Integrated copper-halide activated scintillator fiber array for remote high-resolution X-ray imaging
  • SJ_Zhang
  • Jul. 7, 2025

Introduction

Efficient scintillator arrays are essential for achieving high-resolution X-ray imaging, which is vital for both scientific research and practical applications, such as medical diagnostics, food and equipment quality inspection, security screening, and geological exploration1. The advent of sophisticated industrialization increases the demand for high-resolution imaging applications, stimulating continued efforts in the development of not only scintillators with enhanced resolution but also rigorous standards for the long-distance transmission of image signals1,2,3. Nevertheless, the light penetration and spatial resolution of transparent scintillators (e.g., Bi4Ge3O12 and CsPbBr3) with a thickness of millimeter level remain constrained by the intrinsic limitations of conventional emission mechanisms, namely transverse photon scattering and self-absorption (Fig. 1a)4,5. In subsequent studies, researchers sought to construct scintillator pixel arrays with the objective of improving resolution and transmission distance, wherein scintillating photons are oriented vertically to attenuate undesirable transverse photon leakage or to render guiding photons entirely reflective within a specified area (Fig. 1b)6,7,8. However, most of the contemporary research has concentrated on the processing of thin-film and bulk scintillators, which presents significant challenges in terms of its adaptability to the requirements of long-distance transmission (Supplementary Table 1, and Supplementary Fig. 1)9,10,11,12,13,14,15,16,17. In contrast, active fiber arrays designed with waveguide optical structures represent a streamlined and effective means of transmitting high-resolution imaging signals over long distances (Fig. 1c)18,19. The active fiber array scintillator represents a significant advancement in X-ray imaging, offering high resolution and long-distance transmission capabilities that surpass those of flat-plate and flat-plate array scintillators. Moreover, the high flexibility combined with the robustness of fiber arrays could facilitate deep tissue penetration for immediate remote imaging applications9,20.

 

Fig. 1: X-ray imaging by photon manipulation.

figure 1

Different imaging modalities based on a flat-plate scintillators, b flat-plate array scintillators, and c active fiber array scintillators inside the confined space (e.g., hollow iron spheres). The utilization of high-resolution imaging capabilities of arrays, in conjunction with the efficient signal transmission abilities of optical fibers, has enabled the conception of active fiber array scintillators that exhibit both high-resolution imaging functionality and long-distance signal transmission capabilities.

In order to prepare efficient active fiber arrays with incorporated scintillator crystals, it is first necessary to design uniform and highly transparent materials that minimize inherent light scattering from non-oriented crystal structures, mitigate material self-absorption, and ensure optimal optical transmission efficiency. Copper halide materials like Cs3Cu2X5 are emerging as environmentally friendly and high-performing scintillators, characterized by their low toxicity, ease of preparation via solution methods, high photoluminescence quantum yield (PLQY), significant Stokes shift, negligible self-absorption, and high linear attenuation coefficients21. The fabrication of glass fibers and dense fiber array bundles is currently well-established, while the incorporation of the designed scintillator crystals into glass fibers through in situ crystallization usually results in significant scattering losses caused by the crystals. The loss of optical transparency is generally ascribed to refractive index mismatch between the glass matrix and the scintillator crystals. Although scattering losses can be minimized by reducing the crystal size, in the hot-drawing fiber process the growth of the crystals cannot be easily controlled to obtain nanosized crystals with smaller Rayleigh scattering. Therefore, the development of low-loss glass fibers with the incorporation of designed high-performance nanocrystals for targeted applications remains a tremendous challenge for contemporary fiber fabrication methods.

In this work, we demonstrate the fabrication of glass fibers incorporating Cs3Cu2X5 nanocrystals (NCs), which leads to the development of a pixelated active fiber array scintillator for remote X-ray imaging, which achieves both high spatial resolution and long-distance transmission simultaneously. The waveguide structure is formed by the fiber bundle, with X-ray induced scintillation photons confined within a single pixel region by the lead-free metal halide glass fiber core, thereby enabling propagation along the well-aligned fiber scintillators. It is noteworthy that the fiber scintillator array is comprised of a hexagonal cladding, which allows for the flexible, dense arrangement of scintillator fibers without crosstalk. We design and fabricate a fiber array scintillator, which consists of ~1600 fibers with a single pixel diameter of ~10.4 μm. The fiber exhibits the ability to transmit scintillator emission over a distance of 5 meters, significantly outperforming other scintillator fibers reported in the literature and meeting practical requirements for long-distance transmission (Supplementary Table 2). Furthermore, the flexibility and long-distance transmission capabilities of the scintillator fiber array allows for a large penetration depth within complex structures, enabling functions that are not amenable to conventional flat panel detectors.

Results

Preparation and optical properties of Cs3Cu2X5 glass

To prepare scintillator fiber bundles for long-distance transmission, we employ a melt-quenching process to synthesize borosilicate precursor glasses (PG). These glasses are subsequently heat-treated at temperatures near the glass transition point (Tg) to induce the growth of Cs3Cu2X5 NCs within the glass matrix (Fig. 2a and Supplementary Fig. 2)22. A comprehensive analysis of the factors of transparency and luminous intensity reveals that the optimal annealing condition for Cs3Cu2X5 nanocrystals glass (Cs3Cu2X5 NC glass) is 490 °C for 10 hours. The control of crystal size and distribution enables the Cs3Cu2X5 NC glass to exhibit low light scattering, high transparency, excellent UV-induced emission performance, and high PLQY (Figs. 2b, c, Supplementary Figs. 39). It is noteworthy that the peak of the photoluminescence (PL) band of Cs3Cu2X5 NC glass can be tuned from 445 nm to 526 nm by adjusting the halogen species (Cl, Br, and I), with a large Stokes shift (130–211 nm), which effectively suppresses the reabsorption of the initially emitted visible photons (Fig. 2b)23. The PL spectra of Cs3Cu2I5 nanocrystal glass, obtained under various excitation wavelengths, along with the photoluminescence excitation (PLE) spectra collected at different emission wavelengths, reveal matching spectral patterns, indicating a unified origin for excitation and recombination processes. Cs3Cu2Br5 and Cs3Cu2Cl5 NC glass exhibit the same characteristics (Supplementary Fig. 10)24. Based on single-exponential PL decay dynamics, the average lifetimes of the Cs3Cu2X5 NC glass are determined to be from 1.15 to 113 μs, which are consistent with previous reports (Supplementary Fig. 11)25. As shown by X-ray photoelectron spectroscopy (XPS), only Cu+ is observed in the Cu 2p spectra, indicating that the Cs3Cu2X5 NC glass remains unoxidized during the preparation (Supplementary Fig. 12)21. The exciton binding energy (Eb), exciton-phonon coupling Huang-Rys factor (S), and phonon frequency (?ωphonon) of Cs3Cu2I5 NC glass were calculated by fitting temperature-dependent PL spectra and full width at half maxima (FWHM) curves to be 303.4 meV, 40.33 meV, and 14.52 meV, respectively (Supplementary Fig. 13). These parameters are consistent with previously reported data in the literature, which strongly supports that the emission mechanism of Cs3Cu2I5 originates from self-trapped exciton (STE) emission, as illustrated in the mechanistic diagram shown in Supplementary Fig. 14a. The strong confinement effect on excitons and local lattice distortion lead to the generation of multiple self-trapped excitons. These excitons are further localized in the Cs3Cu2X5 lattice and are difficult to dissociate into free electrons and holes, which enhances the radiative recombination probability and luminescence efficiency of electron-hole pairs. This further confirms the precipitation of Cs3Cu2I5 NCs in the precursor glass26. The unique non-self-absorbing nature of STE emission, the zero-dimensional structure of Cs3Cu2X5 nanocrystal, and their excellent crystallinity resulting from in situ precipitation within the glass matrix collectively contribute to the high PLQY of Cs3Cu2X5 NC glasses (73.46% for Cs3Cu2Cl5, 70.84% for Cs3Cu2Br5 and 87.02% for Cs3Cu2I5) at room temperature. This also lays the foundation for subsequent X-ray luminescence studies.

 

Fig. 2: Fabrication and characterizations of the Cs3Cu2X5 (X = Cl, Br, I) NC glass.

figure 2

a X-ray diffraction (XRD) patterns of the Cs3Cu2X5 NC glass. b PL and PLE spectra, c PLQY, transparency, Stokes shift, fluorescence lifetime, FWHM compose the satellite map of Cs3Cu2X5 NC glass. d Three-dimensional views of the simulated glass network. e Simulated dependences of the fractions of [BO4] on K2O contents (calculation simulation and actual data mapping). f29Si NMR spectra of the Cs3Cu2I5 NC glass are presented before and after annealing at 490 °C for 10 hours. g Model illustrating the migration of copper ions leads to the crystallization of Cs3Cu2X5 NC in glass, h Cu ions lead to the structural changes of the boron-oxygen network and silicon-oxygen network.

We prepare a series of samples to analyze and verify the crystal precipitation mechanism of Cs3Cu2X5 NC glass (Supplementary Table 2). A glass network structure that is too loose could facilitate the self-precipitation of Cs3Cu2X5 NCs, whereas a structure that is too tight could prevent the precipitation of these crystals (Supplementary Fig. 15a). The results of the calculations simulating structural changes in the glass network, the actual changes measured by infrared (IR) absorption transmission spectroscopy, and the changes in Tg are in accordance with this conclusion (Figs. 2d, e, Supplementary Figs. 15 and 16 and Supplementary Tables 3 and 4). Moreover, the distribution of cesium, copper and halide ions within the glass network structure is contingent upon the specific control variables27. Contrary to conventional assumptions, Raman spectroscopy demonstrates that Cs+ ion and I ions are concentrated within the boron-oxygen network, whereas Cu+ ions are distributed across both the boron-oxygen and silica-oxygen networks (Supplementary Fig. 17). To validate this hypothesis28,29, Raman and nuclear magnetic resonance spectroscopy (NMR) analyses were conducted on samples before and after annealing (Fig. 2f and Supplementary Figs. 18 and 19). Following the process of crystallization, the Q2 and Q3 units within the silica-oxygen network undergo a transition to Q4 (Fig. 2f) (Q1, Q2, Q3, and Q4 represent the number of each silicon bridged to the bridging oxygen in [Si-O]), while the manifestation within the boron-oxygen network is a transition from [BO3] to [BO4] (Supplementary Fig. 19). This observation indicates that the crystallization process is not solely the precipitation of Cs, Cu, and I from the boron-oxygen network in a ratio of 3:2:5, but also involves the migration of Cu ions from the silica-oxygen network to the boron-oxygen network. Therefore, the crystallization process can be ultimately simplified as a charge addition from the silica-oxygen network to the boron-oxygen network. Given that the ionic radii of Cs+ and I are considerably larger than that of Cu+, and that the strength of the Cs-O, Cs-I, binding bond is markedly stronger than that of the Cu-O bond, it can be inferred that Cu+ is relatively facile in migrating over long distances within the context of the network structure30. In conclusion, two principal processes have been identified as the driving forces behind the precipitation of Cs3Cu2I5 NCs. First, the migration of Cu+ from the silica-oxygen network into the boron-oxygen container network, followed by its gathering with the Cs and I ions, represents a pivotal mechanism (Fig. 2g). Second, the separation of the enriched Cs, Cu, and I elements from the glass matrix leads to the formation of Cs3Cu2I5 NCs (Fig. 2h).

Cs3Cu2X5 nanocrystal glass scintillation and imaging

We then examine the scintillation properties of Cs3Cu2X5 NC glass. In the samples tested, the RL spectra of Cs3Cu2X5 NC glass align with the PL spectra (Fig. 3a), and the corresponding image is shown in Supplementary Fig. 20a. Among these samples, Cs3Cu2I5 NC glass exhibits the most intense scintillation luminescence. To achieve an exceptionally high level of transparency, we limit the NC concentration the concentration of NCs within the glass matrix. Despite this, the integrated intensity of the X-ray-excited emission peaks observed for Cs3Cu2I5 NC glass is comparable to that of commercial Bi4Ge3O12 (BGO) crystals (Fig. 3a)14, with X-ray absorption coefficients only slightly lower than those of the commercial BGO scintillator (Fig. 3b and Supplementary Fig. 20b)14. When the annealing temperature was increased to 500 °C, the X-ray luminescence intensity rose significantly, and the calculated light output increased from 7087 photons MeV−1 to 14,059 photons MeV−1 (Supplementary Figs. 20cd). These findings indicate that the primary factors influencing the emission intensity of Cs3Cu2I5 NC glass are the content and crystallinity of the embedded NCs. They also suggest that the Cs3Cu2I5 NCs within the glass inherently possess excellent radioluminescence. This intense scintillation has been attributed in previous studies to STE emission, which benefits from a strong confinement effect that enhances luminescence efficiency25. As shown in Supplementary Fig. 21, we measured the temperature-dependent X-ray excitation spectra of Cs3Cu2I5 NC glass and calculated the corresponding Eb (302.3 meV), S (39.3 meV), and ?ωphonon (16.7 meV), all of which are similar to those observed under UV excitation. This indicates that both the radioluminescence and UV excitation processes are governed by STE emission, as shown in Supplementary Fig. 14b. The interaction of X-ray photons with heavy atoms in the Cs3Cu2I5 NCs glass results in the ejection of a large number of energetic electrons through the photoelectric effect and Compton scattering. These electrons are then rapidly thermalized, generating secondary electrons, which ultimately relax into the conduction band, facilitating complexation.

 

Fig. 3: Optical properties, high resolution and dynamic imaging demonstration of Cs3Cu2X5 (X = Cl, Br, I) NC glass scintillator under X-ray excitation.

figure 3

a RL spectra of a standard commercial scintillator (BGO) wafer and Cs3Cu2X5 NC glass. Note that the size and thickness of the wafer and the glass are the same. b Calculated absorption curves of BGO, CsI: Tl, Cs3Cu2X5 NC glass, CsPbBr3 NCs, as a function of X-ray energy. c The recorded SNR values of Cs3Cu2X5 NC glass as a function of dose rate. d Bright-field and X-ray image of the standard X-ray resolution pattern plate with the Cs3Cu2X5 NC glass scintillator (voltage: 40 kV), and e the corresponding MTF curves of Cs3Cu2X5 NC glass scintillator. f Photographs of Cs3Cu2I5 NC glass (scale bars: 10 mm), chips (scale bars: 2.5 mm), and copper mesh (scale bars: 0.5 mm), under sunlight and X-ray irradiation, respectively. g The dynamic, real-time X-ray imaging capability of the Cs3Cu2X5 NC glass during the rotation of a specially designed metal plate (scale bars: 10 mm).

The Cs3Cu2X5 NC glass scintillator displays an excellent linear response to X-ray dose rates across a broad range (Fig. 3c and Supplementary Fig. 22), indicating its significant potential for X-ray dose rate calibration. Moreover, the measured X-ray dose rate shows a detection limit of 22 nGyair s−1, which is considerably lower than that required for current diagnostic procedures (5.5 μGyair s−1)12. It is noteworthy that the Cs3Cu2X5 NC glass exhibits both high stability under X-ray irradiation and excellent resistance to environmental weathering. Supplementary Fig. 23a provides further evidence supporting the stability of the glass under repeated X-ray irradiation, with no significant reduction observed during 100 cycles of on-off irradiation. In addition, the RL intensity remained almost unchanged after 10 weeks of immersion in water (Supplementary Fig. 23b), and a photograph of continuous irradiation for 1 hour is shown in Supplementary Fig. 23c. The exceptional water resistance of the Cs3Cu2X5 NC glass can be attributed to the protection provided by the inorganic glass matrix, which prevents the aggregation and degradation of the Cs3Cu2X5 NCs.

In subsequent experiments, an X-ray imaging system is subsequently constructed, as illustrated in Supplementary Fig. 24, with image acquisition performed using a digital camera. The Cs3Cu2X5 NC glass is clearly discernible when used for imaging the standard X-ray resolution line card at 30 lp mm−1 (Fig. 3d). The resolution limit of the modulation transfer function (MTF) for Cs3Cu2X5 NCs glass, calculated by using the bevel method, ranges from 32 to 33 lp mm−1, which significantly surpasses that of the typical commercial CsI: Tl scintillator (12 lp mm−1) and conventional glass and single-crystal scintillators (Fig. 3e and Supplementary Fig. 25)12. Furthermore, a large-area Cs3Cu2X5 NC glass scintillator of 7 cm in diameter and 3 mm in thickness is fabricated (Fig. 3f and Supplementary Fig. 26). The Cs3Cu2I5 NC glass scintillator reveals detailed structural information of the chip, and the copper mesh, all at high resolution under a 40 kV, 100 μA X-ray source (Fig. 3f). The dynamic, real-time X-ray imaging capability of the Cs3Cu2X5 NC glass was further evaluated by capturing images during the rotation of a specially designed metal plate (Fig. 3g, Supplementary Fig. 27 and Supplementary Movie 1). The X-ray image at the different time points demonstrates a distinct phase difference and precise delineation of the chopper blade edges, without the presence of ghosting effects. These results substantiate the potential of the Cs3Cu2X5 NC glass for long-term utilization in medical and industrial X-ray imaging and detection applications.

Cs3Cu2X5 nanocrystal glass fiber scintillators

Optical fiber facilitates efficient optical signal transmission and supports low-loss interfaces in photonic integrated circuits. The high glass-forming ability of the precursor glass ensures the suitability for manufacturing fiber array substrates. However, the production of the Cs3Cu2X5 NC glass fiber by the traditional in-tube rod fiber drawing method is challenged by the rapid crystallization of Cs3Cu2X5 NCs that occurs at the fiber-drawing temperature, which is higher than the crystallization temperature. Therefore, during the fiber-drawing process, uncontrolled glass crystallization may occur, resulting in high optical loss and diminished scintillation emission. We fabricate Cs3Cu2X5 NC glass fibers using an in-tube melting fiber-drawing technique31. We maintain the drawing temperature at 850 °C. At this temperature, the core glass melts while the cladding glass softens (Supplementary Fig. 28). Following rapid cooling during the fiber-drawing process, the core glass undergoes a transition from a molten state to a glassy state without crystallization, resulting in the production of transparent, uniform, flexible and continuous precursor glass fibers (Figs. 4a, b). The fiber core of our design is composed of Cs3Cu2X5 NCs embedded in glass, while highly transparent borosilicate glass (K2 glass) is used as the cladding. The refractive indices (n) of the core and K2 cladding glasses are 1.52 and 1.49, respectively, resulting in a relatively large numerical aperture (NA) of 0.30, which enables efficient light collection31. The scintillation fiber features a large-core design, with a core diameter of 69 μm and a cladding diameter of 125 μm, which optimizes the X-ray absorption region (Supplementary Fig. 29). After heat treatment, the growth of Cs3Cu2X5 NCs occurs within the fiber core, resulting in the emission of bright blue and green light under ultraviolet (UV) excitation (Fig. 4c). This emission corresponds to the PL and RL spectra, which are consistent with those of Cs3Cu2X5 NCs NC glass (Supplementary Fig. 30). The electron probe microanalyzer (EPMA) spectrum of the Cs3Cu2X5 NC glass fiber demonstrates effective suppression of element migration across the core-cladding interface (Supplementary Fig. 31). Furthermore, we analyze Raman and X-ray diffraction (XRD) spectra of the core region before and after annealing (Supplementary Fig. 32). The Raman spectrum is predominantly characterized by two peaks, at 108 cm−1 and 160 cm−1, which correspond to the I-Cu-I vibration. These results, in conjunction with the XRD pattern, unambiguously support the formation of Cs3Cu2X5 NCs within the fiber core region.

 

Fig. 4: Cs3Cu2X5 NC glass fiber arrays for remote pixel-point imaging.

figure 4

a Photographs of Cs3Cu2X5 (X=Cl, Br, I) NC glass fibers under sun light and b 310 nm UV light illumination (scale bars: 5 cm). c Optical images of the cross-section for the Cs3Cu2X5 NC glass fibers (scale bars: 30 μm). d Optical loss of the Cs3Cu2X5 NC glass fibers by a cutback method under X-ray. e Preparation process of fiber array, multiple ortho-hexagonal fibers are polymerized through a mold and secondary drawn at 800 °C. f Photographs of a fiber bundle with a decreasing diameter (scale bars: 5 mm), the truncated end face of a fiber bundle (scale bars: 50 µm), and a micrograph of a 450 nm single-mode fiber laser (core diameter: 10 µm) coupled at a single pixel (scale bars: 20 μm). gh The propagation of light is simulated using R-soft software to observe the light transmission in fiber bundles of varying diameters with an identical core-to-package ratio, g correspond to r1 = 5.8 μm, d1 = 10.4 μm, h correspond to r2 = 2.9 μm, d2 = 5.2 μm. Preset optical transmission distance of 10 cm. i Photograph of an active fiber optic array in sunlight, with the outer layer consisting of polydimethylsiloxane (PDMS) as a protective coating (scale bars: 10 mm). j Demonstration of long-range high-resolution dynamic imaging of fiber optic bundles using mask plates (scale bars: 40 μm). k Conventional flat plate scintillator (scale bars: 5 mm) and l fiber optic array scintillator is used to detect cracks in the inner wall of a perforated hollow steel sphere using different detection modes (Fine mask plates are placed at positions 1–5 to simulate cracks and scale bars is 20 μm). The direction of X-ray incidence is directed to the other end of the circular hole from different positions of 1-5, respectively.

The optical losses of Cs3Cu2X5 NCs glass fibers were quantified by using a cutback method at 450 nm, yielding values of 0.075 dB cm−1, 0.083 dB cm−1, and 0.089 dB cm−1 for Cs3Cu2I5, Cs3Cu2Br5 and Cs3Cu2Cl5 glass fibers, respectively (Supplementary Figs. 33a and 33b). Similarly, the optical losses of Cs3Cu2I5 NCs glass fibers remain unchanged at 0.076 dB cm−1 and 0.078 dB cm−1 under 310 nm and X-ray excitation, respectively (Fig. 4d and Supplementary Fig. 33). As shown in Supplementary Fig. 33d, the end of the fiber ≈5 m away from the Cs3Cu2I5 NCs glass fibers still clearly detects the switching signal of the radiation source. These values meet the requisite standards for optical transmission31. It is noteworthy that Cs3Cu2I5 NC glass fibers exhibit remarkable stability under immersion in water and no degradation of PL intensity is observed after 60 days of water immersion (Supplementary Fig. 34). These findings indicate that Cs3Cu2X5 NC glass fibers display robust scintillation emission, effective light transmission, remarkable water resistance, and high durability, which collectively position them as a promising candidate material for remote X-ray detection and imaging.

We then experimentally demonstrate the fabrication of fiber array scintillators for X-ray imaging. The array comprises 200 μm Cs3Cu2I5 NCs glass fibers, with optical fibers densely packed within a template. Following the heating and setting process, the preform is redrawn at 800 °C, allowing the fabrication of a fiber bundle consisting of ~1600 optical fibers with a diameter of 10.4 μm (Fig. 4e). The total diameter of the optical fiber bundle is approximately 318 μm (Fig. 4f). All K2 fiber claddings are pre-shaped into regular hexagons to prevent deformation resulting from the dense packing of circular fibers during the secondary drawing process (Fig. 4e and Supplementary Figs. 35-39). It can be theoretically demonstrated that the imaging resolution of the fiber array plate is independent of the fiber length (Supplementary Fig. 40). In the fiber bundle, the total internal reflection within the fiber core guarantees that each scintillating glass fiber operates independently, without optical crosstalk, thus enabling them to serve as individual pixels under ultraviolet light or X-ray excitation. The spatial resolution of the fiber array plate is primarily constrained by the diameter of the fibers, which can be improved to several microns or even sub-microns through optimizing the fiber drawing procedure (Fig. 4f and Supplementary Fig. 41). However, the distance of adjacent cores gradually decreases as the fibers’ cladding thickness is reduced, resulting in optical coupling and signal crosstalk, which affects the resolution of the multi-core fiber32. To theoretically study the coupling effects, we simulate the light propagation by R-soft based on the beam propagation method to observe the optical transmission in the fabricated fiber33. For comparison, we fabricate two samples with different diameters but the same r/d ratio, representing different drawing processes (r represents the core diameter, and d represents the distance of adjacent core, respectively), and simulate the light propagation by exciting the central core with a laser beam. For sample 1, the parameters are defined as r1 = 5.80 μm, d1 = 10.40 μm, and for sample 2 the parameters are defined as, r2 = 2.90 μm, d2 = 5.20 μm, where r1/d1 = r2/d2 = 0.557. The propagation length of the two samples is 10 cm, and the results are shown in Fig. 4g and h. In the case of sample 1, when a laser beam with a wavelength of 450 nm is guided into the fiber core, light is confined within the core region for propagation without coupling, as can be seen from the output facet and the light dynamic evolution (Fig. 4g). However, in the case of sample 2, when the fiber is drawn into a smaller size with geometrical parameters half of sample 1, the results become different. Light is no longer localized, but spread to other cores, and multiple signals are detected at the output facet for a fiber length of 10 cm (Fig. 4h). The light dynamic evolution also confirms that light is coupled to other cores along propagation, as shown in Supplementary Fig. 41. Our calculations further indicate that the imaging resolution cannot be further enhanced by reducing the core diameter below 8.23 µm (Supplementary Fig. 42), which implies that the fiber bundle of this system has an X-ray imaging resolution limit of 60.7 lp mm−1. It is important to note that additional adjustments to the core and cladding refractive index ratio and diameter ratio can effectively mitigate fiber coupling, thereby facilitating the attainment of enhanced resolution imaging for applications in human nerve detection, cell probing, and other related fields.

The point spread function of the fiber array panel’s intensity distribution is calculated based on images captured with an optical microscope. The FWHM of the fiber diffusion functions is ~5.58 µm, demonstrating the uniformity of the fiber core layer diameter and ensuring reliable optical information transmission (Supplementary Fig. 43). Nevertheless, the integrated area of the point spread function for a single fiber’s intensity distribution and the uniformity of the luminescence distribution within the mapped fiber bundle requires improvement, which is anticipated to be resolved in subsequent commercial processing. To more accurately visualize and assess the resolution limitations, the constructed 10 cm-long Cs3Cu2I5 NC glass fiber bundle (Fig. 4i) is imaged using a specially constructed hollow SCUT mask. We dynamically and clearly visualize the shape variations along the edges of the mask template and between different letters (Fig. 4j, Supplementary Fig. 44 and Supplementary Movie 2). The high flexibility of millimeter-scale fiber bundles (Supplementary Fig. 45) can be used without the need for an electromechanical objective system, while simultaneously ensuring the requisite functionality within the optical field system. Fiber-optic imaging probe systems offer advantages in instances where experimental limitations impede the capture of three-dimensional (3D) image data16. To confirm this hypothesis, a hollow iron ball with a thickness of 3 mm and 2.5 mm diameter holes distributed on its left side is prepared (Supplementary Fig. 46).

To confirm this hypothesis, we fabricate a hollow iron sphere with a thickness of 3 mm, featuring holes with a diameter of 2.5 mm distributed on its left side (Supplementary Fig. 46). A small masking plate is placed inside these holes to simulate cracks in a damaged metal casing. When X-rays are projected from left to right, conventional scintillation blocks struggle to provide detailed information about the shapes of tiny cracks, even when observed from different angles due to obstruction by thick metal (Fig. 4k and Supplementary Fig. 47). In contrast, as shown in Fig. 4l and Supplementary Fig. 47, the active fiber array scintillator penetrates directly into the interior, effectively detecting not only the shapes of the cracks (with clearly displayed numbers), but also cracks in multiple orientations (1–5 different directions). These results demonstrate the flexibility of Cs3Cu2I5 NC glass fiber scintillator arrays and their potential for developing remote, high-resolution X-ray imaging applications.

Discussion

In conclusion, we have demonstrated the development of a fiber array made of Cs3Cu2X5 NC glass with highly efficient radiation-induced luminescence and low propagation loss, which are leveraged for remote and real-time radiation detection and pixelated imaging. The Cs3Cu2X5 NC glass achieves high quantum efficiency (87.02%) and transparency (91%) due to judicious selection of the glass matrix and the regulation of the NC precipitation process. The combination of these features with special STE emission (no self-absorption) enables the achievement of a record-breaking high resolution (31–32 lp mm−1) in X-ray imaging of a glass block. Moreover, additional preparation of optical fibers allows for the transmission of optical signals over long distances (5 m). The scintillator fibers can be readily and consistently configured into a compact matrix of active fibers and deployed to traverse narrow passages for X-ray imaging (e.g., within hollow spheres) with a resolution of 48 lp mm−1 (limit: 60.7 lp mm−1) for X-ray imaging. It is anticipated that the scale-up production of these scintillating glass fiber arrays will enable diverse applications, including fine radiation therapy, brain nerve information collection, and industrial 3D-ray tomography.

Methods

Computational section

Molecular dynamics (MD) simulations were carried out for B2O3-SiO2-K2O-Cs2O-BaO glass compositions by a large-scale atomic/molecular massively parallel simulator (LAMMPS) with periodic boundary conditions in all directions. The interactions between particles are described by a pairwise potential function, with its form and parameters extracted directly from the literature. For each composition, the particles are randomly placed in a simulation box with a length of 4 nm, while excessive overlap is avoided to prevent excessively high potential energy. Then, the box is fully melted at 6000 K in the canonical ensemble to lose the arbitrary configuration. After a full relaxation in the microcanonical ensemble, the system is then rapidly cooled to 300 K and sampled in the microcanonical ensemble. Both time and ensemble averaging are employed to ensure statistical significance of the results. This melt-quenching process has been fully adopted from the original paper to minimize possible structural discrepancies in the final glass, and widely adopted and validated by researchers.

Experimental section

Preparation of glass samples: amorphous and transparent glass specimens were synthesized using a conventional melt-quenching approach. High-purity reagents, including SiO2, B2O3, K2CO3, BaCO3, CsCl, CsBr, CsI, CuCl, CuBr, CuI and Cs2CO3, were weighed and blended homogeneously based on the nominal molar composition of 25SiO2-50B2O3-8K2O-3BaO-6CsX-7Cs2O-1CuX (mol%, X=Cl, Br, I). The resulting batch was heated in a furnace under an argon atmosphere at 1150 °C for 20 minutes. The molten mixture was then poured into a stainless-steel mold for quenching, forming precursor glasses. These were subsequently annealed at 400 °C to alleviate internal thermal stress. Finally, the glass specimens were cut and polished for subsequent crystallization and optical analysis. Crystallization was carried out via heat treatment at 470–500 °C (above the glass transition temperature Tg) for 10 hours.

Preparation of glass fibers: precursor glass fibers without crystallization were fabricated by using the in-tube melting method. The core glasses with the composition of 25SiO2-50B2O3-8K2O-3BaO-6CsX-7Cs2O-1CuX (mol%, X=Cl, Br, I) were machined and polished into cylindrical rods with a diameter of 10.0 mm. A borosilicate glass tube was used as the cladding, with the composition of K2 glass, purchased from Nantong Zhenhua Optical-electric Co., Ltd, China). The core glass was inserted into the cladding glass tube to form a preform. The glass fiber preform was suspended in a fiber-drawing tower and drawn into glass fibers at 850 °C. Afterwards, the precursor glass fibers were heat-treated at 470–500 °C for crystallization.

Preparation of glass fiber arrays: Multiple fibers are meticulously bundled using a die. These preforms then undergo a low-temperature annealing process to form the desired scintillator fiber arrays. Subsequently, the preforms are subjected to high-temperature drawing, resulting in the formation of scintillator fiber arrays. Specifically, Cs3Cu2I5 NC glass fibers with a diameter of 200 μm are densely arranged within the template. After heating and curing at 550 °C, the preform is redrawn at 800 °C, yielding a fiber bundle composed of ~1600 fibers, each with a diameter of 10.4 μm.

Sample characterizations: to determine the glass transition and crystallization temperatures, differential scanning calorimetry (DSC) was performed using a STA449C Jupiter thermal analyzer (Netzsch, Bavaria, Germany) under ambient conditions, applying a heating rate of 10 °C min−1. The thermal expansion behaviors of both core and cladding glasses were examined using a DIL 402PC thermal dilatometer (Netzsch, Germany) with the same heating protocol. Raman spectra were acquired with a Renishaw inVia spectrometer (London, UK), utilizing a 785 nm laser as the excitation source. The morphology and particle size distribution of quantum dots (QDs) formed after glass crystallization were characterized via transmission electron microscopy (TEM, JEM-2100 F, JEOL, Japan). X-ray diffraction (XRD) patterns were obtained using a Bruker D8 Advance diffractometer (Faellanden, Switzerland) with Cu Kα radiation (λ = 0.1541 nm). Optical absorption spectra were recorded using a Lambda 900 spectrophotometer (PerkinElmer, Waltham, USA). Photoluminescence (PL) spectra and decay lifetimes of the bulk glass were measured with an Edinburgh FLS 920 spectrofluorometer (Edinburgh Instruments Ltd., Livingston, UK) coupled with a Hamamatsu photomultiplier for detection. A 450 W ozone-free xenon arc lamp and a pulsed xenon lamp were used as excitation sources for PL and decay measurements, respectively. For glass fibers, PL spectra were collected using a QEpro fiber-optic spectrometer (Ocean Optics) under 310 nm UV excitation. Elemental mapping across the fiber cross-section was conducted via electron probe microanalysis (EPMA-1600, Shimadzu, Kyoto, Japan).

To evaluate radioluminescence (RL) performance under X-ray exposure, a compact X-ray source (MagPro X, Motex Ltd., USA) with a maximum output of 12 W (voltage range: 4–60 kV; current range: 0–1000 μA) was employed to generate X-ray radiation. RL spectra from the bulk glass specimens were obtained using an Omni spectrometer (Zolix Ltd., Beijing, China) integrated with a photomultiplier tube. For fiber-based samples, emission spectra were collected using a QEpro fiber-optic spectrometer (Ocean Optics). The X-ray dose rate was calibrated using an ionization chamber dosimeter (10 × 6−180, Radcal Ltd., USA) alongside an Accu-dose digital reader. A Nikon D610 DSLR camera was utilized to capture X-ray projection images. Temperature-dependent RL spectra were recorded using a fiber-coupled spectrometer with a temperature regulation system. Image resolution was quantified by calculating the modulation transfer function (MTF), derived from the edge spread function (ESF) and line spread function (LSF). ESF was extracted from the image edge profile, and LSF was obtained by differentiating the ESF; the MTF was then calculated via the Fourier transformation of the LSF. Photon yield under X-ray irradiation was co-calibrated using standard scintillators, including BGO and CsI: Tl. The number of emitted photons was determined by integrating the RL spectral intensity for each sample, followed by normalization with respect to X-ray attenuation across the entire energy spectrum and sample thickness. This photon output was then benchmarked against that of the reference scintillators. The X-ray photon attenuation data were retrieved from the XCOM database provided by the National Institute of Standards and Technology.

Simulation of light propagation: We simulate the light propagation by R-soft based on the beam propagation method to observe the optical transmission in the fabricated fiber. For comparison, we fabricate two samples with different sizes but the same r/d ratio, representing different drawing processes (r represents the core diameter, and d represents the distance of the adjacent cores, respectively), and simulate the light propagation by exciting the central core with a laser beam. The analog transmission wavelength is 450 nm, and the transmission distance is 10 cm.