Two-photon polymerisation creates micro- and sub-micrometre polymer geometries with a focused laser. The smallest feature in a brochure does not guarantee a usable part: exposure window, voxel anisotropy, resin, scan strategy, development and measurement jointly determine output. This guide gives an application-validation sequence for micro-optics, microfluidics and fine mechanical structures.

1. Start from function

For optics define wavelength, focus, numerical aperture and wavefront or transmission requirement. For microfluidics define channel section, patency, surface chemistry and pressure resistance. For a mechanical lattice define stiffness, buckling, fatigue and the smallest connection. High resolution is not an acceptance criterion by itself.

2. Map the exposure window

Print a dose matrix across laser power and scan speed. Near the lower boundary a line breaks or disappears during development; excessive dose thickens and merges features and polymerises surrounding volume. Select the centre of a robust window rather than the single finest line that survived.

Threshold depends on material, objective, wavelength and depth. Aberration and absorption can change focus in a thick resin layer. If the process uses multiple heights, do not calibrate only next to the substrate.

3. Measure voxel directionality

A focused voxel is usually longer along the optical axis than laterally. An identical CAD feature can therefore print differently horizontally, vertically and diagonally. Print lines, pillars and gaps in several directions and measure both positive and negative features; a 300 nm pillar and 300 nm opening are different manufacturing tasks.

4. Separate contour and infill

A shell can use fine paths while the interior uses a larger voxel or sparse fill. Their interface may show over- or under-cure, stress or refractive-index differences. Prepare a cross-section and check that the shell does not detach and the bulk does not overwhelm the fine surface.

5. Treat development as manufacturing

Removing uncured resin from a narrow channel is a diffusion and flow problem. Too little development leaves residue; aggressive solvent or drying can deform weak structures. Record solvent, time, agitation, temperature, rinse and drying. High-aspect-ratio elements may require a drying route that limits capillary force.

6. Do not confuse roughness with form error

For surface texture record instrument, lateral resolution, field, filtering and whether the metric is Sq, Sa or Ra. Evaluate form separately against a registered nominal geometry. A spherical surface may have low local roughness but the wrong radius; the two results require different corrections.

An optical component also needs functional testing such as focal spot, modulation transfer, wavefront, loss or spectral transmission. A nanometre surface number alone does not prove optical performance.

7. Account for shrinkage and substrate constraint

Polymerisation and development can cause direction- and geometry-dependent dimensional change. Substrate adhesion may constrain one direction while a free surface moves. Measure on-substrate and released components, and apply CAD compensation only to stable distortion repeated across batches.

8. Test repeatability by position and time

Print the same coupon at several field positions, on at least three days and preferably with more than one resin lot. Track ambient temperature, resin age, laser verification and objective cleaning. One striking part demonstrates capability; series variation demonstrates a process.

9. Use a sentinel coupon and control chart

A short reference build should contain fine positive and negative features, a curved surface, a long slender element and the application's critical geometry. Record at least two dimensions, a surface or functional metric and failure cause. Trend can matter more than a single limit: slow drift may indicate optical contamination or material change.

10. Cost total cycle time and accepted output

Add setup, substrate handling, development, drying, release, measurement and failed parts to scan time. A larger voxel or adaptive resolution improves throughput only if inspection and scrap do not rise. Parallel placement of many small parts often matters more than record speed on one path.

Evidence levels

Verified process data are repeated on the user's machine with a fixed recipe and calibrated measurement. Vendor data are a useful starting point but belong to a named material, objective and sample. LAYERTIMES concludes that 2PP resolution is described not by one nanometre number but by exposure window, three-dimensional form error, functional testing and series variation together.