3D nanoprinting of metals by spatiotemporally confined hot electrons via multiple-electron excitations in nanocrystals Authors: Kai Wei, Kai Chang, Xinyi Wang, Shoufeng Lan, M Cynthia Hipwell, Heng Pan Published: 2026 ABSTRACT AbstractFor decades, multi-photon polymerization has been a key technology for fabricating complex 3D micro- and nanoscale structures from polymer materials. However, extending this capability beyond polymers remains a significant challenge. In particular, nanoscale metal printing is challenging because solid metal formation typically requires energy-driven reactions or thermally activated processes that are difficult to spatially localize, resulting in compromised feature resolution. Here, we demonstrate 3D nanoprinting of metals with depth and lateral resolution <250 nm. The method employs femtosecond laser-induced hot electrons spatiotemporally confined in nanocrystals to facilitate nonlinear multi-electron absorption, ligand desorption, and nanocrystal fusion. This approach operates at a pulse energy about 100× lower than simultaneous multi-photon processes, avoids organic additives, and is compatible with free-space or layer-by-layer printing. Printing of multiple metals is demonstrated, achieving mechanical strength comparable to pure metals, along with functional mechanical and optical metamaterials. This technology enables customizable 3D metal nanoprinting for advanced applications in metamaterials, biotechnology, nanorobotics, sensors, and semiconductor manufacturing.Subject terms: Nanoparticles, Laser material processing Introduction Metal 3D printing has rapidly evolved from a prototyping tool to a method for producing end-use components in the aerospace, medical, automotive, and energy industries1–3. However, achieving nanoscale resolution remains a challenge because it requires precise control over nucleation, growth, and coalescence of metal atoms. Many current techniques rely on additives, including polymers, resists, or ligands, to guide deposition and maintain structural integrity. Two-photon polymerization (2PP) of metal-containing polymers4,5, as well as 2PP combined with metallization6 have been developed to fabricate metallic nano-architectures. Laser-triggered assembly of inorganic nanocrystals (NCs) through ligand bonding allows printing 3D nanostructures7,8. Although these additives aid structure fabrication, they hinder pure metal formation and may limit achievable material properties unless removed through subsequent high-temperature pyrolysis. Alternatively, metals can be directly deposited from atomic or ionic precursors in polymer-free processes using focused energetic beams or electrically stimulated reactions. Focused electron beam-induced deposition (FEBID)9 can fabricate complex, maskless metal structures with sub-100 nm resolution. However, its slow growth rate and vacuum requirements limit scalability. Two-photon decomposition10 without polymer additives shows potential for 3D metal nanoprinting, however, an ultrafast laser pulse of 2–8 nJ is required. The required pulse energy is comparable to that used in two-photon polymerization11, which tends to induce excessive heating and crystal growth during metal printing. Localized pulsed electrodeposition12 enables metal printing at sub-micrometer resolutions with high precision and suitability for multi-metal fabrication. The pulsed process offers improved control compared to continuous electrodeposition. However, the resolution is constrained by nozzle size, and questions remain regarding its printing speed and ability to produce complex 3D geometries. Using ligand-free nanoparticles as larger building blocks simplifies the printing process compared to the approaches based on atomic or ionic precursors, while retaining the material versatility and purity. Charged nanoparticles have been assembled to 3D print multiple metals with feature sizes below 100 nm13. While this method offers high resolution, it relies on a mask, which limits its ability to fabricate arbitrary complex structures. Other nanoscale metal printing techniques that employ masks or molds can increase throughput. Nanotransfer printing (nTP) mechanically or chemically transfers metal films from a patterned mold to a target substrate with sub-20 nm resolution14,15. Multi-step nTP stacking has further enabled large-area 3D metal metamaterials16. Nanostencil lithography utilizes a freestanding membrane with nanoscale apertures as a physical shadow mask, enabling scalable deposition of metal nanostructures17–19. These nanoprinting strategies rely on pre-fabricated masks or molds, which constrain the process to predefined geometries and limit the ability to produce arbitrary free-form metal nanostructures. To increase printing throughput while preserving the flexibility of maskless approaches, various laser beam modulation strategies have shown promise, including multi-beam configurations20, structured light fields (e.g., optical vortex beams21), and spatiotemporal modulation22,23. However, these approaches have yet to be demonstrated for metal printing. In summary, nanoscale 3D metal printing is a rapidly evolving field that can be broadly categorized into mask-based approaches with high throughput and maskless approaches offering greater flexibility (Supplementary Table 4 summarizes the key attributes of representative processes). Within maskless approaches, achieving flexible (e.g., material versatility, arbitrary geometries, and low processing temperatures) and high-quality (e.g., high resolution and high purity) metal printing remains an active area of research, as it requires precise control over atomistic coalescence or nanoparticle assembly, while minimizing the use of polymers, additives, or excessive energy that could compromise printing quality. To this end, we introduce a high-precision metal printing scheme that employs spatiotemporally confined hot electrons to enable metal printing with low-pulse energy, providing an additive-free route to flexible and low-temperature nanoscale metal printing. Results Printing mechanism Femtosecond laser excitation of NCs produces a highly nonequilibrium distribution of electrons that are simultaneously confined in space by the NC dimensions and in time by the femtosecond excitation duration. Due to the spatial and temporal confinement, hot electrons are generated in rapid succession and repeatedly excited in metal NCs, leading to nonlinear optical absorption, ligand desorption, and localized fusion and deposition of uncapped NCs, which we term nanoprinting induced by multiple-electron transition (nPIMET) (Fig. 1a). Through sequential multiple-electron excitations, sufficiently energetic hot electrons can be formed that interact with ligand’s lowest unoccupied molecular orbital (LUMO) and induce ligand desorption. The resulting ligand desorption enables localized fusion of the metal NCs, achieving high printing resolution in the depth and lateral direction (Fig. 1e). In comparison, printing induced by linear absorption results in reduced resolution in depth due to the absence of spatially localized excitation (Fig. 1f). Multi-electron 3D nanoprinting mechanism. a Schematic illustration of the nPIMET. b Schematic illustration showing the structures of NC capped with ligands. c Schematic illustration showing re-excitation of hot electrons facilitates nonlinear absorption and ligand desorption. d Schematic illustration showing re-excitation of hot electrons facilitates printing with high resolution. e Schematic illustration showing printing induced by nonlinear absorption. f Schematic illustration showing printing induced by linear absorption. g Schematic showing OAm ligand-capped metal NCs and the mechanism of laser-induced ligand desorption. h SEM images and corresponding schematic representations of Au nanostructures printed using 343 and 515 nm wavelengths, highlighting differences in depth and lateral printing resolution. Printing experiments were independently repeated ≥5 times, yielding similar results. i Height of printed nanostructures fabricated at 343 and 515 nm wavelengths. j Width of printed nanostructures fabricated at 343 and 515 nm wavelengths. i, j Each data point (mean ± standard deviation) represents the average of n = 5 independently fabricated structures under identical conditions. The effect of wavelength on pillar height and width was considered statistically significant based on two-tailed Student’s t tests (p < 0.05). k Z-scan curves of gold NCs under 515 nm wavelength illustrating the nonlinear absorption behavior due to multi-electron absorption. Scale bars, h 500 nm. i, k Source data are provided as a Source Data file. To demonstrate this concept, we chose metal NCs capped with oleylamine (OAm) (Fig. 1g). OAm bonds to metal NCs through electron donation from its nitrogen atom24, forming a binding energy of <0.9 eV across a range of metals (Methods, Supplementary Fig. 12). The OAm-capped NCs of various metals were synthesized in-house or commercially purchased, and NC inks were prepared for printing (Methods, Supplementary Fig. 16). The printing process is depicted in Fig. 1a where the laser beam is focused into the NC ink that is sealed in a printing chamber. Laser wavelength selection is key to our approach. We evaluated the printing resolution using laser wavelengths of 343 and 515 nm. At the laser focal spot, the laser intensity reaches the threshold to induce ligand desorption and form uncapped NCs, which instantly fuse and deposit into nanostructures (Fig. 1b–d). In the case of 343 nm irradiation, after 30 s of laser exposure, the printed Au structure grows axially into a free-standing nanopillar exceeding 2.5 µm in height. In contrast, under 515 nm irradiation, the printing is highly localized, forming a confined nanodot in both the depth and lateral directions (Fig. 1h), suggesting high printing resolution. The height and width of the structures printed at 515 nm are smaller than those printed at 343 nm (Fig. 1i, j). Similar behavior is observed across different metals, including Ag, Pt, Cu, Ni, and Co (Supplementary Fig. 3), indicating that the high printing resolution is applicable for multiple metals. Highly localized printing using 515 nm wavelength originates from nonlinear optical absorption mediated by hot electrons and induced ligand desorption. Z-scan characterization shows pronounced nonlinearities in Au NCs at the 515 nm wavelength (Fig. 1k), where transient hot electrons drive intensity-dependent transitions from saturable absorption (SA) at the lower intensity region to reverse saturable absorption (RSA) at the higher intensity region. The RSA originating from the enhanced transient absorption by hot electrons in the conduction band before thermalization25–28 gives rise to a highly localized absorption region in the center of the laser spot. Due to highly localized absorption, printing is confined to the center of the laser spot and rapidly diminishes away from it, resulting in a confined voxel size, particularly in the depth direction. Time-dependent density functional theory (TDDFT) simulation (Methods, Supplementary Fig. 14) shows at a laser intensity 1 × 1011 W cm−2 (515 nm) the interval time between two successive hot electrons in a 2.5 nm NC is ≈2.5–25 fs (Ratehot electron = 4 × 1014−4 × 1013s−1), which is shorter than the electron-electron relaxation time τe ≈ 100 fs29,30 in metals, allowing for re-excitation of hot electrons, as evident by the RSA. We performed Z-scan measurements of other metal NCs (Cu, Ag, Pt, Ni, Co) and confirm RSA is observed at 515 nm while no observable nonlinear response is detected at 343 nm for laser intensities up to the threshold for printing initiation (Supplementary Fig. 5). The lack of an observable nonlinear response at 343 nm is likely attributable to the employed ink concentration, for which noticeable linear-absorption-induced printing is inevitable. Without nonlinear localization, the absorbed laser energy scales linearly with local laser intensity, leading to an elongated voxel in the depth direction. Transient exchange of photoexcited hot carriers between the metal and chemically bound molecules via hybridized electronic states leads to vibrational excitation of the molecules31 and subsequent reactions32, a phenomenon observed for both plasmonic33,34 and non-plasmonic35,36 metals. Under high-intensity photoexcitation, ligand desorption and fragmentation could occur37. Density functional theory (DFT) calculation indicates the hybridized states between LUMO of amine ligand and various metals are located ≈3–4 eV above Fermi level (Supplementary Fig. 13). Photoexcited nascent hot electrons, reaching energies of up to ≈3.0 eV (Supplementary Fig. 15) or possibly higher through re-excitation as evident by RSA, are transiently coupled to hybridized states, inducing nonlinear ligand desorption. This mechanism enables 3D nano printing of a wide range of metals with nearly isotropic feature resolution and at low-pulse energy (down to ≈25 pJ). If the pulse energy exceeds the threshold (≈25 pJ), a single pulse irradiation can deterministically induce NC fusion (Supplementary Fig. 21), suggesting that the multi-electron process is highly efficient and repeatable. Printing of structures To demonstrate the capability of 3D printing of nanostructures, we synchronized an x-y-z axis piezo stage with a 2D galvo scanner to fabricate arbitrary 3D objects. Both layer-by-layer stacking and free-space direct printing strategies were employed, providing flexibility in printing diverse structures and aligning with the methods commonly used in commercial 3D printers (Methods). We first performed 2D nanopatterning at a high speed of 100 μm s−1, generating a portrait over an area of 22.5 × 35 μm2 (Fig. 2a). The grayscale contrast observed in the SEM image was achieved by tuning the laser dwell time of each pixel location, modulating the printing of nano-dots. Additionally, the 2.5D pixel-art portrait was produced by arranging the space of freestanding nanopillars (Fig. 2b), each with an average diameter ≈288 nm (Fig. 2c). In addition, an array of 10 × 10 nanopillars was fabricated, showing consistent printing behavior (Fig. 2d). Both the nanopatterned features and 2.5D structures exhibit high spatial resolution and precise control over the printing process. Each nanopillar has a smooth surface and dense morphology, highlighting the robustness of the multi-electron nanoprinting approach. The printed features are able to withstand rinsing and air drying without loss or deformation, indicating strong mechanical stability. For 3D nanoprinting, typical parameters include laser powers of 50–270 µW, repetition rates of 1 kHz to 10 MHz, and scan speeds of 1–10 μm s−1 (Supplementary Table 2). A 4 × 4 array of spiral structures was printed, showing excellent uniformity across the array (Fig. 2e). Besides these nanostructured arrays, we fabricated complex volumetric 3D architectures to further evaluate the capabilities of the nanoprinting system. Examples include an Eiffel Tower-like structure (Fig. 2f), a hierarchical mechanical metamaterial (HMM) (Fig. 2g), a dome-shaped architecture (Fig. 2k, l), and a miniature Statue of Liberty (Fig. 2m). These structures are volumetric and preserve fine architectural details. A butterfly-like structure array (Fig.2h) and a C60-inspired structure and its array (Fig. 2i, j), containing many overhanging features, confirm the fidelity of printing arbitrary features. 2D and 3D multi-electron nanoprinting. The following prints all using gold NC as demonstration. a SEM image of the Mona Lisa portrait in grayscale (Mona Lisa by Leonardo da Vinci, c. 1503–1519; Musée du Louvre, Public Domain). b SEM images of Albert Einstein’s portrait in pixel art (Photo of Albert Einstein by Orren Jack Turner, 1947; courtesy of the Library of Congress, Public Domain). c SEM image of a freestanding nanopillar fabricated in the portrait. d SEM image of the 10 × 10 array of nanopillars. e SEM image of the 10 × 10 array of spiral structures. f SEM image of the 3D-printed Eiffel Tower-like architecture (the inset SEM image shows the top view). g SEM image of the HMM (the inset SEM image shows the top view). h SEM image of the butterfly-like structures. i, j SEM image of the C60-inspired structure and its array. k, l SEM image of miniature dome architecture. m SEM image of a miniature Statue of Liberty. Scale bars, a, b, f, g, i, k–m 10 μm; d, e 5 μm; c 500 nm; h, j 20 μm. a–m All printing experiments were independently repeated ≥2 times, yielding similar results. Multiple-material printing We fabricated complex boat-shaped structures with multiple metals to investigate the versatility and reliability of the 3D nanoprinting system. Each structure has dimensions of 15 × 7.7 × 21 μm3 and is printed in Au (Fig. 3a, g, m), Pt (Fig. 3b, h, n), Ag (Fig. 3c, i, o), Cu (Fig. 3d, j, p), Ni (Fig. 3e, k, q), and Co (Fig. 3f, l, r). The boat structure has diverse geometrical features, such as overhanging structures, sharp curvatures, and 90° corners, serving as benchmarks for evaluating printing performance across different metals. Printing parameters were individually optimized for each metal to achieve high-fidelity structures with minimal defects (Supplementary Table 3). Also, the laser pulse energy required for printing all metals is low, ranging from 45 to 85 pJ, indicating highly efficient ligand desorption and nanoparticle fusion. Multi-metal 3D nanoprinting of microstructures. a–l SEM images of the boat structure using Au, Pt, Ag, Cu, Ni, and Co, respectively. m–r EDS mapping images of the boat structure using Au, Pt, Ag, Cu, Ni, and Co, respectively. Scale bars a–r 5 μm. All printing experiments were independently repeated ≥2 times, yielding similar results. Mechanical properties, electrical properties, and density Confined re-excitation of hot electrons initiates the low-temperature printing process, resulting in minimal thermal accumulation within the printed structures. By avoiding thermal perturbation during printing, metallic NCs can be fused into dense features. To demonstrate the robust mechanical properties of the printed structures, we fabricated a body-centered cubic Au HMM architecture and conducted SEM-based in-situ quasi-static uniaxial compression testing. Functional mechanical metamaterials require reproducible and scalable fabrication while supporting complex overhanging and intertwined features at the micro- and nanoscale. Localized distortion of the unit cell or defects could largely reduce the recovery effect. A sequence of SEM images captured the deformation process of the HMM during compression (Fig. 4a–c). At maximum compressive displacement, the architecture shows significant deformation by localized buckling and distortion of the unit cells, while supporting overall structural integrity. During unloading, the architecture releases the stored elastic energy, recovering its geometry without local fracturing or collapsing, indicating resilience of the HMM. The corresponding force–displacement curve (Fig. 4d) shows a peak compressive strain of 25%, followed by 40.7% recovery, reflecting high consistency of the printing and the dense structural integrity. Extrinsic factors such as printing quality and strain rate, together with intrinsic factors including metal type, unit-cell type of the HMM, and feature size, govern the mechanical performance of the structures. To isolate the intrinsic behavior of the printed metals, we fabricated single nanopillars of metals and performed compression testing to reveal their fundamental mechanical properties. Take an Ag nanopillar as an example, the initial and final stage of compression process is captured in Fig. 4e, f. The images revealed obvious plastic deformation without fracture of the nanopillar. The top region was flattened, indicating localized yielding, while the pillar base remained intact. We extended the compression tests to other metals, including Au, Pt, Cu, Ni, and Co. All metallic nanopillars maintained their structural integrity after compression, consistent with the results from stress–strain curves (Fig. 4g). The curves show that the nanopillars initially undergo elastic deformation up to a strain of ≈0.04, followed by plastic deformation. Co reaches the highest peak stress among all metals tested, up to ≈2.4 GPa, while Au exhibits the lowest, at ≈340 MPa. These results are consistent with the bulk mechanical behavior of these metals. The maximum compressive stress range is comparable to the values reported for nanopillars fabricated using alternative techniques38–40. To better resolve the mechanical properties of the printed metals, the Young’s modulus was measured. The measured Young’s modulus for printed Au, Ag, Pt, Ni, Cu, and Co is 33.3, 22.3, 86.0, 65.5, 39.5, and 71.6 GPa, respectively, corresponding to 45%, 26%, 51%, 33%, 31%, and 34% of their respective bulk values (Fig. 4h). The printed structures were constructed from fused NCs, which contain high density of grain boundaries. High density of grain boundaries may promote intergranular shear or local grain-boundary sliding under loading, acting as compliant layers during deformation and reducing the Young’s modulus41. Mechanical density is another factor influencing the overall mechanical performance. Focused ion beam cutting of the overhanging regions provided direct insight into the internal density. The cross-sectional cut revealed no voids or porosity, confirming that the printed feature is dense (Supplementary Fig. 11). To benchmark the mechanical strength enabled by this technology, Fig. 4i compares the yield strength of metal nanopillars fabricated using various micro- and nano-metal 3D printing techniques, including 2PP-based approaches. Ni nanopillars fabricated via conventional 2PP followed by high-temperature pyrolysis have demonstrated high yield strengths in the range of 1–3 GPa42. Electrochemical additive manufacturing (EAM) enables the fabrication of multi-metal structures with feature sizes down to ≈210 nm. Cu nanopillars produced by EAM have shown yield strengths up to ≈2 GPa12,43,44. Similarly, Pt micropillars fabricated via focused electron/ion beam-induced deposition (FEBID/FIBID) exhibit yield strengths >3 GPa, attributed to their dense microstructure45. In addition, nanoimprinted Ag nanopillars with feature sizes ≈150 nm have demonstrated yield strengths above 0.8 GPa46. This developed technology is capable of printing various metals with feature sizes ranging from 170–700 nm and yield strength ranging from 232 MPa to 2.64 GPa. Au nanopillars with a diameter of 170 nm show the highest yield strength (up to 2.64 GPa), which is comparable to 2PP-printed Ni nanopillars. Co, Cu, Ni nanopillars with a feature size of 700 nm show yield strength ranging from 480 to 805 MPa, comparable to EAM-printed Cu nanopillars. In addition to mechanical integrity, electrical performance is also critical for expanding the utility of printed metals. The measured electrical conductivity reaches ≈32% of the bulk value, which is sufficient for most device-level integration scenarios (Methods). Residual carbon is a common concern in organic–inorganic printing systems. By employing a printing process based on ligand desorption instead of relying on polymer additives, the carbon content in the printed structures is significantly reduced. Raman spectroscopy confirms the absence of detectable carbon residue in our printed structures, highlighting the high purity of printed metals (Supplementary Fig. 7). Energy-dispersive X-ray spectroscopy (EDS) analysis further shows the printed structure contains negligible oxygen, indicating a non-oxidized printing process (Supplementary Fig. 17). The concentric rings observed in the selected-area electron diffraction (SAED) pattern, combined with the lattice fringes visible in high-resolution transmission electron microscopy (HRTEM) images, confirm that the printed structures are nanocrystalline and composed of metal (Supplementary Fig. 18). Mechanical properties of printed structures. a–c The HMM structure with dimensions of 10 × 10 × 12 μm3. SEM snapshots show the in-situ compression test: a the initial segment of preloading before contact. b At the segment of reaching maximum displacement. c The final segment of unloading after detaching. d Force-displacement curve of printed HMM. SEM images of Ag nanopillars e before and f after the compression test. g Stress-strain diagram of metals (Au, Ag, Pt, Ni, Cu, and Co). The stress values were calculated as σ = F/A, where F is the applied force, and A is the initial cross-sectional area of the nanopillar measured from pre-compression SEM images. h Young’s modulus value comparison of the printed metals (square shape) and bulk materials (solid line). i Yield strength versus feature size for micro- and nano-printed pillars, comparing this technology with other techniques. Scale bar, a–c 10 μm; e, f 1 μm. a–c, e, f Experiments were independently repeated three times, yielding similar results. For d, g–i Source data are provided as a Source Data file. Optical properties of printed structures The 3D nanoprinting not only produces functional mechanical metamaterials but also enables high-resolution fabrication of optical metamaterials. Optical metamaterials can manipulate the incident electromagnetic field by light-matter interaction of structures, tuning the light response. The geometrical features of the structures strongly govern their interaction with the incident electromagnetic field. Particularly, nanospirals break optical symmetry and enable polarization control by inducing responses to left- and right-handed circularly polarized light (LCP and RCP). Fabrication of metal nanospirals has been reported, showing chiral dichroism (CD) by using focused-ion/electron beam induced deposition (FIBID/FEBID)47–50, glancing-angle deposition (GLAD)51–53, and two-photon polymerization combined metal deposition (2PP+metallization)54–58. FIBID/FEBID remains fundamentally constrained by its reliance on conductive substrates and vacuum operation. Also, the inherent slow deposition rate limits scalability, making high-throughput fabrication challenging. 2PP+metallization enables the fabrication of a larger area of spiral array. However, the feature size is limited to ≈500–1000 nm, and anisotropic resolution of the printing is difficult to avoid. We fabricated 15 × 15 Au nanospirals with a feature size diameter of 220 nm (Fig. 5a, b). Benefiting from the spatial and temporal confined electronic excitation, nearly isotropic resolution of printed features can be achieved. This developed technology offers higher resolution than 2PP+metallization, extending the operational wavelength range into the near-infrared (NIR) regime that remains inaccessible to 2PP+metallization methods, while exhibiting lower resolution than FIBID/FEBID and GLAD (Supplementary Fig. 22a). Moreover, the nanospiral arrays show uniformity across the sample, maintaining coherence in the measured chiral optical response. The Au nanospiral geometry was extracted from SEM measurements and applied to a simulation model (Fig. 5c). Transmission spectra reveal a broadband chiral optical response ranging 800–1500 nm (Fig. 5d). In particular, the nanospiral array shows a clear circular dichroism between 1100 and 1300 nm, where the LCP transmission exceeds the RCP transmission by over 20%. The simulated spectra show a similar transmission contrast but produce a narrower chiral response band, confined to 1250–1350 nm. The circular dichroism was quantified using the dissymmetry g-factor calculated from the transmission spectra, with values ≈ 0.17 (Fig. 5e). Consistent with the transmission results, the calculated g-factor demonstrates a broadband response, indicative of strong chiroptical activity. The simulated g-factor shows sharper and more distinct peaks, which may be attributed to the artificially narrow resonance distribution in the idealized model, leading to an amplified response within a restricted wavelength range (Fig. 5f). In addition, idealized infinite periodicity and perfectly uniform geometry applied in the simulation minimized inhomogeneous broadening that is present in the experimental sample. Overall, the experimentally obtained g-factor trends closely match the simulation, confirming that the printed nanospiral array effectively manipulates circularly polarized light. We benchmark the chiral response of our structures against chiral nanostructures fabricated using alternative approaches (Supplementary Fig. 22b). The g-factor achieved in this work is comparable to that obtained via GLAD at similar visible-to-NIR wavelengths. While 2PP+metallization can attain higher g-factors (up to 1.73), the operation is typically limited to longer infrared wavelengths beyond 3 µm. Optical measurement of printed structures. a SEM of a printed Au nanospiral array. b Zoom-in SEM of printed Au nanospirals. c 3D model of the Au nanospiral with geometrical dimensions (H = 4 μm, D = 1 μm, and d = 220 nm). d Optical transmission spectrum of printed structures, showing LCP and RCP curves. The measured data at 1050–1100 nm have been removed due to light source saturation. e Measured and f simulated g-factor of nanospiral array. Scale bar, a 5 μm; b 1 µm. a, b Printing experiments were independently repeated two times, yielding similar results. d–f Source data are provided as a Source Data file. Nanoprinting induced by multiple-electron transition enables the fabrication of arbitrary and complex 3D metal structures. The spatiotemporal confinement of photoexcited hot electrons induces strong nonlinear absorption in metal NCs, resulting in high-resolution printing of multiple metals with picojoule-level pulse energy. The demonstrated low-pulse-energy metal printing provides a pathway toward scalable, multi-beam, maskless metal printing and establishes a foundation for its development as a nanoscale prototyping platform. The printed architectures exhibit promising mechanical and optical properties, demonstrating diverse functionalities derived from their 3D geometries. This technology enables customizable 3D metal nanoprinting for advanced applications in metamaterials, biotechnology, nanorobotics, sensors, and semiconductor manufacturing. Methods Materials Gold (Au), silver (Ag), and platinum (Pt) NCs inks were purchased from UT Dots Inc (Champaign, IL, USA). Xylene (>75%), toluene (>99.5%), pure 200 proof ethanol (>99.45%), oleylamine (OAm >70%), oleic acid (OA 90%), tert-butylamine borane complex (TBAB > 97%), triphenylphosphine (PPh3 > 99%), copper (II) chloride (CuCl2 99%), cobalt(II) chloride hexahydrate (CoCl2 · 6H2O 98%), nickel(II) chloride hexahydrate (NiCl2 · 6H2O > 98%), lithium borohydride (LiBH4 90%), sodium borohydride (NaBH4 > 98.0%) were purchased from Sigma-Aldrich. Synthesis of metallic NCs The NCs synthesis was performed under a nitrogen atmosphere in the glove box. All the chemicals were handled in a glove box and used as received without purification or degassing. The commercially purchased Au, Ag, and Pt NCs ink was diluted to 0.2 wt.% concentration with xylene for 3D printing. The Cu NCs were synthesized by following a recipe reported in ref. 59. In brief, 7.42 g of PPh3 dissolved in 100 mL ethanol under stirring and heated at 90 °C until forming a homogeneous solution. 0.9 g of CuCl2 was added to the solution, yielding a heterogeneous mixture. The mixture was stirred for 10 min and cooled to room temperature, white powders of CuCl(PPh3)3 were obtained. The powders were centrifuged at 1000 × g for 10 min and washed with ethanol three times, further dried in the glove box as the precursor. 0.25 mmol of CuCl(PPh3)3 precursor was dissolved in 25 mL of toluene in a round flask and heated to 50 °C with heating mantle. Until the temperature stabilized, 0.52 mL of OAm was injected rapidly into the solution under vigorous stirring. Meanwhile, a second solution is prepared by dissolving 2.5 mmol of TBAB complex in 15 mL toluene, then heated to 50 °C. TBAB complex was further injected into the first solution at a slow rate of 1 mL min−1 to minimize bursting nucleation and temperature fluctuation. During the reaction, the color of the mixture gradually turned from yellow to red, indicating the formation of Cu NCs. Keep stirring the mixture for 1 h after the injection to ensure full reaction, then cool down to room temperature. As the temperature of the mixture reached room temperature, 0.4 mL of OA was injected and stirred for 4 min to strengthen the stability of Cu NCs in solution. To remove excess OAm and other impurities, the mixture was washed with ethanol, centrifuged at 1000 × g for 10 min, and then the supernatant was removed. The centrifugation and washing process was repeated 10 times. Finally, the resulting reddish precipitate was collected and dispersed in toluene, yielding a solution with a concentration of 0.2 wt.%. The Ni and Co NCs were synthesized by following a recipe reported in ref. 60. For the synthesis of Ni NCs, 1.2 g of NiCl2 · 6H2O and 5.35 g of PPh3 were added in 50 mL of ethanol. In a separate flask, 0.133 g of LiBH4 was dissolved in 10 mL of ethanol and then slowly injected into the NiCl2/PPh3 solution under vigorous stirring. The reaction was held for 2 h, and the orange precipitate of ClNi(PPh3)3 complex formed. The precipitate was centrifuged at 1000 × g and washed with ethanol three times, further dried in the glove box for use as the precursor. For the thermal decomposition step, 0.25 mmol of ClNi(PPh3)3 complex precursor was mixed with 10 mL of OAm in a round flask, heated from 25 to 190 °C over 1 h. The color of the solution turned from orange to black, indicating Ni NCs formed. The mixture was washed with ethanol, centrifuged at 1000 × g for 10 min, and then the supernatant was removed. The centrifugation and washing process was repeated three times. Finally, the resulting black precipitate was collected and dispersed in toluene, yielding a solution with a concentration of 0.2 wt.%. For the synthesis of Co NCs, 17 mmol of CoCl2 · 6H2O and 40 mmol of PPh3 were added in 200 mL of ethanol. Then, 10 mmol of NaBH4 was gradually added to the mixture in 2 h with vigorous stirring. After the reaction, the brown precipitate of ClCo (PPh3)3 complex formed. The precipitate was centrifuged at 1000 × g for 10 min and washed with ethanol three times, further dried in the glove box for use as the precursor. For the thermal decomposition step, 0.25 mmol of ClCo (PPh3)3 complex precursor was mixed with 10 mL of OAm in a round flask, heated from 25 to 120 °C over 30 min. The color of the solution turned from brown to black, indicating Co NCs formed. The mixture was washed with ethanol, centrifuged at 1000 × g for 10 min, and then the supernatant was removed. The centrifugation and washing process was repeated three times. Finally, the resulting black precipitate was collected and dispersed in toluene, yielding a solution with a concentration of 0.2 wt.%. The photo of synthesized metallic NC inks is shown in Supplementary Fig. 16. 3D nanoprinting setup and printing process Figure 1a and Supplementary Fig. 9 illustrate the 3D nanoprinting system setup. The custom 3D nanoprinting setup consists of a laser with 300 fs pulse duration, tunable repetition rate from 1 Hz to 40 MHz, and nonlinear optics to achieve green (515 nm) and UV (343 nm) wavelength outputs (Satsuma HP2). The laser beam passes through a mechanical shutter (Thorlabs, SH05R), a custom beam expander, a galvo scanner, then focused on the sample by an oil-immersion objective (Olympus, UPlanFL N ×40, NA = 1.3). The galvo scanner is synchronized with the z axis piezo stage (Thorlabs, LPS710E) and the shutter to determine the laser-beam path in 3D space. A CMOS camera with 10 megapixels (Amscope, MU1003) and an LED illuminator are aligned with the laser beam path for monitoring the printing in real time. To prepare the sample, the metallic NCs ink was dropped on the substrate, sealed in a custom chamber with an adhesive spacer (SecureSeal, thickness = 120 μm) and a cover glass to avoid fast evaporation of the solvent. The printing process can be effectively implemented with an ink concentration range of 0.2–3 wt.%. We have conducted printing studies using ink concentrations up to 3 wt.%. Compared to low concentrations, higher NC ink concentrations require the use of a thinner spacer to ensure that the laser beam can penetrate the ink layer and reach the substrate surface to initiate printing. All printing results in this work were obtained using 0.2 wt.% unless otherwise specified. Multi-electron 3D nanoprinting is a substrate-independent technology. It is compatible with substrate materials such as glass, silicon, ITO-coated glass, quartz, and sapphire. However, the laser threshold required to initiate the printing process can vary depending on the substrate type. The adhesion strength of the printed structures is affected by the substrate materials. The laser parameters typically include a repetition rate between 1 kHz and 10 MHz, power ranging from 0.05 to 0.27 mW, and a laser scan speed of 1–10 μm s−1. The combination of these parameters must be optimized based on the structure size and the type of metal to ensure high-quality 3D printing. Using power above this range can cause localized thermal accumulation, resulting in increased surface roughness or damage to the printed features. On the other hand, using power below the optimal level may fail to excite enough NCs for fusion, leading to discontinuous printing. Notably, the minimum power required for printing can be as low as 50 μW at 2 MHz, equal to a pulse energy of 25 pJ, which is beneficial for employing multi-beams to scale up the process. After printing, the sample is immersed in a pure, volatile nonpolar solvent such as toluene or xylene. The chamber is then opened to allow any unfused NC ink to disperse into the solvent. For 2D nanopatterns and 2.5D structures, the samples are rinsed several times with the solvent to remove residual ink and then dried in air. For certain complex 3D nanostructures, critical point drying is recommended to minimize surface tension and prevent collapse of overhanging features. In this case, the samples are first rinsed with 100% nonpolar solvent and then gradually exchanged with ethanol until only ethanol remains. Once fully immersed in ethanol, a standard critical-point drying protocol is applied. The printing can be applied to both layer-by-layer and free-space printing modes with comparable printing quality. The selection between the two approaches depends on the structural requirements of the target geometry. For structures with complex or highly arbitrary geometries, the layer-by-layer mode is generally more suitable. This approach allows greater flexibility in toolpath planning and printing strategy, making it easier to design and fabricate intricate features. The free-space printing mode is suitable for regular or continuous structures, such as spiral arrays or other periodic geometries. One advantage of the free-space method is that it can provide higher positional precision, which helps maintain geometric consistency in such structures. The in-situ compression test of the printed HMM (Fig. 4a–c) was tested in the as-printed state without post-treatment. The printed nanopillars (Fig. 4g) were post-sintered at 350 °C for 5 min prior to testing. Young’s modulus was estimated from the unloading segment of the stress–strain diagram. Optical simulation of Au nanospiral arrays The transmission spectra and g factors were computed using the commercial software Ansys Lumerical FDTD-3D Electromagnetic Solver. The gold nanospiral chiral structures were modeled from cross-sectional slices, with a width of 1 µm, a height of 4 µm, and a wire diameter of 260 nm. These helices were arranged on the substrate in a square lattice with a period of 1.7 µm. The substrate consisted of a 55 nm gold film on top of a semi-infinite SiO2 layer. The excitation source was a linearly polarized plane wave incident from the SiO2 substrate onto the structure. Circular polarization was generated by combining two co-propagating orthogonal linearly polarized beams with relative phase shifts of +90° and –90°, corresponding to left- and right-handed circular polarization, respectively. A discrete Fourier transform monitor was positioned in the vacuum region above the chiral structures to record the transmitted intensity. The FDTD simulation domain was defined as a single unit cell, with periodic boundary conditions along the x and y directions, and perfectly matched layers along the z direction. After obtaining the LCP and RCP transmission spectra, the g-factor spectrum was calculated using the relation where TLCPλ and TRCP(λ) represent the simulated transmission intensity of left- and right-handed circular polarization, respectively. The resulting LCP and RCP transmission spectra, together with the g-factor spectra, are presented in Fig. 5d–f. Z-scan measurement Z-scan measurement setup consists of the femtosecond laser as a given energy source, a plano-convex lens (f = 50 mm) to focus the laser beam, and a photodiode detector to record laser power (Supplementary Fig. 4). The sample is fixed at the translational stage, moving along the z-direction to receive different laser intensity levels. The detector collects the transmitted laser intensity at different z-positions of the samples. Hence, the transmitted laser intensity versus the z-position curve can be recorded. The laser source is operated at a repetition rate of 10 kHz, 300 fs pulse width, 343 nm and 515 nm wavelength. The samples are prepared by dispersing metallic NCs in xylene or toluene, which is contained in the glass cuvette with a thickness of 1 mm. The ink concentration is selected to ensure the linear transmission of the ink is ≈80% at 515 nm wavelength for all materials. The translational distance is 30 mm with a speed of 0.15 mm s−1, and the focal point is at the center. The normalized transmission was calculated by dividing the measured intensity by the intensity measured at z = −15 mm. The z-scan curves measured from metallic NC dispersion (Au, Ag, Pt, Cu, Ni, Co) under different laser pulse energies are shown in Fig. 1g and Supplementary Fig. 5. Characterization techniques SEM and EDS were performed using a Tescan FERA-3, operated at 10 kV. Focused ion-beam milling was conducted with a probe current of 1.0 nA to obtain clean cross-sectional surfaces. In-situ SEM compression testing of metal nanopillars and mechanical metamaterials was carried out using an indenter system (Hysitron PI-85 Picoindenter) under 10 kV acceleration voltage. Flat-end tips with diameters of 5 or 20 µm were selected depending on the specimen dimensions. A constant loading/unloading strain rate of 0.05 s⁻¹ was applied. HRTEM and SAED patterns were performed using a Titan Themis 300 S, operated at 300 kV. SEM image analysis (dimensional measurements) was performed using ImageJ. Raman spectroscopy was performed (Horiba Jobin-Yvon LabRam HR) by using a 785 nm laser and a ×50 objective to focus on the surface structure. Transmission spectra of metal NC inks, including plasmonic (Au, Ag) and non-plasmonic (Pt) materials, were measured by a UV-Vis-NIR spectrophotometer (Hitachi U-4100) and shown in Supplementary Fig. 6. Electrical conductivity of the printed Au structures was measured using a two-probe configuration. Bridge-like Au structures were printed across two isolated Au pads (Supplementary Fig. 8), where the printed structure acted as the conductive interconnect and the pads served as electrodes. The electrodes were formed by spin-coating Au NC ink at 1000 rpm onto the substrate, followed by sintering at 180 °C for 5 min to yield a continuous Au film, which was then patterned into two pads using a 343 nm laser. After printing, the specimen was post-sintered at 350 °C for 5 min to improve electrical contact between the structure and the pads. Two conductive tungsten probes (Micromanipulator) were attached to the pads and connected to a digital sourcemeter (Keithley 2401). The conductivity σ was calculated from the measured resistance R using σ = L/(R×A), where L is the structure length, and A is the cross-sectional area. To ensure accurate extraction of the printed structure’s intrinsic conductivity, the resistance contribution from the Au pad and pad-probe contact was measured and subtracted. The resulting conductivity reached ≈33% of the bulk Au value. The chiral response was measured using a custom optical setup (Supplementary Fig. 10). A non-polarized supercontinuum white-light source (SuperK COMPACT, NKT Photonics) was circularly polarized using a linear polarizer followed by a quarter-wave plate. By rotating the fast axis of the quarter-wave plate to ±45° with respect to the polarizer, left-handed or right-handed circular polarization was obtained. An inverted ×40 objective (NA = 0.6) was used in combination with a pinhole behind the objective to block highly divergent beams, ensuring that only light interacting with the printed structures was collected. The collection cone was limited to ≈2° full angle. The transmitted circularly polarized light was analyzed using a spectrometer (SpectraPro HRS-300, Teledyne) equipped with a NIR detector (PyLoN-IR, Teledyne). The transmission spectra under left- and right-handed circularly polarized illumination were first recorded. The transmittance spectra were obtained by normalizing the transmitted light intensity to the incident light intensity measured in the absence of the sample. The CD g-factor was then calculated using g = 2(IL-IR)/(IL + IR). Where IL and IR represent the transmitted intensities for left- and right-handed circular polarization, respectively. To confirm that the observed CD exclusively comes from chiral effects, the linear dichroism of the sample was also characterized and found to be negligible. DFT and TDDFT simulations To determine the binding energy of OAm with various metals, DFT calculations were performed using the GPAW code, a real-space grid-based DFT implementation with the projector-augmented wave method. To model the adsorbate (ligand)-metal complex, a methylamine molecule was adsorbed to the top site of the (111) surface of metals, as shown in Supplementary Fig. 12. The (111) surface was modeled by a 3 × 3 × 2 slab with 15 Å of vacuum space using 13 irreducible k-points and a grid spacing of 0.2 Å. Spin-polarization was included in all the calculations. A finite temperature Fermi function (kBT = 0.1 eV) was utilized to facilitate the SCF convergence by smearing the band occupation. The exchange-correlation interaction was approximated using the PBE function. The potential energy surfaces (PES) of methylamine on metal were obtained by calculating the free energy as a function of the molecule–metal separation, specifically the N-metal distance. The energy levels HOMO and LUMO of the methylamine molecule and their hybridization with metal atoms are most relevant in understanding the adsorption and desorption behaviors. The total density of states (DOS) of methylamine-metal complex is shown in Supplementary Fig. 13 along with the projected density of states (PDOS) of HOMO and LUMO of the methylamine molecule. The generation of hot electrons in NCs and their energy distribution upon laser irradiation can be calculated by TDDFT32,61 using the GPAW code. Model systems consist of octahedral M489 NCs (M: Au, Cu, Ag, Pt, Ni, Co). DFT ground state calculations are performed using linear combination of atomic orbitals (LCAO) basis sets; the p-valence basis is used to represent Au, Cu, Ag, Pt, Ni and Co. The GLLB-sc XC functional was used for Au, Cu, Ag, and the PBE XC functional was used for Pt, Ni, and Co. A Gaussian pulse of frequency 2.5 eV (wavelength of 500 nm) with a 10-fs pulse width (5 fs FWHM) and peak electric field (5.13 × 103 V cm−1) is used to excite the NC (Supplementary Fig. 14). The peak electric field corresponds to a peak intensity of 3.51 × 104 W cm−2, and the specified pulse shape corresponds to an averaged intensity of 1.75 × 104 W cm−2 over 10 fs pulse duration. The hot electron generated by the pulse in the NC can be calculated by following the procedure described in refs. 32,61. The number of hot electrons is plotted in Supplementary Fig. 14. It can be seen that with the averaged intensity of 1.75 × 104 W cm−2 over 10 fs pulse duration, the total number of hot electrons generated per NC is 6.45 × 10−7, 3.2 × 10−7, 7.15 × 10−8, 2.27 × 10−7, 2.85 × 10−7, 1.94 × 10−7 for Au, Cu, Ag, Pt, Ni, Co, respectively, as evaluated at ≈18 fs. By scaling up to the intensity of 1.0 × 1011 W cm−2, the hot electron generation rate is 3.66 × 1014 s−1, 1.83 × 1014 s−1, 4.08 × 1014 s−1, 1.29 × 1014 s−1, 1.62 × 1014 s−1, 1.1 × 1014 s−1 for Au, Cu, Ag, Pt, Ni and Co. The interval time between two hot electrons is 2.7 fs, 5.4 fs, 24.4 fs, 7.69 fs, 6.14 fs, 9.03 fs for Au, Cu, Ag, Pt, Ni and Co. To obtain the energy distribution of the hot electron above the Fermi level, the number of electrons and holes is evaluated after a 10-fs laser pulse with an averaged laser intensity of 1.75 × 104 W cm−2 are shown in Supplementary Fig. 15. Hot electrons with energy up to ≈3.0 eV can be formed above the Fermi level. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Reporting Summary Transparent Peer Review file Source data Source data Author contributions Conceptualization: H.P. Methodology: H.P., K.W., C.H. Investigation: K.W., K.C., X.W., S.L., C.H., H.P. Visualization: K.W., H.P. Funding acquisition and project administration: H.P., C.P. Supervision: H.P., C.P., S.L. Writing–original draft: KW. Writing–review & editing: K.W., X.W., S.L., C.H., H.P. Peer review Peer review information Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available. Funding The authors gratefully acknowledge support from the National Science Foundation CAREER program under grant 2054098 (H.P.) and the National Science Foundation grant 2213693 (H.P., C.H.). Data availability The data that support the findings of this study are available from the corresponding authors upon request. Unprocessed raw data are provided as Supplementary Data 1. Source files used as printing references in this study were obtained from publicly accessible repositories, and their origins and reproduction rights are detailed as follows: The C60 model in Fig. 2i, j and Supplementary Fig. 11 was obtained from “Bucky Ball” by Thingiverse user Fredbartels (https://www.thingiverse.com/thing:579591), licensed under CC BY-SA 4.0, and used without modification. In accordance with the ShareAlike provision of the source license, Fig. 2i, j and Supplementary Fig. 11 are correspondingly released under CC BY-SA 4.0 and used without modification. The miniature dome architecture model in Fig. 2k, l was obtained from “Radcliffe Camera” by Thingiverse user JC_Workshop (Johnny Chan; https://www.thingiverse.com/thing:3499107), originally licensed under CC BY-NC 4.0. The original author has granted the authors specific written permission for use in this publication and release under CC BY 4.0. The model was used as provided without modification. The 3D Statue of Liberty model in Fig. 2m was obtained from Sketchfab (3Dystopia, https://sketchfab.com/3d-models/statue-of-liberty-04a8e7b1ba9142ba8cebc6108cd5bb97), licensed under CC BY 4.0. The 3DBenchy reference model in Fig. 3a–r was obtained from the official source (https://www.3dbenchy.com). The model was used as provided without modification. Source data are provided with this paper. Competing interests A patent application has been filed related to this work. Patent applicant: the Texas A&M University System. Name of inventors: Heng Pan, Cynthia Mary Hipwell, Kai Wei, and Kai Chang. International patent application no.: PCT/US2025/041442. Status of application: pending. Location and institution: Texas A&M University, College Station, TX. Specific aspects covered in the patent application: claims are directed towards the nanoprinting system, materials, and the method of manufacturing. Apart from this, the authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information M. Cynthia Hipwell, Email: cynthia.hipwell@tamu.edu. Heng Pan, Email: hpan@tamu.edu. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-74926-9. References Associated Data Supplementary Materials Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Reporting Summary Transparent Peer Review file Source data Data Availability Statement The data that support the findings of this study are available from the corresponding authors upon request. Unprocessed raw data are provided as Supplementary Data 1. Source files used as printing references in this study were obtained from publicly accessible repositories, and their origins and reproduction rights are detailed as follows: The C60 model in Fig. 2i, j and Supplementary Fig. 11 was obtained from “Bucky Ball” by Thingiverse user Fredbartels (https://www.thingiverse.com/thing:579591), licensed under CC BY-SA 4.0, and used without modification. In accordance with the ShareAlike provision of the source license, Fig. 2i, j and Supplementary Fig. 11 are correspondingly released under CC BY-SA 4.0 and used without modification. The miniature dome architecture model in Fig. 2k, l was obtained from “Radcliffe Camera” by Thingiverse user JC_Workshop (Johnny Chan; https://www.thingiverse.com/thing:3499107), originally licensed under CC BY-NC 4.0. The original author has granted the authors specific written permission for use in this publication and release under CC BY 4.0. The model was used as provided without modification. The 3D Statue of Liberty model in Fig. 2m was obtained from Sketchfab (3Dystopia, https://sketchfab.com/3d-models/statue-of-liberty-04a8e7b1ba9142ba8cebc6108cd5bb97), licensed under CC BY 4.0. The 3DBenchy reference model in Fig. 3a–r was obtained from the official source (https://www.3dbenchy.com). The model was used as provided without modification. Source data are provided with this paper.