Perovskite CsPbBr3 Single‐Crystal Detector Operating at 1010 Photons s−1 mm−2 for Ultra‐High Flux X‐ray Detection

Semiconductors for detecting hard radiation are confronted with considerable problems when operating at high photon fluxes. A perovskite CsPbBr3 single‐crystal detector capable of operating at X‐ray fluxes of up to 1010 photons s−1 mm−2 with beam area ≤ 0.25 mm2 at 58.61 keV for current‐mode X‐ray detection is reported. The spectrometer‐grade melt‐grown CsPbBr3 detectors show a gamma‐ray energy resolution of ~7.5% at 122 keV for 57Co and dark current as low as 4.3 nA (0.5 nA mm−2) at a reverse bias voltage of 200 V (118 V mm−1). The detector is tested at X‐ray energies of 8.2, 10, and 58.61 keV at a synchrotron light source under a reverse bias voltage of up to 1000 V (588 V mm−1). Under a sufficiently high bias voltage and within several hundreds of seconds X‐ray exposure, good photocurrent linearity (goodness of fit R2 > 0.99) and reproducibility are obtained up to a flux of ≈1010 photons s−1 mm−2 at beam area 0.25 mm2 with Lower Limit of Detection of ≈105 photons s−1 mm−2 and Charge Collection Efficiency of ≈100% for 58.61 keV X‐rays. Accordingly, wide application of CsPbBr3 detectors in high‐flux X‐ray detection is anticipated.


Introduction
Detectors for hard X-ray detection and imaging with high and ultra-high photon fluxes (φ p ) are critical in various areas. Medical X-ray imaging requires φ p > 10 7 photons s −1 mm −2 (abbreviated as p s −1 mm −2 ), [1] while modern Computed Tomography system requires φ p ≈10 9 p s −1 mm −2 . Synchrotron X-rays can produce ultra-high flux of φ p > 10 10 p s −1 mm −2 , which can enable scientific advancements in numerical fields (e.g., material also been considered for X-ray detection but with limited success. Perovskites have recently emerged as promising nextgeneration semiconductor radiation detectors. [1,18] Perovskite thin films can be solution-processed for flat-panel X-ray imagers, [19] and bulk single crystals (e.g., CsPbBr 3 ) can be grown from melt or solution for gamma-ray spectroscopy [20] and X-ray photon counting. [21] Despite various widely reported types of perovskite X-ray detectors, [22] there have been no reports on performance under ultra-high X-ray fluxes (e.g., φ p > 10 8 p s −1 mm −2 ). Such harsh high flux working conditions impose stringent requirements on perovskite crystal quality, materials stability, and device fabrication. These requirements include large perovskite single crystals with low defects, carrier traps, and minimal ion migration. In contrast, many perovskites exhibit prominent ion migration problems. [23] Here, we report the first study of its kind, specifically the performance of an all-inorganic perovskite CsPbBr 3 singlecrystal detector under synchrotron X-ray radiation with ultrahigh fluxes from 5.8 × 10 6 to 1.9 × 10 12 p s −1 mm −2 at beam area ≤ 0.25 mm 2 at X-ray energies of 8.2, 10, and 58.61 keV. CsPbBr 3 single-crystal ingots were grown by the Bridgman growth method, and high-quality spectrometer-grade crystals were selected and fabricated into detectors. The detectors had low and stable dark current, e.g., 4.3 nA (0.5 nA mm −2 ) at a reverse voltage value (Vr) of 200 V (118 V mm −1 ), which enables a superior ability to detect weak X-rays (i.e., a small Lower Limit of Detection) and stable detector operation up to at least Vr = 1000 V (588 V mm −1 ). The excellent crystal quality and high reverse bias voltage allowed a high charge collection efficiency (CCE) of ≈100%, leading to a measured X-ray detection Sensitivity (S) that approximates the theoretical value. Under a sufficiently high bias voltage and within several hundreds of seconds, a linear and reproducible photocurrent response was obtained within a broad dynamic range from ≈10 6 to ≈10 10 p s −1 mm −2 at beam area 0.25 mm 2 for 58.61 keV X-ray. Besides, we further tested the performance of CsPbBr 3 detectors under ultra-high fluxes up to the limit of the synchrotron facility, which could be taken as a proxy to accelerated polarization testing that provides valuable clues to the detector's longterm stability under lower flux applications. Our investigation demonstrates the promising performance of CsPbBr 3 detector for high and ultra-high flux X-ray detection, which could lead to essential scientific advancements in medical and biological imaging and various synchrotron-based X-ray techniques.

Spectrometer-Grade CsPbBr 3 Single-Crystal Detectors
Given the aforementioned polarization mechanism, [6a,10] perovskite crystals with lower defects are expected to suffer less polarization and hence exhibit better device performance. We use photoluminescence (PL) spectroscopy to screen the CsPbBr 3 crystal quality. The as-grown CsPbBr 3 ingot was cleaved at different positions, and PL of the freshly cleaved surface was measured (Figure 1a). Time-resolved photoluminescence (TRPL) decay lifetime (τ ave ) has been associated with defect density in perovskite materials and are widely used as a proxy for crystal quality. [24] Theoretically, higher defect density means more free carriers would recombine through defectassisted non-radiative (Shockley-Read-Hall) recombination, which results in weaker radiative recombination detected by TRPL and hence a shorter τ ave . In fact, it has been frequently observed that lower defect density corresponds to a longer τ ave in perovskite materials. [25] The CsPbBr 3 ingot has a PL lifetime τ ave decreasing from the first-to-crystallize bottom tip to the last-to-crystallize top (Figure 1b), which indicates a lower number of defects in the bottom. The CsPbBr 3 ingot quality gradient derives from varying impurity concentrations at different locations, which is a feature of the Bridgman growth method. Interestingly, we found that the steady-state PL emission peak wavelength λ peak of the CsPbBr 3 crystal is correlated to the TRPL decay lifetime. Statistically, longer τ ave usually corresponds to longer λ peak (Figure 1c). While the detailed mechanism behind the relationship between λ peak and defects in CsPbBr 3 needs further investigation, we can grow and faithfully select high quality CsPbBr 3 crystals, represented by long τ ave and λ peak , for subsequent device fabrication and testing.
A high detector bias voltage is critical to mitigating radiationinduced polarization. [6a,10] However, a high bias voltage could lead to large and unstable dark currents, which degrades the detector's ability to detect weak X-ray signals. [26] To maintain low dark current at a high bias voltage, we fabricated Schottky diode CsPbBr 3 detectors with gold and gallium indium eutectic alloy (denoted as GaIn) planar electrodes, i.e., Au/CsPbBr 3 / GaIn. A Schottky/Ohmic contact is formed between the p-type CsPbBr 3 and the low/high work function metal GaIn/Au. [27] The direction of the electric field at reverse bias is from GaIn to CsPbBr 3 . For gamma-ray and X-ray detection, the GaIn electrode side is irradiated. Given the limited penetration depth of X-rays ( Figure S1, Supporting Information), holes contribute more to the induced signal. Because the hole mobility lifetime product, µ h τ h , exceeds the µ e τ e of electrons in CsPbBr 3 (i.e., µ h τ h ≈ 1.3 × 10 −3 cm 2 V −1 vs µ e τ e ≈ 8.8 × 10 −4 cm 2 V −1 , as previously reported [20b,28] ), having holes contribute more to the signal can result in higher CCE ( Figure S1, Supporting Information shows the setup). We fabricated and tested 6 devices made of CsPbBr 3 crystals with relatively long τ ave and λ peak from the high-quality parts of different ingots with same growth parameters described in our previous work. [20a,28b] (Figure 1a shows the τ ave of devices I and II, Figure S2 Supporting Information shows the TRPL and PL emission spectra of the rest crystals, Figure S3 Supporting Information shows device pictures, Table 1 shows crystal thickness).
We measured the CsPbBr 3 detector I-V curve under different conditions: 1) in the dark, 2) under ambient room light, 3) under X-ray irradiation, 4) dark after detector irradiation. These measurements were performed sequentially in a single session. The voltage was swept from + 200 V to -200 V and it took ≈3 mins for each measurement (voltage step 5 V and ≈2 s per current point). The dark I-V curve of the as-fabricated device and I-V under light illumination ("dark-1" and "light" in Figure 1d) showed current-rectifying behavior due to the Schottky diode structure ( Figure S4 Supporting Information depicts the I-V curves of other devices with similar behavior). Yet, we noticed some interesting results. As forward bias voltage ramps down from +200 V to 0 V, the forward current www.advopticalmat.de first increases and then decreases. Besides, the I-V curve measured under X-ray irradiation and measured after X-ray irradiation in dark ("X-ray" and "dark-2" in Figure 1d) became symmetric. The abnormal I-V behavior may be due to either ion migration or charge trapping. To confirm which factor dominates, we compared the dark I-V curves after different biasing conditions. As shown in Figure 1e, the initial I-V of a freshly fabricated device showed a typical asymmetric diode behavior (voltage sweep from 0 V to 300 V, then from 0 V to −300 V). During reverse 1000 V biasing for 1 min, the dark current kept decreasing and approached a steady value (Figure 1f). After the reverse biasing, the I-V curve became symmetric with small leakage current at both reverse and forward bias directions ( Figure 1e). The CsPbBr 3 detector was then forwardbiased at 1000 V, and the initially small dark current gradually  τ ave is average PL decay lifetime at the peak emission wavelength λ peak . b) Typical relationship between τ ave and crystal location of an CsPbBr 3 ingot. c) Statistical data of τ ave versus λ peak obtained from many crystals of different ingots. d) I-V curves of device I. I-V curve testing sequence was "dark-1" (measured in dark before X-ray irradiation), "light" (under ambient room light), "X-ray" (under constant X-ray irradiation of 3.2 × 10 9 p s −1 mm −2 at 8.2 keV), and "dark-2" (measured in dark right after X-ray irradiation). e) I-V curve of a CsPbBr 3 device measured after different biasing conditions. f) Current as a function of time under constant biasing of forward and reverse 1000 V. g,h) 57 Co gamma-ray energy spectrum of as-fabricated CsPbBr 3 devices (shaping time: 10 µs; spectrum acquisition time: 300 s). www.advopticalmat.de increased ( Figure 1f). Finally, the I-V curve returned to asymmetric diode behavior after forward biasing ( Figure 1e). The I-V curve change under different biasing conditions indicates ion migration in the CsPbBr 3 crystal, similar to the results reported elsewhere. [29] The continuous increase of current under repetitive forward biasing at a fixed X-ray flux further showed that charge trapping is not a major factor leading to the symmetric I-V curve ( Figure S4, Supporting Information). In the latter scenario, we would expect the forward current to decrease under X-ray irradiation. Ion migration occurs because of the presence of vacancy defects and can account for the fact that crystals obtained from different batches can behave differently as they may not contain identical numbers of defects or vacancies.
Regardless of the I-V curve change and ion migration, the CsPbBr 3 detectors work in reverse bias mode. By preconditioning the CsPbBr 3 detectors with reverse biasing, the dark current can reach a steady saturation value. A dark current, "dark-2" as low as 4.3 nA (0.5 nA mm −2 ) at reverse voltage Vr = 200 V (118 V mm −1 ), was achieved for the best device (device III). Different device fabrication aspects (e.g., electrode contact and wire connection) contribute to the variation in the dark current among all detectors (Table 1). High quality of both the CsPbBr 3 crystal and device fabrication is critical for excellent gamma-ray and X-ray detection performance. Five of the six devices can resolve the 57 Co 122 keV gamma-ray photopeak (Figure 1g,h) with the best energy resolution of ≈7.5% (Table 1), demonstrating that the ion migration is not fatal to the spectroscopy performance of the CsPbBr 3 detector.

Linear and Reproducible Current Response up to 10 10 p s −1 mm −2
We first tested the CsPbBr 3 detector under synchrotron X-rays with φ p = 10 8 -10 9 p s −1 mm −2 at energy (E X ) values of 8.2 and 10 keV. A high bias voltage is critical to reducing radiationinduced polarization, [6a,10] so we first tested detector responses at different bias voltages under a fixed photon flux. At φ p = 3.2 × 10 9 p s −1 mm −2 and E X = 8.2 keV, the current response of device I is shown in Figure 2a. The long current response time of several hundred seconds (i.e., the time from the start of the current rising to stabilize) at low bias voltage (e.g., < 100 V) indicates significant radiation-induced polarization. As the X-ray generated free carriers continue to fill the traps in the crystal, more carriers are being collected, which leads to the slowly rising current. Contrarily, the response time is significantly reduced to several seconds or shorter at higher bias voltage, e.g., less than 1 s at 1000 V (588 V mm −1 ) (see Figure S5, Supporting Information for the fast response at 1000 V), which means less radiation-induced polarization. For comparison, we measured a spectrometer CZT detector whose response time is smaller than 1 s at 200 V (108 V mm −1 ) ( Figure S6, Supporting Information), comparable to the CsPbBr 3 detector at high bias voltage. A short response time can enable a high frame rate, important in imaging applications. [30] A photodiode's response speed may be limited by the carrier transit time or the RC time constant of the circuit. [31] We estimate the transit-time-limited 3 dB cutoff frequency to be ≈0.2 MHz for a CsPbBr 3 detector with thickness d = 1.7 mm at a bias voltage of 300 V, much smaller than the RC-limited 3 dB cutoff frequency estimated to be ≈GHz order (see Supporting Information for estimation). Therefore, the CsPbBr 3 detector response speed is typically limited by the charge carrier transit time instead of the RC effect. To increase the response speed, the bias voltage should be increased to reduce the carrier transit time. However, high bias voltage could result in large noise. Device surface passivation or guard ring structures may help to reduce noise.
Besides, a higher bias voltage can also benefit a higher CCE. As can be seen in Figure 2b, the photocurrent density increases as a function of bias voltage and approaches saturation, which means CCE is increased at higher bias voltage ( Figure S5, Supporting Information depicts raw current response at φ p = 0.75 × 10 9 p s −1 mm −2 ). Usually, the Hecht equation can be used to fit the photocurrent versus voltage data to estimate the charge carrier µτ product. [20b,28] However, since significant polarization occurred at low bias voltage (e.g., < 100 V), the Hecht equation fitting would result in µτ value with error. Hence, we did not perform such fitting. Despite the advantage of reduced polarization, higher bias voltage could lead to a large and unstable dark current (see Figure S5 and Figure S7, Supporting Information for the unstable dark current of some devices). As a rule of thumb, the detector should be subjected to as high a bias voltage as possible without developing unstable dark current or breakdown. Following this as a selection rule, we chose a relatively high bias voltage of 300 V (158 V mm −1 ) as the working voltage of device I and performed further testing. As shown in Figure 2c, the detector has a stable dark current and fast response to different X-ray fluxes. The photocurrent density is linear at both E X = 8.2 keV and E X = 10 keV in flux range 10 8 -10 9 p s −1 mm −2 at beam area 1 mm 2 with goodness of fit R 2 > 0.999 (Figure 2d) ( Figure S5, Supporting Information shows the raw current response at E X = 10 keV). The photocurrent is also linear against X-ray beam area at a fixed flux of φ p = 2.98 × 10 9 p s −1 mm −2 (R 2 = 0.9984) (Figure 2e) ( Figure S5, Supporting Information shows the raw current response). Similarly, for device II, the photocurrent density J versus φ p at E X = 8.2 keV showed good linearity at Vr > 300 V (Figure 2f). At Vr < 300 V, the linearity is poor because of the radiation induced polarization ( Figure S8, Supporting Information shows the raw current response at different Vr and a fixed flux).
After testing under softer X-rays of 8.2 and 10 keV, we tested under 58.61 keV X-rays that have a larger penetration depth. A broader photon flux of φ p = 5.8 × 10 6 to 8.1 × 10 10 p s −1 mm −2 were used. The current response at different applied Vr values under a fixed φ p was first measured to determine working Vr values for different devices (see Figure S7, Supporting Information). The effect of bias voltage on polarization was observed with the plot of photocurrent versus bias voltage for device III (Figure 2g). At a lower flux φ p = 3.4 × 10 9 p s −1 mm −2 , the photocurrent quickly increases and approaches saturation. In comparison, at a higher flux φ p = 4.3 × 10 10 p s −1 mm −2 , the photocurrent was suppressed at a lower bias voltage region due to radiation-induced polarization. Different devices may exhibit different photocurrent saturation behavior due to different extents of polarization (see Figure S9, Supporting Information for photocurrent vs bias voltage for devices IV, V, and VI).
Based on the bias voltage selection rule mentioned above, the optimal working Vr value was selected as 1000 V www.advopticalmat.de (588 V mm −1 ) for device III and V, and 300 V (176 V mm −1 ) for devices IV and VI. For device III with Vr = 1000 V, φ p first increased from 5.8 × 10 6 to 8.1 × 10 10 p s −1 mm −2 and then decreased back to 5.8 × 10 6 p s −1 mm −2 with the same flux values (Figure 2h). Stable dark and signal currents were obtained for all fluxes over several hundreds of seconds, indicating negligible radiation-induced polarization. Without polarization, the photocurrent obtained with increasing φ p agrees with that obtained from decreasing φ p (Figure 2i), indicating a good reproducibility of the detector photocurrent at the same X-ray flux and a good linearity (R 2 = 0.99997). In addition, devices IV and V also showed good current linearity in the φ p range 5.8 × 10 6 -3.4 × 10 9 p s −1 mm −2 ( Figure S10, Supporting Information). We note that the photocurrent magnitude may vary considerably between different devices due to the different device testing and polarization history. In summary, under sufficiently high bias voltage (e.g., 1000 V), the best-tested CsPbBr 3 detector (device III) can be operated at φ p as high as 8.1 × 10 10 p s −1 mm −2 at E X = 58.61 keV and beam area 0.25 mm 2 with good current stability, reproducibility, and linearity in a timescale of several hundreds of seconds. In comparison, a CdTe detector shows significant radiation-induced polarization at an energy flux of ≈9 × 10 10 keV s −1 mm −2 , [11] and the highflux capable CZT detector can be operated at ≈10 8 p s −1 mm −2 Adv. Optical Mater. 2023, 11, 2202946 Figure 2. Current response of CsPbBr 3 devices to synchrotron X-ray. E X : X-ray energy; φ p (p s −1 mm −2 ): X-ray photon flux; R 2 : goodness of fit for the linear fitting. a) Current response of device I under different reverse bias voltage (Vr) at a fixed flux (E X = 8.2 keV, φ p = 3.2 × 10 9 p s −1 mm −2 , A beam = 1 × 1 mm 2 ). b) photocurrent density J versus Vr of device I at different flux (E X = 8.2 keV, A beam = 1 × 1 mm 2 ). c) Device I current response to different φ p in decreasing order (A beam = 1 × 1 mm 2 ). d) Device I J versus φ p at E X = 8.2 and 10 keV (A beam = 1 × 1 mm 2 ). e) Device I photocurrent versus X-ray beam area A beam . f) J versus φ p of device II at different Vr (A beam = 1 × 1 mm 2 ). g) Current response of device III to different Vr under fixed φ p values (A beam = 0.2 × 0.2 mm 2 . Photocurrent normalized to that of 1000 V to highlight the trend of current increasing). h) Current response of device III to different φ p values (A beam = 0.5 × 0.5 mm 2 ; Vr = 1000 V). i) Device III photocurrent density obtained during increasing and decreasing flux shown in h.
Sensitivity and Lower Limit of Detection (LLD) are important figures of merit of X-ray detectors. [26] Different from the polychromatic X-ray sensitivity (defined as S = J/Ḋa ir , where J is the photocurrent density, and Ḋa ir is the X-ray dose rate in the air), we calculate the sensitivity of monochromatic synchrotron X-rays by Equation (1), because the photon flux φ p and energy E X of synchrotron X-rays are readily known. Theoretically, the highest sensitivity, S 0 , when CCE = 100% can be calculated using Equation (2) (Supporting Information shows the derivation of S 0 ). Then, the LLD can be calculated according to an established dark current method [26] using Equation (3), where A eff is the effective area of current generation (i.e., the smaller area between X-ray beam and electrode), and dark I σ is the standard deviation of the dark current. The sensitivity S and LLD for devices III -V at respective working bias voltage is shown in Table 2. The measured sensitivities approach the theoretical sensitivity of S 0 = 3 × 10 −8 nC keV −1 (ε pair = 5.3 eV for CsPbBr 3 [28b] ) indicating that ≈100% of the 58.61 keV photons were absorbed by the CsPbBr 3 crystal (i.e., negligible absorption by electrode), and CCE is ≈100%. In addition, the sensitivity remains nearly constant within the linear dynamic range (The sensitivity is the slope of the linear fitting of photocurrent versus X-ray energy flux). Both sensitivity and dark current fluctuation influence the LLD, so LLD reflects the crystal and device fabrication quality. Device V has a much higher LLD due to its large dark current and fluctuation (see Figure S10, Supporting Information).

Accelerated Polarization testing of CsPbBr 3 Detector
So far, we have seen good linearity and reproducibility of the CsPbBr 3 detectors under a time scale of several hundred seconds. We further performed accelerated polarization testing of the CsPbBr 3 detector by increasing the X-ray flux or the irradiation time. Compared to the previous flux range of φ p ≈5.8 × 10 6 to 8.1 × 10 10 p s −1 mm −2 , the devices were tested at a higher flux range of φ p ≈ 3.4 × 10 9 to 1.9 × 10 12 p s −1 mm −2 which is the flux limit of the synchrotron beamline. As shown in Figure 3a for device III, while current drifting is negligible at lower fluxes (e.g., φ p < 1.5 × 10 11 p s −1 mm −2 ), it decreases significantly above 1.5 × 10 11 p s −1 mm −2 . When the dose is further increased by increasing the X-ray beam size from A beam = 0.05 × 0.05 mm 2 to A beam = 0.2 × 0.2 and 0.3 × 0.3 mm 2 (Figure 3b,c), the current drifting became more prominent, which indicates that the polarization is area dependent. Additionally, under repeated tests at the same flux of φ p = 1.9 × 10 12 p s −1 mm −2 (Figure 3b,c), the signal current could not recover its initial value (i.e., value before decreasing). The signal current follows the end value in the last test, indicating significant radiation-induced polarization. Once significant polarization occurred, even at the same lower flux (e.g., φ p < 1.5 × 10 11 p s −1 mm −2 ) where polarization is negligible, the signal current after significant polarization became smaller than its original value before polarization ( Figure 3c). This means that polarization degrades the reproducibility of the detector current. In contrast, good current reproducibility was obtained at lower flux values where no significant polarization occurred (Figure 2g). At larger beam sizes of A beam = 0.4 × 0.4 and 0.5 × 0.5 mm 2 , reducing the highest φ p from 1.9 × 10 12 to 1.5 × 10 11 and 4.3 × 10 10 p s −1 mm −2 , respectively, leads to better current reproducibility than that at φ p = 1.9 × 10 12 p s −1 mm −2 ( Figure S11, Supporting Information shows the raw current response). Despite the previous polarization at the flux limit of the synchrotron facility φ p = 1.9 × 10 12 p s −1 mm −2 , subsequent testing with φ p in a lower range of 5.8 × 10 6 -4.3 × 10 10 p s −1 mm −2 shows good current reproducibility and linearity ( Figure S11, Supporting Information). Since the significant polarization at flux above a certain level is detrimental to detector performance, it is important to understand the higher limit of flux under which the polarization is not a concern in several hundred seconds. However, as shown in Figure 3a, the degree of polarization manifested by the current drifting became progressively prominent, so it is difficult to quantify the higher limit of flux precisely.
Here, we try to find out a higher limit of photon flux by examining the current drifting (see Supporting Information for details). The current drifting is negligible under ≈20 s irradiation of flux of 8.1 × 10 10 p s −1 mm −2 , 3.4 × 10 9 p s −1 mm −2 , and 1.2 × 10 10 p s −1 mm −2 at 58.61 keV, respectively, for device III, IV, and V. These flux values may be taken as the higher limit of flux for respective devices. Despite the current decreasing at higher flux due to polarization, we examined the linearity of photocurrent density J versus φ p with J calculated from the peak photocurrent before decreasing. As shown in Figure 3d, good linearity was obtained for device III. However, practically, the detector should not be operated at φ p > 1.5 × 10 11 p s −1 mm −2 where significant current decreasing occurs due to polarization. In comparison, device IV had photocurrent density deviated from linearity at flux φ p > 1.5 × 10 11 p s −1 mm −2 due to polarization (Figure 3e). Increasing the bias voltage to 1000 V could not improve the linearity for device IV ( Figure S12, Supporting Information). The worse performance of device IV than device III may be due to worse crystal quality (e.g., shorter τ ave ). Device V and VI also showed a photocurrent density deviation from linearity ( Figure S12  A beam = 0.2 × 0.2 mm 2 for devices IV; A beam = 0.5 × 0.5 mm 2 for devices III and V; Because photocurrent is affected by device polarization history, S was calculated using a single point at 3.4 × 10 9 p s −1 mm −2 before any polarization occurred at higher fluxes.

www.advopticalmat.de
Supporting Information). Ideally, the photocurrent density should be independent of the X-ray beam area. As shown in Figure 3d, however, the fact that photocurrent density is different for different beam areas indicates that polarization occurred along the testing track, and the photocurrent density is affected by the testing sequence. Given the progressive behavior of polarization, we further tested the CsPbBr 3 detector under continuous irradiation for several hours at a low photon flux that polarization is not a concern in several hundred seconds. As shown in Figure 3f, the decrease in the signal current of device I is negligible in several hundred seconds at a lower flux φ p = 0.89 × 10 9 p s −1 mm −2 (Vr = 300 V, E X = 8.2 keV, A beam = 1 mm 2 ).
However, at a higher flux of φ p = 2.7 × 10 9 p s −1 mm −2 and a longer time scale of several hours, the current shows significant drift, which suggests that the accumulated continuous irradiation plays a role in detector polarization. CsPbBr 3 device II, III, and V exhibited similar current decreasing behavior in a time scale of several hours (see Figure S13, Supporting Information). Finally, we measured the dark I-V curve and the 57 Co gamma-ray energy spectrum of the devices after testing under synchrotron X-ray. Since the devices were under reverse bias for several hours, the dark I-V curves remained symmetric with a small forward bias current (Figure 3g). The dark currents at reverse 200 V after synchrotron X-ray testing are comparable to  that before testing ( Table 3). The dark current change (several nA mm −2 or less) is much smaller than the photocurrent produced by X-rays (10 1 -10 5 nA mm −2 ), and therefore, does not harm the device performance reproducibility. The energy spectrum of the detectors showed different degrees of degradation due to detector polarization (Figure 3h,i). A polarized detector may recover after remaining under dark ambient conditions for several months, as observed in our previous work. [21] Active depolarization techniques, such as switching the bias voltage direction, [11] infrared LED illumination, [32] and heating, may be effective in accelerating CsPbBr 3 detector recovery, which needs further investigation.

Conclusions
We developed efficient melt-grown CsPbBr 3 crystal and device quality screening procedure with photoluminescence spectroscopy and I-V measurement and found out that the ion migration does not necessarily affect the performance of the CsPbBr 3 detectors. With the spectrometer-grade CsPbBr 3 detectors fabricated from high-quality crystals, a bias voltage of at least 1000 V (588 V mm −1 ) can be applied, which is critical to reducing the polarization and hence enable a good detector photocurrent linearity and stability. Consequently, we demonstrated that the CsPbBr 3 detectors could operate at ultra-high X-ray fluxes of up to 10 10 p s −1 mm −2 with beam area 0.25 mm 2 at 58.61 keV with good photocurrent linearity, stability, and reproducibility over a timescale of several hundreds of seconds. The capability of the CsPbBr 3 detector operating in the current mode under ultra-high flux without significant polarization also has positive implications on its performance in the pulse mode as photon-counting detector [21] because the underlying physics of polarization is the same for the current and pulse modes. Future work will need to focus on a) further improvement of crystal quality and the device fabrication process to further increase the bias voltage that can be applied and b) investigations of the dynamics and mechanisms of CsPbBr 3 polarization and its potential recovery. In future work, we will report on the performance of solution-grown CsPbBr 3 crystals.

Experimental Section
Crystal Growth and Device Fabrication: The Bridgman method was employed to grow CsPbBr 3 single crystals. The growth process parameters and cooling schemes have been previously described in detail. [20a,28b] Ingots of CsPbBr 3 were cut into pieces of appropriate dimensions for detector fabrication. First, the surfaces of CsPbBr 3 crystals were finely polished using sandpaper and then cleaned with toluene. A gold electrode, approximately 100-nm thick, was thermally evaporated onto the crystal surface. Then, a liquid GaIn electrode was brushed onto the crystal surface. We note that the thickness of the GaIn electrode could be non-uniform and may be in the range of several hundred nm to several hundred µm. The thick GaIn electrode could absorb a significant portion of 8.2 keV and 10 keV X-rays (see Supporting Information for an example calculation) and results in different photocurrent magnitude for different devices due to the non-uniform thickness. A copper wire connected the electrodes to the outer circuit, and crystals with a finished electrode and wire connection were placed on a glass holder encapsulated in paraffin wax.
Detector Testing Setup under Synchrotron X-Ray: Detector testing under synchrotron X-rays was performed at the Advanced Photon Source, Argonne National Laboratory. Measurements were performed at beamlines 1-BM-B and 11-ID-B, which provide 8.2 and 10 keV, and 58.61 keV X-rays, respectively. The experimental setup is illustrated in Figure S14 (Supporting Information). An ion chamber that measures the X-ray intensity directly out of the shutter, a set of adjustable slits that controls the beam area, and a CsPbBr 3 detector that is mounted on an X-Y-Z stage were located in sequence downstream of the beam shutter. The detector was first aligned with the X-ray beam using a laser-alignment tool. Then at a low X-ray flux, the detector position was fine-tuned to find the position that generates a maximum current. As shown in Figure S15 Supporting Information, because the GaIn electrode could be very thick at the center region (e.g., several hundred µm), the maximum current was usually produced when the X-ray beam irradiates the edge area. The X-ray flux at beamline 1-BM-B was modified by detuning the second crystal monochromator of the beamline facility, and the X-ray flux at beamline 11-ID-B was varied using a set of attenuators. A calibrated Si PIN diode was inserted into the X-ray beam path to measure the flux received by the CsPbBr 3 detector (Table S2, Supporting Information summarizes the flux calibration for beamline 11-ID-B).
Photoluminescence and Electrical Characterization: An FS5 spectrofluorometer (Edinburgh Instruments) was used to measure the time-resolved PL spectra and PL emission spectra at room temperature. The exciting wavelength for PL measurement is 405 nm generated by a laser. A Keithley 6517 B electrometer was employed to apply bias voltage to the CsPbBr 3 detector and measure current.
Gamma Energy Spectrum Acquisition: The gamma spectroscopy system used for gamma energy spectrum acquisition included an eV-550 preamplifier, a 572A shaping amplifier (ORTEC), and a 927 multichannel analyzer (ORTEC). The energy resolution of the photopeak was calculated as the full width at half maximum divided by the peak channel number. A shaping time of 10 µs is used. Given hole mobility of ≈52 cm 2 V −1 s −1 and thickness d = 1.7 mm, at a bias voltage of 300 V, the charge carrier drift time τ tr = d 2 /(µV) is estimated to be ≈1.85 µs. Since the shaping time is larger than τ tr , the ballistic deficit is not a concern affecting the energy spectroscopy performance. Ideally, the same bias voltage should be used for detector testing and performance comparison. However, the detector noise could change after heavy irradiation. The highest bias voltage without excessive noise was used for energy spectrum acquisition after irradiation.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.