Letter
1038/nature24032
Printable organometallic perovskite enables
large-area, low-dose X-ray imaging
Yong Churl Kim1, Kwang Hee Kim1, Dae-Yong Son2, Dong-Nyuk Jeong2, Ja-Young Seo2, Yeong Suk Choi1, In Taek Han1,
Sang Yoon Lee1 & Nam-Gyu Park2
Medical X-ray imaging procedures require digital flat detectors
operating at low doses to reduce radiation health risks1,2. Solution-
processed organic–inorganic hybrid perovskites have characteristics
that make them good candidates for the photoconductive layer
of such sensitive detectors3–7. However, such detectors have
not yet been built on thin-film transistor arrays because it has
been difficult to prepare thick perovskite films (more than a few
hundred micrometres) over large areas (a detector is typically
50 centimetres by 50 centimetres). We report here an all-solution-
based (in contrast to conventional vacuum processing) synthetic
route to producing printable polycrystalline perovskites with
sharply faceted large grains having morphologies and optoelectronic
properties comparable to those of single crystals. High sensitivities
of up to 11 microcoulombs per air KERMA of milligray per square
centimetre (μC mGyair−1 cm−2) are achieved under irradiation
with a 100-kilovolt bremsstrahlung source, which are at least one
order of magnitude higher than the sensitivities achieved with
currently used amorphous selenium or thallium-doped cesium
iodide detectors. We demonstrate X-ray imaging in a conventional
thin-film transistor substrate by embedding an 830-micrometre-
thick perovskite film and an additional two interlayers of polymer/
perovskite composites to provide conformal interfaces between
perovskite films and electrodes that control dark currents and
temporal charge carrier transportation. Such an all-solution-based
perovskite detector could enable low-dose X-ray imaging, and could
also be used in photoconductive devices for radiation imaging,
sensing and energy harvesting.
The use of digital flat detectors in many diagnostic and interven-
tional medical applications, including in X-ray imaging procedures, has
increased rapidly over the past decades. This results in higher average
radiation doses for individuals. Repeated exposure to such ionizing
radiation can cause cancer. Recent reports of radiation detection by
organometallic perovskites (MAPbX3, where MA =​  CH3NH3 and
X =​ Cl, Br or I) that are triggered by solar cells8,9 suggest that such
materials may enable low-dose X-ray imaging.
At present, two flat detector technologies prevail. The dominant
method is based on indirect X-ray conversion using thallium-doped
cesium iodide (CsI:Tl) scintillators formed on amorphous ­silicon
(a-Si) photodetector arrays with a typical sensitivity10 of about
0.3 μ​C mGyair−1 cm−2. The other method uses direct photoconductors
such as amorphous selenium (a-Se), which is particularly suitable for
high-resolution mammography because of its good X-ray conversion
efficiency of soft X-ray photons (<​40 keV). These direct-conversion
detectors read out charges by an applied electric field, leading to almost
none of the optical crosstalk usually suffered by indirect-conversion
detectors. However, a-Se has limited use because of its poor absorptivity
of hard X-rays, caused by its low-lying K-edge resonance absorption,
an insufficient charge-carrier mobility lifetime (μτ) product value
(about 10−7 cm2 V−1), and the accompanying high operating elect­
rical fields (10−​40 V  μ​m−1) required to overcome the poor charge
1
Samsung Advanced Institute of Technology (SAIT), Samsung Electronics Materials Research Complex, Youngtong, Suwon 443-803, South Korea. 2School of Chemical Engineering and Department
of Energy Science, Sungkyunkwan University, Suwon 440-746, South Korea.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
doi:10.
collection efficiency. a-Se has field-dependent sensitivities of up to
0.44 μ​C mGyair−1 cm−2 at operating electric fields as high11 as 15 V μ​m−1.
We therefore aimed to produce highly sensitive all-solution-based
X-ray detectors based on printable polycrystalline MAPbI3 thick films.
The X-ray detector was built as illustrated in Fig. 1a and the ­individual
thin-film transistor (TFT) pixel is depicted in Fig. 1b (see Methods for
details). Two interlayers of polyimide (PI)-perovskite c­ omposites (see
Methods) were inserted by spin-casting to reduce the dark ­current
(that is, the background ‘leakage’ current, which produces noise in
the X-ray detector). The bottom layer of the PI-MAPbI3 composite
(Fig. 1b) forms a hole-transporting layer and the top PI-MAPbBr3
layer forms a hole-blocking layer (HBL). The high intrinsic resistivity
(1018 Ω cm) of PI also reduces electrical crosstalk among the pixels. The
830-μ​m-thick polycrystalline MAPbI3 (MPC) photoconductor layer
was printed on the coated PI-MAPbI3 using a doctor blade (Fig. 1b).
The top electrode (indium tin oxide, ITO) was biased to a positive
­voltage source to operate the detector in hole-collection mode. Finally,
tape-automated bonding (in which bare integrated circuits are placed
onto a printed circuit board by attaching them to fine conductors in a
polyamide or polyimide film) of readout integrated circuits (ROICs; see
Methods) was performed on the X-ray detectors. The MAPbI3 has not
only a stable three-dimensional perovskite crystal structure, but also
its mass attenuation coefficients (obtained from the NIST database12)
are comparable to (or exceed) that of CdZnTe throughout the medical
X-ray energy domain (10−​200 keV) thanks to the heavy I and Pb atoms
that exhibit resonance K-edge absorption at 33.17 keV and 88 keV,
respectively (Extended Data Fig. 1a). The X-ray-related measurements
and images were recorded using a 3-mm-thick Al-filtered 100-kV (peak
tube voltage, Vp) tungsten bremsstrahlung source with a dose rate range
of 1−​5  mGyair s−1, resulting in a dose range of 10−​200  μ​Gyair by varying
the irradiation pulse width. The measured pixel current (averaged over
a square centimetre of detector area) as a function of applied voltage
(10−​200 V) yielded a sensitivity range of 0.3−​3.8  μ​C mGyair−1 cm−2
(Fig. 1c). These sensitivities are about one-third those measured from a
small-sized diode configuration (1−​11  μ​C mGyair−1 cm−2) as described
below. The μτ product was fitted to 1.0 ×​  10−4 cm2 V−1 by the Hecht
formula13, Q =​  Q0μτVd−2[1 −​ exp(−​d2/μτV)], where V is the applied
bias, d the thickness, and Q0 the asymptotic collected charge. The cor-
responding electron–hole pair generation energy (W±) was e­ stimated
to be 12.5 eV at bias 200 V, while 4.7 eV was obtained at the same bias in
the diode measurement (Fig. 1c and Extended Data Fig. 1b), approach-
ing the theoretically predicted value 3Eg ≈​ 4.5  eV (where Eg is the
optical bandgap of MAPbI3, about 1.5 eV) by Klein’s rule14.
Meanwhile, the spatial resolution of the detector can be determined
with the modulation transfer function (MTF), representing the transfer
ability of input signal modulation of the spatial frequency (in-line pair
per millimetre; lp mm−1) relative to its output. Using the slanted-edge
method15, the MTF of the MPC detector was compared with a state-of-
the art 250-μ​m-thick a-Se detector with the same pitch (Extended Data
Fig. 1c). The MTF of the MPC detector dropped to 0.2 at 3.1 lp mm−1,
5 o c t o b e r 2 0 1 7 | V O L 5 5 0 | N A T U RE | 8 7
 RESEARCH Letter
a X-ray photons b
V+ PI-MAPbI3 on TFT
ITO
PI-MAPbBr3
MPC
PI-MAPbI3
TFT backplane
ITO
d Fitted to the Hecht equation
Signal current (×106 e mGy–1
Figure 1 | Building an X-ray detector based on hybrid perovskites.
a, Illustration of an all-solution-processed digital X-ray detector. b, The
left panel shows an optical image of spin-cast PI-MAPbI3 on an a-Si:H
TFT backplane. The inset shows a single-pixel structure of TFT (scale
bar 30 μ​m) in which the collection electrode (white outline) is connected
to the drain contact of the TFT through a via (circular pad). The right
whereas the MTF of the a-Se detector was 0.2 at 7.1 lp mm−1. The
enlarged resolution test image resolved between 2.8 lp mm−1 and
3.1 lp mm−1. A side-by-side comparison is not clear owing to differ-
ences in the photoconductor layer thickness, material packing and
morphology, applied field, and radiation energy between the two
detectors. However, the most probable cause of the MPC detector’s
lower MTF (than that of the a-Se detector) is the 20−​100  μ​m size of
the MPC crystallites: those crystallites larger than a pixel can overlap
neighbouring pixels, leading to charge sharing, which degrades MTF
substantially. The resolution of the demonstrated MPC detector could
be much improved by uniform material preparation and a greater
understanding of pixel-to-pixel charge collections and trap-assisted
charge behaviours in the MPC detectors. X-ray images of a hand
phantom and a smaller smartphone produced under low doses of
25 μ​Gyair and 10 μ​Gyair, respectively, are shown in Fig. 1d and Extended
Data Fig. 1d.
Though an ideal photoconductor would be a single crystal, such
crystals are difficult and expensive to assemble on large flat panels. In
photo­conductive devices such as solar cells, larger crystal grain size
has been predicted16 to increase the power-conversion efficiency, and
this has been proved experimentally in doctor-bladed MAPbI3 films17
and spin-cast thin MAPbI2Cl films18. In the preparation of viscous
MPC paste, an anti-solvent method was used (see Methods). We chose
γ​-butyrolactone as a primary polar solvent and α​-terpineol ­(relative
dielectric constant εr =​ 2.75 at room temperature) as our anti-­solvent.
The as-grown MPCs have an average grain size of 30 μ​m, and the
­largest grains of up to 100 μ​m are found to have crystal habits ­typical
of a body-centred tetragonal lattice exhibited as sharply ­faceted
­rhombo-hexagonal dodecahedra (Fig. 2a and b). Powder X-ray
­diffraction (Fig. 2c) reveals the well crystallized tetragonal (I4/mcm)
phase of the prepared MPCs, distinguishable from the cubic phase
by the presence of the (211) peak19,20. Additional t­ hermogravimetry
­a nalysis coupled with gas-phase Fourier-transform infrared
­spectrometry did not show any notable mass loss below 230 °C (Fig. 2d),
unlike other reports of MAPbI3, which typically show mass loss in
the temperature range of 80−​130 °C, attributed to dehydration of
8 8 | N A T U RE | V O L 5 5 0 | 5 o c t o b e r 2 0 1 7
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
TFT
Printed MPC I3
APb
PI-M
c
Measured; pixellated detector
4
air per pixel)
X-ray sensitivity (μC mGyair
1,200
µτ = 1.0 × 10−4 cm2 V−1
800
12.5 eV 2
W± (eV)
10
400
–1 cm–2)
1
4.7 eV
1 0 50 100 150 200
Bias voltage (V)
0 0
0 30 60 90 120 150 180 210
Bias voltage (V)
panel shows a photograph of printed MPCs on the PI-MAPbI3. c, Charge
collection and sensitivity characteristics of the MPC detector measured at
100 kVp. The inset shows W± in the pixellated (blue symbols) and diode
(red symbols) detectors. d, A hand phantom X-ray image obtained from an
MPC detector (using 100 kVp and 5 mGyair s−1 for 5 ms exposure, resulting
in a dose of 25 μ​Gyair and a bias voltage of 50 V).
perovskites or possible desorption of residual solvents21. Loss of the
CH3NH2 species followed by loss of the HI was observed between
300 °C and 370 °C, with corresponding infrared absorption peaks
related to CH3NH2. Loss of PbI2 started at around 410 °C by melting.
To characterize the optoelectronic properties, we fabricated MPC
photoconductor devices by printing on the ITO-coated glasses that
form the cathodes and contacting the top layer (on the illuminated
side) by sandwiching other ITO glasses to form anodes. The optical
­properties were compared with those of thin MAPbI3 films (0.5 μ​m in
­thickness) p ­ repared by an adduct method (hereafter referred to as the
adduct film)22. The MPC films showed an absorption band extended
to 840 nm, whereas the thin adduct films had an absorption edge at
760 nm (Fig. 3a). Such absorption extension is similar to the ­behaviour
reported for MAPbI3 single crystals23. The photoluminescence peak
of the MPCs red-shifted from 760 nm (adduct film) to 810 nm. We
note here that the photoluminescence intensity of the MPCs is about
100-fold stronger than that of the thin adduct films (Extended Data
Fig. 2a), which is thought to be due to a reduced number of non-
radiative surface defects in the MPC crystallites. The optical bandgap
estimated by the Tauc method was lowered to 1.50 eV from the 1.60 eV
of the thin adduct film (Fig. 3b). This bandgap reduction is strongly
related to the fewer grain boundaries and corresponding lower locali­
zed defects in our large-grain MPCs. Carrier lifetime τ was estimated
by using time-resolved photoluminescence. We compared the transient
fluorescence from the photoexcitation peaks at 810 nm (MPC) and
760 nm (adduct film) (Fig. 3c). Biexponential fitting (Extended Data
Fig. 2b) revealed both fast and slow dynamics, with slow component
lifetime values of τ2 =​ 1,052 ns for the MPC and τ2 =​ 128 ns for the thin
adduct film. Meanwhile, impedance spectroscopy24 gave a much longer
carrier recombination lifetime of 6−​300  μ​s measured under 0.1 Sun
and 1 Sun (where 1 Sun =​ 100 mW cm−2) illumination conditions
(Extended Data Fig. 2d). The large measured difference in the lifetime
results is also found in other reports4,16. We chose the smaller τ of time-
resolved photoluminescence result (1,052 ns) as the carrier lifetime. The
charge-carrier mobility was measured by a time-of-flight method using
400-μ​m-thick MPC films (Fig. 3d and Extended Data Fig. 2c). The
 Letter RESEARCH

a b

c d
Tetragonal I4/mcm 100 420 °C
(400)

350 °C

Absorbance (arbitrary units)

Derivative of weight (% ºC–1)
0.0
Counts (arbitrary units)

345 °C
80 C–N stretch C–H stretch
for amine (1,143–1,214) (2,965)
Weight (%)
–0.5 335 °C
(200)

N–H bending N–H stretch

60 (3,288)
(211)

(1,600)
–1.0 330 °C
C–N stretch
40 (900–1,200)
–1.5 320 °C
300 °C
150 °C
10 20 30 40 50 60 70 80 90 20
0 100 200 300 400 500 1,000 1,500 2,000 2,500 3,000 3,500
2θ (deg) Temperature (°C) Wavenumber (cm–1)
Figure 2 | MPC perovskite basic properties. a, A scanning electron prepared MPC film (right; scale bar 30 μ​m). c, X-ray diffraction pattern of
microscope (SEM) image of a 400-μ​m-thick printed MPC film using the powdered grains from an MPC film. d, Analysis of thermal behaviour
a doctor blade. Scale bar, 400 μ​m. b, SEM images showing surface of MPC films using thermogravimetry analysis coupled with gas-phase
morphology(left; scale bar 50 μ​m) and a polygon of crystallites of the Fourier-transform infrared spectrometry.

hole mobility was 139 ±​ 14 cm2 V−1 s−1, and the electron mobility was
66 ±​  7 cm2 V−1 s−1. The μτ products of the MPC films estimated for the
demonstrated X-ray detector (1.0 ×​  10−4 cm2 V−1) agrees reasonably
well with the μτ product (1.46 ×​  10−4 cm2 V−1) obtained independently
from the time-of-flight and time-resolved photoluminescence. Though
the μ​τ product measured here is two orders of magnitude lower than
for those record-high single crystals of perovskite4,5, it is comparable
to that of HgI2 single crystals and better than the vapour-deposited
polycrystalline CdZnTe in the range25 10−5 cm2 V−1.
a b
1.2
To make use of MPC film for medical imaging devices, the dark
c­ urrent must be tightly controlled. A high dark current limits the
dynamic range of the detector as well as increases noise levels, thus
degrading the signal-to-noise ratio. The MPC detectors should be
­operated in the hole-collection mode because of the longer carrier
lifetime and hole mobility. In addition to requiring the HBL on the
anode side, a hole-transporting layer on the cathode side providing
good adhesion between the MPC film and the TFT substrate (back-
plane) is required. Using TiO2 for the HBL and Spiro-OMeTAD for the
8
1.2 12
Normalized photoluminescence

MPC Adduct MAPbI3

Normalized absorbance

6
0.8 0.8 9
(arbitrary units)

(arbitrary units)

(αhυ)2 (× 1010)
(αhυ)2 (×107)

4 6
0.4 0.4
Eg = 1.50 eV Eg = 1.60 eV
2 3
0.0 MPC 0.0
Adduct MAPbI3
400 500 600 700 800 900 0 0
1.4 1.5 1.6 1.7 1.4 1.5 1.6 1.7
Wavelength (nm) hυ (eV) hυ (eV)
c d
100 0.9 0.6
Normalized photoluminescence

MPC 0.8
0.7 0.5
Adduct MAPbI3
intensity (arbitrary units)

0.6 0.4
Normalized current

Normalized current

10–1
(arbitrary units)

(arbitrary units)

0.5
0.4 h+ 0.3
10–2 time of flight e–
0.3 30 V time of flight
0.2
40 V –30 V
10–3 50 V –40 V
0.2 60 V –50 V
70 V –60 V
80 V –80 V
10–4 0.1 0.1
0.0 0.5 1.0 1.5 0.2 0.3 0.4 0.5 0.2 0.4 0.6 0.8 1.0
Time (μs) Time (μs) Time (μs)
Figure 3 | Optoelectronic characterization of MPC thick films. frequency. c, Photoluminescence lifetime of the MPC and adduct films.
a, Normalized photoluminescence and absorption spectra of the MPC d, Normalized time-of-flight h+ (hole carrier) and e− (electron carrier)
and thin adduct MAPbI3 films. b, Optical bandgap (Eg) estimation from charge transient curves for an MPC film under CO2 laser illumination
absorption spectra of the MPC and thin adduct MAPbI3 films, where (wavelength 337.1 nm) with various bias voltages.
α​is the absorption coefficient, h is the Planck constant, and υ is the light

5 o c t o b e r 2 0 1 7 | V O L 5 5 0 | N A T U RE | 8 9
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 RESEARCH Letter

a b
ITO / MPC / ITO
102 ITO / TiO2 / MPC / Spiro-MeOTAD / ITO PI-MAPbI3
ITO / MAPbBr3 (adduct) / MPC / PI-MAPbI3 / ITO
Current density (mA cm–2) ITO / PI-MAPbBr3 / MPC / PI-MAPbI3 / ITO
100 ITO / PI-MAPbBr3 (dilute) / MPC / PI-MAPbI3 / ITO

10–2

10–4
MPC d = 830 μm
PI-MAPbBr3
10–6

0 50 100 150 200

Bias voltage (V)
c
200 V
100 kV
10–2 150 V d
1 mGyair s–1, 0.2 s
125 V
X-ray response (mA cm–2)

Device d = 830 μm
75 V
Bias 50 V (0.06 V μm–1)

X-ray response (mA cm–2)

Device 50 V 10–2
10–3 d = 830 μm S1 S10 X-ray 50 ms shuttered
25 V
100 kV, 1 mGyair s–1
10 V

10–3
10–4
Idark (t)
Idark (0)

10–4
10–5

0.0 0.2 0.4 0.6 0 1 2

Time (s) Time (s)
Figure 4 | Diode MPC device characterization. a, Dark-current plots MPC device response to X-ray pulses with varying applied bias voltage.
as a function of applied bias voltage with various interlayers, where d is d, A response under 50 V to a train of short X-ray pulses with a pulse
MPC thickness. b, Typical SEM images of PI-MAPbI3 and PI-MAPbBr3 width 50 ms.
composite films used for interlayers (scale bars, 5 μ​m and 400 μ​m). c, An

hole-transporting layer, as widely used in perovskite solar cells, reduced falling (reaching 10% of the saturated X-ray signal) times both
the dark current by about one order of magnitude (Fig. 4a). decreased as the bias increased between 10 V and 200 V (Extended Data
We developed improved composites of PI and perovskites. These Fig. 4a). This relatively fast response of our device structure is clearly
composites are motivated by the fact that (i) PI films have been reported shown in a train of short X-ray pulses with a pulse width of 50 ms
as effective HBLs in a-Se detectors26, and PI precursors are usually formed using a shuttered X-ray irradiation (Fig. 4d), which indicates
polymerized in polar solvents such as N-methyl-2-pyrrolidone (NMP) relatively lower trap states in our electrode/interlayer/MPC interfaces
and γ​-butyrolactone, which are compatible with perovskite precursors and that any bulk traps existing in the MPC have relatively low detrap-
(Extended Data Fig. 3a); and (ii) other hybrid perovskites having larger ping energy.
bandgaps than MAPbI3, such as MAPbBr3 or MAPbCl3, can be used The temporal performance of the direct-converting X-ray detectors
as HBLs, as confirmed by dark-current measurements using MAPbBr3 can be characterized by ‘lag’ and ‘ghosting’26,28. As analysed using the
synthesized by the adduct method (Fig. 4a)22 as an HBL. Both spin-cast repeated pulse train measurement on the diode structure (Extended
interlayers—the PI-MAPbBr3 composite as HBL and the PI-MAPbI3 Data Fig. 4b and c), the ‘lag’ showed almost the same behaviour for
composite as the hole-transporting layer—reduced dark current to each exposure. Charge trapping in the bulk of the MPC film can
one order of magnitude less than the dark current of the device sand- ­contribute to ‘lag’ by releasing trapped charges after exposures, which
wiched by TiO2/Spiro-OMeTAD. The PI-MAPbBr3 composite further take longer to be released when the trapping energy is larger (deep
reduced the dark current by controlling the concentration of MAPbBr3 traps). Another contribution to ‘lag’ is charge injection through contact
­precursors in the composite system (Fig. 4a and Extended Data electrodes being under a constant bias potential, in which the charge
Fig. 3b). The coated PI-MAPbI3 formed uniform composite films traps generated by X-ray illumination increase the electric field at the
about 2.5 μ​m thick with MAPbI3 polycrystalline grains underneath electrode interfaces, causing a dark current injection that leads to a ‘lag’
the p ­ olymer-like layer (Fig. 4b), which adhered well to the MPC increase. ‘Ghosting’ is represented as a change in sensitivity with each
films as a result of partial dissolution of the PI overlay by the solvent X-ray exposure by the generated charge traps. No apparent sensitivity
(γ​-­butyrolactone) incorporated in the printed MPC. The PI-MAPbBr3 decrease trend is observed with increase in the number of exposures.
composites showed an interesting morphology (Fig. 4b and Extended We consider that in our devices, with relatively high dark current (about
Data Fig. 3a and b) having orange-coloured fourfold symmetric two orders higher than in stabilized a-Se devices), these unusual lag
­dendrite structures. Such dendritic figures are usually found in metal and ghosting behaviours may be attributable to the extremely short
alloys27 or snowflakes. X-ray pulses were illuminated on the diode-­ charge trap release time as a result of the shallowness of the charge
configured devices for the measurement of the X-ray photoresponse vacancy traps in the MAPbI3 crystallites29 or injected hole carriers
(Fig. 4c). Electrical fields in the range 0.01−​0.24 V  μ​m−1 were through blocking contacts effectively recombining with the charge
applied to collect charges through the bottom ITO electrode. X-ray traps and leading to invariant sensitivity behaviours. Our demonstrated
­sensitivities in the range 1−​11  μ​C mGyair−1 cm−2 were observed (versus device protocol could be applied to low-dose inexpensive detectors in
0.3−​3.8  μ​C mGyair−1 cm−2 in the pixelated detector) as discussed in the full medical X-ray range. The remaining technological challenges
Fig. 1. The rising (reaching 90% of the saturated X-ray signal) and are ­further lowering the dark current, and improving the relatively slow

9 0 | N A T U RE | V O L 5 5 0 | 5 o c t o b e r 2 0 1 7
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

decay time to be implemented in computer tomography applications
under a small bias voltage. Both depend strongly on the nature of the
interface traps between the electrodes and MPCs, and the grain bound-
aries of MPCs. Thus, both challenges can be addressed through opti-
mizing the interlayers (by fine-tuning the work function), and through
engineering the grain boundary in the photoconductor layer (by fusing
grains with sintering or by solvent annealing to promote further grain
growth, and by surface/grain-boundary passivation during or after the
coating process).
In summary, a disruptive polycrystalline CH3NH3PbI3 photocon-
ductor layer is printed on a TFT backplane for use in X-ray detectors.
A large charge-carrier mobility-lifetime product value 1.0 ×​  10−4 cm2 V−1
and high absorptivity over the full medical X-ray energy range com-
bine to produce a good X-ray sensitivity of up to 11 μ​C mGyair−1 cm−2.
The low-temperature prepared thick MPC films and their hard X-ray
imaging capability can easily be integrated into established digital X-ray
detectors with low-cost, large-area and low-dose benefits. Further
improvements in dark current, spatial resolution and W± are achievable
by fine-tuning charge-injection interfaces (to reduce dark current),
understanding charge sharing between MPC grains (to improve spatial
resolution), and modifying electric field and charge distribution at the
pixel gaps30 as proposed in a-Se detectors (to reduce W±).
Online Content Methods, along with any additional Extended Data display items and
Source Data, are available in the online version of the paper; references unique to
these sections appear only in the online paper.
received 17 April; accepted 4 August 2017.
1. Brenner, D. J. & Hall, E. J. Computed tomography: an increasing source of
radiation exposure. N. Engl. J. Med. 357, 2277–2284 (2007).
2. Sinnott, B., Ron, E. & Schneider, A. B. Exposing the thyroid to radiation: a
review of its current extent, risks, and implications. Endocr. Rev. 31, 756–773
(2010).
3. Yakunin, S. et al. Detection of X-ray photons by solution-processed lead halide
perovskites. Nature Photon. 9, 444–449 (2015).
4. Wei, H. et al. Sensitive X-ray detectors made of methylammonium lead
tribromide perovskite single crystals. Nat. Photon. 10, 333–339 (2016).
5. Yakunin, S. et al. Detection of gamma photons using solution-grown single
crystals of hybrid lead halide perovskites. Nat. Photon. 10, 585–589 (2016).
6. Wei, W. et al. Monolithic integration of hybrid perovskite single crystals with
heterogenous substrate for highly sensitive X-ray imaging. Nat. Photon. 11,
315–321 (2017).
7. Shrestha, S. et al. High-performance direct conversion X-ray detectors based
on sintered hybrid lead triiodide perovskite wafers. Nat. Photon. 11, 436–440
(2017).
8. Kojima, A. et al. Organometal halide perovskites as visible-light sensitizers for
photovoltaic cells. J. Am. Chem. Soc. 131, 6050–6051 (2009).
9. Lee, M. et al. Efficient hybrid solar cells based on meso-superstructured
organometal halide perovskites. Science 338, 643–647 (2012).
10. Zentai, G. et al. Detailed imager evaluation and unique applications of a new 20
cm ×​25 cm size mercuric iodide thick film X-ray detector. Proc. SPIE 5047,
84–95 (2003).
11. Choquette, M. et al. Direct selenium X-ray detector for fluoroscopy, R&F, and
radiography. Proc. SPIE 3977, 128–136 (2000).
12. Berger, M. J. XCOM Photon Cross Sections Database https://www.nist.gov/pml/
xcom-photon-cross-sections-database (NIST Phys. Lab., 1998).
13. Kasap, S. O. X-ray sensitivity of photoconductors: application to stabilized a-Se.
J. Phys. D 33, 2853–2865 (2000).
14. Que, W. & Rowlands, J. A. X-ray photogeneration in amorphous selenium:
geminate versus columnar recombination. Phys. Rev. B 51, 10500–10507
(1995).
15. Samei, E., Flynn, M. J. & Reimann, D. A. A method for measuring the
presampled MTF of digital radiographic systems using an edge test device.
Med. Phys. 25, 102–113 (1998).
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Letter RESEARCH
16. Huang, J., Shao, Y. & Dong, Q. Organometal trihalide perovskite single crystals:
a next wave of materials for 25% efficiency photovoltaics and applications
beyond? J. Phys. Chem. Lett. 6, 3218–3227 (2015).
17. Deng, Y. et al. Scalable fabrication of efficient organoleadtrihalide perovskite
solar cells with doctor-bladed active layers. Energy Environ. Sci. 8, 1544–1550
(2015).
18. Nie, W. et al. High-efficiency solution-processed perovskite solar cells with
millimeter-scale grains. Science 347, 522–525 (2015).
19. Baikie, T. et al. Synthesis and crystal chemistry of the hybrid perovskite
(CH3NH3)PbI3 for solid-state sensitised solar cell applications. J. Mater. Chem. A
1, 5628–5641 (2013).
20. Lian, Z. et al. High-performance planar-type photodetector on (100) facet of
MAPbI3 single crystal. Sci. Rep. 5, 16563–16572 (2015).
21. Leguy, A. M. A. et al. Reversible hydration of CH3NH3PbI3 in films, single
crystals, and solar cells. Chem. Mater. 27, 3397–3407 (2015).
22. Ahn, N. et al. Highly reproducible perovskite solar cells with average efficiency
of 18.3% and best efficiency of 19.7% fabricated via Lewis base adduct of
lead(II) iodide. J. Am. Chem. Soc. 137, 8696–8699 (2015).
23. Shi, D. et al. Low trap-state density and long carrier diffusion in organolead
trihalide perovskite single crystals. Science 347, 519–522 (2015).
24. Wang, H., Huang, J., Xing, S. & Yu, J. Improved mobility and lifetime of carrier
for highly efficient ternary polymer solar cells based on TIPS-pentacene in
PTB7: PC71BM. Org. Electron. 28, 11–19 (2016).
25. Del Sordo, S. et al. Progress in the development of CdTe and CdZnTe
semiconductor radiation detectors for astrophysical and medical applications.
Sensors 9, 3491–3526 (2009).
26. Abbaszadeh, S. et al. Enhanced detection efficiency of direct conversion X-ray
detector using polyimide as hole-blocking layer. Sci. Rep. 3, 3360–3366
(2013).
27. Galenko, P. K. et al. Dendritic solidification in undercooled Ni–Zr–Al melts:
experiments and modeling. Acta Mater. 57, 6166–6175 (2009).
28. Zhao, W., DeCrescenzo, G. & Rowlands, J. A. Investigation of lag and ghosting
in amorphous selenium flat-panel X-ray detectors. Proc. SPIE 4682, 9–20
(2002).
29. Buin, A. et al. Materials processing routes to trap-free halide perovskites.
Nano Lett. 14, 6281–6286 (2014).
30. Pang, G., Zhao, W. & Rowlands, J. A. Digital radiology using active matrix
readout of amorphous selenium: geometrical and effective fill factors.
Med. Phys. 25, 1636–1646 (1998).
Acknowledgements N.-G.P. acknowledges partial financial support from
the National Research Foundation of Korea (NRF) grants funded by the
Ministry of Science, ICT Future Planning (MSIP) of Korea under contracts
NRF-2012M3A6A7054861 and NRF-2014M3A6A7060583 (Global
Frontier R&D Program on Center for Multiscale Energy System) and NRF-
2016M3D1A1027664 (Future Materials Discovery Program). Y.C.K., I.T.H. and
S.Y.L. are indebted to all the members of SAIT (especially H. Kim and Y. Kim)
and the Healthcare & Medical Equipment Division of Samsung Electronics
(S. M. Yoon) for their help with the material characterization (SEM, X-ray
diffraction, time-resolved photoluminescence and time-of-flight spectroscopy),
X-ray detector fabrication, tape-automated bonding ROIC for the detector
driving for its characterization, imaging processes and interpretation.
Author Contributions I.T.H. initialized this project. Y.C.K. and N.-G.P. organized
the collaboration. Y.S.C. and S.Y.L. planned and performed thermal and
optoelectronic analysis. Y.C.K. and K.H.K. performed the synthesis of the MCP
and PI-composites, and their analysis. D.-Y.S and D.-N.J. performed adduct film
synthesis and analysis. D.-Y.S. and J.-Y.S. performed the impedance analysis.
Y.C.K. and K.H.K. performed X-ray analysis, detector fabrication and imaging.
Y.C.K., I.T.H. and N.G.P. wrote the manuscript. All authors discussed the results
and commented on the manuscript.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare no competing financial
interests. Readers are welcome to comment on the online version of the paper.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations. Correspondence and
requests for materials should be addressed to I.T.H. (intaek.han@samsung.
com) or N.-G.P. (npark@skku.edu).
Reviewer Information Nature thanks J. A. Rowlands and the other anonymous
reviewer(s) for their contribution to the peer review of this work.
5 o c t o b e r 2 0 1 7 | V O L 5 5 0 | N A T U RE | 9 1
 RESEARCH Letter
Methods
Growth of MPC. We used a two-step solvothermal process with addition of
α​-terpineol to grow MPCs, and the resulting mixture is dense and viscous, readily
printed using doctor blades. The thickness of the coated MPC films was controlled
by the guide frames. Equivalent molar ratios of CH3NH3I (prepared as in ref. 22)
and PbI2 (99%, Sigma-Aldrich) made up to be 78 wt% in total were mixed with
22 wt% of γ​-butyrolactone (≥​99%, Sigma-Aldrich) in a closed bottle. This mixture
was put in the 90 °C convection oven for 4−​6 h with periodic vibration every half
hour. Then 30−​40 wt% of α​-terpineol with respect to γ​-butyrolactone was mixed
and put in the same 90 °C oven for 4−​6 h with periodic vibration every half-hour.
The final viscous solution was printed on substrates using a doctor blade. The
coated substrates were cured on a hot plate at 120 °C for 1.5 h. The blend solution
kept under ambient temperature was vibrated and then warmed to 80 °C before
the printing process owing to its large viscosity increase during ambient storage.
Adduct MAPbI3 and MAPbBr3. Adduct MAPbI3 films were prepared by the
method described in ref. 22. A solution of molar ratio 1:1:6.7:2 of CH3NH3Br was
prepared in-house by mixing PbBr2 (≥​98%, Sigma-Aldrich), dimethyl formamide
(DMF) and dimethyl sulfoxide (DMSO). Stirring this mixture for 1 h resulted in the
adduct solution. The final transparent solution was spin-cast for 30 s at 2,000 r.p.m.
with diethyl ether sprayed over it.
PI-MAPbI3 composite. Polyimide (PI) precursors p-phenylenediamine and
biphenyl tetracarboxylic acid dianhydride (BPDA) with a molar ratio of 1:0.975
in total 15 wt% was mixed with 85 wt% of NMP, then the solution was stirred for
5 h in a N2 purged reactor at room temperature, and then the reactor temperature
was increased to 40 °C for additional 12-h reaction. The resultant poly(pyrom-
ellitic dianhydride-co-4,4′​-oxydianiline) (which is called polyamic acid) formed
the PI-NMP solution. 4.61 g of PbI2 and 1.59 g of CH3NH3I were added to the
prepared PI-NMP (4.8 g) with extra addition of 2.4 g DMF (other solvents such
as NMP or γ​-butyrolactone are also acceptable). After overnight stirring of this
mixture at room temperature, the PI-MAPbI3 solution was spin-cast at 2,000 r.p.m.
for 30 s, and then cured for 10−​30 min on a 120 °C hot plate.
PI-MAPbBr3 composite. We used the same process for the synthesis of
PI-MAPbI3. For example, PbBr2 [3.835 g (for dilute)−​7.670 g (for concentrated)]
+​ CH3NH3Br [1.170 g (for dilute)−​2.340 g (for concentrated)] +​PI-NMP (9.6 g)
+​DMF (4.8 g).
a-Si:H TFT backplane. The TFT backplane (active area 25.088 cm ×​ 28.672 cm)
was specifically customized for commercial a-Se detectors supplied from CMO
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Japan, Inc. We used a part of the active layer for the coating of PI-MAPbI3
and MPC photoconductors. The imaging area consists of 1,428 ×​  1,428 pixels
(10 cm ×​ 10 cm) with a pixel pitch of 70 μ​m. The individual pixel capacitance is
0.4 pF with a geometric fill factor of 76%. The geometric fill factor is defined as
the fraction of effective pixel area occupied by the charge collection electrode. The
MAPbI3 may react with metals (for example, aluminium, which is commonly used
in TFT backplanes), resulting in defective pixels and lines forming over hours to
days after manufacture. Therefore, the metal lines (such as the address and data
lines) were protected by a passivation overcoat on which ITO (which is inert to
MAPbI3) collection electrodes were formed. The ITO electrodes are electrically
connected by holes to the drain contacts of the TFT pixels. The TFTs are protected
by electrostatic shield structures to prevent permanent damage by the applied bias
on the photoconductors.
ROIC. The ROICs are supplied from Analogue Devices Inc. The used ROICs
contains a 256-channel/16-bit analogue front end with low noise of 560e (where
e is the electron charge).
X-ray characterization and imaging in the pixellated detector. The dark current
for the prototype MPC detector presented here reached 0.4 ±​ 0.1 pA per pixel
(areal density 80 ±​  20 pA mm−2), which is up to two orders of magnitude higher
than state-of-the-art direct (a-Se) and indirect (CsI:Tl) detectors. Dark-current
subtraction was applied for the image acquisition.
The pixel X-ray signal was saturated in the exposure range 50−​55  μ​Gyair at
50 V bias. Up to 70% of the saturated pixel value, the detector response showed a
good linearity in about 1% deviation. With exposure increase, the deviation from
linearity was exaggerated until the pixel was saturated. The MTF presented here
was measured under 15 μ​Gyair of exposure (1.5 mGyair s−1 for 10 ms), which is
within a good linear response range.
Pixel-to-pixel signal non-uniformity in the prototype MPC detector presented
here was in the range of 5% (high exposure; 20 μ​Gyair) to 12% (low exposure;
1 μ​Gyair) with respect to exposure levels. This level of non-uniformity is larger than
can be accounted for by the variation of MAPbI3 thickness (under 5%) across the
full prepared sample (under 1% in the local region of interest, about 1 cm2). Other
factors such as the interlayer–photoconductor contact and the existence of small
irregular voids in the MPC film can also contribute to this non-uniformity. After
gain correction, the non-uniformity of the pixel response was reduced to 3%−​4%.
Data availability. The data sets analysed during the study are available from the
corresponding authors upon request.

Extended Data Figure 1 | MPC X-ray detector. a, Calculated mass
attenuation coefficient for a-Se, CdZnTe and MAPbI3, and two normalized
bremsstrahlung X-ray spectra used in medical applications as a function of
X-ray energy. The linear attenuation coefficient (in cm−1) can be obtained
by multiplying mass attenuation and mass density (a.u., arbitrary units).
b, An X-ray spectrum of a 3-mm-filtered 100-kVp tungsten source in
1 mGyair (black line), the absorption by 830-μ​m-thick MPC film (blue
line), and the resulting absorbed spectrum by the MPC film (red line).
The integral of the absorbed spectrum equals the total deposited energy
(Eabs) in the MPC film of about 2.91 ×​  1011 keV cm−2 mGyair−1. The W± is
defined as EabsS−1, where S is the X-ray sensitivity. There are various loss
mechanisms that reduce the sensitivity in the pixelated detector. Provided
that all the specific loss factors are understood and corrected by further
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Letter RESEARCH
study, the panel sensitivities should coincide with those measured from
simple diodes. At this stage this discrepancy is probably mainly due to
the incomplete charge collection in the gap between pixels by the limited
geometric fill factor‚ imperfect MPC contact on the pixel electrode, and a
possible reduction in the effective applied bias on the MPC layer caused
by the thicker interlayers formed in the processing of large-area detectors
(rather than smaller diodes). c, The MTF of an MPC detector and a
conventional a-Se direct converting detector with a same pixel pitch of
70 μ​m. The inset shows a magnified 100-kVp X-ray image of a resolution
phantom. The MTF presented here was measured under 15 μ​Gyair of dose
(dose rate of 1.5 mGyair s−1 for 10 ms exposure time). d, A smartphone
X-ray image at dose rate 1 mGyair s−1 with exposure time 10 ms (10 μ​Gyair
of dose).
 RESEARCH Letter

Extended Data Figure 2 | Optoelectronic properties of MPC film.
a, Photoluminescence spectra of the MPC and the thin adduct
MAPbI3 film using an excitation source wavelength of 505 nm.
b, Photoluminescence lifetime of the MPC film. The fitted recombination
lifetime is 1.052 μ​s. c, Full time-of-flight current behaviour for both
charge carriers: left for hole and right for electron. d, Nyquist plots of
impedance (Z) spectra for the 400-μ​m-thick MPC film measured under
1 Sun and 0.1 Sun irradiation conditions. The inset shows the equivalent
circuit for analysis, where Rs is the series resistance of the MPC, Rrec is the
recombination resistance related to the recombination current, and CPE
is the chemical capacitance represented as a constant phase element. Here,
the lifetime is estimated by multiplying Rrec and CPE.

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Letter RESEARCH

Extended Data Figure 3 | Interlayers of PI–perovskite composites. a, Additional images of interlayer films and their precursor solutions. Scale bar
for the upper middle image, 30 μ​m. b, Morphology changes in PI-MAPbBr3 film with decreasing precursor concentration of MAPbBr3. All scale bars,
200 μ​m.

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 RESEARCH Letter

Extended Data Figure 4 | Temporal X-ray response characterization.
a, Rise and fall time response dependent on the applied bias. b, ‘Lag’
measurement as a function of time after each X-ray pulse signal.
c, ‘Ghosting’ measured using the sensitivity change for each X-ray pulse
signal. Lag can be defined as [Idark(t) – Idark(0)]Sn−1, and ghosting as SnS1−1,
where Idark(t) is the dark current measured at time t after each exposure,
Idark(0) is the initial dark current level before any exposures, S1 is the
X-ray sensitivity for the first exposure, and Sn is the sensitivity of the nth
exposure, respectively.

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.