The energy distribution of the electron precipitation responsible for extremely narrow (70 m) and dynamic auroral filaments is found to be sharply peaked at around 8 keV. The events were captured with high resolution low-light optical imagers located near Tromsø, Norway. The method uses imaging in two emissions which have different energy dependent responses to auroral electron precipitation. The key feature of the events was that no difference in the altitude of the two emissions was detected, nor any time-of-flight dispersion, thus leading to the conclusion that the filaments were caused by monoenergetic precipitation. Comparisons with an electron transport and ion chemistry model show that the high energy filaments were embedded in a region of lower energy precipitation of about 4 keV. There is currently no consistent theory to explain the characteristics of the observed auroral structures.
 Aurora exhibits structuring on a range of scales [Galperin, 2002]. Although observations of the smallest structure of the aurora have been reported for over half a century [Maggs and Davis, 1968], the understanding of the processes governing the scales is far from complete [Sandahl et al., 2008]. A number of models have been put forward to account for electron acceleration producing thin arcs and arc elements. While several important advances have been made since the review of Borovsky , there is no consensus about the processes behind the structuring on the subkilometer (transverse to magnetic field) scales.
 Dispersive Alfvén waves have been suggested as important for the smallest scales in aurora [Stasiewicz et al., 2000]. “Alfvénic acceleration” is used to refer to in situ observed particle populations spanning a range of energies from tens of eVs to keVs, often field-aligned and exhibiting time-of-flight (TOF) energy dispersion [e.g.,Arnoldy et al., 1999; Andersson et al., 2002; Chaston et al., 2002a]. The coupling of dispersive Alfvén waves to optical signatures is less firmly established. Modeling starting from satellite data [Chaston et al., 2003] predicts thicknesses in the kilometer range, rather than tens to hundred of meters. On the other hand, the morphology and dynamics of multiple auroral filaments has been related to dispersive Alfvén waves [Semeter and Blixt, 2006; Semeter et al., 2009]. The structures in flickering aurora (caused by electrons with energies over 10 keV) have also been explained in terms of interfering Alfvén waves [Sakanoi et al., 2005; Whiter et al., 2008].
 To relate the optical signatures to acceleration processes, simultaneous observations of aurora at high resolution, and in situ measurements of electric and magnetic fields and particles, preferably at multiple points, are needed. Such observations have been scarce. Recent examples [Hallinan et al., 2001; Lynch et al., 2012] relate Alfvénic acceleration to dynamic rayed structure of an arc, although the observation geometry prevents conclusive identification. Narrow field of view imaging in the magnetic zenith is the best method to study auroral morphology on the smaller scales.
 We use ground based imaging of aurora in two emissions with different energy dependent responses to auroral electrons in order to characterize a highly structured and dynamic auroral display. The strength of this method was recently discussed in Lanchester and Gustavsson . We argue that thin (100 m scale) arc elements are caused by purely monoenergetic (energies over a narrow range) high energy (8 keV) electrons. This conclusion is based on the lack of altitude separation of the two emissions and the lack of TOF dispersion of energy in the short-lived features.
 The optical data presented here were captured on 15 December 2006, with two of the ASK (Auroral Structure and Kinetics) imagers. Each imager consists of one 512 × 512 pixel EMCCD sensor, F/1 lens system with a field-of-view of 4.4° diagonally, and a narrow (FWHM ∼ 1 nm) interference filter. The transmissions of the two filters have their maximum at 562.0 nm and 777.4 nm, for observations of the O2+ 1 N band and atomic oxygen (IO), respectively [Dahlgren et al., 2008]. The frame rate is 32 Hz. The data are dark and flatfield corrected, and a background has been removed, obtained during a quiet period before the event. Widths of auroral filaments are given as the full width at half maximum (FWHM) of a fitted Gaussian profile, and at an assumed emission altitude of 110 km. ASK was deployed at the EISCAT site outside Tromsø (19.2°E, 69.6°N).
 The EISCAT mainland UHF radar was running the tau2pl experiment, providing estimates of the electron density in the ionospheric E and F region at 1.8–5.4 km range resolution and 5 s temporal resolution. The radar was pointing field-aligned.
3. Event Description
Figure 1presents an overview of the event, spanning 45 s of radar data and 10 s of optical data, captured during a geomagnetic storm (Kp-index 7.7). Electron densities show a sharp ionization peak at 110 km from 03:25:00 UT to 03:25:15 UT. This sequence coincides with auroral structure passing through the field of view of the ASK instrument.
 snapshots from the interval show examples of the structure during the event. Thin filaments (with transverse scales in the 100 m range) with complicated shapes evolve very fast in the image. The filaments are seen concurrently with a more diffuse emission which completely or partially covers the field of view of the instruments. The brightness enhancement in the filaments is well above the noise level of this diffuse emission.
 We report here the analysis of two excerpts from the sequence, at 03:25:09 UT and 03:25:12 UT. These subsets are selected to illustrate the characteristic features of the filamentary structure; similar analysis has been done for other parts of the sequence, not shown here.
 The top panel of Figure 2 presents three consecutive images. Filaments form in a deformed ring which develops from a region with diameter of about 2.1°, corresponding to 3.8 km. The edge filaments of the ring have widths as narrow as a few pixels. The evolution is also shown in the supporting material. The morphology is the same in both imagers. We analyze the apparent width of the structure in both cameras, by making cuts transverse to the filament at positions shown in the middle image. The brightness profiles along the cuts (from bottom) are shown in the line plots, together with the fitted Gaussian profiles. Note that IO has been scaled down by a factor of 4. The estimated widths of the filaments are listed in Table 1. The position of the peaks of the fitted Gaussian curves along the slices are also marked in the central ASK image, as a yellow star for and a blue dash for IO. If interpreted in terms of structure transverse to the magnetic fields, the observed widths have a horizontal extent as small as 70 m. Since perspective effects lead to smearing of height-extended features seen at non-zero zenith angles, these widths are maximum values. Furthermore, the perspective leads to separation in the image of emissions from different altitudes, with higher altitude emissions closer to the magnetic zenith position. There is very little separation between the and IO when the images are superimposed, as seen in the line plots of Figure 2. The largest displacement of the peaks is only 3 pixels, for cut 2, which for a peak emission of O2+at 110 km altitude would mean the atomic oxygen emission had its peak at 112 km - a negligible difference. The atomic oxygen emission originates from both direct excitation in the F region and dissociative recombination of molecular oxygen mainly in the E region. The similar emission height of and IO indicates that the atomic oxygen emission is due to the dissociative process, by high energy electron impact.
Table 1. Estimated Widths and Energies of the Auroral Filaments Shown in Figures 2 and 3
 The auroral filaments have brightnesses of up to 2 kR for and 8 kR for IO. They appear within regions of diffuse aurora with brightnesses of down to 0.5 kR for and 2.8 kR for IO. The brightness ratios of the emissions correspond to an energy of up to 10 keV in the filaments.
 In the few seconds of observations, rapidly evolving features are present, some of them lasting only a fraction of a second. Figure 3a shows one such feature captured with the imager at 03:25:12.094 UT. The temporal evolution of the structure within the marked ROI (region of interest) is shown in b. In a sequence of 8 consecutive images lasting 251 ms, the filamentary structure appears and disappears simultaneously in and IO. The structure is dynamic during its lifetime and no periodicity is detected in its appearance. A video of the event can be downloaded from the auxiliary material. For every other frame, the vs IO brightnesses for the pixels within the ROI are plotted, shown in Figure 3c. There is a high degree of correlation between the two emissions. The to I7774 emission ratio is given by the linear fit to the data, and printed in each plot. The emission ratio can be translated to energy, by invoking the results of the combined electron transport [Lummerzheim and Lilensten, 1994] and ion chemistry model [Lanchester et al., 2001] for the ASK emissions, as presented in Figure 3 in Whiter et al. . Running the same model but with atmospheric activity parameters from the night of interest, a slightly modified ratio to energy relation is obtained, where a ratio of 0.25 corresponds roughly to 9.5 keV, and 0.20 to 5 keV. The ratios for each time in Figure 3 indicate energies in this range throughout the times the filament is present. The scatter plots presented here contain noise so that the fitted lines provide a first order estimate of the energy characteristics. Also, the auroral filaments do not completely fill the chosen ROI, which contributes to spread in the data. More detailed analysis will be performed in the future. The widths of filaments taken across two cuts from the second event (marked in Figure 3a as cuts 7 and 8) are listed in Table 1, together with the derived energy of each auroral filament, obtained from the emission ratio of the peak values of the fitted Gaussian profiles to the filaments. All filaments show similar characteristics.
 Precipitating electrons with a lower energy are expected to arrive later in the ionosphere than their high energy counterparts, assuming all the particles were accelerated to different energies at a localized (in time and space) distance above the ionosphere. Since the atomic oxygen emission is more sensitive to low energy precipitation than the O2+ emission, we would expect to see a temporal offset between the two optical signatures, with IO lagging behind if the precipitation contained a distribution of energies. For an acceleration region at 8000 km above the ionosphere, it would take 10 keV electrons 0.13 s to reach the E-region, whereas 4 keV electrons would get there after 0.19 s. With a temporal resolution of ASK of 0.031 s, the corresponding delay between and IO would be 2.5 frames. The lack of such TOF dispersion (Figures 3d and 3e) indicates that the precipitating electrons in the filaments have small energy spread. Similar analysis of the temporal evolution of auroral filaments were also made for other structures during the event. No TOF dispersion could be discerned to the limits of the measurements.
 The O2+ emission can be used as a proxy for incident energy flux, since the brightness does not vary with energy at energies above 1 keV [Whiter et al. 2010, Figure 3]. An enhanced brightness is therefore the result of increased energy flux. The conversion factor derived from the electron transport model is 10 R/mWm−2. The energy flux is found to be moderately high, 230 mW/m2, and the current density resulting from the monoenergetic precipitation is estimated to be 23 μA/m2 in the filament. A ratio analysis of the emissions from the surrounding region give energies of 4 keV, and energy fluxes of up to 100 mW/m2.
 In this study a few seconds of fine scale auroral filaments have been investigated at high temporal and spatial resolution. It is found that the measured O2+ and atomic oxygen emissions occur in a narrow height interval in the ionospheric E region. Comparing the emission ratios with expected ratios derived with an electron transport and ion chemistry model, it is further confirmed that the auroral filaments are the result of high energy (8 keV) precipitating electrons with little energy spread. From the high cadence of the images it is possible to investigate any energy dispersion, in terms of TOF effects between the two optical channels. The images obtained at 32 frames per second showed no signatures of dispersion, instead features in both spectral channels appear and disappear simultaneously.
 Narrow regions of precipitation without any significant velocity dispersion were observed by Boehm et al. , using data from the Freja satellite. They found the structures to originate from transitions in the precipitating flux. A similar case was observed more recently by Dahlgren et al.  from the ground. Lanchester et al. reported fine scale auroral filaments with very high energy and energy flux, embedded in regions of lower energy. The filaments presented here indicate the same characteristics, with mono-energetic structures superimposed on a region of lower energy precipitation. However, there is no clear consensus on how such narrow auroral features are formed. Some contenders are localized acceleration in the topside ionosphere from transient parallel electric fields, sudden pitch-angle scattering into the loss cone or transient variations in the plasma population in the magnetosphere.
 The results are difficult to reconcile with an acceleration by quasi-static potential structures, since the evolution time scale is too short compared with the typical time scales of tens of minutes usually observed for inverted-V precipitation, and the electron energy flux is larger than the few mW/m2 expected from such a structure [Galperin, 2002]. As well, electrons with a range of energies would be created by the sudden appearance of a potential difference.
 Alfvénic acceleration, as has been observed in situ, is also at odds with the present high energies and the lack of TOF dispersions. Pitch-angle scattering would require a sudden and fast process in the magnetosphere to operate on a trapped population of high flux in order to explain our observations, and would reasonably also result in TOF dispersions. A theory is therefore needed which is able to produce high fluxes of energetic electrons, confined to extreme narrow widths, and with energy distributions that are sharply peaked.
 This work was supported by the National Science Foundation under grants AGS-0852850 and AGS-1027247. EISCAT is an international association supported by research organizations in China (CRIRP), Finland (SA), France (CNRS, until end 2006), Germany (DFG), Japan (NIPR and STEL), Norway (NFR), Sweden (VR), and the United Kingdom (STFC).
 The Editor thanks Matthew Zettergren and Donald L. Hampton for their assistance in evaluating this paper.