The EUV Late-Phase: Statistical Results from 15 Years of Solar Dynamics Observatory Observations

TL;DR

This study delivers the largest spatially resolved statistical characterization of the EUV late-phase (ELP) in solar flares using 15 years of SDO/AIA Fe XVI (335 Å) observations. It identifies 467 ELP events from 5335 isolated flares (≈8% occurrence) and shows no strong solar-cycle dependence, with typical onset, peak-to-peak, and duration metrics of ~19 min, ~88 min, and ~93 min, respectively. The analysis reveals a tight coupling between ELP rise and decay rates, a sub-linear relationship between flare and ELP impulsivity, and three principal modes of variability via PCA, suggesting that ELP evolution reflects both flare-loop properties and reconnection-driven energetics within a finite magnetic energy budget. The results provide insight into the Sun–Earth connection by linking late-phase EUV emission to ionospheric response and highlight the value of long-term, spatially resolved solar observations for understanding flare energy transport and space weather forecasting.

Abstract

Since its launch in 2010, the Solar Dynamics Observatory (SDO) has provided continuous high-cadence, multi-wavelength observations of the Sun, capturing thousands of solar flares and offering new insights into coronal dynamics. Among the discoveries enabled by SDO is the EUV late-phase (ELP), characterised by a secondary enhancement in warm coronal emission occurring tens of minutes after the main flare. While recent work has demonstrated the relevance of the ELP for space weather, its statistical behaviour and physical origin remain poorly constrained. Here, we present the most comprehensive review of the ELP to date, based on 15 years of Fe XVI observations from the Atmospheric Imaging Assembly onboard SDO (SDO/AIA). From a sample of 5335 isolated flares between 2010 and 2025, we identify and validate 467 ELP events. The overall ELP occurrence rate is 8 percent, with no significant dependence on the solar cycle and only a modest enhancement in the low-to-mid M-class range. The ELP typically exhibits and onset delay of 19 minutes, a peak-to-peak delay of 88 minutes, and a duration of 93 minutes. Strong correlations are found between ELP rise and decay rates (p=0.76), and between flare and ELP impulsivity (p=0.61), while no significant correlation is observed between flare and ELP phases. A Principal Component Analysis revealed three dominant axes of variation, corresponding to a timescale component, an energy-release intensity axis, and a partitioning of energy between the flare and ELP. These results suggest that ELP evolution is governed by both flare loop properties and reconnection-driven energetics, likely modulated by a finite magnetic energy budget, and highlight the importance of SDO's long-term observations for understanding flare evolution and the Sun-Earth connection.

Figures (17)

• Figure 1: Example of a combined EUV and SXR timeseries for a C1.1 flare corresponding to a positive ELP identification. Each EUV line was derived from integrated SDO/AIA cutout data. From the Fe xvi line (blue), two clear peaks are resolvable corresponding to the flare and ELP, respectively. The grey shaded represents the derived ELP period based on the characteristics outlined in Section \ref{['subsec:elp_detection']}
• Figure 2: Example of the spatial distinction between the main flare and the ELP emission in SDO/AIA Fe xvi (335 Å) images. The left and middle panels show the 500 $\times$ 500 arcsec AIA cutouts at the flare peak and the ELP peak, respectively. The right panel shows the difference image (ELP peak minus flare peak), where the dark region highlights the location of the enhanced late-phase emission, distinct from the brighter main phase region (blue).
• Figure 3: Percentage of events with an associated ELP by GOES class bin. Error bars denote 95 % Wilson binomial confidence intervals. The annotated N-values correspond to the total number of flares from the full sample of 5335 flares.
• Figure 4: Temporal variation of ELP occurrence from 2010 to 2025. The black line indicates the six-month smoothed monthly mean sunspot number used as a proxy for solar activity level and therefore solar cycle. The red line denotes the total number of flares per month, and the blue line shows the number of validated ELP detections in that month. The percentage of ELPs detected in each month (normalised to the total number of flares) is shown in green.
• Figure 5: Comparison between SDO/AIA and SDO/EVE observations of the Fe xvi emission for a representative ELP event. Top Left: Full-disk AIA 335 Å image at the ELP peak time. The red box denotes the 500 $\times$ 500 arcsec cutout region corresponding to the ELP/flare site used as a proxy for the effective field of view of the disk-integrated EVE measurements. Top right: AIA 335 Å cutout image of the region enclosed by the red box in the full-disk image. Bottom: Normalised Fe xvi lightcurves from AIA (blue-red line) and EVE (blue dashed line) for the same event. The ELP signal is clearly visible in the spatially-resolved AIA data but not in the disk-integrated EVE irradiance.