Hard X-ray Emission During Flares and Photospheric Field Changes
O. Burtseva, J. C. Martínez-Oliveros, G. J. D. Petrie, A. A. Pevtsov
aa r X i v : . [ a s t r o - ph . S R ] M a y Hard X-ray Emission During Flaresand Photospheric Field Changes
O. Burtseva , J. C. Mart´ınez-Oliveros , G. J. D. Petrie , and A. A. Pevtsov [email protected] Received ; accepted National Solar Observatory, Tucson, AZ 85719 Space Sciences Laboratory, UC Berkeley, Berkeley, CA 94720 2 –
ABSTRACT
We study the correlation between abrupt permanent changes of magneticfield during X-class flares observed by the GONG and HMI instruments, andthe hard X-ray (HXR) emission observed by RHESSI, to relate the photosphericfield changes to the coronal restructuring and investigate the origin of the fieldchanges. We find that spatially the early RHESSI emission corresponds well tolocations of the strong field changes. The field changes occur predominantly inthe regions of strong magnetic field near the polarity inversion line (PIL). Thelater RHESSI emission does not correspond to significant field changes as theflare footpoints are moving away from the PIL. Most of the field changes startbefore or around the start time of the detectable HXR signal, and they end atabout the same time or later than the detectable HXR flare emission. Some ofthe field changes propagate with speed close to that of the HXR footpoint at alater phase of the flare. The propagation of the field changes often takes placeafter the strongest peak in the HXR signal when the footpoints start movingaway from the PIL, i.e. the field changes follow the same trajectory as the HXRfootpoint, but at an earlier time. Thus, the field changes and HXR emissionare spatio-temporally related but not co-spatial nor simultaneous. We also findthat in the strongest X-class flares the amplitudes of the field changes peak a fewminutes earlier than the peak of the HXR signal. We briefly discuss this observedtime delay in terms of the formation of current sheets during eruptions.
Subject headings:
Sun: flares - Sun: magnetic fields - Sun: photosphere - Sun: X-rays
1. Introduction
Significant, abrupt and permanent photospheric magnetic field changes duringmajor X-class and M-class flares have been reported by several authors (e.g., Wang1992; Wang et al. 1994; Kosovichev & Zharkova 2001; Wang et al. 2002; Sudol & Harvey2005; Petrie & Sudol 2010; Wang et al. 2012; Petrie 2013). No numerical model has yetreproduced these observations and physically related the field changes at the photosphere(where the magnetic field is not force-free) with the flare that occurs in the nearly force-freefield environment in the corona. Fletcher & Hudson (2008) suggested a scenario whenthe Alfv´en wave pulses transport energy rapidly from the flare site in the corona tothe lower atmosphere as a possible explanation of the observed rapid variations to theline-of-sight (LOS) component of the photospheric magnetic field during the flare impulsivephase. Hudson et al. (2008) introduced the loop-implosion scenario, Hudson et al. (2008)and Fisher et al. (2012) provided the method for estimating flare-related Lorentz forceback-reaction, and several studies (e.g., Wang & Liu 2010; Petrie & Sudol 2010; Wang et al.2012; Alvarado-G´omez et al. 2012; Petrie 2012, 2013) estimated the Lorentz forces frommagnetic field observations. Petrie (2014) showed that the method of Fisher et al. (2012)could be used for coherent changes of strong fields within active regions.Johnstone et al. (2012) analyzed the spatial and temporal relationship between thephotospheric field changes and chromospheric ultraviolet (UV) emission. They found thatthe field changes were co-spatial with UV emission. In all studied cases the chromosphericbrightenings began significantly earlier and ended much later than the photospheric fieldchanges. Thus the authors suggested that the chromosphere must be more responsiveto flare energy because more energy from the corona reaches the chromosphere than thephotosphere and because the chromosphere is less dense. 4 –Cliver et al. (2012) analyzed photospheric magnetic flux changes and soft X-ray (SXR)emission data and found a sharp change in the unsigned magnetic flux co-temporal with theonset of the flare impulsive phase and the end of the stepwise change co-temporal with theSXR emission peak. Both conclusions, by Johnstone et al. (2012) and Cliver et al. (2012),are consistent with the scenario proposed by Fletcher & Hudson (2008) when Alfv´en wavesrapidly transport field changes from the reconnection site in the corona to the photosphere.The non-thermal HXR emission in the flaring chromosphere has a strong associationwith magnetic reconnection in the corona (e.g., Fletcher & Hudson 2002). HXR emissionsources most often appear as kernels at the outer edges of H-alpha/UV ribbons (e.g.,Svestka et al. 1982). Recent studies found that both the reconnection and energy releaserates are stronger at the ribbon segments accompanied by HXR kernels than at thosewithout HXR sources (e.g., Asai et al. 2004; Temmer et al. 2007). Also, the footpointmotions away from the neutral line are considered to be indicative of the reconnectionoccurring in different heights of arcade magnetic fields with a displacement speed roughlyproportional to the rate of reconnection.Cliver et al. (2012) also looked at the timing of the total flux change and hard X-ray(HXR) emission during three X-class flares and found that the onsets of the stepwiseincrease in the total flux and the rise in HXR emission are co-temporal. Yurchyshyn et al.(2004) investigated the connection between the HXR emission and LOS magnetic fieldchanges associated with X4.8 flare on July 23, 2002 also finding that the start of thesharp rise in HXR emission and the start of the strong stepwise magnetic field changes areco-temporal. Zharkova et al. (2005) analyzed the same flare and reported that the magneticfield changes started 6-7 minutes before the HXR emission onset and ended 20 minuteslater. Yet, more studies including larger statistical samples, possibly new approaches,and more details on timing of the field changes and flare emissions during different flare 5 –phases would be beneficial for a better understanding the relation between the reconnectionprocesses in the corona and the photospheric magnetic filed changes during flares.In this work, we investigate the spatial and temporal correlation between abruptpermanent changes of magnetic field during five X-class flares observed by Global OscillationNetwork Group (GONG) and one X-class flare observed by Helioseismic and MagneticImager (HMI) instrument on board Solar Dynamics Observatory (SDO), and the locationof HXR emission observed by Reuven Ramaty High Energy Solar Spectroscopic Imager(RHESSI). Our purpose is to explore the likelihood that the photospheric field changesand the chromospheric HXR sources are generally caused by a common physical process inthe corona during a flare, and to provide new observational constraints for models of solarflares. We also propose a scenario connecting the photosphere and a reconnection region inthe corona that could explain our results.
2. Observations
GONG longitudinal magnetic field data were previously used by Sudol & Harvey (2005)to characterize pixel-by-pixel abrupt, stepwise, permanent field changes during 15 X-classflares. Petrie & Sudol (2010) extended this study to a large sample of flares of intensity M5and above. Following these studies, we analyze ∼ ◦ × ◦ field of view, local to a flare, and tracked; (ii) each image 6 –is registered to a reference image formed from the average of the 10 one-minute imagesimmediately preceding the flare.The stepwise field changes are modeled as in Sudol & Harvey (2005) by fitting thestep-like function: B ( t ) = a + bt + c (cid:18) π tan − ( n ( t − t )) (cid:19) (1)where t is time, a and b model the linear background field evolution, c is the -amplitudeof the step, n is the inverse time parameter associated with the slope of the step, and t isthe time at the midpoint of the step. The start time, t s , and the end time, t e , of the fieldchanges are derived from the fit parameters as: t s = t − π/ (2 n ) and t e = t + π/ (2 n ).The magnetic field in each pixel in an active region as a function of time was fittedwith Eq. (1) and spatial maps of the fit parameters were created. In the following we referto these as c -, t s -, t -, and t e -maps.To avoid spurious fits of Equation 1, the parameter maps include only pixels meetingthe following criteria (see Petrie & Sudol (2010) for more details): (i) exhibiting reasonablysized field changes | c | <
500 G; (ii) with steps of reasonably short duration <
40 minutes;(iii) with the time of the step occurring within 20 minutes of the GOES flare start time.We further require that: (iv) the time series of measurements passed a reduced- χ test;(v) the background field values were not unreasonably large, | a | < | a | > | c | >
500 G have emission artifacts. It is possible todetect a small minority of good cases with | a | > | c | >
500 G by eye, butflux integrals are less affected by noise and artifacts with the conditions | a | < | c | <
500 G imposed.The stepwise field changes associated with HXR emission generally have largeamplitudes around 100-200 G and above (e.g., Fletcher & Hudson 2008). Sudol & Harvey(2005) observed 100-200 G abrupt permanent line-of-sight photospheric magnetic fieldchanges co-spatial with flare ribbons. It has been shown in previous studies (seePetrie & Sudol 2010; Burtseva & Petrie 2013) that the amplitude and number of thestepwise field changes are greater for larger flares. Histograms of the field change amplitudesabove 50 G for the five flares observed by GONG are shown in Figure 1. For the threestronger flares in this work, there is a good sample of field changes above 100 G, whilethere are only few of them in the two weaker flares. In order to have a reasonably goodsample of the field changes for each flare in our analysis, we exclude pixels with field changeamplitude | c | <
50 G.We also use HMI 12-minute vector magnetograms for the X2.1 flare on September 6,2011 in the form of Active Region Patches (HARPs), maps of B r , B θ and B φ with pixel size0 ◦ .
03 in heliographic coordinates derived by cylindrical equal-area (CEA) projection (seeHoeksema et al. 2014). A list of analyzed flares with their GOES times, location on thedisk and instrument used is given in Table 1.We compare the location and timing of the magnetic field changes and the HXR flareemission derived from RHESSI observations. RHESSI images were synthesized using theCLEAN algorithm (see, e.g., Dennis & Pernak 2009). The cadence used varied from 8 to 30seconds in an energy range from 30 up to 300 keV. The RHESSI images were remappedto the same grid with the corresponding pixel size and tracked to the same minute as themagnetic field data. The centroid position of each HXR source at each moment of time wasdefined by fitting a two-dimensional elliptical Gaussian above 50% of the local maximumwith an accuracy within the pixel size of the data. The total flux of each HXR source at 8 –Table 1: Date, class, time and location of analyzed flaresDate GOES GOES Start GOES Peak GOES End Location NOAA InstrumentClass Time (UT) Time (UT) Time (UT) Number2003 Oct 29 X10.0 20:37 20:49 21:01 S15W02 10486 GONG2003 Nov 2 X8.3 17:03 17:25 17:39 S14W56 10486 GONG2004 Aug 13 X1.0 18:07 18:12 18:15 S13W23 10656 GONG2005 Jan 15 X2.6 22:25 23:02 23:31 N15W05 10720 GONG2006 Dec 6 X6.5 18:29 18:47 19:00 S05E64 10930 GONG2011 Sep 6 X2.1 22:12 22:20 22:24 N13W18 11283 HMIeach moment of time was computed within the region of 50% of the maximum flux.Also, TRACE 171, 195 and 1600 ˚A and AIA 171 ˚A observations were used for contextinformation giving us an idea of the conditions of the upper chromosphere and coronal loopconfiguration during flares.
3. Spatial correlation between the field changes and HXR sources
It has been reported by several authors (e.g., Krucker et al. 2003; Metcalf et al. 2003;Temmer et al. 2007), that HXR sources display a complex motion pattern, where the flarefootpoints move along the neutral line in the impulsive phase, and later move away fromand perpendicular to the neutral line. Yang et al. (2011) see this two-phase motion of theHXR footpoints as a possible indication of of the two-phase magnetic reconnection process,implying the reconnection electric field could have different properties in the impulsiveand gradual phases of a flare. In an earlier study Bogachev et al. (2005) characterized 9 –observed HXR footpoint motions into three types. The first type is a motion awayfrom and perpendicular to the PIL as more flux is reconnected, in agreement with thestandard flare model. The second and third types are antiparallel and parallel motionsalong the PIL, indicating shear relaxation (see, e.g., Masuda et al. 2001; Ji et al. 2007,for supporting observations) and motion of acceleration region in the corona along theseparator, respectively. We see all of these motion patterns in the flares we have analyzed.The RHESSI HXR footpoint contours near the emission peak time and temporal evolutionof the source centroids positions during the five flares observed by GONG are shown inFigures 2 and 3, and those during the flare observed by HMI are shown in Figure 4.In the three stronger flares in our study, observed by GONG (see Figure 3), we seeantiparallel motion of the main HXR sources (labeled S1 and S2) along the PIL and thenaway from the PIL after the HXR flux has reached its maximum. Besides the two mainsources, a third HXR source, labeled S3, appears in two of the flares, the X8.3 and X10.0.It is seen for a short time only in the beginning of the HXR emission and does not show aclear pattern in its motion. It also could be a part of the S2 HXR footpoint, but appearsas a separate source at the 50% level of the maximum HXR flux. The S1 footpoint in theX6.5 flare is much weaker than the S2 footpoint. It dims very quickly and does not move asmuch as the S2 source moves. This asymmetry in the brightness of the two footpoints couldbe indicative of an asymmetric loop. Also the HXR footpoint in the stronger magneticfield appears to move more slowly than in the weaker magnetic field (see, e.g., Jing et al.2008; Yang et al. 2011). The slower HXR motion in stronger compared to weaker field isconsistent with an approximately steady reconnection rate.In one of the weaker flares, observed by GONG (see Figure 2, bottom), the X2.6 flare,we notice parallel motion of the S1 and S2 HXR sources along the PIL in the impulsivephase of the flare, and an asymmetric motion after the strongest HXR peak: continuous 10 –motion of S1 along the PIL and motion of S2 away from the PIL. The S3 and S4 HXRsources, appearing in the eastern part of the region, also show an asymmetric motion. TheS3 footpoint moves along the PIL towards the south-east and then turns around and movestowards the north-west in the direction opposite to the motion of the S1 footpoint. The S4source moves parallel to the S3 and then away from the PIL. Liu et al. (2010) have alsoanalyzed the asymmetric motion of the HXR footpoints in this flare and interpreted it asan asymmetric eruption and thus magnetic reconnection progressing along the PIL.In the other weaker flare (see Figure 2, top), the X1.0 flare, there is only one detectableHXR source, moving parallel to the PIL and apparently above the PIL. In the movies ofthe TRACE 171 ˚A observations (see Section 4), where HXR sources were over-plotted,it also looks like a motion of a loop top rather than a footpoint motion, however it ishard to say with certainty. Loop-top sources are most easily observed in limb flareswhose loop footpoints are occulted by the limb. Footpoint sources are stronger than theircoronal counterparts during the impulsive phase of the flare. After the impulsive phase thefootpoints fade and the loop-top becomes brighter, nevertheless, coronal sources above 30keV are rarely observed (e.g., Tomczak & Ciborski 2007).In the X2.1 flare observed by HMI two footpoints move along the PIL in the samedirection (see Figure 4) and seem to start turning away from the PIL at the last data point.However, it is hard to say what happened beyond the time frame of the HXR data set wehave for this flare. The spatial location of the two footpoints coincides with the locationof the strongest magnetic field changes in both radial, B r , and horizontal, B h , vector fieldcomponents and corresponds to the strong tilt angle change.In all of the analyzed flares, the early RHESSI emission corresponds well to locations ofstrong field changes. The strong field changes occur predominantly in the regions of strongmagnetic field near the PIL. The later RHESSI emission does not correspond to significant 11 –field changes as the footpoints are moving away from the PIL.
4. Comparison of HXR with UV and EUV observations from TRACE andAIA
In order to visualize the location and evolution of the HXR footpoints during a flarewith respect to the coronal loop configuration, we over-plot the HXR sources on top ofTRACE 171, 195 and 1600 ˚A images (subject to availability of the particular TRACEdata) and run the movies constructed from this data. The comparison suggests that theHXR footpoints seem to be confined to short, bright loops. We see that, in agreement withprevious studies (e.g., Masuda et al. 2001; Fletcher & Hudson 2002; Cheng et al. 2012),the HXR sources are well correlated in space and time with bright kernels of the UV andextreme UV (EUV) emission. GONG intensity and TRACE WL observations (subjectto availability) show the HXR footpoints moving roughly along the umbral-penumbralboundaries in the sunspots, consistent with previous studies (e.g., Temmer et al. 2007).Johnstone et al. (2012) found a pattern linking 1600 ˚A brightenings and longitudinalfield changes: the brightenings occurred wherever field changes took place but not viceversa. In terms of the temporal correlation between field changes and flares, the authorsfound that the UV brightenings began before and ended after the field changes. Weinvestigate the temporal relationship between the HXR emission and magnetic field changesduring the flare in the next Section.
5. Temporal correlation between field changes and HXR sources
In order to determine the temporal relationship between the magnetic field changesand HXR emission during flares, we perform the following analysis. First, we analyze the 12 –step-function fit parameter maps (the field change amplitude, c -, start time, t s -, midpointtime, t -, and end time, t e -maps), derived from GONG magnetic field data using Eq. (1),along with temporal evolution of the HXR flare emission. Second, we examine the temporaland spatial correlation of the field change and the HXR signal by tracking the field changesdirectly from magnetograms with the HXR footprint contour masks at each moment oftime. We also analyze the timing of the total flux change of the whole flaring region and atthe locations of the strongest field changes next to the PIL. Each of the t s -, t - and t e -maps provides information about the field change timing ineach pixel where a strong stepwise change occurred during a flare. We construct a histogramof the field change over time by integrating amplitudes of the field changes from the c -mapsover one-minute intervals around each flare for each of the start, midpoint, and end timefrom t s -, t - and t e -map, respectively. Figure 5 shows the computed total magnetic fluxchange as well as total flux of the HXR signal as functions of time. Most of the field changesstart before or around the start time of the detectable HXR flare emission, and they end atabout the same time or later than the detectable HXR signal. The three strongest flares,the X6.5, X8.3 and X10.0, show a peak in the total magnetic flux change ∼ t of a stepwise field change has a smaller error than the estimated start and end times t s and t e , as the latter ones strongly depend on the duration of the step: the longer the stepduration, the larger the uncertainty of the t s and t e (see Petrie & Sudol 2010, for details).The mean uncertainty in t in our analysis is ∼ .
28 minutes, which is smaller than the timedifferences between the peak in total flux change and the peak in the HXR signal found forthe three strongest flares in this work.The two weaker flares do not seem to have a strong peak in the flux change at a certaintime and the field changes do not show clear temporal relation to the HXR signal. Forthese flares, only very few pixels had field change amplitudes around or above 100G, and ingeneral the number of field changes that passed the rejection criteria and were larger than50G is less than half as many as for the three strongest flares (see Figure 1). The total fluxchanges for the weaker flares are at least 4 times smaller than the total flux changes for thestronger flares.To understand why the stronger X-class flares show a strong peak in the flux change,which does not seem to be the case for the weaker flares, we compute the total backgroundmagnetic flux in the pixels, contributing to the total flux change in Figure 5, at each minute.The background magnetic field is derived from a reference image composed of ten remappedGONG magnetograms immediately preceding the flare. The total unsigned backgroundfield, shown in Figure 6, is highly correlated with the total unsigned flux change that wasalso found in previous studies (see Petrie & Sudol 2010; Burtseva & Petrie 2013).For the larger statistical sample of 77 flares (39 M-class and 38 X-class) studied byPetrie & Sudol (2010), we now analyze the total flux change as a function of time, andcompute their histograms in the same way as we did for Figure 5. Visually we find that all 14 –of the flares above ∼ class X5 show a strong peak in their flux change over a short timespan.Some of the weaker flares also show a peak but the peak is lower and not as sharp relativeto the amplitudes of the flux changes at other times.To generalize the result to the larger set of 77 flares, we compute the relative differencebetween the maximum peak value and median value of the total flux change during flaretime, where the median is the reference. Some of the flare data sets include very few pixelswhose flux changed in a stepwise manner during the flare. Fewer pixels recorded stepwisefield changes in general during weaker flares (see Petrie & Sudol 2010; Burtseva & Petrie2013). We ignore those cases where the number of pixels with significant flux change duringa flare was <
10. The relative difference as a function of the GOES flare class is shownin Figure 7. A linear trend is clearly seen on the plot. The Pearson linear correlationcoefficients for the start, midpoint and end times are 0.65, 0.51 and 0.52 and their P -values,the probability that the observed correlation occurs by chance, computed using the t-test,are 4 . × − , 3 . × − and 1 . × − , respectively, revealing a moderate correlationbetween the relative peak height in the flux change and GOES flare class. This confirmsthat the stronger flares, in comparison with the weaker flares in our sample, in general,show a stronger peak in the flux change. We also analyze the correlation between the field changes and HXR flare emissiondirectly from GONG magnetic field data, tracking the field changes with RHESSI footpointcontours at 50% of the emission peak. We sum the magnetic field inside the footpointcontours at each footpoint location at each moment of time. As a result we have temporalprofiles of the total flux at each location of the HXR sources during each flare. An exampleof the flux profiles for one of the HXR sources of the X10.0 flare is shown in Figure 8. The 15 –profiles are smoothed with a Gaussian kernel. The width of the Gaussian window variesfrom 2 to 3 minutes for different flares depending on the noise level in their magnetic fluxprofiles. We define the midpoint time of the any profile showing a clear step as a maximumgradient of the step. Since often these profiles have a spike (the magnetic transient, seeFigure 8 around 20:45 UT on the x-axes), we also define the time of the spike.The midpoint times of the field changes as a function of timing of the HXR source areshown in Figure 9. We find that most of the field changes occurred earlier than the HXRsignal was detected in the region. Some of the field changes propagate with speed close tothat of the HXR footpoint, but at a later phase of the flare, often after the HXR signalreaches its maximum and the footpoint has begun to move away from the PIL. Thus, thefield changes follow the same trajectory as the HXR footpoint, but at an earlier time. Mostof these field changes are decreases. On the other hand, there are field changes that occurat later times, but at the locations of earlier HXR sources close to the PIL; these tend tobe increases: to see this, refer to the timing for the HXR source S2 of the X6.5 flare and forthe HXR source S1 of the X2.6 flare in Figure 9.The X1.0 flare had only one well detected HXR source. As mentioned in Section 3,this HXR footpoint moved along the PIL, about half of it on the negative polarity sideand about half on the positive polarity side. It appears that the negative polarity unsignedtotal flux decreases and the positive polarity unsigned total flux increases during the flare,both showing a clear abrupt step. The average unsigned total flux over the whole footpointdoes not show a clear step. Thus the times for this event were defined separately for thenegative, S -1, and positive, S +1, polarity part of the HXR footpoint. As seen in Figure9, both the field changes propagate with nearly the same or slightly slower speed as that ofthe footpoint, but most of them, both the increases and the decreases, occured at a latertime, after the footpoint has passed through the region. 16 –Sudol & Harvey (2005) analyzed field changes during the X2.6 flare on December 11,2001 and found that the field changes propagate with the H α flare ribbon at a similarspeed. They do not mention whether the field change occurred earlier or later than the flareribbon. They saw more cases of propagation of the field changes across the active region,but in all of the cases the field changes were too fast and restricted to a small area to beaccurately tracked. Petrie & Sudol (2010) also notice a propagation of the field changesstarting from near the PIL across the region on the t -map of the X6.5 flare on December6, 2006.The so-called magnetic transient (MT) or magnetic anomaly, significant reversiblemagnetic field change (as opposed to stepwise permanent field changes) that appears as aspike in some of our temporal flux profiles during the impulsive phase of a flare, was observedby many authors (e.g., Kosovichev & Zharkova 2001; Qiu & Gary 2003; Maurya et al. 2012;Harker & Pevtsov 2013). Kosovichev & Zharkova (2001) and Harker & Pevtsov (2013)suggest that this phenomenon could be a real change in magnetic field, while Qiu & Gary(2003) and Maurya et al. (2012) conclude that it is a result of a line-profile change duringflare. The subject is still under debate. In this paper we do not attempt to analyze thenature of the MT, but we describe our observations. We measure the time of the fluxminimum in the MT and plot it as a function of HXR time in Figure 10. The MTsassociated with HXR sources S1 and S2 in the X10.0 flare, and source S2 in the X6.5 flaresare co-spatial and co-temporal with the HXR footpoint in the later phase of the flares,after the HXR emission maximum, when the footpoints are moving away from the PIL. Incontrast, the MT associated with source S2 in the X2.6 flare propagates with the footpointearlier in the impulsive phase of the flare. In case of the HXR source S1, S3 and S4 in theX2.6 flare, and all HXR sources in the X1.0 and X8.3 flares, the MTs are not found to bepropagating with the footpoints. They occur along the footpoint trajectory approximatelyat the time of the HXR emission peak. We also note a transient increase in the magnetic 17 –field in two flares, the X1.0 flare in the S -1 location and the X6.5 fare in the S2 laterlocation that appears ahead of the downward spike time. We compute the total unsigned flux over a rectangular domain including the wholeflaring active region cropped to the area of the strongest field changes. For the five flaresobserved by GONG areas of magnetic field integration correspond to the regions shown inFigures 2 and 3. For the flare observed by HMI this area refers to a smaller, 2 degrees inlatitude by 1 degree in longitude, rectangular region around center of the region shown inFigure 4. The magnetic fluxes in all of the flaring regions show a stepwise change associatedwith the flare (Figure 11). In some of the flares the step is clear and abrupt, whereas inothers it is not as strong and obvious on top of the noise or sharp short-time transientfield change during the flare. We see that the step-wise field change is co-temporal withthe the total HXR flux for most of the flares. The field change starts at around the timewhen emission in the HXR sources is first detected, and it ends at around the same timeas the HXR sources. It is consistent with the results of Cliver et al. (2012) who averagedall pixels’ profiles in a flaring region showing strong stepwise change and reported that theonsets of the stepwise rise in the total flux and the rise in HXR emission are co-temporal.Petrie & Sudol (2010) and Burtseva & Petrie (2013) found that significantly moremagnetic flux decreases than increases occurred during the flares, consistent with a modelof collapsing loop structure for flares. Consequently, integrated magnetic field over a flaredactive region should most likely show a decreasing change in the field. In this work the totalunsigned fluxes in all flares, observed by GONG, decrease. In Cliver et al. (2012) unsignedfluxes increase by construction, as all pixels’ profiles were arranged so that they all changedin the same sense, thus preventing the individual pixels’ steps from canceling each other. 18 –The flare observed by HMI shows an increase in total unsigned flux for both vertical, B r , and horizontal, B h , field components, and the average tilt angle. The change in the Bhis much stronger than in the B r as was found also by Petrie (2012). The B r componentshows the magnetic transient in its profile, while B h does not. The 12-minute vector dataseems to resolve the overall flux change over the rectangular 2 × ∼
6. Discussion
We investigate the spatial and temporal connection between photospheric magneticfield changes during six X-class flares and the location of HXR emission observed byRHESSI to relate the field changes to the coronal restructuring and investigate theorigin of the field changes. We interpret these field changes as a response to magneticreconnection in the corona. Although we do not observe the reconnection process itself,the observed spatio-temporal relationship between the footprint HXR and the photosphericfield changes indicates a relationship between the two, and the HXR is widely associatedwith reconnection. Therefore it is natural to associate the photospheric field changes to thereconnection also (see, e.g., Temmer et al. 2007; Jing et al. 2008; Yang et al. 2011), eventhough the precise nature of this relationship is not yet clear.We find that spatially the early RHESSI emission corresponds well to locations ofthe strong photospheric field changes. The strong field changes occur predominantly inthe regions of strong magnetic field near the PIL. The later RHESSI emission does notcorrespond to significant field changes as the flare footpoints are moving away from the PIL.Most of the field changes start before or around the start time of the detectable HXR signal,and they end at about the same time or later than the detectable HXR flare emission. Some 21 –of the field changes propagate with speed close to that of the HXR footpoint at a laterphase of the flare, often after the HXR signal reaches its maximum, when the footpointstarts moving away from the PIL, i.e. the field changes follow the same trajectory as theHXR footpoint, but at an earlier time. Thus, the field changes and HXR emission arespatio-temporally related but not simultaneous. We also find that in the strongest X-classflares amplitudes of the field changes peak a few minutes earlier than the peak of the HXRsignal.Because of the several-minute time delays between the HXR peaks and the largest fluxchanges in some of the flares, and the sub-minute coronal travel times, a simple causallink between the HXR and flux changes does not seem to be possible. The HXR emissionand the field-change effects are not likely to be an indication of the same event, thoughthey must be related in some way. They may be a result of the consecutive process: flareonset, then photospheric magnetic field changes, then particles acceleration resulting inHXR emission. The main photospheric field changes may be due to earlier reconnectionlower in the atmosphere, in contrast to the reconnection higher in the corona associatedwith the HXR peak. However if the main field changes were due to reconnection low inthe atmosphere occurring earlier than the HXR peak, the HXR could not lag the mainphotospheric field changes along common spatial paths as observed. Instead, the HXR peakwould be expected at a later time and in more distant location of the HXR source from thePIL, than the largest field changes.The Alfv´en speed in the corona is usually estimated to be about 1000 km/s, and canbe as high as a few tenths of the speed of light in the lower corona in presence of strong, ∼ Acknowledgments
We thank Hugh Hudson for discussions and helpful comments. This work utilizesthe Global Oscillation Network Group (GONG) data obtained by the NSO IntegratedSynoptic Program (NISP), managed by the National Solar Observatory, which is operatedby AURA, Inc. under a cooperative agreement with the National Science Foundation.The data were acquired by instruments operated by the Big Bear Solar Observatory, HighAltitude Observatory, Learmonth Solar Observatory, Udaipur Solar Observatory, Institutode Astrofsica de Canarias, and Cerro Tololo Interamerican Observatory. The HMI and AIAdata supplied courtesy of the SDO/HMI and SDO/AIA consortia. SDO is a mission forNASA’s Living With a Star (LWS) Program. The Transition Region and Coronal Explorer(TRACE) is a mission of the Stanford-Lockheed Institute for Space Research, and part ofthe NASA Small Explorer program. O. Burtseva, G. J. D. Petrie and A. A. Pevtsov arepartially supported by NASA grant NNX14AE05G. J. C. Martinez-Oliveros was supportedby NASA grant NNX11AP05G for RHESSI. 25 –Fig. 1.— Histograms of the field change amplitudes above 50 G for each of the five flaresobserved by GONG. 26 –Fig. 2.— LOS magnetogram of the flared active region (left), cropped to the area of thestrongest field changes, represented by the fit parameter c (right) for the X1.0 and X2.6flares, the two weaker flares in our analysis observed by GONG. Location of the strong fieldchanges is indicated by gray contours at 50 and 100 Gauss of the field change amplitude.RHESSI 30-70 keV and 40-100 keV HXR footpoint contours for the two flares, respectively,at levels from 35% to 95% of the maximum counts at around the emission peak time aresuperposed in light blue. The temporal evolution of the source centroids is shown with plussymbols, which increasing size represents progressing times during flares and different colordenotes different sources. In addition to that, direction of the HXR footpoints motion forthe X2.6 flare is indicated by arrows for clarity purposes. HXR sources are labeled by ’S’letter with a number. 27 –Fig. 3.— The same as Figure 2, but for the X6.5, X8.3 and X10.0 flares, the three strongerflares in our analysis observed by GONG. RHESSI 50-300 keV, 50-300 keV and 50-100 keVHXR footpoint contours for the three flares, respectively, superposed in light blue are atlevels from 50% to 90% of the maximum counts at around the emission peak time. 28 –Fig. 4.— Vector magnetic field of the flared active region (left) and the vector magnetic fieldchanges (right) during the X2.1 flare observed by HMI. From top to bottom: the verticalfield component, the horizontal field component and the field tilt angle. RHESSI 50-160keV HXR footpoint contours at levels from 50% to 95% of the maximum counts at aroundthe emission peak time are superposed in light blue. The temporal evolution of the sourcecentroids is shown with plus symbols, which increasing size represents progressing timesduring flares and different color denotes different sources. 29 –Fig. 5.— Total flux change as a function of time during each of the five flares observed byGONG. Time axis is in minutes relatively the GOES flare start time. Black solid, dark-greydashed and light-grey dash-dotted lines correspond to the start, midpoint and end times,respectively. Total HXR flux is over plotted in red. 30 –Fig. 6.— Total background flux as a function of time during each of the five flares observedby GONG. Time axis is in minutes relatively the GOES flare start time. The backgroundflux is obtained for the same pixels used to compute flux change shown in Figure 5. 31 –Fig. 7.— Relative difference between the maximum peak value and median value of the totalflux change during flare time as a function of the GOES flare class. Black pluses, light-bluecrosses and magenta squares correspond to the start, midpoint and end times, respectively. 32 – F l u x ( M x ) F l u x ( M x ) F l u x ( M x ) F l u x ( M x ) Fig. 8.— Magnetic flux as a function of time (black line) at each location of the HXR sourceS1, for first 6 min. 20 s. of the HXR signal data, during X10.0 flare on October 29, 2003.The magnetic flux time series smoothed with Gaussian kernel is over plotted in red. X-labelsshow universal time of the magnetic field data, while the titles represent universal time ofthe HXR signal. 33 –Fig. 9.— Midpoint time of the field change as a function of timing of the HXR signal ineach location of the HXR flare sources. Time axes are in minutes relatively the GOES flarestart time. Dashed lines creating a rectangle in each panel indicate HXR peak time for theflare on the both axes. 34 –Fig. 10.— Time of the magnetic transient as a function of timing of the HXR signal in eachlocation of the HXR flare sources. Time axes are in minutes relatively the GOES flare starttime. Dashed lines creating a rectangle in each panel indicate HXR peak time for the flareon the both axes. 35 –Fig. 11.— Total unsigned flux as a function of time computed over a rectangular includingthe whole flared active region cropped to the area of the strongest field changes shown inFigures 2 and 3 for the five flares observed by GONG. The temporal evolution profile of thetotal flux in Br, Bh components of the vector field and average tilt angle over the regioncropped to the area of the strongest field changes (2 degrees in latitude by 1 degree inlongitude rectangular region around center of the region shown in Figure 4) for the flareobserved by HMI is shown in the bottom left panel. The total HXR flux for each of theflares is over plotted in red. 36 –Fig. 12.— GONG magnetograms of the five flaring active regions with the locations of thestrongest magnetic field changes represented by 50 and 100 G contours of field change. Un-signed field decreases are light-blue, increases are magenta, and the PIL is the purple curves.The temporal evolution of the HXR centroids is shown with plus symbols, which increasingsize represents progressing times during flares and different color denotes different sources.The contoured areas showing a stepwise change during flare are labeled with numbers. 37 –
REFERENCES