Solar Neutron Events of 2003 October-November

Abstract

During the period when the Sun was intensely active in 2003 October-November, two remarkable solar neutron events were observed by the ground-based neutron monitors. On 2003 October 28, in association with an X17.2 large flare, solar neutrons were detected with high statistical significance (6.4 σ) by the neutron monitor at Tsumeb, Namibia. On 2003 November 4, in association with an X28-class flare, relativistic solar neutrons were observed by the neutron monitors at Haleakala in Hawaii and Mexico City and by the solar neutron telescope at Mauna Kea in Hawaii simultaneously. Clear excesses were observed at the same time by these detectors, with the significance calculated as 7.5 σ for Haleakala and 5.2 σ for Mexico City. The detector on board the INTEGRAL satellite observed a high flux of hard X-rays and γ-rays at the same time in these events. By using the time profiles of the γ-ray lines, we can explain the time profile of the neutron monitor. It appears that neutrons were produced at the same time as the γ-ray emission.

Full text

SOLAR NEUTRON EVENTS OF 2003 OCTOBER–NOVEMBER
K. Watanabe,
1
M. Gros,
2
P. H. Stoker,
3
K. Kudela,
4
C. Lopate,
5
J. F. Valde
´s-Galicia,
6
A. Hurtado,
6
O. Musalem,
6
R. Ogasawara,
7
Y. Mizumoto,
7
M. Nakagiri,
7
A. Miyashita,
8
Y. Matsubara,
1
T. Sako,
1
Y. Muraki,
1
T. Sakai,
9
and S. Shibata
10
Received 2005 July 9; accepted 2005 September 14
ABSTRACT
During the period when the Sun was intensely active in 2003 October–November, two remarkable solar neutron
events were observed by the ground-based neutron monitors. On 2003 October 28, in association with an X17.2 large
flare, solar neutrons were detected with high statistical significance (6.4 ) by the neutron monitor at Tsumeb,
Namibia. On 2003 November 4, in association with an X28-class flare, relativistic solar neutrons were observed by
the neutron monitors at Haleakala in Hawaii and Mexico City and by the solar neutron telescope at Mauna Kea in
Hawaii simultaneously. Clear excesses were observed at the same time by these detectors, with the significance
calculated as 7.5 for Haleakala and 5.2 for Mexico City. The detector on board the INTEGRAL satellite observed a
high flux of hard X-rays and -rays at the same time in these events. By using the time profiles of the -ray lines, we
can explain the time profile of the neutron monitor. It appears that neutrons were produced at the same time as the
-ray emission.
Subject headings: acceleration of particles — cosmic rays — radiation mechanisms: nonthermal — Sun: flares —
Sun: particle emission — Sun: X-rays, gamma rays
1. INTRODUCTION
Relativistic particles, in particular solar neutrons, give infor-
mation about ion acceleration in solar flares. Several observa-
tions of solar neutrons in solar cycle 23 have been reported using
the international network of neutron monitors ( Usoskin et al.
1997; Lockwood & Debrunner 1999) and solar neutron tele-
scopes (Tsuchiya et al. 2001; Valde
´s-Galicia et al. 2004). Solar
flares that produce neutrons frequently occur when activity is
near its maximum during a solar cycle. In most cases they have
also produced X-class solar flares. More than 100 X-class flares
have been recorded in this solar cycle.
Intense solar activity occurred from 2003 late October to the
beginning of November. The events that occurred in this period
were observed by numerous satellites and detectors and have
been analyzed by many investigators. During the period when
three active regions appeared simultaneously on the Sun, the soft
X-ray flux was very intense and a series of 11 X-class solar flares
occurred in NOAA regions 10484, 10486, and 10488. At this
time, solar neutrons were observed on 2003 October 28 and
November 4, in association with X17.2- and X28-class solar
flares, respectively.
The X17.2 solar flare on 2003 October 28 was a remarkable
event in this solar cycle. Not only was this a large event, but many
phenomena were observed in association with this flare. It is
particularly worth noting the large flux of relativistic particles at
the Earth (Veselovsky et al. 2004; Panasyuk et al. 2004). Among
these particles were solar neutrons that were observed by the
ground-based neutron monitor before the main ground-level en-
hancement (GLE). This solar neutron event has already been
reported and discussed by Bieber et al. (2005). In this paper, we
compare neutron data with -ray data observed by the Interna-
tional Gamma-Ray Astrophysics Laboratory (INTEGRAL) satellite
and derive the energy spectrum of neutrons using these -ray data.
The second event occurred on 2003 November 4, and solar
neutrons were observed by NM64-type neutron monitors located
at different places, one at Haleakala in Hawaii and the other in
Mexico City in Mexico. In addition, solar neutrons were also ob-
served by a solar neutron telescope located at Mauna Kea in
Hawaii. Thus, it is important for a single model to be able to ex-
plain the data of the three detectors to lead to an accurate spectrum
of solar neutrons from the solar neutron event.
Simultaneous observations of solar neutrons have been made
for a few events. In the solar event of 1982 June 3, neutrons were
simultaneously observed by a ground-level detector and by
spacecraft (Chupp et al. 1987). High-energy neutrons were de-
tected by the IGY type neutron monitor installed at Jungfraujoch,
Switzerland, and low-energy neutronsand high-energy -rays were
observed by the Gamma Ray Spectrometer (GRS) on board the
Solar Maximum Mission (SMM ).
On 1990 May 24, solar neutrons were simultaneously observed
by the IGY type neutron monitors located at Climax and several
stations in North America (Shea et al. 1991; Debrunner et al.1997;
Muraki & Shibata 1996). On 1991 June 4, solar neutron signals
were recorded by the neutron monitor and the solar neutron tele-
scope located at Mount Norikura (Muraki et al. 1992; Struminsky
et al. 1994). However, in this event, the energy spectrum of so-
lar neutrons calculated from the data of these detectors was not
1
Solar-Terrestrial Environment Laboratory, Nagoya University, Furo-cho,
Chikusa-ku, Nagoya 464-8601, Japan.
2
Direction des Sciencesde la Matie
`re/ DAPNIA/SAp, Commisariata
`l’Energie
Atomique, Saclay, 91191 Gif-sur-Yvette, France.
3
Potchefstroom Campus, North-West University, Private Bag X6001,
Potchefstroom 2520, South Africa.
4
Institute of Experimental Physics SAS, Watsonova, 47 040 01 Kosice,
Slovakia.
5
University of New Hampshire, Space Science Center, Morse Hall, 39
College Road, Durham, NH 03824.
6
Instituto de Geof ı´sica, Universidad Nacional Auto
´noma de Me
´xico, Ciudad
Universitaria, Del Coyoaca
´nMe
´xico, C. P. 04510, D. F. Me
´xico, Mexico.
7
National Astronomical Observatory of Japan, Osawa, Mitaka, Tokyo 181-
8588, Japan.
8
Subaru Telescope, National Astronomical Observatory of Japan, 650 North
A‘ohoku Place, Hilo, HI 96720.
9
College of Industrial Technologies, Nihon University, 2-11-1, Shinei,
Narashino, Chiba 275-0005, Japan.
10
Collegeof Engineering, Chubu University, Kasugai, Aichi487-8501, Japan.
1135
The Astrophysical Journal, 636:1135–1144, 2006 January 10
#2006. The American Astronomical Society. All rights reserved. Printed in U.S.A.

self-consistent. This discrepancy came from the propagation model
of solar neutrons in the Earth’s atmosphere. By using the same
propagation model, which is the Shibata model (Shibata 1994),
nearly the same spectrum was obtained (Shibata et al. 1993). In this
paper, we report the analysis results of the November 4 event using
data from the neutron monitors, the solar neutron telescope, and the
spacecraft. Our model can explain data from the three detectors
consistently.
2. SOLAR NEUTRON EVENT ASSOCIATED
WITH AN X17.2 FLARE ON 2003 OCTOBER 28
2.1. Observations
An X17.2-class solar flare occurred at 9:51 UT (time observed
at Earth; same definition is used hereafter) on 2003 October 28
located in NOAA active region 10486 at S16, E08. From 10:36
to 11:06 UT, an interval that includes the start time of intense
emission of soft X-rays from the X17.2 flare, the RHESSI (Reuven
Ramaty HighEnergy Solar Spectroscopic Imager) satellite was,
unfortunately, in the South Atlantic Anomaly (SAA). However,
intense emission of high-energy -rays was seen in the data after
11:06 UT, indicating that strong particle acceleration occurred
during this flare.
On the other hand, large fluxes of hard X-rays and -rays were
observed by the INTEGRAL satellite shortly after 11:00 UT. Fig-
ure 1 shows the bremsstrahlung and line -ray time profiles from
INTEGRAL. In the top panel of Figure 1, two peaks of intense
emission of bremsstrahlung -rays are seen at around 11:03 and
11:05 UT. However, there is only one peak (around 11:05 UT) in
line -ray time profiles as shown in the second to fourth panels in
Figure 1. This more or less coincides with the second peak in the
bremsstrahlung -rays.
Figure 2 shows -ray spectra between 11:02 and 11:03 UT and
between 11:03 and 11:15 UT. From 11:02 to 11:03 UT, when the
first peak of bremsstrahlung -rays was seen, there is no line
-ray component. The -ray lines were clearly seen in the -ray
spectrum after 11:03 UT, consistent with the line -ray time
profiles shown in Figure 1. Thus, it appears that the time profiles
of ion and electron acceleration were quite different at this event,
and ion acceleration either did not occur or was quite weak dur-
ing the first peak of the bremsstrahlung -rays. We can assume
that the ion acceleration started only after 11:03 UT.
In Figure 1, note that the 2.2 MeV neutron capture -ray line
peaks around 11:06 UT and has a long decay time. The 4.4 and
6.1 MeV -ray lines of de-excited C and O ions peak around
11:05 UT, giving about a 1 minute gap between the two peak
times. In general, neutron capture -rays are delayed from -ray
lines of de-excited ions, since it takes time for high-energy neu-
trons to slow down and be captured by protons (Wang &Ramaty
1974). Thus, it is evident that solar neutrons were produced at
this flare, and they were probably produced at the same time that
4.4 and 6.1 MeV -ray lines were emitted. Hereafter, we assume
that solar neutrons were produced around 11:05 UT.
At 11:05 UT on 2003 October 28, the Sun was located over
Africa. Among our international network of solar neutron tele-
scopes, Gornergrat in Switzerland and Aragats in Armenia had a
possibility of observing solar neutrons. On the other hand, Tsumeb
observatory (17N6E, 19N1S; 1240 m above sea level [a.s.l.]) was
located just under the Sun at this time. The altitude of Tsumeb
Observatory is a little bit low; however, the air massfor the line of
sight to the Sun was thinner than that of any other observatory
because the zenith angle of the Sun was 9N5. Solar neutrons were
clearly observed by the Tsumeb neutron monitor (Bieber et al.
2005).
The 5 minute counting rate of the Tsumeb neutron monitor is
shown in Figure 3 (top). Clear excesses are seen between 11:05
and 11:15 UT and between 11:20 and 11:25 UT. The statistical
significances of these excesses are 4.8 for 11:05 – 11:10 UT,
4.2 for 11:10–11:15 UT, and 3.4 for 11:20–11:25 UT. The
total significance for the 10 minutes between 11:05 and 11:15 UT
is 6.4 .
At the same time, high-energy protons were produced in as-
sociation with this flare, and large ground-level enhancements
(GLEs) occurred around the world. We can exclude the possibil-
ity that excesses observed at Tsumeb came from energetic ions
by considering the time profile of the neutron monitor at Lomnicky
Stit (20N2E, 49N2N; 2634 m a.s.l.), together with that of Tsumeb’s
neutron monitor (Fig. 3, bottom). The start time of the first excess
of the Tsumeb neutron monitor is about 10 minutes earlier than the
event at Lomnicky Stit, while the second excess at Tsumeb is
consistent with this event at Lomnicky Stit. Thus, it appears that
the second excess at Tsumeb came from energetic ions and the
first excess was solar neutrons.
2.2. Analysis Result
Using observational data presented in x2.1, we can calculate
the energy spectrum of the solar neutrons even though neutron
monitors cannot measure the energy of neutrons. By using the
time of flight ( TOF ) method and assuming the emission time
of solar neutrons, a spectrum can be derived. We assume that
the neutrons were produced at 11:05 UT, when line -ray emis-
sion peaked, and that the energy of the neutrons responsible for
the excesses recorded by the neutron monitor is greater than
100 MeV.
From the time profile of the neutrons, we calculate the energy
spectrum of solar neutrons at the solar surface by the following
formula:
N
SEnP;ð1Þ
where Nis the number of excess counts contributed by solar
neutrons and is the detection efficiency of the neutron monitor.
Here includes the attenuation of solar neutrons through the
Earth’s atmosphere; Sis the area of the neutron monitor, Enis
the energy range corresponding to one time bin, and Pis the
survival probability of solar neutrons traveling from the Sun to
the Earth.
To obtain the of equation (1), we calculated the attenuation
of solar neutrons by the Earth’s atmosphere using the Shibata
model (Shibata 1994). Solar neutrons with energy less than
100 MeV are strongly attenuated by the Earth’s atmosphere, so
the detection of neutrons by the Tsumeb monitor directly implies
that the spectrum extended beyond 100 MeV. For the detection
efficiency of the neutron monitor, we used the result calculated
by Clem & Dorman (2000).
Using these observational and simulation results, we calcu-
lated the energy spectrum of neutrons at the solar surface using
the method used for the solar neutron event observed on 2000
November 24 ( Watanabe et al. 2003). Figure 4 shows the result
from equation (1). By fitting these data points with a power law
of the form C(En=100 MeV), the energy spectrum of solar neu-
trons was obtained. The energy spectrum is fitted by a power law
as
(3:11:0) ;1027 En
100 MeV

3:60:3
MeV1sr1:ð2Þ
WATANABE ET AL.1136

Fig. 1.—Top p ane l: Time profile of bremsstrahlung -rays observed by the INTEGRAL satellite on 2003 October 28. Second to fourth panels: Line -ray time
profiles observed by the INTEGRAL satellite on 2003 October 28. The bremsstrahlung component has been subtracted. Second panel: Time profile of the 2.2 MeV
neutron capture -rays. Third panel: Profile of 4.4 MeV -rays of C nuclei. Fourth panel: Profile of 6.1 MeV -rays of O nuclei. Bottom panel: Sum of the data in
the third and fourth panels.

For this fit, 2/dof ¼7:10/8 ¼0:89, and the 2probability is
53%. The fitting region is above 100 MeV. This power index
is a typical value for solar neutron events observed thus far.
The total energy flux of >100 MeV neutrons emitted by the
Sun was estimated to be 3:1;1025 ergs sr1.
2.2.1. Simulation by Impulsive Model
In the analysis method described above, the energy spectrum is
calculated by dividing the response into several bins, each char-
acterized by a mean energy. For the survival probability of solar
neutrons, as well as the attenuation of neutrons and detection ef-
ficiency of the neutron monitor, the values at these discrete ener-
gies are used. In order to calculate the energy spectrum of the solar
neutrons in detail, we include an assumption about the time pro-
files of solar neutrons but still assume a power-law spectral index
at the solar surface. Using this method, we can investigate whether
the neutronsare producedcontinuously. To clarify the consistency
with the conventional method, we begin by assuming that the
neutrons are produced impulsively.
In this simulation, the power index of the neutron spectrum at
the Sun is changed from 1.5 to 7.0 with a step of 0.1, while
the energy range of the incident neutrons is confined to 50–
1500 MeV. The time profile of neutrons detected by the neutron
monitor is calculated using the neutron attenuation in the
Earth’s atmosphere given by the Shibata model (Shibata 1994)
and the detection efficiency of the neutron monitor as calculated
by Clem & Dorman (2000). The decay of neutrons between the
Sun and the Earth is also taken into account. The result of this
simulation can then be compared with the observational data,
normalizing the simulated counting rate (N)totheobserved
excess counting rate (N0).
Fig. 2.—Spectra of -rays between 1.5 and 10 MeV observed by INTEGRAL at 11:02 – 11:03 UT (left) and 11:03–11:15 UT (right)on2003October28,with
background subtracted. Right: Clear signals of 2.2, 4.4, and 6.1 MeV -rays appear superimposed on the bremsstrahlung component.
Fig. 3.—Top : 5 minute counting rate observed by the Tsumeb neutron moni-
tor on 2003 October 28. The smooth solid line is the averaged background, and
the dashed lines are 1from the background. Bottom: 1 minute counting rate
of the Tsumeb neutron monitor ( black line) and time profile of the Lomnicky Stit
neutron monitor (gray line). The solar neutron event in the Tsumeb data started
well before the GLE event seen at Lomnicky Stit. Fig. 4.—Energy spectrum of neutrons at the solar surface on 2003 October 28.
WATANABE ET AL.1138 Vol. 636

Figure 5 (top) shows the reduced 2-distribution of the fit of
the simulated counting rate to the observed excess of the Tsumeb
neutron monitor obtained from the following formula:
2¼X
n
i¼1
(NiN0i)2
N0i
:ð3Þ
In this fitting, data obtained from 11:05 to 11:15 UT are used.
The 2has its smallest value when the spectral index is around
3.5. When the spectral index is 3.5, the simulated result re-
produces the observed result as shown in Figure 5 (bottom), where
2/dof ¼6:07/4 ¼1:52, which yields the minimum 2for the
simulated time profile. From this fitting, the energy spectrum is
determined as follows:
(3:30:3) ;1027 En
100 MeV

3:5þ0:4
0:2
MeV1sr1:ð4Þ
This is comparable to the result obtained using the simpler method
shown in equation (2), but the total energy flux of solar neutrons
with energy range between 50 and 1500 MeV is 9:8þ1:2
0:9;
1025 ergs sr1, about a factor of 3 higher than the other estimate.
This is because of the lower cutoff energy of the neutron spectra.
2.2.2. Simulation by Neutron Production
with -Ray Time Profile
We next simulated the neutron time profiles detected at Tsumeb
by assuming that neutrons were produced with a time spread,
since extended production of line -rays was observed by
INTEGRAL. For this calculation, we used the summed time pro-
file of 4.4 and 6.1 MeV -rays as shown in the bottom panel of
Figure 1 as the production model of solar neutrons. These are the
-ray lines of carbon and oxygen, which indicate the time profile
of ion acceleration. We used the data observed from 11:02:45 to
11:10:00 UT. The spectral index of neutrons at the Sun is varied
from 1.1 to 7.0 in steps of 0.1. The energy range of the neu-
trons is again taken to be 50–1500 MeV.
The 2for the fit were calculated by using data obtained from
11:05 to 11:15 UT. The 2has its smallest value for the spectral
index 2.9. The simulated result reproduces the observed result
most closely when the spectral index is 2.9 as shown in Figure 6.
When the spectral index is 2.9, 2/dof ¼2:76/4 ¼0:69, which
provides the minimum 2for the simulated time profiles. From
this fit, the spectral index is found to be 2:90:3. The best-
fit spectral index is harder than the index derived on the as-
sumption that the neutrons were produced impulsively, but the
total energy flux of the neutrons is now estimated to be 6:2þ0:5
0:6;
1025 ergs sr1, not very much different from the result for impul-
sive production.
3. SIMULTANEOUS OBSERVATIONS OF SOLAR
NEUTRONS ON 2003 NOVEMBER 4
3.1. Observations
On 2003 November 4, an X28-class solar flare occurred at
19:29 UT, located in NOAA active region 10486 at S19,W83
.
This is the largest solar flare on record. At around 19:42 UT,
intense emission of soft X-rays was detected by GOES (Geosta-
tionary Operational Environmental Satellite) such that the de-
tection was saturated. After 19:42 UT, intense emission of hard
X-rays and -rays was observed by the INTEGRAL spacecraft.
Unfortunately, at this time, the RHESSI spacecraft was on the
night side of the Earth. Figure 7 shows the energy spectrum of
-rays observed by INTEGRAL. In this event, although the
components of the line emission produced by de-excited ions, C
(4.4 MeV) and O (6.1 MeV), were not prominent, the 2.2 MeV
neutron capture line can be clearly seen. Intense bremsstrahlung
X-rays and -rays were also observed. Figure 8 shows the time
profiles of -rays for different energy bins that contain line -ray
components produced as a result of the ion acceleration. There is
a delay of the 2.2 MeV neutron capture -ray emission from that
of the line -ray components produced by excited ions of C and
O. Wecan assume that ion acceleration occurred at the same time
Fig. 5.—Top : Reduced 2-distribution of the fit of the simulated counting
rate to the observed excess of the Tsumeb neutron monitor. A 2 minute counting
rate is used in this calculation. The x-axis represents the power index of the
simulated time profiles, and the y-axis corresponds to the ratio of the simulated
counting rate (N) to the observed one (N0). In this fitting, data obtained during
11:05–11:15 UT are used. Bottom: 2minutecountingrate(solid line)observed
by the Tsumeb neutron monitor on 2003 October 28, together with simulated
time profile ( points) for which solar neutrons are assumed to be produced in-
stantaneously at 11:05 UT, when power index is 3.5.
Fig. 6.— Observed and simulated time profiles of the Tsumeb neutron monitor.
The solid line is the observed 2 minute counting rate,and points indicate the best-fit
simulatedtime profile for solarneutrons assumedto be produced withthe same time
profile as -ray lines.
SOLAR NEUTRON EVENTS OF 2003 OCTOBER– NOVEMBER 1139No. 2, 2006

as the -ray lines were emitted, although the main component of
these -rays is bremsstrahlung. And we can assume that solar
neutrons were produced at the same time.
At 19:45 UT, the Sun was located between Hawaii and South
America. Although the Chacaltaya observatory was the best place
to observe solar neutrons in our international solar neutron tele-
scope network at this time, no data are available because of a data
gap. Sierra Negra would also have been a good place to observe
solar neutrons, but the Mexico solar neutron telescope had not
started continuous observation at that time. Thus, it was necessary
to examine datafrom the Hawaii observatory, which was the third
closest of the observatories able to observe solar neutrons.
At 19:45 UT, the zenith angle of the Sun was 49N9 at Mauna
Kea and 50N5 at Haleakala. The air mass along the line of sight to
the Sun was 947 and 1112 g cm2, respectively. The other suit-
able location was Mexico City, where the zenith angle of the Sun
was 40N52 and the air mass along the line of sight to the Sun was
1026 g cm2. Attenuation of solar neutrons by the Earth’s atmo-
sphere above these observatories is calculated using the Shibata
model (Shibata 1994). The Haleakala and Mexico City observa-
tories have nearly the same attenuation, while Mauna Kea is
located at the best place to observe solar neutrons. However,
simultaneous signals were found in both the Haleakala and the
Mexico City neutron monitors.
Solar neutrons were observed by the 18NM64 neutron mon-
itor at Haleakala, Hawaii (203N7E, 20N7N; 3030 m a.s.l.). Figure 9
(top) shows the 5 minute averages of the counting rate observed
on 2003 November 4. At this time, the sampling interval of the
Haleakala neutron monitor was 10 s. Clear excesses were seen
after 19:45 UT, continuing for 15 minutes. The statistical signifi-
cances of these excesses are 4.5 for 19:46:20–19:51:20 UT,
5.3 for 19:51:20–19:56:20 UT, and 3.1 for 19:56:20–20:01:
20 UT. The total significance for the 15 minutes between 19:46:20
and 20:01:20 UT is 7.5 . Note that this time interval was just
taken to get the maximum significance.
Solar neutrons were also observed by the 6NM64 neutron
monitor at Mexico City (260N8E, 19N33N; 2274 m a.s.l.), as
shown in Figure 9 (bottom). At this time, the sampling interval of
the Mexico City neutron monitor was 5 minutes. Clear excesses
were seen after 19:45 UT, which was the same time as the ex-
cesses observed by the Haleakala neutron monitor. The statistical
significances of these excesses are 2.6 for 19:45–19:50 UT,
3.1 for 19:50–19:55 UT, and 3.3 for 19:55–20:00 UT. The
total significance for the 15 minutes between 19:45 and 20:00 UT
is 5.2 .
One would expect that Mauna Kea (203N7E, 19N8N; 4200 m
a.s.l.) should be a better place to observe neutrons in this event
than Haleakala and Mexico City. This is the location of the Hawaii
solar neutron telescope with an area of 8 m2, constructed from
proportional counters and plastic scintillators, but only a minimal
excess was seen after 19:45 UT in the PMT_L, PMT_H, and
layer1_with_anti channels in this telescope as shown in Figure 10.
The PMT_LandPMT
_H are channels of scintillation counter that
detect neutrons (recoil protons), the energy thresholds of which
are 12 and 20 MeV, respectively. The layer1_with_anti is a pro-
portional counter channel, which is located under the scintillation
counters. This apparent discrepancy between neutron monitors
and the solar neutron telescope is discussed in the next section
(x3.2.1) considering the surrounding environment of the detector.
3.2. Analysis Result
In order to understand the Mauna Kea result, we first use the
other observational data to estimate the energy spectrum of the
solar neutrons. We begin with the data from the Haleakala mon-
itor because it recorded the largest excess with the best time
resolution. We determine the neutron energy by using the TOF
method, assuming that all the solar neutrons were produced at
19:45 UT, the peak of the intense emission of high-energy -rays
observed by INTEGRAL as shown in Figure 8. Under this as-
sumption, the energy of neutrons observed by the Haleakala
neutron monitor between 19:51:20 and 19:56:20 UT ranged
from 59 to 913 MeV.
To derive the energy spectrum of neutrons at the solar surface
from the observed time profile by the neutron monitor, the sur-
vival probability of neutrons between the Sun and the Earth, the
attenuation of solar neutrons passing through the Earth’s atmo-
sphere, and the detection efficiency of the neutron monitor must be
taken into account. Attenuation is calculated using the Shibata
model (Shibata 1994), and we used the detection efficiency calcu-
lated by Clem & Dorman (2000).
Using these observational and simulation results, we calcu-
lated the energy spectrum of neutrons at the solar surface using
the same method as x2.1. The result is shown in Figure 11. This
spectrum was derived from 2 minute averages of the counting
rate, where the vertical errors that are shown are only statistical
errors. The energy spectrum is well fitted by a power law as
Q¼(1:50:6) ;1028 En
100 MeV

3:90:5
MeV1sr1:ð5Þ
The fitting region is chosen as 100 MeVand above because there
the errors from neutron attenuation in the Earth’s atmosphere are
small. For this fit, 2/dof ¼0:92/3 ¼0:31 and the 2probability
is 82%. This spectral index is typical of solar neutron events
observed thus far. The total energy flux of neutrons emitted from
the Sun in the energy range 59–913 MeV is estimated to be
3:4;1026 ergs sr1.
3.2.1. Simulation Usingthe Impulsive Model
By using the same method of x2.2.1, time profiles of solar
neutrons assuming their spectral index at the solar surface were
simulated on the assumption that the neutrons were produced
impulsively. We examine the 2of the fit of the simulated count-
ing rate to the observed excess of the Haleakala neutron monitor
obtained from equation (3). In this fitting, data obtained from
Fig. 7.—Spectrum of -rays between 1.5 and 10MeVobserved by INTEGRAL
from 19:40 to 19:50 UT on 2003 November 4, with background subtracted. A
signal produced by 2.2 MeV -rays appears superimposed on the bremsstrahlung
component. There is weak evidence for 4–7 MeV -ray lines.
WATANABE ET AL.1140 Vol. 636

19:45 to 20:06 UT are used. The 2has its smallest value when
the spectral index is around 3.9. When the spectral index is
3.9, 2/dof ¼10:3/7 ¼1:47, which yields the minimum 2
for the simulated time profile ( Fig. 12, top). From this fitting, the
energy spectrum is determined as follows:
Q¼2:1þ0:2
0:1;1028 En
100 MeV

3:9þ0:1
0:2
MeV1sr1:ð6Þ
This is consistent with the result obtained using the simpler
method shown in equation (5). The total energy flux of solar
neutrons within the energy range 50–1500 MeV is 6:7þ0:5
0:4;
1026 ergs sr1, about the same order as the calculated value from
equation (5).
We have done the same analysis for the Mexico City neutron
monitor. In this fitting, data obtained during 19:45–20:05 UT are
used. The 2has its smallest value when the spectral index
is around 4.3. When the spectral index is 4.3, 2/dof ¼
5:55/3 ¼1:85, which yields the minimum 2for the simulated
time profile ( Fig. 12, bottom). From this fitting, the energy spec-
trum is determined as
Q¼(1:60:2) ;1028 En
100 MeV

4:30:4
MeV1sr1:ð7Þ
Fig. 8.—Time profiles of -ray lines observed by the INTEGRAL satellite on 2003 November 4. The bremsstrahlung component has not been subtracted. Top panel: Time
profileof the 2.2 MeV neutroncapture -rays. Second panel: 4.4 MeV -rays of C nucle i. Third panel: 6.1 MeV-r ays of O nuclei. Bottompanel: Sum of the datain the second
and third panels. Although these time profiles contain line -ray components, which indicate the time profile of ion acceleration, the dominant component is bremsstrahlung.
SOLAR NEUTRON EVENTS OF 2003 OCTOBER–NOVEMBER 1141No. 2, 2006

Although the spectral index is softer than the result of the
Haleakala neutron monitor, it is consistent with equation (5). The
total energy flux of solar neutrons with energies between 50 and
1500 MeV is calculated to be (5:40:7) ;1026 ergs sr1,
about the same order as the result obtained from the analysis of
Haleakala.
We then simulated the time profile of neutrons that should
be observed from the Hawaii solar neutron telescope using
the energy spectrum of incident neutrons obtained from the data of the Haleakala neutron monitor, namely, 1:5;1028 ;
(En/100 MeV)3:9MeV1sr1. We did not attempt to derive a
spectrum from the data because excesses of the solar neutron
telescope are small.
The detection efficiency of the Hawaii solar neutron telescope is
calculated using Geant3, FLUKA-COLOR model. In this calculation,
Fig. 9.—Five minute averages of th e counting rate observed by the Haleakala
(top) and Mexico City (bottom) neutron monitors on 2003 November 4. The
smooth solid line is the averaged background, and the dashed lines are 1
from the background.
Fig. 10.—Five minute averages of the counting rate of PMT_L, PMT_H,
and layer1_with_anti channels (see details in the text) of the Hawaii solar neu-
tron telescope on 20 03 November 4. The solid smooth line is the ave raged back-
ground, and the dashed lines are 1from the background.
Fig. 11.—Energy spectrum of neutrons at the solar surface on 2003
November 4 calculated from the data of the Haleakala neutron monitor.
Fig. 12.—Best-fit simulated time profiles ( points) when the spectral index is
3.9 for Haleakala (top)and4.3 for Mexico City (bottom), superposed on the
observed counting rate. The start time of the simulated time profile is 19:45 UT,
corresponding to the peak time of -ray emission.
WATANABE ET AL.1142 Vol. 636

the Hawaii solar neutron telescope is surrounded by 20 cm con-
crete walls, since it is situated within the building housing the
SUBARU telescope. Figure 13 shows the detection efficiencies of
the Hawaii solar neutron telescope for neutrons and -rays.
The simulated result for the layer1_with_anti channel of the
Hawaii solar neutron telescope, which recorded the largest ex-
cess, is shown in Figure 14. For this fitting, 2/dof ¼1:24/3 ¼
0:41, so the simulation result is consistent with the observed
excess. Because of the high counting rate from the nonhadronic
component (-ray, muon, and so on) in the Hawaii solar neutron
telescope, these excesses are not statistically significant, al-
though they correspond to the same total flux of solar neutrons
observed by the Haleakala neutron monitor. It is however pos-
sible that solar neutrons actually produced the little hump in the
data.
3.2.2. Simulation by Neutron Production
Using the -Ray Profile
Next we simulated the neutron time profiles detected at
Haleakala and Mexico City by assuming that neutrons were
produced with a time spread. The calculation method is the same
as in x2.2.2. For this calculation, we used the -ray time profile
observed by the INTEGRAL satellite during 19:42–19:48:00 UT
as the production time profile of solar neutrons, as shown in the
bottom panel in Figure 8. The 2of the fit between observed and
simulated time profiles of the Haleakala and Mexico City neu-
tron monitors were calculated. In this fitting, data obtained from
19:42 to 20:06 UT are used for the Haleakala and from 19:45 to
20:05 UT for the Mexico City neutron monitor. For the Haleakala
data, when the power index is 3.9 (Fig. 15, top), 2/dof ¼
10:57/7 ¼1:51, giving the minimum value among the simulated
time profiles. The spectral index is determined to be 3:90:2.
For the Mexico City data, when the power index is 4.3 ( Fig. 15,
bottom), 2/dof ¼4:28/3 ¼1:43, giving the minimum value
among the simulated time profiles and a spectral index that is
determined to be 4:3þ0:4
0:5. The best-fit spectral indices are the
same as those derived by assuming that the neutrons were pro-
duced impulsively. The total energy fluxes of neutrons are esti-
mated to be (7:00:5) ;1026 ergs sr1from the Haleakala data
and 5:7þ0:7
0:8;1026 ergs sr1from the Mexico City data.
4. DISCUSSION AND SUMMARY
Relativistic neutrons were detected in association with the
X17.2 solar flare on 2003 October 28 and the X28 solar flare on
2003 November 4. The October 28 event was detected by the
Tsumeb neutron monitor, and the November 4 event was de-
tected simultaneously by the neutron monitors at Haleakala and
Mexico City and also by the solar neutron telescope at Mauna
Kea. During these events, intense emissions of high-energy
-rays were observed by the INTEGRAL satellite.
In order to investigate the production time of solar neutrons,
we compared the solar neutron data with the -ray data obtained
from INTEGRAL. In the October 28 event, -ray lines from
neutron capture and excited ions of C and O nuclei were clearly
observed and were quite different from the time profile of brems-
strahlung -rays. It appears that the time profile of electron ac-
celeration was distinctly different from the time profile of ion
acceleration. From the time profile of neutron capture -rays, it
seems that high-energy neutrons were produced with the same
time profile as -ray lines of de-excited ions. In the November 4
event, the time profiles of -ray lines, which we have assumed
represent the time profile of solar neutron production, cannot be
obtained independently, since the bremsstrahlung component was
strong and the line -ray components were buried in bremsstrah-
lung. However, from the time profile of the 2.2 MeV neutron
Fig. 13.—Detection efficiencies of the Hawaii solar neutron telescope for
neutrons when the detector is surrounded by a 20 cm concrete wall. The black lines
indicate the PMT_L(solid line)andPMT
_H(dashed line) sci ntillator channels.
The gray lines indicate layer channels of layer 1 (solid line), layer 2 (dashed line),
and layer 3 (dash-dotted line) with anticoincidence of the anticounter.
Fig. 14.—Simulated time profile ( points) of 5 minute counting rate of layer1_
with_anti channel of the Hawaii solar neutron telescope on 2003 November 4,
superposed on the observational data.The energy spectrumof the incident neutrons
is 1:5;1028(En/100 MeV)3:9MeV1sr1, which was obtained from the data of
the Haleakala neutron monitor. The start time of this time profile is 19:45 UT,
corresponding to the peak time of -ray emission.
Fig. 15.—Simulated time profiles ( points) when the spectral index is 3.9 for
Haleakala (top) and 4.3 for Mexico City (bottom), superposed on the observed
counting rate of the Haleakala neutron monitor. Points are the simulated time
profile for solar neutrons, assuming that they were produced with the same time
profile as the high-energy -rays shown in Fig. 8.
SOLAR NEUTRON EVENTS OF 2003 OCTOBER– NOVEMBER 1143No. 2, 2006

capture -rays, it appears that the time profile of ion acceleration
was approximately the same as that of bremsstrahlung emissions.
Assuming that solar neutrons were produced at the time when
these -rays were emitted, we could explain the observed excesses.
If we assume that solar neutrons were produced impulsively at
11:05 UT on October 28 and at 19:45 UT on November 4, when
the -ray lines peak, we can derive the energy spectrum of solar
neutrons at the solar surface from the neutron monitors as data by
using equations (2) and (5), respectively. For the November 4
event, in order to examine whether all excesses observed by the
Haleakala and Mexico City neutron monitors and the solar neu-
tron telescope at Mauna Kea can be expressed by one energy
spectrum consistently, and for more detailed analysis, we sim-
ulated time profiles of solar neutrons for these detectors and com-
pared with observed time profiles. All of the simulation results are
consistent with equation (5) within the range of error as shown in
equations (6) and (7). Thus, we could explain all observations
with a consistent spectrum.
Although we can fit the data by assuming that solar neutrons
are produced impulsively, it is more natural to assume that solar
neutrons are produced continuously over a finite time. We there-
fore modeled the time profiles of solar neutrons by assuming that
the neutrons were produced with the same time profile as -ray
lines from excited ions. For the October 28 event, the spectral
indices thus derived are the same as those derived by assuming
that neutrons were produced impulsively, 3.9 for Haleakala
and 4.3 for Mexico City. In other events, the spectral indices
derived by assuming that neutrons are produced continuously
tend to be harder than those derived by assuming that neutrons
are produced impulsively. However, in this event, since the
-rays have a symmetric time profile centering around 19:45 UT,
there is little difference between the two models. For the
November 4 event, the result was that the index obtained using
the line -ray time profile is clearly harder (2.9) than that
obtained using the impulsive model (3.5). Therefore, for these
two events, the observations were explained by assuming that
solar neutrons were produced with the same time profile as -ray
lines.
Although different spectral indices are obtained with a differ-
ent approach for the October 28 event, these spectral indices are
all consistent with the indices calculated from line -ray observa-
tion by the RHESSI satellite (Share et al. 2004) and INTEGRAL
satellite (Tatischeff et al. 2006).
The spectrum of accelerated ions can be calculated from the
neutron spectrum using the spectrum of escaping neutrons pro-
duced by the accelerated ions(Hua & Lingenfelter 1987a, 1987b;
Hua et al. 2002). From the neutron spectra shown in equations (2)
and (5), the number of protons above 30 MeV would be about
1032 sr1under the assumption that there is no turnover of the
spectrum. This is a typical value for solar neutron events observed
thus far.
The authors wish to thank the INTEGRAL team for their sup-
port to the mission and guidance in the analysis of the INTEGRAL
satellite data. In particular, we thank A. Bykov and M. Mendez
for having kindly permitted us to use their data in advance of
publication. We acknowledge the staff member who is manag-
ing and maintaining the Tsumeb, Lomnicky Stit, Haleakala, and
Mexico City neutron monitors. We also thank E. Flu
¨eckiger and
R. Bu
¨tikofer of the Cosmic Ray Group of the Physikalisches
Institut, University of Bern, Switzerland, members of the Armenia
group, especially A. A. Chilingarian and N. Gevorgyan, and the
staff of the Subaru Telescope for managing and maintaining the
Switzerland, Armenia, and Hawaii solar neutron telescopes. We
also thank Paul Evenson for reading this manuscript. We wish to
thank the referee for evaluating this paper and for helping us to
clarify several arguments.
REFERENCES
Bieber, J. W., Clem, J., Evenson, P., Pyle, R., Ruffolo, D., & Sa
´iz, A. 2005,
Geophys. Res. Lett., 32, L03S02
Chupp, E. L., et al. 1987, ApJ, 318, 913
Clem, J. M., & Dorman, L. I. 2000, Space Sci. Rev., 93, 335
Debrunner, H., et al. 1997, ApJ, 479, 997
Hua, X.-M., Kozlovsky, B., Lingenfelter, R. E., Ramaty, R., & Stupp, A. 2002,
ApJS, 140, 563
Hua, X.-M., & Lingenfelter, R. E. 1987a, ApJ, 323, 779
———. 1987b, Sol. Phys., 107, 351
Lockwood, J. A., & Debrunner, H. 1999, Space Sci. Rev., 88, 483
Muraki, Y., & Shibata, S. 1996, in AIP Conf. Proc. 374, High Energy Solar
Physics, ed. R. Ramaty, N. Mandzhavidze, & X.-M. Hua (Woodbury:AIP), 256
Muraki, Y., et al. 1992, ApJ, 400, L75
Panasyuk, M. I., et al. 2004, Cosmic Res., 42, 489
Share, G. H., Murphy, R. J., Smith, D. M., Schwartz, R. A., & Lin, R. P. 2004,
ApJ, 615, L169
Shea, M. A., Smart, D. F., & Pyle, K. R. 1991, Geophys. Res. Lett., 18, 1655
Shibata, S. 1994, J. Geophys. Res., 99, 6651
Shibata, S., Murakami, K., & Muraki, Y. 1993, in Proc. 23rd Int. Cosmic Ray
Conf. (Calgary), 3, 95
Struminsky, A., Matsuoka, M., & Takahashi, K. 1994, ApJ, 429, 400
Tatischeff, V., Kiener, J., & Gros, M. 2006, in Proc. 5th Rencontres du Vietnam,
New Views on the Universe, in press (astro-ph/0501121)
Tsuchiya, H., et al. 2001, Nucl. Instrum. Methods Phys. Res. A, 463, 183
Usoskin, I. G., Kovaltsov, G. A., Kananen, H., & Tanskanen, P. 1997, Ann.
Geophys., 15, 375
Valde
´s-Galicia, J. F., et al. 2004, Nucl. Instrum. Methods Phys. Res. A, 535,
656
Veselovsky, I. S., et al. 2004, Cosmic Res., 42, 435
Wang, H. T., & Ramaty, R. 1974, Sol. Phys., 36, 129
Watanabe, K., et al. 2003, ApJ, 592, 590
WATANABE ET AL.1144