We investigated the Au-H complexes when H lies both at
anti-bonding sites outside the nearest neighbour shell of Si atoms to
the impurity, and inside the shell. Similar sized clusters were used
as in the Au defect discussed above.
For the 
Au-H defect, the H atom lies near a bond centred
site as illustrated in Fig. 2. This configuration is more stable than
the AB one by 
eV. The Au-H and H-Si lengths are 1.78 and
1.57 Å respectively with an Au-H-Si angle of 118 
. The
difference in energy between a configuration where the H atom is
constrained to lie along [100] with 
symmetry is 0.15 eV and
this represents an estimate of the energy barrier to reorientation.
The 
level (Fig. 1) now contains four electrons and is
split and displaced downwards in the gap by the proton. Inevitably,
this will decrease the donor and deepen the acceptor levels -- as
observed for G2 and G4 --, and encourages the possibility of a second
acceptor level. The Slater method locates the acceptor (-/0) level
at 0.63 eV which is the same as that of Au. This agrees with the
assignment of the mid-gap G4 level. A second acceptor (-2/-) level
is found at 
eV and about 0.17 eV deeper that the
(-2/-) G1 level. The donor (0/+) level at 
eV is
slightly lower than that of Au at 
eV and is 0.17 eV above
G2. Given that the levels are generally deeper than about 0.1-0.2
eV, the results support the finding that the G1, G2, G4 levels all
arise from the same defect which contains just one H atom. In
addition, they support the claim that the G3 level at 
eV
cannot arise from this defect.
Table 2: Bond lengths for the different complexes. The bottom row of
the table gives the distance (d) of the Au
atom from the centre of the cluster. Bond lengths in Å.
Adding additional H atoms to the defect will fill the level and
shift it downwards (Fig. 1). The level becomes filled for
Au-H 
. The H atoms prefer to be inside the vacancy as the energy
of this structure is 0.8 eV below that of the outside configuration.
Two H atoms are strongly bonded to the Si radicals with lengths of
1.50 Å whereas the third is 1.67 Å from a third radical and
1.64 Å from Au. The gold atom is displaced off-site by 
Å along [100]. The defect has and the filled
Kohn-Sham levels now are low lying and there are no empty gap levels.
Thus the defect cannot act as an acceptor. The donor level as
calculated by Slater's method is placed at 
eV. Since our
method leeds to deeper levels by 
0.2 eV than observed, the
defect may identify Au-H with the passive PA complex. This
suggests that the Au-H defect is passivated by a direct combination
with a mobile hydrogen molecule.
The Au-H 
and Au-H 
defects are electrically and
magnetically active but will not be discussed here. It seems unlikely
that the G3 level is one of these defects.
Figure 1: One-electron Kohn-Sham eigenvalues in the vicinity of the gap
for the Au 
-H complexes in their relaxed
configurations, with 
. Arrows indicate filled spin states,
boxes indicate empty one-electron states (these are only shown in region of the gap).
In conclusion, our calculations have found:
has no acceptor levels and a shallow donor level. We
identify it with the passive defect first detected by Pearton and
Tavendale; and Au-H are electrically and magnetically active
defects with deep levels.
Figure 2: Relaxed structure of the Au -H in Si, n
=1,2 and 3. Box edges mark the {100} directions. Note the
increasing (Au-H-Si) with increasing number of H atoms.