Processing soft thin films on liquid surface for seamless creation of on-liquid walkable devices

Abstract

Walking on liquid surface is a unique locomotion ability of insects, but engineering on-liquid walkable devices currently requires disjointed, multistage fabrication and delicate deployment. Here, we introduce HydroSpread—a direct processing technology that enables seamless fabrication and patterning of soft films on liquid surface. It leverages the controlled spreading of liquid ink on liquid surface and combines with precise laser engraving supported by rapid heat transfer at the solid-liquid interface. Geometric shapes, including basic forms of straight lines, sharp turns and circles, and complex patterns, were fabricated with exceptional fidelity to design specifications. We propose two heat-driven hydrodynamic locomotion mechanisms, fin-like bending and leg-like buckling. By harnessing these principles, we engineered two walkable devices—HydroFlexor and HydroBuckler—and demonstrated robust on-water locomotion. This work eliminates fragile postfabrication transfers in soft device manufacturing, bridging the gap between soft films and structure fabrication, and establishes a streamlined pathway for designing and deploying functional soft devices directly in liquid environments.

HydroSpread technology enables fabrication and laser patterning of soft films on liquid surfaces for on-liquid devices.

INTRODUCTION

Engineering on-liquid locomotion devices holds immense potential for widespread applications in liquid quality monitoring and aquatic search and rescue. The fundamental ability of floating on liquid requires device structures with low body mass, and soft thin films are rapidly emerging as a promising solution. Their minimal thickness and high mechanical flexibility enables excellent compatibility with liquid fluidity, making them ideal for soft on-liquid robots. Soft films have already been widely used in skin-like electronics (1–3), flexible sensors (4, 5), and soft robots (6, 7). These films are typically fabricated on solid substrates such as silicon or glass wafer using methods such as spin coating (8, 9), dip coating (10, 11), printing (12, 13), and electrospinning (14, 15), followed by mechanical transfer to target substrates (16, 17). Although great successes have been achieved in minimizing the fabrication thickness of soft films, challenges remain in transferring thin soft films without damage. The delamination energy required to detach films from the fabrication substrates is often comparable to or lower than the mechanical deformation energy of soft films, making them prone to damage and breakage (18, 19). Given the nature of application aquatic environments, fabricating soft thin films directly on liquid surface provides a straightforward way that allows seamless creation of on-liquid locomotion devices. Liquid substrates inherently provide an ultrasmooth surface (20, 21), enabling uniform spreading of soft polymeric ink to form films with exceptional thickness uniformity and ultralow surface roughness, surpassing the conventional solid substrate- based fabrication techniques.

Here, we introduce a HydroSpread fabrication process—a cost-efficient fabrication approach on liquid solution surface for creating and laser-patterning polymeric soft thin films directly on liquid—and present its enabled seamless production of on-liquid walkable locomotion devices. We elucidate the spreading dynamic mechanism of polymeric soft inks on immiscible liquid solution surface and prove the experimental feasibility of fabricating geometrically thin, uniform soft films with ultralow surface roughness, in remarkable agreement with theoretical predictions. We demonstrate the high-fidelity and -resolution laser-patterning of diverse geometric structures, from elementary shapes of strips, circles, and sharp turns to complex patterns with fine local features and also eludicate the rapid thermal transport across the solid-liqud interface through finite element analysis (FEA). By applying a global heating stimulus, we experimentally show, together with confirmations by theory and FEA, two distinct deformation modes in the bilayer soft composite films on liquid surface: bending and buckling. Harnessing these film-liquid interaction mechanisms, we fabricated two on-water walkable devices—HydroFlexor and HydroBuckler—and demonstrated their controllable bending- and buckling-driven locomotion. This approach introduced here is compatible with a wide range of polymeric and composite inks and liquid substrate for a breadth of geometric shapes required for on-liquid devices. Its synergy with existing transfer printing technologies such as water transfer will offer a scalable platform for multifunctional soft electronics and systems, beyond on-liquid devices.

RESULTS

Processing of soft thin films on water surface and mechanics mechanism

Figure 1A (i) shows the optical images of a 5-μl polydimethylsiloxane (PDMS) ink droplet (PDMS 184 with 10:1 ratio of base to curing agent unless other stated, dyed for enhanced contrast) spreading on a water surface. Upon deposition at room temperature, a fast and homogeneous spreading behavior is observed along with a faded color change and is driven by its positive spreading coefficient on water (note S1). After ~10 min, the spreading arrives at equilibrium, forming a stable, continuous thin film that floats on water surface after full curing (~8 hours), which is referred to as HydroSpread process. To quantitatively characterize the spreading dynamics, we developed a laser reflection method to track the radial spreading of ink droplets, where the laser line projected onto the interface exhibits a discontinuity at the boundary between the dyed ink PDMS film and the water substrate (fig. S1 and movie S1). The same measurement radii across all edges confirms the uniform, axisymmetric spreading, yielding a circular film geometry on water surface. The peeling test (fig. S2A) reveals weak interfacial adhesion of the cured PDMS film on water surface (~0.05 J/m² independent of film’s thickness, fully dominated by van der Waal interaction), enabling its fast peeling without damage, as shown in Fig. 1A (ii). The structural continuity of the PDMS film is tested by a mechanical puncturing test, where a sharp needle is punctured onto the as-produced PDMS film on water [Fig. 1A, (iii)]. The water substrate’s fluidity distributes stess laterally, delaying the onset of PDMS film’s failure and enhancing the effective failure stress (fig. S2B). When lifted into air, the unsupported film collapses because of thin, soft, and compliant structure of nature [Fig. 1A, (iv)]. In addition, we examine the pixel change of a smile emoji pattern projected through the PDMS film using optical laser profilometry (fig. S3). Nearly the same pattern without pixel distortion is obtained, further confirming the thickness uniformity and homogeneity in the formed PDMS film on water surface.

Fig. 1. HydroSpread fabrication and characterization of soft thin films on water solution surface.

(A) Spreading dynamics and film formation of PDMS ink on water surface. (i) Optical images of a 5-μl PDMS ink droplet (PDMS 184 with 10:1 ratio of base to curing agent, dyed for visibility) on water surface showing a fast and complete spreading to equilibrium at room temperature for forming a continuous thin film. (ii) Peeling test on the cured PDMS film from water surface. (iii) Structural integrity test by puncturing the cured PDMS film on water with a sharp needle. (iv) Uplifting test of free-standing cured PDMS film in air with a steel rod. (B) Comparison of spreading dynamics of different ink droplets on various immiscible water solution surfaces between experimental measurements and theoretical predictions. (C) Thickness (h; plots, left) and surface roughness (3D optical profilometry images, right) characterization of the PDMS films fabricated on various water ethanol solutions. (D) Comparison of thickness and surface roughness of the PDMS films formed by our work of HydroSpread technique and by conventional solid-substrate techniques [spin coating (47–51), casting (52–55), inkjet printing (56–59), electrospinning (60–63), doctor blading (64, 65), spraying printing (51, 66–70), dip coating (71–73), and screen printing (74–77)]. Symbols denote materials: □, PDMS 184; △, Poly(3,4-ethylenedioxythiophene) (PEDOT); ○, Polytetrafluoroethylene (PTFE); and ◇, polymethyl methacrylate (PMMA).

Figure 1B plots the spreading radius for different ink droplets on water solution surface counterparts. Clearly, all systems exhibit fast initial spreading, followed by equilibrium to a steady state radius, consistent well with observations in Fig. 1A (i). In theory, the spontaneous spreading of droplet on an immiscible liquid surface arises from the competition between the system free energy of both ink droplet and ink-liquid substrate interface ( $G$ ) (22, 23) and the viscous flow of ink droplet–induced dissipation energy ( $\mathrm{\Phi }$ ) (24, 25). For an ink droplet volume ( $V$ ) on a liquid surface, the spreading radius ( $r$ ) can be determined by balancing the system free energy release and the viscous energy dissipation, i.e., $-\frac{\mathit{dG}}{\mathit{dt}}=\mathrm{\Phi }$ , and it yields (note S2)

$$S+\frac{1}{2}\mathrm{\Delta }\mathrm{\rho }\mathit{gh}{\left(t\right)}^2+\frac{A}{2\mathrm{\pi }h{\left(t\right)}^2}-\frac{\mathrm{\tau }}{r_f}=0$$ (1)

where $S$ , $\mathrm{\Delta \rho }$, $h\left(t\right)$, $\mathrm{\tau }$ , and $r_f$ are the spreading coefficient, effective density difference between the droplet and liquid substrate (26), maximum height at the spreading time $t$ , contact line tension, and forefront spreading radius, respectively, while $g$ and $A$ are the gravitational constant and effective Hamaker constant (27), respectively. Figure 1B shows remarkable agreements of ink spreading radius between experimental measurements and theoretical predictions, suggesting the robustness of tracking spreading dynamics by the laser reflection method. The formation of thin film relies on spreading ability of ink droplets on liquid surface, and the theoretical diagram of fabricating soft thin films is summarized in fig. S4A. With the same spreading coefficient $S$ , unlike solid substrate, the liquid surface eliminates pinning effects due to inherent smoothness, enabling homogeneous spreading of ink precursor. Besides, the low-friction liquid-liquid interface facilitates precise control over film thickness, both resulting in exceptional thickness uniformity and ultralow surface roughness, as shown in fig. S4D.

To quantify the thickness and surface roughness, we transferred the cured thin films on a silicon substrate using a pull-up transfer method (28). Figure 1C shows the thickness profiles of the PDMS films measured by three-dimensional (3D) optical profilometry. Consistent thickness across scan distance confirms the thickness uniformity, which echoes with the homogenous spreading dynamics of ink droplets. The local fluctuations (maximum ±1.2 nm) reflect the film surface roughness, which remains barely changed for different films’ thicknesses, as evidenced by 3D surface morphology profiles in Fig. 1C (right). Similar results are found for films on other water/ethanol solution surfaces (fig. S5). In addition, given the same ink droplet, reducing ethanol concentrations in the water solution substrate increases the spreading coefficients and results in smaller radii (or higher film thickness) of films, which also agrees very well with theory (fig. S5A). This tenability suggests a straightforward way of tailoring ink spreading and final film dimension by adjusting liquid-substrate interactions. This spreading mechanism of an ink droplet on a liquid surface enables the spontaneous formation of soft films with synchronized ultrathin thickness and ultralow surface roughness, achieving a record-high performance compared with those prepared on solid substrates by conventional fabrication technologies, as summarized in Fig. 1D.

Laser engraving of structures and patterns on liquid

Laser engraving is considered a power technique for etching materials with intricate patterns and structures but is challenged by achieving fine feature resolutions due to localized overheating and uneven thermal dissipation on solid substrates (29, 30). Fabricating soft films by HydroSpread process substantially enhances the heat transfer because of the unique solid-liquid interface, enabling direct laser patterning with high resolution and fidelity. Figure 2A compares the laser-patterned PDMS stripes with varying widths $\left(W\right)$ on both liquid and solid substrates. The stripes engraved on the liquid substrate precisely match the design specifications. In contrast, those on the solid substrate show clear deviations, and even the strip with $W$ = 0.2 mm fails to appear after lase engraving [fig. S6A (i)]. Similar results are observed when engraving the PDMS films into solid circular shape or polyline with a sharp corner. In particular, on solid substrates, the circle with 0.3 mm in radius and the polyline with 2° in corner angle cannot be patterned [Fig. 2, B and C, and fig. S6A (ii and iii)]. Considering a triangular structure, Fig. 2D further examines the scalability of laser engraving on liquid and solid substrates. As the engraving height decreases from 4 to 1 cm, the pattern on the liquid substrate continues to closely follow as-design dimensions in great agreement. In contrast, notable deviations are observed on the solid substrate, and even when the height is 1 cm, the triangular structure cannot be successfully patterned. The edge in patterns may not remain ideally flat in contact with liquid because of localized capillary effects when the engraving characteristic length is smaller than the capillary length (31). FEA results in Fig. 2E show that at a 15-W engraving laser power, the temperature distribution on the solid substrate covers a much larger area than the one on the liquid substrate, where the heat is highly localized to the laser-engraving region. This is further supported by the temperature variation data in Fig. 2F (top), which show a more rapid temperature drop at the interface of the PDMS film on the liquid substrate compared to the solid substrate. In addition, a steeper in-plane temperature gradient (Fig. 2F, bottom) corroborates the faster heat dissipation on the liquid substrate. Similar observations at 20-W engraving laser power are obtained, as detailed in fig. S7C. Figure 2G demonstrates the synchronous and synergetic capability of engraving complex patterns of “UVA” letter logo on a liquid substrate, where the logo is composed of intricate features with various straight and curvilinear segments and sharp angles. The engraved logo exhibits clear, smooth cutting boundaries, in stark contrast to the rough, serrated edges obtained on solid substrates. These findings are also confirmed on other printed complex patterns (fig. S8 and movie S2). Local key transition locations (#1, 2, and 3) highlight these differences, indicating very clear deviations from the design, particularly in regions with high curvatures or sharp transitions. Dimensional measurements, including length ( $d_1$ ), curve radius ( $R_1$ ), and angle ( ${\mathrm{\theta }}_1$ ), are presented in Fig. 2H (top). These results confirm that patterning on the liquid substrate consistently achieves feature dimensions as designs, while clear discrepancies are found on the solid substrate. Similar results are obtained on other complex patterns in fig. S9 (A and B). Furthermore, analysis of line edge roughness further verifies high-resolution and high-fidelity patterning with direct laser engraving on PDMS films on the liquid substrate, as shown in Fig. 2H (bottom) and figs. S6B and S9C.

Fig. 2. Direct high-resolution laser patterning of geometric shapes on liquid-supported thin films.

Characteristic length and resolution to elementary geometries on PDMS films fabricated via HydroSpread (liquid substrate) versus conventional solid substrates for (A) strip with width $W$ , (B) solid circles with radius $R$ , and (C) polyline with a sharp corner angle $\mathrm{\theta }$ . (D) Scalability of patterning triangular structures with fixed apex angle ( $\mathrm{\theta }$ = 10°) and variable height ( $H$ ). Symbol $\otimes$ denotes patterning failure. (E) FEA of temperature distribution in film/substrate at 15-W engraving laser power with side view (top) and top view (bottom), where σ denotes the radius of the laser-heating zone. (F) Temperature decay along the in-plane z direction (top) and out-of-plane y direction (bottom) at the engraving site. (G) Complex patterning (UVA letters) and edge fidelity and resolutions at critical transitions (highlighted in red lines in insets). (H) Quantitative measurement on patterning feature dimensions of the segment length ( $d_1$ ), curve radius ( $R_1$ ), and angle ( ${\mathrm{\theta }}_1$).

Deformation models of soft films on liquid surface

When a bilayer composite film with different thermal expansion coefficients is formed on a liquid substrate by HydroSpread method (see Materials and Methods), as illustrated in Fig. 3A, by applying a heating, the mechanical deformation on liquid surface relies on the liquid surface tension–induced constraints at the contact lines and can be controlled by adjusting the geometric dimensions and/or mechanical and thermal properties of layers. As shown in Fig. 3A (top right), the free end of the composite strip could overcome the liquid surface tension and penetrates liquid, thereby exhibiting a mechanical bending mode. On the other hand, if the liquid constraint at the free end exceeds the adhesion between the liquid and the composite film, the heating-induced thermal mismatch could lead to delamination at the liquid-film interface, while the free end remains constrained, equivalent to a rolling support, and a buckling deformation mode occurs, as shown in Fig. 3A (bottom right). By analyzing the deformation energy (32–35), we have developed a mechanics mode to distinguish these two deformation modes via the critical bending angle ( ${\mathrm{\beta }}_{\text{cr}}$ ) at the free end, as illustrated in Fig. 3A (middle), and it is (note S2)

$$\begin{array}{c}{\mathrm{\beta }}_{\text{cr}}=\frac{\frac{2D{\mathrm{\pi }}^2}{\mathit{bL}}+4{\mathrm{\gamma }}_{\text{fb}}L}{F_m}\end{array}$$ (2)

where $F_m$ denotes the deformation mismatch force of thermal expansion and geometric eccentricity between the top and bottom layers, which induces the thermal moment $M_{\mathrm{\Delta }T}$ as illustrated in Fig. 3A (middle). The parameters $D$ and ${\mathrm{\gamma }}_{\text{fb}}$ represent the bilayer composite film’s flexural rigidity and the interface tension between the film and liquid substrate, respectively. The term $4{\mathrm{\gamma }}_{\text{fb}}L$ contributes to the interfacial moment $M_{\mathit{int}}$ (Fig. 3A, middle) and together with the term $2D{\mathrm{\pi }}^2/\mathit{bL}$ prevents the composite film’s out-of-plane deformation. When the bending angle $\mathrm{\beta }$ is larger than the ${\mathrm{\beta }}_{\text{cr}}$ , the bending deformation mode is expected. For example, for a PDMS/PDMS-5% multiwalled carbon nanotubes (MWCNTs) bilayer film, Eq. 2 yields ${\mathrm{\beta }}_{\text{cr}}$ = 0.082, well consistent with the perturbation-induced beam’s bending and buckling deformation (36). Figure 3B presents a theoretical phase diagram mapping the bending and buckling deformation modes as a function of the geometric dimensions of composite film and component layers (width-to-length ratio, $b/L$ , and the thickness ratio, $h_a/h_b$ ) and mechanical property of layers (modulus ratio, $E_a/E_b$ ). The theoretical predictions closely match experimental measurements on both deformation modes, as shown in fig. S10. In parallel, we conducted FEA of bilayer composite films on the liquid (see Materials and Methods). Figure 3C shows quantitative comparison of both bending and buckling deformations among experimental, theoretical, and FEA results. Both the bending angle ( $\mathrm{\beta }$ ) and the amplitude-to-buckled length ratio ( $\mathrm{\delta }/\mathrm{\lambda }$ ) increase nonlinearly with the heating temperature change ( $\mathrm{\Delta }T$ ), which agrees well with both theoretical predictions and FEA results. This agreement remains when the geometric dimensions of composite film and component layers vary (fig. S10). The deformation snapshots at $\mathrm{\Delta }T$ = 70°C (Fig. 3C, right, and movie S3) and other heating temperatures (figs. S11 and S12) further confirm the excellent agreement between experimental observations and FEA results on these two distinct deformation modes.

Fig. 3. Deformation modes and mechanisms of bilayer composite stripes on liquid surface.

(A) Schematics of the PDMS/PDMS-MWCNTs bilayer composite strip on liquid surface (left), the heating ( $\mathrm{\Delta }T$)–induced mechanics model (middle), and the resultant bending or buckling deformation (right), where one end of the bilayer is clamped and the other is free to deform. $b$ , $L$ (= 40 mm), $h_a$ , and $h_b$ (= 0.01 mm) are the width and length of the composite stripe, thickness of the top layer (PDMS-MWCNTs), and the bottom layer (PDMS), respectively. ${\mathrm{\alpha }}_a,{\mathrm{\alpha }}_{\mathit{b}}$ , and $E_a,E_b$ (=1.1 MPa) are the respective thermal expansion coefficient and Young’s modulus. $M_{\mathrm{\Delta }T}$ and $M_{\mathit{int}}$ reflects the bending moments induced by the bilayer thermal mismatch and solid-liquid interaction, respectively. $\mathrm{\beta }$ is the bending angle, and $\mathrm{\delta }$ and $\mathrm{\lambda }$ are the amplitude and length of buckling. (B) Theoretical phase diagram of bending and buckling deformation modes. Data points (symbols) represent experimental measurements. (C) Comparison of mechanical deformation among experimental measurements, theoretical predictions, and FEA results for $b/L$ = 0.05 (bending mode) and 0.15 (buckling mode), with $h_a/h_b$ = 6.0 and $E_a/E_b$ = 2.6 (left), and snapshots of experimental and FEA deformation at $\mathrm{\Delta }T$ = 70°C (right). (D) Optical images of the bending deformation and recovery during heating-cooling. (E) Evolution of the bending angle $\mathrm{\beta }$ over single heating cycle (left) and multiple heating cycles (right). (F) Optical images of the buckling deformation and recovery during heating-cooling. (G) Evolution of the amplitude-to-buckled length ratio $\mathrm{\delta }/\mathrm{\lambda }$ over single heating cycle (left) and multiple heating cycles (right).

Figure 3D (left) depicts the evolution of the bending angle ( $\mathrm{\beta }$ ) during a single heating cycle. The snapshots show a gradual increase in $\mathrm{\beta }$ as the temperature rises, followed by a full recovery to the original configuration as the temperature decreases back, suggesting the reversible nature of the bending deformation in soft films. No noticeable deformation degradation is observed after multiple cycles of heating (Fig. 3D, right), confirming consistent and repeatable bending behavior in response to cyclic heating. The buckling deformation was also analyzed under cyclic heating conditions, as shown in Fig. 3F. The snapshots capture well the evolution of the amplitude-to-buckled length ratio ( $\mathrm{\delta }/\mathrm{\lambda }$ ) during a single heating cycle, highlighting the onset and relaxation of buckling deformation. Figure 3G presents the corresponding quantitative results of $\mathrm{\delta }/\mathrm{\lambda }$ over both a single (left) and multiple heating cycles (right). The stable $\mathrm{\delta }/\mathrm{\lambda }$ values without clear degradation over multiple thermal cycles confirm the robustness and reversibility of buckling deformation on the liquid surface, similar to that of bending deformation in Fig. 3E.

Applications in on-liquid walkable devices and locomotion

With the elucidation of thermally driven deformation modes of bilayer composite films on liquid surface, we designed two prototypes to demonstrate the locomotion on water surface. In these devices, PDMS and PDMS-5% MWCNT inks were continuously used to form bilayer films on water surface by the HydroSpread method, where their deformation can be actively powered upon a global heating stimulus in air. Figure 4A shows the geometric design with a fin-like propulsion mechanism. The device body serves as a boundary constraint for the two side fins through the large contact area–enhanced adhesion on liquid surface, while the free end of both fins has a smaller contact area and lower adhesion and can deform freely. By applying a heating stimulus, the free end of fins could bend and penetrate water. Once immersed, the heating energy will be rapidly dissipated, leading to bending deformation recovery of fins back to their original configurations at the water surface. Figure 4B (top) shows optical images of three experimental layouts after laser patterning the bilayer films. In addition to bending, the triangular shape of fins will introduce a bending stiffness gradient along their vertical cross section and results in a torque along the longitudinal axis, thereby leading to a twisting deformation. FEA (Fig. 4B, bottom) confirms this integrated bending-twisting deformation. This coordinated deformation enables the fins to alternately penetrate and withdraw from water in response to cyclic heating stimulus, producing a flapping hydrodynamic force to propel the device on water, analogous to a paddling mechanism (37, 38), which we refer to as HydroFlexor locomotion. With asymmetric design of fins’ dimensions in the body, the bending-twisting modes could be adjusted to control both the direction and speed of locomotion. For example, with the symmetric design in design 2, where a balanced propelling force is maintained in the body during the deformation-recovery cycle, the HydroFlexor follows a straight-line trajectory at $\mathrm{\Delta }T$ = 70°C (Fig. 4C and movie S4). The experimental trajectory closely matches the as-designed path indicated in the red dash line. This great agreement across three different orientations further confirms the accuracy and repeatability of the locomotion mode [Fig. 4C, (ii)]. Moreover, as the heating temperature $\left(\mathrm{\Delta }T\right)$ increases, the amplitude and recovery frequency of the fins’ deformation are enhanced, leading to an increased locomotion speed $\left(v\right)$ , as observed in Fig. 4C (iii). This observation is also consistent with the increased bending angles at elevated temperatures in Fig. 3C. When the fins are designed asymmetrically, Fig. 4D demonstrates the turning behavior. A clockwise turning trajectory [Fig. 4D, (i)] is observed with the left fin shorter than the right one ( $L_1<L_2$ , design 1 in Fig. 4B), while a counterclockwise turning motion (fig. S13A and movie S4) is generated with $L_1>L_2$ (design 3 in Fig. 4A). In both cases, the experimental trajectories show remarkable agreement with as-designs indicated by blue dashed lines, as shown in [Fig. 4D (ii)]. Besides, the turning radius ( $R_t$ ) decreases with the increasing of $\mathrm{\Delta }T$ [Fig. 4D, (iii)], indicating an enhanced maneuverability at elevated temperatures, which aligns well with the enhanced out-of-plane twist deformation in longer fins shown by FEA in Fig. 4B. When the thickness ratio $h_a/h_b$ of bilayer films decreases, similar locomotion results are also obtained, as shown in fig. S14, where the locomotion speed decreases because of the reduced bending deformation–induced propulsion forces.

Fig. 4. Application demonstrations of on-liquid walkable devices.

(A to D) HydroFlexor: Bending-driven locomotion. (A) Schematics of the HydroFlexor working mechanism: heating-induced fin-like bending and twisting in PDMS/PDMS-5% MWCNTs bilayer wings, generating hydrodynamic propulsion. (B) Experimental images of fabricated HydroFlexor with three designs (top) and FEA snapshots on bending deformation at $\mathrm{\Delta }T$ = 70°C (bottom). $b/L_1=b/L_2$ = 0.5, $h_a/h_b$ = 6.0, and $E_a/E_b$ = 2.6. (C) Straight-line locomotion with design 2: (i) optical images of trajectory, (ii) comparison of trajectories between experimental measurements and as-designs in three directions, (iii) locomotion speed with $\mathrm{\Delta }T$ . (D) Turning performance with designs 1 and 3: (i) Experimental trajectory snapshots of design 3, (ii) trajectory position in response to different heating stimuli, (iii) turning radius ( $R_t$ ) with $\mathrm{\Delta }T$ . (E and F) HydroBuckler: Buckling-driven walking-like locomotion. (E) Schematics of the HydroBuckler working mechanism: heating-triggered buckling of leg-like PDMS-PDMS 5%-MWCNTs bilayer wings, generating propulsion. (F) Experimental images of fabricated HydroBuckler devices with three designs (top) and FEA snapshots of buckling deformation at $\mathrm{\Delta }T$ = 70°C (bottom). $b/L_4=b/L_6$ = 0.1, $L_3$ = 0.9$L_4$ , $h_a/h_b$ = 6.0, and $E_a/E_b$ = 2.6. (G) Straight-line walking-like locomotion with design 2: (i) optical images of trajectory, (ii) comparison of trajectories between experimental measurements and as-designs in three directions, (iii) locomotion speed with $\mathrm{\Delta }T$ . (H) Turning performance with designs 1 and 3: (i) Experimental trajectory of snapshots of designs 3, (ii) trajectory position in response to different heating stimuli, (iii) turning radius ( $R_t$ ) with $\mathrm{\Delta }T$.

Unlike the fin-based HydroFlexor (bending) locomotion, Fig. 4E introduces a buckling locomotion design by incorporating multiple thermally responsive, slim, leg-like features around the device’s body. The leg lengths and numbers are designed in a manner inspired by the structural features of water strider of Aquarius remigis and are also guided by the deformation phase diagram in Fig. 3B. Three experimental layouts with different leg lengths were fabricated and are shown in Fig. 4F, where the bilayer structures are PDMS/PDMS-5% MWCNTs, the same as those in Hydroflexor devices. Similar to the deformation observations in Fig. 3 (F and G), by applying a heating cycle, the legs undergo a buckling-recovery deformation on water surface, generating a walking-like motion. The direction and speed of locomotion depend on the symmetry of the buckling deformation in the legs. FEA on these three designs (Fig. 4F, bottom) shows that the buckling in the short legs, especially the front pairs, can be negligible, which helps maintain the device’s stability during locomotion, while the pronounced buckling in the long rear legs generates a forward force against the water surface. As the temperature decreases, these long legs recover to their originals, leading to the propulsion force for locomotion, which we refer to as HydroBuckler locomotion. Figure 4G (i) (and movie S5) demonstrates a straight-line walking trajectory for the device with symmetrically distributed legs of equal lengths (design 2), which ensures the same buckling deformation of legs on both sides. Figure 4G (ii) shows the experimental measurement of trajectories at $\mathrm{\Delta }T$ = 70°C, which closely matches the as-design. The walking direction is determined by the initially positioned orientation of HydroBuckler devices. As $\mathrm{\Delta }T$ increases, the walking speed $\left(v\right)$ rises [Fig. 4G, (iii)], similar to that in the Hydroflexor locomotion in Fig. 4C. Further, take the speed of 0.6 cm/s, the drag force is 1.4 μN, consistent with caterpillar–bioinspired soft robotics (39). By introducing asymmetric leg designs, Fig. 4H shows the turning behavior in the HydroBuckler devices. The device with longer left legs (design 3) results in a clockwise turning trajectory [Fig. 4H, (i), and movie S5], while the one with shorter left legs (design 1), a counterclockwise turning motion is observed (fig. S13B and movie S5). Figure 4H (ii) summarizes experimental trajectory measurements. The remarkable agreement with as-design predictions demonstrates buckling deformation–induced locomotion and efficacy of control. Similar to the HydroFlexor, the turning radius ( $R_t$ ) decreases with increasing $\mathrm{\Delta }T$ [Fig. 4H, (iii)], enabling sharp turns at elevated temperatures. In addition, when the thickness ratio $h_a/h_b$ of bilayer films in the HydroBuckler devices is reduced, the propulsion force reduces because of decreased buckling deformation, leading to a lower locomotion speed, as shown in fig. S15.

DISCUSSION

In summary, we present a HydroSpread process for fabricating soft thin films directly on liquid surfaces, seamlessly integrating film formation, patterning, and device creation all on liquid surface for on-liquid locomotion applications. The spreading dynamics of soft polymeric inks on a liquid surface is established to reveal and control the formation of ultrathin, uniform films. Heat dissipation at the solid-liquid interface is conducted to elucidate the underpinned laser-based high-fidelity patterning mechanism of geometric structures with fine local features. Thermally responsive bilayer composite films were created and, upon heating, depending on geometric dimensions of films, exhibit two distinct deformation modes—bending and buckling—on liquid surface, with remarkable agreement among experiment, theory and FEA. As an application demonstration, we designed and fabricated two bioinspired locomotion devices—HydroFlexor and HydroBuckler—by harnessing the respective bending and buckling deformation–induced propulsion on water. Both on-water locomotion modes including directional control and speed modulation are uncovered and show remarkable agreement with as designs. Compared to conventional on-water soft robotic systems that are typically manufactured on the basis of solid substrate and subsequent deployment to liquid, the entire robot body, including functional bilayer fins or wings, is directly printed and engraved on a liquid surface, enabling ultraconformal assembly and contact without mechanical delivery or alignment, and achieving the distinct, unique locomotion modes. Future optimization to these device designs is expected to provide enhanced levels of robotic performance such as speeding and responsive time. When different responsive materials sensitive to other stimuli such as electric, magnetic, or optical field are used, the controlling means to the on-liquid robotic devices could be improved for accuracy. Moreover, this cost-effective, scalable approach is compatible with a wide range of polymetric and composite inks, paving the way for broad applications in fabricating soft electronics, sensors, and actuators.

MATERIALS AND METHODS

Fabrication of films on water solution surface

PDMS Sylgard 184 (PDMS 184 or PDMS unless stated otherwise) or SE 1700 (PDMS 1700) was used at a 10:1 weight ratio of basis to curing agent. Silc-Pig blue dye (1.5 wt %; Smooth-On Inc.) was added to the PDMS solution to enhance visualization during the film’s spreading dynamics and final thickness measurement. MWCNTs (8 to 15 nm in diameter, 10 to 50 μm in length, 95%; NanoAmor Inc.) were used in the fabrication of PDMS-MWCNTs composite films. PDMS was first mixed and then followed by the addition of MWCNTs. Hexane (Thermo Fisher Scientific Inc.) at a 1:5 weight ratio was added into PDMS base solution to achieve high concentrations of MWCNTs in mixture. The PDMS-MWCNTs mixture was placed in an ultrasonicator for at least 12 hours with 40 kHz to ensure homogeneous dispersion of the MWCNTs. Afterward, the PDMS curing agent was added, and the mixture was degassed for ~2 hours before being deposited onto the water solution surface. A 25 cm–by–25 cm square feet tank filled with water solution served as an infinite liquid substrate, providing a free surface for the spreading of the ink solution droplet. A 5-μl droplet of the prepared ink solution was carefully deposited onto the water solution surface using a mechanical single-channel pipette (Statlab Inc.).

We define the equilibrium spreading state as the point at which the spreading radius $r_f$ increases by no more than 0.5% over a 1-min interval. The curing process begins immediately after equilibrium is reached and proceeds at room temperature for at least 8 hours to ensure full crosslinking. On the basis of standard PDMS protocols with base to curing agent of 10:1, this ensures that the resulting film is fully solidified without further spreading. The final difference in $r_f$ after curing is less than 1%. To create PDMS/PDMS-MWCNTs bilayer composite films, a partially cured PDMS film was formed initially on water surface, and then an ink droplet of PDMS-MWCNTs solution was placed on top to create a strongly bounded bilayer film that would not delaminate under external heating. The thickness of films was controlled by adjusting the volume of ink droplets. While water and water-ethanol mixtures used in this work are chemically inert and do not induce any observable contamination to the formed films atop, caution should be exercised when complex polymer liquid such as PDMS base or organic solvents are used as liquid substrate. These liquids may result in chemical diffusion, swelling, or uptake of low–molecular-weight species into the formed film during ink spreading. In such cases, additional protective strategies such as surface pre–cross-linking (40) or interfacial compatibilization (41) may be required to preserve the formed film’s purity and surface functionality.

FEA on laser pattering–induced heat transfer at interface

To simulate the thermodynamic response during laser engraving of PDMS films on liquid (water) and solid (silicon wafer) substrates, FEA was conducted using ANSYS software (ANSYS Inc.). The PDMS film with a side length of 5 cm and a thickness of 0.5 mm was modeled on either a liquid (water) or a solid (silicon wafer) substrate that were treated as semi-infinite domains to minimize boundary effects. The material properties were assigned as follows: for PDMS film, density: 970 kg/m³, thermal conductivity: 0.15 W/(m·K), and specific heat capacity: 1460 J/(kg·K). For water substrate, density: 997 kg/m³, thermal conductivity: 0.6 W/(m·K), and specific heat capacity: 4182 J/(kg·K); for silicon wafer, density: 2330 kg/m³, thermal conductivity: 148 W/(m·K), and specific heat capacity: 700 J/(kg·K). A localized heat flux ( $q$ ) was applied to the PDMS surface in the region of 5 mm in radius, and it is correlated in simulations with the laser power P (=15, 20, and 25 W) via $q=P/\mathrm{\pi }{c}^2$ . For PDMS film/water system, a thermal contact conductance of 3200 W/(m²·K) was defined at the PDMS-water interface, alongside a convection boundary condition with a heat transfer coefficient of 380 W/(m²·K). For PDMS film/silicon wafer system, thermal contact conductance [3000 W/(m²·K)] without convection cooling was considered because no natural convection occurs at the solid-solid interface, and heat conduction dominates the thermal exchange process (42, 43). The top surface of the PDMS film was exposed to air with negligible heat loss. A fine mesh density in the laser heating region (5-mm radius) with the total number of elements of 250,000 was used to capture localized heat distribution. The transient thermal analysis used a Newton-Raphson iterative solver to handle nonlinearities, with large deformation effects disabled.

FEA on mechanical deformation of bilayer composite stripes on liquid surface

FEA was conducted to investigate the thermally induced mechanical deformation of bilayer composite films on a liquid surface. The bilayer composite film with a side length ( $L$ ) of 4 cm was composed of a 0.06-mm-thick PDMS-5% MWCNTs layer (layer A) and a 0.01-mm-thick PDMS layer (layer B). The material properties of PDMS and PDMS-5% MWCNTs are as follows: for the PDMS, the density: 970 kg/m³, elastic modulus: 1.1 MPa, thermal conductivity: 0.15 W/(m·K), and coefficient of thermal expansion (CTE): 300 × 10⁻⁶/K (44), and for the PDMS-5% MWCNT composites, the density: 980 kg/m³, elastic modulus: 2.9 MPa, thermal conductivity: 0.30 W/(m·K), and the CTE: 300 × 10⁻⁶/K (45), the same as PDMS as MWCNTs primarily influence thermal conductivity rather than expansion characteristics. The simulation was performed using the ANSYS Mechanical APDL (Workbench) Thermal-Structural Coupling Module. One edge of the bilayer film was fully constrained, modeling the clamped boundary condition in experiments. The remaining edges were free to deform in response to heating and solid-liquid interaction. The water liquid surface was modeled as a 0.01-mm-thick virtual layer with an elastic modulus of 0.1 MPa. This virtual layer could model the low stiffness of the water surface while ensuring stable contact interactions and accurate force transmission. The interaction between the film and the water surface was modeled using frictionless contact in the tangential direction, allowing the film to slide freely while maintaining normal adhesion. An adhesive contact was defined with a maximum normal traction of 50 Pa to simulate surface tension effects (46), and the contact formulation was configured to allow separation and reattachment using the Augmented Lagrange method. These settings allow to provide realistic normal support meanwhile not introducing excessive computational complexity and cost. Upon a heating temperature ΔT, the convective heat loss to the surrounding environment was assumed negligible. The resultant thermal expansion by the transient thermal transport was solved using the Newton-Raphson iterative solver to capture nonlinear geometric deformations in the structures. The computational domain was discretized using HEX20 quadratic hexahedral elements, with a refined mesh in high-curvature regions near the fixed and free edges, while a coarser mesh was used elsewhere to optimize computational efficiency.

On-liquid walkable locomotion experiments

Three prototype designs were developed for each locomotion mode (HydroFlexor and HydroBuckler devices). The PDMS/PDMS-5%MWCNTs bilayer films were prepared using the HydroSpread fabrication approach above and, after full cure, were directly engraved to the as-design patterns for both HydroFlexor and HydroBuckler devices at the 15-W laser power. For the bending-induced locomotion (HydroFlexor device), the overall length ratio of both fins ( $L_1∕L_2$ ) were set to 0.5, 1.0, and 2.0. The length gradient of each fin was taken the same 0.5, which ensures the bending deformation upon heating. For the buckling-induced locomotion (HydroBuckler device), the overall length ratio of both legs ( $L_3∕L_5$ and $L_4∕L_6$ ) were set to 0.6, 1.0, and 1.66. The length gradient of each fin was taken the same 0.1. The movement and trajectories of both locomotion modes on water surface were recorded using the high-speed camera. A noncontact infrared heating plate (IHS Inc.) was positioned 10 cm above the model to enable rapid and spatially uniform thermal stimulation. To determine the actual surface temperature of the film, a calibrated spot infrared thermometer (Fluke Inc.) was used to measure the temperature at the surface of top layer film and to ensure the exaction of temperature. To ensure continuous movement, the heater was moved in a synchronization manner with the devices to maintain a sustained heating actuation. The recorded videos were analyzed through image analysis using MATLAB to extract deformation characteristics such as bending angle ( $\mathrm{\theta }$ ) or the buckling length ( $\mathrm{\lambda }$ ) and magnitude ( $\mathrm{\delta }$ ), track trajectory positions, and calculate the parameters such as walkable speed and turning radius. To estimate the propulsion force generated by the bilayer film during locomotion, we considered the viscous shear drag as the dominant resistance, given by $F_{\text{drag}}\approx {\mathrm{\mu }}_{\mathrm{w}}·\frac{A_f·v}{\mathrm{\delta }}$ , where we took ${\mathrm{\mu }}_{\mathrm{w}}=1.0\times {10}^{-3}\text{Pa}·\mathrm{s}$ (dynamic viscosity of water), $A_f=4.2\times {10}^{-4}{\mathrm{m}}^2$ (area of the film), $v=6\times {10}^{-3}\mathrm{m}/\mathrm{s}$ , and $\mathrm{\delta }\sim \frac{L}{\sqrt{\text{Re}}}=1.8\times {10}^{-3}\mathrm{m}$ (estimated boundary layer thickness). This gives a drag force of approximately 1.4 μN and which serves as a lower-bound estimate of the actuation force required for propulsion.

Supplementary Materials

The PDF file includes:

Notes S1 and S2

Figs. S1 to S15

Table S1

Legends for movies S1 to S5

Other Supplementary Material for this manuscript includes the following:

Movies S1 to S5