I found
an interesting paper recently (PDF) that discusses the design of an uncooled thermal sensor with 12µm pitch that uses mechanical resonance rather than microbolometer techniques. The authors explore a system where the resonant frequency of a micro-mechanical device varies with temperature. Sensitivity with the early prototype devices is nowhere near as good as a microbolometer - they're talking about 180mK NETD, compared with ~25mK for a good microbolometer array - but it's early days and they may be onto something. The paper includes some nice electron micrographs.