This writeup assumes a knowledge of the basics of Michelson interferometry... While I don't think a node on that exists, I may node it as soon as finals are over and I get some copious free time
Michelson interferometry is a useful tool for analysing emission or absorption spectra of various materials. Interferometry in the far infrared region (wavelengths from 1000 um to 20 um), however, poses significant technical difficulties - few lens materials are sufficiently transparent over this entire range of wavelengths, and traditional beamsplitters tend to work very poorly due to their frequency-dependent reflection/transmission ratios.
Fortunately, there is a commercially available plastic (a form of TPX) with very desirable optical properties for FT FIR spectroscopy. It has a low absorption in the FIR region (about 3 nepers/cm), a nearly flat spectral response over the FIR region, and the added advantage that it is also transparent in the visible - making optical alignment with a He-Ne laser much easier than with polyethylene, which has slighly better properties in the FIR than TPX, but is opaque to visible light. Additionally, TPX's index of refraction (1.447 at 50 /cm and 1.450 at 350 /cm) is nearly the same in the visible as in the FIR. Finally, TPX has the considerable advantage of being relatively simple to extrude and/or injection mold. In short, TPX is about the best possible material for the job.
Now that we have a good optical substrate, we turn to the problem of the beamsplitter. Mylar has been traditionally used as a beamsplitter in FIR interferometers. However, it has the extreme disadvantage of a highly frequency-dependent efficiency - destructive interference within the film periodically reduces the efficiency to zero, meaning it becomes completely transmissive or reflective. Not at all good. The real innovation in the type of FIR interferometer I will be designing and building this summer is the use of polarizers as beamsplitters. The light from the IR source is polarized before it enters the interferometer (We will arbitrarily call the direction of the polarization "vertical," for reasons which will become apparent later). At the center, where the beamsplitter would go, we instead put a polarizer, made of gold deposited on TPX and etched in parallel lines. This polarizer is cunningly set at a 45 degree angle to the polarization of the light source, so that half of the light will pass into the reference arm and half of the light will be reflected to the sample chamber. Of course, if we simply reflected the light straight back up the arm, all the light which was reflected the first time around will be reflected again, and all the light which was transmitted the first time around will be transmitted again, with the result that the interferogram forms not at the detector, but at the light source. This is not helpful. So, instead of using a flat mirror to send the light back, we use a rooftop mirror with the edge in the vertical direction. This has the lovely effect of reversing the polarization, so that all the light gets back together at the detector, where it can do us some good. Incidentally, this also just about doubles the efficiency of the interferometer over one with a traditional beamsplitter - the traditional beamsplitter sends half the light up to the source, where it is lost.
So this could be a pretty important addition to people's data collection toolboxes, especially for examining the secondary and tertiary structure of proteins - FIR light excites some vibrational transitions, which can give some information on how proteins fold up. I don't know about the applications - I'm no biologist or chemist. All I know is I'm psyched to be building this thing. Special order from the University of Chicago, and I get to run the project.
Much information gleaned from Gunnar Stolze's 1999 Master's Thesis, "A New Polarizing Interferometer for the Far Infrared." Support from ARO grant # DAAD19-99-0067 is gratefully acknowledged.