At the cooling end of the experiment you might consider using a can of
cool spray as sold by electronic supply houses.
Radio Shack sells a product that its add says "cools below -62F".
Once the can's cooling power is calibrated against a standard, known
bi-metallic strip, the deflection between room temp and -62F might be
all that is needed to make a decent coefficient of expansion experiment.
No extra heat, no residue from the spray, and it might even work.
Radio Shack's part number is 64-4321. It cost $6.99, which I think is
awful, but if it works, the same sort of product should be available for
less at a wholesale parts house, just a whole lot less convenient to get
to.