Journal of Environmental Radioactivity 208-209 (2019) 105981

Contents lists available at ScienceDirect

Journal of Environmental Radioactivity

journal homepage: www.elsevier.com/locate/jenvrad

Response of the environmental thermal neutron flux to earthquakes T

a,b a c d e c
Yuri Stenkin , Victor Alekseenko , Zeyu Cai , Zhen Cao , Claudio Cattaneo , Shuwang Cui ,
Pavel Firstovf, Elio Girolettie,g, Xuewen Guoc, Huihai Hed, Ye Liuh, Xinhua Mad,∗,
Oleg Shchegoleva, Piero Vallaniai,j, Carlo Vigoritoj,k, Yaroslav Yanina,b, Jing Zhaod
a
Institute for Nuclear Research, Russian Academy of Science, Moscow, Russia
b
National Research Nuclear University MEPhI (Moscow Engineering Physics Institute), Moscow, Russia
c
The College of Physics Science and Information Engineering, Hebei Normal University, Shijiazhuang, China
d
Key Laboratory of Particle Astrophysics, Institute of High Energy Physics, Chinese Academy of Sciences, Beijing, China
e
Dip. Fisica, Università degli Studi di Pavia, Pavia, Italy
f
Kamchtkan Experimental and Methodical Seismological Department, Geophysical Survey of Russian Academy of Science, Petropavlovsk-Kamchatskiy, Russia
g
Istituto Nazionale di Fisica Nucleare, Pavia, Italy
h
School of Management Science and Engineering, Heibei University of Economics and Business, Hebei, China
i
Osservatorio Astrofisico di Torino dell’Istituto Nazionale di Astrofisica, Torino, Italy
j
Istituto Nazionale di Fisica Nucleare, Torino, Italy
k
Dipartimento di Fisica dell’Università di Torino, Torino, Italy

ARTICLE INFO ABSTRACT

Keywords: Some new results were obtained by the array of EN-detectors (Electron and Neutron detectors) developed in the
Thermal neutron frame of the PRISMA (PRImary Spectrum Measurement Array) project for Extensive Air Showers detection. Our
Radon EN-detectors running both on the Earth surface and underground are continuously measuring the environmental
Earthquake thermal neutron flux. Neutrons are partially produced by radioactive gas radon and its daughter decays through
EN-Detector
(α,n)-reactions in soil close to the detectors. Then neutrons thermalize in media and, being in equilibrium with it,
they are sensitive to many geo-dynamic phenomena including earthquakes. In this work the EN-detectors were
measuring the variations of an environmental neutron flux in Tibet (30.11 N, 90.53 E, 4300 m a.s.l) at a distance
of ∼600 km from the collision zone of the Asian-Indian plates subduction zone (Nepal region). We have ob-
served some anomalies in the dynamics of the neutron flux around the time of the catastrophic earthquakes of
magnitude M = 7.8 happened in Gorkha (Nepal) on 25.04.2015 followed by a series of aftershocks of M > 6.
The use of nuclear physics methods can provide novel results in geophysics and this work demonstrates the
sensitivity of the environmental thermal neutron flux to changes in tense-deformed crust conditions caused by
earthquakes with epicentral distances greater than 500 km.

1. Introduction
Radioactive gas radon produced continuously in the Earth's crust
has been often proposed as a possible tool for predicting the occurence
of earthquakes (see for example (Marcoet et al., 2016) (Floreset et al.,
1990) (Friedmann, 2012) (Zmazekaet et al., 2002) (Cicerone et al.,
2009) and references therein). On the other hand, continuous mon-
itoring of radon activity both above and below ground is not trivial and
could be misleading due to the direct influence of microclimatic con-
ditions (as humidity, rain, snow, pressure, droughts and ventilation) on
radon variations.
The Electron-Neutron detectors (EN-detectors) developed for the
PRImary Spectrum Measurement Array (PRISMA) project (Stenkin,
2009) showed a long-term stability and insensitivity to the above pro-
blems because they measure neutrons produced in (α,n)-reactions in-
side the soil or rock, not in air (due to the lack of suitable target nuclei).
Therefore, the neutron flux is proportional to the concentration (flux) of
radon and its daughter (alpha emitters) nuclides in surrounding solid
matter, like soil, rock or concrete.
Strong earthquakes happened in April–May 2015 in the Nepal re-
gion produced a significant response in our detectors located in Tibet
(YangBaJing, People's Republic of China, 4300 m a.s.l.) at a distance of
∼ 600 km from the epicenter. In this work we analyze for the first time
the response of our detectors looking for a reliable correlation between
the earthquakes and the thermal neutron flux variations. The cata-
strophic Gorkha earthquake in Nepal in April 2015 will be deeply

∗
Corresponding author.
E-mail address: maxh@ihep.ac.cn (X. Ma).

https://doi.org/10.1016/j.jenvrad.2019.05.013
Received 6 March 2019; Received in revised form 17 May 2019; Accepted 17 May 2019
0265-931X/ © 2019 Elsevier Ltd. All rights reserved.
Y. Stenkin, et al.
analyzed. Further details on the detectors and our analysis method can
be found elsewhere ((Alekseenkoet et al., 2015) (Bartoliet et al., 2016)
(Stenkinet et al., 2017a)), while our first results on the correlation be-
tween thermal neutron flux and local seismic activity can be found in
(Alekseenkoet et al., 2009).
2. EN-detector
A novel type of thermal neutrons scintillation detector (EN-detector)
based on solid granulated alloy 6LiF + ZnS(Ag) is used to detect
thermal and epithermal neutrons. The scintillator layer thickness is
∼30 mg/cm2 and the sensitive area of the scintillator surface is
0.35 m2, viewed by a 6″-photomultiplier (PMT) FEU-200. A schematic
description of the EN-detector and its performances can be found
elsewhere ((Alekseenkoet al, 2015) (Stenkinet et al., 2017a)). The ef-
ficiency for thermal neutrons capture was found to be ∼20%, while
neutron recording efficiency measured by using pulse selection method
is close to 15%. The efficiency shows a ∼1/v dependence on neutron
velocity v, typical for thin detectors. Due to the thin scintillator layer,
relativistic charged particles as electrons or muons produce very small
signals well below the discrimination threshold of ∼ 3 minimum io-
nizing particles. For this reason our detector is not sensitive to the
crossing of a single minimum ionizing particle, has very low counting
rate and can be used in scaler mode to study the variations of the en-
vironmental thermal neutron flux. Due to the existence of several
scintillating time constants (from 40 ns to several µs ) in ZnS(Ag), the
pulse shape produced by heavy and slowly moving particle (like triton
and alpha emitted after neutron capture on 6Li target) is longer than
that produced by relativistic charged particles or by PMTs noise.
Therefore, we exclude background signals (but still counting them as
“charged” component) by means of pulse shape digitizing and using
pulse shape selection method ((Alekseenkoet et al., 2015)).
3. Diurnal wave
Four EN-detectors (PRISMA-YBJ array) were operated continuously
at the Yangbajing International Cosmic Ray Observatory between
August 2013, and March 2017 ((Alekseenkoet et al., 2015)), as a pro-
totype of the PRISMA-LHAASO project. During this study, continuous
data time series were accumulated and analyzed. Using the super-
imposed epoch method, the diurnal wave S1 was obtained for both
pulse types classified as “neutron” and “charged” (Fig. 1). No correc-
tions were applied to the data. A smooth blue curve shows the tem-
perature daily variation profile in the experimental hall. Fig. 1 shows
that the two types of events (“neutron” and “charged”) are anti-corre-
lated, possibly due to:
i) the “neutron” wave is very close to the air temperature daily var-
iations in the vicinity of the surface because thermal neutrons are in
equilibrium with media and their mean velocity, according to the
Maxwell-Boltzmann distribution, is:
8kT
<v > = , response of soil radon to barometric pressure is delayed compared to
m (1)
where k is the Boltzmann constant, T is the absolute temperature in
Kelvin degrees and m is the neutron mass, while their flux F ∼
C < v > , where C is the neutron concentration in air;
ii) pulses selected as “charged” are partially connected to radon decays
in air ((Stenkinet et al., 2017a)) and they mostly follow the wave of
radon exhalation from soil and its concentration in air, which are
not connected to air temperature but mostly to soil temperature. Air
dust and aerosols also affect this branch ((Stenkinet et al., 2017a)).
Fig. 1 shows that the maximum of “charged” component coincides
approximately with the sunrise when the temperature derivative
2
Journal of Environmental Radioactivity 208-209 (2019) 105981
Fig. 1. Diurnal wave observed in the PRISMA-YBJ EN-detectors (“neutron” and
“charged” channels, see text) and in absolute temperature inside the experi-
mental hall (in K degrees) normalized to 0 ∘C vs astronomical local time. Circles
indicate Sun upper (yellow) and lower (black) culminations. (For interpretation
of the references to color in this figure legend, the reader is referred to the Web
version of this article.)
changes sign sharply and this could cause an increase of radon ex-
halation from soil. It should be also noticed that in the Yangbajing
region, a very high level of radon concentration in the indoor air
could be explained by the high geothermal activity confirmed by the
presence of a nearby geothermal power station. It is well-known that
geothermal water contains a very high radon concentration giving
rise to a high radon concentration in the surrounding air. This latter
is verified by our radon measurements in the experimental hall.
We remark the fact that the two quantities related to the radon
concentration in soil and air measured by the EN-detectors have anti-
correlated mean diurnal waves. This result is shown in Fig. 1 and
published here for the first time. A closer look to Fig. 1 shows a trickier
behaviour. The shape of the temperature wave almost coincides with
the neutron one but its amplitude AT is much bigger than the ex-
pectation from the Maxwell distribution: Aexp = AT 1.012 = 1.006
while the experimental value is about 7 times higher (1.04). Therefore,
even assuming that the outdoor temperature wave has a twice ampli-
tude, the neutron diurnal wave cannot be explained by temperature
influence alone and there should exist other factors affecting the
thermal neutron diurnal wave. The observed difference between the
diurnal wave phase of neutron and “charged” components could be
explained reminding that neutrons are generated by radon decays in
soil at an effective depth of 2–3 m. Therefore, their flux depends on
radon concentration multiplied by the neutron yield probability in the
depth range 0–3 m, while the flux of “charged” events depends on
radon concentration in air. According to (Firstovet et al., 2007), the
radon in air. This delay corresponds to about half a day for a depth of
1.5 m, and its duration depends on the soil parameters (porosity,
geology, etc.). For these reasons, it would be reasonable to have a delay
of 8–10 h between neutron and air radon diurnal waves if both are
sensitive to barometric or thermal diurnal waves. It is clear that the
temperature diurnal wave phases in soil and air are also different.
4. Response of the EN-detectors to the Nepal earthquakes
During the strong Nepal earthquake of magnitude M = 7.8 on April
25, 2015 (see for example (Malovichkoet et al., 2016)) the PRISMA-YBJ
array was running at a distance of ∼ 680 km from the epicenter. The
parameters of the most powerful shocks are presented in Table 1.
Y. Stenkin, et al. Journal of Environmental Radioactivity 208-209 (2019) 105981

Table 1
Main shocks of Nepal earthquakes in April–May 2015.
Date UTC deg., N deg., E depth, km Mw

25.05.2015 06: 11: 26.9 28.24 84.74 15 7.8

25.05.2015 06: 45: 22.2 28.29 84.90 15 6.7
26.05.2015 07: 09: 10.3 27.86 86.08 15 6.7
12.05.2015 07: 05: 19.0 27.89 86.17 10 7.3
12.05.2015 07: 36: 53.7 27.65 86.31 10 6.2

It is believed that radon concentration could be used as a precursor

of earthquakes ((Cicerone et al., 2009)) due to its long (3.8 d) half-life
that allows its migration for longer distances both in air and soil
compared to thermal neutron diffusion length (corresponding to a few
meters in soil and tens of meters in air). Radon could migrate for much
longer distances than neutrons inside the soil (up to 15–20 m, de-
pending on the soil conditions). We should emphasize that neutrons are
produced locally close to the detector and proportionally to the radon
decay rate in the surrounding soil. Note that radon-222 is produced
continuously in the crust being an element in the uranium radioactive
chain and can accumulate in pores and cracks under pressure. There-
fore, its concentration near the detector depends on the radon diffusion
from lower levels, and this in turn is linked to soil porosity and con-
ditions like ground water level, soil vibration, etc.
Our method based on the recording of radon-due thermal neutrons
allows the monitoring of the concentration of soil radon in time and
thus any possible co-seismic, precursor or post-seismic dynamic
anomaly. For this purpose, we use both time series: thermal neutrons
and “charged”. The latter was found to be sensitive to radon decays in
air, while neutrons are sensitive to radon decays in soil ((Stenkinet
et al., 2017a)).
Results are shown in Fig. 2, where S, the sum of normalized “neu-
trons” and “charged”, is defined as:
Nn Nch
S= + 2,
< Nn> < Nch> (2) Fig. 2. Response of the PRISMA-YBJ detectors during the Nepal earthquakes in
April–May 2015: the arrows show the occurrence of the earthquakes with the
where Nn and Nch are respectively neutron and “charged” rates in
corresponding magnitudes (upper panel); aftershocks time distribution taken
5 min, and < Nn > and < Nch > are the average values of neutron and
from the EMSC site: https://www.emsc-csem.org/Earthquake/233/M7-8-
“charged” rates in the same time window. We use the sum S to em- NEPAL-on-April-25th-2015-at-06-11-UTC (lower panel).
phasize the change in phase of the daily time series of neutrons. As
already shown in Fig. 1, these two branches have usually anti-corre-
lated phases of the diurnal waves. However, during the days around a Another event occurred 1.5 years later is shown in Fig. 4 for the
strong earthquake, the two waves have very similar phases that do not same parameters. In that case, a series of weaker earthquakes occurred
compensate each other giving a maximum in their sum clearly visible in in Nepal (October 11–12, 2016) and in Tibet (Oct. 16, M = 5.9, dis-
Fig. 2. Neutron phase shift depends on upper soil layers parameters that tance ≈500 km). Again, a peak of amplitude ∼ 9% was observed the
are believed to depend strongly on local seismic activity able to change day following the main earthquake in coincidence with the aftershock.
it quickly. Fig. 7 in Ref. (Firstovet et al., 2007) confirms our results The single earthquake in Tibet on Oct. 16, 2016 did not produced
showing changes in soil radon diurnal wave phases some days before significant effects.
and during earthquakes of M = 5.5–5.9 at distances of 350–370 km. It is interesting to note that all these events have similar features: all
The lower panel of Fig. 2 displays the aftershocks time distribution. the biggest earthquakes occurred between 14:00 and 16:00 local time
It is noticeable that this effect did not appear on April 25 when the when the neutron diurnal wave usually has its maximum; big after-
strongest earthquake occurred (M = 7.8), but only the next day (April shocks occurred about 24 h later, i.e. at the same local time. Probably
26) when an aftershock earthquake of M = 6.7 was observed at a dis- this fact and the location of the shocks between Kathmandu and Mt.
tance of ≈580 km from PRISMA-YBJ. After 2 weeks, two other strong Everest can explain why we had a significant response from these
aftershocks occurred (of magnitude M = 7.3 and M = 6.7 at about the events with signals one day later. This is shown by the graphs of Fig. 5.
same distance) and again we saw a significant peak of ∼ 6% amplitude. The next day diurnal waves differ significantly from the expected
The epicenters of all these three aftershocks were located between curves shown in Fig. 1, while the waves during the strongest earth-
Kathmandu and Mt. Everest, very close together and about 100 km quakes (25.04.15 and 11.10.16 but 12.05.15) do not show any differ-
closer to YangBaJing than the main earthquake on April 25. The region ence. We are not able to clearly explain this behavior but we believe
map (from https://www.emsc-csem.org) is shown in Fig. 3 and could that it is strictly linked to the earthquake occurrence.
explain the different response of our detectors based on the different To calculate the significance of the observed signals we searched all
regions geology, resulting in different seismic wave propagation from the events with peaks greater than 6σ in both detectors d1 and d4
different epicentre locations. ((Bartoliet et al., 2016)) over the whole 3.5y data set, these two

3
Y. Stenkin, et al. Journal of Environmental Radioactivity 208-209 (2019) 105981

Fig. 3. Map of Nepal earthquakes location in April–May 2015 and in October 2016 taken from the EMSC site. Circles show the earthquakes (with radii proportional to
magnitude M) while the star shows the detector position at Yangbajing (upper panel); geological map of the area (lower panel).

detectors being the most stable ones. Original data have been acquired
every 0.5 h. Then they were smoothed with 3 h adjacent-averaging
method and the mean value for the entire period of 3.5 y was sub-
tracted.
The significance of the two peaks on April 26, 2017 (9.6 and 9.4 σ)
is the biggest one observed through the whole period. The shock pairs
(April 25–26 and October 11–12) occurred at about the same time (15 h
local time) and both aftershocks repeated after ≈ 24 h. These pairs gave
us the largest effects. No single shock (including the Tibetan one of
M = 5.9 on Oct. 17, 2016 at a distance of ≈500 km to N-E) gave sig-
nificant effects. The only exception is probably the event on May 12,
2015 but the situation is tricky due to 5 earthquakes of M > 4 on the
previous day and 6 earthquakes with M > 5 on the same day. A delay
of 1 or 2 days is possible due to the slow radon diffusion in soil, as we
observed in the underground detector ((Stenkinet et al., 2017b)).
However, the main reason could be due to a quasi-resonant behavior of
the effect. For this analysis, we used the same detectors d1 and d4 lo-
cated indoor at a relative distance of 5 m. Fig. 6 shows the distributions
of the S parameter for these detectors over the whole 3.5 y period.
These distributions have a Gaussian behavior except for the long tails
below a probability of ∼10−4. These tails demonstrate that the dis-
tributions of the S parameter are not only statistical and that the peaks
listed in Table 2 cannot be explained by pure statistical fluctuations but
by some physical processes.

4
Y. Stenkin, et al.
Fig. 4. Response of the PRISMA-YBJ detectors during the Nepal earthquakes in
October 2016.
Fig. 5. Neutron diurnal waves for the earthquake days and the following ones.
The colors correspond to the different days. The arrows pinpoint the occurrence
of the strongest earthquakes in the Nepal region and the corresponding mag-
nitudes. (For interpretation of the references to color in this figure legend, the
reader is referred to the Web version of this article.)
5. Response of radon meters to the Nepal earthquakes
Besides the EN-detectors, two Lucas cells located in the centre and
close to the North wall of the ARGO experimental hall monitored the
radon concentration level in indoor air ((Stenkinet et al., 2017a)).
Fig. 7 shows the response of the two radon meters around the Nepal
main earthquake on April 25, 2015. A significant signal is detected at
least by the North radon meter. A preliminary work showed that radon
gas percolates into the YBJ hall from the north side. This could explain
why the North radon meter measures higher variations with respect to
the central one. Fig. 8 shows the response of the North radon meter
during the earthquakes in October 2016. Rather big peaks can be seen
5
Journal of Environmental Radioactivity 208-209 (2019) 105981
Fig. 6. S parameter distribution for detectors d1 and d4 over the whole 3.5
years measurement period. Histograms: experimental data; smoothed curves:
gaussian fits.
Fig. 7. Response of the radon meters during the Nepal earthquakes in April
2015.
not in coincidence with the shocks. To show the non-statistical behavior
of the North radon meter data we plotted in Fig. 9 its counting rate
significance for the whole data set, similarly to what is shown in Fig. 6
for neutrons. As can be seen, the experimental distribution differs from
the gaussian one for probabilities less than ∼10−2, i.e. a factor ∼100
greater than for neutrons.
6. Conclusions
Using the EN-detectors developed at INR RAS for measuring the
variations of environmental neutron flux, we observed some geophy-
sical phenomena connected to the strong earthquakes happened in
Nepal in 2015 and 2016. We hope that the use of nuclear physics
methods applied to geophysics could give us novel interesting results.
Our method gave only 2 false (not identified) signals over 3.5 years of
observation. As shown, the EN-detectors response depends on geolo-
gical region, epicenter distance, soil porosity, etc. According to the
model of EQ preparation zone (Dobrovolskyet et al., 1979), our
Y. Stenkin, et al. Journal of Environmental Radioactivity 208-209 (2019) 105981

Table 2
Parameters of the most significant peaks (S >6 ) in detectors 1 & 4 (sum of normalized neutron and charged particle signals in standard deviations).
Date (day. month.year) Signal in d1 (σ) Signal in d4 (σ) Event Comments

07.02.15 9.2 7.1 ? not understood

26.04.15 9.6 9.4 25.04: Nepal M = 7.8; 3 huge EQs within ∼0.5 and ∼25 h between them and many aftershocks
25.04: Nepal M = 6.7;
26.04: Nepal M = 6.7
12.05.15 6.7 6.3 12.05: Nepal M = 7.3; 2 huge EQs within ∼0.5 h between them
12.05: Nepal M = 6.2
29.03.16 6.2 6.7 ? not understood
08.05.16 9.0 6.3 08.05: Magnetic storm Kp = 7 big magnetic storm
12.10.16 10.9 8.4 11.10: Nepal M = 4.6; 2 EQs within ∼24 h between them
12.10: Nepal M = 4.1

observation of Gorkha shock and aftershocks with M > 6.5 in neutrons

lies inside the radius of preparation, while the Tibetan EQ does not and
indeed it did not produce any measurable effect. As for double quakes
with M < 5, spaced out of ∼ 24 h, they could give visible effects only
assuming a quasi-resonant impact to the crust processes generating the
thermal neutron diurnal wave. Therefore, we hope that a global net of
EN-detectors could give additional information to the geophysicists
contributing to the understanding of the earthquakes origin and their
preparation zone in time for better studying the geo-dynamic phe-
nomena.

Acknowledgments

This work was supported in Russia by RFBR (18-02-00339, 16-29-

13067_ofi_m), “Fundamental properties of matter and astrophysics”
RAS Presidium Program, and in China by NSFC (No. 10975046,
11375052) and by the International Partnership Program of Chinese
Fig. 8. Response of the North radon meters during the Nepal earthquakes in Academy of Sciences (Grant No. 113111KYSB20170055). We also ac-
October 2016. knowledge the support of the ARGO-YBJ and LHAASO collaborations.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://

doi.org/10.1016/j.jenvrad.2019.05.013.

References

Fig. 9. Distribution of the radon data peaks integrated over 3 h measured by the
North radon meter. Histogram: experimental data; smoothed curve: gaussian
fit.
Alekseenko, V.V., et al., 2009. Izv. Phys. Solid Earth 45 (8), 709–718.
Alekseenko, V.V., et al., 2015. Phys. Rev. Lett. 114, 125003.
Bartoli, B., et al., 2016. Astropart. Phys. 81, 49–60.
Cicerone, R.D., Ebel, J.E., Beitton, J.A., 2009. Tectonophysics 476, 371–396.
Dobrovolsky, et al., 1979. Pure Appl. Geophys. 117 (5), 1025–1044.
Firstov, P.P., et al., 2007. J. Volcanol. Seismol. 1 (6), 397–404.
Flores, Humanante, et al., 1990. Radon signals related to seisimc activity in Ecuador.
Pageoph 132 (3), 504–520 March 1987.
Friedmann, H., 2012. Radiat. Protect. Dosim. 149 (2), 177–184.
Malovichko, A.A., et al., 2016. Seism. Instrum. 52 (3), 195–206.
Marco, Neri, et al., 2016. Soil radon measurements as a potential tracer of tectonic and
volcanic activity. Sci. Rep. 6, 24581. https://doi.org/10.1038/srep24581.
Stenkin, Yu V., 2009. Nucl. Phys. B 196, 293–296.
Stenkin, Yu V., et al., 2017a. Pure Appl. Geophys. 174, 2763–2771.
Stenkin, Yu V., et al., 2017b. J. Exp. Theor. Phys. 124 (5), 718–721.
Zmazeka, B., et al., 2002. Appl. Radiat. Isot. 56, 649–657.