The amphoteric behaviour of gold in silicon has been a matter of
controversy for many years. Only recently [4] has it been
unequivocally established that substitutional gold introduces two
levels in the band gap. These deep acceptor and donor levels lie at
and 
eV respectively. The ground state of
the substitutional Au 
impurity has 
symmetry, tetragonally
distorted, spin state 
and configuration
[5],
with 
and 
. The
distortion is in agreement with the predictions of the vacancy model.
The defect can reorientate easily at cryogenic temperatures (T <
4 K) [6] suggesting a very small departure from 
symmetry. Surprisingly, no EPR signal has been attributed to the
defect and this is understood to arise by a rapid tunnelling between
two equivalent configurations [7] driven by a
dynamical Jahn-Teller effect. This makes the 
and
components of the g tensor very small and hence the
microwave 
transitional probability becomes very
low and the EPR spectrum difficult to observe. This model is
supported by recent theoretical work using the same methods as those
described here [8]. The theory finds that Au distorts
along the [100] direction by 
Å with small shifts of
the neighbouring Si atoms. The barrier to the reorientation is then
only 
eV and the defect might be expected to tunnel easily
between symmetric configurations.
We concentrate in this paper on the energy levels of the defect and
the effect of H upon them. The first study on the influence of H on
the gold centre was reported by Pearton and Tavendale [3]
using samples exposed to a plasma between 150 and 350 
C. A
substantial loss of the Au activity was reported which could, however,
be partially reactivated by annealing at 400 C. No effect was
found for samples annealed in hydrogen gas.
Wet-chemical etching is another process commonly used to introduce
hydrogen into the samples. This process has some advantages over the
the use of a plasma. H is introduced at room temperature, without any
damage to the wafer surface, and leads to the formation of Au-H
defects which posses distinct deep levels [9]. These defects
disappear around 
150 C and are transformed to a
passive defect if sufficient H is available. Such H could be released
from the dissociation of P-H defects. The passive defect is destroyed
by annealing beyond 200 C. This temperature is lower than
that found by Pearton and Tavendale, probably because they heated
the sample in the presence of the plasma, which provided a continuous
source of H. Pt-H 
defects appear to be more stable, forming
around 400 C .
The Au-H defects formed in wet-chemical etching have been studied by
Sveinbjörnsson et al [10]. Four new deep levels
named G1-G4 were observed in the DLTS spectra. G1 is an acceptor
level observed in n-type material with an activation energy of
0.19 eV. G2 and G3 are hole traps detected in p-type Si and located
at 
and 
eV respectively. G4 was detected in
both n- and p-type samples and showed similar characteristics to
those of the gold acceptor level. It was suggested that G1, G2 and G4
are levels of the same Au-H defect, probably Au complexed with
a single H atom, but the PA complex contained additional H atoms.
However, the DLTS study did not reveal the number of H atoms in the
defects. Little is known about the G3 defect.
A very recent study on the electron- and hole-capture kinetics of Au-H complexes in Si has been carried out using Minority Carrier Transient Spectroscopy (MCTS) [11]. This technique, combined with DLTS, allowed the determination of acceptor and donor levels using only one type of material and is a useful technique for defects with multiple levels. The main conclusion refers to the nature of the G1, G2 and G4 levels. Davidson and Evans confirmed that these levels are all produced by the same Au-H complex but G1 was shown to be a double acceptor level.