Researchers demonstrate droplet-based volumetric 3D printing for serial production

Researchers at the Ulsan National Institute of Science and Technology (UNIST) in South Korea published a study in Advanced Functional Materials describing a volumetric additive manufacturing (VAM) method in which photocurable resin is dispensed drop by drop from a glass pipette, printed within the pendant droplet using tomographic light projection, and released onto a moving substrate in under a minute per part—with less than three seconds of non-printing time between cycles.

The work addressed a long-standing bottleneck in computed axial lithography (CAL), the most established VAM technique, which solidifies an entire 3D volume simultaneously using rotating resin and computed light fields rather than building layer by layer. While CAL eliminates the height-dependent speed penalty of conventional SLA and DLP, existing CAL systems require manual loading and unloading of resin vials, the use of an index-matching fluid to prevent optical distortion, and physical removal of the cured object between prints—steps that collectively undermine the throughput advantage of volumetric printing.

Droplet as print chamber

In the DVAM configuration, a syringe pre-filled with resin feeds a borosilicate glass pipette mounted on a motorized rotation stage. Applying pressure to the syringe forms a pendant droplet at the pipette tip, which then serves as a self-contained, temporary print volume. A 442-nanometer laser diode illuminates a digital micromirror device (DMD) whose computed patterns are projected onto the rotating droplet, selectively polymerizing the target geometry within seconds. After curing, the solidified structure detaches under gravity and surface-tension forces and is deposited onto a substrate moving beneath the pipette, while the syringe pressure immediately forms the next droplet.

The absence of an index-matching fluid, however, introduced substantial optical distortion: the curved air-resin interface at the droplet boundary acts as a refractive lens, concentrating light toward the droplet center and compressing the effective exposure diameter to roughly 66% of the target dimension. To correct this, the team implemented a real-time inverse ray-tracing algorithm that estimated the droplet’s radial profile from CCD camera images, then computed compensated projection patterns on the DMD to restore uniform dose delivery. Droplet geometry detection was automated using a customized YOLO-based AI framework retrained to track the resin boundary frame-by-frame under varying illumination and droplet shapes.

Serial volumetric fabrication

With refractive correction active and a rotation speed of 24° per second, the team ran DVAM through 10 consecutive prints spanning geometries such as lattice structures, hollow pyramids, arches, chess pieces, the Eiffel Tower, and letterforms. Total fabrication time for all ten parts was approximately 10 minutes, compared with significantly longer runtimes for equivalent FFF, SLA, and conventional VAM workflows.

Print fidelity was quantified using X-ray micro-computed tomography (micro-CT) and the Jaccard index, a metric measuring volumetric overlap between the printed object and its original CAD geometry. Corrected cubes and pyramids achieved Jaccard scores of 92.18% and 88.93%, respectively, compared with 55.89% and 52.04% for the uncorrected equivalents. Geometries containing holes and negative features showed the largest gains: corrected cylinder-hole and thin-layer-hole structures scored 35.76% and 40.66% in the uncorrected case, with correction restoring peripheral feature accuracy that refractive demagnification had otherwise eliminated. The minimum reproducible feature size in the current prototype was approximately 150 micrometers.

The team noted that residual lateral wobble in the rotating droplet—averaging approximately 25 micrometers, with peaks near 50 micrometers—contributed to rounding in fine features and identified improved mechanical stabilization and stage alignment as the primary paths to higher geometric fidelity in future iterations.