Unlocking high photosensitivity direct laser writing and observing atomic clustering in glass

Abstract

The direct laser writing (DLW) of photoluminescent metal clusters is inspiring intensive research in functional glasses. However, understanding the influence of the host structure on cluster formation and visualizing DLW-induced clusters at the atomic scale remains challenging. In this work, we develop a highly photosensitive fluorophosphate glass through fluorine incorporation. The addition of fluorine establishes a conducive environment for Ag⁺ ions before DLW and enhances the availability of reducing agents and diffusion pathways during DLW. These advantages facilitate the formation of Ag clusters under low-energy single-pulsed DLW. Increasing laser energy results in a combination of Ag clusters and glasses defect, forming a dot + ring photoluminescent pattern. Atom probe tomography (APT), a technique capable of mapping the elemental spatial distribution and identifying clustering, is employed to gain more information on laser-induced clusters. Comparison of APT results between samples without and with DLW reveals the formation of Ag clusters after laser writing. The design concept and characterization enrich the understanding of Ag cluster behavior in glasses. This knowledge opens the possibility of rational design of clusters confined in glasses and inspires their synthesis for various applications.

Ag clusters dispersed in glass are critical for optical storage media. The authors present fluorine-modified glass with high Ag⁺ ion solubility, enhancing the formation of Ag clusters under low-energy laser irradiation. The inert network effectively confines Ag clusters uniformly within the glass.

Introduction

Direct Laser Writing (DLW) enables precise modifications of transparent materials such as polymer¹, sapphire², and glass³. This capability allows for the creation of customized 3D microstructures with tailored properties, providing a versatile platform for material engineering and optical device fabrication. Among applications, there is a growing interest in fabricating silver nanoclusters (Ag NCs) inside glass for high-capacity optical data storage (ODS). Consisting of several to tens of atoms, Ag NCs exhibit bright photoluminescence (PL) due to the quantum confinement effect^4,5. In ODS, DLW is applied to inscribe Ag NCs, with each inscribed PL dot serving as an information recording unit⁶. By tailoring the DLW parameters such as laser energy, duration, and pulse count, tunable PL intensity and multiple grayscales storage can be achieved⁷.

Along with laser parameters, the choice of glass matrix also influences ODS performance. An ideal glass for storage application is expected to facilitate the generation of Ag NCs at low energy to reduce writing energy consumption and should be compatible with single-pulse laser processing to enhance the data writing efficiency⁸. The earliest literature on this topic can be traced back to the year 2010⁹, subsequent research has mainly focused on refining the DLW parameter using oxide-based glasses, such as silicate, phosphate, and borosilicate^10–12. Few attempts have been made to explore alternative glass systems and understand the glass's influence on Ag clustering. Recently, fluorophosphate (PF) glass has emerged as a promising host for Ag clusters, showing potential for high-density storage with enhanced writability¹³. Leveraging the low polarizability of F^- ions, PF glasses could inhibit silver ion reduction and consequent cluster formation. This ensures a uniform distribution of silver ions. During DLW, their large band gap and low refractive index mitigate the self-focusing effect and improve laser resistivity¹⁴. PF glasses also offer a lower phonon energy environment compared to their oxide counterpart, which reduces the probability of nonradiative transitions of Ag NCs and enhances their PL efficiency^15–17. These combined properties make PF glass a suitable choice for studying the photochemistry of Ag under laser irradiation and potentially accessing enhanced photosensitivity and PL properties.

Besides understanding glass matrixes, characterizing small clusters poses an additional challenge. Traditional techniques, such as transmission electron microscopy, face limitations due to the electron-beam-induced cluster growth¹⁸. And scattering methods often require specialized beamlines, making them less accessible¹⁹. To obtain both spatial and chemical information, a combination of techniques is usually employed, including fluorescence, X-ray absorption fine structure spectroscopy, mass spectrometry, and single-crystal X-ray diffraction^20,21. Recently, atom probe tomography (APT) has emerged as a powerful high-resolution 3D mapping technique with sub-nanometer precision, offering valuable insights into the distribution of clusters^22–25. APT offers quantitative information on cluster size and density, thus enhancing our understanding of their structural characteristics.

In this study, we began with a facile approach to enhance the glass photosensitivity by incorporating fluorine into the glass matrix. The addition of fluorine brought two benefits. Firstly, the high electronegativity of fluorine created a less reductant environment, preventing the reduction of Ag⁺ ions prior to DLW. This ensured a high concentration of Ag⁺ ions available for cluster formation during laser irradiation. Secondly, fluorine disrupted the glass network, increasing the diffusion channels and the availability of reducing agents under DLW (Fig. 1a, b). As the laser energy increased, the PL pattern changed from dots made of Ag clusters to a pattern by a combination of clusters and defects, resulting in a dot + ring arrangement (Fig. 1c). For characterizing the laser-induced clusters, APT was used to capture the formation of Ag clusters following laser writing (Fig. 1d).

Fig. 1. Designing glass structures.

a large Ag⁺ ions reservoir and (b) favorable diffusion channels ready for DLW. c Single-pulsed DLW transforms Ag⁺ ions into Ag clusters, offering insights into laser-glass interactions, and (d) enabling APT for cluster observation.

Results and discussion

Glass preparation and structural analysis

PFx glasses with fluorine concentrations ranging from PF5 to PF25 were synthesized (Supplementary Fig. 1). The difference between T_g (glass transition temperature) and T_c (crystallization temperature), ∆T, is used to estimate their thermal stability. The increased ∆T from PF5 to PF25 implies enhanced glass thermal stability²⁶, which is favorable for achieving highly durable storage media (Supplementary Fig. 2). The homogeneous dispersion of Ag⁺ ions was confirmed through absorption, PL, and decay measurement (Supplementary Figs. 3–5 and Supplementary Note. 1 and 2).

Analysis of ³¹P NMR spectra revealed the PF glasses network framework comprises metaphosphate chains by PO₄ tetrahedra, Q², where the superscripts indicate the number of bridging oxygens (BOs)²⁷. With increasing fluorine content, the Q² peak transferred into a smaller chemical shift, and a Q¹ signal emerged (Fig. 2a). This transformation suggested that fluorine depolymerized the metaphosphate structure by replacing bridging P-O-P bonds with terminal P-F and P - nonbridging oxygens (P-NBOs) bonds²⁸.

Fig. 2. Structural analysis of PFx glass.

a ³¹P and (b) ²⁷Al NMR spectra and deconvolution of PF5, PF15, and PF25 glasses. c Structures of phosphate (oxide-based, left) and fluorophosphate glass (fluorine-modified, right) by AIMD simulation.

The substitution of Q² by Q¹ indicated a reduction in the connectivity of the phosphate network, which was further supported by observations in Raman spectra (Supplementary Note. 3, Supplementary Fig. 6, and Supplementary Table. 1). This structural change was expected to bring more diffusion pathways for Ag⁺ ions during DLW. Moreover, the incorporation of F⁻ ions along with O²⁻ ions created a mixed-anion environment that exerts asymmetric forces on Ag species and enhances their mobility²⁹.

Compared to oxygen, the high electronegativity of fluorine resulted in a weak reductant environment, which slowed the reduction of Ag⁺ ions and prevented cluster formation before laser irradiation³⁰. This was substantiated by the PL enhancement of Ag⁺ ions from PF5 to PF25 (Supplementary Fig. 7). In terms of Al element, in PF5, most of Al modified the network interspace in the form of AlO₅ (Al⁵) and AlO₆ (Al⁶), with a minor fraction participating in glass backbone in the form of AlO₄ (Al⁴) (Fig. 2b)³¹. Upon increasing fluorine, the Al⁶ gradually converts into Al⁴, indicating a growing portion of Al entering the phosphate network via P-O-Al bond linkage. The transformation from Al⁶ to Al⁴, along with the generation of more chemically durable P-O-Al bonds, counteracted the fluorine-induced depolymerization, contributing to the enhanced glass chemical stability (Supplementary Note. 4 and Supplementary Fig. 8)³².

To gain deeper insight into the structural change, ab initio molecular dynamics (AIMD) simulations were conducted. Upon fluorine introduction, the glass transformed from a tight-bound phosphate network featuring Q² units to an open structure with Q¹ units and P-F bonds (Fig. 2c). At the same time, a growing proportion of Al⁴ units emerged, enhancing its role as a glass former. Moreover, in fluorine-modified glasses (right of Fig. 2c), Ag⁺ ions were found to be homogenously dispersed throughout the matrix.

Combining experimental and simulated results, the fluorine in PF glass brought two benefits. Firstly, it established a weak reductant environment that prevents premature reduction and ensures sufficient Ag reservoir for subsequent cluster formation. Secondly, the presence of fluorine disrupted the connectivity of metaphosphate chains, introducing more diffusion channels. These channels, in turn, facilitate the aggregation of silver species during laser irradiation and enhance laser responsiveness. These synergistic effects make PF glass a compelling candidate matrix for the fabrication of Ag clusters under low-energy DLW.

Creation and mechanism of Ag nanoclusters under DLW

Drawing on the structural analysis, it was hypothesized that PF25 (with the highest fluorine loading) would present the most favorable photosensitivity. To validate this, a series of single-pulsed femtosecond experiments were conducted (Fig. 3a and Supplementary Fig. 9). The PL of inscribed dots gradually gained its intensity with higher laser energy (Fig. 3b). The minimal energy threshold required to induce PL signal (E_min) decreased from 1.0 μJ (PF5) to 0.1 μJ (PF25), indicating enhanced photosensitivity with increasing fluorine.

Fig. 3. Photoluminescence and distribution of DLW-induced Ag clusters.

a Confocal fluorescence image of PFx glasses, λ_ex = 405 nm. b Integrated PL intensity. c Micro-PL spectrum of the dot written with 0.5 μJ in PF25 (inset: its confocal fluorescence image), smoothing applied without losing intrinsic features (orange line). d Emission decay curve monitored at 640 nm (PF25, 0.5 µJ). e Dependence of PL dot size (solid symbol) and ring diameter (hollow symbol) on laser energy. f, g Backscattered electrons image and elemental maps of the irradiated area under 0.5 µJ and 10 µJ energy in PF25.

Micro PL spectra revealed a broad band centered at 640 nm with a smaller shoulder at 520 nm (Fig. 3c and Supplementary Fig. 10). After thermal treatment at 400 °C for 12 h, the primary peak at 640 nm persisted, while the intensity of 520 nm decreased (Supplementary Fig. 11). This indicates that Ag clusters are responsible for the 640 nm emission, while thermally unstable defects contribute to the 520 nm emission³³. Furthermore, the emission at 640 nm displayed a lifetime of 3.68 ns, consistent with the nanosecond decay characteristic of Ag clusters (Fig. 3d)⁹_. These emission features of Ag clusters resemble those of clusters produced by both multi-pulsed femtosecond laser writing and electron/gamma radiation in glasses^34–37. Moreover, with increasing laser energy, the PL spectrum showed a gradual increase in intensity, indicating a growing number of Ag clusters (Supplementary Fig. 12).

As the writing energy reached 1.0 μJ, peripheral PL rings emerged in PF15 to PF25 (Fig. 3a). With increasing laser energy, the diameter of PL ring (D) presented progressive expansion, while the central dot (d) remained constant (Fig. 3e). This difference suggested that the dot + ring pattern likely originates from two different PL centers. A comparison of electron probe microanalysis (EPMA) revealed that under 1.0 μJ, the Ag enrichment in the outer ring (Fig. 3f, g), while the center exhibits depletion of all elements.

The contrast in element distribution suggested that the PL rings are associated with the migration of Ag species. Conversely, the central dots are likely attributed to intrinsic properties of the glass, possibly arising from laser-induced defects. These defects stem from laser-induced perturbations known as phosphorus oxygen hole centers (POHCs). They are characterized by an unpaired electron shared between two non-bridging oxygens bound to a phosphorus atom. In addition, they can be resonated at an electron paramagnetic resonance scan (EPR in Fig. 4a and Supplementary Note. 5)³⁸. Furthermore, these centers are known to be thermally unstable and can be easily erased by heat treatment at 400 °C for 12 h (Supplementary Fig. 13)³⁹.

Fig. 4. Mechanism of Ag clusters formation under DLW.

a EPR spectra and (b) micro-Raman spectra of the region without and with DLW (w/o-DLW and w-DLW) in PF25, the inset shows the mapping of ν_s (P-O-P) band of the dot written by 0.5 µJ in PF5 and PF25. c Schematic diagram of structural evolution upon fluorine modification and increased laser energy.

To understand the mechanism behind the dot + ring pattern, a closer examination of the interaction between the laser and the glass becomes necessary. During DLW, the energy deposition begins with a multiphoton absorption process (approximately four-photon), exciting electrons from the valence band into the conduction band⁴⁰. These excited electrons then work as reductants, reducing Ag⁺ ions into Ag⁰ atoms. Simultaneously, a hole-capturing process takes place by the reaction PO + hole → POHC⁴¹. In this context, the PO refers to phosphorus-oxygen defect precursors that are considered pre-existing faults in the glass network. Such faults can be oxygen vacancies, nonbridging oxygens, and strained bonds⁴².

Recalling PF glass structure, the incorporation of fluorine brings the P-O-P bond breakage, increasing the number of P-NBOs (phosphorus to non-bridging oxygens bonds) and terminal P-F bonds. These NBOs serve as the phosphorus-oxygen defect precursor for POHC formation, capable of capturing holes and inhibiting the recombination of electron-hole⁴³. As a result, the retained electrons effectively facilitate the reduction of the Ag⁺ ions, leading to the formation of Ag clusters even under low-energy laser irradiation.

When exposed to high-energy DLW, however, the intense laser beam would induce a localized temperature surge, accompanied by a strong shock wave and electron rearrangement at the focal point. This heightened temperature gradient and shock wave facilitate the diffusion of Ag species, leading to the formation of ring-shaped Ag clusters, while the electron rearrangements contribute to the central glass defect^44,45. Consequently, ring-shaped Ag clusters emerge at the periphery, and thermally erasable defects (POHC) appear in the center.

Further insights into the laser-induced structural changes were obtained from micro-Raman spectra. In Fig. 4b, consistent with the Raman trend upon increasing fluorine, DLW reduced the intensity of both the ν_s (Q²) peak and ν_s (P-O-P) with a negative Raman shift. The 2D mapping of ν_s(Q²) displayed a comparable intensity decrease in the regions subjected to DLW for both PF5 and PF25 samples (Supplementary Fig. 14). But for ν_s (P-O-P) — the indicator of phosphate connectivity (Q²), its reduction in PF25 covered a much smaller area (inset of Fig. 4b). This suggested that the fluorine exerts a pre-break impact, similar to that caused by DLW.

This pre-break effect generates both diffusion pathways and precursors for POHCs: the former prepares the channels for Ag clusters accumulation ready for laser irradiation, and the latter aids in hole scavenging and defect formation (Fig. 4c). It should be noted that our observation differs from previous reports, where the formation of Ag clusters required a thermally assisted process under multiple pulse irradiation⁴⁶. However, during single-pulsed laser irradiation, thermal accumulation was absent, as evidenced by the lack of obvious melted structures (Supplementary Fig. S9). The formation of clusters is expected to be limited. In this context, the significance of the glass material becomes evident. The generated Ag clusters evidence the beneficial impact of glass photosensitization, which was facilitated by the introduction of fluorine anions. This reaffirms the effectiveness of the design strategy for PF glasses.

Atom probe tomography of Ag clusters under DLW

To characterize the laser-induced Ag clusters, APT was performed (Fig. 5a). Using a dual-beam focused ion beam, two needle-shaped specimens of the sample without DLW (w/o-DLW) and with direct laser writing (w-DLW) were prepared (Fig. 5b, Supplementary Note. 6and Supplementary Figs. 15–18). Upon laser pulsing, ions in the tip were evaporated from the surface. These ions were directed into a position-sensitive detector with their mass-to-charge ratio recorded by time-of-flight mass spectrometry. Here, the potential for overheating of the specimen tip was deemed minimal, as indicated by the absence of thermal tails in mass spectra (Fig. 5c and Supplementary Fig. 19), suggesting rapid cooling^47,48. Subsequently, the recorded ion positions on the detector were utilized to reconstruct the spatial distribution (Supplementary Figs. 20–21).

Fig. 5. Atom probe tomography of Ag clusters.

a Schematic illustration of APT measurement. b FIB-based APT specimen preparation: i SEM image of a protective Pt cap over the ROI; ii an undercut bar ready for extraction; iii, iv tips of w/o-DLW and w-DLW after annular milling. c Mass spectrum of w-DLW tip. d, g 3D map reconstruction of Ag in w/o-DLW and w-DLW. e, h FD analysis of Ag: dotted lines with symbols were the experimental data, and solid lines were the modeled binomial distributions. f, i Ag-Ag pair correlation function, $g\left(r\right)$, inset in (i) shows normalized $g_n\left(r\right)$ and fitted radius result.

Visual examination of the 3D reconstructions of Ag revealed no signs of aggregation (Fig. 5d, g). To assess the Ag distribution, frequency distribution (FD) analysis was applied⁴⁹. This analysis begins with a comparison between the FD of Ag experimental data and the theoretical randomized distribution, the latter characterized by binomial probability. As shown in Fig. 5e, the experimental and binomial curves for w/o-DLW were almost overlapped, suggesting a randomized distribution of Ag. However, in w-DLW, the experimental curve presented a broadened peak (Fig. 5h). Given the conservation of the total number of atoms, this broadening indicated that some blocks contain fewer Ag atoms while others contain more than expected. The long tails indicate blocks with high concentrations of Ag atoms, suggesting Ag aggregation⁵⁰.

Moreover, the FD analysis incorporates χ² statistics to quantify the degree of randomness, assessed by the p-value^51,52. In w/o-DLW, the Ag experimental curve matched with the theoretical binomial distribution, with a p-value of 0.7, larger than the confidence level typically set at 0.01, suggesting the random distribution of experimental Ag. However, in the w-DLW, a p-value of < 0.001 indicated the aggregation of Ag. Following this, the Pearson coefficient (μ) was used to remove the sample size dependence and give a more accurate estimation, where μ = 1 suggests complete non-randomness and μ = 0 suggests complete randomness. Comparing μ values of w/o-DLW (0.7225) and w-DLW (0.2201) reveals a greater deviation from randomness in w-DLW after laser writing.

Beyond the FD analysis and χ² statistics, the Ag aggregation was also reflected in the nearest neighbor distribution (NND) analysis⁵³. The Ag NND for w/o-DLW presented a minimal deviation from the randomized distribution (Supplementary Figs. 22, 23). However, in w-DLW, there was a gradual deviation that suggests Ag aggregation. This was further examined using the maximum separation method (Supplementary Note. 7 and Supplementary Figs. 24–26), where Ag clusters with radii larger than 1 nm in w-DLW were found, while clusters in w/o-DLW were smaller than 1 nm (Supplementary Fig. 27). In addition, the Ag-Ag pair correlation functions, $g\left(r\right)$, in w-DLW showed a more positive Ag correlation, reflecting the Ag clustering tendency (Fig. 5f, i)^54,55. Moreover, the Ag-Ag correlation distance (distance at which the correlation approaches unity) was longer in w-DLW than in w/o-DLW, indicating a higher proportion of Ag atoms participating in clusters. For a sphere of radius (R), the normalized correlation function $g_n\left(r\right)$ can be fitted to determine the R-value, according to Eq. (1) and (2)⁵⁶:

$$g_n\left(r\right)=\left\{\begin{array}{ccc}\hfill 1-\frac{3r}{4R}-\frac{r^3}{{16R}^{3\prime }}\hfill & \hfill \left(ifr\le 2\mathrm{R}\right)\hfill & \hfill \left(1\right)\hfill \\ \hfill 0,\hfill & \hfill \left(ifr>2\mathrm{R}\right)\hfill & \hfill \left(2\right)\hfill \end{array}\right)$$

In the w-DLW region, the radius (R) was determined to be 0.673 nm. These findings collectively confirmed the presence of Ag clusters and offer valuable insights into the aggregation and distribution of Ag clusters in the DLW region.

Optical data storage demonstration

Considering the bright PL and low energy consumption, PF25 and single-pulse laser energy with 0.5 μJ were selected for ODS demonstration (Fig. 6a). A 7 × 7 dot array (Fig. 6b) and the logo of Zhejiang University (Fig. 6c) were written. Figure 6d shows the PL intensity in the array plotted against the horizontal axis. A distinct emissive peak was observed at the inscribed regions, enabling the representation of binary information as emissive dots denoting 1, interspersed with non-emissive intervals representing 0. Each dot occupied a volume of 2 × 10⁻¹¹ cm⁻³, corresponding to a storage capacity of 5.82 GB cm⁻³ (or 197 GB per glass disk in 120 × 3 mm³ form). This capacity exceeds that of traditional DVDs, which typically range from 5 to 25 GB⁵⁷. Notably, there is potential for achieving even higher storage densities by using a more tightly focused laser beam or implementing multilevel encoding within a single data unit⁵⁸. The spatial processing capability of DLW, combined with the homogeneous matrix, enabled the patterning of information at various depths without crosstalk. Three letters, Z, J, and U, were written with a layer interval of 5 μm, demonstrating the potential for high signal-to-noise and multilayer storage (Fig. 6e).

Fig. 6. Demonstration of 3D optical data storage using PF25 glass.

a Photo of PF25, the width was close to that of a paper clip, and the length was one-third of it. b A 7 × 7 information points array. c Zhejiang University logo. d Intensity profile along the horizontal direction of the array, realizing the 0 and 1 state change in binary information. e Three-level stack of Z, J, and U with an interval of 5 μm.

In conclusion, we demonstrated the simultaneous design of photosensitive glass and atomic characterization of DLW-induced Ag clusters by APT. The fluorine-modified fluorophosphate glass facilitated the efficient creation of photoluminescent Ag clusters under low-energy single-pulsed DLW. The introduction of fluorine played a dual role by disrupting the glass network, creating conducive channels for Ag species clustering, and introducing defect precursors that trapped holes, thereby enhancing electron availability for Ag⁺ ion reduction. High laser energy resulted in a dot + ring PL pattern, where the central dot was glass defect, and the peripheral ring was Ag clusters. Moreover, the use of APT supported the formation of DLW-induced Ag clusters. These findings, along with the optimization of glass structure and the use of advanced characterization techniques, enhance our understanding of DLW-induced Ag clusters in glass and offer promising prospects for developing photo-functional materials.

Methods

Materials and reagents

Glass samples with the composition of (95-x)Al(PO₃)₃-xMF₂−5Ag₂O (M = Mg, Ca, Sr; x = 5–30 mol%) were prepared using the melt-quenching method with Al(PO₃)₃, MgF₂, CaF₂, SrF₂, and AgNO₃ as raw materials. These are denoted as PFx, where x represents the mol% of MF2. PF0 refers to glass with the composition 70Al(PO₃)₃−25MO-5Ag₂O, with MO being alkaline earth metal oxides (M = Mg, Ca, Sr). Al(PO₃)₃ (Analytical reagent) were purchased from Thermo Scientific Reagent Co. MgF₂, CaF₂, SrF₂, MgO, CaO, and SrO (Analytical reagent) were purchased from Aladdin Reagent Co. AgNO₃ (Analytical reagent) was purchased from Sinopharm Chemical Reagent Co. A 30 g batch of each glass composition was weighed and mixed homogenously in the agate mortar. Then, the mixture was melted in alumina crucibles with closed lips to minimize fluorine loss at 1450 °C for 60 min. The melt was then poured onto a cold stainless-steel plate, annealed, and then immediately pressed with another plate to increase the cooling rate and promote glass formation. The obtained glasses were then cut, ground, and polished for subsequent measurements.

Direct laser writing in glasses

Single-pulsed direct laser writing was conducted on a regeneratively amplified Yb: KGW laser (PHAROS, Light Conversion) with a central wavelength of 1030 nm, pulse duration of 220 fs, and repetition rate of 1 kHz. The laser was focused 100 μm below the glass surface using a 50 × (NA = 0.8) objective lens. The diameter of the laser spot was about 2 μm.

Materials characterizations

X-ray powder diffraction (XRD) was performed using an X-ray diffractometer (Shimadzu XRD-6000) with a 5 °C min^−1 scanning speed. The UV-Vis absorption spectra were recorded in a UV-Vis spectrophotometer (Hitachi U-4100). The glass transition temperature (Tg) and crystallization temperature (Tc) were collected from the differential scanning calorimeter (DSC, Q100 TA). The heating rate was 10 °C min^−1. Photoluminescence (PL), Photoluminescence excitation (PLE), and lifetime decay spectra were recorded by Edinburgh Instruments FLSP920 spectrophotometer. The pH of the glass leachate solutions was measured using a pH meter (FiveEasy PlusTM, FB28). Optical images of the laser-modified regions were captured by a CCD camera attached to a Nikon microscope (Eclipse 80i). PL spectra of the laser-modified region were taken from a confocal microscope (λ_ex = 405 nm, LSM780, Zeiss). The PL of Ag clusters after DLW was carried out on a homemade confocal microscope. Raman spectra and mapping were collected by Raman spectrometer (LabRAM HR Evolution) with a 532 nm laser excitation. The ²⁷Al nuclear magnetic resonance (NMR) spectra were recorded on an Agilent 600 DD2 spectrometer at a frequency of 156.25 MHz (14.1 T), with 3.6 μs pulses and 10 s recycle time in a 4 mm double-resonance probe. ³¹P NMR spectra were collected at a frequency of 242.76 MHz (14.1 T) using 3.0 μs pulses and 5 s recycle delays. The AlCl₃ and H₃PO₃ were chosen as the chemical shift reference. SEM images and EMPA were carried out on EMPA-1720 (SHIMADZU) at an accelerating voltage of 15 kV and a current of 10 nA. Electron paramagnetic resonance (EPR) spectra were recorded on a Bruker A300 ESR spectrometer in X-ban. The samples were loaded into a quartz tube and cooled to 120 K. A dual-beam SEM/FIB (FEI Helios Nanolab 600i) was used to prepare lift-out samples.

APT experiments were conducted on a CAMECA LEAP 4000H XR operated in laser pulse mode using 355 nm Nd:YVO₄ ultraviolet (UV) laser. The detection efficiency was 37%. Data were acquired in laser pulsing mode at a specimen temperature of 50 K, with a target evaporation rate of 0.5 ions per 100 pulses, pulse duration of 10 ps, pulsing rate of 200 kHz, and laser pulse energy of 60 pJ. The raw data were processed by the commercial IVAS 3.4.4 software to map the position and mass-to-charge ratio of each species.

ab initio Molecular dynamics simulations

ab initio molecular dynamics simulations were carried out on a cubic sample made up of ~ 1000 atoms with random coordinates by Packmol (v20.14.1)⁵⁹. Two compositions, the 70Al(PO₃)₃−5Ag₂O-25CaO (without fluorine) and 70Al(PO₃)₃−5Ag₂O-25CaF₂ (with fluorine) were studied. Notably, Ca was chosen to represent the alkaline earth metals (Mg, Ca, Sr), as they all exert as the glass modifier. These two systems were subjected to structural optimization using the Density Functional Theory method based on DZVP-MOLOPT-SR-GTH. Starting from the initial configurations, Born-Oppenheimer Molecular Dynamics (BOMD) was applied⁶⁰. The high-temperature melt at 2000 K was quenched to 300 K over a period of 300 ps in the NPT ensemble (constant number of atoms, constant pressure, and constant temperature). The glass was relaxed at 300 K under atmospheric pressure for 5 ps. Periodic boundary conditions were applied in all directions. The visualization of local atomic structure was facilitated by VESTA (Ver 3.4.5)⁶¹.

Author contributions

X.Q., D.T., and W.Z. conceived the idea. W.Z. performed the experiments and analyzed the data. W.C. conducted theoretical calculations. Z.W. did the direct laser writing. W.Z., H.L., G.Y., and M.Z. did the APT test. Q.X., J.Z., J.Q., G.Q., and X.F. took part in the discussion and gave important suggestions. W.Z., X.Q., and D.T. wrote the paper. All authors approved the final version of the paper.

Peer review

Peer review information

Nature Communications thanks Yannick Petit and the other anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data represented in Figs. 2a, b, 3b–e, 4a, b, 5c, e, f, h, i, and 6d are provided as Source Data file. All data in Supplementary Information are available from the corresponding author on request. Source data are provided in this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-024-52628-4.