An initially static, plane-parallel, radiation pressure supported exponential atmosphere in a strong magnetic field is found to be unstable to the growth of buoyant low-density regions or ''photon bubbles'' in linear stability analysis. Here we present a series of numerical studies of the photon bubbles in such an atmosphere carried out using a quasi-two-dimensional radiation hydrodynamical code. When a single-mode perturbation is applied to the atmosphere, we find that the growth of these bubbles is in good agreement with the linear theory. When the evolution becomes nonlinear, the growth of the photon bubbles is found to be an efficient mechanism of energy transport. The presence of the low-density regions helps to increase the photon diffusion speed by a factor of several, while the buoyancy of the bubbles serves to transport energy via advection. Multimode studies, consisting of a perturbation with the linear combination of two single modes and a random perturbation, suggest that there is a tendency toward merger of the photon bubbles in their transport properties. The small wavenumber modes eventually dominate in radiation pressure, while the large wavenumber modes, although having higher growth rates, also saturate more quickly, and their contribution to the energy transport can be truncated. A grid study and the multimode calculations indicate that the energy transport through the atmosphere is well represented as long as the photon bubble mode with optical depth of similar to 10 wavelength is well resolved.Possible applications of the photon bubble instability includes the settling layer in the accretion column of a neutron star undergoing super-Eddington accretion. The enhanced energy transport may manifest itself in the emergent spectrum from the accreting pulsars and in the short-time variability in the light curves.

We describe time-dependent, two-dimensional axisymmetric radiation hydrodynamic calculations of locally super-Eddington accretion onto highly magnetized neutron stars appropriate to the flow onto the polar caps of high-luminosity X-ray pulsars. Our calculations show the development and nonlinear evolution of photon bubble instabilities in the settling plasma below the radiation-dominated shock that terminates free fall along the magnetic field. The photon bubbles develop as elongated, very low density, optically thin regions of local outflow forming typically a few kilometers above the neutron star's surface with outflow velocities of order 0.1c. These “holes” in the plasma reduce the total opacity of the accretion mound and through photon advection significantly enhance the efficiency of radiation transport. The number of bubbles varies from a few to 20 and results in significant fluctuations (4%-10%) in the emitted luminosity. These fluctuations appear as reasonably high quality (Q) “photon bubble oscillations” (PBOs) in the power spectrum of the emergent luminosity time series on timescales between 0.1 and 1 ms, a phenomenon not previously observed. The discovery of PBOs would provide a powerful probe into the physical characteristics of super-Eddington flows. We use our results to exhibit power spectra of the predicted luminosity oscillations and discuss the feasibility of their detection by the X-Ray Timing Explorer (XTE). We also use our results to calculate the emergent spectrum and show that it is a highly depleted modified blackbody distribution roughly consistent with known rotation phase-averaged observations in the 10-100 keV range. Thus we have shown that the spectra of X-ray pulsars are a natural consequence of the dynamics of polar cap accretion.

We reexamine both numerically and analytically the collapse of the singular isothermal sphere in the context of low-mass star formation. We consider the case where the onset of collapse is initiated by some arbitrary process which is accompanied by a central output of either heat or kinetic energy. We find two classes of numerical solutions describing this manner of collapse. The first approaches in time the expansion wave solution of Shu, while the second class is characterized by an ever-decreasing central accretion rate and the presence of an outwardly propagating weak shock. The collapse solution which represents the dividing case between these two classes is determined analytically by a similarity analysis. This solution shares with the expansion wave solution the properties that the gas remains stationary with an r(-2) density profile at large radius and that, at small radius, the gas free-falls onto a nascent core at a constant rate which depends only on the isothermal sound speed. This accretion rate is a factor of similar to 0.1 that predicted by the expansion wave solution: This reduction is due in part to the presence of a weak shock which propagates outward at 1.26 times the sound speed. Gas in the postshock region first moves out subsonically but is then decelerated and begins to collapse. The existence of two classes of numerical collapse solutions is explained in terms of the instability to radial perturbations of the analytic solution. Collapse occurring in the manner described by some of our solutions would eventually unbind a finite-sized core. However, this does not constitute a violation of the instability properties of the singular isothermal sphere which is unstable both to collapse and to expansion. To emphasize this, we consider a purely expanding solution for isothermal spheres. This solution is found to be self-similar and results in a uniform density core in the central regions of the gas. Our solutions may be relevant to the ''luminosity'' problem of protostellar cores since the predicted central accretion rates are significantly reduced relative to that of the expansion wave solution. Furthermore, our calculations indicate that star-forming cloud cores are not very tightly bound and that modest disturbances can easily result in both termination of infall and dispersal of unaccreted material.

The presence of a close binary companion can affect the evolution of a massive star through one or more episodes of mass transfer, or by merger in a common-envelope phase. Monte Carlo calculations indicate that ∼20–35% of all massive supernovae are affected by such processes, and that a substantial fraction of these events will be supernovae of Type II. The properties of the progenitor star, the distribution of circumstellar material, the peak supernova luminosity, the shape of the supernova light curve, and other observable features of the supernova event can be affected by prior binary membership. Binary interactions may be the cause of much of the variability among Type II supernova light curves. In particular, many of the peculiarities of SN 1987A and SN 1993J may well have resulted from the prior duplicity of the progenitors.

SUPERNOVA 1993J in the spiral galaxy M81 is the brightest supernova since SN1987A and, like the latter, appears to be another 'peculiar' type II supernova. The available photometry1,2 of the supernova region before the explosion requires the presence of at least two supergiants (one of early spectral type and the other of late type), but the actual progenitor has yet to be identified. Here we show that the explosion of a late-type supergiant can explain the initial sharp peak in the supernova light curve, provided that the star had lost almost all of its hydrogen-rich envelope before the explosion. In our model, the secondary brightening of the supernova, approximately 10 days later, is then a consequence of the radioactive decay of Ni-56 and subsequently Co-56) produced in the explosion. The progenitor could have lost its hydrogen-rich envelope either in a strong stellar wind or, as seems more likely, through mass transfer to a companion star. In the latter case, the companion should reappear after the supernova photosphere has receded, the system having become a binary composed of a neutron star with a massive stellar companion.

We here report the results of a systematic investigation of how binary interaction affects the presupernova evolution of massive stars and the resulting supernova explosions. We summarize the various types of binary interaction and the evolutionary scenarios in which they are realized. We also present the results of a series of hydrodynamical calculations which model the supernova explosion for various progenitor types and discuss their observable characteristics.

We have systematically investigated how binary interaction affects the presupernova evolution of massive close binaries and the resulting supernova explosions, using a Henyey-type stellar evolution code that we have modified to allow its application to binary stellar evolution calculations. With our modified code, we are able to follow the effects of mass and angular momentum loss from the binary, as well as mass transfer within the binary system. We find that a large number of binary scenarios can be distinguished, depending on the type of binary interaction and the evolutionary stage of the supernova progenitor at the time of the interaction. In general, the structure of a massive star can be affected in three fundamentally different ways: by mass loss, mass accretion, or common-envelope evolution.As a result of mass loss by Roche lobe overflow, stars can lose all of their hydrogen-rich envelopes and become helium stars, which are potential candidates for the progenitors of Type Ib supernovae. If the original masses of the binary components are nearly equal, it is possible (for reasonable assumptions about the mass and angular momentum loss rates) that the primary retains part of its hydrogen-rich envelope. In this case, the supernova progenitor would look like a more or less normal red supergiant, even though it may have lost most of its envelope, and the supernova would resemble a classical Type II-L supernova. In most cases, the system remains bound after the explosion, but it may acquire a substantial orbital eccentricity. The system may subsequently become unbound if the original secondary then evolves to become a second supernova.Mass accretion can also significantly alter the structure of the supernova progenitor, if it takes place after the main-sequence phase of the accreting star. The star may then end its life as a blue supergiant instead of a red supergiant. In this case, the resulting supernova explosion would resemble SN 1987A. At the time of the supernova explosion, the presupernova star still has a stellar or (more likely) a neutron star companion. However, since more than half of the total mass of the system is ejected in the supernova explosion, the system is likely to become unbound after the explosion.In the most dramatic case of binary interaction, in which a supernova progenitor captures its companion in a common envelope, two different outcomes are possible, depending on whether the envelope is ejected during the spiral-in phase or remains bound. If the envelope is ejected, the progenitor will become a helium star and the subsequent supernova explosion may be of the Type Ib variety. If the binary components merge completely, the final outcome would be a single star with no trace (except for possible chemical anomalies) of the original secondary. If, during the merger, a significant amount of mass is added to the envelope, the final star may again be a blue supergiant (similar to the results of the accretion scenario), and the resulting supernova would belong to the same class as SN 1987A.In order to assess the importance of the various scenarios, we performed Monte Carlo simulations to estimate the frequencies of occurrence of the individual scenarios. We find that, because of a previous binary interaction, 15%-30% of all massive stars (with initial masses greater than or similar to 8 M.) become helium stars, and another approximately 5% of all massive stars end their lives as blue supergiants rather than as red supergiants.These results may be directly applicable to Type Ib supernovae. Our estimate for the frequency of helium stars is comparable to the observed frequency of Type Ib supernovae, and hence we expect that the explosions of helium stars in binaries account for a substantial fraction of all Type Ib events.Our calculations may also help to answer one of the major puzzles about SN 1987A, namely, the question of why the apparent progenitor (Sk -69-degrees 202) was a blue supergiant rather than a red one, as had been generally expected for the precursors of Type II supernovae. In addition, binary models for SN 1987A may provide plausible explanations for a variety of other anomalies of SN 1987A, ranging from the asymmetric expansion of the ejecta and the variability of the soft X-ray flux to the barium anomaly and the "mystery spot."

A significant fraction of all stars are in binaries. It is thus reasonable to examine the possibility that the progenitor of SN 1987A was a component of a binary system. We here consider some of the ranges of parameter space in which binary evolution produces progenitors that are applicable to models of SN 1987A [1, 2, 3]. Other binary models have been proposed by FABIAN et al. [4] and CHEVALIER and SOKER [5]. Such models may help to explain several of the anomalies of this supernova, including the blue color of the progenitor, the asymmetries in the ejecta, the barium anomaly, and the mystery spot (see [1, 2, 3] for further discussion and references). Unlike many single-star models (see [1, 2, 3] for references), the fits of these models to the observations do not rely on the assumption of low metallicity, a non-standard treatment of convection, and/or the ad hoc stipulation of severe mass loss from the presupernova star. The binary models presented here make definite predictions that can be checked against observations; moreover, the models are sufficiently general that the results may be relevant to other supernovae as well.