Main

A semiconductor can serve as an efficient solid-state gamma detector when it simultaneously exhibits several distinct characteristics: high resistivity (≤1010 Ω cm), a high µτ product for the efficient collection of carriers and a high average atomic number (Z) for sufficient absorptivity of high-energy photons. The µτ product is highest in single-crystalline semiconductors, where carrier transport is not limited by scattering and trapping at grain boundaries. However, the availability of such high-Z semiconductors in sufficiently large, single-crystalline forms that are both chemically and mechanically robust, is limited. Very few materials are known to meet these requirements at room temperature2,6,7,8,9 and so far only ultrapure CdTe and CdZnTe SCs, usually grown from a melt by the Czochralski method, have been commercially deployed1,2,6,10,11. Overall, due to these stringent requirements, the prices of solid-state semiconductor detectors escalate from visible light to hard-radiation detection by two to three orders of magnitude if calculated for array detectors.

Metal halide semiconductors with perovskite crystal structures12 comprise an emerging class of optoelectronic materials. The outstanding tolerance of optical and electronic characteristics to structural defects, found in methylammonium lead halide perovskites (MAPbX3, where X = Cl, Br or I), has enabled a variety of optoelectronic applications, such as inexpensive solar cells13,14,15,16 with certified power conversion efficiencies that exceed 21%, broadband and narrowband photodetectors operating in the ultraviolet–visible–near infrared17,18,19 and soft X-ray spectral regions20,21, as well as light-emitting diodes12,22 and lasers23,24. Using the atomic absorption coefficients for MAPbI3 (ref. 25) we obtained a rendered absorption spectrum over a broad range of photon energies (Fig. 1a). The absorption cross-sections obtained in this way are similar to that of CdTe (http://physics.nist.gov/PhysRefData/XrayMassCoef/tab4.html). Except for resonant absorption at the K, L and M edges, the penetration depths drastically increase up to ∼1 MeV. Soft X-rays (<10 keV) can be absorbed by a layer of MAPbI3 with a thickness of tens of micrometres, as recently shown for the direct photoconductive sensing of X-rays by polycrystalline MAPbI3 films20. A combined benefit of the reduced defect density, high µτ product and reduced surface/interfacial recombination could be demonstrated with MAPbBr3 SCs21, which achieved much higher X-ray photoconductivities for hard X-rays (22–50 keV). A path length of millimetres to centimetres is required for the complete absorption of hard X-rays and γ photons. The gammavoltaic response of 3 mm MAPbI3 SCs to intense γ radiation (source activities of several TBq) has recently been presented5. In this work, we sought to systematically investigate the possibility of single γ-photon counting using SCs of various hybrid lead halide perovskites, focusing on operation at low γ photon intensities (source activities down to 0.4 MBq), as required for potential applications in solid-state gamma detection. Of all aforementioned optoelectronic applications, γ-photon counting is perhaps the most demanding with respect to low dark currents, low noise levels and high µτ products.

Figure 1: MAPbI3 perovskite SCs for gamma detection.
figure 1

a, The attenuation coefficient and corresponding penetration depth of MAPbI3 and CdTe as a function of photon energy, from soft X-rays to γ radiation. b, The bias dependence of the photocurrent generated by Cu Kα X-ray radiation (8 keV) in a SC of MAPbI3 perovskite; the red line indicates a fit with the Hecht model showing a high μτ product of ∼10−2 cm2 V−1. Top inset: Photograph of typical MAPbI3 perovskite SCs grown from a non-aqueous (retrograde solubility) method, placed on a millimetre ruler. Bottom inset: Schematic of the three-dimensional interconnection of PbI6-octahedra in a perovskite lattice (green, Pb; yellow, I; blue, MA). c, Measurement of the decay of the activity from a short-lived 11C isotope (in the form of 11CO2 gas in a bulb) with initial activity A0 = 70 GBq. d, Response of the conductivity of the MAPbI3 SC to the manual placement of a 137Cs source with activity A of 2.2 MBq.

Full size image

In this work, sufficiently large (3–12 mm in size) SCs of MAPbI3, MAPbBr3, MAPbCl3, FAPbI3, FAPbBr3 and I-treated MAPbBr3 (larger for MAPbI3, Fig. 1b, inset) were grown using two methods: from saturated aqueous solutions3,5,18 and from non-aqueous polar organic solvents using retrograde (inverse) solubility4,26,27. In either case, the estimated price of a SC is on the order of just US$ 0.5–1.0 per cm3 because the solutions are prepared from only moderately pure and inexpensive commercial reagents and the precursor-to-crystal yield can reach 50%. Only the results for iodide-based systems—MAPbI3, FAPbI3 and I-treated MAPbBr3—are presented here, because other compositions exhibited much lower sensitivities to γ photons. A commonly reported challenge for thin-film, high-surface-area MAPbI3 devices such as solar cells is the eventual chemical decomposition of the active layer, yielding PbI2 and volatile products. MAPbI3 SCs exhibit remarkable stability for over 8 months of storage and testing as gamma detectors, presumably due to the lower surface-to-volume ratio than thin films or nanostructures. We must also stress that the megabecquerel to gigabecquerel radioactivities used in this work correspond to rather weak energy fluxes of 10−4–10−1 W cm2, well below the solar power density of 100 W cm2 to which perovskite solar cells are usually exposed.

The µτ product is a fundamental figure of merit for gamma detectors that, under a given applied bias, provides an estimate of the ability of charge carriers in a SC of a given size to reach the current collectors before recombination. The bias dependence of the photoresponse (Fig. 1b), using soft X-rays for the excitation of carriers in the bulk of the crystal (Cu Kα, 8 keV), can be evaluated using the Hecht model28,29:

In this equation, Q/Q0 is the carrier collection efficiency, d is the distance between the electrodes and E is the magnitude of the electric field. The fit of the measured data for MAPbI3 gives a high µτ product of 1 × 10−2 cm2 V−1 (with a 20–30% crystal-to-crystal variation), comparable to that of the highest quality CdZnTe SCs30. For comparison, similarly high values of 0.3–1 × 10−2 cm2 V−1 can also be estimated from other reports on the mobility (25–100 cm2 V−1 s−1, at a low density of photoexcited carriers)5 and lifetime (∼100 µs)5 of MAPbI3 in SC form. We further assessed the sensitivity of MAPbI3 to soft X-ray photons (Cu Kα, 8 keV). With a photon flux of approximately 5.6 × 107 cm−2 s−1, a photocurrent of 22 nA can be obtained in a SC of MAPbI3 with a receiving area of 8 mm2, yielding a sensitivity of ∼0.65 µC mGyair−1 cm−2 (see Supplementary Note 1). Taking into account the penetration depth of 30 µm, the specific sensitivity is ∼220 µC mGyair−1 cm−3, which is roughly an order of magnitude higher than the sensitivity of thin, polycrystalline films tested in a previous study using an identical experimental set-up20; this confirms the crucial role of single-crystallinity in achieving an improved charge collection. The detection speed was estimated from the dependence of the photocurrent on the modulation speed, yielding a 3 dB-frequency of 200 Hz. For comparison, operation at half of that speed (100 Hz at 3 dB) was observed in polycrystalline films, despite the films being deposited directly onto the electrode finger structures with charge collection paths that are 100 times shorter, indicating an effective 200-fold improvement of the single-crystalline material over the polycrystalline equivalent.

Successful tests of the direct photoconductive sensing of X-rays by MAPbI3 led us to subject a perovskite SC to radioactive 11C that emits one γ photon of 0.96 MeV due to positron emission, and then two 0.51 MeV photons from the subsequent positron–electron annihilation. A SC of MAPbI3 was placed 10 cm from a bulb that contained radioactive 11CO2 gas with a relatively high starting activity (70 GBq), separated by a mechanical chopper constructed of 1-cm-thick lead plating. By continuously recording the photocurrent (Fig. 1c), we could estimate a mono-exponential decay of activity that is equivalent to the half-life of 11C (20.3 min). Lock-in signal modulation using the lead chopper helps in reducing the strong effect of the observed electronic polarization, presumably caused by the significant mobility of ions within MAPbI3 (ref. 31). The overall efficiency of the photon-to-current conversion was ∼19% (Supplementary Note 2), if fully quantitative carrier multiplication is assumed. The corresponding γ photon-to-charge carrier generation efficiency was estimated to be 2.5 × 104. A high dynamic range of sensitivity also allows the sensing of much lower activities (Fig. 1d), such as that from a portable source of 137Cs with an activity of 2.2 MBq (0.66 MeV, half-life of ∼30 years). The drift of the dark current is again ascribed to ionic motion in the detector. The photocurrent amplitude is ∼40% of that of a commercial CdZnTe (CZT) crystal of a similar volume and tested under identical conditions.

We then used MAPbI3 SC detectors for single-photon counting. Here the current generated by a single γ photon must be sufficiently above the noise level; it can then be amplified by the appropriate electronics. For simplicity, we placed a SC of MAPbI3 in the place of a SC of CZT in a commercial full detector assembly (from eV Products, Fig. 2a). Clear single pulses can be seen from a 137Cs source (2.2 MBq activity, Fig. 2b). The single γ quantum pulses are characterized by a distinct, fast rise. A relatively long pulse decay of 100 µs (Supplementary Fig. 1) allows a relatively simple counting scheme to be applied. We used an inexpensive sound card from a personal computer with a discretization frequency of 144 kHz, which is sufficient according to the Nyquist–Shannon sampling theorem for the analysis of single γ quantum pulses. The software package PRA10 (Bee Research Pty Ltd) was used for the comprehensive analysis of the detector pulses. Standard amplifier and multichannel analyser schemes yielded identical results to the sound card. Figure 2c presents the response of such a detector to manual placement in close proximity to a 137Cs source for pulse-counting in bins of 1 s. Compared with simple photocurrent measurements (Fig. 1d), the signal-to-noise ratio is clearly improved and the drift of the baseline dark current is almost fully eliminated. Spatially resolved detection can be illustrated by a simple two-pixel detector constructed from one perovskite SC (Supplementary Fig. 2a) using three electrodes. Single-channel registration becomes possible because the pulses from each pixel exhibit opposite signs of electrical polarity. By scanning the 137Cs source across the surface of the detector, the spatial profiles with a full width at half maximum (FWHM) of about 15 mm can be registered (Supplementary Fig. 2b), corresponding to the convolution of a pixel size of ∼7 mm and a source aperture of ∼10 mm.

Figure 2: MAPbI3 SC integrated in a γ quanta counting detector.
figure 2

a, Photograph of the detector assembly. b, Signals from the MAPbI3 perovskite SC detector under exposure to a low-activity 137Cs source (red) and noise (black). The curves are offset for clarity. c, Signal-to-background comparison for the MAPbI3 perovskite SC detector with and without a 137Cs source (2.2 MBq, manual placement).

Full size image

We subsequently investigated the potential utility of perovskite gamma detectors in nuclear medicine. In particular, positron-emission tomography (PET, commonly used in oncology, neurology and cardiology32) is a critical application where good sensitivity to low levels of radioactivity is required because of safety concerns. As a demonstrative experiment, we applied a MAPbI3 SC detector in the radio-isotopic purity control of a labelled compound, 18F-fallypride (Fig. 3a), which is regularly used in studies of dopamine receptors in the human brain (linked to Parkinson's disease, schizophrenia, attention-deficit hyperactivity disorder, social phobia and drug/alcohol dependence)33. The decay of 18F occurs via positron emission (with a half-life of 110 min), immediately followed by positron–electron annihilation producing two oppositely-directed γ photons, each of 0.51 MeV in energy (equal to the electron rest energy); the simultaneous detection of these two photons serves as the basis for PET imaging. The radiotracer must be synthesized, purified and characterized shortly before injection into a patient (see details in the Methods). High-performance liquid chromatography (HPLC) instrumentation equipped with a state-of-the-art commercial gamma detector (in this case, GABI* from Raytest GmbH) is commonly used for both purification and quality control. For parallel detection with a SC of MAPbI3, we simply wrapped the HPLC exhaust tubing around our detector assembly (Fig. 3b). The results closely reproduce those of the commercial GABI* device. Furthermore, semi-log intensity plots also illustrate the coincidence of signals from impurities that amount to <1% of the main eluted peak of 18F-fallypride (Fig. 3c,d). As expected, attempts to detect 18F-Fallypride in direct photocurrent measurement mode were unsuccessful even for the highest permitted activities of 70 MBq.

Figure 3: Realization of a HPLC set-up with a MAPbI3 perovskite SC γ quanta counting detector.
figure 3

a, The chemical structure of the well-studied high-affinity dopamine D2/D3 receptor antagonist 18F-fallypride used in PET. b, Schematic of the HPLC experiment, in which the radiopurity of 18F-fallypride is monitored by a MAPbI3 SC counting detector. 18F-fallypride (100 µl, ∼45 MBq of total activity) was dissolved in 45% acetonitrile and 55% aqueous solution of 1% trimethylamine. c,d, Comparison between the signals obtained from a commercial GABI* detector (c) and the the MAPbI3 detector (d). Insets: The same data (smoothed) with intensities plotted on a logarithmic scale to illustrate the impurity peaks.

Full size image

Spectroscopic applications of gamma detectors require fine energy resolution. For this purpose, a high electrical bias is usually applied to maximize the efficiency of charge carrier collection. However, the chemical lability of perovskites, and particularly their ionic dynamics (such as halide-ion migration, which is known to cause hysteresis and other instabilities in photovoltaic solar cells34), seem to be critical for the operation of SC gamma detectors under a high applied bias of 100–1,000 V. As illustrated in Supplementary Fig. 3, state-of-the-art commercial CZT crystals operate at 500–1,000 V and permit the elucidation of energy-resolved spectra of an 241Am source (59.6 KeV peak energy, source activity 0.4 MBq). The room-temperature voltage stability of MAPbI3, and of all of the other perovskite SCs tested herein, is limited to 20–30 V (lowest for MAPbI3, highest for MAPbBr3). Cooling of the crystals improves the voltage stability up to 100–140 V, yet leads to very limited success in resolving the energy of the γ photons, as exemplified in Supplementary Fig. 4. A different behaviour is observed for FAPbI3 SCs (2–3 mm). In this case, although the room-temperature voltage stability is only moderately improved with respect to MAPbI3 (30 versus 20 V), a much lower dark current and electronic noise are observed (compare Figs 2b and 4a), as well as a higher µτ product of 1.8 × 10−2 cm2 V−1. For an 241Am source, FAPbI3 allowed for count rates that are up to 100 times higher and the possibility of energy-resolved spectroscopy at a low bias voltage of 23 V (Fig. 4b) in comparison with MAPbI3 SCs of the same size. This represents, to the best of our knowledge, the first demonstration of gamma energy resolution capability with solution-grown SCs.

Figure 4: Energy resolution with 3 mm FAPbI3 perovskite SCs.
figure 4

a, Real-time signals from the charge-sensitive preamplifier and amplifier-shaper with 137Cs and 241Am γ sources (traces are shifted for clarity). b, Energy-resolved spectrum of 241Am recorded with a FAPbI3 SC.

Full size image

From a practical viewpoint, black α-FAPbI3 SCs have one important shortcoming with respect to MAPbI3: they exhibit a thermodynamic instability towards conversion into a wide-bandgap δ-phase within 24 h4,27. We found that the growth of such SCs from FA-rich solutions allowed for an extended stability of 1–2 weeks, sufficient for the presented testing. Future improvements in the phase stability of such SCs are expected via compositional engineering of the Goldschmidt tolerance factor, such as through the formation of (Cs/FA)PbI3 compounds, as recently demonstrated in thin-film studies35.

In summary, we have presented the first example of a room-temperature, direct-conversion gamma detector obtained from solution-grown SCs. Among standard applications, such low-cost detectors may be uniquely suited for personal devices (such as the miniature solid-state Geiger counters for radiation sensing that can be attached to most smartphones, from FT Lab http://allsmartlab.com/eng/Smart_Geiger.php). The internal sound card of such a device may be used as the read-out electronics, as demonstrated in this work. For SCs with a mixed anion composition that reduces the noise, we demonstrated the possibility of sufficient spatial resolution. With FAPbI3 SCs, which exhibit the highest μτ product and the lowest noise levels and dark currents of all of the tested materials, room-temperature energy resolution is demonstrated. Further work should be directed at the stabilization of the cubic phase of FA-based systems and the exploration of other compositions that are uniquely accessible from solution growth, such as those containing both organic and inorganic cations.

Methods

Growth of MAPbX3 SCs

All of the materials were handled in air. 5–12 mm MAPbX3 SCs were prepared by procedures described elsewhere using both aqueous (75–80 °C)5 and non-aqueous methods (95 °C)26. Methylammonium iodide (MAI) for the non-aqueous method was prepared as described elsewhere14 and washed with diethyl ether. For partial anion exchange, MAPbBr3 SCs were placed into a 0.17 M solution of MAI in isopropanol and held for 2.5–15 h at room temperature. Subsequently, the crystals were rinsed with isopropanol and dried in air at 80 °C for 30 min. All of the SCs were rinsed with acetone and stored in air. SCs were coated with a conductive silver paste on two opposing facets for the subsequent X- and γ-ray response measurements.

Growth of FAPbI3 SCs

The formamidinium iodide (FAI) precursor was prepared in a similar manner to that reported elsewhere36. Specifically, solid FA acetate (15 g, 99% Aldrich) was slowly dissolved in 50 ml hydroiodic acid (HI) (56–58 wt% abcr GmbH, containing <1.5% hypophosphorous acid as a stabilizer) forming a 2.5 M solution, then vacuum-dried at 55 °C, washed with diethyl ether, recrystallized from ethanol and vacuum-dried at room temperature overnight. To grow FAPbX3 SCs, we followed the inverse temperature crystallization methodology4,27 with slight modifications. A mixture of 0.814 mmol of FAI and 0.65 mmol of PbI2 (99% Aldrich) were dissolved in 1 ml of γ-butyrolactone (≥99% Aldrich), filtered (0.2 μm PTFE filter) and placed into a 4 ml vial and inserted into the oil bath (at 80 °C). Subsequently, the temperature was slowly raised to 110 °C. After 2–3 h, several single crystals formed, each 2–3 mm in size.

X-ray response measurements

Cu Kα radiation (from a Bruker D8 Powder X-ray diffractometer) was shaped into a 3 × 10 mm beam. A commercial X-ray scintillation counter (from STOE & Cie. GmbH) was used for the calibration of the X-ray intensity, and then replaced by the MAPbI3 SC biased at 10 V across the interelectrode distance of 4–5 mm by a Keithley 236 SMU. The chopped signal was recorded by a Stanford Research 850 lock-in amplifier.

γ-ray response measurements

For photoconductive measurements (for simply observing the change in the resistance on irradiation, for example) the detector was placed close to the source of radioactivity (10 cm from a source of 11C with ∼70 GBq activity and ∼0.5 cm from a source of 137Cs with ∼2.2 MBq activity). The activities of both sources were estimated by a VDC-505 dose calibrator (Comecer SpA). When monitoring the decay of 11C, the γ flux was chopped with a 10-mm-thick lead plate and the response was measured with a Stanford Research 850 lock-in amplifier. For the 137Cs portable source, the current through the detector was measured with a Keithley 236 SMU without chopping; instead the 137Cs source was manually placed close to and away from the crystal for ‘on’ and ‘off’ measurements, respectively.

For the γ-counting mode of operation, a MAPbI3 SC was inserted in the place of a CdZnTe SC in a commercial gamma detector (eV-Products, model B1758) and biased with 4–10 V. The detector was tested either using a portable 137Cs source with ∼2.2 MBq activity or wrapped with a plastic tube containing a flow of a radiotracer compound (18F-fallypride) solution. The signal from the detector, after passing through a low-pass filter and low-noise amplifier (351A, Analog Modules Inc.), was analysed by a computer with an Asus Xonar DSX sound card at a sampling rate of 144 kHz. The discrimination level was adjusted such that the ratio between the counting rate and background signal was optimized. To construct the 2-pixel detector (∼7 × 7 mm pixel size), a perovskite SC was interfaced with a signal electrode at the middle and two bias electrodes at the edges. The load resistor can then be excluded, as its function is replaced by one pixel (the one not being measured). The 2.2 MBq 137Cs source was scanned over the detector surface at a speed of 0.1 mm s–1. The signal was recorded by the PC sound card in .wav format and post-treated with positive and negative discrimination levels to separate the responses of the individual pixels.

A custom apparatus was used for energy-resolved measurements, connected to a A250CF CoolFET charge sensitive preamplifier (Ametek) coupled with an amplifier-shaper (Model 572, EG&G Ortec) and a digital multichannel analyser MCA-8000D (Ametek). The high bias voltage was applied through a Keithley 236 SMU that was also used for monitoring the current through the SC detector. An 241Am γ source with an activity of 0.4 MBq was used for recording the energy spectra. The MAPbX3 SCs were cooled down to ∼220 K by submerging into a dry-ice bath, whereas FAPbX3 SCs were measured at room temperature. Commercial CdZnTe crystals (4 × 4 × 3 mm, eV Products) were used for comparison.

HPLC experiment with 18F-fallypride

First H18F was produced at the cyclotron facility of the Center for Radiopharmaceutical Sciences at ETH Zurich by irradiation of H218O with 18 MeV protons. H18F was then immediately used as a reagent in the synthesis of 18F-fallypride33, in which fluoride replaces a tosylate group from the precursor molecule in a nucleophile substitution reaction using an automated synthesis apparatus (COSAB, Sweden). The final product was then isolated by semi-preparative HPLC (Agilent 1,100 Series Value System) with a C-18 column (Phenomenex Gemini C18 10 × 250 mm, 110 Å, 5 µm) using an isocratic elution with 45% 0.1 M NaHCO3 and 55% acetonitrile in a flow of 4 ml min–1. The product was eluted approximately 15 min after injection. For quality control, an additional HPLC run was conducted with a GABI* gamma-HPLC-flow-detector (Raytest GmbH) and with a C-18 column (Phenomenex Gemini-NX, 110 Å, 5 µm). The flow of the isocratic elution was 1 ml min–1, and a mixture of 45% acetonitrile and 55% of aqueous NEt3 (0.1%) was used. The product eluted after approximately 13 min. The exhaust tube from the HPLC was wrapped around a MAPbI3 SC gamma-counting detector for parallel measurement.