Optofluidic three-dimensional microfabrication and nanofabrication
Xianglong Lyu, Wenhai Lei, Gaurav Gardi, Muhammad Turab Ali Khan, Shervin Bagheri, Mingchao Zhang, Metin Sitti
TL;DR
A new method for 3D microfabrication uses light and fluid interactions to create complex structures from various materials, enabling advanced microdevices.
Contribution
A universal 3D fabrication strategy using optofluidic interactions for assembling diverse nanoparticles into complex structures.
Findings
The method enables fabrication of 3D microfluidic valves with size-selective sieving functionality.
Microrobots with multimodal locomotion using four different functional materials were successfully created.
The approach works with a wide range of materials including metals, metal oxides, and quantum dots.
Abstract
Three-dimensional (3D) microfabrication/nanofabrication technologies have revolutionized various fields by enabling the precise construction of complex microstructures/nanostructures1–6. However, existing methods face challenges in fabricating intricate 3D architectures from a diverse range of materials beyond conventional polymers. Here we introduce a universal 3D microfabrication/nanofabrication strategy compatible with a broad range of materials by precisely manipulating optofluidic interactions within a confined 3D space, enabling the creation of volumetric, free-form 3D microstructures/nanostructures. A femtosecond-laser-induced heating spot generates a localized thermal gradient, providing precise spatiotemporal control over optofluidic interactions of the nanoparticle-laden dispersions. This enables the rapid and highly localized assembly of nanoparticles with diverse shapes and…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsNear-Field Optical Microscopy · Pickering emulsions and particle stabilization · Laser Material Processing Techniques
Main
Three-dimensional (3D) microfabrication/nanofabrication technologies have attracted tremendous attention as they can create various functional microdevices/nanodevices, such as microrobots^1,2^, microactuators^3^, micrometre-scale metamaterials^4,5,8^ and microphotonic/nanophotonic devices^6^. Over the past two decades, two-photon polymerization (2PP) has emerged as the state-of-the-art 3D microfabrication/nanofabrication strategy owing to its high resolution (up to 100 nm), facile fabrication process and capability of printing intricate free-form 3D microstructures/nanostructures^9–12^. However, the material compatibility of 2PP remains highly limited to cross-linkable polymers. Recent efforts have aimed to expand printable materials beyond polymers, primarily through developing advanced photoresists with tailored chemistries, such as grafting photochemically bonding ligands onto inorganic colloidal nanocrystals^13,14^ or incorporating metal-coordination complexes into cross-linkable monomers^15–21^. Nonetheless, these approaches remain restricted to specific materials and continue to face challenges in achieving broad compatibility with diverse material systems.
As a promising approach to overcome these material limitations, direct assembly of material building blocks has proved effective for 3D microstructures/nanostructures^22^. Among various assembly techniques, optical assembly^23–30^ has been an appealing strategy for the construction of complex material assemblies. It uses the non-specific light–matter interactions (for example, optical gradient forces^31^ and light-controlled electric or temperature fields^32^) to trap microparticles/nanoparticles suspended in a solution. The trapped particles can then be transported and positioned one by one at designated locations, enabling high-precision single-particle assembly. Techniques such as optical tweezers^33^, opto-thermophoretic assembly^34^, opto-thermoelectric assembly^35^ and optothermal-flow-based assembly^24,27,36,37^ have been used to assemble colloidal particles with diverse sizes, shapes and surface properties. However, these approaches remain largely limited to 2D structural configurations and exhibit low assembly efficiencies, typically on the order of 10^1^–10^3^ particles min^−1^. More importantly, establishing a general optical assembly platform with broader material applicability is still challenging, as most techniques impose strict requirements on parameters, such as particle size/surface properties, solvent composition and other environmental conditions. A detailed comparison of various assembly strategies is provided in Supplementary Table 1. Therefore, innovations in general assembly mechanisms are urgently needed to achieve controllable, non-specific guidance and precise assembly of material building blocks at the microscale/nanoscale, enabling the creation of high-quality, truly volumetric free-form 3D architectures with broad material compatibility.
Here we propose a universal 3D microfabrication/nanofabrication strategy compatible with a broad range of materials. This technique uses the optofluidic interaction—light-driven flow—to efficiently assemble diverse micromaterials/nanomaterials into predefined 2PP-printed microtemplates, creating high-quality, truly volumetric free-form 3D microarchitectures/nanoarchitectures from a variety of materials, including diamond, metal, metal oxide, quantum dots and others. By integrating the optofluidic assembly process with 2PP, our optofluidic 3D microfabrication/nanofabrication strategy proposes a generalizable model that overcomes the material limitations of 2PP, opening new avenues for nanotechnology (Supplementary Fig. 1 and Supplementary Table 2).
Concept of optofluidic 3D micro-/nanofabrication
The working process of optofluidic 3D microfabrication/nanofabrication is illustrated in Fig. 1a and consists of the following key steps. First, a 3D hollow polymeric microstructure with an open hole (for example, a cube) is printed on a glass substrate by the 2PP process, serving as the 3D microtemplate. Next, the printed template is immersed in a solution containing uniformly dispersed nanoparticles (or few-micrometre particles). A femtosecond (fs) laser with a beam diameter of 2 µm is then applied near the open hole of the template, generating a sharp temperature gradient that induces a strong convective flow (up to several mm s^−1^), propelling the dispersed particles towards the open hole. As a result, these particles are transported into the hollow microtemplate and accumulate over time, ultimately assembling into the prescribed 3D shape dictated by the template. Following assembly, the polymer template is selectively removed through rational post-treatments (see details in Methods), yielding a free-standing 3D volumetric microarchitecture composed entirely of the targeted materials. For example, our approach enables the fabrication of a series of solid microcubes randomly assembled by SiO_2_ nanoparticles with an assembly efficiency of approximately 10^5^ particles min^−1^, as shown in the scanning electron microscopy (SEM) image in Fig. 1b, its enlarged image in Fig. 1c and Supplementary Fig. 2. The resulting 3D microarchitecture exhibits high structural integrity, as these constituent nanoparticles are strongly bonded and stabilized by van der Waals forces. Even without further chemical bonding or annealing process for interfacial improvement, these 3D structures can be self-supporting and mechanically stable, owing to the strong van der Waals interactions in colloidal nanoparticle assemblies^12^. This is further evidenced by the successful fabrication of a dangling croissant-shaped superstructure with intricate 3D curved surfaces (Fig. 1d,e), demonstrating the robustness of this technique.Fig. 1. Concept of optofluidic 3D microfabrication/nanofabrication.a, Schematic illustration of the optofluidic 3D microfabrication/nanofabrication process, in which a localized temperature gradient induced by femtosecond (fs) laser heating generates a strong convective flow, guiding the 3D assembly of microparticles/nanoparticles within a confined hollow 3D microtemplate, printed by 2PP. b,c, SEM images of a SiO_2_ colloidal-particle-assembled microcube (b) and its zoomed-in view (c). d,e, SEM images of a dangling croissant-shaped microstructure with 3D curved surface assembled from SiO_2_ particles (d) and its enlarged view (e). The inset in d is the 3D model of the croissant structure. f, Simulation result showing the temperature distribution and fluid flow field around a hollow microcube following fs laser heating. g, Schematic illustrations and time-lapse optical images showing the assembly process of SiO_2_ nanoparticles within a hollow microcube. These images are extracted from Supplementary Video 1. Scale bars, 10 μm (b,d); 4 μm (c,e); 20 μm (g).
To gain deeper insights into the assembly process, we conducted simulations and recorded the experimental process to analyse the underlying dynamics. As shown in Fig. 1f, Supplementary Fig. 3 and Supplementary Video 1, when laser heating is applied at the open hole of a hollow microcube, a sharp thermal gradient near the hotspot is induced. This thermal gradient causes variations in fluid density, pressure and surface tension around the hole^26,38^, generating a strong, directed fluid flow towards the open hole, transporting dispersed SiO_2_ particles into the microcube. Furthermore, we note that bubble formation with a diameter up to 100 µm (Supplementary Fig. 4) occurs easily owing to solvent evaporation induced by femtosecond laser heating, leading to an extra flow from the Marangoni effect^39^, and this disturbance further promotes and accelerates particle assembly within the template (Supplementary Video 1). As shown in Fig. 1g and Supplementary Video 1, an assembly speed of around 700 µm^3^ s^−1^ is reached when assembling SiO_2_ colloidal particles with a diameter of 1 µm, which is about twice as fast as the typical 2PP process (Supplementary Fig. 5). By precisely manipulating the light-driven fluid flow through localized laser irradiation, any nanoparticles carried in the flow can be efficiently assembled into 3D microarchitectures. This non-selective mechanism makes the strategy inherently adaptable to a broad range of materials.
Assembly mechanism
The assembly process is primarily governed by two competing physical forces: inter-particle interactions and particle–fluid interactions. We start with a model system in which SiO_2_ colloidal particles (150 nm in diameter) assemble in aqueous solutions with varying ionic strengths. In this system, the inter-particle interactions consist of van der Waals attraction and electric double layer (EDL) repulsion, commonly described as the net DLVO interaction force^40,41^. The inter-particle DLVO force is highly sensitive to particle distance and particle properties and plays a crucial role in colloidal cluster formation. Meanwhile, the particle–fluid interactions are represented by hydrodynamic forces, particularly the Stokes drag force, which reflects how particles are carried by the surrounding flow. In our local assembly region, it tends to disperse the particles and counteracts aggregation (Fig. 2a). To determine whether colloidal particles will form clusters, we analyse the balance between inter-particle potential energy and hydrodynamic effects. The total interaction energy governing the system can be expressed as: Utotal = UDLVO + Wfluid, in which UDLVO represents the interaction potential between particles and Wfluid denotes the work done by hydrodynamic forces. The DLVO potential energy is derived from the integration of the inter-particle DLVO force over the separation distance, whereas the work done by the fluid force depends on the external flow field and particle motion. The tendency of the system to form clusters can then be evaluated by examining changes in Utotal. If ΔUtotal < 0, the inter-particle attractive interactions dominate, leading to energetically favourable conditions for colloidal clustering around the laser spot. If ΔUtotal > 0, particle–fluid interactions prevail, causing particles to move with the fluid flow and remain dispersed (Fig. 2a).Fig. 2. Assembly mechanism.a, Schematic illustration showing the competition between DLVO interactions and Stokes drag force and their influence on colloidal cluster formation following femtosecond laser heating. b, Cluster area (purple spheres) of SiO_2_ particles dispersed in aqueous solutions with varying NaCl concentrations after 60 s of laser heating at a power of 50 mW and a scan speed of 5 μm s^−1^, along with the corresponding zeta potential value (black diamonds) of these SiO_2_ particles. The insets are optical images of SiO_2_ clusters formed in 0.6 M and 1 M NaCl solutions. Data are presented as mean ± s.d. with at least four independent samples/measurements (n ≥ 4). c,d, Schematic illustrations and optical images depicting the assembly process of SiO_2_ particles within a hollow microcube in 1 M (c) and 100 mM (d) NaCl solution. The microcube measures 32 μm in length and width and 10 μm in height, with an opening of 26 μm in length and 10 μm in height on the side. e, Theoretical phase diagram (grey background) and experimental results (open circles and solid diamonds) showing the influence of various concentrations of NaCl and flow speed on SiO_2_ particle clustering. The light-grey region and open circles represent the domain in which SiO_2_ particles remain dispersed and no cluster is formed, whereas the dark-grey region and solid diamonds represent the domain in which SiO_2_ cluster formation occurs. Scale bars, 30 μm (b); 10 μm (c,d).Source data
The template we used plays a crucial role in enabling deterministic 3D fabrication: it defines the overall geometry to ensure well-defined edges and symmetry, channels the optofluidic flow to reproducibly fill complex volumes and provides design versatility by allowing diverse 3D architectures. To effectively assemble dispersed SiO_2_ colloidal particles within a confined 3D space, the system must satisfy the condition ΔUtotal < 0 to promote colloidal cluster formation. This can be achieved through two approaches: (1) increasing the inter-particle attractive interaction or (2) decreasing the speed of fluid flow. We first investigate the effect of varying inter-particle interactions on SiO_2_ cluster formation. In aqueous solution, SiO_2_ particles are highly surface charged, resulting in strong electrostatic repulsion that stabilizes the dispersion^42–44^. By tuning the ionic strength of the solution, the zeta potential of SiO_2_ colloidal particles and the Debye length—a characteristic length scale over which electric fields are naturally screened—can be systematically adjusted, thereby influencing inter-particle electrostatic interactions. As illustrated in Fig. 2b, increasing the NaCl concentration progressively decreases the zeta potential of SiO_2_ particles and compresses the Debye length, thus reducing the electrostatic repulsion within the EDL. Consequently, the attractive component of the DLVO interaction is enhanced, leading to stronger particle aggregation. At higher NaCl concentrations, this effect facilitates the formation of larger SiO_2_ clusters, whereas no notable clustering is observed when the NaCl concentration is less than 200 mM. Correspondingly, as depicted in Fig. 2c, SiO_2_ particles effectively assemble and accumulate within a hollow template in a solution with 1 M NaCl concentration (Supplementary Video 2, part 1), whereas no assembly is observed at a lower NaCl concentration (100 mM; Fig. 2d and Supplementary Video 2, part 2).
Flow speed is another vital factor affecting the cluster formation of the SiO_2_ colloidal particles. Higher flow speeds generate stronger Stokes forces acting on colloidal particles, hindering their aggregation. By comparing the relative values between UDLVO and Wfluid, a critical flow speed of approximately 300 µm s^−1^ is theoretically determined and summarized in a phase diagram (Fig. 2e; see Methods for details). Above this critical speed, the Stokes force dominates and completely overcomes the DLVO attraction, preventing cluster formation regardless of NaCl concentration. Below this threshold, the DLVO force becomes comparable to the Stokes force, allowing cluster formation at high NaCl concentrations (for example, >0.5 M). By adjusting the laser scan speed, different flow speeds can be generated experimentally. Overall, the assembly process was systematically conducted and analysed across various NaCl concentrations and flow speeds. The experimental results (represented as discrete symbols in Fig. 2e) closely align with theoretical predictions. For example, SiO_2_ particles in 1 M NaCl solution aggregate into clusters at flow speeds less than 320 µm s^−1^, whereas they remain dispersed at higher flow speeds of about 900 µm s^−1^ (Supplementary Fig. 6 and Supplementary Video 3). Besides, an extremely weak flow field cannot provide sufficient driving force to continuously transport the particles towards the inside of a template, failing to fill up the template (Supplementary Video 4).
Assembly in various solvent systems
To broaden the compatibility of the optofluidic 3D microfabrication/nanofabrication technique, the 3D assembly of microparticles/nanoparticles can be extended to various solvent systems, as different colloidal particles in different solutions exhibit varying balances between inter-particle and particle–fluid interactions. For instance, replacing a good solvent with a poor solvent can enhance particle–particle attraction, leading to the clustering of the colloidal particles^45,46^ (Fig. 3a). When hydrophilic SiO_2_ particles are redispersed into hydrophobic solvents, such as immersion oil, mineral oil or oleic acid, the strong hydrophobic interaction facilitates continuous cluster growth (Fig. 3b), resulting in a faster growth of SiO_2_ cluster under laser illumination (Fig. 3c). Consequently, SiO_2_ particles can be efficiently assembled and fully filled within a hollow micro-template in these solvent systems (Supplementary Video 5). More importantly, the strong hydrophobic interaction enables efficient 3D assembly at a flow speed of several mm s^−1^, even under substantial Stokes force, greatly improving assembly efficiency. Furthermore, the assembly speed can be modulated by adjusting the laser dosage (the laser power and scan speed; see Supplementary Fig. 7). Notably, SiO_2_ particles dispersed in silicone oil exhibit poor clustering ability and fail to undergo 3D assembly within the hollow template (Supplementary Video 6). This might be because of weak attractive interaction among SiO_2_ particles, as their strong affinity to silicone oil (attributing to their mutual Si–O bonds) prevents clustering.Fig. 3. Assembly optimization using various solvent systems.a, Schematic illustration showing the clustering behaviours of colloidal particles in solutions with different particle–solvent affinities. b, Cluster area of SiO_2_ particles over time in different solvents. Data points are extracted from Supplementary Video 16 at 10-s intervals. c, Cluster area of SiO_2_ particles in various solvents after 60 s of laser heating at a power of 50 mW and a scan speed of 500 μm s^−1^. Insets are optical images of SiO_2_ clusters in oleic acid and silicone oil. d, Schematic illustration depicting the effect of surfactant on colloidal particle assembly in the aqueous solutions. The addition of surfactant reduces the surface tension gradient, attenuating the bubble growth and weakening Marangoni flow, thereby facilitating the particle clustering. e, Cluster area of SiO_2_ particles (purple spheres) and flow speed (black diamonds) in aqueous solutions with different concentrations of CTAB. Insets are optical images of SiO_2_ clusters in pure water and 1 mM CTAB solution. f, Cluster area of SiO_2_ particles in solutions containing different surfactants (1 wt% PF108, 1 wt% PEG and 8 mM SDS). Insets are optical images of SiO_2_ clusters in solutions of 1 wt% PF108 and 8 mM SDS. The laser power and scan speed for e and f are both 50 mW and 5 μm s^−1^, respectively, with a duration of 10 s. Data points are shown as mean ± s.d. with at least three independent samples/measurements (n ≥ 3). Scale bars, 30 μm (c,e,f).Source data
Beyond solvent selection, surfactants can also enhance efficiency in 3D assembly in aqueous solutions. Laser illumination often induces bubble formation owing to high-temperature-driven evaporation, generating Marangoni flow caused by the surface tension gradients at the bubble–solution interface. Moderate flows promote SiO_2_ particle clustering, whereas a strong flow inhibits cluster formation (Fig. 3d,e and Supplementary Video 7), as the bubble formation frequently results in intense fluid flow exceeding 4 mm s^−1^. The addition of a surfactant, such as hexadecyltrimethylammonium bromide (CTAB), effectively reduces the surface tension, limiting bubble growth and weakening the laser-induced Marangoni flow^39^. Moreover, the cationic ions (CTA^+^) dissociated from CTAB molecules absorb onto the surface of negatively charged SiO_2_ particles, neutralizing their surface charge (Supplementary Fig. 8). This reduction in electrostatic repulsion enhances the DLVO attractions among the SiO_2_ particles, thereby promoting their aggregation. As a result, the SiO_2_ clustering is progressively enhanced by increasing CTAB concentration (Fig. 3e), enabling robust 3D assembly in 1 mM CTAB solution (Supplementary Video 8). Beyond CTAB (cationic surfactant), other surfactants, including anionic surfactant sodium dodecyl sulfate (SDS) and non-ionic surfactants such as polyethylene glycol (PEG) and Pluronic F-108 (PF108), exhibit similar effects on SiO_2_ particle clustering (Fig. 3f). Notably, SiO_2_ particles exhibit strong electrostatic repulsion in the solution of SDS (Supplementary Fig. 8), which may hinder the cluster formation. However, other inter-particle attractions, such as opto-thermophoretic force^23,34,47^ and depletion force^48^, may also arise in these surfactant systems, facilitating particle aggregation.
Broad compatibility with versatile materials
Our strategy is superior at constructing complex 3D microstructures using a diverse range of micromaterials/nanomaterials, regardless of their shape, size and surface chemistry. To demonstrate the universality of this approach, we first assembled SiO_2_ colloidal particles of varying sizes and surface chemistries into versatile 3D microstructures. These include a micro-gourd composed of 1-µm-diameter SiO_2_ (Fig. 4a–c), a micro-hexagon made of 600-nm-diameter SiO_2_ particles (Fig. 4d–f) and microcubes constructed from 1-µm-diameter green fluorescent SiO_2_ particles (Supplementary Fig. 9) and 3-µm-diameter and 10-µm-diameter SiO_2_ particles (Supplementary Video 9). Furthermore, particles of different sizes can be co-assembled to form heterogeneous 3D microarchitectures. For instance, a microsphere consisting of both 1-µm and 600-nm SiO_2_ particles is successfully co-assembled (Fig. 4g–i), demonstrating the capability of integrating different components into 3D microstructures.Fig. 4. Wide compatibility with versatile micromaterials/nanomaterials.a–c, 3D model (a) and SEM images (b,c) of a micro-gourd superstructure assembled with 1-μm SiO_2_ particles. d–f, 3D model (d) and SEM images (e,f) of a hexagon-shaped micro-superstructure assembled with 600-nm SiO_2_ particles. g–i, 3D model (g) and SEM images (h,i) of a microsphere co-assembled with 1-μm and 600-nm SiO_2_ particles. j–l, Model (j) and SEM images (k,l) of superstructures of the letters ‘P’ (assembled with 1-μm SiO_2_ particles) and ‘I’ (assembled with 600-nm SiO_2_ particles). m,n, 3D model (m) and SEM image (n) of a TiO_2_ nanoparticle (NP)-assembled screw-like microstructure. o, TEM image of TiO_2_ NPs. p,q, 3D model (p) and SEM images (q (i), (ii)) of the letter ‘E’ composed of Fe_3_O_4_ NPs. r, TEM image of Fe_3_O_4_ NPs. s–z, Model (s), SEM image (t) and EDS mapping (u–z) of a microcube assembled with various materials, including SiO_2_ (t), TiO_2_ NWs (u), WO_3_ NWs (v), diamond NPs (w), Al_2_O_3_ NWs (x), Fe_3_O_4_ NPs (y) and Ag NPs (z). The high-resolution surface morphology of the microstructures assembled with different nanomaterials, along with the corresponding component nanomaterials, can be found in Extended Data Figs. 1 and 2. Scale bars, 10 μm (b,e,h,k,t); 4 μm (c,f,i,l); 5 μm (u–z); 2 μm (n,q (i)); 800 nm (q (ii)); 200 nm (o); 80 nm (r).
Furthermore, architectures at distinct locations can be site-selectively assembled with particles of different sizes or components by locally addressing the fs laser, with no cross-interference among these assembled architectures. For example, the alphabet letters ‘P’ and ‘I’ are sequentially assembled using 1-µm and 600-nm SiO_2_ within a close mutual distance of about 10 µm on the substrate (Fig. 4j–l). Once assembled within a template, strong inter-particle interactions resist disturbances from the violent flow during sequential assembly process, preventing disassembly. This precise control enables the localized integration of several components for a targeted layout, paving the way for the fabrication of microdevices with spatially varying compositions and on-demand multifunctionalities (see details in Fig. 5). Beyond these micrometre-sized SiO_2_ particles, our method is compatible with a broad range of nanomaterials, enabling the fabrication of diverse 3D structures with nanoscale features. As illustrated in Fig. 4m–r, a screw-like microstructure with helical threads of around 320 nm in width (Fig. 4m,n) and an alphabet letter ‘E’ with a height of around 855 nm (Fig. 4p–r) are created using TiO_2_ nanoparticles with 90.2 ± 15.8 nm diameter and Fe_3_O_4_ nanoparticles with 16.7 ± 2.3 nm diameter, respectively. Notably, smaller nanoparticles with a uniform size distribution yield microstructures/nanostructures with smoother surfaces (Fig. 4q and Extended Data Fig. 1). Furthermore, as illustrated in Fig. 4s–z and Extended Data Fig. 2, we successfully assemble microcubes from various nanomaterials, including TiO_2_ nanowires (NWs), diamond nanoparticles (Supplementary Fig. 10 and Supplementary Video 10), Fe_3_O_4_ nanoparticles, WO_3_ NWs, Al_2_O_3_ NWs, Ag nanoparticles and CdTe quantum dots (Supplementary Video 11), respectively. Notably, although these architectures exhibit intrinsic mechanical robustness (as demonstrated by the dangling croissant-shaped superstructure in Fig. 1d,e) owing to strong intrinsic inter-particle interactions, post-treatments such as annealing can further promote intra-particle interfacial welding, substantially enhancing the mechanical properties of the 3D microarchitectures.Fig. 5. On-demand construction of multifunctional microdevices.a–e, Microfluidic sieving devices. a, Schematic illustration showing the capillary-driven separation process of a microfluidic sieving device. b, Time-lapse optical images showing the capillary-driven fluid flow through a microfluidic chip embedded with a colloidal microvalve (20 μm width, 40 μm length) composed of 1-μm SiO_2_ particles. Images are extracted from Supplementary Video 17. c, SEM image of the SiO_2_-assembled microvalve. d, Fluorescent image of 500-nm polystyrene (PS) nanospheres rejected by the microvalve. e, Fluorescent image showing 100-nm PLGA nanoparticles rejected by the microvalve. Inset is the variation of gray value along the microchannel. f–s, Multifield-driven microrobots. f,g, Schematic illustration (f) and trajectory (g) of the magnetic tumbling of a Fe_3_O_4_ nanoparticle (NP)-assembled cylinder microrobot. h–o, Schematic illustrations (h,j,l,n) and motion trajectories (i,k,m,o) of TiO_2_–Au microrobots with different shapes and material distributions. p–s, Schematic illustration (p) of an L-shaped microrobot integrated with four different materials (TiO_2_, Au, Pt and Fe_3_O_4_) and its three motion modes: magnetic pulling (q), light-driven anticlockwise rotation (r) and clockwise rotation in about5 wt% H_2_O_2_ (s). Scale bars, 50 μm (b,d,e); 60 μm (g,i,k,m,o,r,s); 30 μm (q); 10 μm (c).Source data
On-demand creation of multifunctional microdevices
With the aid of precise spatial control and broad material applicability, microstructures spatially encoded with diverse functional materials can be created, highlighting the potential of our technique for developing microdevices with on-demand functionalities. As proof-of-concept demonstrations, microfluidic chips with tailored separation capabilities for tiny objects are first demonstrated (Fig. 5a–e). This microfluidic chip consists of a polymeric microchannel printed by the 2PP process with a colloidal-particle-assembled microvalve embedded inside (Fig. 5b,c). The microvalve, measuring 40 µm in length and 20 µm in width, is entirely assembled from 1-µm SiO_2_ particles, forming complex 3D porous channels (Fig. 5c). When a droplet of nanoparticle suspension is introduced at one end of the microfluidic chip, capillary forces instantaneously drive the solution into the microchannel (Fig. 5a). The colloidal microvalve, with feature porosity of several hundred nanometres, permits the rapid permeation of solvent flow in seconds (Fig. 5b), while efficiently rejecting and retaining nanoparticles (Fig. 5d). As solvent evaporation continues at the opposite end of the chip, nanoparticles progressively accumulate at the inlet side of the microvalves, forming an enrichment zone (Fig. 5a and Supplementary Fig. 11), which holds potential for trace substance detection. By modulating the size of the microvalves, we can selectively separate particles of different sizes. For instance, 100-nm poly(lactic-co-glycolic) acid (PLGA) nanoparticles are successfully sorted out using a SiO_2_ valve of 40 µm in both width and length (Fig. 5e). Furthermore, we also construct a microfluidic chip embedded with several concatenated microvalves composed of different materials, each featuring distinct porosities and cut-off capabilities, achieving the size-selective sorting of a mixture of different particles (Supplementary Fig. 12).
Subsequently, we demonstrate the application of our technique in fabricating multifield-driven microrobots with multimodal locomotion (Fig. 5f–s). As illustrated in Fig. 5f,g, a Fe_3_O_4_ nanoparticle-assembled cylinder microrobot exhibits magnetically controlled tumbling on the substrate (Supplementary Video 12). The magnetic response can be flexibly tuned by regulating the Fe_3_O_4_ nanoparticle content (Supplementary Fig. 13). Beyond magnetic actuation, Fig. 5h–o illustrates light-driven microrobots with controlled motion enabled by tailoring both the shape and spatial distribution of functional materials. For example, a cylinder heterojunction TiO_2_–Au microrobot demonstrates linear propulsion through self-electrophoresis^49^ under ultraviolet (UV) illumination in water (Fig. 5h,i, Supplementary Fig. 14 and Supplementary Video 13, part 1). Shape asymmetry further enables rotational motion: a T-shaped TiO_2_–Au microrobot rotates in tight circles (Fig. 5j,k, Supplementary Video 13, part 2), whereas a V-shaped design produces larger circular trajectories (Fig. 5l,m and Supplementary Video 13, part 3). Moreover, alternative spatial encoding of functional materials allows the same V-shaped microrobot to switch from rotation to linear motion (Fig. 5n,o and Supplementary Video 13, part 4). Notably, our technique also enables the integration of several functional materials into a single microrobot, achieving multistimulus responsiveness. For instance, we fabricate an L-shaped microrobot incorporating Au, TiO_2_, Pt and Fe_3_O_4_ (Fig. 5p–s), which exhibits three distinct motion modes: magnetic pulling, anticlockwise rotation under UV light and clockwise rotation in H_2_O_2_ (Supplementary Video 14).
In conclusion, our technique overcomes the material limitation in conventional 3D microfabrication/nanofabrication, enabling the creation of complex, volumetric 3D microstructures/nanostructures from a diverse range of materials. We believe that this step change in fabrication capability—bridging optical physics, colloidal science, fluid mechanics and device engineering—not only provides new insights into fundamental colloidal assembly but also opens new frontiers in various fields, such as reconfigurable photonics, multifunctional microdevices/microrobots and biologically integrated systems.
Methods
Materials
SiO_2_ colloidal particles with various sizes (140 nm to 5 μm, 5 wt%), green-fluorescent SiO_2_ (1 μm, excitation/emission: 497/530 nm, 2.5 wt%) and red-fluorescent polystyrene nanospheres (500 nm and 2.5 μm, excitation/emission: 530/607 nm, 2.5 wt%) were purchased from microParticles GmbH. Mineral Oil Rotational Viscosity Standard (494.0 mPa s), oleic acid (technical grade, 90%), silicone oil AR20 (viscosity: approximately 20 mPa s), CTAB (98%), SDS (98.5%), PEG (Mn: 400), PF108 (Mn: 14,600), TiO_2_ NWs (powder, 100 nm × 10 μm), WO_3_ NWs (powder, 50 nm × 10 μm), Al_2_O_3_ NWs (powder, 2–6 nm × 200–400 nm), CdTe quantum dots (powder, COOH functionalized, fluorescent emission: 710 nm), iron oxide (Fe_3_O_4_) powder (50–100 nm, 97%), Fe_3_O_4_ (20 nm, 5 mg ml^−1^, dispersed in H_2_O), TiO_2_ nanoparticles (150 nm, 900 nm, 5 wt%, dispersed in H_2_O), Au urchin nanoparticles (50 nm, in 0.1 mM PBS), Pt nanoparticles (powder, 50 nm, 99.9%), silver (Ag) powder (<100 nm, PVP as dispersant, 99.5%), orange fluorescent PLGA nanospheres (excitation/emission: 530/582 nm, 100 nm), diamond nanopowder (<10 nm, 97%), trichloro(1H,1H,2H,2H-perfluorooctyl) silane (97%), sodium chloride (NaCl, 99%), H_2_O_2_ (30% w/w in H_2_O) and propylene glycol monomethyl ether acetate (PGMEA, 99.5%) were purchased from Sigma-Aldrich. Isopropyl alcohol (IPA, 99.9%) was purchased from Carl Roth GmbH. IPS photoresist was purchased from Nanoscribe GmbH. Immersion oil 518 F (viscosity of about 500 mPa s) was purchased from Fisher Scientific.
Fabrication of 3D hollow microtemplates
The 3D hollow polymeric templates were designed using SOLIDWORKS 2021 and fabricated by direct laser writing using a commercial Photonic Professional GT system (Nanoscribe GmbH). The templates were printed with commercial IPS photoresist in oil-immersion mode with a 63× objective and printed on a fused glass substrate that had been treated overnight with trichloro(1H,1H,2H,2H-perfluorooctyl) silane vapour. The laser power and printing speed were 50 mW and 50,000 μm s^−1^, respectively. After printing, the structures were developed in IPA solution for 5 min to remove non-polymerized photoresist. The templates were then directly used for the particle assembly in particle-laden dispersions. For aqueous dispersion, the high thermal conductivity of water can limit photothermal efficiency. To enhance this, a 5-nm Cr layer followed by a 5-nm Au layer was sputtered onto the substrate, improving photothermal performance.
Process of the optofluidic 3D microfabrication/nanofabrication
SiO_2_ particles of different diameters (150 nm and 1 μm) were used in the assembly experiments. To redisperse SiO_2_ particles into various solvent systems, 100 μl of as-received aqueous SiO_2_ solution was first centrifuged at 8,000 rpm for 3 min. The supernatant was discarded and the collected SiO_2_ precipitate was dried by heating at 60 °C for 2 h. The dried SiO_2_ powder was then redispersed in 1 ml of various solvents (for example, immersion oil, oleic acid) and sonicated for 1 h to ensure thorough dispersion. These hydrophilic SiO_2_ particles can remain dispersed in various oils in the experimental timescale owing to the high viscosity of the medium, which helps prevent clogging the template openings by large agglomerations during assembly (Supplementary Fig. 15). The resulting SiO_2_ suspension was used for the assembly experiments. It should be noted that other colloidal materials were also redispersed in appropriate solvents using a similar procedure to perform the assembly experiments. Furthermore, co-assembly of different materials can be obtained by preparing a mixed suspension through physical blending of various particles, as demonstrated in Fig. 4i, in which SiO_2_ microspheres with 600 nm and 1 μm diameter were co-assembled into a single structure. The detailed parameters of colloidal materials used in this study are summarized in Supplementary Table 3.
For experiments conducted in aqueous solutions with varying NaCl concentrations, 150-nm SiO_2_ particles were selected because of their reduced gravitational settling, ensuring system stability before laser-induced assembly. SiO_2_ dispersions were prepared with different NaCl concentrations at a particle concentration of 1 wt%. To study assembly dynamics and ensure clear observation, large SiO_2_ particles (1 μm) were used in various solvent systems, including aqueous solutions containing surfactants and oil-based solvents. A 100-μl aliquot of each prepared solution was placed inside a square PDMS spacer (1 cm side length, about 200 μm height). The spacer was sealed with a cover glass to prevent evaporation. A fs laser beam (780 nm, 80 MHz) with a scanning area of 2 × 2 µm^2^ was continuously applied at the corner of the printed hollow template to guide particle assembly. After the assembly process, the substrate was sequentially washed with IPA and deionized water, followed by mild sonication to remove any residual colloidal particles surrounding the microstructure.
To create 3D structures or devices made from different particles, we used a sequential, multistep assembly procedure. Suspensions of different particles were introduced successively, with a washing step between each stage to remove excess particles and prevent cross-interference. For instance, when creating alphabet letters ‘P’ and ‘I’, a suspension of 600-nm SiO_2_ was first used to assemble ‘I’. The excess suspension was then rinsed away with IPA and water to ensure clean conditions for the next step, after which a suspension of 1-µm SiO_2_ particles was introduced to assemble ‘P’. By repeating these processes, several particle systems can be used to construct distinct structures at predefined locations on a single substrate or to achieve seamless integration within a single microstructure, thereby enabling the creation of multifunctional microdevices.
Finally, it should be noted that excessively strong inter-particle attraction can lead to the formation of large clusters that clog the openings, preventing complete filling of the 3D structures (Supplementary Video 15). This issue can be mitigated by optimizing the size and distribution of template openings, as shown in Supplementary Discussion 1 and Supplementary Figs. 16–18.
Template removal
To completely remove the outer-layer polymer template and maintain the structural integrity, the colloidal-assembled microstructures were subjected to mild O_2_ plasma treatment for 6 h with a plasma cleaner machine (Tergeo-EM, PIE Scientific) at 75 W power and an O_2_ flow rate of 20 sccm. To further enhance interfacial bonding, the microstructures were annealed in air at 600 °C for 2 h with a heating rate of 2 °C min^−1^. For faster template removal, a high-power Plasma Asher (MPR-6) can also be used.
Fabrication of microfluidic chips for particle separation
The microfluidic channel was fabricated using the Nanoscribe system with IPG-780 photoresist. To enhance fabrication efficiency, template models with an asymmetric hole configuration were designed, similar to the one shown in Supplementary Fig. 10. This design enabled direct printing of the microchannel without requiring polymer template removal after particle assembly. Specifically, 100 μl of IPG-780 was dispensed onto the substrate and pre-baked at 100 °C for 1 h. A microchannel (2 mm length, 40 μm width, 15 μm height) was then printed using the Nanoscribe system with a 63× objective. The laser power and printing speed were set to 30 mW and 80,000 μm s^−1^, respectively. After printing, the substrate was developed in PGMEA for 20 min, followed by IPA for 1 min to remove unpolymerized photoresist. For the microfluidic separation experiments, colloidal particle solutions containing 30 v/v% IPA were prepared to improve wettability within the microchannel.
Microrobot motion experiments
The fabricated microrobots were first treated with O_2_ plasma for 5 min to change their wettability before use. Afterwards, these microrobots were scratched off the glass and dispersed in deionized water. The microrobot suspension was then added to a piece of glass slide and observed under an optical microscope (ZEISS Axio Imager 2). The magnetically actuated tumbling was conducted under a uniform rotating field with an amplitude of 10 mT and frequency of 1–120 Hz. The light-driven motion is powered by a UV light-emitting diode with a wavelength of 365 nm (Thorlabs).
Microscopy characterization
The real-time particle assembly process was recorded using a built-in AxioVision Live-View camera in the Nanoscribe system. The recorded videos were analysed using the Manual Tracking plugin in Fiji software to determine the flow speed. The microfluidic separation experiments were recorded using an optical microscope (ZEISS Axio Imager 2). The SEM and energy-dispersive X-ray spectroscopy (EDS) images were obtained using a Gemini field-emission SEM 500 at an acceleration voltage of 5 or 20 kV. The transmission electron microscopy (TEM) images were acquired using a Philips CM200 TEM with an accelerating voltage of 200 kV. The fluorescent images were captured by a SP8 Leica confocal microscope and the raw data were processed and colourized using the built-in software LAX. The excitation wavelength was set to 552 nm with 10% intensity and the emission wavelength was collected at 630–700 nm.
Zeta potential measurement
The zeta potential of SiO_2_ (140 nm and 1 μm) colloidal particles dispersed in various solvents was measured in a Zetasizer Ultra system (Malvern). The viscosity and relative permittivity of solutions were set at 8.9 × 10^−4 ^Pa s and 78.5, respectively. The final zeta potential of SiO_2_ particles was obtained by averaging the results of five independent measurements.
Finite-element simulations
A 3D model was built with the COMSOL Multiphysics package (version 6.0) to simulate the fluid flow around a 2PP-printed hollow microcube. The size of the hollow cube was 40 × 40 × 40 µm^3^ with a hole of 10 × 10 × 10 µm^3^ at one of its corners (Supplementary Fig. 19). The computational domain was a cubic box of 400 × 400 × 400 µm^3^, with the upper surface set as a free boundary at a temperature of T0 = 293.5 K. The four side boundaries were modelled with symmetric boundary conditions, whereas the remaining boundaries were set as no-slip conditions. A laser with a scanning area of 2 × 2 µm^2^ was applied to reach T1 = 2,293 K. The box was filled with water and the density variation induced by heat transfer drove the flow through buoyancy. The temperature distribution and fluid flow field surrounding the cube were computed using the laminar flow and conjugate heat transfer modules within the box.
Theoretical analysis
To explain the competition between inter-particle interactions and particle–fluid interactions during the assembly process, a model system in which SiO_2_ particles assemble in aqueous solutions with various NaCl concentrations was used. On the basis of our experimental observations (Fig. 2b), the particle suspension initially resided in a metastable dispersion state, in which the attractive van der Waals interaction outweighed the electrostatic repulsion. In this system, the work done by the particle–fluid interaction during an infinitesimal displacement ds is represented by
\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta {W}_{{\rm{f}}{\rm{l}}{\rm{u}}{\rm{i}}{\rm{d}}}=6{\rm{\pi }}\mu Rv\times ds$$\end{document}in which μ is the dynamic viscosity of the fluid, R is the particle radius and v is the flow velocity.
The difference in interaction potential between particles is represented by
\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta {U}_{{\rm{DLVO}}}={F}_{{\rm{DLVO}}}\times {ds}.$$\end{document}DLVO force (FDLVO) consists of two parts: van der Waals force (attractive force for the formation of colloidal clusters) and EDL repulsion force among particles.
The van der Waals force between particle and particle (FvdW–pp) can be described by
\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${F}_{{\rm{vdW}}-{\rm{pp}}}(h)=\frac{A}{6}\left(-\frac{4{R}^{2}(2R+h)}{{[(4R+h)h]}^{2}}-\frac{4{R}^{2}}{{(h+2R)}^{2}}+\frac{8{R}^{2}}{h(4R+h)(h+2R)}\right)$$\end{document}in which A is the Hamaker constant and h is the closest distance to the two-particle surface.
The EDL energy^50^ between particle and particle can be described by
\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${W}_{{\rm{E}}{\rm{D}}{\rm{L}}-{\rm{p}}{\rm{p}}}(h)={\rm{\pi }}{\varepsilon }_{0}{\varepsilon }_{{\rm{l}}}R{\zeta }^{2}\left[{\rm{l}}{\rm{n}}(1-{{\rm{e}}}^{-2\kappa h})+{\rm{l}}{\rm{n}}\left(\frac{1+{{\rm{e}}}^{-\kappa h}}{1-{{\rm{e}}}^{-\kappa h}}\right)\right]$$\end{document}in which ζ is the zeta potential on the solid–fluid interface, ε0 and εl are the dielectric permittivities of vacuum and water, respectively, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\kappa }^{-1}=\sqrt{\frac{{\varepsilon }_{0}{\varepsilon }_{{\rm{l}}}{K}_{{\rm{B}}}T}{2{N}_{{\rm{A}}}{{\rm{e}}}^{2}I}}$$\end{document} is the Debye length, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I=0.5\sum {z}_{{\rm{i}}}^{2}{\rho }_{{\rm{i}}}$$\end{document} , ρi is the ion concentration and zi is the number of charges carried by the ion.
The EDL repulsion force (FEDL–pp) is then described by
\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${F}_{{\rm{EDL}}-{\rm{pp}}}(h)=-\frac{{\rm{\partial }}{W}_{{\rm{EDL}}-{\rm{pp}}}(h)}{{\rm{\partial }}h}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${F}_{{\rm{E}}{\rm{D}}{\rm{L}}-{\rm{p}}{\rm{p}}}(h)=-2{\rm{\pi }}{\varepsilon }_{0}{\varepsilon }_{{\rm{l}}}R\kappa {\zeta }^{2}\left[\frac{{{\rm{e}}}^{-2\kappa h}}{1-{{\rm{e}}}^{-2\kappa h}}-\frac{{{\rm{e}}}^{-\kappa h}}{1-{{\rm{e}}}^{-2\kappa h}}\right]$$\end{document}in which h → 0
\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathop{{\rm{l}}{\rm{i}}{\rm{m}}}\limits_{h\to 0}\,{F}_{{\rm{E}}{\rm{D}}{\rm{L}}-{\rm{p}}{\rm{p}}}(h)={\rm{\pi }}{\varepsilon }_{0}{\varepsilon }_{{\rm{l}}}R\kappa 2{\zeta }^{2}\frac{3\kappa h}{2\kappa h}=3{\rm{\pi }}{\varepsilon }_{0}{\varepsilon }_{{\rm{l}}}R\kappa {\zeta }^{2}$$\end{document}Then the ΔUtotal can be described by
\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta {U}_{{\rm{t}}{\rm{o}}{\rm{t}}{\rm{a}}{\rm{l}}}=({F}_{{\rm{E}}{\rm{D}}{\rm{L}}-{\rm{p}}{\rm{p}}}+{F}_{{\rm{v}}{\rm{d}}{\rm{W}}-{\rm{p}}{\rm{p}}}+6{\rm{\pi }}\mu Rv)\times ds$$\end{document}When increasing the ionic strength of the system, the FEDL–pp is gradually screened. Therefore, colloidal particle clustering becomes more favourable.
The tendency of the system to form clusters can then be evaluated by examining changes in Utotal. When ΔUtotal < 0, the attractive interaction dominates in the fluid flow, leading to the cluster formation of colloidal particles; otherwise, the repulsive interaction dominates and particles would remain dispersed. The detailed information of all physical parameters used in this study, including their corresponding symbols and values, are summarized in Supplementary Table 4. The phase diagram in Fig. 2e showing the dispersing and clustering domains of colloidal particles can be obtained with equation (8) through MATLAB.
Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41586-025-10033-x.
Supplementary information
Supplementary InformationSupplementary Discussion 1, Supplementary Figs. 1–20, Supplementary Tables 1–4 and Supplementary References. Supplementary Data (source data Supplementary Figs. 5, 7, 8, 13, 17a and 20). Peer Review file Supplementary VideosSupplementary Videos 1–17.
Source data
Source Data Fig. 2 Source Data Fig. 3 Source Data Fig. 5