Quasi-periodic Pulsations before and during a Solar Flare in AR 12242

Quasi-periodic pulsations (QPPs) are frequently observed in solar flares, which may reveal some essential characteristics of both thermal and nonthermal energy releases. This work presents multi-wavelength imaging observations of an M8.7 flare in active region AR 12242 on 2014 December 17. We found that there were three different QPPs: UV QPPs with a period of about 4 minutes at 1600 Å images near the center of the active region lasting from the preflare phase to the impulsive phase; EUV QPPs with a period of about 3 minutes along the circular ribbon during the preflare phase; and radio QPPs with a period of about 2 minutes at frequencies of 1.2–2.0 GHz around the flaring source region during the impulsive phase. The observations include the radio images observed by the Mingantu Spectral Radioheliograph in China at frequencies of 1.2–2.0 GHz for the first time, microwave images by the Nobeyama Radioheliograph, UV and EUV images by AIA/SDO, and a magnetogram by HMI/SDO. We suggest that the 4 minute UV QPPs should be modulated by the sunspot oscillations, and the 3 minute EUV QPPs are closely related to the 2 minute radio QPPs for their source regions connected by a group of coronal loops. We propose that the intermittent magnetic reconnecting downward and upward plasmoids may be the possible trigger of both the preflare 3 minute EUV QPPs and the impulsive 2 minute radio QPPs. The other possible mechanism is LRC oscillation, which is associated with the current-carrying coronal loops. The latter mechanism implies that the existence of preflare QPPs may be a possible precursor to solar flares.


Introduction
Solar radio bursts, especially in the frequency range of centimeter and decimeter waves, are strongly associated with solar eruptions, such as solar flares, coronal mass ejections (CMEs), eruptive filaments, and various jets. They are believed to be a sensitive signature of magnetic energy release, and the generation, acceleration, and propagation of nonthermal electrons (e.g., Bastian et al. 1998;Gary & Keller 2004;Pick & Vilmer 2008). Detailed observations of different types of solar radio bursts are crucial for understanding the mechanisms of particle acceleration, and the initiation and evolution processes of solar eruptions.
Recently, a new broadband solar spectral radioheliograph, Mingantu Spectral Radioheliograph (MUSER) was established in Inner Mongolia of China (Yan et al. 2009. MUSER is housed at the Mingantu Observing Station of the National Astronomical Observatories of Chinese Academy of Sciences (NAOC), about 400 km northwest of Beijing. MUSER is an aperture synthesis solar radio telescope array that can observe the full Sun in centimetric-decimetric wavelengths with high resolutions and dual polarizations. MUSER includes two subarrays, MUSER-I and MUSER-II, with frequency ranges of 0.4-2.0 GHz and 2.0-15.0 GHz, respectively. In total, the array is composed of 100 antennas distributed over 3 spiral arms, with the longest baseline being 3.0 km. This distribution configuration has many advantages after considering the distributions of UV-coverage, the imaging simulations (Du et al. 2015), and the instrument construction cost. The imaging quality of MUSER-I depends on the stabilities of the amplitude and phase of the signal from the receiving system, which includes an antenna and reflector system, an analog receiver system, a digital processing system, and a data processing system (Li et al. 2015. And it also requires highaccuracy calibrations by measurement on the following aspects: antenna position, antenna pointing, receiver gain, fixed delay system, system phase, polarization, self-calibration, frequency bandwidth effects, and so on (Liu et al. 2014). The spectral imaging observations by MUSER provide new opportunities to determine the detailed physical partition of the solar radio source regions, distinguish their emission mechanisms in different regions, estimate the physical parameters of the solar corona, diagnose the magnetic fields in the corona, and comprehensively investigate the physical processes of solar bursts (Bastian et al. 1998). In this work, the spectrum and imaging of quasi-periodic pulsations (QPPs) overlaid on type IV solar radio continuum emission were observed by MUSER-I.
QPPs are commonly observed, which show periodic pulses in almost all phases of the flare; see reviews by Nakariakov & Melnikov (2009, and Van Doorsselaere et al. (2016). QPPs are detected in all wavelengths (from radio to gamma-rays) with different timescales from sub-seconds to tens of minutes (Kliem et al. 2000;Sych et al. 2009;Liu et al. 2012;Simões et al. 2015;Hayes et al. 2016;Huang et al. 2016).
The physical mechanisms are classified by Aschwanden (1987) into three groups. First is MHD oscillations. MHD waves and oscillations can produce QPPs by changing plasma parameters related to almost all emissions from the flare, including periodic triggering of the flaring energy release, modulation of electron acceleration/injection, and so on (Nakariakov 2007;Nakariakov & Melnikov 2009). Second is cyclic self-organizing systems, which share the principle of self-organization and are governed by an oscillatory phase of wave-wave or wave-particle interactions (Aschwanden 1987;Nakariakov & Melnikov 2009;Aschwanden et al. 2018). Zaitsev (1971) proposed a model in which a stabilized stream of high-energy protons trapped in a magnetic bottle causes a quasi-periodic stream instability and periodically modulates plasma waves. Another self-organizing mechanism is the LRCcircuit model developed by Zaitsev et al. (1998Zaitsev et al. ( , 2000. It illustrates that a current-carrying magnetic loop can be twisted and form an LRC-circuit resonator, which may cause periodic modulation of loop magnetic field, energy release rate, electron acceleration, and the emission of nonthermal electrons (Khodachenko et al. 2009;Tan et al. 2016). Third is modulation of magnetic reconnection may lead to an intermittent energy release and particle acceleration (Kliem et al. 2000;Karlický et al. 2005;Ofman & Sui 2006;Murray et al. 2009;McLaughlin et al. 2012;Zhang et al. 2016). Kliem et al. (2000) simulated the repeated formation and coalescence of magnetic islands in a flare current sheet and concluded the drifting pulsating structures were generated from accelerated electrons trapped in the moving plasmoids. Wu et al. (2016) presented the first microwave observations of a hot flux rope structure at 17 GHz with a 2 minute period during the pre-impulsive stage. The same periodicity with AIA sunward-contracting loops and upward ejective plasmoids could demonstrate that they are caused by the same intermittent reconnection process at a 2 minute timescale .
The features of QPPs are complicated and they may have multiple periods in the same flare (Inglis & Nakariakov 2009;Tan et al. 2010), which may be caused by different mechanisms or a combination thereof. Although it is very hard to conclusively identify the emission mechanism of those QPPs, we attempt to give a reasonable physical mechanism responsible for the generation of QPPs.
In this work, we found three interesting components of QPPs before and during a powerful M8.7 flare in the AR 12242 on 2014 December 17. This paper is organized as follows. In Section 2, we present the results of high-resolution imaging and spectroscopy observations by MUSER-I, the study of the radio bursts related to the flare event, and corresponding overviews of the observations at radio, EUV, and X-ray wavelengths. Section 3 presents a discussion of the physical processes and Section 4 provides a summary.

Instruments and Observational Data
The M8.7 flare starts at 04:25 UT, peaks at 04:51 UT, and ends at 05:20 UT on 2014 December 17 in AR 12242. We use the following data to analyze this event: (1) Radio imaging observations in a frequency range of 1.2-2.0 GHz obtained by MUSER-I.
(2) Radio dynamic spectrum at frequencies of 0.4-2.0 GHz, with temporal resolutions of t 25  = ms and frequency resolutions of f 25  = MHz integrated from the observations of MUSER-I.
(3) The UV and EUV imaging observations at multiple wavelengths observed from the Atmospheric Imaging Assembly/Solar Dynamics Observatory (AIA/SDO; Lemen et al. 2012). (4) Magnetic field data observed by the Helioseismic Magnetic Imager/SDO (HMI/SDO; Schou et al. 2012). The pixel sizes of AIA and HMI are around 0 6, with time cadences of 12 s and 45 s, respectively. (5) The X-ray observations from RHESSI (Lin et al. 2002) at different energy bands between 12 and 50 keV. (6) The flux recorded by Nobeyama Radio Polarimeters (NORP; Nakajima et al. 1985) and microwave images observed by the Nobeyama Radioheliograph at frequencies of 17 and 34 GHz (NORH; Nakajima et al. 1994).

Introduction of MUSER-I
MUSER-I began observations in 2014, and MUSER-II became operational in the summer of 2016. In this work, we focus on a flare event that occurred on 2014 December 17, therefore we only introduce the main specifics and pipeline of data processing for MUSER-I here. For MUSER-I, the longest and shortest baselines are 3 km and 8 m, which give maximum spatial resolutions from ∼63″ to ∼12 6 at different frequencies, respectively, and exact spatial resolutions in different solar radio events depend on the element flagging during data processing. The dynamic range is designed to be 25 dB, the temporal resolution is 25 ms and the frequency resolution is 25 MHz. The measured accuracy of the dual circular polarization is about 10%.
The data processing of MUSER is based on aperture synthesis imaging technology (Taylor et al. 1999;Thompson et al. 2001). Detailed pipelines for data processing of MUSER were introduced by Wang et al. (2015). For the purposes of accumulating massive amounts of observational data, they also implemented a practical high-performance imaging pipeline for data processing based on the Graphics Processing Unit technology Mei et al. 2018).
As for the imaging observations, we obtain the crosscorrelation data from the observations first, which are in complex domains including real and imaginary parts, then correct the transmission delay caused by different lengths of the optical fiber and subtract the initial phase from the phase of fringe stopping. Calibrated sources need to be selected for the phase calibration due to the lack of redundant baselines. MUSER-I observes the Feng Yun satellite to implement the phase calibration, which is reasonably regarded as a point source. Generally, we directly implement the Fourier transform of the visibility data after calibrations, and then obtain the dirty images, which contain the true radio sources and some instrumental artifacts. We may integrate the data from several continuous frames to obtain an integral image with higher quality before generating the dirty images. It is necessary to flag some antennas or baselines with bad data. We developed an automatic program to eliminate the bad data before integration and generate the flagging tables of antennas or baselines by extracting information from the UV-distance plots and the phase curves between every two antennas. Finally, we perform deconvolution between the dirty images and dirty beams to obtain clean images. These clean images are automatically transformed into the "fits" format with detailed parameters in the header file. In addition to imaging observations, we also obtain a dynamic spectrum using the autocorrelation data from the single antennas, which provides different types of spectral structure for solar radio bursts. Figure 1 shows the solar radio image at 1.725 GHz observed by MUSER-I and the EUV image before the flare onset on 2014 December 17. It presents the radio sources of two active regions (AR 12241 and AR 12242) at 1.725 GHz. The relative positions of these two radio sources perfectly match with the EUV images of the quiet Sun. However, because of the intensive radio emission during the flare process and the limited dynamic range of the instrument, the solar disk and the weak radio sources cannot be distinguished from the radio images.
MUSER-I observes the satellite to implement the phase calibration at 1.7 GHz. As for other frequencies, so far, we have no effective calibration sources. Therefore, we perform the selfcalibration specific to the intense solar radio bursts at frequencies beyond 1.7 GHz, which shows nice imaging results for this event. Here, we assume that most of the radio emission is generated from one source on the solar disk at the peak time of the radio bursts, which is reasonable for this event with a powerful flare with one intense radio source. With this assumption, we perform phase calibration from the crosscorrelation data at the peak time and derive the radio sources with respect to the imaging at the peak time in other frequency channels.

Morphology of Different Loop Systems
The temporal evolutions of morphological characteristics in AR 12242 from the preflare phase to the flare impulsive phase are presented in Figure 2. The corresponding movies for Figure 2 are show the evolutions of those EUV structures. There is a bright, skirt-like stripe surrounding the active region from the EUV images and we refer to it as a circular ribbon.
The boundary of the active region is shaped like a circular ribbon at 131 and 94 Å and composed of bright footpoints at 171 Å. They are faint around 04:00 UT (seen in Figure 2(a)), become sharp-edged and extended during the preflare phase, and seemingly remain unchanged around the flare peak time.
Several small loop systems surrounded by a circular ribbon are clearly seen from the images at the three wavelengths (Figures 2(a), (b)). The flare occurs in the region with loops that seemingly cross the magnetic neutral lines, as shown by the small white box in panel (b) of Figure 2. Before the flare starts, the sheared loop structures that are nearly parallel to the neutral line appear at about 04:00 UT in all AIA channels and the shearing motions approximately last for about 10 minutes (see Figure 2(b)). Another group of small loops is also found to be brightened. Simultaneously, the circular ribbon starts to brighten at about 04:10 UT.
In the images of the hot channels at 94 and 131 Å (∼6 and 10 MK respectively), the large overlying loops form arches across the active region and the remote ribbons before the flare, while they are invisible in the images of cool channels at 171 and 304 Å (∼0.8 MK and 50,000 K respectively), as seen in Figure 2(a). It is reasonable to suggest that the large overlying loops are hot and at a high location.

QPPs in the Circular Ribbon
In order to study the emission processes and the motion of the circular ribbon, we perform some slices across the circular ribbon to create time-distance intensity plots. As shown in Figure 3(a), the slice, marked by a white arrow, is nearly perpendicular to the ribbon. The time-distance diagram constructed from the slice using the AIA image sequence at 171 Å is shown in Figure 3(b), which shows several bright stripes with movements. It can be seen that the intensity curve integrated from the diagram (white line in Figure 3(b)) is not smooth, which clearly presents the periodic oscillations and the enhancement before the flare onset. The intensity of the circular ribbon is obviously strengthened at about 04:10 UT, which remains brightened periodically, with a slight movement for We also calculate the UV intensity around the active region at 1600 Å. It is found that the average UV intensity integrated from the white box also shows quasi-periodic pulsations with a period of about 250 s (about 4 minutes), as seen in Figure 3(h). The QPPs appear from 04:00 UT, and last until 04:46 UT, which covers the full process from the preflare phase to the impulsive phase; they become stronger during the flaring impulsive phase.

Multi-wavelength Observations during the Impulsive
Phase of the Flare

EUV Observations
After the brightening of the circular ribbon, the double ribbons in the north part of the active region are brightened at 04:25 UT. At the same time, the circular ribbon continues to expand outward. Then the shearing loops are enhanced impulsively and the flare takes place. The magnetograms of the flaring region observed by HMI/SDO are shown in green (the negative line-of-sight magnetic field) and red (positive) contours in the three enlarged panels of (b). The X-ray sources from RHESSI (12-25 keV, black; 25-50 keV, pink) and the microwave sources from NORH (17 GHz, blue; 34 GHz, red) are close to the two ribbons shown by the contours (50%, 70%, and 95% of the peak flux) overlaid on the 94 Å image at about 04:45:00 UT in the right panel of (c). The AIA/SDO 304, 171, and 131 Å images are presented as animations that run from 04:00 to 05:00 UT.
(An animation of this figure is available.)

Imaging Observations of the Solar Radio Bursts
The solar radio bursts are observed by MUSER-I during the impulsive phase of the M8.7 flare. The images of radio bursts at different frequencies are overlaid on the AIA images in Figure 4. The backgrounds are the composite images of 171 (red), 131 (green), and 94 Å (blue) at different times from The size of the radio source experiences a temporal evolution from small to large and small at a given frequency, which corresponds to the enhancement and weakening of radio emission processes. The centroids from the two-dimensional (2D) Gaussian fitting of the radio source are marked with plus signs. The radio sources are projected over the positive magnetic field in the middle of the active region. An outward movement of the source can be clearly seen from the temporal evolution of the centroid locations. They move in similar directions at all frequencies, which are along the expanding direction of the flaring loop. The radio sources at six frequencies line up very well, seemingly stretching with time from the lower to the higher locations. Figure 5 shows the radio contours at a fixed intensity (top panels) and 50% intensity level (bottom panels) at the start, peak, and the end times of the flare in the frequency range 1.2-2.0 GHz. The radio source sizes increase slightly with decreasing frequencies. Even though the flaring region is close to the solar center, they still address the evolution of the centroid trajectories. The radio sources are found to be cospatial with each other due to the similar morphologies.

QPPs of the Radio Bursts
The solar radio spectrum provides direct overviews of the dynamic processes of the emission source. As seen in Figure 6, a type IV radio continuum and QPPs are recorded in the dynamic spectrum in the frequency range of 0.4-2.0 GHz during the flaring impulsive phase. The figure shows that the lower-frequency cutoff of the type IV radio continuum has a slow drift from a higher to a lower frequency, which is consistent with the upward movements of the radio sources at multiple frequencies. Solar radio type IV bursts are commonly believed to be excited by energetic electrons trapped in the magnetic loops, or ejected plasmoids (Dulk & Altschuler 1971;Smerd & Dulk 1971) associated with the flare impulsive phase or the first stages of CMEs (Bain et al. 2014;Carley et al. 2017). Here, the type IV radio burst at decimeter wavelengths starts at about 04:26 UT, which is almost at the beginning of the flare and lasts for about 18 minutes. It extends to low frequencies near 25 MHz and exists from decimeter to meter wavelengths, and is also accompanied by a type II radio burst observed by HiRAS (Kondo et al. 1995). Figure 6(b) presents the light curves at multiple frequencies: 1.2, 1.7, and 2.0 GHz as observed by MUSER-I, and 4.0 GHz as observed by NoRP from the full solar disk. The flux at 17 GHz is calculated from the microwave images obtained by NoRH. As a comparison, we also integrate the emission intensity of the radio source at 1.7 GHz observed by MUSER-I, which is plotted with a green line in Figure 6(b). It matches well with the light curve at 1.7 GHz from the spectrum. The light curve at 2.0 GHz from MUSER-I observations is similar to that of NoRP but shows more spikes superimposed upon a normal intensity curve. The light curves show clear oscillations during the flaring impulsive phase. Based on wavelet analysis of the radio emission flux at 2.0 GHz, we find that the period of the radio QPPs is about 121 s, very close to 2 minutes (Figure 7(c)). This fact implies that the radio QPPs in the flare impulsive phase are faster than the EUV QPPs in the preflare phase.

Microwave and Hard X-Ray (HXR) Images in the Flaring Region
We analyze the microwave images at 17 and 34 GHz observed by NoRH, and hard X-ray images at the energy bands of 12-25 and 25-50 keV observed by RHESSI. The images at 04:45 UT are seen in Figure 2(c). The microwave source at 17 GHz is almost stable and its projection location is close to the two ribbons before the flare starts. It changes into a microwave loop-top source and lengthens to the ribbons after 04:28 UT. As for the source of 34 GHz, one loop-top and two footpoint sources are clearly seen in the impulsive phase of the radio bursts. They remain almost unchanged after the flare peak time.
We choose the Pixon algorithm (Metcalf et al. 1996) for the RHESSI image reconstruction, which is considered to offer the best accuracy . The integration time for each image is 20 s. The data are available only after 04:41 UT due to the RHESSI night. The 25-50 keV energy band has two HXR sources, while the 12-25 energy band has only one X-ray source. The locations of hard X-ray sources almost coincide with the microwave sources, which indicate that the magnetic reconnection occurs above the loop top. The wavelet analysis has been carried out for the emissions at 17, 34 GHz and 25-50 keV, but no obvious QPPs were found in both emissions.

Topological Configurations of the Magnetic Field
Topological configurations of the magnetic field allow us to analyze the relationships between the flare ribbon geometry and the dynamic processes. The magnetic field distributions of the active region are a positive polarity embedded in an opposite and circumjacent negative polarity. In order to present the different loop systems clearly, we obtain the three-dimensional (3D) data cube of the vector magnetic field corresponding to the active region based on the nonlinear force-free field extrapolation (NLFFF) from the HMI vector magnetogram at 04:00 UT (Wheatland et al. 2000;Wiegelmann 2004; Fleishman et al. 2017) using GX Simulator (Nitta et al. 2015;Nita et al. 2018). The GX Simulator is a powerful and convenient simulation tool in Solar Software, which can automatically download the vector magnetogram data from HMI for a given region and given time. Three toplogical structures are found from the magnetic field extrapolations based on NLFFF modeling, which include three components: the inner spines, fan-shaped field lines, and large overlying spines (see Figure 8). The forced and potential field extrapolations are also adopted to investigate the lower loop system using the line-of-sight component of the magnetic field data (Zhu et al. 2013(Zhu et al. , 2016. All the extrapolations show the "doughnut-like" magnetic field lines at a lower height, which present a crossing structure to the projection plane and a good match with the circular footpoint. Our extrapolation results are consistent with the well known magnetic configurations for the circular ribbon (a)-(c) Different color contours of the radio source with centroids (plus signs) at a fixed intensity, overlaid on the AIA images at the star (04:25 UT), peak (04:32 UT), and end (04:44 UT) times of the radio bursts. (d)-(f) Contours of the radio source with a 50% intensity level at the start, peak, and the end times of the radio bursts. (Masson et al. 2009;Pariat et al. 2009;Wang & Liu 2012;Hou et al. 2016Hou et al. , 2019, which include the null-point and fanseparatrix magnetic configurations.

Location of Source Region
The radio images at frequencies of 1.2-2.0 GHz match well with the EUV images before the bursts start. From the image at 1.7 GHz, there is only one radio source observed at the peak time due to the strong radio radiation and the limited dynamic range of the instrument. This is normal for most intense events observed by radioheliographs. The self-calibration method is carried out for the imaging at other frequencies, with the assumption of one strong radio source during the radio bursts, which is verified by the image at 1.7 GHz. As we know, there is no universal method for phase calibration. Here the results clearly show two radio sources are projected above the two active regions and their relative positions also match well with the EUV observations of the quiet Sun.

Flaring Process
Combining the extrapolated magnetic field configurations and multi-wavelength observations, the relations among the complicated low-loop systems surrounded by the circular ribbon, fan-shaped loops, large overlying spines, and their emissions, could be understood and illustrated as follows.
This is a typical active region that has a circular ribbon around the edge of the active region, several underlying loops inside, and also large overlying loops connecting to the neighboring ribbons. From the magnetic field extrapolation, the fan structure could also be found. The circular ribbon is formed before the flare onset, and the EUV intensity presents QPPs in the preflare phase (from 04:10 to 04:25 UT). Their intensity is enhanced obviously about 10 minutes before 04:25 UT. When the flare takes place, the circular ribbon continues to expand and the EUV QPPs are diminished.
The circular ribbon and the remote ribbons are found to be brightened before the two-ribbon flare starts, which suggests that the null-point reconnection may take place before the main magnetic reconnections. Joshi et al. (2017) investigated a largescale ejective solar eruption on 2014 December 18 in AR 12241. They found that the parallel ribbons began to brighten well before the onset of the circular ribbon, indicating that tether-cutting reconnection initiated this event instead of  breakout reconnection at the null point of a fan dome, which results in formation of the quasi-circular ribbon. However, as stated previously, our event shows a different process than the quasi-circular ribbon that formed before the main reconnection initiated the flare and the null-point reconnection took place before the main flaring process.
The flare occurs in the north part of this active region. Before the flare starts, the sheared loop structure appears at about 04:00 UT in all EUV channels and the shearing motion lasts for about 10 minutes. When the flare takes place, the sheared loops are strongly strengthened at EUV emission and both microwave emissions at 17 and 34 GHz and HXR emission appear in the flaring loops. The radio emissions at 1.2-2.0 GHz recorded by MUSER-I indicate a single source above the active region. It could be distinguished during the flaring impulsive phase (from 04:25 to 04:41 UT). The sources at different frequencies line up well above the active region and all of their intensity profile share the same overall trend: increase, peak, then decrease.
The dynamic process of this flare could be described as follows. Before the flare starts, the small loops under the fan loops are continuously sheared and expanded, which leads to magnetic reconnections in some local and small regions. The circular ribbon region is heated when the small loops root there. After about 04:10 UT, the circular ribbon becomes brightened, sharp-edged, and extended during the preflare phase. They are generally interpreted as the results of the 3D null-point reconnection (Masson et al. 2009;Pariat et al. 2009;Wang & Liu 2012). When the null-point reconnection occurs, a quasicircular ribbon corresponding to the intersection of the fan surface with the chromosphere should be observed (Masson et al. 2009). From our observations, this circular ribbon and the outer ribbons along the spine field lines are almost brightened simultaneously at about 04:10 UT in all EUV wavelengths, which strongly suggests the null-point reconnection has already occurred before the flare starts. In this event, the M-class flare happens in a multi-polar magnetic configuration: the remote positive polarity, which the positive polarity in the middle surrounded by the negative polarity. Before the flare onset, the lower loop system was confined under the fan-shaped loops and the local magnetic reconnection happened continuously, which may cause the lower loops to rise and interact with the fan-dome structure. Those observations are similar to the presence of breakout reconnection (Aulanier et al. 2000;Gary & Moore 2004;Sun et al. 2013;Chen et al. 2016), suggesting that the breakout reconnection may happen and initiate the following CME.
Then, the full magnetic structures are relaxed and a part of the field lines are dragged outward into the heliosphere. The shearing loops in the north part of the circular region continue shearing and expanding. The double-ribbon flare is initiated under the fan loops. And the circular ribbon expands marginally and brightens partly, which may suggest that null-point reconnection still takes place during the flare. The nonthermal electrons that are accelerated in the double-ribbon flare may propagate along the magnetic field lines. The electron beams travel from the magnetic reconnection site along the loop arcades to produce microwave and HXR emissions, which are usually regarded as evidence of the standard two-ribbon solar flare model, also known as the CSHKP model (Carmichael 1964;Sturrock 1966;Hirayama 1974;Kopp & Pneuman 1976).

The Quasi-periodic Oscillations in Preflare Phase and Impulsive Phase
As we mentioned in Section 2, there are three components of QPPs related to the M8.7 flare. Before this flare onset, the EUV circular ribbon presents QPPs with a period of ∼200 s (about 3 minutes), which disappear after the flare starts. During the flare impulsive phase, the radio source at 1.2-2.0 GHz located above the flaring region shows QPPs with a period of ∼120 s (about 2 minutes). For nearly the whole process, including the preflare and impulsive phases of this flare, there are QPPs with a period of ∼250 s (about 4 minutes) from the UV images at 1600 Å in the center of the active region, which is close to the positive magnetic field from 04:00 to 04:45 UT. Then, a question naturally arises: what is the relationship between these QPPs?
It is well known that sunspot oscillations with periods of 3-5 minutes are ubiquitous from multi-wavelength observations. The 3 minute oscillations are widely observed in the umbral chromosphere, transition region, and corona at multiple wavelengths, and the 5 minute sunspot oscillations are mostly observed at the photospheric level and in the chromosphere of sunspots (Thomas 1985;De Moortel et al. 2002;Khomenko & Collados 2015). The UV emission at 1600 Å originates from the bottom plasma of the chromosphere. Therefore, the explanation for the 4 minute UV QPPs might be closely related to the sunspot oscillations. The source region of the 4 minute oscillations is located above the sunspot, which is also cospatial with the nullpoint region, as shown in the extrapolations in Figure 8. The sunspot oscillations change the magnetic field of the fan loops and the spine structures and modulate the UV emission to generate the 4 minute oscillation. Because the sunspot oscillations may occur at any time around the umbral region, the UV 4 minute oscillation therefore may arise in both the preflare and flare impulsive phases.
How do we explain the 3 minute EUV QPPs occurring far from the flare source region (along the circular ribbon) in the preflare phase and the 2 minute radio QPPs occurring near the flare source region (near the null point) in the flare impulsive phase? Are there any relationships between these two QPPs with different periods?
Before we answer this question, we need to know the emission mechanism of the radio bursts. It is very important that we derive the brightness temperature (T b ) from the observation from which we source the emission mechanism of the solar radio bursts. However, the calculation of T b depends on the intensity calibration of the radio observation data. For the 2014 December 17 event, we did not perform the intensity calibration due to the intensity calibration still being in progress around that time. Because in most cases solar radio bursts occurred in the frequency range 1-2 GHz, and the emission mechanisms are believed to be plasma emission, which is generated by the Langmuir turbulence of nonthermal electrons (Dulk 1985;Bastian et al. 1998), we suggest that the possible emission mechanism of the observed radio burst is also the plasma emission.
The magnetic field extrapolations in Figure 8 show that the two regions are connected by large magnetic loops, therefore the two kinds of QPPs might be closely related to each other. They are possibly modulated by the same mechanism. Because the QPPs are observed from both thermal and nonthermal emission, the first possible mechanism is intermittent magnetic reconnection modulation. The intermittent magnetic reconnecting downward and upward plasmoids near the reconnection site may heat plasmas (EUV brightening) and accelerate electrons (radio bursts), which naturally connects the two types of emission, despite their different natures. The intermittent reconnection may take place at a changing pace that is gradual during the preflare stage, with a relatively long period, and faster during the impulsive stage, with a shorter period.
The other possible mechanism is the LRC oscillation modulations in which the coronal loop is twisted and produces electric currents (Zaitsev et al. 1998(Zaitsev et al. , 2000. The currentcarrying plasma loop forms an LRC-circuit resonator, and the circuit oscillations cause periodic modulation of the loop magnetic field, electron acceleration, production of nonthermal electrons, and plasma emission (Khodachenko et al. 2009). The period (P) of the LRC-circuit oscillation is anti-proportional to the electric current (I): P I 10 12 » (s). If we assume that both the preflare 3 minute EUV QPPs and the impulsive 2 minute radio QPPs are triggered by the LRC-circuit mechanism, then we can estimate that the electric current is about 5×10 9 Ampere in the coronal loop during the preflare phase and then increases to about 8×10 9 Ampere during the flare impulsive phase. This estimation seems to conflict with the general picture. However, according to previous research, both the increase and decrease of current densities occur in different flares (Tan et al. 2006;Tan 2007). Tan et al. (2006) found an obvious decrease of electric current occurring in a small compact flare and an increase of electric current occurring in a large powerful flare. After magnetic reconnections of a small compact flare, free energy released during the impulsive phase and results a decrease of the current density. However, in a large flare with complicated magnetic field topological configurations, the magnetic reconnection can produce strong perturbation around the source region, making the magnetic configuration much more complex, and causing an increase of electric current. In our work, the flare is a powerful M8.7 event, which is very similar to the latter regime, with a complex magnetic configuration and an increasing electric current becoming possible. The period decrease of the QPPs from the preflare phase to the impulsive phase indicates an increase of the electric current, and this implies the growth of non-potentiality in coronal loops from the preflare phase to the flare impulsive phase by shearing, twisting, and magnetic reconnections. After the flare peak, because the magnetic non-potentiality becomes exhausted, the electric current diminishes, therefore the LRC oscillations quenched in the flaring loops. Additionally, the 3 minute EUV QPPs are also consistent with the GOES soft X-ray QPPs in the preflare phase reported by Tan et al. (2016).
If the above interpretation is true, then the electric current should be an important link between the preflare evidence and the flaring processes. And the existence of electric current in the preflare phase may indicate that some non-potentiality activities may occur before the onset of the solar flare. Therefore, the preflare QPPs can be regarded as a precursor to solar flares. Other mechanisms can still possibly explain the formation of 2-3 minute QPPs, such as the leakage of slow magnetoacoustic waves of 3-5 minutes sunspot oscillations from the plasma loops, but this explanation cannot easily demonstrate the change of the periods from the preflare to the impulsive phase.
According to the basic characteristics of solar radio emissions, the radio source at higher frequencies arises from the region close to the solar surface (Dulk 1985). From MUSER observations, we find a increase in the size of the radio source with a decrease in frequency. Although there is a projection effect on the imaging, it could also be found that the radio sources line up from lower (higher frequency) to higher (lower-frequency) locations. From the dynamic spectrum, the type IV solar radio continuum is overlaid by QPPs, which show oscillations with a period of about 2 minutes during 04:25-04:41 UT. The light curves are similar but have much finer structures compared to NORP light curves. The average intensity integrated from the brightest radio source shows a similar profile to the intensity curve from the spectrum, which indicates most spectral structures are produced in this active region.

Summary
In this work, we investigated an M8.7 class flare using radio imaging (MUSER-I) combined with EUV/UV spectroscopy (AIA/SDO), microwave (NORH), and X-ray imaging (RHESSI) observations. The main results of these joint analysis are as follows: (1) The radio imaging observations clearly show two radio sources corresponding to the two active regions, and their relative positions match well with the EUV images obtained by AIA/SDO. The radio source covers a region with positive magnetic field, which shows outward movement and stretches from lower to higher locations.
(2) Three components of QPPs are distinguished and confirmed in this flare: 4 minute UV QPPs at 1600 Å images in the center of the active region from the preflare phase to the impulsive phase (04:00 to 04:45 UT), 3 minute EUV QPPs along the circular ribbon during the preflare phase, and 2 minute radio QPPs in the frequency range 1.2-2.0 GHz around the two-ribbon flaring region during the impulsive phase.
(3) Through a joint analysis of magnetic field extrapolation and EUV images, we suggest that the 4 minute UV QPPs should be modulated by sunspot oscillations, and the 3 minute EUV QPPs are closely linked to the 2 minute radio QPPs by connecting their source region from coronal loops and by a similar formation mechanism. We proposed two candidate mechanisms, one related to the downward and upward plasmoids from the intermittent reconnection region, and the other related to the LRCcircuit mechanism that might demonstrate their formation and the implications for flare precursors.
The close relationship between the 3 minute EUV QPPs in the preflare phase and the 2 minute radio QPPs in the impulsive phase shows the evolutionary process of the non-potentiality in the active region. As the existence of an electric current in the plasma loop may excite a series of plasma instabilities, such as kink-mode, tearing-mode, or ballooning-mode instabilities, this may help us to understand the triggering process of solar flares.
In this work, we observed the spectral structure of the quasiperiodic pulsations overlaid on a type IV solar radio continuum and obtained related radio images in the frequency range 1.2-2.0 GHz for the first time using MUSER-I. The radio observations from MUSER allow us to better understand the evolution of the flare eruption and the related electron acceleration mechanism, including the initiation, eruption, and propagation into the corona. However, so far, the calibration of the observation data is not perfect. We have to use the satellite signal as the calibration source at 1.7 GHz and use self-calibration to obtain the radio images at frequencies beyond 1.7 GHz. In a future step we will include two 20 m antennas in the MUSER array to observe some quasars that can be be the calibration source in the full frequency range. When this work and related software are finished in the near future, we will be better equipped to study the complete solar eruption process.