Major solar energetic particle events of solar cycles 22 and 23: Intensities above the streaming limit

Transcription

1 Click Here for Full Article Major solar energetic particle events of solar cycles 22 and 23: Intensities above the streaming limit D. Lario, 1 A. Aran, 2 and R. B. Decker 1 SPACE WEATHER, VOL. 6,, doi: /2008sw000403, 2008 Received 13 April 2008; revised 9 July 2008; accepted 20 August 2008; published 3 December [1] Large solar energetic particle (SEP) events constitute a serious radiation hazard to astronauts and spacecraft systems. It is essential to determine the highest particle intensities reached in SEP events, especially at the energies that pose serious risks to human health and spacecraft performance. It has been argued that the highest particle intensities measured during large SEP events occur in association with the passage of shocks driven by coronal mass ejections known as the energetic storm particle (ESP) component. Furthermore, it has been argued that the intensities measured early in the SEP events (known as the prompt component) are bounded by a maximum intensity plateau that results from wave-particle interactions that restrict the free streaming of particles (also called the streaming limit ). We analyze proton intensities measured by the GOES spacecraft at the energy channels P5 ( MeV) and P7 ( MeV) during solar cycles 22 and 23 and examine whether the highest intensities were measured during the prompt or the ESP components of the SEP events. We find three (one) SEP events in which the highest proton intensities measured during the prompt component at the energy channel P5 (P7) exceeded by a factor of 4 or more the previously determined streaming limit. Arguments to explain intensities during the prompt components exceeding this limit invoke interplanetary conditions that inhibit the amplification of waves resonating with the streaming particles and/or the presence of interplanetary structures able to confine and/or mirror energetic particles. We analyze these possibilities for each one of these events. Citation: Lario, D., A. Aran, and R. B. Decker (2008), Major solar energetic particle events of solar cycles 22 and 23: Intensities above the streaming limit, Space Weather, 6,, doi: /2008sw Introduction [2] Energetic particle events generated by solar activity pose serious threats to both spacecraft components and operations as well as to astronauts in either extravehicular activity or behind a protective shielding [Lanzerotti, 2004]. Solar energetic particle (SEP) events lead to high radiation doses in short time intervals, having important effects on technology used in interplanetary space. There is a general agreement on the necessity to characterize the evolution of such events and to specify the highest particle intensities reached during major SEP events, especially at the energies that pose serious risks to human health (i.e., protons above 30 MeV) [Feynman and Gabriel, 2000]. [3] Reames [1999] proposed that the highest proton intensities in large SEP events are reached in association 1 Applied Physics Laboratory, Johns Hopkins University, Laurel, Maryland, USA. 2 Departament d Astronomia i Meteorologia, Universitat de Barcelona, Barcelona, Spain. with the arrival of interplanetary shocks driven by the coronal mass ejections (CMEs) that generate the particle events. The particle intensity enhancements associated with the passage of CME-driven shocks have been historically termed energetic storm particle (ESP) events, because of the close association with the geomagnetic storms observed simultaneously when the shocks reach the Earth s magnetosphere [Bryant et al., 1962]. The ESP events are due to both the fraction of shock-accelerated particles trapped by wave-particle interactions near the propagating shock and the fraction of particles locally accelerated by the shock when it arrives at the spacecraft [e.g., Lario and Decker, 2002, and references therein]. [4] Reames and Ng [1998] suggested that energetic particle intensities measured early in a SEP event are bounded by a maximum intensity plateau known as the streaming limit. The physical process of wave generation by particles streaming outward from an intense source near the Sun was suggested as the mechanism providing a selfregulation of the particle intensity [Ng and Reames, 1994]. Copyright 2008 by the American Geophysical Union 1of25

2 Energetic particles propagating along interplanetary magnetic field (IMF) lines reach a maximum intensity plateau because the scattering processes produced by selfgenerated waves restrict their streaming. Therefore, according to this scenario, particle intensities measured early in a large SEP event (known as the prompt component of the SEP event) are bounded by a certain upper limit. This streaming limit intensity can only be exceeded when the source of particles (i.e., the propagating CMEdriven shock) reaches the spacecraft. In the case that the source intensity rises above the level imposed by the streaming limit, additional particles are diffusively trapped near the expanding source. Thus, in this scenario, added acceleration of particles at the traveling shock only serves to increase the intensity at the ESP peak, but not that measured on the early plateau. The consequence is that this proposed streaming limit applies only to the intensity of particles streaming from a distant source and not to those accelerated locally or convected out from a source. [5] By analyzing the largest particle events observed by the National Oceanic Atmospheric Administration (NOAA) Geostationary Operational Environmental Satellites (GOES) during solar cycle 22, Reames and Ng [1998] determined, observationally, the streaming limit intensity at several energy levels. The questions addressed in this paper are whether this limit intensity has been exceeded during the last two solar cycles and, if so, if it is possible to discern the mechanisms that lead to the exceeding of this limit. [6] In the context of the scenario proposed by Reames and Ng [1998], the following two factors may account for the possible exceeding of this empirically determined limit during the prompt component of major SEP events: [7] 1. The level of wave amplification may have not been high enough to limit the streaming of energetic particles. This level is determined by the conditions in the solar wind plasma where particles propagate. Reames et al. [2001a] suggested that plasma regions with a high plasma b p parameter (computed as the ratio between the solar wind thermal proton energy density to the magnetic energy density) favor the rapid growth of waves because the intensity of ambient MHD waves in this region is already large. By contrast, low-b p plasma regions, where the ambient wave intensity is much lower, favor the free streaming of particles [Reames et al., 2001a]. Thus, propagation of energetic particles in a low-b p region may not generate sufficiently intense waves to restrict the particle streaming, and hence increase the probability of observing particle intensities exceeding the streaming limit determined under nominal transport conditions. [8] 2. Interplanetary plasma structures may modify the propagation of particles along IMF lines by confining, trapping or mirroring the energetic particles. Kallenrode and Cliver [2001a] analyzed the SEP events that contribute the most to the total fluence measured in a solar cycle and concluded that interplanetary circumstances (such as the existence of two converging shocks propagating in the interplanetary medium) may lead to the observation of long-lasting high particle intensities. Particle mirroring (and possible reacceleration) between two propagating shocks or particle confinement in the IMF lines distorted by the presence of the interplanetary counterpart of coronal mass ejections (ICMEs) were suggested as the mechanisms to produce these enhanced intensities as the converging structures move across the observer [Kallenrode and Cliver, 2001b; Kallenrode, 2002]. Reflection of energetic particles by solar wind plasma structures formed beyond the spacecraft location (such as interplanetary shocks and/or ICMEs) has also been suggested as a possible mechanism for the production of enhanced intensities observed at 1 AU [Bieber et al., 2002]. [9] In this paper we examine proton observations collected at 1 AU in large SEP events during solar cycles 22 and 23. We explore the highest particle intensities measured in these two solar cycles and in particular those intensities exceeding the previously determined streaming limit. Since only the largest SEP events in a solar cycle reach the highest intensities, the number of events used by Reames and Ng [1998] to establish this streaming limit was very limited, especially at the highest-energy channel used in their study (i.e., MeV protons; see their Figure 4). Because of both the short flight time of the energetic particles constituting the prompt component of a SEP event (from their source near the Sun up to 1 AU) and our inability to predict large SEP events before their onset, the values of the proton intensities established as streaming limit by Reames and Ng [1998] were proposed as a measure of the minimal radiation hazard that astronauts have to tolerate before the arrival of CME-driven shocks and their associated ESP components [Reames, 1999; Neal et al., 2005]. The study of the events exceeding the previously determined streaming limit allows us to determine both the highest intensities reached in a solar cycle (and hence the highest radiation hazards) and the mechanisms leading to these high intensities. We find three (one) SEP events where the MeV ( MeV) proton intensity measured at the prompt component exceeded by a factor of 4 or more this previously determined streaming limit. We analyze the mechanisms proposed to explain the exceeding of this limit in these events by studying either (1) the possible inhibition of wave generation or (2) the existence of interplanetary structures able to confine energetic particles as they move across the observer. 2. Solar Cycle Observations [10] Figure 1 shows hourly averages of the proton intensities measured by the energetic particle sensor (EPS) [Sauer, 1993a] on board GOES-6 (green trace), GOES-7 (blue traces), GOES-8 (red traces) and GOES-11 (black traces) in the energy channels P5 (Figure 1a) and P7 (Figure 1b), the number of X-ray flares per day classified as M- or X-class and reported in the Solar Geophysical Data (available at (Figure 1c), and 2of25

3 Figure 1. Hourly averages of the proton intensities measured by GOES-6 (green trace), GOES-7 (blue traces), GOES-8 (red traces) and GOES-11 (black traces) at the energy channels (a) P5 and (b) P7. (c) Daily number of X-ray flares classified as M- or X-class (purple trace) and averages over Carrington rotation of the daily number of flares (red trace). (d) Monthly sunspot number. The vertical dashed lines indicate the start and end of the solar cycle 22 as identified by Harvey and White [1999]. The horizontal dashed lines in Figures 1a and 1b indicate the streaming limit intensity determined by Reames and Ng [1998]. the monthly sunspot number from 1 January 1986 to 31 December 2007 (available at sidc.oma.be) (Figure 1d). According to the documentation available at spidr.ngdc. noaa.gov describing the energetic particle monitors on board the series of GOES spacecraft, the energy passbands of the channels P5 and P7 were MeV and MeV for GOES-6 and GOES-7. These energy passbands were redefined to MeV and MeV for GOES-8 and GOES-11. Assuming that the energy spectrum of the differential proton intensity follows a power law in energy /E 4, the differential intensity measured in the energy window MeV ( MeV) is a factor 1.02 (2.97) larger than the differential intensity measured using the energy window MeV ( MeV). This factor reduces to 1.00 (1.17) if the energy spectrum is /E 1. [11] The energetic particle monitors on GOES were designed to handle large count rates without including any anticoincidence system. Therefore, the EPS data have been regularly corrected for background counts because of galactic cosmic rays and their secondaries, for the outof-aperture responses, and for counts due to particles entering through secondary energy passbands as described in spidr.ngdc.noaa.gov (R. Zwickl, spidr.ngdc. noaa.gov, 2001). The applied correction algorithm removes these extra counts in each energy channel by assuming that (1) the background is given by the minimum count rate measured within the previous 10 days, (2) the incident proton spectrum is a power law in energy from one energy channel to the next, and (3) the secondary energy passbands that were determined during calibration are responsible for all the secondary count rates. The correction algorithm works best when the energy spectrum of the differential proton intensity is close to a power law in energy /E 3 [Vainio et al., 1995] and fails during the onsets of SEP events at times when velocity dispersion effects dominate the energy spectrum [Posner, 2007]. Recently, Mottl and Nymmik [2007] cautioned against the use of corrected GOES/EPS data since for the highest-energy channels the corrected intensities are larger than those directly measured (i.e., uncorrected), when, in principle, the correction algorithm should remove the extra count rates. In the appendix, we reproduce the analyses presented in this article but using uncorrected GOES/EPS data instead of the corrected GOES/EPS data. [12] Figure 1 provides a complete perspective of the occurrence of the most intense and energetic particle events during the last two solar cycles. The vertical dashed lines in Figure 1 indicate the start and end of solar cycle 22 as determined by Harvey and White [1999]. The horizontal 3of25

4 dashed lines in Figures 1a and 1b indicate the values of the streaming limit determined observationally by Reames and Ng [1998] using corrected GOES/EPS data during solar cycle 22. A value of protons (cm 2 sr s MeV) 1 was inferred by Reames and Ng [1998] for the streaming limited intensity using the proton channel P5, and protons (cm 2 sr s MeV) 1 for the proton channel P7. Figure 1 shows that the intensity value established as the streaming limit was exceeded on several occasions during both solar cycles (see section 3). Note that the change of the energy passbands from GOES-6 and GOES-7 to GOES-8 and GOES-11 implies that the differential proton intensities measured by GOES-8 and GOES-11 plotted in Figures 1a and 1b (red and black traces, respectively) would have been larger (depending on the energy spectrum at each time) if they were measured using the old energy passbands of MeV and MeV. Thus the proton intensities measured by GOES-8 and GOES-11 in solar cycle 23 that exceeded the previously determined streaming limit would have also exceeded this limit if they were measured using the old energy passbands of GOES-6 and GOES-7. [13] An inspection of the particle intensities, sunspot number and X-ray flares observed over the two solar cycles (Figure 1) yields the following points: [14] 1. The two solar cycles have a different duration. Whereas solar cycle lasted 10 years from 17 October 1986 to 11 September 1996 [Harvey and White, 1999], solar cycle 23 lasted at least 11 years with a longer decay in the sunspot number. For the purpose of the present study we have set the end of solar cycle 23 on 31 October Note, however, that the first sunspot of solar cycle 24 (NOAA active region 10981) did not emerge until January 2008 ( stories2008/ _sunspot.html). [15] 2. The distributions of SEP events in each solar cycle were vastly different. Whereas in solar cycle 22 major SEP events tend to concentrate in the years of maximum activity (from 1989 to 1992), in solar cycle 23 major events were more spread out through the solar cycle (with the exception of 1999 and most part of 2006 and 2007). The first event in solar cycle 22 with proton intensities in the energy channel P7 above protons (cm 2 sr s MeV) 1 occurred on 25 July 1989 and the last one on 30 October 1992 (a time span of 1194 days), whereas in solar cycle 23 the first occurred on 6 November 1997 and the last one on 13 December 2006 (a longer time span of 3325 days). [16] 3. More SEP events occur during the maximum in solar activity than during the remaining portion of the solar cycle, but significant SEP events can occur at any time of the solar cycle, especially during solar cycle 23 with major events in November 1997 and December [17] 4. The maximum daily number of X-ray flares of class M or above was larger on specific days of solar cycle 22 than in solar cycle 23. The first M-class (X-class) flare in solar cycle 22 was observed on 5 April 1987 (2 January 1988) and the last M-class (X-class) flare on 10 July 1996 (9 July 1996); that corresponds to a time span of 3384 (3111) days. The first M-class (X-class) flare in solar cycle 23 was observed on 29 November 1996 (4 November 1997) and the last M-class (X-class) flare on 9 June 2007 (14 December 2006); that corresponds to a time span of 3843 (3327) days. The 5-year interval around the sunspot maximum from 1988 to 1992 in solar cycle 22 encompasses the 90% of X- class flares occurring throughout solar cycle 22. By contrast, the 5-year interval includes only the 59% of X-class flares in solar cycle 23. [18] Shea and Smart [2007] found that solar cycle 23 was the most active cycle since 1954 in terms of both the value of the particle fluence integrated over the entire solar cycle and the number of significant SEP events (their definition of a significant SEP event is that with >10 MeV proton intensities above 10 particles (cm 2 sr s) 1 ). The number of significant SEP events during solar cycle 23 (as per their definition) was 37% higher than the average computed over the cycles (103 versus 75), and the >10 MeV proton fluence integrated over the entire solar cycle 23 was the highest of the last five solar cycles [Shea and Smart, 2007]. [19] In order to estimate the frequency distribution of time intervals with elevated proton intensities, we have computed the number of hourly averaged proton intensity data points in both energy channels falling within a particular intensity range. We have divided the proton intensities in equally logarithmic spaced bins and computed the number of data points in each bin. Since the time coverage (total number of points) is not the same in the two energy channels and for each solar cycle, we have normalized the number of hours spent in each specific bin to the total coverage of each solar cycle. The solar cycle 22 extends from 17 October 1986 to 11 September 1996 [Harvey and White, 1999] with a total of and data points in the energy channels P5 and P7, respectively. Although the precise analysis used by Harvey and White [1999] has not been performed to determine the end date of solar cycle 23, we have arbitrarily considered the period from 11 September 1996 to 31 October 2007 as solar cycle 23 corresponding to a total of and data points in the energy channels P5 and P7, respectively. We assume that SEP events after this date will be either considered the new SEP events of the solar cycle 24 or will not add significantly to the intensity distributions shown below (mainly focused on the largest particle intensities). [20] Figure 2 shows the number of hourly averaged data points from the GOES time series plotted in Figure 1 that fall in each of the equally logarithmic-spaced intensity bins for solar cycle 22 (Figures 2a and 2c) and 23 (Figures 2b and 2d) using the energy channels P5 (Figures 2a and 2b) and P7 (Figures 2c and 2d). The dashed vertical line indicates the value of the streaming limit intensity previously determined by Reames and Ng [1998] in each energy channel. The hatched bins indicate the time intervals with intensities clearly above the streaming limit. During solar cycle 22 the hourly averaged intensities 4of25

5 Figure 2. Distribution of hourly averaged proton intensity data points in intensity bins for solar cycles (a and c) 22 and (b and d) 23 as measured in the energy channels P5 (Figures 2a and 2b) and P7 (Figures 2c and 2d). The vertical dashed lines indicate the previously determined streaming limit. Hatched bins indicate the number of data points above 10 5/4 and 10 0 particles (cm 2 sr s MeV) 1 in the energy channels P5 and P7, respectively. The number of data points has been normalized to the total number of data points in each time interval and energy channel. The error bars associated with each point are the estimated standard deviations based upon a Poisson statistical distribution. measured in the energy channel P5 showed values above 10 5/4 protons (cm 2 sr s MeV) 1 for a total of 26 data points (i.e., 26 hours), whereas the proton intensities measured in the energy channel P7 showed only 3 data points (i.e., 3 hours) above 10 0 protons (cm 2 sr s MeV) 1. The 26 hourly averaged data points of the proton channel P5 above 10 5/4 protons (cm 2 sr s MeV) 1 occurred during the intense ESP events on 20 October 1989 (8 hourly averaged data points) [Lario and Decker, 2002] and 24 March 1991 (11 hourly averaged data points) [Shea and Smart, 1993], and during the prompt component of the 12 August 1989 SEP event (7 hourly averaged data points) 5of25

6 [Richardson et al., 1994]. The 3 hourly averaged data points of the proton channel P7 above 10 0 protons (cm 2 sr s MeV) 1 were found exclusively in the ESP event on 20 October 1989 [Lario and Decker, 2002]. During solar cycle 23, the hourly averaged proton intensities in the energy channel P5 showed 75 data points above 10 5/4 protons (cm 2 sr s MeV) 1. These elevated proton intensities were observed during the ESP event on 6 November 2001 (13 data points) [Reames, 2004; Shen et al., 2008], and during the prompt component of the SEP events on 14 July 2000 (26 data points), 8 November 2000 (16 data points) and 28 October 2003 (20 data points). The hourly averaged proton intensities in the energy channel P7 showed only 2 data points above 10 0 protons (cm 2 sr s MeV) 1 exclusively observed during the prompt component of the SEP events on 14 July 2000 (1 data point) and 20 January 2005 (1 data point). In section 3 we study in detail the events whose prompt component exceeded by a factor of 4 or more the previously determined streaming limit. [21] As first noted by Reames et al. [2001b], below the streaming limit at each energy interval, the distributions shown in Figure 2 are well fit by a power law. A slight excess is observed immediately below the streaming limit for the proton intensities at the energy channel P5 in both solar cycles 22 and 23, but only in solar cycle 22 for the proton intensities at the energy channel P7. Above the streaming limit (hatched bins), the distributions fall rapidly below the power law dependence with the exception of the highest proton intensities at the energy channel P5 in solar cycle 23 (see discussion in section 3). We have fit a power law N / I -- g to those values indicated by black dots in Figure 2 weighted by their associated statistical error based upon Poisson statistics (error bars in Figure 2). The black straight lines resulting from this fit and the values deduced for the index g are indicated in Figure 2. In the respective energy channels, the indexes g are always larger in solar cycle 22 than in solar cycle 23 indicating that higher intensities were reached more often in the latter than in the former. [22] Reames et al. [2001b] suggested that these power law dependences are a pure consequence of the power law distributions observed in the peak fluxes of SEP events. The time when the peak intensity in a SEP event is reached depends on the time-intensity profile of the SEP event in particular. These time-intensity profiles may be distorted by either (1) features in the interplanetary medium (such as interplanetary shocks, interplanetary counterparts of CMEs, and corotating interaction regions) [see, e.g., Lario et al., 2008] or by (2) multiple particle injections from the Sun [see, e.g., Lario et al., 2000]. SEP events observed at 1 AU may be simplistically characterized by a rapid rise in intensity to a maximum value and a slow exponential decay to the background level, especially those events generated from western longitudes and at the high proton energies analyzed here [Shea and Smart, 1990]. Assuming that (1) most of the time with elevated proton intensities corresponds to the decay phase of major SEP events and (2) these SEP events have similar characteristic decay times, then the time spent by the particle intensity in each intensity bin plotted in Figure 2 depends on (1) the number of events observed in each energy channel, (2) the maximum intensity attained in each one of these events, and (3) the characteristic decay time of each one of these events. [23] In order to analyze the distribution of peak intensities attained in SEP events, several authors [e.g., Belov et al., 2007, and references therein] have used the differential distribution function Y(I), defined as Y(I) = dnðþ I di, where dn is the number of events with peak intensity value I on the interval di. It has been found that Y(I) follows a power law Y(I) / I -- h with h ranging from 1.13 to 1.5 using either differential energy channels or integral energy channels. The lower values of h correspond to the differential channels. For example, van Hollebeke et al. [1975] obtained h = 1.15 ± 0.15 using MeV proton data and Cliver et al. [1991] h = 1.13 ± 0.04 using MeV proton observations for well-connected events. Studies using either >10 MeV or >100 MeV proton integral channels obtained values of h always above 1.25 [Belov et al., 2007, and references therein]. For integral channels, the highest intensities result from those events that are not only intense per se but also have a hard energy spectrum. By contrast, for differential energy channels, the highest intensities result from events that are intense only within a specific energy range. Because of this double requirement when using integral channels, the distributions of peak intensities in integral channels present a larger deficit of the most intense events with respect to those with lower peak intensities than the distributions using differential channels. Hence the lower values of h for Y(I) distributions when using differential energy channels. [24] We have used hourly averages of the proton intensities plotted in Figure 1 to estimate the proton peak intensities of all SEP events observed in solar cycles 22 and 23 above a threshold peak intensity of 10 6/4 and 10 10/4 particles (cm 2 sr s MeV) 1 for the P5 and P7 proton channels, respectively. These threshold intensities are a factor of 30 above the background instrumental level (see Figure 1) and guarantee a significant number of events with distinctive peak intensities above this level. A total of 57 events (27 events) were identified in the energy channel P5 (P7) during solar cycle 22, whereas the number increased to 67 events (34 events) during solar cycle 23. Figure 3 shows the peak size distribution in the different energy channels and the different solar cycles. The value identified as peak intensity is the absolute maximum intensity of each SEP event regardless of whether it occurred at the prompt or the ESP component. Successive particle intensity enhancements due to multiple solar particle injections are considered as different events as described by Belov et al. [2005, Figure 1]. Note that there is a limit in this distinction since we use use only data collected at 1 AU from the Sun and multiple SEP 6of25

7 Figure 3. Peak size distribution of SEP events at the energy channels (a and b) P5 and (c and d) P7 for solar cycles 22 (Figures 3a and 3c) and 23 (Figures 3b and 3d). The error bars associated with each point are the estimated standard deviations based upon a Poisson statistical distribution. injections from the Sun may appear as just a single SEP event at 1 AU [see, e.g., Wibberenz and Cane, 2006, Figure 3]. [25] Above the previously determined streaming limit (dashed vertical lines in Figure 3) the statistics are low (hatched bins in Figure 3). The number of events with peak intensity in these bins are indicated in Figure 3. Specifically, the events with proton intensities in the energy channel P5 above 10 5/4 particles (cm 2 sr s MeV) 1 are the SEP event on 12 August 1989 [Richardson et al., 1994], and the ESP events on 20 October 1989 [Lario and Decker, 2002] and 24 March 1991 [Shea and Smart, 1993] in solar cycle 22; whereas in solar cycle 23 the SEP events on 14 July 2000, 8 November 2000, 28 October 2003 (see section 3) and the ESP event on 6 November 2001 [Shen et al., 2008] had peak intensities above 10 5/4 particles (cm 2 sr s MeV) 1. The events with proton intensities in the energy channel P7 above 10 0 particles (cm 2 sr s MeV) 1 are the ESP event on 20 October 1989 [Lario and Decker, 2002] in solar cycle 22, and the SEP events on 14 July 2000 and 20 January 2005 in solar cycle 23 (see section 3). [26] Broadly speaking, the peak intensity distributions shown in Figure 3 can be approximated by a power law; 7of25

8 Figure 4. Peak size distribution of SEP events at the energy channels (a) P5 and (b) P7 for solar cycles 22 and 23. The error bars associated with each point are the estimated standard deviations based upon a Poisson statistical distribution. however, the number of events with high peak intensities is low and some of the highest intensity bins do not even have an event. Following the same procedure as in Figure 2, we have fit a power law Y(I) / I -- h to those values indicated by black dots in Figure 3 weighted by the associated statistical error based upon Poisson statistics (error bars in Figure 3). The black straight lines resulting from this fit and the values of h are indicated in Figure 3. Under the assumptions that (1) all the SEP events have a rapid rise in intensity to a maximum followed by a slow exponential decay, (2) all the SEP events have the same characteristic decay time, and (3) all the SEP events are isolated from other particle injections, one would expect g = h The values of h and g in Figures 2 and 3 show that this is not the case, specially at the energy channel P7. Probable reasons for these discrepancy include the following: [27] 1. SEP events are not always characterized by a rapid rise in intensity and a slow decay. In fact, the most intense SEP events in a solar cycle are typically generated from longitudes close to central meridian and have strong ESP components [Shea and Smart, 1996] clearly differing from the assumption of rapid rising and slow exponential decaying intensities. [28] 2. Major SEP events typically occur in periods of intense level of solar activity when multiple injections take place modifying the assumed simple time-intensity profile. An estimation of the number of events that appear as grouped (we have considered two events grouped when the particle intensity of the first event starts increasing again in a new event before it returned to the preevent intensity) shows that out of the 57 (67) events in the energy channel P5 during solar cycle 22 (23), 25 (33) were grouped; in the energy channel P7 out of the 27 (34) events during solar cycle 22 (23), 14 (8) were grouped. Some wellstudied cases of these grouped SEP events include the series of events in August 1989 [Richardson et al., 1994], October 1989 [Cane and Richardson, 1995], May 1990 [Sauer, 1993b], June 1991 [Chen et al., 1994], November 1997 [Lario et al., 2000], July 2000 [Smith et al., 2001], October-- November 2003 [Lario et al., 2005b], January 2005 [Mewaldt et al., 2005] and December 2006 [Cohen et al., 2007]. They usually occur when one or two powerful active regions appear on the solar disk and trigger a series of CMEs as they rotate with the Sun. [29] 3. The characteristic intensity decay time may differ from event to event (as clearly shown in Figure 5). [30] In order to increase the statistics of the distributions shown in Figure 3, we have added the peak intensity distributions observed in the two solar cycles. Figure 4 shows the resulting distribution with the corresponding values of the index h. Note that the combination of data from both solar cycles involves the inclusion of peak intensities measured using different energy passbands. As described above, the differences may be up to a factor of 1.02 (2.97) in intensity for the energy channel P5 (P7) if the energy spectrum is / E 4. This factor diminishes for events with harder spectra. With this caveat in mind, we note that the number of events with peak intensity above the streaming limit in the energy channel P7 is low, whereas in the energy channel P5 it seems that there is an excess of events in the highest intensity bin. [31] In order to discern whether the power laws deduced from the events with low peak intensities (indicated by black dots in Figure 4) extend continuously to the 8of25

9 events with the highest peak intensities, we have performed the following c 2 test. The null hypothesis to test is whether the bins with the largest peak intensities (those without black dots in Figure 4 and below the maximum intensity bin with a number of events different from zero; i.e., 5 bins in Figure 4a and 4 bins in Figure 4b) obey the distributions Y(I) / I -- h deduced from the lower peak intensity bins (indicated by the black dots in Figure 4). We find that, for the energy channel P5, the null hypothesis is disproved to a statistical significance of 2%; whereas for the energy channel P7 the null hypothesis can only be disproved to a statistical significance of 20%. We conclude that the bins without black dots in Figure 4a are not consistent with the distribution Y(I) / I -- h determined using the low intensity bins, whereas it is not strictly proven that the bins without black dots in Figure 4b extend the deduced power law Y(I) / I -- h, since other distributions (e.g., Y(I) / e ( I/I 0), I I 0 ) could do a better job describing the distribution of peak intensity bins. [32] Both, the fact that the events with elevated peak intensities in the energy channel P5 do not continuously prolong the distributions deduced using the low peak intensities (black dots in Figure 4a) and the excess of events observed in the highest peak intensity bin suggest that the mechanisms leading to these high intensities may be different from the mechanisms operating in the events with lower intensities. Apart from the existence of an intense source of particles, the processes by which the highest intensity peaks are reached may result from unusual transport conditions leading to the odd distribution in the highest peak intensity bins shown in Figure 4a. On the other hand, the fact that the number of events with peak intensities above the streaming limit in the energy channel P7 is low indicates that either (1) the Sun has not produced enough events with elevated peak intensities at these energies during the last two solar cycles, (2) these events have not been observed from near Earth, (3) there is an actual limit to the highest intensities of the SEP events observed at 1 AU, and/or (4) the CME-driven shocks are not able to generate intense ESP components at these energies when they arrive at 1 AU. Instrumental effects in both the excess of events in the highest intensity bin of the energy channel P5 and the lack of events in the highest intensity bins of the energy channel P7 are considered below in section Most Intense SEP Events [33] Figure 5 shows the proton intensities measured during selected major SEP events at the energy channels P5 (Figures 5a and 5b) and P7 (Figures 5c and 5d). The events in Figures 5a and 5c are approximately consistent with the scenario depicted by Reames and Ng [1998] where the intensities of particles that arrive early in the SEP events are limited by (or very close to) the previously determined streaming limit, but later in the event, near the time of the shock passage (for those events generated close to central meridian) the intensities may increase by 1 order of magnitude or more. That was the case for the ESP events on 20 October 1989 [Lario and Decker, 2002] and 24 March 1991 [Shea and Smart, 1993] in solar cycle 22, and the ESP event on 6 November 2001 in solar cycle 23 [Shen et al., 2008] (Figure 5a). The specific particle flux increase of the ESP event on 20 October 1989 exceeded also the previously determined proton streaming limit in the energy channel P7 (Figure 5c). Lario and Decker [2002] showed that this specific ESP event was due to the arrival of a complex plasma structure formed in front of the associated CMEdriven shock and the high >39 MeV proton intensities were not due to local shock acceleration but to a population of particles confined in a plasma structure with low density and weak magnetic field that formed in front of the shock [Lario and Decker, 2002]. The same phenomenon consisting in the confinement of energetic particles within an interplanetary structure propagating in front of a traveling shock has been proposed by Shen et al. [2008] to explain the ESP event on 6 November [34] Figures 5b and 5d show those SEP events where the previously determined value of the streaming limit at the energy channels P5 and P7 was amply exceeded (by a factor of 4 or more) early in the event. Although the peak intensity of the 14 July 2000 event at the energy channel P7 exceeded the previously determined streaming limit by only a factor of 2, we have included this event in Figure 5d because it was one of the most intense events observed at these energies and the intensities measured at the energy channel P5 amply exceeded the streaming limit. In fact, the 20 January 2005 and the 14 July 2000 events were the two SEP events with the highest prompt components measured in the energy channel P7 during the last two solar cycles. Note that the two events in Figure 5d were measured by GOES-8 (14 July 2000) and GOES-11 (20 January 2005) where the energy passband of the channel P7 does not exactly coincide with that used to determine the streaming limit. If they were measured using the same energy passband used by Reames and Ng [1998] to determine the streaming limit (i.e., MeV), these events would have reached a higher intensity by a factor that depends on the energy spectrum at each given time. By using only a three-point measurement at the energy channels P5 ( MeV), P6 ( MeV) and P7 ( MeV) at the time of the peak intensity at the energy channel P7, we deduce an energy spectrum / E 1.81 ± 0.91 and / E 0.96 ± 0.72 for the 14 July 2000 and the 20 January 2005 events, respectively. These energy spectra would imply an increase by a factor of 1.42 and 1.16 in the peak intensity plotted in Figure 5d for the 14 July 2000 and the 20 January 2005 events, respectively if they were measured using the energy window MeV instead of MeV. [35] Comparison of the corrected GOES intensities shown in Figures 5a and 5b with corresponding data from the Charged Particle Measurement Experiment (CPME) on the IMP-8 spacecraft [Sarris et al., 1976] for those events occurring prior to October 2001 shows similar behavior 9of25

10 Figure 5. Five-min averages of the proton intensities measured during selected major SEP events at the energy channels (a and b) P5 and (c and d) P7. Events in Figures 5a and 5c consistent with the fact that particle intensities only exceed the previously determined streaming limit (horizontal dashed lines) on their ESP component (when present). The events in Figures 5b and 5d exceeded the streaming limit by a factor of 4 or more during their prompt component. We have also included the 14 July 2000 event in Figure 5d although the proton streaming limit at this energy was only exceeded by a factor of 2. The times have been shifted to the occurrence of the parent X-ray flare. The horizontal solid lines in Figures 5a and 5b and Figures 5c and 5d indicate the value 10 2 and 10 0 particles (cm 2 s sr MeV) 1, respectively, to show the level of particle intensity enhancement reached in these major events. with the same events showing the highest IMP-8/CPME MeV proton intensities [Lario et al., 2001]. [36] As a summary, Table 1 lists those events whose peak intensities (measured either during the ESP or the prompt component) exceeded the previously determined streaming limit by a factor of 4 or more. The events with prompt components exceeding the streaming limit were observed only in solar cycle 23, whereas in solar cycle 22 the intensities exceeding the streaming limit by a factor of 4 or more were observed exclusively in the ESP component of the events. [37] Figure 5b shows that the time-intensity profiles of the events with the highest intensities measured at the energy channel P5 seem to flatten around a value below 10 2 particles (cm 2 sr s MeV) 1 suggesting the possibility of saturation on the GOES/EPS detector. Such a flattening was also observed in the 14 July 2000 event using MeV proton observations by IMP-8/CPME, but not in the 8 November 2000 SEP event where this flattening was observed only in the MeV proton channel of IMP-8/CPME [Lario et al., 2001]. We also note both the excess of hourly averaged data points in the time distribution of particle intensities measured in the energy channel P5 at the highest intensity level (Figure 2b), and the elevated number of events with peak intensities in the highest intensity bin (Figures 3b and 4a). If this flattening is 10 of 25

11 Table 1. SEP Events With Peak Intensity Exceeding by a Factor of 4 or More the Previously Determined Streaming Limit Solar Cycle Energy Channel Event Solar Flare Reference 22 P5 20 Oct (ESP component) X13/4B S27E10 Lario and Decker [2002] 22 P5 24 Mar (ESP component) X9/3B S26E28 Shea and Smart [1993] 22 P7 20 Oct (ESP component) X13/4B S27E10 Lario and Decker [2002] 23 P5 14 Jul (prompt component) X5/3B N22W07 Smith et al. [2001] 23 P5 8 Nov (prompt component) M7/3F N10W77 Lario et al. [2001] 23 P5 6 Nov (ESP component) X1/3B N06W18 Shen et al. [2008] 23 P5 28 Oct (prompt component) X17/4B S16E08 Lario et al. [2005b] 23 P7 20 Jan (prompt component) X7/2B N12W61 Gopalswamy et al. [2005] an instrumental effect, then it is possible that the actual MeV proton intensities in these large SEP events may have been even larger than those shown in Figure 5b. If the flattening is a real physical effect, however, the mechanisms proposed to explain the exceeding of the streaming limit in these events and at these energies must favor the observation of flat time-intensity profiles (see discussion below). The intensities in the energy channel P7 do not show any evidence of a flat time-intensity profile or a maximum intensity line even for the most intense events (Figure 5d). [38] In the following we describe the events shown in Figures 5b and 5d and attempt to explain, in the context of the scenario proposed by Reames and Ng [1998], why the previously determined streaming limit was exceeded. Since solar wind and magnetic field data are local measurements taken at the spacecraft location, the transport conditions sampled by the energetic particles throughout their journey from their source up to the observer cannot be directly determined. The magnetic field sampled by the energetic particles when they are injected close to the Sun remains obscure to our observations if only data from 1 AU are available. Let us suppose that (1) the source of the solar wind sampled at 1 AU remains steady and corotating with the Sun, (2) the properties of the solar wind do not evolve with time and heliocentric distance, and (3) the magnetic field connection between the observer at 1 AU and the Sun is established through a perfect Parker spiral IMF. Then, if SEPs were injected close to the Sun from a longitude of, e.g., W58 the first solar wind conditions sampled by these particles would have been similar to the properties of the solar wind observed at 1 AU at the time of the onset of the SEP event (assuming a constant solar wind speed of 400 km s 1 ). Furthermore, if particle injection occurred from longitudes east (west) of W58, then the first solar wind conditions sampled by the particles would have been similar to those observed at 1 AU some time after (prior to) the onset of the SEP event (assuming a constant solar wind speed of 400 km s 1 ). However, since major SEP events usually occur during periods of intense solar activity, the sources of the solar wind rarely remain steady, and multiple transient structures such as ICMEs distort the steady structure of the solar wind, especially during solar maximum. Therefore, by using only solar wind and magnetic field data from 1 AU it is not possible to determine the plasma conditions sampled by the energetic particles throughout their travel time to the spacecraft. [39] Consider the two possible processes pointed out in the introduction that may contribute to produce SEP intensities in excess of the streaming limit (namely the inhibition of wave generation and/or the existence of interplanetary structures able to modify the nominal SEP transport conditions). By using solar wind data from 1 AU we can only infer the large-scale structure of the interplanetary medium where solar wind transients may have modified the nominal transport conditions sampled by the energetic particles, but we cannot determine the presence or absence of self-generated waves in the inner heliospheric regions where SEPs start propagating. For each one of the SEP events discussed below, we comment on our inability to determine the SEP transport conditions by using the solar wind data recorded at 1 AU prior to or after the onset of the SEP event. Another question not addressed in this paper is whether both the density of energetic particles throughout the SEP event and the solar wind conditions sampled by the particles are conducive for sustaining self-generated waves of large enough amplitude to restrict the streaming of energetic particles all along the IMF lines. In fact, these waves have only been observed in a few specific ESP events [Kennel et al., 1986; Sanderson et al., 1985; Lario et al., 2005a] but not during the prompt component of the SEP events, even using data from the two Helios spacecraft at heliocentric distances >0.3 AU [Alexander and Valdés-Galicia, 1998]. Probably, missions designed to sample the interplanetary medium closer to the Sun such as Solar Orbiter or Solar Probe may provide observational constraints on the amplitudes of these waves and thus verify or reject the predictions of their observation made by models [e.g., Ng et al., 2003] The 14 July 2000 Event [40] This event, also known as the Bastille Day event, has been widely studied by several authors [e.g., Smith et al., 2001; Reames et al., 2001c]. The solar event that originated the SEP event was associated with a fast (1674 km s 1 ) halo CME observed first by SOHO/LASCO at 1054 UT on 14 July 2000 and temporally associated with an X5/3B flare at 1021 UT on the same day at N22W07 [Smith et al., 2001]. Figure 6 shows proton intensities measured by GOES-8 at the energy channels P5 and P7 (Figure 6a); the number of CMEs per day (dashed line) 11 of 25

12 and the number of halo CMEs per day (thick line) observed by the LASCO coronagraph on board SOHO and reported in cdaw.gsfc.nasa.gov/cme_list/, together with the occurrence of X-ray flares above C-class (vertical arrows) (Figure 6b); solar wind speed (Figure 6c); solar wind proton density (Figure 6d); solar wind proton temperature (Figure 6e); plasma b p parameter (Figure 6f); magnetic field magnitude (Figure 6g); and magnetic field orientation in the RTN coordinate system (Figures 6h and 6i). Solar wind data were measured by the SWEPAM instrument on board ACE [McComas et al., 1998]. During periods of large SEP intensities we plot 33-min data from the search mode of the SWEPAM instrument provided by R. M. Skoug (private communication, 2007). Magnetic field data were measured by the MAG instrument on board ACE [Smith et al., 1998]. The two horizontal dashed and dotted lines in Figure 6a indicate the previously determined streaming limit intensity identified by Reames and Ng [1998] in the proton channels P5 and P7, respectively. The vertical solid lines in Figure 6 indicate the passage of interplanetary shocks and the gray bars the passage of ICMEs as identified by Smith et al. [2001]. Signatures of interplanetary shock passages are discontinuities in the solar wind speed, density, temperature and magnetic field magnitude; whereas signatures of ICMEs include smooth evolution of the magnetic field magnitude, magnetic field rotation, low b p and low solar wind temperature (note that not all these signatures are present in each single ICME [Neugebauer and Goldstein, 1997]). The presence of the two ICMEs prior to and after the onset of the 14 July 2000 SEP event prevents us from using the solar wind observed at 1 AU as an indicator of the solar wind conditions sampled by the SEPs throughout their journey to 1 AU. [41] The period of proton intensities in the energy channel P5 above the streaming limit occurred between the passage of two interplanetary shocks and within an ICME. The trapping of particles within the ICME and/or the mirroring of particles between the two shocks may have been the cause of this prolonged elevated proton intensity plateau [Kallenrode and Cliver, 2001a; Reames et al., 2001c]. By contrast, the proton intensities in the energy channel P7 exceeded the streaming limit by a factor of 2 during the prompt component of the SEP event just after the passage of the prior ICME. Figure 7 shows in detail the time interval with proton intensities in the energy channel P7 above the previously determined streaming limit indicated by the horizontal dashed line (note here the use of a linear vertical scale). The gray bars and vertical solid line in Figure 7 indicate the same interplanetary structures as in Figure 6. After a slow rise of 1 hour, the proton intensities in this energy channel reached a maximum at 1200 UT ± 5 minutes. Analyses of the neutron monitor observations during this event by Bieber et al. [2002] show a first anisotropic increase with peak intensity reached between 1041 and 1150 UT (depending on the neutron monitor station) followed by a rapid decrease of the anisotropy. In order to model the intensity-time and anisotropy-time profiles of the relativistic solar protons observed in this event, Bieber et al. [2002] invoked a magnetic barrier located at 0.3 AU beyond Earth s orbit. According to their model, this barrier reflected a major fraction (85%) of the relativistic protons back toward the Earth. Similarly, mirroring of protons measured in the energy channel P7 by this magnetic barrier may have caused the intensity to exceed the streaming limit; an observer at 1 AU would have detected not only those particles injected from the Sun but also those reflected from beyond 1 AU. Unlike neutron monitor observations, GOES data do not allow us to analyze anisotropies, thus the arrival direction of protons at the GOES spacecraft remains obscure. We suggest that the effects of the magnetic barrier were similar for both the relativistic solar protons and the MeV protons observed by GOES-8 and hence the elevated intensities observed at 1 AU The 8 November 2000 Event [42] Figure 8 shows with the same format as Figure 6 data for the SEP event on 8 November Gray vertical bars indicate the passage of ICMEs as identified by Cane and Richardson [2003]. A fast (>1738 km s 1 ) partial halo CME first observed by SOHO/LASCO at 2306 UT on 8 November 2000 and temporally related to a M7/3F flare at 2304 UT on the same day at N10W77 was associated with the solar origin of this event. [43] Local solar wind and magnetic field measurements (Figures 8c--8i) do not show evidence of the passage of any transient interplanetary structure during the prompt component of the SEP event. Prior to the occurrence of the parent solar event, an ICME was observed at 1 AU with leading and trailing edges 49 and 20 hours prior to the onset of the SEP event, respectively. By using the solar wind speed measured at the leading (580 km s 1 ) and trailing (440 km s 1 ) edges of the ICME we estimate that the ICME expanded from 0.7 AU (its leading edge) to 0.2 AU (its trailing edge) beyond the orbit of the Earth. The presence of this ICME prevents us from using the solar wind observed at 1 AU as an indicator of the solar wind conditions sampled by the particles injected close to the Sun presumably from a longitude of W77. As pointed out by several authors [e.g., Hundhausen, 1972], the presence of an ICME in interplanetary space modifies the IMF topology as field lines get draped around the traveling interplanetary structure. This modification of the IMF topology has also been predicted by MHD simulations of flux rope propagation in the interplanetary medium [Vandas et al., 1996; Lario et al., 1999]. When a new SEP injection occurs, the distortion of the IMF lines by the ICME affects the transport of the energetic particles, with the distortions acting as magnetic mirrors and storing the bulk of particles behind or sunward of the ICME [Lario et al., 1999; Kallenrode, 2002]. It is possible that particles reflected by the ICME observed at 1 AU prior to the onset 12 of 25