What is the difference between using a cube beamsplitter and a plate (mirror-type) beamsplitter?

A cube beamsplitter is essentially two identical right angle prisms where the hypotenuse of one of the prisms is coated with a semi-reflective coating before the two prisms are cemented together to form a cube. Cube beamsplitters offer several advantages over plate beamsplitters. They are easier to mount, since the 45 degree reflecting surface is within the cube and it can be mounted flush. Since a plate beamsplitter is mounted from the edges of a thin glass, it cannot handle as much deformation from mounting stresses without affecting the performance. The reflective coating is within the cube, so it is naturally more durable and has a longer lifetime than for a plate beamsplitter. Also there are no ghost images, since most of the reflections within a cube are reflected back in the direction they came and thus do not effect the output beams/images. However because converging and diverging beams add considerable image quality errors, cube beamsplitters should only be used in collimated beams. Additionally, since cube beamsplitters are heavier, they should be considered carefully for weight sensitive applications. Just like all beamsplitters, (the prism with) the reflective coated surface should face the input beam or source. Our cube beamsplitters are available with dielectric coatings as dichroic cubes for 50% reflection at 550nm or as polarizing cubes with 92% reflection for S-polarization at 632.8nm. All other polished outside surfaces are anti-reflection coated to minimize ghost images.