PUBLICATIONS

Geophysical Research Letters

RESEARCH LETTER Gamma Ray Signatures of Neutrons From a 10.1002/2017GL075071 Terrestrial Gamma Ray Flash Key Points: G. S. Bowers1 , D. M. Smith1 , G. F. Martinez-McKinney1, M. Kamogawa2 , S. A. Cummer3 , • Gamma rays producing a neutron 4 5 6 7 flash with a ground fluence of J. R. Dwyer , D. Wang , M. Stock , and Z. Kawasaki 2 1,000 n cm and 100 ms duration 1 2 3 was observed following a University of California, Santa Cruz, CA, USA, Tokyo Gakugei University, Tokyo, Japan, Duke University, Durham, NC, USA, 4 5 6 strike University of New Hampshire, Durham, NH, USA, Gifu University, Gifu, Japan, Networks, Germantown, MD, USA, • This observation is consistent with 7Osaka University, Osaka, Japan simulations of photoneutron production from a terrestrial gamma ray flash (TGF) with typical brightness • fi Abstract Following a lightning strike to a wind turbine in Japan, we have observed a large burst of These measurements de nitively 2 show that low altitude atmospheric neutrons lasting 100 ms with a ground fluence of ~1,000 n cm , thousands of times greater than the electricity can drive nuclear processes peak neutron flux associated with the largest ground level solar particle event ever observed. This is the first detection of an unequivocal signature of neutrons from a terrestrial gamma ray flash, consisting of a 2.223 MeV gamma-ray spectral line from a neutron-capture on hydrogen reaction occurring in our detector, 12 13 Correspondence to: and is shown to be consistent with the production of 10 –10 photoneutrons from a downward terrestrial G. S. Bowers, gamma ray flash (TGF) at 1.0 km, with a gamma ray brightness typical of upward TGFs observed [email protected]; [email protected] by . Plain Language Summary Using detectors to measure gamma-rays, we have observed a cloud of Citation: neutrons following a lightning strike to a wind turbine in Japan. Bowers, G. S., Smith, D. M., Martinez- McKinney, G. F., Kamogawa, M., Cummer, S. A., Dwyer, J. R., … Kawasaki, 1. Introduction Z. (2017). Gamma ray signatures of neutrons from a terrestrial gamma ray Terrestrial gamma ray flashes (TGFs) are submillisecond bursts of gamma ray emissions up to 40 MeV asso- fl ash. Geophysical Research Letters, 44, ciated with lightning discharges and have routinely been observed by satellites over the last 25 years 10,063–10,070. https://doi.org/10.1002/ 2017GL075071 (Briggs et al., 2010; Fishman et al., 1994; Marisaldi et al., 2010; Smith et al., 2005; Ursi et al., 2017). Current observations support the idea that TGFs are an acceleration and avalanching of relativistic electrons in thun- Received 9 AUG 2017 derstorm electric fields that produce gamma ray bremsstrahlung (Dwyer et al., 2012). Neutron production Accepted 11 SEP 2017 associated with lightning has been reported (Ishtiaq et al., 2016; Gurevich et al., 2012, 2016; Martin & Alves, Accepted article online 18 SEP 2017 Published online 13 OCT 2017 2010; Shah et al., 1985; Shyam & Kaushik, 1999), and both thermonuclear fusion in the lightning channel (Libby & Lukens, 1973) and photonuclear reactions (Babich, 2007) have been proposed as possible generation mechanisms, but a definitive observation distinguishing these two cases has not been made. Fusion has been shown to be too inefficient to account for observations (Babich, 2006). In this work we present a detailed observation from a suite of scintillation detectors, coincident with a lightning strike to a nearby wind turbine caught on high speed camera, of a sudden-onset increase in scintillator counts lasting ~100 ms with a spectroscopic and temporal signature unique to the presence of thermal neutrons. These observations are compared to Monte Carlo simulations and shown to be consistent with photonuclear production from TGF gamma ray interactions in the ground and atmosphere. The derived TGF brightness is comparable to those seen from (Mailyan et al., 2016).

2. Materials and Methods 2.1. Gamma Ray Instrument The Gamma ray Observations During Overhead Thunderstorms (GODOT) instrument consists of three scintil- lation detectors and two additional electronics chains without detectors for noise monitoring. Each scintillator is encapsulated with and mounted to a photomultiplier tube (PMT) inside a MuMetal shielded housing. The three scintillators are cylinders with equal length and diameter, one 2.5 cm in each dimension, made of BC-408 plastic (“SmPl”), one 12.5 cm of BC-408 (“LgPl”), and one 12.5 cm of NaI(Tl) crystal (“NaI”). For noise monitoring we monitor with the same readout electronics a 30 cm long 50 Ω terminated

©2017. American Geophysical Union. coaxial cable and a small blank detector, which is identical to the small plastic detector system, but without All Rights Reserved. a scintillator coupled to its PMT. During the neutron flash, the multichannel analyzer (MCA) connected to the

BOWERS ET AL. GAMMA RAYS FROM A TGF NEUTRON FLASH 10,063 Geophysical Research Letters 10.1002/2017GL075071

50 Ω terminated cable did not register any counts. The MCA connected to the blank detector was not opera- tional as it was writing its buffer to file. Each detector is connected to a Bridgeport Instruments eMorpho MCA that uses a time-tagged event mode to record the integrated pulse area (with 16-bit resolution) and arrival time (with 32-bit/12.5 ns resolution) of the detector output. Each eMorpho can store 5461 events, with the exception of the eMorpho connected to the 50 Ω terminated cable, which can only store 341. During the buffer readout the eMorpho is not capable of recording new counts. In the winter of 2015, GODOT was deployed to the city of Uchinada along the western Ishikawa prefecture of Japan, 300 m from the base of a wind turbine and its lightning protection tower (Figure 1). The coast of Japan is an ideal place to look for TGFs from the ground as coastal winter thunderstorms have very low charge centers (Krehbiel, 1986) where TGF production might occur.

2.2. Atmospheric Electric Field Instrument and Broadband Observation Network for Lightning and Thunderstorms Sensor A Boltek Co. Ltd. EMF-100 atmospheric electric field (AEF) monitor was installed next to GODOT to measure variations (with 2 Hz sampling) in the AEF during thunderstorm and lightning activity. Additionally, a Broadband Observation network for Lightning and Thunderstorms (BOLT) sensor ~1.25 km east of GODOT recorded fast electric field changes (with 4 MHz sampling) due to charge motion from lightning discharges.

2.3. High-Speed Camera A high-speed camera, NAC MEMRECAM GX-8, operated at 30,000 frames per second and colocated with the BOLT sensor had been previously installed to observe lightning strikes to the Uchinada wind turbine and its lighting protection tower (Wang & Takagi, 2012).

3. Observation of an Intense Neutron Flash During a winter thunderstorm on 3 December 2015, the lightning protection tower was struck by lightning at ~20:20:29 UTC, initiated by a positive upward leader from the tower resulting in a negative cloud-to-ground (CG) event (Figure 1, bottom). During the lightning flash, GODOT recorded a burst of counts in all three scintillation detectors, coincident with a large change in the electric field consistent with the removal of negative charge overhead (Figure 2a). The NaI(Tl) detector recorded the most counts, but we do not consider those data here due to calibration difficulties at high count rates. The LgPl and SmPl detectors (capable of recording energies from 300 keV to 25 MeV and 100 keV to 6 MeV, respectively) saw a burst of events that slowly faded over a ~100 ms timescale (Figure 2b). During the event, the LgPl detector’s memory buffer filled up and it was inactive while the buffer was read out (Figure 2b). There is ~500 μs uncertainty in the absolute timing of the GODOT data, but the event was consistent with a brightening at the tip of the tower observed by the high speed camera, and a large negative charge motion observed by the fast electric field antenna (Figure 2c). The LgPl detector initially registered a burst of counts before experiencing a period of dead time lasting for ~100 μs and then registered ~5000 events over the next 70 ms before filling up its event buffer. The SmPl detector registered ~25 events during the 100 μs dead time interval in LgPl, consistent with a very bright event “turning on” the comparatively insensitive SmPl detector while simultaneously paralyzing the LgPl detector. pffi The feature seen in LgPl data consisting of a clustering of events roughly tracing out a radical sign “ ”from 0.375 to 0.384 s that approaches 2 MeV is the primary neutron signature. It is consistent with a process in which neutrons are captured by hydrogen nuclei (protons), mainly in the plastic scintillator itself, immediately releasing a gamma ray at 2.223 MeV and forming deuterium, denoted by the reaction H(n,γ)D. The gamma ray Compton scatters once in the scintillator before exiting (this being by far the most likely interaction). The 2 MeV signal represents the energy deposited by a scatter of 180°, which has a high probability of occurrence and deposits the most energy of any scatter angle. The initial dip in the energy of this feature at ~0.375 s is consistent with the detector’s high voltage power supply sagging from the large current drawn by the PMT. Similar PMT sagging and dead time effects have been observed in a plastic and NaI(Tl) scintillator during a

BOWERS ET AL. GAMMA RAYS FROM A TGF NEUTRON FLASH 10,064 Geophysical Research Letters 10.1002/2017GL075071

Figure 1. (top) Overview of experiment. GODOT and the electric field mill shown located on a rooftop (far right) 300 m away from the Uchinada wind turbine and its lightning protection tower. (bottom) Frames from high-speed video showing the progression of the upward leader from the tower.

ground observation of a TGF at the International Center for Lightning Research and Testing (ICLRT) in North Central Florida (Hare et al., 2016). To test the hypothesis that the spectral feature in the LgPl detector is consistent with this neutron process, and that a TGF could explain the production of these neutrons, reproducing both their intensity and the time- scale of their decay, we simulated the response of our instrument under the beam of a downward TGF in a model atmosphere using GEANT4, a simulation toolkit to model the passage of particles through matter (Agostinelli et al., 2003; Allison et al., 2006, 2016).

4. Monte Carlo Simulations of Photoneutron Production from a TGF 4.1. Model TGF The spectra of TGFs observed by are consistent with bremsstrahlung radiation from relativistic run- away electron avalanches (RREA) (Dwyer & Smith, 2005; Østgaard et al., 2008), and the inferred gamma ray spectrum follows closely the distribution

fEðÞexpðÞE=E0 =E (1)

where E is the electron energy and E0 is the e-folding energy (Dwyer et al., 2012).

We simulate a TGF by inputting gamma rays up to 40 MeV according to equation (1) with E0 = 6.5 MeV. This is an accurate representation of the electron bremsstrahlung spectrum resulting from a TGF electron avalanche with energy distribution fe (E) ~ exp(E/E1) with E1 = 7.3 MeV. 4.2. Three Stage Simulation To simulate the instrument response to a TGF and infer the brightness of a TGF from observations, we perform a three-stage Monte Carlo simulation, for purposes of computational economy, where the output of one stage is collected and passed to the input of the next stage in such a way that there is an effective multiplication of the number of particles being simulated. This is possible because the underlying processes are linear, and therefore, the simulation outputs are proportional to the inputs.

BOWERS ET AL. GAMMA RAYS FROM A TGF NEUTRON FLASH 10,065 Geophysical Research Letters 10.1002/2017GL075071

Figure 2. (a) The GODOT scintillation detectors (top and middle) recorded a large burst of counts coincident with a large negative change in the atmospheric electric field (AEF) (bottom). Each dot corresponds to a high energy count recorded in the detector showing the arrival time and energy. Note that a negative AEF is defined as a downward, fair- direction. (b) The 600 ms detail of counts recorded by the small plastic (SmPl) and large plastic (LgPl) detectors during the event, and simulated counts in mass models of the detectors (*SmPl, *LgPl) from a TGF at 1 km above the turbine from neutrons and gamma rays. (c) The 20 ms detail of the beginning of the event, and pixel intensity of the lightning channel at the tip of the lightning protection tower (black) and the waveform from the fast electric field antenna station (blue). In the simulations, events originating from neutron production are shown in black and from the original TGF in red.

In the first stage a number of gamma rays, denoted by N1,in, are input into a mass model of the atmosphere at a given altitude and the properties (energy, location, momentum vector, and elapsed propagation time) of all photoneutrons produced are recorded at their moment of creation; then the neutrons are removed from the simulation. In the second stage, each photoneutron from stage 1 is input 1,000 times (with the same energy, location, momentum vector, and elapsed propagation time) into the same model atmosphere and ground used in stage 1 and allowed to propagate and produce secondaries. The number of particles that arrive at the ground

at the correct horizontal distance within some annulus of area A2 is recorded, denoted by N2,out.

In the third stage, a number of particles, denoted by N3,in, representative of the particles collected at the ground within the radial annulus in stage 2, are input onto the surface of a sphere with cross-sectional area

A3 centered around a mass model of the instrument, and particles that deposit energy in the detector are recorded and the simulated number of counts N3,out determined.

The number of TGF gamma rays, NTGF, can be inferred from the observed number of counts, NDetector, and the simulated response (N1,in, N2,out, N3,in, N3,out, A2, A3) by the relation

BOWERS ET AL. GAMMA RAYS FROM A TGF NEUTRON FLASH 10,066 Geophysical Research Letters 10.1002/2017GL075071

ab

Figure 3. Difference in distribution of photoneutrons and secondaries produced at the ground by a downward beam of gamma rays from a TGF at 1.5 km, (a) omitting and (b) including thermalization in the ground.

A2 1000N1;in N3;in NTGF ¼ cosθ3;in NDetector (2) A3 N2;out N3;out

where θ is the incident angle of the particle collected at the surface of stage 2, and cosθ3;in is the arithmetic average of cos? over all particles input into stage 3, a necessary flux correction when collecting particles on a flat surface and then inputting them onto the surface of a sphere. It is important to account for the production of thermal neutrons when fast neutrons produced in the atmo- sphere are captured in the ground and thermalize (Yamashita et al., 1966). In stage 2, we define a collection volume above the ground with the same height as our instrument (~1 m) and record all particles that cross the upper or lower boundary of this volume. This means that single particles can be recorded many times by scattering back and forth through the collection volume. We argue that this is the correct way to count; each time a particle moves through the collection volume is another opportunity to interact in the detector. Figure 3 shows distributions of the kinetic energy and arrival times for TGF neutrons and their secondaries that have arrived at the ground produced by a downward TGF at 1.5 km, for cases where thermalization in the ground has and has not been modeled. Figure 3a shows the particle distribution for simulations consid- ering only thermalization in the air. Figure 3b shows the same distribution for simulations considering both thermalization in the air and in the ground. The broad horizontal feature at high energies consists primarily of gamma rays from neutron-induced nuclear reactions. By including ground thermalization, a large population of thermal neutrons is produced earlier and for longer than when ground thermalization is ignored.

Figure 4. Comparison of observed spectrum to simulations. Only the (left) ~2,000 LgPl events shown in red where the gain is reasonably well behaved are used to calculate (right) the observed spectrum, which is compared to the simulated spectra for a TGF at 0.5, 1.0, and 1.5 km.

BOWERS ET AL. GAMMA RAYS FROM A TGF NEUTRON FLASH 10,067 Geophysical Research Letters 10.1002/2017GL075071

5. Comparison of Observed Neutron Flash to Simulated Neutron Flash For our simulation we created a cylindrical volume of atmosphere, 4 km in diameter and 3 km in height, coarsely divided into 15 slabs following the International Standard Atmosphere (International Organization for Standardization, 1975). We simulated a downward TGF at 0.5 km, 1.0 km, and 1.5 km, based on the likely range of altitudes for the main negative charge center in Japanese winter thunderstorms (Saito et al., 2009), assum- ing that the TGF is produced just as the upward positive leader approaches this concentration of opposite charge (Hare et al., 2016). We input an RREA gamma ray spectrum with a 15.3° half-angle Gaussian beam shape (Hazelton et al., 2009) and tracked all photoneutrons and their secondaries Figure 5. Observed (black) and simulated (red, gold, and blue) count rates produced through the photonuclear reactions (γ,n) down to the ground versus time in the large plastic detector as a function of time (see text). and into the instrument using the three-stage method discussed above. The output of the second stage was collected in an annulus of 250–350 m around the nadir of the downward TGF to match the distance from the detector to the tower. A comparison of the simulated response to observations is shown in Figure 2c. The simulation reproduces four distinguishing features of the observation: the timescale (Figure 2b), the observed ratio of events in the LgPl and SmPl detectors after the initial paralysis of the LgPl (~75:1), the 2 MeV Compton edge in the LgPl detector from the capture of thermal neutrons on hydrogen (Figure 2c), and counts below 10 MeV corresponding to prompt gamma ray emission from neutron capture on atmo- spheric nitrogen and oxygen (Figure 2c). The observed and simulated spectra are compared from 1 to 10 MeV in Figure 4, showing agreement between the observed and simulated 2.223 MeV neutron-hydrogen capture line which appears as a 2 MeV Compton edge. The straight line drawn in blue in Figure 4 (left) traces the observed 2 MeV Compton edge from 2.29 MeV to 1.75 MeV on the unperturbed energy scale and is used introduce the same amount of gain variation into the simulated data. This gain shift dominates the LgPl’s normal resolution of 15% full width at half maximum at 2 MeV. An additional spectral component from 1.1–1.3 MeV is likely due to other induced radioactivities from materials not included in our simulation. The small feature from 3 to 4 MeV may have a similar origin or may be due to pileup (adding) of two counts in the 1–2 MeV range. Our background signal in LgPl is negligible; as can be seen in the second panel of Figure 2b, there was only a single count >1 MeV in the 60 ms before the TGF. The typical brightness of a TGF observed from space is within an order of magnitude of1017gamma rays (Mailyan et al., 2016), and this is used as our basis of comparison. By comparing the simulated to the observed count rates in LgPl (Figure 5), we can estimate the TGF brightness. The solid colored lines show the count rates per 5 ms per 1017 simulated TGF gammas for simulated TGF photonuclear product interactions in the detector, excluding direct TGF gammas. The dashed colored lines show the same count rate for direct gammas only. The observed count rate per 5 ms is shown in black. The direct gamma rays are mostly unobserved due to the (expected) paralysis of the LgPl at large count rates.

6. Conclusions For a downward TGF at 1.0 km, the observed count rate from photoneutrons is consistent with a TGF having produced ~1017 gamma rays. The total fluence of gamma rays at the ground was calculated to be approxi- mately ~100,000 photons/cm2, much larger than the total combined fluence of all TGFs observed by satellites since 1994. This would have produced such a large count rate in our large plastic detector that the MCA would have been paralyzed, consistent with our observation of counts SmPl coincident with paralysis in LgPl. This shows the importance of employing small detectors that can survive large fluxes when observing TGFs from suborbital altitudes. The total number of photoneutrons produced for this event is calculated to be ~1012 – 1013, consistent with estimates from TGF simulations (Babich et al., 2008; Carlson et al., 2010). 3 In previous observations of neutrons associated with lightning, the detectors were large BF3 and He filled proportional counters that have been considered sensitive only to thermal neutrons but have been

BOWERS ET AL. GAMMA RAYS FROM A TGF NEUTRON FLASH 10,068 Geophysical Research Letters 10.1002/2017GL075071

recently been shown to be susceptible to contamination from high energy gamma rays (Tsuchiya et al., 2012). We show that by using relatively small scintillation detectors it is possible to observe unique spectral and temporal gamma ray signatures of neutron production. Our simulations show that ground thermalization is important when considering the instrument response and effective radiation dose from neutron flashes. Additionally, the ejection of neutrons from O and N by gamma rays in TGFs implies the production of positron-emitting radioisotopes (13N, 15O) associated with TGFs in a quantity equal to the estimated number of neutrons produced, ~1012 – 1013. Two recent papers (Dwyer et al., 2015; Umemoto et al., 2016) discussed transient positron annihilation in the atmosphere at timescales longer than that of a TGF, possibly from these isotopes, with timescales of ~200 ms and ~30 s, respectively. In Dwyer et al. (2015), it was stated that the positrons from the radioactivity were repelled from the aircraft from which the observation was made by its positive charge, which was temporarily removed by the discharge during the short time of the observa- tion. The ~30 s decay of the signal observed by Umemoto et al. (2016) in the wake of a probable TGF on the ground could have been caused by the wind blowing the radioisotopes away from the vicinity of the detector. While both of these explanations are tentative, our clear detection of neutron signatures supports them by implying the presence of the radioisotopes left behind by the ejection of the neutrons from N and O nuclei. The absence of a clear annihilation signal in our large NaI detector in the seconds after our neutron event suggests that the event seen by Umemoto et al. (2016), whose detector was completely paralyzed during the gamma and neutron stages, was either much closer, intrinsically much brighter, or both.

Acknowledgments References We thank Shusaku Takahashi and Akiko … — Ishikawa for helping deploy and main- Agostinelli, S., Allison, J., Amako, K., Apostolakis, J., Araujo, H., Arce, P., Zschiesche, D. (2003). GEANT4 A simulation toolkit. Nuclear – tain GODOT in Japan and Michael Instruments and Methods in Physics Research A, 506, 250 303. … McCarthy for discussions on multistage Allison, J., Amako, K., Apostolakis, J., Araujo, H., Arce Dubois, P., Asai, M., Yoshida, H. (2006). Geant4 developments and applications. IEEE – Monte Carlo simulations. Deployment Transactions on Nuclear Science, 53, 270 278. … of GODOT to Japan was made possible Allison, J., Amako, K., Apostolakis, J., Arce, P., Asai, M., Aso, T., Yoshida, H. (2016). Recent developments in GEANT4. Nuclear Instruments and – through the FY2015 Japan Society of Methods in Physics Research A, 835, 186 225. the Promotion of Science (JSPS) Babich, L. P. (2006). Generation of neutrons in giant upward atmospheric discharges. Soviet Journal of Experimental and Theoretical Physics – Postdoctoral Fellowship (short term) for Letters, 84, 285 288. – North American Researchers (PE 15757). Babich, L. P. (2007). Neutron generation mechanism correlated with lightning discharges. Geomagnetism and Aeronomy, 47, 664 670. fl This research was funded by National Babich, L. P., Kudryavtsev, A. Y., Kudryavtseva, M. L., & Kutsyk, I. M. (2008). Atmospheric gamma-ray and neutron ashes. Soviet Journal of – Science Foundation grants Experimental and Theoretical Physics, 106,65 76. … AGS-1160226, AGS-1613028, and Briggs, M. S., Fishman, G. J., Connaughton, V., Bhat, P. N., Paciesas, W. S., Preece, R. D., Chekhtman, A. (2010). First results on terrestrial fl AGS-1612466. Our GEANT4 software is gamma ray ashes from the Fermi Gamma-ray Burst Monitor. Journal of Geophysical Research, 115, A07323. https://doi.org/10.1029/ available at https://github.com/ 2009JA015242 fl gsbowers/TGF-Neutron-Flash, its output Carlson, B. E., Lehtinen, N. G., & Inan, U. S. (2010). Neutron production in terrestrial gamma ray ashes. Journal of Geophysical Research, 115, at http://www.godot.jp/notes/151203/ A00E19. https://doi.org/10.1029/2009JA014696 fl 202029/simulations/, and GODOT’s data Dwyer, J. R., & Smith, D. M. (2005). A comparison between Monte Carlo simulations of runaway breakdown and terrestrial gamma-ray ash at http://www.godot.jp/notes/151203/ observations. Geophysical Research Letters, 32, L22804. https://doi.org/10.1029/2005GL023848 fl 202029/observations/. Dwyer, J. R., Smith, D. M., & Cummer, S. A. (2012). High-energy : Terrestrial gamma-ray ashes and related phenomena. Space Science Reviews, 173, 133–196. Dwyer, J. R., Smith, D. M., Hazelton, B. J., Grefenstette, B. W., Kelley, N. A., Lowell, A. W., … Rassoul, H. K. (2015). Positron clouds within thunderstorms. Journal of Plasma Physics, 81. https://doi.org/10.1017/S0022377815000549 Fishman, G. J., Bhat, P. N., Mallozzi, R., Horack, J. M., Koshut, T., Kouveliotou, C., … Christian, H. J. (1994). Discovery of intense gamma-ray flashes of atmospheric origin. Science, 264, 1313–1316. Gurevich, A. V., Almenova, A. . M., Antonova, V. . P., Chubenko, A. . P., Karashtin, A. . N., Kryakunova, O. . N., … Zybin, K. . P. (2016). Observations of high-energy radiation during thunderstorms at Tien-Shan. Physical Review D, 94(2), 023003. Gurevich, A. V., Antonova, V. P., Chubenko, A. P., Karashtin, A. N., Mitko, G. G., Ptitsyn, M. O., … Zybin, K. P. (2012). Strong flux of low-energy neutrons produced by thunderstorms. Physical Review Letters, 108(12), 125001. Hare, B. M., Uman, M. A., Dwyer, J. R., Jordan, D. M., Biggerstaff, M. I., Caicedo, J. A., … Bozarth, A. (2016). Ground-level observation of a ter- restrial gamma ray flash initiated by a triggered lightning. Journal of Geophysical Research: Atmospheres, 121, 6511–6533. https://doi.org/ 10.1002/2015JD024426 Hazelton, B. J., Grefenstette, B. W., Smith, D. M., Dwyer, J. R., Shao, X.-M., Cummer, S. A., … Holzworth, R. H. (2009). Spectral dependence of terrestrial gamma-ray flashes on source distance. Geophysical Research Letters, 36, L01108. https://doi.org/10.1029/2008GL035906 Ishtiaq, P. M., Mufti, S., Darzi, M. A., Mir, T. A., & Shah, G. N. (2016). Observation of 2.45 MeV neutrons correlated with natural atmospheric lightning discharges by Lead-Free Gulmarg Neutron Monitor. Journal of Geophysical Research: Atmospheres, 121, 692–703. https://doi.org/ 10.1002/2015JD023343 International Organization for Standardization (1975). Standard atmosphere. Geneva, Switzerland: International Organization for Standardization. Krehbiel, P. R. (1986). The electrical structure of thunderstorms. In E. P. Krider & P. R. Krehbiel (Eds.), The Earth’s Electrical Environment (pp. 96–113). Washington, DC: The National Academies Press. Libby, L. M., & Lukens, H. R. (1973). Production of radiocarbon in tree rings by lightning bolts. Journal of Geophysical Research, 78, 5902–5903. https://doi.org/10.1029/JB078i026p05902 Mailyan, B. G., Briggs, M. S., Cramer, E. S., Fitzpatrick, G., Roberts, O. J., Stanbro, M., … Dwyer, J. R. (2016). The spectroscopy of individual terrestrial gamma-ray flashes: Constraining the source properties. Journal of Geophysical Research: Space Physics, 121, 11,346–11,363. https://doi.org/10.1002/2016JA022702

BOWERS ET AL. GAMMA RAYS FROM A TGF NEUTRON FLASH 10,069 Geophysical Research Letters 10.1002/2017GL075071

Marisaldi, M., Fuschino, F., Labanti, C., Galli, M., Longo, F., Del Monte, E., … Salotti, L. (2010). Detection of terrestrial gamma ray flashes up to 40 MeV by the AGILE satellite. Journal of Geophysical Research, 115, A00E13. https://doi.org/10.1029/2009JA014502 Martin, I. M., & Alves, M. A. (2010). Observation of a possible neutron burst associated with a lightning discharge? Journal of Geophysical Research, 115, A00E11. https://doi.org/10.1029/2009JA014498 Østgaard, N., Gjesteland, T., Stadsnes, J., Connell, P. H., & Carlson, B. (2008). Production altitude and time delays of the terrestrial gamma flashes: Revisiting the Burst and Transient Source Experiment spectra. Journal of Geophysical Research, 113, A02307. https://doi.org/ 10.1029/2007JA012618 Saito, M., Ishii, M., Kawamura, H., & Shindo, T. (2009). Location of negative charge associated with continuing current of upward lightning flash in winter. IEEJ Transactions on Power and Energy, 129, 929–934. Shah, G. N., Razdan, H., Bhat, C. L., & Ali, Q. M. (1985). Neutron generation in lightning bolts. Nature, 313(6005), 773–775. Shyam, A., & Kaushik, T. C. (1999). Observation of neutron bursts associated with atmospheric lightning discharge. Journal of Geophysical Research, 104, 6867–6870. https://doi.org/10.1029/98JA02683 Smith, D. M., Lopez, L. I., Lin, R. P., & Barrington-Leigh, C. P. (2005). Terrestrial gamma-ray flashes observed up to 20 MeV. Science, 307, 1085–1088. Tsuchiya, H., Hibino, K., Kawata, K., Hotta, N., Tateyama, N., Ohnishi, M., … Makishima, K. (2012). Observation of thundercloud-related gamma rays and neutrons in Tibet. Physical Review D, 85. https://doi.org/10.1103/PhysRevD.85.092006 Umemoto, D., Tsuchiya, H., Enoto, T., Yamada, S., Yuasa, T., Kawaharada, M., … Makishima, K. (2016). On-ground detection of an electron- positron annihilation line from thunderclouds. Physical Review E, 93. https://doi.org/10.1103/PhysRevE.93.021201 Ursi, A., Guidorzi, C., Marisaldi, M., Sarria, D., & Frontera, F. (2017). Terrestrial gamma-ray flashes in the BeppoSAX data archive. Journal of Atmospheric and Solar-Terrestrial Physics, 156,50–56. Wang, D., & Takagi, N. (2012). Characteristics of winter lightning that occurred on a windmill and its lightning protection tower in Japan. IEEJ Transactions on Power and Energy, 132, 568–572. Yamashita, M., Stephens, L. D., & Patterson, H. W. (1966). Cosmic-ray-produced neutrons at ground level: Neutron production rate and flux distribution. Journal of Geophysical Research, 71, 3817–3834.

BOWERS ET AL. GAMMA RAYS FROM A TGF NEUTRON FLASH 10,070