Joint X-ray, EUV and UV Observations of a Small Microflare

We present the first joint observation of a small microflare in X-rays with the Nuclear Spectroscopic Telescope ARray (NuSTAR), UV with the Interface Region Imaging Spectrograph (IRIS) and EUV with the Solar Dynamics Observatory/Atmospheric Imaging Assembly (SDO/AIA). These combined observations allows us to study the microflare's hot coronal and cooler chromospheric/transition region emission. This small microflare peaks from SOL2016-07-26T23:35 to 23:36UT, in both NuSTAR, SDO/AIA and IRIS. Spatially this corresponds to a small loop visible in the SDO/AIA Fe XVIII emission, which matches a similar structure lower in the solar atmosphere seen by IRIS in SJI1330{\AA} and 1400\AA. The NuSTAR emission in both 2.5-4 keV and 4-6 keV, is located in a small source at this loop location. The IRIS slit was over the microflaring loop, and fits show little change in Mg II but do show intensity increases, slight width enhancements and redshifts in Si IV andO IV, indicating that this microflare had most significance in and above the upper chromosphere. The NuSTAR microflare spectrum is well fitted by a thermal component of 5.8MK and $2.3\times10^{44}$ cm$^{-3}$, which corresponds to a thermal energy of $10^{26}$ erg, making it considerably smaller than previously studied X-ray microflares. No non-thermal emission was detected but this could be due to the limited effective exposure time of the observation. This observation shows that even ordinary features seen in UV and EUV, can remarkably have a higher energy component that is clear in X-rays.


INTRODUCTION
Microflares are small releases of stored magnetic energy in the solar atmosphere that heat material and accelerate particles. Energetically, they are down to a million times smaller than the largest events, yet still demonstrate similar properties (Hannah et al. 2011). The smaller microflares range down to GOES A-Class events, with 1-8Å flux < 10 −7 Wm −2 , and are considerably more frequent than the largest flares. The frequency distribution of flares is a negative power-law with index α ∼ 2 (Hudson 1991). However, it is still not clear down to what energy scales this rate persists, a crucial fact to determine the overall contribution of micro-, or even smaller, nanoflares, to heating the solar corona.
X-ray observations of microflares provide clear diagnostics of the energetics of the heated material and accelerated electrons. Above a few keV this is predominantly bremsstrahlung continuum emission and RHESSI  showed that even the smallest microflares exhibit non-thermal footpoints, at the ends of coronal loops containing material > 10MK (Krucker et al. 2002;Hannah et al. 2008b). A large statistical study of RHESSI microflares (Christe et al. 2008;Hannah et al. 2008a) showed that these events are exclusively in active regions, last for a few minutes, are not necessarily spatially small, have emission > 10MK, and over the initial impulsive period have median thermal energy 10 28 erg and non-thermal 10 27 erg.
Going beyond RHESSI, the Nuclear Spectroscopic Telescope ARray NuSTAR (Harrison et al. 2013), is an astrophysics telescope with two direct focusing optics modules, and a higher effective area than RHESSI. It has observed solar emission several times since the first pointing in late 2014 , and has observed active region microflares as well as quiet Sun brightenings. These microflares are about an order of magnitude smaller than RHESSI could observe, down to an estimated GOES emission of ∼A0.1, showed heating up to about 10MK and thermal energy of 10 27 erg (Wright et al. 2017;Glesener et al. 2017). These events were also well observed at longer wavelengths, in terms of softer X-rays with Hinode/XRT and the Solar Dynamics Observatory/Atmospheric Imaging Assembly (SDO/AIA ). X-ray brightenings were also observed with NuSTAR outside of active regions in the quiet Sun, with temperatures of about 3-4MK, estimated GOES emission of A0.01 and thermal energy of about 10 26 erg (Kuhar et al. 2018).
These microflares and brightenings did not show any non-thermal emission however this is likely due to an observational constraint of using NuSTAR to observe the Sun. NuSTAR has a detector throughput of only 400 counts s −1 telescope −1 , which even quiet solar emission could swamp . This results in a detector livetime fraction considerably less than unity, and a greatly reduced effective exposure time. Given the steep nature of X-ray spectrum, these short effective exposure times limits the spectral dynamic range, producing few, or no, counts at higher energies, the range in which non-thermal emission is expected (Hannah et al. 2016). The effective exposures of these NuSTAR observations were short because there were also other bright sources on the solar disk. Unfortunately even regions outside the NuS-TAR field of view of 12' × 12', can be detected and exacerbate the throughput issue . Even with higher livetime NuSTAR observations there is still a limit to the sensitivity arising from the inherent short duration of these small flares.
Non-thermal emission is expected from small microflares as particle acceleration often features during magnetic reconnection, the energy release mechanism that is thought to be behind flares of all sizes. Even the smallest RHESSI microflares could show considerable non-thermal emission from accelerated electrons (Hannah et al. 2008b). Small scale events accelerating electrons is thought to be behind radio noise storms (Mercier & Trottet 1997;Shibasaki et al. 2011) however it is considerably more difficult to obtain the electron energetics from the radio data, compared to X-rays, due to the non-linear nature of the emission mechanism processes. The presence of accelerated electrons in small events has also been inferred from UV observations with IRIS (De Pontieu et al. 2014). Rapid brightenings (over 10s of seconds) were observed at the footpoints of hot coronal loops (Testa et al. 2014). The observed blue-shifts (upflows) of the Si IV 1403Å line in these "moss" brightenings (Berger et al. 1999) are consistent with RADYN numerical simulations of chromosphere/transition region heating by a beam of accelerated electrons (a power-law of non-thermal energy 6×10 24 erg, with spectral index δ = 7, above a cutoff of E C = 10keV). Thermal conduction and Alfven waves could not reproduce the line shift, nor the intense brightening (Testa et al. 2014). This combination of RADYN simulations and UV observations were further shown to provide constraints to the properties of the non-thermal electrons (Polito et al. 2018). They found that the blue-shifts were dependent on both the nonthermal energy and the low energy cutoff (E C ≥ 5keV for 10 24 erg, and E C ≥ 15keV for 10 25 erg) otherwise red-shifts were produced. This work showed that Mg II could also be used to help constrain the electron beam properties.
In this paper, we present a small microflare that was observed on SOL2016-07-26T23:35 in X-rays with NuSTAR , UV with IRIS and EUV with SDO/AIA , allowing us to study the heating of both the chromosphere/transition region and corona. In §2, we give an overview of the event, before going into detail about the spatial and temporal behaviour of the microflare in §3. Then in §4 we derive properties about the emission from both the IRIS and NuSTAR spectra. The thermal properties found from the NuSTAR spectra are compared to the emission observed by SDO/AIA and GOES/XRS  The region that produced the microflare was observed by NuSTAR on 2016-Jul-26 between 23:27 and 23:37UT, towards the south-eastern limb. The region was never given an NOAA ID, but was identified as SPoCA 19717. This particular NuSTAR solar pointing 1 had spent 3 hours focused on brighter active regions on the opposite western limb as they rotated off the visible disk, occulting the brighter emission from the lower solar atmosphere. By the time NuSTAR targeted the microflaring region for 10 minutes, it was the brightest X-ray region on the disk, with the other regions being well-occulted, determined from GOES/SXI images. Therefore, what is seen in the GOES/XRS full-disk emission should be dominated by the target region. Both the 1-8Å and 0.5-4Å channels of GOES/XRS show a small microflare between 23:35 to 23:36UT, see the top panel of Figure 1. This event peaks at GOES A8-level, but is only about an A1 excess above the pre-flare emission. The full time profile and spatial behaviour of the microflare in X-ray, EUV, and UV are shown in Figures 1, 2 3 and will be discussed in the next section, §3.

SPATIAL & TEMPORAL ANALYSIS
The SDO/AIA images of the region were processed to level 1.6 data, using the standard software to prep, as well as deconvolving the point spread function. Most of the EUV channels showed only weak emission from the region of interest, which barely changed over the 10 minutes. The 94Å channel did show a small loop and this brightens at the same time as the X-ray emission. We removed the cooler component of the 94Å channel via the approach of Del Zanna (2013), leaving just the emission above 3 MK from Fe XVIII . Overview images of the whole region in SDO/AIA 94Å, 171Å and 211Å, as well as Fe XVIII are shown in Figure 2 for the microflare time, 23:35 to 23:36UT. The region has extensive cooler emission, but the hotter emission, as indicated by Fe XVIII, is more compact with the small loop that brightens during the microflare time. A zoomed-in view of this emission, shown in the left-hand panels of Figure 3, gives both the pre-flare (23:28 to 23:29UT) and microflare (23:35 to 23:36UT) times, where the loop and the brightening becomes clearer. In both Figures 2 and 3, the SDO/AIA images shown here have been summed over 1 minute to improve the signal-to-noise. The full 12s cadence images were used to determine the time profile of the EUV emission from just the brightening loop region, the resulting lightcurves (shown in the second top panel of Figure 1) clearly show a peak of emission, and just slightly after the X-ray microflare seen with GOES/XRS. The   304Å channel had some features that brightened over the observation time but were not spatially or temporally correlated with the 94Å nor Fe XVIII emission. IRIS co-observed the region from 2016-Jul-26 21:53:26 to 2016-Jul-27 02:47:17UT with 17 large sparse 64-step rasters (OBSID 3600110059) with steps of 1 , exposure times of 15s, and a factor 2 for spatial and spectral summing. All SJI filters (1330Å, 1400Å, 2796Å, 2832Å) were used, giving a cadence of 65 s for each SJI filter. For this paper, we analyze the 6th raster, taken from 23:19:57 to 23:36:59UT. We verified the remaining orbital variation to be below 0.2 km s −1 in NUV during this raster, which is below our desired accuracy, and therefore we use the original raster with the newest calibrations (L12-2017-04-23) for the analysis without additional corrections. We align the SJI1400Å data to the SDO/AIA 1600Å data, which includes a 0.6 • roll and a <2 shift. There are multiple little bright loops in the IRIS SJI1400Å and SJI1330Å images but there is one at exactly the same location, and of the same size and orientation of the microflaring loop seen in SDO/AIA Fe XVIII. This is shown in the middle panels of Figure 3. Considering just the emission from this small UV loop, the resulting lightcurve (shown in Figure 1), also brightens but slightly before the time of the X-ray and EUV microflare. Crucially the IRIS slit moves across the loop during the time of the microflare, from 23:35 to 23:36UT, and this spectral analysis is detailed in §4. The NuSTAR emission from the region is taken only from the FPMA telescope, as the FPMB telescope has the detector gap directly through the microflare. The NuSTAR data was filtered 2 to remove bad pixels, and non-grade 0 events to minimise detector pileup . These observations were all made under a single Camera Head Unit (CHU3), so we can apply just one correction to the pointing. The NuSTAR pointing was aligned to the whole region seen in SDO/AIA Fe XVIII, not just the microflaring loop. The resulting lightcurve for the NuSTAR emission in 2.5-4 keV and 4-6 keV were found over this larger region (shown in Figure 2) and both energy ranges brighten during the microflare time. It also peaks slightly earlier in the higher energy channel, consistent with flare heating followed by cooling. The NuSTAR images for both energy channels, and the pre-flare and microflare times, are shown as maps in Figure 2 and zoomed in contours in Figure 3. These images have been processed to deconvolve the point spread function and show a structure with peak brightness close to the SDO/AIA Fe XVIII source. At both times the NuS-TAR images in each energy range overlap, although the location of peak brightness are slightly mis-aligned. However this is not significant as it is within the spatial resolution of NuSTAR .

SPECTRAL ANALYSIS
Hinode/EIS was also targeting this region, but at the time of the NuSTAR observations the slit was further to the East of the microflare, missing the hot material NuSTAR was detecting. The IRIS slit was however over the region during the time  of the NuSTAR observation, and moved across the microflaring loop during the 1 minute it brightened in X-rays and EUV, but faded in UV.

IRIS spectra
To obtain the features of the spectral lines we use the iris get mg features lev2 routine for Mg, which derives the position and intensities of the blue, red and central peaks. We mainly focus on Mg k at 2796.35Å, because the h line shows an identical behavior. For the FUV lines, we perform Gaussian fitting to the Si IV 1393.76Å, Si IV 1402.77Å, and O IV 1401.16Å lines, and obtain the Doppler shifts, Doppler widths, and line intensities. Examples of such fits are shown in Figure 4 for a pixel in the microflare (top row) and a quiet Sun pixel (bottom row). The other O IV lines (1399.78Å, 1404.78Å) are below the noise limit in most pixels and therefore cannot be used as a density diagnostic. The coronal Fe XXI line, which usually only appears in flares, is below the detection limit in this microflare.
Maps of the fit results are shown in Figure 5. The box formed by the white lines indicates the location of the microflare. The Mg II line core Doppler velocity around the microflare shows a weak redshift of less than 10 km s −1 , which occurs everywhere in the field of view. The Mg peak ratio is defined in Eq. 2 of Leenaarts et al. (2013) and it correlates with the average velocity in the upper chromosphere. Black areas indicate fitting issues, i.e. locations where the Mg line profiles do not show their typical shape, but rather a single peak. The peak ratio around the microflare is zero indicating that it does not influence the apparent upper chromospheric dynamics. Similarly, it is invisible in the Mg intensity. Si IV 1394Å and 1403Å are very similar, therefore only Si IV 1394Å is shown in the plots, as it has a higher absolute intensity. Note that the ratio of the total intensity in these two Si IV lines for the microflare loop are approximately 2, the expected value for optically thin emission (Kerr et al. 2018). In Si IV, the microflare is clearly visible in intensity. The Doppler width of Si IV is slightly enhanced (0.2Å), but such enhancements also occur in other parts of the FOV and can therefore not be attributed solely to the microflare. The velocities of Si IV are generally higher than those of Mg II, as can be expected, because Si IV forms at higher temperatures. At the location of the microflare downflows of the order of 20 km s −1 are prevalent, which are commonly found in the quiet Sun. The microflare is also visible in the O IV intensity. The O IV 1401Å line is often weak, which explains the lack of fits (black locations) in its FWHM plot. Similarly to Si IV, the O IV FWHM and velocities are enhanced, but it is unclear if this is related to the microflare because similar enhancements are seen throughout the quiet Sun. The fact that the small loop is visible in the O IV and Si IV lines suggests material that is heated to logT = 4.8 and logT = 5.2 (or 0.06 and 0.16MK) respectively. Because the loop is invisible in the Mg II line, it means that there is little material at logT = 4.0 in the loop. It seems that plasma below the upper chromosphere is not significantly affected by this microflare.

NuSTAR spectrum
The fitted NuSTAR spectrum for the microflare time is shown in Figure 6. Here we show the spectrum from NuSTAR FPMA over the region shown in Figure 2 during 23:35 to 23:36UT. The data was rebinned before fitting so that there were at least 10 counts in each bin, and that there were no zero count bins in the fit range. Bad pixels and non-zero grade events where filtered out of the eventlist used to make the spectrum. The spectrum was fitted in XSPEC using an APEC thermal model, with coronal abundances manually set using the values from Feldman et al. (1992), not using the default solar ones (which are photospheric and not coronal). The minimum fit energy used was 2.7keV, as below this energy there is a discrepancy in the instrumental response arising from uncertainty in the detection threshold (Grefenstette et al. 2018). The best fit parameters were found using the Cash statistic (Cash 1979). Given the bright emission from the rest of the region a pre-flare component was used to represent the background. This fixed thermal model of 3.2MK and 6.9 × 10 46 cm −3 was found by fitting the NuSTAR spectrum over 23:28 to 23:29UT, which was well fitted by this single thermal component. To fit the microflare excess above the pre-flare emission we added a second thermal component and found a good fit to the data with an additional component of 5.8MK and 2.3 × 10 44 cm −3 . The 1σ uncertainty range for these parameters are 5.1MK and 5.0 × 10 44 cm −3 to 6.7MK and 1.0 × 10 44 cm −3 . There are no significant counts above 5keV in this event, from hotter or non-thermal emission, but this observation did have a short effective exposure (about 3.1s) and was only seen in one of the two telescopes, limiting the spectral dynamic range. Using the observed SDO/AIA Fe XVIII loop, of about 8 pixels long by 4 pixels wide, we get a volume estimate of 8.3 × 10 24 cm 3 , assuming a filling factor of unity. Combined with the NuSTAR emission measure, this gives a density of 5.3 × 10 9 cm −3 , with uncertainty range of 3.4 × 10 9 cm −3 to 7.8 × 10 9 cm −3 . From this we can calculate the instantaneous thermal energy (Hannah et al. 2008a) of the microflare over the minute it is seen above the pre-flare emission, finding 1.0 × 10 26 erg, with an uncertainty range of 7.9 × 10 25 erg to 1.4 × 10 26 erg. This means that this microflare is about an order of magnitude smaller in energy than those previously seen with NuSTAR (Wright et al. 2017;Glesener et al. 2017). The density and thermal energy of whole region during the pre-flare time can also be estimated by assuming the volume of the region is related to the observed SDO/AIA Fe XVIII area as V = A 3/2 , giving a density of 1.1 × 10 9 cm −3 and 1.5 × 10 29 erg. So the microflaring loop contains  only about 0.07% the thermal energy of the whole region and is not contributing substantially to the overall heating of the region.

Comparison of NuSTAR and SDO/AIA
Using the NuSTAR count rate over 2.7-4 and 4-5 keV, and by calculating the thermal response 3 of NuSTAR in these energy channels for this observation, we can produce the EM loci curves. These are useful for confirming the temperature and emission measure found from the spectral fitting, as well as checking whether the NuSTAR and SDO/AIA are coming from the same thermal source. We choose these two energy bands so they match the energy range fitted in the NuSTAR spectrum shown in Figure 6. The resulting NuSTAR EM loci curves are shown in Figure 7 for both the pre-flare and the microflare excess (subtracting the emission during the pre-flare from the microflare) times. For each time interval, where the curves of the different energy channels cross gives the isothermal thermal values. The EM loci values give slightly higher temperatures, but with lower emission measures compared to the spectral fits. This consistency between the EM loci and spectral fitted values is despite the APEC thermal model being using for the fitting, and CHIANTI atomic database for the EM loci curves. Also shown in Figure 7, is the EM loci curves for the loop SDO/AIA Fe XVIII emission. For the pre-flare time there is a mismatch between the SDO/AIA and NuSTAR curves but that is likely due to the NuSTAR observed emission being at the edge of Fe XVIII temperature response range. Also the calculation of the Fe XVIII emission is an empirical approach and does not perform well when the emission is weak, such as we have in this region. During the microflare the emission is brighter due to the presence of hotter material, and the Fe XVIII loci curve overlaps where the two NuSTAR curves also overlap, showing that both NuSTAR and the Fe XVIII observed emission is coming from the same loop material about 6MK.
For a clearer comparison of the observed NuSTAR and SDO/AIA emission, we take the temperature and emission found from fitting the NuSTAR spectrum and fold this through the SDO/AIA temperature response for each different channel. We then compare the observed SDO/AIA emission in each channel to the one derived from the NuSTAR spectral fit, which we call the NuSTAR synthetic emission. The resulting plot for the emission during the pre-flare and microflare times are shown in Figure 8. As expected, the hotter emission observed by NuSTAR is only contributing a tiny fraction to the observed emission in most of the SDO/AIA channels. The only channels in which the majority of the observed emission is coming from the temperatures NuSTAR observed are, as expected, 94Å and Fe XVIII. This helps confirm why the microflaring loop is only clearly visible in those SDO/AIA channels, as there appears to be no significant change in the amount of material at cooler temperatures.

Comparison of NuSTAR and GOES/XRS
Using the thermal parameters found from fitting the NuSTAR spectra we can estimate the GOES/XRS flux that should have been produced. For the emission from the whole region during the pre-flare time we estimate the GOES/XRS flux using the standard routine 4 as 3.9 × 10 −9 Wm −2 . The observed GOES/XRS from the full-disk over this time was actually 6.6 × 10 −8 Wm −2 , about a factor of 16 higher. Similarly, using the NuSTAR temperature and emission found for the microflare excess we obtain a GOES/XRS flux of 1.0 × 10 −10 Wm −2 , equivalent to 0.01A-class. The observed flux was 1.0×10 −8 Wm −2 , about 100 times higher. It could be that there was emission coming from elsewhere on the disk, however close examination of both GOES/SXI and SDO/AIA Fe XVIII full disk images show that the NuSTAR region was the main and brightest one on the disk and certainly cannot explain such large discrepancies. Although there is a substantial difference between the calculated and observed fluxes it should be noted that GOES/XRS is poorly calibrated in this flux range, as it is designed to monitor in an operational mode the flux from large flares. This is highlighted in the recent comparison of GOES/XRS emission with the softer X-ray spectrometer MinXSS-1 (Mason et al. 2016). The MinXSS-1 spectrum gives a more robust irradiance measure compared to the broader channel used by GOES/XRS and showed deviations below fluxes of 10 −6 Wm −2 , which became even more substantial once below 10 −7 Wm −2 (Woods et al. 2017). MinXSS-1 was operational when these NuSTAR observations were made but unfortunately nothing was discernible above the noise levels, which may have been due to it operating in a "non-fine pointing" mode during this time range (Moore, private communication).

DISCUSSIONS & CONCLUSIONS
In this paper, we presented the smallest microflare seen yet with NuSTAR , about an order of magnitude smaller than those previously observed with NuSTAR (Wright et al. 2017;Glesener et al. 2017) and well beyond the microflares observed with RHESSI (Hannah et al. 2008a). In this microflare we saw emission at about 6MK, which gave an instantaneous thermal energy of around 10 26 erg. It is remarkable that even in this small X-ray microflare we were still able to see corresponding emission in EUV and UV, allowing us to study both the coronal and upper chromospheric/transition region response. The small loop seen with IRIS in UV and SDO/AIA in EUV by itself was unexciting, but this changes with the unexpected addition of emission seen at higher energies with NuSTAR . In this microflare no higher temperature (closer to 10MK) or non-thermal emission was observed but that is could be due to limited effective area from only one of the two telescopes observing the flare and as well as the short exposure time. Only about 3s was achieved over an on-time of 60s, due to emission elsewhere on the solar disk. Subsequent NuSTAR observations of microflaring active regions have successfully caught the emission with both telescopes and been observed during times with less activity across the Sun, corresponding to higher livetimes. So with these event currently under study, the presence of higher temperature and non-thermal emission is more likely to be detected. Observations of small flares have the inherent problem that these are short duration events, so long exposures are not possible and require instruments with higher sensitivity from larger detector effective area.
It is surprising that this microflare is only seen at the hotter coronal temperatures and lower chromospheric/transition region ones, but not in the few MK range that would have produced emission in the majority of the SDO/AIA EUV channels. It could have been that there was more background material in this temperature range so the small increase due to the microflare was hidden, rendering it effectively invisible. Or it may have been that hotter material seen by NuSTAR and SDO/AIA Fe XVIII cooled too rapidly to be seen, or that the ionisation timescale was longer than the cooling timescale. This event did not present the moss brightenings reported in previous IRIS small flare work (Testa et al. 2014), so it could be that this event is even smaller, with faster rastering required to catch velocities clearly associated with the microflare, or possibly a different type of event.
Although the microflare is seen as a brightening in GOES/XRS, it is difficult to trust the observed flux measure given that this is at the limit of the instruments sensitivity and the substantial uncertainties in the calibration (in terms of the spectral distribution of these small events relative to the instruments response function). But again it should be noted that GOES/XRS was not designed to be useful for these small fluxes. Future observations with NuSTAR that overlap with other softer X-ray spectrometers, such as MinXSS-2 (Moore et al. 2018) or MaGIXS (Kobayashi et al. 2018), might help to resolve the true emission of these small microflares over this energy range.
The NuSTAR observations of this small microflare have shown that even fairly ordinary features seen in UV and EUV can remarkably have a higher energy Xray component. This shows that there is substantial potential for studying weaker solar activity at higher energy X-rays, either occasionally with NuSTAR or with an optimised solar spacecraft such as the proposed FOXSI (Christe et al. 2017).