A photodetector (10) of the non-equilibrium kind incorporates three successively disposed sections (14, 16, 18) of like layer construction. Each of the sections (e.g. 14) contains three layers (14A, 14B, 14C) of semiconductor materials of the Cd.sub.x Hg.sub.1-x Te alloy system (CMT). The central layer (14B) of each section (14) is of narrow bandgap CMT, i.e. x=0.19 or 0.265 for absorption at 3-5 .mu.m or 8-12 .mu.m, and has very low doping (10.sup.15 cm.sup.-3) providing intrinsic conductivity. It is 1.5 .mu.m thick, less than one third of an optical absorption length. The outer layers of each section (14A, 14B) are 10 .mu.m thick and are of wide bandgap CMT, i.e. x=0.4. They have respective n and p type dopant concentrations of 3.times.10.sup.16 cm.sup.-3 providing extrinsic conductivity. Each central layer (14B) is therefore bounded by an excluding contact (14AB) and an extracting contact (14BC), which depress its carrier concentration to an extrinsic level under the action of electrical bias. This simulates cooling to low temperature. The central layers (14B to 18B) have a collective thickness (4.5 .mu.m) approaching an optical absorption length (6 .mu.m). A mirror (20) is arranged to return through the photodetector (10) radiation transmitted by it. This presents a total active region thickness six times that of an individual central region (14B) and greater than an optical absorption length. The photodetector (10) consequently has high quantum efficiency despite the deficiencies of n-type CMT material in this regard.