JournalofGeophysicalResearch: SpacePhysics

Transcription

1 JournalofGeophysicalResearch: SpacePhysics RESEARCH ARTICLE Key Points: We classify the SPEs into four groups We studied temporal properties of 42 SPEs Flare-associated events start to increase the flux during flare increasing Correspondence to: R.-S. Kim, Citation: Kim, R.-S., K.-S. Cho, J. Lee, S.-C. Bong, and Y.-D. Park (204), A refined classification of SPEs based on the multienergy channel observations, J. Geophys. Res. Space Physics, 9, doi:. Received 4 JUL 204 Accepted 2 NOV 204 Accepted article online 26 NOV 204 A refined classification of SPEs based on the multienergy channel observations R.-S. Kim, K.-S. Cho,J.Lee 2, S.-C. Bong, and Y.-D. Park Korea Astronomy and Space Science Institute, Daejeon, Korea, 2 Department of Astronomy and Space Science, Chungnam National University, Daejeon, Korea Abstract We have investigated characteristics of solar proton events (SPEs) and their association with other types of solar eruption using 42 SPEs observed with SOlar and Heliospheric Observatory/Energetic and Relativistic Nuclei and Electron detector from 997 to 202. A velocity dispersion analysis was performed to correctly estimate the onset times of proton flux increase at the solar vicinity. These SPE onset times were compared with those of associated flares, coronal mass ejections (CMEs), and interplanetary type II radio bursts. We found that (i) the proton flux of 3 SPEs (3%) increase during the flare X-ray intensity is increasing, and the rest 29 SPEs (69%) show onsets well coincident with the first appearance of CMEs in Large Angle and Spectrometric COronagraph field of view. (ii) All flare-associated SPEs show the flux enhancements starting from the lower energy, while the CME-associated SPEs show the flux enhancements starting from either the higher or the lower energies. In the other events, the flux enhancement occurs simultaneously at all energies within min. (iii) For the former, large flux enhancements occur in a short time, while the latter tend to show relatively weak and slow flux enhancements. Our classification uses two criteria, SPE onset timing relative to flares and energy-dependent flux enhancement, unlike the conventional classification of SPEs based on whether the flux time profile is impulsive or gradual. Nevertheless our classification scheme refines the distinction between the flare-associated SPEs and the CME-associated SPEs in terms of the onset timing. Additional information on the proton acceleration as implied by the energy-dependent patterns of flux enhancement is briefly discussed. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.. Introduction Solar proton events (SPEs) are generated when protons are accelerated to high energies either close to the Sun or in the interplanetary (IP) space. Understanding where and how solar protons are accelerated to the high energies is one of the hottest topics of space weather, since SPE is a significant radiation hazard to spacecraft and especially astronauts. Main drivers of particle acceleration are magnetic reconnections in solar flares and fast mode MHD shocks formed by coronal mass ejections (CMEs). Solar flares as a driver of SPEs were a popular idea in the past, but there have been increasing number of observational evidence supporting CME-driven shocks as the driver of SPEs [e.g., Kahler et al., 978; Cliver, 982]. Several particle acceleration mechanisms have been presented for solar flares [Ohsawa and Sakai, 988; Mori et al., 998; Bombardieri et al., 2008] and CMEs [Klein and Trottet, 200;Roussev et al., 2004;Cane et al., 986], and modeling of SEP events have been pursued by a number of groups [Ng et al., 2003; Lee, 2005; Luhmann et al., 2007, 20]. The Particle Acceleration and Transport in the Heliosphere model, which is a time-dependent model of shock wave propagation in the solar wind [Zank et al., 2007; Li et al., 2003, 2005; Verkhoglyadova et al., 2009, 20, 202], has been applied to individual SEPs. Recently, Reames [2009] and Gopalswamy et al. [202] examined the onset times and release heights of energetic particles using the ground-level enhancement events. However, physical understanding of SPEs origin and mechanism still remains incomplete. We believe that timing analysis of SPEs should be a key step in correctly identifying source of proton acceleration. To perform such a timing analysis, we have to consider the travel times of the energetic protons from the Sun to the Earth along the Parker spiral [Krucker and Lin, 2000]. Note that the energetic protons are traveling at certain speeds corresponding to their energies until detected in situ whereas other solar eruptive phenomena such as flares, CMEs, and type II radio bursts are the electromagnetic radiations and their detection is only delayed by the speed of light. In several studies such time delays have been KIM ET AL The Authors.

2 Journal of Geophysical Research: Space Physics 08:30 UT 07:2 UT (c) (d) Figure. Observations of a SPE and related eruptive solar phenomena occurred on 4 November 997. Enhancement of proton flux (E > MeV) detected by GOES at 07:2 UT, X-class flare which started to increase the X-ray flux at 05:52 UT observed by 95 Å SOHO/EIT, (c) white-light running difference image of the halo CME which appeared on LASCO/C2 field of view at 06: UT, and (d) IP type II radio burst observed by Wind/WAVES at 06:00 UT. compensated by subtracting the typical travel times from SPEs observing time. For example, Kocharov et al. [2007] used 39 min for the travel time of relatively lower energetic protons (7 22 MeV) and min for the higher energetic protons ( 40 MeV). For this kind of study, the velocity dispersion analysis (VDA) is generally used to calculate travel times for energetic particles of respective energy levels [Krucker and Lin, 2000; Kocharov et al., 2007]. To apply VDA, we should consider the location (longitude and height) of acceleration site and path length of accelerated particles. However, it is not so easy to obtain those two parameters together; therefore, we assume that all particles with different energies are traveling the same path length, to show if any systematical trends which vary independently to the path length based on statistical approach. In addition to the timing analysis, we consider temporal variation patterns of flux enhancement of SPEs as they can give a clue to the mechanisms for proton acceleration and triggering of SPEs, as well as their association with flares and CMEs. Using multienergy channel observations of energetic protons, we can identify at which energy the proton acceleration occurs first. There are theoretical models for proton acceleration during flares that indicate rapid acceleration of solar protons to high energies can occur within a timescale as short as observed at hard X-ray light curves [Miller et al., 990, 996]. On the other hand, particle accelerations by the CME-driven shock are considered to be gradual under the diffusive shock acceleration mechanism, in which energetic particles are dominant in the initial stage, and low-energy particles arise later as the shock propagates further into the solar wind [Zank et al., 2000]. In this paper we study the onset times of SPEs and patterns of proton flux enhancement to identify the origin of SPEs and derive information on the proton acceleration mechanisms. The paper is organized as follows. Data and methodology of our study are given in section 2. We examine the characteristics of SPEs according to classification based on the onset timing of SPEs and related solar eruptive phenomena, and the temporal pattern of flux enhancement in section 3. A brief summary and discussion are presented in section Data and Methodology 2.. SPEs and Related Solar Eruptive Phenomena A SPE is defined as the event where the number of incoming energetic (> MeV) protons exceeds particle flux unit (pfu) ( pfu = proton per cm2 sr s) at geosynchronous satellite altitudes. More than a hundred events have been observed after the launch of the SOlar and Heliospheric Observatory (SOHO: 997 ). We have selected 42 SPEs, which have clear information of associated flares, CMEs, and IP type II radio bursts during , from the list published by the National Oceanic and Atmospheric Administration/Space Weather Prediction Center ( Figure shows the first SPE and related solar eruptive phenomena in the SOHO era. Figure a shows the flux of > MeV protons (red) detected by Geostationary Operational Environmental Satellite (GOES) starts to increase at 07:2 UT on 4 November 997, and Figure b shows the accompanying X-class flare KIM ET AL The Authors. 2

3 Table. Timing of SPEs and Related Solar Eruptive Phenomena SPE (GOES) Flare CME Type II SPE (ERNE) # Date Start Peak Onset Peak Location Appearance Onset Onset Low High Group :30 :20 05:52 05:58 S4W33 06: 06:00 06: 06:5 06: D :05 02:55 a :49 :55 S8W63 3:00 2:20 2:8 2:4 2:9 B :00 2:05 a 09:38 :2 S43W90 :07 :25 :37 :43 :0 C :20 6:50 3:3 3:42 S5W5 4:06 4:25 3:32 3:32 3:44 A :45 09:45 07:58 08:09 SW65 08:29 08:25 07:45 07:45 08:2 A :55 09:30 a 5:2 5:4 N6W66 6:32 5:45 5:29 5:32 5:44 A :35 09:40 a 4:58 b 5:25 b N20E8 5:54 b 5:20 b 7:28 b 7:5 b 7:28 b B :05 20:45 6:40 7:02 N22W38 7:08 7:5 6:56 6:56 7:4 A :45 2:30 a :03 :24 N22W07 :54 :30 :07 :07 :23 A :20 4:05 :7 :34 N4W56 2:30 :45 :42 2: :47 B :55 03:40 a :3 2:3 S7W09 3:3 2:00 2:24 2:40 2:24 B :25 8:40 06:40 07:28 N04W90 07:27 07: 07:07 07:07 07:9 A :50 5:55 a 22:42 23:28 N05W77 23:06 23:20 22:43 22:43 23:08 A :20 20:30 c 04:55 05:02 N20W05 05:30 05: 05:3 06:02 05:3 B :25 06:55 a 5:40 6:00 S04W59 5:54 5:45 6:22 6:55 6:22 B :35 06: a 09:57 :5 N4W2 :26 :2 :36 :40 :20 C :40 07:45 a 2:32 2:5 N8W82 22:00 22:05 22:08 22:08 22:08 D : 9:20 3:9 3:50 S20W85 4:30 4:05 3:2 3:2 3:47 A :5 :45 02: 02:4 S20W90 02:30 02:55 02:30 02:34 02:34 D :35 4:55 :04 :28 S2W49 :54 :50 :5 :5 2:08 C :5 22:35 a 09:32 :38 S6E23 :30 :45 :44 :49 :06 C :25 22:35 6:3 6:30 N5W29 6:50 6:45 6:59 7:7 7:00 B :05 02:5 c 6:03 6:20 N06W8 6:35 6:30 6:8 6:8 6:34 A :05 :5 04:32 05:40 N08W54 05:30 05:20 05:9 05:9 05:32 A :20 3:20 a :2 :4 S20W87 :06 :30 :55 :55 3:09 C :30 5:40 07:46 08:24 S4W34 08:26 08:30 08:46 08:46 :23 C :25 23:20 00:43 0:5 S4W84 0:27 0:30 0:07 0:07 0:27 A :00 6:20 0:47 02:2 N09W54 02:06 02:20 0:48 0:48 02:03 A :40 08:35 00:49 0:2 S08W90 0:27 0:45 0: 0: 0:8 D :20 05:40 a 3:08 3:23 S2W29 3:3 3:20 4: 4:20 4:8 D :40 06:45 02:3 02:24 S07W65 02:30 03:00 02:34 02:44 02:39 D :25 22:35 7:2 8:9 N02W38 7:54 7:45 7:20 7:20 7:38 A :5 06:5 a 09:5 : S6E08 :54 : :2 :23 :35 C :05 7:30 09:40 09:48 S9W89 :26 :00 :8 :8 :32 C :35 8:45 03:54 04:9 S4W47 04:30 04:20 04:38 04:44 04:38 D :55 22:50 a 4:9 5:4 N08W33 5:30 5:00 5:24 5:38 5:28 B : 7:50 a 22:25 b 23:02 b N5W05 23:06 b 23:00 b 23:00 b 23:45 b 23:00 b B :25 02:40 a 6:3 b 6:57 b N2E 7:22 b 7:00 b 7: b 7:3 b 7:40 b C :45 03:45 a 4:0 b 4:49 b NW80 4:30 b 4:5 b 4:56 b 5:07 b 4:56 b B :40 :45 a 6:46 7:27 S2W60 7:30 7:5 7:36 8:5 7:42 B : 09:25 02:4 02:40 S05W23 02:54 02:45 02:37 03:05 02:4 B :05 08:00 9:43 b 20:2 b N24W59 20:00 b 20:00 b 20:37 b 2:05 b 20:39 b B a Time corresponds to next day. b Time corresponds to previous day. c Time corresponds to next next day. KIM ET AL The Authors. 3

4 Spiral Structure of IMF SUN Travel Time (min.) Travel Time 0 0 Energy (MeV) Figure 2. Path of solar energetic protons along the spiral structure of the IMF and proton travel time as a function of energy from the Sun to the Earth. (X2.) observed at 95 Å by the Extreme ultraviolet Imaging Telescope (EIT) on board SOHO at 05:52 UT. A halo CME with a linear speed of 785 km s appears in the Large Angle and Spectrometric COronagraph (LASCO)/C2 field of view at 06: UT (Figure c), and IP type II radio burst is detected by WAVES instrument on board Wind satellite at 06:00 UT (Figure d). Table summarizes the timing of 42 selected SPEs and related phenomena. The third and fourth columns of the table show the start times of SPEs and peak times of the proton flux in the GOES observation near the Earth. Note that the SPE start time does not always coincide with the onset time, because the SPE start time refers to the moment when the proton flux exceeds the criterion for SPE ( pfu) in the course of continuous enhancement. For instance, see Figure a where the proton flux of MeV (red) starts to increase at 07:2 UT but can be regarded as a SPE only after until 08:30 UT, when the flux exceed the pfu as marked by arrows. From the fifth to seventh columns are the times of flare onset and peak observed by GOES X-ray flux detector, and its location on the solar surface. Although the locations of flaring site are scattered in between E23 and west limb, the mean value is W45. More detailed examination of the flaring site will be performed in our further study. We also list the first appearance time of CMEs in LASCO field of view, and the onset times of type II radio bursts detected by Wind/WAVE instrument from Coordinated Data Analysis Workshops data center ( in the eighth and ninth columns SPE s Travel Time and its Onset in the Solar Vicinity To compare the SPE onsets with those of other related phenomena, it is crucial to compensate proton travel time from the Sun corresponding to their energies. We perform the VDA of the data obtained from the High Energy Detector (HED) of Energetic and Relativistic Nuclei and Electron (ERNE) on board the SOHO. The ERNE/HED consists of energy channels from 3 to 30 MeV and simultaneously detects the enhancements of proton flux at different energy levels [Torsti et al., 2004]. Therefore, onset time measured using ERNE/HED data may vary with the proton energy. After measuring the onset times in the multichannel in situ observations, we remove the proton travel time to retrace the onset of proton acceleration in the solar vicinity. Krucker and Lin [2000] showed that the path length of scatter-free SPE is about.2 AU, which is the length of an Archimedean spiral corresponding to a solar wind speed of 400 km s [Cane and Lario, 2006] as shown in Figure 2a. We assume that protons at all energy channels travel the same path length (L) of.2 AU along the spiral structure of the interplanetary magnetic field (IMF). Then the travel time (t tr ) for a proton of energy E can be calculated as, t tr (E) =L v(e) () where v(e) is the proton velocity of corresponding to each energy channel. Figure 2b shows the travel times, t tr (E) for the protons corresponding to the energy channels of ERNE/HED indicated by red diamonds. For the highest energy band of 0 30 MeV, the t tr is about 22 min, and for the lowest band of 3 6 MeV, it is about 57 min. The energy-dependent onset time, T os (E) of a SPE at its origin in the solar vicinity is then expressed as T os (E) =T ob (E) t tr (E)+t c (2) where T ob is the observed time near the Earth and t c is the light s travel time from the Sun to the Earth corresponding to 8.3 min. We will compare T os with the onset times of flares, CMEs, and IP type II radio burst. Figure 3 shows the SPE observed on 4 November 997, which is the same event with Figure. Figure 3a shows in situ SOHO/ERNE observation of the energy channels from 3 to 30 MeV, and Figure 3b shows KIM ET AL The Authors. 4

5 In situ measurement 3 MeV Travel time corrected MeV :00 07: 08:24 09: :47 05:59 07: 08:23 Figure 3. In situ SOHO/ERNE observation of the SPE on 4 November 997 and the corrected proton flux time profiles where the proton travel time is removed. Colors represent the energy channels of SOHO/ERNE. the corrected proton flux according to different travel times of corresponding energy channels. As shown in Figure 3a, the enhancement of proton flux at different energy bands do not always coincide with each other. For this event, the flux of higher energetic protons of 0 30 MeV is detected first at 06:24 UT, 4 min earlier than the flux enhancement of lower energetic protons (07:05 UT at 3 6 MeV). After the travel time correction, the onset times of proton accelerations in the solar vicinity, T os at the highest and lowest energy channels are 06: UT and 06:5 UT, respectively. They only have 5 min difference. Among those onsets of multichannels, we set the representative onset time, T os for each 00 Type II onset Proton Onset CME Height GOES X-ray 00 Energy (MeV) /05/02 3: Hours after flare onset 2000/07/22 : Hours after flare onset 5 00 (c) (d) 00 Energy (MeV) /04/7 07:46 200/04/02 2: Hours after flare onset Hours after flare onset 5 Figure 4. Onset times of SPEs and related eruptive phenomena including flares, CMEs, and IP type II radio bursts. Proton acceleration is preceding the flare peak, and the acceleration started from the lower energy channel. (b d) The proton accelerations occur after the flare peaks but well coincident with CME appearances. The accelerations for those events started from the higher energy channel (Figure 4b), or from the lower energy channel (Figure 4c), or simultaneously in all energy channels (Figure 4d). KIM ET AL The Authors. 5

6 25 25 A B C D Event No. 5 Event No SPE-Flare Onset (hr) SPE-CME Appearance (hr) Event No (c) Mean time difference (hr) (d) SPE-Flare onset SPE-CME onset SPE-Type II onset SPE-Type II Onset (hr) -.5 A B C D Group Figure 5. Histograms of time differences between the onsets of SPE and those of flares, appearance of CMEs on LASCO/C2 field of view, and (c) type II radio bursts. (d) The average time differences per group. The positive values in the figure mean that the SPE onset is following the onsets of other phenomena, while the negative values mean that SPE is preceding others. The horizontal dotted line in Figure 5d corresponds to the onset time of SPEs. event as the first onset, regardless of its energy as listed in the tenth column of Table. We also summarize the onset times at the lowest energy channel (mainly 3 6 MeV) and the highest energy channel (mainly 0 30 MeV) in the eleventh and twelfth columns of the table, respectively. 3. Results 3.. SPE s Onset Times and Flux Enhancement Patterns We first compare the onset times with flare onset, flare maximum, CME first appearance, and IP type II radio burst onset. If we take the onset time as the only single criterion, then we can classify the 42 SPEs into the following two groups: (i) for 3 events (3%), the proton acceleration starts during the flare intensity is increasing (namely the acceleration starts prior to all of the flare peak, CME first appearance, and type II radio burst onset). This means that the proton acceleration is related to flare as shown in Figure 4a. (ii) For the other 29 events (69%), their onset times well agree with the CME s first appearances time and/or the onset of the type II radio burst. If we consider the temporal pattern of flux enhancement as well as the onset times for the criteria, we find four different groups as shown in Figure 4, for example. In the figure the red closed circles denote the retraced SPEs onset times near the Sun for all energy channels, the black closed circles indicate the CME observation times and heights, the black solid lines are scaled GOES X-ray flux, and the blue vertical dotted lines indicate the type II radio burst onset times. All flare-related events (3, 3%) show a single pattern that the acceleration is started from only the lower energy channel as shown in Figure 4a. KIM ET AL The Authors. 6

7 Figure 6. Onset time differences between the highest- and the lowest energy channels for all events. The positive value means the proton acceleration starting from the lowest energy channel, and the negative value means the acceleration starting from the higher energy channel. In contrast, the CME-related events show three different acceleration patterns (which means): the proton accelerations started from either lower or higher energy channels in addition to the case of simultaneous commencement in all energy channels. In specific, among the 29 CME-related SPEs, 3 events (3%) show the proton acceleration starting from higher energy channels (Figure 4b). Nine events (2%) show the acceleration starting from lower energy channels (Figure 4c), and seven events (7%) show the simultaneous accelerations in all energy channels within min (Figure 4d). We call these four groups of the SPEs A D, respectively, and list them in the last column of Table Comparison of SPEs and Other Phenomena To further find other characteristics of each group of SPEs, we then examine how they are associated with the related phenomena based on the closeness of their onset times, as shown in Figure 5. Note that the positive values in Figure 5 mean that the SPE onset followed the onsets of other phenomena, while the negative values mean that SPE onset preceded others. As shown in Figures 5a 5c, the SPEs in the group A (blue) tend to precede other phenomena, such as flare (Figure 5a), CME (Figure 5b), and type II radio burst (Figure 5c), while the SPEs in the group C (orange) tend to follow others. This tendency is more clearly seen in terms of the mean time difference plotted in Figure 5d. The horizontal dotted line in Figure 5d indicates the onset time of SPEs (mean time difference = 0 h). Only the SPEs in group A have onset times closer to those of flares than the CME s first appearance time. For the groups B D, SPEs onset times are closer to the CME appearance times than the flare onsets. The SPEs in group B (yellow) show relatively long time difference between the CME and type II radio burst, and even the type II radio bursts are preceding CME s appearance, which implies the source location of type II radio burst in the low corona. The SPEs in group C have very delayed onsets with respect to CMEs and type II radio bursts. The SPEs in group D (red) show relatively short time differences of each phenomenon Evolution of Proton Energy Spectrum We also measure the time differences between onsets in the highest and lowest energy channels, (T H T L ) for each event. This tells how fast the accelerations are processing in respective energy channels. For instance, the time differences for group A are positive, which means the proton acceleration started from the lowest energy, and the acceleration took place in the highest energy within 30 min (mean: 7 min) as shown in Figure 6. Based on the mean values for each group, we can see that for the SPEs in group B, the proton acceleration started earlier in the higher energy and later in the lower energy; the time difference can be about 26 min. SPEs in group C have the acceleration pattern similar to those in group A, but the length of time difference (35 min) in group C is longer than that of group A. Group D shows very short time differences with the mean value being less than 2 min (.8 min). To investigate the acceleration processes of SPEs in more detail, we examine the temporal flux changes in the energy spectrum. An example is shown in Figure 7a, which is the same event shown in Figure 3. As illustrated by vertical colored lines, we plot the energy spectra from 3 to 30 MeV at time steps, which are start from min before and end 40 min after the first flux increase with 5 min intervals. Figure 7b shows the time evolution of the energy spectrum, that is, time delays as a function of energy. After the SPEs onset (yellow), the flux of all energy channels are increased. Figure 8 shows the temporal variations of the average proton flux spectrum for the four groups. In order to comprehend the typical behavior of the spectral variation per group, we averaged the proton fluxes from all events in each group into the time bins as exemplified by Figure 7. The same color table used in KIM ET AL The Authors. 7

8 Travel time corrected Group: D :47 05:59 07: 08: Energy (MeV) Figure 7. Construction of energy spectrum of the proton flux. Time profiles of the same SPE as shown in Figure 3. The vertical lines denote the time bins, each with 5 min interval, for which the energy spectra are constructed. Energy spectra at the time intervals constructed from the data shown in Figure 7a. The rainbow color table is used to denote the times of individual spectra. Figure 7 for representing the time bins is used again in Figure 8 so that the color of a spectrum represents the time of the spectrum increasing from red to dark blue. The black solid line in all panels is the reference spectrum that is obtained by averaging the proton flux spectra from the 42 SPEs in all time bins. This figure reveals how the spectral variation pattern differs from a group to another. The SPEs in group (c) (d) Figure 8. Temporal evolution of the proton flux spectrum during proton acceleration. The colors indicate the change of the time, as same as denoted in Figure 7. The black solid line in each panel illustrates the mean values for all events for all time steps. KIM ET AL The Authors. 8

9 Table 2. Characteristics of Four SPE Groups Acceleration Energy Spectrum Group Event # Association Direction T H T L (Average) Dominant Energy Band Enhancement A 3 (3%) Flare Low High 7 min Low and Middle 4 B 3 (3%) CME High Low 26 min High > 2 C 9 (2%) CME Low High 35 min Low 3 D 7 (7%) CME Simultaneous.8 min Middle and High 2 A show the flux enhancement by the largest amount ( 4 ) among others, especially in the low and middle energy channels. Those in group B show the flux enhancement by the smallest amount in overall. Especially they exhibit almost no enhancement with time in the lowest energy channels. In contrast, the SPEs in group C show relatively large flux enhancements at low energy channels and least enhancements at higher energy channels. Finally, the SPEs in group D behave similar to group A, but with modest flux enhancements (increase by a factor of 2 ) in all channels. These characteristics of the group-averaged spectra are summarized in Table Summary and Discussions We studied temporal properties of 42 SPEs observed with the ERNE/HED on board the SOHO from 997 to 202. The main task was to estimate the onset times of the SPEs and compared them with the onsets of other solar eruptions including flares, CMEs, and IP type II radio bursts. The comparison of the onset times of the SPEs was made after the proton travel time correction through the VDA in the multienergy channel data. As a result, we could classify the SPEs into four groups according to their onset timing relative to solar eruptions, and temporal pattern of proton flux enhancement. In specific we summarize the result as follows.. In terms of the onset timing, the SPEs are classified into two groups: (i) 3 flare-associated SPEs (3%), which start before the flare peak time and (ii) 29 CME-associated SPEs (69%) whose onset times are coincident with the times of the first appearances of CMEs in LASCO field of view and the start times of IP type II radio bursts. 2. For all flare-associated SPEs, the proton flux enhancement starts from the lower energy channels and occurs in the higher energy channels in relatively a short duration of 7 min (group A). 3. For the CME-associated SPEs, we could further classify them into three subgroups depending on their temporal pattern of proton flux enhancement: starting earlier at the higher energy channels (group B: 3 events, 3%), the lower energy channels (group C: nine events, 2%), and all energy channels simultaneously (group D: seven events, 7%). For the groups B and C, the energy-dependent onset proceeds over relatively a longer period (26 min and 35 min, respectively) than the flare-associated events. On the other hand, group D shows immediate flux enhancement in all energy channels within 2 min. 4. This classification scheme is more obvious when we look at the temporal flux changes in the multichannel energy spectrum. The flare-associated SPEs tend to show a large flux enhancement by a factor of 4 in a short time, while the CME-associated SPEs tend to show relatively a weak flux enhancement by a factor of 2. Our classification of SPEs differs from the conventional one in that we utilize, as criteria, the energy-dependent onset timing and flux enhancement patterns. Nonetheless, it recovers the conventional notion of impulsive and gradual events: the rapid flux increase in a broad energy range seen for the flare-associated SPEs (group A) resembles the property of impulsive SPEs, and the relatively weak and slow flux enhancements in the CME-associated SPEs (groups B D) would correspond to the gradual SPEs, respectively. The energy-dependent pattern of flux enhancement in groups in B and C clearly differs from that of group A, indicating that the proton acceleration mechanisms operating in CMEs differ from that in solar flares. For group B, the acceleration is more efficient at high energies, and the diffusive shock acceleration mechanism [Zank et al., 2000] may be more appropriate to explain their flux enhancement pattern. The spectral variation in group C would suggest a different acceleration mechanism that is more effective at low energies with a limiting high energy such as the acceleration by finite electric field. KIM ET AL The Authors. 9

10 The apparent similarity in the spectral variation between groups A and D is rather unexpected. On the one hand, this may imply a common acceleration mechanism operating in both flares and CMEs but with very different efficiencies or interplanetary medium conditions. For instance, in the proton acceleration model developed for solar flares [Miller et al., 990], the acceleration efficiency is related to the energy density in either magnetosonic or Alfven waves. In this case, the difference between group A and D may thus be due to the lower wave energy density in CMEs than in flares. On the other hand, the largely different acceleration efficiency could mean entirely different mechanisms for these two classes. Quantitative assessment of the acceleration mechanisms adequate for these spectral behaviors would certainly deserve a further effort. Relating to the interplanetary medium conditions, a recent model that has been discussed heavily is the twin-cme model [e.g., Ding et al., 203]. They presented the observation of the correlation between the presence of preceding CMEs and large SPE events. On the other side, Kahler and Vourlidas [204] suggested that the large SPE with preceding CMEs may not be due primarily to CME interactions, such as the twin-cme scenario, but are explained by a general increase of both background seed particles and more frequent CMEs during times of higher solar activity. A factor, we did not examine in this study, yet important is the location of the flare (solar eruption center) with respect to the foot point of the IP magnetic field line connected to the spacecraft. For a detector onboard the near-earth spacecraft, a nearby IP magnetic field line is connected to the heliolongitude of west on the Sun. Therefore, the particles from the flare or CME core located there will first arrive at the spacecraft without extra delay, whereas particles from the eastern part of solar disc or from beyond the west limb may arrive with extra delay of min. In addition, the path length fixed for all events as.2 AU in this study may be refined although it is commonly accepted for the particles propagation [Cane and Lario, 2006]. The onset times measured without considering the location of solar eruption and the path length may be subject to a certain degree of uncertainty. Another factor that may help improving this study result is the formation height of CME-driven shock and physical environment of proton acceleration for each group. The current study may improve if the above information are considered in further studies. In spite of the uncertainties of the onset times, our statistical results show clear differences between the groups. We hope that this more detailed classification of SPEs will help identifying dominant physical mechanisms for proton acceleration in future. Acknowledgments The data for this paper are available at Space Research Laboratory of Turku University ( data/). We thank SOHO/ERNE team for multichannel energetic proton flux data available on the internet. We especially thank L. Kocharov and the referees for valuable suggestions and comments that led to significant improvement of the paper. J.L. was supported by the Brainpool program 204 of KOFST. This work was supported by the Construction of Korean Space Weather Center as the project of KASI, the KASI Basic Research Fund, and Research Fellowship for Young Scientists of KRCF. Yuming Wang thanks the reviewers for their assistance in evaluating this paper. References Bombardieri, D. J., M. L. Duldig, J. E. Himble, and K. J. Michael (2008), An improved model for relativistic solar proton acceleration applied to the 2005 January 20 and earlier events, Astrophys. J., 682, Cane, H. V., and D. Lario (2006), An introduction to CMEs and energetic particles, Space Sci. Rev., 23, Cane, H. V., R. E. McGuire, and T. T. von Rosenvinge (986), Two classes of solar energetic particle events associated with impulsive and long-duration soft X-ray flares, Astrophys. J., 30, Cliver, E. W. (982), Prompt injection of relativistic protons from the September, 97 solar flare, Sol. Phys., 75, Ding, L., Y. Jiang, L. Zhao, and G. Li (203), The Twin-CME scenario and large solar energetic particle events in solar cycle 23, Astrophys. J., 763, 30, doi:.88/ x/763//30. Gopalswamy, N., H. Xie, S. Yashiro, S. Akiyama, P. Mäkelä, and I. G. Usoskin (202), Properties of ground level enhancement events and the associated solar eruptions during solar cycle 23, Space Sci. Rev., 7, Kahler, S. W., and A. Vourlidas (204), Do interacting coronal mass ejections play a role in solar energetic particle events?, Astrophys. J., 784, 47, doi:.88/ x/784//47. Kahler, S. W., E. Hildner, and M. A. I. van Hollebeke (978), Prompt solar proton events and coronal mass ejections, Sol. Phys., 57, Klein, K.-L, and G. Trottet (200), The origin of solar energetic particle events: Coronal acceleration versus shock wave acceleratios, Space Sci. Rev., 95, Kocharov, L., O. Saloniemi, J. Torsti, E. Riihonen, J. Lehti, K.-L. Klein, L. Didkovsky, D. L. Judge, A. R. Jones, and R. Pyle (2007), High-energy protons associated with liftoff of a coronal mass ejection, Astrophys. J., 659, Krucker, S., and R. P. Lin (2000), Two classes of solar proton events derived from onset time analysis, Astrophys. J., 542, L6 L64. Lee, M. A. (2005), Coupled hydromagnetic waves excitation and ion acceleration at an evolving coronal/interplanetary shock, Astrophys. J. Suppl. Ser., 58, Li, G., G. P. Zank, and W. K. M. Rice (2003), Energetic particle acceleration and transport at coronal mass ejection-driven shocks, J. Geophys. Res., 8(A2), 82, doi:.29/2002ja Li, G., G. P. Zank, and W. K. M. Rice (2005), Acceleration and transport of heavy ions at coronal mass ejection-driven shocks, J. Geophys. Res.,, A064, doi:.29/2004ja0600. Luhmann, J. G., S. A. Ledvina, D. Krauss-Varban, D. Odstrcil, and P. Riley (2007), A heliospheric simulation-based approach to SEP source and transport modeling, Adv. Space Res., 40, Luhmann, J. G., S. A. Ledvina, D. Odstrcil, M. J. Owens, X.-P. Zhao, Y. Liu, and P. Riley (20), Cone model-based SEP event calculations for applications to multipoint observations, Adv. Space Res., 46, 2. Miller, J. A., N. Guessoum, and R. Ramaty (990), Stochastic Fermi acceleration in solar flares, Astrophys. J., 36, KIM ET AL The Authors.

11 Miller, J. A., T. N. LaRosa, and R. L. Moore (996), Stochastic electron acceleration by cascading fast mode waves in impulsive solar flares, Astrophys. J., 46, Mori, K, J. Sakai, and J. Zhao (998), Proton acceleration near an X-type magnetic reconnection region, Astrophys. J., 494, Ng, C. K., D. V. Reames, and A. J. Tylka (2003), Modeling shock-accelerated solar energetic particles coupled to interplanetary Alfven waves, Astrophys. J., 59, Ohsawa, Y., and J. Sakai (988), Prompt simultaneous acceleration of proton and electrons to relativistic energies by shock waves in solar flares, Sol. Phys., 332, Reames, D. V. (2009), Solar energetic-particle release times in historic ground-level events, Astrophys. J., 706, Roussev, I. I., I. V. Sokolov, T. G. Forbes, T. I. Gombosi, M. A. Lee, and J. Sakai (2004), A numerical model of a coronal mass ejection: Shock development with implications for the acceleration of GeV protons, Astrophys. J., 605, L73 L76, doi:.86/ Torsti, J., E. Riihonen, and L. Kocharov (2004), The 998 May 2-3 magnetic cloud: An interplanetary highway for solar energetic particles observed with SOHO/ERNE, Astrophys. J., 600, L83, doi:.86/ Verkhoglyadova, O. P., G. Li, G. P. Zank, Q. Hu, and R. A. Mewaldt (2009), Using the PATH code for modeling gradual SEP events in the inner heliosphere, Astrophys. J., 693, Verkhoglyadova, O. P., G. Li, G. P. Zank, Q. Hu, C. M. S. Cohen, R. A. Mewaldt, G. M. Mason, D. K. Haggerty, T. T. von Rosenvinge, and M. D. Looper (20), Understanding large SEP events with the PATH code: Modeling of the 3 December 2006 SEP event, J. Geophys. Res., 5, A23, doi:.29/20ja0565. Verkhoglyadova, O. P., G. Li, G. P. Zank, Q. Hu, and R. A. Mewaldt (202), Radial dependence of peak proton and iron fluxes in solar energetic particle events: Application of the PATH code, Astrophys. J., 757, 75, doi:.88/ x/757//75. Zank, G. P., W. K. M. Rice, and C. C. Wu (2000), Particle acceleration and coronal mass ejection driven shocks: A theoretical model, J. Geophys. Res., 5(A), 25,079 25,095, doi:.29/999ja Zank, G. P., G. Li, and O. Verkhoglyadova (2007), Particle acceleration at interplanetary shocks, Space Sci. Rev., 30, KIM ET AL The Authors.