This experimental study asks how conformal thin films can be patterned on complex three-dimensional surfaces. Scaffold-Architected Lift-Off (SALO) uses lithographically defined sacrificial scaffolds, usually printed by two-photon polymerization, followed by thin-film deposition and mechanical removal through sonication in isopropanol. The authors demonstrate patterned metal and atomic-layer-deposited films on curved, stepped and vertical surfaces, selective and sequential multi-material patterning, a conducting nanowire crossing a slope, and a fiber-tip diffraction grating. Scaffold geometry a

Nature Communications · 2026

How do you draw a tiny metal pattern around a corner? Print a scaffold, coat everything, then let the scaffold go.

Scaffold-architected Lift-off (SALO) brings thin-film lithography onto three-dimensional surfaces.

An interactive guide to Liu et al.

Print what you want to remove. The scaffold is temporary. The thin-film pattern stays.

A negative mask makes a positive possibility: functional coatings that follow a 3D surface.

01 / A virtual experiment

Make a pattern around a corner.

Choose a surface. Build its mask. Discover what remains after lift-off.

The mask will follow its steps and vertical faces.

Advance one step at a time, or revisit any stage.

The target is bare. Print a temporary polymer scaffold where you want gaps in the final film.

Conceptual geometry; film thickness and process speed are exaggerated. Gold denotes the deposited film.

The laser makes the mask.

A focused femtosecond laser polymerizes IP-Dip resist through two-photon absorption. The GT2 system writes a scaffold directly onto the existing 3D surface.

Deposition makes the film.

The scaffold and exposed target receive a coating. Conformal atomic layer deposition can reach surfaces that directional deposition cannot.

The bath makes the pattern.

In an isopropanol (IPA) bath, ultrasonic cavitation produces microjets and shockwaves. The scaffold and its coating detach, revealing bare substrate at the mask footprints.

On polymer target structures, the researchers used a 10 nm SiO₂ intermediate layer to limit bonding between the target and mask.

02 / Let geometry do the work

Tall enough to release. Strong enough to stand.

Pulling off a coating is a mechanical problem. Adjust the scaffold’s height-to-width ratio and compare with the tested samples.

h is height; w is the attachment width.

Ready to let go.

At h/w = 4.0, this scaffold sits within the reported 0.5–8 lift-off range for 100 nm Ag.

Reference: square-pillar samples, Supplementary Table S1. These are observed ranges for the tested coatings. The drawing and stress indicator illustrate the beam relation σmax = 3p(h/w)² at fixed pressure, rather than a detachment probability.

More height means more leverage. At fixed lateral pressure, doubling h/w gives four times the model’s basal bending stress. Extremely slender masks can collapse before the release step.

Thin shells can cover large areas. Hollow scaffolds attach only at their rim, reducing the footprint that needs to detach.

03 / From design to measurement

How closely does the film follow the mask?

Explore 48 source-data measurements of line features patterned in a 50 nm sputtered nickel film.

Select a width to inspect all three measurements.

Magnified comparison

Bars share a scale within each selection.

Every measurement, in context.

Measured source data from Figure 3b, Liu et al. (2026). Three entries per designed width; signed differences calculated from the reported values.

Approximate direct two-photon scaffold scale. The paper’s 27 nm feature used oxygen-plasma shrinking of a two-photon-printed mask. Its 50 nm example used an electron-beam-defined mask. Those routes extend the method beyond the direct printing scale.

04 / Beyond the flat surface

One process. Many ways to make it useful.

Solid masks, hollow shells and sequential release broaden what can be patterned. The researchers also built working electrical and optical examples.

Hollow-shell masks

A large footprint, a smaller attachment.

A hollow shell contacts the surface at its rim. The supplement shows a letter T patterned in a 20 nm nickel film, with elemental mapping confirming the result.

Sequential and selective release

Pattern more than one material.

Successive SALO steps pattern tantalum, nickel and hafnium on one structure. Deposition geometry can also make a cap release first and its pillar later.

Demonstrated devices / Figure 6

Cross an edge. Pattern a fiber.

The researchers patterned a 900 nm wide gold nanowire across a 54.7° slope. Its total length was 1080 µm, and the measured current–voltage response was linear.

3.01 kΩ measured resistance 2.58 kΩ on the planar reference

Device drawing: conceptual schematic, based on Fig. 6. Measurements are from the paper.

Patterned conductors, optical elements and sensing surfaces can be built directly on complex topography. The paper proposes richer 3D photonic and electronic architectures and parallel printing for greater throughput.

The researchers demonstrate sequential Ni, Ta and Hf patterning. With a ZnO conformal coating and directional Ti deposition, a cap can release first and a pillar later. Choosing the mask structure and deposition geometry controls the release sequence.

See Figure 5 and the “Multi-material patterning by sequential SALO and by selective SALO” section of the paper.

Explore the research. Follow the evidence.

The original research

Xinxin Liu, Zifan Che, Zofia Maj, Lee-Lun Lai, Kristinn B. Gylfason, Valentin Dubois, Shyamprasad N. Raja, Göran Stemme & Frank Niklaus. Nature Communications 17, 6201 (2026). DOI: 10.1038/s41467-026-75538-z.

Article figures, supplementary panels and source measurements © the authors, licensed under CC BY 4.0. Figures rescaled; supplementary panels cropped as indicated. Interactive visualizations are explanatory adaptations. The graphical abstract is an AI-generated conceptual illustration inspired by Figs. 1 and 4.

Inspect the source figure