3D Printing of Multilayer Magnetic Miniature Soft Robots with Programmable Magnetization

Magnetically driven miniature soft robots exhibit fast and dexterous responses to an applied external magnetic field. With remote manipulation, controlled navigation of robots can be realized within hard‐to‐access spaces for potential use in the human body. Existing magnetic miniature soft robots using digital light processing are fabricated from planar sheets, and thus have limited shape transformations and locomotive behaviors. Herein, a multilayer 3D printing method is reported for patterning magnetic nanoparticles in ultraviolet (UV)‐curable polymer matrix. Various multilayer 3D structures within 10 mm in overall size are fabricated with controlled volumes at different parts, which outperform 2D folded shapes in terms of robustness and kinematic flexibility. By programming heterogeneous magnetization within discrete multilayer robot segments, magnetic torque‐induced shape changes including gripping, rolling, swimming, and walking are induced by a global actuation field. Stacked design features with minimum dimension of 200 μm and encoded magnetization with resolution of 350 μm can be realized in the printing process. Meanwhile, enhanced deformation flexibility and formation of orientation‐anchoring mechanisms are created by integrating multiple materials with distinct mechanical and magnetic properties, respectively, which enables the creation of versatile 3D multi‐material actuators.

directly patterning magnetic composite with MNPs loaded in the polymer matrix using DLP.Meanwhile, 3D structures can be produced at a higher throughput using DLP than mold casting [21][22][23] or jig-assisted printing. [5]For DIW, structural instability arises in filament stacking, so the rheological properties of the material should be tuned and supporting parts may need to be used as buttresses.Similarly, drop-on-demand 3D printing also requires tuning of the material's rheological properties.With DLP, no rheological modifications or external supports of materials are needed during layer stacking as it uses patterned light to cure layers of photoresponsive polymer.Therefore, multilayer 3D magnetic structures on millimeter scale can be produced with high efficiency and reproducibility by directly patterning magnetic composite via the digitalized process of DLP, which provides a novel way of fabricating sophisticated multilayer magnetic actuators, in the meantime, enabling the production of versatile miniature structures that are based on photocurable stimuli-responsive materials.
The structural complexity of magnetic robots also gets enhanced in this work using the technique of DLP.Small-scale 3D structures that are manufactured using DLP in previous studies approximate each actuating part as a single magnetic point dipole, and use concentrated fields generated by an external permanent magnet to actuate the magnetoactive segment. [28,29]Such actuation schemes rely on the magnetic force generated by the permanent magnet in close proximity, where complex field changes such as oscillatory or periodic signals cannot be achieved, thus limiting the maneuverability for printed robots.Other than the magnetic-force-induced deformation, shape morphing with higher complexity can be realized through magnetic-torque-induced deformation.27] With the techniques of fabrication and magnetization encoding discussed earlier, programmed shape transformation and controlled locomotion such as gripping, [27,28] crawling, [19,29,30,37] swimming, [2][3][4][7][8][9]13,14,22,38] metachronal waving, [23,39] or multimodal behaviors [40] can be realized under magnetic actuation or manipulations combined with light, [2,41] thermal, [18,42,43] or radio frequency [10] signals.
The complexity of structural design and magnetic polarity impacts the functionality of printed microrobots.With DLP, we have previously demonstrated the fabrication of 2D folded magnetic structures with programmed discrete magnetization, [27] and we also improved the design complexity by integrating multiple materials with different magnetic or mechanical properties. [44]However, single-layer robot shapes have limited performance in terms of structural flexibility and compliance compared with multilayer 3D structures.For instance, volumes of joints and actuating parts of multilayer structures can be adjusted by stacking features vertically during the printing process to get higher bending flexibility and greater actuating torques, respectively, which is unrealizable in single-layer fabrication.Building a volume through vertical stacking can also prevent the printed robot from capillary trapping which single-layer structures may suffer from during actuation.
In addition, the structural and magnetic characteristics also depend on the types of materials the robot contains.The structural compliance and maneuverability can be enhanced by integrating rigid and flexible materials as well as incorporating magnetically soft and hard materials.Since the magnetic and mechanical characteristics are controlled by embedded MNPs and polymer matrices, respectively, such properties can be engineered in parallel to enhance the functionalities for printed robots.
In this work, multilayer magnetic shapes including stoolshaped, gripper-shaped, capsule-like, helical, and walking robots were fabricated with encoded discretized magnetization patterns.Independent deformation of each actuating part and an overall complex shape changes were triggered by magnetic torque with a global actuating field generated by electromagnetic coils.Enhanced shape morphing and locomotion capabilities were observed by integrating materials with different magnetic and mechanical properties for the walking robot.

Multilayer Printing with Magnetization Encoding
Extended from the method of printing 2D folded magnetic shapes using DLP in our previous work, [27] multilayer miniature soft robots were fabricated by vertically stacking layers of magnetic patterns in this paper.A bottom-up printing method was used, where the resin is cured through a window in the bottom of the resin vat by a light source from below.As is common in standard (nonmagnetic) DLP-based 3D printers, the build platform within the vat is raised sequentially after each layer is completed, to enable multilayer structure printing.The magnetization of individual regions within layers is encoded by orienting embedded MNPs during selective curing of discrete zones using projected UV light patterns.

Fabrication System
Figure 1a shows the physical system for printing 3D magnetic shapes with discrete magnetization profiles.The magnetic composite, comprising NdFeB nanoparticles within a UV photopolymer, is filled into a small resin container.The resin container has a transparent base made of acrylic, which allows the light pattern projected from UV light source to be transmitted through.During printing, the light pattern travels through a series of lenses and mirrors and gets focused on the curing zone within the vat which is located directly above the base window.
To align the embedded MNPs within the resin to allow for patterned magnetization of a target region, a large magnetic field is induced at the build site by a permanent cubic magnet which is placed below the prism mirror.The advantage of placing this cubic magnet below the resin container (as opposed to other arrangements such as placing this magnet to the side which would allow for easier optical paths) is that this magnetic location keeps the MNPs from being pulled laterally during curing that would distort the MNPs' uniformity within the cured structures.Placing the permanent magnet directly below the resin container would block the optical path of UV light as the light is projected from below either.However, the permanent magnet needs to be placed in close proximity to the curing zone to ensure large alignment field, hence a small-sized prism mirror is inserted in between to bring the projected light pattern in from the side.The relation between this alignment field magnitude and separation distance as well as the relation between field angle and orientation of magnet is shown in Figure S2, Supporting Information.To ensure a large alignment field as well as prevent collision between the permanent magnet and the prism mirror, the smallest separation distance that can be achieved from magnet's center of mass to the curing zone is set to 3.25 cm.With this separation distance, the projected alignment field measured at the curing zone ranges from 48 to 84 mT depending on the orientation of the magnet, which is shown as Figure S2c, Supporting Information.
To define a single-layer thickness during the printing process, a small gap is formed between the base of the container and the build plate.The build plate is mounted on a linear motion stage and can be lifted by adjusting the vertical position.The cured structures need to be detached from the base when lifting the build plate for multilayer stacking, thus, a PDMS film is coated on top of the base to reduce such adhesion.
Figure 1b shows the process of manufacturing multilayer magnetic structures with magnetization encoding.The pre-magnetized embedded MNPs were first oriented by the alignment field, at which time the UV light was projected to mechanically fix the magnetization direction by polymerizing the selected regions and freezing nanoparticles' orientations.
After the selected region was cured, the build plate was lifted for a small distance to create a new gap with the cured structure attached.The surrounding magnetic slurry flowed in and filled the gap due to its weight.Subsequent layers were printed by repeating the same procedure discussed earlier.

Cross-Linked Thickness
Because the polymer matrix is only partially transparent (due to embedded MNPs), the UV light is attenuated as it travels through the slurry, leading to a cross-linked polymer thickness which is dependent on the powder loading ratio as well as the curing time.This cross-linked thickness is crucial to determining the layerlifting distance in multilayer fabrication.A lifting distance which is too large can cause fracture of printed structures, whereas a small lifting distance may result in programming interference between adjacent layers.Therefore, the lifting distance is set to 75-85% of the calibrated cross-linked thickness during layer stacking, which has been observed to maintain the bonding of adjacent layers while preventing programming interference between layers.
The magnetic material is composed of MNPs and UV photopolymer.NdFeB or iron particles were used as MNPs, which served as magnetically hard (permanent) or soft (nonpermanent) particles, respectively.UV polymers with different flexural modulus including FormLabs Clear V4 (%6.91 GPa), FormLabs Elastic 50 A (%0.30 GPa), and GC3D-EBE Flexible (%0.12 GPa) [44] were chosen to meet different structural requirements under various experimental scenarios.Three types of magnetic composites were studied here with different combinations of MNPs and photopolymer, which include two flexible hard magnetic composites (NdFeB nanoparticles mixed with FormLabs Elastic 50 A or GC3D-EBE Flexible) and one rigid soft magnetic composite (iron particles mixed with FormLabs Clear V4).
The cross-linked thickness varies for different polymer matrices.Meanwhile, the transparency of the composite varies for different mass fractions of MNPs, which affects the cross-linked thickness of the structure layer when being exposed to UV light.Another factor that impacts the cross-linked thickness is the orientation of embedded nanoparticles, whether "in-plane" or "out-of-plane" with respect to the horizontal plane.Figure 2a shows the cross-linked thickness with varying mass fractions of MNPs under different exposure time.Here, the cross-linked thickness was tested for plain UV polymer as well as magnetic composites with different mass ratios and orientations of embedded MNPs.Square samples with size of 4 by 4 mm were printed for the thickness measurement using a micrometer.
Based on the measured result, the cross-linked thickness increases as the exposure time increases.For the same type of composite, the thickness decreases as the mass ratio of embedded MNPs increases.In addition, the "in-plane" orientation shows lower cross-linked thickness than the "out-of-plane" orientation under the same condition.

Evaluation of Magnetization Encoding
The magnetization of the robot segments is controlled by orienting embedded MNPs using the alignment field in sequence.However, features that were already printed can themselves locally generate a magnetic field, which in some cases affect subsequent MNPs' reorientation processes in nearby features within the same layer and segments in subsequent layers.Such magnetic interaction can cause inaccuracy in magnetization encoding.To assess the degree of such influence, the encoding result of magnetization is evaluated by reading the magnetic polarity using magneto-optical sensor (MagView S, Matesy GmbH, Germany) for both single-layer and double-layer structures.
For a single-layer structure, we project a 6 Â 6 mm square pattern of micro-QR code to the magnetic composite and encode black and white phases by orienting embedded NdFeB particles in 0°and 180°"out-of-plane" magnetization directions, respectively.The projected pattern is shown in Figure 2b, showing a magnetization resolution of 350 μm.By visualizing the magnetization pattern of fabricated square samples using the MagView, a  clear pattern of micro-QR code is generated and the character string of "UofT" can be successfully scanned.
To test the ability to accurately program a two-layer structure, the first layer is given a uniform 0°"out-of-plane" magnetization across its entire surface, while the magnetization of the second layer is encoded with the same micro-QR code as single-layer structure.By observing each side with the MagView, magnetization profiles for both layers of the double-layer structure can be displayed as Figure 2c.In the visualization process of the second layer, the first layer also projects magnetic fields that can be captured by the MagView, thus generating a net magnetization direction toward 0°and shifting the displayed pattern toward darker spectrum.Figure S4, Supporting Information, shows the magnetization profiles for both single-layer and double-layer structures under the same field range.It is observed that the QR code pattern of the double-layer structure is darker than the single-layer structure.Therefore, the field range of the QR code pattern is shifted by -2 kA m À1 in Figure 2c, and the character string of "UofT" can still be successfully scanned after field shifting.
The high resolution of magnetization encoding for both single-layer and double-layer structures indicates that the external alignment field is dominant in orientating the embedded MNPs, and the magnetic effect of formerly printed features on subsequent features is negligible.Therefore, the magnetic polarity across multiple layers can be created with high precision using such magnetization approach.

Integration of Multiple Materials
Multiple materials with different stiffness and magnetic properties were integrated to enhance the functionality and complexity of fabricated robots.The integration of multiple materials can be achieved by switching resin type during fabrication.Locking docks were added on the positioning stage to secure the build plate during the switching process, which ensures the consistency of build plate's position during material switching and the alignment of projected light patterns for different materials.The build plate was extracted out of the workspace and cleaned by flushing away uncured composites when changing materials.The build plate with cured structure was then placed back for the printing of next material.
Soft magnetic material (SMM) such as iron exhibits lower coercivity and higher permeability than hard magnetic material (HMM) such as NdFeB.While SMM cannot be programmed with a fixed magnetization direction, application of an alignment field during curing will align the particles of SMM in chains along the applied field axis.][47] This easy axis of magnetization can be switched forward/backward by applying the actuating field in an oscillating manner.Based on such easy-axis property, a segment made of SMM can be used as orientation-anchoring mechanism to fix the heading angle of the robot while the segment made of HMM can be used to exhibit reciprocal motions under oscillatory field signals.

3D-Printed Multilayer Structures
Compared with 2D folded shapes, 3D structures have better performance in terms of robustness and kinematic flexibility.For example, capillary force plays an essential role in the actuation of 2D folded soft robots, which may cause the robot to adhere to the inner surface of the workspace.However, the actuating part of 3D structures can assume a larger volume and more complex geometry, thus having greater force and smaller contact area to overcome the capillary effect.
Figure 3 shows printed 3D structures including stool-shaped, gripper-shaped, capsule-like, and helical robots which use the magnetic composite made of NdFeB nanoparticles and Elastic 50 A photopolymer.The stool-shaped robot consists of three layers, where the top and bottom layers are encoded with anisotropic magnetization and the middle layer is used for conjunction.The structure gets deformed into a stool shape when actuated by the external magnetic field, which is shown as Figure 3a.
The gripping arms of gripper-shaped robot are printed by stacking layers of rectangle patterns with the smallest width of 200 μm, which demonstrates the capability of aligning and stacking small features along vertical direction using such fabrication method.Meanwhile, the dimension of the feature can be precisely controlled during stacking.As Figure 3b shows, the arm is thinner at the joint part for higher flexibility and wider at the link part to get a greater actuating torque.

Capsule-Like Robot
Capsule-like robots are promising robotic tools in untethered drug delivery or biological sampling. [48]We print a capsule-like robot by stacking ring shapes along the height of the cylinder with two solid circles at both ends.Small rectangular patterns are stacked vertically to form the backbone.The volume of the backbone needs to be small to ensure the bending flexibility when opening and closing the caps.One layer of ring is skipped in the middle to create the gap between two caps.A small supporting beam is added in between to prevent the structural delamination during build plate lifting.
Figure 3c shows the magnetization profile of the robot.The net magnetic moment is pointing along the height of the cylinder.In the idle state, two caps attract each other due to magnetic interaction, which forms its "closed" state.When deployed into actuation field, the caps align themselves with the external field due to magnetic torque and shift into "open" state.The relation between the opening angle and the applied field is shown in Figure 4a.It is observed that the opening angle of the capsule is zero when the external field is lower than 3 mT.Therefore, a small field can be applied to steer the robot without opening the caps.The opening angle increases as the applied magnetic field increases and becomes stable when the field strength is greater than 35 mT.
Figure 4b shows the actuation of the capsule-like robot in a maze map.The robot can achieve the procedure of cargo gripping, transporting, and releasing.During the process of gripping and releasing, a uniform magnetic field of 25 mT within horizontal plane was applied to open the caps of the robot.A small field along the vertical axis was superimposed to generate tumbling motion for the capsule to approach the cargo.The robot got into "closed" state with two caps attracting each other when the external field was removed, thus gripping the cargo tightly.Then, a rotating magnetic field of 2 mT was applied to rotate the capsule and transport the cargo to the other end of the maze map.The uniform field of 25 mT was applied again to open the caps of the robot for cargo releasing.Such procedure can also be used to sample and transport swarm of small particles.

Helical Robot
][9]14] In this work, helical robot with height of 4 mm and diameter of 2 mm was fabricated by stacking 20 layers of rectangles along its  height.The thickness of each layer is 0.2 mm.The helix angle is controlled by the twisting angle between adjacent layers.Figure 3d shows the fabricated helical robots with helix angles of 30°, 45°, and 60°, respectively.All layers have "in-plane" magnetization with the same direction; thus, the magnetic moment of the helical robot is along its geometric short axis.By applying a rotating field with high frequency, the helical robot can rotate along its long axis and generate propulsive force in liquid environment.
We measured the swimming speed of helical robots by navigating the helical robots in a 3D printed channel.The helical robots were deployed into the electromagnetic coil system and actuated by a rotating magnetic field with the amplitude of 16 mT.Figure 4c shows the measured swimming speeds of helical robots under rotational frequencies of 8, 10, and 12 Hz in liquids with different viscosities.The helical robot exhibits tumbling motion under low rotational frequency and liquid viscosity, which prevents it from propulsion.When the liquid viscosity or the rotational frequency increases, with a tumbling motion, the fluidic drag on the robot becomes larger, thus transforming its rotation along the geometric short axis to the long axis.In this way, the propulsive force is generated to propel the robot for swimming.The "step-out" occurs when the rotational frequency and liquid viscosity become high, thus impeding propulsion.
Figure S5, Supporting Information, shows the swimming patterns of helical robots in the liquid with viscosity of 5 mPa s.The robots with different helix angles were actuated under three rotational frequencies.The swimming speed increases as the rotational frequency increases.It is observed that the helical robot with helix angle of 45°has higher swimming speeds than the other two angles.Therefore, we deployed such robot into a vascular model and successfully navigated it from one end to the other using rotating field with rotational frequency of 12 Hz, which is shown as Figure 4d.The direction of robot's propulsion is controlled by steering the plane where the rotating field lies.

Walking Robot
A walking robot with four legs was fabricated by integrating SMM and HMM, as well as rigid and flexible polymer matrices.As discussed previously, SMM has low coercivity and can induce a large magnetic moment in an applied field, which makes it a great candidate as orientation-anchoring mechanism under oscillating field signals.As Figure 5a shows, the main body of the walking robot is made of rigid SMM while the legs are made of flexible HMM.For each leg, the width of the link is greater than the joint for a larger deformation.Meanwhile, the volume of the soft magnetic segment needs to be large to ensure the induced magnetic moment of soft magnetic segment is greater than that of deformed hard magnetic segments, thus making the net magnetic moment align with the external field and exhibit the anchoring characteristic.The induced magnetic moment points along the alignment direction of particle chains, so the embedded iron particles are reoriented to align along the geometric long axis of the soft magnetic segment, which is shown as Figure 5b.
Cooperative torque can be generated on the soft magnetic segment to correct the heading angle if it deviates from the direction of applied field.The shoulder parts are made of plain UV photopolymer GC3D-EBE Flexible for better deformation performance due to its low flexural modulus.
The oscillating magnetic field to actuate the walking robot is applied within the horizontal plane and has the pattern of sinusoidal wave with the amplitude of 40 mT.Fields with magnitude lower than 10 mT are filtered because small fields cannot induce large magnetic moment on soft magnetic segment but generate magnetic torques on deformed hard magnetic segments, which may cause the reversal of robot's orientation.
By applying such oscillating field along the geometric long axis of soft magnetic segment, the forelegs and rear legs made of HMM swing toward and away from each other while the heading angle remains consistent.To make the robot walk, small field components along the vertical axis are superimposed to tilt the body of the robot periodically.A walking gait is formed as Figure 5d shows, which induces the robot to walk toward the right.The heading angle of the robot can be steered by rotating the plane where the oscillating field lies.Figure 5e measures the walking speed of the robot under different oscillatory frequencies.The "step-out" occurs when the frequency is higher than 3 Hz.

Discussion
In this article, we developed a fabrication approach to 3D print millimeter-scale magnetic structures with encoded magnetization profiles.Structures with complex shapes can be produced within minutes using stereolithography.The digital process also enables the manufacturing of 3D miniature robots with high degree of automation and reproducibility.Table S1, Supporting Information, summarizes the features of fabricated robots discussed in this work.
Compared with single-layer structures, multilayer printing explores a wider spectrum for robot designs and greatly enhances the structural complexity.Meanwhile, large magnetic torques can be generated by stacking thick features or multiple layers for deforming parts, which is not achievable in single-layer fabrication, thus contributing to a higher strength for actuation.Moreover, integration of multiple materials with different stiffness and magnetic properties enhances the functionality and maneuverability of printed structures.
However, the presented fabrication approach has limitations in terms of the intensity and continuity of magnetization.The magnetization is formed by orienting pre-magnetized NdFeB nanoparticles using alignment field with moderate strength (48-84 mT).Thus, the intensity of magnetization is lower compared with the single-pass magnetization using saturated field (2-3 T).Consequently, the magnitude of structural deformation for printed shapes under actuation is limited.Moreover, compared with single-pass magnetization to form continuous magnetization on pre-deformed shapes, the presented method requires a longer time to form a smooth and continuous magnetization profile as the magnetization is discretely patterned for each individual segment.
Future work will explore the incorporation of multiple materials with different stimuli.Meanwhile, biocompatibility of the material will be studied combined with in vitro or ex vivo tests using printed robots, such as transporting drug particles to the wound of biological tissues using capsule-like robot or navigating helical robots coated with thrombolytic drugs to dissolve blood clots at target locations.

Experimental Section
Materials: Various types of magnetic materials were prepared by mixing NdFeB or iron particles with different polymer matrices.For flexible HMM, the permanent magnetic particles (D50 = 5 μm, MQFP-15-7, NdPrFeB, Magnequench [Tianjin] Co. Ltd., Tianjin, China) were first magnetized in a strong uniform magnetic field (2.4 T) generated by an impulse magnetizer (9 Tesla Pulse Magnetiser, Magnetic Measurements Ltd., Lancashire, UK) to get the saturation remanence.They were then mixed with flexible UV resin (Elastic 50 A, FormLabs Inc., MA, USA, or GC3D-EBE Flexible Beige, Smart Your Life, Taiwan).For rigid SMM, the iron particles (Iron Metallic Powder À325Mesh, Sculpture Supply Canada, Canada) were mixed with rigid UV resin (Clear V4, FormLabs Inc., MA, USA).Fumed silica of 1% w/w (S5130-100 G, Sigma-Aldrich Canada Co., Canada) was added to the magnetic composite to homogenize the distribution of embedded MNPs and prevent particle aggregation during printing.The mixed composite was then stirred by a glass rod for 5 min to form a homogeneous magnetic slurry.After removing bubbles in a vacuum degassing chamber for 1 min, the magnetic slurry was transferred into a cylindrical resin container.The container was composed of 3D printed walls and a base made of clear acrylic plastic (Clear Scratch-and UV-Resistant Cast Acrylic Sheet, McMaster Carr, OH, USA).A polydimethylsiloxane (PDMS) (SYLGARD 184, The Dow Chemical Company, USA) film with thickness of %1 mm was coated on top of the container base to reduce the adhesion between cured parts and the base when detaching layer during lifting.The polymeric base and curing agent of PDMS were thoroughly mixed with the mass ratio of 10:1.The mixture was then poured into the container and took 24 h to get cured and form the coating.
Magnetic composite made of NdFeB nanoparticles and FormLabs Elastic 50 A was used for the fabrication of stool-shaped, gripper-shaped, capsule-like, and helical robots.For the walking robot, the rectangular body was first printed using SMM made of iron particles and FormLabs Clear V4 and followed by printing the shoulder parts using plain photopolymer GC3D-EBE Flexible.Then, the legs were printed using flexible HMM made of NdFeB nanoparticles and GC3D-EBE Flexible.
Fabrication: The resin container filled with magnetic composite was placed above the prism mirror in the fabrication system.The UV light pattern was generated from a DLP projector (DLP LightCrafter 4500, modified with a 405 nm light engine, Texas Instruments Inc., TX, USA).The light beam was transmitted through a series of convex lenses (LA1608-A/LA1509/LA1986, Thorlabs Inc., NJ, USA) and focused on the curing zone.The embedded MNPs were oriented by the alignment field generated by a cubic permanent magnet below the prism mirror.After that, the region of interest was exposed to the blue light (405 nm, 80 lx) to fix the magnetization direction.The duration of exposure time depended on the desired cross-linked thickness.The reorienting and curing steps were alternated for more cycles to create the customized magnetization profile.The build plate was extracted out of the vat after all curing steps and then washed by isopropyl alcohol (IPA) (IPA 99.5%þ, Caledon Laboratories Ltd., ON, Canada) to remove uncured materials.For multi-material printing, the material in resin container was first changed, and then the build plate was mounted back on the vertical motion stage for the printing of next material.Finally, the build plate with the finished structure was immersed in IPA for 2 min and the cured structure was then detached from the build plate by needles.
Actuation: Electromagnetic coil system with 8 degrees of freedom (DOF) was used for the robot actuation, which is shown in Figure S6, Supporting Information.The printed robots were actuated in the plane 10 cm above the tabletop.The magnetic field was modulated by a joystick.The videos and images were captured by two complementary metal-oxidesemiconductor (CMOS) cameras mounted on top and to the side.

Figure 1 .
Figure 1.a) Schematic representation of fabrication system.The inset shows the sectional view of build plate, resin container, and prism mirror.The vertical motion stage can be adjusted to move along z axis to lift the build plate.The image of the real apparatus is shown in Figure S1, Supporting Information.b) Process of multilayer printing.The yellow arrows within magnetic composite indicate the magnetization directions.The cured and uncured segments of magnetic composite are shown in black and gray, respectively.Red arrows represent the directions of alignment field B, and gray arrows in step (iv) represent the lifting direction of the build plate.

Figure 2 .
Figure2.a) Cross-linked thickness of magnetic composite.Three types of magnetic composites are studied here, of which the types of magnetic nanoparticles (MNPs) and polymer matrix are shown in the title box of each legend.The "in" and "out" represent "in-plane" and "out-of-plane" magnetization, respectively.Zero thickness implies the printed sample is not cured or fully cured.b) Single-layer square structure with encoded magnetization of microquick-response (QR) code.The black and white pixels represent the magnetization directions of 0°and 180°, as indicated by the red and blue arrows.c) Double-layer square structure with magnetization of 0°on the first layer and micro-QR code on the second layer.The magnetization patterns of first and second layers are scanned using the MagView from top side and bottom side, respectively.Scale bar, 2 mm.

Figure 3 .
Figure 3.The 3D-printed robots with encoded magnetization profiles and shape transformation under magnetic actuation.Yellow arrow represents the magnetization direction at each segment and red arrow represents the direction of actuation field.The magnitude of actuation field is 30 mT. a) Stool-shaped robot.b) Gripper-shaped robot.c) Capsule-like robot.d) Helical robots with helix angles of 30°, 45°, and 60°.Scale bar, 1 mm.

Figure 4 .
Figure 4. a) Opening angle of capsule-like robot under applied magnetic field.The insets illustrate the patterns of capsule opening under magnetic fields of 0, 15, and 30 mT. b) Navigation of capsule-like robot in a maze map for cargo manipulation including gripping, transporting, and releasing.c) Swimming speed of helical robots with different helix angles.The robots are deployed into fluids with different viscosities and get actuated under different rotational frequencies.Occurrence of tumbling motion results in zero swimming speeds at viscosity of 1 mPa s, and the "step-out" phenomenon causes zero speeds at viscosity of 50 mPa s. d) Navigation of helical robot in a vascular model.Scale bar, 2 mm.

Figure 5 .
Figure 5. Walking robot made of soft magnetic material (SMM) and hard magnetic material (HMM).a) Schematic representation of walking robot in the idle state.Segments made of SMM and HMM are in gray and black, respectively.The white segment represents the nonmagnetic part made of plain UV photopolymer.Yellow arrow represents the magnetization direction at each segment.Scale bar, 1 mm.b) Illustration of soft magnetic segment with embedded iron particle chains aligned along the geometric long axis.The bidirectional yellow arrow indicates the alignment direction of iron particle chains.Cooperative torque is generated when the external field B is applied.(Note: Iron particles in the figure are only for the illustration purpose of particle chain and are not drawn here to scale.)c) Deformation of robot's legs under magnetic field.The forelegs and rear legs bend toward and away from each other while the orientation of the robot keeps constant as the field direction switches by 180°.Red arrow represents the direction of applied magnetic field.The magnitude of applied field is 50 mT.d) Demonstration of walking gait in a gait cycle from top view.Step (i) shows the gait in the idle state, and the robot moves toward right by circulating step (ii) to step (v).e) Walking speed of the robot in oscillating magnetic field with different frequencies.f ) Motion of the robot under actuation field with oscillatory frequency of 2 Hz.Scale bar, 10 mm.