​ Hi YJ,

Thank you for sharing your file. zar is unfortunately not supported by our platform so we usually send zip files. 

Just looking at the editor, we know that the geometric rays are likely to be different from POP as the beam is propagated over a long distance where the beam will diffract with POP but will stay unchanged with rays. So I agree that POP is a good choice in that case.

I then looked at the POP ZBF. One important thing to note since your beam is polarized is that when we use Polarization, 2 beam arrays will be propagated, one each for the X and Y polarizations of the beam. 
The phase on surface 2 is clearly under sampled:

As a first step, I can use rays to go through the lenses. I have ticked “Use Rays to Propagate to Next Surface” on all the lenses but not on the long distance.

• “Use Rays to Propagate to Next Surface” is selected for surfaces 2,3,4,7,8,9. It is quite a useful option here since each individual lens adds aberrations where the phase will be hard to sample, but are then compensated by the following lens.
• On surface 6, I increased the sampling because it was under sampled on surface 10:

  

   

• On surface 11, I also increased the sampling:

  

   

It gives me an enough sampled beam. The diffraction is only considered from surface 1 to before the 1st lens, on the long distance and from after the last lens to the image plane.

So this result doesn’t consider the diffraction that happens at the lenses but considers the main diffraction which is on the long distance. I have attached my file.
The process is quite tedious since I am checking the phase at each surface and then adjusting the sampling manually until I see that it is visually properly sampled. Let me know your thoughts.

Just as an extra note here, we can also tell that how much diffraction happens in the long distance. POP tells me that the beam size changes from a radius of 5mm to a radius of 15mm (I checked visually). The ray approach is not the right one here since the beam stays constant with rays.