[CIG-SEISMO] PML in SPECFEM3D

[Dimitri Komatitsch]

Hi Ivan,

In the current implementation we have Cartesian PMLs, i.e. the PML 
layers must be aligned with (parallel to) the X, Y or Z axis.
Curvilinear PMLs (e.g. as needed for a cylinder) are not implemented.

Best regards,
Dimitri.

On 13/08/2016 17:13, Ivan Karpov wrote:
> Dear all,
>
> My name is Ivan Karpov, I am a PhD student at NTNU, Norway. I am trying
> to model seismic waves in media pierced by thin boreholes. The most
> convenient model in my work is bounded by a cylinder (i.e. surfaces
> bottom and zmax are flat and horizontal, and surfaces xmin, ymin, xmax,
> ymax are parts of the cylinder and are merged with no corners between
> them). In order to suppress the waves reflected off the model limits, I
> used Stacey boundary conditions. But the absorption is not sufficient
> enough, so I need to put PMLs there. As far as x- and y- layers smoothly
> continue each other, I do not have PMLs elements belonging to both x-
> and y-layers (i.e. I do not have elements with flags 4 and 7). In this
> case, specfem prompts errors of the following kind:
>
> Error in slice            0 : number of PML elements        17750  but
> only        16598  flags set
> STOP Error C-PML array has invalid number of PML flags set
>
> Does it mean that I can use only cuboid models (or models with corners)?
>  Is it possible to overcome this issue?
>
> With kind regards,
> Ivan Karpov
>

-- 
Dimitri Komatitsch
CNRS Research Director (DR CNRS), Laboratory of Mechanics and Acoustics,
UPR 7051, Marseille, France    http://komatitsch.free.fr