Search

Observations indicate that turbulent motions are present on most massive star surfaces. Starting from the observed phenomena of spectral lines with widths much larger than thermal broadening (e.g. micro- and macroturbulence) to the detection of stochastic low-frequency variability (SLFV) in the Transiting Exoplanet Survey Satellite photometry, these stars clearly have large scale turbulent motions on their surfaces. The cause of this turbulence is debated, with near-surface convection zones, core internal gravity waves, and wind variability being proposed. Our 3D grey radiation hydrodynamic (RHD) models characterized the surfaces' convective dynamics driven by near-surface convection zones and provided a reasonable match to the observed SLFV in the most luminous massive stars. We now explore the complex emitting surfaces of these 3D RHD models, which strongly violate the 1D assumption of a plane parallel atmosphere. By post-processing the grey RHD models with the Monte Carlo radiation transport code SEDONA, we synthesize stellar spectra and extract information from the broadening of individual photospheric lines. The use of SEDONA enables the calculation of the viewing angle and temporal dependence of spectral absorption line profiles. Combining uncorrelated temporal snapshots together, we compare the broadening from the 3D RHD models' velocity fields to the thermal broadening of the extended emitting region, showing that our synthesized spectral lines closely resemble the observed macroturbulent broadening from similarly luminous stars. More generally, the new techniques we have developed will allow for systematic studies of the origin of turbulent velocity broadening from any future 3D simulations., Comment: 16 pages, 9 figures, 1 table

Increasing main sequence stellar luminosity with stellar mass leads to the eventual dominance of radiation pressure in stellar envelope hydrostatic balance. As the luminosity approaches the Eddington limit, additional instabilities (beyond conventional convection) can occur. These instabilities readily manifest in the outer envelopes of OB stars, where the opacity increase associated with iron yields density and gas pressure inversions in 1D models. Additionally, recent photometric surveys (e.g. TESS) have detected excess broadband low frequency variability in power spectra of OB star lightcurves, called stochastic low frequency variability (SLFV). This motivates our novel 3D Athena++ radiation hydrodynamical (RHD) simulations of two 35$\,$M$_\odot$ star envelopes (the outer $\approx$15$\%$ of the stellar radial extent), one on the zero-age main sequence and the other in the middle of the main sequence. Both models exhibit turbulent motion far above and below the conventional iron opacity peak convection zone (FeCZ), obliterating any ``quiet" part of the near-surface region and leading to velocities at the photosphere of 10-100$\,$km$\,$s$^{-1}$, directly agreeing with spectroscopic data. Surface turbulence also produces SLFV in model lightcurves with amplitudes and power-law slopes that are strikingly similar to those of observed stars. The characteristic frequencies associated with SLFV in our models are comparable to the thermal time in the FeCZ ($\approx$3-7$\,$days$^{-1}$). These simulations, which have no free parameters, are directly validated by observations and, though more models are needed, we remain optimistic that 3D RHD models of main sequence O star envelopes exhibit SLFV originating from the FeCZ., Comment: 7 pages, 5 figures

Common variable star classifiers are built only with the goal of producing the correct class labels, leaving much of the multi-task capability of deep neural networks unexplored. We present a periodic light curve classifier that combines a recurrent neural network autoencoder for unsupervised feature extraction and a dual-purpose estimation network for supervised classification and novelty detection. The estimation network optimizes a Gaussian mixture model in the reduced-dimension feature space, where each Gaussian component corresponds to a variable class. An estimation network with a basic structure of a single hidden layer attains a cross-validation classification accuracy of ~99%, on par with the conventional workhorses, random forest classifiers. With the addition of photometric features, the network is capable of detecting previously unseen types of variability with precision 0.90, recall 0.96, and an F1 score of 0.93. The simultaneous training of the autoencoder and estimation network is found to be mutually beneficial, resulting in faster autoencoder convergence, and superior classification and novelty detection performance. The estimation network also delivers adequate results even when optimized with pre-trained autoencoder features, suggesting that it can readily extend existing classifiers to provide added novelty detection capabilities., Comment: 9 pages, 3 figures, 2 tables. Accepted to ApJ Letters