15:33:20 , which is about half of the peak energy calculated via the UFC method in § III.5 and shown in Figure 20 b. The power lost through conduction or radiation, P_{c} or P_{r} respectively, is computed by dividing U_{\rm th} by the corresponding time ( Longcope et al. 2010 ) , as shown in Figure 22 b. The evaporative phase is dominated by conduction ( P_{c}\gg P_{r} ) driving the evaporation which carries the energy back into the loop — no energy is lost. The radiative cooling phase is characterized by P_{c}\simeq P_{r} (see Figure 22 b) both of which represent losses from the system: P_{r} is lost to radiation from the corona while P_{c} is conducted out of the corona to the feet where it is also radiated. Figure 22: Evolution of energy (a) and power (b) and (c), as computed from GOES light curves. (a) shows the evolution of thermal energy, U_{\rm th} , in red. The diamond and triangle mark the beginning and end of the evaporative phase, as in Figure 21 . The shaded region integrates the power plotted as shaded in (c), and is added to the peak of U_{\rm th} . (b) shows the power lost to radiation (blue) and thermal conduction (green). During the evaporative phase these combine to a net loss, shown in violet. (c) plots the change in thermal energy (red) and the losses as a negative contribution. The grey region between the curves is the power that must be supplied to keep the thermal energy from falling too rapidly during radiative cooling. This difference is plotted as a black curve. During the radiative cooling phase, the combined losses exceed the observed decrease in thermal energy, as indicated by the gray region in Figure 22 c. This difference must be somehow supplied to the flare plasma even during the so-called cooling phase to account for the slow rate at which U_{\rm th} is observed to decrease. This extra heating accounts for the fact that the loop is asymptotically approaching a very high temperature, T_{\infty}=8.0 MK, as shown in Figure 21 a. The net power required to maintain this temperature is \displaystyle P_{\rm tail} \displaystyle= \displaystyle h\,V~=~3.6\,\kappa_{0}\,T_{\infty}^{7/2}\,{V\over L^{2}} (8) \displaystyle\simeq \displaystyle~0.16\times 10^{28}\,{\rm erg\,s^{-1}}~~, after inverting Equation 6 and using V/L^{2}=3 Mm. This power is consistent with the observed difference, plotted as a black solid line on Figure 22 c. When that power is integrated we obtain a persistent contribution to the energy shown as a shaded region in Figure 22 a. This contribution continues to grow, even as U_{\rm th} decreases. As of 16:40 UT about 0.3\times 10^{31} ergs has been supplied, raising the total flare energy by 50% to 0.9\times 10^{31} ergs. Even at this late time the energy appears to be continually growing. The cooling-phase contribution inferred from the GOES light curves ( 0.3\times 10^{31} ergs, Figure 22 a), is comparable to that derived from the UFC method in § III.5 ( 0.5\times 10^{31} ergs, Figure 20 a). Indeed, the qualitative and quantitative similarity of Figures 20 a and 22 a, derived using independent analyses, gives support to their energy profiles, including the tail heating component, which contributes to the energetics of each loop. Interestingly, these values are about one order of magnitude greater than the energy in the non-thermal electrons during the decay phase as estimated from the Fermi GBM data. III.6.2 differential emission measure approximation As a second method of calculating the thermal energy, we estimate the temperatures and emission measures of the event by calculating the differential emission measure (DEM) of the flaring region using imaging data. To compute DEMs we use both AIA EUV passbands (94, 131, 171, 193, 211, 335 Å) and the high-dynamic-range SUVI passbands (94, 131, 171, 195 Å) in frames where the AIA observations showed signs of major saturation. We used the regularized inversion method ( Hannah & Kontar 2012 ) to calculate the DEMs. In Figure 23 we show the flare observed in AIA 193 Å during the decay phase (panel a) and the corresponding inverted emission measure (EM) (panel b) and DEM-weighted mean temperature (panel c). Figure 23: Panel (a): The flare arcade in AIA 193 Å. Panel (b): the emission measure of the region for 5~40 %, are good candidates – though even a reduction by a factor of two would still support the conclusion of Emslie et al. 2012 that CMEs represent the majority of the energy released by an eruptive flare. There is evidence that EUV waves are CME-driven, and that more energetic CMEs are associated with faster, more impulsive EUV waves ( Muhr et al. 2014 ) . Thus the strong, fast EUV wave associated with the event is also consistent with the energy distribution for an event in which the CME carries the majority of the total energy. Additionally, it is possible that our magnetic free energy is an underestimate (see discussion in Section III.1 .) Warmuth & Mann 2020 point out that the bolometric energy is a useful check on the total energy contained in the flare part of the eruption. We did not measure bolometric energy directly, but it can be estimated using the previous result of Emslie et al. 2012 and Kretzschmar 2011 . Using this method we found a bolometric energy in the range 1.6–3.0 \times 10^{31} ergs. Additionally, the UFC method has been found to be a reasonable proxy for the bolometric energy ( Qiu 2021 ) , and using this method, we find an energy of 1.2 \times 10^{31} ergs at the peak of the flare, with another 0.5 \times 10^{31} ergs in the tail, consistent with the estimate of the bolometric energy within errors. For our event, the total energy in the nonthermal particles is estimated to be \sim 2.25 \times 10^{31} ergs, which is within the calculated range of bolometric energies, thus verifying that our estimates of the nonthermal energy are reasonable. Previous estimates of the ratio of nonthermal energy to the bolometric energy have been in the range of 10-60% ( Warmuth & Mann 2020 ) . An outlier is the study of Aschwanden et al. 2017 , which finds that the nonthermal energy is over 5 times the bolometric energy. However, the nonthermal energies for the events in this sudy were revised significantly downward in a subsequent paper ( Aschwanden et al. 2019 ) , leading to more typical ratios of \sim 35%. Our estimates put the ratio of bolometric energy to thermal energy for this event somewhere between 1.5 and 6, depending on which end of the bolometric energy range is included and whether only the peak or the peak + tail is considered. Aschwanden et al. 2017 find that the bolometric energy is approximately equal to the thermal energy and Emslie et al. 2012 report bolometric energies that are between 1.6–11 times greater than the thermal energy, with the mean falling somwehere around 5. Our ratio of bolometric to thermal energy is thus more consistent with ratios reported by Emslie et al. 2012 . Warmuth & Mann 2020 argued that DEM-based methods tend to overestimate the amount of high temperature plasma, and thus yield unrealistic levels of X-ray emission for many events. This overestimation may explain why Aschwanden et al. 2017 concluded bolometric and thermal energies should be approximately equal, while Emslie et al. 2012 concluded that the energy radiated by the SXR plasma is often only 20% of the bolometric energy or less. Here we have estimated the thermal energy of the Oct. 28 eruption in two different, independent ways, including one that directly leverages the SXR emission. That we arrive at consistent results suggests our thermal energy estimates may be sufficiently reliable to conclude that – for at least some events – bolometric energy is likely to exceed thermal energy by a factor of at least a few. From the in situ analysis, we find that the leading energy fluxes are the gravitational and kinetic terms at both SolO and Wind . The magnetic energy flux and the Poynting flux are of the same order of magnitude at both heliocentric distances. The enthalpy flux is an order of magnitude larger than both the magnetic and Poynting flux, or the electromagnetic wave energy, in both cases. Similar to the solar wind at large heliocentric distances, under energy conservation, the thermal, magnetic, Poynting flux energies have translated the majority of their energy to accelerating the CME plasma against gravity via thermal pressure gradients and magnetic forces. Despite the CME carrying a large amount of kinetic energy, the CME shock does not produce a large number of accelerated particles. The 2021 October 30 shock at SolO exhibits a conversion of upstream kinetic energy into mostly thermal energy, with the fraction of energetic particle components much higher than later observed at Wind , as shown in Figure 30 . The partial conversion of energy flux into energization of protons and electrons is surprisingly small compared to other events. Previous cases of IPs inefficiently accelerating protons have been reported at 1 au ( Lario et al. 2003 ) . Although a specific geometric configuration could explain the lack of > 100 keV protons as a local effect at a single spacecraft ( Dimmock et al. 2023 ; Trotta et al. 2023 ) , the less efficient energization at Wind (corresponding to a smaller M_{fms} , as shown in Table 2 ), reported here for this event, requires that a shock region with comparable plasma properties is observed at 1 au by Wind and the efficiently accelerating shock regions do not cross either spacecraft 10 10 10 STEREO-A measured a <200 keV ion enhancement in the Solar Electron Proton Telescope ( Müller-Mellin et al. 2008 , SEPT,) , sunward (anti-sunward) looking direction started at \sim 00:00 UT ( \sim 06:00 UT) of 30 October 2021 (not shown here), unlikely related to the eruption discussed herein. . Full 3D simulations of the expansion of the CME-driven shock will be necessary to fully interpret the joint SolO - Wind measurements. The sum total of the evidence presented here indicates that the majority of the energy released in the 28 October 2021 eruption is partitioned into the kinetic and gravitational potential energy of the CME. This energy division continues to hold even as the CME evolves in the heliosphere, where in situ measurements (Figure 30 ) show that the kinetic energy still represents a majority of the total energy flux in the shock, and is not redirected into energetic particles to a significant extent. The overall partition in this event is similar to the distribution reported by Emslie et al. 2012 , where the CME kinetic energy represents about 3/4 of the total event energy. This project was made possible by the wide variety of measurements available in the Heliophysics System Observatory (HSO) suite of missions and instruments, and the NASA HSO Connect Program that brought together experts from many different fields to do the analysis. Projects of this scope can only be realized by dedicated investments across multi-disciplinary research fields. The authors would like to thank the anonymous referee for insightful comments that improved this paper. This work was supported by the NASA Heliophysics System Connect Program, grant number 80NSSC20K1283. PH acknowledges additional support from the Office of Naval Research. DT acknowledges funding from ISRO/RESPOND for the project “Solar Flares: Physics and Forecasting for Better Understanding of Space Weather” ISRO/RES/2/438/22-23. DP acknowledges the support by the National Natural Science Foundation of China (Grant Nos. 42188101, 42521007, 42474221). SY acknowledges additional support from NASA grant 80NSSC24K1242 to NJIT. We acknowledge the use of AIA and HMI data. AIA and HMI are instruments onboard SDO, a mission for NASA’s Living With a Star (LWS) program. Hinode is a Japanese mission developed and launched by ISAS/JAXA, with NAOJ as domestic partner and NASA and STFC (UK) as international partners. It is operated by these agencies in co-operation with ESA and NSC (Norway). The STEREO/SECCHI data are produced by an international consortium of the NRL, LMSAL, NASA GSFC (USA), RAL and the University of Birmingham (UK), MPS (Germany), CSL (Belgium), and IOTA and IAS (France). The SOHO/LASCO data used here are produced by a consortium of the Naval Research Laboratory (USA), Max-Planck-Institut fuer Aeronomie (Germany), Laboratoire d’Astronomie (France), and the University of Birmingham (UK). SOHO is a project of international cooperation between ESA and NASA. Solar Orbiter is a mission of international cooperation between ESA and NASA, operated by ESA. The Solar Orbiter magnetometer was funded by the UK Space Agency (grant ST/T001062/1). The Solar Orbiter STIX instrument is an international collaboration between Switzerland, Poland, France, Czech Republic, Germany, Austria, Ireland, and Italy. This paper uses Solar Orbiter EPD data, generated and maintained by the EPD team. The Solar Orbiter Solar Wind Analyser (SWA) scientific sensors, SWA-EAS, SWA-PAS, SWA-HIS, and the SWA-DPU have been designed and created, and are operated under funding provided in numerous contracts from the UK Space Agency (UKSA), the UK Science and Technology Facilities Council (STFC), the Agenzia Spaziale Italiana (ASI), the Centre National d’Etudes Spatiales (CNES, France), the Centre National de la Recherche Scientifique (CNRS, France), the Czech contribution to the ESA PRODEX programme, and NASA. The authors thank the National Space Science Data Center of the Goddard Space Flight Center for the use permission of Wind data and the NASA CDAWeb team for making these data available at http://cdaweb.gsfc.nasa.gov/istp_public/. Appendix A Treatment of the Flare as a Single Loop Interpreting the phase portrait of GOES-derived T vs. EM , shown in Figure 21 b, as a single impulsively-heated loop allows us to derive properties of the flare without recourse to imaging. Impulsive energy input drives the temperature to a peak ( T=15.0 MK marked by a diamond) too rapidly for density – i.e. EM – to change. After impulsive input, the radiative time, \tau_{\rm r} , remains longer than the conductive cooling time \tau_{\rm c} , so thermal conduction drives evaporation which returns the energy to the loop, along with additional mass. As a consequence the energy and volume of the loop remain roughly fixed, as does its pressure ( Antiochos & Sturrock 1978 ; Cargill et al. 1995 ) , p~\simeq~2\,k_{\rm b}\,T\,n_{e}~\simeq~2\,k_{\rm b}\,T\,{\sqrt{EM\over V}}~~, (A1) where V is the flare volume. This evaporative phase, characterized by T\sim EM^{-1/2} (black dashed line in Figure 21 b), persists until the density reaches a level sufficient to shorten the radiative cooling time to match the conductive time. It appears that some additional energy release occurs during this phase, and the actual evolution arcs away from the constant- p curve over the interval from point A to B . Evaporation ends at approximately 15:41 UT, with T_{B}=10.9 MK and EM_{B}=3.7\times 10^{49}\,{\rm cm}^{-3} , shown by the triangles in Figure 21 . After this the loop remains in mechanical equilibrium, \partial p/\partial\ell\simeq 0 , cooling radiatively to reduce p . We assume classical thermal conduction, F_{c}=\kappa_{0}T^{5/2}(\partial T/\partial\ell) , where \kappa_{0}\simeq 10^{-6} is the Spitzer-Härm coefficient. Balancing this with radiative losses while maintaining mechanical equilibrium yields a loop described by the scalings of Rosner, Tucker and Vaianna ( Rosner et al. 1978 , RTV,) p~\sim~{T_{\rm pk}^{3}\over L}~~, (A2) where T_{\rm pk} is loop’s the apex temperature, and L is its full length. Mechanical equilibrium is maintained during the cooling, which proceeds according to eq. ( 5 ). Fitting the time evolution of T , shown in Fig. 21 a, provides an estimate of the effective loop length of the flare. An RTV loop is characterized by ( Serio et al. 1991 ) \tau_{c}~=~\tau_{r}~\simeq~2\times 10^{-3}\,L\,T^{-1/2}~~, (A3) when all quantities are expressed in CGS units. Setting this to \tau_{c}=92 min at T=T_{B}=10.9 MK, yields a full loop length L=102 Mm. The fit to the evaporative leg of the phase portrait can be used to find the flare volume without recourse to imaging. To remain in mechanical equilibrium, the cooling loop must drain thereby decreasing EM along with T EM~\simeq~C_{\rm rtv}\,{V\over L^{2}}\,\left({T\over 10^{7}\,{\rm K}}\right)^{\nu}~~. (A4) applying the RTV scaling of eq. ( A2 ) yields \nu=4 with C_{\rm rtv}=0.86\times 10^{41}\,{\rm cm}^{-4} . This relation, adjusted to pass through point B , is shown as the dashed magenta curves in both Fig. 21 a and 21 b. Both fit extremely well until about 16:00, when some change to the equilibrium loop seems to occur. Applying Equation ( A4 ) to the values T_{B}=10.9 MK and EM_{B}=3.7\times 10^{49}\,{\rm cm}^{-3} yields V/L^{2}\simeq 3.0 Mm. Using the length, L=102 Mm found above from time evolution, yields a flare volume V=3.1\times 10^{28}\,{\rm cm}^{3} . The net thermal energy is given in eq. ( 7 ) and plotted as a red curve on Figure 22 a. The conductive and radiative cooling times are ( Aschwanden & Alexander 2001 ) \displaystyle\tau_{c} \displaystyle= \displaystyle 2.7\times 10^{-10}\,T^{-5/2}\,L^{2}\,\sqrt{EM\over V} (A5) \displaystyle\tau_{r} \displaystyle= \displaystyle 2.3\times 10^{3}\,T^{3/2}\,\sqrt{V\over EM}~~, (A6) respectively, when all quantities are expressed in CGS units. The power lost through either process is computed by dividing U_{\rm th} by the corresponding time ( Longcope et al. 2010 ) , as shown in Figure 22 b. In Figure 22 b these curves fall on top of one another because we match an RTV equilibrium when we set V/L^{2}=3 Mm. Appendix B Analysis of EUV Wave Kinematics The EUV wave associated with the 28 October 2021 eruption appears as an expanding circle both from the GOES and STEREO-A vantage points, which are separated by roughly 37.5\degree longitude at the time of the event (see Figure 31 , top panels). The fact that the wave appears circular independent of viewpoint for much of its evolution strongly suggests it is essentially an expanding spherical shell, at least until parts of it begin to interact significantly with the ambient corona and magnetic field. Figure 31: Circular fits to the wave in SUVI (top left) and EUVI-A (top right) observations; color indicates time, with blue earliest and red latest. Bottom panels show the corresponding reconstructed position in latitude and longitude (bottom left; color indicates time as above) and radius and center height above the photosphere (bottom right). We used SUVI 195 Å images to estimate the evolution of the wave’s radial propagation velocity away from its origin at the flare site in 20^{\circ} bins spanning all azimuthal angles around the center of the flare itself. Figure 32 shows a summary of the wave’s propagation both on the disk (red + symbols) and above the limb (blue \square ). The wave first becomes clearly visible at a radial distance of about 1.5\times 10^{5} km from the flare site, and it rapidly expands and accelerates until it reaches a speed of 910\pm 200\;\mathrm{km\,s^{-1}} at a distance of 2-2.5\times 10^{5} km from the flare site. At this point, the part of the wave that travels off-the disk continues to propagate outwards at a steady velocity, while the part of the wave observed on-disk begins to interact with the ambient corona and slows and dissipates. Figure 32: Wave radial velocity as a function of distance from the center of the flare site as observed in SUVI’s 195 Å passband. Red + symbols indicate points at which the wave is observed on-disk, while blue \square symbols indicate points where the wave is observed above the limb. The dashed line indicates the wave’s final average velocity of \sim 910\;\mathrm{km\,s^{-1}} , while the shaded band indicates the standard deviation of speeds, which is about \pm 200\;\mathrm{km\,s^{-1}} . To understand how the wave propagates in 3D space, we fit circles to corresponding images of the wave’s evolution in time from both STEREO-A and GOES perspectives and used epipolar geometric methods ( Inhester 2006 ) to reconstruct the size and location of a spherical shell resulting from these fits. Figure 31 shows these fits and the reconstructed positions of wave in 3D. Our reconstruction indicates that the wave is simultaneously rising and traveling towards the south as it expands outwards. Fitting a line to the successive center point positions of the sphere, we find that the center is rising from the solar surface at about 530\pm 40 km s -1 , while the sphere’s early expansion is consistent with the velocities shown in Figure 32 . This result is not entirely straightforward to understand, but might be explained if the wave were expanding within an outflowing region, perhaps as the result of a large-scale flow related to the outgoing CME itself. The wave’s early evolution appears to be insensitive to any variation in the magnetic field of the surrounding corona. Such behavior appears to support the interpretation that EUV waves are initiated as blast waves due to the rapid expansion of the erupting structures. This early behavior, plus the wave’s later interactions with the low corona, appear to be consistent with the conclusion of Long et al. 2015 and Long et al. 2017 , that EUV Waves likely evolve from blast waves to fast-mode large-amplitude waves or shocks over their lifetimes.