Scientists 3D Print World's Smallest Wine Glass From Quartz Glass

• 1
  Key takeaways
• 2
  What actually happens when scientists 3D print world's smallest glass parts
• 3
  Why quartz instead of polymer or borosilicate
• 4
  Where the process stops working
• 5
  What this means for a metal and glass prototype
• 6
  Quartz glass vs borosilicate vs polymer for small optical parts
• 7
  When to print glass and when to machine it

The team did not melt quartz and squirt it. They started with a photocurable resin loaded with amorphous silica nanoparticles. A laser or projector writes a 3D pattern into that resin, voxel by voxel, exactly like a standard two-photon or DLP printer. The output at this stage is not glass. It is a green body, a soft plastic shape holding silica powder in place.

The second step is thermal. The green body goes into a furnace where the organic binder burns out and the silica particles sinter together. Furnace temperature sits well below the 1,700–2,000 °C needed to melt quartz in bulk. That lower temperature is the point of the whole method. It saves energy and it keeps the shape from slumping under its own weight.

The third step is where the wine glass appears. The sintered part is transparent quartz glass, and it is smaller than what came off the printer. Linear shrinkage runs roughly 30–40% depending on silica loading. Engineers scale the CAD model up by the inverse of that factor before printing, so the final resonator or lens lands on the target dimension.

A rim thinner than a human hair is the demo, not the product. The useful output is the optical resonator: a tiny ring or torus that traps light at specific wavelengths. Those parts are hard to machine because they are small, curved, and need low surface roughness. Printing builds them in one piece instead of assembling them.

• 1
  Green body firstPrinted resin holds the silica; it is not glass yet.
• 2
  Furnace secondBinder burnout plus sintering below bulk melting point.
• 3
  Shrinkage thirdScale the model up 30–40% to compensate.

Quartz glass has a low thermal expansion coefficient, roughly 0.5 × 10⁻⁶ /°C. That number matters for resonators. A ring that changes size with temperature drifts off its resonance wavelength. Polymer optics move far more. Borosilicate sits between the two, but quartz wins when the part has to hold a wavelength across a temperature swing.

Quartz also transmits deep into the ultraviolet. Standard glass absorbs below about 350 nm. Fused quartz keeps transmitting down past 200 nm. If the application is UV lithography, fluorescence detection, or a sensor window, that window of transmission decides the material choice before anything else does.

The trade is cost and cycle time. A printed quartz part spends hours in a furnace, and the furnace run is not a five-minute operation. For a part that fits on a CNC table and is larger than a few millimeters, machining a quartz blank or molding it is usually faster and cheaper. Printing earns its place when the geometry is too small, too curved, or too internal to reach with a tool.

• 1
  Low expansionKeeps resonance stable across temperature.
• 2
  UV transmissionUseful below 350 nm where standard glass blocks.
• 3
  Slow cycleHours in the furnace, so cost climbs fast.

Feature size has a floor. Walls thinner than roughly 1–2 μm tend to collapse or ball up during binder burnout, before the silica has fused enough to support itself. The wine glass demo pushed near that edge. It is a useful picture of the limit, not a promise that any 0.5 μm wall will survive.

Overhang and enclosed volume are the other constraints. Sintering pulls material inward, so a long unsupported bridge sags. A fully sealed hollow chamber traps binder gas and can blister or crack. Designers add drain paths or split the part, then bond it after firing.

Surface finish after sintering is not optical grade by default. Sintered quartz often needs polishing or a light fire-polish to reach the roughness an optical surface needs. That finishing step can be more expensive than the print itself on small runs. Plan for it in the quote, not after the first part fails inspection.

Shrinkage is predictable in X, Y and Z only if the furnace profile is repeatable. Change the ramp rate or the hold time and dimensions move. A production run needs a fixed thermal recipe plus a first-article measurement, not just a scaled CAD file.

• 1
  Thin walls failBelow about 1–2 μm the wall collapses.
• 2
  Sealed voids trap gasAdd vents or split the part before firing.
• 3
  Post-polish is normalBudget for it on optical surfaces.

Most of the parts we quote are not pure glass. They are hybrids: a machined aluminum or stainless housing that holds a small glass element, a lens, a window, or a resonator. The glass does the optical work. The metal does the mounting, the threads, and the heat path.

That split is where tolerances get interesting. The metal seat has to locate the glass without crushing it. Quartz is brittle and does not yield, so a press fit that works on aluminum will crack a glass ring. We usually machine the seat to a slip fit and hold it with a compliant element or adhesive, then measure the assembly.

On the metal side, our 5-axis centers hold ±0.005 mm on features like a lens seat or a resonator pocket. A Ø400 mm rotary table handles concentric bores in one setup, which keeps the seat and the bore coaxial. Machined finishes land at Ra 0.8–1.6 μm as a standard, and Ra 0.2–0.8 μm when the drawing calls for it.

A typical build starts with a drawing and a function note. We run DFM within 12 hours, flag any glass-to-metal fit that will crack, and quote the metal parts, the finishing, and an assembly check. Uploads stay confidential, and an NDA is available before you send files.

• 1
  Slip fit, not press fitQuartz cracks before it deforms.
• 2
  One-setup boresKeeps the glass seat coaxial with the housing.
• 3
  12-hour DFMFit and tolerance issues flagged before cutting.