Femtosecond Laser Processing: True Three-dimensional Nanofabrication Technology

• 1
  Why a 100 fs pulse behaves differently from a 10 ns pulse
• 2
  How femtosecond laser processing reaches inside a transparent part
• 3
  Feature size, aspect ratio, and the limits of the method
• 4
  Which materials femtosecond laser processing actually suits
• 5
  Femtosecond laser processing versus the alternatives

A femtosecond is 10^-15 s. A 100 fs pulse is roughly five orders of magnitude shorter than the nanosecond pulses used in most marking and drilling lasers. That gap is not a small improvement. It changes which physical process removes the material.

With a long pulse, energy lands in the electron system, electrons collide with the lattice, and heat spreads outward for nanoseconds. The result is melting, recasting, micro-cracking, and a heat-affected zone measured in tens of micrometres. Feature edges drift, and thin walls distort.

Shorten the pulse below the electron-lattice relaxation time and the energy never has time to diffuse as heat. Electrons are stripped from the lattice in a Coulomb explosion and the material leaves as ions and vapor. The lattice stays cold at a distance of a few nanometres. Engineers call this cold ablation.

The practical payoff is clean edges on materials that normally crack or char. Polymers, glass, ceramics, and hardened tool steel all cut without a recast layer. The trade is speed. Removing a cubic millimetre of steel with 100 fs pulses takes far longer than a 20 ns fibre laser doing the same job.

That single trade-off decides most sourcing questions. Femtosecond laser processing is chosen for feature size and edge quality, not for throughput. If the part is 40 mm across with 0.2 mm features, a mill will almost always beat it on cost and lead time.

A transparent material such as fused silica or PMMA has a band gap far larger than a single 800 nm photon can bridge. One photon passes straight through. Nothing happens. This is why you can see through the workpiece at all.

At the focal point of a high-numerical-aperture objective, photon density is high enough that two photons arrive within the same femtosecond window and their energies add. A 1.55 eV photon plus another 1.55 eV photon now crosses a 3.1 eV gap. Absorption becomes possible only where the beam is focused.

Outside that focal volume, intensity falls off as the square of the distance from focus. Two-photon absorption scales with the square of intensity, so the effective writing volume is much smaller than the optical spot. Voxels of 100 to 200 nm are routine with a 1.4 NA oil objective.

The consequence is genuine three-dimensional writing. Move the focus in X, Y, and Z and you write a continuous path inside a solid block. No layer stacking, no support structures, no sealed internal channels that need to be glued shut afterwards.

This is what separates femtosecond laser processing from stereolithography and from two-photon 3D printing that cures a resin. The workpiece itself is the resist. Glass, crystal, and some polymers change refractive index or dissolve selectively after exposure, and the pattern is developed in a later etch step.

Resolution is set by the objective, the pulse energy, and the material. With a 1.4 NA objective and pulse energies near threshold, written lines of 150 to 300 nm width are achievable in fused silica. Push the energy higher and the voxel grows, often to 1 µm or more, because the squared intensity profile widens with it.

Aspect ratio is the harder constraint. Writing a channel 20 µm deep and 1 µm wide is possible. Writing one 2 mm deep and 1 µm wide is not, because the focus cannot stay tight through that much material. Spherical aberration alone will blur the spot before you reach that depth.

Surface finish inside a written channel is rough by machining standards. Etched glass channels land around Ra 0.5 to 2 µm depending on etch chemistry and annealing. If a mating surface needs Ra 0.2 to 0.8 µm, that surface has to be polished or machined after the laser step.

Write speed is typically 0.1 to 10 mm/s along the path, and each voxel is a separate exposure. A 10 mm long micro-channel with a 200 nm voxel pitch means tens of thousands of exposure points and a raster time measured in hours. Batch several parts on one stage to amortise that.

So the method makes sense for microfluidic masters, photonic waveguides, optical diffractive elements, and MEMS-scale features. It does not make sense for a 200 mm aluminium housing with a few tapped holes. That part belongs on a 5-axis mill, and we will say so during DFM review.

Transparent dielectrics are the native territory. Fused silica, borosilicate, quartz, sapphire, lithium niobate, and PMMA all absorb at the focus through two-photon or multiphoton processes. Waveguides, micro-optical elements, and lab-on-chip channels are the standard applications.

Metals behave differently. Above the ablation threshold of a few tenths of a joule per square centimetre, a femtosecond pulse still removes steel, titanium, and aluminium, and it leaves almost no recast layer compared with a nanosecond laser. This matters for stents, fuel injector nozzles, and pacemaker electrodes.

Polymers are softer targets. PEEK, polyimide, and PMMA are prone to melting with long pulses. With short pulses you get clean cuts and clean drilled holes down to 10 to 20 µm diameter, which is useful for nozzle plates and micro-sieves.

Hard and brittle materials crack under thermal load. Silicon carbide, alumina, and glass-ceramic are all cut with less chipping when the pulse is short, but the removal rate stays low. Expect to trade throughput for yield on these materials.

What the method cannot do: bulk removal at production rates, deep bores with a high depth-to-diameter ratio, or anything that needs a specific Ra below 0.2 µm straight off the machine. Those go to grinding, EDM, or CNC with a polishing step.