Femtosecond laser writing of ant-inspired reconfigurable microbot collectives

Design, fabrication and characterization of ant microbot

The schematic of the design and fabrication process of the ant microbot with multi-materials is illustrated in Fig. 2 (Supplementary Fig. 1), and the whole procedure consists of three steps. In the first step, the 3D magnetic photoresist body of an ant is fabricated by TPP process based on a predesigned model (Fig. 2a). Subsequently, after the sample is developed with ethanol and washed with DI water, the thermal stimuli-responsive hydrogel is dropped onto the microstructure. In the secondary TPP process, the two asymmetrical bilayer hydrogel joints are fabricated to connect the mandibles to the body (Fig. 2b). Specifically, the hydrogel joint is composed of a low crosslink density portion and a high crosslink density portion, where the low crosslink density part has a larger shrinkage ratio. SEM images provide a direct comparison of the ant microbot before and after joints integration, and detailed SEM images show the distribution of the low and high crosslink density parts, where their contours are marked in light and dark red, respectively (Fig. 2b). Finally, after the removal of the unpolymerized hydrogel, the silver precursor is added to the microbot, and the Ag NPs coating tightly attached to the surface of the ant microbot head is reduced by two-photon reduction (Fig. 2c and Supplementary Fig. 2), in which the Ag NPs with average roughness of 42.3 nm (Supplementary Fig. 3) are reduced from silver ammonium ions absorbing photons. The differences between before and after reduction of Ag NPs coating are also presented by SEM images, and the contours of Ag NPs are marked in yellow (Fig. 2c). Energy dispersive spectrometer (EDS) images demonstrate the presence of the Fe and Ag elements. In this way, an assembling micro unit incorporating various materials with distinct properties and metal coating can be manufactured (Supplementary Fig. 4).

The ant microbot consists of a magnetic photoresist body, an Ag NPs coating and two hydrogel joints. Among them, the hydrogel is mainly composed of N-Isopropyl acylamide (NIPAM) monomer^40,41, Methylene-Bis-Acrylamide (MBA) crosslinker, and Diphenyl (2,4,6-tri-methyl benzoyl) phosphine oxide (TPO) photoinitiator. The deformation mechanism of thermo-responsive hydrogel is thought to be a reversible phase transition from a contracted hydrophobic state at temperatures above its lower critical solution temperature (LCST) to a hydrophilic swollen state at temperatures below its LCST, which is almost between 32 °C and 33 °C^42,43. Since the mechanical strength of printed NIPAM structures is relatively poor⁴⁴, the polyvinylpyrrolidone (PVP, K30) is dissolved in the pre-gel solution to improve the mechanical strength of the hydrogel microstructures (Supplementary Fig. 5), and it has a near-negligible effect on the photothermal sensitivity of the hydrogel²¹. In this work, to quantitatively investigate the expansion and contraction properties of this hydrogel, micro circular plates with a diameter of 20 μm and a thickness of 3 μm are fabricated and measured and full contraction in 8 ms is observed when the microplate is illuminated. We define the shrinking ratio (ε) of the hydrogel structures as the ratio of the length change to the original length, and the processing parameters that affect the shrinkage have been reported in detail in our previous work²¹. In this study, taking advantage of the programmable fabrication capability^45,46, we keep the processing laser power and scanning point step constant at 33 mW and 320 nm, respectively, while choosing scanning repeat time (SRT) as the processing variable to control the shrinking ratio of the hydrogel, and the maximum shrinking ratios for SRT = 1 ms and SRT = 3 ms are ∼0.39 and ∼0.1, respectively (Supplementary Fig. 6). For hydrogel joints, we systematically investigate the effect of the difference and the width ratio of different crosslink density layers on their bending angle (Supplementary Figs. 7–8). Based on these experimental results, we design a micro asymmetrical double-layered joint (8 μm in length and 4 μm in width) to achieve bending deformation, where the width ratio of the SRT = 1 ms region to the total width is 0.7 and the remaining part of SRT = 3 ms. In the composite ant microbot, the Ag NPs function as the photoreceptor and photothermal agent, which absorb energy from the laser irradiation and then convert it to thermal energy to heat thermo-responsive hydrogel joints. The hydrogel joint consists of two layers with different SRT, which result in distinct degrees of crosslinking, and consequently, the hydrogel joints can exhibit asymmetric deformation. Specifically, as soon as the temperature of the joint is higher than the LCST of the hydrogel via surface incorporated photo absorbers, it immediately bends toward the side of the fewer SRT due to the larger shrinking ratio of the fewer SRT than the more SRT. In this way, the mandibles initially close together, and once the light field perpendicular to the surface of the head is applied, the mandibles will be actuated by the joints to open. When the external illumination is removed, the mandibles return to their original closed condition (Fig. 2d). To obtain insight into the degree of bending deformation caused by different SRTs, theoretical simulation is conducted to verify the bending shape of the joints, and the simulated bending direction and angle match well with the experimental results, indicating the feasibility of this design strategy (Fig. 2e). The dynamic deformation of the joints under 20 mW laser indicates a rapid response of the joints (Fig. 2f and Supplementary Movie 1). It takes only 8 ms to reach a maximum opening angle of ∼86°, whereas the recovery to the initial state takes 12 ms after the switch off of the laser. This is understandable because the heating process of hydrogel joints that benefit from high energy density is faster than the cooling process (Supplementary Fig. 9). In addition, the opening/closing behavior of the mandibles is stable and reversible. Figure 2g shows the opening angle of the ant microbot during 1000 times of switching on/off the illumination, and no significant fatigue is observed after repeated stimulation, indicating the stable responsiveness (Supplementary Fig. 10 and Supplementary Movie 2).

The component of the ant microbot body is magnetic photoresist, which is intended to be controlled remotely by introducing an external magnetic field. The magnetic photoresist is prepared by mixing Fe₃O₄ magnetic nanoparticles surface-modified with oleic acid with commercial photoresist SZ2080⁴⁷. The vibrating sample magnetometry (VSM) test shows that the remanence and coercivity are all almost zero, indicating the ant microbot is superparamagnetic, which is desirable for movement manipulation (Fig. 2h). A permanent ferromagnet is used for remote control (Supplementary Fig. 11), which applies a magnetic torque and force on the magnetized ant microbot. The magnetic force associated with the magnetic field gradient leads to the translation of the ant microbot, and the magnetic torque induces rotational motion of the ant microbot to be aligned along the magnetic field (Fig. 2i). We test the ant microbot’s rotational motion capability by using a Helmholtz coil that can generate a uniform magnetic field (∼45 mT), and the direction of the magnetic field is changed by 45° each time. The results show that the response time, standard deviation and overshoot are 13.05 s, ∼0.71°, and 2.35°, respectively. (Fig. 2j and Supplementary Fig. 12).

Reversible, multiple and robust assembly of magnetic ant microbots

The on-demand assembly and separation of the ant microbots are achieved by coordinated control of the magnetic and light fields. As the schematics show in Fig. 3a, there are two ant microbots, and one of them is fixed to the substrate while the other is free to move. When a handheld permanent magnet is carefully approached, the ant microbot in the free state will be aligned along the direction of the magnetic field and will be able to translate under the magnetic force. By adjusting the relative distance and angle between the units as well as leveraging the light field to control the opening and closing of the mandibles, we can controllably accomplish two stable assembly states by gripping the tail with the mandibles, i.e., tending to 180° assembly when the angle between two units in long axis direction is 180° and tending to 90° assembly when the angle is 90°. Benefiting from the non-contact triggering and structural design, the deformation-based mechanical assembly between units can tolerate large angular and positional errors, and the assembled structure no longer requires a continuous supply of external stimuli to be maintained. On the other hand, benefiting from the hydrogel joints having a stable deformation capability, the assembled units can be easily separated under the combined control of magnetic and light fields. However, since the assembled units are sufficiently stable, they will remain in their original state if they are only subjected to the magnetic field, which is the reason that the new units can be further assembled normally.

To intuitively demonstrate the effectiveness of assembly and configurations of ant microbots, we employ two ant microbots to achieve 180° assembly and 90° assembly sequentially (Supplementary Movie 3-4). As the snapshots shown in Fig. 3b, one ant microbot is anchored to the substrate while the other is randomly distributed. Then, the magnetic field actuates the free unit to approach the fixed unit. When the latter mandibles contact with the former tail, a light field is introduced to drive the latter mandibles open. Afterward, the latter continues to move forward driven by the magnetic field so that the latter mandibles encircle the former tail. Finally, the two ants achieve 180° assembly after turning off the light field. Conversely, we can use magnetic and light fields to disperse the assembled structure into two individual units. Similarly, using this strategy the two ant microbots can also achieve 90° assembly and separate on-demand (Fig. 3b). Notably, the 90° and 180° assembly methods are combinable and extendable to complete multiple configurations. As the demonstrations illustrated in Fig. 3c, by changing the sequence and number of the two assembly modes, numerous configurations of ant microbot can be realized. Here, the 4 units are sequentially assembled into an initial C shape by a 90° assembly, a 180° assembly and a 90° assembly, then the assembly pattern between two neighboring units is adjusted on demand by the control of magnetic and optical fields for further morphological reconfiguration. Benefiting from the controlled and selective assembly, these independent units are reassembled into a J shape. Similarly, with repeated controllable separation and selective assembly, micro units can be reversibly reconfigured into the O, V, S, and W shape by combining two assembly modes. In addition to the ant microbot, our assembly strategy is equally applicable to other shaped microbots with multiple claws and gripping points. Thanks to the flexibility of our fabrication approach, we have also designed other shaped microbots and realized more complex assembly forms (Supplementary Fig. 13). For example, rectangular-shaped microbots with two claws and six gripping points can be constructed into bifurcated tree-like patterns (Supplementary Fig. 14), and disk-shaped microbots with two claws and two gripping points can be assembled into shapes dispersing from the center outward (Supplementary Fig. 15). To manipulate multiple ant microbots at the same time, we can introduce a spatial light modulator to generate multifocal light fields to programmatically control the opening and closing of the mandibles of many ant microbots as well as many claws on a single microbot (Supplementary Fig. 16). In addition, by controlling the concentration of magnetic nanoparticles in the magnetic photoresist, multiple ant microbots can achieve synchronized or asynchronized motions driven by the same magnetic field (Supplementary Figs. 17–18).

The controllable and selective assembly strategy can also enable the assembly between multiple non-magnetic material units with the assistance of a magnetic unit (Supplementary Movie 5-6). Continuous sequence images of this process are shown in Supplementary Fig. 19, a magnetic unit that clamps a non-magnetic unit is guided by a magnetic field close to the other non-magnetic unit. When the mandibles of the second non-magnetic unit contact to the tail of the first non-magnetic unit, a light field is turned on to actuate the mandibles of the second non-magnetic unit to open. After that, the magnetic unit continues to move forward with the non-magnetic unit so that the mandibles of the second non-magnetic unit can encircle the tail of the first non-magnetic unit. Finally, the three units achieve 180° assembly after turning off the light field. In addition, focusing the light spot on the head of the third magnetic unit induces its mandibles to open, which in combination with the control of the magnetic field allows it to separate from the whole, thus realizing the 180° assembly between two non-magnetic units. Similarly, using this method the two non-magnetic units can also achieve 90° assembly and separate on-demand (Supplementary Fig. 19). To illustrate the dynamic control capability of the magnetic unit to manipulate different non-magnetic material units in a single coding sequence, we pattern different material units into various configurations that closely resemble the tetrominoes in the popular game Tetris (Supplementary Fig. 19). Therefore, this controllable and selective assembly strategy provides more possibilities for the assembly of micro unit collectives.

To evaluate the reliability of assembly, we employ a horizontal vibration generator (VG) to investigate the bonding stability between units with 90° and 180° connection respectively. Note that the first unit in both assembly patterns is anchored to the substrate to vibrate with the VG, while the remaining two units are assembled by our proposed strategy. Figure 3d shows the stabilization of the assembled structure under different combinations of vibration frequency and amplitude (Supplementary Movie 7–8 present the stabilization of two assembled modes at a vibration frequency of 5 Hz with increasing amplitude). The green dots represent that the assembled microstructures remain intact and the red cross symbols indicate that the assembled microstructures are destroyed (Supplementary Fig. 20). It can be seen that the amplitude needed to disperse the assembled architecture gradually decreases as the applied frequency increases for both assembly methods and the stability of the C shape is larger than that of the I shape. Note that the assemble reliability is sufficient to guarantee that the ant microbot collectives can move and stably transport cargo as shown in the following sections.

Gap traversal with a single ant microbot and assembled ant microbots

In this work, we systematically investigate the locomotion performance of single and assembled microbots steered by a magnetic field in a gap environment, where the size of the gap is of the same order of magnitude as the body length of an ant microbot. In detail, the ant and the gap lengths are 97 µm and 120 µm, respectively, while the ant and the gap heights are both 20 µm (Fig. 4a). For a single robot performing forward locomotion and attempting to cross the gap, most failures occur after the middle pair of legs fall into the gap because the center of gravity (CG) of the microbot is located above the middle pair of legs. From this, we deduce that the maximum length of the gap that a single microbot can traverse is approximately equal to the distance D₁ from its CG to the front of the mandibles, which is 50 µm. Using the same principle, we derive that the maximum gap lengths that two and three 180° connected microbots can traverse are 93 µm and 145 µm, respectively (Fig. 4b). That is, a single ant microbot and two assembled microbots cannot pass through this gap, while three connected microbots are able to traverse successfully. The experimental results confirm our speculation, as shown in Fig. 4c–e (Supplementary Movie 9). When an individual microbot attempts to traverse the gap guided by the magnetic field, the robot falls completely into the gap because the length of the gap is larger than the body length of the robot (Fig. 4c). For the two connected microbots attempting to cross the gap, they fail because the distal edge of the gap is beyond the reachable space of the front legs thus causing the front legs to fall into the gap. The whole rotates on their latter unit legs and the former unit tilts forward, which results in a sloped body posture. Even so, they remain as one complete individual, indicating a high degree of assembly stability (Fig. 4d). By contrast, the three assembled microbots can pass through without difficulty. Three connected microbots will shift the CG far enough to the rear and provide a proper weight distribution to traverse the gap and the pitch angle of the body is always maintained close to zero during the crossing process. This demonstrates that the assembled microbots have a stronger ability to traverse the gap compared to an individual microbot (Fig. 4e).

Maze navigation and drug delivery by microbot collectives with different assembled morphologies

Ant colonies in nature contain many individuals, which can collaborate to accomplish various tasks. In contrast, it is difficult to use our assembly strategy for large numbers of individuals. Our proposed method can accomplish the reconfigurable assembly of several microbots, and it is suitable for manipulating particles, cells, and other micro-objects whose sizes are concentrated in the range of a few micrometers to tens of micrometers. Benefiting from the proposed strategy of reversible assembly, we achieve cargo transportation and maze navigation by switching between I shape and C shape consisting of three magnetic units (Supplementary Fig. 21 and Supplementary Movie 10). The width of the entrance and exit of the maze allow only an I-shaped three-unit microbot to pass through and the entrance and exit of the maze have a gripping point to facilitate the switching of the assembly pattern. As illustrated in Supplementary Fig. 21, first, the three-unit microbot is manipulated into the maze in an I shape by a magnetic field, after which a light field is combined to enable it to clamp the gripping point. Here, due to the global actuation of the magnetic field in a small area, the purpose of clamping is to selectively manipulate one unit without affecting neighboring units to adjust the assembly pattern of the three units. Next, again under the coordination control of the magnetic and optical fields, the three assembled units are transformed from I shape to L shape and then to C shape to be able to carry a cargo. Then, with the cargo in their grasp, the C-shaped three-unit microbot is transported magnetically to the exit of the maze. Afterwards, the C-shaped three-unit microbot is reconfigured to an L shape and then to the initial I shape under the coordinated control of the magnetic and optical fields to release the cargo as well as to be able to exit the maze. Finally, the three-unit microbot assembled in the I shape is successfully disengaged from the gripping point and maneuvers out of the maze under the manipulation of the magnetic and light fields. These results indicate that the environmental adaptability and functionality of microbots could be significantly improved by morphological reconfiguration, which is expected to facilitate the development of biomedical engineering.

Next, we use the drug carrier as delivery object to further investigate the transportation capabilities of the ant microbot collectives in a maze environment. Thanks to the proposed assembly strategy confers environmental adaptability and portability to the connected units, our microbots can perform drug delivery to Hela cells in the targeted region. As depicted in Fig. 5a, a photoresist square maze (600 µm in side length) with a square micro-fence (150 µm in side length) at each end of one edge is prepared by TPP process. The nutrient medium containing Hela cells is dropped onto the maze and cells outside the micro-fences in the maze are removed as much as possible. One of the micro-fences containing Hela cells is treated as the experimental group, i.e., the area where the drugs will be delivered to and function, while the other micro-fence serves as the reference group. In this process, the shape reconfiguration of microbot collectives and the transportation of the drug carrier are the same as for the cargo described in previous section. we manipulate the three-unit microbot to transport the cubic drug carrier (20 µm in length) from the entrance of the maze to the experimental group by coordinating the magnetic and light fields, where the drug carrier is prepared by immersing a cubic hydrogel block processed by TPP in doxorubicin hydrochloride (DOX) solution (5 mg mL⁻¹) for 1 h to load drugs onto the carrier. Figure 5b shows time-lapse images of drug carrier delivery, and the images from each location have been stitched together to demonstrate the full path. After the carrier is delivered to the targeted position, the extent and intensity of the red fluorescence gradually increases over time and needs ∼20 min to fill the entire area of the experimental group, indicating that the drugs had diffused rapidly from the carrier (Supplementary Fig. 22). During the 12 h of culture in the incubator, the green fluorescence of the cells in the experimental group gradually fades, while the cells in the control group maintain a strong green fluorescence all the time (Fig. 5c). Quantitative analysis reveals 98% attenuation of cells fluorescence intensity in the micro-fence containing the doxorubicin hydrogel block, whereas only 18% fluorescence attenuation is observed in the control group (Fig. 5d), demonstrating a significant inactivating effect of the drugs on the Hela cells.

Materials

The Iron (II) chloride tetrahydrate (FeCl₂·4H₂O), Iron (III) chloride hexahydrate (FeCl₃·6H₂O), Ammonia water (NH₃·H₂O, 25 wt.%), Oleic acid and n-Octane were purchased from Sinopharm Chemical Reagent Co., Ltd. N-isopropylacrylamide, N, N’-methylenebis (acrylamide), Diphenyl (2,4,6-trimethylbenzoyl) phosphine oxide, ethylene glycol, Polyvinylpyrrolidone K30, Rhodamine 6 G, trisodium citrate, aqueous ammonia, and silver nitrate were purchased from Aladin.

Preparation of magnetic photoresist

FeCl₃·6H₂O (2.69 g, 0.1 mol) and FeCl₂·4H₂O (0.99 g, 0.05 mol) were dissolved in DI water (100 mL) and then stirred for 30 min. Next, NH₃·H₂O (25 wt.% 4 mL) was added dropwise to the mixed solution under vigorous stirring, resulting in the formation of Fe₃O₄ nanocrystals in weak alkaline solution. After stirring for 3 h, the resulting precipitate was separated from the mixture using a strong magnet and washed three times with DI water. The surfactant Oleic acid (0.61 g, 2.15 mmol) was added dropwise to the Fe₃O₄/water suspension at 80 °C under nitrogen protection with continuous stirring for 30 min. The prepared surface-modified Fe₃O₄ nanocrystals were then uniformly dispersed in n-Octane (3.2 mL, 20 mmol). Finally, after commercially available photoresist SZ2080 was added, the mixture was ultrasonic stirred for 1 h at room temperature to obtain magnetic photoresist.

Preparation of silver precursor

Firstly, 14.1 mg of silver nitrate aqueous and 13.4 mg of trisodium citrate were added to 1 mL of DI water and stirred to mix the mixture well. Subsequently, NH₃·H₂O (25 wt.%) was added dropwise to the mixture under stirring until a clarified solution was formed.

Preparation of pNIPAM precursor

The precursor was prepared by dissolving 400 mg of N-isopropylacrylamide, 30 mg of N, N’-methylenebis (acrylamide), 30 mg of Diphenyl (2,4,6-trimethylbenzoyl) phosphine oxide, and 50 mg of Polyvinylpyrrolidone K30 into 450 μL of ethylene glycol, followed by the addition of 2 mg of Rhodamine 6 G for fluorescence imaging.

Fabrication of 3D microstructures

A mode-locked Ti: sapphire ultrafast oscillator (Chameleon Vision-S, Coherent) with an oil immersion objective (60 ×, NA = 1.35) was used for both photopolymerization and photoreduction. The magnetic photoresist was first added dropwise to a coverslip and heated at 100 °C for 1 h. After that, the coverslip was placed on a high-precision platform, where the scanning oscillator in the XY plane and the Z-directional nano-position stage can be controlled to achieve point-by-point scanning processing in the magnetic photoresist. The processing files were obtained by converting the 3D models (SolidWorks) imported into the home-built software, which was able to control coordinately the position of the XY plane scanning oscillator and Z-directional piezoelectric stage. Subsequently, the sample was immersed in the developing solution (1-propanol) for 15 min to remove the uncured photoresist. In the second step, the hydrogel precursor was drop-added onto the same coverslip where there were previously processed microstructures and the bilayer hydrogel joints was fabricated to connected the ant mandibles and body. The complete ant microbot was obtained after 15 min of development with ethanol. In the final step, the silver precursor was dropped on the microstructures and the Ag layer can be deposited on any desired surface of the micro-ant by the photoreduction process.

Cell culture

HeLa-EGFP cells were obtained from American Type Culture Collection and cultured in a humidified atmosphere at 37 °C with 5% CO₂. The cells were cultured in normal Dulbecco’s modified eagle medium (Gibco, Thermo Fisher Scientific, Grand Island, NY) supplemented with 10% fetal bovine serum (HyClone, Logan, UT) and 1% penicillin/streptomycin (Gibco, Life Technologies, Grand Island, NY). For cells therapeutic experiment, Hela cells were detached from the culture dish by trypsinization with 0.25% Trypsin-EDTA (Gibco, Thermo Fisher Scientific, Grand Island, NY) at 37 °C for 30 s and centrifuged at 150 g for 5 min. Then, the cells were resuspended in culture medium at a density of 2 × 10⁶ mL⁻¹ for use.

Characterization of samples

The SEM images were taken by using a secondary-electron SEM (EVO18, ZEISS), and the samples were subjected to supercritical drying and coated with a gold film (a thickness of ∼3–5 nm) beforehand. The optical microscopy images were obtained using an inverted microscope (DMI 3000B, Lecia). The energy dispersive spectrometer (EDS) analysis spectroscopy images were obtained by a COLD FESEM (Hitachi, SU8220).

Simulation

The numerical simulation was performed in Comsol Multiphysics 5.4. The process of ant microbot mandibles opening in response to the light field involved heat transfer, mass diffusion and large deformation, which was proposed by He et al.⁵². Details of the simulation methods were reported in Supplementary Information.

Supplementary information

Source data

Peer review information

Nature Communications thanks Wenqi Hu, Sabine Hauert and Aaron Becker, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.