In 1896 Guillaume found that iron with about 36% nickel expands roughly a tenth as much as the
steel it is made from. He took the 1920 Nobel Prize in Physics for it, and the citation is about
metrology rather than metallurgy: before Invar, the length of a metre bar depended on the room.
It works by cancellation. The lattice expands with temperature like any lattice; the alloy's
spontaneous magnetostriction shrinks over the same range by almost exactly as much. Two large
effects with opposite signs. Which is why the property dies at the Curie point — above
550 K there is no magnetisation left to cancel with, and Invar expands like the steel it is.
Every number below comes from js/invar.js, which
has one fitted parameter, set at room temperature. Everything else is a prediction, and the hero
above is the same model driving two rulings against each other.