Solar ultraviolet bursts

The term"ultraviolet (UV) burst"is introduced to describe small, intense, transient brightenings in ultraviolet images of solar active regions. We inventorize their properties and provide a definition based on image sequences in transition-region lines. Coronal signatures are rare, and most bursts are associated with small-scale, canceling opposite-polarity fields in the photosphere that occur in emerging flux regions, moving magnetic features in sunspot moats, and sunspot light bridges. We also compare UV bursts with similar transition-region phenomena found previously in solar ultraviolet spectrometry and with similar phenomena at optical wavelengths, in particular Ellerman bombs. Akin to the latter, UV bursts are probably small-scale magnetic reconnection events occurring in the low atmosphere, at photospheric and/or chromospheric heights. Their intense emission in lines with optically thin formation gives unique diagnostic opportunities for studying the physics of magnetic reconnection in the low solar atmosphere. This paper is a review report from an International Space Science Institute team that met in 2016-2017.

Imager (SJI) yields simultaneous images in four bandpasses. The one at 1400Å is of particular interest as it is often dominated by two Si IV lines, making IRIS the first mission to routinely obtain high-resolution images in TR lines (Appendix A). When combined with simultaneous spectroscopy of these lines, the environment of TR phenomena, including their history, can be identified, facilitating interpretation of their UV signatures in terms of other solar fine structures found with other diagnostics.
In most energetic events, the two Si IV line profiles have factor-two peak and profile ratios indicating optically thin line formation. This is a valuable asset as their intensities therefore represent a more direct measure of energy release than the complex, opticallythick, source function mapping of lines such as the C II and Mg II h & k resonance lines sampled by IRIS. Their dominance of the IRIS 1400Å slit-jaw passband over the continuum contribution wherever there is substantial heating delivers direct, high-cadence, wide-field imaging of such events (example in panel b of Fig. 1).
An early discovery from IRIS (Peter et al. 2014) was a type of very intense, compact brightening in active regions with highly complex Si IV profiles called a bomb ("IRIS bombs", IB) in view of striking similarities with Ellerman bombs (EB;Ellerman 1917). This discovery inspired the first author to convene an International Space Science Institute (ISSI) International Team to study these bombs and other intense UV brightenings detected in IRIS spectra and images. The team met twice, in 2016 January and 2017 March, with the subject "Solar UV bursts-a new insight to magnetic reconnection". The term UV burst was chosen to denote the full scale of IRIS brightening events. Their discussions included the following major issues: How are UV bursts related to EBs and the magnetic field? Can underlying patterns be recognised in different types? Can complex burst line profiles be reproduced with numerical simulation codes?
The present article summarizes these discussions and the conclusions reached at these meetings, and also provides observational criteria to classify UV bursts in other IRIS data.
Section 2 presents our definition of a UV burst. Section 3 summarizes transient TR phenomena observed with previous UV instrumentation. Section 4 presents a summary of observational UV-burst results obtained so far from IRIS, while Section 5 describes efforts in modeling these. A final summary is given in Sect. 6.

UV burst definition
Most solar phenomena are identified from their morphology in images or image sequences. For example, relatively stable structures such as prominences, sunspots and coronal loops are easily identified on individual images, while transient features such as flares and jets are identified from fast-cadence image sequences. Do UV bursts possess distinct signatures that can be readily employed for their identification? Figure 1 shows near-simultaneous snapshots from 10-minute active-region image sequences obtained with IRIS and the Solar Dynamics Observatory (SDO). A movie version covering the whole sequence in the same format accompanies this manuscript. The IRIS 1400Å slit-jaw image (SJI) in panel (b) shows a number of compact bright grains (black in the images) that are much brighter than their surroundings. Viewing the full sequence in the movie shows that their brightness flickers, that some are present for the entire duration, and that others are short-lived. We call such features UV bursts. Figure 1 also shows the co-temporal line-of-sight (LOS) magnetic field from the Helioseismic and Magnetic Imager (HMI; ) on board SDO, and UV images from SDO's Atmospheric Imaging Assembly (AIA; Lemen et al. 2012). The AIA 171Å filter is dominated by Fe IX λ 171.1, formed at 0.8 MK. The 304Å filter is dominated by the He II Lyα-like resonance line and is formed around 80 kK. The AIA 1600Å and 1700Å channels sample the upper photosphere, with magnetic concentrations appearing bright from their evacuation. If chromospheric heating is sufficiently large then the C IV λ λ 1548, 1550Å resonance lines can dominate over the continuum in the 1600Å channel, and this is often the case for UV bursts hence the stronger contrast compared to the 1700Å channel. Note that the 1600Å channel has a significantly broader filter width than the SJI 1400Å channel and the continuum is more intense, thus the SJI channel is much more effective for TR imaging.
Our name "burst" denotes the brief duration of the brightening events. As such it is a standard astronomy term, as in "gamma-ray bursts" and "fast radio bursts". This name has also been used before to describe solar TR phenomena (Section 3.4).
We append "UV" to burst instead of "transition region" to keep closer to observation and because "region" in the latter is somewhat a misnomer. In the actual solar atmosphere every radial column must naturally contain a steep temperature rise, but its height varies tremendously both spatially and temporally and does not define a layer or shell as in traditional one-dimensional equilibrium models such as the classical ones of Vernazza et al. (1981). A sharper and physically sounder description is to define the chromosphere-corona transition at any location and instant as the height where hydrogen reaches 50% ionization-generally a non-equilibrium quantity making up a highly warped, dynamic, and history-dependent surface. There are probably no phenomena restricted solely to this instantaneous transition surface; more likely heating events with increasing ionization come from below (such as the bursts here) whereas cooling events with increasing recombination come from above (such as coronal rain). The "TR" phenomena reported in the older literature so must represent spectral UV signatures of wider-origin events. Our goal is to re-classify them into the latter by exploiting IRIS' UV imaging capability.
Thus, we prefer the term "UV bursts". The UV wavelength range 465 to 1550Å (bounded by Ne VII λ 465 and C IV λ 1550) is indeed dominated by chromosphere-corona transition lines, although there are also some chromospheric and coronal lines in this range. Figure 1 and its movie version give a quick impression of UV bursts; now we define them more formally. They are identified in image sequences sampling ultraviolet passbands that are dominated by a chromosphere-corona transition emission line, and have the following properties: 1. Compactness. Core brightenings 2 in size, typically ≤ 1 . A burst may appear with extended structure (jet, fibril, loop) connected to it, but these are typically less bright. The burst itself may also appear spatially extended into one direction ("flame"), but remains 2 . Two or more bursts may appear close together but remain associated with the same magnetic feature in the photosphere; 2. Duration. Bursts can have lifetimes ranging from tens of seconds to over an hour. For long-lived ones the intensity is not constant, but may flicker by about a factor two on timescales around a minute. At high angular resolution such bursts can also appear as a sequence of intermittent, repetitive flarings, possibly with a migrating footpoint; 3. Intensity. The burst is significantly brighter than the surroundings in the pertinent UV passband. For Figure 1(b) the blue contours correspond to a factor 24 above the image median value. Peak SJI 1400Å intensities of individual spatial pixels of the brighter events can reach factors of 100-1000 higher than the median.
4. Motion. UV bursts show only small proper motions, typically ≤ 10 km s −1 . That is, they generally track dynamics of photospheric magnetic features and their interactions, rather than traveling-front or wave motions along extended structures such as loops; 5. UV bursts are not directly related to flares. This condition distinguishes UV bursts from the compact, intense kernels that appear along flare ribbons and otherwise look similar.
We do not recommend setting a specific intensity threshold for UV bursts, as there is no physical reason to exclude a wide spectrum of events with different energies and temperatures defining the signal in the pertinent wavelength bands. Instead, a dataset-specific threshold suits better to let the observer obtain a manageable number of events for study, for example by specifying a factor above the median intensity value (such as 24 for panel b of Figure 1), or a factor kσ above the median value with σ the standard deviation of the intensity over an image (k ≈ 7 for the 1400Å data in Figure 1). Also note that integrated intensity over the burst area may define a more appropriate threshold than the intensity of the brightest spatial pixel in the burst.
The above criteria also include most IBs identified in the literature, but with the important difference that the latter are defined through their spectroscopic signatures discussed in Section 4. At least one type of event, the narrow line bursts of Hou et al. (2016), do not satisfy the IB criteria while satisfying the UV burst criteria. Therefore we envision IBs to be a subclass of UV bursts, although probably the dominant one.
Section 4 describes key results from the IRIS papers in more detail; we summarize these here: -Si IV line profiles of UV bursts mostly have a complex shape that can include multiple peaks, large excess broadening beyond the thermal width (at temperature T max ), and very extended wings. They often indicate the presence of bi-modal jets; in slanted limbward viewing these may also be mapped into profile differences along different LOSs to the burst. However, some events show simple, Gaussian-shaped profiles not much wider than the thermal width; -UV bursts generally overlie or are directly adjacent to small magnetic features in the photosphere that interact (usually cancelation following convergent proper motion) with similar or larger opposite-polarity features; -Most UV bursts do not show significant co-spatial brightening in the AIA 171Å channel (dominated by Fe IX, formed at 0.8 MK), nor in the other coronal AIA channels, and only rarely in the He II 304Å passband; -EBs are more common than UV bursts, with around 10-20% of EBs having a burst signature while 30-60% of bursts show an EB signature.
These should be considered properties of UV bursts, but are not part of their definition above.
The general consensus from the literature and our ISSI team is that UV bursts are smallscale magnetic reconnection events, and that their complex Si IV line profiles possibly display dynamics associated with a current sheet; for example fast-moving, dense plasmoids 1 , or turbulence induced in surrounding plasma by reconnection outflow jets. The reconnection occurs somewhere in the photosphere-to-upper chromosphere regime, rather deep in the atmosphere. The particular location and also the line of sight to the reconnection site may be responsible for differences in burst signatures such as variations in Si IV profiles. Our UV burst definition requires transition region imaging capability, which was either not available with instruments prior to IRIS or infrequently used (Appendix A). If a slit spectrometer is raster-stepping to build an image scan, only infrequent snapshots of the burst will be obtained, and just one if it is short-lived. If there is only one snapshot the evolution and proper motion cannot be analyzed. However, it may be possible to estimate at least the latter from supporting data, for example, magnetograms (HMI or better) or magneticconcentration monitoring in AIA 1600 and 1700Å images. Also, if the burst intensity happens to be low at the time(s) of snapshot sampling it may fail high-intensity threshold criteria. Therefore raster mode data tend to underestimate the number of UV bursts, but some may get identified properly.
In sit-and-stare mode following mean solar rotation (10 /hr at disk center) the time evolution can be accurately monitored as long as the burst remains within the slit, but if the burst has low intensity during this sampling period it may not be flagged at all. Also, the spatial extent of a burst cannot be established with fixed-slit data, which may lead to erroneous feature classification.

Previous event types
The various types of TR events discovered in the past with UV spectrometry depended to some extent on the capabilities of the instruments, in particular their spatial resolution, spectral resolution and coverage. Table 1 summarizes the characteristics of the instruments treated in this section. Note that due to multiple configurations or changes in time, the listed parameters are not complete; the listed values are typical for TR measurements by these instruments. The table shows that IRIS has better spatial and spectral resolutions compared to other instruments by factors of three and two, respectively.
The first report of strongly broadened TR lines was Brueckner et al. (1976) who presented spectra of TR instabilities observed as very broad features in the C IV λ λ 1548, 1550 doublet observed with the Skylab S082B instrument. This spectrometer had effectively no spatial resolution along the slit, so that further analysis awaited flights of the High Resolution Telescope Spectrograph (HRTS) rocket experiment. HRTS was the first solar instrument to observe the TR at high spectral and spatial resolution. It was flown ten times on rockets between 1975 and 1997 and once as part of the Spacelab 2 payload in 1985 2 . were the next major spectrometers with TR coverage, observing for almost 20 years since 1996. The EUV Imaging Spectrometer (EIS) on board the Hinode spacecraft was launched in 2006 and, although mostly focused on the corona, it observes a number of upper TR emission lines. Finally, IRIS was launched in 2013. It observes UV lines with significantly higher spatial and spectral resolution than previous instruments, and adds highly-valuable slitjaw imaging.
We now discuss different types of TR features that were identified with these instruments in the light of our UV bursts. The emission lines that were most commonly used are given in Table 2, with T max values derived from the CHIANTI database (Dere et al. 1997;Del Zanna et al. 2015;. We suggest that especially the so-called SMM bursts (Section 3.4) and possibly also active region blinkers (Section 3.5) directly correspond to UV bursts, based on the emission measure comparison in Appendix B.

Jets
In recent solar physics the term jet is usually applied to transient, elongated and extending structures that are seen in emission in image sequences. The term surge pre-dates the use of jet in the solar physics literature (e.g., Zirin 1966) and is generally used to describe similar structures that are mostly seen in absorption in chromospheric lines (particularly Hα).
The lack of TR imaging capability prior to IRIS means that the analogs of coronal or chromospheric jets could not be directly identified. Instead TR spectra have been reported of jets identified from other imaging data. For example, Schmieder et al. (1983) studied SMM/UVSP spectra of a surge identified from Hα images, and they measured blueshifts and line broadening in the C IV lines. Similarly, Madjarska (2011) captured SOHO/SUMER spectra of a coronal jet identified from EUV and X-ray imaging data, and found blueshifts of up to 300 km s −1 in several TR lines.
The term jet was applied specifically to a TR spectral feature by Brueckner and Bartoe (1983) who identified intense emission in Si IV and C IV lines that always appeared blueshifted and extended out to 400 km s −1 . Subsequent work showed that these features are quite rare, however (Dere 1994).
Since the launch of IRIS, jets can now be identified directly from TR imaging data, and examples include the quiet Sun network jets of Tian et al. (2014b) and the active region jets of Cheung et al. (2015).
A jet may correspond to a UV burst if there is an intense brightening at the jet base that passes the criteria of Section 2. The column of the jet, where large velocities are expected, generally would not pass the criteria because the emission is typically weak and extends over several or more arcseconds.

Penumbral microjets and sunspot dots
Penumbral microjets were first described by Katsukawa et al. (2007) based on Hinode Solar Optical Telescope filtergram images centered on the Ca II H line. They are small jets occurring within a sunspot's penumbra with lengths typically 1 to 4 Mm and lifetimes ≤1 minute. Many can be identified at any one time within a penumbra. They often have a different orientation than the adjacent penumbral filaments. Cospatial observations with IRIS and the Swedish Solar Telescope (SST) in Ca II  demonstrated that these features also appear in Si IV as a compact brightening at the top of the jet with an intensity enhancement of around five compared to the surroundings. Tian et al. (2014a) found brightness features in IRIS SJI 1330Å and 1400Å images that also occur mainly in sunspot penumbrae, which they termed bright dots. They have a progressively weaker signal in lower atmosphere layers, with most events having no signature in Hα ) Tian et al. (2014a suggested that bright dots have a connection to the Ca II penumbral microjets which was largely confirmed by , but Samanta et al. (2017) identified a distinct class of bright dots without Ca II signature. Penumbral bright dots were also found in 193Å images from the Hi-C rocket flight (Alpert et al. 2016), which were not considered to be coronal features but contributed by cooler TR emission lines in the 193Å passband. Bright dots in a sunspot umbra were identified by Chitta et al. (2016); they lie at the footpoints of coronal loops.
We note that sunspot bright dots are excluded from our UV burst definition due to their relatively weak intensity enhancement in SJI 1400Å images and their large proper motions of 10-40 km s −1 (Tian et al. 2014a). However, since penumbral microjets and some bright dots are found at locations of convective intrusion into strong field, the physics of these events may have similarities to the bursts found in umbral light bridges (Section 4.2).

Explosive events
This name was introduced by Dere et al. (1989) for events found with the third HRTS rocket flight in 1979; earlier they were named "turbulent events" by Brueckner and Bartoe (1983). They were identified as large broadening of C IV λ 1548, on one side of the line only or on both sides but with asymmetrical wings. Wherever both wings were enhanced there was often a spatial separation along the slit of up to 2 . The maximum Doppler shift was not found to vary with the position of the events on the solar disk, suggesting that the corresponding mass motions are isotropic. This led the authors to regard the events as explosions, hence the name "explosive event".
Most of the HRTS-era explosive events were measured in quiet areas, with perhaps the most significant active region event reported by Brueckner et al. (1988) with the HRTS on Spacelab 2 in 1985. This was reported to be only five times more intense than nearby active region plage. Although difference in spatial resolution must be accounted for when comparing intensities of compact features from different instruments, this relatively low enhancement suggests that the various HRTS flights did not observe anything directly equivalent to the IBs of Peter et al. (2014) since those are characterized by much larger intensity enhancements of the Si IV lines.
There were a number explosive event studies from SUMER data (see Huang et al. 2014, for a summary of results), but almost all were focused on quiet Sun or coronal holes as active regions on the disk were rarely observed due to concerns over instrument degradation. One exception is the event presented by Brekke et al. (2000) that showed high velocity features in ions ranging from N IV to Ne VIII (log T max (K) = 5.2-5.8) although, based on the images presented in the paper, it did not show the strong intensity enhancement characteristic of IBs.
Since explosive events were interpreted as reconnection events that generate jets with velocities ≈ 100 km s −1 (Dere et al. 1989), it may be expected that instruments with higher spatial resolution than HRTS and SUMER would be able to image these jets. It is possible that some of the quiet-Sun network jets discovered by Tian et al. (2014b) from IRIS SJI images do correspond to these jets.

Bursts
The use of the word "burst" with regard TR phenomena dates back to the Skylab period, with Emslie and Noyes (1978) and Widing (1982) referring to impulsive solar bursts. These showed sudden increases in UV emission-line intensities, with two events associated with flares and two not. Of direct relevance to the IRIS observations is the study by Hayes and Shine (1987) who reported observations from the SMM/UVSP instrument with the Si IV λ 1402.77 and O IV λ 1401.16 lines that are also observed by IRIS. They defined bursts as events for which the Si IV λ 1402.77 intensity increased by a factor two in one minute. The line width and velocity were not part of this definition. The emission measure analysis presented in Appendix B suggests that they observed equivalents of IBs.
We also note that Innes et al. (1997) used the term burst to refer to a contiguous train of explosive events, lasting for up to 30 minutes. Studies of some IRIS bursts suggest similar behavior (e.g., Gupta and Tripathi 2015).

Blinkers
The term blinker was introduced by Harrison (1997) to describe small-scale brightenings seen in quiet-Sun areas with the Coronal Diagnostic Spectrometer (CDS) on board SOHO. More detailed studies followed by  and , who applied an automatic detection method. They were identified from peaks in light curves for individual emission lines by requiring that the peak be a factor P times the noise level of nearby light-curve minima. In addition, this criterion must be satisfied for N neighboring spatial pixels. Ranges of P and N values were investigated, with P = 5 and N = 3 chosen for statistical analysis. The O V λ 629.7 line was the strongest TR line observed by CDS and was the main reference line used in these studies.
Unlike SUMER, the CDS instrument routinely observed active regions on the solar disk so, although it lacked the spectral resolution required to study broadened profiles of explosive events or IBs, it was well capable of detecting large UV-line enhancements. An early study by  found two intense UV brightenings in the core of a small, recently-emerged active region that were factors of ≈ 50 brighter than the average quiet Sun. The study of active region blinkers of  yielded a wide spectrum of blinker events, with the most intense being around a factor 100 times stronger than average activeregion emission. The emission measure analysis of Appendix B shows that O V blinkers are about an order of magnitude weaker than UV bursts in Si IV, but we consider them to be consistent with the UV burst definition. Young (2004) studied a handful of the most intense active region blinker events and found densities as high as 10 11 − 10 12 cm −3 , although only one event exhibited significantly enhanced non-thermal broadening.
The Hinode/EIS instrument mostly observes coronal emission lines, but a number of lines with T max values from 0.1 to 0.7 MK are also observed, as highlighted by Young et al. (2007), although they are significantly weaker than the strong TR lines observed by CDS, SUMER and IRIS. An intense UV brightening was identified in Young et al. (2007) and considered analogous to CDS active-region blinkers. Despite this, we are aware of only one other comparable brightening reported in the literature, that of Guglielmino et al. (2010). A systematic search for these brightenings in existing EIS data-sets and additional joint EIS-IRIS observing campaigns are required to determine whether EIS events are rarer than IBs or can be equated to the latter.

Ellerman bombs
Ellerman (1917) referred to these as "solar hydrogen bombs", but Ellerman bomb (EB) is the preferred name today. EBs are photospheric features with sizes ≤ 2 and lifetimes of several minutes that have traditionally been identified only in complex emerging active regions, but similar phenomena occur in sunspot moats and were also recently found in quiet-Sun areas (Rouppe van der Voort et al. 2016; Nelson et al. 2017). They are defined by their signature in the Hα line, consisting of strong enhancements in the red and blue wings. Crucially, these wing intensity enhancements are significantly larger than for the much more common quiescent magnetic flux concentrations that cause wing brightening by the hot-wall effect, a phenomenon different from EBs but sometimes causing confusion . Several studies Nelson et al. 2013;Nelson et al. 2015;Reid et al. 2015Reid et al. , 2016 suggest that enhancement of about 50% above the mean at ±1Å from Hα center is a good criterion to separate EBs from flux concentrations. When observed towards the limb at high angular resolution (at least of 0.2 ) EBs are found to have distinctive flame-like appearance in images obtained in the wing of Hα (Hashimoto et al. 2010;Watanabe et al. 2011). Recent 3D MHD simulations have been able to reproduce this tell-tale characteristic remarkably well (Danilovic 2017;. EBs occur where bi-polar small-scale magnetic fields move together and cancel, typically at locations of emerging flux in active regions as stipulated by Ellerman (1917), but also at moving magnetic features (MMF) in moats around sunspots. They are interpreted as magnetic reconnection events occurring below 1000 km height (Georgoulis et al. 2002;Pariat et al. 2004;Watanabe et al. 2011;. Emerging-flux EBs were discussed in the review article of Schmieder et al. (2015).
EBs are sometimes found at the base of Hα surges-large jet-like structures seen in absorption in filtergrams taken in a Hα wing-but less often than claimed by Roy (1973). He also reported surges from EB-like brightenings in light bridges. These different magnetic environments are further discussed in Section 4.2.

UV burst results from IRIS
Most IRIS studies in this context focused on IRIS bombs (IBs) that were introduced by Peter et al. (2014) who named these for similarities with Ellerman bombs. Other terms have also been used, such as hot explosions (Kim et al. 2015), explosive events (Gupta and Tripathi 2015;, compact brightenings (Grubecka et al. 2016) and flaring activeregion filaments . In the present discussion we assume that they are all UV bursts under the definition in Section 2 above.
To illustrate the properties of IBs and UV bursts in this section, we show four examples from the literature in Figure 3. Bursts (a) and (b) correspond to bomb events 4 and 1, respectively, of Peter et al. (2014); burst (c) is the event studied by Gupta and Tripathi (2015); burst (d) is an event from the data-set studied by Toriumi et al. (2015b). For each burst we show the spectral image for Si IV λ 1402.77 (leftmost panels), with the nearest-intime IRIS slit-jaw (SJI) 1400Å image (middle panels) and the corresponding LOS magnetograms from HMI (rightmost panels). The images were spatially aligned using the AIA 1600Å channel images, which also reveal the bursts (yellow contours in right panels). We emphasize the high intensity of the Si IV lines by displaying values of the lines' maximum specific intensities in the left panels. For comparison, an average quiet Sun λ 1402.77 profile has a peak specific intensity of about 800 erg cm −2 s −1 sr −1Å−1 and a full-width at half-maximum of 0.2Å. Average active region line profiles typically have peaks in the 5000-20 000 erg cm −2 s −1 sr −1Å−1 range.
The IBs were identified by Peter et al. (2014) from a 140 × 170 arcsec 2 IRIS raster scan taken on 2013 September 24 during 11:44-12:04 UT. The authors did not explicitly state criteria that defined IBs, but they formulated the four properties given below. We consider criteria (2) to be too strict, as noted in the following text. 1. the Si IV lines are very wide with wings reaching out to ≈ 200 km s −1 separation from line center; 2. the Si IV intensities are enhanced by a factor ∼1000 compared to the surrounding active region; 3. narrow atomic and once-ionized absorption lines are superimposed as absorption blends on the IRIS lines, including the Si IV lines; 4. the ratio of Si IV λ 1402.77 to O IV λ 1401.16 becomes much larger than usual for areas where both appear, suggesting very high densities 10 13 cm −3 .
Whereas very wide lines are typical of explosive events (Sect. 3.3), properties (2)-(4) were not previously reported for such events. Yan et al. (2015) demonstrated that recognizing property (3) requires the high spectral resolution of IRIS.
The line width criterion (1) is required to ensure the line is broad enough to display the narrow absorption blends. Note that it does not refer to the full-width at half-maximum of the line, as is often used for describing Gaussian-shaped lines. The strongest absorption line (Ni II λ 1393.33) lies at −91 km s −1 from the center of Si IV λ 1393.76-see the red profiles in the left panels of Figure 3(a)-(c).
The IB intensity criterion above applies only to the most extreme events, and many of the IBs in the literature do not reach this intensity. Also, since IBs are defined from their spectroscopic signature, then the intensity at the instant the IRIS slit crosses the event is unlikely to be the highest value. This is particularly relevant when trying to associate IBs with features identified from imaging data (e.g., Zhao et al. 2017). Gupta and Tripathi (2015) demonstrated that IB criteria (1), (3) and (4) are valid for most of the lifetime of the event they studied, which was captured with a sit-and-stare observation. If only a single snapshot had been obtained with a raster scan, then the event may have failed criterion (2). Assessing the feature intensity by measurement on IRIS SJI 1400Å images is to be preferred. It has not yet been demonstrated whether events exist that satisfy IB criteria (1), (3) and (4) but have only a weak intensity enhancement, say a factor 5 to 10. To put the IB intensities in context, we first use an IRIS quiet Sun dataset from 2013 October 3 04:20-04:37 UT to measure an average quiet Sun Si IV λ 1402.77 intensity of 160 erg cm −2 s −1 sr −1 . The average intensity 3 of the active region studied by Peter et al. (2014) was 2800 erg cm −2 s −1 sr −1 , and the intensity of the brightest spatial pixel of IB number 1 of Peter et al. (2014) was 5.2 × 10 5 erg cm −2 s −1 sr −1 . We caution that the sensitivity of IRIS has degraded significantly since launch, with version 4 of the radiometric calibration (as implemented through the IDL code iris_get_response) showing a decrease in sensitivity of a factor 4 at 1402.8Å between 2013 August and 2017 August. It is therefore recommended to always quote intensities in calibrated units rather than data number (DN) units in order that bursts at different times can be compared.
One feature of UV bursts is the flickering of their intensities over time. Light curves constructed from the IRIS/SJI 1400Å images are shown by Peter et al. (2014),  and Kim et al. (2015) and reveal that the intensity evolution of a UV burst consists of many individual peaks that are often close to the SJI temporal resolution. The cadence per SJI channel is usually lower than for the spectrometer exposures due to cycling between different SJI channels. The fastest possible cadence of SJI imaging is 1.7 seconds, and we have identified one data-set beginning at 09:08 UT on 2016 October 16 that has this cadence for the SJI 1400Å channel and contains bursts. Figure 2(a) shows an image of one burst from this dataset, captured at the peak brightness. Panel (b) shows the light curve of the burst and the inset plot shows a close up of the variability near the peak intensity. Individual spikes at the sequence cadence can be identified, which suggests that the full variability of the burst is not being captured at this cadence. Gupta and Tripathi (2015) presented sit-and-stare observations that captured a UV burst with a five second cadence. Figure 7 of this work shows that the Si IV emission line light curve exhibits strong variability on this timescale. These authors applied wavelet analysis to identify periods of 30 and 60-90 seconds in the C II and Si IV spectral lines. We note that flickering has previously been reported for EBs, for example by Pariat et al. (2007) and Watanabe et al. (2011).  Chitta et al. (2017). In the latter case a MMF was studied and redshifts were found on the forward side of the MMF, blueshifts on the rearward side, which was interpreted as reconnection jets tilted with respect to the line of sight.
A defining feature of IBs is the presence of atomic or once-ionized blends superimposed on the broad Si IV emission lines. The deepest of the absorption lines is usually Ni II λ 1393.33, superimposed on Si IV λ 1393.76. Sometimes the latter is not returned in the IRIS telemetry stream, however, and absorption is less easy to identify in Si IV λ 1402.77 (compare profiles in the leftmost column of Figure 3). When present, the narrow absorption dips have relatively small Doppler shifts (generally less than 10 km s −1 ) and betray the presence of relatively cool gas (hydrogen and helium predominantly neutral) along the line of sight to the burst that produces the wide profiles.
Occasionally absorption features near the rest velocities of the Si IV lines can be identified, and examples have been shown by Yan et al. (2015) and . We note that the latter paper suggested the absorption occurs in an overlying layer rather than due to the burst emission itself being optically thick.
In addition to the complex line profile patterns, bursts with very narrow, Gaussianshaped Si IV line profiles have been reported by Hou et al. (2016). They were found above sunspots and the authors suggested a connection with sunspot plumes. The most intense event had an intensity of 3.5 × 10 5 erg cm −2 s −1 sr −1 in the brightest spatial pixel, comparable to the IBs of Peter et al. (2014). These events thus appear to satisfy the UV burst criteria, but the narrow line profiles disqualifies them from being IBs. The ISSI team found other examples of intense, narrow-line bursts, and Figure 4 shows an example from 2013 October 23, although this one is not above a sunspot. A key requirement is to distinguish narrow-line bursts from mini-flare ribbons (Section 4.4) and dynamic loops (Section 4.5), that may also show narrow line profiles. A wider survey of narrow-line bursts and a study of their underlying magnetic field structure will be worthwhile.
The complex UV burst profiles suggests that some details of the reconnection physics are being revealed in the UV lines; a first modeling effort was performed by Innes et al. (2015) who considered how plasmoids in a current sheet may lead to line profile shapes as observed, in particular triangular profiles. New high spatial resolution chromospheric images from the CHROMIS instrument (Rouppe van der Voort et al. 2017) show tiny brightenings of sizes close to 0.1 that are associated with a burst and may correspond to plasmoids. The authors also simulated Si IV emission using a 2.5D radiative MHD simulation and found that plasmoids could be responsible for the complex Si IV profiles if the line-of-sight passes through multiple plasmoids along the current sheet.

Magnetic environments and signatures
The UV bursts reported in the literature are generally characterized by rapid evolution of small-scale magnetic elements on the photospheric surface, often evidenced in the 45 second cadence LOS magnetograms from HMI but sometimes requiring better resolution and sensitivity. Three types of magnetic environment are recognized to harbor UV bursts: 1. Emerging flux regions (EFRs) in complex active regions. Small flux elements in these display fast streaming motions; cancellation occurs regularly. 2. Moving magnetic features (MMF) in sunspot moats. These can have opposite polarity to the spot in sea-serpent patterns (Harvey and Harvey 1973) with cancellation against same-polarity features. 3. Light bridges (LB). Elongated features with highly-sheared magnetic field that cross sunspot umbrae, or occur close to them (Solanki 2003). HMI magnetograms are shown in the rightmost column of Figure 3. Events (a)-(c) are located within 1 of small negative polarity features with vertical flux densities of around 200-400 G. Event (d) from Toriumi et al. (2015b) is located to the north side of a light bridge running east-west and does not show an obvious compact magnetic feature, but the authors noted a continuous supply of field with LOS flux densities of 200-400 G within the light bridge that kept driving brightenings. Hinode Spectropolarimeter data showed that this field was highly inclined with horizontal flux densities of around 1000 G. This highlights the importance to obtain the full vector magnetic field when studying UV burst evolution.
Light bridges are believed to be convective intrusions into strong background magnetic field and have mostly horizontal field (Leka 1997). Reconnection between this field and the surrounding vertical field generates EB-like phenomena and surges of cool plasma (Roy 1973). Recent, high-resolution Hα data from the SST show that the surges resolve into a "fan" of many jets side-by-side (Robustini et al. 2016). IRIS data have demonstrated that UV bursts are rooted in LBs at the bases of the surges (Toriumi et al. 2015a; Hou et al.  The line-of-sight (LOS) magnetograms from HMI are valuable for studying the magnetic field evolution of UV bursts, but the precise magnetic field topology requires vector magnetograms and/or extrapolation of the field into the corona. Vector magnetograms are available from HMI at 720 second cadence, but the noise level for the transverse field is about 100 G compared to about 10 G for the LOS magnetograms, not sufficient for the small field concentrations that typically underly or cause UV bursts.
A particular scenario believed important for EBs and UV bursts is U-loop reconnection, in which reconnection takes place in the arms of a U-shaped magnetic field line ( Figures 5  and 6). Such a field configuration can occur where flux emerges through the photosphere and mass-loading at the center of the loop prevents it from rising into the atmosphere. When viewed in a LOS magnetogram, the field shows two distant footpoints and between them an apparent, close bipole where the field is U-shaped rather than having the Ω shape of normal field configurations. Such configurations are sometimes referred to as "bald patches"-see the magnetic field lines at the photosphere in Figure 5. The difference with Ω sea-serpent configurations can only be recognised from vector magnetograms of sufficient quality to establish that the field connects the two polarities underneath rather than above. This was done by Georgoulis et al. (2002) for an emerging flux region observed in the Flare Genesis balloon flight, leading them to suggest the U-shaped loop reconnection scenario for EBs ( Figure 5) that was later adopted also by Peter et al. (2014) for IBs ( Figure 6) Due to this nature of the field configuration the reconnection necessarily occurs in the low atmosphere.
A different scenario was presented by Chitta et al. (2017), who were able to identify a fan-spine magnetic topology for a MMF UV burst by applying a magnetofrictional relaxation technique to HMI LOS magnetograms. This technique begins with a potential field extrapolation from an initial magnetogram, which is then evolved to a series of non-linear force-free field states by inputting the subsequent magnetograms as new boundary conditions. The null point of the fan-spine system was found to be at 500 km, and the reconnection driving the UV burst was suggested to be due to shearing at the null point, driven by the motion of the MMF. Zhao et al. (2017) and Tian et al. (2018b) combined vector magnetogram data from HMI with a MHD relaxation technique (Zhu et al. 2013(Zhu et al. , 2016 to investigate the magnetic environment of UV bursts. Zhao et al. (2017) distinguished UV bursts associated with bald patches and those with flux cancellation. Tian et al. (2018b) also found that some UV bursts were associated with bald patches for a different emerging flux region, but most of the bursts were not. A better correspondence was found with locations with a high squashing factor that are often associated with currents and magnetic reconnection with regard to the generation of flares.
The evidence from the magnetic field extrapolations is that reconnection takes place at heights of 0.5-1.0 Mm (Chitta et al. 2017;Tian et al. 2018b). Further evidence for these low heights comes from the weakness or absence of coronal emission (Section 4.7) and the appearance of the narrow cool-gas blends on the IB profiles. Note that these blends do not require the presence of cool gas on top of (and part of) the bomb as suggested by Peter et al. (2014), but may originate in cool gas along a slanted LOS to the bomb base as in Figure 2 of Rutten (2016). Such gas then also absorbs any coronal EUV radiation from the burst at wavelengths below the hydrogen photoionization edge at 912Å.
High-quality vector magnetic field measurements are highly desirable for UV bursts, both high spatial resolution and high sensitivity. Although current ground-based instruments are capable of much better measurements than HMI, the polarization data are much more susceptible to bad seeing so that homogeneous good-quality sequences covering the entire duration of UV bursts have not yet been obtained. This is one area where the 4 meter Daniel K. Inouye Solar Telescope (DKIST) should offer large improvement. Unipolar magnetic field regions show a different type of intensity enhancement, with examples presented by Toriumi et al. (2017) for a small EFR. In addition to the bursts in the center of the EFR, which were interpreted as bald patch events, a set of brightenings above the unipolar patches at the edge of the EFR were identified. These were interpreted as shockheating events driven by plasma flowing down the legs of the rising arch filament system. The unipolar region events showed weaker Si IV intensities than the bald patch bursts, with an enhancement of only around five compared to quiet regions; these may not generally qualify as UV bursts.

Ellerman bombs
Peter et al. (2014) recognized the connection between IBs and EBs by naming the events "bombs". Their U-loop reconnection scenario was partly based on the similar one for EBs by Georgoulis et al. (2002). However, without access to simultaneous Hα observations, the authors could not make the IB-EB connection. (The later work of Grubecka et al. 2016, on the same data-set could, however, make a connection to EBs for IBs 3 and 4 based on signatures in the Mg II lines-see below.) Direct IB-EB connections were demonstrated by , Kim et al. (2015) and , who each had access to simultaneous ground-based Hα data and found examples of EBs that had a clear IB signature, i.e., intense, strongly-broadened Si IV emission lines with superimposed narrow cool-gas blends.  identified 10 IBs based on the presence of broadened Si IV lines with such superimposed absorption features. Three IBs were clearly matched with EBs, four had no EB signature, the remaining three were ambiguous. They also noted that while the IRIS slit crossed about 30 EBs only 6 had an IB signature.
As simultaneous, high-resolution, spectroscopic Hα data are rarely available for IRIS observations some authors have sought to find EB identifiers in the IRIS data themselves. Grubecka et al. (2016) found that 1D radiative transfer models that fit the Hα wings of EBs may also fit the wings of the Mg II lines, and therefore used Mg II h line wing emission to find EBs in their IRIS data-set. They found that samples at −3.5 and +1.0Å (−374 and +107 km s −1 ) are suited for identifying EBs, and subsequently identified 74 events that were a factor two brighter than their surroundings. Around 10% of these events had a signature in the SJI 1400Å channel.  used narrow-band filtergrams in the wings of Hα to identify EBs and found that averaging images obtained at ±1.33Å (±143 km s −1 ) from the line core of Mg II k is a good proxy for Hα EBs. Another method for detecting EBs with Mg II was suggested by Hong et al. (2017b) who integrated a region ±0.25Å around the two members of the Mg II triplet lines at 2798.75 and 2798.82Å, and found a good correlation with Hα wing enhancements found with the Fast Imaging Solar Spectrograph on the Goode Solar Telescope.
The ubiquitous availability of AIA full-disk images offers another avenue for finding EB signatures, with the AIA 1700 and 1600Å passbands the most promising. ,  and  compared Hα observations from SST/CRISP with AIA and found good correlations between Hα EBs and AIA mid-UV brightenings. The last paper set as EB criterion ≥ 8σ above the mean 1700Å intensity over the whole active region. Further work by G. Vissers has suggested that precise one-to-one correspondence is not possible but, as noted by , the 1700Å brightenings are of interest in their own right. Therefore, a statistical study of bursts identified through intensity-thresholding in the AIA 1600Å and 1700Å channels would be valuable, particularly as these full-disk data mean that active regions can be tracked continuously from their initial emergence. Surveys of the magnetic features (e.g., MMFs, EFRs and LBs) underlying these bursts can also be made using the co-spatial and co-temporal HMI magnetograms.
Finally we emphasize that the results of Grubecka et al. (2016) and  suggest that 10-20% of EBs have an IB signature.  have also suggested that 30-60% of UV bursts are co-spatial with EBs. Since Section 5 below shows that it is hard to achieve both EB signatures and UV burst signatures from a single reconnection event when modeling EBs and IBs, then obtaining improved statistics on simultaneous EBs and bursts is desirable.

Mini flare ribbons
Our definition of UV bursts in Sect. 2 excludes bright kernels in flare ribbons, but active regions can also display what we call "mini flare ribbons" that are small in size (say, 10 or less) and have either no or very weak signal in the 1-8Å channel of the Geostationary Operational Environmental Satellite (GOES) X-ray monitor. The Si IV intensities of such events are comparable to UV bursts, but a key distinguishing feature is that the lines are mostly red-shifted, consistent with the profiles seen in most normal flare ribbons. The ribbons probably correspond to coronal nano-or micro-flares that would be visible at X-ray or EUV wavelengths.
The only example in the IRIS literature is the event described by Bai et al. (2016) which took place in the penumbra of a small sunspot. Multiple transient brightenings took place along a ribbon of length 10 , with the brightest comparable with the Peter et al. (2014) IBs. The Si IV profiles were mostly Gaussian shaped and red-shifted by 15-20 km s −1 . The ribbon also exhibited coronal emission in the AIA channels, which is not typical of UV bursts (Sect. 4.7). Similar events have been found by the ISSI team and an example is shown in Figure 7. The IRIS slit crossed a very intense, compact brightening that is visible in panel (a) and could be interpreted as a UV burst. Note the "chain" of weaker brightenings extending from the main brightening that is reminiscent of a flare ribbon. The Si IV λ 1402.77 line profile is predominantly red-shifted and significantly narrower than the profiles of Figure 3. The line's amplitude, however, is comparable to those of Figures 3 and 4. About two minutes after the IRIS image was taken, the AIA 94Å bandpass (panel c) shows a loop running northwards from the location of the intense IRIS brightening to another strong brightening (not shown in panel a). This loop is clearly filled with plasma at ∼ 10 MK due to chromospheric evaporation, as expected from the standard flare model. Panel (d) shows the GOES X-ray flux at this time, where a small peak is seen shortly after the time of the AIA image. It is not clear if this peak arises from the event as the GOES flux is an average over the entire solar disk, but if it is then the energy of an A-class flare is implied. (We note that this event was taken from the same IRIS data-set used by Gupta and Tripathi 2015, but they selected a different feature for study that was discussed earlier.) To distinguish a UV burst from a mini flare ribbon, the key features to check are (i) an extended ribbon structure, (ii) a predominantly red-shifted line profile, and (iii) loop emission in the "hot" AIA channels at 94 and 131Å that would imply chromospheric evaporation has taken place. Bai et al. (2016) referred to their event as a nanoflare ribbon after estimating the thermal energy from AIA imaging. We prefer the term "mini flare ribbon" in order to cover a wider range of energies, and suggest that there is a need for a survey of such events to investigate how they compare with larger flares.

Dynamic loops (arch filament systems)
Another class of events in SJI 1400Å images that show strong UV-line emission but not considered by us to be UV bursts are dynamic TR loops. These are mostly seen in emerging flux regions, and seem to be analogous to the fibrils of arch filament systems that are seen in absorption in Hα (Bruzek 1967). Examples have been presented by Yan et al. (2015 and . The loops span the emerging flux region and are typically 5-40 long. Propagating intensity fronts in SJI 1400Å image sequences suggest flows along these loops are common. The line profiles often show a relatively narrow central component with weaker, but very extended wings. We show an example from the data-set studied by Yan et al. (2015) in Figure 8. These authors showed in their Figure 3 a more intense loop spectrum with a complex line profile. We consider Figure 8 to be a more representative dynamic loop spectrum, and Figure 5 from  shows a similar example from a different data-set. Post-flare loop arcades can also exhibit similar profiles, such as those shown in Figure 4 of Brannon (2016). The Si IV intensities of dynamic loops are generally lower than those of the most intense UV bursts. For example the peak amplitude of Si IV λ 1402.77 for the event shown in Figure 8 is 9000 erg cm −2 s −1 sr −1Å−1 , which can be compared with the numbers shown in blue in Figures 3 and 4. Zhao et al. (2017) applied magnetic field extrapolation to an emerging flux region (see also Section 4.2), attributed the low-lying field lines in their extrapolation to an arcade of dynamic loops seen in SJI 1400Å, and associated the latter with reconnection between the emerging flux and the overlying magnetic field at quasi-separatrix layers (QSLs). Previously, Georgoulis et al. (2002) had identified the connection between QSLs and EBs for an emerging flux, and they also identified bright fibrils in the TRACE 1600Å channel that they interpreted as due to bright C IV emission. These were likely the equivalent of the Si IV dynamic loops.

Flaring fibrils
A study of EBs identified from the Flare Genesis Experiment revealed co-spatial brightenings and dynamic loops in the TRACE 1600Å channel (Georgoulis et al. 2002;). The loops were referred to as "flaring arch filaments", and  used this notation to describe compact brightenings in AIA 1600Å images, similar to EBs but with obvious elongated morphology and proper motion along filamentary strands that they interpreted as due to C IV line emission lines in this passband. Rutten (2016) subsequently proposed that "flaring active-region fibrils" (FAFs) is a better name as it avoids confusion with the dynamic loops of arch filament systems referred to in the previous section.  compared AIA FAFs with IRIS data and found intense, IB-like Si IV profiles at the locations of the 1600Å brightenings, with jets or loops in the SJI 1400Å images and shell-like fronts expanding away from them in the hotter AIA passbands. To highlight the difference between the dynamic loops described in the previous section with FAFs, we show a SJI 1400Å image from the 2014 June 15 dataset studied by  in Figure 9a. There are many dynamic loops in this image, and they can be seen to have a generally smooth variation in their intensities from footpoint to footpoint. The FAFs are the brightest loops in the image which appear to be connected to compact brightenings. Note that the brightenings can be identified as discrete features-the inset image of Figure 9(a) shows one of the FAF brightenings on a linear intensity scale, revealing a distinct brightening. In particular, the brightenings are not simply small segments that brighten during the loop's evolution. Si IV λ 1402.77 spectra from three locations are shown in panels (c)-(e). (c) and (e) correspond to FAF brightenings, and show the broad, very intense profiles of IBs (compare with Figure 3), although (e) does not show the cool absorption blends. Panel (d) shows a profile from a fibril that connects to the FAF brightening, and it is more comparable to the dynamic loop profile shown in Figure 8, although the intensity is an order of magnitude larger. The FAF brightenings we consider to be UV bursts, and a plausible interpretation is a low-lying reconnection event that is able to connect to the overlying arch filament system and deposit heated plasma in certain loops. Another event in the literature that we consider to be a FAF is the one shown in Figure 13 of , which was described as the footpoint of two interacting loop systems that exhibited explosive event line profiles.
To summarize the terminology, the FAF is the combination of the loop (fibril) and the burst that evolve together. Peter et al. (2014) noted that the four IBs they observed did not have a signature in the coronal AIA passbands. This seems to be a common feature of UV bursts, but not an absolute rule. For example, the event of Gupta and Tripathi (2015) showed weak AIA signals, allowing the authors to derive a differential emission measure curve, and  described thin long arcs seen in the hotter AIA diagnostics spreading away from FAF sites.

Coronal signatures
The weakness or absence of co-spatial coronal emission either implies that the bursts are not heated beyond about 10 5 K, or that the EUV emission is blocked by overlying cool plasma with sufficient neutral hydrogen and helium that all lines below 912Å are strongly attenuated. IRIS does observe a coronal line above 912Å-Fe XII λ 1349.40, with T max = 1.6 MK-but this line is very weak and has not yet been reported from a UV burst observation.

UV burst modeling
Our interpretation of UV bursts as magnetic reconnection events in the low atmosphere means there are three components to modeling that must be considered: (1) the atmospheric evolution that leads to reconnection at low heights in the atmosphere; (2) the physics associated with the reconnection current sheets; and (3) the effect of dynamics and heating on spectral line profiles. Our definition of a UV burst requires the simulations to produce a compact, very intense, flickering brightening in a TR line (with T max ∼ 100 kK). Since most UV bursts also show the complex line profiles of IBs and many also the cool-line absorption blends, then these are additional tests for the models to pass.
UV bursts are relatively new discoveries so simulations directly focused on them are few, but a number of works have considered how simulations relevant to EBs may produce UV bursts. In particular, over the past 20 years increasingly sophisticated codes for modeling flux emergence have been performed and demonstrated to produce the U-loop reconnection that has been suggested for EBs and IBs (Figures 5, 6). The physics of magnetic reconnection current sheets has also been studied extensively, and may be particularly relevant to the often complex line profiles of UV bursts. Another focus of modeling has been on whether 1D radiative transfer models that are able to reproduce the profiles of strong chromospheric lines from EBs are consistent with UV bursts.
1D modeling of EBs has a history going back to Kitai (1983), with the most basic aim to reproduce the wings-only brightening of Hα. The method involves the perturbation of a plane-parallel atmosphere, usually by the insertion of a hot component near the temperature minimum region (heights of ≈ 450 km). NLTE radiative transfer is included and profiles of strong lines such as Hα, Ca II H & K and Mg II h & k are modeled. Recent work in this vein includes Berlicki et al. (2010), Berlicki and Heinzel (2014), Grubecka et al. (2016), andFang et al. (2017), and the more sophisticated 2D NLTE modeling by Bello González et al. (2013). Cloud modeling introduces an additional plasma component to simulate the effect of absorption by overlying chromospheric fibrils on the line core, and a recent example is the work of Hong et al. (2017b). Comparisons of the models with observations typically constrain the hot component to be about 100-3000 K above the background temperature (e.g., Grubecka et al. 2016). Note that, since the models are focused on effects at the temperature minimum region, the atmospheres are usually truncated at 20 000 K or lower, well below the T max of Si IV.
The discovery of IBs led Fang et al. (2017),  and Hong et al. (2017a) to investigate whether the EB models could be modified to produce Si IV emission. Fang et al. (2017) considered temperatures of the hot component up to 15 000 K, but then the chromospheric signatures were not consistent with the EB observations. Both  and Hong et al. (2017a) used the RADYN code (Carlsson and Stein 1997), which couples radiative transfer with hydrodynamics, to allow the plasma to respond to the heating. Again the authors found that it was not possible to reconcile Si IV emission with the chromospheric line profiles of EBs. We highlight that, as discussed in Sect. 4.3, only 10-20% of EBs have been found to have a hot UV burst signature, thus the failure of the 1D models to account for the hot emission does not invalidate the models for the much more common, cooler events. In addition, it is possible that non-equilibrium effects such as κ electron distributions can cause Si IV to be formed at much lower temperatures (Dudík et al. 2014). A similar effect may also arise if EBs are formed at sufficiently high density that Si IV has near-Saha-Boltzmann opacity, giving a formation temperature of 10-20 kK (Rutten 2016). However, within this cooler regime no attempt has been yet made to reproduce the non-visibility of EBs in the Na I D and Mg I b lines, already mentioned by Ellerman (1917) and confirmed by , which may be difficult to reconcile with 1D temperature humps covering the formation heights of these lines.
There are a number of simulations that focus solely on the physics of magnetic reconnection current sheets in the solar atmosphere. They do not investiate the wider plasma evolution that leads to the sheet formation, but they do resolve the current sheet at higher resolution than large-scale MHD codes. Ni et al. (2015Ni et al. ( , 2016 investigated plasma heating in current sheets located at atmospheric heights of 100 to 1000 km, and with horizontal and vertical orientations. Plasmoids are formed in the current sheet, and heating occurs where slow-mode shocks from the reconnection site interact with the plasmoids. They found that the plasma β (ratio of plasma pressure to magnetic pressure) is crucial in determining if UV burst temperatures are produced: β 0.1 enables heating to 80 kK (the temperature of formation of Si IV). The further work of Ni et al. (2018) found that the inclusion of nonequilibrium ionization in current sheet simulations results in a faster reconnection rate but smaller temperature increases, potentially affecting the formation of Si IV. Innes et al. (2015) also studied current sheet dynamics in relation to burst-like events recorded from IRIS. They highlighted that the break-up of a current sheet into plasmoids separated by magnetic field x-points at which particles are accelerated can give rise to complex velocity distributions. Modeling the Si IV emission lines integrated along the current sheet was found to give realistic, IB-like profiles. They also highlighted that a relatively small change to the line-of-sight had a significant effect on the profiles.
Another 2D simulation that produced current sheet plasmoids, but this time as part of the wider evolution of solar atmospheric structures was presented by  and Rouppe van der Voort et al. (2017). The interaction of an emerged bipole with pre-existing magnetic field was simulated, leading to a surge and a current sheet. The latter occurred where the "dome" of the emerged flux pushed against opposite polarity field, and the simulation resolution was sufficient to resolve plasmoids. By choosing a line-of-sight along the current sheet, Rouppe van der Voort et al. (2017) demonstrated that complex, IBlike Si IV profiles can be reproduced. Note that the reconnection in this scenario is not the U-loop reconnection discussed in Sect. 4.2, but that more commonly associated with jets (e.g., Figure 8 of Shimojo and Shibata 2000). This is also the scenario inferred by Chitta et al. (2017) for a MMF burst.
There are many time-dependent simulations of emerging flux in the literature (see review of Schmieder et al. 2015), and a particular issue relevant to UV bursts and EBs is the occurrence of U-loop reconnection (Sect. 4.2 and Figures 5,6). This has been demonstrated to arise naturally on account of the Parker buoyancy instability acting on the flux and giving rise to so-called serpentine field lines. A 2D resistive MHD simulation of an emerging flux sheet was presented by Isobe et al. (2007), and it showed plasma heating and flows at the sites of U-loop reconnection that were identified with EBs. A similar result was found by Archontis and Hood (2009) but for a fully-3D resistive MHD simulation of flux sheet emergence. Further advances were made by Tortosa-Andreu and Moreno-Insertis (2009) and Cheung et al. (2010), who performed 3D simulations for the emergence of twisted cylindrical and semi-torus flux tubes, respectively, demonstrating U-loop reconnection in both cases. The latter work effectively demonstrated that EBs can occur during the formation of a bipolar active region. These latter two simulations both used the MURaM code (Vögler et al. 2005) and included simplified radiative transfer effects. More recently the code has been applied to the case of an emerging flux sheet by Danilovic (2017), but with a more sophisticated treatment of radiative transfer, including detailed Hα synthesis. The model produces realistic EB signatures, but the atmosphere does not extend beyond the chromosphere and so UV bursts are not produced.
The Bifrost code of Gudiksen et al. (2011) does extend to the chromosphere and corona, allowing the evolution of flux emergence to be studied in these layers. This was done in full 3D by  who explicitly address EBs, UV bursts including IBs, and microflares. It includes spectral synthesis of Hα, Mg II h & k, and the Si IV 1393.7Å line including the Ni II blend for selected simulation snapshots. There are many striking agreements between the simulated and observed appearances of both EBs and UV bursts. In the simulation the EBs are photospheric in origin and remain generally under 10 000 K; the IBlike UV bursts are chromospheric and reach higher temperatures although their total energyrelease content is far less. The microflares go higher up and reach coronal temperatures. In this simulation they are all the result of reconnection between the legs of dragged-down U-loops and expanding Ω -loops that are part of serpentine flux emergence. Unlike the 2D codes discussed earlier, the simulation resolution is not sufficient to resolve plasmoids in the current sheets, but broad Si IV line profiles are produced from bi-directional flows at the reconnection sites. A remaining discrepancy is that the simulated EBs show Si IV emission only barely or none at all, whereas this has been observed for 10-20% of EBs. This version of the Bifrost code did not yet account for ion-neutral separation (ambipolar diffusion) whereas tests have shown in the meantime that this can substantially increase the rate of reconnection , which may impact the predicted emissions of EBs.

Summary
The IRIS instrument has revealed a set of intense, transient, compact brightenings that occur in active regions and are visible in the resonance lines of Si IV. In this work we advocate that these events be referred to as UV bursts. The definition of UV bursts is given in Section 2 and it includes the set of events that have been called IRIS bombs in the literature. The latter were introduced by Peter et al. (2014), and are defined through their spectroscopic signature in the Si IV lines.
Studies of UV bursts are ongoing, but they are widely believed to be small-scale reconnection events occurring at heights from the photosphere to the upper chromosphere. As the events reach TR temperatures then they have special advantages for investigating reconnection in the solar atmosphere. The TR is thin (typically 10's of kilometers) and so the large energy output from reconnection gives a very strong signal in TR lines that overwhelms the background emission and allows fine details in the line profiles to be studied. In contrast a reconnection event in the corona takes place within a volume that is 10's of megameters thick, and thus gives a relatively subdued signal unless it is a significant size flare. The TR has a number of very strong lines in the far ultraviolet where instrument sensitivity, spectral and spatial resolution are generally higher than for the EUV and X-ray regions, where most coronal lines are found. In the chromosphere, the lines are mostly formed optically thick, and thus reconnection events in this region are mostly studied only in the lines' wings. However, the Fabry-Pérot spectroscopic imaging instruments commonly used have relatively narrow wavelength coverage and thus do not cover the full spectral extents of the lines. The TR lines in the UV therefore offer excellent opportunities for studying the physical processes of magnetic reconnection and heating in the solar atmosphere.
The early work on UV bursts has suggested certain common features: they mostly have weak or non-existent signal in the coronal imaging channels of AIA; they generally overlie small-scale, dynamic magnetic features seen in LOS magnetograms from HMI, and emerging flux regions, moving magnetic features and light bridges are the most common locations; and about 10 to 20% of UV bursts are cospatial with EBs, which are generally interpreted as photospheric reconnection events.
A number of modeling efforts have already been applied to UV bursts with some success (Section 5), but one common feature is that it has not been possible to achieve both an EB signature and a UV burst signature from the same event.
Despite the progress already made in describing and understanding UV bursts, we identified some areas for further work with the IRIS data in the preceding text that we summarize here.
-Coordinated studies with Hinode/EIS and IRIS to determine the maximum temperatures reached by the events. The next solar UV spectroscopic capability for solar physics is the SPICE instrument (Fludra et al. 2013) that will be flown on Solar Orbiter. This will observe two wavelength bands at 704-790 and 973-1049Å, which give excellent coverage from the chromosphere through the transition region to the corona, and thus will allow the UV burst heating to be tracked through the layers of the atmosphere. The spatial and spectral resolutions will be somewhat lower than those of IRIS, but will be sufficient to see UV bursts.
The Atacama Large Millimeter/submillimeter Array (ALMA; Wedemeyer et al. 2016) promises superb new capability for UV-burst studies when the (far from trivial) technological problems of achieving long-baseline, many-telescope observations including full calibration become solved. Since hydrogen ionizes in UV bursts, their free-free hydrogen opacity becomes very large, even exceeding Lyman-α at high temperature, and since free-free extinction is an LTE process ALMA will directly measure temperatures of UV bursts wherever these are optically thick, at outstanding angular resolution.
DKIST is expected to become operational in late-2019/early-2020 and will resolve structures down to 25 km size and enable high quality vector magnetograms through higher polarimetric sensitivities compared to other ground-based telescopes. UV bursts can not be directly studied as there is no transition region capability, but EBs will be a prime target and offer the opportunity for the reconnection process to be studied in unprecedented detail.
Longer term, a new UV/EUV spectroscopic mission with improved spatial resolution compared to IRIS and a more complete temperature coverage would be ideal for UV bursts. The rapid variability of UV bursts makes sub-second temporal cadence desirable, which is currently limited by the readout times of UV-sensitive CCDs. The recent report 4 of the Next Generation Solar Physics Mission Science Objectives Team (NGSPM-SOT) recommended a coronal/transition region spectrograph with 0.3 spatial resolution, and Teriaca et al. (2012) presented an instrument concept called LEMUR (Large European Module for solar Ultraviolet Research) that is compatible with the NGSPM-SOT recommendations. Such an instrument would provide new advances for UV burst studies, particularly when combined with DKIST.

A Transition region imaging
Imaging the solar atmosphere through narrow bandpass filters enables particular temperature layers to be isolated and studied at high temporal resolution. At visible wavelengths, Fabry-Pérot filters have long been used and they enable high cadence spectral imaging in chromospheric lines such as Hα and Ca II H & K. For shorter wavelengths (≈ 100-500Å) multilayer coatings on optical surfaces enable good reflectivity over a relatively narrow wavelength range (typically 10% of the central wavelength). If the filters are centered on strong, isolated emission lines then the images can be a good means of isolating a particular temperature regime in the atmosphere. This has been exploited by several space-based instruments with the most advanced being SDO/AIA, which obtains full-disk solar images in seven EUV filters between 94 and 335Å with around 1 arcsec spatial resolution and a 12 second cadence. The 100-500Å region contains mostly coronal emission lines, with the only strong, isolated cooler lines being He II λ 304 and Ne VII λ 465. The former is observed with AIA and is formed around 80 kK, however the formation of this line is optically thick, is complex and so the emission not easy to interpret. The Ne VII line is from the upper transition region (0.5 MK), but has not been successfully observed with a multilayer instrument yet.  (2007) The lack of filter imaging for the TR means that imaging has been restricted to spectrometers that have the option of a wide slit (usually referred to as a slot) that allows monochromatic imaging in a small spatial region but at the risk of overlapping images from neighboring emission lines. SOHO/CDS had the option of a 90 × 240 slot that yielded transition region images in O V λ 629.7 although only at 6-10 spatial resolution (e.g., Brković et al. 2000). More recently Hinode/EIS has the options of 40 and 266 slots, although there are no strong transition region lines. Useful data have been obtained from the upper TR lines Mg VI λ 269.0 and Si VII λ 275.4, however (Ugarte-Urra et al. 2009). Since using a narrow or a wide slit is an either/or option and spectrometer science is typically focused on emission line parameters, then TR imaging has been rare. This has had the consequence that transition region phenomena have generally been classified based on their spectroscopic signature. For example, explosive events (Sect. 3.3) are principally distinguished by their line profiles.
The first filtergraph imager for the transition region is the IRIS slitjaw imager (SJI), which has a filter centered at 1390Å that picks up the Si IV lines at 1393.8 and 1402.8Å, as well as the FUV continuum. For intense UV bursts the Si IV emission dominates the signal in the 1400Å images and so the bursts can be confidently assigned to the transition region.

B Emission measure comparison
The large Si IV intensities of UV bursts are their most striking feature, and they imply a large emission measure (EM). Here we compare a typical UV burst EM with values derived from the earlier features that seem most similar to them, namely the SMM/UVSP bursts (Sect. 3.4) and active region blinkers (Sect. 3.5).
The expression relating an observed intensity, I, to a volume emission measure, EM V , for an isothermal plasma of temperature T is where G(T, N e ) is the contribution function for the observed emission line that is computed using the CHI-ANTI gofnt IDL software routine. For the lines considered here, G has a weak dependence on density and we compute the function at a density of 10 12 cm −3 and the temperature at which the function peaks. G contains the element abundance of the emitting ion and we use the "Caffau" solar photospheric abundance set from CHIANTI (see Caffau et al. 2011;Landi et al. 2013). A is the area of the event, and the interpretation varies depending on the spatial resolution of the instrument, as described below. For IRIS we consider event 1 of Peter et al. (2014) -see also panel (b) of Figure 3. For computing the EM V value we sum the intensity over a 3 × 7 block of spatial pixels centered on the peak intensity value, and we sum across the spectral line profile to obtain the intensity listed in Table 3. The intensity profile across the burst in the Y -direction takes a Gaussian shape with a full-width at half-maximum of 0.62 . We thus assume that the event has a spatial size of 0.62 × 0.62 , which we use as A in Eq. 1. Hayes and Shine (1987) give Si IV λ 1402.7 intensities for nine bursts in their Table 1, and we use the median value for Table 3. The UVSP pixels for this observation had a size 4 × 4 and we assume the bursts were fully contained within these pixels, so A =4 × 4 .
For SOHO/CDS we choose the event reported by Young (2004) from the data-set s7616r00 from 1997 April 15, beginning at 21:12 UT. This was a sit-and-stare observation, and the intensity was measured from the exposure with the peak brightness in the O V λ 629.7 line. The burst produced a Gaussian distribution of intensity along the CDS slit, with a FWHM consistent with the PSF of the instrument (Young 2004). Therefore A was set to the CDS pixel size, which in this case was 4.00 × 1.68 .
The Hinode/EIS event of Young et al. (2007) was chosen for deriving an emission measure and, to enable direct comparison with CDS, the intensity of O V λ 192.9 was measured. The value of G was calculated by including the two transitions at 192.904 and 192.911Å. As the intensity distribution along the EIS slit had a width consistent with the instrument point spread function, then we considered the event to be unresolved, so the intensity was summed in the Y direction and A was set to 2 × 1 , the pixel size for this observation.
Comparing the EM values in Table 3 we see that the Si IV values are at least an order of magnitude higher than the CDS O V value. We caution that Si IV belongs to the sodium isoelectronic sequence, and there is a well-known problem that ions from the sodium and lithium-like isoelectronic sequences generally yield EM values higher than other species at similar temperatures (Dupree 1972). However, this is unlikely to account for an order of magnitude difference and it suggests a significantly smaller EM in the upper transition region. The EIS event is a further order of magnitude weaker than the CDS event and this, coupled with the apparent lack of intense active region blinkers seen by EIS, suggests that the high-excitation λ 192.9 line is not as sensitive to the events as the CDS λ 629.7 line.
Based on the EM, it appears that the SMM/UVSP bursts are consistent with the UV bursts. The most intense of the CDS active region blinkers are weaker than the IRIS and UVSP events, but likely still qualify as UV bursts.