3D printing of high-precision and ferromagnetic functional devices

The development of projection-based stereolithography additive manufacturing techniques and magnetic photosensitive resins has provided a powerful approach to fabricate miniaturized magnetic functional devices with complex three-dimensional spatial structures. However, the present magnetic photosensitive resins face great challenges in the trade-off between high ferromagnetism and excellent printing quality. To address these challenges, we develop a novel NdFeB-Fe3O4 magnetic photosensitive resin comprising 20 wt.% solid loading of magnetic particles, which can be used to fabricate high-precision and ferromagnetic functional devices via micro-continuous liquid interface production process. This resin combining ferromagnetic NdFeB microparticles and strongly absorbing Fe3O4 nanoparticles is able to provide ferromagnetic capabilities and excellent printing quality simultaneously compared to both existing soft and hard magnetic photosensitive resins. The established penetration depth model reveals the effect of particle size, solid loading, and absorbance on the curing characteristics of magnetic photosensitive resin. A high-precision forming and ferromagnetic capability of the NdFeB-Fe3O4 magnetic photosensitive resin are comprehensively demonstrated. It is found that the photosensitive resin (NdFeB:Fe3O4 = 1:1) can print samples with sub-40 μm fine features, reduced by 87% compared to existing hard magnetic photosensitive resin, and exhibits significantly enhanced coercivity and remanence in comparison with existing soft magnetic photosensitive resins, showing by an increase of 24 times and 6 times, respectively. The reported NdFeB-Fe3O4 magnetic photosensitive resin is anticipated to provide a new functional material for the design and manufacture of next-generation micro-robotics, electromagnetic sensor, and magneto-thermal devices.

The authors contributed equally to this work. * Authors to whom any correspondence should be addressed.
Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.

Introduction
Miniaturized magnetic functional devices (actuators and robots) being of a range of benefits, such as non-contact interactions and fast, reversible actuations, can be widely used in biosensing, micromanipulation, and targeted drug delivery [1][2][3][4][5]. Conventional manufacturing techniques such as photolithography [6][7][8], electrodeposition [9,10], and template self-assembly [11][12][13] are pretty common to be used for fabricating 2D and quasi-3D magnetic devices. However, these processes are difficult to manufacture a miniaturized magnetic functional device with complex 3D spatial structures (i.e. limited by template shape, magnetic coating thickness, material type, and assembly accuracy). Therefore, it is of great interest to develop new manufacturing process to fabricate magnetic functional devices with complex 3D structures.
Additive manufacturing, or 3D printing, is of outstanding advantages for manufacturing 3D magnetic devices, for example, it is able to create complex structures being of high precision as well as the integration of functional components with a wide range of materials. Therefore, this technique is capable to provide notable design flexibility, high accuracy and efficiency in comparison to conventional techniques, especially for fabricating 3D magnetic functional devices [14][15][16][17]. Basically, there are three commonly used 3D printing technologies for manufacturing 3D magnetic functional devices, namely, fused deposition modeling (FDM), direct ink writing (DIW), and stereolithography (SLA). The manufacturing accuracy of FDM and DIW is generally greater than 100 µm restricted by nozzle diameter [18]. On the contrary, in the family of micro-SLA (µSLA), projection µSLA and micro continuous liquid interface production (µCLIP) are considered to be the most promising solutions for both considering printing quality and fabrication time [19][20][21][22][23]. It can achieve a submicron resolution and centimeter-scale overall shape accuracy, which meets the requirement for fabricating magnetic functional devices with high-precision. Additionally, the photopolymerization reaction occurs at the surface of a liquid photosensitive resin, allowing for curing magnetic photosensitive resins with strong absorbance.
In addition to manufacture process, photosensitive resin is also an important factor to be considered for improving printing quality. Magnetic photosensitive resin is an essential composite material produced by uniformly mixing liquid photosensitive resin with micro/nanoscale magnetic particles [22]. The magnetic photosensitive resin is of the advantage of fabricating complex 3D microstructures compared to other printing materials. In particular, the magnetic properties and printing quality of the printed device can be directly adjusted by playing with the micro/nanoscale magnetic particles in resin. In the existing literature, the common micro/nanoscale magnetic particles are Fe 3 O 4 [24][25][26][27][28], Fe [29], and SrFe 12 O 19 [30]. However, the solid loading and printing quality of these resins can be affected by the absorption and settlement of magnetic particles, making it difficult to achieve a uniform material distribution, high printing quality as well as exceptional magnetic properties. To date, Fe 3 O 4 nanoparticle-based resin is found to be the most effective magnetic photosensitive resin for 3D printing, due to its solid loading of 30 wt.%, long-term stability of up to 72 h, as previously reported in our previous work [24]. This resin is of excellent printing quality and has been used to print magnetic grippers [26], micro actuators [30], and magnetic flexible deformation structures [25]. However, it is still not widely used due to its weaker magnetism in comparison to hard magnetic materials, which are commonly used in electric motors and generators, magnetic separation processes, magnetic storage devices, and electromagnetic shielding. There are few studies on hard magnetic photosensitive resins. This is because it is very difficult to incorporate hard magnetic materials into 3D printing techniques, as a result of the low surface area large particle size, and interaction between magnetic particles of magnetic photosensitive resin, resulting in an inferior stability and light absorption of magnetic photosensitive resin [31].
NdFeB is a well-known hard magnetic material with exceptional ferromagnetic properties, commonly used in fabricating hard magnetic devices. It is generally believed that the optimal particle size for NdFeB powders produced using current methods is within a few microns, as this size range allows for a balance between magnetism and stability. Although NdFeB microparticles have been doped into photosensitive resins for printing magnetic robots in literature, it can only fabricate 2D planar structures with a minimum feature size greater than 200 µm [32]. The major challenges for fabricating high precision 3D structures using NdFeB microparticles are how to achieve precise curing and great stability of high solid magnetic resin. In particular, a large penetration depth is caused by a low absorbance of NdFeB microparticles, leading to overcuring or inaccurate curing of magnetic photosensitive resins. Additionally, the magnetic resin with large particle size can cause sedimentation and layering, resulting in an inferior stability of the prepared magnetic photosensitive resin and hence reducing printing efficiency. Therefore, conventional magnetic photosensitive resins are unable to accomplish great printing quality and ferromagnetism simultaneously. In light of this, it is necessary to develop a new type of magnetic photosensitive resin for fabricating high-precision and ferromagnetic functional devices to be used in robotics, electronic devices as well energy storage area.
In this paper, we develop a novel NdFeB-Fe 3 O 4 magnetic photosensitive resin comprising 20 wt.% solid loading of magnetic particles, which is synthesized by combining ferromagnetic NdFeB microparticles and strongly absorbing Fe 3 O 4 nanoparticles to achieve ferromagnetism and incredible printing quality simultaneously. This resin can be efficiently used to fabricate ferromagnetic functional devices with high precision via µCLIP process. We also investigate the highprecision forming capability and ferromagnetic capability of the NdFeB-Fe 3 O 4 magnetic photosensitive resin through theoretical modeling and experimental analysis. To optimize the performance of the resin, we compare and evaluate the properties of five different ratios of magnetic particles. Finally, the printing quality and ferromagnetic capability of optimized magnetic photosensitive resin are discussed.

Printability and magnetism of NdFeB-Fe 3 O 4 magnetic photosensitive resins
The printability of NdFeB-Fe 3 O 4 magnetic photosensitive resins involves using mathematical models to further investigate the effect of the two types of particles on the curing characteristics and stability of resin. Figure 1(a) shows the composition of NdFeB-Fe 3 O 4 magnetic photosensitive resins and the characteristics of the printed samples. A novel magnetic photosensitive resin with great printability and ferromagnetism is prepared by incorporating micron-sized NdFeB with ferromagnetic and nanometer-sized Fe 3 O 4 with strong absorbance into a mixture of NdFeB-Fe 3 O 4 micronanoparticles. Figure 1(b) shows scanning electron microscopy (SEM) images of magnetic particle powders with five different ratios. The SEM images are taken to analyze the morphology and size distribution of the magnetic particles. Additionally, the absorbance and magnetism of the five powders are measured, as shown in figure 1(c). The results indicate that the NdFeB-Fe 3 O 4 micro-nanoparticles are of great magnetic and absorbance properties. However, the curing model for magnetic photosensitive resins remains unclear at present. Figure 1(d) illustrates the schematic diagram of the curing process for NdFeB-Fe 3 O 4 magnetic photosensitive resin. The penetration depth directly affects the curing depth in µCLIP process. Therefore, the penetration depth model of NdFeB-Fe 3 O 4 magnetic photosensitive resins have been established to analyze curing characteristics. The theoretical model of penetration depth investigates the effect of solid content and particle size of two particles on curing characteristics. Here, three assumptions have proposed according to published literature [33]. Frist, we assume that the NdFeB microparticles are prolate ellipsoid (a = b < c) and the Fe 3 O 4 nanoparticles are sphere with a radius of r. Second, we assume that the magnetic particles absorb almost all projected light while being unaffected by shadows and light scattering according to optical measurement results (1 < absorbance < 2; 90% < light absorption < 99%). Finally, we assume that the magnetic particles in the printable magnetic photosensitive resins are uniformly dispersed in the resin. The magnetic resin is considered to have denatured and failed if aggregation and settling of the magnetic particles occurs. It follows that the number of NdFeB-Fe 3 O 4 particles per unit volume (N N−F (w) ) can be represented by equation (1), which is derived from equations (S1)-(S4), where n N−F is the number of magnetic particles, V N−F is the volume of magnetic photosensitive resin, w N andρ N are the content and density of NdFeB microparticles, w F and ρ F are the content and density of Fe 3 O 4 nanoparticles, ρ R is the density of resin. In this work, the solid loading of the magnetic particles is 20 wt.%, so w N + w F = 20. Penetration depth is defined as the thickness at which the outgoing light attenuates to the incident light intensity 1/e ≈ 36.8%. The relationship between the absorbance coefficient (α N−F ) and penetration depth (D PN−F ) can be represented by where D PR is the penetration depth of liquid resin, α PN−F is the absorbance coefficient of NdFeB-Fe 3 O 4 micro-nanoparticles, S N is the projected area of the ellipsoid and S N ∈ . It is assumed that the projected area of the ellipsoid follows the normal distribution. Therefore, the average valuē S N = π(ac+a 2 ) 2 and equation (1) are taken into the equation (2). The penetration depth of NdFeB-Fe 3 O 4 magnetic photosensitive resins (D PN−F ) can be represented by To simplify the equation (3), we let A = a 2 +ac+2r 2 . It is calculated that A ≪ B, so the penetration depth is proportional with w N and inversely proportional with w F at constant size of magnetic particle. Moreover, the penetration depth of NdFeB microparticles magnetic photosensitive resin (D PN ) and Fe 3 O 4 nanoparticles magnetic photosensitive resin (D PF ) can obtained as a particular case when w F = 0 (equation (S6)) and w N = 0 (equation (S8)), respectively. The particle absorption coefficient in the formula is proportional to the measured absorbance, ordered as absorption coefficient of NdFeB microparticles (α PN ) < α PN−F < absorption coefficient of Fe 3 O 4 nanoparticles (α PF ). The three penetration depths are sorted as follows: In particular, some of the parameters in the theoretical model are affected by the purity, surface morphology and irregular particle size of the magnetic particles. This theoretical model developed can qualitatively predict the trend of the penetration depths for different material ratios by varying each parameter. It follows that the penetration depth is proportional with particle size at a constant content of magnetic particle. Therefore, the doping of Fe 3 O 4 nanoparticles will reduce the penetration depth and improve printing quality.
Although the magnetic particles are evenly mixed and distributed within the liquid resin when the preparation of magnetic photosensitive resin is completed, they will eventually settle and break the original uniform distribution with time. In particular, micron-sized NdFeB particles and nanoscale Fe 3 O 4 particles are selected as magnetic particles. It is difficult to quantify uniformly the aggregation and sedimentation of both particles for such large wide ranges. Therefore, the force analysis of a single magnetic particles is proposed to evaluate NdFeB-Fe 3 O 4 magnetic photosensitive resins stability.
The forces acting on a single magnetic particle mainly include gravity (F g ), buoyancy (F f ), kinetic viscous resistance (F v ), and the collision force of particle Brownian motion. To simplify the calculation, the resultant forces of NdFeB microparticles (F ZN ) and Fe 3 O 4 nanoparticles (F ZF ) in the vertical direction are represented by equation (4), where f N = 6πηaσ N and v N are drag coefficient and particle settling velocity of NdFeB microparticles, σ N is ellipsoid drag parameters (equation (S9)), f F = 6πηr and v F are drag coefficient and particle settling velocity of Fe 3 O 4 nanoparticles, η is liquid resin viscosity. According to equation (4) and the disparate particle size, it can be concluded that F ZN is significantly greater than F ZF . Therefore, Fe 3 O 4 nanoparticles magnetic photosensitive resin (N0-F1) is more stable than NdFeB microparticles magnetic photosensitive resin (N1-F0) in unit volume, and NdFeB-Fe 3 O 4 magnetic photosensitive resins stability should be lie in between. In addition, Brownian motion is characterized by the movement and diffusion of magnetic particles from high concentration to low concentration. It is obvious that the average displacement of the Brownian motion of Fe 3 O 4 nanoparticles is larger than that of NdFeB microparticles, which effectively improves the diffusion effect. Thus, the doping of Fe 3 O 4 nanoparticles will effectively improve the stability of NdFeB-Fe 3 O 4 magnetic photosensitive resins.
Magnetism is determined by the magnetic particles in the liquid resin. NdFeB is a well-known hard magnetic material that has significantly stronger magnetism than Fe 3 O 4 soft magnetic material. However, there will be exchange spring behavior after mixing NdFeB microparticle powders and Fe 3 O 4 nanoparticle powders. The exchange spring behavior has been extensively studied in the literature [34]. Therefore, the room-temperature hysteresis loop of NdFeB-Fe 3 O 4 magnetic powders and printed samples (20 wt.%) with five different ratios were characterized by vibrating sample magnetometer (Dynacool-14T, Quantum Design Inc.). Figure 1(e) shows that the intersection of the hysteresis loop with the X-axis and Yaxis (coercivity and remanence) keeps getting closer to the zero point with the increase of Fe 3 O 4 nanoparticles due to the exchange spring behavior. It follows that coercivity and remanence are proportional to the solid loading of NdFeB microparticles.
To summarize, the doping of Fe 3 O 4 nanoparticles improves printing quality and stability of NdFeB-Fe 3 O 4 magnetic photosensitive resin, but at the expense of reduced magnetism. The printability and magnetism will be verified and optimized the NdFeB-Fe 3 O 4 magnetic photosensitive resin ratio in the following experiments.

Stability and curing characteristics of NdFeB-Fe 3 O 4 magnetic photosensitive resins
The stability is a crucial property of magnetic photosensitive resin, as it can affect the curing characteristics of the material significantly. The stability of the magnetic photosensitive resins can be evaluated by storing them at room temperature and recording the precipitation over the duration of 2.5 h at a 0.5 h interval (figure S1). The precipitation speed is found to be inversely proportional to the solid loading of Fe 3 O 4 nanoparticles. Therefore, experimental result shows that the doping of Fe 3 O 4 nanoparticles significantly improves the stability of NdFeB-Fe 3 O 4 magnetic photosensitive resin and also corroborates the results obtained from theoretical analysis.
The curing characteristics of the magnetic photosensitive resin are measured experimentally using the well-established speed working curve method for µCLIP process. In this work, the solid loading of the NdFeB-Fe 3 O 4 magnetic photosensitive resin is 20 wt.%, so the same printing method (>15 wt.%) is used as the previous work [24]. Correspondingly, the curing depth of µCLIP process could be determined by the speedworking curing method in our previous work [22], where C D is the curing depth of a single exposure, v c is the critical stage moving speed and v s is the actual stage moving speed.
A ladder-like model with 100 µm-thick crossbeams is designed and sliced into 10 µm-thick layers for the characterization of magnetic photosensitive resins in our µCLIP system. The above printed ladder-like model is used to be a magnetic standard sample and its computer aided design (CAD) model is shown in figure 2(a). The standard samples are printed by NdFeB-Fe 3 O 4 magnetic photosensitive resins with five different ratios (figure S2), and thickness of the printed crossbeams is measured using 3D microscope. Figure 2(b) shows that fitted curing curve of NdFeB-Fe 3 O 4 magnetic photosensitive resins with five different ratios in µCLIP system with a fixed ultraviolet (UV) power density at 2.10 mW cm −2 and various stage speeds. The experimentally measured curing depth is linearly proportional to the stage speed in the linear logarithmic X-axis plot, which agrees well with the underlying curing model. According to the fitted curing curve, optimal stage speeds of NdFeB- the structural reliability of the manufacturing process and the consistency of the magnetic property. In addition, the penetration depth of the NdFeB-Fe 3 O 4 magnetic photosensitive resin can be represented by the slope of the linear fitting in figure 2(b). Figure 2(c) shows that the penetration depth is proportional to the solid loading of NdFeB microparticle. It is found that the theoretical trend and the trendline of experimental results are consistent. Thus, it can be concluded that an increase in content of Fe 3 O 4 nanoparticles can effectively increase the attenuation of the incident UV light and reduce the penetration depth, thereby improving the printing quality.

Characterization of high precision forming capability in
3D printed magnetic samples. Under the optimized process condition obtained from section 3.1, the high precision forming capability of 3D printed standard samples using NdFeB-Fe 3 O 4 magnetic photosensitive resins with five different ratios is compared. We have measured surface quality and observed differences in the fracture surface and film contact surface among five printed magnetic standard samples. Figure 2(d) shows that the fracture surface of five printed magnetic standard samples using a 3D microscope and SEM images. These images are taken to analyze the microstructural characteristics of the fracture surfaces. The 3D microscope images indicates that the flatness of the fracture surface is significantly improved as the Fe 3 O 4 nanoparticle content increases. It is noteworthy that the flatness of the fracture surface is an important factor in evaluating the high precision forming capability of 3D printed magnetic samples. SEM images are then used to examine the adhesion state of particles and resins in the fracture surface. It is observed that the adhesion state of NdFeB microparticles and liquid resin is significantly improved as the Fe 3 O 4 nanoparticle content increases, due to the high specific surface area of the nanoparticles. This suggests that the Fe 3 O 4 nanoparticles can enhance the bonding between the microparticles and resin, resulting in improved adhesion and potentially leading to improved mechanical properties. A threedimensional white light interferometer is used to measure the surface roughness of five printed magnetic standard samples ( figure S3). The average surface roughness (S a ) of the NdFeB-Fe 3 O 4 magnetic samples is found to decrease by 6.07% to 21.23% as the content of Fe 3 O 4 nanoparticles increased, in comparison to the NdFeB magnetic sample. Figure S4 shows that the film contact surface of five printed magnetic standard samples with SEM. The aim of this experiments is to compare the settling of magnetic particles in the five resins during the printing process. It was found that an increase in the content of Fe 3 O 4 nanoparticles leads to a significant decrease in the number of magnetic particles on the film contact surface, leading to improved stability of the magnetic photosensitive resin during the printing process. This suggests that the incorporation of Fe 3 O 4 nanoparticles significantly improves the printability of NdFeB-Fe 3 O 4 magnetic photosensitive resins, as evidenced by both theoretical analysis and curing depth experiments.
We further demonstrated and compared the ability of 3D printed magnetic microstructures using NdFeB-Fe 3 O 4 magnetic photosensitive resins with five different ratios. The CAD model and printed samples of hollow cube are shown in figure 2(e). The dimensions of the hollow cube are 5 mm in length, while the radii of the large and small hollow spheres are 0.8 mm and 0.3 mm, respectively. The smallest feature size achieved is 80 µm. All five NdFeB-Fe 3 O 4 magnetic photosensitive resins can be used to print the hollow cubes under the optimal parameters, but there are obvious differences in the feature structure. In particular, the printing quality of the printed hollow cube is improved as Fe 3 O 4 nanoparticles content increased. Furthermore, printed hollow cubes with NdFeB microparticle content exceeded 50% (N1-F0 and N3-F1) have blockages within porous structure. In addition, the compactness of the solid structure part is proportional to the solid load-

Characterization of ferromagnetic capability in 3D prin-
ted magnetic samples. The ferromagnetic capability of 3D printed samples using NdFeB-Fe 3 O 4 magnetic photosensitive resins is evaluated under in both direct current (DC) and alternating current (AC) rotating magnetic fields. Firstly, we design and print a 10 mm long magnetic beam structure magnetized along the direction shown in figure 3(a), and depict its bending deformation under a DC magnetic field in a schematic diagram. The end of this magnetic beam structure is fixed, while the top is capable of being attracted by the DC magnetic field intensity (H DC ), resulting in deformation at the angle. Figure 3(b) illustrates the deformation of the magnetic beam structure with increasing H DC . At H DC = 0, the magnetic beam structure is in a static state. As H DC increases to a value less than the maximum field intensity (H max ), the magnetic beam structure undergoes deformation. Upon reaching H max , the magnetic beam structure reaches its final state and the resulting rotation angle is recorded. Figure 3(c) shows that the deformation angles of the five printed magnetic beam structures are measured with increasing H DC (5, 10, 15, 20, 25, and 30 mT). Experimental results indicate a proportional relationship between the deformation angle and H DC for all magnetic beam structures, with the exception of those printed using Fe 3 O 4 nanoparticle resins (N0-F1). For example, in the magnetic field intensity of 30 mT, the deformation angles of the five printed magnetic beam structures are 8.56 • , 6.79 • , 4.40 • , 3.05 • and 0 • , respectively. The reason is that the magnetism of Fe 3 O 4 nanoparticles is much lower than that of NdFeB microparticles. Therefore, the deformation angle is proportional to the solid loading of NdFeB microparticles in the same H DC .
Furthermore, a magnetic gear with an addendum circle diameter of 5 mm and tooth width of 1.5 mm is designed and magnetized. As shown in figure 3(d), the rotation of a magnetic gear is submerged in water and placed in a variable AC rotating magnetic field. At an AC magnetic field (H AC ) of 2 mT, a digital camera (60 fps) is used to record the rotation of magnetic gear under different magnetic field frequencies (video S1). The angular velocity is then measured and calculated frame by frame. Figure 3(e) shows a photographic illustration of the rotation of a magnetic gear in an alternating current AC magnetic field (H AC = 2 mT) at various periods, in which the U x and U y components alternate with a period of 1/4 T, resulting in counterclockwise rotation of the gear. Figure 3(f) shows that the angular velocity of the five printed magnetic gears is measured with increasing magnetic field frequency (10,20,30,40,50, and 60 Hz). The peak and average angular velocity of each magnetic gear is proportional to the solid loading of NdFeB microparticles. The peak angular velocities of five magnetic gears are 32.45 rad s −1 , 26.98 rad s −1 , 23.39 rad s −1 , 10.13 rad s −1 , and 5.53 rad s −1 , respectively. Meanwhile, the printed samples of five magnetic gears are placed individually to the point of peak angular velocity for each curve. Thus, it is also demonstrated that surface quality of the magnetic gear is inversely to the solid loading of NdFeB microparticles. In addition, a peak frequency f p (corresponding to the peak speed) exists when the magnetic field frequency increases from 10 to 60 Hz. The required magnetic torque is proportional to the solid loading of NdFeB microparticle, leading to a decrease in peak frequency f p with increasing amounts of Fe 3 O 4 nanoparticles. This mechanism is primarily because the rotation motion of the gear has frequency hysteresis [35]. In summary, these results demonstrate that the magnetism of 3D printed samples using NdFeB-Fe 3 O 4 magnetic photosensitive resins are mainly determined by the solid loading of NdFeB microparticle. Moreover, it can be concluded that the effect of the doping of Fe 3 O 4 nanoparticles can be regulated through the selection of an appropriate material ratio.

Comparison of NdFeB-Fe 3 O 4 magnetic photosensitive resins
This section provides a comprehensive comparison of the printability and magnetism of NdFeB-Fe 3 O 4 magnetic photosensitive resins with five ratios based on the above experimental results. Figure 4(a) shows the quantitative indicators of printability and magnetism through radar charts for  Table S1 shows that the measured values and classified levels of five quantitative indicators. According to the quantitative comparison of radar charts, the printability and magnetism of the NdFeB-Fe 3 O 4 magnetic photosensitive resins are the most balanced when the ratio of NdFeB microparticle content and Fe 3 O 4 nanoparticles content is 1:1 (N1-F1). Further, experimental result has demonstrated that NdFeB-Fe 3 O 4 magnetic photosensitive resins (N1-F1) can be printed with a minimum feature size of 40 µm (figure S6).
In order to assess the performance of the NdFeB-Fe 3 O 4 magnetic photosensitive resin, it is necessary to compare with previously published research in the field. This can be achieved by comparing the properties and performance of the resin to those described in the literature cited in the introduction. Figure 4(b) shows the data on the minimum feature size (X-axis) and remanence per unit of solid loading (Y-axis) of the magnetic photosensitive resin. Trends in magnetic photosensitive resins have always demanded high remanence per unit of solid loading (red trend line) and high printing quality (blue trend line). According to the data presented in figure 4(b), the particle types of magnetic photosensitive resins in the existing literature are primarily composed of soft magnetic particles. The minimum feature size of a printable 3D structure with a hard magnetic photosensitive resin (SrFe 12 O 19 ) in the recent literature is 300 µm. The NdFeB-Fe 3 O 4 magnetic photosensitive resin proposed in this article achieved a reduction of 87% in the minimum feature size compared to hard magnetic photosensitive resins, reaching a size of 40 µm or below. Additionally, the remanence of this resin is six times higher and the coercivity is 24 times higher than that observed in our previous work. This work greatly improves the print quality of hard magnetic photosensitive resins while maintaining magnetism and a high solid loading (20 wt.%).

Conclusion
In summary, we developed a novel NdFeB-Fe 3 O 4 magnetic photosensitive resin with high ferromagnetism and great printing quality. This resin can be used to manufacture highprecision and ferromagnetic functional devices via µCLIP process comprised of 20 wt.% solid loading of magnetic particles. Theoretical analysis is found to be in good agreement with experimental result. It is demonstrated that the developed NdFeB-Fe 3 O 4 magnetic photosensitive resin show excellent performance for high precision forming and ferromagnetic capabilities simultaneously compared to both existing hard and soft magnetic photosensitive resins, affected by the solid loading of Fe 3 O 4 and NdFeB. It is also found that the desired resin composition for specific functional requirements can be achieved by adjusting the ratio of NdFeB and Fe 3 O 4 . Specifically, the optimized resin (NdFeB:Fe 3 O 4 = 1:1) is able to print centimeter-size samples but with a sub-40 µm fine feature, reduced by 87% compared to existing hard magnetic photosensitive resin. In addition, this resin shows enhanced coercivity and remanence in comparison with existing soft magnetic photosensitive resins, showing by increase of 24 times and 6 times, respectively. We envision the NdFeB-Fe 3 O 4 magnetic photosensitive resin offer great potentials for the design and manufacture of future anisotropic enhancement devices, micro-robots, electromagnetic sensor, and magnetothermal devices.

Preparation of NdFeB-Fe 3 O 4 magnetic photosensitive resins
Three monomers of different viscosities can be used as the base materials to adjust the viscosities of the magnetic photosensitive resin. The primary monomers being used consisted of CN981 NS (Shkingchem Inc. viscosity of 5000 cps) monomer, which is an aliphatic polyester/polyether based on polyurethane diacrylate oligomer. The secondary monomer was the tripropylene glycol diacrylate (TPGDA) (Curease chemical Inc.) with relatively low viscosity of 12 cps. The tertiary monomer was the poly (ethylene glycol) diacrylate (PEGDA; average Mn 700) (Sigma-Aldrich Inc.) with relatively low viscosity of 55 cps. Additional composition of magnetic photosensitive resins consisted of 2 wt.% Irgacure 819 (BASF Inc.) as photoinitiator and 20 wt.% magnetic particles as ferromagnetic additives. The magnetic particles are mixed by NdFeB microparticles (1-2 µm) and Fe 3 O 4 nanoparticles (10 nm) (Shanghai Naiou Nano technology Inc.) according to the five different ratios. Figure 2 . The ratio of CN981 NS and TPGDA is mixed as 4:1 (w/w) and this mixture is mixed with PEGDA according to the ratio of 5:3 (w/w). CN981 NS and TPGDA are first mixed at 50 • C and 400 rpm by magnetic stirrer for 2 h to get the monomer. Moreover, this monomer (CN981 NS-TPGDA) and PEGDA are mixed at the same mixing parameters. Irgacure 819 is then added into the monomer and mixed in ultrasonic for 2 h for dissolving. NdFeB microparticles and Fe 3 O 4 nanoparticles are manually mixed into the resin for 2 min for primary dispersion, followed by mixed in ultrasonic for 2 h, mechanical stirring for 15 min, and vacuum for 30 min to get the final magnetic photosensitive resin. Especially, the operating temperature in the final mixing should be controlled below 40 • C to avoid thermal polymerization of magnetic particles.

3D printing system
The homemade µCLIP system was illustrated in our previous work [22,24]. It features 405 nm UV light and 1920 × 1080 pixel resolution, providing a building area of 11.6 × 6.5 mm 2 with spatial resolution of 6.04 × 6.04 µm 2 pixel −1 , and linear printing speed in the range from 1 µm s −1 to 100 µm s −1 . All the 3D printed samples were firstly soaked in acetone for 2 min to remove the residual liquid resin, then transferred into ethanol and ultrasonic cleaned for 5 min. After drying the samples in ambient environment for 30 min, they were post-cured in UV light featuring 405 nm wavelength for 10 min.

Magnetic field generation
Magnetic field in the experiments was generated by a threedimensional Helmholtz coil (figure S7). The Helmholtz coil was driven in DC or AC controlled by a DC power supply or a signal generator. AC magnetic field includes two modes of rotating magnetic field and oscillating magnetic field. Angles and distances were measured in the recorded figures using the open-source image processing program ImageJ (National Institutes of Health, NIH).