• numerical simulations
• stellar winds
• mass loss
• stellar evolution
• massive stars

Mass loss is of paramount importance for the lives of massive stars. It influences
their evolution, and changes their final fate (successful supernova explosion of
different types, or collapse), and remnant (neutron star or black hole). Two broad
categories of mass loss mechanisms are expected for massive stars: (i) radiatively
driven stellar winds; (ii) extreme events (e.g. eruptions, pulsational instabilities,
wave driven mass loss). The latter could, in principle, strip away large fractions
of a star’s mass in a very short time. In a binary system, Roche Lobe Overflow is
another mass loss mechanism from the point of view of the primary star.
However, because of its intrinsically dynamical nature, and because of the
strong non-linearity of the responsible physical processes driving it, mass loss is
one of the largest sources of uncertainty in the simulation of massive star evolu-
tion.
In most stellar evolution simulations, only wind mass loss is included through
parametric algorithms obtained as combinations of different formulae for each
phase of the evolution. These formulae express the mass loss rate as a function of
an (arbitrarily) chosen set of stellar parameters, dM/dt ≡ dM/dt( L, T_eff , Z, ...), and have
empirical or theoretical grounds. Moreover, it is common practice to use an ef-
ficiency factor η, whose (positive) value is not unique (and lacks a direct phys-
ical interpretation derived from first principles). η modifies the rate to account
for possible biases in the mass loss rate determination (e.g. potential overesti-
mation due to the assumption of homogeneity in the observed wind structures).
Although it may be dominant in terms of the total mass shed, mass loss from
eruptive events is commonly neglected. These events are particularly difficult to
model both because of the uncertainties in the driving process(es) and because of
their inherent short timescales and multidimensionality.
This thesis attempts to understand and constrain the uncertainties connected
to mass loss in the evolution of massive stars in the (initial) mass range between
15Msun and 30Msun. This is done by computing a grid of stellar models with the
open-source stellar evolution code MESA, modified to include some mass loss
algorithms, customized stopping criteria and stricter timestep controls. I carried
out several simulations with varying initial masses and differing only in the wind
algorithms and efficiencies, and compare the output. I then perform a simplified
numerical experiment to simulate mass stripping in a 15Msun star, to study how
its structure and evolution change with the removal of portions of its envelope
at different moments in its evolution. In a simplified way, this mass stripping
procedure mimics eruptive, violent events, or possibly Roche Lobe Overflow in a
binary.
The use of different wind mass loss scheme produces significant disagreement
among the simulated evolutionary tracks. I show that, although the mass loss al-
gorithms compared are just, in principle, different parametrization of the same
physical phenomenon, they are not equivalent. The uncertainty on mass loss in-
creases at higher initial masses and the η factor has a strong influence on the
results.
The stripped models obtained by artificially removing different portions of the
hydrogen-rich envelope of a star at different moments of its evolution all have
the same core structure (i.e. the mass stripping at the moment chosen does not
influence the core structure), but significant differences are found at intermedi-
ate masses. I also find significant variation in the density gradient within the
hydrogen-rich envelope that remains at the onset of core collapse.
In Chapter 1, I review the importance of massive stars in the broad context of
astrophysics, and outline the standard picture of their evolution. Then, I present
the current challenges encountered in the numerical modelling of these stars with
one-dimensional stellar evolution codes. Finally, I discuss the possible mass loss
channels for massive stars and review the mass loss algorithms commonly used
to model steady wind mass loss. Signatures of stellar winds which allow obser-
vational determinations of the mass loss rate are discussed too.
In Chapter 2, I review the basics of the MESA code, and present the cus-
tomized routines that I implemented for this work. I also describe the setup of
the simulations in detail, with the explicit aim of providing all the information
needed to reproduce the results. Chapter 2 also contains the description of the
simplified procedure to simulate a violent, short, eruptive mass loss event or an
envelope stripping caused by a companion star, and the advantages and short-
comings of this method.
In Chapter 3, I present my grid of models to compare, in a systematic way, the
various wind mass loss algorithms. I compare separately the algorithms for each
evolutionary phase, and I discuss the different resulting evolutionary tracks and
final characteristics of the stellar structures.
In Chapter 4, I discuss the results of the simplified model for an envelope
shedding mass loss event in a 15Msun star, such as may arise from pulsational
or wave driven mass loss events, or in the evolution of a binary system when
the primary star increases its radius. I compare the outcome of the simplified
procedure adopted to remove the envelope to the unstripped reference model,
and discuss the pre-supernova structures resulting from these simulations.
Chapter 5, summarizes the main results and suggests possible observational
probes and further research directions to better understand the uncertainties in
massive stars mass loss.
Appendix A contains the routines implemented for this work, and the MESA
parameter files used. Appendix B is a discussion of some of the computational
issues encountered in this work.