Observations of narrow bipolar events reveal how lightning is initiated in thunderstorms

A long-standing but fundamental question in lightning studies concerns how lightning is initiated inside storms, given the absence of physical conductors. The issue has revolved around the question of whether the discharges are initiated solely by conventional dielectric breakdown or involve relativistic runaway electron processes. Here we report observations of a relatively unknown type of discharge, called fast positive breakdown, that is the cause of high-power discharges known as narrow bipolar events. The breakdown is found to have a wide range of strengths and is the initiating event of numerous lightning discharges. It appears to be purely dielectric in nature and to consist of a system of positive streamers in a locally intense electric field region. It initiates negative breakdown at the starting location of the streamers, which leads to the ensuing flash. The observations show that many or possibly all lightning flashes are initiated by fast positive breakdown.

This did not affect the ability of the INTF to determine the source directions, as the phase of the signals was unchanged by the clipping. b, Exponential increase in the VHF power during the first 1.5 µs of the NBE, prior to being clipped. The red line corresponds to a 0.3 µs initial rise time constant. c, Expanded plot of the first few µs of the NBE, showing i) its rapid onset, ii) the simultaneous occurrence of the FA sferic and VHF radiation, indicative of the current being produced by the breakdown, and iii) the lack of detectable activity prior to the NBE onset. Both the VHF and fast electric field signals were quiet down to the ambient noise level of the INTF site, 66 dB below their full scale amplitudes. The transient ripple in the FA waveform is due to periodic local interference source, seen in Fig. 4d and in Supplementary Fig. 2b. d, Exponential rise and fall of the VHF power obtained from the attenuated VHF waveform present on the FA signal, digitised at the same 180 MHz rate as the INTF signals. The overall rise and fall time constants were 1.0 and 4.7 µs, respectively. ...  Figure 3: Current waveforms. a, Double exponential current waveforms that simulated the sferics for NBE1 and NBE3, showing the difference in duration of the two currents. b, Current versus propagation distance at successive 4 µs time intervals, illustrating how the current for NBE3 was spatially more compact while the current for NBE1 appeared to be spatially spread out relative to the overall extent of the breakdown, consistent with the NBEs being produced by a succssion of breakdown events.
...  b, Expanded view of the initial 7 µs of the NBE, showing i) the relatively gradual onset of the VHF activity, ii) the retrogressive upward development of the initial breakdown, iii) several attempted downward events as the discharge intensified, and iv) the lack of activity prior to the NBE's onset. c,d, simulation of NBE2's sferic and the two-pulse current waveform used in the simulation (see main text).
...   Fig. 6b (orange circles). These and the other circled PC events were captured in the preflash intervals of INTF recordings triggered by the subsequent flashes. PC1 occurred immediately above the periphery of a horizontally extensive CG flash that followed 2 s later (red sources). PC3 occurred 1 s before a bilevel IC flash (red sources), a short distance away from the IC's upward channel and immediately below upper positive charge discharged by the IC. Precursor 'd' was an uncommon low-altitude event that occurred 1.5 s before and immediately below midlevel negative charge discharged by the IC.
... Figure 12: Detailed observations of the screening discharge. a, Expanded time series data for the temporally-separated upward positive breakdown events at the beginning of the 6 ms-duration screening discharge of Fig. 6c, showing the retrogressive downward development of the successive events (dashed arrow). b, Additionally expanded data for the first three events, showing the apparent speeds of the first and second discharges. Each successive event had decreased vertical extent but slightly stronger peak VHF activity. c,d, Spatial development of the VHF radiation sources in azimuth-elevation format during the initial 50 and 250 µs of the discharge, respectively. Although the discharge was at 11.7 km plan distance, due to its great height (14.2 km), it was near 45 • elevation angle relative to the INTF at 3.2 km MSL. Uncertainties in the INTF source locations are shown in Supplementary Fig. 20.

Supplementary
...  2 The intersection of the lines gives the direction of the source. Geometrically, the direction cosine plane corresponds to the equatorial plane of a unit celestial hemisphere centered on the antenna array. 3,4 The inscribed circle corresponds to the horizon and its center to the zenith. The dashed circles correspond to 30 • and 60 • elevation angles and shows that the interferometer locates sources less well below 30 • elevation. The azimuth and elevation angles are obtained by projecting the solutions in the direction cosine plane up onto the celestial hemisphere. b, Zoomed-in view of the intersection. Noise and/or multiple sources cause the lines not to have a common intersection, but to outline a triangular region. The solution is obtained from a least-square fit of the lines and lies within the triangle. The size of the triangle provides a measure of the uncertainty in the source direction. For the example shown, the time differences of arrival at the three baselines were (τ 12 , τ 23 , τ 31 ) = (+0.0294, −0.0152, −0.0149) µs. The maximum and minimum azimuth angles were 45.408 • and 46.452 • and the elevation angles were 51.968 • and 53.187 • , corresponding to 'peak-to-peak' uncertainties of 1.04 • in azimuth and 1.22 • in elevation around the least-square value. For point source and noise-free observations, the three time differences (i.e., the 'closure delay' τ 123 going around the array) add to zero. For the example shown, the time differences add to τ 123 = −0.0007 µs, or -0.7 ns. The uncertainties provide a measure of the combined effects of noise and non-localized radiation sources. The observations indicate the uncertainty is primarily due to multiple sources rather than background noise (e.g, Supplementary Figs. 16 and 17). . Sources with centered dots in panel a) have a small uncertainty. The noticeable increase in the elevation uncertainty following the NBE (panel a) and also in azimuth (panel c) is due primarily to the negative breakdown having multiple sources, rather than to their reduced power and signal to noise ratio. This is seen from the first two sources at the beginning of the NBE being as weak or weaker than the post-NBE activity, yet having very small elevation uncertainties ( 0.2-0.3 • ). It indicates the initial breakdown of the NBE was highly localized, and illustrates the limiting accuracy of the INTF measurements. The elevation uncertainties remain relatively small during the initial part of the downward NBE, but increase for some of the sources toward the end of the NBE, consistent with the activity occurring at multiple altitudes. The sudden elevation increase just after 0.02 ms (20 µs) is significant uncertainty-wise and coincides with the beginning of the partial field recovery, further supporting the result that the recovery signals the current having died out beyond this time.
...  Supplementary Fig. 16, except for NBE2. Again, the uncertainties of the weak radiation at the beginning of the NBE are small, indicating the initial breakdown was highly localized, and providing a baseline for the uncertainty caused by background noise. The relatively small uncertainties of the sources during the downward NBE indicate that it was primarily a monotonic event, consistent with the NBE's narrower current pulse ( Supplementary  Fig. 3).
Experimental studies of positive streamers in air measure their speeds to be one or two orders of magnitude below the observed speeds of the fast positive breakdown. The highest reported speed is 4×10 6 m s −1 , measured by Briels et al. 5 with a 96 kV pulse across a 4 cm point to plane gap at ambient pressure. An approximate average value for the non-uniform electric field would be 2.5×10 6 m s −1 , comparable to the breakdown strength of air (3.2×10 6 V m −1 ).
Allen and Mikropoulos 6 made detailed observations of the speed of positive streamers in a uniform electric field, and investigated how the propagation speed varied with the ambient field E. The uniform field was provided by a parallel plate arrangement having a 12 cm gap. They expressed the observations in terms of the minimum field E st required to initiate a stable streamer, called the 'stability' field, and the corresponding stability velocity v st of the resulting streamer. An electric field E st = 491·δ kV m −1 was required to produce a streamer of minimal initial energy, where δ is the fractional air density relative to standard atmospheric conditions. The associated propagation speed was v st = 1.25 × 10 5 m s −1 . For stronger fields, the streamer speed v str was well-fitted by a cubic power law dependence on E, according to (1) The measurements were made at ambient pressure with electric fields between 450 and 800 kV m −1 . The speed corresponding to 800 kV m −1 was between 5 and 10×10 5 m s −1 , depending on the amplitude and duration of the pulse used to initiate the streamers from a recessed point in the ground plane (2 and 4 kV, and 135 and 270 ns, respectively).
The question of interest is whether at stronger fields the cubic relation would contine to hold, and whether the propagation speed can reach the observed values obtained for the fast positive breakdown. To answer the question it is instructive to evaluate (1) for the streamer speed that would be predicted if E is assumed to be the breakdown field E k = (3.2×10 6 ) · δ V m −1 . Calling this speed v k , the answer is that the density dependence cancels, giving v k = (1.25 × 10 5 ) 3.2 × 10 6 4.91 × 10 5 3 = 3.5 × 10 7 m s −1 .
Remarkably, this is the same as the speeds of NBEs 1 and 3, and is close to the speeds estimated for the lesser-power discharges. It is based entirely on empirical data from laboratory observations made at ground level and over a restricted range of E field values. Agreement with the observations of the present study suggests that (1) is valid for fields up to and possibly beyond breakdown, and that the fast positive breakdown occurs at or near E k , independent of altitude. Taken at face value, E/E k would be slightly above unity when v str > v k , and slightly less than unity for v str < v k . In the super-critical case (e.g., for NBE2 or for the screening discharge), the breakdown might be expected take the form of an ionization wave rather than discrete streamers. While this might happen in some circumstances, it should be noted that the value of v st in (1) applies to streamers having minimal initial energy, and that the speeds for a given value of E are increased by a factor of two for initial energies corresponding to relatively small potentials of 2-4 kV (Fig. 10 of Allen and Mikropoulos). The observed speeds would then be attained at fields somewhat below breakdown.
Overall, the picture that emerges is that positive streamers at the beginning of a discharge intensify the electric field up to or near breakdown, and then continue as a result of self-generated field enhancement ahead of the streamer system 7 . Allen and Mikropoulos showed that the streamer speed stabilized within a few cm of being initiated, and noted particularly that speed invariance through the remainder of the gap indicated that branching had negligible effect on the results.