2. The sessions

Session 1, Large Surveys and Key Observations gave an overview of new data being acquired and analyzed, highlighting their extraordinary quality and quantity. Fundamental properties, as derived by comparing these data to standard single-star model predictions were shown to compare favorably in many cases, but in many others some form of binary interaction effects were thought to be needed.

The large quantity of data makes the use of automated methods necessary. Attention was drawn to machine learning tools being developed and some of the spectacular results being obtained from using, for example, neural networks. However, it was pointed out that at least in the case of spectroscopic surveys of stellar objects, fitting theoretical models to the observations is far from straightforward. Spectral fitting routines using machine learning methods need to account for binarity, micro-turbulence and line-profile variability associated with pulsations, surface convective regions and wind clumping. It was emphasized that a careful one-by-one analysis of sub-samples and peculiar objects cannot be neglected. Surveys alone can not solve all problems and need to be supported by detailed assessments of individual stars.

In Session 2 we heard results on High-z stars: observed and synthetic. One of the many highlighted results is the discovery of very highly redshifted spectra displaying a strong FUV rest-wavelength spectral energy distribution and prominent emission lines of nuclear processed material such as nitrogen, carbon and oxygen. These properties at such high redshifts are not predicted by population synthesis models based on standard stellar evolutionary tracks. An extraordinary example of these discoveries is a JADES NIRSpec object at a z=11 and believed to have formed $\sim$ 430 Myr after the Big Bang (Bunker et al. Reference Bunker2023). It exemplifies the tension between current theoretical models and observations of early universe objects as it implies that a lot of the light we see at high redshift must originate in massive stars. These stars must have very efficient mixing processes, such as in chemically homogeneous evolutionary models, a top-heavy initial mass function, and very efficient star formation rates.

Sessions 3 and 4 dealt with Stellar evolution and atmosphere models and observations. Recent results on the interior structure and evolution of massive stars were presented, several making use of Kippenhahn-diagrams to illustrate the results because they provide a useful birds-eye view of the physical location of nuclear burning, convective, and radiative zones over time from the ZAMS to the pre-SN phases. The primary takeaway of these presentations is that there is a lot of Physics involved in constructing the theoretical models! The processes involved include: nuclear reaction networks and rates, equations of state, opacity, neutrino losses, convection, mass-loss, and magnetic fields. Most of these have a dependence on metallicity and they are not all independent of each other. Tweaking only one of these processes can significantly affect the fate of massive stars and their products.

The metallicity dependence that is found in large 1D stellar structure grids was discussed, and it was noted that the emerging 3D simulations can be used to improve the prescriptions used in 1D models, especially important for parameters such as the convective overshoot. A striking novel result of multi-dimensional calculations is that late-phase nuclear-burning shells suffer instabilities that can lead them to merge (Rizzuti et al. Reference Rizzuti2024). This effectively replaces the idealized onion-ring picture of the pre-SN internal structure.

Observational constraints on the interior structure depend on another layer (both literally and figuratively!) of models. Specifically, once the internal structure has been modeled, stellar atmosphere models are required to deal with the large temperature and density gradients in the atmosphere, as well as with large scale mass motions such as winds. Important new developments include results from multi-dimensional models showing the presence of a turbulent atmospheric region below the photosphere, and a highly structured developing wind region above it (see N. Moens, these Proceedings). Comparison with observations requires detailed radiative transfer calculations which, however, are challenging in multi-dimensions even for isotropic, smooth flows.

Quantitative spectroscopy is a powerful tool for obtaining strong constraints on fundamental parameters (effective temperature, surface gravity, metallicity and projected rotation velocity). Full advantage of this technique requires observations obtained with sufficiently high signal-to-noise (S/N $\geq$ 200) and spectral resolution (R $\sim$ 100,000). Such observations are extremely challenging to obtain for even relatively nearby objects (due to limited availability of telescope time) but are important for ascertaining the uncertainties of results derived only from lower quality data.

All of the above comments refer stellar properties viewed through the lense of single-star models. Incorporating the manner in which internal and surface physical processes respond to binary-induced perturbations during the main sequence is mostly unchartered territory. The focus of presentations in this Symposium, with only a few exceptions, was on mass-transfer and orbital angular momentum processes leading to stripped stars, all within scenarios that assume the binary system to be in equilibrium. I will comment on this assumption and possible implications for non-equilibrium systems in Section 3.

Session 5 focused on Asteroseismology, Spectropolarimetry, and Interferometry. We heared how these methods can be used to significantly constrain the properties of massive stars. For example, the internal rotation and mixing characteristics can be gleaned through asteroseismology, while interferometry allows spatially resolving binary companions and circumstellar material, and spectropolarimetry yields information on magnetic fields and large scale departures from symmetry. The talks in this session were highly informative and were delivered in such a way that the the general methodolgy was easy to understand and the impact of the results clearly evident. Detailed results were also found in the posters.

In Session 6, Final products of massive stars, it was pointed out how critically important for the final stages of massive star evolution are the processes that preceded them, particularly the interior mixing efficiency and mass-loss. The uncertainties in main sequence models propagate to later phases and are amplified as a star reaches the pre-supernova phase. Thus, connecting final products with progenitors is far from straightforward. The uncertainties are even greater when trying to relate the properties of mergers deduced from gravitational wave sources with their progenitors because of the long timescale between the final stellar evolutionary phase and the occurence of the merger. In this context, the caption of the xkcd cartoonFootnote † shown by F. Broekgaarden is quite relevant:

I don’t know how to propagate error correctly, so I just put error bars on all my error bars.

Session 7 dealt with Population Synthesis models and feedback. Population synthesis models are key to gleaning the physical properties and evolutionary histories of distant galaxies. They use an assumed initial mass function and stellar structure, evolution and atmosphere models. They include feedback processes and, in some cases, products of binary evolution. The models constructed with all these ingredients are used to predict observables, such as spectral energy distributions, which are compared with observations (Bruzual this Symposium; Walcher et al. Reference Walcher2011; Stanway Reference Stanway2020).

It was noted during the discussions that the inherent uncertainties in the individual input models are compounded when dealing with very distant galaxies. In particular, the unknowns in late stage massive star evolution introduce significant uncertainties. Despite all the uncertainties, it is interesting to note that population synthesis models successfully match spectral energy distribution of even the most distant galaxies and provide insight on the total mass and ages (Chevallard & Charlot Reference Chevallard and Charlot2016). However, models still face challenges regarding both feedback predictions (specially ionizing fluxes) and being able to reproduce all strong stellar UV features seen in integrated spectra simultaneously.

Nearby, low-metallicity populations are useful testing grounds for the models, particularly because they allow for high signal-to-noise and high spectral resolution observations (see, for example, Asa’d et al. Reference Asa’d2025). Thus, the importance of studying local low-metallicity analogues, such as NGC 346 and POX 186, and producing integrated properties to compare with population synthesis models of high-redshift objects was highlighted several times during the meeting. It was also pointed out that, in addition to the feedback processes due to winds and SN ejecta, wind-wind collisions in binaries and clusters also inject material into the ISM with potentially different properties from those contemplated as standard feedback processes (see A. Kashi, this Symposium).

Finally, in Session 8, Looking to the future, we got a glimpse of what the future may hold in terms of observational facilities which are currently in different phases of development. We heard about new and powerful MOS and other instruments for the very large telescopes; new spectropolarimeters; new initiatives for large baseline interferomenters; CubeSpec, a dedicated space mission for asteroseismology and others being considered for launch in the 2040s; and Hubble Space Telescope’s health and outlook into the 2030s. There was ample agreement that the POLSTAR project is essential for bridging the gap between Hubble’s UV capabilities and the next UV space-based facility. The importance of the UV spectropolarimetric data sets it will provide for understanding massive stars was even highlighted by the lyrics of a delightful song that closed J. Drake’s presentation.

3. Reflections on binary interactions

The daunting complexity of binary evolution has caused some people to decide that this problem is simply a cleverly disguised opportunity for theorists (and enterprising observers) to ascend into free-parameter heaven. (J. Gallagher Reference Gallagher1989).

A recurring theme throughout the Symposium was the importance of binary stars for understanding the integrated light from distant objects. Binary interactions were invoked as a possible source of very large inferred heavy element abundances in high redshift objects and their strong UV continua, both of these anomalous compared to current predictions. In this section, I would like to reflect on the assumptions adopted by most binary interaction models, and comment on some of the physical processes that are often omitted, including some that are not yet well understood and require significant advances before they can be included in grids of models.

It is first important to clarify that the term binary interactions encompasses a wide range of different and sometimes unrelated physical processes: dynamical ejection of a star during a supernova event or as a consequence of a third-body encounter; periodic variability in spectra and light curves; mass-transfer which strips away the hydrogen-rich external layer exposing the nuclear processed region below; and processes involving very efficient internal mixing.

I focus on the latter connotation because mixing not only transports nuclear processed material toward surface layers but also produces larger nuclear-processed stellar cores that live longer and can produce more massive end products (de Mink et al. 2009). In the limit of very large mixing efficiency, the transition from H-burning to He-burning in the core does not involve the expansion of outer layers and thus, when in a binary system, does not lead to the extensively invoked Roche Lobe Overflow mass transfer process. Instead, the star remains compact and evolves blueward on the Hertzprung-Russell Diagram. Thus, very efficient mixing leads to properties that could explain recent observations of early-universe objects.

According to current scenarios, efficient mixing is achieved by maintaining a star in rapid rotation over a significant portion of its main sequence lifetime (Heger et al. Reference Heger, Langer and Woosley2000; Meynet & Maeder Reference Meynet and Maeder2000). These theoretical models predict an increase in surface abundances of nuclear processed materials with increasing stellar rotation speed (Brott et al. Reference Brott2011). As summarized below, this expectation is not always met.