wiki上查到一些，搜索Polarization Beam Splitter就可以，不过是英文的
相关部分如下：
Beam-splitting polarizers split the incident beam into two beams of differing polarization. For an ideal polarizing beamsplitter these would be fully polarized, with orthogonal polarizations. For many common beam-splitting polarizers, however, only one of the two output beams is fully polarized. The other contains a mixture of polarization states.
Unlike absorptive polarizers, beam splitting polarizers do not need to absorb and dissipate the energy of the rejected polarization state, and so they are more suitable for use with high intensity beams such as laser light. True polarizing beamsplitters are also useful where the two polarization components are to be analyzed or used simultaneously.
[edit] Polarization by reflection
A stack of plates at Brewster's angle to a beam reflects off a fraction of the s-polarized light at each surface, leaving a p-polarized beam. Full polarization at Brewster's angle requires many more plates than shown.When light reflects at an angle from an interface between two transparent materials, the reflectivity is different for light polarized in the plane of incidence and light polarized perpendicular to it. Light polarized in the plane is said to be p-polarized, while that polarized perpendicular to it is s-polarized. At a special angle known as Brewster's angle, no p-polarized light is reflected from the surface, thus all reflected light must be s-polarized, with an electric field perpendicular to the plane of incidence.
A simple polarizer can be made by tilting a stack of glass plates at Brewster's angle to the beam. Some of the s-polarized light is reflected from each surface of each plate. For a stack of plates, each reflection depletes the incident beam of s-polarized light, leaving a greater fraction of p-polarized light in the transmitted beam at each stage. For visible light in air and typical glass, Brewster's angle is about 57°, and about 16% of the s-polarized light present in the beam is reflected for each air-to-glass or glass-to-air transition. It takes many plates to achieve even mediocre polarization of the transmitted beam with this approach. For a stack of 10 plates (20 reflections), about 3% (= (1-0.16)20) of the s-polarized light is transmitted. The reflected beam, while fully polarized, is spread out and may not be very useful.
A more useful polarized beam can be obtained by tilting the pile of plates at a steeper angle to the incident beam. Counterintuitively, using incident angles greater than Brewster's angle yields a higher degree of polarization of the transmitted beam, at the expense of decreased overall transmission. For angles of incidence steeper than 80° the polarization of the transmitted beam can approach 100% with as few as four plates, although the transmitted intensity is very low in this case[3]. Adding more plates and reducing the angle allows a better compromise between transmission and polarization to be achieved.
[edit] Birefringent polarizers
Other polarizers exploit the birefringent properties of crystals such as quartz and calcite. In these crystals, a beam of unpolarized light incident on their surface is split by refraction into two rays. Snell's law holds for one of these rays, the ordinary or o-ray, but not for the other, the extraordinary or e-ray. In general the two rays will be in different polarization states, though not in linear polarization states except for certain propagation directions relative to the crystal axis. The two rays also experience differing refractive indices in the crystal.
A Nicol prismA Nicol prism was an early type of birefringent polarizer, that consists of a crystal of calcite which has been split and rejoined with Canada balsam. The crystal is cut such that the o- and e-rays are in orthogonal linear polarization states. Total internal reflection of the o-ray occurs at the balsam interface, since it experiences a larger refractive index in calcite than in the balsam, and the ray is deflected to the side of the crystal. The e-ray, which sees a smaller refractive index in the calcite, is transmitted through the interface without deflection. Nicol prisms produce a very high purity of polarized light, and were extensively used in microscopy, though in modern use they have been mostly replaced with alternatives such as the Glan-Thompson prism, Glan-Foucault prism, and Glan-Taylor prism. These prisms are not true polarizing beamsplitters since only the transmitted beam is fully polarized.
A Wollaston prismA Wollaston prism is another birefringent polarizer consisting of two triangular calcite prisms with orthogonal crystal axes that are cemented together. At the internal interface, an unpolarized beam splits into two linearly polarized rays which leave the prism at a divergence angle of 15°-45°. The Rochon and Sénarmont prisms are similar, but use different optical axis orientations in the two prisms. The Sénarmont prism is air spaced, unlike the Wollaston and Rochon prisms. These prisms truly split the beam into two fully polarized beams with perpendicular polarizations. The Nomarski prism is a variant of the Wollaston prism, which is widely used in differential interference contrast microscopy
大概看了一眼，好像就是用了反射折射为偏振光的原理。。。
温馨提示：答案为网友推荐，仅供参考