Invar was the first successful attempt at making a metal alloy that exhibits a nearly zero coefficient of thermal expansion (CTE). One of its first uses was in precision mechanical clocks, most notably in the very successful Shortt master clock which was only replaced by quartz technology. Today Invar is still used in precision clocks, and for making scientific instruments. It is also used in color CRT shadow masks, and as one half of the bimetal strip in mechanical thermostats. Surveying tapes used to use Invar but laser rangefinding rendered them obsolete. VHF/UHF/microwave components sometimes use invar, such as in duplexers and other resonant cavities.
Note that Invar maintains low CTE over "room" temperature and then some, but not at wild extremes such as liquid air or red-hot. For example, one source implied that annealed Invar should never be heated above 100 C.
A related metal is Elinvar, that maintains a relatively constant elasticity (springiness) over temperature, and was used in watch hairsprings.
Invar is basically steel with 36% nickel, and other smaller amounts of other elements for machinability (and because a pure mixture is hard to obtain!)
I have heard that Invar should not be roughly handled, such as dropped. Apparently this will disturb the grain structure and/or cause it to become magnetized to a small extent by the earth's magnetic field.
There have been specific specimens of Invar that exhibited zero or even negative thermal expansion. Trouble is, you can't reliably make it that way, it's more of an accident when it happens.
There are three common types of Invar. Invar 36 is good stuff but a "first class sumbitch" to machine; about like stringy stainless steel. CTE is about 1.6*10^-6/K. There is a variant called "FM" (Free Machining) that has selenium and more manganese alloyed in, and is nice to work with; if it matters, though, FM has twice the carbon content, which apparently contributes to long-term instability. Then there is Super Invar which has some Cobalt added, has about 3 to 6 times lower thermal expansion, around .63*10^-6/K, but once again is a sumbitch to work with. There are some reports that Super Invar is not as dimensionally stable as ordinary Invar, and will spontaneously change size (very slightly, of course!) more often than Invar. This problem can be reduced by proper annealing, I believe. You can get lucky and buy Invar 36 and get a piece that is as good as Super Invar; the specs are more of a worst-case value, not a guarantee.
There is a fellow at Carpenter Technology Corporation, Les Harner, called "Mr. Invar"
who recommends the following annealing procedure to maximize stability:
1. Anneal at 1500 F for 1 hour, then water quench.
2. Stress relieve at 600 F for 1 hour, then slow air cool.
3. Age at 200 F for 24 hours, then slow air cool.
Another treatment I've read (Berthold et. al.) used the same temperatures
except the times were 30 min, 1 hour, and 48 hours.
Update: Another, rare, type of Invar was recently developed by NASA/JPL. Dubbed HP (High Purity) Invar 36, it has much improved CTE and dimensional stability over time. They made one batch for a spacecraft camera (Cassini) using powder metallurgy. Very pure elemental iron and nickel, in powder form, were weighed, mixed, pressed into a mold, and sintered in a controlled atmosphere. This was then processed and cut; 1/2 was extruded and 1/2 was hot hammered. I doubt if you can just go out and buy this stuff. The mixing/sintering was done by Spang Specialty Metals, in Butler PA. The post-processing was done by Scientific Alloy Inc, in Westerly RI. The exceptional properties were attributed to the high purity of the result, especially the low carbon content (under .01%). See the NASA Technical Support Package "Temporally and Thermally Stable Iron/Nickel Alloy" for the August 1995 issue of NASA Tech Briefs.
"Specialty metals since 1981" "Rare alloys & non-standard sizes". Invar 36, Invar 36 FM, Super Invar, among others.
Substitutes for Invar include certain glasses and ceramics. I was told that Coors may make a machinable ceramic with low CTE. Machine it, then cook it, then use it. I have not verified this.
Another suggestion was Macor from Corning Glass Works, CTE = 9.4*10^-6/K. (I don't think you need to bake Macor after machining and before use.)
For really low CTE, Zerodur from Schott Glass Technologies, Duryea, PA CTE = 0 +/- 0.10*10^-6/K 273K < T < 323K. This is a glass ceramic.
Quartz may also be good depending on the application, such as Corning's 7940, or Heraeus-Schott's Homosil.
Owens-Illinois Cer-Vit C-101 (glass ceramic)
Corning ULE 7971 (Titanium Silicate).
Also Graphite-epoxy and graphite-cyanate, not the ordinary stuff but special fiber/resin ratios (try Composite Optics, Inc.). WATCH THE OUT OF PLANE CTE! You can buy laminate with 0 CTE in both in-plane directions, but out of plane it always has the CTE of the resin. This is so big that a single thickness can dominate the CTE of an assembly. Careful joint design can eliminate any thickness in the thermal expansion path.
Articles:
"Dimensional Instability of Invars", Applied Optics, letter to the editor, V 23 n. 20 (15 Oct 1984), page 3500-3502.
"Dimensional Stability of Fused Silica, Invar, and Several Ultra-Low Thermal Expansion Materials" Berthold, et. al., Metrologia, V 13, 9-16 (1977), pages 9-16.
"Pendulum Rods" by Robert Matthys, Horological Sciences Newsletter issue 1993-5, 27 Dec 1993, pages 8-11. Pages 10-11 are a reproduction of the Carpenter Steel data sheet for their free-cutting Invar 36. This is the newsletter of the Horological Science Chapter #161 of the National Association of Watch and Clock Collectors (NAWCC).