Chemistry Reference
In-Depth Information
According to Chazalviel
et al.,
the oxide is always being formed with high quality, but
can become low quality under a high field. The essential sequence at a constant poten-
tial is as follows: (i) fast chemical dissolution of the top low-quality oxide until the
high-quality oxide is exposed; (ii) slower dissolution of the high-quality oxide leading
to oxide thinning and increasing the field inside the oxide; (iii) breakdown of the oxide
under the high field which turns the oxide into the low-quality form accompanying a
large current flow; and (iv) fast dissolution of the low-quality oxide occurs along with
the growth of the high-quality oxide underneath the low-quality oxide which results in
a decrease of current and brings the system back to step (i).
Carstensen
et al
.
1127,1136,1142
proposed a detailed model based on measurements of
current transients at different time of a current oscillation period. The macroscopic
oscillation is postulated to result from the synchronization of the microscopic oscilla-
tors associated with the growth and dissolution of the oxide. Thus, this model is similar
to that proposed by Ozanam
et al
.
949,951
in its physical nature, that is, the anodic oxide
film on the silicon surface is not uniform laterally but consists of small domains with
different thicknesses. At any given time, some of these domains with a thin oxide act
as the active channel to conduct current resulting in increase of the oxide thickness to
reach a maximum thickness, growth then stopping; some other domains with a thick
oxide do not conduct current until the oxide is thinned by the dissolution to a minimum
value when a breakthrough current occurs. These local domains act as microscopic
oscillators. Macroscopic oscillation occurs when the events associated with these
microscopic domains on a macroscopic scale are synchronized. On the other hand, no
oscillation occurs when they are not synchronized. In order for this model to agree with
the experimental data, a number of assumptions were made: (1) The size of the domains
is on the same order as the thickness of the oxide. (2) There are two types of oxides
on a macroscopic scale, one being relatively homogeneous and existing at the minimum
of the current oscillation and the other being an oscillating oxide. The dissolution rates
of the two oxides are constant in the thickness direction with the rate of the oscillation
oxide being about 1.7 times that of the homogeneous oxide. (3) The growth of oxide
at the local domain is a noncontinuous event regulated by the field; oxide will start to
grow when the maximum field, is obtained, where
V
is the applied poten-
tial and is the minimum thickness; the growth will stop when the minimum field,
is reached.
In all of the models described above, the processes involved in the growth and
dissolution of the anodic oxide film are recognized to be responsible for the oscillation.
The physical and chemical properties of anodic oxides change with time during oscil-
lation, thus resulting in the variation in the oxide growth rate and the chemical etch
rate. The models differ from each other only as to the cause of the periodic change of
the oxide properties. These models can be grouped according to the assumptions made
on the physical origin of the oscillation: those attributing it to the microdomains asso-
ciated with the lateral inhomogeneity in the oxide film, and those attributing it to a
sudden change in the properties of the oxide film during its growth. For the
microdomain models, the physical basis for the occurrence of these domains is not
clear. A related problem with the microdomain model is its implied micro surface rough-
ness. However, the silicon surface after the dissolution in the potential range, in which
oscillation occurs, is most smooth.
38
As far as the models based on a sudden change of
