Computed-tomography imaging spectropolarimeter (CTISP)

Abstract

A complete Stokes imaging spectropolarimeter has been developed based on the principles of computed-tomography, spectrometry and polarimetry. The Computed-Tomography Imaging SpectroPolarimeter (CTISP) is a polarization extension to the Computed Tomography Imaging Spectrometer (CTIS)¹. Imaging spectrometers estimate the object cube (x,y, λ), whose smallest subdivision is a voxel, while Stokes imaging spectrometers estimate four Stokes object cubes (x,y, Sp(λ); p = 0,1,2,3), one for each Stokes parameter. CTISP uses a two-dimensional disperser to diffract the image in the field stop into a 5-by-5 array of diffraction orders. As in computed tomography, each focal plane array (FPA) pixel effectively integrates a different path through the object cube, and when all pixels are recorded, a significant portion of the object cube's information is obtained. The frequency space representation of the object cube, however, indicates that two conical regions of information are not recorded, thereby limiting the reconstruction accuracy. CTISP scans only in the polarization domain (not spectral or spatial domains), acquiring four FPA frames, one behind each of the four polarization analyzers. Currently, CTISP's resolution is 33 by 33 spatially over a 3.5 degrees full angle field of view with 16 spectral bands of 20nm width covering 440nm-740nm. CTISP acquisition is modeled using the linear imaging equation gₐ=Hₐfₐ+ξₐ, which is inverted using the iterative expectation-maximization algorithm to solve for fₐ, the object cube as seen through analyzer a. The recorded diffraction images gₐ and the empirically determined calibration matrices Hₐ, are each acquired using analyzer a. The nth voxel reconstruction result is extracted from each of the four fₐ vectors to form a four element vector f(n) which is then multiplied by the inverse of the voxel characteristic matrix W(n) to obtain the estimate of the Stokes vector S(n). W(n) is derived from the four Hₐ matrices. A fully computer-controlled calibration facility and a suite of programs are used to calibrate CTISP. CTISP was validated using synthetically generated and real objects. Spectral agreement is consistent with CTIS, while Stokes parameter polarization errors were typically 0.04-0.07 for this instrument. Errors are most significant at the spectral limits of CTISP. An object dependent correction reduces these errors to below one percent.