The authors of this white paper met on 16-17 January 2020 at the New Jersey Institute of Technology, Newark, NJ, for a 2-day workshop that brought together a group of heliophysicists, data providers, expert modelers, and computer/data scientists. Their objective was to discuss critical developments and prospects of the application of machine and/or deep learning techniques for data analysis, modeling and forecasting in Heliophysics, and to shape a strategy for further developments in the field. The workshop combined a set of plenary sessions featuring invited introductory talks interleaved with a set of open discussion sessions. The outcome of the discussion is encapsulated in this white paper that also features a top-level list of recommendations agreed by participants.

The He I 10830 \AA\ triplet is a very informative indicator of chromospheric activities as the helium is the second most abundant element in the solar atmosphere. Taking advantage of the high resolution of the 1.6 m Goode Solar Telescope (GST) at Big Bear Solar Observatory (BBSO), previous observations have shown clear evidence of the enhanced absorption, instead of typically-observed emission, for two M-class flares. In this study, we analyze the evolution of the He I 10830 10830 \AA\ emission in numerical models and compare it with observations. The models represent the RADYN simulation results obtained from the F-CHROMA database. We consider the models with the injected electron spectra parameters close to observational estimates for the 2013-August-17 flare event ($\delta=8$, $E_c = \{15,20\}$ keV, $F=\{1\times 10^{11}, 3\times{}10^{11}\}$ erg cm$^{-2}$) in detail, as well as other available models. The modeling results agree well with observations, in the sense of both the maximum intensity decrease (-17.1%, compared to the observed value of -13.7%) and the trend of temporal variation (initial absorption phase followed by the emission). All models demonstrate the increased number densities and decreased ratio of the upper and lower level populations of He I 10830 10830 \AA\ transition in the initial phase, which enhances the opacity and forms an absorption feature. Models suggest that the temperatures and free electron densities at heights of 1.3-1.5 Mm should be larger than $\sim 10^4$ K and $6\times 10^{11}$ cm$^{-3}$ thresholds for the line to start being in emission.

We perform an analysis of the line-of-sight (LOS) observables of the SDO/HMI for models of the solar atmosphere heated by precipitating high-energy electrons during solar flares. The radiative hydrodynamic (RADYN) flare models are obtained from the F-CHROMA database. The Stokes profiles for the Fe 6173 A line observed by SDO/HMI are calculated using the radiative transfer code RH1.5D assuming statistical equilibrium for atomic level populations and imposing uniform background vertical magnetic field of various strength. The SDO/HMI observing sequence and LOS data processing pipeline algorithm are applied to derive the observables (continuum intensity, line depth, Doppler velocity, LOS magnetic field). Our results reveal that the strongest deviations of the observables from the actual spectroscopic line parameters are found for the model with a total energy deposited of $E_{total}=1.0\times{}10^{12}$ erg cm$^{-2}$, injected with power-law spectral index of $\delta=3$ above a low-energy cutoff of $E_{c}=25$ keV. The magnitudes of the velocity and magnetic field deviations depend on the imposed magnetic field, and can reach 0.35 km/s for LOS velocities, 90 G for LOS magnetic field, and 3\% for continuum enhancement for the 1000 G imposed LOS magnetic field setup. For $E_{total}\geq{}3.0\times{}10^{11}$ erg cm$^{-2}$ models, the velocity and magnetic field deviations are most strongly correlated with the energy flux carried by $\sim$50 keV electrons, and the continuum enhancement is correlated with the synthesized $\sim$55-60 keV hard X-ray photon flux. The relatively low magnitudes of perturbations of the observables and absence of magnetic field sign reversals suggest that the considered radiative hydrodynamic beam heating models augmented with the uniform vertical magnetic field setups cannot explain strong transient changes found in the SDO/HMI observations.

We present a statistical analysis of properties of Soft X-Ray (SXR) emission, plasma temperature (T), and emission measure (EM), derived from GOES observations of flares in 2002-2017. The temperature and emission measures are obtained using the TEBBS algorithm (Ryan et al. 2012), which delivers reliable results together with uncertainties even for weak B-class flare events. More than 96% of flares demonstrate a sequential appearance of T, SXR, and EM maxima, in agreement with the expected behavior of the chromospheric evaporation process. The relative number of such flares increases with increasing the SXR flux maximum. The SXR maximum is closer in time to the T maximum for B-class flares than for >=C-class flares, while it is very close to the EM maximum for M- and X-class flares. We define flares as "T-controlled" if the time interval between the SXR and T maxima is at least two times shorter than the interval between the EM and SXR maxima, and as "EM-controlled" if the time interval between the EM and SXR maxima is at least two times shorter than the interval between the SXR and T maxima. For any considered flare class range, the T-controlled events compared to EM-controlled events have: a) higher EM but lower T; b) longer durations and shorter relative growth times; c) longer FWHM and characteristic decay times. Interpretation of these statistical results based on analysis of a single loop dynamics suggests that for flares of the same class range, the T-controlled events can be developed in longer loops than the EM-controlled events.

We present a statistical study of chromospheric evaporation in solar flares using simultaneous observations by the RHESSI X-ray telescope and the IRIS UV spectrograph. The results are compared with radiation hydrodynamic flare models from the F-CHROMA RADYN database. For each event, we study spatially-resolved Doppler shifts of spectral lines formed in the transition region (C\,II\,1334.5\,{\AA}) and hot coronal plasma (Fe\,XXI\,1354.1\,{\AA}) to investigate the dynamics of the solar atmosphere during the flare impulsive phase. We estimate the energy fluxes deposited by high-energy electrons using X-ray imaging spectroscopy and assuming the standard thick-target model. Using the RADYN flare models, the RH 1.5D radiative transfer code and the Chianti atomic line database, we calculate C\,II and Fe\,XXI line profiles and compare with the observations. While the RADYN models predict a correlation between the Doppler shifts and deposited energy flux for both lines, this was only observed in the C II data. Several quantitative discrepancies are found between the observations and models: the Fe\,XXI Doppler shifts are substantially stronger in the models than in the data; the C\,II mean blueshifts are absent in the observations but predicted by the models. The transition energy between `gentle' and `explosive' evaporation regimes estimated from the observations ($2-8\times{}10^{9}$~erg~cm$^{-2}$~s$^{-1}$) and derived from the models ($2.2-10.1\times{}10^{9}$~erg~cm$^{-2}$~s$^{-1}$) are comparable with each other. The results illustrate relationships among the processes of chromospheric evaporation, response of the colder layers, and the flare energy flux deposited by high-energy electrons, although demonstrating discrepancy between analyzed observations and RADYN models.

We analyze the relationship between the flare X-ray peak flux, and characteristics of the Polarity Inversion Line (PIL) and Active Regions (AR), derived from line-of-sight (LOS) magnetograms. The PIL detection algorithm based on a magnetogram segmentation procedure is applied for each AR with 1 hour cadence. The PIL and AR characteristics are associated with the AR flare history and divided into flaring and non-flaring cases. Effectiveness of the derived characteristics for flare forecasting is determined by the number of non-flaring cases separated from flaring cases by a certain threshold, and by their Fisher ranking score. The Support Vector Machine (SVM) classifier trained only on the PIL characteristics is used for the flare prediction. We have obtained the following results: (1) the PIL characteristics are more effective than global characteristics of ARs, (2) the highest True Skill Statistics (TSS) values of 0.76$\pm$0.03 for $\geq$M1.0 flares and 0.84$\pm$0.07 for $\geq$X1.0 flares are obtained using the "Sigmoid" SVM kernel, (3) the TSS scores obtained using only the LOS magnetograms are slightly lower than the scores obtained using vector magnetograms, but significantly better than current expert-based predictions, (4) for prediction of $\geq$M1.0 class flares 74.4 % of all cases, and 91.2 % for $\geq$X1.0 class, can be pre-classified as negative with no significant effect on the results, (5) the inclusion of global AR characteristics does not improve the forecast. The study confirms the unique role of the PIL region characteristics in the flare initiation process, and demonstrate possibilities of flare forecasting using only the line-of-sight magnetograms.

We study flare processes in the solar atmosphere using observational data for a M1-class flare of June 12, 2014, obtained by New Solar Telescope (NST/BBSO) and Helioseismic Magnetic Imager (HMI/SDO). The main goal is to understand triggers and manifestations of the flare energy release in the photosphere and chromosphere using high-resolution optical observations and magnetic field measurements. We analyze optical images, HMI Dopplergrams and vector magnetograms, and use Non-Linear Force-Free Field (NLFFF) extrapolations for reconstruction of the magnetic topology and electric currents. The NLFFF modelling reveals interaction of two magnetic flux ropes with oppositely directed magnetic field in the PIL. These flux ropes are observed as a compact sheared arcade along the PIL in the high-resolution broad-band continuum images from NST. In the vicinity of PIL, the NST H alpha observations reveal formation of a thin three-ribbon structure corresponding to a small-scale photospheric magnetic arcade. The observational results evidence in favor of location of the primary energy release site in the chromospheric plasma with strong electric currents concentrated near the polarity inversion line. In this case, magnetic reconnection is triggered by the interacting magnetic flux ropes forming a current sheet elongated along the PIL.

Solar flares are complicated physical phenomena that are observable in a broad range of the electromagnetic spectrum, from radiowaves to $\gamma$-rays. For a more comprehensive understanding of flares, it is necessary to perform a combined multi-wavelength analysis using observations from many satellites and ground-based observatories. For efficient data search, integration of different flare lists and representation of observational data, we have developed an Interactive Multi-Instrument Database of Solar Flares (https://solarflare.njit.edu/). The web accessible database is fully functional and allows the user to search for uniquely-identified flare events based on their physical descriptors and availability of observations by a particular set of instruments. Currently, the data from three primary flare lists (GOES, RHESSI and HEK) and a variety of other event catalogs (Hinode, Fermi GBM, Konus-Wind, OVSA flare catalogs, CACTus CME catalog, Filament eruption catalog) and observing logs (IRIS and Nobeyama coverage) are integrated, and an additional set of physical descriptors (temperature and emission measure) is provided along with an observing summary, data links, and multi-wavelength light curves for each flare event since January, 2002. We envision that this new tool will allow researchers to significantly speed up the search of events of interest for statistical and case studies.

The X-class solar flare of October 23, 2012, generated continuum photospheric emission and a strong helioseismic wave ("sunquake") that points to an intensive energy release in the dense part of the solar atmosphere. We study properties of the energy release with high temporal and spatial resolutions, using photospheric data from the Helioseismic Magnetic Imager (HMI) onboard Solar Dynamics Observatory (SDO), and hard X-ray observations made by the Ramaty High-Energy Solar Spectroscopic Imager (RHESSI). For this analysis we use level-1 HMI data (filtergrams), obtained by scanning the Fe I line (6731~\AA) with the time cadence of $\sim 3.6$ s and spatial resolution of $\sim 0.5^{\prime\prime}$ per pixel. It is found that the photospheric disturbances caused by the flare spatially coincide with the region of hard X-ray emission, but are delayed by $\lesssim 4$ seconds. This delay is consistent with predictions of the flare hydrodynamics RADYN models. However, the models fail to explain the magnitude of variations observed by the HMI. The data indicate that the photospheric impact and helioseismic wave might be caused by the electron energy flux substantially higher than that in the current flare radiative hydrodynamic models.

We study flare processes in the lower solar atmosphere using observational data for a M1-class flare of June 12, 2014, obtained by New Solar Telescope (NST/BBSO) and Helioseismic Magnetic Imager (HMI/SDO). The main goal is to understand triggers and manifestations of the flare energy release in the lower layers of the solar atmosphere (the photosphere and chromosphere) using high-resolution optical observations and magnetic field measurements. We analyze optical images, HMI Dopplergrams and vector magnetograms, and use Non-Linear Force-Free Field (NLFFF) extrapolations for reconstruction of the magnetic topology. The NLFFF modelling reveals interaction of oppositely directed magnetic flux-tubes in the PIL. These two interacting magnetic flux tubes are observed as a compact sheared arcade along the PIL in the high-resolution broad-band continuum images from NST. In the vicinity of the PIL, the NST H alpha observations reveal formation of a thin three-ribbon structure corresponding to the small-scale photospheric magnetic arcade. Presented observational results evidence in favor of location of the primary energy release site in the dense chromosphere where plasma is partially ionized in the region of strong electric currents concentrated near the polarity inversion line. Magnetic reconnection may be triggered by two interacting magnetic flux tubes with forming current sheet elongated along the PIL.

Chromospheric evaporation is observed as Doppler blueshift during solar flares. It plays one of key roles in dynamics and energetics of solar flares, however, its mechanism is still unknown. In this paper we present a detailed analysis of spatially-resolved multi-wavelength observations of chromospheric evaporation during an M 1.0 class solar flare (SOL2014-06-12T21:12) using data from the NASA's IRIS (Interface Region Imaging Spectrograph) and HMI/SDO (Helioseismic and Magnetic Imager onboard Solar Dynamics Observatory) telescopes, and VIS/NST (Visible Imaging Spectrometer at New Solar Telescope) high-resolution observations, covering the temperature range from 10^4 K to 10^7 K. The results show that the averaged over the region Fe XXI blueshift of the hot evaporating plasma is delayed relative to the C II redshift of the relatively cold chromospheric plasma by about 1 min. The spatial distribution of the delays is not uniform across the region and can be as long as 2 min in several zones. Using vector magnetograms from HMI we reconstruct the magnetic field topology and the quasi-separatrix layer (QSL) and find that the blueshift delay regions as well as the H-alpha flare ribbons are connected to the region of magnetic polarity inversion line (PIL) and an expanding flux rope via a system of low-lying loop arcades with height < ~4.5 Mm. This allows us to propose an interpretation of the chromospheric evaporation based on the geometry of local magnetic fields, and the primary energy source associated with the PIL.

Dynamics of hot chromospheric plasma of solar flares is a key to understanding of mechanisms of flare energy release and particle acceleration. A moderate M1.0 class flare of 12 June, 2014 (SOL2014-06-12T21:12) was simultaneously observed by NASA's Interface Region Imaging Spectrograph (IRIS), other spacecraft, and also by New Solar Telescope (NST) at the BBSO. This paper presents the first part of our investigation focused on analysis of the IRIS data. Our analysis of the IRIS data in different spectral lines reveals strong redshifted jet-like flow with the speed of ~100 km/s of the chromospheric material before the flare. Strong nonthermal emission of the C II k 1334.5 A line, formed in the chromosphere-corona transition region, is observed at the beginning of the impulsive phase in several small (with a size of ~1 arcsec) points. It is also found that the C II k line is redshifted across the flaring region before, during and after the impulsive phase. A peak of integrated emission of the hot (1.1 MK) plasma in the Fe XXI 1354.1 A line is detected approximately 5 minutes after the integrated emission peak of the lower temperature C II k. A strong blueshift of the Fe XXI line across the flaring region corresponds to evaporation flows of the hot chromospheric plasma with a speed of 50 km/s. Additional analysis of the Reuven Ramaty High-Energy Solar Spectroscopic Imager (RHESSI) data supports the idea that the upper chromospheric dynamics observed by IRIS has features of "gentle" evaporation driven by heating of the solar chromosphere by accelerated electrons and by a heat flux from the flare energy release site.

We present results of non-LTE modeling of H-alpha 6563 A and Mg II k&h 2796 A and 2803 A lines. This modeling is important for interpretation of coordinated observations from the recently launched NASA's IRIS mission and from the New Solar Telescope at Big Bear Solar Observatory. Among available codes for the non-LTE modeling, the RH code is chosen as the most appropriate for modeling of the line profiles. The most suitable Hydrogen and Magnesium atomic models are selected by performing several tests of the code. The influence of the ionization degree on the line profiles is also studied. Radiative-MHD simulations of the solar atmosphere, obtained with the Bifrost code, are used as input data for calculation of synthetic spectra of the H-alpha and Mg II h&k lines for particular locations evolving with time. The spectral line variations reveal the presence of MHD waves in the simulation results. We construct oscillation power spectra of the line intensity for different wavelength, and compare these with the corresponding height-dependent power spectra of atmospheric parameters from the simulations. We find correlations between the power spectra of intensities of the line profiles at certain wavelengths and the power spectra of the atmospheric parameters at the tau-unity heights for these wavelengths. These results provide a new diagnostic method of chromospheric oscillations; however, larger amounts of data are needed to confirm these correlations.