Temperature mapping from molecular absorption tomography

Simulations were performed to demonstrate temperature tomography using circular 32 pixel-radius temperature and (randomised but smoothed) concentration phantoms. The simulations were run on a sensor head configuration constituting 36 sets of 32 parallel beams with equal angular spacing over 180°. For each beam, data from the HITRAN 2008 database was used to simulate direct absorption measurements of a pair of H₂O transitions at 5263.7 and 6476.1 cm^-1 at discrete frequencies (42 in total) near the line centres. The phantom temperature ranged from 1600 - 2000 K, the concentration of H₂O, from 0.92% to 7.86% and pressure was uniform for each simulation, though it was varied between simulations from 1 to 25 bar. At each frequency, Landweber iteration was used to reconstruct local attenuation coefficient images from line integrals, and the Levenberg-Maquardt-Fletcher algorithm was used to fit Voigt profiles to the data at each pixel, from which the temperature was calculated. Comparison with phantoms yielded a range of root mean square errors from 16.6 K at 1 bar to 50.0 K at 25 bar.