Introduction

Scintillation, the process by which certain materials emit light upon exposure to high-energy particles such as X-rays, is paramount to numerous technologies1. Its role is particularly prominent in X-ray imaging and characterization, where scintillators are crucial for converting X-ray energy into visible light, which can then be detected and analyzed2. Advances in bulk scintillator processing have been key to their widespread adoption in X-ray imaging applications. Techniques like the Czochralski and Bridgman methods3,4 are highly scalable and have been adapted to produce large, high-quality scintillator crystals in bulk. Other established techniques such as thermal evaporation5,6 and sol-gel methods7 have also been successfully tailored for large-area polycrystalline scintillator manufacturing.

An emerging approach in scintillator research—coined “nanophotonic scintillators”—consists of structuring scintillator materials at the scale of their emission wavelength8,9 to control their emission properties, such as their light yield8,10, emission directionality11, detection efficiency12, imaging performance13, and timing14. Such enhancements open up new possibilities for more precise and efficient X-ray imaging technologies15,16,17.

Despite these promising results, the widespread adoption of nanophotonic scintillators is hindered by challenges in scalable fabrication techniques. Current top-down fabrication methods, which rely on sophisticated lithographic techniques, offer nanometer-scale resolution and repeatability but are often complex, costly, difficult to scale to large areas18, and not transferable to “unconventional” substrate materials such as scintillator crystals. On the other hand, bottom-up approaches such as laser printing19,20, chemical or self-assembly methods21,22, and topographical control of particle positioning23 have been explored as alternatives. However, these methods come with their own set of limitations, including throughput constraints, surface defects, roughness, and restricted material choices, which also hinder their practical application. To maximize the technological impact of nanophotonic scintillators, it is imperative to develop fabrication techniques that are scalable to industry-standard detector dimensions (~cm), while preserving enhancements obtained from nanophotonic patterning.

Here, we demonstrate a large-scale nanophotonic scintillator fabrication method that realizes a six-fold nanophotonic enhancement in light yield over centimeter scales. Our method is based on the self assembly of chalcogenide glass photonic crystals. With the devised method, we realize nanophotonic scintillators over an area of 4 cm  × 4 cm (e.g., comparable in length scale to commercial flat panel detectors) with six-fold light yield enhancements compared to a reference bare scintillator. We demonstrate this enhancement in a conventional, and widely used X-ray scintillator material (cerium-doped yttrium aluminum garnet, YAG:Ce). We then elucidate the influence of surface disorder in our nanophotonic scintillator’s imaging performance and obtain large-area scans of biological and inorganic specimens. Our results are poised for rapid integration into industrial applications, enabling a family of optimized nanophotonic-enhanced scintillators for use in medicine, defense, and beyond.

Results

Nanophotonic scintillation enhancement over centimeter scales

The depicted fabrication method seamlessly integrates a thin chalcogenide photonic crystal coating atop a scintillating substrate (Fig. 1a, b). Chalcogenide glasses represent a particularly relevant class of materials for nanophotonic scintillation enhancement due their high refractive index (2 ≤ n ≤ 4) and low optical losses from the infrared to the visible spectrum24. Furthermore, thin (sub-100 nm) chalcogenide glass films exhibit viscous behavior during annealing over an extended processing window, making them ideal candidates for templated dewetting processes25.

Fig. 1: Fabrication scheme of self-assembled nanophotonic scintillators.
figure 1

a Fabrication method: (1) soft nanoimprinting of polymer coating (blue) spin-coated over scintillator (yellow); (2) chalcogenide film deposition (orange); (3) annealing and dewetting. b Cross-sectional schematic of nanophotonic scintillator. c Photograph of large scale sample after fabrication of nanostructured layer on a 0.5 mm-thick YAG:Ce scintillator substrate (IL sample). d Top view scanning electron micrograph of the dewetted nano-array (IL sample). e Zoomed-in top view (IL sample). f False color cross-sectional view of a single nanoparticle. From bottom to top, the cross-section shows the polymer coating (blue), a single chalcogenide nanoparticle (orange), and air.

Full size image

This process begins with the fabrication of a silicon master mold by traditional lithographic techniques (see Methods). Both interference lithography and electron beam lithography are used to make distinct molds, enabling different trade-offs between patterned area and resolution. The sample prepared with electron beam lithography extends over an area 4 mm × 4 mm ("EL” in the rest of this work), while the sample prepared with interference lithography extends over a much larger area of 4 cm × 4 cm ("IL” in the rest of this work). In the next process step, nanoimprint lithography is leveraged to reproduce the master texture onto the bulk scintillator. A polydimethylsiloxane stamp, replicated from a silicon master mold, is pressed onto a UV-curable polymer layer directly on the scintillator substrate and exposed to ultraviolet light (Fig. 1a-1). In a final process step, physical vapor deposition and dewetting is used to obtain a high-index nanoparticle array. A thin layer of chalcogenide glass is deposited using thermal evaporation (Fig. 1a-2). A final glass annealing step above its glass transition temperature induces the re-arrangement of the film into an array of highly ordered nanospheroids (Fig. 1a-3). The precise manipulation of interfacial tension as well as film-texture interaction is instrumental in achieving defect-free nanostructures, as exemplified in Fig. 1d, e. The photonic crystal exhibits a subwavelength period of 450 nm, covering a total patterned area of 4 cm × 4 cm, therefore counting around 10 billion nanoscale spheroids on the chip. Finally, a 15 nm SiO2 layer is sputtered over the resulting chalcogenide nanoparticle array to avoid further oxidation.

Using a recent framework to model scintillation emission in nanophotonic structures, we anticipate an enhancement in nanophotonic scintillation by amplifying light yield due to better in/out-coupling of light (which maps to an enhancement in non-equilibrium optical absorption in the scintillator layer, via Lorentz reciprocity)8, as experimentally observed in Fig. 2a, c. Taking the bare scintillator as reference (no coating nor pattern), our simulations indicate an increase in scintillation light yield of 3.48-fold (EL) and 6.96-fold (IL). These enhancement values are confirmed experimentally, with X-ray line scans showed in Fig. 2c, d: 3.00 ± 0.21 (EL) and 6.62 ± 1.60 (IL). Since the fabricated structures are spheroids, our numerical simulations consist of a multi-step process that combines finite element methods and rigorous coupled wave analysis26 (see Methods).

Fig. 2: Large-area nanophotonic enhancement of scintillation.
figure 2

a, c Intensity line scan (EL sample with t1 = 1 mm (a) and IL sample with t2 = 0.5 mm (c)) under X-ray exposure across unpatterned, coated, and patterned areas of the sample. The coating refers to the polymer coating and silica cladding depicted in Fig. 1, without chalcogenide. The pattern refers to the polymer coating, chalcogenide nanoparticle, and silica cladding. b, d Corresponding theoretical predictions for unpatterned and patterned scintillators. Data is normalized to the mean signal value from the unpatterned signal. e Large-area nanophotonic scintillator benchmark: light yield enhancement and patterned area for all comparable devices reported in the literature. The dot color corresponds to the ratio of wavelength λ to photonic crystal periodicity P. Subwavelength designs correspond to λ/P > 1. Works appearing on this plot are refs. 10,44,45,46,47,48,49,50,51,52.

Full size image

The emission enhancement is associated with the coupling of guided modes to free space plane waves, which creates guided resonances. Consequently, the emission enhancement is proportional to the number of guided resonances, and the subwavelength periodic pattern is critical in achieving optimal enhancement27,28. Furthermore, different emission angles may experience different enhancements due to their varying coupling efficiencies. For example, in the region of very large off-normal emission angles, the coupling is weaker and depends on the azimuthal coordinate (decreased enhancement), while at normal and near normal directions, a stronger coupling results in maximum enhancement (see Fig. S2)28,29. Further discussion of the enhancement mechanism and angular dependence is provided in the SI, Section S1.

We attribute the difference in nanophotonic scintillation enhancement between the two samples to several factors. First, we find in our simulations that thicker scintillators generally exhibit less nanophotonic enhancement, which is consistent with an analysis based on density of states27,30 (the samples’ thickness are t1 = 1 mm and t2 = 0.5 mm). Second, we also observe experimentally the influence of different dewetting schemes and mold quality (Compared to the bare scintillator, the coating on the EL sample (respectively, IL sample) slightly reduces (respectively increases) the light out-coupling efficiency. This can be attributed to random pattern formation that occurs when dewetting flat surfaces31.).

Compared to previous works that aimed at realizing large-area micro or nanostructures on scintillators, our work realizes a six-fold nanophotonic enhancement over a patterned area of  >1600 mm2 (which is only limited by the size of the available scintillator substrates). An overview of the state of the art is shown in Fig. 2e. Previous work with comparably large patterned areas realized enhancements  ~1.3. We mainly attribute this five-fold improvement over the state of the art by the use of subwavelength nanophotonic structures. Periodic nanophotonic structures of period P such that λ/P ≳ 1 are known to be optimal in terms of density of states enhancement8,27. The fact that our method is compatible with patterning of high index, subwavelength structures is therefore key in scaling up nanophotonic scintillator technology to areas required for X-ray imaging applications. While we focus our comparison on the patterned areas, the demonstrated scintillators also improve upon previous nanophotonic scintillators8,32 by preserving comparable light yield enhancements while increasing thickness by almost an order of magnitude.

Large-area X-ray imaging

Using the larger area IL sample (4 cm × 4 cm), we realized X-ray scans of inorganic and organic specimens. The X-ray imaging parameters (source voltage and power, as well as geometric and objective magnifications) can be found in the Methods and in the Supplementary Information (SI), Section S4. Each X-ray scan is taken in conjunction with an X-ray flat field image used for post processing, and the final brightness and contrast are digitally adjusted (as would be done in a commercial X-ray scanner for industrial or medical applications).

We first image parts of a chicken foot (tarsometatarsus and digits, shown in Fig. 3a–e. We clearly see several phalanges separated by interphalangeal joints: two proximal and middle phalanges in Fig. 3b, c, and two intermediate and distal phalanges in Fig. 3d, e. We then image a USB stick in Fig. 3f–h and can distinguish multiple levels of printed circuits overlayed on top of each other. These images demonstrate the potential of our scintillators to realize X-ray scans of centimeter-large objects with nanophotonic enhancement.

Fig. 3: X-ray imaging with large-area nanophotonic scintillators.
figure 3

a, f Photograph of different objects imaged through our customized imaging setup. The colored square denotes estimated field of views. b, d, g Corresponding raw X-ray images. c, e, h Corresponding flat-field corrected and contrast-adjusted X-ray images.

Full size image

Controlling spatial resolution with disorder in photonic crystal scintillators

Next, we elucidate the influence of fabrication disorder on the spatial resolution of nanophotonic scintillators.

When comparing atomic force microscopy (AFM, shown in Fig. 4c, d) images of the two samples, we observe various levels of disorder which we attribute to different lithography methods used to realize the nanoimprinting mold. A disorder distribution is extracted for both samples using Fourier analysis (see SI, Section S3 and Fig. 4b). We also experimentally characterized each sample’s spatial resolution by X-ray imaging the sharp edge of a razor blade, with a setup shown in Fig. 4a. In particular, we fitted an error function to the image of the razor blade and extracted its full width at half maximum (FWHM) σ as a metric of the spatial resolution, although due to the shape of the razor blade’s edge, it should not be interpreted as the “true” spatial resolution of the imaging system (further discussion can be found in the SI, Section S4D).

Fig. 4: Controlling spatial resolution with disorder in photonic crystal scintillators.
figure 4

a Experimental setup for X-ray imaging with large-area nanophotonic scintillators. OBJ = object; LENS = microscope objective; CAM = CMOS camera. b The transfer function of the disordered photonic crystal is modeled as a convolution between that of an ordered photonic crystal and that of a disordered height map. c, d Atomic force micrographs (AFM) for EL (c) and IL samples (d). e, f Corresponding spatial resolution measurements in the presence of a razor blade to block part of the incoming X-rays. We note that our imaging system was focused by hand without tracking the position of the focal plane. Therefore, the focal plane when imaging the EL sample may have been closer to the front (patterned) face of the sample than in the case of the IL sample, resulting in, effectively, lower blur and higher measured spatial resolution. g Disorder distribution of the two samples. h Relative decrease in spatial resolution (increase in blur) as a function of the disorder root mean square (RMS). Full lines are calculated for fields propagating through the whole scintillator (thickness t), dashed lines for its effective thickness teff taking into account X-ray absorption. The dots represent measurements for the two samples (EL and IL) in this paper.

Full size image

Generally, we observe a correlation between greater amounts of nanofabrication disorder and a decrease in the scintillator’s spatial resolution (blur increase). Specifically, the nanofabricated pattern on the EL sample (Fig. 4e) has a spatial resolution of σ = 0.286 ± 0.026 mm, compared to that of a bare, exposed surface on the same sample of σ = 0.270 ± 0.013 mm. This indicates no significant influence on the spatial resolution of the scintillator (with a slight relative decrease within experimental uncertainty). However, the IL sample (Fig. 4f) exhibits a decrease in relative spatial resolution by a factor of 2.508 ± 0.099 (from σ = 0.993 ± 0.029 mm for the patterned surface to σ = 0.396 ± 0.004 mm for the bare surface). Dependence on the X-ray energy of the spatial resolution is analyzed in the SI, Section S5. A limitation of our measurement of spatial resolution is that our imaging system was focused by hand to achieve the sharpest possible image, without precisely tracking the position of the focal plane. Therefore, it is possible that the focal plane when imaging the EL sample was closer to the front (patterned) facet than the case of the IL sample, resulting in, effectively, lower blur. In the future, this can be mitigated by either tracking the focal plane or keeping it constant between measurements.

To account for the influence of disorder on spatial resolution, we adapted the framework of stochastic surface transfer functions33 to X-ray scintillation imaging. The disorder distribution from AFM measurements is modeled as a Gaussian-distributed surface transfer function that blurs optical waves incident from within the scintillator. We calculated the relative decrease in spatial resolution for both samples (corresponding to an increase in optical blur, shown in Fig. 4g), with scintillation emission happening on the front facet of the scintillator (propagation through thickness t) and at the mean X-ray absorption position (propagation through thickness teff). The relative decrease in resolution is defined as the ratio of full widths at half maxima of the line spread functions (for a disordered photonic crystal vs. a flat scintillator surface) – a value  > 1 corresponding to a decrease in resolution, greater blur, and coarser features of the X-ray scan. These ab initio disorder simulations agree within 7% (comparing data from the EL sample and simulations for t1,eff) and 15% (comparing data from the IL sample and simulations for t2,eff). The remaining discrepancy originates from the uncertainty in the depth at which the imaging objective is focused. More details on the spatial resolution measurement method, disorder modeling, AFM analysis, and calculation of the effective thickness, can be found in the SI, Section S2-3.

Overall, the 0.5-mm patterned scintillator matches the photon yield of a 3-mm unpatterned scintillator (σ = 2.78 mm) yet achieves far better resolution (σ = 0.993 ± 0.029 mm). Even with disorder-induced degradation, the resolution still exceeds (about 3-fold) that of a 3-mm unpatterned scintillator (see Fig. S5), overcoming the intrinsic trade-off between light yield and resolution11,16.

Discussion

We have presented a platform for nanophotonic scintillation that combines the following key features: (1) compatibility with greater than centimeter-scale fabrication methods, to enable large-area X-ray imaging; (2) subwavelength feature sizes, to maximize nanophotonic outcoupling8,27; (3) absence of residual layer, which may reduce nanophotonic enhancement due to impedance mismatch34; (4) ability to realize low-loss high-index (n > 2) nanostructures, for greater nanophotonic control25,35; (5) high repeatability since a master mold can be used to generate thousands or tens of thousands of large-area nanophotonic scintillators36,37. The combination of these factors allowed us to demonstrate scintillation enhancement over scales commensurate with that of commercial X-ray flat panel detectors and provides a path towards their mass production. Unlike bulk nanomaterial-based scintillators, where self-absorption may limit light collection, our approach minimizes this effect by restricting patterning to a very shallow surface layer while preserving bulk scintillator absorption efficiency.

We have also established a clear correlation between nanofabrication disorder and spatial resolution decrease. Improvements in mold quality realized with interference lithography will lead to lower amounts of nanofabrication disorder and little to no decrease in the scintillator’s spatial resolution, while improving scintillation light yield six-fold.

Our work is also a first step towards the realization of large-scale metaoptics on scintillators, since the realized method is amenable to local control of the spheroid’s shape and induced phase shift25. Proper design and optimization of metasurface masks could lead to coincident enhancements in spatial resolution and light yield11, as well as local control of the surface’s transfer function35,38,39,40.

Furthermore, the inherent scalability of our fabrication technique opens the door to patterning entire rolls of scintillator materials, paving the way for rapid and affordable mass manufacturing of nanophotonic scintillators. Our method is in principle substrate agnostic, as long as patterns with subwavelength periodicity27 (compared to the scintillator emission wavelength) can be realized with the master mold. By enabling nanoscale control over electromagnetic properties with controllable disorder, our method ensures both high performance and robustness in practical applications. These developments hold the potential to advance X-ray imaging, as they allow for the integration of complex nanophotonic structures over large areas without compromising quality or significantly increasing costs. These improvements may facilitate the wider adoption of nanophotonic scintillators in various fields requiring high-resolution, high-sensitivity detection of X-rays and other high-energy particles.

Methods

Templated dewetting of chalcogenide on scintillator materials

Mold Fabrication

Two separate molds are prepared for nanoimprinting. A first mold (Mold 1) is prepared by electron beam lithography over an area of 4 mm × 4 mm. Following oxygen plasma cleaning and HMDS monolayer deposition, a negative electron beam resist (maN 2403, MicroResist Technology GmbH, Germany) is spun at 3000 rpm on a silicon wafer covered with a 30 nm native oxide. The resist is soft baked at 90 °C for \(2\,\min\). An inverted grid pattern is written using a 50 keV electron beam and 10 nA current. The resist is developed using a TMAH-based developer (AZ-726, MicroChemicals GmbH, Germany) for \(2\,\min\). The native oxide is selectively etched using ICP reactive ion etching (RIE 230iP, Samco, Japan) with fluorine chemistry. A second mold (Mold 2) is prepared by interference lithography over an area of 40 mm × 40 mm. A HeCd laser source (λ = 325 nm) is directed at a pinhole placed approximately 60 cm away from a Lloyd’s mirror setup. An antireflective coating layer is spun at 3500 rpm onto a silicon wafer with a 30 nm native oxide (AZ Barli II 90, iMicroMaterials, Germany), and baked at 180 °C for 1 min. A layer of positive photoresist (AZ 3312, iMicroMaterials, Germany) is spun at 5000 rpm over the antireflective coating, followed by a soft bake at 110 °C for 1 min. The second mold is exposed with an MLA 150 Advanced Maskless Aligner (Heidelberg Instruments, Germany) at 135 μC/cm2, followed by a post-exposure bake at 110 °C for 2 min. The photoresist is developed in AZ 726 for 2 min. The native oxide and antireflective coating are etched using ICP reactive ion etching (RIE 230iP, Samco, Japan) with fluorine chemistry. Both resulting molds are then stripped with an oxygen plasma cleaning step (e3511 Plasma Asher, ESI, USA). The resulting silicon wafer is placed in a 25% KOH solution at 60 °C for anisotropic silicon etching for \(2\,\min\). The native oxide hard mask is stripped with a \(1\,\min\) dip in diluted HF 10: 1.

Nanoimprint Lithography

The resulting silicon mold is treated with an anti-sticking layer (Trichloro(1H,1H,2H,2H-perfluorooctyl)silane, Millipore Sigma, USA) in a vacuum desiccator following a short oxygen plasma surface activation. A (poly)dimethylsiloxane layer (PDMS Sylgard 184, Corning, USA) is drop-casted on the silanized mold and cured at 80 °C for 2 h. Upon curing, the PDMS layer is peeled off from the mold.

Scintillator Patterning

A thin layer of diluted UV-Curable polymer (Ormocer, MicroResist Technology, Germany) is spun onto the YAG:Ce bulk scintillator. The PDMS nanoimprint mold is pressed directly onto the thin polymer layer. A UV light source (λ = 375 nm) is shone through the PDMS to cure the polymer film, with a dose  >1500 mJ/cm2. A sub-100 nm thin chalcogenide layer is deposited by thermal evaporation (PVD Products, USA) directly onto the patterned polymer film. The chalcogenide is annealed above its glass transition temperature to dewet according to the underlying texture. A 15 nm SiO2 layer is sputtered over the resulting chalcogenide nanoparticle array to protect it from the surrounding oxidative environment (ATC Sputtering System, AJA International, USA).

X-ray imaging experiments

All experiments, including imaging and measurements of scintillation enhancement and spatial resolution, were done using a custom-built experimental setup inside of a ZEISS Xradia Versa 620 micro-CT. Images were captured using a Hamamatsu ORCA-Fusion C14440-20UP CMOS camera along with a wide-field-of-view camera lens (Edmund Optics 33-304). A narrow bandpass filter centered at 550 nm (AVR Optics FF01-560/14-25, 14 nm bandwidth) was placed in front of the camera lens to minimize unwanted background from other wavelengths. In all experiments, the source was ds = 150 mm away from the scintillator, while the object distance do depended on the desired geometric magnification of each image, defined as Mg = ds/do. Here, we summarize the most important details of each experiment; further details of the experimental setup, methods, underlying theory, and data processing can be found in the SI.

Scintillation enhancement

Scintillation enhancement was measured by directly imaging the scintillator under excitation from X-rays, comparing patterned and unpatterned areas of the same sample. The line profiles shown in Fig. 2 were extracted from these images. The EL sample was excited by X-rays at 60 kVp and 6.5 W, while the IL sample was excited by X-rays at 150 kVp and 23 W. The scintillation enhancement is calculated using the formula (IP − IB)/(IU − IB), where IP, IU, IB are the average patterned, unpatterned, and background intensities from the line profile.

Imaging

To capture an image, an object was placed between the source and the scintillator. For each object, the focal plane—usually somewhere inside the scintillator rather than coplanar with the patterned surface—was adjusted by slightly moving the camera lens back and forth to capture the sharpest image. Brightness and contrast were adjusted by carefully tuning the X-ray energy (i.e., kVp), exposure time, and pixel binning and subarray. Figure 3b, d were captured under the following conditions: X-ray energy and power of 30 kVp and 2 W, exposure time of 10 s, binning of 4, and subarray of 144 × 144, and geometric magnification of 1.2. Finally, Fig. 3g was captured under the following conditions: X-ray energy and power of 60 kVp and 6.5 W, exposure time of 7 s, binning of 1, subarray of 576 × 576, and geometric magnification of 2. Post-processing included flat-field correction and digitally adjusting brightness and contrast (Fig. 3c, e, and h). The flat-fields were captured by simply removing the objects without changing any of the aforementioned parameters. The chicken foot used in this study was obtained from a local grocery store and prepared for X-ray imaging following standard procedures, including vacuum sealing in a plastic bag to prevent contamination. The USB stick was a 3.0 USB Flash Drive Pen from OneSquareCore.

Spatial resolution

The spatial resolutions of the nanophotonic scintillators were estimated by imaging a carbon steel razor blade. Ideally, a region of interest drawn across the edge of the razor blade looks like a blurred, two-dimensional step function that can be reasonably approximated by an edge spread function (ESF) of the form \({{{\rm{ESF}}}}(x,y)=(A/2){{{\rm{erf}}}}\left[(x-\mu )/\sqrt{2}\sigma \right]\), where the spatial resolution is \(2\sqrt{2\ln 2}\sigma \approx 2.3548\sigma\). By fitting the raw data (not the flat-field corrected data, which may remove the effects of blurring due to thickness), σ can be estimated. In Fig. 4e–f, the measured ESFs were captured under the following conditions: X-ray energy and power of 90 kVp and 12 W, exposure time of 3 s, binning of 1, and geometric magnification of 2. The EL sample (Fig. 4e) used a subarray of 144 × 144, while the IL sample (Fig. 4f) used a subarray of 576 × 576.

Modeling nanophotonic enhancement

Finite element + rigorous coupled wave analysis modeling

The nanophotonic scintillation enhancement is modeled using a three-step approach. The large thickness (~mm) of the scintillator and the complex geometry of the nanophotonic structure make it challenging to use a single computational tool with high efficiency and low computational cost. Therefore, a combination of finite element (FE) and rigorous coupled wave analysis (RCWA) methods is employed. The nanophotonic scintillator consists of a chalcogenide spheroid with a diameter of about 395 nm, a subwavelength period of 450 nm, a polymer coating of 450 nm thickness, and scintillator substrate. The refractive index of SiO2 is obtained from literature41, while that of the Ormocer polymer (1.5) and YAG:Ce are given by the suppliers. The refractive index of chalcogenide is obtained from in-house ellipsometry measurements (see SI, Fig. S2). All calculations were carried out in the wavelength range of 540 to 560 nm, with the YAG:Ce emission peak centered at 550 nm.

We utilize a commercially available FE solver, COMSOL Multiphysics®, to model the electromagnetic response of the spheroid nanophotonic structure. Initially, we simulate the superstrate without YAG:Ce (polymer coating and chalcogenide spheroid), then combine it with a thinner (1 μm) YAG:Ce scintillator to optimize the geometry. Next, we replace the spheroids with multiple stacked cylinders that approximate the electromagnetic response (transmission and reflection) of the full spheroid structure calculated in the first step (still with an FE solver). This is so that we can shift from FE simulations—where simulating a thick substrate would be difficult to model—to RCWA simulations, where we can use our approximated structure and a thick substrate since it is a semianalytical method. In the third step, we use an automatically differentiable RCWA solver26, to simulate the approximate geometry with the thick substrate. The absorption/emission (via reciprocity) within/from the volume of the scintillator is calculated for both polarizations (transverse electric (TE) and transverse magnetic (TM)) and averaged to mimic unpolarized light. Finally, we calculate the enhancement factor as the ratio of the spectrally integrated emission of the patterned scintillator to the unpatterned scintillator.

Influence of disorder on spatial resolution

The image degradation due to disorder within the nanophotonic structure is analyzed using surface scatter theory based on the linear shift-invariant system formulation42. This is executed in several steps, following traditional image formation theory. First, the modulation transfer function (MTF) of the nanophotonic spheroids is calculated using the transfer function retrieved directly from the transmission as a function of angle (using RCWA). Consequently, the disorder is modeled using the surface scatter theory to calculate the surface transfer function (STF). The scattering property by the disorder (defects) is considered in transmission mode to formulate the STF (see SI, Section S2). Here, we utilize a Gaussian autocovariance function with surface roughness root mean square (RMS) retrieved from AFM images using Fourier analysis (see SI, Section S3).

Further, we define the system modulation transfer function (MTFsys), which incorporates the influence of nanophotonic spheroids and disorder, via the two transfer functions: MTF and STF (multiplication in the Fourier space). MTFsys provides a complete linear system formulation of image quality as degraded by surface scatter effects due to disorder from residual optical fabrication errors. Once the MTFsys is derived, we proceed to obtain the line spread function (LSF) through an inverse Fourier transform to real space. This is followed by propagation through the optical imaging setup, encompassing free space and optical components. The full width at half maximum (FWHM) of the LSF serves as the defining measure of the system’s resolution. These procedures are consistently applied to both EL and IL samples in both effective (0.76, 0.36 mm) and real (1, 0.5 mm) thicknesses (see definition of the effective thickness in SI, Section S2). Subsequently, the FWHM of the nanophotonic structure is compared with that of a smooth unpatterned scintillator to calculate the relative spatial resolution as a function of RMS (1–11 nm). The calculations above are performed by considering the peak emission wavelength (550 nm) of the scintillator.