Properties of positive and negative leaders developing in air gaps ranging from 4 to 10 m that were subjected to 100/7,500‐μs voltage impulses were examined using a two‐frame, high‐speed video camera with image enhancement. Abrupt extension (stepping) that culminated in a bright and structured corona streamer burst was observed for both negative (expected for the “classical” stepping process) and positive (expected for the so‐called restrike process) leaders. Selected high‐quality images of five negative and four positive leaders with pronounced corona streamer bursts are presented here. The morphology of corona streamer bursts was essentially independent of polarity. Streamer bursts exhibiting nearly spherical symmetry were observed. For the four positive leaders, the newly added channel sections (steps) were almost straight and had lengths ranging from about 50 to over 120 cm. For the five negative leaders, most of the steps were curved and their 2‐D lengths were some tens of centimeters. It is generally thought that positive leaders in both long sparks and lightning extend continuously or exhibit optically unresolvable steps whose length is comparable to the leader tip size (1 cm or less) and that for sparks only when the absolute humidity is relatively high (>10 g/m3 or so) or voltage rise time is relatively long (around 1 ms or more) can larger steps occur. In this study, both modes of propagation for different branches of the same positive leader were observed.

We have observed unusual plasma formations (UPFs) in artificial clouds of charged water droplets using a high‐speed infrared camera operating in conjunction with a high‐speed visible‐range camera. Inferred plasma parameters were close to those of long‐spark leaders observed in the same experiments, while the channel morphology was distinctly different from that of leaders, so that UPFs can be viewed as a new type of in‐cloud discharge. These formations can occur in the absence of spark leaders and appear to be manifestations of collective processes building, essentially from scratch, a complex hierarchical network of interacting channels at different stages of development (some of which are hot and live for milliseconds). We believe that the phenomenon should commonly occur in thunderclouds and might give insights on the missing link in the still poorly understood lightning initiation process.

Based on experimental results of recent years, this article presents a qualitative description of a possible mechanism (termed the Mechanism) covering the main stages of lightning initiation, starting before and including the initiating event, followed by the initial electric field change (IEC), followed by the first few initial breakdown pulses (IBPs). The Mechanism assumes initiation occurs in a region of~1 km 3 with average electric field E > 0.3 MV/(m•atm), which contains, because of turbulence, numerous small "E th volumes" of~10 −4-10 −3 m 3 with E ≥ 3 MV/(m•atm). The Mechanism allows for lightning initiation by either of two observed types of events: a high-power, very high frequency (VHF) event such as a Narrow Bipolar Event or a weak VHF event. According to the Mechanism, both types of initiating events are caused by a group of relativistic runaway electron avalanche particles (where the initial electrons are secondary particles of an extensive air shower) passing through many E th volumes, thereby causing the nearly simultaneous launching of many positive streamer flashes. Due to ionization-heating instability, unusual plasma formations (UPFs) appear along the streamers' trajectories. These UPFs combine into three-dimensional (3-D) networks of hot plasma channels during the IEC, resulting in its observed weak current flow. The subsequent development and combination of two (or more) of these 3-D networks of hot plasma channels then causes the first IBP. Each subsequent IBP is caused when another 3-D network of hot plasma channels combines with the chain of networks caused by earlier IBPs.

Detailed observations of the connection between positive and negative leaders in meter‐scale electric discharges generated by clouds of negatively charged water droplets are presented, and their possible implications for the attachment process in lightning are discussed. Optical images obtained with three different high‐speed cameras (visible range with image enhancement, visible‐range regular, and infrared) and corresponding current recordings were used. Two snapshots of the breakthrough phase of the leader connection, showing significant leader branching inside the common streamer zone, are presented for the first time. Positive and negative leader speeds inside the common streamer zone for two events were found to be similar. Higher leader speeds were generally associated with higher leader currents. In the case of head‐to‐head leader connection, the infrared brightness of the junction region (probably representing the gas temperature and, hence, the energy input) was typically a factor of 5 or so higher than for channel sections either below or above that region. In 16% of cases, the downward negative leader connected to the upward positive leader below its tip (attached to the lateral surface of the positive leader), with the connection being accomplished via a channel segment that appeared to be perpendicular to one or both of the leader channels.

Detailed infrared (2.7-5.5 μm) images of bidirectional leaders produced by the cloud of small (typical radius of 0.5 μm), positively charged water droplets are presented. The leader was composed of the downward extending positive part and the upward extending negative part, these two parts (both branched, although in different ways) being connected by the single-channel middle part. The downward extending part included the tortuous positive leader channel (similar to its upward extending counterpart observed when the cloud polarity was negative) that was often accompanied by much less tortuous but often equally bright downward extending plasma formations of unknown nature. Very faint positive streamer zone was also observed. Either the positive leader channel or the unusual plasma formation (UPF) can come in contact with the grounded plane. The upward extending part is associated with a large network of faint channels, mostly fanning out of the upper part of the usually much brighter leader channel and apparently pervading the entire upper part of the cloud. Some of those faint channels could be unusually long and bright negative streamers, while others could be similar to UPFs. The IR luminosity along the brightest part of the bidirectional leader channel is often nonuniform. Some variations in channel brightness are localized and suggest the involvement of space leader-type processes at multiple positions along the channel, changes in channel orientation, or variations in channel radius.

To date, a non-controversial mechanism for the initiation and development of lightning in thunderstorm clouds has not been formulated as a sequential chain of events that begins with the first electron avalanches and ends with the appearance of a negative step leader. Based on experimental results of recent years, this article presents a qualitative description of a possible mechanism (termed the Mechanism) covering the main stages of lightning initiation, starting before and including the initiating event, followed by the initial electric field change (IEC), followed by the first few initial breakdown pulses (IBPs). The later transition from IBPs to a negative stepped leader is briefly described. The Mechanism assumes initiation occurs in a region of a thundercloud of ~1 km 3 with electric field E > 0.3 MV/(m•atm), which contains, because of turbulence, numerous small "E th -volumes" of ~0.001 m 3 with E ≥ 3 MV/(m•atm). The Mechanism allows for lightning initiation by two observed types of initiating events: a high power VHF event called an NBE (narrow bipolar event), or a weak VHF event. According to the Mechanism, both types of initiating events are caused by a group of relativistic runaway electron avalanche particles passing through many of the E th -volumes, thereby causing the nearly simultaneous launching of many positive streamer flashes. Due to ionization-heating instability, unusual plasma formations (UPFs) appear along the trajectories of the streamers. These UPFs combine into three-dimensional (3D) networks of hot plasma channels during the IEC, resulting in its observed weak current flow. The subsequent development and combination of two (or more) of these 3D networks of hot plasma channels then causes the first IBP. Each subsequent IBP is caused when another 3D network of hot plasma channels combines with the chain of networks caused by earlier IBPs.

The streamer zone of positive leader during the breakthrough phase of long sparks was experimentally investigated with two methods. One of the methods is the analysis of streamer‐zone images obtained with a high‐speed framing camera with image enhancement. This method allowed us to estimate the spatial distribution of streamer density and low‐frequency conductivity in the streamer zone. The other method is the microwave probing, which we applied for the first time to long sparks. The attenuation of microwave beam in the streamer zone in our experiments is proportional to the total number of free electrons inside the microwave beam. Experimental data on the microwave attenuation combined with the streamer density found using the first method allowed us to estimate the total number of free electrons in one streamer. The following parameters were obtained. The streamer density in the center of the streamer zone is (0.6–1) · 105 m−3, and the total number of streamers in the streamer zone is 4 · 105–106. The average total number of free electrons in one streamer is about 3 · 1010. Low‐frequency conductivity on the axis of streamer zone was estimated to be typically 2 · 10−5 S/m, which is similar to that estimated for corona sheath in lightning. Both methods are based on the assumption of constancy of electric field and similarity of all streamers inside the streamer zone. The overall results of this study are generally consistent with this assumption.

No abstract