Peak-to-valley figure is a length. λ/10 at HeNe (632.8 nm) is 63 nm PV. λ/10 at 10.6 µm is 1.06 µm PV — sixteen times looser in nanometres. An optic that is a comfortable λ/10 at the CO₂ line can be worse than λ/2 at 633 nm. If your print says only “λ/10”, the fabricator is entitled to pick the wavelength that makes the part pass. That is not malice. That is how the unit works.
What the interferogram is actually counting
A Fizeau or Twyman–Green at 632.8 nm maps optical path difference (OPD) in the test cavity. One fringe is λ/2 in air for a double-pass Fizeau (round-trip), depending on how the software is scaled — always read the instrument note. The reported PV is the difference between the highest and lowest points on the fitted surface after removing piston and usually tilt. Power (focus error) is a separate Zernike term. Irregularity is what remains after power is removed. If you need a window that does not add focus, specify irregularity separately from power.
ISO 10110-5 notation is the grown-up way to write this: 3/ A(B) where A is PV sagitta error in fringes or nm and B is irregularity. Name the wavelength in the note. “3/ 0.5(0.2) @ 632.8 nm” is a specification. “λ/10” in a title block is a wish.
IR parts tested in the visible
Most IR shops still qualify figure with a visible interferometer because 10.6 µm interferometers are rare and expensive. That is valid if two conditions hold: the transmitted wavefront at the use wavelength is what you care about, and you have converted units. For a transmissive ZnSe window, visible figure on the surfaces plus index homogeneity of the blank both contribute to transmitted WFE at 10.6 µm. Homogeneity of CVD ZnSe is typically good enough that surface figure dominates — until you get into thick, large-diameter windows for kW beams.
Germanium is opaque at 633 nm. You cannot transmission-test a Ge lens on a HeNe Fizeau. You test surfaces in reflection, or you use an IR interferometer. A “λ/10 Ge lens” with no test method is undefined.
RMS vs PV, and why 0.1-wave RMS is not λ/10
PV is a two-pixel statistic. One dust speck or a turned-down edge can dominate PV while RMS (the quantity that predicts Strehl for small random errors) looks fine. High-energy laser windows often specify RMS figure over a clear aperture, plus a separate edge-exclusion. State the clear aperture. A 50 mm part with λ/10 over 80 % CA is not the same as λ/10 to the chamfer.
RFQ checklist
- Figure: PV and/or RMS, over which clear aperture, at which test wavelength.
- Power and irregularity called out separately if the optic is in a collimated path.
- Test method: reflection Fizeau, transmission, or IR interferometer.
- Surface quality (scratch-dig 40-20 / 20-10) is not figure. Do not substitute one for the other.
InfraSpec Optics inspects figure interferometrically and reports the test wavelength on the data pack. If your print still says “λ/10”, we will ask which λ before we quote — that question is the inspection.
References. ISO 10110-5, surface form tolerances (fringes; DIN/ISO historically tied to the Hg e-line 546.07 nm). Shop practice and Laseroptik substrate notes: λ/10 @ 632.8 nm is the common HeNe callout — name it, because it is not the ISO default wavelength. Zygo / Fizeau scaling: one fringe in a double-pass Fizeau is λ/2 in OPD unless the software says otherwise.