Connected three-dimensional polyhedral frames for programmable liquid processing

CPF designThe CPFs are wettable to the processed liquid (Fig. 1a,b). When the frames are slowly lifted out of the liquid, the capillary force can overcome gravity if the frame size is within the capillary length of the processed liquid, enabling the frames to capture the liquid. Interesting phenomena emerge when the frames are connected in different ways. Taking water as an example, when lifting the CPFs from water, single-rod connected frames capture and retain water (Fig. 1c and Supplementary Video 1); however, water drains from double-rod connected frames (Fig. 1d and Supplementary Video 1). Therefore, frames above the single-rod connection function as a capturer (the frame unit that captures and retains liquids), while the frames above the double-rod connection serve as a releaser (the frame unit imbibes but releases liquids). This phenomenon can be extended to a 3D array of CPFs (Fig. 1e,f and Supplementary Video 1). The last row of frames in Fig. 1f serves as the capturers; the captured liquid can be completely released by adding double-rod structures as shown in Supplementary Fig. 1. Furthermore, the steerable liquids can be extended from aqueous solutions to biologically relevant fluids, hydrogels, organic solvents, polymer solutions and silicone oils by designing the size of the polyhedral frame (PF) within the capillary length of the corresponding liquid (Extended Data Fig. 1 and Supplementary Fig. 2). CPFs printed with other materials, including stainless steel, biocompatible polylactic acid (PLA) and polyurethane (PU), all showed the capture and release of corresponding liquids (Supplementary Fig. 3). These findings show the versatility of CPFs to different liquids and frame materials.Fig. 1: The CPFs.a,b, Schematic diagram of the CPFs. When the CPFs are slowly lifted out of the liquid, the single-rod-connected capturers capture and retain liquid (a), while double-rod-connected releasers release their imbibed liquid (b). c, Liquid can be captured and retained as the capturers are lifted from the liquid. d, The releasers release their liquid when they are lifted from the liquid. e, The 3D liquid array is prepared using a capturer array. f, The 3D releaser array still releases the liquid. The liquid in c–f is blue-dyed water. Scale bars, 2 mm.CPF capture and release mechanismsUnderstanding the mechanism of liquid capture by the CPFs is straightforward: when they are lifted from the liquid, the liquid-philic frames provide a capillary driving force to capture the liquid, while the single rod between the frames acts only as a mechanical connection without providing a pathway for liquid drainage between frames10. Understanding their mode of liquid release, however, requires experiments, numerical simulations and analysis. Taking the cubic PF as an example (Fig. 2a), we analyze the release process in detail (Fig. 2b and Supplementary Video 2). When the CPFs are slowly lifted from the liquid, a liquid film forms between the double-rod connection and, thus, connects the liquid in the frame with the bulk liquid below. Thereafter, when the liquid in the frame rises to a critical height, gravity-dominated liquid drainage occurs in the frame. We tracked the evolution of liquid height (h), shown as h–t (height–time) curves in Fig. 2c. Figure 2c indicates that h varies periodically with t between a maximum (hmax) and a minimum height (hmin).Fig. 2: The working mechanism of liquid capture–release by CPFs.a, Structural diagram of the releaser. a and r are the side length and rod radius of the PFs, respectively, and d is the diameter of the circumscribed sphere of the PFs. b, The release process of releasers. During this process, a liquid film is formed between the double-rod connection. Scale bar, 1 mm. c, When the releasers are slowly lifted from the liquid, the liquid height (h) in the CPFs varies periodically with time between a maximum (hmax) and minimum (hmin). d, The h reaches its maximum when the curvature radius of the rising meniscus in the frame becomes minimum (Rmin). e, Three-dimensional schematic diagram of Rmin and the critical state of liquid rupture when two adjacent menisci with Rmin contact. f,g, The key factors that affect the value of Rmin for releasers with different geometries include the radius of the inscribed circle of the polygonal faces (X), the dihedral angle of the polyhedral frame (α), the contact angle of liquid on the frame surface (θ), the polyhedral side length (a) and the frame rod radius (r). h, Experimentally measured and theoretically calculated \(\frac{{h}_{\max }}{a}\) of releasers with different geometry and size parameter \(\frac{r}{a}\) (data are presented as mean ± s.d. of n = 3 replicates). i, By adjusting the ratio of hmax over d, the number of liquid-filled PFs above the bulk liquid can be controlled. j, Three stages of the release process, the height-dominated release (stage I), free fall release (stage II) and film release (stage III). The water is dyed blue to facilitate visualization. In i and j, the scale bars are 2 mm.Source dataTo understand the liquid-height changes in rising CPFs, an analogy can be drawn to a tube with a nonuniform cross-section (Supplementary Discussion 1). As shown through numerical simulations of capillary rise in uniform and nonuniform capillary tubes (Supplementary Video 3), when a nonuniform tube is slowly lifted from the liquid, the height h of the capillary rise first gradually increases to its maximum hmax and then suddenly decreases; hmax appears where the curvature radius R of the rising meniscus reaches its minimum11. This simulation corresponds to one cycle of the periodic changes of liquid height (h) in the CPFs. Therefore, as the CPFs are pulled out of the liquid (Fig. 2b1,b2), a liquid film forms between the double-rod connection and, thus, connects the liquid in the frame with the bulk liquid below. The pressure difference (PB − PA in Fig. 2d, with PB and PA being the pressure at points B and A, respectively) is generated by the rising liquid in the frame and equals ρgh. As h increases, the liquid in the frame shrinks inward, decreasing R in supporting the greater liquid weight. Meanwhile, the decrease in liquid volume shows the release (Fig. 2b2,b3). When the liquid shrinks to its critical state, the state at which the adjacent menisci contact, the curvature radius R of the corresponding meniscus reaches its minimum Rmin (Fig. 2d,e) and the liquid height h reaches its maximum hmax (Fig. 2b3). Subsequently, the liquid ruptures in the frame, h returns to hmin (Fig. 2b4) and the process repeats in the next frame. The lifting process of the CPFs from water into air was simulated by a pseudopotential two-phase lattice Boltzmann model. Extended Data Fig. 2 shows the simulated liquid-pressure changes in the CPFs during the frame rising process (Supplementary Video 4), further confirming the working mechanism of the CPFs that serve as the capturer and the releaser, respectively.By analyzing the static equilibrium at the critical state (Fig. 2d,e), hmax can be readily obtained. At this state, the liquid pressure at point A (PA; Fig. 2d) is PB − ρghmax from hydrostatic analysis and Pa − 2σ/Rmin from the application of the Laplace equation to the meniscus. Here, ρ, σ and g are the liquid density, surface tension and gravitational acceleration, respectively. Pa is the ambient air pressure (a constant), and PB is the liquid pressure at Point B (Fig. 2d) that equals Pa. Therefore,$${h}_{\max =}\frac{2\sigma }{\rho g{R}_{\min }},$$
(1)
which indicates that, for a given liquid, hmax is mainly determined by Rmin and is independent of the connecting rod’s parameters. This is experimentally verified: specifically, hmax keeps constant when changing the length of the connecting rod, and different ways of double-rod connection induce only a small difference in hmax (Supplementary Fig. 4).The theoretical general expression of Rmin reads as follows for releasers with different geometries and size parameters (Supplementary Discussion 2):$$\begin{array}{l}{R}_{\min }=\\\frac{X}{\sin \frac{\alpha }{2}}+r\cos \theta {\cot }^{2}\frac{\alpha }{2}-\sqrt{{r}^{2}{\cot }^{4}\frac{\alpha }{2}({\cos }^{2}\theta +1)+\frac{2{rX}\cos \theta }{\sin \frac{\alpha }{2}}{\cot }^{2}\frac{\alpha }{2}},\end{array}$$
(2)
where X is the radius of the inscribed circle of the polygonal faces of diverse PFs as shown in Fig. 2f, α is the dihedral angle of the polyhedral frame and θ is the contact angle of the frame surface with the liquid. a and r are the polyhedral side length and frame rod radius, respectively. The general expression of Rmin is derived from the geometric relationship between the liquid film and the frame at the critical state when the adjacent meniscus of the minimum curvature radius of the releaser contacts on the bisector of the corresponding polyhedral dihedral angle (Fig. 2g). The detailed derivation and analysis process is included in Supplementary Discussion 2. Nondimensionalizing Rmin with a yields$$\begin{array}{l}\frac{{R}_{\min }}{a}=\frac{{C}_{1}}{\sin \frac{\alpha }{2}}\\\qquad\quad+\frac{r}{a}\cos \theta {\cot }^{2}\frac{\alpha }{2}-\sqrt{{\left(\frac{r}{a}\right)}^{2}{\cot }^{4}\frac{\alpha }{2}({\cos}^{2}\theta +1)+\frac{r}{a}\cos \theta \frac{2{C}_{1}{\cot }^{2}\frac{\alpha }{2}}{\sin \frac{\alpha }{2}}},\end{array}$$
(3)
where C1 is a constant determined by the geometry of the releaser which is defined by X = C1a (the fixed relationship between X and side length a), and their values are listed in Fig. 2f. The specific dimensionless expressions of Rmin are derived and summarized in Supplementary Discussion 2 for releasers with different geometries by substituting the values of the geometric parameters C1 and α into equation (3).Substituting equation (3) into equation (1) yields$$\begin{array}{l}\frac{{h}_{\max }}{a}=\frac{2\sigma }{\rho g\frac{{R}_{\min }}{a}}\times \frac{1}{{a}^{2}}\\\qquad=\frac{2\sigma }{\rho g\left[\frac{{C}_{1}}{\sin \frac{\alpha }{2}}+\frac{r}{a}\cos \theta {\cot }^{2}\frac{\alpha }{2}-\sqrt{{\left(\frac{r}{a}\right)}^{2}{\cot }^{4}\frac{\alpha }{2}({\cos}^{2}\theta +1)+\frac{r}{a}\cos \theta \frac{2{C}_{1}{\cot }^{2}\frac{\alpha }{2}}{\sin \frac{\alpha }{2}}}\right]}\times \frac{1}{{a}^{2}}.\end{array}$$
(4)
To verify the validity of equation (4), we compared the theoretically calculated and experimentally measured \(\frac{{h}_{\max }}{a}-\frac{r}{a}\) curves for releasers with different geometries. In the experiment (Extended Data Fig. 3), the contact angle of water on the solid surface of the releasers is 65°, so θ is taken as 65° for the theoretical calculations. As shown in Fig. 2h, the experimental results are in good agreement with the theoretical calculations, supporting that hmax can be controlled by designing the frame’s shape and size. By adjusting the ratio of hmax over d (the diameter of the circumscribed sphere of the polyhed ral frame), the number of liquid-filled frame units above the bulk liquid can be programmed as shown in Fig. 2i, and hmax < d is the condition for no liquid capture.Based on equation (3), we can isolate the effect of the contact angle θ and size parameter \(\scriptstyle\frac{r}{a}\) on \(\scriptstyle\frac{{R}_{\min }}{a}\). The typical \(\scriptstyle\frac{{R}_{\min }}{a}-\cos \theta\) curves and the \(\scriptstyle\frac{{R}_{\min }}{a}-\frac{r}{a}\) curves are included in Supplementary Discussion 2, showing that \(\scriptstyle\frac{{R}_{\min }}{a}\) increases with the increase of θ, which is associated with the decrease of the frame liquid-philicity, and \(\scriptstyle\frac{{R}_{\min }}{a}\) decreases with the increase of \(\scriptstyle\frac{r}{a}\) owing to the increase in solid surface area and capillarity. Both the \(\scriptstyle\frac{{R}_{\min }}{a}-\cos \theta\) and the \(\scriptstyle\frac{{R}_{\min }}{a}-\frac{r}{a}\) curves shed light on how the geometry of the releaser affects its \(\frac{{R}_{\min }}{a}\): under the same contact angle θ and size parameter \(\frac{r}{a}\), the releaser with the icosahedron geometry provides the minimum \(\frac{{R}_{\min }}{a}\), followed by the cube and dodecahedron; the releaser with geometry of the truncated icosahedron has the maximum \(\frac{{R}_{\min }}{a}\). Using numerical simulation and experiments, we corroborate the effect of the frame surface wettability on hmax (Extended Data Fig. 4). Both show that hmax gradually decreases with an increase in contact angle (~5–85°).Next, we studied the release process in the frame that consists of three stages (Fig. 2j and Supplementary Discussion 3). Stage I starts from the beginning and ends at h = hmax, the liquid in the frame gradually shrinks inward and drains downward. The releasing volume flow rate at this stage (QI) is defined as \(\scriptstyle{Q}_{{\rm{I}}}=-{F}^{{\prime} }\left(h\right)\frac{{{\mathrm{d}}h}}{{{\mathrm{d}}t}}\), where F(h) is the function of liquid volume at different heights in the frame. Because this release stage is mainly governed by the rising height, it is called the ‘height-dominated release’. Stage II refers to the moment when the liquid ruptures and falls rapidly (<0.6 s). After the liquid ruptures, the meniscus in the frame flattens instantly, making the Laplace pressure difference across the meniscus close to zero. Without the support of the Laplace pressure difference, the liquid falls almost in a free fall12. Therefore, stage II is named ‘free-fall release’, and the releasing volume flow rate (QII) is given as \({Q}_{{\rm{II}}}=\sqrt{\frac{g}{2{h}_{\max }}}F({h}_{\max })\). Experiments have confirmed that release stages I and II can release most of the liquid in the frame, but their release proportions vary in frames with different geometries (Supplementary Video 5). For example, stage I occupies a larger release proportion in cubic and pentahedral frames (Fig. 2b and Supplementary Fig. 5), while stage II contributes more to icosahedral and truncated icosahedral frames (Supplementary Fig. 5). Stage III is the final release process until the liquid film between the double-rod connection ruptures. At this stage, the remaining liquid releases through the liquid film, which is a viscous flow process driven by the pressure gradient in the film13,14. Hence, stage III is named the ‘film release’ stage.Switching between capture and releaseThe release mechanism shows that the liquid continuity constructed by the liquid film between the double-rod connection is the key to triggering liquid drainage, which inspired us to achieve switching between liquid capture and release by establishing or breaking this liquid continuity. Destroying the liquid continuity can switch release into capture; conversely, the establishment of liquid continuity will convert capture into release (Supplementary Fig. 6a). We demonstrated two approaches to breaking the liquid continuity by destroying the liquid film between the double-rod connection, that is, directly piercing the liquid film with a swab or imposing a very small lifting speed to promote spontaneous liquid film rupture induced by evaporation (Supplementary Video 6). Likewise, two simple methods for establishing liquid continuity are provided. One way is to create droplet-to-droplet contact, which can release the captured liquid (Supplementary Video 6). Alternatively, a soap film is used to connect the dispersed liquids in the capturers, which also releases the captured liquid (Supplementary Video 6). To completely release the liquid, the size of the soap film needs to meet the condition as illustrated in Supplementary Fig. 6b; when a soap film with longitudinal length L contacts the capturer, the distance from the contact point to the bottom of the film (L1) should be larger than hmax (Supplementary Video 7). Following this principle, soap films with predesigned patterns are applied to release liquids at designated locations to obtain designed liquid patterns (Supplementary Fig. 6c and Supplementary Video 7). In addition, the liquid in the capturer array can be fully released by the soap film array (Supplementary Fig. 6d and Supplementary Video 7), which renders a facile and efficient method for recovering liquids from CPFs. The function (liquid capture or release) of each CPF is readily switched back after being switched as well simply by triggering the ‘liquid continuity change’, a feature termed as ‘reversible’ (Supplementary Table 1c). The capability to reversibly capture and release liquids will potentially benefit many applications including biological sampling, medical diagnostics and chemical synthesis.Programmable 3D liquid patterning and applicationsProgrammable 3D patterning of liquids is of importance for the 3D spatiotemporal control of biological and chemical processes15,16 but still remains challenging with current fluidic devices. Here, we overcome this limitation with switchable capturing and releasing CPFs, achieving programmable 3D patterning of liquids. The basic coding principle of the programmable liquid patterning is shown in Fig. 3a,b, that is, releasers above the double-rod connection are denoted as ‘0’, and capturers above the single-rod connection are denoted as ‘1’. By designing the frame connections, the 3D spatial arrangement of capturers and releasers can be programmed to acquire the desired liquid patterns (Supplementary Video 8). To prepare the two-dimensional (2D) liquid pattern, the pattern is first disassembled into digital codes of ‘1’ and ‘0’ pixel by pixel, that is, pixels in the target pattern area are coded as ‘1’ and pixels in the blank background area are coded as ‘0’. Based on the digital codes, we can deduce the spatial arrangement of single-rod and double-rod connections between the frames, and the expected liquid pattern can be obtained after printing the predesigned model. Two-dimensional letter patterns of ‘U’, ‘W’ and ‘O’ are successfully produced using preprogrammed CPFs (Fig. 3c and Supplementary Video 8), as an example. This principle can readily be exploited to achieve 3D liquid patterning by first disassembling 3D patterns into 2D patterns layer by layer. Building on this concept, we prepared hourglass-like and other predesigned 3D liquid patterns as exhibited in Fig. 3d and Fig. 3e, respectively (Supplementary Video 8).Fig. 3: Programmable 3D liquid patterning.a, Coding principle for programmable 3D liquid patterning: releasers above the double-rod connection are denoted as ‘0’, and capturers above the single-rod connection are denoted as ‘1’. b, The corresponding liquid patterns obtained based on the coding principle. Scale bar, 1 mm. c, Two-dimensional letter patterns of ‘U’, ‘W’ and ‘O’ created using preprogrammed CPFs. d,e, The hourglass-like (d) and other predesigned (e) 3D liquid patterns fabricated with preprogrammed CPF arrays. f, Basic principles of programmable 3D binary liquid patterning. g, Using 3D alternating CPFs to sequentially capture red-dyed gellan gum solution and blue-dyed water, the 3D binary liquid pattern with alternating red and blue pixels was prepared. h, Vitamins B2 and B12 were patterned in predesigned CPFs consisting of a truncated icosahedron frame and 34 cubic frames, where vitamin B12 is encapsulated in the truncated icosahedron frame and vitamin B2 is captured in the cubic frames. i, Demonstration of the release process of vitamins B2 and B12 in water. j, Corresponding release profiles of vitamins B2 and B12 (data are presented as mean ± s.d. of n = 3 replicates). In c–e and g–i, the scale bars are 3 mm.Source dataCompared with 3D unary liquid patterning, 3D patterning of binary liquids can open broader applications because it allows 3D parallel multivariate functions. Combined with interfacial reactions, here we accomplished programmable 3D patterning of binary liquids using the switchable capturing and releasing of CPFs. The detailed patterning principle is elucidated in Fig. 3f. The CPFs first capture liquid A and then encapsulate it through interfacial solidification (for example, polymerization or gelation). Subsequently, these solidified liquid interfaces will break the liquid continuity and block the release process of the adjacent releasers, converting the releasers into capturers to capture liquid B, finally forming the 3D patterns of binary liquids. Employing interfacial gelation, we elaborate the 3D patterning of water, hydrogel and drug-containing aqueous solutions. The 3D pattern of gellan gum solution and water is illustrated in Fig. 3g. By designing alternately adjacent single-rod and double-rod connections, 3D alternately distributed capturers and releasers were obtained. After capturing red-dyed gellan gum solution and blue-dyed water in sequence, a 3D binary liquid pattern with alternating red and blue pixels was prepared. The 3D pattern of two gellan gum solutions is exhibited in Supplementary Fig. 7. By designing layer-by-layer alternating single-rod and double-rod connections, 3D layer-by-layer alternating capturers and releasers were generated. After capturing red- and green-dyed gellan gum solutions in sequence, a 3D binary liquid pattern with alternating red and green layers was created. The liquid-patterning processes in Fig. 3g and Supplementary Fig. 7 both take advantage of the gelling properties of gellan gum at low temperatures (Supplementary Video 9). Because a wide range of biomaterials and chemicals can be incorporated into aqueous solutions and hydrogels, their 3D patterns of binary liquids can be used for multipart cell or bacterial culture17,18, 3D chemical-gradient generation16,19 and multimaterial structure fabrication20 in cell–cell communication21, bacterial ecology22, chemical engineering23 and hybrid bioprinting.To demonstrate the practical utility of 3D patterning of binary liquids, we show its application in 3D spatiotemporal control of multimaterial concentration distributions through numerical simulations and experiments. The simulated 3D spatial concentration distribution of the two substances is shown in Extended Data Fig. 5 in their different binary liquid patterns during diffusion in water. By tuning the diffusion coefficient and designing diverse binary liquid patterns, the 3D concentration distribution of these two substances can be spatiotemporally controlled. We validated this by realizing controlled multidrug release using the interfacial gelation of CaCl2 in sodium alginate. Vitamins B2 and B12 were chosen to represent the two different drugs because of the high color contrast of their aqueous solutions. Their 3D binary liquid pattern is displayed in Fig. 3h, where the red vitamin B12 solution is encapsulated in the truncated icosahedron frame and the yellow vitamin B2 solution is encapsulated in cubic frames. The specific patterning process is recorded in Supplementary Fig. 8. The capturers first capture the CaCl2 solution dissolved with vitamin B12 and complete the liquid encapsulation in sodium alginate. Then the release-blocked releasers capture the viscous gellan gum solution dissolved with vitamin B2. Their release process in water and the corresponding release profiles are presented in Fig. 3i and Fig. 3j, respectively, implying that vitamin B2 in gellan gum solution is rapidly released, while the encapsulated vitamin B12 shows slow release due to the membrane barrier. The thickness of the membrane can be flexibly adjusted by changing the interfacial gelation time, thereby controlling the release rate ratio of the two different drugs. By designing the spatial arrangement of capturers and releasers, the spatial concentration distribution of the two drugs can be further manipulated. Therefore, 3D spatial and temporal control of multidrug release can be successfully achieved.Combined with interfacial polymerization, multiliquid patterning can readily be achieved and enable multimaterial manufacturing. Three-dimensional patterning of two different photopolymers was achieved with CPFs as shown in Extended Data Fig. 6, where a solid structure consisting of three kinds of photopolymer was obtained via photopolymerization. In addition, just using captures to capture different photopolymers, combined with photopolymerization, a variety of multimaterial manufacturing is implemented (Extended Data Fig. 7). As a simple and straightforward multimaterial manufacturing method, CPFs with switchable functions of capturing and releasing may provide a facile route to fabricate complex multifunctional structures composed of multimaterials.Applications in interfacial processesSwitchable liquid capture and release enables a versatile interfacial processing platform as shown in Fig. 4a. To increase the fluid interfacial area, the reactant liquid is first dispersed into a 3D liquid array using the CPF array, under the condition that the size of the PF is smaller than the capillary length as well as the distance between adjacent frames24,25 (Supplementary Discussion 4). After the interfacial processes, by converting the capture into release, the dispersed liquid can be completely recovered to facilitate the concentration, purification and detection of the products. This platform is generally applicable to both gas–liquid and liquid–liquid interface processes.Fig. 4: Applications in interfacial processes.a, Versatile interfacial process platform based on CPFs. The reactant is first dispersed into a liquid array using capturers. After the interfacial processes, by contacting the capturers with liquid films, the dispersed liquid can be completely recovered to facilitate product processing. b, Reversible sampling and release demonstration. The insulin release curve of the CPFs, cotton swab and flocking swab in water (data are presented as mean ± s.d. of n = 3 replicates). c, Influenza virus sampling and detection results of the CPFs, cotton swab and flocking swab. d, Schematic illustration of bacterial encapsulation. After encapsulation by interfacial gelation, the Acetobacterium are confined in the semi-permeable membrane, the reactants can enter the membrane timely to be replenished and the products can diffuse out. e, Concentrations of acetate and suspended bacteria in solution versus time. f, Evaporative humidifier demonstration. The dry air passes through the water-captured CPF array, and the humidified air is output as the water droplets evaporate. The output humidity can be adjusted by the airflow rate (data are presented as mean ± s.d. of n = 3 replicates). g, Demonstration of CO2 capture. The CPF array first disperses the absorbent, and after absorbing CO2, the droplets are recovered by the soap film array for CO2 storage. Subsequently, heating the recovered liquid can release CO2 on demand. Scale bars, 3 mm.Source dataWe first explored liquid–liquid interfacial processes (that is, diffusion and mixing) with the application of switchable sampling and releasing. Owing to their simplicity in absorbing liquids, sampling tools such as cotton swabs and flocking swabs are extensively used in diagnosis, dressing and clinical medicine. However, these tools suffer from sample residues during their sample release, greatly reducing their detection accuracy and leading to material waste. CPFs can potentially address this challenge because their frame structure renders free liquid–liquid interfaces for the full release of samples. Moreover, the release process will not cause shear-induced sample rupture associated with traditional pipette tools26, an issue in processing objects such as cells, enzymes and proteins. The quantitative sampling and release experiments of insulin (Fig. 4b) and vitamin B12 solutions (Supplementary Fig. 9a) are performed with the CPF, cotton swab and flocking swab. Their release curves verify that the CPFs can completely release the sample in a few seconds, but both flocking and cotton swabs can only release part of the sample. The improved release capacity of the CPFs was also corroborated by a qualitative dye-release experiment (Supplementary Fig. 9b), showing complete release of dye with the CPF even without applying any stirring, but obvious dye residues associated with the cotton and flocking swabs. Using the influenza virus as an example, we demonstrate the practical utility of CPFs as sampling tools with improved release performance. As shown in Fig. 4c and Supplementary Fig. 9c, when the virus concentration was low (1 × 104 to 5 × 104 plaque-forming units (PFU) ml−1), only the CPFs detected the virus, while both the flocking swab and cotton swab showed false negative results. Besides, compared with cotton swabs, the CPFs exhibited comparable liquid absorption capacity for both hydrophilic and hydrophobic surfaces and can maintain sufficient mechanical strength even under repeated large deformations (Supplementary Video 10). Therefore, as supplementary sampling tools with improved release capacity, CPFs can improve the detection accuracy of bioassays such as coronavirus disease 2019 testing, while reducing the required sample volume.In addition, the use of interfacial gelation enables the application of CPFs in biomaterial encapsulation. As shown in Fig. 4d and Extended Data Fig. 8, when an Acetobacterium solution containing CaCl2 captured in the CPFs contacts with the sodium alginate solution, a semi-permeable hydrogel membrane will be formed to encapsulate the Acetobacterium in the CPFs. The microscopic morphology of the semipermeable membrane presents a porous structure with a pore diameter of about 50 nm. Therefore, during a microbial electrosynthesis reaction, the micrometer-scale Acetobacterium are confined to the membrane, the gas reactants including H2 and CO2 can enter the membrane in time to be replenished, and the acetate product can diffuse out. This is confirmed by the experimental results in Fig. 4e, where the acetate concentration in the reactor continued to increase while the bacteria concentration in the reactor (optical density measured at 600 nm, OD600) was almost constant, indicating that the bacteria were successfully encapsulated in the CPFs. This can facilitate the separation of bacteria and reaction products, simplify the microbial reaction process and enhance the utilization rate of bacteria. In addition, CPFs can also potentially be used to encapsulate other biological materials such as algae and cells.Next, we investigated gas–liquid interfacial processes including evaporation, adsorption, diffusion and reaction, using an evaporative humidifier and CO2 capture as examples. Figure 4f illustrates the evaporative humidifier. By introducing dry air (humidity 10% at 21 °C) into a sealed chamber equipped with a water-captured CPF array, the humidified air can be output as the droplets evaporate (Supplementary Fig. 10). The output air humidity can be adjusted from 40% to 80% at 21 °C depending on the airflow speed. Notably, experiments have shown that directly introducing dry air into water confined in the bottle cannot effectively humidify the air due to the limited gas–liquid contact area/time. To further demonstrate the practicability of CPF-based evaporative humidifiers, we prepared a commercial-scale humidifier prototype as shown in Extended Data Fig. 9. The experimental results show that the humidification volume is 760 ml h−1, humidification efficiency is 12.6 ml h−1 W−1 and humidified air output is about 2,000 m3 h−1. Compared with commercial paper-based evaporative humidifiers, our CPFs-based humidifiers have a higher water storage capacity and require less water flow, making them potentially more energy efficient. Furthermore, unlike paper-based materials, CPFs can be easily customized through additive manufacturing, and the structure is reusable. In addition, compared with the continuous liquid formed by the reported 3D capillary structure27, the CPFs allow large-scale 3D liquid dispersion to form a larger surface area with less resistance to airflow, leading to greater airflow flux and higher humidification efficiency. An ideal CO2 cycle process performed with our platform is exhibited in Fig. 4g, which includes carbon capture and storage and CO2 reutilization28. Na2CO3 solution dissolved with thymol blue serves as the absorbent. The absorption efficiency of the same amount of absorbent in the bottle and the CPF array are compared. During the experiment, CO2 is directly introduced into the absorbent in the bottle, and the absorbent in the array absorbs the overflowed CO2. Absorbents in the array exhibit noticeable color changes earlier and faster, showing higher capture efficiency (Extended Data Fig. 10). After the absorption, the absorbent in the array can be fully recovered by a soap film array for CO2 storage. Subsequently, heating the recovered absorbent can release CO2 for applications such as enhanced oil and gas recovery, winemaking, artificial rainfall and so on29, while both the array and absorbent can be reused for carbon capture and storage. Altering the liquid composition and gas environment, the platform is also applicable to other gas–liquid interfacial processes, such as the toxicity of e-cigarettes to lung tissue cells, harmful gas absorption and environmental monitoring.