We use the Space Telescope Imaging Spectrograph on board the Hubble Space Telescope to obtain new high-quality spectra covering the 1900 to 2360 Angstrom wavelength range for two metal-poor stars, HD 108317 and HD 128279. We derive abundances of Cu I
I, Zn II, As I, Se I, Mo II, and Cd II, which have not been detected previously in either star. Abundances derived for Ge I, Te I, Os II, and Pt I confirm those derived from lines at longer wavelengths. We also derive upper limits from the non-detection of W II, Hg II, Pb II, and Bi I. The mean [As/Fe] ratio derived from these two stars and five others in the literature is unchanged over the metallicity range -2.8 < [Fe/H] < -0.6, <[As/Fe]> = +0.28 +/- 0.14 (std. dev. = 0.36 dex). The mean [Se/Fe] ratio derived from these two stars and six others in the literature is also constant, <[Se/Fe]> = +0.16 +/- 0.09 (std. dev. = 0.26 dex). The As and Se abundances are enhanced relative to a simple extrapolation of the iron-peak abundances to higher masses, suggesting that this mass region (75 < A < 82) may be the point at which a different nucleosynthetic mechanism begins to dominate the quasi-equilibrium alpha-rich freezeout of the iron peak. <[CuII/CuI]> = +0.56 +/- 0.23 in HD 108317 and HD 128279, and we infer that lines of Cu I may not be formed in local thermodynamic equilibrium in these stars. The [Zn/Fe], [Mo/Fe], [Cd/Fe], and [Os/Fe] ratios are also derived from neutral and ionized species, and each ratio pair agrees within the mutual uncertainties, which range from 0.15 to 0.52 dex.
The temperature in the crust of an accreting neutron star, which comprises its outermost kilometer, is set by heating from nuclear reactions at large densities, neutrino cooling, and heat transport from the interior. The heated crust has been thought
to affect observable phenomena at shallower depths, such as thermonuclear bursts in the accreted envelope. Here we report that cycles of electron capture and its inverse, $beta^-$ decay, involving neutron-rich nuclei at a typical depth of about 150 m, cool the outer neutron star crust by emitting neutrinos while also thermally decoupling the surface layers from the deeper crust. This Urca mechanism has been studied in the context of white dwarfs and Type Ia supernovae, but hitherto was not considered in neutron stars, because previous models computed the crust reactions using a zero-temperature approximation and assumed that only a single nuclear species was present at any given depth. This thermal decoupling means that X-ray bursts and other surface phenomena are largely independent of the strength of deep crustal heating. The unexpectedly short recurrence times, of the order of years, observed for very energetic thermonuclear superbursts are therefore not an indicator of a hot crust, but may point instead to an unknown local heating mechanism near the neutron star surface.
