Increasing the Efficiency of Thermoresponsive Actuation at the Microscale by Direct Laser Writing of pNIPAM

Thermoresponsive hydrogels such as poly(N‐isopropylacrylamide) (pNIPAM) are highly interesting materials for generating soft actuator systems. Whereas the material has so far mostly been used in macroscopic systems, here it is demonstrated that pNIPAM is an excellent material for generating actuator systems at the micrometer scale. Two‐photon direct laser writing is used to precisely structure thermoresponsive pNIPAM hydrogels at the micrometer scale based on a photosensitive resist. This study systematically shows that the surface‐to‐volume ratio of the microactuators is decisive to their actuation efficiency. The phase transition of the pNIPAM is also demonstrated by nanoindentation experiments. It is observed that the mechanical properties of the material can easily be adjusted by the writing process. Finally, it is found that not only the total size and surface structure of the microactuator play an important role, but also the crosslinking of the polymer itself. The results demonstrate for the first time a systematic study of pNIPAM‐based microactuators, which can easily be extended to systems of microactuators that act cooperatively, e.g., in microvalves.


Introduction
Active materials for stimuli-controlled microactuation can strongly improve the function of many applications related to lab-on-a-chip devices, from the manipulation of single cells to controlling microfluidic systems. [1][2][3][4] Due to its thermoresponsive properties, poly(N-isopropylacrylamide) (pNIPAM) is an excellent material for generating actuators in aqueous environments. [5][6][7] The thermoresponsive properties of pNIPAM are

Results and Discussion
Thermoresponsive pNIPAM hydrogel microstructures were fabricated by a high-precision DLW approach. [17] In brief, a photosensitive resist containing N-isopropylacrylamide monomers, N,N′-methylenebis(acrylamide) comonomers, and lithium phenyl(2,4,6-trimethylbenzoyl)phosphinate (LAP) photoinitiator is produced according to the work of Hippler et al., which can be polymerized by two-photon absorption. [16] A focused femtosecond pulsed near infrared (NIR) laser (λ = 780 nm) is used Figure 1. Direct laser writing of pNIPAM microactuators. a) The DLW fabrication process is depicted schematically. A pulsed and highly focused NIR laser (λ = 780 nm, t pulse = 100 fs) is used to initiate two-photon polymerization (2PP) of a monomer resin in a very locally confined volume. By scanning the laser both in lateral directions (XY) and in different slices (Z), it is possible to fabricate 3D microstructures based on a computer model. b) Models of different architectures constructed via computer aided design (CAD) are shown (top left) as well as 3D reconstructions of measured confocal fluorescence image stacks of the fabricated structures, each including a system of nine pNIPAM microactuators. On purpose, the three images to the right in each actuator system show the cross section of the actuators. These images prove the applicability of the DLW fabrication process to yield the desired hydrogel microstructures. Scale bars are 20 µm.
to initiate the polymerization in a very small volume (<1 µm 3 ). Figure 1a describes the fabrication process schematically, in which a droplet of the pNIPAM resist is cast onto a glass substrate and the focused laser is scanned in lines and slices to write the hydrogel microstructures based on a computer design. Different pillar designs are depicted in Figure 1b, including cylinders, hollow cylinders, and star-shaped pillars. In addition, 3D reconstructions of confocal fluorescence image stacks of the fabricated microstructures are presented, which demonstrate the success of the DLW process. To reduce shrinking restraints of the micropillars at the hydrogel/glass interface, the structural design contains a small pedestal at the bottom of the pillars, which links the hydrogel to the substrate.
Various sizes and architectures of pNIPAM micropillars were fabricated to investigate the effect of surface-to-volume ratio on the shrinking and swelling properties of the hydrogel microstructures. In a previous work we have demonstrated that structuring pNIPAM at the microscale by introducing an interconnected network of microchannels can drastically enhance the in-and outflow of water and hence the responsivity of the hydrogel even in bulk samples. [10] As DLW allows for a high freedom in 3D design, several material properties were analyzed to study the performance of the microactuators prepared by DLW in this work.
First, the thermoresponsive shrinking and swelling properties of the material were investigated by fluorescence confocal imaging of the pNIPAM microstructures at different temperatures. These results are summarized in Figure 2, where a 2D comparison of cross-sections and a 3D comparison of volumes of the hydrogel structures before and after shrinking at 22 and 45 °C, respectively, are presented. In a first experiment, cylindrical micropillars of various sizes (aspect ratio 2:1, height:diameter) were analyzed, where the cross-section areas in the center of each pillar and the volumes were measured at 22 and 45 °C (Figure 2a). The percentual shrinkage of each microstructure was calculated according to Equation (1) for the 2D case and Equation (2) for the 3D case, where A and V refer to the cross-section areas and volumes at a specific temperature, respectively.
In a second experiment, different architectures of microstructures of the same size were evaluated in the same manner ( Figure 2b). Finally, the percentual shrinkage is presented as a function of perimeter-to-cross-section ratio (2D) and surface-tovolume ratio (3D) in Figure 2c,d, respectively.
From these results a significant increase in shrinkage can be observed as the microstructure dimensions get smaller. Whereas the 50 µm diameter pillar shrinks very little, the 10 µm diameter pillar reduces its volume by 41 ± 3%. A similar trend is observed for the different architectures of micropillars, where the shrinkage increases with larger surface-to-volume ratio. The increase in shrinkage can be related to a facilitated outflow of water from the microstructure, the thinner the sample dimensions get. [18] This means controlling the shape of the microstructures by the fabrication process can drastically improve the actuation capabilities.
To investigate the reversibility of shrinking and swelling, the microstructures were cooled down to 22 °C after the initial shrinking at 45 °C, and the sample dimensions were evaluated again. These results are depicted in Figure S1 (Supporting Information) and prove that the initial dimensions can be restored after cooling, which is in agreement to previous results on pNIPAM microgels in surface coatings. [9] In addition to the temperature-induced volume change, also the mechanical properties of the pNIPAM hydrogel change during the material's phase transition, as they mainly depend on the degree of hydration. The intrinsic mechanical properties of the hydrogel at constant temperature are especially important for the use in soft microactuators, where commonly mechanical work is generated by a volume change. The material stiffness strongly depends on the degree of crosslinking and is reduced for a lower degree, while stiffness increases with a higher crosslinking degree. [19] Therefore, pNIPAM microstructures fabricated by different writing speeds and fixed laser power were studied in nanoindentation experiments. For the DLW process used in this work, the degree of crosslinking increases with higher doses of laser light initiating the polymerization. A higher dose will result in a polymer network exhibiting a higher crosslinking density and therefore different material properties, despite using the same resin. The exact crosslinking degree in the final microstructures remains unknown though. A comparison of writing speed, relative crosslink density and shrinkage can be found in the supporting information in Figure S2 (Supporting Information), which concludes that higher writing speeds result in higher shrinkage and eventually less crosslinking. These findings agree with other studies in this field. [20] Figure 3a depicts the nanoindentation process schematically, where a cantilever with spherical tip is used to indent into the surface of pNIPAM microcubes at several locations. A microscope image showing an array of identical test structures and the cantilever is displayed in Figure 3b. Representative loadindentation curves of microstructures fabricated with a writing speed of 30 mm s −1 are presented in Figure 3c,d for 22 and 45 °C, respectively. From these plots a significant change in surface stiffness is observed before and after the phase transition of pNIPAM microcubes, a 3D structure specifically chosen for the indentation tests. For a moderate writing speed of 30 mm s −1 the effective Young's modulus increases from 19 ± 5 kPa at 22 °C to 475 ± 74 kPa at 45 °C. This increase in stiffness is related to the dehydration of the hydrogel above 32 °C, in which water is flowing out of the material and the polymer network densifies. [21] The results are in agreement with Schmidt et al., who investigated the mechanical properties of pNIPAM microgel films and observed a transition in elastic modulus from less than 100 kPa at 25 °C to above 600 kPa at 47 °C. [22] Furthermore, a relation between writing speed and material stiffness is observed. Increasing the writing speed results in a decreased surface stiffness, both for 22 and 45 °C. This can be explained by a lower dose of laser light for higher writing speeds, which provides less energy to initiate polymerization and thus results in fewer crosslinks. The direct relation  In both cases, the shrinkage increases with increasing surface-to-volume ratio (3D) and perimeter-to-cross-section ratio (2D), respectively. This is explained by the ease of water flowing out of the hydrogel for larger surfaces. c,d) Summary of these findings in plots, where a clear trend of increased shrinkage is observed. Error bars denote standard deviations calculated from nine tested samples each.
between writing speed and material stiffness allows to simply control the mechanical properties of the microstructures and thus highlights the great potential of DLW as a fabrication technology for soft microactuators. [23] For the polymer resist used in this work, a range of writing speeds from 10 to 50 mm s −1 was suitable. Slower writing speeds resulted in microexplosions due to excessive energy input, while faster writing speeds did not cause polymerization at all. However, these limits can also be influenced by other writing parameters, such as laser power, hatching and slicing distances, and the chemical composition of the resist. Altogether these factors offer many degrees of freedom to adjust the material properties and highlight the versatile utilization of the DLW technology presented in this work. To demonstrate the actuation capabilities and design freedom of the actuators fabricated with our DLW technology, two examples of soft microactuator systems were designed. First, temperature controlled microvalves were fabricated, which are depicted in Figure 4a. The valves have a diameter of 60 µm and consist of a ring-shaped base with a circular lid on top. The lid is made of several triangular sections arranged in circular fashion such that one end of each section points in the valve center to create a round opening. The base is fabricated with a low writing speed and thus shrinks less, while for the lid a high writing speed is used, which results in a significant shrinkage. Therefore, the opening of the microvalves becomes larger upon heating and smaller upon cooling (cf. dashed red circle). Furthermore, the magnitude of shrinkage and opening increases with higher writing speeds, which is related to a less crosslinked polymer network for high writing speeds. This observation confirms the previously mentioned benefit of the DLW technology to adjust the material properties easily by, e.g., varying the writing speed in situ. The microvalve structures shown here can find application in capturing and release of microobjects, microparticles, and bioactive cargos like cells or bacteria.
Another possible application is presented in Figure 4b, which illustrates a system of pNIPAM micropillars that can be actuated by temperature. Here, a square array of 100 micropillars is heated to 50 °C to induce a shrinkage and a change in gap size between the pillars. This temperature-controlled variation in gap size could be utilized to design a dynamic hydrogel brush for cell sorting applications, having the potential to tremendously increase the flexibility of the current nondynamic lab-on-a-chip sorting devices. [24] For all microstructures presented in this work, the actuation speed was limited by the heating and cooling rates of the microscope chambers, which was in the range of minutes, rather than the hydrogel material itself. We believe that the microstructures adapt to the environment much faster due to their high surface-to-volume ratios compared to larger bulk samples. Other studies found that such structures can even be actuated within the millisecond range, via locally heating with a laser beam. [16] These fast actuation capabilities are essential factors for the application range of the responsive material.
In addition to these two examples of microactuators, various other designs could be realized with high precision and high flexibility of adjusting material properties by the DLW process presented. Therefore, this approach of engineering responsive soft microactuators with improved actuation capabilities offers great potential for other fields of manufacturing soft microrobotic devices, or even humidity sensing. [25][26][27]

Conclusion
This work paves the way toward the precise fabrication of micrometer-sized actuator systems by a robust method based on DLW. Our results emphasize that pNIPAM is a promising material for printing micrometer-sized actuating objects. The dependency of actuation efficiency on the size and design of the printed microobjects suggests the surface-to-volume ratio of the object is a critical parameter for efficient microactuation function. To this end we propose that DLW is a great technology to prepare pNIPAM based microactuators not only as single microactuators, but even microactuator systems. The ability to assemble microactuators with micrometer resolution in almost arbitrary 3D shapes and to position them at a defined place on a surface may find a wide range of applications in future microrobotic systems that are based on thermoresponsive materials.
pNIPAM-Resist Formulation: The resist formulation was adapted from Hippler et al. [16] For the preparation of the pNIPAM-resist 400 mg N-isopropylacrylamide, 40 mg N,N′-methylenebis(acrylamide), and 10 mg lithium phenyl(2,4,6-trimethylbenzoyl)phosphinate were dissolved in 450 µL of ethylene glycol at room temperature and under constant magnetic stirring. To label the material for fluorescence imaging, 4 mg of acryloxyethyl thiocarbamoyl Rhodamine B was added. The solution was kept in dark to protect the photosensitive compounds from light.
Fabrication of pNIPAM Hydrogel Microstructures: 3D microstructures of pNIPAM hydrogel were fabricated with a commercially available direct laser writing setup (Photonic Professional GT2, Nanoscribe GmbH & Co. KG) using oil immersion configuration with a 25×, NA = 0.8 immersion objective. All microstructures were fabricated with 100% laser power (power scale = 1), slicing distance of 0.3 µm, and hatching distance of 0.2 µm. The laser scanning speed was varied between 10 and 50 mm s −1 to yield microstructures with different degrees of crosslinking. After writing, the microstructures were developed in 40 mL deionized water for 10 min and subsequently transferred into a fresh beaker of 40 mL deionized water for 1 min. The samples were kept in deionized water at 22 °C for storage. To facilitate adhesion of the microstructures to the glass substrates (borosilicate glass, 130-160 µm thickness), coverslips were functionalized with 3-(trimethoxysilyl)propyl methacrylate: First, the coverslips were cleaned in MilliQ water under sonication for 5 min and subsequently sonicated for another 5 min in ethanol. The substrates were dried with a heat gun and incubated in 3-(trimethoxysilyl)propyl methacrylate (15 × 10 −3 m in ethanol). After rinsing with MilliQ water and ethanol the substrates were dried with a heat gun and kept in an oven at 80 °C for 2 h.
Computer models of the sample geometries were designed with Autodesk Inventor Professional 2019 and imported as stl-files into the slicing software Describe, which defines the writing parameters.
Confocal Imaging: A laser scanning fluorescence microscope (Nikon A1R, Nikon Imaging Center Heidelberg) was used to study the sample response to temperature. The acquisition of the images was performed with a Nikon Plan Fluor 40× NA 1.3 objective. The glass coverslip with the sample was analyzed at room temperature and, after acclimatizing the sample for 15 min, in a temperature-controlled chamber (Tokai Hit chamber). The chamber temperature was regulated by an external controller (STXG Tokai controller) that was set up as following: top heater 65 °C, bath heater 50 °C, stage heater 50 °C, sample temperature 40 °C. The sample was always kept hydrated using ultrapure water filtered 0.22 µm.
Confocal fluorescence images were analyzed with ImageJ. First, an intensity threshold was defined and applied to all samples to create binary images. These binary images were then processed using binary operations Erode, Dilate, Open and Close. For the 2D images a cross-section of the micropillars was chosen and using the measure particle function the area and the perimeter were analyzed. For the 3D images the ImageJ Plugin 3D objects counter was used to evaluate surface area and volume. [28] Micromechanical Characterization: To quantify the mechanical properties of the printed materials, nanoindentations at the microstructure surfaces were performed with a commercially available fiber-optics based nanoindenter (Pavone, Optics11, Netherlands). For this purpose, pNIPAM microcubes with a side length of 50 µm were fabricated in square arrays of 3 × 3 cubes (cf. Figure 3b). Five arrays were fabricated with various laser scanning speeds ranging from 10 to 50 mm s −1 in steps of 10 mm s −1 . A precalibrated indenter probe with a cantilever spring constant of k = 0.23 N m −1 and a spherical glass tip of 3 µm radius was used for all experiments. For all arrays three cubes were tested in a matrix scan manner, which means that for each cube a square matrix of 3 × 3 indentation points with a lateral distance of 2 µm from each other was chosen in the center of each cube (cf. Figure 3a). For each location an indentation profile of 500 nm indentation at a speed of 1 µm s −1 followed by a hold time of 0.5 s and subsequent retraction at 1 µm s −1 was applied. The samples were immersed in deionized water throughout all experiments and equilibrated for at least 24 h before the measurements. Load-indentation curves were recorded both at 22 and 45 °C by using the internal temperature-controlled chamber of the device.
Data analysis was done using the Data Viewer (V2.5.0) software supplied by the device manufacturer. To determine the Young's modulus from each load-indentation curve, Hertzian contact model was applied according to a publication of Huth et al., using a constant indentation speed. [29] The contact point of each load-indentation curve was found by using the software integrated contact fit up to 20% of the maximum load. The Hertz fits were applied in the range between contact point (0 nm) and 300 nm for 22 °C or 0 and 150 nm for 45 °C, respectively.

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