3D Printing of glasses with tunable UV–VIS–IR photoluminescence via low-temperature nanoscale engineering

Abstract

Recent breakthroughs in three-dimensional (3D) printing of glass have enabled precise shaping of glass structures from macro to microscale while retaining their excellent physical and chemical properties. However, current functional developments in 3D printed glass primarily focus on leveraging its intrinsic optical transparency, with limited exploration of photonics application. 3D-printed glass functionalized by quantum dots (QDs) will hold great promise for creating desired photonic functional macro- and microstructures. However, voxel-precision customization of photonic functions in 3D printed glass by the monodispersed and size-controlled QDs is fundamentally conflict with the high-temperature sintering process required in 3D printed glass. Here, we present a simple, cost-effective method for creating a library of QDs functionalized 3D-printed glasses with tunable UV-VIS-IR photoluminescence (PL) by printing nanoporous glass (NG) followed by low-temperature nanoscale engineering of QDs growth. The integration of uniform and size-controlled QDs imparts 3D printed glass with fine tuning in composition, architecture, and photonic functionalities, while nanoscale engineering of the micro-environment enables precise tailoring of various QDs with enhanced optical performance and stability. This universal approach establishes a benchmark for 3D-printed functional glasses with advanced photonic capabilities.

Here authors report 3D printed glass with tunable UV-to-IR emission. The method overcomes thermal limits, integrating various quantum dots into 3D printed complex glass structures to enable advanced photonic functionalities.

Introduction

Glass stands as one of the most important materials known to mankind, with exceptional optical transparency, thermal/chemical stability, mechanical properties, and composition tunability, and has become a cornerstone of advanced photonic systems. In the realm of glass manufacturing, 3D printing, also known as additive manufacturing, holds the potential to combine the automation of modern industrial methods with a level of geometric complexity that exceeds the limitations of traditional artisanal techniques^1–5. Recent advances, including stereolithography (SLA)³, digital light processing (DLP)⁶, two-photon polymerization (TPP)⁷, and computed axial lithography⁸ have demonstrated success in fabricating silica-based optical components through photopolymerization of resin–glass composites. These technologies enable the integration of glass’s outstanding physical and optical properties into the creation of arbitrary macro- and microstructures^2,3. Further photonic functionalization is highly desirable.

Photonic functional glasses, such as glass containing metal and metal oxide nanocrystals⁹, rare-earth ions¹⁰, or QDs¹¹, have played crucial and central roles in the birth and development of photonics, which deals with the generation, transmission, modulation, amplification, conversion, and detection of light^12–15. Among them, semiconductor QDs exhibit quantum-confined emission tunable from ultraviolet to visible to infrared (UV–VIS–IR), coupled with the highest photoluminescence quantum yields (PLQY) up to 100%, narrow band emission, and a broad range of tunable chemical and physical properties due to their large surface-to-volume ratio^16–20. 3D printing of QD composite offers a fabrication route to engineer devices with advanced functional properties. The challenge is that current 3D printing technologies for QDs typically utilize photocurable resins, which suffer from low mechanical strength and degraded properties²¹. In contrast to resins, glass possesses exceptional optical transparency and physical properties. However, the high-temperature treatment required to remove organic compounds in glass 3D printing technology will cause severe aggregation due to Ostwald ripening and surface ligand decomposition of semiconductor QDs^9,22,23, which will prohibit the integration of QDs with 3D-printed glass. Moreover, uniform monodispersed distribution and precise size control of QDs in 3D-printed glass are hindered by uncontrollable growth kinetics and weak confinement effect in the solid medium. In addition, given the rich chemistry of QDs, a universal methodology is highly desired for the variety of QD systems in 3D-printed glass.

Here, we overcome these challenges by developing a universal nanoscale post-functionalization for 3D-printed glass, which offers a low-temperature, precise micro-environment chemistry strategy enabling the integration of QDs accessible to modern 3D printing glass techniques. By 3D printing NG with desired metal ion precursors, a library of 3D printed glasses functionalized by QDs were constructed relying on the spatial and chemical dual nanoconfined growth of diverse QDs (Fig. 1). The method is to exploit the controllable diffusion of metal ions in the confined environment, which tailors the delivery of precursors to the reaction and constrains the nanocrystal size in the quantum confinement regimes. This post-functionalization strategy circumvents the thermal limitations that lead to QD degradation. In addition to its convenience and energy-saving benefits, the nano-reaction space provided by this low-temperature approach also effectively prevents the aggregation of QDs, as it avoids the Ostwald ripening process that typically occurs during the melt-quenching route. Considering the inherent flexibility in incorporating QDs in 3D-printed glass via precise micro-environment chemistry, this will inspire the development of approaches to 3D printing of functional glass with the precise control over complex geometry, broad chemical compositions, and tunable photonics properties. These QDs functionalized 3D-printed glasses exhibit broad tunable PL ranging from UV (300 nm) to near-IR (2 μm) with luminescence lifetime from 16 to 699 ns and demonstrate voxel-level control over the QD bandgap. Benefiting from the multi-functionalization of the 3D-printed glass, various promising applications, such as photocatalysis, information encryption, and decryption, could be realized.

Fig. 1. 3D printing of QD-functionalized glass structures.

a Schematic fabrication of a light-emitting Oriental Pearl Tower through DLP 3D printing and subsequent QDs in-situ nanoconfined growth. b Illustrative picture of a complex-shaped object along different stages of the process. Photographs of QD glass structures: c 3D-printed Loong with complex structure and texture, containing CsPbBr₂I QDs. d Photographs and corresponding PL spectra of 3D-printed glasses structures functionalized by various QDs, including ZnS, CsPbBrCl₂, CsPbBr₃, MAPbBr₃, FAPbBr₃, CdS, CdSe, CsPbI₃, AgInS₂, CdTe, Ag₂S, PbS, and PbSe. Here, MA refers to methylammonium and FA refers to formamidinium. The photo shown uses CsPbBr₃ as the representative for the lead halide perovskite. e 3D-printed multi-color light-emitting Great Wall containing CsPbX₃ (X = Cl, Br, and I) QDs (top) and CsPbBr₃ QDs with different QD size (bottom). All samples were photoexcited at 375 nm, except for ZnS, which was excited at 280 nm. Infrared-emitting structures were imaged using an infrared-sensitive camera. The scale bars for a–c and e are 5 mm, and the scale bar for d is 2.5 mm.

Results

3D printed glasses functionalized by QDs

Figure 1a systematically details the manufacturing paradigm through four evolutionary stages: sol–gel ink formulation, vat photopolymerization 3D printing of green body, 3D-printed metal ions-doped NG (M-NG) formation, and final QD post-functionalizing. The engineered sol–gel ink comprises a photopolymerizable silica–alumina glass system optimized for vat photopolymerization printing, with functional additives enabling dual-phase control of nanoscale confinement environment and QD crystallization. Key components include: Tetraethyl orthosilicate (TEOS) as silica precursor, 3-(Trimethoxysilyl)propyl acrylate (APTMS) as covalent network reinforcement (Table S1). Metal salts [Pb(NO₃)₂, Cd(NO₃)₂, Zn(NO₃)₂, AgNO₃, and In(NO₃)₃] were added as precursors for Pb based [CsPbX₃ (X = Cl, Br, I), PbS, PbSe)], Cd based (CdS, CdSe, CdTe), ZnS, Ag₂S, and AgInS₂ QDs, respectively. Sudan Orange G was added as a photo-absorber to enhance vertical resolution and maintain dimensional accuracy. Besides, aluminum lactate is selected as a nanopore former in sol–gel-derived M-NG. TEOS, aluminum lactate, and APTMS undergo hydrolysis and condensation reactions, jointly contributing to the formation of a spatial network structure modified with acrylate groups^5,6,24.

A moderate drying step transformed the printed wet gel structure into xerogel, followed by low-temperature treatment at 650 °C in air², which converted the green body into 3D-printed M-NG, resulting in linear shrinkage of ~40%. The heat treatment decomposed organic compounds and removed carbon through oxidation. The thermal gravimetric analysis (TG) curve of the green body shows no significant weight change above 650 °C, indicating the complete conversion to inorganic glass (Supplementary Fig. 1). Differential scanning calorimetry (DSC) analysis of NG reveals a typical glass transition temperature (T_g) of 747 °C (Supplementary Fig. 1). The composition of NG is confirmed by the energy-dispersive X-ray spectroscopy (EDS) spectra, exhibiting an atomic percentage of all elements consistent with the designed SiO₂–Al₂O₃ network and no detectable carbon contamination (Supplementary Fig. 2). Raman spectroscopy further confirms the complete removal of organic components after sintering, showing spectra that closely resemble those of commercial fused silica (Supplementary Fig. 3). The as-prepared 3D-printed NG showcases the nanoporous structure with a pore size centered at 4 nm with smooth surface (R_ms ~ 1.1 nm) and excellent transparency over 90% in visible range (Supplementary Figs. 4–6), which is highly suitable for optical devices. Due to the presence of nanopores, NG has a slightly lower refractive index compared to commercial fused silica glass (Supplementary Fig. 7).

All-inorganic perovskite (CsPbX₃) QDs functionalized 3D-printed glasses are used here as a representative system. The doped metal ions (Pb²⁺) in M-NG primarily exist as glass modifiers, distributed uniformly in the form of single atoms or atomic clusters to facilitate QD crystallization, as shown in high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images (Supplementary Fig. 8). Then, the 3D printed M-NG was saturated in QD precursor solution to facilitate the in-situ growth of QDs. Ultimately, post-functionalization of 3D-printed glass by embedding QDs subsequently enabled the fabrication of arbitrary QD glass structures with complex spatial architectures and rationally designed optoelectronic properties (Fig. 1b and Supplementary Fig. 9). The voxel size is 180.0 × 180.0 × 5.6 μm, with QDs uniformly distributed within a single voxel (Supplementary Fig. 10). The optical image (Fig. 1c) shows the printed Loong-embedded CsPbBr₂I QDs, revealing intricate spatial details and texture, emitting orange light under UV excitation.

This low-temperature nanoconfined strategy for 3D-printed glass is generally applicable for various semiconducting QDs, including lead halide perovskite, CdS, CdSe, CdTe, PbS, PbSe, ZnS, AgInS₂, and Ag₂S (Fig. 1d and Supplementary Fig. 11). Corresponding transmission electron microscopy (TEM), HAADF-STEM micrograph, and EDS mapping of these QDs confirm their composition (Supplementary Figs. 12–20). These glasses exhibit emission with tunable peak from UV (380 nm) to near-infrared region (1.6 μm), with lifetimes varying from 16 ns for CsPbBr₃ QDs to 699 ns for PbS QDs (Supplementary Fig. 21). In addition, this method establishes two-dimensional bandgap engineering capabilities within monolithic glass objectives, permitting independent control of both QD composition and physical dimensions at micron resolution. By adjusting the QD composition, specifically the halide ions (from CsPbBr_xCl_3−x to CsPbBr_yI_3−y), this approach enables an expected emission peak tunability spanning from 450 to 710 nm (Supplementary Fig. 22a). Furthermore, by regulating the pore size of M-NG (from 2.6 to 7.3 nm) through tuning the Al content, the size of CsPbBr₃ QDs can be precisely controlled, enabling fine tuning of the emission peak from 465 to 510 nm (Supplementary Figs. 22b, 23). Within a single 3D-printed glass object, a controllable spatially resolved red-to-blue color gradient across the structure is achieved through spatial sectionalized treatment with different QD precursor solutions, and green-to-blue dual emission confined to designated regions is realized through selective tuning of QD size (Fig. 1e). Overall, our method successfully enables the QDs functionalized 3D-printed glass with high spatial resolution and complex geometries, allowing for the synthesis to a variety of QDs with controllable emission and lifetime. This would provide insights for the functionalization of the 3D-printed glass.

Low-temperature nanoscale engineering of QDs growth kinetics

Achieving monodisperse QDs with uniform sizes within a glass matrix has long been a central challenge for QD-functionalized glass, and rational growth kinetics control has proven to be an effective strategy for achieving monodispersed QDs^25,26. To address this, we propose a precisely engineered micro-environmental chemistry strategy. Our approach incorporates two key considerations. First, we utilize metal ion dopants in the M-NG matrix as the QD precursor source. These ions are gradually released into the confined nanochannel and become enriched, forming spatially localized heterogeneous nucleation centers (Fig. 2a). Unlike conventional glasses, where atomic diffusion is largely frozen below the T_g, the M-NG exhibits notable ion mobility even at room temperature (Supplementary Fig. 24). This behavior stems from its exceptionally high specific surface area (~564 m²/g), which exposes a substantial fraction of atoms at the internal pore surfaces²⁷. Based on molecular dynamics (MD) simulations, this surface contribution was estimated to correspond to over 60% of the total atoms (Supplementary Fig. 25). The high specific surface area and the nanoscale thin pore walls of the porous glass matrix result in elevated surface energy, which significantly facilitates the diffusion of Pb ions within the glass. As a result, the spatial distribution of the synthesized QDs could inherit the uniform distribution of Pb²⁺ introduced during the sol–gel process, resulting in uniform monodispersed QDs within NG. Second, during the reaction with precursor solution, the precursors undergo rapid reaction upon reaching the nucleation threshold²⁵. The QD experiences rapid growth while being spatially confined by the nanochannel of the M-NG matrix. Meanwhile, the precursor depletion drives the system rapidly into the Ostwald ripening regime, in which smaller nuclei that fail to achieve sustainable growth undergo selective dissolution²⁸. Under the dual regulation of precisely controlled precursor release and nanoconfined growth, a natural size-focusing process is achieved within the nanochannel, ultimately ensuring the exceptional size uniformity of the resulting QDs in M-NG.

Fig. 2. Mechanism in the kinetic control of QDs growth.

a Schematic of low-temperature nanoconfined growth of QDs within M-NG. b In-situ SAXS data and c in-situ recorded absorption spectra of CsPbBr₃ QDs glass collected during the 3000 s reaction. d The variation of the QD concentration derived from in-situ SAXS and in-situ absorption spectra, and Monte Carlo fitting results of SAXS in Fig. 2b, reveals the size variation of QDs. In-situ recorded absorbance of QDs at e various Pb content in M-NG and f various CsBr concentration showing QD concentration variation during the reaction. g Schematic representation of the CsBr diffusion-limited model within NG. h Theoretical predictions and experimental observations regarding the dependence of the reaction saturation time on the CsBr concentration.

We used perovskite CsPbBr₃ QDs (PQDs)—a representative system with notoriously fast growth dynamics²⁶—as a model to clarify our precisely engineered micro-environmental chemistry strategy. We monitored the formation process of PQDs through a combination of synchrotron in-situ small-angle X-ray scattering (SAXS) and in-situ absorption spectroscopy (Fig. 2b, c). SAXS analysis reveals the time-dependent evolution of PQD concentration, which shows a rapid initial increase followed by a gradual slowdown, eventually reaching saturation at ~1500 s (Fig. 2d top). Ex-situ XPS analysis further reveals the prolonged reaction process (Supplementary Fig. 26). Complementary in-situ absorption spectroscopy (Fig. 2c, d bottom) provides a similarly resolved concentration profile, enabling efficient and facile parametric screening. Based on the evolution of SAXS patterns, the average size of synthesized QDs has no significant change during the reaction and is estimated to be 4 nm (Fig. 2d right and Supplementary Text S1), which corresponds to the average nanochannel diameter of 4 nm. In addition, the TEM image shows the size distribution of PQDs ranging from 2 to 6 nm (Supplementary Fig. 27), which is consistent with the in-situ SAXS results. These results highlight the effective spatial confinement imposed by the nanoscale reaction environment on QD growth. In contrast to the extraordinarily rapid reaction kinetics typically observed in PQD synthesis, our nanoconfined synthetic strategy results in a significantly prolonged and controllable formation process, generating uniform and monodispersed QDs.

To gain further insight into the factors affecting the diffusion-regulated growth process, we systematically investigated the influence of precursor conditions, specifically, the Pb²⁺-doping level in the glass matrix and the concentration of CsBr in solution, on the reaction kinetics. Obviously, the Pb²⁺-doping level primarily determines the final absorbance, which is directly related to the concentration of QDs, while an increase in CsBr concentration significantly shortens the time required for the reaction to reach saturation (Fig. 2e, f). These observations imply that CsBr diffusion serves as the key rate-limiting factor governing the reaction kinetics in this nanoconfined system. We employed a one-dimensional transient diffusion model based on Fick’s second law to describe the time-dependent concentration profile of CsBr diffusing into the NG matrix (Fig. 2g)^29,30. Detailed procedure and discussion are shown in Supplementary Text S3. Once the CsBr concentration in the NG reaches a critical threshold (C_c), CsPbBr₃ rapidly crystallizes and fills the nanopores, forming the QD formation region. As CsBr continues to diffuse inward, the QD formation region propagates from the glass surface toward the center, and the reaction is saturated when this region reaches the core of the glass matrix. We observed good agreement between theoretical predictions and experimental results concerning the dependence of reaction saturation time on CsBr concentration (Fig. 2h), confirming that CsBr diffusion is the primary rate-limiting factor governing QD growth in this heterogeneous system and leading to a diffusion-mediated growth mode.

Physical and chemical dual confinement of QDs

The 3D printed PQD glass of the Oriental Pearl Tower model exhibits bright blue-green emission under UV excitation (Fig. 3a inset), which can be attributed to the high transparency of 3D-printed glass (Supplementary Fig. 29), benefiting from uniform QD distribution inside and the sub-diffraction limit size of the nanopore, as shown in Fig. 3a. The high-resolution transmission electron microscopy (HRTEM) image of PQD glass reveals a lattice spacing of 2.92 Å, assigned to (200) crystal facet of CsPbBr₃³¹. The X-ray diffraction (XRD) and wide-angle X-ray scattering (WAXS) of PQD glass (Fig. 3b and Supplementary Fig. 30) show diffraction peaks indexed to the cubic CsPbBr₃. Importantly, the uniform distribution of PQDs throughout the M-NG is supported by cross-sectional SEM and EDS elemental mapping, as well as PL mapping from both the surface and interior, all of which indicate homogeneous QD formation and emission (Supplementary Figs. 31–34). This is attributed to the spatially confined and gradual growth mechanism in our precise micro-environment chemistry strategy. The transient absorption (TA) also shows a single bleaching peak without any energy transfer process, indicating that few QD aggregations occur (Supplementary Fig. 35). These results validate the efficient and homogeneous nanoconfined growth of PQDs within M-NG.

Fig. 3. Physical and chemical dual confinement of QDs in glass for improved emission and stability.

a TEM and HRTEM images of the PQD glass. The inset is a photograph of a 3D-printed structure containing PQDs. b XRD patterns of PQDs glass and M-NG, bars represent standard diffraction data of CsPbBr₃ (PDF#54-0752). c EXAFS in R-space and d counter plot Morlet (10,1) WT amplitude of M-NG and PQDs glass. PDOS and the wavefunction spatial distribution of e and f CsPbBr₃ with surface Br defect and gand h CsPbBr₃ with Pb–O modified. The VBM is set at zero energy with a vertical gray dotted line as a guide to the eye. Atomic color codes are Cs (red), Pb (purple), Br (gray), and O (orange). The yellow and blue isosurface represents the spatial distribution of the defect-state wavefunction and its elimination after modification. i PL stability for PQD glass under ambient conditions.

In-situ confined-grown crystals typically exhibit enhanced interactions with the substrate material, which can improve the overall stability and homogeneity of the material^24,32,33. We further investigate the interfacial interaction between confined PQD and M-NG through X-ray absorption spectroscopy (XAS) and high-resolution XPS. X-ray absorption near edge structure (XANES) spectra show that the Pb L₃-edge absorption of M-NG and PQD glass is shifted to higher energies compared to that of Pb foil, indicating that lead in these samples exists in an oxidized state (Supplementary Fig. 36). The k²-weighted Fourier transform extended X-ray absorption fine structure (EXAFS) and wavelet transform (WT) analysis (Fig. 3c) reveal a Pb–O peak for M-NG at 1.6 Å (phase uncorrected). The Pb–O bonds could facilitate the bonding of QDs with the glass matrix at the interface. In contrast, the peaks for the PQD in 3D-printed glass are primarily observed at 2.5 Å after precursor reactions, which can be assigned to Pb–Br, with an additional peak at low R region. WT analysis (Fig. 3d) reveals that the additional feature originates from a light element, suggesting that oxygen in the glass is the likely backscatterer. To exclude the influence of unreacted PbO in the QD growth, we conducted inductively coupled plasma (ICP) tests (Table S2). The result reveals that the 3D-printed glass has an almost equal atomic ratio of Cs to Pb, which confirms that most of the Pb element can be converted into PQDs. Furthermore, comparing high-resolution XPS spectra of Pb4f in M-NG and PQDs glass (Supplementary Fig. 37), the Pb4f peaks exhibit a discernible shift to lower binding energy due to the conversion from Pb–O to Pb–Br. These values are lower than those of colloidal CsPbBr₃ QD due to the transfer of electron density from Pb atoms to SiO₄ tetrahedra within the glass. These results imply the formation of Pb–O bonds at the interface between PQDs and the M-NG matrix (Supplementary Fig. 38). Density functional theory (DFT) calculations reveal that interfacial Pb–O bonding can effectively passivate surface defects of QDs, thereby enhancing their optical performance. Br vacancy leads to the formation of a trap state that is located at ∼0.6 eV below the conduction band with a localized charge density within the lattice (indicated by yellow symbols), acting as an electron-trapping center for nonradiative recombination (Fig. 3e, f). Once the Br vacancy on the surface is modified by O, it removes the midgap states through contributions from the glass matrix, resulting in a significant reduction of wavefunction localization within the materials (Fig. 3g, h). Additionally, DFT calculations show that the interaction between Pb and O at the CsPbBr₃/SiO₂ interface leads to a higher adhesion energy, indicating a more stable interface for the QDs³⁴ (Supplementary Fig. 39).

The physical and chemical dual nanoconfined environment of M-NG effectively regulates the emission properties of QDs. The temperature-dependent (77–300 K) PL of the PQDs glass revealed a higher E_A value (exciton activation energy from thermal dissociation) of 165 meV, twice that of colloidal QDs (79 meV) (Supplementary Fig. 40 and Supplementary Text S4). The significant E_A value and minimal change in PL intensity suggest that the exciton in PQD glass has a higher energy threshold for thermal dissociation due to the enhanced spatial confinement of electrons and holes in M-NG³⁵. Owing to interfacial defect passivation and enhanced exciton binding energy, the PQD glass exhibits a high PLQY up to 82%. Moreover, the physical isolation by the nanopores along with the interfacial chemical bonding effectively enhances the stability of the QDs. As shown in Fig. 3i, PQD glass maintains 80% of its initial PL intensity after 180 days in ambient conditions of 25 °C and 50% relative humidity (RH), highlighting its excellent long-term stability. In contrast, the PL intensity of colloidal QDs decreased to 23% of the initial intensity after 30 days. Even under high humidity conditions (25 °C, 80% RH), the PQD glass retains 74% of its initial PL intensity after 600 h, while the pure QDs only retain 15% after 117 h (Supplementary Fig. 41). Moreover, we assess PQD glass photo-stability under continuous 405 nm, 120 mW laser irradiation in ambient conditions. The PQD glass retains 86% of its PL intensity after 16 h, while the colloidal QD film decreases to 44% (Supplementary Fig. 42).

Demonstration of 3D multifunctional glass architectures

This approach for 3D printing of QD glass can be used to fabricate multifunctional photonic and optical devices, which was first demonstrated by printing QD glass architecture for photocatalytic CO₂ reduction (Fig. 4a). The 3D design freedom offered by this additive manufacturing approach enables simultaneous optimization of multiple performance-determining parameters in photocatalytic systems. Specifically, the integration of QDs into a transparent, M-NG matrix provides both efficient photon harvesting and spatially confined photocatalytic reaction environments, while 3D printing allows for deterministic control over hierarchical geometric complexity that surpasses conventional fabrication, including the realization of bioinspired fractal surface features. These structures emulate nature’s efficient light-harvesting designs, such as tree-like or forest architectures³⁶, and are known to enhance photon scattering, increase illuminated surface area, and promote mass transfer efficiency in photocatalytic processes. Finally, we fabricated three types of 3D-printed architectures with different surface fractal micro-features: smooth surface (#1), cylindrical microarray (#2), and channel-like microarray (#3) (Fig. 4b and Supplementary Fig. 43). As shown in the SEM images, the 3D-printed QD glass exhibits well-defined microscale fractal surface features, characteristic of the designed hierarchical architecture. Control experiments using pristine M-NG without PQD loading show negligible photocatalytic activity, confirming that the CO₂ reduction functionality originates from the post-functionalized PQDs (Fig. 4c). Importantly, the introduction of surface microstructures significantly enhanced the photocatalytic performance: dome structures with increasing geometric complexity exhibited progressively higher gas production rates compared to the planar QD glass. The CH₄ evolution rate increases from 2.9 to 3.2 μmol g⁻¹ for #1, 4.6 μmol g⁻¹ for #2 and 7.4 μmol g⁻¹ for #3, representing an improvement up to 2.6-fold. Similarly, the CO evolution rate rises from 4.5 μmol g⁻¹ to 4.8 μmol for #1, 9.9 μmol g⁻¹ for #2 and 13.4 μmol g⁻¹ for #3, achieving an increase up to 3-fold. Moreover, to showcase the potential of this 3D-printing technique as a multifunctional photonic platform, we printed QD-embedded glass architectures as building blocks for optical information encryption and decryption (Supplementary Fig. 44). By spatially doping different metal ions and triggering selective luminescence through precursor-mediated reactions, we realized chemically responsive optical information encryption and decryption.

Fig. 4. 3D printed QD-functionalized glass architectures for enhanced photocatalytic CO₂ reduction.

a Schematic diagram of photocatalytic CO₂ reduction based on 3D-printed glass. b 3D-printed hemispherical PQDs glass with different 3D microarchitectures. The scale bar for the second column image is 500 μm, and the scale bar for the third column magnified image is 200 μm. c CH₄ and CO generation from CO₂ photoreduction using different printed structures.

Discussion

In summary, we establish 3D printing of NG as a promising platform for digitally architecting functional glass structures with voxel-precision in composition, architecture, and photonic properties. Our low-temperature post-functionalization strategy avoids the QD degradation during the high-temperature sintering process in 3D printing of glass. The uniformly doped metal ions and diffusion-limited reaction in nanoscale confined space precisely control the growth kinetics of QDs with desired chemical composition, size, and spatial distribution in the nanopores of 3D-printed glass, which achieves rational tunable PL across the UV–VIS–IR full-spectrum in 3D-printed glasses. As exemplified by CsPbBr₃ perovskite QDs, we achieve uniform and monodispersed QD formation within the 3D-printed glass matrix, with high PLQY up to 82% and narrow size distribution. Advanced spectroscopic and theoretical investigations reveal the formation of an interfacial Pb–O bond between the QDs and the glass matrix, which effectively passivates surface defects and significantly enhances QD stability. The resulting functional glasses exhibit broad spectral coverage and high design freedom in geometry and composition, setting a benchmark for the fabrication of 3D photonic structures with multifunctional optical properties. As nanoconfined nanocrystal growth represents a fundamental phenomenon, we anticipate its applicability extends to diverse QDs and additive processes such as inkjet and TPP.

The integration of QDs with 3D-printed functional glass offers a paradigm shift in glass fabrication towards voxel-level customization of photonic properties. Compared to rare-earth ions or other conventional luminophores with fixed energy levels, semiconductor QDs feature size-dependent bandgaps, enabling precise and continuous tuning of their emission across a broad full-spectral range. This full-spectral agility facilitates the rational functional design of glass-based freeform optics, which is highly promising for the next generation of photonic, sensing, and imaging technologies, such as glass-based functional metasurface or on-chip optical amplifier. Beyond functional glass, this approach also achieves the scalable additive manufacturing of QD-based photonic elements, laying the foundation for a new class of architected materials that seamlessly combine quantum-level control with macro-scale manufacturability.

Methods

Chemicals

The materials were used as received without further purification: Aluminum lactate (L-Al, 95%) and Diphenyl (2,4,6-trimethyibenzoyl) phosphine oxide (TPO, 98%), 3-(trimethoxysilyl) propyl Acrylate (APTMS, 99%), tetraethyl orthosilicate (TEOS, 99%) were purchased from Adamas-Beta. Sudan orange G (96%), ethanol (99.9%), N,N-dimethylformamide (DMF, 99.5%), Tris–hydrochloric acid buffer (0.05 M, pH = 7.4), 1-butanol (99.5%), methyl acetate (99%), octylamine (99%), oleic acid (99%), and HCl (AR, 37 wt%) were purchased from Sigma-Aldrich. CsBr (99.99%), CsCl (99.99%), CsI (99.99%), Na₂S (99.9%), Pb(NO₃)₂ (98.5%), Cd(NO₃)₂·4H₂O (99%), In(NO₃)₃ (99.9%), AgNO₃ (99%), Zn(NO₃)₂·4H₂O (AR), selenium (99.9999%), tellurium (99.999%), sodium sulfite (98%) were purchased from Aladdin. MABr (99.5%) and FABr (99.5%) were purchased from Xi’an Yuri Solar Co., Ltd. Ammonium acetate (99.9%) and PbBr₂ (99%) were purchased from Macklin.

Preparation of UV-curable sol ink

In our case (as can be calculated from Table S1), the sol–gel ink was prepared as follows: first, a solution (M1) was prepared by dissolving L-Al, a metal salt, and deionized water. The metal salts included one or more of the following: Pb(NO₃)₂, Cd(NO₃)₂·4H₂O, AgNO₃, Zn(NO₃)₂, and In(NO₃)₃. L-Al and metal salts were weighed in proportion and dissolved in deionized water, then stirred with a magnetic stirrer for 30 min. Second, a mixture of TEOS and ethanol (EtOH) (denoted M2) was added to M1 during the stirring process by slowly adding dropwise, and stirring the mixture until TEOS was fully reactive. Then, the M3 composed of 3-(Trimethoxysilyl) propyl Acrylate (APTMS) and Diphenyl (2,4,6-trimethyibenzoyl) phosphine oxide (TPO) was added, then stirred with a magnetic stirrer for 30 min. Finally, NH₄Ac solution (denoted M4) was added and sealed in a dark vessel and stirred for ≈45 min until the formed sol could form a gel matrix upon exposure to UV radiation with a wavelength of 405 nm. The weight percentage L-Al:TEOS:H₂O:EtOH:APTMS:TPO:NH₄Ac at all stages of the process is presented in Table S1.

3D printing of M-NG

The 3D models of a sample are designed by Blender and printed using a commercial DLP 3D printer (MaxUV, Asiga). The printing parameters include using a UV-LED light source (405 nm) with a light intensity of 15 mW cm⁻². After printing, the structures were placed in a closed vessel with a tiny hole and aged, dried in a 50 °C oven for 7 days. Then, dried samples were sintered at 650 °C (0.5 °C/min to 650 °C and kept for 120 min) in a muffle furnace to obtain 3D-printed M-NG.

In-situ confined growth of QDs

Precursor preparation. For perovskite precursors: CsX (X = I, Br, Cl): Dissolved in anhydrous methanol (5 mg/mL) with magnetic stirring at 70 °C for 2 h in a glove box. MABr/FABr: Dissolved in n-butanol (5 mg/mL) with magnetic stirring at 70 °C for 2 h in a glove box. For chalcogenide precursors: Na₂S: 0.1 M solution in Tris–HCl buffer (pH 7.4, 20 mL) with magnetic stirring for 30 min. Na₂SeSO₃: Synthesized by refluxing Se powder (5 mmol) and Na₂SO₃ (6 mmol) in deionized water (50 mL) at 80 °C for 3 h. NaHTe: Generated via NaBH₄ reduction (2.5 g in 5 mL H₂O) of Te powder (0.127 g) in a Schlenk flask under N₂ flow (40 °C, 20 min vigorous stirring) and stored in a glove box. Desired M-NG was activated by vacuum at 100 °C for 1 h to remove adsorbed gas and moisture.

The synthesis of QD-functionalized glass. Perovskite QDs: CsPbX₃ (X = Cl, Br, I): Pb-M-NG immersed in methanolic CsX solutions (60 °C, 2 h). MAPbBr₃/FAPbBr₃: Pb-M-NG immersed in methanolic MA/FABr solutions (60 °C, 2 h). Chalcogenide QDs: Cd-based systems: CdS: Cd-M-NG immersed in 0.1 M Na₂S (Tris–HCl buffer, pH 7.4) for 1 h. CdSe: Cd-M-NG reacted with 0.08 M Na₂SeSO₃ for 1 h. CdTe: Cd-M-NG treated with freshly prepared NaHTe (0.05 M in degassed H₂O) for 1 h in a glove box. Pb-based systems: PbS: Pb-M-NG in Na₂S solution (1 h, static immersion). PbSe: Pb-M-NG in Na₂SeSO₃ solution for 1 h. Ag-based QDs: Ag₂S: Ag-M-NG in Na₂S solution (1 h, dark condition). AgInS₂: In/Ag-M-NG (1 molar ratio) reacted with Na₂S solution for 1 h. Zn-based QDs: ZnS: Zn-M-NG in Na₂S solution for 1 h. After QD growth, vacuum drying was performed at 50 °C to remove the solvent.

Preparation of colloidal CsPbBr₃ QDs. Colloidal CsPbBr₃ QDs were prepared using the hot-injection method³⁷. Cs oleate precursors were prepared by reacting Cs₂CO₃ with oleic acid in 1-octadecene. The mixture was degassed at 130 °C for 1 h in a three-necked flask, then heated to 150 °C under nitrogen. Then, PbBr₂ precursor solution (PbBr₂, oleylamine, and oleic acid mixed in 1-octadecene) was dried under vacuum at 120 °C for 1 h in a three-necked flask. The temperature was raised to 160 °C, followed by the rapid injection of Cs oleate precursor. The reaction was quenched with an ice water bath, and CsPbBr₃ QDs were collected by centrifugation (9056×g, 0 °C), washed three times with hexane and methyl acetate, and dispersed in toluene (~100 mg/mL). 50 μL of QD solution was drop-cast onto a glass slide and dried at 60 °C until the solvent evaporated, forming a uniform film.

In-situ absorption and PL measurements

The reactions were carried out in a modified quartz cuvette. The absorption spectra were recorded with an Ocean Optics deuterium–tungsten light source (DH2000-BAL-TTL-24V) and a Thorlabs CCS200/M spectrometer. For in-situ PL and PL mapping measurements, the excitation beam was introduced to a confocal system and focused on the samples through an objective (×10). The PL signals were collected by the objective and detected by a spectrograph with 1200 mm⁻¹ grating and a silicon-charged coupled resolution of 0.04 nm.

Photocatalytic CO₂ reduction tests

The photocatalytic CO₂ reduction performance was explored using Labsolar-6A system (Perfect Light Co., Ltd., China) and a gas chromatograph. A 300 W Xe lamp was used as the photocatalytic light source to simulate sunlight. QD glass samples were added to the reaction chamber. Before the test, the system was dehydrated to remove O₂ and other gases, and then CO₂ was introduced. The content of Cs element in the glass matrix was determined by ICP testing, thus obtaining the content of CsPbBr₃ QDs.

Small and wide-angle X-ray scattering (SAXS/WAXS) measurements

The SAXS experiment was conducted at beamline BL19U2, and the WAXS experiment was carried out at BL17B1 in Shanghai Synchrotron Radiation Facility (SSRF). For in-situ SAXS, X-ray photon energies were 12 keV with a Pilatus 2M detector. Initially, the M-NG was confined within a cylindrical reactor. To eliminate the influence of methanol diffusion within the M-NG pores during the reaction, the M-NG was pre-soaked in methanol to ensure complete infiltration of the nanopores. Following this pre-treatment, an excess amount of precursor solution was introduced by an electrically driven automatic injector to initiate the reaction. For WAXS, X-ray photon energies were 10 keV with a Pilatus 2M detector. The q-range and the scattering intensity were calibrated with LaB₆.

Morphology and composition measurement

N₂ adsorption–desorption isotherms were measured at 77 K using a Quanta Chrome Autosorb iQ-MP gas sorption analyzer. Specific surface area was calculated using the Brunauer–Emmett–Teller (BET) method in the relative pressure range of 0.1–0.3. Pore size distribution was derived from the desorption branch using the non-local density functional theory (NLDFT) model. The refractive index was measured using a GAERTNER L116 ellipsometer. High-resolution TEM and HAADF-STEM imaging were performed using a Thermo Fisher Spectra Ultra scanning transmission electron microscope. Scanning electron microscopy (SEM) images were acquired using a ZEISS Gemini 500 field-emission SEM. Atomic force microscopy (AFM) was performed using a Bruker Dimension Icon system in tapping mode. Elemental analysis was performed using an Agilent 5110 ICP-OES system. XAS measurements at the Pb L-edge were performed at the BL16U1 and BL17B1 in SSRF. Data was collected from 100 eV below the Pb L-edge to ∼800 eV above the edge.

Computational methods

The dense glass matrix was modeled using classical molecular dynamics (MD) simulations with a Morse potential^38,39. Long-range Coulombic interactions were computed using Ewald summation with a 10 Å cutoff. A simulation box containing 17,997 atoms and periodic boundary conditions (PBC) in all directions was employed. The system was heated to 3000 K over 0.57 ns, equilibrated at this temperature for 1.0 ns, then cooled to 300 K over 0.57 ns and further equilibrated at 300 K for 0.2 ns under the NPT ensemble (0 bar), using LAMMPS⁴⁰. This process yielded a dense aluminosilicate glass structure with a cubic box size of 63.3 Å × 63.3 Å × 63.3 Å. Eight idealized spherical cavities with a diameter of 2.8 nm were introduced at the center of the box by removing the corresponding glass atoms. The porosity and specific surface area were designed based on experimental Brunauer–Emmett–Teller (BET) analysis.

Theoretical methodologies based on DFT calculations were performed using the Vienna ab-initio simulation package (VASP)^41–43. The generalized gradient approximation (GGA) with Perdew–Burke–Ernzerhof (PBE) function was used as the exchange-correlation function⁴¹. The projected augmented wave (PAW) method was used to describe the interactions between ions and electrons⁴⁴. Grimme’s D3 correction was used to consider the van der Waals interactions⁴⁵. Monkhorst–Pack mesh and kinetic energy cutoff were set to be a grid of (3 × 3 × 1) k-points and 500 eV, respectively⁴⁴.

For the computation of adhesive energy, the bottom atoms of the SiO₂ substrate were constrained to maintain the bulk environment. To evaluate the stability of the CsPbBr₃/SiO₂ interfaces, we calculate the interfacial binding energy (E_b), which is defined as

$$E_{\mathrm{b}}=E_{interface-}{E}_{Si{O}_2-}{E}_{perovskite}$$ 1

where E_interface, E_SiO2, and E_perovskite denote the total energies of the heterointerface supercell, the SiO₂ substrate, and the perovskite slab.

Author contributions

J.D., J.H., and Z.H. conceived the idea of this study. J.D. supervised and coordinated the project. F.Z. and P.W. conducted initial studies on M-NG synthesis that inspired the current project. F.Z. printed the QD glasses and took material characterization with help from K.F., Y.Y., and R.L. Optical measurement was conducted with help from Z.H., Z.Z., and Z.L. D.W. carried out the photocatalytic performance tests. Z.X., Z.O., and K.S. carried out the theoretical calculations. Y.Y., J.H., and J.D. provided financial support. All authors discussed the results and their implications and revised the manuscript at all stages.

Peer review

Peer review information

Nature Communications thanks Dezhi Tan and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data generated in this study are provided in the manuscript and Supplementary Information. The data supporting the findings of this study are provided as a Source Data file. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68523-z.