Features Comphy (short for compact-physics) is a Python framework developed by TU Wien and imec for the unified simulation of BTI
for 1-dimensional MOS structures and is based on the non-radiative multi-phonon (NMP) theory.Comphy V1.0
The following list gives a brief overview of the features in Comphy V1.0,
while a detailed description can be found in the related publication
G. Rzepa et al. "Comphy - A Compact-Physics Framework for Unified Modeling of BTI", Microelectronics Reliability, 2018:

Simulation of BTI for negative (NBTI) and positive (PBTI) gate voltages

As a main feature, Comphy V1.0 offers the possibility to simulate the threshold voltage shift due to charge trapping
for arbitrary time-dependent temperature and voltage profiles. During the simulation various device parameters like the
flatband voltage, threshold voltage, surface potential, or the potential across the stack can be calculated.

Self consistent operation mode

Comphy V1.0 provides a self-consistent operation mode (SCP) that treats charges of oxide defects self-consistently in the Poisson equation which
allows for accurate simulations at high charge densities. For example this is important for accurate prediction of the degradation
at device end of life (typically 10 years).

Fast self consistent operation mode

Comphy V1.0 also offers a fast self-consistent mode (FSCP) which accounts for the reduction in electric field due to the oxide charges, but does not consider the impact on the occupancy of this reduction self-consistently.
Since this mode is already very accurate and much faster than the SCP mode, it is used as default in the Comphy framework.

Monte Carlo defect sampling

In order to simulate the combined effect of multiple defects, the defects have to be sampled according to their underlying distribution.
This can be done by a Monte Carlo approach, i.e. by drawing random samples from the underlying distribution.
On small devices with a limited number of defects, this approach reflects the native variability due to oxide defects and enables studies thereof.

Deterministic defect sampling on a grid

Alternatively, to compute the mean degradation compared to sampling a large number of defects in a Monte Carlo fashion, Comphy V1.0 offers the possibility
to sample defects on an equidistant grid.

Fast AC mode

For fast computation of high-frequency long-term periodic AC signals, Comphy V1.0 features a dedicated AC
mode to simulate the degradation of the two BTI components. This mode allows to simulate trillions of duty cycles in seconds.

Comphy V2.0
Comphy V2.0 was an internal version in which mainly the structure of the Python code was revised.
This version was never released and is therefore not available to the public. A description of the changes compared to Comphy V1.0 is therefore omitted.
Comphy V3.0
Compared to previous versions, several newly developed models and features have been added in Comphy V3.0 in order to meet the current demands for nanoscale device modeling.
The following list gives a brief overview of the most relevant new features in Comphy V3.0, while a detailed description is
provided in the related publication
D. Waldhoer et al. "Comphy v3.0 -- A Compact-Physics Framework for Modeling Charge Trapping Related Reliability Phenomena in MOS Devices":

Coupling to TCAD

Although the 1D compact device modeling approach employed by Comphy provides a fast way to assess device reliability, it clearly
has its limitations for more recent device architectures like FinFETs or gate-all-around (GAA) FETs. Since a general
treatment of such geometries is too cumbersome to deal with in a compact manner, Comphy V3.0 can be interfaced with Minimos-NT
in order to obtain critical device quantities like the carrier concentration in the channel or the surface
potential from highly accurate 3D TCAD simulations. In this workflow, the general device physics is treated within
TCAD, whereas the impact of oxide defects is calculated using the compact-physics model of Comphy.

Effective single defect decomposition (ESiD)

Due to limited measurement resolution, most degradation experiments are conducted on
large-area devices where only the collective response of a whole defect ensemble can be observed and the underlying defect parameter distribution has to be inferred
from those observations. Previous studies simply assumed Gaussian distributions for the parameters of the defects and their
mean and sigma values were obtained from an optimization loop by fitting experimental data. However, such
a procedure can lead to extraction artifacts due to enforcing the shape of the distribution.
In order to circumvent these issue, Comphy V3.0 offers a novel method of parameter extraction named Effective Single Defect Decomposition (ESiD),
which allows for a semi-automatic extraction of defect parameters from experimental measure-stress-measure traces without the aforementioned assumptions about their distribution.

Charge trapping at cryogenic temperatures

For applications at room temperature and above, the full quantum mechanical NMP model can be reasonably well
approximated by its classical version. However, the classical model predicts a complete freeze-out of charge transfer towards cryogenic temperatures, whereas charge trapping can still be observed at low temperatures. In order to efficiently model charge trapping under cryogenic conditions, e.g. for stuying BTI in emerging quantum information applications, a
Wentzel–Kramers–Brillouin (WKB) approximation to the full quantum mechanical model has been incorporated into Comphy v3.0.

Trap-assisted tunneling (TAT)

While the most pronounced effects of charge trapping at defects in the oxide are electrostatic shifts manifesting as BTI, RTN and hysteresis, the same mechanism can also result in a parasitic gate leakage current by conductance over defects.
Although several different models have been proposed in the literature to describe this phenomenon, it is usually treated separately from charge
trapping in the context of BTI. Comphy V3.0 provides a recently developed unified approach for TAT and BTI,
where both are treated on the same footing within the NMP framework.