The degraded
in small-area transistors with only a few defects relaxes in
discrete steps. Each step reveals a hole emission event at the emission time
of a particular defect [115, 111]. Larger devices contain
a larger number of defects, which in combination with a nearly uniform
distribution of the activation energies
yields a log-like recovery behavior as
displayed in Fig. 7.12. As there are many different pairs of
and
within the device, their extraction from the experimental data is discussed
first.

like
in small-area transistors [115, 111], the relaxation after BTI exhibits
discrete jumps. Enlarging the area (more defects) and assuming a uniform
distribution of them adds up to a
behavior, instead.
By subtracting two recovery traces after stress times
and
, the
fraction of defects with capture time constants with
is
determined first [116], which is shown in Fig. 7.13. By dividing the difference
trace into intervals
, the fraction of defects having
and
is obtained.

and
yields the fraction of defects with capture time
constants with
These ranges of capture time constants of
certain defects are depicted as function of
. The contour lines below
the three graphs emphasize the amount of defects contributing to
.
For NBTI with an
of
, the characteristics of
are
not changed with increasing
, despite some shift along the positive
-axis. The maximum
values for all
-ranges are obtained
for small values of
. This implies fast relaxation. On the contrary,
PBTI (
) yields a larger degradation and additionally moves the
characteristics of
towards increasing
. For the largest available
,
which covers time constants between
and
, the maximum of
is moved away from the minimum
. This maximum marks the
beginning of the change of emission time constants
depicted in Fig. 7.15
and is even more pronounced for
.
To be able to describe the frequency of occurrence of capture time constants
and emission time constants
properly, a large set of long recovery traces
with varying
is needed. The experiments performed cover
from
up to
and
intervals between
and
. This
allows for an extraction of the time constants as exemplarily depicted in
Fig. 7.14.
It is now possible to explain the above mentioned effect with the varying
oxide electric field on the basis of Fig. 7.15, where the fraction of
due to
defects with
and
is plotted as smoothed surface over
and
.
For NBTI with an
of
the surface shows two peaks.
One peak covers
and
smaller than
, while the other more
pronounced one clearly illustrates that the largest part of the degradation was
due to defects with
larger than
, which is highlighted by the
contour lines below the graph. When comparing the different
for
PBTI for
covering time constants between
and
, the
peak of
mainly consists of
, while it is widened for
towards smaller
. This supports the hypothesis of decreased
for higher
after PBTI stress, which appears as faster long-term
recovery.

and
is depicted for three different oxide electric fields. The
contour lines below the graphs highlight the biggest changes of
. Both
surface and contour lines are smoothed for a better visualization. It is shown
that the oxide electric field is related to the magnitude of
. Increasing
yields a shift of the peak towards smaller
, which corresponds to our
monitored increased recovery at larger
. Note that only for
a
full set of
and
is available and therefore the map has to be truncated
in order to be comparable with the case
.