Specifying a dimensional stability test for Invar parts

By Invar Machining editorial team· Published September 26, 2026 · How we write these guides

An Invar thermal cycling test is an acceptance test for dimensional stability, not a test of expansion coefficient. The part is measured at 20 °C, cycled a set number of times between two temperatures that bracket its service range, then measured again at 20 °C on the same CMM with the same program. It passes when the change between the two measurements stays inside a limit you set, usually a small fraction of the feature tolerance.

What the test checks and what it does not

A machined Invar 36 part can move after it leaves the shop. Cutting leaves residual stress and a thin cold worked layer. Clamping leaves strain. A stress relief between roughing and finishing removes most of it, but not all, and a skipped or short cycle leaves more. At room temperature that stress relaxes slowly, over weeks. Warming and cooling the part speeds the process up, so a few cycles between, say, 0 and 50 °C reveal in a day what the part would otherwise do in a year on the bench.

That is what the test checks: whether the machining sequence, described in stress relief in the Invar machining sequence, actually produced a stable part. Dimensional stability, in the sense used in the glossary, means the part returns to the same size and shape when the temperature returns. It is a test of the process applied to this part, which is why a passed coupon from a previous lot does not transfer.

It does not measure the coefficient of thermal expansion. CTE is measured on a small coupon in a dilatometer to ASTM E228 or by TMA to ASTM E831, and the result belongs on the material certificate, not on the part inspection report. If your drawing needs both, write both. The two tests answer different questions: the dilatometer says how much the Invar alloy grows per degree, the cycling test says whether the part comes back. The Invar 36 CTE versus temperature guide covers the first. This one covers the second.

Choosing the two temperatures

The range should bracket the service range with some margin and stay well below any temperature that would change the part. Two limits matter.

The upper limit must stay far below the stress relief temperature. A typical Invar 36 stress relief runs at 315 to 345 °C. Cycling to 150 °C would not anneal the part, but it would start to relax stress in a way the service environment never will. Most acceptance tests stay at or below 80 °C. For Super Invar keep the upper limit under about 90 °C, because its expansion rises sharply above that band and the numbers stop meaning what you think.

The lower limit should reach below the coldest service or shipping temperature. For a lab instrument that is rarely below 0 °C. For an airborne or space structure the drawing may call for a cold soak, and that becomes a different, larger test with its own fixture and chamber.

The table below lists typical ranges. Your service conditions govern.

Application	Service range, typical	Cycling range, typical
Lab optical bench or metrology frame	20 ± 5 °C	0 to 50 °C
Semiconductor tool frame in a cleanroom	22 ± 0.5 °C	5 to 45 °C
Composite layup tool	20 to 180 °C	Cycle to cure temperature, separate spec
Shipped part, uncontrolled transit	-20 to 60 °C	-20 to 60 °C

Typical values for planning. The drawing or the purchase order sets the actual range.

How many cycles, and how fast

Most of the movement happens in the first cycle. If a part is going to shift, it shifts when stress first relaxes, and later cycles add smaller amounts. That is why a useful test measures the trend, not only the end point. Three cycles are the minimum that shows a trend. Five is the common choice. Ten is used when the drift limit is a micron or less.

Dwell and ramp matter as much as the count. Invar has low thermal conductivity, roughly a quarter that of carbon steel, so a fast ramp sets up a temperature gradient across a thick section, and that gradient is itself a stress. The dwell has to be long enough for the whole part to reach the chamber temperature. A 5 mm sheet part gets there in minutes. A 40 mm plate takes an hour or more.

Parameter	Typical value	Why
Number of cycles	3 to 10, usually 5	Enough to see whether shift per cycle is decreasing
Ramp rate	1 to 3 °C/min	Limits gradients in a low conductivity alloy
Dwell at each limit	30 min for thin parts, 1 to 2 h above 25 mm thick	Whole section reaches temperature
Support in chamber	Three point, unclamped	No fixture strain added during the test

Typical values. State the ones you want on the drawing, because a shop that guesses will guess short.

Measuring before and after

The measurement is where most thermal cycling tests go wrong, because the drift you are looking for is often smaller than the day to day repeatability of a CMM.

Measure at 20 °C. ISO 1 sets 20 °C as the reference temperature for all geometrical product specifications, and the CMM scales, the fixture and the granite all expand more than the Invar part does. Soak the part in the metrology room before both measurements. Four hours is typical for a small part, 12 to 24 hours for a plate over 20 mm thick. Record the room temperature and the part temperature at measurement time.

Use the same machine, the same program, the same probe and the same datum scheme for both runs. Support the part the same way, ideally on three points at the positions marked on the drawing. A part measured on a flat granite before the test and on three points after it will show a flatness change that is fixture, not Invar.

Know your repeatability. Run the program twice before the cycling starts. The spread between those two runs is the noise floor. If the drift limit is 0.003 mm and the CMM repeats to 0.002 mm, the test cannot tell pass from fail. A common rule is that measurement uncertainty should be no more than a quarter of the limit. The Invar CMM inspection report guide covers what a usable report contains.

What drift is acceptable

There is no standard number. The limit comes from your error budget, and it is usually written as a fraction of the feature tolerance or as an absolute value on the features that matter. Typical practice looks like this.

Feature	Tolerance on drawing, typical	Drift limit, typical
Bore position on a mount, ±0.010 mm	0.020 mm zone	0.003 to 0.005 mm
Flatness of a mounting face	0.005 mm	0.001 to 0.002 mm change
Length between reference faces, 300 mm	±0.010 mm	0.002 to 0.003 mm
Parallelism of two faces	0.010 mm	0.002 to 0.003 mm change

Typical values for a precision optics or metrology part. Your budget governs.

Two other things belong in the acceptance criterion. First, the trend: if the shift between cycle 4 and cycle 5 is as large as the shift in cycle 1, the part is still moving and should not pass even if the total is inside the limit. Second, the direction: a uniform shrink of every dimension points to stress relaxation, while a twist or a change in flatness alone points to a clamping or support problem.

The Invar machining tolerances guide explains what shops hold in the first place, which helps you set a drift limit that a good part can pass.

Worked example, hypothetical

Take a hypothetical Invar 36 reference plate, 250 × 150 × 20 mm, with four bores on a 200 × 100 mm pattern at position ±0.010 mm and a mounting face flat to 0.005 mm. It sits in a lab held at 20 ± 3 °C.

A reasonable test specification would read: five cycles between 0 and 50 °C, ramp 2 °C/min, dwell 1 h at each limit, part unclamped on three points. Soak 12 h at 20 °C before each measurement. Measure bore positions from datums A, B and C and flatness of datum A on the same CMM program before and after. Limits: bore position shift 0.003 mm maximum on any bore, flatness change 0.002 mm maximum.

Hypothetical results might look like this: bore shifts of 0.0012, 0.0018, 0.0009 and 0.0021 mm, flatness before 0.0041 mm and after 0.0047 mm, CMM repeatability 0.0007 mm from two pre-test runs. Every number is under its limit and the largest shift is three times the noise floor, so the result is meaningful. The plate passes. Had the largest shift been 0.0029 mm with a 0.0020 mm noise floor, the honest answer would be “inconclusive” and the fix would be a better measurement, not a rejected part.

What to write on the drawing or in the PO

Put the test on the drawing as a note, or in the purchase order as a quality clause, and say the same thing in both places. A workable note contains seven items:

1. Purpose: “Dimensional stability acceptance test after final machining and stabilization, before coating.”
2. Cycle: number of cycles, the two temperatures, ramp rate and dwell.
3. Support during cycling: unclamped, three point.
4. Measurement: at 20 ± 1 °C after a stated soak, same CMM program before and after, datum scheme per the drawing.
5. Features measured: list them by balloon number. Do not write “all dimensions”.
6. Limits: per feature, and a trend condition if you want one.
7. Report: what it must contain and that it ships with the parts.

Also say who runs the test. Many shops can run the cycles in a small oven and a freezer and measure in house. Say whether the test applies to first article only or to every part, because that changes the quote. And say what happens on a fail: a second stabilization and retest is the usual answer, and it is better agreed before the order than argued after.

What the report should contain

A report you can act on has these parts:

• Part number, serial number, drawing revision and heat lot of the material.
• The test parameters as run, with a chamber temperature log, not just the values from the note.
• Both measurement runs in full, plus the pre-test repeatability runs.
• Room and part temperature at each measurement.
• A table of each feature with before, after, difference and limit, plus a pass or fail against each.
• Any intermediate measurements, with the per cycle trend if taken.
• The CMM identity and its last calibration date.
• A signature and a date.

What to put on the RFQ

A shop cannot quote a stability test it has not seen. Attach the drawing with the test note, or paste the note into the RFQ text. State the alloy and spec, Invar 36 to ASTM F1684 or Super Invar, plus the heat treatment sequence you expect, because the test only makes sense after stress relief and stabilization. Say whether the test is on first article or every part, and whether you want measured CTE data from the heat lot as well. When you send the RFQ, we match you with up to three shops that cut the alloy, run the sequence and can measure at 20 °C to the repeatability the test needs. For the specs named above, see the Invar specifications reference and the Invar 36 machining page.

Frequently asked questions

Reviewed against

• ASTM E228, Standard Test Method for Linear Thermal Expansion of Solid Materials With a Push-Rod Dilatometer
• Wikipedia, Invar

Values in this guide are typical handbook figures for planning. The mill certificate for your heat lot and the shop's quote for your geometry govern.

Keep reading

• Stress Relief for Invar Machined Parts: When and How Shops Do It · Process
• Reading an Invar CMM Inspection Report: What Each Number Means · Process
• Long-Term Dimensional Stability: Does Invar Age After Machining? · Materials
• Invar Machining Tolerances: What a Shop Can Hold and What It Costs · Process

All choosing a supplier guides