Electric Discharge in Erupting Mud
Abstract
Self-ignition during the explosive eruption of mud volcanoes can create flames that in some cases reach heights that exceed hundreds of meters. To study the controls on electrical discharge in natural mud, we performed laboratory experiments using a shock-tube apparatus to simulate explosive eruptions of mud. We vary the water content of the mud and proportions of fine particles. We measure electric discharge within a Faraday cage and we use a high-speed video camera to image the eruption of mud and some of the electric discharge events. We find that (a) decreasing the proportion of fine particles and (b) increasing water content each suppress the number and magnitude of electric discharge events. Experimentally observed mud volcano lightning occurs where particles exit from the vent and within the jet of erupting particles.
Key Points
-
Laboratory experiments show that jets of erupting mud can create electric discharge
-
Decreasing the proportion of fine particles and increasing water content suppress number and magnitude of electric discharge events
-
Electric discharge is one mechanism that may enable self-ignition of hydrocarbons during explosive mud volcano eruptions
Plain Language Summary
Vigorous mud volcano eruptions are often accompanied by large flames, which are the result of the self-ignition of natural gas blasted with the erupting mud. The mechanisms for this gas ignition are not yet understood. We study a potential self-ignition mechanism: volcanic lightning. We erupt mud in a laboratory setting and measure the conditions that promote lightning. We find that dry mud and mud with abundant fine particles create more lightning. For safety reasons, we studied mud samples with argon as the carrier gas for the eruption but expect that if a flammable gas was used, the lightning could ignite the gas.
1 Introduction
Mud breccias that erupt explosively at mud volcanoes are sometimes accompanied by flames produced by the self-ignition of methane and other hydrocarbons. Flames can reach heights up to one km (Planke et al., 2003), though are most often 60–100 m high (Bagirov et al., 1996). While most of these flaming eruptions are short-lived (Judd, 2005), once ignited, methane seeps can burn for millennia (e.g., Etiope et al., 2004) and in some cases burn hot enough and long enough to melt rock (Grapes et al., 2013) causing pyrometamorphism (Sokol et al., 2008). The spectacular flames garner widespread media attention and speculation about the reasons for the self-ignition. Sparks, other electric discharge, or spontaneous combustion are possible processes.
Self-ignition has long been attributed to collisions between rock fragments. Arnold and MacReady (1956) wrote that “sparks were probably formed by the striking together of some of the rocks, and the petroleum gas thus ignited” and Snead (1964) wrote that in “violent eruptions enough friction is exerted between rock fragments to ignite methane gas escaping with the liquid mud.” Electric discharge from erupting solids is, in fact, well documented. Volcanic lightning has been reported at least since Pliny's account of the 79 AD eruption of Vesuvius. It accompanies nearly all styles of explosive eruption (e.g., Cimarelli & Genareau, 2022), from the small eruptions at Stromboli, Italy (Vossen et al., 2022) to almost 600,000 detected lightning strikes during the 2022 Hunga Tonga-Hunga Ha'apai eruption that reached a height of >50 km. Electrical activity can occur both near the vent and in the plume, with the former promoted by high velocities and the latter by high volume fluxes (Smith et al., 2021). Jetting of fine ash is thought to be key for charging (McNutt & Thomas, 2015).
Not all mud eruptions self-ignite. Only “the most vigorous of such outbursts uproot trees, ignite methane” (Kopf, 2003). The Shikhzarli mud volcano, Azerbaijan, has flames in 74% of its eruptions (Kokh et al., 2017). Lokbatan, Azerbaijan, has a number of well documented explosive eruptions. The 5 January 1887 eruption produced a flame that reached a height greater than 600 m (Dimitrov, 2002). An example flaming eruption from Lokbatan on 20 September 2012 is shown in Figure 1 (Mazzini et al., 2021). Lokbatan erupted on 11 August 2022, with conflicting reports about whether the eruption was accompanied by flames. Investigations of the parameters which govern ignition can yield insights into eruption initiation and dynamics. Here, we study here the influence of grain size distribution and mud humidity (e.g., Springsklee, Scheu, et al., 2022; Stern et al., 2019) on electrical discharge in erupting mud.

Mud eruption of Lokbatan, Azerbaijan on 20 September 2012. Two vents are active at this time. The flame in the foreground may have been ignited by flames at the vent in the background (adapted from Mazzini et al. (2021)). Scale is obtained by triangulating the known vent locations and estimated location of the video recording.
Explosive eruptions are generally attributed to sealing of the vent between eruptions so that large overpressures are needed to initiate eruptions and unblock feeder channels and vents (Dimitrov, 2002; Mazzini & Etiope, 2017). Bagirov et al. (1996) compiled statistics of mud eruptions in Azerbaijan and found that gas self-ignites in 42% of eruptions. In a global assessment, Judd (2005) found that 30% of methane emitted at mud volcanoes ignites to produce flames. While explosive mud eruptions accompanied by flames are most abundantly documented in Azerbaijan, examples have been reported in Trinidad (Anderson, 1911), New Zealand (Ridd, 1970), and Pakistan (Snead, 1964). Self-igniting gas explosions from permafrost have also been documented (e.g., Bogoyavlensky et al., 2022) and may share processes in common with mud volcanoes.
Here we use laboratory shock-tube experiments to study the conditions that lead to electrical discharge events in explosively ejected mud. In these experiments, mud held in an autoclave is subjected to a rapid decrease in pressure and it erupts through a vent to form an expanding jet. This simulates the brecciation of the seal that confines overpressured sediments (Mazzini et al., 2021). The jet is imaged with a high-speed video camera, and electric discharge is measured with a Faraday cage. Similar experiments have been used to simulate volcanic lightning to understand how properties of the erupted materials and environmental variables control electrical activity (Cimarelli et al., 2014; Gaudin & Cimarelli, 2019; Stern et al., 2019). For the case of mud, we first document that electrical discharge can occur. We then assess how particle size and water content affect electrical activity.
2 Materials and Methods
2.1 Samples
Mud was collected from the Davis-Schrimpf hydrothermal field in the Imperial Valley, California, USA. The mud extrudes from gryphons (cone-shaped vents) 1–2 m in height. Bubbles that burst in the vents or pools of mud can also eject spatter. This mud is composed of 41% quartz, 27% clay minerals (illite, montmorillonite and kaolinite), 24% feldspar, and 9% carbonates (Tran et al., 2015). Figure S1 in Supporting Information S1 shows the particle size distribution measured with a Bettersizer S3 Plus laser particle size analyzer. The sub-10-micron fraction comprises 30.7 vol% of the mud.
For the experiments in which the mud is not fully saturated with water, the mud was initially dried and then hydrated in an environmental chamber at controlled humidity and temperature. In the process of drying, the mud became consolidated and was thus mechanically disaggregated with a vibratory disc mill prior to use in experiments, leading to an increase in the volume fraction of fine particles (<10 microns) (Figure S1 in Supporting Information S1) from 30.7 to 41.7 vol%. The proportion of fines is a crucial parameter in the number and intensity of discharges in experiments (Cimarelli et al., 2014; Gaudin & Cimarelli, 2019; Stern et al., 2019) and modeling based on these experiments (Rayborn & Jellinek, 2022).
The mineralogy and particle size distribution of the mud from the Davis-Schrimpf vents are similar to those of other mud volcanoes (Tran et al., 2015) which generally tend to erupt sediment deposited in deltaic and lacustrian settings. More deeply rooted mud volcanoes may have a greater proportion of illite among the clay minerals. The Davis-Schrimpf vents cannot ignite because the gas driving their eruptions (CO2) is not flammable and there is no record of violent eruptions (Mazzini et al., 2011; Svensen et al., 2009).
To evaluate the potential role of particle size, we performed three additional experiments in which we added 0.17 mm diameter quartz sand. The grain size distribution of this sand is provided in Figure S1 in Supporting Information S1.
2.2 Experimental Facility
The experimental apparatus is shown in Figure 2 and is adapted from the system used in previous studies (Alidibirov & Dingwell, 1996; Cimarelli et al., 2014; Gaudin & Cimarelli, 2019; Stern et al., 2019). Samples were placed in a cylindrical steel autoclave, with sample chamber dimensions of 26 mm in diameter and 160 mm in length; the entire autoclave was electrically grounded. The autoclave was completely filled in all experiments. At this constant volume, variable densities of the mud + sand + water mixtures resulted in differing sample masses.

Photograph and illustration of the experimental apparatus. The upper part (light gray color) represents the Faraday cage that is insulated by a plastic flange (dark green color) from the decompression system including the nozzle and the autoclave. The autoclave is completely filled with the sample and grounded. Discharges from the jet to the nozzle are recorded by the datalogger and captured by a high-speed camera. The additional resistor and capacitor enable the calculation of the electrical current from the recorded voltage.
The samples were pressurized with argon gas to 10.2 ± 0.1 MPa. The sample was then subjected to a rapid decompression by the controlled disruption of iron rupture diaphragms that sit between the autoclave and the capture tank. Upon diaphragm rupture, the sample was free to expand as a gas-particle jet into the capture tank which was instrumented to record optically the expanding jet as well as flashes of light from electrical discharges using a V711 Phantom high-speed video camera. The entire capture tank was electrically insulated and served as a Faraday cage such that changes in charge over the course of the experiment could be recorded and quantified (Stern et al., 2019).
The frame rate of the camera was set to 15,000 frames/s. The video camera was employed in three different modes. The shutter was left open for either 4 µs (6% of the time between individual frames), 10 µs (Experiment 382—completely water-saturated mud), or 66.66 µs (continuous mode). In the first case, the images captured sharply the ejection of mud fragments, sand particles, as well as the dynamics of the gas phase. However, the probability of capturing visible electrical discharges was greatly reduced as the shutter remained closed for 94% of the experiment. In the third case, all electrical discharges visible in the field of view of the camera were captured at a high focal depth but the long exposure times did not allow us to capture the motion of the initially rapidly expanding jet thus causing trails and streaks from the gas phase and particles and resulting in more blurred images. Owing to the high concentration of particles, only electrical discharge events near the outside of the jet and where the jet emerges from the vent were visible to the camera.
3 Results
Figure 3 shows the evolution of electric current and charge with time over the course of an experiment with mud containing 2.92 wt% water (Experiment 378). Equivalent times series for all the experiments are plotted in Figure S2 in Supporting Information S1.

Example time series (Experiment 378, mud containing 2.92 wt% water): (a) decrease in pressure due to decompression, (b) measured raw current signal, and (c) the electric discharges obtained from the deconvolved signal. The photos show snapshots of the erupting mud during the progression of the experiment. The photos show the initial gas pulse that accompanies decompression experiments, followed by the gas-particle mixture. The high-speed camera detected discharges within the jet as well as discharges from the jet to the nozzle (circled). A complete set of images is in Movie S1.
Following the rupture of the diaphragm, pressure in the autoclave drops (Figure 3a). Synchronized acquisition of pressure, electrical discharges, and video recording is triggered as soon as the pressure in the autoclave drops below 6.5 MPa (falling edge trigger) and represents time = 0 in the raw data (shifted by 1 ms in the processed data to include potential discharges from the very beginning of decompression). A pre-trigger (∼30%) ensures that we capture the decompression event in its entity. The decompression of the sample accelerates the mud upwards through the vent and into the capture tank, creating an expanding jet of gas and particles. Figure 3b shows the raw electrical current recorded by the Faraday cage. Each peak combined with an exponential decay identifies an electrical discharge event. The charging of the detector decays approximately exponentially with time. By fitting an exponential decay model to each event, it is possible to deconvolve the instrument response from the raw data in order to detect and count individual events and compute their magnitudes (Gaudin & Cimarelli, 2019).
Discharge events begin with the arrival of particles at the vent. Most, but not all, events generate a negative charge (Figure 3 and Figure S2 in Supporting Information S1). During the process of triboelectrification, large particles tend to be positively charged while small particles tend to acquire a negative charge (Forward et al., 2009; Lacks & Levandovsky, 2007; Lee et al., 2015). In our experiments, the discharges are positive as well as negative.
Table 1 summarizes for all 13 successful experiments the properties of sample used, the number and magnitude of discharge events, the total charge neutralized during each experiment and the detectable particle front velocity, observable by increased obscuration within the jet.
EXP | Material | Water [wt%] | Sand [wt%] | Total mass [g] | Autoclave pressure [MPa] | Positive discharges 0–5 ms [nC] | Negative discharges 0–5 ms [nC] | Number of positive discharges 0–5 ms | Number of negative discharges 0–5 ms | Particle front velocity [m/s] |
---|---|---|---|---|---|---|---|---|---|---|
377 | Mud | 0.00 | 0.0 | 72.97 | 10.08 | 855.6 | −1068.0 | 178 | 246 | 291.1 ± 7.8 |
378 | Mud | 2.92 | 0.0 | 70.01 | 10.16 | 5.5 | −400.2 | 1 | 28 | 304.4 ± 15.1 |
379 | Mud + Sand | 0.00 | 10.0 | 60.34 | 10.15 | 397.0 | −621.6 | 56 | 152 | 397.3 ± 14.2 |
380 | Mud + Sand | 0.00 | 90.0 | 113.86 | 10.46 | 13.2 | −100.4 | 6 | 22 | 292.6 ± 13.8 |
382 | Mud | 31.12 | 0.0 | 76.01 | 10.20 | 0.0 | 0.0 | 0 | 0 | 367.6 ± 2.9 |
383 | Mud | 5.06 | 0.0 | 62.03 | 10.23 | 0.0 | −5.8 | 0 | 3 | 285.1 ± 5.1 |
385 | Mud | 0.00 | 0.0 | 78.95 | 10.30 | 226.5 | −729.1 | 58 | 155 | 178.1 ± 44.1 |
386 | Mud | 0.00 | 0.0 | 60.77 | 10.14 | 75.8 | −576.4 | 19 | 104 | 212.5 ± 14.7 |
387 | Mud + Sand | 0.00 | 50.0 | 82.93 | 10.20 | 24.4 | −188.2 | 12 | 47 | 166.9 ± 8.0 |
388 | Mud | 0.62 | 0.0 | 57.01 | 10.30 | 38.1 | −553.2 | 11 | 61 | 208.0 ± 39.1 |
389 | Mud | 0.62 | 0.0 | 57.60 | 10.11 | 4.3 | −623.8 | 2 | 68 | 282.6 ± 19.8 |
390 | Mud | 3.49 | 0.0 | 47.03 | 10.35 | 0.0 | −101.5 | 0 | 14 | - |
391 | Mud | 3.49 | 0.0 | 49.44 | 10.26 | 0.0 | −96.9 | 0 | 11 | 286.2 ± 3.5 |
- Note. For Experiment 390, no particle front velocity was measured because the video recording was not triggered.
Figure 4a shows the number of detected discharge events and Figure 4c shows the total charge accumulation, as a function of water content. Also shown for dry samples are the effects of adding sand (Figures 4b and 4d). The dry mud (without sand added) exhibits the greatest charging and the highest number of events (with a measurable variability in both parameters being observed in repeat experiments). The observations of the high-speed video reveal discharges within the upper part of the jet (Figure S3 in Supporting Information S1).

Total number of flashes and the total magnitude of discharge (both divided by sample mass) as a function of (a, and c) water content and (b and d) weight fraction of sand for dry samples. Numbers next to the open circles indicate the weight percent of sand in the experimental run.
When sand or water has been added to the mud, both the number of discharge events and the total charging decrease. The addition of 2.9 wt% of water yielded a decrease of 93% of both the number of discharges observed, and the magnitude of the largest discharge event recorded. The addition of 5.1 wt% of water decreases the number of discharges observed by > 99%. A fully water-saturated mud experiment (31.1 wt% of water) produced no observed discharge events.
4 Discussion
Our experiments document that dry mud particles can become electrically charged and induce discharges during explosive mud volcano eruptions. Increasing water content of the mud, and decreasing fine fraction of mud particles by adding sand, both yield decreases in the number and magnitude of observed electrical discharge events.
The first stage in creating electric discharge is to charge particles. There are several particle charging mechanisms in natural particle-laden jets, and quantifying these processes remains the subject of active research (Arason et al., 2011; Björnsson et al., 1967; Cimarelli et al., 2014; James et al., 2000; Méndez Harper et al., 2021; Nicoll et al., 2019; Prata et al., 2020; Thomas et al., 2010; Van Eaton et al., 2020). Ion transfer between colliding particles, triboelectrification, is generally thought to be the dominant process in thunderstorms (Mather & Harrison, 2006) and volcanic plumes (Cimarelli et al., 2014; Méndez Harper & Dufek, 2016). Fracto-electrification, in contrast, generated by fracturing particles (e.g., James et al., 2000) is likely insignificant in our experiments as we do not observe or expect significant fragmentation in these experiments since the particles are already very small and mostly non-porous (sand), and both of these particle attributes suppress further fragmentation during rapid gas expansion and/or particle collisions.
Prior to electrostatic discharging, charged particles need to be separated and surrounded by an electrically insulating medium to prevent charge transfer. In our experiments gas is the low-conductivity medium. The turbulence in jets and plumes, and the differential velocity of particles of different sizes, enable the charge separation (Cimarelli et al., 2014; Gaudin & Cimarelli, 2019). In more detail, the dynamics of particles depends on whether their motion is well-coupled to the surrounding fluid. The inertial response time of particles with diameter d and density
in a fluid with density
and viscosity
is
; the time characterizing eddy circulation is
, with
and
the vent radius and initial jet velocity, respectively (Jessop and Jellinek, 2014). The ratio of these time scales is the Stokes Number,
. With our chosen definitions,
kg/m3 (Tran et al., 2015), measured
m/s (±61 m/s), and
cm, then St < 1 for d < 10 microns and hence our fine particles will circulate in eddies. While details of particle-gas coupling depend on particle shape and drag coefficient, St is of order unity or less for particles smaller than 10
m (Figure S4 in Supporting Information S1).
In analogous experiments with volcanic ash it has also been observed that the number of discharge events increases with the fine fraction of volcanic ash (Gaudin & Cimarelli, 2019). The magnitude of individual discharge events in erupting volcanic ash is approximately linearly proportional to the erupted mass and the initial overpressure (Gaudin & Cimarelli, 2019). Thus, larger and more explosive eruptions will produce larger electrical discharges, and presumably this will increase the probability of ignition in the presence of methane or other flammable hydrocarbons. We note that natural muds and source rocks for mud breccia eruptions contain abundant fine particles (Dimitrov, 2002).
Our experiments show that even small amounts of water can suppress electric discharge. This may seem counterintuitive considering the common occurrence of lightning in clouds where ice particles and liquid water droplets are the charge carriers, but liquid water on particles can also enhance rapid charge leakage (Lorenz, 2018) and reduce charge accumulation (Méndez Harper et al., 2020). Wet volcanic eruptions also produce abundant volcanic lightning, such as the recent 2016–2017 Bogoslof, Alaska eruption (Van Eaton et al., 2020) and the 2022 Hunga Tonga-Hunga Ha'apai eruption. Lightning in the plumes of these large, wet explosive volcanic eruptions may be enabled by the formation of ice above the atmospheric freezing level and hence electrification through mechanisms that operate in thunderstorms (Van Eaton et al., 2020). In fact, controlled laboratory experiments of volcanic particle jets reveal that as relative humidity and water content increase, the electrification of volcanic ash decreases (Méndez Harper et al., 2020). A mere 1.8 wt% water in volcanic ash has been observed to significantly decrease electrical discharge (Stern et al., 2019), quantitatively similar to our present findings with natural mud.
There are some limitations to generalizing from our experimental results. First, for safety reasons, we employed argon rather than methane as the compressing gas phase. Thus, in the absence of flammable gases we could not assess whether the electrical discharge events actually generate ignition and combustion of the gas. Further, gas composition and pressure both affect the minimum breakdown distance for electric discharge and therefore our results are not quantitatively applicable to discharges within a methane-dominated jet. Hackam (1969) compares the electrical breakdown for gases including N2, CH4, and CO2 and emphasizes the significance of the gas composition for electric discharge. The role of gas properties is also the subject of ongoing research into electrical discharge in extraterrestrial atmospheres, for example, for brown dwarfs (Helling et al., 2013). Second, we have not yet systematically investigated the role of mineralogy on this phenomenon. Finally, we have not explored the effects of overpressure that initiates the eruption. The value of 10 MPa was chosen as it is similar to the tensile strength of rocks, though even larger overpressures may develop in the source region of mud volcanoes at depths of a few kilometers (e.g., Blouin, Imbert, et al., 2019; Blouin, Sultan, et al., 2019). Complete conversion of the experimental overpressure ∆P into kinetic energy and then gravitational potential energy would lead to eruption heights of ∆P/ρg where ρ is mud density and g is gravitational acceleration. This estimate neglects frictional losses and drag during eruption which would reduce height, but also acceleration of particles from gas expansion that increases height. If ∆P = 10 MPa, the ejection height would be 0.6 km for a bulk mud density of 1,600 kg/m3 (Tran et al., 2015); the mean ejection speed of 262 m/s in our experiments leads to ballistic heights of 3.5 km; these estimates are similar to heights of a few kilometers reported in large explosive mud volcano eruptions (Dimitrov, 2003).
5 Conclusions
Explosive eruption of dry mud creates electrical discharge events, and thus may be one mechanism to self-ignite methane in erupting mud. The conditions that promote electrification include the presence of fine particles and a limited amount of water. Although the mud and mud breccias that erupt during sustained eruptions generally have higher water contents that may dampen the propensity toward ignition, it may be an initial explosive venting through dry ground which provides the window of opportunity for self-ignition and initiation of prolonged flames.
Acknowledgments
B.S. and C.S. gratefully acknowledge financial support the Deutsche Forschungsgemeinschaft (DFG) through the TRR 235 Emergence of Life (Project-ID 364653263). D.B.D. was supported by 2018 ERC ADV Grant 834255 (EAVESDROP). CC was supported by DFG grant CI 254/2-1 and ERC CON Grant 864052 (VOLTA). M. M. is supported by the Humboldt foundation, CIFAR Earth 4D, and NSF 2042173. Special thanks go to the technical staff of the LMU workshop, especially Markus Sieber.
Open Research
Data Availability Statement
The data of the performed experiments is online at the GFZ Data Services (Springsklee, Manga, et al., 2022). The obtained data was analyzed using the Pandas Data Analysis Library (McKinney, 2010; Reback et al., 2022) and visualized using the Matplotlib Library (Caswell et al., 2021; Hunter, 2007).