High power beam splitting I'm trying to split a high power laser (1064nm, 20W, beam diameter @ 1/e^2 intensity: 3.2mm) into two beams and couple them after some additional optics into two SMF's (single mode fiber). So far I tried a PBS (polarizing beam splitter) cube and a NPBS (non polarizing beam splitter) cube. I can split the beam but the cubes seem to destroy the mode in the reflected port at high power. I conclude this because I couple approx. 20% less efficient in the SMF. 
Does anybody know how to split(50:50 would be sufficient although polarisation dependent would be preferred) a high power beam without aggravating the mode?
Note1:
Plate beam splitters are not preferable as they usually have the same dielectric beam splitter coating as cubes and the reflected port tends to interfere with reflections from the back side of the plate
Note2:
I'm aware of Glan-Laser Calcite, Glan-Thompson and Glan-Taylor polarizers but as they are, similar to PBS/NPBS cubes, two cemented prisms and I personally suspect this layer to be the reason. I would like to know if anybody has experience with them concerning high power.
 A: Well do we know if the high powered laser is itself polarized, or not??  Using any ordinary 50-50 beam splitter cube at 20 Watts, is likely to create problems with damage to the polarizing splitter medium of such a cube.
It is possible to make beam splitter cubes, specially at well controlled wavelengths, using a Frustrated TIR cube, but they aren't cheap.  The FTIR cube operates, because of the evanescent field leakage beyond the face of a TIR (45 deg.) mirror.   The leakage field won't propagate beyond the order of a wavelength in air or vacuum, but if you bring another glass prism or medium with higher than 1.0 index, near to the first mirror face, such that there would not be TIR, if they were in optical contact, then the field can then enter the new medium, but at a reduced amplitude, because of the gap.   So the split ratio, between transmitted, and reflected rays, is determined by the small spacing between the two separated glass faces, and the wavelength of the laser, along with the index of the prisms at that wavelength.  And since your 1064 laser wavelength is well known, then the splitter prism can be designed for just that wavelength.  In principle, you can use a spacing material, with the right Temperature coefficient of expansion, so that the 50-50 split is Temperature compensated; but expect to pay well for such a prism; but people know how to make them.
And then good luck on getting 10 Watts into the fiber.   You will have to fully understand the laws of optics, and coupling lens design, to do it, and those lenses won't be cheap either .
A: It turned out that a Glan-Laser polarizer does the required job. The transmitted and reflected beam can be coupled into a SMF equally well as the original beam without splitting. I guess that the absence of any cement or dielectric layer in the Glan-Laser polarizer conserves the mode of the beam.
