Non-Periodic Phenomena in Variable Stars IAU Colloquium, Budapest, 1968 FLARES OF UV CETI TYPE STARS Introductory Paper by R. E. GERSHBERG Crimean Astrophysical Observatory, USSR The term "flare stars" is used sometimes as a synonym to "eruptive stars" and in that case the term "flare" covers a wide range of phenomena of stellar variability. I intend to give a review of observational and theoretical results bearing on the classical flare stars of UV Ceti type only and I shall use the words "flare" and "flare stars" only in that limited sense. Owing to the restricted time, I have no possibility to give the detailed history of the investigations of the UV Ceti type stars. This history can be found in Joy's (1960), Oskanjan's (1964) and Haro's (1968) reviews - therefore I shall submit the state of the problem only for the present moment. That is why I shall not refer to a number of investigations which were important for their times but were surpassed by following studies. The dMe-objects with quick flares of brightness are attributed to classical flare stars of UV Ceti type. Today no spectral or photometric criteria are known which would permit to establish the relation of a dMe to the UV Ceti type by observing it in a quiet state. About 25 UV Ceti type stars are at present known, and they make nearly a quarter of the known dMe stars and about 5 per cent of all the dM objects; because M dwarfs represent to about 80% per cent of galactic stellar population, one may suppose that the flares of UV Ceti-type stars are the most wide-spread kind of stellar variability. Among the 25 UV Ceti type stars 19 are known as binaries. 3 of them are spectroscopic binaries, 2 of them have distances less than 1" between the components; in the other 14 visual binaries the fainter components are flare stars. The masses of flare stars are small: the mass of UV Cet itself is equal to 0.04M sun, that is less than a minimum mass of a main sequence star; the mass of EQ Peg is equal to 0.13M sun, and that of DO Cep to 0.16M sun (Petit, 1961). The diameters of flare stars are about 3 times less than that of the Sun (Lippincott, 1953). The luminosities of these objects are low, and the absolutely faintest star, van Biesbroeck's object, BD +4 4048 B, M_v = 18.6m, is a flare star. But we are not certain that there exist systematic differences in masses, sizes, luminosities and percentage of binaries between flare and normal M-dwarfs. The dispersion of the peculiar velocities of dMe and UV Ceti type stars is 2-2.5 times less than that of normal dM stars (Gliese, 1958). After giving this short stellar statistical characterization of UV Ceti type stars, we may pass to discuss the flares themselves. OBSERVATIONS In accordance with the topic of our Colloquium it is necessary to begin with the time features of flares. Time distribution of flares For nearly 20 years there had been a belief that the flares of UV Ceti type stars occurred irregularly. But Andrews (1968) found some recurrence in the time distribution of 9 flares of YZ CMi: 2 intervals between flares were near to 122h, 3 near to 73h and 3 near to 47h; later Andrews found the same effect with a characteristic interval near 48h for flares of V 1216 Sgr. A closeness of all these quasiperiods to values wick are divisible by 24h supposes a possible effect of observational selection. The most detailed consideration of a possible periodicity of flares has been carried out by Chugainov: he has studied the time distribution of 28 flares which, were registered during a cooperative observation of UV Cet organized by Lovell at several observatories. Chugainov has found as the best periodic representation of maximum flare moments: T_max = const + 0.1821d X E. 11 periods of this cycle are equal to 48.1h. But the deviations, O-C, are large: mean(O-C) = 43m and (O-C)_max = 99m. The registered flares have occurred not in all, but only in 70 per cent of "critical moments"; but that is not a contradiction to the hypothesis of periodicity of flares: the remaining 30 per cent of flares could have small amplitudes or occurred on the opposite side of the star. The arguments against the periodicity hypothesis are the large mean(O-C), a value which is close to 1/6 of the period proposed, and a possibility to represent the observable time distribution of flares as a Poisson distribution. This year the Working Group on Flare Stars organized several cooperative observations of UV Ceti type stars with attemps to realize a 24^h photometric patrol. We hope to receive an important information on the time-distribution of flares from these observations, but their discussions have not yet been finished. Nearly 15 years ago Oskanjan (1964) has found variations of the flare activity level of UV Cet from season to season. A list of photoelectric observations of this star made by the end of 1967 is given in Table 1 (Gershberg and Chugainov, 1968). It is seen that the mean monitoring time per flare spent by different observers varies from 4.1h to 47h. But this table does not permit to reach a final conclusion: first, using different telescopes and different spectral bands we have different thresholds of flare detection; second, it is not clear whether a mean monitoring time per flare can characterize a flare activity level. In order to clear up these points, let us consider Chugainov's, observations of UV Ceti which were carried out for 4 years with the same instrumental and photometric system. Three different criteria of the flare activity level are given in Table 2: the mean monitoring time per flare, the mean radiative energy of a flare and the ratio of the radiative energy of flares to the radiative energy of the star calculated by integrations over the monitoring time. These data show the reality of the flare activity level variations and detect some correlation between different criteria of this level. Table 1. List of photoelectric observations of UV Ceti Total Mean Spectral monitoring Number monitoring Observer Season Telescope region time of flares time per flare (hours) registered (hours) Roques 1952 12" refractor without 94 2 47 filter Chugainov 1963 64 cm meniscus V 25 3 8.4 Chugainov 1964 telescope V 47 4 12 Chugainov 1965 V 70 17 4.1 Chugainov 1966 70 cm reflector H_beta 49 12 4.1 Eksteen 1966 16" reflector V 24 3 8.0 Chugainov 1967 64 cm meniscus V 35 8 4.4 telescope Table 2. Different criteria of the flare activity level of UV Ceti Ratio of radiative Number of flares Mean monitoring Mean radiative flare energy in Season registered time per flare energy of flare V-region to the stellar (hours) in V-region (ergs) radiation in V during monitoring 1963 3 8.4 9.3 X 10^30 0.0060 1964 4 12 3.2 X 10^30 0.0014 1965 17 4.1 9.0 X 10^30 0.012 1967 8 4.4 8.3 X 10^30 0.012 Before finishing the discussion of time characteristics of UV Ceti type star flares and going to photometric characteristics, it is necessary to note, that observations carried out by different instrumental methods give us results which are difficult to compare. As seen from Table 3, even experienced visual observers overestimate systematically the amplitudes of flares registered and miss small flares. On the other hand, observations in UV region have a threshold of flare detection three times lower than those in blue and 9 times lower than those in visual region (Kunkel, 1967). This point complicates the statistical discussion of flare features. Table 3. Comparison of the results of simultaneous visual (Odessa) and photoelectric (Crimea) monitoring of the brightness of UV Ceti Date U. T. M_vis m_v 19.9.65 21h 03m 0.9 20.9. 00 18 2.1 1.0 00 52 0.35 22.9. 22 59 0.4 23.9 23 46 1.1 0.65 24.9 00 32 2.9 1.9 26.9 00 47 4.0 >=1.5 28.9 00 11 0.4 21 09 2.1 1.15 1.10. 21 12 2.3 1.4 2.10. 21 57 0.4 23 54 4.2 1.7 Table 4. Comparison of the observed and calculated Balmer decrements according to Kunkel (1967) The observations of EV Lac flare on 11.12.1965 U. T. H_beta H_gamma H_delta H_zeta H_eta H_10 H_11 3h 55m 1.0 1.24 1.48 1.22 1.17 0.94 0.80 4 00 1.0 1.04 1.16 0.92 0.63 0.64 0.47 4 03 1.10 1.28 1.10 0.90 0.67 0.59 4 08 1.0 1.13 1.06 0.76 0.54 0.52 0.38 4 56 1.0 1.15 0.90 Photometric characteristics and energetics of flares Light curves of UV Ceti type star flares are very asymmetrical: as a rule, after a very quick increase of brightness there is a sharp, momentary maximum which is followed by a smoother decay (see Figs. 4 and 6). According to statistics (Gershberg and Chugainov, 1968) which is based on the discussion of about 100 photoelectric light curves, the time of flare growth is 10 to 30 sec for the half of the flares and 3 to 100 sec for 90 per cent of the flares. The time of photometric decay of flares is 10 to 100 times as large as that of flare growth, but, as a rule, the rate of increase of energy output just before the maximum is only 2 to 3 times as large as the rate of decrease of energy output immediately after the maximum. Then the flare decay slows down and such details as secondary maxima and steps of constant brightness appear on the light curve. Strong secondary maxima occur usually 5-10 min later than the main maximum and the light curve of secondary maximum is more symmetrical; this photometric feature can be regarded as a criterion to distinguish two close flares from a flare with a secondary maximum. As a rule, on the ascending branch of the light curve - in contrast to the descending branch - no deviations from a monotone growth of brightness are seen. Often, but not always, a slow brightening appears some minutes before the sharp beginning of the flare and the amplitude of such a slow brightening amounts to several tenths of a stellar magnitude. Flares of UV Ceti type stars are known with amplitudes up to 3-4 magn. Of course, the lower limit of flare amplitudes is determined with the precision of photometric observations. The behavior of UV Ceti type stars outside the flares is not clear up to now; the observers, who were monitoring the brightness of these stars visually and photographically, sometimes noted small and slow variations of brightness with amplitudes up to 0.3-0.5 magn. and with a characteristic time close to half an hour; but such secondary brightness variations were not confirmed by special photoelectric observations. According to Gershberg and Chugainov (1968) and Kunkel (1967) the total radiation of flares of the most active UV Ceti type stars amounts to 0.1-1 per cent of the energy of the radiation of these stars outside the flares. For the best studied 4 flare stars the distributions of flares according to their energy of radiation (L) are given in Fig. 1. One sees that the total energy of flare radiation in blue region amounts to 3 X 10^(31+-2) ergs and more than half of the flares radiate 10^(31+-1) ergs. Fig. 1 permits to conclude that an absolutely brighter star shows stronger flares on the average and certainly this conclusion can not be due to the observational selection effect. For the same 4 stars the distributions of flares according to their absolute rates of increase of energy output before maximum (dl/dt) are given in Fig. 2. In all investigated cases these rates were within the limits 10^27 and 3 X 10^28 ergs/sec^2. The narrowness of these histograms should be noted, they are 2-3 times narrower than the previous ones. It is suspected that the brighter the star is, the slower are the flares on an average, but we did not find any correlation between the total radiative energy of individual flares and their rate of increase. Fig. 1. Flare distributions according to their total radiative energy for 4 UV Ceti stars. Non-dashed districts are less certain data. Fig. 2. Flare distributions according to their absolute rates of energy output increase before maximum for 4 UV Cet stars. Non-dashed districts are less certain data. Intrinsic colors of flares The most certain and complete information on the intrinsic colors of UV Ceti type star flares was obtained by Kunkel (1967). By using his data a two-color diagram of flares is drawn in Fig. 3: the location of several flares of three UV Ceti type stars near their maxima are marked with different symbols and three broken lines represent the tracks of flares which could be studied colorimetrically for a long time. This diagram gives a good idea of the intrinsic colors of flares near their maxima (B - V approx. 0.0m +- 0.3m, U-B approx. -1.1m +- 0.2m) and of the character of a flare drift on the two-color diagram (to the right and slightly downwards) during their decay. Spectral features of flares In 1948 Joy and Humason (1949) took the first slit spectrogram of an UV Ceti flare. The examination of this unique plate taken with the exposure of 144 min has shown that during the flare the emission hydrogen lines became much stronger, CaII emission intensified, but to a less extent emission lines of HeI and HeII appeared which had not been seen in the quiet state star spectrum. Absorption lines almost disappeared, being veiled by a continuum which was very strong in UV spectral region; the spectrophotometric temperature of that continuum exceeded 10000 deg K, widths of the emission hydrogen lines amounted to 2 A, and the decrement was not steep. Fig. 3. Two-color diagram for UV Cet star flares. In the left and upper part of the plot there are the colors of hot ionized hydrogen clouds of different temperatures and optical thickness. Having used a high sensitive receiver (image tube), the spectral observations of flares were carried out in Crimea in 1965 and nearly 30 spectrograms of 10 flares of AD Leo and UV Cet were obtained with a time resolution from 20 sec up to 1 - 2 min (Gershberg and Chugainov, 1966, 1967); simultaneously the brightness of the star was being monitored photoelectrically. The light curve of the strongest. AD Leo flare registered by us is given in Fig. 4, the time intervals of flare spectrographying are marked too. Five spectra of this flare are reproduced in Fig. 5. During the strong flare the stellar spectrum transformed beyond recognition in the photographic region, but the changes were not so striking in green and red. The most prominent feature in all flare spectra is an intensification of Balmer emission lines. The quantitative treatment of the spectrograms shows that the equivalent widths of the emission hydrogen lines during the flares approach tens of angstroms, the augmentations of the widths at a half intensity level amount to 3-5 A. With the flare decaying, the continuum radiation decreases first of all and line emission decreases more slowly; sometimes line emission is still visible when a wide-band photoelectric photometry does not find a trace of a flare. The widths of emission lines return to the normal state more quickly than the intensities of these lines do. The helium lines were found only near flare maxima, the maximum in CaII takes place later than that in hydrogen lines. The veiling of intensity jumps near the TiO-band limits and that of the absorption line lambda 4227 A give a possibility to evaluate a part of flare continuum in the whole continuum radiation for moderate intense flares. Later Chugainov (1968) carried out two sets of photoelectric observations of EV Lac and UV Cet flares; he used narrow-band interference filters and confirmed time variations; he determined absolute values of equivalent widths of the H_beta-line spectrographically with the image tube device. The same year important spectral studies of UV Ceti type star flares were carried out by Kunkel (1967). Kunkel's essential success is an investigation on the UV spectral region of flares and spectrophotometric measurements at a wide wave-length interval. Kunkel has found and measured the emission jump near the Balmer limit, he has confirmed the fact of a quick disappearance of the strong continuum radiation and the quick narrowing of emission lines after the flare maximum and found the Balmer decrement at several stages of the flare. He has shown that the relative rate of the flare decay in the lines is nearly one half of the rate in the continuum, and the decay in CaII is the slowest one. At present we do not possess any information on flare line profiles, their Doppler shifts and possible anomalies in abundances of elements and isotopes. Nowadays such investigations are on the very limit or beyond the limit of instrumental power. Fig. 4. Light curve of the AD Leo flare on 18.5.1965. Numbered rectangles mark time intervals of spectrographying the flare. Fig. 5 Spectrograms of AD Leo during the strong flare on 18.5.1965 and in the quiet state on 4.6.1965. Numbers on the left correspond to the numeration in Fig. 4. Polarimetric studies Attempts to measure the polarization of UV Ceti type flare radiation were undertaken more than once. But as it was shown by Efimov (1968), all those observations had been made without due regard to the extreme rapidity of UV Ceti type flares. It is clear that when studying polarimetrically a variable source, the whole cycle of consecutive measurements must be made during the time which is small in comparison with the characteristic time of the source variations. But UV Ceti type flares are weak, therefore the quantum fluctuations of flare radiation flux turn out to be an essential obstacle when using small or moderate telescopes and small time averages for polarimetric measurements. It is necessary to use a large telescope and a special rapid-acting polarimeter (may by, similar to the device which was constructed by Oskanjan, Kubichela and Arsenijevich in Jugoslavia some years ago) in order to obtain reliable results on polarization features of flares. Radio emission of flares Excluding the sun, the UV Ceti stars are the only stellar bodies from which radio emission is certainly registered up to now. Radio emission of UV Ceti flares was found by Lovell (1964) with Jodrell Bank radio telescopes in 1958. First the radio emission of flares was found statistically by superposition of the radio records during 23 small optical flares (Lovell, 1963). But now we have more than two dozens of individual radio flare records at wave lengths in the range from 20 cm to 15 m. Fig. 6. Radio and optical flare of UV Cet on 19.10.1963. The radio emission of flares varies as quickly as the optical radiation. The radio emission is characterized by a high brightness temperature: a moderate UV Cet radio flare distributed over the whole stellar disk corresponds to Tb approx. 10^15 deg K. A typical radio flare record and its light curve are given in Fig. 6. According to a rough estimate, a flare radiates 100 times less energy in the radio wave-length range than in the optical flare energy. Lovell (1964) found a certain delay in radio emission at the lower frequencies when observing the UV Cet flare on 25.10.1963 at two frequences. (Fig. 7) The radio emission of the V 371 Ori flare on 30.11.1962 was studied in the fullest detail (Fig. 8): Australian investigators found the flare radio emission at 3 frequencies and at 410 MHz sharp and deep fadings were observed (Slee et al. 1963). Fig. 7. Radio emission drift over frequencies during the UV Cet flare on 25.10.1963. Fig. 8. V 371 Ori flare on 30.11.1962. a) optical observations: solid line - photographical, dashed line - visual monitoring of brightness; b) radio observations: solid line is the smoothed record at 410 MHz, segments mark time intervals when the flare is recorded at other frequencies; c) the flare-record at 410 MHz. INTERPRETATION AND HYPOTHESES Let us go to phenomenological interpretations and the physical hypotheses related to UV Ceti flares. It is known that in 1924 Hertzsprung found for the first time a stellar flare, similar to UV Ceti flares, and in accordance with the spirit of the 20th years astronomy he supposed that the falling of an asteroid on the star could be regarded as a cause of the flare. Now this idea may be considered only as a historical curiosity. Since the beginning of intensive study of flare stars in 1948 nearly a dozen hypotheses have appeared. Today, the so-called nebular or chromospheric flare model has the closest contact with observations. Therefore, I shall give an account of this scheme and then shall describe other models and hypotheses in short. Nebular or chromospheric model The main supposition of the nebular model is that an optical flare is connected with a quick appearance of a hot ionized gaseous cloud above the photosphere of a cold star; this cloud is deprived of external sources of ionization and radiates due to irreversible recombinations. Let us compare this scheme with the observations. If during a flare the mass of the cloud is constant and its optical thickness is small, then it is not difficult to calculate the expected light curve. The comparison of 10 observed EV Lac flares with theoretical curves, which have been calculated for the simplest isothermal radiative process, are given in Fig. 9. (Gershberg, 1964). In half of the cases we have an agreement. Later calculations were carried out taking the cooling effects into account (Gershberg, 1967), and now we calculate the theoretical curves making allowance for self-absorption in the Balmer lines; as a result, the theoretical curve-family enriches and a possibility to fit the theory to the observations increases. But one ought not to undergo a delusion: on one hand, a rich theoretical curve-family makes the comparison of the theory with observations non-critical; on the other hand, no theoretical curves calculated for a homogeneous and uniformly expanded cloud are able to explain such details of light curves as secondary maxima and time intervals of constant brightness, and to interpret the ascending branches of flares. Therefore, the observed light curves are not in contradiction with the nebular model but this model is too primitive to give a complete theory of the observed light curves. It should be noted that many observers have represented the observed light curves of UV Ceti flares as one or two exponential curves, and this representation is not worse than the nebular one; but the exponential representation is not substantiated physically, it is an erroneous conclusion of a wrong hypothesis on a hot spot (see below). The nebular interpretation of color features of flares is given in Fig. 3 according to Kunkel (1967). The colors of a hot ionized hydrogen cloud, which has the optical thickness tau_H alpha = 0-10^5, are located in the left and upper part of the plot. From the relative positions of flares and nebular models on the two-color diagram one can conclude that the flare radiation at maxima has the same colors as the hot gas in the case of T_e >= 30000 deg K and tau_H alpha approx. 10^2 to 10^3. The approximate character of Kunkel's calculations (a stationary gas in LTE-conditions, self-absorption effects in coherent approximation) does not permit to insist on the values T_e and tau_H alpha obtained, but the agreement between observations and the nebular model is at hand. With the same certainty, Fig. 3 shows a flare drift according to the nebular model during the flare decay. Kunkel interprets this drift as an increasing contribution of an equilibrium hot photospheric spot - a photosphere's "burn" - to the surplus stellar radiation. Fig. 9. Nebular representations of 10 light curves of EV Lac flares. Qualitative spectral features of a UV Ceti type flare - the appearance of strong continuum emission and strong intensification of line emission - were known from previous observations by Wachmann (1939), Joy and Humason (1949), and Herbig (1956) and naturally they fit to the nebular model. Our observations (Gershberg and Chugainov 1966; 1967) and those of Kunkel (1967) permit to make a quantitative comparison. Fig. 10. Comparison of the observed equivalent widths of Balmer lines in the AD Leo flare on 18.5.1965. and corresponding equivalent widths calculated in accordance with the flare brightness at different times for the optically thin flare model. The comparison of the equivalent widths of Balmer emission lines observed in the AD Leo flare on 18.5.1965 with the equivalent widths calculated in accordance with the brightness of the flare and in supposition of an optically thin gas is given in Fig. 10. It is seen that at the beginning of the flare the equivalent widths observed are ten times less than the calculated ones. The same results follow from other spectrograms of AD Leo and UV Cet flares. Since our calculations have been carried out for T_e <= 80000 deg K, the obtained disagreement can be explained either by a higher temperature, or by self-absorption effects in the lines. Kunkel's observations decide this dilemma: the considerable magnitude of the emission Balmer jump found by him means T_e < 25000 deg K. It should be noted that this upper limit of temperature may not be far from the real temperature because of the existing HeII emission. Comparing the observed and calculated Balmer decrements (see Table 5) Kunkel has found an independent and decisive argument for the chromospheric flare model: observed line intensity ratios from H_beta to H_11 correspond to radiation of the gas of the temperature T_e = 20000 to 25000 deg K and of the same optical thickness what has been obtained from colorimetric studies. Physical hypotheses The phenomenological nebular model of flares permits several physical interpretations. The possible existence of a characteristic time between flares suggests the assumption of active regions on the stellar surface and of a rather quick rotation of the star. The origin of a hot cloud above the photosphere can be connected either with a shock wave appearance, or with a hot "bubble" coming to the surface (Gorbatzkij, 1964), or with solar chromospheric flare type processes. Table 5 Calculated Balmer decrements for LTE-conditions, coherent re-emission in lines, n_e = 30x 10^13 cm^-3 and V_turb. = 20 km/sec. lg tau_H alpha H_alpha H_beta H_gamma H_delta H_xi H_eta H_10 H_11 T_e = 20000 deg K 2.0 0.55 1.00 1.06 0.74 0.32 0.22 0.16 0.12 2.5 0.63 1.00 1.32 1.27 0.68 0.49 0.36 0.27 3.0 0.76 1.00 1.31 1.55 1.26 0.97 0.73 0.56 3.5 0.81 1.00 1.21 1.45 1.65 1.49 1.26 1.02 4.0 0.84 1.00 1.13 1.29 1.57 1.64 1.60 1.46 T_e = 25000 deg K 2.0 0.51 1.00 1.09 0.78 0.35 0.24 0.17 0.13 2.5 0.58 1.00 1.37 1.35 0.74 0.53 0.39 0.29 3.0 0.68 1.00 1.36 1.65 1.37 1.06 0.81 0.62 3.5 0.75 1.00 1.25 1.54 1.79 1.64 1.39 1.12 4.0 0.78 1.00 1.17 1.36 1.70 1.80 1.76 1.61 An analogy between solar and UV Cet type flares was suspected by Greenstein and Whipple nearly 20 years ago, then it was spoken about by the Burbidges, Schatzman, Struve and al., but now this analogy becomes clearer and deeper. Indeed, both types of events have a strongly pronounced explosive behavior, they have no clear periodicity, but there are epochs of different activity level; the same emission lines appear in the spectra of both types of flares and there is similarity in the sequence of the flaring of different lines and in the character of intensity and line width variations during the flare decay; both types of optical flares are accompanied by strong radio flares, and the physical parameters of the hot gas responsible for the optical flares are similar. Therefore we have a good reason to suspect the similarity in intrinsic physical causes of both types of flares. As it is known, the energy source of chromospheric flares is the magnetic field, and the solar activity is determined by magneto-hydrodynamic phenomena which are mostly caused by convective motions. As UV Ceti stars are bodies of small mass and their inner structure is completely convective, it is natural to expect strong magneto-hydrodynamic motions and, as a result, a flare activity in these stars. The decisive confirmation of this conception has been obtained recently. Poveda (1965) showed that the convection must be very strong in stars of low luminosity up to K1-stars and Haro (1968) found a high flare activity of stars (in clusters of different age) up to the same spectral class. The main but the only parameter of the modern theory of stellar evolution - the mass - is used in this conception; therefore, in order to explain the differences between dMe and UV Ceti type stars and between dMe and normal M dwarfs, it is necessary to appeal either to evolutionary considerations or to additional parameters, as rotation, anomaly of element abundance, binary systems' features etc. We may note that the approximate characteristic time of flare activity level variations is a few months for UV Cet and the orbital motion characteristic time is tens of years for the L726-8AB-system (UV Cet = L726-8B), therefore a close connection between these phenomena can not exist. Alternative models and hypotheses Finally we submit some critical comments on other phenomenological models and physical hypotheses. Recent results on UV Ceti-type star flares permit to reject some previous models of flares. For instance, the ideas about a flare as an appearance of a hot equilibrium photospheric spot, and hypotheses on pure synchrotronic or pure Compton nature of flare radiation must be rejected in view of the discovery of the Balmer emission jump and the strong line emission in flares. It is also necessary to reject different hypotheses of external excitation of flares as the flare frequency and the drift of radio emission to long wavelength range indicate inner causes of these events. Ambarzumjan (1954) supposed that the UV Ceti type flares were connected with ejections of an unknown "pre-stellar matter" from the stellar interior. Recent experimental data do not confirm this hypothesis but they do not disprove it either. Since 1956 Schatzman (1967) has developed a stellar vibrational instability theory and applied it to different stellar flares; but this mechanism cannot be responsible for UV Ceti type flares as its basis is a strong dependence of the thermonuclear reaction rate on temperature and density of matter and such a reaction is not maintained in small mass M dwarfs. Kolesnik (1966) has used a maser effect in nonequilibrium plasma to interpret the flares. The necessity of such a hypothesis is not obvious today. For the last years Gurzadjan (1965, 1966) has been developing a theory of UV Ceti type flares; all nebular features of flares are regarded as secondary effects while the primary one is supposed to be a fast electron ejection and their Compton interaction with the photospheric radiation. According to this hypothesis the electron energy must exceed 10^5 to 10^7 times the optical energy of the flare and does not excite any observable consequences on the star; this is the main difficulty in Gurzadjan's theory. It is my pleasure to thank Dr. W. E. Kunkel, who courteously sent me his Dissertation. Prof. Haro for making available a preprint of his review on flare stars and Dr. P. F. Chugainov, Chairman of Working Group on Flare Stars, for the information on premilinary results of cooperative observations. All these data were very useful while preparing this report. REFERENCES Ambarzumjan, V. A., 1954, Sobbsch. Bjurak. Obs. 13. Andrews, A. D., 1966, Publ. astr. Soc. Pacific 78, 324; 542. "Non stable stars". Ed, by. Arakeljan, M. A., Erevan 1957. Chugainov, P. F., 1966, Astr. J. 43, 1168. Chugainov, P. F., 1968, Izv. Krym. astrofiz. Obs. 38, 200, 40. Efimov, YU. S. 1968, Krym. astrofiz. Obs. 41. Gershberg, R. E., 1964, Izv. Krym. astrofiz. Obs. 32, 133. Gershberg, R. E., 1967, Izv. Krym. astrofiz. Obs. 36, 216. Gershberg, R. E., and Chugainov, P. F., 1966, Astr. Zu. 43. 1168. Gershberg, R. E., and Chugainov, P. F., 1967, Astr. Zu. 44, 260. Gershberg, R. E., and Chugainov, P. F., 1968, Izv. Krym, astrofiz. Obs. 40. Gliese, W., 1958, Z. Astrophys. 45, 293. Gorbatzkij, V. G., 1964, Astr. Zu. 41, 53. Gurzadjan, G. A., 1965, Astrofizika 1, 319, C. R. Ac. Sci. USSR. Gurzadjan, G. A., 1966a, Astrofizika 2, 217. Gurzadjan, G. A., 1966b, C. R. Ac. Sci. USSR, 166, 53. Haro, G., 1968, Flare stars. "Clusters and binaries", Ed. by B. Middlehurst, Chicago. Non-stable stars. Symposium No. 3. IAU. Ed by Herbig, G. H. Cambridge, 1957. Herbig, G. H., 1956, Publ. astr. Soc. Pacific 68, 531. Joy, A. H. and Humason, M. L., 1949, Publ. astr. Soc. Pacific 61, 133. Joy, A. H., 1960, "Stellar Atmospheres". Ed. by Greenstein, J. L. Chicago, p. 666. Kolesnik, L G., 1966, Dissertation, Kiev. Kunkel, W. E., 1967, An optical study of stellar flares. Austin, Texas. Lippincott, S. L., 1953, J. astr. Soc. Can. 47, 24. Lovell, B., Whipple, F. L., and Solomon, L. H., 1963, Nature 198, 228. Lovell, B., Whipple, F. L., and Solomon, L. H., 1964, Nature 201, 1013. Lovell, B., 1964. Scient. Am. 211, No. 2, 13. Oskanjan, V., 1964, Publ. Obs. astr. Beograd No. 10. Petit, M., 1961, J. Observateurs 44, 11. Poveda, A., 1964, Nature 202, 1319. Slee, O. B., Solomon, L. H., and Patston, G. E., 1963, Nature, 199, 991. Wachmann, A. A., 1939, Beob. Zirk. 21, 25. COMMENT Lortet-Zuckerman: I did not understand whether the nebular model may account for the high observed ratio of radio to optical energy. The ratio radio to optical energy is two or three orders greater for flare stars than for the sun.