Remotely Controlled Colloidal Assembly of Soft Microrobotic Artificial Muscle

Herein, a methodology for the directed self‐assembly of untethered microactuators and soft robotic microdevices from nanoscale building blocks is presented. The building block is a multifunctional stimuli‐responsive nanoactuator that consists of a magnetized gold nanorod encapsulated by a thermoresponsive hydrogel. The metallic core serves as a photonic nanoheater that transduces thermal energy from near‐infrared (NIR) light and a magnetic nanomotor that generates motion while driven by magnetic fields. Rapid control of temperature enables collective manipulation of nanoactuators through thermocapillary flows. In addition, catalytic activity of the nanorod instantiates a chemical reaction that covalently binds amine groups displayed on the surface of the surrounding soft gel capsule. A combination of optical and magnetic excitation realizes both reversible and permanent in situ assembly of microactuators within seconds that can perform both spatiotemporally controlled muscle‐like contraction (up to 30% strain) and motion. It is demonstrated that by linking nanoactuators with rationally designed compliant microstructures, more complex devices such as micromanipulators can be both fabricated and operated remotely. Colloidal assembly of microactuators ensures homogenous distribution of materials and functionality, thus preserving high performance provided by nanotechnology at multiple scales.

DOI: 10.1002/aisy.202000062 Herein, a methodology for the directed self-assembly of untethered microactuators and soft robotic microdevices from nanoscale building blocks is presented. The building block is a multifunctional stimuli-responsive nanoactuator that consists of a magnetized gold nanorod encapsulated by a thermoresponsive hydrogel. The metallic core serves as a photonic nanoheater that transduces thermal energy from near-infrared (NIR) light and a magnetic nanomotor that generates motion while driven by magnetic fields. Rapid control of temperature enables collective manipulation of nanoactuators through thermocapillary flows. In addition, catalytic activity of the nanorod instantiates a chemical reaction that covalently binds amine groups displayed on the surface of the surrounding soft gel capsule. A combination of optical and magnetic excitation realizes both reversible and permanent in situ assembly of microactuators within seconds that can perform both spatiotemporally controlled muscle-like contraction (up to 30% strain) and motion. It is demonstrated that by linking nanoactuators with rationally designed compliant microstructures, more complex devices such as micromanipulators can be both fabricated and operated remotely. Colloidal assembly of microactuators ensures homogenous distribution of materials and functionality, thus preserving high performance provided by nanotechnology at multiple scales.
that effectively transduced heat into linear actuation (Figure 1a). The size of the nanorod (%30 nm Â 95 nm) was specifically chosen to observe the localized surface plasmon resonance (LSPR) peak at 785 nm because near-infrared (NIR) light has high penetration depth under physiological conditions and allows the utilization of dyes in the visible range for fluorescence imaging. Ensembles of nanorods were coated with pNIPMAM polymer en masse using in situ free radical polymerization. [19,21] The surfaces of the nanogels were decorated with functional amine groups to initiate covalent cross-linking during colloidal assembly. This way, we ensured strong coupling between adjacent mNAs, a property that was required to transmit forces effectively and stay connected during multiple deformation cycles.
We introduced magnetization through controlled growth of a nickel film on the encapsulated nanorods. [31] Prior to the growth of the Ni layer, the Au core was covered with a small amount of Pt, which served as a catalyst for the decomposition of a Ni-hydrazine complex. Elemental analysis using energydispersive X-ray (EDX) spectroscopy confirmed the existence of a Pt-Ni shell, whereas Au was the predominant element at the particle center ( Figure 1b). Line scanning microanalysis across a single nanoactuator showed that the thickness of the Ni layer was on the order of 10 nm ( Figure S1, Supporting Information). We checked whether the magnetic coating influenced the thermoresponsive behavior of the pNIPMAM nanogel using dynamic light scattering (DLS) measurements ( Figure 1c). The mNAs exhibited a drastic decrease in hydrodynamic size above the lower critical solution temperature (LCST) of 42 C, from 400 to 300 nm, and reached a steady-state value of 220 nm at 65 C. Despite the small volume of magnetic material, the mNAs readily responded to externally applied magnetic field gradients at room temperature by accumulating in the vicinity of a permanent magnet that was placed next to the glass vial ( Figure 1d). Furthermore, application of a homogenous rotating magnetic fields led to the formation of particle chains (Figure 1e and Movie S1, Supporting Information), whose size depend on the particle concentration, magnetic field strength, and frequency of rotation. The chains broke up and particles formed smaller clusters at higher frequencies (at B ¼ 40 mT, transition frequency was 1 Hz) due to hydrodynamic effects. [6] The overall directed colloidal assembly process is shown in Figure 1f. Rotating magnetic fields program mNAs into transient clusters that discharge as soon as we turn off the magnetic field. In contrast, photothermal assembly process produces permanent assemblies that sustain physical attachment after the removal of the laser illumination. Localized nanoscale heating leads to ultrafast hydrogel volume-phase transition that is on the order of nanoseconds. [32] In our experiments, activation of mNAs at High-angle annular dark field (HAADF) shows the overall morphology and different metal layers. Scale bar, 50 nm. c) Hydrodynamic diameter, D H , of the mNAs measured at different temperatures. D H is calculated from the measured value of the translation diffusion coefficient using the Stokes-Einstein equation. d) Magnetic concentration of mNAs in a suspension. A permanent magnet was kept next to the mNA suspension for 5 min (left). After the removal of the magnet, an agglomerate of mNAs with a shape that followed the magnetic field lines was visible (right). e) Reversible assembly of mNAs under uniform rotating magnetic fields at different frequencies. Scale bar, 100 nm. f ) Schematic representation of the programmable assembly process. Left: Nanoactuator suspension on a glass slide. While magnetic manipulation drives reversible assembly of mNAs into chains (top right), photothermal effects generated by NIR illumination form permanent assemblies of mNAs (bottom right).
www.advancedsciencenews.com www.advintellsyst.com their LSPR frequency generated a strong and steep thermal gradient inside the solution that led to the formation of thermocapillary convection. Particles started to move with the flow from cold to hot regions as prescribed by the location of the laser beam. Physical and chemical interactions among agglomerating mNAs resulted in irreversible attachment and the formation of permanent microscale assemblies. Figure 2a shows a sequence of time-lapse images taken from a representative movie (Movie S2, Supporting Information) recorded during the in situ colloidal assembly of the soft microrobotic artificial muscle (artificial μmuscle). Upon NIR exposure, mNAs that were inside the exposure area started to serve as nanoscale heaters and contributed to the generation of directed thermocapillary flows that brought more particles into the assembly site. The nucleation of the artificial μmuscle started as soon as mNAs came sufficiently close to each other, which could be detected from the phase change in the workspace. The delay between the activation of the laser and beginning of the nucleation process depended on the laser power density. We observed a delay of about 2.5 s at a power density of 5.5 μW μm À2 , whereas increasing the power density to 7 μW μm À2 reduced the delay to less than a second. The growth rate, measured as area versus time, was controlled by the laser power density, reaching as high as 93 μm 2 s À1 at 7 μW μm À2 (Figure 2b). Once the size of the assembly exceeds the illumination area, the part of the actuator that resides outside the illumination area stops generating heat. In addition, at microscale, heat transfer takes place quite rapidly. As a result, the temperature of the outer edge of the structure drops below LCST of the hydrogel, which leads to local swelling. The resultant nonlinear expansion along with asymmetry in the convective flow prescribed the printing of actuators in the form of filaments even though the laser beam was held stationary. The increase in the size of the assembly due to swelling must be considered during the design process. We generated streamlines by tracking 1 μm-diameter fluorescent polystyrene beads in consecutive measurements to visualize the fluid flow (Movie S3, Supporting Information), as shown in Figure 2c. The thermal gradient was powerful enough to transport beads that were up to 360 μm away from the laser spot at a power density of 5.6 μW μm À2 . Particle image velocimetry (PIV) showed that the flow velocity was maximum around the laser spot and decreased in the radial direction ( Figure 2d). The magnitude of instantaneous velocity was as high as 223 μm s À1 at a power density of 7 μW μm À2 and 70.7 μm s À1 at 5.6 μW μm À2 .
After the completion of the assembly phase, structures were resuspended in water to study the actuation performance. They readily responded to low-intensity laser illumination (<4 μW μm À2 ) and low-strength magnetic fields (<30 mT). Each nanogel undergoes a volume-phase transition with plasmon heating that collectively led to rapid and controllable contraction of artificial μmuscle, as shown in Figure 2e. The amplitude, duration, repetition, and loading rate of mechanical output of the actuator can be effectively controlled by NIR pulse train (Movie S4, Supporting Information). Deformation was monitored by measuring the actuation strain, which was defined as percentage change in radius normalized with respect to the initial radius ( Figure 2f ). To demonstrate frequency modulation, we applied optical signals with a fixed pulse width of 100 ms at varying frequencies and recorded the contraction-relaxation cycles of the microactuator (Figure 2g). At low frequencies, artificial μmuscle could reach 65-70% of the maximum actuation strain and had time to relax back to the fully swollen state. With increasing frequency, structures went into a tetanus-like state, where the gel did not have time to swell and stayed contracted. Likewise, adjusting the input laser power for a given pulse width and signal period provided control over actuator deformation through amplitude modulation (Figure 2h). The optical energy density absorbed by the artificial μmuscle was estimated to be 3.06 Â 10 11 kJ m À3 (Supporting Information).
Spatiotemporally controlled laser illumination can assemble mNAs into artificial μmuscle with arbitrary shapes. We instructed the mNAs to form a fiber-shaped artificial μmuscle to demonstrate multiple degrees of freedom (DOF) actuation. The fiber behaved as a bending actuator, and its curvature was actively and reversibly controlled by tuning the location and power of NIR illumination (Figure 2i). The actuator was continuously deformable without showing any signs of detachment or plastic deformation, presenting unlimited DOF for robotic manipulation tasks. Laser writing of actuators can be done sequentially by moving the laser beam as long as there are free mNAs in the workspace. As an experimental demonstration, we fabricated an actuator in the form of a square frame in four writing steps, as shown in Figure 2j and Movie S5, Supporting Information. Controlling the thickness of the frame was challenging specifically at the joints because elongated or repeated laser exposure causes overpolymerization. As a result, printing parameters must be optimized for a given shape. We probed the magnetic response of the assemblies through application of magnetic torque. An artificial μmuscle with asymmetric shape was printed to be able to demonstrate magnetic control of orientation and translation. The μmuscle oriented along the direction of externally applied homogenous magnetic field with field strength as low as 10 mT (Movie S6, Supporting Information), and followed the rotating field up to 10 Hz at 40 mT before reaching step-out frequency. In addition to orientation control, we demonstrated remote positioning of μmuscle using magnetic field gradients (Movie S7, Supporting information).
We performed a series of experiments to gain insight on the mechanism of the assembly process.
Surprisingly, nonmagnetic nanoactuators with only Au nanorod core did not attach to each other, although they generated strong thermocapillary convection. We hypothesized that the disappearance of permanent attachment may be explained by the following effects: 1) extra metallic layers somehow provide the activation energy for a chemical reaction or physical polymer entanglement; and 2) magnetism plays an important role in the binding process. The absorbance spectra of particles with varying metallic coatings provided evidence for the first hypothesis ( Figure 3a). With the deposition of Pt and Ni layers, the signal flattened and the LSPR peak shifted toward the red end of the spectrum by 23 and 72 nm, respectively. [33] The broadening and damping effects are expected to facilitate the generation of more heat at a given laser power density. [34] Not surprisingly, PIV reported lower flow velocity around the illuminated area in the absence of Pt-Ni coating. To probe the potential role of a chemical reaction, we checked whether the amine groups presented on the surfaces of the mNAs are required for the assembly. Notably, mNAs without amine groups did not attach www.advancedsciencenews.com www.advintellsyst.com  to each other (Movie S8, Supporting Information), while exchanging the amine group with alkyne group sustained connectivity but with reduced cohesion (Movie S9, Supporting Information). We next explored whether magnetic interactions among mNAs led to attractive forces that held the particles together. Nonmagnetic nanoactuators with Au-Pt core (without Ni coating) also clicked to each other upon laser illumination. Furthermore, cryo-SEM images of magnetic assemblies showed no evidence of alignment among the nanorods (Figure 3b). We concluded that magnetism did not play a major role in the laser-induced assembly process. Finally, we checked whether Pt layer was indispensable for the catalysis of a specific reaction. We deposited a silver layer on Au nanorod by reducing Ag salt, [35] which was verified using microelemental EDX analysis (Figure 3c). Nanoactuators with Au-Ag core also formed permanent assemblies under NIR exposure. The UV-vis spectroscopy showed that Ag and Pt coating resulted in similar absorbance spectra; thus, the heat losses are expected to be comparable (Figure 3a). As a final test, we resuspended assemblies into acetone and applied gentle pipetting to check whether the particles stayed together due to an unspecific adhesion process. The fact that they survived this treatment reinforced the theory that mNAs attached to each other due to the covalent cross-linking of amine groups, and this reaction required reaching a certain temperature and/or catalytic activity that was achieved by the additional metal coating.
The isotropic contraction of the artificial μmuscle may be converted to linear or rotary motion with desired mechanical advantage using compliant mechanisms. We chose poly(ethylene glycol) diacrylate (PEGDA) as the hydrogel for manufacturing the mechanisms due to its tunable stiffness, ease of polymerization, and temperature-independent swelling properties. Upon laser exposure, the mNAs accumulated between the arms of a lever mechanism that was designed to amplify the deformation (Figure 4a). Figure 4b shows time-lapse images from an experimental assembly process (Movie S10, Supporting Information). The particles not only attached to each other but also firmly adhered to the surface of the PEGDA structure. The contraction of the in situ assembled artificial μmuscle moved the arms toward each other (flexion), whereas opening (extension) was achieved with the removal of the actuator force due to the stored energy ( Figure 4c and Movie S10, Supporting Information). Contraction of the actuator that gives 20% closing of the lever was completed in 750 ms, whereas relaxation took seconds due to the diffusion process (Figure 4d). The contraction-relaxation curves were identical for several actuation cycles (Figure 4e). Long-term performance of the micromachines was tested with the application of 1000 actuation cycles, each consisted of laser pulses with 0.5 Hz frequency, 50% duty cycle, and 7 μW μm À2 power density. The angular displacement was recorded as 9.8 AE 0.7 for the first 5 cycles and 9.7 AE 1.3 for the last 5 cycles. This test verified the robustness of the actuator. The strong adhesion of nanoactuators to the PEGDA structures also provides a versatile route to form layered actuators. By depositing an actuation layer at the base of an elastic beam, we engineered an active hinge that could rotate the beam in both clockwise and counterclockwise directions in an antagonistic fashion (Figure 4f ). Figure 4g shows snapshots from an experimental assembly process. By simply changing the location ( Figure 4h) and intensity (Figure 4i) of laser exposure, we precisely controlled the angular displacement of the beam. Finally, we asked whether the actuators would heat or show signs of fatigue during elongated exposure. The rotating beam device was excited continuously at various power densities for 10 min during which deformation angle remained the same (Figure 4j).
We harnessed laser power to accomplish three distinct goals: optofluidic transport and manipulation of nanoactuators by thermocapillary convection, catalyzing a local reaction for phase transition and permanent assembly of agglomerating nanoactuators into artificial μmuscle, and localized heating for fast hydrogel volume-phase transition and actuation of the self-assembled artificial μmuscle. Methods that utilize hydrodynamic forces to manipulate particles offer several advantages such as favorable force scaling, low power requirement, and large workspace. Other optothermal effects such as thermophoresis have been reported in the literature in the context of particle assembly. [23,36] Thermophoresis is not expected to play an important role in our manipulation protocol because, when we repeated the experiments in closed chambers where evaporation cannot take place, www.advancedsciencenews.com www.advintellsyst.com the particles showed very limited motion. Furthermore, both mNAs and fluorescent polystyrene beads moved from cold to hot regions in our experiments and we do not have charged surfactants in the solution. The colloidal assembly approach may be pushed to the next level if, instead of photopolymerization, we could self-assemble mechanisms from nanoscale building blocks. This concept may be realized by synthesizing thermally nonresponsive nanogels with gold nanorod core.
Tuning the plasmon resonance frequency of the active and passive components using nanorods with different aspect ratios will enable coordinated assembly of mechanisms and actuators. Finally, the assembly process may become reversible by the initiation of covalent bonds that can be enzymatically removed or using DNA linkers. [37] Figure 4. Controlled self-assembly and characterization of soft micromachines actuated by artificial μmuscle. a) Schematic representation of in situ assembly of artificial μmuscle between the arms of a lever mechanism. b) Snapshots from a time-lapse movie (Movie S10, Supporting Information) showing the growing actuator at the designated location. The compliant lever mechanism was fabricated from a hydrogel that does not respond to changes in temperature. c) Optical actuation with 7 μW μm À2 laser power density (Movie S10, Supporting Information). D 0 denotes the reference distance between the arms at the fully swollen state of the actuator. d) Actuation strain over time graph for a single actuation cycle at a laser power density of 7 μW μm À2 . e) Actuation strain over time under NIR illumination with 3 s pulse width and 0.1 Hz frequency at 7 μW μm À2 laser power density. f ) Schematic representation of the self-assembly of antagonistic actuators with bidirectional rotation. g) Time-lapse images showing the self-assembly of the actuators on both sides of the beam. h) Optical actuation upon NIR illumination with 7 μW μm À2 laser power density. Images show the relaxed and fully contracted states of the machine. The angular displacement, θ, is measured with respect to the reference state. i) The change in θ as a function of laser power density. The trend is linear for the given power density values. j) Angular displacement of the beam over time at different laser power densities. The angular displacement stays the same for minutes under continuous NIR illumination. Scale bars, 50 μm.
www.advancedsciencenews.com www.advintellsyst.com Although the complexity of the shapes that can be achieved by laser writing is unprecedented, the technique may not be suitable for in vivo applications where penetration of NIR is an issue and Marangoni flows may not be generated effectively. In contrast, magnetic fields are suitable for particle manipulation and powering actuators under these conditions. The magnetic properties of the mNAs pave the way for the formation of dynamically stable morphologies using rotating and oscillating magnetic fields. The heat that is required for chemically crosslinking the particles and driving the sol-gel transition can be generated using radiofrequency (RF) oscillating magnetic fields. RF signals can also be used for powering the self-assembled artificial μmuscle. Previous work has shown that bulk thermoresponsive hydrogels fabricated from magnetic nanocomposites could be actuated using magnetic hyperthermia. [38,39] However, this strategy will come with an important trade-off, the spatial resolution and frequency selective activation provided by plasmon resonance heating will be lost. Intuitively, the resolution of printing depends on the laser dose and the concentration of mNAs, and precise control over laser spot would enable printing of structures with arbitrary complexity. Future work will address the optimization of parameters for user-defined 3D printing of artificial μmuscle.
Synthesis of Nanoactuators: The synthesis of nonmagnetic nanoactuators was done following a protocol we reported elsewhere. [21,40,41] Briefly, the gold nanoparticles were synthesized using a two-step procedure. Seed nucleation was performed with CTAB as surfactant, and the growth solution had AgNO 3 as well as the surfactants CTAB and NaOL. After purification, the gold nanorods were coated with pNIPMAM using in situ free radical thermal polymerization. The monomers NIPMAM and BIS were added in a mass proportion of 10:1, and AAPH was used as the thermal initiator. For amine functionalization of the polymer chain, allylamine was added after 30 min of reaction time, and for alkyne functionalization, propargyl was added after 90 min. The final product was centrifuged at 3500 relative centrifugal force (rcf ) for 30 min at 10 C, 3 times and resuspended in Milli-Q water.
The protocol for magnetization was adapted from ref. [31]. Briefly, prior to Ni coating, a catalytic shell of Pt was grown on Au to promote Ni reduction in the presence of hydrazine. Throughout the process, molar ratios Au 0 /Pt II /Ni II were kept as unity. To a 10 mL of 0.1 M CTAB, 5.6 μmol of K 2 PtCl 4 (78.5 μL, 0.071 M) and pNIPMAM-coated gold nanorods were added. The suspension was sonicated for 30 min and subsequently 157 μL of 0.1 M ascorbic acid was added. The resulting suspension was mixed for 2 min and left overnight at 40 C. Next day, Pt-coated nanoparticles were cleaned 3 times by centrifugation for 10 min at 5000 rcf. If Pt-coated particles were the final product, the pellet was redispersed in ultrapure water (Milli-Q, Merck Millipore). To deposit an extra Ni layer, the particles were resuspended in 6 mL of 0.1 M CTAB solution. In separate, 8.7 mL of ultrapure water, 22.4 μL of 0.25 M of NiCl 2 , and 179 μL of 2.5 M hydrazine were added and mixed thoroughly. To this solution, Pt-coated nanoparticle suspension was added and incubated at 40 C for 150 min. The resulting product was purified by centrifugation at 3500 rcf for 10 min at room temperature, 3 times, and resuspended in ultrapure water.
The protocol for silver coating was adapted from previously published reports. [35,42] Briefly, hydrogel coated nanorods were resuspended in a 5 mL solution of 0.01 M CTAB and 0.1 mM AgNO 3 to obtain a final gold concentration of 9.6 Â 10 À4 M. The reaction solution was left undisturbed at room temperature for 1 min, and subsequently 100 μL of 0.1 M ascorbic acid and 50 μL of 10 M NaOH solution were added. The solution was incubated for 3 h at room temperature with vigorous stirring. Finally, the product was purified by repeated centrifugation for 10 min at 3500 rcf at 10 ºC.
Characterization of Nanoactuators: Several techniques were used to characterize the morphology and structure as well as thermal and optical properties of the materials. Transmission electron microscopy (TEM) and scanning TEM with EDX (STEM-EDX) were used to directly measure the overall size ( Figure S2, Supporting Information) and elemental mapping of the nanoparticles, respectively. The samples were prepared by depositing a 15 μL droplet of the nanoparticle suspension onto a carbon grid. To be able to visualize the pNIPMAM coating, a negative staining with 2% uranyl acetate was performed prior to imaging through exchanging the solvents. The samples were dried at room temperature. TEM microscope (Tecnai Osiris, FEI) was operated with a voltage of 200 kV. UV-vis measurements were performed with V-670 UV-vis-NIR spectrophotometer (Jasco). DLS measurements were performed using a Malvern Zetasizer Nano. Twelve measurements were taken for each temperature point from three different samples. The internal structure of the microscale assemblies was visualized using cryo-SEM (Zeiss Gemini 500) with an acceleration voltage of 5 kV, working distance of 9.6 nm, magnification of 14,260X, field of view (FOV) of 8.016 μm, and an imaging mode with variable pressure and a detection of electron backscatter diffraction (VP BSD). The assemblies were encapsulated inside 10 mM gelatin and cut into small cubes in the micrometer range. The gelatin cubes containing the samples were then transferred into 2 M sucrose solution and kept chilled at 4 ºC overnight. The following day, the cubes were placed on SEM holders and dipped into liquid nitrogen for rapid freezing. The frozen samples were prepared for imaging using Leica EM ACE600, where they were sublimated at À90 ºC for 10 min to remove impurities and coated with a 5 nm layer of carbon at À150 ºC using sputtering. Focused ion beam (FIB) was used to remove remaining impurities. Leica EM vacuum cryo manipulation was used for holder mounting.
Fabrication of Mechanisms: The compliant mechanisms were fabricated using a maskless projection photolithography method. For this purpose, a programmable digital micromirror device (DMD) module was coupled to the motorized inverted microscope, Nikon Ti-Eclipse, used as our work station. The DMD module projects light from an ultraviolet LED source (365 nm) through the microscope objectives according to computer-aided design drawings. The prepolymer solution is composed of PEGDA (700 kDa) and 20% (v/v) of the UV photoinitiator DAROCUR.
Experimental Setup: All experiments were performed on a motorized inverted microscope (Nikon Ti-Eclipse) coupled with a 785 nm laser source (120 mW, Thorlabs) that provided NIR illumination. The laser illumination was controlled with a custom LabView program. For magnetic manipulation, a circular Halbach array, attached to a rotational stage and optimized for uniform magnetic field, 40 mT, was mounted on the inverted microscope. The rotation stage was controlled by a program written in MATLAB. Images and videos were captured using an ORCA-Flash4.0 CMOS camera (Hamamatsu). Videos were recorded with frame rates ranging from 33 to 100 fps depending on the experimental conditions. Image Processing: Fiji from ImageJ [43] was used to analyze images and some videos. To study the deformation properties of the assemblies, videos with 100 fps were obtained and processed with a program based on edge-detection algorithm in MATLAB. [44] To study the flow rate generated in our experiments, 1 μm-diameter fluorescent particles were added to the colloidal suspension and fluorescent imaging was performed while the laser was on. Videos were recorded with high frame rate and processed in MATLAB with the PIV tool, [45] and plotted with the plugin quiverc. [46] Measurement of Beam Stiffness: A commercial microelectromechanical systems force sensor (FT-S1000, FemtoTools) with a resolution of 0.05 μN was mounted on a micromanipulator with a custom-machined mount.
The force sensor was powered by a programmable linear power supply (Keysight E3631A), and output was measured using a precision multimeter (Keysight 34465A). Force measurements were obtained using the factory calibration.

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