Vapour-deposited high-performance tin perovskite transistors

Tin (Sn²⁺)-halide perovskites—such as caesium tin triiodide (CsSnI₃)^1,2, methylammonium tin triiodide (MASnI₃)³ and formamidinium tin triiodide (FASnI₃) (refs. ^4,5)—are of potential use in the development of high-performance p-channel thin-film transistors (TFTs)^6,7,8,9. Such devices could, in particular, find practical application in next-generation complementary electronics when integrated with commercial n-channel metal-oxide technology. The materials have intrinsic p-type semiconducting properties, as well as low effective mass (m*), low Fröhlich interaction, weak ion migration and intrinsically high hole mobilities (µ) (ranging over hundreds of square centimetres per voltage second)^10,11,12. To apply the perovskites as channel layers in TFT applications, it is crucial to modulate the excessive hole concentration and control the crystallization process to extend the scattering time (τ) (µ = qτ/m*, where q is the elementary charge)^13,14. Composition engineering methods have been used to regulate the nucleation and crystallization of solution-processed Sn²⁺-halide perovskite precursors, allowing TFTs to achieve high hole field-effect mobilities (µ_FE) over 50 cm² V⁻¹ s⁻¹, rivalling that of low-temperature polysilicon devices^{5,15,16,17,18}.

Solution processing has been the primary deposition technique for Sn²⁺-halide perovskite thin films. This method can provide rapid optimization trials in a cost-effective manner. However, it has low compatibility with conventional manufacturing processes for flat-panel displays and semiconducting devices^19,20,21. An alternative method is vapour deposition, which is at the forefront of commercialized thin-film electronics, providing precision, simplicity and compatibility with existing production infrastructures^22,23,24. Vapour deposition can allow precise manipulation of thin-film thickness and morphology, which have a critical impact on device performance. Its simplicity in terms of thin-film deposition can provide notable batch-by-batch reproducibility, as well as uniformity in large-scale production^25,26,27. However, owing to the difference in crystallization processes between solution processing (fast ionic reactions) and vapour deposition (slow solid reactions), it is still a challenge to achieve high-quality, high-mobility Sn²⁺-perovskite channels with suitable hole density by means of vapour deposition.

In this article, we report vapour-deposited inorganic CsSnI₃-based p-channel TFTs using lead chloride (PbCl₂) as an additive. The volatile chloride can initiate solid-state reactions which extend to complete perovskite phase formation, promoting densely packed enlarged grains in a cascaded manner. Our optimized TFTs exhibit µ_FE of more than 33.8 cm² V⁻¹ s⁻¹ and on/off current ratios (I_on/I_off) greater than 10⁸, as well as long-term stability. We also fabricate large-scale uniform Sn²⁺-halide perovskite TFT arrays, overcoming the technical challenges previously identified in solution processing.

To deposit the CsSnI₃-based perovskite channel layer, a sequential deposition method was used for each component—PbCl₂, SnI₂ and CsI—to avoid contamination and ensure precise control of the atomic ratio (Fig. 1a). PbCl₂ was deposited as the initial layer on the 100-nm SiO₂ dielectric substrate, followed by the sequential evaporation of SnI₂ and CsI layers. The deposition of PbCl₂ in the bottom layer plays a key role in achieving high-quality film deposition, which will be discussed later. The multilayer films were then annealed at 320 °C to merge and form a single CsSnI₃ layer (Fig. 1b and Supplementary Fig. 1). The sample with PbCl₂ will hereafter be referred to as the CsSnI₃:PbCl₂ film. The final step involved the deposition of Au source/drain electrodes to construct bottom-gate, top-contact TFTs (Fig. 1c). For the pristine CsSnI₃ TFT, a metallic behaviour with poor field-effect modulation was observed (Fig. 1d,e). This phenomenon indicates a high hole concentration (n) in the CsSnI₃ channel, measured to be ~10¹⁸ cm⁻³ using Hall measurements.

a–c, Fabrication of bottom-gate, top-contact vapour-deposited CsSnI₃-based TFT. Sequential evaporation of perovskite compounds (a), post annealing treatment of the perovskite multilayers (b) and evaporation of Au electrodes (c). d, Field-effect mobility and on/off current ratio as a function of PbCl₂ atomic ratio. The average (centre) and standard deviations (error bars) were obtained from sample size n = 10. e, Transfer characteristics of CsSnI₃ TFTs and optimized CsSnI₃:PbCl₂ TFTs (Cs_1.3SnPb_0.2I₃Cl_0.15). I_G indicates gate leakage current. Channel length/width = 100 µm/200 µm (V_DS = −40 V). f, Output characteristics of optimized CsSnI₃:PbCl₂ TFTs. g, Consecutive forward scans of transfer characteristics for 100 cycles (V_DS = −40 V). h, Negative bias-stress measurements for CsSnI₃ and CsSnI₃:PbCl₂ TFTs (V_DS = V_GS = −40 V) for 2,000 s. i, Benchmark of representative µ_FE and I_on/I_off values of reported vapour-deposited perovskite TFTs based on different channel materials.

Full size image

Owing to the low formation energy of tin vacancy (V_Sn) in Sn²⁺-halide perovskite, the deposited three-dimensional Sn²⁺-halide perovskite films typically contain excessive hole concentrations, making the field-effect modulation challenging for transistor applications^28,29. To address this issue, the use of tin fluoride (SnF₂) (~10 mol.%) has been developed as an indispensable additive in solution-processed Sn²⁺-halide perovskite optoelectronic devices^5,30. The low solubility of SnF₂ provides a synergistic effect, enhancing nucleation for better crystallization and V_Sn passivation³¹. However, we found that the efficacy of SnF₂ differs in vapour deposition, where it does not beneficially affect the modulation of hole concentration or crystallization (Supplementary Fig. 2). This discrepancy can be attributed to the different crystallization behaviours in vapour deposition: the thermally driven solid-state reactions of sequentially deposited precursor layers, without a solvent to aid in molecular dissociation^32,33,34,35. Thus, bond dissociation energy is critical to initiate these solid-state reactions^36,37. The high electronegativity of fluorine makes Sn–F bonds particularly strong, which makes them difficult to dissociate during the solid-state reaction³⁸. Therefore, a different compound with sufficiently low bond dissociation energy is needed to actively participate as a hole suppressor and/or to aid crystallization. A promising candidate is Pb–X (X = I, Br, Cl), which has a lower bond dissociation energy than Sn–X due to its larger atomic size, longer bond length and thus weaker bond energy. In addition, Pb is considered an optimal alternative to Sn owing to its similar coordination environment. The greater number of inert 6s electrons of Pb can also increase the V_Sn formation energy and effectively lower the hole concentration³⁹. PbCl₂ has been introduced as a beneficial additive in thermal evaporated Pb-based perovskite solar cells. Partial substitution of PbI₂ with PbCl₂ induced formation of Cl-alloy, facilitating the perovskite phase transition and enhancing film crystallinity by promoting the diffusion and evaporation of Cl in the form of organic ammonium chloride^34,40. Furthermore, studies on Cl-alloyed tin oxide as an electron-transporting layer have demonstrated that evaporation of Cl during annealing promotes uniform growth of large, vertically oriented grains^41,42,43.

We then tested the effect of PbI₂, PbBr₂ and PbCl₂ on the electrical performance of CsSnI₃ TFTs (Fig. 1 and Supplementary Fig. 3). Hereby, the atomic percentage (at.%) of each element in the final film (CsSnI₃:PbX₂) was calculated from the area integration throughout the film thickness by X-ray photoelectron spectroscopy (XPS) depth profile, relative to tin (Supplementary Table 1). Interestingly, although PbI₂ shows a slight hole suppression effect, the device failed to reach a reliable off-state current (I_off) even with 30 at.% addition of PbI₂. By contrast, PbBr₂ and PbCl₂ demonstrated notable hole suppression properties, achieving a reliable I_off with the addition of 20 at.% PbCl₂ and 30 at.% PbBr₂. Because PbCl₂ demonstrates the highest efficiency as a hole suppressor and device performance enhancer, we continued to investigate the association of varied PbCl₂ amounts on the electrical properties of CsSnI₃:PbCl₂. Increasing PbCl₂ from 10 at.% to 20 at.% resulted in a reliable I_off, elevating the I_on/I_off from ~10² to ~10⁸ (Fig. 1d). However, further increasing the PbCl₂ molar ratio to 25 at.% degraded the TFT performance by lowering the current level and the µ_FE. The optimal addition of 20 at.% PbCl₂ achieved high TFT electrical performance, including a high µ_FE of 33.8 cm² V⁻¹ s⁻¹, a high I_on/I_off of 8.2 × 10⁸, and a small subthreshold swing value of 0.5 V dec⁻¹ (Fig. 1e). The corresponding output characteristics of the optimized CsSnI₃:PbCl₂ TFT exhibit good linearity at low drain-source voltage (V_DS) and current saturation at high V_DS, indicating a low charge-carrier injection barrier and ohmic contact between the channel layer and electrodes (Fig. 1f). The contact resistance (R_c) was measured at ~57 Ω cm, which is lower compared with previously reported solution-processed CsSnI₃-based TFTs at ~1 kΩ cm, suggesting enhanced quality of the channel/contact interface (Supplementary Fig. 4). Reliable device operation was also verified by the negligible difference in transfer curves after scanning current–voltage characteristics for 100 repetitive cycles (Fig. 1g). To evaluate the PbCl₂ additive effect on device stability, a negative-bias-stress test was performed on CsSnI₃ TFTs and CsSnI₃:PbCl₂ TFTs. The results clearly show that the CsSnI₃:PbCl₂ TFT has much less performance degradation compared with CsSnI₃ TFTs (Fig. 1h), which indicates improved perovskite channel quality and lower defect density. To benchmark the electrical parameter of our vapour-deposited CsSnI₃:PbCl₂ TFTs compared with previously reported perovskite TFTs, we listed representative studies in Supplementary Table 2 and summarized the key data in Fig. 1i. The achieved µ_FE and I_on/I_off represent the state-of-the-art values for the vapour-deposited perovskite device and are much superior to the majority of solution-processed devices.

We then performed film characterizations from the surface to bulk to understand the mechanism behind the formation of CsSnI₃ in vapour deposition and the unique role of PbCl₂. The scanning electron microscopy (SEM) images of CsSnI₃:PbX₂ films illustrate a distinctive difference from pristine CsSnI₃ to PbX₂-incorporated films (Fig. 2a). The pristine CsSnI₃ film exhibited 1–2-µm sized grains with grain-sized pinholes and sharp white aggregates, indicating uneven reaction of the as-evaporated perovskite compounds. CsSnI₃:PbBr₂ and CsSnI₃:PbI₂ films showed improved uniformity with reduced pinholes, although notable surface particles remained. Energy dispersive spectroscopy mapping of CsSnI₃, CsSnI₃:PbBr₂ and CsSnI₃:PbI₂ films identified the white aggregates to be unreacted precursor compounds of CsI (Supplementary Fig. 5). By contrast, the CsSnI₃:PbCl₂ film exhibited much improved morphology with high film homogeneity and large grain sizes exceeding 5 µm. The grain boundaries were vaguely visible, connecting each grain in a cascade style. The corresponding surface root mean square value of CsSnI₃:PbCl₂ film was 9.95 nm, which was much lower than those of pristine, and Br- and I-incorporated films, which were 42.8, 16.7 and 15.2 nm, respectively (Fig. 2b and Supplementary Fig. 6).

a, SEM images of of CsSnI₃, CsSnI₃:PbCl₂, CsSnI₃:PbBr₂ and CsSnI₃:PbI₂ thin films. b, Root mean square values extracted from atomic force microscopy (AFM) images. The centre represents the mean value and the error bars represent the standard deviation of 10 AFM image samples (n = 10). c, XRD spectra of CsSnI₃, CsSnI₃:PbCl₂, CsSnI₃:PbBr₂ and CsSnI₃:PbI₂ thin films. Inset shows zoomed-in XRD spectra of CsSnI₃ (202) peak. Scale bars, 2 µm.

Full size image

X-ray diffraction (XRD) patterns of the pristine CsSnI₃ and CsSnI₃:PbX₂ films exhibited dominant three-dimensional orthorhombic phase (β-γ-CsSnI₃) peaks at (101) and (202) and the CsSnI₃:PbCl₂ film demonstrated the highest film crystallization peak intensity (Fig. 2c)¹¹. In addition, the Br- and Cl-incorporated samples showed minor shifts in peak positions from 29.10° (CsSnI₃) to 29.14° (CsSnI₃:PbCl₂) and 29.40° (CsSnI₃:PbBr₂). The incorporation of Br into the iodide-based perovskite lattice is well documented owing to the close ionic radius^44,45,46. The minor shift in CsSnI₃:PbCl₂ suggests that only a portion of the Cl was incorporated into the lattice. The incorporation of Br⁻ and residue Cl⁻ caused a slight shift in XRD spectra towards a higher angle, corresponding to lattice contraction. Slight contraction can increase orbital overlap and decrease charge scattering to increase carrier mobility^47,48.

To characterize and trace the Cl incorporation in the CsSnI₃-based thin film, we conducted XPS analysis and depth profiling (Supplementary Fig. 7). The deposited CsSnI₃ films exhibited evenly distributed elements of Cs, Sn and I throughout the bulk, indicating successful reaction among precursor compounds (Fig. 3a). In addition, a uniform distribution of Pb was detected throughout the bulk, whereas a minor signal of Cl was identified at the film surface, indicating that most of the Cl was sublimated out of the film during annealing (Fig. 3b,c). This is in contrast to the CsSnI₃:PbBr₂ and CsSnI₃:PbI₂ thin films, which showed a uniform distribution of Br and I from the surface to the bulk (Supplementary Fig. 8). From XPS analysis, the relative atomic ratios of the final CsSnI₃:PbX₂ film are listed in Supplementary Table 1, with the optimized CsSnI₃:PbCl₂ final film as Cs_1.3SnPb_0.2I₃Cl_0.15, representatively.

a,b, XPS depth analysis of deposited CsSnI₃ (a) and CsSnI₃:PbCl₂ (b) thin films on bare Si substrates. Etch rate was 0.5 nm s⁻¹. c, Cl 2p core level XPS spectra of CsSnI₃:PbCl₂ thin film with etch time 1–30 s. d, TGA of as-evaporated CsSnI₃:PbCl₂ thin film. Onset temperature and weight loss at points A–D: A (26.5 °C, 2%), B (143.3 °C, 8.3%), C (278.1 °C, 19.3%) and D (617.6 °C, 35.2%). e, Illustration of post annealing treatment on PbCl₂ effect of evaporated perovskites. f, Schematic of final vapour-deposited CsSnI₃:PbCl₂ lattice composition.

Full size image

To identify the form of vaporized Cl, we conducted powder thermogravimetric analysis (TGA) collected from the as-evaporated pristine CsSnI₃ and CsSnI₃:PbCl₂ thin films (Fig. 3d). Both films demonstrated comparable trends, but a notable difference was identified between the weight loss below 100 °C, where the CsSnI₃:PbCl₂ film showed 2% weight loss since the beginning. To identify the origin of this weight loss, we conducted TGA of each CsI, SnI₂, PbCl₂, SnCl₂ and CsCl powder, and only SnCl₂ demonstrated an initial 2% weight loss, identical to the CsSnI₃:PbCl₂ film (Supplementary Fig. 9). Thus, we speculated the thermally activated initiation of Cl vaporization through SnCl₂-related and/or generated species and the start of solid-state reactions. The vaporization of Cl can promote uniform grain growth in the vertical direction, triggering perovskite conversion of sequentially deposited precursor layers and consequently enhanced film crystallinity^34,49. The low reaction efficiency of sequentially deposited layers relies on thermally driven layer-to-layer solid melting and merging process^32,33,34,35. Here a thermally activated volatile species can initiate its sublimation from the bottom, diffusing through the entire layer and transporting reactive species, to enhance the reaction kinetics at each interface, and accelerate and refine overall perovskite conversion (Fig. 3e,f). This reaction kinetics corresponds to the disappearance of unreacted CsI aggregates identified in scanning electron microscopy–energy dispersive X-ray spectroscopy analysis, through volatile Cl-related species-driven perovskite conversion (Supplementary Fig. 5). The residue Cl on the film surface may passivate the iodide vacancy as aligned with the positive peak shift in the XRD spectrum.

Comparing the field-effect modulation between CsSnI₃ and PbX₂ reveals that Pb did not play a crucial role in hole suppression because similar amounts of PbI₂ did not effectively reduce the I_off. In comparison with I and Br, the small ionic size of Cl makes it difficult to dominantly alloy into the iodide-based perovskite matrix. Instead, only a minimal amount of Cl remains whereas vaporization of dominant amount of Cl plays a key role in increasing perovskite phase crystallization^34,40. The enhanced perovskite film quality and uniformity can greatly reduce the number of dominant defects and the associated hole source that are heavily concentrated in grain boundaries, that is, V_Sn, and thus successfully modulate the high intrinsic conductivity^50,51,52. To verify the unique effect of the buried PbCl₂ layer, we deposited PbCl₂ as the top layer of the multilayer precursor compounds (Supplementary Fig. 10). Interestingly, we observed that the transfer characteristics of CsSnI₃:PbCl₂ TFTs with PbCl₂ as the top layer exhibited almost metallic characteristics with the absence of reliable off-current, indicating that the effect of PbCl₂ is not present when it is not the initial layer. This supports the notion that Cl from the buried layer enhances the reaction kinetics and film quality, improving field-effect modulation and reducing hole concentration.

The epitome of thermal evaporation is the ease of processing and scalability of thin films and devices. We therefore demonstrate large-area fabrication of CsSnI₃:PbCl₂ TFTs on a 49-cm² wafer (Fig. 4a). The transfer characteristics of 75 evenly selected devices out of 576 CsSnI₃:PbCl₂ TFTs show high uniformity, with a Gaussian distribution of µ_FE averaging 34.2 cm² V⁻¹ s⁻¹ and a standard deviation of 2.0 cm² V⁻¹ s⁻¹ (Fig. 4b and Supplementary Fig. 11). This is a demonstration of a large-scale three-dimensional Sn²⁺-halide perovskite TFT array. It should be noted that fabricating high-performance and uniform Sn²⁺-perovskite TFTs over a large area has remained an unsolved challenge when using the typical spin-coating approach. This difficulty arises from the necessity of adding an antisolvent dripping to promote nucleation for highly crystalline films during spin-coating, which results in varying perovskite film quality from centre to edge⁵³.

a, Image of 49-cm² wafer with 24 × 24 = 576 devices. b, Transfer characteristics of 75 devices evenly selected from the wafer-scale array. Channel length/width = 100 µm/200 µm (V_DS = −40 V). Inset shows the distribution of field-effect mobilities of the evenly selected 75 TFTs. c, Current–voltage transfer curves as a function of storage time in N₂-filled glove box of vapour-deposited (left) and solution-processed (right) CsSn_0.9Pb_0.1I₃ TFTs without encapsulation. Channel length/width = 100 µm/200 µm (V_DS = −40 V). d, Corresponding normalized µ_FE and I_on/I_off of transfer curves as a function of storage time. e, Film thickness and density. The dashed line represents the average height value. f, SEM images of vapour-deposited and solution-processed CsSnI₃-based perovskite films. Scale bars, 1 µm.

Full size image

More importantly, vapour-deposited CsSnI₃:PbCl₂ TFTs exhibit much improved storage stability in an N₂-filled glove box. By contrast, earlier works on solution-processed CsSnI₃-based perovskite TFTs could only achieve long-term stability with glass or polymer encapsulation¹². Without encapsulation, solution-processed CsSnI₃ TFTs lasted only a few hours, even when stored in the glove box (Fig. 4c). However, our vapour-deposited CsSnI₃ perovskite TFTs demonstrate superior stability under the same conditions, maintaining their electrical properties for over 150 days without any encapsulation (Fig. 4d). The fragile nature of solution-processed CsSnI₃-based thin films compared with their vapour-deposited counterparts can be traced back to the film fabrication, final thickness and density. Through AFM measurement, we confirmed that the thickness of solution-processed CsSnI₃-based film was substantially smaller than that of vapour-deposited film, that is, 16 nm and 48 nm, respectively (Fig. 4e). The film densities of the solution-processed and vapour-deposited film was measured through X-ray reflectometry, which were 3.74 cm⁻³ and 4.68 cm⁻³, respectively. The high density of vapour-deposited film arises from the dense packing of grains through high-temperature post annealing at 320 °C, whereas solution-processed films are annealed at 120 °C. The classical nucleation theory suggests that high temperature enables decreased nucleation rate, increased growth rate and, consequently, formation of densely packed large grains when applied to the crystallization of sequentially deposited perovskite precursor layers (Fig. 4e,f)^54,55. The XRD spectrum of the vapour-deposited film illustrates preferred vertical orientation with high crystallinity induced by vaporization of Cl, in comparison with the lower crystallinity of solution-processed film (Supplementary Fig. 12). Therefore, vapour-deposited CsSnI₃-based films and devices demonstrate stronger structural and environmental stability compared with solution-processed counterparts.

We have reported high-performance, stable p-channel CsSnI₃-based TFTs using a commercially compatible vapour-deposition approach with PbCl₂ as an additive. The volatile chloride triggers solid-state reactions and the conversion of as-evaporated precursor compounds. This facilitates the formation of high-quality and uniform perovskite films, and also modulates the high hole density, making them suitable for use as channel layers. The optimized CsSnI₃:PbCl₂ TFTs delivered average µ_FE of around 34 cm² V⁻¹ s⁻¹, I_on/I_off of around 10⁸ and storage stability of more than 150 days. We also demonstrated a large-scale Sn²⁺-halide perovskite TFT array that overcomes the technical challenges faced in the solution process. Our vapour-deposited TFTs could be used in backplanes for organic light-emitting diode displays, or in logic devices and circuits for monolithic three-dimensional integration, where low process temperatures are required.

Thin-film fabrication

CsI (99.9%) and PbCl₂ (99.99%) were purchased from Xi’an Polymer Light Technology Corp., and SnI₂ (99.99%), PbBr₂ (98%) and PbI₂ (99%) were purchased from Merck. These powders were directly used as vapour-deposition sources. The CsSnI₃:PbX₂ films were deposited using a thermal evaporator through the following procedure. Each powder was loaded in an alumina crucible for optimal conditions. The substrate temperature was maintained at room temperature, and vacuum pressure before evaporation was ~2.7 × 10⁻⁷ Torr. The distance between the powder-loaded alumina crucible and the substrate holder was ~40 cm. The thickness of each source was monitored during deposition. The perovskite precursor materials were evaporated in the following order: PbX₂/SnI₂/CsI. The thickness of each precursor source was monitored during deposition, and the thickness value of each as-evaporated precursor layer was measured by AFM to be 47.5 nm (CsI), 45 nm (SnI₂), 7.5 nm (25 at.% PbCl₂), 9.5 nm (30 at.% PbBr₂) and 11 nm (30 at.% PbI₂) (Supplementary Fig. 13). The multilayer of perovskite compounds was annealed at 320 °C for 10 min in an N₂-filled glove box. The final thickness value of the optimized Cs_1.3SnPb_0.2I₃Cl_0.15 film was measured at 48 nm (Supplementary Fig. 14).

Device fabrication and characterization

A heavily doped Si wafer (resistivity: 1–100 Ω cm) with a 100-nm thermally grown SiO₂ layer was used as the gate electrode and the dielectric layer. The CsSnI₃:PbX₂ channels were deposited on SiO₂ as channel layers using the aforementioned procedure. The shadow mask was covered on the substrate to obtain the patterned channel layers with width W = 750 µm and length L = 550 µm. Au source/drain electrodes (20 nm) were deposited with a shadow mask using thermal evaporation to construct a bottom-gate top-contact TFT. The channel length and width were 100 µm and 200 µm, respectively. All TFTs were characterized using a Keithley 4200SCS at room temperature.

The value of µ_FE and the threshold voltage (V_TH) were extracted at the saturation regime from the |I_DS| versus V_GS curve, following equation (1) (refs. ^56,57). C_i is the capacitance of the gate oxide. The extraction of V_TH and µ_FE is further described in Supplementary Fig. 15.

Film characterizations

The crystal structures of the films were analysed using XRD with MAX-2500 (Rigaku). The XPS depth profile and ultraviolet photoelectron spectroscopy analysis was performed using a PHI 5000 VersaProbe instrument (Ulvac-PHI, Tokyo, Japan). The Hall measurements of the films were performed using the van der Pauw method with a 0.51-T magnet (HMS-3000, Ecopia). The SEM images were measured using high-resolution FE-SEM (JSM 7800 F Prime). The AFM images were measured using NX-10 (Park Systems). The TGA analysis was measured using SDT Q-600 (TA Instruments), performed in an N₂-filled environment. The film samples were collected and measured in powder form, by scraping the as-evaporated film from glass substrates.