In the original model paper (https://doi.org/10.1063/1.166311), the diffusion coefficient is expressed in $cm^2/s$, while $dx$ is given in centimeters and $dt$ in milliseconds.
In the current implementation, however, the units are not specified and the diffusion coefficient is simply set to $1.0$.
Omitting units in this case is not trivial: as an example, using the parameters from Figure 5 of the paper
-
$dx = 0.0262\text{cm} $,
-
$dt = 0.17\text{ms} $,
- $D = 1\text{cm}^2/\text{s} $
leads to a violation of the CFL condition if units are ignored, resulting in non-convergent simulations.
I don’t think users should be expected to look up the original paper to determine the correct units. To avoid hidden pitfalls, I suggest the following:
- Set the diffusion coefficient to $0.1 \text{mm}^2/\text{ms}$.
- Explicitly specify that $dx$ is in millimeters and $dt$ in milliseconds.
With these conventions, users can safely omit unit conversions without risking instability.
I am not sure if the Fenton–Karma model is the only case where units cannot safely be ignored, but might also be worth discussing a general policy for handling and documenting physical units, especially if external contributors are expected to add new models.
In the original model paper (https://doi.org/10.1063/1.166311), the diffusion coefficient is expressed in$cm^2/s$ , while $dx$ is given in centimeters and $dt$ in milliseconds.$1.0$ .
In the current implementation, however, the units are not specified and the diffusion coefficient is simply set to
Omitting units in this case is not trivial: as an example, using the parameters from Figure 5 of the paper
leads to a violation of the CFL condition if units are ignored, resulting in non-convergent simulations.
I don’t think users should be expected to look up the original paper to determine the correct units. To avoid hidden pitfalls, I suggest the following:
With these conventions, users can safely omit unit conversions without risking instability.
I am not sure if the Fenton–Karma model is the only case where units cannot safely be ignored, but might also be worth discussing a general policy for handling and documenting physical units, especially if external contributors are expected to add new models.