How did an unusual nickel–steel alloy reduce temperature errors in precision measurement?
Prepare alloys with different compositions
Measure length against temperature
Select a low-expansion range
Control treatment and long-term drift
A metal ruler does not keep exactly the same length when the room warms. If its marked interval expands, comparing an object with that interval changes the result even when the object's own size is stable. Precision measurement therefore needs both a reproducible length reference and a way to account for temperature. Invar helped because its length changed unusually little over a useful temperature range.
Charles-Édouard Guillaume worked on nickel–steel alloys in the 1890s in the setting of international length metrology. In his 1920 lecture, he recalled the precision and expense of the platinum–iridium metre prototypes distributed in 1889. Ordinary applications needed practical materials whose behaviour could be characterized and reproduced. The decisive result was not a new definition of the metre. It was a measured alloy anomaly that reduced a persistent source of error in physical standards, surveying equipment and instrument structures. A material property could make a measurement easier before the microscopic reason for that property was fully understood.
Mixing two metals does not generally make a property interpolate smoothly between their pure values. Guillaume's nickel–steel measurements revealed an especially strong counterexample. Near the composition now associated with Invar 36, roughly 36 percent nickel by mass with iron as the principal balance, the expansion is much smaller than a simple mixture estimate would suggest. Small additions and metallurgical treatment also matter.
Read the evidence as a family of curves. For each composition, measure length as temperature changes, then compare the slopes over the same temperature interval. Plotting those slopes against nickel content reveals the low-expansion region. A single unusually flat trace could be an instrumental accident; a reproducible composition-dependent minimum is a material pattern that must be explained. Modern manufacturer data still specify a grade, a treatment and a temperature interval. The quoted coefficient is often a mean over an interval, not the instantaneous slope at every temperature. An alloy selected for stability near room temperature can expand much more rapidly at higher temperature.
The essential experiment compares a known interval on a specimen at several temperatures. The specimen needs reproducible support and time to approach a uniform temperature. A length comparator can locate its marks against a reference; a modern instrument can locate marks with a microscope while a laser interferometer measures the carriage displacement. The thermometer must describe the specimen, not merely the surrounding air. If the two ends are at different temperatures, one number cannot fully describe the bar.
The linear expansion coefficient describes the fractional increase in length per kelvin. Over a small temperature interval where that coefficient changes little, multiply the initial length, the coefficient and the temperature change to estimate the extension. In the equation these are L with subscript zero, alpha and delta T. The approximation assumes free expansion and uniform temperature. Clamping the bar turns part of the expansion into stress. Comparing heating and cooling traces, repeating reference readings and allowing time to settle help distinguish reversible expansion from thermal lag and instrument drift.
Consider two illustrative one-metre bars with constant expansion coefficients of twelve and one micrometre per metre per kelvin. A ten-kelvin rise lengthens them by 120 and 10 micrometres respectively. These chosen rounded values show the scale of the advantage; they are not certificates for a particular steel or Invar specimen. The smaller change can still be much larger than the error allowed in precision work.
A known expansion can be corrected, but the correction has uncertainty. If the specimen's temperature is uncertain, its inferred reference-temperature length is uncertain. Multiplying reference length, the magnitude of expansion coefficient and temperature uncertainty estimates the first-order length contribution. This assumes that the expansion coefficient and reference length are otherwise known; other uncertainty sources must be evaluated separately. For the illustrative coefficients, a 0.1-kelvin temperature uncertainty contributes about 1.2 micrometres for the steel-like bar and 0.1 micrometre for the low-expansion bar. Uncertainty in the coefficient becomes more important as the measuring temperature moves farther from the reference. Low expansion reduces sensitivity and makes a given temperature uncertainty less costly.
Atoms in a solid vibrate more strongly as temperature rises. Because the forces between them are anharmonic, the mean spacing commonly grows. Invar's magnetic state also couples to volume, so the lattice-vibration contribution is not the whole story. Over the low-expansion range, magnetic and vibrational contributions can largely cancel. The atomic structure remains active; a small net expansion does not mean that thermal motion has stopped.
In 2023, Lohaus and colleagues used X-rays to investigate atomic vibrations and compared magnetic behaviour at different pressures. Thermodynamic analysis separated the contributions to expansion. They nearly cancelled in the low-pressure range studied, but the cancellation disappeared beyond it. Including the interaction between magnetic behaviour and atomic vibration improved the explanation. This later evidence shows why composition, pressure and temperature all matter when describing the low-expansion range.
Sources: [5]
A low slope of length against temperature describes one kind of stability. A permanent change of length with time describes another. Forming, machining and heat treatment leave a material with a particular stress and structural history. Relaxation can move its reference length even if a short temperature cycle shows very little expansion. Manufacturer instructions for dimensional stabilization therefore discuss stress relief and ageing as well as the nominal alloy composition.
A convincing qualification test has to return to the same reference temperature after treatment and compare the length again. If the original interval has changed, that displacement cannot be explained by a reversible coefficient alone. Recalibration and repeated measurements over time reveal the difference. The practical problem is especially clear for a surveying tape: temperature response, tension, support and handling can each change the distance assigned to its marks. Substituting low-expansion metal solves an important part of that problem, and leaves several mechanical parts to be measured and controlled.
A clock contains more than a length reference. A pendulum’s period depends on its length, so a low-expansion rod reduces one temperature effect. A balance wheel and spring respond differently: the wheel’s resistance to angular acceleration and the spring’s restoring torque both affect the rate. Even if dimensions change very little, temperature can change the material’s elasticity, making the spring softer or stiffer and changing how fast the clock runs.
Guillaume's Elinvar work addressed an unusually small temperature dependence of elasticity. Invar and Elinvar therefore name different useful anomalies. They are not interchangeable promises that every dimension and every mechanical property stay fixed. The 1920 prize recognized the service of nickel–steel anomalies to precision measurement. The historical lesson is concrete: identify which measured property enters an instrument's response before choosing a material to stabilize it. Support bars call for stable lengths; springs also call for stable elasticity.
Light can make a tiny extension readable. A laser interferometer recombines two light paths. A change in one path moves the bright and dark interference fringes, making displacements too small for an ordinary ruler accessible by counting fringes. Record displacement together with specimen temperature, a reference length and known supports, and the material’s expansion becomes measurable.
Low-expansion material reduces dimensional change with temperature, while stabilization treatments help control slower drift. Precision measurement still requires a reference length, actual sample temperature and known mechanical support. Guillaume’s nickel-steel work brought material selection into that measurement chain: accuracy depends on the readout and on the material carrying the length.
Distinguish small expansion from small temperature dependence of elasticity; composition and treatment matter.
Official attribution and award year. The observation or theoretical paper may be much earlier.
Contemporary compilation of composition-dependent expansion, heat treatment and slow length changes.
Nominal composition, temperature-dependent mean expansion and dimensional-stability treatment for a specified modern grade.
Nuclear resonant scattering and pressure measurements separate phonon and magnetic contributions; modern mechanism evidence is distinct from Guillaume’s discovery.
Temperature and expansion-coefficient uncertainties both affect the length corrected to a reference temperature.
A microscope locates scale marks while a laser interferometer measures displacement; temperature, alignment and air refractivity enter the uncertainty.
You are hereHow Invar reduces temperature errors in precision measurement
Condensed matter · Reading path
Each experiment from this article that another series also uses, followed by the articles there.