Plant-inspired TransfOrigami microfluidics

Abstract

The healthy functioning of the plants’ vasculature depends on their ability to respond to environmental changes. In contrast, synthetic microfluidic systems have rarely demonstrated this environmental responsiveness. Plants respond to environmental stimuli through nastic movement, which inspires us to introduce transformable microfluidics: By embedding stimuli-responsive materials, the microfluidic device can respond to temperature, humidity, and light irradiance. Furthermore, by designing a foldable geometry, these responsive movements can follow the preset origami transformation. We term this device TransfOrigami microfluidics (TOM) to highlight the close connection between its transformation and the origami structure. TOM can be used as an environmentally adaptive photomicroreactor. It senses the environmental stimuli and feeds them back positively into photosynthetic conversion through morphological transformation. The principle behind this morphable microsystem can potentially be extended to applications that require responsiveness between the environment and the devices, such as dynamic artificial vascular networks and shape-adaptive flexible electronics.

Inspired by nastic plants, a transformable origami microfluidic chip is designed for environmentally adaptive photosynthesis.

INTRODUCTION

Plants have very rich and complex vasculature that transports water and nutrients through their tissues to maintain normal metabolism. For example, in leaves, which represent the main organ of plants for photosynthesis, their veins can deliver the nutrients produced by photosynthesis to the rest of the body while also transporting water throughout the leaflet for transpiration. This vasculature has inspired us to develop artificial systems with embedded fluidic channels, such as biomimetic microfluidic devices (1–5). Moreover, plants have evolved the ability to respond to environmental changes to allow their vasculatures to function more healthily even in an ever-changing natural environment. This ability of individual systems to sense the external environment and adjust themselves accordingly is sometimes known as “environmental adaptability.” In particular, the nastic movement of plants is the fastest way to adapt their shapes to environmental changes in light, temperature, and humidity (6, 7). For example, the genus Oxalis deploys the leaflets during the sunny day to promote photosynthesis and folds them at night to retard energy dissipation due to transpiration (8). However, this plant-like ability to interact with the environment is rarely mentioned in synthetic microfluidic systems (9, 10). Endowing microfluidic systems with these stimuli-responsive shape-changing functions can pave the way for sophisticated, multifunctional, or intelligent fluidic systems with dynamic biomimetic design or environmental adaptability (3, 11–28).

Now, microfluidic systems respond to the environment by relying on linked or onboard electronics and computer programming. This can result in a tethered, complex, and overall cumbersome system (26, 29, 30). In contrast, their natural counterparts contain veins as fluid-transporting microchannels and pulvinus as stimuli-responsive actuators in their thin, lightweight, and flexible leaflets (6, 7). Besides, along with the nastic movement, the leaflets tend to fold or unfold into a regular geometry that has specific purposes. Two main aspects that restrict the development of stimuli-responsive morphing microfluidics are as follows: First, the mainstream materials used to fabricate microfluidic devices are inert materials without environmental responsiveness, such as polydimethylsiloxane (PDMS) and poly(methyl methacrylate) (PMMA); second, the design of microfluidic devices is usually based on considering the fluidic flow to design the embedded channel shape rather than the overall shape of the microfluidic device. Consequently, most common microfluidic devices do not show environmental responsiveness and targeted shape change, let alone the nastic movement, similar to those in leaves. To gain ability of nastic movement, the microfluidic device has to exhibit stimuli responsiveness at the device level, and the overall shape of the device has to be programmable. These microfluidic devices will enable previously unfeasible applications: For instance, the resultant environmentally adaptive photomicroreactors self-regulate the photosynthetic conversion rate according to weather changes (23, 24, 26).

Stimuli-responsive materials—such as shape memory polymers, liquid crystal elastomers, and responsive hydrogels—are capable of changing their shapes in response to external temperature, light, or humidity (29, 31–33). They have functioned as both sensors and actuators synchronously in many morphable structures. Nonetheless, current microfluidic devices use responsive materials only as localized components, such as light-controlled valves or flow-switching channels, rather than in the overall morphing of a microfluidic device (34–39). To implement a preset overall three-dimensional (3D) morphing, the device’s dimensions, the positions of the responsive materials embedded in the device, and the target operation in response must be engineered specifically (40). Combining with the ancient art of origami, the desired 3D structures not only can be constructed from precursors via 2D processing techniques but also can mutually convert between a fully deployed 2D plane and a compact 3D form by folding and unfolding (29, 31, 41–44). Therefore, by fusing responsive materials into a microfluidic device designed with an origami geometry, it can transform between 2D and 3D states based on the preset structure via stimuli-triggered responses.

In this work, the nastic plants inspire our concept of a microfluidic device with environmental responsiveness, and the stimuli-responsive structures offer us the conditions to realize this concept. On the basis of these, we have developed a transformable microfluidic chip by integrating stimuli-responsive materials with a thin and foldable microfluidic chip (Fig. 1). The entire device can respond to changes in temperature, humidity, and light irradiance by transforming along the preset origami folds. Thus, we name this transformable origami microfluidic approach TransfOrigami microfluidics (TOM). We demonstrate that TOM can be applied as an environmentally adaptive photomicroreactor. The transformation reconstructs the reaction channels, changes their light-harvesting capability, and eventually regulates the photosynthetic conversion. Positive feedback control is built into interactions between TOM and the environment. When the external conditions are favorable for the photoreaction, the feedback results in an enhanced photosynthetic conversion and vice versa. As the first of its kind, stimuli-responsive morphing microsystems, our TOM will inspire applications in energy, robotics, or biomedicine that require environmental adaptations, such as artificial vascular networks or flexible electronics with adaptive rhythmic movements (40, 45, 46).

Fig. 1. Plant-inspired TOM.

O. corniculata at its (A) open state during the day time and (B) close state during night time. Left insets: Top view of its open and close states, respectively. Right insets: Unfolded and folded states of the corresponding origami structure, respectively. (C) Schematic drawing of the TOM at its unfolded state when the temperature is high and light irradiates; folded state when the temperature is low and humidity is high. The insets show the schematic dehydration and hydration in the pNIPAM hydrogel layer of the actuating unit at the unfolded and folded states. (D) Bright-field images (channel is filled with blue dye) and dark-field images (channel is filled with fluorescent dye) of TOM’s two states. Scale bars, 10 mm. (E) Origami design of the TOM. Solid line indicates the folding line; dashed line indicates the edge line. Scale bar, 10 mm.

RESULTS

Plant-inspired transformable origami microfluidics

Oxalis corniculata horizontally deploys its leaflets during the daytime to expose more of its top surface to the sunlight and closes entirely by folding the leaflets at night (Fig. 1, A and B, and movie S1). Its nastic movement results from the actuation of the pulvinus to light (7, 47). Because of their inherent symmetrical appearance, the leaflets fold and unfold in a shape similar to the origami square base. When unfolding the square base, its upper plane is totally exposed. Conversely, when folding, the square folds along with the diagonal lines and its four corners would gather to collapse the upper plane. The origami square base can thus adjust the upper surface area exposed in the horizontal plane via transformation. These inspirations suggest that our device should adopt a thin and foldable origami structure. In addition to the overall structure, we also need to consider the choice of responsive materials. Silicone elastomer (PDMS) has a well-established soft lithography technology to support its microfabrications. Although the virgin PDMS is inert, it can be further functionalized through functional material doping and surface modifications (48). Hydrophobic nanomaterials can be dispersed in PDMS prepolymer homogenously, and polyacrylamide hydrogel can be grafted onto the PDMS surface firmly through photopolymerization. The prepared hydrogel-elastomer bilayer is a typical stimuli-responsive soft actuator (48). The morphing behavior of the hydrogel-elastomer actuator is induced by different swelling ratios of hydrogel and elastomer under stimuli. The hydrogel layer acts as an active layer, swelling and shrinking in response to hydration and dehydration. The elastomer layer works as a passive layer, forming a stress difference with the active layer and thus causing the overall shape to bend. A thermal-responsive hydrogel [poly (N-isopropylacrylamide) (pNIPAM)] is used as the active hydrogel layer to conduct this bending deformation based on the temperature. Photothermal dopants, such as graphene nanoplatelets (GNPs), are added to PDMS to realize the photothermal responsive actuation of the system in the TOM devices (Fig. 1C).

We integrate the stimuli-responsive morphing origami into a thin microfluidic chip to fabricate the TOM (Fig. 1D and movie S2). The microfluidic chip, in particular, is manufactured by conventional soft lithography with some modifications: (21, 49, 50) (i) To facilitate morphing, the designed device is thinner (<1 mm in thickness) than conventional microfluidic devices; (ii) during chip fabrication, we in situ dope the GNPs and modify pNIPAM at specific areas of the device, patterning the actuating units at the diagonal and centerline positions. Driven by external stimuli, the diagonal actuators perform the mountain fold, while the median actuators induce the valley fold. The two types of fold synergistically transform the microreactor into an origami square base (Fig. 1E and fig. S1). This origami structure, allowing flexible, mutual conversion between its 2D flat state and 3D compact form, is used to realize 3D fluid motion via a simple 2D device. Compared to other responsive morphing devices reported to date, this is the first microreactor that mimics nastic plant morphing with mechanical motion triggered by environmental stimuli.

To demonstrate the concept, a thin microfluidic device is fabricated according to a modification of previous reports (21, 49, 50). A mold with microchannel patterns is first produced by photolithography (fig. S2). Then, a thin layer of PDMS prepolymer is spin-coated on top of the mold and cured. Afterward, a similar step is repeated to coat a layer of benzophenone (BP)–doped PDMS (BP-PDMS) on top of this cured PDMS layer to obtain layer A (Fig. 2A and fig. S3). For layer B preparation, a layer of BP-PDMS is first spin-coated on a clean wafer without a pattern. After curing, the predesigned patterns that needed to be doped with GNPs are hollowed out in this BP-PDMS layer. In this case, the hollow patterns are rectangular and distributed on the center line and diagonal line of the square area reserved for the device (fig. S2). GNP and BP–codoped PDMS (G-PDMS) are in situ assigned to the hollowed-out regions. One more layer of PDMS is coated on top of the layer with G-PDMS patterns. Subsequently, the well-prepared layer A and layer B are aligned and bonded (fig. S3). In the resulting G-PDMS–patterned thin microfluidic device, the designed microchannels are embedded in the middle PDMS layer without dopants to avoid additional effects of the dopants on the subsequent photoreactions. The outer surfaces of the device are all doped with BP as the photoinitiator for the following surface hydrogel modification. The surface modification of the hydrogel is carried out by sandwiching the semifinished device between two masks that are hollowed out according to the desired modification patterns (Fig. 2A and figs. S2 and S4). The pNIPAM grafting is conducted by adding the NIPAM monomer solution to the wells on the mask on both sides under ultraviolet (UV) treatment. NIPAM is polymerized and grafted onto BP-PDMS and G-PDMS surfaces by chemical bonding (51). Last, the device is released from the masks, washed with deionized water, and trimmed to the designed shape (Fig. 2B and figs. S2 and S4). The total thickness of the completed device is around 615 μm, which is thinner than most microfluidic devices (Fig. 2, C and D). The thin dimension allows our device to fold easily with the swelling of the pNIPAM. The device has customized pNIPAM patterns on both sides (sides A and B), as shown in Fig. 2B. On side B, pNIPAM patterns are grafted at the diagonal positions, whereas on side A, they are grafted at center positions. The position, size, and distribution of these spacing and alternating pNIPAM patterns are designed to match the folding lines of the desired origami structure as closely as possible, resulting in the subsequent mountain and valley folds in the TOM (Fig. 1E and fig. S1). Specifically, the diagonal actuators provide mountain folds, and the center actuators offer valley folds. The cooperativity of the two folding approaches enables the device as a whole to form a square base origami.

Fig. 2. Fabrication and characterizations of TOM.

(A) Fabrication of TOM. (B) Schematics show both sides of the final device and the exploded view of TOM. (C) Schematic cross sections of three regions on TOM. (D) OM and SEM of three regions. Scale bar, 500 μm. (E) SEM imaging on pNIPAM top surface and the interface between pNIPAM and G-PDMS coupled with EDX mapping. Scale bars, 100 μm.

We partition the device into three regions based on their functions: (i) diagonal actuators, (ii) light-harvesting panels, and (iii) center actuators (Fig. 2B). The pNIPAM is grafted on the top and bottom surfaces of the regions (i) and (iii), respectively. The average thickness of pNIPAM is around 65 μm, while the thickness of G-PDMS on the top layer is 130 μm (Fig. 2, C and D). Since GNPs are black powders, the G-PDMS layer is apparently nontransparent under the optical microscope (OM). However, under the scanning electron microscope (SEM), we can clearly identify the interface between pNIPAM and G-PDMS by their different SEM signals in the nontransparent area that optical microscopy cannot view (Fig. 2, D and E). Energy-dispersive x-ray (EDX) mappings are used to confirm their elements; it can be seen that although both layers contain carbon, only the bottom layer is rich in silicon. This indicates that the lower layer is PDMS, and the upper layer is pNIPAM. In addition, from the top view in Fig. 2E, creases can be found because of the surface instability of the hydrogel (52, 53). These results are in agreement with the literature and show the successful modification of the pNIPAM on the PDMS surface (54, 55). The strong adhesion between pNIPAM and PDMS allows the device to be used in the air for long periods of time without loss of function. The highly light-transmissive region (ii) consists of most of the microchannels for photoreactions as light-harvesting panels. These serpentine channels have a cross-section of 400 μm in length and 120 μm in height. This geometric design serves two purposes: The serpentine structure facilitates the loading of longer reaction channels in a limited space, and the high aspect ratio allows more photons directly above to enter the channel in the unfolded state of the device.

Environmentally responsive morphing of the TOM

To quantitatively analyze the folding and unfolding performance of the TOM, we have to define the degree of its folding or unfolding. In our subsequent experimental settings, the variation of the horizontally projected area of the device not only reflects the folding or unfolding deformation of TOM but also allows estimation of its light-collecting capability in the subsequent photoreaction experiments. Hence, the horizontally projected area is adopted in quantifying the degree of folding or unfolding. The camera is set right above the TOM to record the variation of the horizontally projected area of the device, as illustrated in Fig. 3A. The unfold percent X for the transition stage is calculated using the following equation: X(%) = (A_X − A_0%) ÷ (A_100% − A_0%) × 100%, where A_X is the horizontally projected area of a morphing device and A_0% and A_100% denote the horizontally projected areas of a wholly folded and unfolded device, respectively.

Fig. 3. Temperature, light, and humidity responsiveness of TOM.

(A) An illustration shows the definition of unfold percent. (B) The unfolding and folding of TOM under light irradiation [temperature (temp.) ≈ 25°C; relative humidity (R.H.) ≈ 50% RH; illumination (illum.) ≈ 30 klx] and humidified environment (temp. ≈ 6°C; R.H. ≈ 100% RH; illum. ≈ 0 klx). Scale bars, 10 mm. (C) Change in unfold percent as a function of processing time at a different temperature or light illuminance. (D) Photothermal testing of the GNPs-doped TOM and undoped TOM (illum. ≈ 30 klx). (E) Outdoor experiments monitoring the changes of unfold percent, illuminance, temperature, and humidity.

When exposed to light irradiation, TOM gradually unfolds over time, while in a low-temperature and high-humidity environment, it folds gradually (Fig. 3B and movie S3). To quantitatively analyze the environmental responsiveness of the TOM, we measure the unfold percent as a function of processing time under different situations at different temperatures or light illuminances, as shown in Fig. 3C. pNIPAM-based TOM primarily responds to changes in temperature. The pNIPAM’s lower critical solution temperature is around 32°C (56, 57). Temperature switching below and above 32°C will cause pNIPAM to undergo a reversible phase transition between a swollen hydrated state and a shrunken dehydrated state (Fig. 1C, insets). That is, the temperature determines which state the pNIPAM is in, while the humidity determines the ability to hydrate and swell or to dehydrate and shrink in the corresponding phase and the rate of its swelling or shrinking. As a result, TOM does not deploy when the temperature is 20°C but only at 40°C and above. In addition, the higher the temperature, the faster it unfolds (70% unfolding change in 10 min with 80°C heating). Even in high-humidity environments, temperature also plays a role in determining the rate of folding. Although the permeation effect makes it fold at 40°C, the folding efficiency is low, and only at a low temperature of 6°C can it return from 70 to 10% unfolding in 10 min. Theoretically, the thicker the pNIPAM as the actuating layer and the thinner the PDMS as the passive layer, the higher the degree of the deformation of TOM would be (fig. S5). To achieve an apparent deformation, the current pNIPAM and PDMS thicknesses in the TOM prototype are designed to be the thickest and thinnest, respectively, based on our fabrication techniques and facilities accessible to us. In the design conception, we expect the TOM to be responsive to light, so GNPs with photothermal response are integrated into the corresponding actuator. We adopt the halogen lamp as a sunlight simulator because it covers most of the wavelength range (400 to 1000 nm) consistent with sunlight (300 to 900 nm) except for the UV region (fig. S6). This wavelength band not only provides the wavelength required for subsequent photochemical reactions but also contains the near-infrared region with thermal effects that contribute to the photothermal conversion of the device (58, 59). Hence, the graph of light response shows that the higher the light intensity, the faster the TOM unfolds, and at an intensity of 30 klx, it can turn the device from 10 to 90% unfold percent in 10 min. The maximum unfold percent is around 95%, as a small amount of residual water likely remains deep in the pNIPAM and causes a slight deformation even when the TOM is in its fully unfolded state. Furthermore, we observe the unfolding of the GNP-doped device and the undoped device under light irradiation via thermal imaging (Fig. 3D and movie S4). The GNP pattern areas warm up noticeably in the doped device. The undoped device, however, maintains the same temperature as the ambient. The quantitative data thus extracted can be referred to in fig. S7, where the average temperature of the GNP-patterned region increases from 15° to 40°C during the irradiation and eventually exceeds 10°C above the control group and the ambience. With the similar irradiation intensity of light, the doped device has a higher heating rate (R_dope = 2.5°C/min; R_undope = 1.5°C/min) and unfolding rate (R′_dope = 4.5%/min; R′_undope = 2.5%/min) compared to the undoped device.

In addition to the controlled laboratory environment, we monitor the change in unfold percent on sunny and rainy days to demonstrate the response of TOM in a natural outdoor climate. During the outdoor light exposure period, the device’s initially unfold percent is kept at about 40% due to cloudy weather with light intensity below 20 klx (light yellow background in the graph), as shown in Fig. 3E, fig. S8, and movie S5. Afterward, the detected illuminance soars to 90 klx as the clouds disperse. The device then starts to unfold, and the increase in the unfold percent is 25% during this 8-min period of stronger light irradiation (dark yellow background in the graph). In the last 3 min, the light intensity drops and the unfolding rate stops increasing correspondently. Similar monitoring is repeated on a rainy day. The outdoor relative humidity (R.H.) on a rainy day can reach 100% RH, so the device starts folding at the beginning of our recording. The unfold percent gently declines for the first 25 min when the rain is mild (light blue background in the graph). As the rain increases, the folding speeds up accordingly (dark blue background in the graph). Throughout the 60-min recording, the unfold percent drops from 80 to 20%. Although the natural conditions of light, temperature, and humidity are unstable and uncontrollable, our device still adapts to the environment well. In the sunny day outdoor experiment, the actual sunlight intensity is stronger than those used in the laboratory, so the intensity required for triggering the morphing of the device can be met. In the rainy day outdoor experiment, although raindrops fail to provide particularly high water pressure to facilitate the penetration of water into the hydrogel, and thus the response rate is low, the high humidity provided is still sufficient for the morphing. Hence, the results of these experiments suggest its potential for outdoor applications.

Self-regulated light collection and channel structure based on morphing

The morphing of TOM caused by environmental stimuli is expected to further regulate the photoreaction in the microreactor. We hypothesize that this self-regulation in TOM is synergistically enabled by its light-collecting capability and the reshaped structure of the channels. The inclination of the light-harvesting panel would weaken the power of light refracted into the microreactor, and the area exposed to light would alter as it folds or unfolds. Specifically, the folding of TOM leads to the inclination of the light-harvesting panel (P denotes the top surface of this panel) on it, which, in turn, enlarges the incident angle θ on the interface of air and PDMS, as shown in Fig. 4 (A and B). According to Fresnel equations, the larger the incident angle on panel P, the more light energy is reflected by the interface and less is transmitted (60). As a result of the TOM folding, the energy of the light transmitted to the embedded reaction channel decreases, weakening the photoreaction (Fig. 4B, inset). Furthermore, we simulate the change in refracted light energy after incident rays pass through the three morphing models of TOM (Fig. 4C and movie S6). As seen in the projection spot diagram of the primary rays (released rays and refracted rays), the area where the ray energy diminishes is also the horizontally projected area. This is the reason that, in the earlier investigations, we quantify the device’s morphing via the unfold percent calculated by the horizontally projected area. When the TOM folds, the area of low-energy rays shrinks, indicating that less light passes through the device and thus less is used for photoreaction. Another key finding from the spot diagrams is that as the unfold percent drops, the number of blue spots grows, implying a loss in the energy of the transmitted rays and a rise in the energy of the reflected light. This result is consistent with the analysis in Fig. 4 (A and B).

Fig. 4. Self-regulation of light collection capability and channel structure based on morphing.

(A) Reflection (R) and transmission (T) on inclined and horizontal device surface are depicted schematically. θ is the angle of incidence (I), and θ_incl. is the inclination angel. Subfigure i and ii represent the reaction zones before and after the unfolding of TOM, respectively. (B) 3D models of TOM with different unfold percents (100, 50, and 0%) and illustrations of the θ, I, R, and T of the top surface P on the light-harvesting panel at their corresponding unfolding states. Inset: The calculated θ, T, and R values correspond to the P at various unfold percents. (C) Light ray trajectories and projection spot diagrams of the primary rays. Scale bars, 10 mm. (D) Stimuli-responsive morphing of TOM can be reflected by the variation of the liquid pressure inside the microchannel. Insets: Elastic strain distribution on TOM with unfold percents of 100, 50, and 0%, respectively. NA, no additional stimuli are applied.

In addition to its light collection capability, another influence of morphing on the photoreaction is that it alters the volume of the reaction microchannel. We visualize the elastic strain distribution in TOM by 3D finite element simulation (Fig. 4D, inset, and movie S7). The strain in the actuating regions increases pronouncedly when the device is folded. Thus, the structure of some microchannels running through these regions can be reshaped. To further investigate how the overall morphing changes the channel structure, we connect a highly sensitive pressure gauge near the inlet of the microfluidic system to monitor the real-time pressure change under different morphing states (fig. S9) (1). We can deduce that the pressure difference ∆P is inversely proportional to wh³, where w and h denote the width and height of the channel for rectangular cross section, respectively (see Materials and Methods for deduction details) (1, 61). Thus, the cross-section size is increased by unfolding, causing pressure drops. The initial state of the TOM is unfolded around 20% for the experiment, and the velocity of the water is set at 2000 μl/hour, as shown in Fig. 4D. Without any additional stimuli (temp. ≈ 25°C; R.H. ≈ 50% RH; illum. ≈ 0 klx), the detected pressure remains relatively stable at 16 kPa for the first 10 min. Afterward, when applying an illuminance of 30 klx to the device, the unfold percent is raised to 80%, and ∆P decreases to 12.6 kPa at 20 min, suggesting that the R_hyd is reduced by 21%. After 10 min of buffering without additional stimuli again, the humidity is elevated to ≥95% RH, and ∆P climbs to 16.7 kPa at 40 min. The change of pressure suggests that morphing would regulate the sectional area of the microchannel in a microfluidic system with inner fluid velocity. When TOM is folded, its volume decreases and vice versa. Because less reagent may flow into the folded TOM, the reaction efficiency per unit time is lowered.

Adaptive photosynthesis via TOM

To experimentally verify the effect of TOM morphing on photoreaction, we design a set-up including a syringe pump, a reaction chamber with a solar simulator, and an optical flow cell with an absorption spectrometer for monitoring the photosynthesis flow through the TOM (Fig. 5A). The cycloaddition of 9,10-diphenylanthracene (DPA) to the corresponding endoperoxide is chosen as a model reaction because it converts by visible light, displays light-limited kinetics, and is almost entirely independent of temperature (14–16). Methylene blue (MB) is used as a photocatalyst with a maximum absorption peak of 654 nm, as shown in Fig. 5B. The excited MB interacts with oxygen to form singlet oxygen, promoting this cycloaddition reaction. The absorption at 372 nm of endoperoxide obtained after the photoreaction of DPA is greatly weakened, so it can be used as a quantitative parameter for the conversion rate of the photoreaction (see Materials and Methods for calculation details). The reaction chamber is the only area of the system where the photoreaction can take place. Another of its functions is to simulate changes in TOM’s external environment. It is difficult to replicate complex natural environments in the laboratory, so we will only discuss the simplified representative cases in our experiments. Without considering light, two conditions are set to trigger the TOM folding and unfolding: (i) normal temperature and humidity (i.e., temp. ≈ 25°C; R.H. ≈ 50% RH) and (ii) low temperature and high humidity (i.e., temp. ≈ 6°C; R.H. ≈ 100% RH). The purpose is for TOM to reach its relatively maximum and minimum unfold percent (~90 and ~10%) as quickly as possible during the real-time test. To demonstrate that the change in conversion at constant light intensity depends only on the morphing of the TOM, but not on other factors, we compare the conversions by placing the unfolded and folded devices, whose shapes are fixed, in the two conditions mentioned above, respectively. Under the same light intensity (illum. ≈ 6 klx), although the external medium and stimuli are different, no obvious difference in the conversions is obtained from the same morphing state (Fig. 5C). Yet, a 20% higher conversion from the unfolded state than the folded state can be found. Therefore, it is reasonable to consider morphing as the only factor that determines the conversion rate in the subsequent experiments with equal light intensities. Afterward, we compare the change of real-time conversion for TOM’s two states under various light intensities, as shown in Fig. 5D. When the structure of this photoreactor is fixed at a certain folded or unfolded state, we can see that the difference in conversion rates between TOM’s two morphing states will be amplified as the light becomes stronger (Fig. 5D, inset). For instance, under the illumination of 8 klx, about 30% conversion gap exists between its unfolded and folded states. This also implies that a small deformation of the light-harvesting panel under more intense natural light (10 times stronger than the 8 klx used in this case) can lead to a substantial difference in the end product of photosynthesis.

Fig. 5. Adaptive photosynthesis via TOM.

(A) Schematic illustration of the setup and photoreaction applied in TOM. (B) Absorption spectrum of the MB, DPA, MB and DPA, and MB and DPA treated with light. a.u., arbitrary units. (C) Comparison of conversion of TOM in folded/unfolded state in air and in water after 30-min irradiation with 6-klx illuminance, respectively. (D) The variation of the photosynthetic conversion for two states of TOM with progressive light intensity. Gray background indicates no light irradiation. The gradient from yellow to red background represents a gradient of light intensity from 2 to 8 klx. Inset: The difference between the conversions of the two morphing states (∆C) increases with higher illuminance. (E) The real-time monitoring of the conversion in two switching conditions: The normal temperature and humidity with relatively high light intensity (i.e., temp. ≈ 25°C; R.H. ≈ 50% RH; illum. ≈ 6 klx; yellow background) and low temperature and high humidity with a relatively low light intensity (i.e., temp. ≈ 6°C; R.H. ≈ 100% RH; illum. ≈ 2 klx; blue background). Gray background indicates no light irradiation. (F) Schematic illustration of the adaptive photosynthesis.

We expect that the TOM morphing response to the environment would give real-time feedback on the photoreaction, just as plants respond to changes in the natural environment. Therefore, we adjust the humidity, temperature, and light irradiation in the reaction chamber during real-time photosynthesis (Fig. 5E). We match the various light intensities to the two simplified conditions described above: (i) normal temperature and humidity with relatively high light intensity (i.e., temp. ≈ 25°C; R.H. ≈ 50% RH; illum. ≈ 6 klx) and (ii) low temperature and high humidity with relatively low light intensity (i.e., temp. ≈ 6°C; R.H. ≈ 100% RH; illum. ≈ 2 klx). Besides, the changes in conversion for the two fixed morphing states are used as control groups, indicating the nonmorphing devices (dashed lines in Fig. 5E). When the external environment is unsuitable for photosynthesis (e.g., low light intensity, low temperature, and high humidity), the conversion from TOM drops to a lower level (C ≈ 10%). However, it will automatically adjust the conversion to a higher level (C ≈ 70%) by unfolding when the external environment is suitable for the photoreaction (e.g., high light intensity, high temperature, and low humidity). In particular, during the transition between conditions (i) and (ii), the difference between maximum and minimum conversions for the unfolded state (∆C_unfold) is calculated to be 48%, and for the folded state, ∆C_fold is 25%, whereas ∆C_TOM is 55% during the same transition (Fig. 5E). The conversion difference in the shape-fixed TOM results only from the external light intensity at different conditions. In contrast, the free TOM obtains a higher conversion difference than its fixed states because the device morphs to different shape according to the light intensity profiles. In this process, TOM senses ambient circumstances and tunes the conversion to the highest or lowest value possible, allowing an increase in the magnitude of the regulatable conversion rate. Thus, TOM’s stimuli-responsive morphing feature offers a positive feedback control that facilitates the adaptative photosynthesis in our system.

An adaptive system is able to adjust its behavior in response to changes and uncertainties in the environment or in the system itself (62). In our case, the external environment is simplified as the expected environment for photosynthesis and uncertainties according to the values of illuminance, temperature, and humidity (Fig. 5F). Under different circumstances, the system can adapt itself to get a different output. The more favorable the environment is for the photoreaction, the higher rate the conversion will be reached, and vice versa. This positive feedback is distinct from the negative one that maintains stability (e.g., maintenance of body temperature in biological systems) and divides the output toward the two poles (the highest possible or lowest possible conversion). However, it is this feature that plays a crucial role in a self-sustainable system to harvest, conserve, manage, and use the limited energy. Once the feedback mechanism is established, the management of energy and resources can be achieved. Hypothetically, the products obtained with high conversion (60 to 100%) can be defined as qualified products, while the “products” from low conversion (0 to 20%) can be regarded as acceptable raw materials for recycling without changing any other parameters in the production flow. In terms of energy and raw material management, the former converts the light energy to chemical energy, positively driving the production; the latter preserves the raw materials, pausing the production.

DISCUSSION

The reconfiguration and regulation of our morphing microfluidic device can be transformed from 2D to 3D or between different 3D structures. The dynamic switching between different 3D structures adds a temporal dimension, making our device de facto 4D. Inside the 4D microfluidic device, regulation of the fluid behavior is conducted by the reconstruction of microchannels with certain properties, for example, the orientation, mixing efficiency, and flow rate. When fluids interact with the external environment, such as during the photosynthesis reaction in our demonstration, this fluid regulation amplifies the effect of the interaction even more. The trigger for this series of reconfigurations is a change in stimuli in the environment. Therefore, fundamentally, our approach endows the microfluidic system with a plant-inspired way to self-adapt to the environment through morphing. It will be functionally closer to the biological vascular system with enhanced environmental adaptability than traditional microfluidic systems with fixed channel structures and relying on external devices for regulation. This biomimetic concept can be extended to soft microsystems that need to adapt to the environment in the real world, such as wearable electronics that self-adapt to the body’s environment and bionic soft robots that work in a changing environment.

Although 4D-printed microfluidics, a concept similar to transformable microfluidics, has been proposed, it relies on 3D printing of stimuli-responsive materials (63, 64). Two issues would hamper this: (i) Most of the responsive materials developed using traditional soft lithography lack a set of well-established and compatible micromachining for microfluidics; (ii) 4D printing is still in its infancy stage, where it is not yet involved in morphing microtubing at a resolution that is compatible with soft lithography (65, 66). Moreover, our soft devices made by in situ doping and surface modification are all-in-one systems. This avoids not only the errors and matching problems that may occur during the assembly but also the potential detachment caused by instability at the interface of different components during use and eventually increases the robustness of the whole system. The current TOM still has room for improvement. For instance, the resulting 3D microchannel structure can only be formed via the folding of a predesigned 2D channel structure and cannot be transformed into an arbitrary 3D structure. Besides, the response speed of the current TOM is hindered by the slow swelling or shrinkage rate of the hydrogel active layer and the thick PDMS microfluidic layer as the passive layer. These could be improved by developing advanced fabrication techniques and incorporating high-performance responsive components into the material. Thus, further optimization for TOM will be implemented by (i) introducing alternative methods for designing 3D structures, such as krigami; (ii) selecting actuators with higher photothermal efficiency and faster response, such as coupling gold nanorods with liquid crystal elastomer; and (iii) fabricating a thicker active layer and a thinner passive layer in TOM.

In general, microfluidics combined with smart materials such as functional hydrogels or elastomers has seen progress during the last decade. However, the study has focused largely on inserting smart materials in the microchannels or modifying the overall matrix material of the device; the programmable morphing of the microfluidic device as a whole is not taken into account. For the TOM system, we propose three potential developments for future research: (i) a better correlation between morphing and fluid in TOM, such as morphing-induced fluid channel switching or fluid-induced morphing; (ii) a wider range of hybrid TOM systems, such as integrating biomaterials into TOM to achieve an organ-on-chip with adaptive rhythmic movements; and (iii) a fully autonomous TOM system, such as a more comprehensive plant-like system combining nastic movement, capillary action, and transpiration.

In conclusion, we pioneer plant-inspired morphing origami microfluidics that truly fulfills adaptive photosynthesis. The device coordinates stimuli-responsive morphing materials with a microfluidic chip, which is based on self-actuating elastomer responses to ambient temperature, humidity, and light irradiance, morphing following the preset origami folds. This morphing is further applied to regulate photosynthetic conversion with a built-in positive feedback control in the system. The morphing microfluidics is a smart matter–based intelligent system that could open up a pathway toward the development of intelligent soft devices and artificial vasculature in industrial and biomimetic applications.

MATERIALS AND METHODS

Materials

A Sylgard 184 silicone elastomer kit was purchased from Dow Corning (DOWSIL, USA). GNPs were purchased from Suzhou TANFENG graphene Tech Co. Ltd. (China). N,N′-methylene-bis-acrylamide (BIS), MB, acetonitrile (ACN), and trichloro(1H,1H,2H,2H-perfluorooctyl)silane were purchased from Sigma-Aldrich (Merck KGaA, Germany). NIPAM, BP, and DPA were purchased from Aladdin Industrial Corporation (China). Chloroform was purchased from Thermo Fisher Scientific. Deionized (DI) water was produced by a DI water system (Merck KGaA, Germany). All chemicals were used as received.

PDMS was prepared by mixing Sylgard 184 silicone elastomer base with 10 weight % (wt %) of its curing agent. NIPAM solution was prepared by dissolving 9 wt % of NIPAM and 0.2 wt % of BIS in DI water. BP solution was prepared by dissolving 40 wt % of BP in chloroform. BP-PDMS was prepared by mixing 7 wt % BP solution in PDMS prepolymer. G-PDMS was prepared by mixing 7 wt % BP solution and 2 wt % GNPs in PDMS prepolymer.

Preparation of thin microfluidic device

A negative mold with a channel of 120-μm height and 400-μm width was fabricated by a maskless photolithography machine (SF-100 Xcel, Intelligent Micro Patterning LLC, UK) using SU-8 (2075, KAYAKU, USA) on a silicon wafer (N100, University Wafer, USA). The mold was placed in a box with a piece of Kimwipes tissue wetted with 4 μl of trichloro(1H,1H,2H,2H-perfluorooctyl)silane for 48 hours. Afterward, it was taken out and placed on a hot plate at 100°C for 3 hours to evaporate the unbonded silane molecules. As a result, a uniform perfluorosilane coating was applied to the mold, increasing the surface’s hydrophobicity. To produce a thin microfluidic device, PDMS was first spin-coated (Model WS-650MZ-23NPPB, Laurell, USA) on top of the mold at a speed of 300 rpm for 1 min before being placed in a vacuum box to remove air trapped inside the mold (fig. S3). Afterward, it was placed on a hot plate at 85°C for 15 min. After curing, a BP-PDMS layer was spin-coated at 500 rpm for 1 min, and a frame of filter paper was embedded into the wet BP-PDMS. It was placed on a hot plate to cure PDMS and then was peeled off from the mold. In the next step, the bottom layer membrane was fabricated by spin-coating BP-PDMS on a clear wafer at a speed of 500 rpm for 1 min and was cured at 85°C for 15 min. The designed patterns were hollowed out by a cutter and an equivalent amount of G-PDMS was injected back into the empty spaces. The wet G-PDMS was leveled for 10 min and then was cured on a hot plate. Afterward, a layer of PDMS was spin-coated on the cured BP-PDMS membrane with G-PDMS patterns at a speed of 500 rpm for 1 min. Following curing, an additional thin layer of PDMS was spin-coated on their top at a speed of 8000 rpm for 1 min to serve as an adhesive layer (an uncross-linked PDMS layer of about 2 μm). After punching the inlet and outlet on the top layer, the top layer and bottom layers were brought into contact and left at room temperature overnight. Curing overnight allowed the PDMS to diffuse into both sides and formed a very strong bond. Last, it was placed on the hot plate at 85°C for 30 min to cure completely. The thin microfluidic device was peeled off from the wafer, and the paper frame was cut away.

Hydrogel surface modification on the device

Photomasks (2-mm-thick PMMA plate covered by UV-resistant polyethylene protection tape) and protective films (300-μm-thick silicone film) with hollowed-out patterns were first prepared by laser cutting (fig. S4). The thin device was aligned, stacked, and clamped following the top-to-bottom sequence shown in fig. S4. To modify the hydrogel pattern on the surface of the thin microfluidic device, NIPAM solution was filled into the pattern wells and then placed the assembly in a UV chamber (ABM-USA Inc., USA) for 60-min irradiation. Afterward, the assembly was flipped over and given a similar UV treatment. Eventually, the hydrogel surface-modified device was released and was washed by DI water three times. Then, it was trimmed on the basis of the designed alignment marks. Flexible microtubing (0.20 mm inside diameter × 0.36 mm outer diameter) was inserted into the microchannels in the distal end of the device and sealed by a small quantity of PDMS.

Microscopy

The cross sections of three regions on the device were imaged by using an SEM (Hitachi S3400N, Japan). EDX scattering was used to obtain the elemental mapping of various elements in the cross section.

Responsiveness testing

For all kinds of responsive morphing of TOM, the digital camera (EOS 70D, Canon, Japan) was set directly above the TOM device to monitor the horizontally projected area change during the response. In particular, for the temperature responsiveness in the air, the folded TOM was placed on a heating plate set at different temperatures (20°, 40°, 60°, and 80°C), and its unfolding transformations were recorded by the camera. For temperature responsiveness in high humidity environment (in water), the unfolded TOM was immersed in water at different temperatures (6°, 27°, and 40°C), and its folding transformation processes were recorded while the whole device was submerged in water. For light responsiveness, the folded TOM was placed around 20 cm away from the solar simulator (Halogen lamp, PHILIPS, Netherlands). Then, we adjusted the power of the lamp to change the light intensity at the device position (0, 10, 30, and 50 klx) and recorded the unfolding processes of TOM under different light intensities, respectively. During the experiments, the TOM was used at a temperature far below the pNIPAM decomposition temperature of 350°C to avoid severe shrinkage of pNIPAM that could possibly cause reverse deformation of the device. In this study, we defined the state of the TOM device as “wholly folded” after 20 min of treatment in iced water (temp. ≈ 6°C), while the state of the TOM device was defined as “wholly unfolded” after 20 min of light irradiation (illum. ≈ 30 klx) and subsequent flattening on a flat platform. The unfold percent X for the transition stage is calculated using the following equation: X(%) = (A_X − A_0%) ÷ (A_100% − A_0%) × 100%, where A_X is the horizontally projected area of a morphing device and A_0% and A_100% denote the horizontally projected areas of a wholly folded and unfolded device, respectively. The unfold percent was used as a quantitative parameter for device folding and unfolding.

Infrared thermal imaging

The light-triggered photothermal morphing of the devices was determined by using a thermal camera (ETS320, FLIR, USA). The folded TOM devices with GNPs doping and without GNPs doping were placed at a distance of 20 cm from the solar simulator (illum. ≈ 30 klx). Their unfolding transformations were recorded by the digital camera and thermal camera, respectively. The temperature values of the devices can be calculated with the software FLIR Tool.

Outdoor experiment setup

The outdoor experiment was performed on the campus of the University of Hong Kong on 13 September 2020 between 1:10 p.m. and 1:30 p.m. (sunny day) and 15 April 2021 between 1:05 p.m. and 2:05 p.m. (rainy day), respectively. A camera was used to capture the top-view images of the device’s folding and unfolding periods from directly above. Meanwhile, humidity, temperature, and illuminance meters (GM1361, BENETECH, China; GM1030C, BENETECH, China) were applied to monitor changes during the experiments under the same conditions. In the sunny day test, the device was first folded by being immersed in water at 6°C for 10 min for initialization, and then its unfolding process was recorded under sunlight. Similarly, in the rainy day test, the device was unfolded in advance by 30-klx irradiation for 10 min for initialization, and then its folding process was recorded in the rain.

Finite element analysis of elastic strain and ray optics

The elastic strain simulation was conducted on commercial finite element analysis software (ANSYS, USA). The pNIPAM parts and one PDMS part were modeled as distinct parts and bonded together with no interfacial displacement. To enhance the stability of the solving process, the geometry model was meshed with four-node linear elements, and the adaptive meshing was applied during solving. To accurately represent the large deformation of the TOM, the pNIPAM material and the PDMS material are defined as hyperelastic material. At the initial state, the TOM was in an unfolded state with no internal stress. During the folding process, the pNIPAM parts were set to swell increasingly and isotropically as time increased, while the PDMS part was under passive deformation in response to the pNIPAM expansion. The deformed shapes of the whole TOM and the elastic strain distribution at different time steps were solved. In the simulation, the effect of the microchannels on the deformation of TOM is negligible due to their smaller sizes, and only one channel embedded in each actuator region.

The ray optics simulation was conducted by using the Ray Optics Module of a commercial finite element analysis software (COMSOL Multiphysics, USA). We first imported the 3D models (unfold percent = 100, 50, and 0%) formed during the elastic strain simulation into COMSOL. Then, to simplify the model material to focus on studying the transmission of light in the TOM, we defined the whole device as PDMS with a refractive index of 1.4 and set its extinction coefficient to be 0. In addition, the rays were set to release from the grid (length × width = 50 mm by 50 mm, interval = 1 mm) downward to the 3D models with a wavelength of 654 nm and an initial intensity of 1000 W/m². The rays exited the starting plane, went through the plane where the TOM models were placed (40 mm from the starting plane), and arrived at a terminal plane 80 mm from the starting plane. In the results, we studied the trajectories of the primary rays (i.e., incident rays and transmission rays) and the spot diagrams of on the terminal planes.

Liquid pressure testing

As drawn in fig. S9, syringe pumps, T-pipe, fiber optic sensor (LifeSens, Opsens Inc., Canada), pressure gauge (LifeSens, Opsens Inc., Canada), TOM, and liquid waste container were assembled. The sensor was inserted into the T-pipe after zeroing out the reading at atmospheric pressure. Water was injected into the tubing and microchannel to drive the air out of the system and then inserted the sensor into the T-pipe. A flow rate of 2000 liters/hour was used during the test. A solar simulator and humidifier were applied to change the external light irradiation, temperature, and humidity so as to trigger the morphological transformation of the TOM.

The Reynolds number in our case is estimated to be 2.14, so the fluid flows in the laminar regime in our pressure-driven microfluidics (61). Hence, the pressure difference ∆P = p_in − p_out = p_in – p₀ = Q∙R_hyd, where the hydraulic resistance R_hyd ≈ αμL/wh³; Q represents the flow rate; p_in, p_out, and p₀ represent the pressures at the inlet, outlet, and in air, respectively; μ represents the viscosity of the fluid; L, w, and h represent the length, width, and height of the channel for rectangular cross section, respectively; and α is a coefficient related to the shape of the cross section (1, 61). In our case, Q, α, μ, and L are constant. Therefore, we have ∆P = Q∙αμL/wh³, i.e., ∆P is inversely proportional to wh³. The cross-section size is increased by unfolding, causing a reduction in hydraulic resistance and, eventually, pressure drops.

Spectrum testing

The spectrum of sunlight and the solar simulator were determined by using a high-resolution fiber optic spectrometer (HR2000 + CG-UV-NIR, Ocean Optics Inc.). The absorption spectra of the solutions were tested with a UV-visible spectrophotometer (UV-2600, SHIMADZU).

Design and working method of an optical flow cell

The optical flow cell was designed using AutoCAD (Autodesk) and fabricated via precision micromilling (LPKF ProtoMat S100) of black PMMA as described previously (67). Briefly, the flow cell is composed of two interlocking micromilled PMMA pieces with a channel cut into both pieces such that the 0.7-mm OD tubing (polytetrafluoroethylene, thin wall tubing, UT5, Adtech Polymer Engineering Ltd., UK) can fit into it. The light-to-voltage converter (photodiode, TSL257, ams AG, Austria) and light-emitting diode (LED; 370 nm; SST-10-UV-B130-E365-00, Luminus Devices, UK) are placed in the micromilled parts in a way that light emitted from an LED travels to an opposing photodiode via tubing (path length of 0.5 mm). It has a hole in both parts to provide a light path from LED to photodiode through the tubing. A signal is collected using Arduino nano that relays the signal to the computer, and data are recorded with LabVIEW (National Instruments, USA) as described here (68).

Photosynthesis

For the cycloaddition of DPA with singlet oxygen, solutions consisting of 0.5 mM MB and 0.5 mM DPA in ACN were used. Stock solutions (20 ml) of the two reagents were prepared. The two solutions are mixed and loaded in a 10-ml syringe connected with tubing. The solution of DPA in the syringe and from the syringe to the microreactor was wrapped with aluminum foil, as well as the tubing between outlet and detector, to prevent the reaction from occurring outside the device. The conversion was monitored with an optical flow cell (detector). In particular, the DPA peak was measured at 372 nm. The absorption with 0% conversion was measured immediately after TOM without any light irradiation, while the full conversion value was acquired by exposing a closed glass vial of the reaction mixture to direct irradiation from the solar simulator for at least 30 min. The conversion of the starting material was determined from the absorption peak with the following equation: C(%) = [Abs_DPA+MB(0) − Abs_DPA+MB(χ)] ÷ [Abs_DPA+MB(0) − Abs_MB(0)] × 100%, where the C(%) is the conversion rate, and Abs_DPA+MB(0) and Abs_DPA+MB(χ) are the initial and residual absorbance of the mixed substrate solution, respectively. The Abs_MB(0) is the initial absorbance of the MB solution.

Real-time photosynthesis experiment setup

In this study, we used the same setup in all photosynthesis experiments, as shown in Fig. 5A. The TOM device was placed in a light-proof chamber with an opening at the top. The inner wall of the chamber is covered with black sandpaper to reduce the reflected light. A solar simulator was set directly above the TOM device. The inlet of TOM was connected with syringe (flow rate = 1000 liters/hour), while the outlet was linked with the designed optical flow cell by tubing. The syringe, tubing, and optical flow cell were all protected with aluminum foil to avoid additional photoreactions occurring somewhere other than the TOM device during the experiments. The real-time absorbance in the optical flow cell can be recorded and processed by a computer, while the photosynthesis products can be collected from the outlet of the optical flow cell.

There are two real-time photosynthesis experiments in this study:

1) Monitoring the photosynthetic conversion by changing the light intensity in unfolded and folded states of TOM respectively. For the unfolded state of the TOM, we kept the device folded by immersing it in iced water throughout the experiment. For the folded state of the TOM, we allowed it to unfold completely by preirradiating it with 30 klx for 10 min and then kept it open in air for the whole reaction process. During the experiment, the light intensity started from 0 klx and was raised by 2 klx every 10 min until it reached 8 klx.

2) Monitoring the photosynthetic conversion when changing the external environment. Two external environments were designed for this experiment. Situation 1: the normal temperature and humidity with relatively high light intensity (i.e., temp. ≈ 25°C; R.H. ≈ 50% RH; illum. ≈ 6 klx); situation 2: low temperature and high humidity with a relatively low light intensity (i.e., temp. ≈ 6°C; R.H. ≈ 100% RH; illum. ≈ 2 klx). This experimental environment started with 0-klx light intensity, then changed to situation 2 lasting 10 min, then to situation 1 for 30 min, then back to situation 2 for 30 min, and, lastly, returned to 0-klx light intensity. The following three states of TOM were examined: transformable, unfolded, and folded.

Supplementary Materials

This PDF file includes:

Figs. S1 to S9

Other Supplementary Material for this manuscript includes the following:

Movies S1 to S7