Optical flats
Using optical flats to compare polished surfaces, interpret interference fringes and control reference, support and uncertainty limits.
Tags: inspection, dimensional-metrology, optical-flats
On this page
An optical flat is a precision reference surface used to inspect another surface through optical interference. It is useful on suitable polished or lapped faces, including gauge blocks, setting standards and micrometer anvils. The observed pattern compares the two surfaces under the measurement conditions; it does not directly establish thickness, parallelism or every aspect of surface texture. Mitutoyo identifies reflective gauge blocks and setting standards as applications for its Series 158 flats. Mitutoyo Asia Pacific, Optical Flats Series 158, undated product page
Illustrative local estimate: assume 589.0 nm monochromatic light, near-normal viewing and a suitable calibrated reference. If a fringe bows by 0.30 of the adjacent fringe spacing relative to its straight reference chord:
relative surface departure ≈ 0.30 × 589.0 / 2 = 88.35 nm ≈ 0.088 µm
This estimates departure in the evaluated region under those assumptions. It is not automatically a whole-aperture peak-to-valley result or a complete uncertainty statement. A large total band count with almost straight fringes can give a smaller local departure than a few strongly bent bands.
What the fringes represent
With monochromatic illumination, reflections from the reference and test surfaces interfere across the thin air gap. A deliberate small wedge makes the pattern easier to inspect. Straight, parallel bands indicate a uniform wedge between surfaces; bending or distortion indicates changes in their relative separation. Edmund Optics describes the air-wedge method and estimating departure by comparing fringe curvature with fringe spacing. Edmund Optics, Optical Flats, undated application note
For near-normal observation across an air gap, one interval between adjacent dark fringes, or adjacent bright fringes, corresponds approximately to half the illumination wavelength in surface separation. It is not one full wavelength of surface height. NPL explains estimating a surface departure from the distortion of a fringe as a fraction of its spacing, rather than counting all the bands across the part. More bands can simply mean more relative tilt. NPL, Fundamental Good Practice in Dimensional Metrology, GPG80, optical-flat discussion, July 2005; corrected reprint October 2012
Do not assign “convex” or “concave” from an isolated pattern without establishing its sign. Lapmaster notes that the initial band pattern alone does not determine the direction of departure; additional technique is required. Preserve the orientation and use the applicable interpretation procedure rather than guessing which way the surface curves. Lapmaster Wolters, Optical Flat Use, undated technical guidance
Choosing the reference
| Choice | Information to specify |
|---|---|
| Aperture and thickness | Required inspected area, usable reference area and support arrangement |
| Reference quality | Numerical flatness, reference wavelength, evaluated area and error metric |
| Single or double surface | Which face is the calibrated reference; whether both faces are specified |
| Material and handling | Thermal behaviour, mass, durability and compatibility with the test setup |
| Calibration | Surface identification, map or profiles, uncertainty and mounting conditions |
A wavelength fraction needs its wavelength. Edmund's application note defines its λ/20 reference at 632.8 nm, corresponding to 31.64 nm peak-to-valley. That is a reference-surface specification, not the uncertainty of every subsequent shop measurement. Edmund Optics, reference-flat specification convention
For a bounded product example, Kemet 361225 is listed as a 75 mm diameter, 15 mm thick, single-sided, 1/10-light-band (λ/20) flat. The double-sided counterpart is 361240. This demonstrates that a fraction of a light band differs from the same fraction of a wavelength; confirm the source convention before comparing catalogues. A specified pair of flat faces also does not by itself establish their parallelism. Kemet separately offers optical parallels for micrometer-anvil flatness and parallelism checks. Kemet, Optical Flat product table and optical-parallel description, undated
Edmund's TECHSPEC Single Surface Optical Flats specify the secondary face as window-quality for viewing; it should not be assumed equivalent to the reference face. The family page states that λ/20 flats 76.2 mm and larger include a calibration certificate. Confirm the actual supplied item and certificate rather than assuming every size or grade includes one. Edmund Optics, Single Surface Optical Flats
Preparing and handling the surfaces
Both surfaces must be clean and suitable for producing visible fringes. Kemet instructs users to clean with lens tissue or a soft lint-free cloth, place the flat carefully without sliding, and clean and return it to its protective case after use. Its sodium-light unit needs warm-up before use. Follow the actual light source and flat instructions rather than imposing a universal warm-up or solvent recipe. Kemet, How to keep optical flats in great condition?, undated
Poor contrast, crowded bands or unstable patterns are reasons to inspect cleanliness, illumination and seating before accepting a result. Do not polish or otherwise alter a finished inspection surface merely to obtain a prettier pattern unless that preparation is part of the authorized measurement process. Record any preparation that changes the surface being evaluated.
Loading, support and environment
An optical flat's weight can distort a compliant part. Lapmaster describes a reflection stand that places the component on top of the flat, allowing inspection without loading the component with the flat's weight; it highlights thin carbon-graphite components as an example. Even then, the component remains subject to its own weight and support conditions. “No contact force from a probe” does not mean an undeformed part. Lapmaster Wolters, Optical Flat Reflection Stand
A reference comparison alone cannot separate both surfaces' unknown errors. NIST describes three-flat calibration using pairwise combinations and rotation to determine individual reference errors. The reference can also deform: its 2008 three-flat research shows that mount forces and changes during rotation affect optical-flat calibration, particularly for larger, heavier flats. Its demonstrated 300 mm fused-silica arrangement is a specific study, not a correction to apply to every flat. Griesmann and colleagues, Stationary and non-stationary deformations in three-flat tests, July 7, 2008
NIST's historical Handbook 150-2F, September 2003, section 2.10, identifies thermal equilibrium, reduced air turbulence and calibration in the intended mounting orientation as technical considerations. It distinguishes diagonal profiles from a full surface map and calls for reporting the support arrangement. These remain useful measurement distinctions; this dated handbook is not presented as a current accreditation requirement. NIST Handbook 150-2F, optical reference planes
Interpreting and reporting a result
Define the inspected area and evaluation method before judging conformance. Account for reference error, illumination wavelength, reading repeatability, surface preparation, support and thermal effects. A manufacturer's flatness grade and a visually repeatable fringe estimate do not together establish the complete measurement uncertainty.
Retain flat identity and reference face, calibration scope, source wavelength, part orientation/support, evaluated area, fringe image or readings and the acceptance rule. Escalate ambiguous patterns or near-limit decisions to a validated interferometric or other suitable form-measurement method.