1. Introduction
The fabrication of complex three-dimensional (3D) micro/nano-structures is fundamental to numerous cutting-edge fields. In particular, with the development of advanced domains such as metamaterials, micro/nano-optics, microelectronics, and biomedical engineering, the limitations of existing precision manufacturing technologies have become increasingly evident. This has led to a growing demand for ultrahigh-precision micro- and nano-additive manufacturing technologies.
Currently, numerous additive manufacturing techniques are available, including electron-beam-induced deposition, selective laser sintering, electrohydrodynamic printing (oxidation–reduction printing), direct ink writing, fused filament fabrication, inkjet 3D printing, and photopolymerization-based 3D printing. Photopolymerization-based 3D printing technology involves light-induced polymerization triggered by the interaction between light and the material, enabling the manufacture of complex 3D structures through selective light exposure. This process entails the absorption of photons by initiator molecules in the photosensitizer, resulting in the generation of free radicals or cations that catalyze polymerization reactions with monomers or oligomers. This technique has attracted significant attention due to its high flexibility and resolution.
Recent advancements in techniques such as continuous liquid interface printing (CLIP)
[1], computational axial lithography (CAL)
[2], multi-photon lithography (MPL)
[3],
[4],
[5],
[6],
[7], “xolography” volumetric 3D printing
[8], and two-step absorption lithography
[9],
[10] have notably improved the speed and precision of photopolymerization-based 3D printing. Among these, MPL is based on nonlinear absorption, where the absorption rate correlates directly with the
nth power of the light intensity (where
n is the number of absorbed photons). This imparts remarkable spatial selectivity and superior lateral and axial resolutions. Two-photon polymerization (TPP) is the most common and well-developed MPL technology based on two-photon absorption. In practice, higher-order multi-photon absorption may also occur during TPP processes
[11],
[12]. In 2001, Kawata et al.
[13] printed nanobull structures with a resolution of 120 nm, signifying the introduction of MPL into nano-scale printing. Subsequently, researchers have continually enhanced the printing resolution and speed, developing novel photoresist materials
[14],
[15],
[16] and light field engineering techniques
[6],
[17],
[18],
[19],
[20], such as spatial and temporal focusing and 4π focusing technology
[21],
[22],
[23],
[24],
[25]. Furthermore, studies on photoinhibition lithography (PIL)
[26],
[27],
[28],
[29],
[30], including stimulated emission depletion (STED) lithography derived from STED microscopy
[31],
[32], have incorporated both field control and novel photosensitive materials. Over the past two decades, MPL has been used for a wide range of applications. For instance, in optics, MPL can overcome the limitations of traditional manufacturing methods by enabling the printing of optical lenses with sub-nanometer features, offering extensive design flexibility for applications in imaging, illumination, and beam shaping
[33],
[34],
[35],
[36],
[37],
[38],
[39],
[40],
[41],
[42],
[43],
[44],
[45],
[46],
[47],
[48],
[49],
[50]. In the biomedical field, this technique is employed in cell growth, differentiation, adhesion, and migration
[51],
[52],
[53],
[54],
[55],
[56],
[57],
[58]. It also facilitates the rapid production of drug delivery devices by manufacturing templates
[59],
[60]. Additionally, with the development of new intelligent responsive materials, metals and their compounds, and composite materials, MPL has extensive applications in micro/nano-electronics (sensors
[5],
[61],
[62],
[63],
[64],
[65], memristors, and capacitors, etc.) and micromechanics (motion control
[66],
[67],
[68] and deformable micromechanics
[69],
[70],
[71],
[72],
[73],
[74],
[75],
[76], etc.).
Several companies have introduced commercial printing systems based on MPL, such as Photonic Professional GT2 (Nanoscribe GmbH, Germany), MPO100 (Heidelberg Instruments, Germany), MicroFAB3D (Microlight3D, France), and NanoOne (Upnano, Austria). These printing systems achieve nano-scale feature sizes and resolutions while maintaining a volumetric yield of about 0.04 mm
3·h
−1, which caters primarily to laboratory requirements. However, current MPL technology still struggles to meet the demands of large-scale industrial production and applications. This paper discusses the latest research progress in processing speed and material systems, and summarizes the current challenges and development trends of multi-photon 3D nanoprinting for large-scale industrial production and application. Overview of advances and challenges of multi-photon 3D nanoprinting is shown in
Fig.1.
2. Processing speed and throughput
Insufficient processing speed has persistently been a key factor limiting the large-scale production and application of MPL. Due to their nonlinear absorption characteristics, femtosecond lasers are typically preferred as light sources, resulting in higher equipment costs compared to other 3D printing devices. Moreover, the low printing throughput exacerbates the cost issue, making it challenging to achieve industrial-scale production. Consequently, the pivotal issue is enhancing the processing speed of MPL for large-scale production.
2.1. Projection-based processing technology
Projection-based processing technology spatially shapes lasers using projection devices, projecting the corresponding patterns of a sliced model onto the interior of a photoresist. Integration with a 3D stage enables projection-based processing
[6],
[77],
[78],
[79]. Alternatively, it entails exposing the complete volume elements within the photoresist, thereby enabling 3D volumetric fabrication
[80],
[81]. Common projection techniques include mask-based and holographic-based projections.
2.1.1. Mask-based projection technology
Mask-based projection technology commonly involves digital mask generation systems such as spatial light modulators (SLMs) and digital micromirror devices (DMDs). Through dynamic digital masks, these systems modulate the optical field, enabling simultaneous exposure of entire model slice patterns and facilitating a layer-by-layer printing approach. In 2019, Saha et al.
[6] proposed a projection MPL based on spatial and temporal focusing, as illustrated in
Fig. 2(a). This system directs femtosecond laser beams onto the DMD, where the exposure pattern masks and diffracts the femtosecond laser. A light sheet with angular dispersion is then projected onto the photoresist, achieving spatial and temporal focusing at the focal plane behind the objective and enabling single-layer pattern exposure. By combining the high-speed refresh rate of the DMD with
z-axis stage movement for layer scanning, the processing speed is significantly enhanced. The volumetric yield of this system can reach 8.7 mm
3·h
−1, with lateral and axial resolution of 151 nm and 1–4 μm, respectively. In 2021, Somers et al.
[79] developed a similar MPL system capable of printing with lateral and axial resolution of 380 nm and 2 μm, respectively, resulting in an overall volumetric yield of 3.88 mm
3·h
−1. In the same year, Liu et al.
[78] developed a two-dimensional (2D) MPL system employing a DMD, achieving a minimum linewidth of 32 nm, as shown in
Fig. 2(b). Nevertheless, due to the absence of a function for performing optical slicing in the axial direction, this system is restricted to processing 2D patterns exclusively. The throughput of the system is 1.1 mm
2·min
−1.
2.1.2. Holographic-based projection technology
Holographic-based projection technology commonly involves the generation of holograms using binary holography algorithms to project modulated light fields onto a photoresist for layer or volumetric exposure. For layer projection, Yang et al.
[77] in 2019 developed an
x–
z directional layer projection MPL system based on the Gerchberg–Saxton algorithm utilizing SLM. This system exhibited a scanning speed of 6 μm·s
−1 in the
y direction, allowing single-exposure multi-photon polymerization (MPP) for small-volume structures, as illustrated in
Fig. 3(a). Nevertheless, due to limitations in laser power and SLM refresh rates, this system requires an extended time to expose large-area structures and lacks high-speed scanning capability in the
y-axis direction.
Additionally, researchers have projected specialized 3D optical fields using methods such as complex amplitude modulation. Several studies have achieved the single-exposure fabrication of specific feature structures
[80],
[81],
[82],
[83],
[84],
[85],
[86]. In 2009, Bautista et al.
[81] introduced helical-phase elements into a laser using an SLM, projecting holographic optical vortices within the photoresist, as depicted in
Fig. 3(b). By manipulating the topological charge and maximum phase level of the optical vortices, they controlled the shape of the microgears, ultimately enabling one-step fabrication of arbitrary microgear structures through MPL. Similarly, in 2014, Zhang et al.
[80] employed SLM-based phase modulation to generate optical vortex beams with orbital angular momentum (OAM). These beams exhibit internal structures instead of elliptical intensity distributions upon focusing. Utilizing superimposed Gaussian–Laguerre beams and 2D phase modulation, they achieved intricate 3D focusing, eventually enabling the one-step fabrication of 3D hollow double-helix structures. The 3D objects printed using the system are shown in
Fig. 3(c). In 2023, Wang et al.
[87] modulated a Gaussian beam into an Airy beam by loading a predesigned computer-generated hologram (CGH) onto an SLM. A 3D curved micropillar was fabricated via a single exposure to a focused Airy beam, as depicted in
Fig. 3(d). This technique allows the single-exposure fabrication of characteristic structures. While these technologies offer relatively high processing speeds, they lack the ability to fabricate arbitrary 3D structures, thus relinquishing the inherent flexibility and freedom of additive manufacturing.
Projection-based processing techniques exhibit exceedingly high processing speeds. However, due to the proximity effect, this technology faces difficulties in generating a uniformly distributed light field along the vertical z-axis when exposing intricate patterns. Consequently, this technology generally suffers from relatively low vertical resolution, making it challenging to fabricate high-precision 3D structures of arbitrary shapes. This is one of the prevailing challenges in current MPL, presenting a conflict between processing accuracy and efficiency.
2.2. Point scanning-based processing technology
Point-scanning-based processing technologies can be categorized into two main types based on their scanning paths: raster and random-access scanning. Raster scanning commonly involves the use of mechanical scanning devices such as a galvanoscanner and a translational stage for point scanning. In contrast, random-access scanning employs diffractive optical components, such as SLM and DMD, combined with holographic algorithms to generate holograms for multifocus scanning
[88],
[89],
[90],
[91]. In this method, the position of the laser focus is determined by the hologram of the diffractive device rather than by mechanical scanning devices. This allows for random scanning paths that can increase the duty factor, particularly when processing hollow structures such as outlines or frameworks
[92],
[93]. Compared to projection-based processing technology, point-scanning-based processing technology generally has higher resolution but slower processing speed. The scanning speed of the system is predominantly influenced by the light-scanning devices, including the mechanical inertia, refresh rate of diffractive elements, and diffractive efficiency. Consequently, since the inception of MPL, numerous researchers have explored methods to enhance the scanning speed.
2.2.1. Random-access scanning processing technology
Enhancing processing speed commonly involves augmenting parallel processing efficiency by increasing the number of foci. Additionally, because the scanning path of the foci is based on holograms, the scanning speed of random-access scanning technology is highly correlated with the refresh rate of the diffractive components. Common random-access scanning techniques include the following
(1) SLM-based random-access scanning. In 2014, Vizsnyiczai et al.
[88] devised an MPL system that employed an SLM to generate multifocus random-access scans. The system directed the femtosecond laser onto the SLM, implementing a weighted Gerchberg–Saxton (GSW) algorithm to generate multiple foci to follow distinct trajectories for parallel processing, as depicted in
Fig. 4(a). The voxel size of the system was set to 150 nm. However, the single-point scanning speed is only 9 μm·s
−1 due to the SLM’s 60 Hz refresh rate. It should be noted that the random-access scanning method can achieve not only focal point scanning but also projected pattern scanning. In 2017, Yang et al.
[89] investigated the focusing characteristics of superposed Bessel beams (SBBs) under high-numerical aperture (NA) objectives, combined with the Debye vector diffraction theory, to design a multifocus MPL system based on SLM, as illustrated in
Fig. 4(b). This system rapidly produces holograms of multifocus or annular light spots using SBBs to fabricate 3D structures. However, due to the characteristics of SBBs, this system is restricted to processing particular shapes and is incapable of achieving arbitrary 3D structures.
(2) DMD-based random-access scanning. The refresh rate of commonly used SLMs is only 60 Hz, which significantly limits the enhancement of processing speed. In contrast, the DMD has an impressive refresh rate of up to dozens of kilohertz, making it suitable for rapid projections. In 2019, Geng et al.
[90] developed an MPL system based on DMD to generate binary holograms, achieving a scanning speed of 5 mm·s
−1 for a single focus. However, limited by a 3% energy utilization rate, the system has four foci with a yield of 0.04 mm
3·h
−1. In 2023, Ouyang et al.
[94] developed an MPL system with 2000 individually programmable laser foci using a similar approach, as depicted in
Fig. 5. However, with an increase in the number of foci, there may be a decline in processing resolution due to the influence of the holographic algorithm.
2.2.2. Raster scanning processing technology
Raster scanning processing technology relies on the periodic scanning of the deflector to accomplish point scanning. Consequently, researchers have focused on enhancing the performance of scanning devices to improve processing speeds. Common scanning devices include the following:
(1) Galvanoscanner-based processing technology. The galvanoscanner is a common scanning device known for its simple structure and cost-effectiveness. It is employed not only for raster scanning but also for random-access scanning, finding wide applications in the field of MPL. In addition to enhancing the scanning speed of the galvanoscanner, researchers frequently integrate various beam-splitting devices to generate multifocusing for parallel scanning. Examples include microlens arrays
[95], diffractive beam splitters (DBSs)
[96], SLM
[97], and diffractive optical elements (DOEs)
[42], among others. In 2020, Hahn et al.
[42] developed a rapid multifocus MPL system by combining a DOE with a high-speed galvanoscanner, as depicted in
Fig. 6. They used a Photonic Professional GT (Nanoscribe, Germany) to fabricate a DOE, generating a 3 × 3 array of periodic uniformly diffracting beams. This setup achieved a periodic array scan of 9 foci, with a minimum feature size of 406 nm laterally and 1.01 µm axially. The scanning speed of each focus reached 400 mm·s
−1, resulting in a throughput of 9 × 10
6 voxels·s
−1, which was the highest throughput for a galvanoscanner-based MPL. They printed a cross-scale metamaterial with a volume of 2.4 mm × 2.4 mm × 9.6 mm, containing the largest number of units available at present (108 000 3D unit cells). Their work significantly advanced the development of cross-scale MPL. As we were writing this review, Kiefer et al.
[98] used a DOE to generate a 7 × 7 beam array, increasing the throughput to 7.1 × 10
7 voxels·s
−1. Notably, Zhang et al
. [99] utilized a liquid crystal on silicon-SLM (LCoS-SLM) to generate multi-spot holograms for parallel multifocus processing, achieving up to 448 foci and a printing throughput of 1.49 × 10
8 voxels·s
−1. Nonetheless, the mechanical inertia of galvanometric mirrors, combined with mirror distortion during high-speed scanning, makes it challenging to achieve higher-speed scanning with acceptable optical aberration. Additionally, because the DOE affects the uniformity of the laser beam, the processing resolution is reduced.
(2)Resonant-scanning processing technology. Resonant scanners operate reflective mirrors driven at resonant frequencies, allowing for higher scanning speeds than galvanoscanners. In 2019, Pearre et al.
[100] used resonant scanners to develop a high-speed single-point MPL system, as illustrated in
Fig. 7. The scanning frequency of the resonant scanner was 8 kHz (compared to approximately 1.5 kHz for a regular galvanoscanner), resulting in a scan speed of 3300 mm·s
−1 for the system. The minimum feature size was 1900 nm (average) in the
x direction, 350 nm in the
y direction, and 450 nm in the
z direction. However, because of the non-constant scanning speed of resonant scanners, which is distributed sinusoidally across the scan angles, the resolution of the system is not uniform. Additionally, owing to the limited switching frequency of the Pockels cell (3.33 MHz), the minimum feature size along the
x direction at 1900 nm was far from the diffraction limit.
(3) Polygon-scanner-based processing technology. Unlike the reciprocating movement of the galvanoscanner, the polygon scanner scans by rotating unidirectionally at a constant speed, as shown in
Fig. 8. This process does not involve direct dynamic motion, allowing the scanning frequency to exceed 30 kHz. In 2022, Wang et al.
[101] developed a high-speed parallel MPL system employing a polygonal scanner, an SLM, and a multichannel acoustic optical modulator (AOMC), as depicted in
Fig. 8(b). The polygonal scanner enables a writing speed of 7.77 m·s
−1 per channel. They utilized an SLM and AOMC to achieve multipoint splitting and independent control of each channel, thus reaching a system speed of 46 m·s
−1. However, limited by the switching frequency of the AOMC, the resolution of the technique in the scanning direction is lower than the feature size (150 nm) and cannot exceed the diffraction limit. Additionally, due to manufacturing defects, the polygonal mirrors have angular errors ranging from 25 to 150 μrad, leading to pyramid errors during scanning that cause scan line deviations and diminish the processing resolution.
(4) Acousto-optical scanning processing technology. By combining existing MPL methods, projection-based processing exhibits excellent processing speed but relatively low resolution. In contrast, raster scanning technology provides more precise resolution but at a slower processing speed, making large-scale production challenging. Even when the scanning frequency is increased, the inertia and mechanical distortion of the scanning device limit both processing speed and accuracy. Notably, in 2023, Jiao et al.
[102] proposed the use of acousto-optic deflectors (AOD) instead of traditional inertial mechanical scanning devices, establishing an inertial-free acousto-optic scanning spatial-switching (AOSS) MPL, as shown in
Fig. 9. The AOD modulates the diffracted wavefront of the laser through the acousto-optic effect of the crystal to enable laser scanning, as illustrated in
Fig. 9(b). Thus, the AOD does not face inertial limitations, enabling higher scanning frequencies with a scan speed of 3.57 m·s
−1 per point. Additionally, as shown in
Fig. 9(a), they integrated DOE to achieve parallel processing with eight focal points, further enhancing the processing speed. By leveraging the advantages of projection-based processing, they used a DMD as a spatial optical switch to control multifocus switching individually, as depicted in
Fig. 9(c). While achieving improved processing speed, it can print arbitrary non-periodic structures. This method, which combines the merits of both raster scanning and projection-based processing, achieves a voxel size of 212 nm while achieving a record-high throughput of 7.6 × 10
7 voxels·s
−1, nearly 8.4 times faster than previously reported mechanical scanning methods
[42] and nearly equivalent to the processing speed of projection-based technology. Moreover, the scanning angular speed
ω of AOSS is related to the AOD bandwidth Δ
f, which is represented by
, where
represents the wave length of laser. Therefore, the printing throughput of the AOSS method can be further increased by enhancing the AOD bandwidth, thereby offering a viable technical route for large-scale MPL 3D nanoprinting.
Finally, we summarize the existing high-speed MPL printing methods according to their processing modes, resolution, and throughput, as shown in
Table 1 [6],
[42],
[77],
[78],
[94],
[97],
[100],
[102],
[103].
3. Materials
Expanding the range of photosensitive materials significantly broadens the applications of multi-photon 3D nanoprinting, with material formability and functionality being critical factors in determining its practical applications in manufacturing. Many resin materials used in MPL, such as the widely utilized organic polymer systems, are derived from traditional ultraviolet (UV) lithography or stereolithography. Although functional materials such as metals and their compounds, ceramics, glass, and composites can now be processed through MPL, challenges persist in their formability and performance. Additionally, the processing strategies for these materials, particularly post-processing strategies, are crucial for enhancing their formability and performance. Therefore, exploring new formation mechanisms and developing processing strategies to enhance the performance and formability of materials are key research focuses for advancing multi-photon 3D nanoprinting.
3.1. Organic polymer
Organic polymer materials have the advantage of low cost and can be developed from material systems used in classical UV lithography or stereolithography. Common photopolymerization mechanisms include free-radical and cationic polymerization. In radical photopolymerization, reactive species are generated as free radicals by the homolytic cleavage of a bond in the excited state of the photoinitiator, with monomers mainly being acrylates and methacrylates. In cationic photopolymerization, diazonium, onium, iodonium, sulfonium, or pyrylium salts are used to generate charged species that initiate the polymerization of specific monomers, usually from the epoxy resin family
[104]. Organic polymers are largely transparent at both visible and near-infrared wavelengths. Despite potential issues such as shrinkage, these materials exhibit good mechanical, optical, and biological properties. They have been widely utilized in various fields, including microfluidics, micro-photonics, biomedical applications, microelectromechanical systems (MEMS), and micro-robotics.
Researchers have adopted additional strategies to address issues such as shrinkage. For example, thermal baking and UV exposure during post-processing are used to increase the degree of crosslinking, and supercritical drying is employed to reduce liquid surface tension.
3.1.1. Acrylates
Organic polymers were the earliest and most widely used materials for MPL. The initial materials employed in MPL are acrylates
[105], which are known for their stable performance. They have been extensively used as substrates to explore the interactions between microenvironments and biology. For example, Pennacchio et al.
[106] used TPP to directly write 3D cage-like structures for cell scaffolds. Otuka et al.
[107] utilized TPP to create microstructural arrays for cultivating
Komagataeibacter xylinus bacteria by simulating a natural environment. However, due to the relatively singular functionality of acrylate materials, designing and processing 3D shapes often require additional modifications to the structural surface to achieve specific functional applications. For example, Berwind et al.
[108] used plasma polymerization of hexamethyldisiloxane (HMDSO) to coat substrates fabricated from acrylic photopolymers, resulting in wetting states ranging from slightly hydrophilic to superhydrophobic. Additionally, researchers have often applied metallic coatings to the surfaces of acrylate-printed structures to achieve magnetic properties and enhance biocompatibility. This approach has extensive applications in micro- and nano-robotics
[76],
[109],
[110],
[111].
3.1.2. Epoxy
Another class of commercially utilized photoresists in MPL processing is SU-8 (Kayaku Advanced Materials, Inc., USA)
[112],
[113], which is based on epoxy groups and can be polymerized using cations. SU-8, with its eight epoxy groups, allows for a spin-coating thickness ranging from less than 1 micron to over 300 microns, making it suitable for processing high aspect ratio structures. In contrast to most MPL processing materials, SU-8 requires spin-coating and prebaking steps to remove unwanted solvents. Laser irradiation generates hexafluoroantimonic acid, which activates the resin monomers. After post-baking, the epoxy groups begin to crosslink, resulting in higher mechanical strength
[114]. This material also exhibits good thermal stability
[115] and a relatively high refractive index (1.57)
[116], making it widely used in various micro/nano-functional devices. For example, Williams et al.
[117] developed a melt reflow process to print SU-8 optical devices, including plano-convex lenses, cylindrical lenses, composite lens systems, and chiral photonic crystal structures at the optical fiber end face, as shown in
Fig. 10(a). Li et al.
[111] used SU-8 to manufacture mico “rocket” robots, as illustrated in
Fig. 10(b). They processed the structure of the robot using MPL, followed by sputtering a layer of gold metal film on its surface. The robot exhibited thermal motion when stimulated with an 808 nm infrared laser.
3.1.3. Hydrogels and protein
Hydrogels and protein materials have rapidly emerged as a new category of MPL materials because of their flexibility, stimulus responsiveness, and biocompatibility. Many proteins and hydrogels exist in nature, and the modification of these natural materials can produce hydrogels better suited for TPP, such as bovine serum albumin (BSA)
[118],
[119], fibrin protein
[120], gelatin
[121],
[122], chitosan
[123], and polydopamine
[124]. These materials not only retain biodegradability but also improve affinity through surface modification, making them more suitable for applications such as cell scaffolds in biomedical tissue cultivation. Other classes of synthetically engineered hydrogels, such as humidity-responsive poly (ethylene glycol) diacrylate (PEGDA) and temperature-responsive poly(
N-isopropylacrylamide) (PNIPAM), have gained considerable attention in recent years due to their stimulus-responsive properties. Control over the local material mechanical properties and structurally predefined deformation can be achieved through preprogrammed laser-writing parameters
[74],
[75] (e.g., scanning speed and laser power) or the design of asymmetric microscale mechanical metamaterial unit cells
[125], as shown in
Figs. 10(c) and
(d). However, most intelligent micromachines are composed entirely of polymers, which limits their functionality. Although MPL technology is capable of heterogeneous integration, it requires complex post-processing steps. The crosslinking density and strength of the material are crucial for structure formation. Insufficient strength during formation can cause the structure to collapse or lose integrity during development. This often necessitates increasing the laser power or enlarging the structure size for compensation, which in turn reduces the processing resolution. Many hydrogels lack high strength, which can hinder their ability to achieve finer processing. With the enhancement of heterogeneous integration capabilities, more practical intelligent micromachines are expected to be fabricated.
3.2. Inorganic compound materials
Inorganic compound materials exhibit optical, electrical, mechanical, and thermal characteristics that are distinct from those of organic materials, making them indispensable in the field of micro/nano-fabrication. Most inorganic materials, including metals, metallic compounds, ceramics, and glass, often adopt the structure of organic materials as a frame. In addition, annealing is typically used for post-processing, which can affect the purity, conductivity, and formability of the processed materials. After annealing, as the organic components are removed, the structure may shrink or the feature size may decrease. Researchers have developed various methods to enhance formability and functionality based on different material types and requirements.
3.2.1. Metals and metallic compounds
With the continuous advancement of MPL processing, metals and metallic compounds, as an important category of materials, have been subjected to femtosecond laser direct writing through various methods, primarily categorized into four approaches: direct reduction of metal ions
[126],
[127],
[128], doping of metal nanoparticles (NPs)
[4],
[129], template adsorption
[130],
[131],
[132],
[133], and metal-ion resin
[134],
[135],
[136], as illustrated in
Fig. 11. The direct reduction of metal ions involves laser irradiation of a metal-ion solution, causing a reduction reaction of metal ions by absorbing multiple photon energies. This method has been employed to fabricate patterned micro/nano-structures of gold
[128], platinum
[126], and palladium
[126], as shown in
Fig. 11(a). For example, Long et al.
[137] and Xu et al.
[138] achieved the rapid
in situ synthesis and patterning of edge-unsaturated MoS
2 using this approach to prepare highly sensitive gas sensors. Although this method is a one-step process and possesses high metal purity, it is challenging to 3D-form and extend it to elements with stronger metal activity. The doping of metal nanomaterials involves the directional assembly of metal nanowires through multi-photon absorption. This method was used to fabricate the 3D micro/nano-structures of Ag
[129] and ZnO
[4], as shown in
Fig. 11(b). However, the doping of metals and their compounds can lead to light scattering and micro-explosions, resulting in poor resolution and processing quality. Template adsorption method uses resins to manufacture polymer templates through MPL, which are immersed in a metal-ion solution for adsorption, followed by annealing to remove the organic template
[131],
[132],
[133]. Saccone et al.
[133] implemented the template adsorption method using digital light processing as depicted in
Fig. 11(c), which was subsequently adapted for MPL
[139]. This method is applicable for fabrication of various metal elements, but due to the low content of metal elements, severe shrinkage occurs after annealing, making it difficult to achieve heterogeneous integration. Metal-ion resin methods involve dissolving metal salt solutions in a polymeric resin for direct writing and annealing. For example, Vyatskikh et al.
[134],
[136] and Liu et. al
[62] utilized an exchange reaction between acrylic acid and alcohol salts to process 3D micro/nano-structures of nickel, titanium, and zinc dioxide, as shown in
Fig. 11(d). Each of these methods has its strengths and weaknesses but collectively expands the material range for MPP processing. However, the processing of metal materials often requires a higher laser power. Therefore, the problem of insufficient power in high-throughput MPL can become severe.
3.2.2. Glass and ceramic
Glass and ceramic materials are among the most crucial materials in modern engineering applications owing to their excellent optical transparency and high thermal, chemical, and mechanical resilience. For glass materials, traditional fabrication methods often rely on particle-loaded binders and sintering methods. High-temperature sintering is required to remove the binder and fuse the silica particles into a solid structure, making it unsuitable for preparing low-melting-point materials. Additionally, particle loading can affect the direct writing process; excessive particles may cause laser scattering and micro-explosions, whereas insufficient loading can lead to significant annealing shrinkage, limiting their application in microsystem technology
[140],
[141],
[142],
[143]. Bauer et al.
[140] introduced a low-temperature 3D printing route for binder-free structures using a hybrid organic–inorganic polymer resin as the raw material, as shown in
Fig. 12(a). This approach enables the processing of complex silica nanophotonic devices in the range of hundreds of micrometers, offering a novel method for the direct writing of glass materials. For ceramic materials, scattering and annealing shrinkage pose significant challenges for direct writing. To address these issues, Fraunhofer-Institut für Silicatforschung, Germany, introduced a class of ORMOCER organic modified ceramics
[144]. This material, which has strong covalent bonds between the ceramic and polymer components, prevents separation into different phases, thereby providing outstanding chemical and thermal stability. This allows the manufacture of high-resolution 3D structures with enhanced mechanical properties and tunable optical characteristics for photonic applications. SZ2080
TM (IESL-FORTH, Greece), a common material prepared by sol–gel and organic–inorganic hybrid technology, has a low-shrinkage rate, is compatible with calcination technology, and its refractive index can be adjusted by changing the inorganic content of the sol–gel
[145],
[146],
[147],
[148],
[149]. Additionally, Desponds et al.
[150] proposed a method based on a composite resin combining ultra-small (5 nm) zirconium oxide-stabilized NPs with a zirconium acrylate precursor, as illustrated in
Fig. 12(b). This method utilizes nanoparticles as both crystal seeds and structural stabilizers, significantly reducing light scattering and structural shrinkage during processing.
3.3. Composites
Micro/nano-structures fabricated using organic polymers often possess limited functionality, primarily retaining only their shape. To address this limitation, incorporating organic/inorganic functional nanomaterials uniformly dispersed within a photocurable organic matrix offers the potential to enhance specific properties or introduce new characteristics into composite materials. It is worth noting that NPs in composite materials can cause laser scattering. Combined with the inherent size of the particles, this may lead to a decrease in processing resolution. In this section, we retrospectively examine advancements in composite materials for MPL, focusing on functional device aspects, including mechanical, conductive, optical, and magnetically driven properties. We present methods for uniformly dispersing nanoparticles into an organic substrate and enhancing the functionality of the materials.
3.3.1. Mechanically reinforced composites
Generally, soft organic materials are used for multi-photon 3D nanoprinting. The mechanical properties of these structures can be improved by introducing nanocomposites, which is beneficial for enhancing the complexity and stability of printed structures and alleviating structural deformation caused by capillary forces and environmental disturbances during the development and drying processes
[151]. This approach also enhances the complexity and stability of printed structures. Some research teams have demonstrated that incorporating nanofiber materials into a polymer matrix significantly enhances the mechanical performance of the objects produced
[152],
[153],
[154]. For example, Xiong et al.
[154] utilized thiol-modified acid-purified multiwalled carbon nanotubes (MWNTs) to prepare an MWNT-thiol-acrylate (MTA) composite resin and used MPL technology for printing. This approach significantly improved the mechanical performance of the polymer structure, and the volume shrinkage of the structure gradually decreased with increasing MWNT concentrations.
3.3.2. Electrically conductive reinforced composites
The construction of micro-nano 3D functional devices based on MPL requires compatible, high-performance conductive composites. The introduction of organic/inorganic conductive nanofillers can impart conductivity to insulating photoresins.
Table 2 [129],
[154],
[155],
[156],
[157],
[158],
[159],
[160] lists representative studies on MPL-conductive composite materials
[129],
[154],
[155],
[156],
[157],
[158],
[159],
[160]. Generally, the nanofiller concentration must reach a certain percolation threshold to significantly enhance the conductivity of the polymer matrix
[161]. Materials such as carbon nanotubes, metal nanowires, and long-chain conductive polymers can significantly increase conductivity at lower dispersion concentrations. Staudinger et al.
[156] achieved approximately 9.1 S·m
−1 conductivity in a Femtobond 4B commercial photoresin (Laser nanoFab, Germany) with 0.05 wt% single-walled carbon nanotubes (SWNTs). But dispersion by ultrasonic struggles to uniformly disperse carbon nanotube (CNTs) at high concentrations, leading to structural deformation and lower conductivity
[156]. Xiong et al.
[154] used thiol modification to achieve a uniform and stable dispersion of MWNTs up to 0.2 wt%, processing 3D micro/nano-structures using MPL, as shown in
Figs. 13(a)–(f). The optimal conductivity reached 46.8 S·m
−1. Due to the oriented arrangement of the MWNTs induced by direct laser writing during photopolymerization, the conductivity in the parallel scanning mode was three orders of magnitude higher than that in the vertical scanning mode. Intrinsic conductive polymer (ICP) materials face challenges in terms of effective dispersion and strong absorption in the near-infrared range, making direct dispersion into acrylate materials for MPL challenging. To overcome this problem, Dadras-Toussi et al.
[157] successfully dispersed poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) material in acrylate systems through interactions with dimethyl sulfoxide (DMSO). They achieved a high conductivity of up to 2.7 × 10
4 S·m
−1, as depicted in
Figs. 13(g)–(m).
3.4. Optical reinforced composites
Printing 3D micro/nano-structures with high refractive indices and customization capabilities has diverse applications in photonics. However, it’s challenging for the refractive index of organic polymer materials to exceed 1.7, which seriously limits their applicability in photonics
[162]. By appropriately introducing nanocomposite materials, it is possible to increase the refractive index of a product while maintaining its high transparency
[163],
[164], thereby expanding its range of applications. Additionally, the introduction of composite materials can confer certain fluorescence capabilities to MPL products. Sun et al.
[165,166] utilized methyl acrylic acid cadmium as the CdS precursor, achieving size control of
in situ synthesized CdS NPs by controlling the crosslinking density. This approach resulted in fluorescence effects of different colors, as illustrated in
Figs. 14(a) and
(b). Peng et al.
[167] incorporated three different quantum dots into a pentaerythritol triacrylate (PETA) resin to achieve red, green, and blue light-induced luminescence, as depicted in
Figs. 14(c)–(e).
3.5. Magnetic reinforced composites
Magnetic actuation provides non-contact control of materials in most environments, excluding a few ferromagnetic settings, and has extensive application potential in MEMS, micro-robotics, and other fields. By dispersing magnetic nanoparticles into a photoresist system to construct a TPP-compatible organic–inorganic composite photosensitive resin, 3D micro-nano structures that are magnetically driven can be fabricated using TPP. Xia et al.
[130], Wang et al
[168], and Tian et al.
[169] dispersed surface-modified Fe
3O
4 NPs into an acrylate photoresist, enabling the MPL fabrication of remotely controllable microsprings and micro-turbine structures, as shown in
Figs. 14(f) and
(g). Ceylan et al.
[122] dispersed iron oxide nanoparticles into gelatin methacryloyl (GelMA), using MPL to produce biodegradable microrobots. Through structural design and the utilization of a rotating magnetic field, efficient motion of the microrobots was achieved. These microrobots can swell and biodegrade in the presence of matrix metalloproteinase 2 (MMP2), leaving no toxic residues after degradation.
In composites, the organic components facilitate the control of the architecture, mechanical properties, and porosity of the material system. The inorganic components impart thermal and mechanical stability to the material and alter the refractive index of the hybrid material. Although significant progress has been made in research on MPL-compatible composite materials in terms of mechanics, conductivity, optics, and magnetic actuation, challenges remain in achieving a highly concentrated, uniform dispersion of dopant materials while avoiding issues such as light scattering and structural deformation caused by micro-explosions during the MPL process.
3.6. Heterogeneous integration
The above-mentioned materials used for multi-photon 3D nanoprinting provide organic polymers with enhanced functionalities. However, practical processes often require materials with complex and diverse properties, and a single material cannot satisfy all these requirements. The use of multiple materials allows for the integration of additional functionalities into 3D micro/nano-structures
[38],
[170]. For example, Hippler et al.
[171] utilized trimethylolpropane ethoxylate triacrylate (TPETA), a protein inhibitor, as the wall of a host–guest hydrogel. They incorporated PETA, another protein, at the top of the walls to create protein-adhesive arms, thereby constructing a composite cell scaffold, as shown in
Fig. 15(a). Researchers introduced 1-adamantanecarboxylic acid (1-AdCA) molecules into water to expand the host–guest hydrogel, pushing the walls and stretching cells to study cellular responses to mechanical stimuli. Ma et al.
[172] developed a spider-like microrobot using SU-8 as a rigid skeleton and BSA as a flexible muscle. The bending and stretching of the spider legs were controlled by changing the pH of the aqueous solution. By combining materials with different properties, they created a stimulus-responsive microgripper capable of capturing, controlling, transporting, and releasing micro-objects, as shown in
Fig. 15(b). Zeng et al.
[69] developed the first liquid-crystal elastomer as a muscle for a micro/nano-scale walker, demonstrating various movements such as walking and rotation. They used light-responsive liquid-crystal materials and a commercial photopolymer IP-Dip (Nanoscribe, Germany), paving the way for the manufacture of micro/nano-robots with liquid-crystal elastomers, as shown in
Fig. 15(c).
Currently, sequential printing is commonly used for the heterogeneous integration of MPL. For instance, Klein et al.
[173] printed a structure composed of PEGDA and PETA, then added a biocompatible photoresist, Ormocomp (micro resist technology, Germany), by dropping it onto a previously printed scaffold to create a secondary structure.
However, this printing method requires extremely high positioning accuracy and is cumbersome when processing multiple materials. To address these challenges, Mayer et al.
[174] employed microfluidic channel technology, as illustrated in
Fig. 15(d). After printing with one photoresist, the development liquid was injected with nitrogen, and following
in situ development, another fluorescent photoresist was injected. This approach enabled printing with five different materials without the need for repeated positioning. However, this method introduces complexity into the setup when switching frequently between various viscous photoresists.
4. Challenges and perspectives
Multi-photon 3D nanoprinting technologies generally face a reciprocal constraint between processing precision and printing efficiency.
Fig. 16 [6],
[42],
[79],
[94],
[95],
[96],
[97],
[98],
[99],
[102],
[175],
[176] illustrates the performance of different multi-photon 3D nanoprinting methods in terms of precision and throughput developed in recent years
[42]. Each multi-photon 3D nanoprinting method has its own set of merits and limitations, and the printing method selected depends on a range of requirements, including printing output, resolution, surface smoothness, repeatability, and economic considerations
[177]. However, reflecting on the evolution of MPL since its invention in 1997, most commercially available processing equipment is geared toward laboratory research, whereas the availability of equipment for large-scale industrial production remains relatively limited. Currently, notable applications have been realized primarily in areas such as the production of micro- and nano-master plates. Based on the literature review and analysis, we broadly attribute these circumstances to the following factors.
4.1. Throughput and cost
Firstly, productivity is a major bottleneck faced by MPL processing in achieving large-scale industrial production
[178]. Guided by application needs, we discuss the productivity of MPL, such as the nanoprinting throughput, required in the future to meet the demands of large-scale nano-manufacturing. For example, in the medical field, using TPP to print drug-loaded inhalable microcapsules is a feasible and meaningful idea. However, loading 5 mg of medication into a capsule requires approximately 10
12 voxels
[179]. Given that the current highest throughput of TPP devices is approximately 10
8 voxels·s
−1, processing still requires several hours, making mass production impractical. We believe that if the throughput can be increased by two orders of magnitude to reach 10
10 voxels·s
−1, it would be suitable for mass production. Currently, there are two main approaches for improving the processing speed: increasing the serial speed, such as using devices with higher scanning speeds, or increasing the parallel efficiency, such as multipoint processing or pattern projection. As processing throughput continues to increase, laser power becomes a new limiting factor. Under the same laser power, the exposure received by the photoresist per unit volume decreases with increasing processing speed. When the laser power does not reach the threshold required for photopolymerization, the photoresist may fail to polymerize. Generally, there are two methods to solve this problem: increasing the laser power or lowering the photoresist polymerization threshold. The production of high-power, high-repetition-rate, narrow-pulse-width femtosecond lasers is challenging and typically requires custom production, adding fixed costs and posing challenges to optical components. Increasing the photoresist sensitivity is also an effective method, such as reducing inhibitors in the photoresist and adopting more efficient photoinitiators. While reducing the inhibition effects may lower the threshold to some extent, it could unfortunately lead to a reduction in resolution. Comparatively, using more efficient photoinitiators is more suitable for fine MPL processing
[180],
[181],
[182],
[183],
[184]. Although separate designs for different polymerization systems are required, it is a relatively effective method and is expected to become a mainstream solution to power and threshold issues in future large-scale production.
Secondly, the high cost has raised the threshold for large-scale industrial applications of MPL technologies. The nonlinear effects of the absorption rate and virtual intermediate energy levels in MPL generally require femtosecond and high-power lasers to reach the photopolymerization threshold, leading to high hardware costs. Moreover, the current processing rates cannot meet the demands of large-scale direct writing. Processing large-scale samples often takes several tens of hours, which significantly increases the time cost. It is worth mentioning that two-step processing
[9],
[10] also has an absorption probability proportional to the square of the light intensity, achieving a resolution close to that of MPL. Compared to MPL, the intermediate energy levels in two-step absorption are real energy levels, giving it much higher sensitivity and not requiring femtosecond lasers. The use of continuous lasers significantly reduces processing costs. 3D printing using upconversion nanocapsules can be achieved with a power density that is several orders of magnitude lower than that required for TPP-based 3D printing
[185]. These technologies are expected to become more promising micro/nano 3D printing techniques than femtosecond laser MPL processing.
4.2. Optimization of cross-scale processing
Cross-scale processing presents a complex challenge involving not only limitations in processing speed but also issues related to slicing, focusing, and stitching. Rational slicing and layer planning, as well as the scanning path strategy, are crucial for minimizing processing time and enhancing processing quality without altering the processing voxel
[186]. For instance, in regions with varying slopes, adjusting the slice thickness based on the slope can effectively reduce staircase effects and prevent surface roughness
[187],
[188]. Additionally, for structures such as microlenses with spherical profiles, employing circular scanning and variable-thickness slicing methods can achieve high-quality and smooth surface profiles
[189]. Employing voxels of different sizes for direct writing in different regions can reduce processing time when printing large volumes
[91]. Technologies such as Nanoscribe grayscale lithography, which is based on variable power, enable rapid and accurate surface patterning. When utilizing a precision displacement stage with a larger stroke for processing a large model, stitching in the spliced areas becomes a crucial problem that affects processing quality. Researchers have proposed methods to address this problem, including strategies such as linear stage and scanner synchronization and optimization of splicing positions and overlap
[190],
[191]. In addition, seamless splicing generally requires integration with automatic focusing technology to minimize vertical splicing errors
[192],
[193]. Furthermore, MPL processing technology can be integrated with other manufacturing techniques
[194],
[195], leveraging the advantages of precision and speed to achieve cross-scale manufacturing and production. In summary, low-cost, cross-scale, and rapid manufacturing is poised to become a primary developmental direction in the future of multi-photon 3D nanoprinting.
4.3. Material limitations
Constrained by material processing capabilities and performance, many devices produced using MPL do not exhibit ideal performance. Initially, the materials used for MPL were organic polymers. Despite the subsequent development of various methods to enrich the material system for MPL processing, allowing for the processing of inorganic and composite materials, these materials are still reliant on the framework of organic materials. Final shaping and purification are achieved through subsequent methods such as doping, adsorption, conversion, and pyrolysis. Therefore, the physicochemical properties of organic polymers directly affect the performance of the micro/nano-devices produced using this processing method.
Organic polymers produced through direct writing commonly face challenges such as shrinkage and distortion, which affect the performance of sensitive devices like micro- and nano-optical devices. Compensatory methods, such as pre-compensation design and the use of low-shrinkage materials, may be effective but increase the complexity of material and equipment design. Additionally, the limited refractive index and transparency of these polymers restrict the focusing performance and tolerable optical power of processed optical devices. Moreover, the mechanical strength of polymers written using MPL is generally low, limiting their application in micromechanical systems with high-intensity and complex motion control. Although modifying the functional groups and crosslinking structures can adjust the properties of the polymer, achieving optimal control requires further research. Notably, the use of metal nanoclusters as photoinitiators has shown promise for significantly enhancing the mechanical properties of directly written materials, providing a novel approach for improving their performance
[196]. Additionally, certain photoresists can undergo polymerization via femtosecond laser irradiation without a photoinitiator. Polymers without a photoinitiator exhibit no side effects such as coloration, absorption, or toxicity
[197],
[198].
The use of organic polymers as a framework for metals and their compounds in MPL direct writing has limitations compared to methods such as meniscus-confined electroplating. This approach involves sintering or chemical etching of organic materials, resulting in a limited range of writable metals, lower purity, relatively poor conductivity, low aspect ratios, and significant shrinkage
[4],
[127],
[129],
[130],
[131],
[132],
[133],
[135]. Mutual interference among various metal ions makes it difficult to achieve integrated structures with different metal types, restricting the design and application of micro/nano electronic devices
[5],
[62],
[63],
[64]. Increasing the metal-ion content in a polymer is crucial for enhancing the purity and electrical properties of the final product and addressing issues such as shrinkage. However, current methods are insufficient to meet the demands of micro/nano electronic device performance. Recent studies have explored new polymerization mechanisms. Liu et al.
[199] proposed a method using photoexcitation-induced chemical bonding for the 3D nanoprinting of semiconductor quantum dots. This alternative approach, which does not rely on organic material polymerization, allows printing resolutions beyond the diffraction limit for arbitrary 3D quantum dot structures. In a subsequent study, they expanded the printable materials to more than 10 semiconductors, metal oxides, metals, and their mixtures by utilizing colloidal nanocrystals bonded through their native ligands
[200]. Wang et al.
[201] used a method called threshold tracking and lock-in (TTL) to induce atomic point defects in damaged crystals, thereby achieving manufacturing with feature sizes smaller than 5 nm. In summary, refining material systems and improving material properties can significantly advance the application and development of multi-photon 3D manufacturing in future industries.
Although the research field of MPL has made significant progress in processing and materials over the past two decades, overcoming challenges in large-scale industrial production may require rethinking its fundamental mechanisms. While MPL excels in spatial precision, its reliance on organic polymerization imposes many limitations. Therefore, proposing a new additive manufacturing reaction principle that retains nonlinear absorption characteristics while possessing excellent shaping and performance capabilities, along with high reaction sensitivity, is one of the most effective pathways to achieve large-scale industrial production. The realization of MPL 3D nanoprinting is expected to create disruptive new structures and functional devices, significantly advancing fields such as mechanics, optics, biology, and other disciplines in the micro/nano domain.
CRediT authorship contribution statement
Fayu Chen: Writing – review & editing, Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Shaoxi Shi: Writing – review & editing, Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Songyan Xue: Writing – original draft, Formal analysis, Data curation. Huace Hu: Writing – original draft, Formal analysis, Data curation. Zexu Zhang: Writing – original draft, Formal analysis, Data curation. Xuhao Fan: Writing – original draft, Data curation. Mingduo Zhang: Writing – original draft, Investigation. Xinger Wang: Writing – original draft, Investigation. Zhe Zhao: Investigation. Hui Gao: Validation, Supervision, Resources, Project administration, Funding acquisition, Formal analysis, Conceptualization. Wei Xiong: Validation, Supervision, Resources, Project administration, Funding acquisition, Formal analysis, Conceptualization.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This research was financially supported by the National Key Research and Development Program of China (2021YFF0502700), the National Natural Science Foundation of China (52275429 and 62205117), the Innovation Project of Optics Valley Laboratory (OVL2021ZD002), the Young Elite Scientists Sponsorship Program by CAST (2022QNRC001), the West Light Foundation of the Chinese Academy of Sciences (xbzg-zdsys-202206), the Knowledge Innovation Program of Wuhan-Shuguang, and the Hubei Provincial Natural Science Foundation of China (2022CFB792).
PDF
(7075KB)
