On-Chip Silica Origami

• The paper demonstrates how controlled laser and capillary reflow techniques convert planar silica on silicon into complex 3D microstructures.
• It achieves ultrahigh optical quality with surface roughness as low as ~0.5 nm and Q-factors reaching up to 1.7×10^8, enabling advanced photonic devices.
• The field spans various methods—including elastocapillary folding, DNA silicification, and sacrificial-template replication—broadening design options for chip-scale silica structures.

On-chip silica origami denotes a set of fabrication paradigms in which silica structures prepared on planar substrates are transformed into three-dimensional forms while retaining the material advantages of silica, notably optical smoothness, chemical stability, and compatibility with silicon-based processing. In the most direct usage, the term refers to controlled laser-driven folding or reflow of lithographically defined silica on silicon into curved photonic elements such as whispering-gallery-mode microresonators, concave micromirrors, polylines, and helices (Ohana et al., 2024, Malhotra et al., 6 Jul 2025). Closely related literature extends the idea to elastocapillary self-folding of oxide-compatible micro-objects, silica reinforcement of DNA origami through silicification, and sacrificial-template replication of arbitrary three-dimensional hollow fused-silica geometries (Legrain et al., 2014, Ober et al., 2022, Kotz et al., 2018). Taken together, these works define a broader field concerned with turning planar or soft templates into high-quality silica-based three-dimensional microstructures on or for chips.

1. Conceptual scope and physical basis

The central premise is that silica can be reconfigured in three dimensions without forfeiting the smoothness and purity established by planar cleanroom processing. In laser-driven implementations, a CO2 laser locally heats suspended silica until viscosity drops sufficiently for surface tension to reshape the material, after which rapid cooling freezes the new geometry (Ohana et al., 2024, Malhotra et al., 6 Jul 2025). In this sense, “origami” is not a literal hinge-and-fold analogue alone; it also includes surface-tension-driven reflow of suspended silica disks into toroids or spheroids, and bending of released silica bars into knees, polylines, and helices.

The physical drivers are consistently capillarity, viscous flow, elasticity, and heat transport. For laser-folded silica bars, the mechanism is described as laser absorption in silica, rapid local heating, viscous reflow, and surface-tension-driven shape change, with the molten region reaching nearly $3000\ \mathrm{K}$ on the heated side while the cooler side is just above the glass transition temperature $(\sim 1500\ \mathrm{K})$ (Malhotra et al., 6 Jul 2025). For chip-based whispering-gallery-mode resonators, silica is heated to near the “mobility” temperature $T_{\mathrm{mobility}} \approx 2400\,^\circ\mathrm{C}$, where viscosity drops to $\approx 10^3\ \mathrm{Pa\cdot s}$ and the flow becomes surface-tension-driven (Ohana et al., 2024). In both cases, the silicon support plays an important thermal role by acting as a heat sink.

A recurring motivation is to overcome the roughness penalties of additive silica printing. The photonic-origami work explicitly contrasts folding of pre-fabricated lithographically smooth silica with additive 3D printing routes that form discrete voxels and typically yield rough surfaces that scatter light (Malhotra et al., 6 Jul 2025). The microresonator work similarly relies on CO2 reflow and surface-tension smoothing to achieve ultrahigh optical quality at visible wavelengths (Ohana et al., 2024). This suggests that the defining feature of on-chip silica origami is not merely three-dimensionality, but three-dimensionality combined with preservation or recovery of optical-grade surfaces.

2. Laser-induced folding of silica on silicon

A direct realization of on-chip silica origami is the laser-induced folding of suspended silica microbars on silicon chips into discrete and continuous three-dimensional forms (Malhotra et al., 6 Jul 2025). The fabrication begins with thermally grown amorphous silica on silicon, followed by dry XeF2 undercut to form cantilevers while leaving an anchor region. Wet etching is explicitly avoided because liquid–solid interfacial tension in wet processes can fracture ultra-thin, long sheets (Malhotra et al., 6 Jul 2025).

The laser system is a CO2 source at $11\ \mu\mathrm{m}$, focused to a $\approx 60\ \mu\mathrm{m}$ FWHM spot. A top-view microscope is bore-sighted to the laser for targeting, and a side-view microscope tracks the folding dynamics and angle. Snap-motion $90^\circ$ folding is achieved with $4\ \mathrm{ms}$ pulses at $26.5\ \mathrm{kW/cm^2}$, with net bend completed in $<1\ \mathrm{ms}$. Fine-angle control is obtained with low-power pulse trains, for example $(\sim 1500\ \mathrm{K})$0 pulses at $(\sim 1500\ \mathrm{K})$1, giving an angle resolution of $(\sim 1500\ \mathrm{K})$2 milliradian per pulse. Continuous helical folding is demonstrated by scanning at $(\sim 1500\ \mathrm{K})$3 through an intensity of $(\sim 1500\ \mathrm{K})$4 (Malhotra et al., 6 Jul 2025).

The reported dynamics are unusually fast. The molten zone has an estimated thermal time constant of $(\sim 1500\ \mathrm{K})$5, the folding can occur in less than $(\sim 1500\ \mathrm{K})$6, accelerations reach $(\sim 1500\ \mathrm{K})$7, and the tip speed is $(\sim 1500\ \mathrm{K})$8 (Malhotra et al., 6 Jul 2025). The same work reports $(\sim 1500\ \mathrm{K})$9 alignment accuracy, minimal curvature radius $T_{\mathrm{mobility}} \approx 2400\,^\circ\mathrm{C}$0, and preservation of lithographic smoothness at RMS $T_{\mathrm{mobility}} \approx 2400\,^\circ\mathrm{C}$1.

The mechanics are framed with standard capillarity and elasticity relations. At the molten interface, the Young–Laplace pressure jump is

$T_{\mathrm{mobility}} \approx 2400\,^\circ\mathrm{C}$2

and the elastic resistance of a beam is described by

$T_{\mathrm{mobility}} \approx 2400\,^\circ\mathrm{C}$3

A standard scaling argument further writes the capillary line force as $T_{\mathrm{mobility}} \approx 2400\,^\circ\mathrm{C}$4, leading to a torque balance

$T_{\mathrm{mobility}} \approx 2400\,^\circ\mathrm{C}$5

which captures the observed dependence on thickness and molten geometry (Malhotra et al., 6 Jul 2025).

A notable quantitative result is the achieved slenderness ratio $T_{\mathrm{mobility}} \approx 2400\,^\circ\mathrm{C}$6 for a $T_{\mathrm{mobility}} \approx 2400\,^\circ\mathrm{C}$7 long, $T_{\mathrm{mobility}} \approx 2400\,^\circ\mathrm{C}$8 thick silica bar (Malhotra et al., 6 Jul 2025). The paper identifies gravity as an upper bound at larger lengths or thicknesses, and self-shadowing as a practical issue in multi-bend geometries, mitigated by increasing laser power during later bends.

3. Reflowed silica microresonators and the whispering-gallery-mode regime

In a more specialized usage, on-chip silica origami describes the controlled reflow of lithographically defined suspended silica disks into ultrahigh-$T_{\mathrm{mobility}} \approx 2400\,^\circ\mathrm{C}$9 whispering-gallery-mode microresonators on silicon (Ohana et al., 2024). The process starts with a planar geometry: a silica disk of radius $\approx 10^3\ \mathrm{Pa\cdot s}$0 and thickness $\approx 10^3\ \mathrm{Pa\cdot s}$1 suspended on a silicon pillar of radius $\approx 10^3\ \mathrm{Pa\cdot s}$2. Representative values given are $\approx 10^3\ \mathrm{Pa\cdot s}$3, $\approx 10^3\ \mathrm{Pa\cdot s}$4, and $\approx 10^3\ \mathrm{Pa\cdot s}$5 or $\approx 10^3\ \mathrm{Pa\cdot s}$6 (Ohana et al., 2024).

The fabrication has two stages. First, planar microfabrication defines the suspended disk using thermal oxidation, photolithography, buffered oxide etch to produce a wedge profile, resist removal and cleaning, and isotropic SF6 plasma undercut of silicon. Second, CO2-laser reflow folds the disk into a toroid, spheroid, or cup geometry (Ohana et al., 2024). The beam waist is significantly larger than the disk so that heating is approximately uniform; for $\approx 10^3\ \mathrm{Pa\cdot s}$7 disks, a Gaussian beam divergence of $\approx 10^3\ \mathrm{Pa\cdot s}$8 is reported.

A key feature is self-limiting folding. As the disk folds inward, the illuminated cross-section shrinks, reducing heat input, while thermal dissipation through the pillar increases. Reflow terminates when net heating falls below dissipation, cooling the silica below $\approx 10^3\ \mathrm{Pa\cdot s}$9. Illumination times of $11\ \mu\mathrm{m}$0–$11\ \mu\mathrm{m}$1 are typical to reach steady state (Ohana et al., 2024). At high intensities, evaporation is non-negligible and can remove up to tens of percent of the volume, further affecting the final resonator geometry.

The thermal model is radially symmetric and defines a “radius of mobility” $11\ \mu\mathrm{m}$2 as the radial boundary where the steady-state temperature reaches $11\ \mu\mathrm{m}$3. The governing relations include

$11\ \mu\mathrm{m}$4

and an energy balance including laser heating, conduction, radiation, convection, and evaporation (Ohana et al., 2024). The evaporation rate is modeled by an Arrhenius form fitted to fused-silica measurements: $11\ \mu\mathrm{m}$5

Given $11\ \mu\mathrm{m}$6 and the evaporated volume, the mobile silica redistributes into one of three families—cup, torus, or spheroid—selected by minimal surface energy subject to mass conservation and pillar contact constraints (Ohana et al., 2024). The phase diagram shows spheroids for small $11\ \mu\mathrm{m}$7 at given $11\ \mu\mathrm{m}$8 and higher intensities, toroids at intermediate conditions, and cups otherwise, with good agreement between theory and experiment.

For whispering-gallery modes, the free spectral range is approximated by

$11\ \mu\mathrm{m}$9

with $\approx 60\ \mu\mathrm{m}$0 near $\approx 60\ \mu\mathrm{m}$1. For $\approx 60\ \mu\mathrm{m}$2, the paper gives $\approx 60\ \mu\mathrm{m}$3 (Ohana et al., 2024). The mode volume is defined by

$\approx 60\ \mu\mathrm{m}$4

The work notes that toroids typically achieve smaller $\approx 60\ \mu\mathrm{m}$5 than spheres of similar outer size because of the tight cross-section and rim localization.

The measured optical performance reaches $\approx 60\ \mu\mathrm{m}$6 at $\approx 60\ \mu\mathrm{m}$7 in air, corresponding to $\approx 60\ \mu\mathrm{m}$8 and photon lifetime $\approx 60\ \mu\mathrm{m}$9 (Ohana et al., 2024). In UHV, devices show $90^\circ$0–$90^\circ$1 at $90^\circ$2. Surface roughness measured by AFM is RMS $90^\circ$3, supporting negligible scattering loss in the reported regime. The total quality factor is expressed as

$90^\circ$4

4. Cavity-QED motivation and integrated photonic devices

The most demanding application discussed for reflowed silica origami is single-atom cavity QED (Ohana et al., 2024). The objective is simultaneous attainment of ultrahigh $90^\circ$5 and very small mode volume $90^\circ$6. High $90^\circ$7 reduces cavity decay according to

$90^\circ$8

while small $90^\circ$9 increases the vacuum field and the atom–cavity coupling rate

$4\ \mathrm{ms}$0

The cooperativity is

$4\ \mathrm{ms}$1

For the Rb D2 transition, the atomic decay rate is given as $4\ \mathrm{ms}$2, and with $4\ \mathrm{ms}$3 at $4\ \mathrm{ms}$4, the cavity linewidth $4\ \mathrm{ms}$5 lies below $4\ \mathrm{ms}$6, making strong coupling contingent on achieving sufficiently large $4\ \mathrm{ms}$7 through small $4\ \mathrm{ms}$8 (Ohana et al., 2024).

The same chip-based silica platform supports stable coupling by tapered fibers or waveguides and is compatible with UHV operation for cold atoms (Ohana et al., 2024). The loaded quality factor is described by

$4\ \mathrm{ms}$9

with critical coupling at $26.5\ \mathrm{kW/cm^2}$0. An anisotropic silicon etch can increase pillar height to ease tapered-fiber coupling while maintaining mechanical robustness.

A parallel photonic-origami route realizes other integrated optical components on the same material system. The 2025 work reports concave micromirrors formed by evaporation and convex spherical microresonators formed by local reflow (Malhotra et al., 6 Jul 2025). The concave mirror has focal length $26.5\ \mathrm{kW/cm^2}$1, aperture diameter $26.5\ \mathrm{kW/cm^2}$2, and numerical aperture $26.5\ \mathrm{kW/cm^2}$3. The spherical microresonator, measured with a tapered fiber coupler, achieves $26.5\ \mathrm{kW/cm^2}$4 at $26.5\ \mathrm{kW/cm^2}$5 (Malhotra et al., 6 Jul 2025). Interferometric mapping verifies the parabolic mirror figure, and the microresonator performance is attributed to the preserved RMS roughness of $26.5\ \mathrm{kW/cm^2}$6.

A common misconception is that three-dimensional silica folding necessarily sacrifices optical quality. The cited photonic and resonator studies indicate the opposite under controlled reflow conditions: the surface-tension process is used specifically to maintain or enhance surface smoothness, with reported RMS values of $26.5\ \mathrm{kW/cm^2}$7 in folded silica bars and $26.5\ \mathrm{kW/cm^2}$8 in reflowed WGM resonators (Malhotra et al., 6 Jul 2025, Ohana et al., 2024).

5. Elastocapillary self-folding and hinge-based transfer to silica

A distinct but related lineage comes from elastocapillary origami of micromachined plates with flexible hinges (Legrain et al., 2014). In that work, rigid silicon-nitride flaps connected by thin hinges are folded on-chip by the surface tension of a water droplet dispensed through a $26.5\ \mathrm{kW/cm^2}$9-diameter hydrophobic micro-pipette. Folding proceeds over $<1\ \mathrm{ms}$0–$<1\ \mathrm{ms}$1, can reach $<1\ \mathrm{ms}$2, and is monitored with top and side cameras while electrical continuity through Pt/Cr bi-layer hinges is measured in situ (Legrain et al., 2014).

The mechanics are summarized by the bending stiffness of a thin plate per unit width,

$<1\ \mathrm{ms}$3

the elastocapillary length

$<1\ \mathrm{ms}$4

and the curvature relation $<1\ \mathrm{ms}$5 with $<1\ \mathrm{ms}$6 (Legrain et al., 2014). The reported structures survive extremely small bending radii of $<1\ \mathrm{ms}$7, corresponding to $<1\ \mathrm{ms}$8, without loss of conductivity. Conductive yield, however, depends strongly on hinge length: $<1\ \mathrm{ms}$9 for $(\sim 1500\ \mathrm{K})$00, $(\sim 1500\ \mathrm{K})$01 for $(\sim 1500\ \mathrm{K})$02, and $(\sim 1500\ \mathrm{K})$03 for $(\sim 1500\ \mathrm{K})$04 (Legrain et al., 2014).

The same paper explicitly discusses implications for “On-Chip Silica Origami” by transferring these elastocapillary design rules from SiRN to $(\sim 1500\ \mathrm{K})$05 (Legrain et al., 2014). Using room-temperature material properties $(\sim 1500\ \mathrm{K})$06 and $(\sim 1500\ \mathrm{K})$07 for silica, it gives for $(\sim 1500\ \mathrm{K})$08 and water $(\sim 1500\ \mathrm{K})$09: $(\sim 1500\ \mathrm{K})$10 For $(\sim 1500\ \mathrm{K})$11, it estimates

$(\sim 1500\ \mathrm{K})$12

These values support hinge lengths in the $(\sim 1500\ \mathrm{K})$13–$(\sim 1500\ \mathrm{K})$14 range and target radii $(\sim 1500\ \mathrm{K})$15–$(\sim 1500\ \mathrm{K})$16 for silica-based capillary folding (Legrain et al., 2014).

The significance of this literature is methodological rather than demonstrative for silica itself. It provides a practical blueprint: thin silica hinges of approximately $(\sim 1500\ \mathrm{K})$17–$(\sim 1500\ \mathrm{K})$18, conductive hinge lengths below $(\sim 1500\ \mathrm{K})$19, stable Pt adhesion strategies on $(\sim 1500\ \mathrm{K})$20, and careful post-release cleaning to avoid metal degradation (Legrain et al., 2014). This suggests that the term on-chip silica origami can also encompass mechanically compliant, droplet-actuated architectures, not only laser-molten reflow.

6. Silicified origami and template-based fused-silica architectures

A broader extension of the topic appears in two additional directions: silicification of pre-formed origami nanostructures and sacrificial-template replication of arbitrary hollow fused-silica microstructures (Ober et al., 2022, Kotz et al., 2018).

In DNA-origami silicification, the starting object is not a planar silica layer but a DNA origami scaffold that is reinforced by silica growth. The mechanism begins with electrostatic priming of the phosphate backbone by TMAPS, followed by TEOS hydrolysis and co-condensation. In situ SAXS shows that silica forms on both outer and inner surfaces of the origami, replacing internal hydration water and driving strong condensation before later re-expansion (Ober et al., 2022). For 24-helix bundles, the outer cylinder radius changes from $(\sim 1500\ \mathrm{K})$21 to $(\sim 1500\ \mathrm{K})$22 within the first $(\sim 1500\ \mathrm{K})$23, while the interhelical spacing decreases from $(\sim 1500\ \mathrm{K})$24 to $(\sim 1500\ \mathrm{K})$25 by $(\sim 1500\ \mathrm{K})$26. At $(\sim 1500\ \mathrm{K})$27, the outer radius recovers to $(\sim 1500\ \mathrm{K})$28, but the internal lattice remains more condensed than the native state (Ober et al., 2022).

The contrast-matching analysis uses scattering length densities $(\sim 1500\ \mathrm{K})$29, $(\sim 1500\ \mathrm{K})$30, and $(\sim 1500\ \mathrm{K})$31, yielding a silica volume fraction

$(\sim 1500\ \mathrm{K})$32

Thus, more than $(\sim 1500\ \mathrm{K})$33 of internal hydration water is replaced by silica at the contrast-match point (Ober et al., 2022). The maximally condensed state is stable at $(\sim 1500\ \mathrm{K})$34 for $(\sim 1500\ \mathrm{K})$35, whereas the bare structures melt at $(\sim 1500\ \mathrm{K})$36. The upper bound for the outer shell thickness is $(\sim 1500\ \mathrm{K})$37 (Ober et al., 2022). For flat DNA origami, aggregation during silicification is pronounced, which the authors attribute to the same entropic forces that cause condensation.

Sacrificial-template replication addresses a different scale and geometry class. In that method, a polymer template is fully embedded in a UV-curable silica nanocomposite, followed by thermal debinding at $(\sim 1500\ \mathrm{K})$38 in air and sintering at $(\sim 1500\ \mathrm{K})$39 in vacuum $(\sim 1500\ \mathrm{K})$40 with a $(\sim 1500\ \mathrm{K})$41 heating rate to produce full-density, transparent fused silica (Kotz et al., 2018). Template fabrication can use two-photon polymerization, melt electrowriting, PEGDA microlithography, or nylon threads. The resulting structures include DNA double helices, intertwined spirals, out-of-plane mixers, inverse micromeshes, and enclosed channels with cross-sections as small as $(\sim 1500\ \mathrm{K})$42 and lengths in the centimeter range (Kotz et al., 2018).

This template route is not a folding method in the strict sense, but it contributes an important adjacent interpretation of silica origami: arbitrary three-dimensional silica geometries produced from a sacrificial precursor. The internal surface roughness is reported as mean roughness $(\sim 1500\ \mathrm{K})$43 for DLW-templated channels, and the authors state that “surfaces of optical quality are therefore achievable” (Kotz et al., 2018). The method avoids HF etching, taper, and debris blockage, and it supports circular, triangular, trapezoidal, and rectangular channel profiles with aspect ratios from $(\sim 1500\ \mathrm{K})$44 to $(\sim 1500\ \mathrm{K})$45.

7. Design trade-offs, comparisons, and current limitations

Across the cited literature, several trade-offs recur. In resonator origami, shrinking the toroid minor radius lowers mode volume but increases curvature-induced radiation loss, reducing $(\sim 1500\ \mathrm{K})$46 (Ohana et al., 2024). Evaporation assists volume reduction but can also cause excessive shrinkage or geometry drift. In photonic bar folding, thinner bars are easier to bend because $(\sim 1500\ \mathrm{K})$47, but below $(\sim 1500\ \mathrm{K})$48, residual stress in thermally grown silica can induce unwanted bending (Malhotra et al., 6 Jul 2025). In capillary hinge-based folding, longer conductive hinges reduce electrical yield even if the folding mechanics remain viable (Legrain et al., 2014). In silicified DNA origami, stronger silica growth improves robustness but also drives re-expansion and, for flat objects, aggregation (Ober et al., 2022).

The different approaches also occupy distinct application regimes.

Approach	Characteristic mechanism	Representative outcome
CO2 reflow of suspended silica disks	Surface-tension-driven folding and evaporation	Toroids, spheroids, cups; $(\sim 1500\ \mathrm{K})$49 at $(\sim 1500\ \mathrm{K})$50 (Ohana et al., 2024)
Laser folding of silica bars	Local liquefaction and capillary bending	Polylines, helices, concave mirrors, spheres; $(\sim 1500\ \mathrm{K})$51 alignment accuracy (Malhotra et al., 6 Jul 2025)
Elastocapillary hinge folding	Water-droplet actuation of flexible hinges	$(\sim 1500\ \mathrm{K})$52 rotation, $(\sim 1500\ \mathrm{K})$53; silica-transfer design rules (Legrain et al., 2014)
DNA origami silicification	Internal and external silica growth from TMAPS/TEOS	$(\sim 1500\ \mathrm{K})$54 size reduction, $(\sim 1500\ \mathrm{K})$55 internal water replacement (Ober et al., 2022)
Sacrificial-template replication	Burn-out of polymer template in silica nanocomposite	Enclosed fused-silica channels down to $(\sim 1500\ \mathrm{K})$56 (Kotz et al., 2018)

One controversy implicit in the literature concerns what should properly count as “origami.” The laser-folding and microresonator papers use the term directly for surface-tension-driven shape change of silica on silicon (Malhotra et al., 6 Jul 2025, Ohana et al., 2024). The capillary-hinge, DNA-silicification, and sacrificial-template papers connect to the topic by extension, offering mechanisms and process rules for silica-based three-dimensional self-assembly or transfer (Legrain et al., 2014, Ober et al., 2022, Kotz et al., 2018). This suggests that the field is currently heterogeneous: the phrase identifies a common design ambition—high-quality three-dimensional silica on chip—more reliably than a single universal mechanism.

The broader implication is that on-chip silica origami has become a convergence point between ultrahigh-$(\sim 1500\ \mathrm{K})$57 photonics, micro-optomechanics, capillarity-driven microsystems, and silica-enabled structural reinforcement. The most mature optical demonstrations emphasize the combination of chip integration, ultrasmooth surfaces, and three-dimensional geometry, while adjacent approaches expand the accessible design space toward compliant hinges, biomolecular templates, and arbitrary hollow fused-silica networks (Ohana et al., 2024, Malhotra et al., 6 Jul 2025, Legrain et al., 2014, Ober et al., 2022, Kotz et al., 2018).