This UCLA PhD dissertation asks how two-photon lithography (TPL), a laser 3D-printing method with sub-micron resolution, can overcome three limits: printing one material at a time with low throughput, being unable to deform parts while printing them, and being unable to assemble and join micro-parts. It builds and experimentally demonstrates three hybrid systems. (1) Scanning two-photon continuous flow lithography steers the femtosecond laser to follow particles in a flowing resin; it printed ~50 um bi-material cell-carrier particles at 10 particles per second, with a bead-labelled shelter reg

A laser that hardens liquid only at its focus — taught to chase a flowing stream, to weld what light is bending, and to glue spheres that light carries.

One colour writes, one colour holds

Ordinary light-curing hardens resin all along the beam. Two-photon lithography uses a femtosecond laser whose energy is absorbed only when two photons arrive at once, which happens only in the tight focus. Move the focus and you draw in 3D, with features far smaller than a cell.

The dissertation adds a second laser. Its resin cures under the 760 nm (or 690 nm) femtosecond writing beam but ignores 532 nm green light. That green light can therefore grip and move objects (optical tweezers) without printing anything. Printed polymer (refractive index 1.4912) stands out enough from the liquid resin (1.3918) for the tweezers to grab freshly printed parts.

Longer exposure makes the two-photon voxel grow taller, but it stays at the focus. Chapter 2 uses exactly this to set particle heights without moving the focus up or down.

A focused Gaussian beam (waist 0.35 µm, Rayleigh range 0.9 µm) in a 12 µm resin film. Each point receives dose = exposure × intensity for one-photon absorption, or exposure × intensity² for two-photon absorption. Resin hardens where the dose passes a fixed threshold. illustrative

Write in a river

Steering the laser along with flowing resin prints particles made of two materials.

Bend it, then pin it

Tweezers bend freshly printed parts, then welds lock spring energy inside them.

Carry, then glue

An automated system carries silica spheres with light and glues them into crystals.

Write in a river

Galvanometer mirrors swing the femtosecond spot across the channel at 15,000 points per second. Before each run, the system measures the flow and shifts every scan point downstream. The resin keeps moving and the laser keeps up, so each particle prints as designed.

The cell carriers are dumbbell-shaped. Their asymmetric shape is designed to line up by itself in flow. They are printed where three streams flow side by side. The narrow middle stream (about 10 µm wide) carries fluorescent beads, so the carrier's sheltered groove ends up in a different material from the rest of its body.

A groove without moving the focus

The carrier is cut into three stacked slices, but the focus never moves up or down. A region scanned with fewer points gets less light and grows less tall. Scanning the groove more sparsely carves a shallow shelter where a cell could sit.

Confocal measurements: shelter 18–24% shorter than the neighbouring region (depending on laser power), ~1.5 µm deep , ~10–15 µm long to match adherent cells. measured The slider uses height ∝ exposure. illustrative

Flow is fastest in the middle

Liquid drags along the walls, so speed varies with depth (a parabola, called Poiseuille flow). Print near a wall and the top of a particle travels faster than its bottom, so its slices slide apart. The fix was wider (1200 vs 200 µm), deeper (130 vs 50 µm) channels, printing at the mid-plane.

Slip = difference in flow speed across a 7 µm-thick particle × 0.1 s print time, at a mean speed of 900 µm/s. illustrative

Bend it with light, then pin it

Printed parts usually come out relaxed. Some of the most interesting microsystems need springs that are already loaded: switches that snap, lattices that hold energy, joints that roll. This system prints a part, bends it with several green traps while it floats in the same resin, then prints new features that weld it in the bent shape.

This was the first system to combine two-photon lithography with holographic tweezers, which create many independently steered traps from one laser. No washing step separates printing and bending, so the whole sequence can run automatically.

The writing laser prints the structure. G1 sits on five base layers, so it is fixed to the glass. G2 and the shuttle S float.

Bistable means two resting positions: up or down, like one bit of memory with no electronics. The dissertation demonstrated an optical trap driving the welded switch between its two stable positions (flexures 800 nm thick). demonstrated

Grip at the edges

The dissertation extended a ray-optics simulation to accept any CAD (.stl) shape. It showed that a trap pulls hardest where light bends at a part's edges, and grips best where two edges sit close together. Very thin features grip steeply but only over a tiny range, so the parts were designed with 3 µm-wide handles. Traps were focused slightly above the writing plane: about 375 nm above a part's mid-plane, the vertical force is zero and the part moves without tilting.

The paper's simulation used a 50 mW, 532 nm trap. simulated in the paper The curve here only reproduces its shape. illustrative

One resin, two lasers

Resin: 38.4% ethoxylated trimethylolpropane triacrylate, 1.1% TPO-Li photoinitiator, 60.5% water. It cures by two-photon absorption at 760 nm and stays unreactive at 532 nm.

Writing: femtosecond laser steered by galvanometer mirrors. CAD files are sliced into G-code scan paths.

Holding: 532 nm continuous-wave laser on a 1920 × 1152-pixel phase-only spatial light modulator, which makes many traps from one beam.

Both lasers share one 100× oil-immersion objective (NA 1.45). Camera images locate parts between steps.

Two traps push the smiley disk into its box and one closes the lid. When the traps switch off, the spring pops the face out and knocks the lid open. demonstrated · Fig 3.3

Four traps compress V-shaped flexures, four welds fuse each cell pair, and a second pair is printed, pushed and fused on top. demonstrated · Fig 3.4 Releasing the energy on impact proposed

Six traps pull on 4 µm handle disks, and the lattice widens as it lengthens. Average Poisson's ratio −0.54 (sideways change per unit stretch; negative means it widens), over 3 cycles within 30% strain. measured · Fig 3.5

Straps are printed straight, then bent around their cams by tweezers. The finished joint can roll with nearly constant strain energy. demonstrated · Fig 3.9

Carry spheres with light, glue them with light

Granular crystals are ordered packings of touching spheres. The way stress waves travel through them depends on exactly where each sphere sits. Self-assembly only makes a few standard packings. Before this work, the largest crystal built with optical tweezers was 27 spheres (3×3×3), placed by hand one at a time.

Here, a single laser beam reflects off a spatial light modulator to make several green traps. They carry 4.86 µm silica spheres through a water-based resin, and the femtosecond laser cures small polymer joints between them.

Five traps, never three in a line

The phase pattern on the modulator is computed with the Gerchberg–Saxton algorithm. The dissertation's simulations found that traps placed along a line come out with very uneven power, plus unwanted extra "ghost" traps. Patterns with no more than two traps on any line come out uniform. Drag the rings, or try the presets.

Challenge: find a power that lifts all five spheres without boiling any. Lifting needs 22 mW per trap (the system slid spheres at 11 mW and lifted them at 22 mW). Above 60 mW, a trap boils the resin. illustrative limit

Live, simplified model: a 256 × 256 phase-only modulator lit by a Gaussian beam. It starts from the sum of one tilted wavefront per trap, then runs 5 Gerchberg–Saxton iterations with equal target weights. The image is the trap-plane intensity, with the undiffracted spot left out (the real system blocks it). live model It reproduces the trend in the dissertation's Fig 4.5. It is not the authors' code.

Why lines are special, in this model: a phase-only hologram leaks light into mixing points such as 2a − b, or a + b − c, of the trap positions a, b and c. With evenly spaced traps in a line (or four traps forming a parallelogram), those points land exactly on other traps or just beyond the ends of the line. Light then piles up unevenly. In most irregular patterns the leaked light scatters into empty space instead.

1 Pump delivers spheres

2 Check 1: discard clumps & misfits

3 Plan paths to a holding pattern

4 Lift: check 2 finds defects

5 Place & weld with X-shaped joints

6 Dispose of failures, refill

Step 3 uses the Hungarian algorithm (minimum total travel) and collision-free paths at 11 mW per trap. Lifting doubles the power to 22 mW. A sphere that stays on the glass shows a bright centre and dark ring, so the camera can spot it. Joints are cured with 30 mW of 100 fs pulses at 690 nm, for 180 ms each.

The real 5×5×5 build took about an hour. Here it is compressed, and the 6-neighbour layout is drawn as a schematic. schematic timing

How fast could this ever go?

The demonstrated system handles five spheres at a time and builds about 14 × 10⁻⁶ mm³ of crystal per hour. The dissertation estimates the ceiling for an ideal holographic system: laser force limits how fast a sphere can be dragged through the liquid, and the field of view limits how many traps can work at once. estimate in the paper

The conclusions put the ceiling at up to ~3.7 mm³/h, if the liquid-crystal modulator were replaced by a micro-mirror array.

Advancing Two-photon Lithography by Integrating Microfluidics and Optical TweezersSamira Chizari · PhD dissertation in Mechanical Engineering, University of California, Los Angeles, 2020 · UCLA Electronic Theses and Dissertations (eScholarship). Committee co-chairs: Jonathan B. Hopkins and Yong Chen.

Chapters of the dissertation appear as journal versions: Ch. 2 in Optics Express 28(26):40088 (2020). Ch. 3 in Materials Horizons 6(2):350 (2019), with case studies from Nature Communications 10:882 (2019) and 9:4594 (2018). Ch. 4 in Small 2000314 (2020).

All visuals on this page are original interactive re-drawings based on the dissertation's text and figures. No figures or videos from the dissertation are reproduced. Interactive models are simplified illustrations unless marked as measured.

Colour code: red = femtosecond writing laser (near-infrared, 690–760 nm, drawn red) · green = 532 nm optical tweezers · blue = flowing resin and fluorescent beads · amber = cured welds.