Laboratory investigation of earthquake lightning due to landslide

Several laboratory experimental studies of photoemission characteristics of rocks as they fracture have been conducted to elucidate the mechanism of earthquake lightning (EQL). In most of these studies, granite, which exhibits remarkable photoemission, is widely used to explain the mechanism attributed to the exoelectron emission effect or piezo-induced effect of quartz, a constituent mineral of granite. Photoemission induced by rock fracturing has been observed prominently during landslides caused by earthquakes. According to the literature on historical earthquakes, landslide-induced EQL was witnessed even at locations where quartz was not present. In this study, we therefore studied the physicochemical characteristics of photoemissions associated with impact shear fractures in various rocks, selected based on historical earthquake records of landslide EQL. As a result, we could highlight various physicochemical processes associated with the frictional heating generated during shear impacts between rocks.


Introduction
Fracto-emission is a phenomenon in which charged particles and photons are emitted from a newly created fracture surface upon the rupture of hard and brittle solids, such as ceramics, polymers, and rocks (e.g., Dickinson et al. 1981Dickinson et al. , 1988Enomoto and Chaudhri 1993). In the area of earthquake research, photoemission from fracturing rocks has been observed in relation to earthquake lightning (EQL), which is often observed before, during, or after an earthquake (Musya 1932;Enomoto 1998). In laboratory experiments, granite containing quartz, which exhibits a piezoelectric effect, has been widely used due to the ease of detecting its photoemission (Brady and Rowell 1986;Kawaguchi 1998;Kato et al. 2010). These results led to proposed models for the exoelectron bombardment mechanism (Brady et al. 1986) or piezoelectricinduced effect (Ikeya and Takagi 1996).
Lightning from fracturing rocks can be directly observed not when rock rupture occurs at a deep focal zone, but when landslides due to mountain collapse occur in association with earthquakes. Because the stratum of the Japanese archipelago is complex and has many active faults, historical earthquake records in Japan reveal several cases where EQL was eye-witnessed in association with landslides. However, these mountains were not necessarily composed of granite, but other kinds of rock. Therefore, we conducted impact fracture experiments on the rocks that composed the mountain strata based on the evidence of landslide EQL found in historical earthquake records, and investigated photoemission characteristics in terms of the physicochemical nature of these rocks.

Historical records on landslide EQL
We surveyed the descriptions of EQL in the Japan Historical Earthquake Archives (e.g., ERI 1981ERI -1994Musya 1932) to find at least 5 eye-witness accounts out of 55 reports of EQL caused by landslides (Enomoto 1998). In consideration of the geology and the lightning characteristics for these seven events, the three landslide EQL events listed below were selected to clarify the mechanism. Figure 1a shows the locations of three selected Open Access *Correspondence: enomoto@shinshu-u.ac.jp 1 Fii, Ueda Campus, Shinshu University, 3-15-1 Tokida, Ueda, Nagano 386-8567, Japan Full list of author information is available at the end of the article Enomoto et al. Earth, Planets and Space (2020) Fig. 1b. Because it was witnessed as a severe pyroclastic flow, people at that time thought Mayuyama had a volcanic eruption. Later studies, however, indicated that this event was due to landslide collapse of Mt. Mayu-yama (Ohta 1969;Katayama 1974). 2. In the 1847 M7.4 Nagano (Zenkoji) earthquake, which occurred at about 21:30 LT on 8 May, the quake caused many landslides and collapses, mainly in the western mountains of Nagano City. In particular, a collapse that occurred on the southwest side of Mt. Iwakura (764 m above sea level), about 12 km southwest of Nagano City, was the largest. Iwakura is mainly sandstone and conglomerate, and rhyolite tuff breccia is sandwiched in irregular forms (Nagano-ken 1997). It should be noted that when the collapse occurred, every corner inside the houses at the foot of Aniwa village were illuminated as bright as daytime. 3. In the 1909 M6.7 Anegawa (Kono) earthquake in Shiga prefecture, a limestone cliff collapsed on the west side of Mt. Ibuki (1377 m above sea level), dust flew up, the mountain summit could not be seen from the foot, and the area became dark, even in the daytime, 15:31 LT. At that time, many of the residents at the foot witnessed EQL. They, therefore, thought that the earthquake was caused by a volcanic eructation of Mt. Ibuki. The survey report by Koto (1910) mentioned that the cause of EQL was not well understood. His sketch of a geological structure, that is, Fig. 1  limestone of the collapsed west side of Mt. Ibuki, is shown in Fig. 1d.
Of the 5 landslide EQL records mentioned above, the other two are landslide EQL of Mt. Takasaki (628 m, Oita) in the 1 Sept. 1596 M7.0 ± ¼ Keicho-Bungo earthquake and that of Mt Maru-yama (226 m, Hokkaido) in the 29 May 1938 M6.8 Kussyaro-ko earthquake. Both mountains are lava domes with a history of volcanic activity in the past. The mechanism of the landslide EQL of these mountains might, therefore, be similar to the that of Mayu-yama in the 1972 Unzen earthquake as noted in 1).

Charpy impact tests
Because we expect landslide EQL to be caused by photoemission that occurs due to collisional shear fractures between rocks as they fall down a slope, an impact shear fracture type test is more appropriate as an experimental method to simulate the process than an uniaxial compression fracture test or a three-point bending fracture test. We therefore used a Charpy impact tester as shown in Fig. 1e-g. This test in industry is a standardized high strain-rate test which measures a material resistance for fracture (e.g., Leis 2013). The present Charpy impact test involves striking a rock sample fixed in a bottom holder with a pendulum containing the same kind of rock in its holder at a predetermined height. The as-received rock samples were cut to an appropriate size so that they could be fixed to both holders of the Charpy impact tester. As shown in Fig. 1g, the two rock samples were set to engage to a depth of 2-5 mm wide at impact. The arm length was 400 mm, and the swing angle a was 150°. The speed v and the kinetic energy ½Mv 2 , where M is pendulum mass, at impact were 4.0 m/s and 25 J, respectively. The contact time during impact was about 1.3 ms. To carry out photography and spectroscopy, the test was performed in the room air with the entire test apparatus enclosed by a dark curtain so as to prevent light leakage.
Light photoemitted during the collision was photographed by a digital camera (Canon EOS Kiss X7i) with an ISO sensitivity of 25,600 and the shutter was opened for 1 s from just before to just after the collision. Meanwhile, the spectrum measurements were conducted by means of a spectroradiometer (CS-2000A, Konica Minolta) with a close-up lens for measurement of smaller areas where photoemission occurred, in the wavelength range from 380 to 780 nm and at a wavelength resolution of 0.9 nm/pixel. The spectrometer was also used to measure the spectrum of thermoluminescence from limestone.
To examine flammable gases such as hydrocarbons and H 2 S contained in the rock samples, we employed thermal desorption spectroscopy-mass spectrometry (TDS-MS). A rock sample was cut to a size of 15 × 15 × 4 mm and heated at a rate of 20 °C/min using an infrared heating furnace in a vacuum vessel. Then, the released gases were analyzed using a quadrupole mass spectrometer. Note that since the impact period, comparable to photoemission period, is 1.3 ms which is much shorter than the exposure period of 1 s, the intensity distribution of photoemission is less affected by the impact hammer speed. Granite showed an intense whitish spark streak during impact (Fig. 2a). Pyroclastic rock produced a glowing gaseous red lightning around the hotspot (Fig. 2b). Biotite-bearing rhyolite showed an orangecolored lightning streak (Fig. 2c). Limestone showed a weak red color on the newly created fracture surface (Fig. 2d). Serpentinite showed no photoemission, and thus it is not shown in the figure. The red, green, and blue (RGB) intensities on a scale of 0 to 255 for photoemissions from the brightest part at impact for each rock and from a position off the hotspot for granite and rhyolite are also shown in these figures. The order of the photoemission intensity was granite > biotite rhyolite > pyroclastic rock > limestone, with no emission from serpentinite (not shown In Fig. 2). The difference in the shape, color, and RGB ratio of the photoemission as seen in Fig. 2 suggests that the physicochemical mechanism differs from rock to rock. Next, the mechanisms of photoemission from each rock were investigated, and are discussed in detail.

Granite
Photoemission from granite was the brightest among the tested rocks, and was the only one that could be captured upon impact using a spectroradiometer, as shown in Fig. 3a. The spectral intensity increased continuously as the wavelength became longer, which is characteristic of black-body radiation: the peak of the spectrum should be in the far-infrared region, but the temperature can be inferred from the intensity ratio for two wavelengths, λ(740 nm)/λ(620 nm) = 3.6, assuming that the tail can be expressed by Planck's law (Fig. 3b). We used the results in Fig. 3c to estimate the frictional temperature during collision to be 1750 K (= 1477 °C).
The TDS-MS analysis showed that any flammable gases are less contained in the granite; therefore, photoemission due to degassing is not possible. At such a high temperature, minute fragments of non-stoichiometric oxides of rock constituents would be ignited when the fragments underwent oxidation reaction and show a continuous radiation spectrum.
In contrast, many spectral lines appeared in the short wavelength range around 380-420 nm, which might be attributed to photoemission from vibrationally or electrically excited states of the constituent elements of the granite; that is, when the temperature rises to 250-400 °C, lone-pair electrons captured by lattice defects in the granite are thermally released. In a previous experimental study , measurements of thermally stimulated exoelectrons (TSEEs) emitted from constituent minerals, including quartz, feldspar, and biotite in granite (Inada, Ibaraki), indicated notable emissions from biotite. The energy of TSEEs should be as low as about 1 eV, but when the rock breaks, an intense electric field is likely created in the small gap between both cracked new faces due to charge separation (Enomoto and Hashimoto 1990;Enomoto and Chaudhri 1993). The electric field can accelerate exoelectrons to gain enough energy to ionize atmospheric molecules and other species in the rock minerals, resulting in the generation of a plasma streak (Brady and Rowell 1986;Enomoto and Hashimoto 1990). That is, according to Paschen's law, which gives the voltage necessary to form a discharge between two electrodes in a gas as a function of pressure p and gap length d, the minimal breakdown voltage V for pd = 7.5 × 10 −6 m atm, where V is 327 V in air at standard atmosphere pressure (p = 1 atm) at a distance of 7.5 μm. This condition could be achieved at the instant when the rock is fractured by the present Charpy impact experiment (Enomoto and Chaudhri 1993).
Japan has many mountains composed of granite, such as Mt. Kaikoma, Mt. Yakushi, and Mt. Tsukuba in (see Fig. 1a). However, our survey showed that there are few Japanese historical earthquake records on EQL due to granite landslides on such mountains (e.g., ERI 1981ERI -1994Musya 1932;Enomoto 1998).

Pyroclastic rock
The pyroclastic rock was subjected to the TDS-MS analysis. The typical mass spectrum is shown in Fig. 4a for pyroclastic rock, showing H 2 O, lower hydrocarbons showing complex cracking patterns like as C, CH, CH 4 , C 2 H 2 , C 2 H 4 , etc., H 2 S, and CO 2 peaks, which are components of typical volcanic gases (Sakata et al. 1996). It is noted that the amount of hydrocarbon outgassing increases about 20 times as the temperature rises from 250 to 450 °C, which indicates that it is promoted by heating. Both hydrocarbons and H 2 S are flammable gases, but the ignition temperature for lower hydrocarbons is 537-630 °C, which is higher than that for H 2 S, 260 °C. However, these temperatures could easily be reached by frictional heating during sliding impact. Therefore, H 2 S first burns due to frictional heating and heats lower hydrocarbons to release carbon and become hot: the carbon produces a red flame, as confirmed in the laboratory test, and shown in Fig. 1b.

Biotite-bearing rhyolite
Rhyolite has the second-highest photoemission intensity after granite (see Fig. 3c). The B value is slightly smaller than the R and G values. The TDS-MS spectrum of this rock is shown in Fig. 4b, showing that even when the heating temperature rises from 250 to 500 °C, the peak intensity for lower hydrocarbons remains almost unchanged, unlike those for pyroclastic rock. This may indicate that hydrocarbons were chemisorbed by the rock minerals. Furthermore, the peaks appear periodic with m/z differences of 14 (CH 2 ), so this rock may contain a small amount of mixed hydrocarbons. However, the amount of photoemission from the rhyolite due to the effect of degassing and burning of these hydrocarbons might be small, unlike that for pyroclastic rock. Rather, because the rhyolite contains biotite and quartz with high TSEE activity, the cause of the photoemission is likely to be similar to that of granite, as described above.  Enomoto et al. Earth, Planets and Space (2020) 72:108 Limestone Limestone is a sedimentary rock composed of calcium carbonate (CaCO 3 ) formed mainly from sea shells. Limestone does not contain quartz, biotite, or flammable gases that can cause photoemission. However, it is well known that natural radiation for extended geological ages creates electronically excited (trapped) states in limestone.
Heating allows these trapped states to interact with thermally activated lattice vibrations and rapidly decay into lower energy states, causing the emission of photons, that is, thermoluminescence, TL (Kalita and Chithambo 2019). The fracture surface, newly created during impact tests, shows red emission due to TL, as seen in Fig. 5c. The red light associated with the collapse of limestone in the 1909 Anegawa earthquake may be the reason why the villagers assumed that the volcano had exploded. To confirm this hypothesis, we conducted TL experiments on the limestone sample.
A thin limestone plate sliced from the as-received sample was heated on a hot plate at a rate of 5 °C/min in air in a dark room, and the color and the spectra of the light emitted from the sample were observed by photography and spectroradiometry, respectively. At a temperature of 212 °C, an orange spectrum around 613 nm was seen, as shown in Fig. 5a, but the red emission at 780 nm became more intense at around 260 °C. The activation energies for photons released from the trapped sites at wavelengths of 613 nm and 780 nm were determined from the Arrhenius plot to be 0.42 eV and 0.27 eV, respectively, as shown in Fig. 5b. A temperature of 260 °C could be easily reached by frictional heating in the present Charpy tests.

Serpentinite
No photoemission from serpentinite during impact could be confirmed with the present camera sensitivity, although we were able to measure emissions of negatively charged particles (fracto-emissions) accompanying fracture in previous experiments during a uniaxial compression fracture test (Enomoto et al. 1994). It is known that normally the shear-fractured surface of serpentinite contains clay minerals such as montmorillonite and smectite, and thus the rock is often smooth, such that a fault may have a mirror-like luster capable of specular reflection. In the present Charpy test, when the serpentinite was subjected to impact, the impacted surface seemed as smooth as a cleaved surface. Therefore, physicochemical mechanisms might not occur to produce photoemission by thermal stimulation.
What can be predicted from the present laboratory experiments is that no landslide EQL can be witnessed in mountains like as Mt. Tanigawa and Mt. Shibutsu, which are composed of serpentinite. But this prediction cannot be verified yet as there was no such event in historical reports (e.g., ERI 1981ERI -1994Musya 1932;Enomoto 1998).

Conclusion
In the present impact fracture tests of rocks, photoemission was observed from granite, pyroclastic rock, rhyolite, and limestone, and the intensity and spectral characteristics depended on the type of rock. Photoemission is affected by the temperature rise due to frictional heating during impact, which leads to discharge by a coupling interaction between TSEEs, high electric potential generation due to charge separation and/or the piezoelectric effect, desorption and combustion of flammable gases, and TL due to electronic transitions from trapped sites. The critical temperature at which these physicochemical processes are operated is in a range of 250-450 °C, and it could be sufficiently excited by frictional heat generated by the present impact tests. The experimental results are consistent with eye-witness testimony of landslide EQL.