High sensitivity organic inorganic hybrid X-ray detectors with direct transduction and broadband response

X-ray detectors are critical to healthcare diagnostics, cancer therapy and homeland security, with many potential uses limited by system cost and/or detector dimensions. Current X-ray detector sensitivities are limited by the bulk X-ray attenuation of the materials and consequently necessitate thick crystals (~1 mm–1 cm), resulting in rigid structures, high operational voltages and high cost. Here we present a disruptive, flexible, low cost, broadband, and high sensitivity direct X-ray transduction technology produced by embedding high atomic number bismuth oxide nanoparticles in an organic bulk heterojunction. These hybrid detectors demonstrate sensitivities of 1712 µC mGy−1 cm−3 for “soft” X-rays and ~30 and 58 µC mGy−1 cm−3 under 6 and 15 MV “hard” X-rays generated from a medical linear accelerator; strongly competing with the current solid state detectors, all achieved at low bias voltages (−10 V) and low power, enabling detector operation powered by coin cell batteries. X-ray detectors based on low-cost organic semiconductors have inherently low detector sensitivity due to poor X-ray to charge conversion and charge collection. Here, the authors demonstrate a flexible, high-sensitivity X-ray detector based on an organic bulk heterojunction-nanoparticle composite.

X -rays are widely used in homeland security, therapeutic and diagnostic healthcare and industrial process control (e.g. pharmaceuticals) with each application necessitating specific detector requirements 1 . For example, direct conversion detectors based on materials such as amorphous selenium 2 are currently used in mammography, but are limited by their low Xray attenuation for energies higher than 50 keV. Detectors based on p-type silicon with its high radiation-damage resistance are used in radiotherapy for dose measurement or beam imaging 3 . However, their propensity to damage from accumulated dose and drift due to environmental effects makes these less useful for beam calibration 4 . High-quality single crystal Cd(Zn)Te 5 is used for homeland security screening, but suffers from being limited to small dimensions, high cost, charge carrier trapping and highvoltage operation (>500 V). Similarly, X-ray detection in the nondestructive evaluation sector is currently dominated by CsI (Tl) scintillator screens coupled to a-Si which, despite their high stopping power and spatial resolution, are limited to sizes less than 60 × 60 cm 2 . Therefore, there is a demand for broadband, high sensitivity, low-cost radiation detectors, which current inorganic detectors fail to fulfil.
Organic semiconductors can be fabricated over large areas in a flexible format, enabling conformability to complex structures at low cost and are now commercialized for photovoltaics, displays etc 6 . There is increasing attention given to organic photodetectors for X-ray detection 7,8 . This often involves the coupling of scintillator screens with organic photodiodes 9 , insertion of highatomic number (Z) nanoparticles (NPs) 10 , quantum dots 11 or scintillator particles into organic diodes 12 , or the use of thin film organic semiconductors 8 or crystals 13 . Of these, the use of X-ray scintillators is often preferred as this enables the already mature organic photodetector technologies to be adapted for X-ray detection. However, the absorption of light by the organic semiconductor forms bound electron-hole pairs (excitons), which need to be dissociated 12 resulting in significant losses, limiting detector sensitivity as opposed to a direct conversion process.
Here, we introduce a broadband, direct, X-ray detector concept based on a thin film, hybrid semiconductor diode consisting of an organic bulk heterojunction (BHJ)-bismuth oxide (Bi 2 O 3 ) NP composite. These direct X-ray detectors demonstrate high sensitivities of 1712 µC mGy −1 cm −3 under 50 kV soft X-rays and~30 and 58 µC mGy −1 cm −3 under 6 and 15 MV hard X-rays. Furthermore, we also demonstrate a flexible detector based on the same device concept which offer a high sensitivity of 280 µC mGy −1 cm −3 . More importantly, these sensitivities are achieved at −10 V.

Results
X-ray response of BHJ-NP detectors. The device compromises of a diode architecture where the BHJ-NP composite is sandwiched between indium tin oxide (ITO) and aluminium (Al) electrodes (Fig. 1a). Here, the introduction of the Bi 2 O 3 (Z = 83 for Bi) is utilised to increase the X-ray attenuation 14 . We have chosen Bi 2 O 3 from the many metal oxides available based on its direct conversion of X-rays and lower environmental impact and health risks when compared to, for example, high Z Pb-based semiconductors. Given its existing use as a non-toxic dental material such as in the case of hydraulic silicate cements 15 with an opacity to X-rays makes it an ideal candidate for our application.  c An X-ray imager based on the hybrid X-ray detector and 70 kV X-ray image of a bolt taken using the X-ray imager Regioregular poly(3-hexylthiophene-2,5-diyl) (P3HT) and [6,6]-Phenyl C 71 butyric acid methyl ester (PC 70 BM) were selected as the BHJ system. The formation of nanoscale diodes throughout the volume of the BHJ, in close proximity to the NPs leads to an in-built depletion region, with local electric fields as high as~200 V µm −116 , which has been experimentally quantified with Fourier-transform IR-absorption spectroscopy for the P3HT:PCBM system. This is further enhanced by dielectric inhomogeneities in the material 17,18 . The above factors, in combination with the high crystallinity of P3HT:PC 70 BM enables efficient electron and hole extraction from the entirety of the depleted active layer under low reverse-bias voltages (<10 V). The above factors enable an X-ray detector with high sensitivity that strongly competes with all existing solid state X-ray detector technologies, under low voltages as well as over a broad X-ray energy range ( Fig. 1b and Supplementary Figure 1) with potential applications in X-ray imaging as shown from a prototype imager developed through this work (Fig. 1c). For optimisation of the detector, the NP loading within the device active volume was increased in order to increase the X-ray attenuation by varying the Bi 2 O 3 content in the parent solution (0, 10, 20, 30, 40, 50, 60, 70 and 80 mg ml −1 : X mg ml −1 is noted as Bi 2 O 3 -X). It is worth to note that the highest NP-loaded device which could be fabricated using the given procedure is Bi 2 O 3 -80, due to the formation of cracks during the annealing process at higher loadings beyond Bi 2 O 3 -80. However, with the appropriate selection of the organic bulk heterojunction and with tuning of the solvent used, much higher Bi 2 O 3 -loaded device fabrication maybe possible. Assuming a periodic structure for the NPs within the BHJ, this enables the unit cell dimensions to be reduced from 72 nm for  Figure 4a). We note recent work in the literature where the dark current can be tuned to meet industrial requirements through appropriate BHJ selection as a promising route for further improvements 19 .
Visible light photocurrent measurements are a useful tool in determining whether NP incorporation disrupts the BHJ phase separation thereby impeding charge transport. The lack of significant variation in the visible light photocurrent response for different Bi 2 O 3 loadings indicates that the phase separation within the BHJ remains undisturbed (Supplementary Figure 4b). The X-ray photocurrent response of the detectors tested under a 50 kV X-ray source at −10 V bias, demonstrates a linear increase with increasing NP loading from Bi 2 O 3 -0 to Bi 2 O 3 -40 ( Fig. 2a, b), followed by a non-linear increase for Bi 2 O 3 -60 and upwards. The X-ray sensitivity (S) depends on the amount of X-rays stopped, which depends on both the device cross section and its thickness, and hence, the sensitivity of the detector is calculated by: where, I X-ray and I dark are the current under X-ray irradiation and in the dark respectively, D is the dose and V, the detector volume.
For the detectors studied herein, S increases from 41 (Bi 2 O 3 -10) to 1712 (Bi 2 O 3 -80) µC mGy −1 cm −3 (Fig. 2c). We note that Ciavatti et al. 20 recently reported Bi 2 O 3 NP-loaded PFO polymerbased diodes for X-ray detection where the highest S observed was 160 µC mGy −1 cm −3 which is slightly higher than the S of Bi 2 O 3 -40 (105 µC mGy −1 cm −3 ). Despite the similarity in magnitude of these S values, we note that the former work employed a 35 kV Mo target as opposed to the 50 kV W target used in this work. As stated and shown schematically (Fig. 1b), the X-ray attenuation significantly improves by approximately one order of magnitude as the X-ray spectrum shifts to lower energies which is expected to result in the nearly similar S values for the Bi 2 O 3 NP-loaded PFO polymer-based diodes and the Bi 2 O 3 -40 diodes. We expect the S values for the Bi 2 O 3 -40 diodes to be higher if measured under low kV sources due to the enhanced X-ray attenuation as a result of increase in the mass attenuation, especially under high NP loading as well as due to the use of the BHJ thick films which enables a more balanced carrier transport. The combination of the direct conversion of Xray photons to charge carriers as well as the very high electric fields at the hybrid interfaces adjacent to the NPs, is a necessity for the much improved currents observed. The number of charges extracted (as calculated from the X-ray photocurrent response) vs. the number of X-ray photons absorbed (calculated based on the bulk X-ray attenuation model-Supplementary Figure 5) indicates a~× 10 3 enhancement in charge collection efficiency (CCE) than expected from the bulk attenuation model for Bi 2 O 3 -10 to Bi 2 O 3 -70, and a~× 10 5 CCE increase for Bi 2 O 3 -80 (Fig. 2d).
In comparison, the Bi 2 O 3 -0 system shows nearly an order of magnitude lower CCE. The voltage dependence of the BHJ-NP X-ray detectors display a linear behaviour for low reverse-bias voltages (<10 V) with sensitivity of 0.13 µC mGy −1 cm −3 for low-voltage bias (−2.5 V) indicating its suitability for portable, real time radiation monitoring, powered by coin cell batteries or by indoor lighting. An X-ray photocurrent response was also obtained under shortcircuit conditions (at 0 V) (Fig. 2e, f). A non-linear X-ray current response is observed for higher biases due to space charge-limited conduction with a three orders of magnitude improvement observed under −50 V. Further, as the voltage increases, the Xray response increases while retaining the dark currents of 6 and 41 nA cm −2 (at −0.1 and −1 V, respectively). The sensitivity of 0.13 µC mGy −1 cm −3 under reverse-bias voltages of <3 V indicates the suitability of this technology for portable, real time radiation monitoring, with power provided by coin cell batteries or by indoor light harvesting photovoltaic cells. Figure 2g shows that the rise time (to 90% of the maximum signal) and the fall time (to 10% of the maximum signal) increases as the Bi 2 O 3 loading increases. The slow response under reverse bias is due to the trap states generally present in metal oxide surfaces 21 and between the metal/semiconductor interfaces 22 . Further development in passivation of these defects is expected to significantly improve the response time of these detectors. We note that the rise and fall times are significantly reduced to less than 100 ms when the diodes are operated under 0 V. Another important detector metric is the linearity of the detector response, which enables the rapid determination of the X-ray dose, especially for dosimetry. Supplementary Figure 6 shows a representative X-ray photocurrent response curve for the Bi 2 O 3 -40 device in which an excellent linear response is observed.
An area of significant interest of X-ray imaging is the development of flexible, conformable imagers, a feature not enabled by current digital flat panel X-ray detectors. In order to evaluate the performance of the BHJ-NP detectors under deformation, we fabricated the optimized Bi 2 O 3 -40 X-ray detectors on a flexible substrate. The detectors demonstrate a nearly unchanged, high sensitivity of 280 µC mGy −1 cm −3 (Fig. 2h) both prior to, and after undergoing 10 bend cycles of 3 mm bend radius (Fig. 2i). A slight variation in the rise and decay constant was observed possibly due to the poor mechanical properties of the contact materials or the use of a fullerene-based acceptor, as opposed to an all polymer-based BHJ, which might have deteriorated the mechanical properties of the BHJ 23 . Based on our concept of hybrid organic-inorganic materials for X-ray detectors, many new and more suitable combinations can now be examined for future detectors.
A major potential application for such conformable X-ray detectors is, as dosimeters to be used in combination with a medical linear accelerator (LINAC), which are widely used for cancer therapy. The use of a conformable dosimeter as an in vivo detector on the surface of the patient or within a body cavity is highly likely to enable a more accurate X-ray delivery to patients thereby minimizing additional normal tissue damage as well as potential risks related to secondary cancer induction. A recent review in this field has recommended that all the radiotherapy treatments with curative intent should be verified through in vivo dose measurements in combination with pre-treatment checks 24 . As such, our optimized Bi 2 O 3 -40 detectors were tested under 6 and 15 MV X-rays from a medical LINAC (Fig. 3). Under a 114 µGy s -1 dose rate and −10 V reverse bias, the detector delivered a sensitivity of 30 and 58 µC mGy −1 cm −3 for 6 and 15 MV X-rays, respectively. These values are nearly ×100 higher than those reported for organic photodetectors tested under 6 and 15 MV LINAC X-rays using Gd 2 O 2 S:Tb as a scintillator 9,25 , therefore demonstrating that the BHJ-NP architecture developed here is a promising dosimeter concept for accurate dose delivery. It is noted that the Bi 2 O 3 -40 device did not show a noticeable performance degradation when exposed to 6 and 15 MV X-rays over several X-ray exposure cycles which results in a cumulative exposure dose of 0.15 Gy.
Structural characterisation. We examined the impact of the Bi 2 O 3 NP inclusion on the structural properties of the P3HT: PC 70 BM through Grazing Incidence Wide Angle X-Ray Scattering (GI-WAXS) (Fig. 4a)   Energy dispersive X-ray analysis carried out on the BHJ-NP film cross sections (Fig. 4b) shows a uniform Bi 2 O 3 NP distribution throughout the device thickness, indicating a homogeneous distribution of NPs and minimal dead volume without any NP sedimentation unlike previous reports 10 . Such distribution of the NPs offer an efficient X-ray to charge direct conversion, throughout the entirety of the device thickness. SEM and atomic force microscopy topographical analysis (Supplementary Figure 9) of the films suggests NP aggregation on the film surfaces which we have previously reported to enable efficient charge extraction through enhancements in the electric field via metal electrode structuring 26 .
Charge transport analysis. Transport of both electrons and holes was analysed by photo-induced time-of-flight (TOF), a technique widely used to measure the charge carrier transport in various low mobility semiconductors 27 . The mobility (μ) of the charges is determined from 28 μ ¼ d 2 =ðt trans VÞ where d is the sample thickness and V is the biased voltage. Figure 5a and b gives an example of TOF characteristics for electrons in a Bi 2 O 3 -20 device. Calculation of the carrier mobilities from the TOF transients indicates balanced electron and hole mobilities of~10 −3 cm 2 V −1 s −1 , with minor variation occurring for different electric fields (Fig. 5c, d). These values are comparable to those observed for thinner P3HT:PC 70 BM photovoltaic devices 29  Electric field (V µm -1 ) Electric field (V cm -1 ) 1.2×10 -4 2.0×10 -5 4.0×10 -5 6.0×10 -5 8.0×10 -5 1.0×10 -4 which is in agreement with the high crystallinity observed in the GI-WAXS as explained previously, as well as through DSC given in Supplementary Information 8 where a high crystallinity in excess of 45% is observed.
An important point for consideration in the development of Xray detectors is their CCE, described by the Hecht equation 14 ; Here, Q and Q 0 are the number of charges extracted and generated respectively at a distance x from the anode, E is the electric field and µ h , µ e , τ h and τ e are the hole and electron mobilities and transit times, respectively. Figure 5e exhibits the collected electron charge for different devices with a fit for the Hecht equation. For the BHJ-NP detectors, the CCE of electrons and holes separately are~63% with relatively similar μτ product of μτ %10 −7 cm 2 V −1 .
Origins of high sensitivity. As discussed previously, a thousandfold increase in charges extracted is observed with respect to the absorbed X-rays for Bi 2 O 3 -10 to Bi 2 O 3 -70 devices which increases to~× 10 5 for Bi 2 O 3 -80 devices (Fig. 2d). An important effect for enhanced sensitivity is impact ionization due to hot X-ray photoelectrons (PEs) and holes which enables~10 3 free carriers to be generated per photon 30 . However, the CCE of~63% points towards the possibility to an additional mechanism for sensitivity enhancement. One contributing factor towards this enhancement is photoconductivity gain, which results from trapping either electrons or holes 14 . The emergence of rise decay times well exceeding 100 ms from the device Bi 2 O 3 -50 upwards suggests photoconducting gain as a possible sensitivity enhancement mechanism for Bi 2 O 3 -50 devices and upwards. However, the <100 ms rise times for the lower NP loading is indicative of a different gain mechanism that enables fairly high rise and decay times to be observed. In astrophysical environments, it is well known that dust and ice NPs efficiently scatter X-rays due to their grain geometric sizes being on the same scale as the photon wavelengths 31,32 . In such studies, the classical Mie model is used as an approximation in determining differential and total scattering cross sections for nano-sized particles 32 . Mie-scattering results when electromagnetic radiation (EMR) interacts with particles with dimensions larger than the EMR wavelength. Due to the sub-nanometre wavelength of X-rays, NPs are extremely efficient Mie scatters of X-rays where the ratio of scattered X-ray intensity I s λ ð Þ, to the irradiated X-ray intensity I R λ ð Þ; can be found through 33 , where R and r d are the distance from the X-ray source to the detector and the detector thickness. a is the NP radius, θ is the Xray scattering angle, φ is the Azimuthal angle with respect to the axis source-detector and λ is the radiation wavelength. The differential cross section for the scattering of unpolarised radiation is given in Supplementary Note 4. In order to assess the impact of the NP size on the Mie scattering of X-rays, we simulated the Mie-scattering process for spherical Bi 2 O 3 NPs with diameters (d) of 20, 40 and 100 nm. The NP size is crucial for detector sensitivity as NPs in the quantum dot regime (<10 nm) lead to indirect X-ray detection 34 , while large nanoparticles (≥100 nm) reduce the extraction probability for charges generated within the NP and also reduce the differential scattering cross section as the particle size becomes very much greater than the X-ray wavelength 31 . These simulations do not take into consideration the effect of an ensemble of NPs, particle size distribution of the NPs, the formation of NP clusters, or deviations in the aspect ratio from a simple spherical geometry. Figure 6a shows the angular dependent differential cross sections for photon energies ranging from 1 to 30 keV for d = 40 nm NPs, where a high scattering effect can be seen at small angles. Considering a simplistic scenario where the Bi 2 O 3 NPs are arranged in a periodic structure throughout the detector volume, an incident X-ray photon (on a NP) with energy of 8 keV (E i ) has a high scattering cross section within 2°( Supplementary Figure 10 and Supplementary Note 5). Under elastic scattering this photon will be scattered by~45 NPs within the detector volume. However, as Mie scattering is an inelastic process, the scattered photon will have a lower energy (E s ) than E i . Based on the X-ray attenuation curve for Bi 2 O 3 , this results in the deposition of more energy (equivalent to E s ) as the scattered photon interacts with a subsequent NP (Fig. 6b, c). Therefore, tuning the NP loading such that two NP scattering sites are placed in close proximity to each other enables a significant enhancement (×100) in the X-ray path length due to the X-ray scattering process, and energy deposited even within layers that are dimensionally "thin" for X-ray detection (i.e. <100 µm).
Following the above simulations, we proceeded to confirm the predicted enhancement in X-ray scattering through Grazing Incidence-Small Angle X-ray Scattering (GI-SAXS) for BHJ-Bi 2 O 3 NP films (same composition) consisting of three different NP sizes (d = 20, 40 and 100 nm) (characterization of the Bi 2 O 3 NPs used in this study is given in Supplementary Figure 11a, c and Supplementary Note 6). The comparison between the simulation and experimental scattering (at 8 keV) is given in Fig. 6d. The GI-SAXS results are in good agreement with the predictions based on the Mie-scattering effects, with approximately an order of magnitude difference in X-ray scattering at angles < 0.75°for the d = 20 and 40 nm NP dimensions. However, as the NP diameter increases, up to 100 nm, the X-ray scattering is reduced by two orders of magnitude as compared to the d = 40 nm device. While this observation is not supported by the d = 100 nm simulation carried out for a single sphere, GI-SAXS analysis (Supplementary Figure 11b) indicates that the X-ray scattering in this film takes place from~15 nm sized features. This is indicative of the importance of optimum NP packing in these films to achieve the desired Miescattering effects.
In order to assess the impact of the Mie-scattering process on the detector sensitivity, X-ray detectors were fabricated using the three different NP sizes (d = 20, 40 and 100 nm) under the Bi 2 O 3 -40 condition, and tested under a 50 kV X-ray source (Fig. 6e, f). The d = 40 nm NP detectors show a three times higher response compared to the d = 20 and 100 nm NP devices, which is in good agreement with the enhanced X-ray scattering theoretically predicted for the d = 40 nm NPs. We note that while quantum mechanical effects due to nanoscopic features can also influence the scattering processes this would require more in-depth theoretical and experimental studies, which are outside the scope of this work.

Discussion
In conclusion, we have developed a hybrid 'inorganic in organic', direct transduction X-ray detector that delivers outstanding sensitivities of 1712 and 58 µC mGy −1 cm −3 for soft and hard X-rays respectively. Furthermore, flexible detectors also show a high sensitivity approaching 300 µC mGy −1 cm −3 , highlighting the promise of the technology for dosimetry and imaging in non-planar architectures. The improved X-ray sensitivity is a result of impact ionization, and an enhanced path length due to Mie scattering and the efficient separation, and transport of these by the BHJ-NP architecture resulting in high-charge collection efficiencies (>60%). Based on this concept, a preliminary flat panel imager has also been demonstrated. The method of direct detection and imaging, combined with low cost, flexibility and scalability for large-area manufacture, improves on current solid state X-ray detector performance by 2-3 orders of magnitude, under low voltages, while delivering novel attributes suitable for a range of current and previously unexplored detection and imaging applications. 8.20 Incoming X-ray  X-ray irradiation and characterization. Six devices were tested for each X-ray irradiation measurement. Three different X-ray beam sources were employed for the characterisation of the detectors: 1. A tungsten tube X-ray beam (Seifert RP-149 Semiconductor Irradiation System) with accelerating voltage of 50 kV under a dose-rate range of 27-131 µGy s −1 . Dose calibration was completed using an ion beam chamber (Radcal; 10 × 6-6). A Keithley 2400 source measure unit was used for recording the electrical characteristics.
3. 6 and 15 MV X-rays from a multi-mode linear accelerator (Clinac iX, Varian USA) located at the Royal Surrey County Hospital. Dose rates from 100 to 400 cGy min −1 were provided by the LINAC and a Keithley 2400 source measurement unit was used for recording the electrical characteristics of the devices.
Image readout and processing. An X-ray imager was fabricated using a method reported elsewhere 7  X-ray beam source (2) was used for the characterisation. Images were taken by custom-made driving and readout electronics with a commercial available readout IC (ROIC) (ISC9717 from Flir). The backplane of the imager consists of a Borosilicate glass substrate with an array of 256 × 256 a-Si:H TFTs with 98 μm pixel pitch. The signal at the input was simultaneously integrated, amplified, low pass filtered and converted from analog to digital with a 14-bit converter. The integration time was 10 ms and the integrator feedback capacitance C f was 4 pF. To eliminate fixed pattern noise and image inhomogeneity, a flat fielding has been performed.
GI-SAXS. Devices for GI-SAXS were fabricated as stated above, but with 20, 38 and 100 nm Bi 2 O 3 NPs in Bi 2 O 3 -40 devices. Measurements were completed on a XEUSS 2.0 (Xenocs, France) equipped with a Cu K α microfocus source and a Pilatus 300k detector (Dectris, Switzerland). The scattering vector (q) range of the data was calibrated using a silver behenate standard material. The detector was positioned 2.495 m away from the sample to utilise the lowest q range available. The sample was aligned such that the surface was in the centre of the beam and parallel with the beam. For the GI-SAXS measurements, the sample was tilted by 1°. This moved the specular reflection sufficiently far away in q that the scattering from the particles on the surface could be measured. The scattering was recorded for 30 min.
GI-WAXS. X-ray measurements were performed with a Xeuss 2.0 (XENOCS, France) system. The system is equipped with a MetalJet (Excillum, Sweden) liquid gallium source, providing a 9.24 keV X-ray beam collimated to a beam spot of 400 µm laterally at the sample position, measuring the full sample length. X-ray diffraction patterns were acquired with a Pilatus3R 1M 2D detector (Dectris, Switzerland) placed at~311 mm from the sample. The distance between the sample and the detector was measured using a silver behenate calibrant in transmission geometry. Samples were measured in GI-WAXS geometry at an incident angle of 0.3°(calculated to probe the entire film thickness) under vacuum atmosphere. Diffraction images were then remapped from pixel to scattering vector coordinates with the calibration equations reported 35 and by using MATLAB software.
SEM/AFM/EDX. SEM was carried out using a FEI Quanta 200F Environmental scanning electron microscope and AFM was carried out using a VEECO Dimension 3000 atomic force microscope. EDX analysis was carried out using an x-act Oxford Instrument system coupled with the SEM.
Time-of-flight. Devices were fabricated as stated above, but with a 20 nm thick Al contact for irradiation through the Al (hole transport analysis). As an excitation source, a 6 ns-pulsed Nd:YAG laser (Quantel, 532 nm, 45 mJ) was used, and the bias voltage was applied to the sample using a Keithley 2400. The transient current was measured as the voltage drop over a load resistor (10 kΩ) and recorded with an oscilloscope (Agilent infiniium 1 GHz, 4 GSa s −1 ) Simulation. FLUKA, a Monte Carlo simulation programme designed for the interaction and transport of articles and nuclei in matter, was used for simulation of the X-rays interaction with the active material. In order to simulate the energy depletion of our devices, the geometry of the P3HT:PC 70 BM BHJ with embedded Bi 2 O 3 NPs (the thickness variation of the active layer was obtained experimentally) were modelled and irradiated with 50 kV X-ray photons to be processed by the FLUKA programme. Mie-scattering differential cross section simulations were carried out using MATLAB software, using a Mie function.