1 Introduction

Solar flares are powerful bursts of energy release, often followed by coronal mass ejections (CMEs). These eruptions accelerate particles and eject plasma from the solar atmosphere (Benz, 2008; Gopalswamy, 2016). However, the relationship between flares and CMEs is not always clear (Kawabata et al., 2018). The most energetic flares are known to be associated with fast and wide CMEs, but the relation between CMEs and small and less energetic flares, or with flares missing altogether, is more complicated. It has been suggested that the magnetic reconnection process that happens beneath the CME affects the CME dynamics most (Vrs̆nak, Sudar, and Ruz̆djak, 2005).

As flares and CMEs accelerate particles, some of their processes can be observed in radio emission; see reviews by Nindos et al. (2008) and Pick and Vilmer (2008). Particle streams, i.e. accelerated electrons, cause radio emission at the local plasma frequency as they propagate through the solar atmosphere. The radio emission is then also subject to scattering by density inhomogeneities and other propagation effects (Kontar et al., 2019). The most often observed solar radio emission types are type III bursts (caused by accelerated electron beams), type II bursts (caused by shock-accelerated electrons, typically ahead of propagating CMEs), and type I noise storms (electrons trapped in the magnetic field).

Type III bursts are most often associated with flares, and they can appear as isolated bursts or in groups. There is now direct evidence that semi-relativistic electrons energized in magnetic reconnection regions produce type III bursts (Cairns et al., 2018). At starting frequencies higher than 100 MHz, type III bursts have been found to be associated mostly with GOES B- and C-class flares, but the X-ray emission at 6 keV usually lasts much longer than the groups of type III bursts (Reid and Vilmer, 2017). Type III bursts can be observed as fast-drifting emission features in the radio dynamic spectra, some only at decimetric–metric wavelengths (coronal heights) and some continuing to kilometer waves (near Earth distances).

Shocks can form in the solar corona when an ejected projectile propagates at speed higher than the local magnetosonic speed. Also, flare blasts can create propagating shocks that accelerate electrons (Warmuth, 2007). The frequency drift in a radio type II burst is much slower than in type III bursts, as the drift compares to the propagation speed of the shock. Observationally, type II bursts can often be identified from their harmonic emission (Roberts, 1959).

Type I noise storms are associated with active regions and sometimes with flares and CMEs, but the typical duration of a noise storm continuum is significantly longer than that of the associated flare (Iwai et al., 2012). The broad-band continuum also contains short-duration, narrow-band bursts. The generally accepted scenario is that the emission is due to non-thermal electrons trapped in closed magnetic fields. The origin of these fast electrons is not very clear, but small-scale reconnection and weak shocks associated with newly emerging flux have been suggested; see, for example, Mercier et al. (2015) and references therein. Mondal and Oberoi (2021) recently suggested that small-scale reconnection may produce electron beams, which quickly get collisionally damped, and therefore the plasma emission occurs only within a narrow bandwidth.

In this study, we analyse a solar event that consisted of a small, GOES B-class flare and a compact CME that had a very narrow width and relatively high initial speed and was associated with groups of metric type III bursts and type I noise storm bursts. Our aim is to find out the reason for the continuous particle acceleration and the generation of repeated type III bursts. We also investigate the CME formation and its propagation as a compact, bullet-like form.

2 Data and Analysis

We have used in the analysis extreme ultraviolet (EUV) solar images from the Atmospheric Imaging Assembly (AIA) instrument onboard the Solar Dynamics Observatory (SDO: Lemen et al., 2012) and from the Extreme Ultraviolet Imager (EUVI) onboard the Solar Terrestrial Relation Observatory (STEREO) spacecraft (Wuelser et al., 2004). Coronagraph images and associated data products were obtained from the CDAW LASCO CME Catalog at cdaw.gsfc.nasa.gov. The Gamma-ray Burst Monitor (GBM) onboard the Fermi Gamma-ray Space Telescope (Meegan et al., 2009) provided X-ray flux observations for the event.

For decimetric–metric radio emission, we used radio spectral data from the Kilpisjärvi Atmospheric Imaging Receiver Array (KAIRA: McKay-Bukowski et al., 2015) located in Finland, from the various CALLISTO instruments in the e-Callisto Network (Benz, Monstein, and Meyer, 2005; Monstein, Csillaghy, and Benz, 2023), and from the Nançay Decameter Array (NDA: Lecacheux, 2000) and ORFEES radio-spectrograph (Hamini et al., 2021) located in France and provided by the Radio Monitoring website at secchirh.obspm.fr. For longer radio wavelengths, we used data from the WAVES instrument on the Wind spacecraft (Bougeret et al., 1995). Radio imaging at selected frequencies in the 150 – 445 MHz range were provided by the Nançay Radioheliograph (NRH: Kerdraon and Delouis, 1997).

2.1 Active Region and Flare

A small, GOES B4.5-class flare was observed on 27 September 2021 (SOL2021-09-27T11:46) in NOAA active region (AR) 2871, located at S29W35 (Figure 1). The GOES flare was listed as starting at 11:40 UT, peaking at 11:46 UT, and ending at 11:50 UT. However, a post-burst increase was observed to last until 12:50 UT in the 1.5 – 12.5 keV energy range (1 – 8 Å). The potential magnetic field is presented in Figure 2, created using both SDO/AIA (Earth view) and STEREO-A/EUVI (separation angle between them 39.8 degrees) observations and the potential-field source-surface (PFSS) model.

Figure 1
figure 1

On 27 September 2021 a GOES B4.5-class flare was observed to start at 11:40 UT, with maximum flux at 11:46 UT, at location S29W35 in active region AR 12871.

Full size image
Figure 2
figure 2

Top: Spacecraft locations on 27 September 2021. The STEREO-A spacecraft longitude separation to Earth was 39.8 degrees, with SOHO, SDO, Wind, and Fermi spacecraft at Earth’s longitude. The reference longitude is for the B4.5-class flare located at S29W35. (Plot prepared with Solar-MACH, The Solar MAgnetic Connection HAUS tool.) Bottom left: STEREO-A/EUVI image rotated to SDO/AIA field of view, with PFSS magnetic field lines. Blue and red lines indicate open field with opposite polarities and white lines indicate closed field. Bottom right: SDO/HMI magnetogram of the eruption region (from SolarMonitor.org).

Full size image

Fermi/GBM also observed the flare with its NaI detectors that cover the energy range from a few keV to about 1 MeV. In the 4 – 15 keV energy range the maximum count rates were observed during 11:43:50 – 11:44:40 UT, but count-rate enhancements were also visible at 11:43 UT, 11:44 UT, 11:46 UT, and 11:50 UT at energies 4 – 50 keV (Figure 5). Flare emission was observed until 11:51 UT, after which the spacecraft turned away from the Sun (Fermi is in low-Earth orbit). The comparison between X-rays and radio emission is discussed more in Section 2.2.

A plasmoid ejection was observed to start from the flare site at 11:42 UT. A narrow filamentary structure was ejected from the AR, and it was observed to change direction and turn to follow the fan-like magnetic field. The bending of this long filamentary structure along the magnetic field lines is shown in the EUV images from SDO and STEREO-A; see Figure 3. The bending was best observed in the STEREO-A view, over the limb and against the sky, which enabled us to better estimate and compare the source heights.

Figure 3
figure 3

SDO/AIA and STEREO-A/EUVI images of the AR and its surroundings at 11:49 UT. Arrow points to the ejected filament/plasmoid structure. Note the bending of the structure along the magnetic field lines.

Full size image

2.2 Radio Emission

The main radio features associated with the flare–CME event were type III bursts, i.e. fast electron beams propagating at speeds ≈ 0.3 c. Figure 4 shows how the type III bursts occurred in groups. We could identify five separate groups, labelled 1 – 5 in Figure 4. The approximate burst group durations were 11:42:20 – 11:44:50 UT (150 seconds), 11:45:00 – 11:47:35 UT (155 seconds), 11:49:30 – 11:51:20 UT (110 seconds), 11:53:50 – 11:55:20 UT (90 seconds), and 11:57:30 – 11:58:55 UT (85 seconds).

Figure 4
figure 4

Radio Monitoring composite showing observed radio emission in the dynamic spectra for the 27 September 2021 event. Numbers 1 – 5 refer to the type III burst groups. The locations of bursts labelled 1a and 4a are shown in Figures 7 and 6, respectively. Comparison between NDA and Wind/WAVES spectra is also shown.

Full size image

The type III burst groups appear to have some temporal association with the X-ray flux measured by Fermi/GBM (Figure 5). Intense type III burst emission is observed at the time of the first X-ray peak near 11:43 UT, and the end time of the first type III burst group is also the time when the X-ray burst intensity decreases. There is a small increase in X-rays during the type III burst group number 3. However, we find no clear one-to-one correspondence between the X-ray count rate peaks and individual type III bursts.

Figure 5
figure 5

Fermi/GBM background-subtracted X-ray counts from the most sunward-pointing detector at 11:42 – 11:52 UT (bottom). The energy ranges are 4 – 15 keV, 15 – 25 keV, and 25 – 50 keV. Radio dynamic spectrum from e-CALLISTO Glasgow is shown on top, at frequency range 45 – 81 MHz. Numbers and lines indicate the type III burst groups and their durations.

Full size image

Bursts within the first two type III burst groups reached interplanetary space and propagated to Earth distances (emission continues down to kHz frequencies), but burst emission in the later three groups fades away at 5 – 2 MHz; see the Wind/WAVES dynamic spectrum in Figure 4.

Some isolated type III bursts were observed at 500 – 200 MHz near the flare start time, but most of the type III burst groups had a start frequency near 200 MHz (Figure 4). The start frequency, 200 MHz, corresponds to an atmospheric height of 0.15 R⊙ when calculated with the 2-fold Newkirk atmospheric density model (Newkirk, 1961), which describes well the lower altitudes in the solar corona. Another often used atmospheric model, the so-called hybrid model by Vrs̆nak, Magdalenic̆, and Zlobec (2004), gives heights close to the 2-fold Newkirk model in the low corona but works well also at larger distances. The differences between atmospheric density models and the calculation of radio source heights is explained in, e.g., Pohjolainen et al. (2007).

Individual type III bursts could be best identified in the 80 – 20 MHz frequency range, and many of them could be imaged near their start frequency, at 150 MHz with NRH. These bursts were all located south-west of the AR, and some of them reached the solar limb; see an example of radio imaging in Figure 6, the type III burst is marked 4a in the dynamic spectrum in Figure 4. The calculated source heights at 150 MHz are in agreement with the imaged radio source distances from the AR. As the imaged radio sources are observed on the solar disc in projection from the Earth view, their heights are only estimates and can also suffer from various plasma effects (Chen et al., 2020).

Figure 6
figure 6

Temporal evolution of burst locations at 150 MHz, observed with NRH. The sources are shown near burst start (black contours) and burst end (white contours), overplotted on an EUV image (left). The spatial evolution is also shown in the color images, from 11:54:11 to 11:54:13 UT (right). This spectral feature is marked 4a (within group 4 bursts) in Figure 4. The radio source locations are typical for all type III bursts in the type III burst groups observed at 150 MHz.

Full size image

2.3 Specific Radio Spectral Features

Near the flare start, at 11:43 UT, a high-frequency type III burst was observed, and it could be imaged at 228 and 270 MHz. Two separate radio source locations appeared then, one south-west and one north of the AR (Figure 7). This type III burst is marked 1a in the dynamic spectrum in Figure 4. The flux measured from the northern source region showed an enhancement at 270 MHz, and it peaked later than the 228 MHz flux. This suggests that the type III burst was reversed, i.e. the electron beam was moving down in the solar atmosphere. We note that the northern source did not appear again at later times.

Figure 7
figure 7

NRH source contours at 270 MHz (black) and 228 MHz (white) at 11:43:08 UT, plotted over an EUV image (left). The spectral feature is marked 1a within group 1 bursts) in Figure 4. The flux curves (on the right) of the northern region show an enhancement at 270 MHz, which peaks later than the emission at 228 MHz. This suggests that the electron beam was moving down in the solar atmosphere. The later type III burst groups did not show this northern region again.

Full size image

Some of the type III bursts showed enhanced plasma emission features along the electron-beam propagation path; the two features are indicated with arrows in Figure 8. The features appeared at ≈ 60 MHz around 11:46 UT. They were observed by e-Callisto stations and KAIRA, with different temporal resolutions of 0.25 and 0.01 seconds, respectively. The enhancements had a narrow bandwidth, only a few MHz, and on closer look they seem to be separated from individual type III emission lanes and have a curved appearance.

Figure 8
figure 8

Dynamic spectrum taken during the second group of type III bursts show narrow-band enhanced plasma emission features, which are indicated with arrows in the KAIRA observations at 75 – 25 MHz (top). Narrow-band spikes (type I noise storm features) were observed in between the type III burst groups (KAIRA observations, middle). e-CALLISTO GLASGOW dynamic spectrum at 80 – 60 MHz shows the spikes and spike burst groups in more detail (bottom).

Full size image

In between the type III bursts, we could also observe narrow-band radio spikes, identified as type I noise-storm structures (Figure 8). In high-resolution dynamic spectra, type I bursts appear as spot-like features, whereas vertical stripe-like features are type III bursts; see, for example, Mugundhan et al. (2018). The bandwidth of individual spikes was only 1 – 3 MHz. The spikes appeared in groups within a wider frequency range, between 80 and 50 MHz. These groups also drifted in frequency, typically towards higher frequencies.

Noise storm spikes are visible in the spectra also before our flare, appearing and growing stronger in intensity around 9:00 UT and lasting up to around 15:00 UT in the 80 – 50 MHz range. A GOES B3-class enhancement was observed at 09:00 – 09:45 UT, a possible sign of small-scale reconnection event(s). As type I noise storms are generally thought to be excited by plasma waves caused by non-thermal electrons trapped in closed magnetic-field lines, this suggests that there were a large number of non-thermal electrons present in the AR loops even before the start of our flare–CME event.

Near the start time of the first type III burst group, at 11:43 UT, some of the type III burst lanes seemed to split into two separate lanes. Figure 9 shows two of these V-shaped bursts in detail, observed by KAIRA (top spectrum, the slower-drift and turning lanes are indicated with arrows). The separation of electron-beam paths happens near 40 MHz, and the later, slower beam lanes disappear near 30 MHz. These frequencies correspond to solar atmospheric altitudes of 0.7 – 1.0 R⊙, which are reasonable for large-scale loop heights and agree with the loops obtained with the PFSS model of the magnetic field, shown in Figure 2. Beam path curvature back towards higher frequencies, i.e. to higher densities, indicates that the electron beam follows magnetic field lines back to the Sun.

Figure 9
figure 9

KAIRA observations of V-shaped type III bursts at 11:43 UT (top). Arrows point to the change in emission lane direction, visible at 35 – 27 MHz. The bursts were also observed by several e-Callisto stations; the dynamic spectrum shown here is from the SWISS-Landschlacht e-CALLISTO at 80 – 20 MHz (bottom left). Cartoon shows a possible configuration for the creation of V-shaped type III bursts. Star indicates reconnection, and arrows show the directions of two particle-beam paths, one along open field lines (emission with constantly decreasing frequencies) and the other following closed loops back to the Sun (emission lane turning back towards lower frequencies).

Full size image

2.4 Coronal Mass Ejection

After the plasmoid ejection, a narrow dimming region started to appear in the EUV images, elongated towards the limb in the south-west direction. The SDO/AIA difference images are shown in Figure 10. A CME was first observed by the SOHO/LASCO-C2 coronagraph at 12:12:05 UT, when the CME leading front was located at a height of 2.57 R⊙. The CME had a narrow angular width, 25 – 20 degrees (Figure 11), and it was propagating towards the south-west. There were no streamers located nearby the CME.

Figure 10
figure 10

SDO/AIA 193 Å  base-difference images (base at 11:32 UT). The narrow-width dimming region, located from the AR towards the south-west limb, matches with the type III burst regions imaged at 150 MHz, and also with the CME propagation direction.

Full size image
Figure 11
figure 11

SDO/AIA 193 Å  and SOHO/LASCO-C2 difference image at 12:36 UT (left), and direct AIA and C2 images at 12:48 UT (middle) and 13:25 UT (right). The CME was very compact and had a narrow width. No streamers were visible along the CME propagation path.

Full size image

The CME speed from the first LASCO observations was ≈ 500 km s−1, but the velocity decreased after the first hour and then stabilized to about 380 km s−1, based on the linear fit for the CME front heights available from the CME Catalog. The observed solar wind speed near Earth was 300 – 330 km s−1, which explains some of the decrease in the CME speed.

We converted the CME heights to frequencies with the hybrid atmospheric-density model (Vrs̆nak, Magdalenic̆, and Zlobec, 2004), which works well for the larger distances in the corona. These heights and frequencies are listed in Table 1, and they are shown in Figure 12.

Figure 12
figure 12

Top: CME leading front heights (boxes), the heights are from the LASCO CME Catalog, flare start time (cross), and metric type III burst group durations (filled boxes). Bottom: Wind/WAVES dynamic spectrum where white circles mark the CME leading front heights converted to frequencies with the hybrid atmospheric-density model.

Full size image
Table 1 Heliocentric CME heights from LASCO-C2 observations converted to frequencies with the hybrid atmospheric-density model.
Full size table

3 Summary and Conclusions

The observed long-duration, metric type III burst emission that appeared as type III burst groups indicates continuous electron acceleration. If no high-frequency emission is observed at the same time, then it suggests that the acceleration region is located high in the corona (the higher the radio frequency, the lower the atmospheric height). The type III bursts that appeared in groups were formed at 200 MHz or nearby frequencies (corresponding to ≈ 0.15 R⊙ or 100 Mm, heights above the solar surface as plasma frequency depends on electron density in the atmosphere), and each group had a duration of 2.5 – 1.5 minutes. These bursts did not look to have a direct, one-to-one correlation with hard X-ray count peaks (Fermi/GBM), but enhanced X-ray emission was observed during the type III burst group periods. As X-rays are the result of electrons colliding with denser structures, this can mean that part of the X-ray emission came from heating and part from particle acceleration.

The unchanged start frequency (formation height) of each radio burst group suggests that the acceleration process of the burst particles was not due to reconnection in rising structures, but that the reconnecting fields remained at the same heights. A decrease in start frequency would indicate that the reconnection region is moving up in the atmosphere, where the density is lower. For example, Reiner et al. (2008) have suggested that in complex type III bursts the later electron acceleration can be the result of coronal reconfiguration, caused by, for example, an erupting CME.

In Figure 9, we presented a possible configuration for the creation of type III bursts that show a split into two separate lanes at metric wavelengths. The observed split happened near 40 – 50 MHz, which corresponds to heights of ≈ 0.6 – 0.8 R⊙ above the solar surface. The later, curved beam path can be explained by the large-scale loop shape and its slower rate of change in height, which becomes almost vertical later on. The “split” does not necessarily mean a physical split of the electron beam, but that part of the accelerated electrons gets access to field lines that curve to different directions, compared to those that propagate directly out along the open field lines. A similar split into two, “fast” and “slow” drifting burst parts, was reported also by Kallunki, McKay, and Tornikoski (2021). The burst they described was most probably associated with a CME that originated from the far side and had a narrow width in Earth view, and the true GOES-class was unknown and listed as a B-class flare.

Type III bursts with inverted-U and N-shaped emission lanes, observed at coronal heights, have been discussed by Démoulin et al. (2006). These burst shapes require large-scale loops and closed magnetic field lines, as in them the electron beams first travel upward along the field lines, then come back to lower heights, and in N-bursts the particles are mirrored back to larger heights. In our split V-shaped bursts, we did not find any evidence of mirroring, as the later branch simply faded away and ceased to be observed.

This high-altitude scenario resembles those presented by Heyvaerts, Priest, and Rust (1977) and Sterling and Moore (2001). In their models, there is a rising flux rope on the side of a coronal-hole open field, which reconnects (“external reconnection”). In our case, we have a large-scale loop on the side of open active-region field lines, with a rising but very narrow filament/plasmoid in between them. Due to the changes in the outflowing material, reconnection and particle acceleration are quasi-periodical. In this sense, the scenario is similar to loop–loop interactions, where the loop movements cause reconnection leading to oscillation in the radio emission. The many different possibilities for modulated reconnection have recently been described in Cattell et al. (2021).

The active region that produced the small, GOES B4.5-class flare on 27 September 2021 was surrounded by strong fields that guided the outflow of the ejected plasmoid. The CME had a very narrow width due to the small volume of plasma and the narrow tunnel-like exit path from the solar surface. This was observed in Figure 3, where the propagation direction of the long filamentary CME structure was changed, and it was bent along the fan-like field lines. The PFSS model also predicted large-scale loops that could come into contact with the ejected plasmoid. The metric type III burst groups could have been created by periodic reconnection, caused by coronal reconfiguration that affected the large-scale loops and the open field lines nearby. The type III bursts ended when the CME had lifted off and the material outflow had ceased.