Hysteresis-free perovskite transistor with exceptional stability through molecular cross-linking and amine-based surface passivation

Please note that technical editing may introduce minor changes to the text and/or graphics, which may alter content. The journal’s standard Terms & Conditions and the Ethical guidelines still apply. In no event shall the Royal Society of Chemistry be held responsible for any errors or omissions in this Accepted Manuscript or any consequences arising from the use of any information it contains. Accepted Manuscript View Article Online


Introduction
Organo-metal halide perovskites have attracted tremendous interest for high-performance, large area, low-cost, optoelectronic devices. [1][2][3][4][5][6] Besides their great success in photovoltaics (PVs) and light-emitting diodes (LEDs), 7-10 they are also highly promising for field-effect transistors (FETs) due to their very large theoretically predicted mobility values that are comparable with those of inorganic semiconductors, such as GaAs. 11,12 However, the demonstration of room temperature operating perovskite FETs that are free of hysteresis and instability issues still remains elusive. [13][14][15][16][17][18][19] The overall performance of perovskite transistors is poor due to several factors that can be divided into intrinsic and extrinsic ones. 20 The intrinsic factors primarily include the migration of ions produced by the imperfections in perovskite stoichiometry and film growth/morphology (i.e. structural defects/grain boundaries). The main extrinsic factors causing perovskite degradation are ambient humidity and oxygen, UV irradiation and elevated temperature. 21,22 The large hysteresis of perovskite FETs occurs due to ions migration and modulation of charge transport by carrier trapping and detrapping at deep energetic sites located at grain boundaries of the polycrystalline film. This also hampers the lateral charge transport over a long distance between contacts through the semiconductordielectric interface reducing charge carrier mobility. 23,24 Additionally, the source-drain contacts are usually aggravated by unavoidable deposition-induced defects and chemical imperfections that largely dictate the interface properties. 25 The intrinsic inability of perovskite to resist attack by moisture induces significant instability during FET's operation, 26 with no published work on stable and hysteresis-free perovskite FETs reported to date. Many recent efforts to solve these issues have been mainly focused on the use of perovskite microplates, 6,8 the optimization of the synthetic procedures, 27 the use of single crystal perovskite layers, 28 the application of solvent vapor annealing, 29 the surface passivation using self-assembled monolayers (SAMs), 30 and interface engineering to improve charge injection. 31 Despite the great merits of these works, the room temperature measured field-effect mobilities remain as low as 0.18 (0.17) cm 2 V -1 s -1 for holes (electrons) for trihalide perovskites and 1.24 (1.01) cm 2 V -1 s -1 for mixed-halide 3D perovskites. Note that, hole mobilities up to 15 cm 2 V -1 s -1 and electron mobilities up to 10 cm 2 V -1 s -1 have been recently reported for FETs based on the 2D layered (C 6 H 5 C 2 H 4 NH 3 ) 2 SnI 4 , 32 and the 3D CH 3 NH 3 PbI 3-x Cl x perovskites, 33 respectively. However, these transistors were measured under high vacuum of at least 10 -4 Torr. For practical device applications in complete circuits, efficient transistors capable of operating in ambient conditions are necessary. More important, steady improvements in the device hysteresis and stability issues have not been achieved to date thus still puzzling the community. Arguably, a lack of sustained stability and the presence of hysteresis are the biggest obstacles on the further development of perovskite transistors. Recently, our group demonstrated the first perovskite FET with balanced room temperature ambipolar transport by using a triple cation perovskite of the chemical structure Cs x (MA 0.17 FA 0.83 ) 1-x Pb(Br 0.17 I 0.83 ) 3 , where Cs is caesium, MA and FA are methylammonium and formamidinium, respectively, as the semiconducting channel material and a PMMA film as the gate dielectric. 34 These FETs featured both hole and electron mobilities around 2 cm 2 V -1 s -1 as the outcome of improvements in crystallinity linked to stabilization of the perovskite crystal structure upon the incorporation of an optimized Cs content. 35 However, hysteresis and instability issues still governed the operation of these transistors. Here, we report the first hysteresis-free perovskite FET with exceptional stability under bias stress and upon aging in air. Moreover, we obtained a nearly 100% improvement compared to our previous work in the measured hole (electron) mobilities up to 4.02 (3.35) cm 2 V −1 s −1 and a low subthreshold-swing (SS) of 267 mV dec -1 which is the best values reported to date for ambient conditions operating perovskite transistors. This became possible through the implementation of a molecular cross-linking strategy of the perovskite grains by the simple addition of a hydrophobic crosslinker namely diethyl-(12-phosphonododecyl)phosphonate (DPP) in the perovskite precursor followed by effective surface passivation of perovskite film with a thin amine-bearing polyethylenimine ethoxylated (PEIE) interlayer. A tank of experiments and density functional theory (DFT) simulations confirmed the molecular passivation principle in the form of strong hydrogen bonding interactions and high adsorption energy between cross-linker and perovskite.

Experimental section
Materials. The organic cations were purchased from Dyesol; the lead compounds from TCI; CsI from abcr GmbH. The "mixed" perovskite precursor solutions were deposited from a precursor solution containing FAI (1 M), PbI 2 (1.1 M), MABr (0.2 M) and PbBr 2 (0.2 M) in anhydrous DMF: DMSO 4:1 (v:v). Then CsI, pre-dissolved as a 1.5 M stock solution in DMSO, was added to the mixed perovskite precursor to achieve the desired triple cation composition. Fabrication and characterization of FET devices. 1 M of DPP was added into the triple cation composition and kept stirred for 2 hrs. A DPP modified-CsMAFA was deposited by two-step spin coating method on an n-doped silicon substrate (ρ~2 Ω cm) which was covered with 300 nm thick SiO 2 in a two steps program at 1000 and 6000 rpm for 10 and 30 s, respectively in a nitrogen-filled glove box. During the second step, 100 μL of chlorobenzene (CB) was poured on the spinning substrate 15 s prior to the end of the program, to assist the perovskite film crystallization. The thickness of modified-CsMAFA layer is approximately 250 nm. For the as-spun transistors, no annealing treatment of the samples was performed. For the annealed transistors, the samples were annealed at 100 °C for 20 min. Prior to these processes, silicon substrates were sequentially cleaned in acetone, toluene and isopropyl alcohol for 15 min each, later dried in an oven at 120 °C for 15 min, then exposed for 1.5 h to UV light in air before finally the surface was treated with 3 mM noctadecyltrimethoxysilane (OTS) self-assembled monolayer (SAM) for 0.5 h by spin-coating on the piranha-solution-cleaned wafer at 3000 rpm for 30 s. Then, the wafer was exposed to ammonia vapor for ~12 h, followed by sonication cleaning, sequential washing, and drying. After the perovskite deposition, a thin layer of PEIE was spin-coated at 4000 rpm for 45 s and later annealed at 100 °C for 10 min. The source and drain electrodes made of yttrium fabricated by e-beam lithography and evaporation, followed by a standard lift-off process. The yttrium oxide film with a thickness of 60 nm was deposited by reactive e-beam evaporation. Lastly, Ti/Au (10 nm/30 nm) metal stack was evaporated as the top gate. To check the moisture resistance ability, two water molecules are adsorbed over the relaxed surface of unmodified-CsMAFA and modified-CsMAFA, as shown in Figs. 5b and c. In addition, the interaction of O 2 with these modified and unmodified surfaces, two O 2 molecules are adsorbed over the relaxed surfaces of unmodified-CsMAFA and modified-CsMAFA, shown in Figs S23 a,b. Device measurements. All electrical measurements were monitored by an Agilent 4156 C semiconductor parameter analyzer. FETs' current-voltage characteristics are measured in the dark and at room temperature under ambient pressure, where the voltage scan rate was at 5 mV s -1 . The field-effect mobility (μ) was determined using , where I DS is the drain-to-source current, μ is the mobility, and V G and V T are the gate voltage and threshold voltage, respectively. Measured samples had channel lengths L=50 μm and channel widths W=1000 μm. Characterization tools. The scanning electron microscopy was conducted using a SEM instrument, Hitachi S-4700. X-ray diffraction (XRD) measurements were carried out using X'PERT PRO of PANalytical Diffractometer with a Cu Kα source (wavelength of 1.5405 Å). Photoluminescence (PL) spectra were recorded using a lock-in technique with JASCO FP-6500 composed of two monochromators for excitation and emission, a 150 Watt Xe lamp with shielded lamp house and a photomultiplier as a light detector. Absorption spectra were collected with a Varian Cary 300 UV-Vis spectrophotometer with an internally coupled integrating sphere. The film thicknesses were measured using a Dektak AlphaStep Profilometer. Contact angle measurements were performed using an Attention Theta optical tensiometer with automated liquid pumping system was used for the contact angle measurements. Purified (Milli-Q) and degassed water was used as the probe liquid. Proton nuclear magnetic resonance ( 1 H NMR) was performed using Bruker Advance 300 ( 1 H: 300 MHz) spectrometer at 298 K using partially deuterated solvents as internal standards.

Results and discussion
Molecular cross-linking of perovskite grains. For the fabrication of our FETs we used the bottom-contact/top-gate (BC/TG) configuration ( Figure 1a) despite the fact it is generally difficult to achieve high performance using this geometry. This is because, charge carriers injected from the source (S) and drain (D) electrodes into the semiconductor channel need to travel typically several tens of nanometers normal to the substrate before they reach the semiconductor/gate dielectric interface where controlled charge transport takes place. In addition, the bulk resistance on the path between the semiconducting channel and the S-D contacts is very large thereby significantly increasing the total resistance. We, however, adopted this transistor configuration because it is more realistic for practical applications. The semiconducting channel material consisted of the previously introduced triple cation perovskite Cs x (MA 0.17 FA 0.83 ) 1-x Pb(Br 0.17 I 0.83 ) 3 (CsMAFA) where x=0.05 with the stark difference that it was subjected to our molecular cross-linking methodology. In particular, the perovskite film was prepared by spin-coating a mixture of the triple cation perovskite and a well-known hydrophobic cross-linker namely diethyl-(12phosphonododecyl)phosphonate (DPP) (Figure 1b and Figure S1, Supporting Information). Motivated by the molecular passivation and molecular cross-linking approaches successfully applied to perovskite PVs and LEDs, [36][37][38][39][40][41] we hypothesized that it should be possible to address the perovskite FET issues by cross-linking the perovskite grains within a framework of a uniform material. To achieve that endeavor, we viewed a DPP molecule bearing phosphonic groups as a potentially suitable cross-linker due to the well-established strong hydrogen bonding coordination of the above groups to the unsaturated bonds. 42 The phosphonic groups present in both sides of each molecule are expected to bond to the periphery of the PbX 6 octahedra forming hydrogen bonds with the halide anions of the perovskite (Figure 1c) and, therefore, crosslinking neighbouring perovskite grains. To scrutinize the formation of hydrogen bonds between the perovskite surface and the crosslinker we performed 1 H NMR measurements. The phosphonic ester group in the pristine DPP solution appears at around 3.95 (CH 2 ) and 1.25 ppm (CH 3 ) (Figure 1d). It is well-known that hydrogen bonding produces upfield chemical shifts caused by shielding. The phosphonic signals of DPP appear in the mixed DPP-CsMAFA solution where they undergo a small upfield shift suggesting the formation of hydrogen bonds between the molecular additive and the perovskite. Our conjecture that the pendant phosphonic groups are integrated at the surface of the perovskite grains was further probed by Fourier transform infrared (FTIR) spectroscopy. The main stretching bands of DPP in the DPP+perovskite mixed film ( Figure  S2) exhibit a significant downward shift with respect to those of the pure DPP, indicating strong interaction between the perovskite matrix and the cross-linking additive.
Structural and morphological studies. The powder XRD pattern (PXRD) taken in DPP modified-CsMAFA material ( Figure S3) indicates that the nanocrystals derived from the mixed solution are composed of neat perovskite phase suggesting that the cross-linker molecules coordinate exclusively at the surface of perovskite acting as "glue" that cross-links the neighbouring grains and do not occupy the perovskite lattice. Such coordination bonding also results in excellent defect passivation as evidenced by the large enhancement in photoluminescence (PL) of the DPP-CsMAFA film compared to the pristine perovskite ( Figure S4). Our experimental data provide evidence for the coordination of pedant phosphonic groups to the surface of perovskite grains through the formation of hydrogen bonds. As a result, effective defect passivation was realized. Moreover, molecular cross-linking of neighbouring perovskite results in crystallinity enhancement of the DPP modified perovskite film compared to the unmodified one as indicated from XRD and differential scanning calorimetry (DSC) measurements ( Figure S5 and S6 and Table S1). We also performed elemental analysis of the mixed DPP-perovskite film by scanning transmission electron microscopy (STEM). In Figure 1e the high-angle annular dark-field (HAADF) image and EDS mapping in STEM mode are presented confirming the presence of DPP in the sample. The 3 mol % DPP molecules used here easily led to the detection of P in our sample. From STEM-EDS mapping of Pb and P, we infer that these two elements exhibit similar distributions in the mixed sample.
Top-view ( Figure S7) and cross-sectional SEM images (Figure 1f) revealed that the pristine perovskite film exhibits a large number of dispersed grains that are loosely bound and randomly distributed within the film. On the contrary, the modified-CsMAFA films are very uniform and smooth with tightly connected grains. This also allowed the formation of a high-quality interface between the perovskite film and the gate dielectric deposited directly on top of the former (Figure 1g). These findings are of great importance as the structural uniformity of the perovskite layer dictates charge transport ability and interface quality that are highly important for superior FET operation.  (Table S2). As shown in Fig. 2b and 2d, the DPP molecules are packed almost perpendicular to the surface of perovskite, through phosphonic anchoring groups. The electronic total charge density slices (Figures 2c,d) predict that the anchoring groups of DPP strongly interact with the FA + and MA + , besides the halide groups. The strong hydrogen bonding interaction between perovskite and DPP (Figure 2b) not only cross-link the perovskite grains but improves its charge mobilities as the result of a passivation effect. This can be also concluded from its simulated bandgap, UV-Vis spectra and effective masses of charge carriers (Figure 2c). More specifically, the bandgap of unmodified-CsMAFA decreases from 1.58 to 1.44 eV, upon adsorption of DPP molecules (modified-CsMAFA) in accordance with the experiment (Figure S8), and the effective mass of electrons substantially decreases from 2.98 to 0.16 m e , thus resulting in high charge carrier mobility. Effective masses of electrons and holes are estimated from the conduction

Please do not adjust margins
Please do not adjust margins band minimum (CBM) and valence band maximum (VBM) of the band structure, respectively, which are listed in Table S3. The lighter the masses of these charge carriers, the higher will be their charge mobilities slowing down charge recombination. Our theoretical results (Figure 2e,f) are in good agreement with the experimental data demonstrating that effective molecular cross-linking of perovskite grains causes desirable changes in optical and electronic properties of the modified perovskite. Despite the fact that not all of these changes (i.e. improvement in UV-Vis absorption) have direct impact on the FET performance, they demonstrate the universal impact of our approach in improving the optoelectronic properties of perovskite materials for several types of device applications (i.e. perovskite PVs).

Improvements in mobility and elimination of hysteresis.
Except for the semiconducting channel material, the choice of the gate dielectric and source-drain contacts is of paramount importance for the satisfactory FET operation. The polymer dielectrics commonly used in perovskite FETs undergo severe physical aging deteriorating the device stability. In addition, their dielectric constants (k) are not sufficiently high (4.9 for PMMA and 2.1 for Cytop) to increase the device capacitance at low thicknesses and thus to scale-down the device dimensions. Moreover, the commonly used gold (Au) contacts exhibit chemical interactions with the perovskite film thus altering the interface properties. We therefore replaced the reactive gold with inert yttrium (Y) contacts and the 550 nm thick PMMA dielectric with a 60 nm thick yttrium oxide (Y 2 O 3 ). The latter possesses a high-k value equal to 16-19, a defect-free electronic structure and an ultra-fine nanomorphology. 43,44 Actually, Y 2 O 3 has been considered as an ideal gate dielectric especially for transistors using carbon containing materials due to the excellent wetting behavior of yttrium on sp 2 carbon framework. 45 Initially, we performed measurements in a perovskite FET using unmodified CsMAFA channel. The output characteristics of this unmodified transistor in both hole (Fig. S9 a) and electron (Fig. S10 (Table S4). There is a significant unbalance between hole and electron mobilities. The transistor subthreshold swing (SS) values were found equal to 532 and 502 mV dec -1 for hole and electron operation, respectively. These are much lower than the SS values usually reported for perovskite transistors irrespective of the configuration used (above 1000 mV and usually few V dec -1 ). 46 We next fabricated FETs using perovskite channels embedding 3% per mol DPP to act as the cross-linker. The hole output and transfer characteristics of the modified FET are shown in Fig. S9 b and Fig.  3b. The electron output and transfer characteristics are presented in Fig. S10 b and Fig. 3d, respectively. The device performance metrics are summarized in Table S5. An increase in hole mobility up to 3.97 cm 2 V −1 s −1 and a more remarkable enhancement in electron mobility up to 3.20 cm 2 V −1 s −1 is achieved upon the addition of DPP cross-linker within the perovskite (Table S5). Note that this increase was dependent on the DPP concentration ( Figure S11). To understand these improvements in mobility values we have to take into account that except for the externally applied gate voltage an additional internal one is also applied in the semiconducting channel due to the mobile ions present on the perovskite film. This internal field depends on the intensity and the polarity of the gate voltages that were applied during previous measurements. Due to the random movement of ions through the lattice, this internal field is also random and combined with imperfections at the dielectric/perovskite interface (e.g., dangling bonds and interface roughness) will cause the formation of a fringing field, which causes significant scattering to the carriers moving across the channel thus reducing the measured mobilities. The primary culprit is the movement of ions through the perovskite films; under the application of a negative V GS , these positive ions gradually move through grain boundaries and accumulate at the perovskite-gate dielectric and perovskite-contacts interfaces. They thereby act as scattering centers and reduce the number of charges within the channel through Coulomb repulsion resulting in significant hysteresis and loss of stability. The ion movement can also explain the dependence of threshold voltage (V t , in both, p-and n-channel) on the annealing temperature of the perovskite film. A different thermal annealing result in different degree of structural transformation which influences the production of mobile ions in the perovskite film thus affecting the intensity of the internal field and, therefore, shifts the measured V t . However, even in the optimized annealing of 100 o C these mobile ions were not eliminated thus reducing the performance of unmodified transistor. By applying our cross-linking modification approach on the perovskite we effectively reduced the density of defects present at grain boundaries hence gaining a significant enhancement in the device performance. Notably, the electron mobility is more strongly affected by the concentration of the cross-linker verifying that upon the addition of DPP the positive ions that may repel electrons are highly suppressed. Additionally, the SS values were found even lower and equal to 288 and 267 mV dec -1 for the p-and n-channel, respectively. These are the lowest values ever reported for perovskite transistors and the closest to the room temperature Boltzmann theoretical limit of 60 mV dec -1 . However, from the transfer characteristics in Fig. 3a-d, it becomes evident that, even though decreases compared to the unmodified transistor, the hysteresis of the modified FET is still present.
Previous reports indicate that hysteresis in perovskite FETs is due to a combined effect of charge trapping and ion transport, and thus decreasing structural and ionic defects in bulk perovskite films could both decrease hysteresis. [27][28][29] Sirringhaus's group also demonstrated that the perovskite surface defects also contribute to hysteretic issues. 17 We, therefore, applied a perovskite surface passivation approach by spin coating an amine-containing nonconjugated polymer namely polyethylenimine ethoxylated (PEIE) to reduce the density of surface defect states and dangling bonds due to well established passivation ability of amine groups. Indeed, it has been previously demonstrated that PEIE can effectively modify the surface of perovskites by forming strong hydrogen bonding interactions with surface defects. 47 Fig. 3e,f presents the hole and electron transfer characteristics for the DPP-modified and PEIE surface treated perovskite FET where complete elimination of hysteresis is observed. In fact, this is the first hysteresis-free perovskite transistor reported in the literature. The complete elimination of hysteretic issues was irrespective of the perovskite post-annealing temperature. Moreover, the room temperature mobility was further improved to 4.02 (hole) and 3.35 (electron) cm 2 V −1 s −1 (Table S6). These are the highest mobilities reported thus far for any perovskite transistor operating under ambient pressure at room temperature (Table S7). This is expected to have a profound impact on the use of perovskite transistors in complementary logic elements requiring n-and p-type semiconductors.
Stability study. Furthermore, an outstanding increase in both aging and operational stability under stress was obtained in the DPP modified and PEIE passivated FET as seen from the variation of hole and electron mobility of the unmodified and DPP-modified FETs for over sixteen days as the aging period ( Figure S12). Remarkably, the mobility (both hole and electron) of the modified FET remains nearly unchanged. However, a large decrease in hole and mainly in electron mobility is obtained for the aged unmodified FET which is attributed to the continuous accumulation of positive ions at the surface or grain boundaries of the perovskite film upon aging. The dramatic increase in the intensity of the PbI 2 peak in the XRD spectrum of the aged unmodified perovskite (Figure 4a) provides evidence for the well-known degradation process of the perovskite film through hydrolysis that produces PbI 2 upon the release of gas phase HI and CH 3 NH 2 species. 48,49 On the contrary, the PbI 2 peak does not appear in the XRD pattern of the aged modified perovskite (Figure 4b) providing evidence for the strong resistance of the molecular cross-linked perovskite on attack of moisture as expected due the hydrophobic nature of the DPP-modified film. Indeed, upon the addition of the hydrophobic crosslinker within the perovskite film, the latter becomes less hydrophilic ( Figure S13) hence being more resilience to environmental water molecules that cause the undesired perovskite hydrolysis and device degradation. SEM images of the aged unmodified and modified perovskite films (Fig.  4c,d) reveal the stronger resistance to environment induced degradation of the modified-CsMAFA film compared to the unmodified one.
Please do not adjust margins Please do not adjust margins Bias stress driven electrical stability of FETs is extremely important as it dictates the normal operation of an FET-based electronic circuit. Figure 4e presents the channel current of the modified FET under a continuous bias stress of -20 V for over 80,000 seconds (the inset shows the I SD -V GS curve taken before and after bias stress) whereas Figure S14 presents the bias stress stability of the unmodified transistor.  Figure 4f shows the cycle stability of the device, where a train of gate voltage pulse (-20 V) was applied and the device was switched between on and off for 50,000 cycles (1 Hz). Inset shows the I SD response from the 15,000 th until 30,000 th seconds. Such measurements were not possible in the unmodified FET due to its large instability attributed to charge carrier trapping and ion migration inside the device originated from uncontrolled microstructure of the perovskite film. By forming a cohesive and compact and passivated perovskite layer, charge carrier trapping was successfully eliminating thereby allowing for the achievement of unprecedented bias stability of the DPP-modified PEIE passivated device.
DFT calculations on hysteresis and stability. Severe hysteresis is seen for unmodified-CsMAFA devices which is largely alleviated after inclusion of DPP molecules. As structural defects and ion migration are responsible for the hysteretic behavior, 2% Idefective (vacancy) unmodified-CsMAFA and modified-CsMAFA are employed as models for theoretical simulations ( Figure S15). The simulated formation energy of modified-CsMAFA is 0.46 eV higher than that of unmodified-CsMAFA, thus validating our experimental results (Figure 5 a,b and Table S8). It is known that water adsorption on the perovskite surface can be hardly prevented, followed by water infiltration into the subsurface of the perovskite network, which begins the hydrolysis process in bulk perovskite. A DFT calculation was again used to evaluate the interaction between water molecules and the perovskite lattice. In the modeling of atomic structures, the surface Cs(Ma/FA)(Br/I) molecules were replaced with DPP, and the stabilized molecular configurations are shown in Figure a,b. We place two water molecules between the DPP molecules, unmodified-CsMAFA and modified-CsMAFA framework. The corresponding formation energy is -0.49 eV per molecule for modified-CsMAFA, and -0.01 eV per molecule for unmodified-CsMAFA; confirming the thermodynamic stability of the DPP molecules. To verify whether the molecular crosslinking possibly decreases water adsorption, we calculate the water adsorption energy (E ad ) on the surface of the perovskite films. An adsorption energy of -107 kcal mol -1 between water and unmodified-CsMAFA (Figure 5c), confirmed its poor stability against moisture. However, modified-CsMAFA has lower interaction energy with water (E ad -36.44 kcal mol -1 ), which is almost one-third of the unmodified-CsMAFA-water system, given in Figure 5d and Table S9.
This lower E ad further validates the DPP passivation principle in the form of improved moisture resistance as can be seen from the electron total charge density slices (Figure 5e,f and Table S10), which implies that DPP molecules efficiently keep water molecules away from the perovskite lattice. Analysis of Figure 5c,d led us to conclude that the unmodified-CsMAFA has attracted water molecules through sharing electronic cloud density, compared to that of modified-CsMAFA where the water molecules are repelled.
The notably reduced E ad in modified-CsMAFA indicates that the DPP molecules have better water repelling capability as observed from the electron total charge density slices (Fig. 5e,f). As we know, the extrinsic factors causing perovskite degradation are not only the moisture, but oxygen as well. We have also simulated the interaction of oxygen molecules with unmodified-and modified-CsMAFA, as can be visualized from Figure S16a, b. along with their electronic total charge density slices ( Figure S16c, d). In case of unmodified-CsMAFA, the oxygen atoms of O 2 have about 3.30 Å, inter-atomic distance with the surface atoms. However, this distance is longer in modified-CsMAFA, which is about 2.44 Å. The per O 2 molecule interaction (adsorption) with unmodified-and modified-CsMAFA is -286.53 and -104.58 kcal mol -1 , respectively (see Table S2). Again, it is pointed out that DPP-passivated CsMAFA repel both water and O 2 molecules which further validate and confirm the stability of our modified perovskite.

Conclusions
In this work, we demonstrate the first hysteresis-free, extremely stable perovskite FET with balanced ambipolar transport under ambient pressure at room temperature. Added to the merits, our transistor exhibits high hole and electron mobilities and the lowest SS value of 267 mV dec -1 reported to date. To achieve those goals, we applied a combinatorial strategy including molecular crosslinking of the perovskite grains that enabled the formation of a cohesive and compact film and effective surface passivation of the perovskite channel. DFT calculations confirmed the molecular crosslinking and passivation principle. High adsorption energy and strong hydrogen bonding between DPP and perovskite confirm their excellent interaction and stability. That ingenious approach successfully addresses the severe hysteresis and instability issues of perovskite transistors and paves the way for the fabrication of enhanced performance perovskite FETs that would be highly suitable for the future realization of FET-based products and CMOS circuits.