Lead halide perovskite semiconductors are in general known to have an inherently high X-ray absorption cross-section and a significantly higher carrier mobility than any other low-temperature solution-processed semiconductor. So far, the processing of several-hundred-micrometres-thick high-quality crystalline perovskite films over a large area has been unresolved for efficient X-ray detection. In this Article, we present a mechanical sintering process to fabricate polycrystalline methyl ammonium lead triiodide perovskite (MAPbI3) wafers with millimetre thickness and well-defined crystallinity. Benchmarking of the MAPbI3 wafers against state-of-the-art CdTe detectors reveals competitive conversion efficiencies of 2,527 µC Gyair−1 cm−2 under 70 kVp X-ray exposure. The high ambipolar mobility–lifetime product of 2 × 10−4 cm2 V−1 is suggested to be responsible for this exceptionally high sensitivity. Our findings inform a new generation of highly efficient and low-cost X-ray detectors based on perovskite wafers.
Due to their outstanding physical properties, hybrid organic–inorganic perovskites (HOIP), and most notably CH3NH3(Pb,Sn)(I,Br)3, have received extraordinary attention from the research community. Especially significant are the charge-transport properties of MAPbI3, which has demonstrated long minority-carrier lifetimes and diffusion lengths that are comparable with single-crystalline covalent semiconductors1,2,3,4,5.
A new application for this material class is the sensitive detection of high-energy radiation such as X-ray6,7 and γ radiation8. The search for an ideal X-ray-sensitive photoconductor (PC) is of ongoing interest because most semiconductors with good charge-transport properties do not absorb high-energy radiation effectively. The latter, however, is an intrinsic property of HOIPs due to the presence of heavy metal and halide ions. Current state-of-the-art materials for direct X-ray detection9 include stabilized amorphous Se (a-Se)10, PbI2 (ref. 11), HgI2 (refs 12, 13), CdTe (ref. 14) and CdZnTe (ref. 15). a-Se detectors have been successfully commercialized but are limited to mammography applications due to their low absorption coefficient in the spectral regime higher than 50 keV. PbI2 and HgI2 detectors face stability issues that hinder their widespread application. Significant effort is being directed at Cd(Zn)Te (ref. 16), but upscaling to larger wafers and charge carrier trapping remain issues yet to be overcome.
Generally, for efficient X-ray absorption, the PC layer should be approximately three times larger than the attenuation length, which is the length at which 63% of photons are absorbed17. As discussed later in detail, HOIPs have attenuation lengths on the order of hundreds of micrometres for the X-ray energies commonly used in medical applications. Although the widely reported solution-process protocols for fabricating thin films of HOIPs are efficient in the sub-micrometre regime, producing several-hundred-micrometres-thick, high-quality films covering a large area is extremely difficult. In this Article, we present a room-temperature sintering process for MAPbI3 microcrystals to form thick (200 µm to 1 mm) rigid wafers of virtually any size.
MAPbI3 wafer sintering process
In contrast to covalent semiconductors such as Si and Cd(Zn)Te, which require high-temperature crystallization processes, ionic crystals can be processed from solution at low temperatures. An easily overlooked property of ionic crystals is their plasticity, which results in pressure-induced flow processes18. Although HOIPs are not pure ionic crystals, this material class does exhibit pressure-induced agglomeration effects.
The MAPbI3 microcrystals presented here were synthesized by precipitation from a precursor solution at room temperature19 (see Methods). The precipitated microcrystals were irregular in shape, with sizes ranging from 50 nm to 1 µm (Supplementary Fig. 1). By applying a pressure of 0.3 GPa to the microcrystals for 5 min using a hydraulic press, a compact MAPbI3 layer (which will be referred to as a wafer) was formed. A series of wafers with 1/2 inch diameter and thickness ranging from 200 µm to 1 mm were prepared. It is remarkable that these free-standing MAPbI3 wafers have a mirror-like reflective surface with a root-mean-square roughness of only ∼75 nm (Fig. 1a; for atomic force microscopy topography see Supplementary Fig. 3). Scanning electron microscopy (SEM) of a wafer cross-section shows that the wafer is homogeneous and dense (Fig. 1b). The density of the wafer is 3.76 g cm–3, which is close to the density of single-crystalline MAPbI3 (4.15 g cm–3), as calculated from lattice parameters20,21.
It is evident from the SEM cross-section that the grain boundaries of the microcrystals remain well defined in the wafer. Furthermore, grain boundary interconnections can be easily recognized (circles in the inset of Fig. 1b).
Structural and mechanical investigations
Nanoindentation has emerged as the method for evaluating the hardness and Young's modulus of small material volumes and thin films22,23. Measurements were performed with a Berkovich tip, which was pressed 2.5 µm into the MAPbI3 wafer surface using a constant strain rate of 0.05 s−1. Single-crystal MAPbI3 synthesized in house served as a reference for this measurement. To provide a measurement with continuous stiffness, a sinusoidal oscillation was superimposed to the load to evaluate the hardness and Young's modulus as a function of indentation depth24. Hardness is defined as the ratio of the load on the tip to the projected tip contact area.
The steep decrease in hardness at shallow indentation depths shown in Fig. 2a is typical for such measurements and is attributed to uncertainties in the geometry of the tip apex25. At higher indentation depths, the hardness of the MAPbI3 wafer and that of the single crystal levels off at ∼0.36 GPa and ∼0.47 GPa, respectively. The latter value is in good agreement with previously reported values26. The mechanical investigations were complemented by strain rate jump tests, which are a measure of the dynamic mechanical behaviour of a material, that is, the dependence of its strength on the imposed deformation rate27. With this technique, a strain rate sensitivity m value of 0.065 was estimated (Supplementary Figs 4 and 5). This value is in the upper range of m values found for metallic nanocrystals28 and indicates that time-dependent deformation mechanisms such as diffusion, dislocation activity and plastic flow due to grain boundary sliding are dominant. This viscoplasticity is interpreted as the underlying mechanism for the mechanical rigidity of the wafers.
To investigate the structural properties we performed X-ray diffraction (XRD) measurements. Figure 2b clearly shows that the diffraction peaks of the MAPbI3 wafer, microcrystals and single crystal are at the same positions. Major diffraction peaks at angles of 14.0, 23.4, 24.4, 28.0 and 28.3° correspond to (110), (211), (202), (004) and (220) lattice planes, and are consistent with the tetragonal I4cm space group previously reported by Stoumpos et al. and Kawamura and co-authors21,29. The full-width at half-maximum (FWHM) of the diffraction peaks from the wafer and microcrystals are comparable. Also, the maximum of the photoluminescence spectrum at 770 nm (Supplementary Fig. 6) is in good agreement with the literature30,31,32,33.
To conclude, the observed time-dependent flow process triggers a pressure-induced agglomeration of the microcrystals due to deformations on their grain boundaries, and photoluminescence, XRD and SEM investigations confirm that the crystallinity of the microcrystals is retained. These results are consistent with in situ XRD measurements performed on MAPbBr3 and MAPbI3 crystals, which indicate that amorphization is observed at a much higher pressure of 2 GPa, and it is reversible up to 34 GPa (ref. 34) and 50 GPa, respectively33.
In polycrystalline materials, the macroscopic charge-carrier mobility and the free charge-carrier lifetime are dominated by the nature of the grain boundaries; that is, boundaries will scatter carriers being transported from grain to grain, surface defects may lead to deep trap states, and space-charge layers will act as potential barriers35. To investigate the charge-carrier transport properties of the polycrystalline MAPbI3 wafers, the time-of-flight (TOF) technique was used. TOF is an elegant method frequently used for organic36 and inorganic semiconductors37. Generally, a laser pulse generates a carrier reservoir Q in a thin layer near the illuminated surface, which is subsequently extracted by the applied electric field. The laser pulse intensity was kept low to prevent a distortion of the electric field (Q < CV, where C is the sample capacitance and V is the bias voltage). At too high laser intensities the excess charge-carrier concentration distorts the electric field and the temporal response features an overshoot38,39. Consequently, the laser intensity was reduced until the temporal response of the transient current remained unaltered. The device stack consisted of glass/indium tin oxide (ITO)/poly(methyl methacrylate) (PMMA)/MAPbI3 wafer/Au. The 0.5-µm-thick PMMA layer was used to suppress current injection from the ITO electrode. The sample was optically excited from the glass side using a nanosecond laser emitting at 532 nm. The transient photocurrent response for holes, shown in Fig. 3a, represents a characteristic and well-resolved TOF transient, in which the transit time ttr of the fastest charge carriers is defined as the intersection of asymptotes from the plateau and tail in a double logarithmic I(t) plot40,41. The TOF transients show a clear trend towards non-dispersive transport at higher electric fields.
The mobility is calculated from ttr using μ = d/(Ettr), where d is the sample thickness and E is the applied electric field. For holes, the mobility is in the range of 0.53–0.70 cm2 V−1 s−1, with a negative field dependency (that is, due to increased scattering at higher E fields, Fig. 3b). Under the same experimental conditions but with reversed bias polarity, a slightly lower mobility of 0.50–0.56 cm2 V−1 s−1 and a more dispersive transport for electrons (Supplementary Figs 7 and 8) is measured.
This observed ambipolar transport is remarkable, as very low or even missing charge transport of either electrons or holes is commonly observed in amorphous and polycrystalline systems11,42. The field dependency of the charge transport indicates potential barriers (that is, at the grain boundaries), while quenching of free charge-carrier lifetime and of the macroscopic mobility via deep traps is unlikely to be present. These findings are consistent with recent theoretical investigations that show that MAPbI3 is a defect-tolerant semiconductor and that vacancy-type defects (that is, iodine vacancies) are resonant in the bands43.
Direct X-ray to current conversion
The device stack of the X-ray detector is shown in Fig. 4a. The electric contacts of the 1-mm-thick MAPbI3 wafer are formed by poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) and phenyl-C61-butyric acid methyl ester (PCBM) and ZnO as hole-selective and electron-selective contact and buffer layers, respectively. A bottom ITO/glass substrate and a silver top electrode finish the device stack.
The sample was irradiated with X-ray radiation of 70 kVp from the silver electrode side for 2 s. The most intense emission was at 38 keV (see Supplementary Information, pages 7 and 8 for the set-up with the X-ray source and the simulated X-ray photon-density spectrum). A simulation using the XCOM database estimates an attenuation depth in the MAPbI3 wafer at 38 keV of ∼125 µm (Supplementary Fig. 11). Consequently, the fraction of incident X-ray photons that are attenuated approaches unity17 (Supplementary equation (3)).
The photocurrent response under constant reverse bias (the electrodes are selectively e/h extracting) and increasing dose rates is presented in Fig. 4b. The extracted charge shows a nearly linear X-ray dose dependency up to 5 mGyair (Fig. 4d). The slope of a linear fit (red line in Fig. 4d) is defined as the sensitivity of the detector and we obtained a conservatively estimated sensitivity of ∼2,527 µC Gyair−1 cm−2 at an electric field of 0.2 V µm–1. Figure 4c shows the dependence of the extracted charge from the MAPbI3 sample on the electric field at a constant dose rate of 6.72 mGyair s−1. A fit with the Hecht equation gives a tentative μτ product of 2 × 10–4 cm2 V−1 (red line in Fig. 4c).
To compare and validate the X-ray response of the MAPbI3 wafer-based detector, we measured the charge released from a 1-mm-thick single-crystalline CdTe sensor layer on the photon-counting pixel detector ‘Timepix’ (see ref. 44 and the Supplementary Information, page 12, for a technical description). For comparison, the CdTe sensor layer has the same thickness as the MAPbI3 wafer and was biased with the same electric field. As is evident from Fig. 4d, the amount of extracted charge Q from the MAPbI3 sample is comparable to the commercial CdTe ‘Timepix’ reference detector. For doses <5 mGyair, Q is almost identical to that of the CdTe reference (∼5% higher for the CdTe sample). For doses >5 mGyair, Q deviates from the CdTe reference (due to enhanced recombination).
Another essential parameter is the ionization energy W±, which is the amount of radiation energy absorbed to create a single free electron and/or hole. Stated differently, it is the total adsorbed energy divided by the total number of extracted electrons. For the MAPbI3 wafer, an ionization energy of ∼5 eV at 0.2 V µm–1 electric field is calculated (see Supplementary equation (7)). This value is in close agreement with the Que-Rowlands45 and Klein rules46, which predict the lowest limit of W± according to W± = 2.2Eg + Ephonon ≤ 3.9 eV and W± = 3Eg = 4.65 eV, respectively, where Eg = 1.55 eV is the bandgap and Ephonon ≤ 0.5 eV is the phonon energy. We conclude that the low W± is due to an excellent collection efficiency (CE) and a low geminate recombination rate. This finding is in agreement with recent investigations on MAPbI3 single crystals claiming a CE of 75% using 20–35 keV soft X-ray irradiation with an electrode distance of 2.5 mm (ref. 47).
Despite these excellent electronic properties, one important drawback of these devices is the high and rather unstable dark-current density of 6 µA cm–2 at 0.2 V µm–1 reverse bias (see baseline drift in Fig. 4b). In MAPbI3, the dark conductivity is dominated by ionic conduction (free iodine migration48) and the current instability at reverse bias is caused by the non-ideal e/h selective electrodes. Carrier-selective electrodes featuring improved e/h selectivity are currently an area of intense research. Improvements in this direction are expected to further advance the performance of perovskite-based X-ray converters.
A comparison of previously reported sensitivities, as shown in the Supplementary Information (Supplementary Fig. 14), reveals that the sensitivity value, the μτ product9 and the ionization energy W± are on par with the best available technologies, such as HgI2 and Cd(Zn)Te (refs 9, 13). In addition to its toxicity, the electronic parameters of HgI2 are strongly dependent on the deposition process and can vary from sample to sample49. Conversely, crystalline Cd(Zn)Te wafers offer close to ideal properties but require energy-demanding production processes such as zone melting, the Bridgman method, epitaxial growth, or the travelling heater method (THM)50. Furthermore, wafer inhomogeneities due to Cd,Te inclusions51 and upscaling to larger areas remain issues with these technologies16. Many of theses shortcomings are resolved by the presented sintering method for MAPbI3 microcrystals, which allows the formation of rigid wafers of virtually any size.
In conclusion, in this Article, we present a sintering process to manufacture rigid, several-hundred-micrometres-thick MAPbI3 wafers. The wafer conserves the structural and optical properties of the microcrystalline starting material. Ambipolar charge transport is demonstrated with a mobility of 0.45–0.7 cm2 V−1 s−1. Under X-ray exposure, a μτ product of ∼2 × 10–4 cm2 V−1 is measured with an excellent conversion sensitivity of 2,527 µC Gyair−1 cm−2 at 0.2 V µm–1 electric field.
MAPbI3 single-crystal and microcrystal synthesis
The MAPbI3 microcrystals were synthesized by a precipitation reaction as reported in ref. 19. A 1 M precursor solution was prepared by dissolving an equimolar ratio of methylammonium iodide CH3NH3I (synthesized in house) and lead(II) iodide (Sigma Aldrich) in γ-butyrolactone (GBL) at 60 °C and stirred for 3 h. Chloroform was added to the filtered precursor solution, which immediately precipitated CH3NH3PbI3 microcrystals. The microcrystals were washed twice with chloroform in a centrifuge, and then dried in a vacuum oven at 40 °C overnight. The synthesis of the single crystals followed procedures in refs 52 and 53. In brief, the precursor was heated at 100 °C for 3 h, which produced small MAPbI3 crystals. One small crystal was used as a seed in 1 M precursor solution and allowed to grow overnight at 100 °C.
A hydraulic press (Specac) with a 1/2-inch die was used for sintering. A polished stainless-steel cylinder was placed in the bore of the cylinder body followed by 250–500 mg of CH3NH3PbI3 powder. Next, a second polished cylinder and a plunger were inserted into the cylinder body. A vacuum membrane pump was connected to the die. After about 2 min of turning on the vacuum pump, a pressure of 0.3 GPa was applied to the plunger for about 5 min to form the MAPbI3 wafer.
Fabrication of the MAPbI3 wafer based photodetector
Glass substrates with sputtered ∼100-nm-thick ITO were cleaned in an ultrasonic bath with acetone and isopropanol for 10 min and then placed in an oxygen plasma oven for 2 min. PEDOT:PSS (VP Al 4083 from H.C. Starck) was spin-coated onto the cleaned substrate to form a 100-nm-thick thin film. MAPbI3 wafers were pressed onto the PEDOT:PSS/ITO substrates at about 15 MPa for 2 min using a hydraulic press. PCBM (20 mg ml–1) then ZnO were subsequently spin-coated onto the wafers. The Ag top contacts (6.6 mm2 in area and 70–100 nm thick) were deposited by thermal evaporation.
SEM measurements were performed on MAPbI3 wafers and the microcrystalline particles supported on standard adhesive carbon pads using an ULTRA 55 Carl Zeiss AG SEM under 10 kV acceleration voltage. For the image of the cross-section, the wafers were ‘broken’ with a hammer.
Atomic force microscopy measurements
An NT-MDT nano-educator AFM in contact mode was used.
Photoluminescence spectra were measured using a 450 nm laser diode as the excitation source in a back-scattering configuration at room temperature. The laser beam was focused to a spot with a diameter of ∼10 µm. The emission was recorded with a Si charge-coupled device camera (Syncerity, Horiba Jobin-Yvon) attached to an iHR320 monochromator (Horiba Jobin-Yvon) with a resolution of 5 nm. The spectra were corrected for the spectral sensitivity of the system, determined with the help of a calibrated halogen lamp (Stellarnet).
Nanoindentation measurements were performed using a Keysight G200 Nanoindenter. A Berkovich tip was pressed 2.5 µm into the sample surface, using a constant strain rate of 0.05 s–1. Nanoindentation was carried out in continuous stiffness mode (CSM), where a small oscillation is superposed on the indentation load. A 2 nm sinusoidal oscillation was superposed to the loading signal. The hardness values were average over 25 measurements taken at least 100 µm apart. Strain-rate jumps tests were implemented during the nanoindentation loading segment. The corresponding strain rates ranged between 0.025 s–1 and 0.005 s–1. Strain-rate sensitivity was determined as the slope of a linear fit in a double logarithmic plot of hardness versus strain rate.
PMMA (Sigma Aldrich) in chlorobenzene (100 mg ml–1) was spin-coated on cleaned ITO substrates to form a 500-nm-thick layer. MAPbI3 wafers were pressed into the PMMA/ITO substrates at ∼50 MPa for 2 min using the hydraulic press. The Au top contacts (70 nm thick and 0.2 cm2 in area) were thermally evaporated to provide the charge-collecting electrode. As excitation, a nanosecond pulse laser (532 nm, CryLaS) was used. The bias voltage was applied to the sample using a function generator (Agilent 33500B) and a voltage amplifier (Falco Systems WMA 300). The transient current was measured as the voltage drop over a load resistor and recorded with a digital oscilloscope (Tektronix DPO 3034).
Photocurrent under X-ray exposure
A 70 kV X-ray source (Siemens MEGALIX Cat Plus 125/40/90, 124 GW) with a tungsten anode was used. The X-ray spectrum was filtered with a 2.5-mm-thick Al plate. The dose rate was changed by changing the X-ray tube current and was calibrated with a PTW Diados T11003-001896 dosimeter. A Keithley 2400 SMU measuring at 10 Hz was used to record the photocurrent. See Supplementary Information for additional details.
The data that support the plots within this paper and other findings of this study are available from the corresponding author upon reasonable request.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The Cluster of Excellence Engineering of Advanced Materials (EAM) at the FAU University Erlangen and the Gradko 1896 ‘in situ Microscopy’ (DFG) is acknowledged for support. The authors thank C.O. Quiroz and M. Salvador for reading the manuscript.