Composition‐Dependent High‐Performance Phototransistors Based on Solution Processed CH3NH3PbI3/ZnO Heterostructures

Recently, hybrid organic−inorganic perovskites have emerged as promising photo‐sensing materials for next‐generation solution‐processed phototransistors, achieving high responsivity, detectivity, and fast response. Here, a phototransistor that can detect visible light using a low‐cost, solution processed methylammonium lead iodide/zinc oxide (CH3NH3PbI3/ZnO) heterostructure is reported. While typical ZnO thin‐film transistors (TFTs) do not show any photocurrent under visible light illumination, CH3NH3PbI3 (MAPbI3) coated ZnO TFTs exhibit substantial photocurrent. Additionally, the optical, morphological, and structural characteristics of the light‐absorbing layers are further controlled by altering the precursor ratio of methylammonium iodide and lead (II) iodide (MAI:PbI2), which in turn affects the photosensitivity. Stoichiometric composition (MAI:PbI2 = 1:1) of MAPbI3 demonstrates optimal characteristics with a responsivity of 234 A/W and a high detectivity of 3.74 × 1013 Jones under white light illumination. The high photo‐responsivity and detectivity result from the combination of the suitable optoelectronic properties of the stoichiometric MAPbI3 film and its smooth interface with the ZnO channel.


Introduction
In recent years, high-performance phototransistors have attracted extensive attention due to gate bias control functionality, DOI: 10.1002/adom.202300367pixel array integration, and the ease of on-panel processability. [1,2][8] In particular, film fabrication using a solution process allows uniform integration of solar cells, [9] photodiodes, [10,11] photoconductors, [12] and phototransistor devices [8] over a large area into a hybrid device.Hybrid perovskite materials, represented by the structural form of ABX 3, have excellent light absorbing characteristics and are strong sensitizers in the range from 300 to 800 nm, especially ≈500 nm, with exciton diffusion length of up to 100 nm. [13]It has been reported that for perovskite-based devices, the absorption rate of light could exceed 90%. [14]Moreover, the defect density in the energy band of perovskites is very low, resulting in excellent device performance. [13]n addition to excellent optoelectronic properties, it is also possible to tune device performance by varying the ratio of the organic (MAI) and inorganic (PbI 2 ) components in the MAPbI 3. One method that is commonly used to maximize the performance of perovskite solar cells is adding a small amount of excess of PbI 2 in the final perovskite absorber layer.[17] However, other reports have also claimed that excess PbI 2 can accelerate device degradation in the presence of oxygen and humidity, [18,19] or heat and light. [19,20]Furthermore, some studies found that excess PbI 2 has a detrimental effect on device performance and can cause parasitic absorption. [10,21]Therefore, in order to draw a conclusion regarding these apparent contradictions, it is essential to further investigate the role of the precursor ratio on the performance of MAPbI 3 based optoelectronic devices.
On a different note, oxide semiconductor (OS) materials are also considered for future optoelectronics applications due to their high electron mobility, stability, and optoelectronics properties. [22,23][26] However the majority of these reports use vacuum deposition techniques to fabricate devices.[27][28] Moreover, reports have suggested that devices fabricated with the most commonly used OS, IGZO, suffer from persistent photoconductivity (PPC), which is induced by photoionization of metastable oxygen vacancy sites in its amorphous structure. [2,29]This consequently increases the response time of the photodetector to optical stimuli, resulting in poor transient response characteristics.
Among the commonly used OSs, ZnO is considered as an alternative to IGZO for future optoelectronic applications. [30]36][37][38] To utilize OSs in visible light detectors, several attempts have been made to generate photocurrent in oxide TFTs under visible light illumination by adding polymers or metal nanoparticles to the surface of the active layer. [39,40]However, due to its bandgap tunability and excellent optical properties, MAPbI 3 has been proven as a suitable photo absorber, explored by numerous researchers for photodetection applications. [41,42]Moreover, the crystallographic anisotropy of perovskites could drastically improve the response time and stability of photodetectors. [43]urthermore, the low-temperature facile solution-processing of MAPbI 3 has opened up bright prospects for the fabrication of low cost and sustainable optoelectronic applications.Given that fabrication procedures of conventional two-terminal photodetectors (avalanche photodiodes and photomultipliers) is complicated, expensive, and that these bulky devices consume a lot of power, a facile gated photodetector (phototransistor) could be an excellent solution. [44,45]][48] However, their performance still requires improvement and the majority of demonstrations employ complicated device structures, necessary to boost light sensitivity.Furthermore, they require inert conditions for processing due to ambient air instability.
Here, we demonstrate phototransistors (Figure 1a) fabricated in ambient air using low-cost and low-temperature solutionprocess techniques, including spin coating and spray pyrolysis.For a broader absorption spectrum, a smaller bandgap organicinorganic hybrid perovskite film (MAPbI 3 ) was deposited on top of the ZnO channel layer as a light-absorbing layer.Due to the large absorption coefficient and long diffusion lengths of charge carriers in perovskites, in addition to the high electron mobility and hole-blocking nature of ZnO, the photo-induced carriers can be separated efficiently.Compared with typical ZnO TFTs, the performance of phototransistors coated with the MAPbI 3 perovskite films significantly improved.Furthermore, we varied the precursor ratio (MAI:PbI 2 ) to understand its impact on phototransistor performance.The optimized device demonstrates excellent characteristics for detection of weak light; with a responsivity of 234 A/W and a high detectivity of 3.74 × 10 13 Jones under white light illumination at a low drain voltage.The high photoresponsivity and detectivity results from the combination of suitable optoelectronic properties of MAPbI 3 films with the high mobility, low dark current noise of spray coated ZnO films.
active layer, deposited via spray pyrolysis (substrate temperature 350 °C).ZnO films were patterned and wet etched to form active islands, followed by patterning and wet etching of the GI.Next, a sputtered Mo layer was patterned and dry etched to form source and drain contacts.Finally, MAPbI 3 was deposited via spin coating (baked at 100°C for 10 min in ambient air to remove solvents).Inorganic-(I-), stochiometric-(S-), and organic-(O-)rich MAPbI 3 were created by varying MAI and PbI 2 ratios, yielding MAPbI 3 with MAI:PbI 2 of 0.8:1 for I-MAPbI 3 , 1:1 for S-MAPbI 3 , and 1.2:1 for O-MAPbI 3 .Patterning was performed via contact photolithography (see Experimental Section for full process and device characterization details).We first present results of film characterization before assessing heterostructure photo-TFT performance.

CH 3 NH 3 PbI 3 and ZnO Film Characterization
The optical, surface and structural properties of a ZnO monolayer and MAPbI 3 films for the three different MAI:PbI 2 ratios (I-MAPbI 3 , S-MAPbI 3 , and O-MAPbI 3 ) on glass substrates are presented in Figure 2. We begin with the measured ultravioletvisible (UV-vis) absorption spectra of these films, for which the S-MAPbI 3 film had the highest absorbance, followed by the O-MAPbI 3 film, with the I-MAPbI 3 film having the lowest .Figure 2a-d shows Tauc plots deduced from ultraviolet visible (UV-vis) absorption spectra of the ZnO, I-MAPbI 3, S-MAPbI 3, and O-MAPbI 3 films on glass.The optical bandgap of the spray coated ZnO film is 3.20 eV, while those of the spin coated I-MAPbI 3 , S-MAPbI 3 , and O-MAPbI 3 films are 1.58, 1.56, and 1.59 eV, respectively.The narrow band gap of all three perovskite films supports the appropriateness of MAPbI 3 as a wide range visible light photo-absorber in the heterostructure. [49]he interaction between the organic and inorganic components dynamically determines the bandgap of MAPbI 3 .][52][53][54] Y. Wang et.al. showed that changing the ratio of precursor content led to an evolution of the unit cell parameters, absorption, photoluminescence, and highest occupied molecular orbital (HOMO) level, emphasizing the effect of precursor stoichiometry on perovskite material properties. [53]dditionally, K. P. Ong et.al., showed that the energy band gap of MAPbI 3 can vary from 1.02 to 1.55 eV, depending on MA molecule orientation, corresponding change in the lattice constant, and volume of the unit cell. [54]Results presented herein are thus consistent with excess MA, favoring orientations with larger band gaps in the organic rich films.For the inorganic-rich films, excess PbI 2 can also contribute to larger sizes of the measured bandgap.
X-ray diffraction (XRD) patterns (Figure 2e-h) confirm the successful fabrication of ZnO, I-MAPbI 3 , S-MAPbI 3 , O-MAPbI 3 films.The X-ray diffraction pattern of the spray coated ZnO film (Figure 2e) indicates c-axis oriented growth of a smooth nanocrystalline surface, which promotes uniform interfaces in TFTs between the active layer and GI, as well as the active layer and source/drain contacts.Interface uniformity is essential to reduce surface defects and facilitate charge transfer from the perovskite films to the ZnO channel of the TFT.With respect to XRD spectra of all MAPbI 3 films, sharp peaks at 2 values of 13.9°, 28.2°, and 31.7°(Figure2f-h, respectively) can be observed, which indicates high crystallinity in the tetragonal phases of the ( 110), (220), and (310) crystal planes of the MAPbI 3 film. [55,56]The intensity of the PbI 2 peak (at the 2 value of 12.67°) is absent in the S-MAPbI 3 , and O-MAPbI 3 films, whereas its presence in I-MAPbI 3 indicates excess PbI 2 in the film. [55,56]urface morphology of the spray coated ZnO and the MAPbI 3 films was also investigated by atomic force microscope (AFM) and scanning electron microscope (SEM) and the images are shown in Figure 2i-l and Figure 2m-p, respectively.As expected, the ZnO formed a smooth film with root-mean-square roughness (R RMS ) of only 1 nm.On the contrary, the surfaces of the perovskite films were rough with R RMS of 34, 27, and 16 nm for I-MAPbI 3 , S-MAPbI 3 , and O-MAPbI 3 films, respectively (Figure 2j-l).This agrees with previous reports that higher PbI 2 concentration generally increases the roughness with microfibers developing on the surface, possibly due to the crystallization of the perovskite and/or PbI 2 . [57,58]Such a nonuniform interface may hinder charge carrier transport and thus requires further improvement. [59]onsistent with the XRD and AFM, the ZnO film shows a smooth nano-crystalline like structure in the SEM image (Figure 2m).Top-view SEM images of ambient air-processed MAPbI 3 films show high surface coverage (Figure 2n-p).The films are polycrystalline in nature and consist of microstructures, characterized with grains and grain boundaries.The S-MAPbI 3 has the largest grains among the three perovskite films, followed by the O-MAPbI 3 .The surface of the I-MAPbI 3 film appears rough, with a large amount of perovskite particles and voids (Figure 2n), which were also observed in previous reports. [10,11,58]urthermore, consistent with the XRD spectrum of the I-MAPbI 3 film in Figure 2n, bright PbI 2 crystals can be seen on the surface.PbI 2 crystals are brighter than perovskite crystals due to the higher average atomic number (selected PbI 2 crystals are circled in red of Figure 2n). [11,58]Cross-sectional SEM images (inset of Figure 2n-p) reveal that the bottom surface of the S-MAPbI 3 film forms a continuous smooth interface with the underlying layer, whereas those of both the I-MAPbI 3 and O-MAPbI 3 films are uneven and discontinuous with their underlying layers.The microstructures of MAPbI 3 films have been found to have profound effects on the properties, performance, and stability of fabricated devices. [60,61]Grain size is also likely to affect crucial properties, such as conductivity and dielectric constant. [61]In the context of high-performance photodetectors, it is important for the grains to extend across the full thickness of the film, as horizontal grain boundaries may impede carrier transport and extraction in the vertical device setting.In addition, grain boundaries can serve as recombination centers due to their high trap densities, resulting in shorter carrier diffusion lengths and lifetimes, which are detrimental to device performance. [15,62,63]n addition, we performed surface composition analysis of the I-MAPbI 3 , S-MAPbI 3 , and O-MAPbI 3 films by X-ray photoelectron spectroscopy (XPS, Figure 3).For the three films shown in Figure 3a-c, the binding energy (BE) of the Pb 4f 7/2 level remained unchanged, indicating that all the films are MAPbI 3 .The Pb 4f 7/2 core level comprises of two components, represented by two peaks centered at 137.9 eV and 136.3 eV, respectively. [64]The former is associated with the Pb-I bond in the MAPbI 3 structure, whereas the latter is associated with the residual metallic Pb (Pb 0 ). [65,66]Lindblad et al. observed similar peaks for the Pb 4f 7/2 core level, with an energy difference of 1.7 eV between them, where metallic Pb 0 is separated from the basic components as an impurity. [66]The peak associated with the residual Pb 0 is also largest in the PbI 2 -rich film (Figure 3a), which is also in good This may introduce recombination centers, thereby degrading device performance by parasitic absorption and consequently limiting the long-term stability of the devices. [10,11,67]Thus I-MAPbI 3 -based phototransistors would suffer from higher levels of nonradiative recombination, a reduction in absorption and charge extraction, and generation of defects than S-MAPbI 3 and O-MAPbI 3 based phototransistors.

Effect of Organic-inorganic Composition CH 3 NH 3 PbI 3 /ZnO Phototransistor Performance
Figure 1a shows a schematic of the bottom gate top contact MAPbI 3 /ZnO photo-TFT under white light illumination, with Figure 1b illustrating the energy level and working mechanism of the MAPbI 3 /ZnO heterostructure.The energy band diagram was deduced from UV-vis absorption and ultra-violet photoemission spectroscopy data, where the perovskite layer absorbs light, which generates electron-hole pairs under illumination and augments the field-effect injection process of the TFT channel, governed by gate-source bias V GS . [68,69]Due to the energy band offset between the conduction band (CB) of the MAPbI 3 and the ZnO films, flow of holes is blocked by the ZnO and only electrons from the MAPbI 3 move along the back-channel of the ZnO layer, resulting in a higher electron concentration, which increases the drain current I D .The high electron mobility of the ZnO film accelerates the electron transfer to the drain contact under drainsource bias V DS , while the holes remain in the valence band of the MAPbI 3 film. [47,48]This characteristic of ZnO has great influence on the performance of the devices, as it keeps the amount of carrier recombination negligible and thereby resulting in a higher photoresponse.
The graphs in Figure 4 present transfer characteristics with forward and reverse sweeps, as well as output characteristics of ZnO and MAPbI 3 /ZnO phototransistors in dark conditions; ZnO monolayers (Figure 4a  I D than phototransistors with ZnO monolayers (Figure 4b).During the formation of the MAPbI 3/ ZnO heterojunction, band potentials at the interface of the two semiconductors are aligned in such a way that the conduction band of MAPbI 3 is more electronegative than that of the ZnO film (Figure 1b). [70,71]As a result, electrons can easily flow from the MAPbI 3 film to the ZnO film, thereby increasing the back-channel component of I D .
Under light illumination, the performance of the three types of phototransistors were evaluated by increasing the power of a white light emitting diode (LED) light source, from 0 (dark) to 4.02 × 10 −4 W cm 2 .Due to the narrow bandgap of perovskites, photo-generation of carriers occurs mostly in the perovskite layer.The band alignment at the interface enables the injection of these photo-generated carriers from the perovskite film into the conduction band of the ZnO film. [71]Due to the energy band offset between the conduction band (CB) of the MAPbI 3 and the ZnO films, it is expected that only photo-generated electrons in the MAPbI 3 film will move along the lateral direction in the backchannel of the ZnO layer, located at the MAPbI 3 /ZnO interface.When the light illumination is combined with positive V GS (i.e., TFT on-state), electrons injected into the top surface of the ZnO film (back-channel) migrate to the bottom surface of the ZnO film (front channel) under the influence of the vertical gate field, induced by applying V GS .Upon reaching the front channel, these electrons drift towards the drain electrode, owing to the presence of the drain to source field (by applying V DS ), resulting in higher electron injection, and consequently, an increase in photocurrent.As shown in Figure 5a-c, all MAPbI 3 /ZnO heterojunction phototransistors exhibited excellent optical characteristics under white light irradiation.However, a slight negative threshold voltage V TH shift is apparent at high illumination power, which is consistent with defect generation.
The small shoulder in the transfer characteristics under illumination indicates the existence of two logical channels in one physical device.The second (parasitic) channel can be caused by many leakage paths in the device.However, given that the ZnO transistor without the perovskite coating does not exhibit this shoulder under illumination (Figure S1a, Supporting Information), it is reasonable to attribute the parasitic channel to the MAPbI 3 film.The parasitic channel occurs near the MAPbI 3 /ZnO interface when the MAPbI 3 film is highly conductive during illumination, and it turns on first before the main ZnO channel, thereby causing the small shoulder in the transfer characteristics.[74][75][76] The effect is more severe in the TFTs with the non-stoichiometric MAPbI 3 films -consistent with the presence of more defects at their MAPbI 3 /ZnO interfaces.Also, defects in the bulk, interface, and grain boundaries of MAPbI 3 , as well as the morphology of the thin film and the interface between the MAPbI 3 /ZnO, can significantly affect their electronic properties.The ideal structure of MAPbI 3 has twelve intrinsic point defects, including vacancies of MA, Pb, and I, interstitial MA, Pb, and I, and six substitutions: MA Pb , Pb MA , MA I , Pb I , I MA , and I Pb , which can induce shallow transition energy levels. [77]I-MAPbI 3 and O-MAPbI 3 have a higher density of defects than S-MAPbI 3 due to excess or deficient cations or anions, respectively.An interesting observation is that off-current (I OFF ) barely changes after light illumination in all devices.This is a clear indication of the absence of hole current in the ZnO channel.The lack of mobile hole carriers in ZnO films has been previously attributed to the presence of a large amount of oxygen vacancy (V O ) defects located just above the valence band.These V O states are intrinsic n-type dopants in ZnO films.They also pin the Fermi level when the TFTs are driven with negative gate voltage biases or trap most of the photogenerated holes during light illumination, resulting in negligible hole current. [33,76]It is important to note that the I OFF measured for these TFTs is mainly due to gate leakage current.
The high photosensitivity of the MAPbI 3 /ZnO phototransistors may be related to the direct bandgap and large absorption coefficients of the MAPbI 3 film, given that the ZnO monolayer yielded negligible photocurrent under white light illumination (Figure S1a, Supporting Information).For an illumination power of 4.02×10 −4 W/cm 2 , I-MAPbI 3 , S-MAPbI 3 , and O-MAPbI 3 -based devices show current levels of ≈10 −6 A at V GS of -0.5 V (i.e., TFT off-state), which is almost 6 orders of magnitude higher than the dark current at the same V GS value (Figure 5a-c).As the MAPbI 3 Perovskite film absorbs a wide range of visible light, the heterojunction phototransistors with MAPbI 3 /ZnO showed high current level characteristics under white light.The phototransistor employing the S-MAPbI 3 film yielded the largest photocurrent, as it contains the highest crystal content, smoothest surface and thus the least number of defects.The photocurrent (I PH ), taken as the difference between the I D in dark (I DARK ) and the I D under light (I LIGHT ) illumination, increased with increasing power intensity of the light source.
Response time of the phototransistor is a highly critical parameter which measures the efficiency of charge transport and collection in the device.Here, we measured the time dependent I D for the three phototransistors fabricated with I-MAPbI 3 , S-MAPbI 3 , and O-MAPbI 3 films (Figure 5d-f).For measurement of I PH , the light source was turned on and off at 10 s intervals with V GS and V DS values of -2.5 and 0.1 V, respectively.The optical response of the devices was stable and reproducible for all three phototransistors.In all cases, the photocurrent quickly increased and returned to the original value as soon as the light was turned on and off (Figure 5d-f), demonstrating that the phototransistors operate as excellent light-activated switches.For the I-MAPbI 3 /ZnO phototransistor, the rise and fall times were 1.2 and 1.4 s, respectively.For the O-MAPbI 3 /ZnO phototransistor, they were 0.8 and 1 s, respectively.The fastest response was observed in the S-MAPbI 3 based phototransistor, reaching record rise and fall times of 0.7 and 0.8 s, respectively.Due to the low-temperature solution processing of the ZnO and MAPbI 3 films, the response time is overall a bit slower than that of standalone MAPbI 3 -based phototransistors, which are processed in vacuum at high temperatures. [8]or the ZnO transistor without a perovskite coating, the rise and fall times were found to be ≈20 and 40 s, respectively (Figure S1b, Supporting Information).The longer fall time is closely associated with a large number of oxygen vacancies and persistent photoconductivity (PPC) effect, which is quite common in solutionprocessed wide band gap OS. [2,27] Given that the response time is a measure of the ZnO channel, these results indicate that the transfer of electrons from MAPbI 3 to the ZnO channel is fastest in transistors with the stoichiometric MAPbI 3 film.Variation in precursor ratio has an impact on crystal formation, defect creation, morphology, and subsequently optoelectronic properties. [77,78]By changing the precursor ratio, there is a high likelihood of introduction of cation or anion vacancies, resulting in defects such as trap states, non-radiative recombination centers, poor charge transport, and slow response.[79] As a result, both I-MAPbI 3 and O-MAPbI 3 have higher defect concentrations than S-MAPbI 3 , which can reduce response times.In addition, excess PbI 2 has a detrimental effect on the performance of optoelectronic devices and can cause parasitic absorption in I-MAPbI 3 based phototransistors. [10,23]Moreover, grain boundaries can negatively affect the performance of perovskite thin films as a barriers toward charge transport by introducing extrinsic defects that negatively affect performance by scattering or trapping of the electrons migrating from the MAPbI 3 into the ZnO. [79]As the transport of photo-excited charge carriers mainly occurs in bands originating from the lead halide framework, any disruption of that framework, by enrichment of grain boundaries with organic ions in the O-MAPbI 3 /ZnO device, may result in slower response. [10,23]S-MAPbI 3 has the best morphology, with a smoother surface (Figure 2j-l), narrower grain boundaries (Figure 2n-p), smallest bandgap (Figure 2b-d), and better crystallinity (Figure 2f-h).Defects, including excess MA and PbI 2 , can scatter electrons migrating from the MAPbI 3 film into the ZnO channel, thereby increasing the response time in TFTs with the non-stoichiometric MAPbI 3 films. [10,23]he main photo-parameters of a phototransistor are the responsivity, photosensivity, and the specific detectivity.In order to assess the uniformity of the performance of MAPbI 3 /ZnO phototransistors employing I-MAPbI 3 , O-MAPbI 3 and S-MAPbI 3 perovskite films, we extracted parameters from 20 different devices for each case.Figure 6 shows the responsivity, photosensivity, and detectivity of the heterojunction phototransistors under increasing illumination power.Upon white light illumination, we observed drastically increases in I D with increasing illumination power, due to the increased carrier concentration in the ZnO channel under illumination.
The responsivity (R), detectivity (D*) and photo-detectivity (P) plotted as a function of the incident light power density shown in Figure 6a-i were extracted from the transfer curves using the following equations: Here, q is the elementary electron charge, E e is the incident illumination power density; and A is the device active area (50 μm wide and 10 μm long).All devices were confirmed to yield excellent photo-parameter characteristics.Phototransistors employing I-MAPbI 3 yielded responsivity ≈128 A/W, photosensivity of ≈10 5 , and detectivity of ≈7.84×10 12 J (Figure 6a-c).The S-MAPbI 3 /ZnO phototransistors, reached a responsivity of 234 A/W, photosensivity of 1×10 5 , and detectivity of 3.74×10 13 J (Figure 6d-f).Finally, the O-MAPbI 3 based phototransistor responsivity, photosensivity, and detectivity reached 146 A/W, 1×10 5 , and 1.83×10 13 J, respectively (Figure 6g-i).The results demonstrate that the best phototransistor performance was achieved with S-MAPbI 3 , with excellent photo parameters, as well as providing reduced variance in device performance, indicating higher uniformity.

Conclusion
We have developed low-cost, ambient air, solution processed MAPbI 3 /ZnO heterostructure TFTs with good optical sensing performance.Precise tuning of the sensitivity can be achieved by controlling the composition of the MAPbI 3 perovskite by varying precursor ratios.We varied the precursor ratio (MAI:PbI 2 ) and found that bandgap and top surface roughness of the MAPbI 3 film decreases with PbI 2 .However, film analysis shows that the S-MAPbI 3 perovskite has the smoothest bottom surface, which is crucial to form a defect free interface with the underlying ZnO layer.While excess PbI 2 increases defect sites for recombination, excess MAI leads to the disruption of the lead halide framework, affecting charge transport.The high mobility of the ZnO channel and the superior optoelectronic properties of the S-MAPbI 3 film enable the achievement of detectivity as high as 3.74×10 13 J under white light illumination.Further improvement in response time can be achieved by reducing the number of defects and the morphology of the films through techniques such as longtime annealing, use of plasma treatments, and passivation layers.Optimizing the device architecture can also improve response time by using thicker MAPbI 3 films for more absorption and smaller channel lengths.We believe that the MAPbI 3 /ZnO phototransistor will promote the development of low-cost and highperformance photodetectors and pave the way for large-scale fabrication of low-cost, low-power, and high performance photodetectors., 99.99%), toluene (anhydrous, 99.8%), 2-methoxy-ethanol (2ME, anhydrous, 99.8%), ethylene glycol (EG, anhydrous, 99.8%), acetonitrile (ACN, anhydrous, 99.8%), and dimethylformamide (DMF, 99.8%) were purchased from Sigma-Aldrich.All chemicals were used without any further purification.

Experimental Section
Solution Preparation: MAPbI 3 solutions were prepared by dissolving precursors MAI and PbI 2 into N,N-Dimethylformamide (DMF) in: 0.8:1 ratio for inorganic rich I-MAPbI 3 ; 1:1 ratio for stoichiometric S-MAPbI 3 ; and 1.2:1 ration for organic rich O-MAPbI 3 .The solutions were then stirred for six hours at 60 °C.The 0.2 M ZrO x solution was prepared by dissolving zirconyl chloride octahydrate into a mixed solvent of ethylene glycol (65%) and acetonitrile (35%).Zinc acetate dihydrate and ammonium acetate were dissolved in 2-Methoxyethanol (2-ME) to prepare a 0.2 M ZnO precursor solution.
Heterostructure Phototransistor Fabrication: To fabricate an array of TFTs on glass substrates, first Molybdenum (Mo) with thickness of 40 nm was sputtered and patterned to form the gate electrode.Next, the solution processed ZrO x (≈24 nm) gate insulator (GI) layer was spin coated and annealed at 350 °C for 2 h in ambient air.Afterward, the ZnO film was deposited by spray pyrolysis at 350 °C.For a uniform film, the nozzle's movement was controlled both horizontally and vertically.≈8 cm distance between the spray nozzle and the substrate was maintained.
Precursor solution flowed at a rate of 3 mL min −1 , and each spray cycle over a substrate of 15 cm by 15 cm lasted 60 s.To obtain the 25 nm thick ZnO film, the procedure was carried out five times.Finally, ZnO film was patterned by photolithography and wet-etched by oxalic acid (C 2 H 2 O 4 ) to form an active island.Subsequently, the ZrO x GI layer was patterned and wet etched with Hydrofluoric acid (HF) to open the contacts.To finish, a 40-nm-thick Mo film was sputtered, patterned by photolithography, and dry-etched to form source/drain electrodes.The channel width and length of fabricated TFTs were 50 and 10 μm, respectively.The MAPbI 3 solution was spin coated at 6000 rotations per minute (rpm) for 70 s in ambient air (relative humidity of 50 ± 5% RH and temperature of 25 ± 2 °C) on top of the ZnO TFT structures.The residual solvents were removed by annealing the samples at 100 °C for 10 min in ambient air for crystallization.No passivation was added to the as fabricated phototransistors.
Film Characterization: For film surface analysis, SEM imaging was conducted, using a Hitachi S-4700 field emission scanning electron microscopy (FE-SEM) (Hitachi High Technologies, Japan).The microscope was operated under the following conditions: voltage 15 kV, current 8-10 nA, beam diameter 6 m, and decreased vacuum in the chamber with the pressure of 50 Pa.The AFM was conducted using XE-7 (Park Systems, Korea).The microscope operated under the following conditions: decoupled XY and Z scanner, where scanning range of XY scanner was 50 μm and Z scanner was 25 μm.Sample scanning area was 5 × 5 μm 2 , mode: noncontact cantilever (NCHR).XRD measurements were conducted using X'PERT PRO of PANalytical Diffractometer with a copper K alpha (Cu K) source (wavelength of 1.5405 Å).X-ray photoelectron spectroscopy (XPS) to determine surface chemical composition was performed using an X-ray source (Al = 1486 eV) and the PHI 5000 VersaProbe spectrometer with peak fitting, performed using Kratos software.Current and voltage were 10 mA and 15 kV, respectively, with an operating pressure of 10 −9 bar.Film thicknesses were measured using a Dektak AlphaStep Profiler.
Device Characterization: For all measurements, the phototransistors were stored in ambient air without a passivation layer, and all currentvoltage (I-V) characteristics were measured by an Agilent 4156C precision semiconductor parameter analyzer at room temperature.The turn-on voltage (V ON ) was taken from the gate-source voltage (V GS ) at which the drain current (

Figure 1 .
Figure 1.Schematics of a) heterostructure phototransistor cross-section, and b) energy level diagram of the MAPbI 3 /ZnO bilayer, deduced from UV-vis absorption and adjusted to the work function of molybdenum (Mo).Photo-generated electrons in the perovskite film can easily transfer to the conduction band of ZnO due to the shallower conduction band minimum of the perovskite.

Figure 4 .
Figure 4. Transfer curves and output characteristics of TFTs with a,b) exclusively ZnO active layer TFT, c,d) ZnO/Inorganic-rich perovskite active layers, e,f) ZnO/Stoichiometric perovskite active layers, and g,h) ZnO/Organic-rich perovskite active layers.The arrows indicate the scan direction for the I D -V GS measurements.The increase of the drain current upon perovskite coating results from the charge transfer from the perovskite to the ZnO.For all devices, the TFT channels are 50 μm wide and 10 μm long.

Figure 6 .
Figure 6.Box and whiskers plots comparing performance of MAPbI 3 /ZnO photo-TFTs with a-c) inorganic-rich, d-f) stoichiometric, and g-i) organicrich perovskite films under increasing illumination power from 0 to 4.02 × 10 −4 W cm −2 .Responsivity, detectivity, and photosensitivity are extracted from the transfer curves of 20 of each kind.
I D ) begins to monotonically increase.The field-effect mobility (μ FE ) and threshold-voltage (V TH ) were derived from the slope and intercept of the linear regression line of the |I D | 1/2 (V DS ) plot in the saturation region (|V DS | > |V GS |-|V TH |), where V DS is the drain-source voltage.The subthreshold swing (SS) was taken as the minimum of (dlog(I D )/dV GS ) −1 .Photoresponsivity was measured with white LED light at intensities of 0.11 × 10 −4 , 1.24 × 10 −4 and 4.02 × 10 −4 W cm 2 .