About the PML

[guoketing]

Dear Alan,

I'm sorry to disturb you again. After reading the source code of finite difference wave equation discretization. I have some questions about PML, I don't know why you set the auxiliary wavefield like this. I wonder the paper to support the auxiliary wavefield and the PML. Thank you very much!!

-Keting

[ar4]

Dear Keting, I am glad that you were able to find and understand the code for the propagators. I believe I used the PML implementation that I found in PySIT. I remember that at the time when I implemented it I also struggled to understand where this came from, but it seemed to work and I think I remember that it was simpler than other published PML methods that I found, so I decided to use it. I have just verified that my 2D implementation does indeed seem to be equivalent to the 2D implementation in PySIT: https://github.com/pysit/pysit/blob/master/pysit/solvers/constant_density_acoustic/time/scalar/constant_density_acoustic_time_scalar_2D_4.h . Do you think that a different formulation might be better?

…

-Alan

[guoketing]

Dear Alan,

Thank you for your reply, I would see the PML method in the PySIT, and I will try to implement other PML methods in published papers, if they have better results compare to this method. I'll connect to you to improve the deepwave.

-Keing

[ar4]

The PML that was used in the previous versions of Deepwave was, as I mentioned, based on the one in PySIT. The author of PySIT discusses the origin of this PML here: pysit/pysit#34

I changed the PML in v0.0.10 of Deepwave to use the one proposed by Pasalic and McGarry, which I find to work better at near-grazing incidence (which is especially important if you do not have a free surface).