Copyright by Athena Ranice Stacy 2011 - The University of Texas at ...

stars.
We also note that, despite their very similar numerical setups, the disk
evolution and accretion rate of our ‘no-feedback’ case differs somewhat from
that described in Stacy et al. (2010). Several distinctions between the sim-
ulations, however, explain this. The high-density cooling and chemistry is
updated from that used in Stacy et al. 2010 (see Chapter 3.2.2). We also
use an adaptive softening length instead of a single softening length for all
gas particles as in Stacy et al. (2010), and our criteria for sink accretion are
slightly more stringent. The main contribution to the difference, however, is
likely the stochastic nature of the sink particle dynamics. While no sink was
ejected in Stacy et al. (2010), the sink ejection and subsequent rapid velocity
of the main sink in our ‘no-feedback’ case altered the disk structure, and the
main sink would likely have grown to a higher mass otherwise. Nevertheless,
the final sink masses in both simulations were still the same to within a factor
of two.
The radiative feedback seen here is much stronger than the analytical
prediction of McKee and Tan (2008). They found that a Pop III star could
grow to over 100 M⊙ through disk accretion, as disk shadowing allowed mass
to flow onto the star even while the polar regions became ionized. Our lack
of resolution prevents this disk shadowing to be modeled properly on sub-
sink scales, and the ionizing photon emission emanating from the sink edge
is likely overestimated. McKee and Tan (2008) furthermore assumed disk
axisymmetry, which does not describe the disk in either of our test cases.
At various points in the ‘with-feedback’ simulation, the lack of axisymmetry
allowed the sink to ionize the more diffuse parts of the disk, which would not
have occured in the McKee and Tan (2008) model. Nevertheless, shielding does
91