Perspiring Soft Robotics Skin Constituted by Dynamic Polarity‐Switching Porous Liquid Crystal Membrane

Secretion of functional fluids is essential for affecting surface properties in ecosystems. The existing polymer membranes that mimic human skin functions are limited to secreting, either apolar or polar, liquid. However, the development of membranes that grant exchange liquid with different polarities remains a grand challenge. This process is prohibited by the mismatch of the polarity between the carrier polymer and the loaded liquid. To conquer this limitation, an innovative strategy is reported to dynamically switch the polarity of the porous membrane, thereby empowering the exchange of apolar liquid with polar liquid and vice versa. This approach incorporates a benzoic acid derivative into the original apolar polymer network. The benzoic acid dimerizes and forms hydrogen bonds, which supports the molecular alignment, but can be broken into the ionic state when subjected to alkaline treatment, changing the polarity of themembrane. Consequently, the apolar liquid can be replaced with a more polar one. This polar liquid is ejected upon safe‐dose UV illumination from the membrane. Reabsorption occurs on demand by illumination of visible light or when left in contact with the membrane, spontaneously in the dark. Based on this, the consumed membrane is replenished with the same or different exchanging liquid.


DOI: 10.1002/adma.202211143
intriguing liquid secretion phenomena of plants have boosted the development of artificial systems. [3,4] The field of intelligent liquid release research has expanded considerably in the past decade. Various rational designs aimed at structures, chemical compositions, and surface chemistry have been extensively investigated to realize dynamic polymer membranes. For synthetic systems demonstrated so far, liquid-based porous coatings showcase outstanding utilitarian functions in, among others, antifouling, [6] surface adhesion control, [7,8] and cargo release, [9,10] owing to the infusing liquid's flexibility, liquidity, and surface tension property.
Previously, supramolecular polymer gels with self-regulated liquid secretion based on the reorganization of hydrogen bonds were reported. [4] Nevertheless, the secretion is spontaneous and non-controllable. Lately, our group developed liquid crystal polymer skins which enable ondemand perspiration and reverse-perspiration of the same liquid, controlled by external stimuli. [11][12][13][14] In this work, we present a new generation of the dynamic polarityswitching membrane which empowers the exchange of liquids even with properties at the opposite of the polarity spectrum. To this end, we utilize liquid crystal polymer networks (LCNs) as the liquid reservoir. LCNs are a typical class of materials that have anisotropic characteristics which have been extensively employed in optical devices, [15][16][17] soft actuators, [18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33] and dynamic surfaces. [5,[34][35][36][37][38][39][40][41] We take advantage of the reduction of molecular order, upon the actuation of the LCNs, [42,43] to eject the initially stored primary liquid to the coating surface. Subsequently, we alternate the polarity of the porous LCNs to match the secondary liquid to be exchanged (Figure 1a). To achieve this, we covalently bind a benzoic acid derivative to form an integral component of the LCNs. [44] The hydrogen-bonded benzoic acid dimers change into an ionic state upon base treatment, simply by a short immersion step, thus, the membrane's capillary properties from hydrophobic to hydrophilic. This method is so efficient that only requires a small amount, as low as 5%, of the co-polymerizable benzoic acid, added to the LCN.
We fabricated our "perspiring" porous liquid crystal polymer skins by introducing phase separation of non-reactive porogen and reactive mesogens during the photopolymerization process. [45,46] The photocuring kinetics determine the pore sizes of the membrane. [47] For the best secretion properties including secretion amount and rate, we have chosen a smectic molecular Secretion of functional fluids is essential for affecting surface properties in ecosystems. The existing polymer membranes that mimic human skin functions are limited to secreting, either apolar or polar, liquid. However, the development of membranes that grant exchange liquid with different polarities remains a grand challenge. This process is prohibited by the mismatch of the polarity between the carrier polymer and the loaded liquid. To conquer this limitation, an innovative strategy is reported to dynamically switch the polarity of the porous membrane, thereby empowering the exchange of apolar liquid with polar liquid and vice versa. This approach incorporates a benzoic acid derivative into the original apolar polymer network. The benzoic acid dimerizes and forms hydrogen bonds, which supports the molecular alignment, but can be broken into the ionic state when subjected to alkaline treatment, changing the polarity of themembrane. Consequently, the apolar liquid can be replaced with a more polar one. This polar liquid is ejected upon safe-dose UV illumination from the membrane. Reabsorption occurs on demand by illumination of visible light or when left in contact with the membrane, spontaneously in the dark. Based on this, the consumed membrane is replenished with the same or different exchanging liquid.

Introduction
Nature always finds its way to survive. [1,2] Guttation, first discovered by Bergerstein in 1887, is a conceptually similar phenomenon to sweating for the removal of excess moisture and numerous chemicals including enzymes and minerals. Typically, vascular plants and fungi exchange fluids with the associated soil/air environment via water absorption at roots and guttation at leaves to maintain a delicate balance of water and indispensable nutrients for their healthy growth. Discoveries of alignment in which the mesogens are aligned homeotropically in layers (Figure 1a). The LCN coating with such alignment contracts along the coating thickness and thereby repels the initially stored liquid upon external stimulus application. [14] The materials we employed are shown in Figure 1b. Molecule 1, a non-reactive liquid crystal, served as porogen and established the desired smectic molecular alignment. Reactive liquid crystal mesogens 2 and 3 were the building blocks to form a smecticordered polymer network. We further added azobenzene derivative 4, to provide light responsiveness. To access polarity con-version after the formation of the polymer network, we introduced a benzoic acid moiety functionalized monoacrylate 5, viz. OBA. To prevent premature isomerization of the azobenzene compound, photoinitiator Irgacure 819 was used to generate free radicals under exposure to light above 400 nm which initiates the chain-addition polymerization reaction. Prior to photocuring, the smectic homeotropic alignment was promoted by a thin alignment layer. This alignment was frozen in upon photopolymerization, as shown in Figure 1c. More details of the coating fabrication are shown in the Methods section. The  structure was confirmed by optical microscopy and grazing-incidence small-angle X-ray scattering (Figures S1, S2, Supporting Information).
In the first experiment, to check on the membrane's porosity properties, we exchanged the initially locked-in liquid crystal porogen 8CB (4′-Octyl-4-biphenylcarbonitrile) with a different liquid crystal 5CB (4′-Pentyl-4-biphenylcarbonitrile), molecule 7, (shown in Figure 1b) that has a shorter aliphatic chain (Figure 1a). 8CB was removed by solvent extraction, as confirmed by the disappearance of characteristic absorption of the nitrile group peaked at 2227 cm −1 measured by Fourier transform infrared (FTIR) spectroscopy (blue line in Figure 1d). Subsequently, the membrane was refilled with 5CB by capillary action, and recorded by the appearance of the re-established nitrile band at 2227 cm −1 (red line in Figure 1d). We successfully exchange liquid with the same polarity, still, the exchange of liquid with an opposite polarity remains unachieved ( Figure S3, Supporting Information). Therefore, we advance the membrane system to the next level by exchanging the liquid with distinct polarity. To achieve this goal, we create a dynamic polymer network that inverses its polarity between the apolar state and polar state by introducing OBA, molecule 5. In its processing stage, the benzoic acid was dimerized through intramolecular hydrogen bonds. The dimers in their monomeric state are liquid crystalline and support the processing of the monomer mixture in its liquid crystal state. After polymerization, the benzoic dimer still behaves moderately apolar. However, upon breaking the hydrogen bonds via alkaline treatment, carboxylic acid salts are generated, which switches the polarity of the polymer network. The process is reversible as by washing the ions through treatment with a lower pH solution, we can restore the initial low polarity state. Experimentally, we developed the following procedures ( Figure 1e). After the initial formation of the porogen containing LCN, we removed the porogen by solvent extraction in cyclohexane. The emptied coating was analyzed by scanning electron microscopy (SEM). Cross-section and top-view images exhibit a clear porous structure of the coating with pore sizes ranging from 500 nm to 1 µm (Figure 1f, Figure S4, Supporting Information). Due to the porous structure, the elastic surface indentation modulus as measured by atomic force microscopy (AFM) is 3 MPa on average (Figure 1g,h), compared to the densely crosslinked polymer network with a modulus of several GPa. To alter the polarity, we broke the hydrogen bonds of the benzoic dimers upon subjection to a base solution. Results were confirmed by various means, FTIR, SEM, and UV-vis measurement. The FTIR spectra clearly show that the featured absorption of hydrogen bridges peaked at 1680 cm −1 , which corresponds to the stretching of CO group of the benzoic dimers, vanishes and two new absorption bands peaked at 1384 and 1550 cm −1 appearing due to the symmetric and anti-symmetric stretching of COO − group (Figure 1i). Since the hydrogen bonds are broken, the crosslink density is reduced changing the LCN rigidity, and the pores collapse and close the membranes, as shown in Figure 1j, Figure S4, Supporting Information. Simultaneously, this leads to the order parameter reduction which can be estimated by measuring the UV absorption of the azo dye parallel to the molecular direction. The absorption of azo dye peaked at 365 nm and increases after hydrogen bond breaking, indicating order parameter decrease ( Figure S5, Supporting Information).
After the conversion of the benzoic dimers into free carboxylic anions, the membrane with inversed surface energy is now ready to be refilled with polar liquid by bringing the surface in contact with the liquid. We chose ethylene glycol as the exchanging polar liquid for its relatively low viscosity and volatility. The enhanced interaction of ethylene glycol and polar group at the internal pore surfaces stimulates rapid capillary enforced replenishing to a new equilibrium (Figure 2a-d).
As the total in-plane membrane area does not change during depletion and refilling, we can calculate the (re)filling efficiency from the coating thickness change, assuming that in the empty state, all pores are fully closed ( Figure S3, Supporting Information). Results show that the overall refilling efficiency, irrespective of sample composition, is above 80%. To stimulate liquid secretion and uptake, we used light as the trigger. [48] To quantify the liquid release, we converted absolute surface height change, as defined by h i − h a s (Figure 2a), by a simple calculation into the volume of ejected liquid, given the conservation of volume under the assumption that the unoccupied voids are filled by the polymer after repelling liquid. We reached the secretion efficiency, which is defined as the volume of secreted liquid over the total refilled liquid volume in the coating, up to 10%. To ensure UV exposure at a safe level, we limited the UV dose below 1.0 J cm −2 (Figure 2e). [49] We monitored the liquid secretion and uptake process at the surface by optical microscopy and digital holographic microscopy (DHM) (Videos S1, S2, Supporting Information). Figure 2a illustrates the secretion and uptake of ethylene glycol upon light illumination. Figure 2b,c depicts a dry and flat surface before UV irradiation. When exposed to UV light, ethylene glycol secretes and coalesces forming big droplets, uniformly distributed at the surface. As a result, the surface height decreases accordingly (Figure 2f). When shining blue light, the released liquid is reabsorbed, as seen from the fully recovered surface height. Results show that the liquid secretion and uptake are completely reversible, and there is no considerable liquid loss during activation in the first seven cycles (Figure 2g).
Next, we systematically investigated the influence of several design parameters, including the concentration of OBA, porogen, and azobenzene derivative, on secretion. When adjusting the content of each parameter, we kept the ratio of the rest of the components constant. Figure S6, Supporting Information and Figure 2h plot the uptake and secretion quantity against OBA concentration, respectively. In the absence of OBA, it is obvious that no secretion occurs. This is because without alkaline ions the low surface-energy LCN fails to refill with polar liquid. In the presence of the OBA, liquid secretion takes place. These results demonstrate that polarity is the key to liquid exchange. We vary the concentration of OBA from 5% to 20%. We observed that at 5% the polarity can already be modified significantly. With increasing OBA concentration, secretion quantity and efficiency increase, resulting from the decrease in surface modulus of the coating, as the refilling efficiency is above 90% despite OBA content (Figure 2i, Figure S7, Supporting Information). Because the secretion efficiency is the highest, while the LCN still maintained its robustness, we choose 20% OBA for further investigations. The initial porogen concentration also impacts the secretion as it is directly related to the porosity of the membrane. As shown in Figure 2j, the secretion maximum appears at 70% of the porogen. Also here there is a trade-off in the uptake of ethylene glycol and the mechanical properties of the membrane. Above a value of 70%, the LCN is not mechanically strong enough to generate the contraction forces to repel liquid. To proceed further with LCN obtained from a 70% porogen concentration, we varied the azobenzene concentration. In general, the ejection quantity increases with the azobenzene concentration, due to the large contraction force generated from trans-to-cis photoisomerization of covalently bonded azobenzene moiety (Figure 2k). Above 8% this effect levels off, presumably because of reduced penetration due to the azobenzene's light absorption.
Besides exchanging and secreting a single component, we also studied various liquid blends. For example, we blended ethylene glycol with water, as water is the most essential element for living organisms, and further investigated their influence on the secretion. Results show that the secretion occurs when the water concentration is below 50%, and the secretion efficiency decreases with increasing the water content (Figure 2l, Figure S8, Supporting Information). This is attributed to low refilling efficiency resulting from the leaching-out of potassium ions from the polymer network and dissolving into filling liquid during solvent-refilling treatment. [50] We also studied polyethylene glycol (PEG), dimethyl sulfoxide (DMSO), and distilled water as exchanging liquid blends due to their wide use in the medical field. We observed the secretion after UV light exposure both microscopically and macroscopically (Figure 3a, Video S3, Supporting Information). We analyzed the liquid release and uptake for various volume ratios of liquid blends of PEG-water and PEG-DMSO, respectively. For the PEG-water mixture, the secretion rises with increasing water percentage, as PEG is diluted, peaks at 20% of water, and ceases at higher water concentrations which results from the low refilling efficiency caused by leaching-out of ions from the polymer network (Figure 3b). In terms of the PEG-DMSO blend, the secretion maximum occurs at 90% of DMSO, below which, the secretion quantity increases with increasing DMSO due to its low molecular weight; above which, the secretion decreases, as liquid evaporation plays a dominant role (Figure 3c). We also observed secretion loss in both blends over secretion-uptake cycles, stemming from evaporation (Figure 3d, Figure S9, Supporting Information). Compared to the PEGwater blend, the PEG-DMSO blend exhibits higher evaporation rates. This is ascribed to the presence of formed intermolecular hydrogen bonds in the PEG-water. Moreover, we precisely quantified the components of the secreted PEG-DMSO blends via confocal Raman microscopy analysis. In the Raman spectra of liquid blends, we focused on the characteristic shift of DMSO peaked at 676 cm −1 (Figure 3e). Calibration measurements, presented in Figure S10, Supporting Information, reveal the relationship between the intensity at 676 cm −1 and the actual concentration of DMSO. As indicated in Figure 3f, the actual concentration of DMSO in the blends highly aligns with the theoretical one, suggesting homogeneity of the secreted liquids. PEG-water blend cannot be analyzed due to the overlapping of their featured absorption peaks ( Figure S11, Supporting Information).
With the ethylene glycol-filled membrane, we now explore a few potential applications demonstrated in  placed in a simulated frozen environment with a fixed temperature of −20 °C and relative humidity of 90% ( Figure S12, Supporting Information). As shown in Figure 4a,b, the membrane surface without exposure to UV light appears icing after 1 h freezing treatment with 100% ice coverage on the surface, the same as bare glass. In contrast, the UV-exposed membrane surface stays liquified and does not show icing during the same freezing treatment (Figure 4c). Our proposed anti-icing approach provides an alternative solution to the reported ones which are based on complex micro-or nanoscale hierarchical structures. [51] Furthermore, the released ethylene glycol adjusts surface friction. The in situ friction force, with/without light irradiation, was measured by using the same experimental setup as reported. [14] As seen from Figure 4d, the friction force increases by up to 1.7 N under UV exposure, as a result of the released liquid forming a capillary bridge between the glass and coating surface. When irradiating with blue light, the force further increases due to the negative pressure effect induced by the reabsorption of released liquid. Moreover, we explored potential applications of our membranes in biomedical fields. [52] For this purpose, we first estimated the hazards of the LCN membrane ( Figure S13, Supporting Information). We performed residual analysis, particularly on azobenzene derivative, given that other acrylates are fully converted into the polymeric state after the photocuring procedure. Results indicate that all the azobenzene diacrylates are converted and no residual is detected, within the resolution of the FTIR test around 300 ppm. Specifically, for wound infection treatment and even prevention, we showcased the release of 10% vitamin C (VC) carried by the PEG-DMSO blend. Raman spectra exhibit the featured scattering of VC peaked at 1697 cm −1 (Figure 4e). Based on calibration measurement displaying the intensity at 1697 cm −1 as a function of actual VC concentration ( Figure S14, Supporting Information), we estimated that the actual concentration of the released VC is 7.64%. Combined with DHM analysis on quantification of liquid release ( Figure S15, Supporting Information), we determined the amount of the released VC over time (Figure 4f), indicating a sustained release in a controlled manner. Besides rigid glass plates as membrane support, we also used flexible substrates, for example, polyethylene terephthalate (PET). We can see from Figure 4g and Figure S16, Video S4, Supporting Information that liquid ejection successfully takes place on the bent PET film. Based on this, we proposed a new generation of adhesive tapes with tunable adhesion (Figure 4h), strong fixation, and easy-release dressing enabled by secretion/uptake of biocompatible liquid. To achieve this, we spray-coated and photocured a thin layer of pressure-sensitive adhesive around 5 µm in thickness on the LCN membrane using the monomer mixture shown in Figure 4i (Figure 4j,k, Figure S17, Supporting Information). We performed a 180° peel strength test to evaluate the peel strength before and after liquid ejection (Figure 4l, Figure S19, Video S5, Supporting Information). As shown in Figure 4m and Figure S18, Supporting Information, peel force decreases by ≈50% with UV illumination.

Conclusions
In conclusion, we developed a dynamic polarity-switching LCN membrane skin that enables the exchange of apolar liquid to polar liquid. In addition to a single liquid component, various liquid blends, for example, ethylene glycol, PEG, DMSO,  and/or water, were also exchanged and refilled for multifunctional uses. We demonstrated a number of applications based on this system ranging from anti-icing, surface friction tuning, and drug release, to surface adhesion control. Special attention was given to drug delivery and easy-release dressing. Our membranes solve the problems of strong adhesion of the existing bio-adhesives, such as Lumina film, by deactivating dressings for painless release using UV or visible light as an external stimulus without exposing them to solvents. Furthermore, existing commercial products or technologies are not reusable, and even if a patient's wound or catheter must be secured for several days, access may be required multiple times each day. Our membranes address this demand for multiple-use and recyclable adhesive tapes by triggering multiple rounds of switching on and off enabled by the secretion/absorption of biocompatible liquid. We anticipate that our active liquidexchanging and transporting membranes will pave the way for the development of soft robotics, biomedical science and technology, and circular economy.  received. Potassium hydroxide (Sigma-Aldrich) solution (0.05 m) was prepared by dissolving solid material in distilled water. 2-Ethylhexyl acrylate, acrylic acid, isobornyl acrylate, and poly(ethylene glycol) methyl ether methacrylate were purchased from Sigma-Aldrich.

Experimental Section
Sample Preparation: Coatings with smectic homeotropic alignment perpendicular to the substrate were prepared by photopolymerization of a liquid crystal monomer mixture in an empty cell. To enhance surface adhesion, a primary plate, glass plate or flexible indium tin oxide coated PET sheet (surface resistivity 60 Ω sq −1 , Sigma Aldrich), was coated with 3-(trimethoxysilyl) propyl methacrylate. The cell was formed by gluing a secondary glass plate treated with a polyimide layer (SE5661, Nissan Chemicals) that provided a homeotropic alignment. The thickness of the coating was confined to 20 µm by using a spacer.
The liquid crystal monomer mixture was filled in the cell by capillary force at isotropic phase and held for 20 min to avoid thermal convection, then slowly cooled down to desired smectic phase of liquid crystal monomer mixture and polymerized under UV light using a UV lamp (Omnicure EXFO S2000). A cut-off filter (Newport FSQ-GG400 filter) was placed between the UV lamp and the sample during the polymerization to filter the light < 400 nm to avoid the premature isomerization of azobenzene moiety. After photopolymerization, the glass slide coated with a polyimide layer was detached by using a razor blade.
Adhesive monomer composition dissolved in ethanol (1:1 in volume) was mixed with ethylene glycol with a volume ratio of 50%. The mixture was spray-coated onto the PEG-DMSO refilled LCN membrane and cured with UV light above 400 nm.
Characterization: Phase-transition temperatures of liquid crystal monomer mixtures were measured by DSC (Q2000, TA Instruments) at a rate of 10 °C min −1 . The thickness of the coating was measured by an interferometer (S Neox Profiler, Sensofar). Dynamic surface height changes were monitored by a digital holographic microscope (DHM-R and DHM-T, Lyncée Tec). The porous structure of the coatings was measured by scanning electron microscopy (FEI Quanta 3D FEG). The nanoscale elastic surface modulus and topography were measured by Multimode atomic force microscope in the PeakForce Quantitative Nanomechanical Mapping (PF-QNM) mode (Bruker Corporation) using HQ: NSC19/No Al (MikroMasch) cantilevers. Liquid secretion was visualized by an optical microscope (Nikon Ci Eclipse). Removal of molecule 1 after solvent extraction, breaking-down of hydrogen bonds after base treatment, and residual of azobenzene monomer after photopolymerization was checked by Fourier-transform infrared spectroscopy with an attenuated total reflectance attachment (Varian 670-IR, FTIR Spectrometer). Refilling of function liquids was confirmed by using a confocal Raman microscope. Confocal Raman measurements were performed at room temperature using the Witec α-300 R µ-Raman system. In Raman maps, 50× objective lens with a numerical aperture of 0.55 was utilized together with 2 s integration time, 300 mm g −1 grating, 633 nm continuous laser with 10 mW laser power. The penetration depth of the laser is 3.8 µm. Prior to Raman data analysis, spike (or cosmic ray) removal, and background subtraction was performed by using Project 5 software. Flat tip indentation experiments were performed using an MTS nano-indenter XP (MTS Nano-Instruments, Oak Ridge, Tennessee), equipped with a tungsten, 300-micrometer diameter, flat punch. The samples were loaded at a (displacement-controlled) rate of 100 nm s −1 to a final depth of 10% of the thickness of the coating. Subsequently, the maximum load was removed from the samples at an unloading rate of 0.5 mN s −1 . For each sample, indents were placed on a 3 × 3 grid with a distance of 3 mm between indents in order to check repeatability and sample homogeneity. The simulated frozen environment was built by constant moisture input driven by airflow and freezer chambers. The temperature and relative humidity of the environment was measured by SEK-SensorBridge.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.