M-type Stars Research Articles

New M-type stars found in the BAO Plate Archive centered at α = +15h δ = +580

The BAO Plate Archive low-resolution spectral plate centered at α = +15h δ = +580 is analyzed to find new late-type stars. 23 new late-type stars have been detected. New objects were cross-correlated with GAIA DR3, 2MASS, AllWISE, TESS and SDSS catalogues. For the detected objects, we present luminosity classes estimated from Gaia DR3 and 2MASS photometry and available proper motions. The majority of the objects are M dwarfs. Two objects are M giants.

Dual-band Observations of the Asymmetric Ring around CIDA 9A: Dead or Alive?

While the most exciting explanation of the observed dust asymmetries in protoplanetary disks is the presence of protoplanets, other mechanisms can also form the dust features. This paper presents dual-wavelength Atacama Large Millimeter/submillimeter Array observations of a large asymmetric dusty ring around the M-type star CIDA 9A. We detect a dust asymmetry in both 1.3 and 3.1 mm data. To characterize the asymmetric structure, a parametric model is used to fit the observed visibilities. We report a tentative azimuthal shift of the dust emission peaks between the observations at the two wavelengths. This shift is consistent with a dust trap caused by a vortex, which may be formed by an embedded protoplanet or other hydrodynamical instabilities, such as a dead zone. Deep high-spatial-resolution observations of dust and molecular gas are needed to constrain the mechanisms that formed the observed millimeter cavity and dust asymmetry in the protoplanetary disk around CIDA 9A.

Polarization Criterion in Targeted SETI Observation

In this paper, we propose a novel method for distinguishing extraterrestrial intelligence (ETI) signals from radio frequency interference (RFI) by leveraging polarization features. We exploit the sinusoidal variation of the linearly polarized components of Stokes parameters with the parallactic angle as a characteristic signature of ETI signals, while such linearly polarized components remain relatively stable for terrestrial RFI. Typically, a minimum of 4–8 hr of observation time is required to detect these sinusoidal variations. The polarization approach in the search for extraterrestrial intelligence also enables us to study the radio stellar bursts emitted by M-type stars as ancillary science, which is relevant to assessing the habitability of exoplanets. Compared to the frequency drift method, the polarization method effectively reduces the required observation time for signal identification while improving the signal identification process.

Self-consistent simulation of photoelectrons in exoplanet winds: Faster ionisation and weaker mass-loss rates

Context. Close-in exoplanets undergo extreme irradiation levels leading to hydrodynamic atmospheric escape and the formation of planetary winds. The planetary mass-loss is governed by several physical mechanisms, including photoionisation that may impact the evolution of the atmosphere. The stellar radiation energy deposited as heat strongly depends on the energy of the primary electrons following photoionisation and on the local fractional ionisation. All these factors affect the model-estimated atmospheric mass-loss rates and other characteristics of the outflow in ways that have not been clearly elucidated. Moreover, the shape of the XUV stellar spectra strongly influences the photoionisation and heating deposition in the atmosphere. Substantial changes are to be expected in the planetary mass-loss rate. Aims. We study the effect of secondary ionisation by photoelectrons on the ionisation and heating of the gas for different planet-star systems. We focus on the local and planet-wise effects, to clearly demonstrate the significance of these interactions. Methods. Using the PLUTO code, we performed 1D hydrodynamics simulations for a variety of planets and stellar types. We included planets in the size range from Neptune to Jupiter, and stars from M dwarfs to Sun-like. Results. Our results indicate a significant decrease in the planetary mass-loss rate for all planetary systems when secondary ionisation is taken into account. The mass-loss rate is found to decrease by 43% for the more massive exoplanets to 54% for the less massive exoplanets orbiting Sun-like stars, and up to 52% for a Jovian-like planet orbiting an M-type star. Our results also indicate much faster ionisation of the atmosphere due to photoelectrons. Conclusions. We built a self-consistent model including secondary ionisation by photoelectrons to evaluate its impact on mass-loss rates. We find that photoelectrons affect the mass-loss rates by factors that are potentially important for planetary evolution theories. We also find that enhanced ionisation occurs at altitudes that are often probed with specific atomic lines in transmission spectroscopy. Future modelling of these processes should include the role of photoelectrons. For this purpose, we make a simple but accurate parameterisation for atomic hydrogen atmospheres available.

Statistical and Radio Analysis of Exoplanets and Their Host Stars

As of February 2022, over 4900 exoplanets have been confirmed. In this study, we conducted statistical analyses on both the exoplanets and their host stars’ parameters. Our findings suggest that the radius and true mass distribution of the exoplanets remain largely unchanged compared to prior research. However, we observed a correlation between the average eccentricity and the number of planets in a system, and fluctuations in the “size” of the planets may contribute to such variation. Moreover, we discovered that, among planets with precise measurements of radius, true mass, and semi-major axis, the true mass-radius relationship follows a power–law distribution. Interestingly, the power–law index tends to decrease from super-Earths to cold Jupiters, potentially due to atmospheric composition. We also revised the radius valley, and determined that M-type host stars with low mass and metal abundance exhibit high planetary ownership rates or harbor large-mass planets, suggesting a different planet formation mechanism than GK-type stars. Lastly, we assessed the possibility of detecting exoplanets using FAST and found that there are three planets in FAST sky that may be detected, namely CoRoT-3 b, GPX-1 b, and TOI-2109 b.

Correlations between laboratory line lists for FeH, CrH, and NiH and M-star spectra collected with ESPaDOnS and SPIRou

Context. Molecular bands of metal oxides and hydrides dominate the optical and near-infrared spectra of M dwarfs. High-resolution spectra of these bands have immense potential for determining many properties of these stars, such as effective temperature, surface gravity, elemental abundances, radial velocity, or surface magnetic fields. Techniques are being developed to do this but remain limited by the current availability and accuracy of molecular data and spectral line lists. Aims. This paper reports metal monohydride line lists selected from near-infrared and visible laboratory data to show that specific bands in several electronic transitions can be used to identify CrH, NiH, and FeH in M stars and to determine radial velocities from Doppler shifts. The possibility of measuring magnetic fields is also investigated for FeH and CrH. Methods. We used systematic cross-correlation analysis between unpolarised spectra from a selection of M stars and state-specific laboratory line lists. These lists were generated from a combination of existing data and new laboratory laser-excitation spectra recorded at Doppler-limited resolution, in zero-field conditions or in magnetic fields up to 0.6 tesla. Results. We show that transitions at visible wavelengths in FeH and NiH, usually neglected in the analysis of the spectra of M-type stars, do in fact contribute to the spectra, and we demonstrate the influence of magnetic sensitivity on selected transitions in CrH and FeH. Conclusions. Although the new line lists focus on transitions recorded at temperatures significantly lower than those of stellar objects, they remain pertinent because they cover some band-head regions of high spectral density. FeH bands can provide a useful supplement to atomic lines for the analysis of high-resolution optical and near-infrared spectra of M dwarfs. We demonstrate the influence of a magnetic field on CrH signatures around 862 nm.

The orbital period of the recurrent nova V2487 Oph revealed

ABSTRACT We present the first reliable determination of the orbital period of the recurrent nova V2487 Oph (Nova Oph 1998). We derived a value of 0.753 ± 0.016 d (18.1 ± 0.4 h) from the radial velocity curve of the intense He ii λ4686 emission line as detected in time-series X-shooter spectra. The orbital period is significantly shorter than earlier claims, but it makes V2487 Oph one of the longest period cataclysmic variables known. The spectrum of V2487 Oph is prolific in broad Balmer absorptions that resemble a white dwarf spectrum. However, we show that they come from the accretion disc viewed at low inclination. Although highly speculative, the analysis of the radial velocity curves provides a binary mass ratio q ≈ 0.16 and a donor star mass M2 ≈ 0.21 M⊙, assuming the reported white dwarf mass M1 = 1.35 M⊙. A subgiant M-type star is tentatively suggested as the donor star. We were lucky to inadvertently take some of the spectra when V2487 Oph was in a flare state. During the flare, we detected high-velocity emission in the Balmer and He ii λ4686 lines exceeding −2000 km s−1 at close to orbital phase 0.4. Receding emission up to 1200 km s−1 at about phase 0.3 is also observed. The similarities with the magnetic cataclysmic variables may point to magnetic accretion on to the white dwarf during the repeating flares.

New Cases of Superflares on Slowly Rotating Solar-type Stars and Large Amplitude Superflares in G- and M-type Main Sequence Stars

In our previous work, we searched for superflares on different types of stars while focusing on G-type dwarfs using entire Kepler data to study statistical properties of the occurrence rate of superflares. Using these new data, as a by-product, we found 14 cases of superflare detection on 13 slowly rotating Sun-like stars with rotation periods of 24.5–44 days. This result supports the earlier conclusion by others that the Sun may possibly undergo a surprise superflare. Moreover, we found 12 and seven new cases of detection of exceptionally large amplitude superflares on six and four main sequence stars of G- and M-type, respectively. No large-amplitude flares were detected in A, F or K main sequence stars. Here we present preliminary analysis of these cases. The superflare detection, i.e., an estimation of flare energy, is based on a more accurate method compared to previous studies. We fit an exponential decay function to flare light curves and study the relation between e-folding decay time, τ, versus flare amplitude and flare energy. We find that for slowly rotating Sun-like stars, large values of τ correspond to small flare energies and small values of τ correspond to high flare energies considered. Similarly, τ is large for small flare amplitudes and τ is small for large amplitudes considered. However, there is no clear relation between these parameters for large amplitude superflares in the main sequence G- and M-type stars, as we could not establish clear functional dependence between the parameters via standard fitting algorithms.

Stratospheric dayside-to-nightside circulation drives the 3D ozone distribution on synchronously rotating rocky exoplanets

ABSTRACT Determining the habitability and interpreting future atmospheric observations of exoplanets requires understanding the atmospheric dynamics and chemistry from a 3D perspective. Previous studies have shown significant spatial variability in the ozone layer of synchronously rotating M-dwarf planets, assuming an Earth-like initial atmospheric composition. We simulate Proxima Centauri b in an 11.2-d orbit around its M-type host star using a 3D coupled climate-chemistry model to understand the spatial variability of ozone and identify the mechanism responsible for it. We document a previously unreported connection between the ozone production regions on the photochemically active dayside hemisphere and the nightside devoid of stellar radiation and thus photochemistry. We find that stratospheric dayside-to-nightside overturning circulation can advect ozone-rich air to the nightside. On the nightside, ozone-rich air subsides at the locations of two quasi-stationary Rossby gyres, resulting in an exchange between the stratosphere and troposphere and the accumulation of ozone at the gyre locations. The mechanism drives the ozone distribution for both the present atmospheric level (PAL) and a 0.01 PAL O2 atmosphere. We identify the hemispheric contrast in radiative heating and cooling as the main driver of the stratospheric dayside-to-nightside circulation. An age-of-air experiment shows that the mechanism also impacts other tracer species in the atmosphere (gaseous and non-gaseous phase) as long as chemical lifetimes exceed dynamical lifetimes. These findings, applicable to exoplanets in similar orbital configurations, illustrate the 3D nature of planetary atmospheres and the spatial and temporal variability that we can expect to impact spectroscopic observations of exoplanet atmospheres.

Three-component modelling of O-rich AGB star winds

Stellar winds of cool and pulsating asymptotic giant branch (AGB) stars enrich the interstellar medium with large amounts of processed elements and various types of dust. We present the first study on the influence of gas-to-dust drift on ab initio simulations of stellar winds of M-type stars driven by radiation pressure on forsterite particles. Our study is based on our radiation hydrodynamic model code T-800 that includes frequency-dependent radiative transfer, dust extinction based on Mie scattering, grain growth and ablation, gas-to-dust drift using one mean grain size, a piston that simulates stellar pulsations, and an accurate high spatial resolution numerical scheme. To enable this study, we calculated new gas opacities based on the EXOMOL database, and we extended the model code to handle the formation of minerals that may form in M-type stars. We determine the effects of drift by comparing drift models to our new and extant non-drift models. Three out of four new drift models show high drift velocities, 87–310 km s−1. Our new drift model mass-loss rates are 1.7–13 per cent of the corresponding values of our non-drift models, but compared to the results of two extant non-drift models that use the same stellar parameters, these same values are 0.33–1.5 per cent. Meanwhile, a comparison of other properties such as the expansion velocity and grain size show similar values. Our results, which are based on single-component forsterite particles, show that the inclusion of gas-to-drift is of fundamental importance in stellar wind models driven by such transparent grains. Assuming that the drift velocity is insignificant, properties such as the mass-loss rate may be off from more realistic values by a factor of 50 or more.

Hubble Space Telescope imaging of the compact elliptical galaxy M32 reveals a dearth of carbon stars

ABSTRACT We present new Hubble Space Telescope Wide-Field Camera 3/infrared medium-band photometry of the compact elliptical galaxy M32, chemically resolving its thermally pulsating asymptotic giant branch stars. We find 2829 M-type stars and 57 C stars. The carbon stars are likely contaminants from M31. If carbon stars are present in M32 they are so in very low numbers. The uncorrected C/M ratio is 0.020 ± 0.003; this drops to less than 0.007 after taking into account contamination from M31. As the mean metallicity of M32 is just below solar, this low ratio of C to M stars is unlikely due to a metallicity ceiling for the formation of carbon stars. Instead, the age of the asymptotic giant branch (AGB) population is likely to be the primary factor. The ratio of AGB to red giant branch stars in M32 is similar to that of the inner disc of M31 which contain stars that formed 1.5–4 Gyr ago. If the M32 population is at the older end of this age then its lack of C-stars may be consistent with a narrow mass range for carbon star formation predicted by some stellar evolution models. Applying our chemical classifications to the dusty variable stars identified with Spitzer, we find that the x-AGB candidates identified with Spitzer are predominately M-type stars. This substantially increases the lower limit to the cumulative dust-production rate in M32 to > 1.20 × 10−5${\rm M}_{\odot } \, {\rm yr}^{-1}$.

Unravelling the period gap using LAMOST chromospheric activity indices

ABSTRACT In our recent catalogue of BY Draconis (BY Dra) variables based on Zwicky Transient Facility data, we found traces of a period gap in the period–colour diagram. We combined our BY Dra data base with catalogues from the Kepler and K2 surveys, revealing a prominent period gap. Here, we use this combined ZTF–Kepler–K2 data set to investigate the origin of the period gap observed for BY Dra stars using chromospheric activity indices. We use low- and medium-resolution spectra from the LAMOST Data Release 7 to derive magnetic activity indices for the Ca ii H and K and Hα emission lines. We find a strong dependence of chromospheric activity on both stellar mass and rotation period. For partially convective K–M-type stars, the activity decreases steeply up to an age of ∼700–1000 Myr, subsequently evolving to the type of low-level saturation associated with spin-down stallation. In contrast, F–G-type stars with thinner convective envelopes exhibit constant activity with increasing age. We suspect that the observed steep decrease for partially convective stars is driven by core–envelope coupling. This mechanism reduces differential rotation at the core–envelope transition, hence leading to decreased magnetic activity. Moreover, we derive activity indices for previously known star clusters and find similar trends as regards their activity levels as a function of age. In particular, very low-level activity is observed around the location of the period gap. Therefore, we conclude that the period gap, defined by the non-detection of variable sources, is driven by a minimum in chromospheric activity.

A Catalog and Statistical Analysis for Magnetic Stars

Magnetic fields are significant in the structure and evolution of stars. We present a comprehensive catalog of 1784 known magnetic stars, detailing their identifications, HD numbers, precise locations, spectral types and averaged quadratic effective magnetic fields among other important information. The group comprises 177 O-type stars, 551 B-type stars, 520 A-type stars, 91 F-type stars, 53 G-type stars, 61 K-type stars, 31 M-type stars and an additional 300 stars whose spectral classification remains indeterminate. Our analysis examines the statistical properties of these magnetic stars. The relative integrated distribution function and number distribution function for all magnetic stars of the same spectral type can be effectively approximated using an exponential function of the averaged quadratic effective magnetic field. The analysis further reveals that A and B-type stars possess the strongest mean magnetic fields, indicating an easier detection of their magnetic fields.

Evidence of Stellar Oscillations in the Post-common-envelope Binary Candidate ASASSN-V J205543.90+240033.5

ASASSN-V J205543.90+240033.5 (ASJ2055) is a possible post-common-envelope binary system. Its optical photometric data show an orbital variation of about 0.52 days and a fast period modulation of P 0 ∼ 9.77 minutes, whose origin is unknown. In this Letter, we report evidence of the stellar oscillation of the companion star as the origin of the fast period modulation. We analyze the photometric data taken by the Transiting Exoplanet Survey Satellite, the Liverpool Telescope, and the Lulin One-meter Telescope. It is found that the period of the 9.77 minutes signal measured in 2022 August is significantly shorter than that in 2021 July/August, and the magnitude of the change is of the order of ∣△P 0∣/P 0 ∼ 0.0008(4). Such a large variation will be incompatible with the scenario of the white dwarf (WD) spin as the origin of the 9.77 minutes periodic modulation. We suggest that the fast periodic signal is related to the emission from the irradiated companion star rather than that of the WD. Using existing photometric data covering a wide wavelength range, we estimate that the hot WD in ASJ2055 has a temperature of T eff ∼ 80,000 K and is heating the oscillating M-type main-sequence star with T eff ∼ 3500 K on its unirradiated surface. The stellar oscillation of the M-type main-sequence star has been predicted in theoretical studies, but no observational confirmation has been done. ASJ2055, therefore, has the potential to be a unique laboratory for investigating the stellar oscillation of an M-type main-sequence star and the heating effect on stellar oscillation.

Main Sequence Star Super-flare Frequency based on Entire Kepler Data

We wrote and used an automated flare detection Python script to search for super-flares on main sequence stars of types A, F, G, K and M in Kepler’s long-cadence data from Q0 to Q17. We studied the statistical properties of the occurrence rate of super-flares. For the G-type data set, we compared our results with the previous results of Okamoto et al. by splitting the data set into four rotational bands. We found similar power-law indices for the flare frequency distribution. Hence, we show that inclusion of a high-pass filter, sample biases, gyrochronology and completeness of flare detection is of no significance, as our results are similar to those of Okamoto et al. We estimated that a super-flare on G-type dwarfs with energy of 1035 erg occurs on a star once every 4360 yr. We found 4637 super-flares on 1896 G-type dwarfs. Moreover, we identified 321, 1125, 4538 and 5445 super-flares on 136, 522, 770 and 312 dwarfs of types A, F, K and M, respectively. We ascertained that the occurrence rate (dN/dE) of super-flares versus flare energy, E, shows a power-law distribution with , where α ≃ 2.0 to 2.1 for the spectral types from F-type to M-type stars. In contrast, the obtained α ≃ 1.3 for A-type stars suggests that the flare conditions differ from those of the other spectral-type stars. We note an increase in flare incidence rate in F-type to M-type stars and a decrease in A-type to F-type stars.

Opacity for realistic 3D MHD simulations of cool stellar atmospheres

Context. Realistic three-dimensional time-dependent simulations of stellar near-surface convection employ the opacity binning method for the efficient and accurate computation of the radiative energy exchange. The method provides several orders of magnitude of speedup, but its implementation includes a number of free parameters. Aims. Our aim is to evaluate the accuracy of the opacity binning method as a function of the choice of these free parameters. Methods. The monochromatic opacities computed with the SYNSPEC code were used to construct opacity distribution function (ODF) that was then verified through detailed comparison with the results of the ATLAS code. The opacity binning method was implemented with the SYNSPEC opacities for four representative cool main-sequence stellar spectral types (F3V, G2V, K0V, and M2V). Results. The ODFs from SYNSPEC and ATLAS show consistent results for the opacity and bolometric radiative energy exchange rate Q in the case of the F-, G-, and K-type stars. Significant differences, coming mainly from the molecular line lists, are found for the M-type star. It is possible to optimise a small number of bins to reduce the deviation of the results coming from the opacity grouping with respect to the ODF for the F-, G-, and K-type stars. In the case of the M-type star, the inclusion of splitting in wavelength is needed in the grouping to get similar results, with a subsequent increase in computing time. In the limit of a large number of bins, the deviation for all the binning configurations tested saturates and the results do not converge to the ODF solution. Due to this saturation, the Q rate cannot be improved by increasing the number of bins to more than about 20 bins. The more effective strategy is to select the optimal location of fewer bins.

Wapiti: A data-driven approach to correct for systematics in RV data

Context. Recent advances in the development of precise radial velocity (RV) instruments in the near-infrared (near-IR) domain, such as SPIRou, have facilitated the study of M-type stars to more effectively characterize planetary systems. However, the near-IR presents unique challenges in exoplanet detection due to various sources of planet-independent signals which can result in systematic errors in the RV data. Aims. In order to address the challenges posed by the detection of exoplanetary systems around M-type stars using near-IR observations, we introduced a new data-driven approach for correcting systematic errors in RV data. The effectiveness of this method is demonstrated through its application to the star GJ 251. Methods. Our proposed method, Weighted principAl comPonent reconsTructIon (referred to as Wapiti), used a dataset of per-line RV time series generated by the line-by-line (LBL) algorithm and employed a weighted Principal Component Analysis (wPCA) to reconstruct the original RV time series. A multistep process was employed to determine the appropriate number of components, with the ultimate goal of subtracting the wPCA reconstruction of the per-line RV time series from the original data in order to correct systematic errors. Results. The application of Wapiti to GJ 251 successfully eliminated spurious signals from the RV time series and enabled the first detection in the near-IR of GJ 251b, a known temperate super-Earth with an orbital period of 14.2 days. This demonstrates that, even when systematics in SPIRou data are unidentified, it is still possible to effectively address them and fully realize the instrument’s capability for exoplanet detection. Additionally, in contrast to the use of optical RVs, this detection did not require us to filter stellar activity, highlighting a key advantage of near-IR RV measurements.

Imaging the innermost gaseous layers of the Mira star R Car with GRAVITY-VLTI

Context. The mass-loss mechanisms in M-type asymptotic giant branch (AGB) stars are still not well understood; these include, in particular, the formation of dust-driven winds from the innermost gaseous layers around these stars. One way to understand the gas-dust interaction in these regions and its impact on the mass-loss mechanisms is through the analysis of high-resolution observations of the stellar surface and its closest environment. Aims. We aim to characterize the inner circumstellar environment (~3 R*) of the M-type Mira star R Car in the near-infrared at different phases of a pulsation period. Methods. We used GRAVITY interferometric observations in the K band obtained during two different epochs over 2018. Those data were analyzed using parametric models and image reconstruction of both the pseudo-continuum and the CO band heads observed. The reported data are the highest angular resolution observations on the source in the K band. Results. We determined sizes of R Car’s stellar disk of 16.67 ± 0.05 mas (3.03 au) in January 2018 and 14.84 ± 0.06 mas (2.70 au) in February, 2018, respectively. From our physical model, we determined temperatures and size ranges for the innermost CO layer detected around R Car. The derived column density of the CO is in the ~9.18×1018–1×1019 cm−2 range, which is sufficient to permit dust nucleation and the formation of stable dust-driven winds. We find that magnesium composites, Mg2SiO4 and MgSiO3, have temperatures and condensation distances consistent with the ones obtained for the CO layer model and pure-line reconstructed images, which are the dust types most likely to be responsible for wind formation. Our reconstructed images show evidence of asymmetrical and inhomogeneous structures, which might trace a complex and perhaps clumpy structure of the CO molecule distribution. Conclusions. Our work demonstrates that the conditions for dust nucleation and thus for initialising dust-driven winds in M-type AGB stars are met in R Car, and we identify magnesium composites as the most probable candidates. We find structural changes between two observing epochs (which are separated by ~10% of the full pulsation period of the star) and evidence of the effects of asymmetries and clumpiness. This observational evidence is crucial to constraining the role of convection and pulsation in M-type stars.

Photometric Observations of Flares on AD Leo from GWAC-F30 and TESS

We observed active M dwarf star AD Leo for 146 hr in photometry by GWAC-F30 and also analyzed 528 hr of photometric data of the star from the Transiting Exoplanet Survey Satellite (TESS). A total of 9 and 70 flares are detected from GWAC-F30 and TESS, respectively. Flare durations, amplitudes and energies are calculated. The distributions of the three properties and flare frequency distributions (FFDs) are given. Within the same energy range of flares, the FFDs of AD Leo obtained in this research and the previous study are basically consistent, which suggests that the magnetic activity of this star has not significantly changed compared to decades ago. Comparing with the average FFD of M-type stars, AD Leo’s FFD is twice higher, indicating that its magnetic activity is more active than that of the average level of the M-type. Based on the TESS light curve, AD Leo’s rotation period is calculated as days, supporting the result given in previous research. During the decay phase of the most energetic flare from TESS, we identified quasi-periodic pulsations and determined a 26.5 minute oscillation period, which is currently the longest period for AD Leo, suggesting that long periodic physical processes existed during the flare of this star.

Spectral classification of young stars using conditional invertible neural networks

Aims. We introduce a new deep-learning tool that estimates stellar parameters (e.g. effective temperature, surface gravity, and extinction) of young low-mass stars by coupling the Phoenix stellar atmosphere model with a conditional invertible neural network (cINN). Our networks allow us to infer the posterior distribution of each stellar parameter from the optical spectrum. Methods. We discuss cINNs trained on three different Phoenix grids: Settl, NextGen, and Dusty. We evaluate the performance of these cINNs on unlearned Phoenix synthetic spectra and on the spectra of 36 class III template stars with well-characterised stellar parameters. Results. We confirm that the cINNs estimate the considered stellar parameters almost perfectly when tested on unlearned Phoenix synthetic spectra. Applying our networks to class III stars, we find good agreement with deviations of 5–10% at most. The cINNs perform slightly better for earlier-type stars than for later-type stars such as late M-type stars, but we conclude that estimates of effective temperature and surface gravity are reliable for all spectral types within the training range of the network. Conclusions. Our networks are time-efficient tools that are applicable to large numbers of observations. Among the three networks, we recommend using the cINN trained on the Settl library (Settl-Net) because it provides the best performance across the widest range of temperature and gravity.