Inorganic Cs₃Bi₂I₉ lead-free halide perovskite film for large-area X-ray detector via low-cost ambient spray coating

Abstract

Lead-free Cs₃Bi₂I₉ single crystals have been demonstrated to be promising materials for direct X-ray detectors with remarkable performance. However, their application for 2D X-ray imaging is hindered by their time-consuming preparation and limited crystal size. In this paper, a thick Cs₃Bi₂I₉ perovskite film fabricated via facile spray coating at a low processing temperature, which increases the area of the photoactive film, reduces the processing time, decreases the energy budget and the production cost, and enhances the production yield due to high material utilization, has great potential for commercial applications. Careful control of the processing temperature and intervals during spray coating results in a dense and thick perovskite film with well-stacked perovskite domains. The compact perovskite film enhances the charge transport capability of the Cs₃Bi₂I₉ perovskite film and reduces the dark current density of the X-ray detector. The resultant X-ray detector, prepared through a two-step spray coating process, exhibited a sensitivity of 127.23 μC Gy_air⁻¹ cm⁻² and a detection limit of 7.4 μGy_air s⁻¹. In addition, the device delivers long-term stability with a consistent photoresponse when exposed to consecutive X-ray pulse irradiation.

Introduction

X-ray detection has extensive applications in medical diagnostics^1,2, nondestructive industrial radiography^3,4, and security inspection. However, the ionizing radiation generated by high-dose X-rays poses a threat to human DNA, increasing the risk of cancer for patients. Therefore, having a high-sensitivity X-ray detector with a low detection limit is crucial in medical imaging to mitigate cancer risk. Recently, metal halide perovskite semiconductors have attracted enormous attention for their potential application in X-ray detectors owing to their promising optoelectronic properties^{5,6,7,8,9,10,11}. Halide perovskites exhibit a remarkable attenuation coefficient when exposed to high-energy X-ray irradiation, attributed to their composition containing high atomic number (Z) elements, such as Pb and I¹². Moreover, the large mobility-lifetime product of perovskites (ranging from 10⁻⁴ to 10⁻² cm² V⁻¹ for single crystals and 10⁻⁷ to 10⁻³ cm² V⁻¹ for thin films, compared to 10⁻⁷ cm² V⁻¹ for amorphous Se (a-Se)) enables efficient transport of photogenerated carriers within the perovskite layer, ensuring effective charge collection by the electrodes^5,6,7,13. Their easy, versatile, and low-temperature fabrication processes make them promising candidates for direct X-ray sensing.

Various halide perovskites, such as methylammonium lead iodide (MAPbI₃)¹⁴, methylammonium lead bromide (MAPbBr₃)^15,16, cesium lead bromide (CsPbBr₃)¹⁷, and 2D layered PEA₂PbBr₄ perovskite (PEA: phenethylammonium)¹⁸, have been employed in X-ray detectors, delivering superior sensitivity compared to that of commercial a-Se-based direct X-ray detectors. The low mobility-lifetime product of a-Se (on the order of 10⁻⁷ cm² V⁻¹)⁶ results in a high device bias to effectively collect charge, leading to reduced detection sensitivity (20 μC Gy_air⁻¹ cm⁻²) and an increased detection limit (5.5 μGy_air s⁻¹). However, organic‒inorganic hybrid perovskites suffer from degradation upon exposure to moisture and oxygen^19,20.

Moreover, the use of toxic lead remains a concern for the commercial application of X-ray detectors. Lead-free perovskites, which replace the B-site lead cations with nontoxic metal cations such as Sn, Bi, Sb, In, Ag, Ti, and Na, offer an eco-friendly alternative^21,22,23. Among the lead-free perovskites, the Bi-based lead-free perovskites exhibit comparable X-ray attenuation to that of lead-based perovskites owing to Bi having the largest atomic number among the nonradioactive elements^8,9,10. Consequently, X-ray detectors employing Bi-based lead-free perovskites demonstrate a comparable performance to those using lead-based perovskites^24,25.

High-performance X-ray detectors using halide perovskite single crystals have been demonstrated via the synthetic routes of solution temperature-lowering crystallization (STLC), inversion temperature crystallization (ITC), antisolvent vapor-assisted crystallization (AVC), and solvent volatilized crystallization (SVC) methods²⁶. For instance, Zhang et al. introduced the ITC method to demonstrate a durable and high-performance X-ray detector using inorganic Cs₃Bi₂I₉ single crystals as the photoactive layer. Moreover, the device exhibited long-term stability in ambient conditions and outstanding thermal stability at 100 °C²⁷. The constant change in temperature for Cs₃Bi₂I₉ single crystals grown by the ITC method resulted in a variation in the crystal growth rate, which induced the formation of defects and thereby reduced the quality of the Cs₃Bi₂I₉ single crystals. To address this issue, Wei et al. conducted the liquid diffused separation-induced crystallization (LDSC) method at a low temperature to produce Cs₃Bi₂I₉ single crystals for X-ray detection. This approach minimized the temperature fluctuations and thermal convection to promote ordered growth and reduce the defect density²⁸.

Yang et al. employed the refined evaporation method, which is based on the SVC method, with a constant temperature of 50 °C to produce a constant and slow growth rate for synthesizing high-quality Cs₃Bi₂I₉ single crystals. The as-grown Cs₃Bi₂I₉ single crystal exhibited a low defect density, a high resistivity, a large mobility-lifetime product (ranging from 10⁻⁴ to 10⁻³ cm² V⁻¹), and good stability. The device with a lateral structure achieved a record sensitivity of 59,464.4 μC Gy_air⁻¹ cm⁻² and a low detection limit of 1.35 × 10⁻² μGy_air s^{−1 29}. Although the SVC method ensures a constant growth rate, its requirement of a relatively low temperature (<60 °C) to avoid the formation of redundant crystals prolongs the growth period.

To improve the quality of Cs₃Bi₂I₉ single crystals, Zhang’s group introduced the top-seed solution (TSS) method and controlled the temperature gradient to regulate the growth rate. When the temperature gradient was in the range of 70–80 °C, the as-grown single crystal for the X-ray detector exhibited a sensitivity of 4382 μC Gy_air⁻¹ cm⁻² and a low detection limit of 7.93 × 10⁻³ μGy_air s^{−1 30}. During the crystal growth, the different solubilities of CsI, BiI₃, and small Cs₃Bi₂I₉ crystals in the precursor solution led to changes in the concentration profiles of Cs⁺, Bi³⁺, and I⁻ in the solution, changing with the temperature field, which further caused a deviation in the stoichiometric ratio for the as-grown Cs₃Bi₂I₉ single crystals. The composition deviation gave rise to vacancy defects in the Cs₃Bi₂I₉ single crystals and deteriorated their electronic properties. The same group optimized the temperature gradient and the CsI:BiI₃ ratio of 3.3:2 in the precursor solution to produce high-quality Cs₃Bi₂I₉ single crystals whose stoichiometric ratio was close to the theoretical ratio. Increasing the CsI content in the precursor solution effectively reduced the number of vacancy defects in the Cs₃Bi₂I₉ single crystals. The resultant X-ray detector achieved a remarkable sensitivity of 29,386.2 μC Gy_air⁻¹ cm⁻² and an extremely low detection limit of 3.6 × 10⁻⁴ μGy_air s^{−1 31}.

A low defect concentration in single-crystalline perovskites can reduce carrier trapping within the perovskite bulk, leading to superior X-ray detector performance. However, the time-consuming synthesis (up to a month) and limited crystal size (<1 mm²) hinder practical applications in direct X-ray detection. To address this issue, Bu et al. applied a hot-pressing process to fabricate a high-quality Cs₃Bi₂I₉ polycrystalline wafer with a low defect density, a high resistivity, a large mobility-lifetime product (on the order of 10⁻⁴ cm² V⁻¹), and a diameter of 2 cm. The fabricated X-ray detector with an active area of 3.14 mm² delivered a sensitivity of 230.46 ± 19.86 μC Gy_air⁻¹ cm⁻² and a detection limit of 6.13 × 10⁻² μGy_air s⁻¹ under 40-kV X-ray irradiation and an electric field of 40 V mm^{−1 32}.

Wei et al. further incorporated MXene nanoflakes, which exhibit excellent metallic conductivity, high electron mobility, and mechanical flexibility, in Cs₃Bi₂I₉ perovskite to produce an MXene/Cs₃Bi₂I₉ composite wafer by a mechanical pressing process. The introduction of MXene nanoflakes into the polycrystalline Cs₃Bi₂I₉ perovskite wafer improved the mobility-lifetime product (on the order of 10⁻³ cm² V⁻¹) due to the effective carrier transport from the Cs₃Bi₂I₉ perovskite to the high-conductivity MXene nanoflakes. An X-ray detector based on the MXene/Cs₃Bi₂I₉ composite wafer showed a sensitivity of 368 μC Gy_air⁻¹ cm⁻² and a detection limit of 0.23 μGy_air s^{−1 33}. Although the above results indicated the promise of Cs₃Bi₂I₉ perovskite for direct X-ray imaging, the limited size of the single crystals and wafers restricted the ability of the device to image small objects.

There is a critical need for a time-effective fabrication process to produce large-area perovskites that can meet the demands of real applications. The development of an upscaling method for preparing large-area inorganic perovskite films is necessary for X-ray detectors. Various methods for scaled-up fabrication, including inkjet printing³⁴, tablet pressing³⁵, bar coating³⁶, blade coating^16,37, and spray coating³⁸, have been employed to prepare large-area perovskite films. Among these methods, the spray coating method stands out as a cost-effective approach for depositing large-area and thick perovskite films due to its simplicity, high precursor utilization rate, and compatibility with various substrates³⁹. In this work, for the first time, we applied a low-cost spray coating to fabricate a thick Cs₃Bi₂I₉ perovskite film as an active layer for direct X-ray detection under ambient conditions. The device structure consists of FTO/TiO₂/SnO₂/Cs₃Bi₂I₉/Au. We optimized the processing temperature and time intervals during spray coating to improve the perovskite domain stacking, produce a compact perovskite film, reduce the dark current density, and enhance the sensitivity of the X-ray detector. The fabricated X-ray detector based on the Cs₃Bi₂I₉ perovskite presented a sensitivity of 127.23 μC Gy_air⁻¹ cm⁻² along with a detection limit of 7.4 μGy_air s⁻¹ and it maintained durability after consecutive X-ray pulse irradiation. Notably, we achieved a record-large active area of 1.5 cm² for the Cs₃Bi₂I₉-based X-ray detector.

Materials and methods

Fabrication of the X-ray detector

The structure of the X-ray detector was composed of FTO/TiO₂/SnO₂/Cs₃Bi₂I₉ perovskite/Au. The pre-patterned FTO on the glass substrate was sequentially cleaned via ultrasonic treatment in deionized (DI) water, acetone, and isopropanol. An ultrathin TiO₂ layer was deposited onto the cleaned FTO substrates by spraying a solution of titanium diisopropoxide bis(acetylacetonate) diluted in ethanol (20 mM). During the spray pyrolysis, the FTO substrate was maintained at 470 °C on a hot plate, and each spray was pumped out using a pulverizer (Glaskeller) with a carrier gas of oxygen. Ten milliliters of the diluted solution were spray-coated onto the FTO substrates with dimensions of 10 × 20 cm². After completing the spray deposition, the substrate was sintered at 500 °C for 30 min. Subsequently, a SnO₂ colloidal dispersion (tin(IV) oxide, 15% in H₂O colloidal dispersion; Alfa Aesar) diluted in DI water at a 1:1 volume ratio was spin-coated onto the TiO₂ layer at 5000 rpm for 30 s. Following the spin-coating process, the substrate was annealed at 150 °C for 30 min under ambient conditions (RH = 50–60%). The thickness of the as-deposited TiO₂/SnO₂ bilayer was approximately 50 nm, as seen in the cross-sectional SEM image shown in Fig. S1.

The perovskite precursor solution for Cs₃Bi₂I₉ was prepared by dissolving bismuth iodide (BiI₃) (0.2 M) and cesium iodide (CsI) (0.3 M) in a mixed solvent system of dimethylformamide (DMF) and dimethyl sulfoxide (DMSO) (v:v = 4:1). Prior to spraying the precursor onto the substrate, the substrate was treated with UV–ozone to enhance the wettability of the perovskite precursor. The perovskite precursor was then spray-coated onto the preheated substrate. Approximately 0.3 mL of the precursor being pumped out for each spray using a pulverizer (Glaskeller) and a carrier gas of nitrogen. The handheld pulverizer moved in a zigzag trajectory to spray-coat the perovskite precursor onto four 2.5 × 2.5 cm substrates in a single round. The perovskite film thickness was controlled by adjusting the number of deposition cycles.

Two fabrication processes, referred to as Method-1 and Method-2, were conducted to deposit the Cs₃Bi₂I₉ perovskite on the TiO₂/SnO₂-coated FTO substrate. The deposition of perovskite on the substrate was conducted within a fume hood. To minimize the risk of harm from DMF/DMSO solvent evaporation, operators were required to wear goggles and a half-facepiece reusable respirator equipped with a vapor canister (such as 3 M 6200 or 3 M 6006). Following the deposition of the perovskite film, a 60 nm thick gold electrode was thermally evaporated onto the top of the perovskite film under a pressure of 5 × 10⁻⁶Torr. The device’s active area was defined by the gold electrode with an active area of 1.5 cm².

Characterizations

The grazing-incidence X-ray diffraction (GIXRD) profiles of the perovskite film surface were characterized by a grazing X-ray diffractometer using a Bruker D8 Advanced ECO XRD spectrometer with Cu Kα radiation (λ = 1.54184 Å) at 40 kV and 40 mA. A grazing-incidence angle of 1°, scan rate of 3° per minute, and step angle of 0.1° were used for the experiment. The X-ray diffraction (XRD) profiles of the bulk perovskite were collected using Bruker D8D Plus-TXS with the θ-2θ method and Cu Kα radiation. The in-tube voltage was 45 kV, the tube current was 100 mA, and the scanning rate was 2.67° per minute with a step angle of 0.02°. The surface and cross-sectional morphology of the perovskite film were measured using a multifunction environmental field emission scanning electron microscope (SEM) (SU5000, HITACHI). The valence band maximum (VBM) and work function of the perovskite film were determined through ambient pressure photoemission spectroscopy and Kelvin probe measurements (APS04, KP Technology), respectively. The absorption spectrum of the perovskite film was recorded using a UV‒Vis spectrophotometer (U-4100, HITACHI). The current density‒voltage (J–V) characteristic measurements were conducted by irradiating the device with X-rays, and the device was connected to a source meter (Keithley 2401) for tracing the J–V data. An integrated X-ray table system (MyVet Table X500) with an X-ray tube (E7239X, Toshiba) served as the X-ray irradiation source. The X-ray dose was calibrated using a dose meter (Cobia Flex R/F).

Results and discussion

The fabrication process for the Cs₃Bi₂I₉ perovskite film and the X-ray detector is presented in the experimental section. In this study, we introduced two fabrication methods, Method-1 and Method-2, to prepare a Cs₃Bi₂I₉ perovskite photoactive layer on a TiO₂/SnO₂-coated FTO substrate. Figure 1a, b illustrate the process of spray coating perovskite films using Method-1 and Method-2, respectively. In Method-1, we sprayed the perovskite precursors onto the substrate, which was preheated to 130 °C, for a total of 200 cycles, with a time interval of 20 s between each spray. The perovskite precursor was pumped out using a pulverizer with nitrogen as the carrier gas for each spray. Following spray coating, the substrate covered with perovskite film was annealed at 150 °C (close to the boiling temperature of DMF, which is 153 °C) for 30 min to remove the mixed DMF/DMSO solvent.

Fig. 1: Fabrication process and morphology of spray-coated Cs₃Bi₂I₉ perovskite films.

Process flow of spray-coating perovskite films fabricated by a Method-1 (single process: 200 cycles of spray coating at 130 °C with 20 s intervals) and b Method-2 (two-step process; first step: 20 cycles of spray coating at 110 °C with 40 s intervals; second step: 180 cycles of spray coating at 130 °C with 20 s intervals). Cross-sectional SEM images of Cs₃Bi₂I₉ perovskite films prepared by c Method-1 and d Method-2. The scale bars in c and d are 30 μm. The insets in c, d are the corresponding top-view SEM images. The scale bar in the insets is 10 μm.

For Method-2, a two-step spray coating process was employed to enhance the quality of the perovskite layer. In the first step, we sprayed the perovskite precursor 20 cycles onto the substrate, which was preheated to a lower temperature of 110 °C compared to Method-1, with a time interval of 40 s. After completing the first 20 cycles of spray coating, the second step of 180 cycles of spray coating with a time interval of 20 s was implemented on the substrate heated to 130 °C. After the second step, the spray-coated perovskite film was annealed at 150 °C for 30 min.

Figure 1c, d present cross-sectional SEM images of Cs₃Bi₂I₉ perovskite films spray-coated onto TiO₂/SnO₂-coated FTO substrates using Method-1 and Method-2, respectively. The corresponding top-view SEM images are presented in the insets of Fig. 1c, d. The pinhole ratio within the top-view SEM framework is estimated to be 2.23% and 0.17% for the perovskite films prepared by Method-1 and Method-2, respectively, using ImageJ software. Similarly, within the cross-sectional SEM framework, the pinhole ratio is estimated to be 3.64% and 2.32% for the films prepared by Method-1 and Method-2, respectively, using ImageJ software.

In the perovskite film prepared by Method-1, an uneven surface and apparent pinholes are evident from the top-view SEM image. The cross-sectional SEM image of the perovskite film prepared by Method-1 reveals multiple columnar perovskite domains of varying sizes and disordered orientations arranged closely at the bottom of the film, as shown in Fig. 1c. Additionally, chunky perovskite domains with uneven sizes further stack on top of these multiple columnar perovskite domains. This stacking of uneven perovskite domains leads to the formation of large pinholes, affecting the continuity of the perovskite film and hindering charge transport⁴⁰. This disordered stacking of perovskite domains is presumably attributed to the rapid growth of perovskite crystals when the substrate is heated to a high temperature of 130 °C. During spray coating of the perovskite precursor, this elevated temperature leads to fast solvent evaporation, resulting in rapid crystallization of the perovskite. Furthermore, the solvent of the subsequent spray cannot fully dissolve the previously formed perovskite crystals to undergo recrystallization and grain-merging processes. Consequently, we observe disordered stacking of the perovskite domains³⁸.

To enhance the film quality, we implemented a two-step process, referred to as Method-2, with some modifications to the spray coating conditions. In the first step, a lower temperature of 110 °C was used for the initial 20 cycles of spray coating. This lower temperature facilitates the redissolution of preformed perovskite domains. Because the lower temperature does not entirely evaporate the solvent, surface melting of the as-deposited perovskite film occurs. Additionally, we extended the time interval from 20 s to 40 s. This adjustment provided sufficient time for the pregrowth perovskite crystals to undergo recrystallization and grain merging, ultimately resulting in the formation of a smooth and densely packed surface, as shown in the top-view SEM image, and columnar perovskite domains, as shown in the cross-sectional SEM image.

The cross-sectional SEM image shows multiple columnar perovskite domains with uniform sizes that are well-arranged at the bottom of the film, as illustrated in Fig. 1d. Following the first 20 cycles of spray coating, a second step of 180 cycles of spray coating with a time interval of 20 s was performed on the substrate, which was heated to 130 °C (consistent with the processing temperature in Method-1). During this second spray-coating process, the well-arranged perovskite domains at the film bottom facilitate the stacking of additional perovskite domains on top. It is worth noting that chunky perovskite domains with even sizes consecutively stack atop the multiple columnar perovskite domains. Due to this orderly stacking of perovskite domains, the perovskite film prepared via Method-2 presents a compact arrangement of perovskite domains, resulting in smaller pinholes compared to those prepared via Method-1. This is advantageous for charge transport within the perovskite film.

To further determine the impact of the first-step processing temperature on the spray-coating perovskite film, we fabricated a Cs₃Bi₂I₉ perovskite film using two-step processing conditions similar to Method-2, except for the first-step processing temperature. In the first step of spray-coating, we sprayed the perovskite precursor solution on the substrate preheated at 80, 110, and 130 °C for 20 cycles with a time interval of 40 s. Subsequently, a second step of 20 cycles of spray-coating the perovskite precursor, with a time interval of 20 s, was conducted on the as-deposited substrate, which was heated at 130 °C. All perovskite films were ultimately annealed at 150 °C for 30 min under ambient conditions (RT ~ 25 °C with RH = 50–60%).

The Cs₃Bi₂I₉ perovskite films prepared with the first-step spray coating at temperatures of 80, 110, and 130 °C and the second-step spray coating at a constant temperature of 130 °C are denoted as T₁/T₂ = 80/130, 110/130, and 130/130, respectively, for convenience. Figure S2 presents the top-view and cross-sectional morphology of the Cs₃Bi₂I₉ perovskite films with T₁/T₂ = 80/130, 110/130, and 130/130. To quantitatively compare the quality of the perovskite films, the pinhole ratios for the films prepared with T₁/T₂ = 80/130, 110/130, and 130/130 were estimated to be 3.23%, 0.42%, and 7.15%, respectively, based on the top-view SEM images (Fig. S2a, S2c, and S2e). Additionally, the pinhole ratios from the cross-sectional SEM images of perovskite films prepared with T₁/T₂ = 80/130, 110/130, and 130/130 (Fig. S2b, S2d, and S2f) were estimated to be 4.37%, 2.01%, and 4.44%, respectively.

The spray-coated Cs₃Bi₂I₉ perovskite film with T₁/T₂ = 80/130 displays a smooth surface morphology, as shown in Fig. S2a. The corresponding cross-sectional SEM image of the Cs₃Bi₂I₉ perovskite film shows a bulk-like perovskite domain at the bottom of the film, which is in the region underneath the yellow dashed line in Fig. S2b, while chunky perovskite domains with a lateral size of several micrometers are arranged on the upper layer (region above the yellow dashed line). The clear boundary highlighted by the yellow dashed line is presumably due to the large difference in processing temperature between the two spray-coating processes (first step: 80 °C; second step: 130 °C). This significant temperature difference could lead to different perovskite domain sizes and affect the cross-sectional morphology.

Notably, some large pinholes are observed in the upper layer of the perovskite film. After the first-step spray coating, the substrate was heated to 130 °C, and the second-step spray coating was immediately carried out. Heating to 130 °C promoted the evaporation of residual solvent from the underlying perovskite film that was predeposited on the substrate at 80 °C. During the second-step spray coating, the vaporized solvent from the first-step spray-coated perovskite could enter the second-step spray-coated perovskite film, leading to the formation of pinholes in the upper perovskite film.

When the first-step processing temperature was increased to 110 °C, the Cs₃Bi₂I₉ perovskite film with T₁/T₂ = 110/130 (consistent with the processing temperature condition in Method-2) also displays a smooth film surface with fewer pinholes, as depicted in Fig. S2c. The Cs₃Bi₂I₉ perovskite film with T₁/T₂ = 110/130 demonstrates a cross-sectional morphology similar to that prepared by Method-2, especially at the bottom of the perovskite film prepared by Method-2 (refer to Fig. S2d and Fig. 1d). Figure S2d shows that the columnar perovskite domains are well packed in the lateral direction, and fewer pinholes with smaller sizes are observed compared to the cross-sectional SEM image of the perovskite film with T₁/T₂ = 80/130. No clear boundary is visible, presumably due to a reduced temperature difference between the two spray coating processes (first step: 110 °C; second step: 130 °C). When the first-step processing temperature increased to 130 °C, many pinholes and cracks appeared on the perovskite film surface, as shown in Fig. S2e, primarily due to the rapid evaporation of the solvent. Moreover, the cross-sectional SEM image of the perovskite film with T₁/T₂ = 130/130 shows multiple chunky perovskite domains instead of columnar domains stacked in the vertical direction, as shown in Fig. S2f.

Fig. 2: Mechanism of formation of spray-coated Cs₃Bi₂I₉ perovskite films.

Schematic illustration depicting the growth of perovskite crystals during spray coating and the formation of perovskite domains during subsequent spray coating on substrates heated at different temperatures of a 80 °C (F_in > F_out), b 110 °C (F_in ≈ F_out), and c 130 °C (F_in < F_out).

To determine the influence of the first-step processing temperature on the preferred orientation in the perovskite films, we conducted a series of experiments. Initially, we prepared perovskite films by spray coating at 80, 110, and 130 °C for 20 cycles, followed by annealing at 150 °C. The corresponding XRD profiles are presented in Fig. S3a. At a lower processing temperature of 80 °C, the perovskite film exhibits a predominant peak corresponding to the (006) facet. Increasing the processing temperature to 110 °C results in a more pronounced XRD peak intensity for the (220) facet than for the (006) facet. Further elevating the temperature to 130 °C leads to a higher XRD peak intensity for the (006) facet than for the (220) facet. These results indicate that the preferred orientation of (220) is formed for the perovskite film processed at 110 °C in the first step.

Second, Fig. S3b compares the XRD profiles of perovskite films with T₁/T₂ = 80/130, 110/130, and 130/130. The perovskite film with T₁/T₂ = 80/130 presents a dominant XRD peak for the (006) facet, probably due to a low perovskite growth rate. The low growth rate allows a good arrangement of perovskite crystals along the c-axis. When the first-step processing temperature increases to 110 and 130 °C, the spray-coated perovskite films exhibit a second dominant XRD peak ascribed to the (220) facet, which is beneficial for charge transport. Although the perovskite film with T₁/T₂ = 80/130 presents a smooth perovskite film with the largest perovskite domains, the preferred orientation along the (006) facet hampers the charge transport and the resultant device photoresponse. The hexagonal structure of Cs₃Bi₂I₉ has distinct dimer [Bi₂I₉]^3- anions, which are composed of two face-sharing [BiI₆]^3- octahedra placed in a layered arrangement along the ab plane. The layered structure of Cs₃Bi₂I₉ perovskites leads to strong anisotropy in their electronic and optical properties^41,42.

Sun et al. synthesized a Cs₃Bi₂I₉ single crystal and explored its anisotropic properties along the (100), (101), and (001) facets. The Cs₃Bi₂I₉ perovskite along the (100) facet has the greatest mobility-lifetime product, mainly due to having the smallest effective mass. On the other hand, the layered arrangement of the [Bi₂I₉]^3- dimer hinders charge carrier transport along the c-axis⁴³. These results indicate that carrier transport along the ab-plane of the Cs₃Bi₂I₉ perovskite is more effective than that along the c-plane of the Cs₃Bi₂I₉ perovskite.

Figure 2 schematically illustrates the growth mechanism of spray-coated perovskite films. When the precursor solution is sprayed onto a preheated substrate, an inward flux (F_in) of the sprayed perovskite solution and an outward flux (F_out) due to solvent evaporation are created³⁸. Controlling the F_in-to-F_out ratio by adjusting the processing temperature effectively regulates the perovskite film morphology and domain size. When the precursor solution is spray-coated onto the substrate heated at a temperature of 80 °C, this processing temperature, which is far below the boiling temperature of the DMF/DMSO mixed solvent, gives rise to F_in being much larger than F_out (F_in > F_out), as presented in Fig. 2a. During the spray coating process, the previously formed polycrystalline perovskite film on the substrate melts and redissolves. The slow evaporation rate of the solvent facilitates the recrystallization and grain-merging process, allowing the perovskite grains to laterally coalesce and form a bulk-like perovskite domain, as observed in Fig. S2b (the region below the yellow dashed line).

When the perovskite film is deposited on the substrate heated at a modest temperature of 110 °C, a balanced flux between F_in and F_out (F_in ≈ F_out) promotes the formation of columnar perovskite domains (see Fig. 2b). A higher temperature of 110 °C results in an increase in F_out, which enhances the crystallization rate of the newly deposited perovskite solution but reduces the grain-merging/recrystallization rate within the as-deposited perovskite film. As evaporation occurs in the exposed area, the solvent evaporation rate in the vertical direction is much faster than that in the lateral direction. Consequently, the grain-merging/recrystallization process in the vertical direction is much faster than that in the latter direction, thereby producing columnar perovskite domains.

For the perovskite spray-coated on the substrate at a high temperature of 130 °C, the elevated processing temperature, which is close to the boiling temperature of the DMF/DMSO mixed solvent, leads to F_out being much larger than F_in (F_out > F_in) (refer to Fig. 2c). A higher processing temperature leads to a significantly higher solvent evaporation rate, inducing a much faster crystallization rate of the newly deposited perovskite solution but a much lower grain-merging/recrystallization rate within the perovskite film. The grain-merging/recrystallization process could terminate before the crystallization of newly deposited perovskite crystals. Therefore, the resultant perovskite film consists of small perovskite domains.

Figure 3a displays the XRD profiles of the Cs₃Bi₂I₉ perovskite films prepared using Method-1 and Method-2. The results clearly indicate that multiple characteristic XRD peaks in both films correspond to the crystalline facets of Cs₃Bi₂I₉ perovskite (ICSD 410726), and the Cs₃Bi₂I₉ perovskite films prepared using Method-1 and Method-2 have similar XRD profiles. The processing temperature during spray coating primarily influences the compactness of the perovskite film rather than the regulation of the crystalline orientation of the Cs₃Bi₂I₉ perovskite. Due to the compact morphology of the perovskite film prepared using Method-2, the X-ray detector exhibited a superior photoresponse compared to that of the perovskite film prepared using Method-1 under the same X-ray dose rate, as depicted in Fig. S4. Figure S5a compares the XRD profiles of perovskite films prepared using Method-1 and Method-2 with the XRD peaks of BiOI (ICSD 118143) and Bi₂O₃ (ICSD 28443). The results indicate that the XRD peaks of the spray-coated perovskite films correspond well to the facets of Cs₃Bi₂I₉ (ICSD 410726), with no evidence of an oxide-based phase.

Fig. 3: Material properties of spray-coated Cs₃Bi₂I₉ perovskite films.

a XRD profiles of Cs₃Bi₂I₉ perovskite prepared by Method-1 and Method-2. b UV‒Vis absorption spectrum of the Cs₃Bi₂I₉ perovskite prepared by Method-2. c Absorption coefficients of Bi-based perovskites, Si, a-Se, and CdTe as a function of photon energy. d Attenuation efficiency of Bi-based perovskites, Si, a-Se, and CdTe as a function of film thickness under X-ray irradiation with a photon energy of 50 keV.

We further conducted X-ray photoelectron spectroscopy (XPS) to determine whether an oxide-based composition was formed on the perovskite film surface since the perovskite film was spray-coated under ambient conditions⁴⁴. XPS is a surface-sensitive method for examining the surface of perovskite films with thicknesses less than 10 nm. In Fig. S5b, the survey scans of the XPS spectra for perovskite films prepared by Method-1 and Method-2 display primary peaks corresponding to Cs (3d), I (3d), Bi (4d), and Bi (4f). Narrow scans of XPS spectra for the 3d-Cs, 4f-Bi, 3d-I, and 1s-O elements are presented in Fig. S5c–f, respectively. For the 3d-Cs narrow XPS spectra (Fig. S5c), one peak at ~738.50 eV is attributed to 3d_3/2-Cs, and the other at ~724.50 eV is attributed to 3d_5/2-Cs, which is consistent with standard Cs element characteristics. These two peaks exhibit an energy separation of ~14.0 eV. The Bi core-level peaks at binding energies of ~164.30 and ~158.90 eV are attributed to Bi 4f_5/2 and Bi 4f_7/2, respectively. The measured energy separation between these two peaks was ~5.4 eV, which is characteristic of Bi³⁺ species (Fig. S5d). Additionally, two satellite peaks at ~162.2 and 156.9 eV suggest the presence of metallic Bi (marked by * in Fig. S5d) on the film surface, which can potentially serve as defects to trap carriers^45,46. The peaks at ~630.50 and ~618.9 eV can be assigned to 3d_3/2-I and 3d_5/2-I, respectively, indicating the existence of monovalent iodine atoms with an energy separation of ~11.4 eV (Fig. S5e). Figure S5f shows the O-1s signals at binding energies of 531.9 and 529.3 eV, which are assigned to I-O and Bi-O bonding, respectively, suggesting the existence of BiOI on the perovskite film surface.

Figure 3b shows the UV‒Vis absorption spectrum of the Cs₃Bi₂I₉ perovskite prepared using Method-2. Based on the Tauc plot resulting from the UV‒Vis absorption spectrum, we can estimate the optical bandgap of the Cs₃Bi₂I₉ perovskite film, which is 1.91 eV. This result is in agreement with the reported value³⁰ and aids in establishing the band diagram of the Cs₃Bi₂I₉ perovskite. In Fig. 3c, we compare the absorption coefficients of various Bi-based perovskites, including MA₃Bi₂I₉, (NH₄)₃Bi₂I₉, Cs₂AgBiBr₆, and Cs₃Bi₂I₉, with those of commonly used semiconductors for X-ray detectors, such as Si, a-Se, and cadmium telluride (CdTe), under X-ray irradiation with photon energy ranging from 0.01 to 1 MeV (https://physics.nist.gov/PhysRefData/Xcom/html/xcom1.html). Due to the composition of large Z atoms (such as Bi and I), these Bi-based perovskites present a comparable absorption coefficient but a higher absorption coefficient than that of Si and a-Se under X-ray irradiation.

Figure 3d displays the attenuation efficiency of Bi-based perovskites, Si, a-Se, and CdTe under X-ray irradiation at 50 keV, plotted against the film thickness. Notably, at equivalent thicknesses, Bi-based perovskites exhibit significantly greater attenuation efficiency than a-Se and Si semiconductors. It is worth mentioning that the X-ray attenuation efficiency of the Cs₃Bi₂I₉ perovskite is slightly superior to that of CdTe. Figure S6 further compares the absorption coefficient and attenuation efficiency of various perovskites, including ABX₃, A₂B’BX₆, A₂BX₄, and A₃B₂X₉. Among these perovskite materials, Cs₃Bi₂I₉ perovskite has the largest average atomic number of 57.71. Consequently, the Cs₃Bi₂I₉ perovskite absorbs 99% of the 50-keV X-ray irradiation within a 1 mm thickness. These results strongly suggest that the Cs₃Bi₂I₉ perovskite is a potential candidate for high-performance X-ray detectors.

Figure 4a shows a photograph of the spray-coated Cs₃Bi₂I₉ perovskite film with a size of 2.5 × 2.5 cm², and Fig. 4b shows a photograph of the Cs₃Bi₂I₉-based X-ray detector, whose active area is 1.53 cm², as indicated by the red dashed line, with a device structure of FTO/TiO₂/SnO₂/Cs₃Bi₂I₉ perovskite/Au, as illustrated in Fig. 4c. Figure 4d shows the corresponding energy diagram of each individual constituent. We conducted ambient pressure photoemission spectroscopy to determine the valence band maximum (VBM) of the Cs₃Bi₂I₉ perovskite film fabricated by Method-2, as revealed in Fig. S7. The measured VBM of the Cs₃Bi₂I₉ perovskite film is −5.63 eV⁴⁷, while the conduction band minimum (CBM) is −3.72 eV, considering the bandgap of 1.91 eV resulting from the Tauc plot. The CBM of the Cs₃Bi₂I₉ perovskite film is slightly greater than that of SnO₂ and TiO₂. When the Cs₃Bi₂I₉ perovskite film is exposed to X-ray irradiation, the X-rays are attenuated to generate electron–hole pairs. An energy cascade between the Cs₃Bi₂I₉ perovskite film and the SnO₂/TiO₂ bilayer facilitates electron collection by SnO₂/TiO₂. Moreover, the deep VBM of SnO₂ and TiO₂ effectively block hole transport to the SnO₂/TiO₂ bilayer, reducing the noise current of the devices.

Fig. 4: Photograph of spray-coated Cs₃Bi₂I₉ perovskite film and its application for X-ray detector.

a Photograph of the spray-coated Cs₃Bi₂I₉ perovskite film. b Photograph, c schematic illustration, and d band diagram of the X-ray detector using the Cs₃Bi₂I₉ perovskite film prepared by Method-2.

Figure 5a shows the dark current density of X-ray detectors using Cs₃Bi₂I₉ perovskite films prepared by Method-1 and Method-2. The results indicate that the device fabricated via Method-2 exhibits a much lower dark current density than that fabricated via Method-1. In Method-1, a high processing temperature with a short time interval during spray coating leads to fast growth of the perovskite domains. Owing to the fast growth rate, many defects could be formed within the perovskite domains. These defect states with energy levels in the perovskite bandgap may provide thermal stepping stones for the trapped carriers to detrap, leading to an increase in the dark current⁴⁸. However, in Method-2, a low processing temperature and a prolonged time interval in the first-step spray-coating process can slow the perovskite growth rate due to balanced F_in and F_out and allow the pregrowth of perovskite crystals to undergo a recrystallization/grain-merging process into a large perovskite domain. As a result, the crystal defects can be markedly reduced, and the leakage of dark current caused by these defects is further suppressed⁴⁹.

Fig. 5: Performance of Cs₃Bi₂I₉-based X-ray detector.

a Dark current density of devices prepared by Method-1 and Method-2. b J–V characteristic curves of the Cs₃Bi₂I₉-based X-ray detector prepared using Method-2 under X-ray irradiation at different dose rates. The dark current of the Cs₃Bi₂I₉-based X-ray detector is shown for comparison. c Sensitivity of the Cs₃Bi₂I₉-based X-ray detector as a function of the external electric field with different thicknesses of the active layer. d Sensitivity of the Cs₃Bi₂I₉-based X-ray detector as a function of dose rate for different thicknesses of the active layer. e I–t response of the Cs₃Bi₂I₉-based X-ray detector under X-ray irradiation with different thicknesses of the photoactive layer and different dose rates. f I–t curve of the Cs₃Bi₂I₉-based X-ray detector in response to multiple X-ray pulse irradiation. The thickness of the perovskite film is 50 μm, the applied electric field is 0.12 V μm⁻¹, the dose rate is 799 μGy_air s⁻¹, and the time interval of X-ray on/off irradiation is 10 s.

Figure 5b shows the photocurrent density‒voltage (J‒V) characteristic curves of devices prepared via Method-2 under X-ray irradiation at different dose rates. When the X-ray dose rate increases, the photocurrent generated by the device also increases under the same external electric field. An X-ray detector exposed to a high X-ray dose rate generates more charge carriers that are effectively extracted by the applied electric field. The lowest current density in the J–V curves does not occur at zero-bias voltage. This phenomenon is presumably attributed to the asymmetric device structures. Our structures solely include n-type semiconductors such as TiO₂ or SnO₂ but lack p-type semiconductors such as 2,2′,7,7′-tetrakis[N,N-di(4-methoxyphenyl)amino]-9,9′-spirobifluorene (Spiro-OMeTAD) or the polymer poly[bis(4-phenyl)(2,4,6-trimethylphenyl)amine] (PTAA). This asymmetry in the device structure causes unbalanced electron/hole extraction rates from the perovskite film to the electrode during the voltage sweep. Moreover, the presence of metallic Bi as a surface defect at the perovskite/Au interface can impede hole transport toward the top Au electrode. These effects result in the lowest current density in the J–V curves not aligning with the zero-bias voltage, as shown in Fig. S8¹⁶.

Figure 5c shows the sensitivity of devices with different thicknesses of the photoactive layer as a function of the applied electric field under the same X-ray dose rate. As shown in Fig. 1d, the perovskite film prepared via Method-2 with 200 cycles of spray coating has a thickness of approximately 50 μm. To enhance the sensitivity of the X-ray detector, a two-step process involving 300 cycles of spray coating (first step: 20 cycles; second step: 280 cycles) was conducted, resulting in an increase in the thickness of the perovskite film to 80 μm. The sensitivity (S) of the X-ray detector is given by the following formula:

$$S=\frac{{I}_{p}-{I}_{d}}{D\times A}$$

(1)

where D is the exposure dose rate of X-ray irradiation, A is the active area of the device, I_p is the photogenerated current of the device under X-ray irradiation, and I_d is the dark current. Increasing the applied electric field effectively drives the electron and hole carriers to be collected by their corresponding electrodes. As a result, both the photocurrent and the sensitivity are simultaneously enhanced. A thicker active layer can prolong the interaction path between the perovskite and the X-ray photons; therefore, increasing the thickness of the perovskite film can significantly improve the photocurrent of the device, as presented in Fig. S9.

Figure 5d compares the sensitivity of X-ray detectors with different thicknesses of the active layer under varying X-ray dose rates. The sensitivity of the X-ray detector decreases as the X-ray dose rate increases. At high X-ray dose rates, the photoactive layer exposed to the elevated X-ray photon flux generates more electron–hole pairs and increases scattering events, including Rayleigh and Compton scattering. Subsequent collisions between charge carriers and between charge carriers and scattered photons hinder carrier transport and reduce the carrier lifetime, thereby limiting the charge collection efficiency (CCE) of the X-ray detector. The CCE is defined as follows¹¹:

$$CCE=\frac{\mu \tau V}{{L}^{2}}[1-{e}^{(-\frac{{L}^{2}}{\mu \tau V})}]$$

(2)

where μ is the carrier mobility, τ is the carrier lifetime, V is the applied bias voltage, and L is the thickness of the active layer. A decrease in the $\frac{\mu \tau V}{{L}^{2}}$ ratio decreases the CCE. As the number of collision events increases with increasing incident photon flux, the reduced carrier lifetime leads to a decrease in the CCE, subsequently reducing the photocurrent and sensitivity of the X-ray detector. Additionally, collision events increase with the thickness of the perovskite film in the X-ray detector, further diminishing the CCE, as observed in Eq. (2). Consequently, the sensitivity of the X-ray detector with an 80 μm perovskite film significantly decreases with increasing X-ray dose rate compared to that with a 50 μm perovskite film. For devices with two thicknesses of perovskite film (50 and 80 μm) exposed to the same X-ray dose rate, the sensitivity of the devices increases with the thickness of the active layer due to the increased X-ray attenuation ratio. The corresponding sensitivities with different electric fields and X-ray dose rates are shown in Fig. S10.

Figure 5e shows the photocurrent-time (I–t) curves of devices with different thicknesses of the active layer exposed to different X-ray dose rates. An increasing X-ray dose rate promotes the generation of electron–hole pairs, thereby enhancing the photocurrent of the devices. Similarly, as the thickness of the photoactive perovskite film increases, the photocurrent of the X-ray detector also increases due to an enhanced interaction path between the perovskite and the X-ray photon, resulting in increased attenuation efficiency^50,51. Figure 5f shows the stability of the response of the device by measuring the I–t response under consecutive X-ray pulse irradiation. The thickness of the perovskite film, applied electric field on the device, X-ray dose rate, and pulse duration of X-ray irradiation were 50 μm, 0.1 V μm⁻¹, 799 μGy_air s⁻¹, and 10 s, respectively. The consistent photocurrent indicates that the as-fabricated device remains stable after exposure to consecutive X-ray irradiation pulses.

Figure 6a shows the detection limit of the X-ray detector using the Cs₃Bi₂I₉ perovskite spray-coated by Method-2 with a thickness of 80 μm. For the device under a constant external electric field, the photocurrent density–dose rate curve is linearly fitted (as indicated by the solid red line), and the corresponding slope is the sensitivity. As defined by the International Union of Pure and Applied Chemistry (IUPAC), the detection limit represents the lowest dose rate at which the X-ray detector still achieves a signal-to-noise ratio (SNR) of 3. Pan et al. further applied a statistical model to establish methods for determining detection limits that extend the well-known Currie formulas. The reliability factor for the SNR, set as 3.29, is used to determine the detectable signal of the photocurrent. The device delivered this detectable signal when exposed to a specific dose rate, which corresponds to the detection limit. By considering the standard deviation of the dark current as noise, the detection limit can be estimated by Eq. (3)⁵²:

$${\rm{Detection}}\,{\rm{limit}}=\frac{3.29\times {\rm{Standard}}\,{\rm{deviation}}\,{\rm{of}}\,{\rm{dark}}\,{\rm{current}}}{{\rm{Area}}\times {\rm{Sensitivity}}}$$

(3)

The resultant detection limit of the Cs₃Bi₂I₉-based X-ray detector was estimated to be 7.4 μGy_air s⁻¹.

Fig. 6: Comparison of sensitivity and detection limit for A₃Bi₂I₉-based X-ray detectors.

a Photocurrent density plotted against dose rate for the Cs₃Bi₂I₉-based X-ray detector. b Sensitivity as a function of the detection limit for the A₃Bi₂I₉-based X-ray detectors (A-site cation: MA⁺ or Cs⁺). The performance of X-ray detectors using a-Se, HgI₂, and CZT is shown for comparison.

Figure 6b compares the performance of X-ray detectors using A₃Bi₂I₉-based perovskites, including MA₃Bi₂I₉ and Cs₃Bi₂I₉ perovskites, in terms of sensitivity and detection limit. The performances of commercially used a-Se, HgI₂, and cadmium zinc telluride (CZT) X-ray detectors are shown for comparison. The corresponding sensitivity and detection limit of X-ray detectors using Bi-based perovskites, a-Se, HgI₂, and CZT are summarized in Table 1. The device performances of X-ray detectors using different perovskite materials under different working conditions, including different X-ray photon energies, electric fields, and dose rates, are further summarized in Fig. S11 and Table S1. In contrast to the X-ray detector using CZT, our device offers a lower sensitivity and a much lower detection limit. Although the CZT-based X-ray detector has a higher sensitivity than our detector, the high processing temperatures for CZT fabrication can increase the energy budget and fabrication cost. In this work, a Cs₃Bi₂I₉ perovskite was fabricated through a facile process of spray coating at a low temperature, which is critical for the fabrication of low-cost X-ray detectors.

Table 1 Summary of the sensitivity, detection limit, and active area of X-ray detectors using Cs₃Bi₂I₉, MA₃Bi₂I₉, a-Se, HgI₂, and CZT.

On the other hand, although our device has a lower sensitivity than the device using HgI₂, the toxic nature of HgI₂ can cause serious harm to human health. The nontoxic nature of Bi-based perovskites makes them potential candidates for commercial X-ray detectors because of their environmentally friendly composition. The commercial a-Se-based direct X-ray detector exhibited a low detection sensitivity of 20 μC Gy_air⁻¹ cm⁻² and a high detection limit of 5.5 μGy_air s⁻¹ under a remarkably high electric field of 80 V μm⁻¹ due to the low mobility-lifetime product of a-Se. In contrast, our device delivers a higher sensitivity of 127.23 μC Gy_air⁻¹ cm⁻² and a comparable detection limit of 7.4 μGy_air s⁻¹ compared to the a-Se-based X-ray detector. Compared with devices using MA₃Bi₂I₉ prepared by blade coating and spray coating, our device achieves superior sensitivity but a higher detection limit. Notably, the active area of the Bi-based X-ray detector in our work is the largest among those in the published literature.

Among the X-ray detectors utilizing Bi-based perovskites, devices with single-crystal perovskites exhibit the best performance with a low detection limit and high sensitivity due to their low defect density and large mobility-lifetime product. However, their limited size and time-consuming preparation render them unsuitable for practical applications in X-ray imaging. For example, the dimensions of the X-ray detectors used for medical diagnostics, such as chest radiography, are approximately 40 × 40 cm². The size of the spray coating can be precisely controlled by moving the pulverizer in a zigzag trajectory. Optimizing the translation step improved the uniformity of large-area perovskite films produced via spray coating. Moreover, the perovskite film thickness can be regulated by adjusting the deposition cycle. As a result, a sensitive, cost-effective, and low-temperature processed X-ray detector can be achieved with spray-coated perovskite films covering an area of approximately 40 × 40 cm². The demonstration of spray-coated Cs₃Bi₂I₉ perovskite for X-ray detectors paves the way toward scalable, easily produced, low-cost, stable, and eco-friendly X-ray detectors.

Conclusion

We have successfully employed a spray-coating process to produce a large area and lead-free thick Cs₃Bi₂I₉ perovskite film for X-ray detector applications. The processing temperature and time intervals during spray coating play pivotal roles in determining the morphology of the Cs₃Bi₂I₉ perovskite film. A total of 200 spray-coating cycles at a high processing temperature of 130 °C with time intervals of 20 s led to fast perovskite crystallization and disordered stacking of the perovskite domains. A modified spray coating with a two-step process was conducted to improve the morphology of the Cs₃Bi₂I₉ perovskite film. An initial 20-fold spray coating at a low processing temperature of 110 °C, with an extended time interval of 40 s, can decrease the perovskite growth rate, facilitate the recrystallization/grain merging of perovskite domains, and effectively arrange the perovskite domains. Because of the orderly stacking of perovskite domains at the substrate/Cs₃Bi₂I₉ perovskite film interface, the subsequent 180 cycles of spray coating produced densely packed perovskite grains and a compact perovskite film. This modified method for preparing Cs₃Bi₂I₉ perovskite film facilitates charge transport and decreases the leakage current of the X-ray detector. The as-fabricated device achieved a large active area of 150 mm², a sensitivity of 127.23 μC Gy_air⁻¹ cm⁻², a detection limit of 7.4 μGy_air s⁻¹, and durability after long-term X-ray pulse irradiation.