An FWMAV uses a small brushless motor (Lanxin 2807, China), a two-stage gear drive system, and two spatial 4-bar mechanisms to realize synchronous flapping of wings. One steering engine (Skye 14 mg, China) was used to change the wings’ angle of attack, and two steering engines were used to realize the pitching and rolling of the tail. The mass of an FWMAV is 430 g (without battery). The wingspan and wing chord are 1.1 and 0.2 m, respectively. The length of an FWMAV with the tail is 0.65 m. The length and width of the tail are 0.2 m.

The 5-, 8-, 10-, 12-, and 15-μm SiO₂ microspheres were first washed with piranha solution. The microspheres were then placed on a glass slide and deposited in a 100-nm nickel layer using ion-sputtering equipment (K575XD, Emitech, England) at a 90° incident angle. After a brief sonication in ultrapure water, the Janus microspheres were released from the slide and dispersed in ultrapure water. The Janus microspheres were immersed in a 2-mg/ml PAH solution containing 0.5 M NaCl for 15 min with continuous shaking. The PAH-adsorbed Janus microspheres were centrifuged and washed three times using 0.1 M NaCl solution. The PAH-adsorbed Janus microspheres were then immersed in a 2-mg/ml PSS solution containing 0.5 M NaCl for 15 min with continuous shaking, followed by three-time repeated centrifugation/washing steps. The (PSS/PAH)₅-coated Janus microspheres were obtained by repeating the aforementioned deposition procedure.

The (PSS/PAH)₅-coated Janus microspheres were modified with the silane coupling agent, noctyltriethoxysilane. A combination of 1 ml coupling agent, 1 ml deionized (DI) water, 1 ml ammonium hydroxide (25 wt%), and 10 ml ethanol were mixed in a three-necked round-bottom flask by electrical stirring. The mixture was placed in a 50°C environment for 30 min to complete hydrolysis reaction. Afterward, the dehydrated microspheres were added to the mixture and stirred vigorously for 3 h. Finally, the modified microspheres were washed with ethanol three times and dried in a vacuum oven at 40°C for 12 h. Such functionalization involves the formation of a hydrophobic layer by self-assembly of long alkanethiol chains on the rough surfaces of microspheres.

Scanning electron microscopy (SEM) analysis was performed with a Hitachi S-4300 instrument at an operating voltage of 10 keV. Mapping analysis was conducted by using an Oxford energy-dispersive X-ray spectroscope (EDX) attached to an SEM instrument and operated by Inca software. Videos of Janus microrobots were captured at 25 frame s⁻¹ by using an inverted optical microscope (IX73, Olympus, Japan) coupled with a ×20 objective and a Point Grey CCD camera. The video data were analyzed using ImageJ and MATLAB to obtain the trajectories and velocities of the microrobots.

The setup for generating an external rotating uniform magnetic field consists of Helmholtz coils with three degrees of freedom, a multifunction data acquisition, and a three single-channel output power amplifier. Based on controlling the current and voltage of the Helmholtz coils, an external rotating uniform magnetic field can be generated in any plane of the 3D space to actuate the microrobots in different motion modes. The external magnetic field setup was placed on the observation platform of the microscope to achieve real-time observation of the Janus microrobots.

To create a highly efficient platform for pollutant adsorption, magnetically propelled Janus microrobots combined with an FWMAV were used to remove the residual oil and organic pollutants from aqueous media. As shown in Figure 1A, the FWMAV transports microrobots to the polluted environment, and then spray the microrobots by changing the angle of attack. The microrobots were composed of silica microspheres, a nickel layer polystyrene sulfonate (PSS)/polyallylamine hydrochloride (PAH) multilayer, and hydrophobic layer. The multilayered polymer was a cost-effective microscale or nanoscale adsorbent for the removal of organic pollutants in water. The hydrophobic layer was used to adsorb and carry the residual oil. Hence, the surface-modified magnetically propelled microrobots and FWMAV systems are essential to achieve this goal. The FWMAV has a length of 0.65 m and a width of 1.1 m, as shown in Figure 1B. The depot, which was used to store 80 g microrobots, was fixed in the barycenter of the FWMAV. The fabrication of Janus microstructures with five bilayers of PSS/PAH and modification of oil-sorption hydrophobic layers are presented in Figure 1C. These Janus microspheres were fabricated by half-coating silica microspheres with a thick nickel layer using electron beam evaporation. Five bilayers of PSS/PAH were deposited on the surfaces of Janus SiO₂ microspheres via layer-by-layer self-assembly. Such functionalization involves the formation of a hydrophobic layer by self-assembly of long alkanethiol chains on the rough surfaces of microspheres. The surface morphology of these Janus microrobots with a diameter of 5 μm was characterized by optical microscopy, scanning electron microscopy, and energy-dispersive X-ray spectroscopy map analysis to examine the nickel composition of oil-sorption Janus microrobots (Figures 1D, E).

To investigate the mechanism of the FWMAV, a simple model was used to simulate the wind tunnel experiment simulation by XFlow. XFlow adopts the particle tracking method based on the three-dimensional Boltzmann model of particles, and the evolution equation is f i ( x + e i δ i , t + δ i ) − f i ( x , t ) = 1 τ ( f i ( x , t ) − f i e q ( x , t ) ) , (1) where   f i is the particle velocity distribution function,   f i e q is the local equilibrium function, τ is the relaxation time of dimension one,   e i is the discrete velocity, and δ i is the time step.

The equilibrium distribution function is f i e q ( x , t ) = ρ w i [ 1 + e i ⋅ u c s 2 + ( e i ⋅ u ) 2 2 c s 4 − u 2 2 c s 2 ] , (2) where w i is the weight coefficient and c _s is a parameter related to the speed of sound; the fluid macroscopic density ρ and velocity u can be shown as follows: ρ = ∑ i f i e q , (3) ρ u = ∑ i c i f i e q . (4)

The size of the wind tunnel is 5 m*2 m*2 m, and the incoming velocity is 5 m/s. The flapping frequency is 10 Hz. The angle of attack, the upward flapping amplitude, and the downward flapping amplitude are 25°, 45°, and 15°, respectively. The motion function is θ x = 30 ° cos ( 2 π ft ) + 15 ° . (5)

The vorticity and velocity near the wings are shown in Figures 2A, B. Flapping wings can produce a series of Karman vortex streets on the trailing edge, thereby affecting the thrust of flight. Figure 2C presents the simulated lift force of the FWMAV over 1s. During the upward flapping of the wings, the lift force produces a trough, which is a negative lift. During the downward flapping of the wings, the lift force produces a wave crest, which is a positive lift. The flapping time difference improves the lift force. Because of the flapping motion of the wings, the FWMAV could achieve unique aerodynamic advantages over the traditional fixed wing or rotary flight when the feature size is on a small scale.

Figure 3A, along with the corresponding Supplementary Video S1, displays the continuous locomotion of the FWMAV in air. The FWMAV controls flight speed by changing the flapping frequency of the wings. The tail is used to control the pitch of the FWMAV by changing the angle of attack. If the FWMAV needs to yaw, it will roll the tail and lean the wings to one side at the same time. The maximum and minimum speeds of the FWMAV are 31.6 and 7.3 km/h, respectively. As shown in Figure 3B, as the FWMAV transports microrobots to the polluted environment, microrobots can be sprayed by changing the angle of attack. The effect of battery capacity on the locomotion distance of the FWMAV has also been investigated experimentally. As shown in Figure 3C, the locomotion distances of the FWMAV were 3.3, 6.0, 8.8, and 11.3 km with the battery capacities of 400, 800, 1,200, and 1,600 mAh, respectively. Because of its attractive performance, the FWMA can be used for long-distance transportation and delivery of microrobots.

The abilities of remote and precise actuation are highly attractive features for microrobots in the application of environmental remediation. Here, we demonstrated the remote navigation strategy of Janus microrobots. Figure 4A shows a three-dimensional rotating magnetic field generator composed of three-degrees-of-freedom Helmholtz coils. The schematic of the controllable movement of a single Janus microrobot is presented in the inset. First, a circularly polarized rotating magnetic field is applied in the x-z plane and the microrobot is rolled along the x axis. The magnetic field is governed by H ( t ) = H 0 [ cos ( ω t ) e x + sin ( ω t ) e z ] , (6) where H ₀ is the magnitude of H(t), ω is the angular frequency of the magnetic field, t is time, and e _x and e _z are the unit vectors along the x and z axes, respectively (hereafter, e _y is that along the y axis).

When the rotating magnetic field was changed and applied in the y-z plane, the direction of the microrobot motion changed to the y axis. H ( t ) = H 0 [ − cos ( ω t ) e y + sin ( ω t ) e z ] . (7)

The propulsion direction of the microrobot can be altered by changing the direction of the rotating magnetic field, which can be achieved by controlling the input current manually. It was essential to investigate the velocity of Janus microrobots since the efficiency of oil-sorption is affected by the motion of microrobots. When a microrobot was exposed to a rotating magnetic field, the torque induced by the rotating magnetic field and the viscous drag due to the surface caused it to roll forward along the surface. The dependence of the velocity of Janus microrobots with different sizes on the driving frequency was characterized, as shown in Figure 4B. The velocity of the 5-μm Janus motor increased from 9.0 to 48.1 μm/s (∼9.6 body length/s) upon increasing the driving frequency from 2.5 to 40.0 Hz with a magnetic strength of 20 mT. The 8-, 10-, and 12-μm Janus microspheres presented similar speed trends, and their speeds were high up to 71.8 μm/s (∼9.0 body length/s), 92.0 μm/s (∼9.2 body length/s), and 108.7 μm/s (∼9.1 body length/s), respectively. A linear relation is presented between the velocity of the Janus microsphere and driving frequency. The speeds of larger Janus microspheres are higher than those of smaller ones at the same driving frequency, while their relative speed is only frequency dependent. Notably, for 15-μm Janus microspheres, the speed increased linearly with the driving frequency and reached a maximum velocity of 109.0 μm/s (∼7.3 body length/s) at the frequency of 30 Hz and magnetic strength of 20 mT. Further increasing the frequency reduced the velocity. This maximum synchronization frequency is called the step-out frequency. We speculated that the reason for the decrease in speed is the occurrence of an out-of-step phenomenon and the increase in resistance caused by the increase in speed. The efficiency of pollutant purification is directly affected by the speed of the microrobot. Based on the results shown in Figure 4B, the 5-μm Janus microspheres were used as the preferred microrobots in the subsequent experiments. Controllable movement was the basis of the microrobot working in the micropores. Based on the aforementioned sensitive magnetic orientation, a microrobot walked along a predefined star-shaped trajectory as shown in Figure 4C (taken from Supplementary Video S2). The corners of the “star” track line were achieved by adjusting the angle of the magnetic field.