Low-noise X-ray PIN photodiodes made of perovskite single crystals by Solution-processed doped epitaxial growth

X-ray photodiodes made of metal halide perovskites (MHPs) which directly convert X-ray photons into electron-hole pairs have shown advantages in low-cost and high X-ray detection sensitivity. However, devices fabricated by spin-coating and evaporation methods suffer from high traps density near poor interfacial layers (n-type/intrinsic and p-type/intrinsic) which lead to high dark current and noise current under large reverse bias. In this work, solution-processed doped epitaxial growth is employed to limit these traps through epitaxially growing n-type MHPs (bismuth-doped) and p-type MHPs (silver-doped) on opposite faces of intrinsic CH 3 NH 3 PbBr 2.5 Cl 0.5 MHP single crystals. Through energy structure design, effective electron/hole blocking layers could decrease the noise and dark current, these X-ray PIN photodiodes work under a large external electrical field, which enables a state-of-art response speed (fall) of 750 ns and a lowest detectable dose rate of 17.7 nGys -1 (40 kVp). This work will motivate new strategies to fabricate high-performance devices based on perovskites using solution-processed methods. These founding also explore a new generation of low dose and high dynamic X-ray detectors based on MHPs.

However, to capture hard X-ray photons (>100 keV), the thickness of the MHPSCs must be at least 1 mm. Consequently, a long penetration path requires a large electrical field to drift the photoinduced charges for high-efficiency charge collection. However, the dark current increases with the applied bias in MHPSC-based photodiodes. Meanwhile, the noise current increases and floods the weak signal current induced by X-ray photons. Therefore, the X-ray sensitivity, response speed, and detection limit must be balanced against the noise current and dark current when setting the value of the external bias.
In addition to the crystallization quality improvements [11][12][13] , several photodiode structures have been investigated to decrease the dark current of MHPSC-based photodiodes, such as sandwiched MHPSC-based PIN photodiodes fabricated using two deposited charge-blocking layers [14][15][16] , MHPSC-based PN junctions formed by doping 17,18 , and MHPSC-based Schottky diodes realized using asymmetric metal electrodes 10,19,20 . However, most MHPSC-based PIN photodiodes use organic semiconductors 21 , nano-oxides 22 and carbon derivatives 23 as blocking layers that are not lattice-matched with MHPSCs, and the traps near interfacial layers (n-type/intrinsic and p-type/intrinsic) lead to the leakage of dark current. MHPSC-based PN junctions realized by growing n-type doped MHPSCs (Bi 3+ , Se 2+ ) on undoped p-type MHPSCs suffer from narrow depletion layers, rendering X-ray detection inferior compared to that with MHPSC-based PIN photodiodes 17,18 . MHPSC-based Schottky diodes can effectively decrease the dark current, but metals such as Al, Ga, Ag, and Cr could react with halide ions in days due to their low work function, leading to performance degradation 10,19,20 . As a result, PIN photodiode structure is suitable for X-ray detectors and low traps interfaces could further decrease the noise. Thus, MHPSC-based PIN photodiodes fabricated by lattice-matched blocking layers epitaxially grown [24][25][26][27][28][29][30] on intrinsic MHPSCs may further decrease the traps near the interfaces and lead to low dark current and noise.
The traps near the interfaces can be suppressed by solution-processed doped epitaxial growth (SPDE) grown n-type and p-type lattice-matched MHPSCs on the opposite face of the intrinsic layer. Differing from the spin-coating and evaporation fabrication strategy of nanometer-scale thickness films 8 , p-type and n-type layers, tens of micrometers thick, can effectively block the injected charges from the large voltage source without leakage current from pinholes. This work not only demonstrates that SPDE methods are able to fabricate high-quality PIN photodiodes for X-ray detection, but also provides an important guide to improve solutionprocessed and perovskite-based devices.  Based on the key information( Fig. 1f and Fig. 1g) of the band structure, the energy structure of these layers is shown in Fig. 2a. (where ni is the intrinsic carrier concentration~10 7 cm -3 , k is the Boltzmann constant and T is the temperature) 16 .

Results:
When these MHPSCs contact and the Fermi levels have the same value, the VBM difference between n-type and intrinsic layers is obtained as 0.76 eV, while the CBM difference between p-type and intrinsic layers is 0.41 eV. Additionally, the large energy barriers of 1.91 eV and 1.56 eV prevent hole and electron injection.
As shown in Fig. 2b, the MAPbBr2.5Cl0.5 MHP single crystals were first grown by inverse temperature crystallization (ITC) 6   The long-term stability of the dark current under a -200 V bias on the X-ray PIN photodiode was measured (Fig. 2e). It was found that ion migration inside bulk MAPbBr2.5Cl0.5 under bias requires approximately 1.5 h to achieve stability, and the dark current density drifts approximately ~3 pAcm -2 , which indicates a very low noise. The stability may be attributed to the ultralow rate of charge injection. Fig. 2f shows the magnified forward and reverse current density-voltage from -150 V to -50 V with step 1Vs -1 . The current density varies by 0.072 nAcm -2 and 0.041 nAcm -2 in the forward and reverse scans, respectively. Thus, the current-density-voltage hysteresis 44 is only 0.031nAcm -2 under -150V bias which is much smaller than spin-coating based PIN photodiodes and Schottky diode ( Supplementary Fig. 5). Consider the same intrinsic layer, the tiny low traps density of lattice-matched interfacial layers leads to this tiny current-density-voltage hysteresis.  In addition to the electrical properties, the space charge region (SCR) plays an important role in PIN junctions. Here, Kelvin probe force microscopy 47 (KPFM) was used to measure the contact potential difference (CPD) between the n-type/intrinsic and p-type/intrinsic layers. As shown in Fig. 3a, the distance between the KPFM probe and the polished surface is set at 150 nm, which is greater than the roughness of the samples of 96 nm and 51 nm, respectively, as shown in Supplementary Fig. 7. Fig. 3b shows the potential distribution of the scan areas, the CPDs are 0.86 mV and -0.55 mV, respectively, which correlate with the energy band differences of 0.83 eV and 0.52 eV calculated from the XPS (Fig. 1f) and UPS profiles (Fig. 1g). Fig. 3c shows the line profile of the CPD, the SCR is located near the interfaces of n-type/intrinsic and p-type intrinsic layers. Due to the intrinsic layer being over 4-mm-thick, it is impossible for electrons in the n-type layer to diffuse into the p-type layer without a bias voltage. The width of the SCR in both cases is below 100 nm under illumination (650 nm).
In addition to the zero-bias situation, the SCR is more important when the PIN junction operates at hundreds of volts; however, it is not feasible to apply such a large bias to the KPFM system and to scan at the millimeter scale using electron beam-induced conductivity methods. There is a high charge density difference across the SCR under reverse bias, while the excited secondary electrons in scanning electron microscopy (SEM) are sensitive to the electron density of the surface. Here, a probe station is installed inside the SEM cavity ( Supplementary Fig. 8) to apply voltage to the photodiode, and the SEM images can reveal the counts of secondary electrons of the photodiode surface, as shown in Fig. 3d [48][49][50] . respectively. The p-type layer is slightly darker than the n-type layer owing to the lower electron density, and the high-resolution transmission electron microscope (HR-TEM) shows the lattice structure of the n-type and p-type layers which remains single crystalline as shown in Fig. 3f.
Typical HR-TEM photos and electron diffraction image are shown in Supplementary Fig. 9. The elemental distribution was investigated by energy-dispersive X-ray spectrometry (Supplementary  is the applied voltage, and d is the thickness of the X-ray PIN photodiode, as shown in Fig. 4b. The large penetration depth of X-ray photons requires a strong electrical field to drift to the electrodes. Consequently, to realize high charge collection efficiency X-ray detection, large reverse is necessary and -200V is chosen as operating bias. The product measured here is in agreement with the product calculated from independent TOF and TRPL measurements (1.39 × 10 +G G +? , Fig.   1e and Fig. 2d). Fig. 4c shows the photocurrent response to low-dose-rate X-ray photons; the anode voltage is set at 40 kVp, while varying the anode current controls the dose rate; aluminum of thickness 2 mm is used as a filter. The voltage bias is set at -200 V for efficient charge collection. The lowest measured dose rate from our X-ray tube is 41 nGys -1 limited by the minimum anode current of our X-ray tube (40 kVp, 0.1 mA), which leads to a photocurrent density of 0.28 nAcm -2 . The signal noise of the X-ray PIN photodiode for 40kVp X-ray under -200 V was measured as ~0.005nAcm -2 . By fitting the photocurrent density versus the dose rate, the detection sensitivity for 40 kVp X-ray is measured as 9.62 µC mGy +? +G and the lowest detectable 40kVp X-ray dose rate with signal-tonoise ratio of 3 was 17.7nGys -1 as shown in Fig. 4d.The X-ray sensitivity of the X-ray PIN photodiode is related to the photon energy, mainly due to the different absorption coefficients. Therefore, the X-ray detection sensitivities with 50 kVp, 100 kVp, and 120 kVp were then measured, as shown in Fig. 4e. The X-ray detection sensitivity decreases from 8.0 µC mGy +? +G to 4.03 µC mGy +? +G , and 3.15 µC mGy +? +G , while the X-ray photons become harder.
In addition to the X-ray detection sensitivity, higher response speeds can shorten the exposure time, which is suitable for high frame counts and low-dose radiography. The response speed was investigated by an X-ray photodiode-external resistance (0. Subsequently, the X-ray PIN photodiode was used for X-ray imaging, as shown in Fig. 4f; two screws were used as the object and a two-direction slide table was used to move the X-ray PIN photodiode to realize space scanning. In the high-current-threshold image, the screw thread is clear from the background owing to the dose difference. In the low-current-threshold image, owing to the low noise current, the X-ray PIN photodiode could recognize the screw thread on the screws where the transmitted X-ray dose was extremely weak. This image suggests that this low-noisecurrent X-ray PIN photodiode could possibly realize better contrast in objects. In addition, ion exchange and migration 55-58 may degrade perovskite-based devices; here, the device is placed at temperatures from 15 (room temperature) to 90 °C in air to accelerate this phenomenon (Fig. 5a). The dark current density increased with the temperature (nearly 1 nAcm -2 /℃), indicates this device could be applied in thermal detection 59 . Then, the device was maintained 90 °C for 12 h as shown in Fig. 5b, the dark current density maintains ~104 nAcm -2 . After that, when the device was cooled back to 15 °C and the dark current density recovered without degradation (Fig. 5a). This proves that the ion exchange and migration speeds are extremely low in this X-ray photodiode and this X-ray PIN photodiode shows a better thermal stability than amorphous selenium. Further, as shown in Fig. 5c, the dark current density of this X-ray photodiode was investigated after 9 months in air without protection and after irradiation by high-dose X-rays in air without shielding; the drift of the dark current density is less than 0.1%, which demonstrates the high stability of the X-ray PIN photodiode.

Discussion
In conclusion, the SPDE method for growing X-ray PIN photodiodes is demonstrated in this work.
Based on the doping concentration and device stricture design, the X-ray PIN photodiodes exhibit remarkably low dark current and noise current under large bias. A record response time of 750 ns was achieved and a 40 keV X-ray sensitivity of 9.62 µC mGy +? +G was observed in this work, which could provide high-contrast and high-speed X-ray images. This work not only demonstrates that the SPDE method is capable of fabricating high-quality PIN photodiodes for X-ray detection, but also provides important insights to guide the future development of solution-processed and perovskite-based devices.