Fast micron-scale 3D printing with a resonant-scanning two-photon microscope

Abstract

3D printing allows rapid fabrication of complex objects from digital designs. One 3D-printing process, direct laser writing, polymerises a light-sensitive material by steering a focused laser beam through the shape of the object to be created. The highest-resolution direct laser writing systems use a femtosecond laser, steered using mechanised stages or galvanometer-controlled mirrors, to effect two-photon polymerisation. Here we report a new high-resolution direct laser writing system that employs a resonant mirror scanner to achieve a significant increase in printing speed over current methods while maintaining resolution on the order of a micron. This printer is based on a software modification to a commercially available resonant-scanning two-photon microscope. We demonstrate the complete process chain from hardware configuration and control software to the printing of objects of approximately 400 × 400 × 350 μm, and validate performance with objective benchmarks. Released under an open-source license, this work makes micron-scale 3D printing available at little or no cost to the large community of two-photon microscope users, and paves the way toward widespread availability of precision-printed devices.

Keywords: 3D printing; Additive manufacturing; Direct laser writing; Lithography; Resonant scanning; Two-photon microscopy.

Fig. 1.

Overview of the resonant rDLW printer. (a) Schematic of the optical path from laser source to printed object. Solid-outline box shows components comprising a standard resonant-scanning 2-photon microscope. Dash-outline box shows zoomed view of build plate stack. (b) The raster scanner rapidly sweeps the laser focus across the X-axis of the printing workspace; a slower galvanometer scans the laser focus across the Y-axis. (c) Top: laser power is modulated above (red line) and below (grey dotted line) the polymerisation threshold (green dashed line) throughout the X-axis sweep. Bottom: by applying this pattern of laser modulation over the workspace, solid features can be built up line by line and layer by layer. (d) Left: 3D model of a chess pawn. Center and Right: SEM micrograph of the chess pawn (192 × 350 μm) printed in IP-Dip photoresist (center) and Ormocomp (right) with our rDLW printer. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 2.

Rulers for calibrating the rDLW system. (a) Ruler for measuring X- and Y-workspace dimensions. (b) Cubes used to calibrate object size and uniformity of power delivery. The cubes shown have widths 300, 200, and 100 μm. The printing parameters were 2.2 × (i.e., 302 × 302 μm FOV), 3.3 × and 6.6 × magnification (zoom), respectively. Each X–Y plane was built with 152 × 1024 voxels, and the vertical spacing between the planes was 0.5 μm for all three cubes. (c) Ruler for Y-axis calibration. The printing parameters are the same as for the 300-μm cube in (b). The horizontal line spacing on the ruler is 5 μm. (d) Vertical ruler for Z-axis calibration. Each row along the X axis contains 11 steps with 1-μm height difference. Adjacent steps along the Y axis have 10-μm height difference. The total height of the ruler is 300 μm. The printing parameters were the same as for the 300-μm cube in (b).

Fig. 3.

Sinusoidal laser velocity over the X-axis results in nonuniform voxel size. Both that and optical nonuniformities such as vignetting require corrective laser power compensation. (a) Laser focal point velocity as the resonant scanner sweeps across the X-axis. (b) Laser focus position varies sinusoidally with time (blue line). The active scanning region is restricted to a portion D of the sweep, with X-axis voxel positions shown as black horizontal dashes. For clarity, we show where voxels would be defined for an 8-kHz resonant scanner with a 1-MHz control system, which yields only 45 voxels. In order to maintain uniform energy deposition across the workspace, laser power is modulated by two factors: it is scaled along the X-axis by the focal point's speed cos(t), and along the X–Y plane by a learned model of the inverse of optical darkening due to polymerisation. (c) Cross-section of the power compensation along X, in which y = 0, x ∈ [−208, 208] μm (1.6× zoom on our rDLW system). (d–g) 400 × 400 × 100-μm bricks used to measure and calibrate energy deposition. The upper image shows the print power mask over the 208 × 208-μm workspace; the middle image shows an actual printed object (normalised using a baseline fluorescence image); and the bottom image shows brightness data gathered by sweeping the object over the lens so that the same set of pixels in the imaging system may be used for each measurement in order to bypass optical nonuniformities therein. Shown: (d) constant power (note that (1) at this zoom optical vignetting comes close to compensating for X-axis nonuniformity due to varying beam speed; and (2) this image was printed at lower nominal power than the others in order to avoid boiling; for the other images, the speed compensation appropriately reduces power); (e) only (X-sinusoidal) speed power compensation; (f,g) two iterations of adaptive power compensation over the visual field (see Section 2.5). The images and data were obtained with ScanImage on our rDLW system. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4.

Complex geometric objects printed with our rDLW printer. (a–c) Woodpile structure with design dimensions 60 × 60 × 60 μm. Along the X-axis, bar thickness was 2 voxels (0.8 μm) and bar spacing 4 voxels (1.6 μm). Bar thickness and spacing on the Y-axis were 13 and 26 voxels respectively in order to be the same size as the X-axis beams, and on the Z-axis bars are 1 voxel thick with 6-μm spacing. The focal plane resolution was 152 × 1024 voxels, and the focal plane (Z) spacing was 0.2 μm. (d) A torus knot design printed at 100 × 100 × 150 μm (top right) and 50 × 50 × 75 μm (bottom left). The inset shows details within the circumscribed region of the bottom left structure. Both knots were printed with focal plane resolution 152 × 512 voxels and focal plane spacing 0.2 μm. (e) SEM micrograph of a Charles Darwin statuette printed with our rDLW printer.

Fig. 5.

Objects with one- and two-voxel features printed on our rDLW system. (a) Object used to estimate minimum voxel size on the Y and Z-axes. All bridges have single-voxel height (Z), and increasing width on the Y-axis. The bottom bridge has one-voxel width; thus, it gives an idea of the thinnest suspended structure that can achieved with the parameters and photoresist used here. The object was printed with 2.2× magnification for a 302 × 302 μm field of view. The resolution of each focal plane is 152 × 1024 voxels. The vertical distance between Z planes in the support structure is 0.5 μm (the bridges span only a single Z plane). (b) Object used to estimate the voxel size on the X- axis. The printing parameters are the same as in (a). The bridges were designed to be two voxels wide on the X-axis, so their size follows a sinusoidal distribution due to the cosinusoidal speed profile of the laser beam. (c) Top view of the central bridge of (b), which represents the largest value in the workspace of double-voxel X resolution at this zoom level. (d) Top view of the lowest bridge of (a). (e) View of the lowest bridge of (a) at 60° from the top view.