Text from the interactive explanation
Seamless Mechanical Programming Through Hybrid Epoxy–Acrylate Printing for Soft–Rigid Robots
David C Bershadsky, Franz A Stolpen, Zachariah A Page, Lillian Chin
Original paperSoft robots · 3D printing with light
The only difference is where violet and ultraviolet light touched the liquid during printing.
All soft → stretches
Stiff stripe → bends
Stiff spiral → twists
Simplified simulation · real printed actuators further down
Nature joins hard and soft without seams. A turtle’s shell grows smoothly into its body. Locusts store energy in graded structures, and centipedes walk on bodies that mix soft and rigid parts. Stiffness changes gradually, inside one continuous material.
Our machines mostly do the opposite. We bond separate pieces together: a rubber O‑ring on a plastic casing, a rubber grip molded over a handle, a soft gripper glued to a rigid mount. The junction is where stress piles up, and in a soft robot that flexes thousands of times, it is where things tear.
Could we print robots that are soft and hard in one continuous piece, the way bodies are?
A team at the University of Texas at Austin did it with one vat of liquid and two colors of light. Here is how, step by step.
1 · The seam problem
Pull it until it breaks
To test a junction, you pull on it. Each test piece is a sandwich: stiff, soft, stiff. Like the samples in the paper, it is 8 mm wide, 24 mm long and 1.5 mm thick. Clamp the stiff ends and stretch. If the junction is the weak point, the strip tears there long before the soft middle reaches its own limit.
Choose a printing method, then pull a sample.
Where your samples broke (0 samples)
Average stretch compared with soft alone: –
Measured in the paper: interface 12.3% · bulk 87.7% · strain ratio 95.9% (73 samples)
Outcomes are drawn at random from the paper’s measured failure rates; the animation is illustrative. The inkjet rates are pooled across the various Vero blends the team printed, so changing the blend changes the picture, not the odds; the blend percentages shown are illustrative. Strain ratio: how far a multi-material strip stretched before breaking, divided by how far a uniform strip of the same soft material stretches.
In the real tests, only 12% of the HEAP (one-pot resin) strips broke at the junction. On average they stretched 96% as far as a uniform strip of the same soft material, across 73 samples. The team even printed a deliberate notch right at the interface, yet the strips still tore through the soft bulk instead.
Filament prints failed at the junction every time and stretched only about 1% as far. Inkjet printers make an in-between hardness by stippling: the whole soft region is a mosaic of soft and hard droplets mixed in different ratios, so every droplet boundary is a tiny internal seam. Those samples broke mostly at the junction, and some cracked through the stippled droplets.
Hardness range each method could reach (Shore D)
Figure 2 from Bershadsky et al., Adv. Mater. 2026 (CC BY-NC-ND 4.0).
2 · One resin, two lights
Two kinds of hooks on one molecule
The resin is mostly one molecule, ECA , and it has two reactive ends: an acrylate hook (circle) and an epoxy hook (square). The liquid also contains two light-triggered starters.
Violet light (405 nm) starts only the acrylate reaction. Acrylate ends link into long chains while the epoxy squares dangle unused, so the result is a soft rubber. Ultraviolet light (365 nm) starts that reaction too, and also releases an acid that stitches epoxy squares to each other. The dense cross-links lock the network into a hard plastic.
Wider settings fade the projected gray level linearly from all UV to all violet across the middle.
Split: UV on the left, violet on the right. Acrylate chains grow straight across the middle, so the two regions form one network.
Illustration: molecules, photons and linking rules are simplified to show the idea.
UV 2.0 mW/cm² · violet 18.0 mW/cm²
Measured trend, redrawn approximately from Figure 2A (printed test cubes). Top axis: violet intensity; bottom axis: UV intensity.
Shore A 46 · like a pencil eraser
The everyday objects are only a rough comparison.
Each projector pixel gets a gray level N between 0 and 255, which sets how much of the exposure time its tiny mirror stays switched on. The paper writes the doses as D365 = (N/255)·I365·t365 and D405 = ((255−N)/255)·I405·t405, where I is intensity and t is exposure time. Each printer was calibrated experimentally.
The key point: the acrylate chains run through soft and hard regions alike. A soft-to-hard boundary is the same network, more or less cross-linked, so there is no seam. Count the bonds that cross the middle in the split view, widen the interface to make a gradient, then try “Glued.”
3 · The printer
Two projectors, one vat
A resin printer projects a picture into the bottom of a vat to harden one thin slice, lifts the part, and repeats. This one has two projectors, at 365 nm and 405 nm, and a dichroic mirror merges their beams into one path. Every slice gets two pictures: a UV mask for whatever should be rigid, and a violet mask for everything else.
Schematic with time compressed; the hand is stylized. On the large printer each 50 µm layer got about 3 s of UV and 1 s of violet light, at up to 20 mW/cm².
4 · Program motion
The Bellows Lab
A bellow is a corrugated tube: pump air in and it lengthens. A soft wall stretches; a stiff wall barely does. So where you put the stiff material decides how the tube moves. Paint your own pattern by dragging across the tube’s unrolled skin below, or start from a preset, and watch the 3D tube respond.
Paint the tube’s skin
Pick a brush below: UV for rigid, violet for soft.
Drag across the skin to paint your own pattern.
Watch the tube bend, twist or stretch.
at full pressure · now: 0% pressure
Stretches straight: the soft wall lengthens evenly all the way around.
Bend it past 90°
Twist it at least 90° while bending less than 30°
Keep it straight but stretch it less than half as much as the all-soft tube
Simplified, illustrative model: each patch stretches according to how stiff you painted it. Uneven stretch around the tube bends it toward the stiffer side, and a stiff band that winds around the tube unwinds as the soft wall lengthens, which twists the tip. Constants are tuned so the stripe preset reaches the paper’s measured 137° bend and the spiral its roughly 140° twist.
With light, the spiral is just a different picture on the projector. The tube’s shape never changes; only the stiffness map inside it does.
In a pressurized tube, the stress running around the wall (hoop stress, σh = P·r/t) is twice the stress along it (σl = P·r/2t). So when the bellows failed, they split vertically within the soft wall, and wider tubes burst at lower pressure: 112 ± 6 kPa for 5 mm bellows versus 56 ± 5 kPa for 8 mm ones. None broke at the junction with the rigid barb.
Shorter corrugation periods and larger amplitudes gave more stretch. As a middle ground, the team chose an 8 mm diameter, 6 mm period and 1 mm amplitude, the proportions used in this lab.
5 · Robots that need both
Printed in one piece, tested for real
Every robot below came out of the printer as one part, with soft and rigid regions already joined. The movies are the team’s recordings, and the numbers are measured.
Endurance
0 cycles
The counter animation is decorative.
Five optimized bellows were inflated 10,000 times each while a camera tracked their length (the graph in the movie). Four survived the full test with minimal drift. The fifth failed at a thin spot in its soft wall, left over from removing print supports. None showed damage at a soft–rigid interface.
Movie S13 (motion-capture excerpt)
A finger with a thinner base
50% thinner base
A bending finger needs a base that resists stretching. Stiffening it (Shore A 64 / D 34, against Shore A 8 for the soft parts) let the team cut it from 4 mm to 2 mm thick, with comparable bending at 21 kPa.
Movie S7
A hand with bones
5 independent fingers
A rigid palm holds the air-tube connections. Each soft finger hides stiff, bone-like plates that confine bending to the joints, much as finger bones do.
Movie S9
A three-axis platform
1 print, 3 bellows
Three bellows stand between rigid plates, and the lower plate has printed connectors for the air hoses. The platform squats, extends and tilts. The optimized design reached a considerably larger workspace than the first version.
Movie S6
A walker with toenails
2.3× faster walking
This 8.9 cm walker has five zones and a rigid tower with barbs for its air tubes. Rigid internal “armour plates” let leftover liquid resin be flushed out without puncturing the thin walls. The upgraded version adds rigid toe tips, like fingernails, and directional friction ridges.
The soft-legged version did well on smooth felt but lost ground on rough sandpaper. With toe tips and ridges, the walker gripped instead: on fine 3000-grit sandpaper the paper reports 2.3× higher displacement per gait than the soft-legged baseline. A comparison walker printed entirely from a commercial flexible resin barely deformed at the same pressure; at 41.4 kPa it moved modestly and ruptured after several gait cycles.
Movie S11
With toe tips and ridges: the back foot grips, so each cycle nets forward.
Illustration of the ratchet idea, not a simulation of the real walker.
6 · Back to the opening question
Built more like a body
A turtle doesn’t glue its shell on. This work brings machines a step closer to that. Stiffness no longer comes from shape alone: it is painted in, pixel by pixel, during a single print, using standard multi-wavelength DLP printing hardware, with no adhesives and no swapping between resins.
The authors point to wearable assistive devices, soft grippers for fragile objects, and terrain-adaptive or bio-inspired walkers. They also mention mounting electronics, compliant energy storage and programmable metamaterials. Other pairs of chemistries, such as thiols, isocyanates or vinyl ethers, could work the same way if their light-triggered reactions can be kept selective.
One challenge remains. Long-term stability currently relies on soaking prints in solvent after printing, which can be costly and creates solvent waste. Greener solvents, less solvent and shorter washes are next on the list.
Source & credits
David C. Bershadsky, Franz A. Stolpen, Zachariah A. Page, Lillian Chin. “Seamless Mechanical Programming Through Hybrid Epoxy–Acrylate Printing for Soft–Rigid Robots.” Advanced Materials 38(63), e75168 (2026). doi:10.1002/adma.75168
Movies S1, S6, S7, S8, S9, S11, S13 and Figures 1–5 from Bershadsky et al., Adv. Mater. 2026, e75168, © 2026 The Authors, published by Wiley-VCH, licensed under CC BY-NC-ND 4.0. Movies re-encoded for web playback (resized, audio removed; Movie S13 shortened to a 40-second excerpt); no other changes.
What is measured and what is illustrated
Simulations and illustrations: the three tubes at the top, the tensile-test animation, the molecule vat, the printer schematic, the Bellows Lab model, the cycle counter and the walker ratchet.
Measured data: the failure rates and strain ratios used in the tensile lab, the hardness trend (redrawn approximately from Figure 2A), the hardness ranges, all numbers on the robot cards, and the movies.
Figures from Bershadsky et al., Adv. Mater. 2026 (CC BY-NC-ND 4.0), unmodified. Figure 2 appears in section 1.
3D graphics use three.js (MIT license, © 2010–2025 three.js authors).
Topics in this iPaper
Engineering
This capability enables single‐shot fabrication of soft–rigid robotic systems that undergo large, cyclic deformations
Abstract
Chemistry
wavelength selective radical and cationic photocuring using (3,4‐epoxycyclohexyl)methyl acrylate (ECA, 71.66 wt%) as the hybrid monomer
Section 4.1, Hybrid Epoxy–Acrylate Printing (HEAP)
soft robotics
highlighting local hardness programming as a powerful strategy for next generation soft robotics
Section 1, Introduction
digital light processing
Selective activation of the two networks is achieved via a DLP 3D printer with two independent DLP projectors
Section 4.1
functionally graded materials
biological systems rely on functionally graded materials to distribute stress and optimize performance
Abstract
multi material additive manufacturing
multi‐material additive manufacturing, one‐pot DLP printing, soft robotics
Keywords
photopolymerization
the BAPO photoinitiator triggers a radical polymerization of the acrylate network
Section 4.1
pneumatic actuators
HEAP was next applied to fabricate single‐shot pneumatic bellows actuators for soft robotic motion
Section 2.2, Soft–Rigid Actuators With HEAP