There are two likely sites for H next to the TM impurity. Either H lies inside the vacancy or outside with approximate trigonal symmetry. The energy difference between these structures is a sensitive function of basis size and previously [#!icds97-1!#] we reported preliminary results using a modest sized basis. We have now repeated these with a much larger basis described above and find, in contrast with the previous results, that the most stable structures are generally with the H atoms outside the vacancy. However, the energy differences are not large. For example, in the case of Au-H, the energy of H anti-bonded to a Si neighbour (AB site) is lower by 0.23 eV than when H lies inside the vacancy at an AB site. Structural details are reported in [#!resende!#]. Here we concentrate on the electrical activity of the various defects.
The electrical levels are determined by the shift and splitting of the t2 manifold. In each case we assumed that a low spin state results from the addition of H. Since the wavefunction of the added electron is distributed over several shells of atoms, it is unable to completely screen the additional proton. This results in a downward shift of the t2 manifold and opens the possibility for a second acceptor level. However, the t2 manifold is split by the presence of H and this can result in an upward movement of the higher levels.
In the case of Ag, the addition of one H atom, with approximate trigonal symmetry, leads to downward shift in the donor level and an upward shift in the acceptor level. This is because the t2 level is split into an a1 (filled) level lying below a half-filled e level. The lowering of the t2 manifold is then compensated by the splitting. A second H atom results in an additional electron occupying the e level and the shift in the level is dominated by a downward shift in the manifold although this is reduced by an upward shift caused by a symmetry induced splitting of the e level. Adding a third H fills the e level and the t2 manifold is pushed below the valence band top. Thus AgH3 and AuH3 do not possess any donor levels. However, it appears that an empty level, due, we believe to the 5s and 6s levels of Ag and Au, creeps into the band gap. We place the resulting (-/0) levels of AgH3 and AuH3 at 0.13 and 0.26 eV below Ec. It may be that these levels are in reality much more shallow and thus substitutional Ag and Au can be passivated by three H atoms.
Experimentally, the levels for Ag [#!icds97-2!#] and Au
[#!svein-95!#,#!davidson-96!#] hydrogen defects are rather
similar. For the former impurity,
the H2 level
at 0.28 eV (0/+) and E2 at 0.45 eV (-/0) have been assigned to Ag-H,
while
the G2 level at 0.21 eV and G4 at
0.53 eV
have been assigned to the (0/+) and (-/0) levels of Au-H. Although the
calculated values for AgH and AuH are in fair agreement with
these assignments, we cannot rule out an assignment to AgH2 and AuH2.
However, the calculated (-/-) levels for AgH and AuH are reasonably
close to
E3 (0.09 eV) and G1 (0.19 eV) respectively, while AgH2 and AuH2 do not
appear to possess (-/-) levels.
This is probably
the result of the filling of the t2 manifold.
We thus conclude that the observed levels are due to
a single H atom complexed with the
impurity.
The H3 level at 0.38 eV has been assigned to the (0/+) level of AgH2
while the G3 level at 0.47 eV to a Au-H2 defect. Although this is
consistent with
our calculations, there are no reports of associated
(-/0) levels. We place these close to E2 and
G4. The absence of these levels places doubts on the assignments.
Both defects Ag and Au can be passivated and our calculations suggest that
AgH3 and AuH3 are candidates given our errors can be around 0.2
eV. However, we describe below a different defect which is electrically
inactive. AuH4 and AgH4 defects possess deep acceptor levels reminiscent
of H.