Downward Terrestrial Gamma-Ray Flashes

• Downward terrestrial gamma-ray flashes are brief, intense bursts of MeV gamma radiation produced by bremsstrahlung from high-energy electrons during lightning breakdown.
• Observations reveal microsecond-scale temporal structures and asymmetric spatial patterns that shed light on lightning leader development and discharge dynamics.
• Multi-instrument methodologies combining ground arrays, Cherenkov detectors, and orbital data provide critical insights into source altitude, beam geometry, and particle acceleration mechanisms.

Searching arXiv for recent and foundational papers on downward terrestrial gamma-ray flashes. {"query":"Downward terrestrial gamma-ray flashes Pierre Auger Telescope Array TGFs", "max_results": 10} Downward terrestrial gamma-ray flashes (TGFs) are brief, intense bursts of MeV gamma radiation produced inside thunderstorms, with the photon beam directed toward the ground. In ground and near-storm observations they appear as sub-millisecond to millisecond high-energy transients associated with lightning activity, especially negative cloud-to-ground flashes and downward negative leader development, and are generally interpreted as bremsstrahlung from high-energy electrons accelerated in strong thunderstorm electric fields (Abbasi et al., 2017, Colalillo et al., 21 Jul 2025). Although the best-known TGF population was established by satellite observations of upward-directed emission, downward TGFs have become a distinct observational and theoretical subject because they provide direct access to the source region, resolve fine temporal structure that is blurred in orbital measurements, and probe the low-altitude branch of thunderstorm particle acceleration (Wada et al., 2019, Belz et al., 2020).

1. Observational definition and phenomenology

Ground-based and near-storm measurements define downward TGFs as intense, downward-beamed bursts of MeV gamma radiation produced in the first milliseconds of lightning development and observed below the source region. In the Telescope Array Surface Detector (TASD), the showers arrived in a sequence of 2–5 short-duration ($\le 10~\mu\mathrm{s}$) bursts over time intervals of several hundred microseconds, and originated at an altitude of $\simeq 3\text{–}5$ kilometers above ground level during the first 1–2 ms of downward negative leader breakdown at the beginning of cloud-to-ground lightning flashes (Abbasi et al., 2017). In the Pierre Auger Observatory, downward TGFs are treated as high-energy bursts of particles during thunderstorms, with durations from tens of microseconds up to milliseconds and signals in water-Cherenkov detectors lasting $>10~\mu\mathrm{s}$, which is described as an order of magnitude longer than typical extensive air shower signals (Weitz, 2024, Colalillo et al., 21 Jul 2025).

A particularly detailed winter-thunderstorm event in Japan showed a strong burst of gamma rays with energies up to $\sim 10~\mathrm{MeV}$ coincident with a lightning discharge. Its leading part was resolved into four intense gamma-ray bunches, each coincident with a low-frequency radio pulse, separated by 0.7–1.5 ms, with a duration of $\ll 1$ ms each (Wada et al., 2019). Earlier rooftop observations by TETRA established one of the first systematic ground-level catalogs of similar bursts: after 2.6 years of observation, twenty-four events with durations 0.02–4.2 msec were detected associated with nearby lightning, including three coincident events observed by detectors separated by $\sim 1000$ m (Ringuette et al., 2013). Across these datasets, the recurring observational signature is a compact, lightning-correlated, high-fluence burst whose timescale is much shorter than gamma-ray glows and whose hardness exceeds that of ordinary thunderstorm background radiation.

2. Lightning context and source geometry

The strongest direct evidence places downward TGFs in the earliest stages of negative lightning breakdown. In TASD data, all 10 TGF-producing flashes examined in detail were negative cloud-to-ground flashes, starting with downward propagating negative leaders, and the gamma-ray source heights were inferred as $\sim 3\text{–}5$ km above ground (Abbasi et al., 2017). Close, high-resolution observations later showed that downward-directed TGFs occur during strong initial breakdown pulses in the first few milliseconds of negative cloud-to-ground and low-altitude intracloud flashes, and that those initial breakdown pulses are produced by fast negative breakdown (Belz et al., 2020). In the Kashiwazaki-Kariwa winter event, Monte Carlo simulations constrained the source altitude to $2500 \pm 500~\mathrm{m}$, again locating the source in the lower part of a thundercloud (Wada et al., 2019).

Ground arrays have also begun to reveal the lateral morphology of the beam. At Auger, many stations are triggered nearly simultaneously over a footprint of order $200~\mathrm{km}^2$, and reconstructed source heights are consistent with thundercloud bases at 1–2 km above ground (Colalillo et al., 21 Jul 2025). A notable result is that Auger TGFs are asymmetric about their vertical axis and present two peaks in the azimuthal direction, suggesting a complex source different from the initially hypothesized downward beam (Colalillo et al., 21 Jul 2025). This does not by itself specify a unique source geometry, but it disfavors a simple axisymmetric cone.

The spatial relationship to lightning is correspondingly tight. In the winter-thunderstorm event, the first gamma-ray bunch occurred during the last stage of stepped-leader development and before the return stroke reported by the lightning network, while the later bunches coincided with additional low-frequency pulses (Wada et al., 2019). In Auger, one burst coincided with a negative cloud-to-ground lightning flash and with a small peak in the return-stroke waveform, called a reflection pulse (Colalillo et al., 21 Jul 2025). These patterns support the view that downward TGFs are coupled to specific leader or initial-breakdown episodes rather than to the entire lightning flash in a generic sense.

3. Radiation physics and source mechanisms

The baseline physical interpretation is that TGFs are bremsstrahlung from high-energy electrons propagating through air. Instrument and theory papers describe these electrons as arising either from relativistic runaway electron avalanches (RREAs) driven by large-scale thunderstorm electric fields or from local intense fields near lightning leaders (Østgaard et al., 2019). In the RREA description, the source photon spectrum is commonly written as

$$\frac{dN_\gamma}{dE} \propto \frac{1}{E}\,\exp\!\left(-\frac{E}{7.3\ \mathrm{MeV}}\right),$$

which was used in detailed AGILE end-to-end simulations and is treated as the standard RREA bremsstrahlung form (Marisaldi et al., 2021). In a complementary formulation, the avalanche condition is tied to a threshold electric field $\simeq 3\text{–}5$0, with practical runaway requiring fields roughly 20–30% above that nominal threshold because of Coulomb and Møller scattering (Chilingarian, 16 May 2025).

Close lightning measurements argue for a specifically leader-linked realization of this physics in downward TGFs. The 2020 origin study reports that the RREAs responsible for producing the TGFs are initiated by embedded spark-like transient conducting events within the fast streamer system, and potentially also by individual fast streamers themselves (Belz et al., 2020). That paper places the bursts during strong initial breakdown pulses generated by fast negative breakdown, with downward propagation speeds $\simeq 3\text{–}5$1, and estimates potential drops of order $\simeq 3\text{–}5$2 across the fast-breakdown region (Belz et al., 2020). In parallel, 2.5D cylindrical Monte Carlo streamer simulations show that streamers in preionized and perturbed air accelerate more efficiently than in non-ionized and uniform air, with air perturbation dominating the streamer acceleration, and that in perturbed air the production rate of runaway electrons varies from $\simeq 3\text{–}5$3 to $\simeq 3\text{–}5$4 with maximum electron energies from some hundreds of eV up to some hundreds of keV (Köhn et al., 2019). This directly supports a streamer-corona contribution to the seed-electron problem.

The relative roles of leader-tip physics and large-scale RREA remain debated. One proposed synthesis treats TGFs, gamma-ray glows, and thunderstorm ground enhancements as different manifestations of the same runaway process at different atmospheric depths, and explicitly argues that “downward TGFs” is a misleading label for intense lower-dipole runaway activity (Chilingarian, 16 May 2025). A plausible implication is that some events classified observationally as downward TGFs may overlap physically with short, intense TGE-like bursts. Even so, the observation that downward TGFs occur during strong initial breakdown pulses in negative flashes remains a robust empirical constraint (Belz et al., 2020).

4. Temporal structure, spectral properties, and photonuclear signatures

Downward TGFs resolve fine temporal structure with unusual clarity because the source-to-detector path is short. In TASD, each flash exhibits 2–5 distinct bursts, with overall TGF activity spanning 87–551 $\simeq 3\text{–}5$5, and composite waveforms show that each trigger resolves into multiple individual pulses, mostly $\simeq 3\text{–}5$6 wide (Abbasi et al., 2017). In the Kashiwazaki-Kariwa event, the four bunches were separated by 0.7–1.5 ms and were embedded in a sub-second burst whose exponential tail had decay constants of 59, 47, and 48 ms in the three BGO detectors (Wada et al., 2019). More generally, general TGF timing studies indicate that the peak in a RHESSI TGF light curve can occur 230 $\simeq 3\text{–}5$7s before the WWLLN time, while gamma and optical emissions remain simultaneous within about $\simeq 3\text{–}5$8 ms (Gjesteland et al., 2017). This suggests that the high-energy burst is tightly coupled to the main lightning development but can slightly precede the dominant VLF pulse.

The spectral evidence shows that downward TGFs are multi-MeV phenomena. In TASD, the observed shower waveforms and detector response imply that the showers consist primarily of downward-beamed gamma radiation, and GEANT4 simulations indicate primary source fluxes of $\simeq 3\text{–}5$9 photons for $>10~\mu\mathrm{s}$0 half-angle beams (Abbasi et al., 2017). The paper adopts the standard RREA photon spectrum

$>10~\mu\mathrm{s}$1

for those simulations (Abbasi et al., 2017). TETRA detected ground-level bursts in the range 50 keV to over 2 MeV, and the composite deposited-energy spectrum for the lightning-associated events is fit between 200 keV and 1.2 MeV by $>10~\mu\mathrm{s}$2 with $>10~\mu\mathrm{s}$3 (Ringuette et al., 2013). In the Japanese winter event, the inferred number of avalanche electrons above 1 MeV was $>10~\mu\mathrm{s}$4, with a total number of bremsstrahlung photons above 1 MeV of $>10~\mu\mathrm{s}$5, comparable to upward TGFs (Wada et al., 2019).

Photonuclear signatures are especially important because they demonstrate photons above the $>10~\mu\mathrm{s}$6 MeV threshold for atmospheric photonuclear reactions. In the winter-thunderstorm case, the sub-second tail was interpreted as neutron capture gamma rays, and a $>10~\mu\mathrm{s}$7 s afterglow was associated with positron annihilation from $>10~\mu\mathrm{s}$8-decaying nuclei created by photonuclear reactions (Wada et al., 2019). A separate modeling study shows that the fraction of TGF photons above 10.5 MeV is $>10~\mu\mathrm{s}$9 for $\sim 10~\mathrm{MeV}$0, and that a TGF with $\sim 10~\mathrm{MeV}$1 photons above 1 MeV can generate $\sim 10~\mathrm{MeV}$2 neutrons through atmospheric photonuclear reactions (Diniz et al., 2021). That paper further argues that $\sim 10~\mathrm{MeV}$3 and $\sim 10~\mathrm{MeV}$4 produced by those reactions can inject positrons over minutes, potentially reseeding RREAs and linking a prompt TGF to later glows or TGEs (Diniz et al., 2021). A plausible implication is that some downward TGFs are not isolated flashes but the first stage of a longer high-energy atmospheric chain.

5. Detection techniques and multi-instrument methodology

Downward TGF research is fundamentally instrument-driven because the phenomenon spans microsecond gamma pulses, millisecond radio structure, and storm-scale geometry. Ground arrays have provided the clearest downward detections. TASD uses 507 surface detector units on a 1.2 km grid over $\sim 10~\mathrm{MeV}$5, with two 3 m$\sim 10~\mathrm{MeV}$6 plastic scintillator planes per unit and 50 MHz waveform sampling (Abbasi et al., 2017). The Pierre Auger Surface Detector consists of 1600 water-Cherenkov detectors over $\sim 10~\mathrm{MeV}$7, and its 40 MHz FADC readout makes it sensitive to long, thunderstorm-related signals that differ from ordinary cosmic-ray showers (Colalillo et al., 21 Jul 2025). TETRA uses four rooftop detector boxes forming an approximate rectangle of $\sim 10~\mathrm{MeV}$8 m, each box containing three NaI(Tl) scintillator plates, and triggered on 2 ms bins exceeding the daily mean by $\sim 10~\mathrm{MeV}$9 (Ringuette et al., 2013). At Kashiwazaki-Kariwa, BGO scintillators saturated on the prompt burst, but ionization chambers at nine monitoring posts preserved the integrated dose information needed to infer the source fluence (Wada et al., 2019).

Dedicated mountain and airborne systems are being built specifically for local and downward events. Gamma-Flash is installed at the Climatic Observatory “O. Vittori” on Mt. Cimone at 2165 m above sea level and combines five $\ll 1$0-ray detectors, three neutron detectors, a lightning antenna with 480 km detection capability, and a waveform-based data acquisition system that produces almost 40 GB of DL0 raw waveforms per day for each detector, reduced to 2 GB at the DL1 data level (Bulgarelli et al., 2023). The same project plans an aircraft payload to observe thunderstorms in the air, explicitly targeting the regime where downward TGFs and glows can be sampled near the source (Bulgarelli et al., 2023).

Multi-instrument lightning mapping is now a central methodological requirement. The Pierre Auger Observatory is adding an interferometric lightning detection array using 11 modified AERA stations operating in the bandwidth between 30–80 MHz, with baselines from 58 m to 66 km, in order to precisely measure lightning stepped leaders in 3D (Weitz, 2024). The 30–80 MHz bandwidth is stated to yield a spatial resolution in the meter range, and long-trace readout of up to 1 second per station has been implemented for lightning work (Weitz, 2024). These radio data are intended to be combined with the direct location of downward TGFs by the water-Cherenkov detectors, potentially providing the first quantitative relation between a TGF source position and the triggering lightning (Weitz, 2024).

Orbital instruments remain essential, even though they do not directly observe the downward branch. ASIM/MXGS is mounted on the ISS and directly measures only the component that escapes upward, but by combining localization, spectral hardness, and source altitude it constrains the downward fluence indirectly (Østgaard et al., 2019). AGILE localized $\ll 1$1 MeV TGF photons to within $\ll 1$2 km of the subsatellite point, showing that high-energy photons reach orbit with limited scattering or attenuation and providing direct constraints on beaming and source geometry (Marisaldi et al., 2010). This orbital context is important because downward TGF models must remain consistent with the upward component seen from space.

6. Population statistics, observational biases, and open controversies

The observed downward-TGF rate is low, but every available dataset is strongly biased by geometry, dead time, and trigger logic. TETRA reported 24 lightning-associated events in 2.6 years, with nine of them occurring within 6 ms and 3 miles of negative polarity cloud-to-ground lightning strokes with measured currents in excess of 20 kA (Ringuette et al., 2013). At Auger, 22 downward TGF events were collected before 2017, corresponding to a detection rate of less than 2 events/year, even though the known lightning rate and a lightning/TGF ratio of approximately $\ll 1$3 imply an expectation of at least $\ll 1$4 downward TGFs per year; the discrepancy is attributed to the SD electronics, triggers, CDAS buffer, and post-acquisition processing, which were optimized for cosmic-ray showers, not TGFs (Colalillo et al., 21 Jul 2025). This strongly suggests that current ground-based samples are incomplete.

Spaceborne statistics reinforce the same point from the upward side. A stacking analysis of RHESSI data identified at least 141 and probably as many as 191 weak TGFs that were not part of the second RHESSI data catalogue, supporting the view that a new population of weak TGFs exists below standard search thresholds (Østgaard et al., 2016). That paper argues that lower-altitude events, more distant events, and intrinsically weak events are all strongly attenuated before reaching orbit, which is directly relevant to downward TGFs because downward or low-altitude events are pushed into the weak tail of the satellite fluence distribution (Østgaard et al., 2016). A related RHESSI study showed that correcting for source altitude and distance softens the inferred source fluence distribution from $\ll 1$5 to $\ll 1$6, emphasizing how severely atmospheric column depth biases source inference (Nisi et al., 2016). A plausible implication is that many downward TGFs remain invisible to satellites and that many weak upward TGFs remain invisible to ground systems unless the geometry is favorable.

There are also active interpretive controversies. One concerns the dominant acceleration site: leader-tip and streamer-corona models are supported by the close association with initial breakdown pulses and fast negative breakdown (Belz et al., 2020, Köhn et al., 2019), whereas a different “change of paradigm” paper argues that downward TGFs should be reclassified as short, intense lower-atmosphere manifestations of the same runaway process that produces TGEs and glows, and explicitly proposes to reject “downward TGFs” as a distinct physical category (Chilingarian, 16 May 2025). Another unresolved issue is timing at the sub-millisecond level: gamma-ray peaks can precede the main VLF pulse by hundreds of microseconds, but optical timing remains uncertain at the $\ll 1$7millisecond level, so the exact ordering of gamma, radio, and optical signatures is not yet settled (Gjesteland et al., 2017). Finally, beam geometry remains uncertain: Auger’s asymmetric azimuthal structure and two peaks suggest a complex source, but current simulations with normal, wide, or isotropic cones do not fully reproduce the observed radial trend (Colalillo et al., 21 Jul 2025).

Downward TGFs therefore occupy a strategically important position in atmospheric high-energy physics. They are empirically established as intense, lightning-correlated, downward-directed gamma-ray bursts, yet they also expose the limits of present models of leader development, streamer energetics, avalanche seeding, and atmospheric transport. Their study is increasingly defined by combined gamma-ray, neutron, radio, electric-field, and orbital measurements, because only that combination can resolve how much of the phenomenon belongs to local leader microphysics and how much reflects the larger-scale runaway structure of the thundercloud.