## Introduction

The development of high-performing, thin, and flexible sensors for the detection of ionizing radiation in real-time at affordable costs is of increasing interest, as the technology currently available still fails to address the requirements of large-area, conformability and portability, lightweight, and low-power operation1. The fields of application range from medical diagnostics and therapy to astrophysics, high energy physics, as well as industrial testing, including civil radiation safety2. Up to now, the research focused greatly on X-ray detection, but the development of innovative charged particles detectors able to retain the above-described peculiarities (e.g., flexibility, scalability to large-area, and low power operation) would be particularly attractive, especially for real-time hadron therapy dosimetry. Proton therapy is currently one of the most advanced medical therapy tools for cancer treatment. Compared to X-ray therapy, where the delivered dose of radiation decreases exponentially with depth, a proton beam allows delivering a conformal radiation dose by tuning and focusing the peak of the maximum released energy (the Bragg peak) on the selected volume3, sparing the surrounding healthy tissues. However, toxicities may still occur within or very close to the target, due to the combination of suboptimal beam arrangement, organ motion, and range uncertainty4,5. A real-time beam monitoring and dose-measuring by a flexible, thin, and human-tissue-equivalent detector is therefore needed to verify the actual dose delivered during the proton-therapy session to healthy tissues close to the target cancer, preventing long-term toxic effects. Current therapy treatment plans are conceived by means of sophisticated simulation codes and designed experiments on customized phantoms6. An experimental real-time and in situ dose monitoring during therapy, carried out by means of a thin wearable device, would be a game shifter, providing highly beneficial7 improvements in cancer patients' care quality and precision.

Fully organic detectors are promising systems for advanced dosimetry, as their mechanical features allow devices to be adapted to complex contoured and large area surfaces, with outstanding portability and lightweight. Further, their human-tissue-equivalent properties, thanks to their density and composition, make them ideal candidates for medical dosimetry applications, as demonstrated in last years for the direct detection of ionizing radiation by organic thin films8,9,10,11,12,13 and single crystals14,15,16,17,18. The indirect detection of ionizing radiation is another effective tool, implemented via a two-step process: (i) the conversion of ionizing radiation into visible photons, carried out by a scintillator, and (ii) the conversion of the visible photons emitted by the scintillator into an electrical signal by a properly matched photoelectric device. With respect to the direct detection approach, in the indirect detectors, a scintillation layer with a thickness ranging from 0.1 to a few mm, allows the absorption of a larger amount of energy from the impinging radiation. The possibility to improve the stopping power of the indirect detector leads to very good detection performance at high radiation energies, more challenging to detect by thin-film-based direct detectors. On the other hand, indirect detectors typically exhibit high radiation tolerance at low energies, i.e., when the whole particle energy is released into the scintillator, thanks to the sparing of the photodetector from the impinging beam. To implement an indirect detector for ionizing radiation, the scintillation-produced UV–vis photons need to be converted into an electronic signal by a properly matched photoelectric device. This approach, applied to an organic electronic device, has been explored only for X-ray detection so far19.

In this work, we report on a fully organic, flexible proton detector operating in the indirect mode based on the combination of a polysiloxane-based flexible scintillator with a flexible organic phototransistor (OPT). Polysiloxane-based scintillators are intrinsically flexible, are highly resistant to gamma-ray and ion irradiation20,21, to high temperatures, to moisture and organic solvents besides offering optimal light output, with discrimination capabilities between different types of ionizing radiation22. These characteristics make them more suitable than commercial plastic scintillators and inorganic rigid crystals23 for applications involving radiation monitoring in contact with human tissues or in complex environments. Besides, OPTs combine high sensitivity with intrinsic signal amplification limiting noise issues to a simplified external circuitry. Preliminary reports in the early 2000s on organic UV-vis photodetector for indirect ionizing radiation detection24,25 lead to recent results on organic photodiodes in X-rays image sensor arrays26 on flexible plastic substrates27, working at low voltages (−1 V) and targeting dosimetry for medical application. An OPT active matrix has been proposed for indirect X-ray imaging application and the first indirect X-ray detector, based on OPT arrays coupled with an organic scintillation layer, was only recently reported28.

Here, we demonstrate the suitability of a flexible organic detector based, on the coupling of polysiloxane scintillators with OPT fabricated on thin plastic foil, for the real-time monitoring of low energy (5 MeV) proton beams. The detector has been assessed for flexibility and operation at low power supply, assuring high comfort and low electric risk for its employment as a personal dosimeter. Further, we present a kinetic model, developed to describe the detector response mechanism, able to precisely reproduce the dynamic response of the device under proton irradiation and to provide further insight into the physical processes controlling its response.

## Results

### Fully organic indirect detector configuration

The schematic cross-section of the here presented fully organic indirect radiation detector is shown in Fig. 1a, and optical images are shown in Fig. 1b. First, the OPT, based on dinaphtho[2,3-b:2′,3′-f]thieno[3,2-b]thiophene (DNTT) as the active layer, was fabricated onto a 100 μm thick polyethylene-naphthalate (PEN) foil, which allows combining flexibility with a carrier-free large-area compatible process. The OPT was developed to work at the low power supply, combining high bias stress tolerance and bending stability with high sensitivity at weak light intensities at a suitable optical absorption wavelength. Then, the 500 μm thick scintillating plastic film is integrated directly on top of the passivation layer. The scintillator is synthesized using a polysiloxane-based matrix, a primary dye (2,5-diphenyl oxazole PPO), and a wavelength shifter (Lumogen Violet®). Two different matrices were considered and compared, the homopolymer polymethylphenylsiloxane (PSS100), whose response under irradiation with ions, gamma-rays, and fast neutrons has been extensively reported and commented22, and the co-polymer, polyvinylphenyl-co-phenylmethyl siloxane (PVP-MPS), whose response as a base material for scintillators is herein reported. This matrix has a chemical structure similar to PSS100, with a monophenyl unit as a substituent on silicon, either methylphenyl or vinylphenyl, and a very close molar concentration of phenyl groups, hence high scintillation yield is expected29. On the other hand, in PVP-MPS the vinyl reactive groups involved in the cross-linking process are present as substituents along the chain of the base resin, as well as terminal groups, and their amount per unit volume is much higher as compared to PSS100 (Supplementary Table 2). This could enhance the density of cross-linkages and, in turn, the resistance to damage. Therefore, both formulations have been considered for the tests. Hereafter the matrix material acronym (PSS100 or PVP-MPS) is used to indicate the corresponding scintillating film.

The motivation of the work we carried out was to assess our device for the challenging application of the in-situ and real-time monitoring of the radiation dose impinging onto healthy tissues during a session of proton therapy, e.g., in prostate tumor treatment. To meaningfully address this task, we need to know with which energy range scattered protons irradiate healthy tissues. For a complete analysis of the problem, the main parameters are the dose released and the linear energy transfer (LET) distribution. A representative treatment plan for a prostate cancer patient is shown in Fig. 1c, d. The dose (Fig. 1c) was calculated with the Monte Carlo (MC) algorithm of the commercial system Raystation, while the dose-average LET (Fig. 1d and Supplementary Fig. 1a) was retrieved with a previously validated version of the TOPAS MC code30. The energy range of incoming protons was (162 ÷ 197) MeV, and two opposing fields were employed. The plan was calculated with a schedule of 2 Gy × 39 fractions, assuming a constant relative biological effectiveness RBE = 1.1 for protons, according to current clinical practice. Due to partial overlap with the target volume, the rectum receives medium-high doses. At the same time, the highest LET values are distributed at the edge of the target as well as in the rectum. This corresponds to low-energy protons (few MeV) characterizing the distal edge as well as the lateral penumbra of the field.

We tested the devices at the LABEC ion beam center (Laboratory of Nuclear Techniques for the Environment and Cultural Heritage, National Institute for Nuclear Physics INFN Firenze, Italy), where the above-described conditions can be reproduced (e.g., beam energy at the end of the path and deposited energy per unit volume). The final coupled full-organic indirect detectors were thus tested for responsivity, dose rate linearity and measurements repeatability using a 5 MeV beam. The proton penetration depth of 5 MeV proton beam inside the PSS100 and PVP-MPS siloxane scintillator films was calculated through the Stopping and Range of Ions in Matter (SRIM) Monte Carlo code31, resulting in both cases below 400 µm. In Fig. 1e the simulated LET released by the 5 MeV proton beam in both materials in terms of [keV µm−1] varies between 8 up to 40 keV µm−1, thereby matching the LET of a therapeutic beam close to being stopped by the material (Supplementary Fig. 1a). The 5 MeV proton beam releases entirely its energy and full stops inside the 0.5 mm scintillating films (in Supplementary Fig. 2 the proton range in the two films is reported), thus maximizing the scintillation efficiency, preserving the underlying OPT from possible damage due to direct proton exposure and granting a proper assessment of the decoupled/independent response of both the scintillator and the OPT.

Figure 1f reports the variation of the detector response obtained mimicking a typical operating condition during a proton therapy session. During the test, the device is exposed to 10 s radiation cycles with different intensities, while the control measurements are collected after 28 and 50 min under a pulse with irradiance 6.5 × 107 H+ s−1 cm−2. The device operates at a very low bias voltage (Vds = Vgs = −1 V), hence it is compliant with basic safety constraints for electrical hazards. The results show that the device response is reliable and stable until the end of the session and beyond.

### Scintillator performance

Two main critical features must be considered for the development of an indirect ionizing radiation detector. The first concerns the proper matching between the scintillation light wavelength and the optical absorption of the OPT active layer. The second one is to employ as OPT active layer an organic semiconductor with high field-effect mobility (µFE), which is directly proportional to the OPT responsivity32,33. The most suitable solutions at the state-of-the-art are summarized in Table S1. The optimal coupling between the chosen blue emitting scintillator and the air-stable, high-mobility DNTT semiconductor is shown in Fig. 2a. The ion beam-induced luminescence (IBIL) spectra of the scintillators under irradiation with 2 MeV protons and the UV–vis absorbance spectrum of the DNTT film show a good overlapping, assuring spectral matching between the light emitting sensor (siloxane) and the photoconverter (DNTT sensitized OPT).

The herein-used PSS100 and PVP-MPS-based scintillating films present an optimal combination of mechanical and optical properties, which make them suitable candidates to be coupled with a light detector fabricated onto a flexible substrate. The elastomeric nature of polysiloxanes is well known, having glass transition temperature well below room temperature (−125 ÷ −90) °C. They are soft and pliable, with an elongation at break in the range (100 ÷ 500) % tensile strength around (5 ÷ 10) MPa and Young modulus variable between 1 MPa and 3 MPa, although the presence of phenyl side groups might affect these features34. It is evident that well-performing plastic commercial scintillators (e.g., EJ-212) have much more limited flexibility, due to their rigid base polymers, i.e., polystyrene or polyvinyltoluene (Fig. 2b).

Figure 2c reports the pulse height spectra of 0.5 mm thick samples of PSS100 and PVP-MPS scintillators exposed to 241Am source (emitting 5.443 MeV alpha particles) as compared with EJ-212 commercial scintillator with the same thickness. Siloxane-based scintillators display optimal light output, with values around 70% of the benchmark scintillator.

To assess the reliability of PSS100 and PVP-MPS scintillators under high fluxes of charged particles, their light emission was recorded and compared with a standard EJ-212 scintillator, as a current variation of a power meter during irradiation with alternated high and low proton beam fluxes, fixing the irradiation time to 10 s (Fig. 2d). The output is negligibly affected by consecutive shots on the same spot, irrespectively of the intensity of beam flux, and the measurement results highly repeatable, as for the same beam flux the same output current is measured. The almost constant response in terms of emitted power, irrespectively of the total received dose, which has been estimated as 0.02 Gy (low) and 20 Gy (high), proves that the siloxane base polymer is almost inert to the deposited energy. Further, it testifies that the bond strength of the Si–O–Si unit guarantees the collection of a reliable and stable signal, in the range of total proton doses here applied, i.e., much higher than the ones expected for an effective proton therapy treatment plan. These results indicate that the response of PSS100 and PVP-MPS scintillators is reliable and stable, showing no evident radiation damage effects in light emission.

### OPT performance

The configuration of the developed OPT is staggered bottom gate top contact (Fig. 3a) realized onto a flexible substrate (Fig. 3b). Typical transfer characteristics in dark and air environments are reported in Fig. 3c. These do not highlight any hysteretic behavior in forward and reverse scans. The devices have threshold voltage (VTH) around −13 V, onset voltage ranging from −6 to 3 V, subthreshold slope down to 1.8 V dec−1 and the log(ION/IOFF) is between 5 and 6 high (transistor figures of merit are calculated according to35,36). Gate leakage current is Igs < 10 nA at Vg = −30 V. The OPT field effect mobility in linear regime is μFE = (1.1 ± 0.2) cm2 V−1 s−1, this value is in class with the highest mobility for polycrystalline DNTT-based transistors on rigid substrates37,38,39,40. The low variability, combined with a device yield as high as 100%, confirms the high reproducibility of the OPT performance.

A relevant feature to be considered is the stability under bias stress, which is necessary to discriminate light and gate bias effects, to ensure enduring performance reliability, and to measure a significant photocurrent even at very weak light intensities due to the absence of compensation effects between the two mechanisms41,42. The devices have shown high stability, as previously reported43, with maximum variation ΔVTH ≈ 260 mV after stress time and biases higher than those in the expected operating conditions (Supplementary Fig. 6a), without significant mobility variation (ΔμFE, = 0.4%).

The OPTs were tested under repeated bending cycles, as reported in Fig. 3c, showing no relevant hysteresis or variation in μFE. Such results are significantly improved compared to the degradation reported for conformable phototransistors44,45.

A spin-coated fluoropolymeric film was used as an encapsulation layer, able to protect the active layer from environment43,46, prevent damage to the OPT during the coupling with the scintillator, and offer a suitable adhesion interface, transparent to the wavelengths of the scintillator light. Adhesion features were confirmed experimentally (Fig. 1b and Supplementary Fig. 5), and the required optical properties are ensured by the polymer high light transmittance ratio and internal transmittance (95% and > 99%, respectively). The efficacy of such an encapsulation layer for the OPT/scintillator coupling is confirmed by the good overlapping of the OPT transfer curves recorded before and after the coupling step (Supplementary Fig. 6b).

The dynamic photoresponse of the OPT was characterized under irradiation at 460 nm light pulses at different irradiances (Fig. 3d). A light pulse width of 10 s has been chosen to emulate the operating conditions used during the measurements under proton beam irradiation, which fits the lowest exposure times of protontherapy treatments47,48. The increase of the drain current (Ids) is defined as photocurrent $$I_{{{{\mathrm{ph}}}}} = I_{{{{\mathrm{ds,light}}}}} - I_{{{{\mathrm{ds,dark}}}}}$$, where Ids,dark is the Ids before illumination and Ids,light is the Ids under light exposure. A good linearity is observed for pulse width up to 0.5 s while, for longer exposure times, the response of the organic semiconductor to visible light shows an increasing non-linearity at high irradiances (see the inset of Fig. 3d), a feature that is well recognized in literature32.

The lowest detectable optical power density in these experimental conditions, also known as the limit of detection (LoD), was evaluated following the IUPAC definition49, as the minimum radiation flux that provides the signal-to-noise ratio of 3 (SNR = 3). The LoD value was obtained from a linear fit of the photoresponse curve for pulse widths of 10 s in a low optical power density range, as LoD = 3σ/slope, where σ is the noise floor of the setup (Supplementary Fig. 7). The estimated value is ≈18.5 nW cm−2.

In order to quantify the time performances of the OPT response under pulsed illumination, both formation and decay of the Iph vs. time waveforms were fitted with the equations:

$$I_{{{{\mathrm{ph}}}}}^{{{{\mathrm{formation}}}}}\left( t \right) = A_1\left( {1 - \exp \left( { - t/\tau _1} \right)} \right) + A_2\left( {1 - {{{\mathrm{exp}}}}\left( { - t/\tau _2} \right)} \right)$$
(1)
$$I_{{{{\mathrm{ph}}}}}^{{{{\mathrm{decay}}}}}\left( t \right) = A_3\exp ( - t/\tau _3) + A_4\exp ( - t/\tau _4)$$
(2)

where Ai is the amplitudes of exponential functions, and τi are the pertaining time constants. According to50,51,52, the use of two-time constants allows an accurate reproduction of the photocurrent dynamics in the time range considered (20 s), corresponding to the largest Iph variation, and permits quantifying the detector response-time performance. The fits clearly show a fast and a slow component (Fig. 3e). In the pulse formation, the fast component has time constants from about half a second to a few tenths of seconds, whereas the slow component ranges from about 6 s to a few seconds. During the pulse decay, the fast component time constants range from about 16 s to nearly 10 s, while the slow one spans from about 1.5 s to about 0.5 s. For both components, the time constants decrease with the pulse optical power. Figure 3f shows the OPT photoresponse at 2200 nW cm−2 with superimposed fitted curve according to Eq. 1 and Eq. 2. The figure also shows the detail of the fast component contribution to the formation and decay process. For optical power higher than 1 μW cm−2 nearly half of the pulse swing is due to the fast components, both during formation and decay. Such behavior makes it possible to deploy this OPT as a detector in practical applications requiring a bandwidth in the order of Hz.

### Characterization of integrated fully organic indirect proton detector

Fully organic coupled indirect detectors were tested under cycles of exposures to a 5 MeV proton beam, with fluxes of particles in the range of (106 ÷ 1010) H+ s−1 cm−2, employing the experimental setup reported in the ‘Methods’ section. The exposure time was kept constant at 10 s. During irradiation, the OPT was polarized in a subthreshold regime, slightly above the onset (Vds = Vgs = −1 V), to achieve the best compromise between responsivity and photosensitivity of the sensor43. The dynamic response of the detector to the different proton fluxes is reported in Fig. 4a, where the yellowish area indicates the proton irradiation window. The photocurrent shows a steep increase upon irradiation and follows the typical OPT response dynamics reported in Fig. 3d, assessing the effective coupling between the scintillating film and the underlying OPT fabricated on a plastic substrate. Figure 4b reports the photocurrent values normalized to the dark current (i.e., the current flowing within the OPT channel in the absence of proton irradiation) as a function of the impinging proton flux, recorded with detectors employing the two proposed scintillator films: PSS100 and PVP-MPS. In both cases, the sensor response results proportional to the proton flux and, as expected from the scintillation pulse height spectra under alpha particle irradiation reported in Fig. 2c, the signal amplitudes are comparable for both scintillators at the same proton flux. The LoD values of the here reported full organic indirect proton detectors are as low as 3.4 × 104 H+ cm−2 s−1 (0.043 Gy min−1) and 1.9 × 104 H+ cm−2 s−1 (0.026 Gy min−1) for PSS100 and PVP-MPS, respectively (Supplementary Fig. 8). These values are well in line with the proton rate required for our target application, the excessive dose detection in prostate cancer proton therapy, which is typically (0.012 ÷ 0.07) Gy min−1. In the Supplementary Information section, Supplementary Table 3 reports the comparison of the operative ranges of the most recent proton detector reported in the literature to the best of the authors’ knowledge.

The bendability of the whole detector (i.e., scintillator coupled with the OPT on the flexible substrate), was assessed under proton flux on two detectors with almost identical OPT transfer characteristics, one kept in a flat configuration and the other bent at a radius of 0.5 cm (Supplementary Fig. 9), a value chosen to be conformable to most of the human body curves in view of possible medical dosimetry applications. The comparison of the proton-induced photocurrent for flat and bent samples as a function of the impinging proton flux is shown in Fig. 4c. The detectors exhibit comparable performance when exposed to similar proton fluxes in the range (2 × 107 ÷ 4 × 108) H+ s−1, with slightly higher values (about 15%) of proton induced photocurrent for the bent device. Additional characterization of the unvaried detector response at different bending radii, upon optically exciting the scintillator, is reported in Supplementary Fig. 10. Similar behavior was reported for organic-based flexible direct detectors under X-ray exposure, with an even higher deviation of bent device response from flat condition8,53. To further evaluate the stability of the operation of the detectors, their tolerance to high proton irradiation up to 3 × 1010 H+ was tested, resulting in almost unvaried OPT transfer characteristics (Fig. 4d), with a moderate increase of the off current due to the long recovery time of the OPT response to UV-Vis light emitted by the scintillator, as will be discussed in detail in the next section. The detector exhibits high stability of the response under 6.5 × 107 H+s−1 cm−2 (56.7 Gy min−1) pulsed irradiation, as reported in Fig. 2d. The time response of full detectors under proton irradiation was analyzed by using Eq. 1 and Eq. 2. In the range of proton fluxes used in these tests, the time constants have comparable values to the ones obtained under direct UV–vis photons exposure (Fig. 4e).

In order to assess a method able to predict the OPT performance in the indirect detector configuration starting from the optical response, the response of the OPT under light and of the full detector under protons were compared. In Fig. 4f the optical power density produced under proton irradiation by PSS100 and PVP-MPS scintillators and the optical power densities needed to generate the experimentally observed Iph, as derived from the λ = 460 nm LED stimulus, are shown. The light power generated by the scintillator when irradiated with protons (same energy and flux as in the previous tests) was measured by directly coupling it to a calibrated power meter. The resulting photo-response trends are clearly parallel, and the observed shift could be ascribed to two main systematic effects: (i) LED and scintillator-emitted lights have different spectral distributions, and optical coupling with the OPT; (ii) the actual size of the light spots produced by the LED and by the scintillator is not identical. On such a basis, the two photo-response data sets could be considered in reasonable agreement and allow our setup to predict the trend of the whole detector response to the proton beam, starting from the OPT optical characterization.

### Kinetic model

Photocurrent in organic semiconductors has been related by several authors to an increase of trapping of minority carriers induced by light exposure8,54,55,56,57. The measurements reported in this work for the OPTs are consistent with this interpretation. As shown in Fig. 4d, the electrical transfer characteristics of the OPT in the detector, measured before and after exposition to a proton flux, show a Iph/Ids,dark that is maximum in low Vgs operation (subthreshold and off-regime). The corresponding current variations that are observed in our devices can be reproduced (Supplementary Fig. 11a) by a field-effect 1D simulation58,59 introducing, as a consequence of light exposure, a deep trapping acceptor level in the HOMO–LUMO energy gap, located at 0.4 eV from the HOMO level. We note that light-induced creation of deep traps for charge carriers in the energy range (0.3 ÷ 0.7) eV from HOMO level has already been evidenced experimentally in the analysis of several organic semiconductors used in different optoelectronic applications60,61,62. Moreover, 1D simulations, for the range of visible radiation exposures (on OPTs) and proton beam exposures (on detectors) used in this work, show that the photocurrent can be assumed linearly dependent upon the total amount of generated traps N(t):

$$I_{{{{\mathrm{ph}}}}}\left( t \right) = c_IN\left( t \right)$$
(3)

where the proportionality factor cI = ∂Iph/∂N depends on the specific polarization used for the OPT (see Supplementary Fig. 11b for an estimation at Vgs = Vds = −1 V). Several microscopic mechanisms have been proposed to account for a photo-induced increase of deep traps, and, among the others, the creation of structural defects by photo-oxidation63, cross-linking64 and hydrogen abstraction (migration)65. The precise identification of the microscopic mechanism responsible for the photocurrent behavior is out of the scope of this work, nonetheless, it is important to develop a dedicated kinetic model able to properly describe the photocurrent behavior in our device to gain a better understanding and control of their application as proton detectors.

The dynamic photocurrent measurements on our organic devices (Fig. 5a, b, experimental curves) show two distinctive features:

1. (1)

The relaxation in the dark of the photocurrent is well described by a stretched exponential (Supplementary Fig. 12a) with an exponent β almost constant (around 0.5), and a characteristic time τs that strongly depends on the exposure conditions ranging in the interval (2 ÷ 8) s (Supplementary Fig. 12b). This kind of dynamical response has been already evidenced in different organic- and inorganic-based devices8,43,66 supporting the presence of a distribution of activation energies for the recovery of trapping centers often related to the lack of crystallinity of the semiconductor thin-film. By applying the theory described in67, these values of τs and β can be associated with the presence of defects with recovery activation energies distributed as a gaussian with an expected value ε1 ≈ 0.7 eV and a variance of δ1 ≈ 50 meV. We note that a stretched exponential behavior can be effectively reproduced, on short time intervals, by a superposition of a few exponentially decaying components, thus confirming the validity of the approach adopted in the section ‘OPT performance’ to characterize the dynamic performances of the detector on short time ranges.

2. (2)

Dynamic photocurrent measurements under repeated exposures show a systematic drift due to the buildup of a persistent component of the photocurrent54,55,68. The resulting dark current increase is clearly noticeable in the measurements made on the OPT under light exposure (Fig. 5a) as well as in the measurements on the detector under proton irradiation (Fig. 5b, c). The persistent photocurrent component has a recovery time of the order of 105 s (more than one day in the dark), which is much longer than the characteristic time τs of the swift component discussed above. It must be associated with defects with a distribution of activation energies with a mean value ε2 considerably higher than ε1.

The above experimental evidence suggests that there are at least two kinds of photo-induced defects determining the photoresponse: both have distributed recovery activation energies, in order to account for the stretched exponential behavior, but the mean values of these distributions must be different, as they relate to the swift and the persistent photocurrent components, respectively. To model these effects, the rate equation proposed by Street et al.60 was modified by introducing two kinds of defects, instead of only one, with distributed recovery activation energies, instead of a fixed value (see Fig. 5d).

The proposed model (described in detail in Supplementary Information, Note 5) is able to quantitatively reproduce the dynamical photoresponse for a wide range of exposure intensities under both photon and proton fluxes. As seen in Fig. 5a–c, the curves computed by the model with the fitted parameters nicely superimpose the experimental ones, correctly reproducing the rise and fall dynamic of the exposure-response and the progressive build-up of a persistent photocurrent. The swift and persistent contributions to the photocurrent given by the fast- and slow-recovery defects, respectively, are also reported in the three panels. The slow-recovery defects determine a relatively small increase of the photocurrent during the exposure phase that, on the time scales of Fig. 5, does not recover appreciably during the dark phase, thus determining the systematic drift of the photocurrent baseline. On the other hand, it is apparent that the fast-recovery states give the bigger contribution to the photocurrent variation during the exposure, and they recover almost completely in a few tens of seconds during the dark. Hence, the slow recovery time of the persistent component does not affect the reproducibility of the device response for pulsed low light intensities, as shown in Fig. 5c and already reported for the phototransistor elsewhere43.

The model gives also some valuable physical insight in explaining some features of the dynamical photoresponse. In particular, the reduction of the characteristic fall time τs with increasing fluxes (Supplementary Fig. 12b) is well reproduced and can be related to the shift of the steady state distribution of the defect activation energies towards lower values as the flux values are increased.

## Discussion

The herein presented results demonstrate a bendable indirect fully organic proton detector able to quantitatively monitor in real-time the dose released by MeV proton beam irradiation. The detector was developed for this application, testing the device in order to assess its reliability for the excessive dose monitoring in MeV proton therapy, but the versatility of the organic materials and of the adopted device architecture used allows us to easily tune its properties for other applications. The scintillator emission wavelength properly matches the optical absorption of the OPT active layer. The herein proposed PSS100 and PVP-MPS-based scintillating films present an optimal combination of mechanical and optical properties, and their response under proton exposure is reliable and stable, not showing degradation in light emission nor other evident radiation damage effects. The light response of both formulations is quite similar, and the variation in reactive vinyl group concentration does not affect the radiation resistance, in the adopted experimental conditions. The developed OPTs have high field-effect mobility, μFE = (1.1 ± 0.2) cm2 V−1 s−1, and high stability under bias stress. The transistors were proved to be suitable for flexible applications, being unaffected by repeated bending cycles. The encapsulation layer efficiently protects the active layer from the environment and aging, while granting an efficient coupling with the scintillator.

The full device shows response stability and reproducibility under pulsed proton flux, in the typical operating conditions of a proton therapy session. We detect a minimum dose rate down to 0.043 Gy min−1 and 0.026 Gy min−1, respectively, for the OPT coupled with PSS100 and PVP-MPS-based scintillators. The detector response trend is unaffected by bending stress and the proton-induced photocurrent shows only a slight increase in bent devices. The detector operates at a very low bias voltage (Vds = Vgs = −1 V), hence it is compliant with basic safety constraints for electrical hazards and with a low power supply.

The kinetic model presented here allows getting a proper physical insight into the detector response mechanism. The model is able to quantitatively reproduce the dynamic response of the detector under proton irradiation, which was compared with the behavior of the OPT under UV–vis light. The computed curves nicely superimposed the experimental ones for both conditions, confirming that the OPT response is entirely due to the effect of the light emitted by the scintillator, as expected by the proton penetration depth calculated through SRIM simulations to terminate within the scintillator. The response in the operating range of the detector, the rise and fall dynamic, and the progressive build-up of a persistent photocurrent are correctly reproduced. The two components identified in the response under proton flux (a swift and a persistent one) were attributed to two kinds of photo-induced defects with different mean values of the distribution of the recovery activation energies. It was demonstrated experimentally and confirmed by the computational analysis that the fast response is recurring, independently from the persistent current drift, thus assessing the suitability of the here proposed devices as a real-time proton detector. Finally, the quantification of the two components in the detector response can be exploited to operate in two simultaneous modes, as recently proposed for direct organic proton detectors with a similar behavior69: (i) the real-time monitoring of proton irradiation and (ii) the monitoring of the total received dose.

This work demonstrates the potential of fully organic thin-film flexible devices for a variety of applications within the proton detection field, from experimental scientific research to innovative theranostics.

## Methods

### Proton irradiation and detection tests

The detectors were characterized using a 5 MeV proton beam provided by the 3 MV Tandetron accelerator of the LABEC ion beam center (INFN Firenze, Italy)70. The beam is extracted into ambient pressure through a 200-nm-thick Si3N4 membrane; the sample is typically installed at a distance of 8 mm from the extraction window. The proton beam current employed during the experiment was in the range (0.03 ÷ 94) pA, corresponding to (1.09 × 106 ÷ 3.5 × 109) H+ cm−2 s−1 proton fluxes. The intensity of the beam is monitored and measured by a rotating chopper71, placed between the Si3N4 window and the sample, that intercepts the beam. The chopper is a graphite vane covered with thin nickel evaporation, and the Ni X-ray yield is used as an indirect measurement of the beam current. To determine the actual energy of the protons impinging onto the top siloxane scintillating layer, the energy lost by the protons passing through the several layers interposed between the beam and the sensor has to be calculated, including 200 nm of Si3N4 for the beam extraction window, 8 mm of mixed air–He (50–50%) atmosphere in the gap between the extraction window and the metal box, 14 µm of Al for the entrance window of the box, where the sensor was enclosed, and 14 mm of air inside the box. After passing through these layers, protons lose about 390 keV, as calculated with the SRIM Monte Carlo code31 During proton irradiation tests, the electrical photoresponse of the devices was measured by using a Keithley 2614 SourceMeter, controlled by a custom-made Labview software. All measurements were carried out keeping the device in the dark, in a Faraday cage, to reduce electrical noise and avoid light-induced photogeneration in the organic semiconductor.

The measurement of the detector response under bending was performed by using a customized 3D printed box with a curved bottom in order to place the detector in the bent configuration, as shown in Supplementary Fig. 9c. The measurement of the detector’s response in a flat configuration and at different bending radii upon optically exciting the scintillator reported in Supplementary Fig. 10 has been performed a Thorlabs SLS400 Xe lamp powered at 150 W coupled a SPEX 500 spectrometer.

### Scintillator fabrication, characterization, and modeling

Polysiloxane scintillators were produced using vinyl-terminated polymethylphenyl-co-phenylvinyl siloxane or polymethylphenylsiloxane as starting resins (PVP-MPS and PMPS), which undergo vulcanization by Pt-catalyzed addition with a silane containing cross-linking resin as described elsewhere72. The structures of the base resins used herein and other peculiar features are reported in Supplementary Table 2.

After the addition of proper additives and cross-linker, the viscous precursor is cast in the form of 0.5 mm thin sheets, using a film applicator and a motorized stage (Erichsen, model Unicoater 509). The resin is spread over a glass plate, previously treated with a thin layer of Teepol, as a release agent. Then, the thin layer is left to dry and cross-link overnight at 60 °C, prior to easy detachment by immersion in water to produce a self-standing film, as shown in the photo of Supplementary Fig. 4.

To collect the IBIL spectra, the samples were irradiated with 2 MeV protons and beam current (1 ÷ 2) nA at the AN2000 accelerator (INFN-LNL). The measurement is done into a vacuum chamber located along the beam line, and the sample is directly exposed to the beam, while an optical fiber is fixed at 45° with respect to the sample surface to collect scintillation light73, as shown in Supplementary Fig. 3. An optical spectrometer is connected to the fiber through proper vacuum/air feedthrough and gathers one spectrum every 5 s during the irradiation, which lasts on the whole about 600 s The total charge is gathered by a Faraday cup and measured over the entire period of irradiation; then, the value of dose rate in ions/s can be derived as an average value. The spectra herein reported are referred to the normalized scintillation intensity collected after the first 5 s irradiation.

The power meter used for the scintillation light measurement is a 818-SL by Newport.

The polysiloxane-based scintillating layer was applied on the top of the OPT device with a thin layer of optical cement EJ-500 (Eljen Technology), which assures a fast and robust bonding between the polysiloxane-based film and the Cytop™ passivation layer. After the application and adhesive curing, the bi-layer was bent to the required curvature radius and fixed on the curved support using Kapton tape, as shown in the photo of Supplementary Fig. 5.

### OPT fabrication and characterization

The OPT fabrication details are reported in43. In brief, the process was carried out in a cleanroom environment at a low temperature, within 100 °C, directly on the free-standing plastic foil, a PEN (100 μm thick, Teonex® Q65FA, DuPont Teijin). The dielectric and encapsulation layers are a fluoropolymer-based material (Cytop™, AGC Chemicals), deposited by spin-coating, 600 nm and 240 nm thick, respectively. All the other layers were thermally evaporated through shadow masks (Stencils Unlimited). The gate is Al 70 nm thick, the semiconductor is dinaphtho[2,3-b:2′, 3′-f]thieno[3,2-b]thiophene (DNTT), 50 nm thick (assay 99%, as purchased, Merck), the source and drain (S/D) contacts, Au 30 nm thick, the connections and pads, Al 70 nm thick. The via-holes were obtained via plasma oxygen. Electrical characterization of the devices was carried out at room temperature in ambient environment, using a probe station equipped with two Keithley 236 source meter. The bending test setup is reported in74. The bending radius was 5 mm, the frequency 0.5 Hz, while the bending direction was chosen to be perpendicular to the channel length to test the device in the worst working conditions. The tests for the photoresponse at 460 nm were carried out by sending light pulses from a calibrated LED source, emitting at 460 nm (Broadcom). The LED driver allows spanning the optical power density range from nW cm−2 to a few μW cm−2.