Ultrasensitive Heterojunctions of Graphene and 2D Perovskites Reveal Spontaneous Iodide Loss

Abstract

•Spontaneous iodide loss is revealed as an important degradation path of perovskites•A graphene/perovskite heterojunction reveals n-type doping caused by iodide loss•Covering perovskites with graphene suppresses iodide loss and improves stability•The potential of graphene as a diagnostic tool for degradation studies is demonstrated Metal halide perovskites have achieved remarkable success in lab-scale solar cells and light-emitting devices. However, instability issues impede their practical use. Various degradation paths of polycrystalline perovskites have been revealed under a range of external environmental stresses such as light, heat, and moisture. However, the understanding of the intrinsic stability of perovskites is far from complete. Here, we reveal spontaneous iodide loss as an important degradation path of 2D perovskite single crystals, enabled by an ultrasensitive, exfoliated 2D perovskite single-crystal sheet/graphene heterostructure device. Furthermore, covering perovskites with a graphene overlayer can suppress iodide loss, significantly improving perovskite stability. Our work provides important insights for future stable perovskite optoelectronic device development and demonstrates the potential of graphene as a promising diagnostic tool for device and material degradation. Despite the demonstrated high efficiency of perovskite solar cells and light-emitting devices, the understanding of the intrinsic stability of perovskites is far from complete. In this work, using an ultrasensitive, exfoliated 2D perovskite single-crystal sheet/graphene heterostructure device, we reveal spontaneous iodide loss as an important degradation pathway of 2D perovskite single crystals, which n-dopes the perovskite semiconductor by generating positively charged iodide vacancies. Furthermore, we show that covering perovskites with graphene can suppress the iodide loss, significantly improving perovskite stability. A perovskite phototransistor is demonstrated with a graphene/2D perovskite/graphene structure, which shows no degradation after 75 days. Our work not only provides important insights for future stable perovskite optoelectronic device development, but also demonstrates the potential of graphene as a promising sensitive diagnostic tool for device and material degradation studies. Despite the demonstrated high efficiency of perovskite solar cells and light-emitting devices, the understanding of the intrinsic stability of perovskites is far from complete. In this work, using an ultrasensitive, exfoliated 2D perovskite single-crystal sheet/graphene heterostructure device, we reveal spontaneous iodide loss as an important degradation pathway of 2D perovskite single crystals, which n-dopes the perovskite semiconductor by generating positively charged iodide vacancies. Furthermore, we show that covering perovskites with graphene can suppress the iodide loss, significantly improving perovskite stability. A perovskite phototransistor is demonstrated with a graphene/2D perovskite/graphene structure, which shows no degradation after 75 days. Our work not only provides important insights for future stable perovskite optoelectronic device development, but also demonstrates the potential of graphene as a promising sensitive diagnostic tool for device and material degradation studies. Organic-inorganic hybrid perovskite materials are an emerging class of low-cost, solution-processable semiconductors with desirable optoelectronic properties,1Sutherland B.R. Sargent E.H. Perovskite photonic sources.Nat. Photonics. 2016; 10: 295-302Crossref Scopus (1180) Google Scholar and have recently achieved remarkable success in lab-scale optoelectronic devices such as solar cells, light-emitting diodes (LEDs), and lasers.2Burschka J. Pellet N. Moon S.-J. Humphry-Baker R. Gao P. Nazeeruddin M.K. Grätzel M. Sequential deposition as a route to high-performance perovskite-sensitized solar cells.Nature. 2013; 499: 316-319Crossref PubMed Scopus (7875) Google Scholar, 3Jeon N.J. Noh J.H. Kim Y.C. Yang W.S. Ryu S. Seok S.I. Solvent engineering for high-performance inorganic-organic hybrid perovskite solar cells.Nat. Mater. 2014; 13: 897-903Crossref PubMed Scopus (5305) Google Scholar, 4Xiao Z. Kerner R.A. Zhao L. Tran N.L. Lee K.M. Koh T.-W. Scholes G.D. Rand B.P. Efficient perovskite light-emitting diodes featuring nanometre-sized crystallites.Nat. Photonics. 2017; 11: 108-115Crossref Scopus (1011) Google Scholar, 5Zhao L. Yeh Y.-W. Tran N.L. Wu F. Xiao Z. Kerner R.A. Lin Y.L. Scholes G.D. Yao N. Rand B.P. In situ preparation of metal halide perovskite nanocrystal thin films for improved light-emitting devices.ACS Nano. 2017; 11: 3957-3964Crossref PubMed Scopus (135) Google Scholar, 6Jia Y. Kerner R.A. Grede A.J. Rand B.P. Giebink N.C. Continuous-wave lasing in an organic-inorganic lead halide perovskite semiconductor.Nat. Photonics. 2017; 11: 784-788Crossref Scopus (292) Google Scholar Recently, the family of layered two-dimensional (2D) perovskites has gained growing attention due to their tunable quantum-well structures,7Gélvez-Rueda M.C. Hutter E.M. Cao D.H. Renaud N. Stoumpos C.C. Hupp J.T. Savenije T.J. Kanatzidis M.G. Grozema F.C. Interconversion between free charges and bound excitons in 2D hybrid lead halide perovskites.J. Phys. Chem. C. 2017; 121: 26566-26574Crossref PubMed Scopus (105) Google Scholar unique optical properties,8Straus D.B. Kagan C.R. Electrons, excitons, and phonons in two-dimensional hybrid perovskites: connecting structural, optical, and electronic properties.J. Phys. Chem. Lett. 2018; 9: 1434-1447Crossref PubMed Scopus (218) Google Scholar improved moisture stability, and promising device performance.9Tsai H. Nie W. Blancon J.-C. Stoumpos C.C. Asadpour R. Harutyunyan B. Neukirch A.J. Verduzco R. Crochet J.J. Tretiak S. et al.High-efficiency two-dimensional Ruddlesden-Popper perovskite solar cells.Nature. 2016; 536: 312-316Crossref PubMed Scopus (2340) Google Scholar, 10Chen Y. Sun Y. Peng J. Tang J. Zheng K. Liang Z. 2D Ruddlesden-Popper perovskites for optoelectronics.Adv. Mater. 2018; 30: 1703487Crossref Scopus (494) Google Scholar Interfacing 2D perovskites with other 2D materials such as graphene and transition-metal dichalcogenides have significantly broadened the application range of 2D materials.11Tian H. Zhao L. Wang X. Yeh Y.-W. Yao N. Rand B.P. Ren T.-L. Extremely low operating current resistive memory based on exfoliated 2D perovskite single crystals for neuromorphic computing.ACS Nano. 2017; 11: 12247-12256Crossref PubMed Scopus (220) Google Scholar, 12Shi E. Gao Y. Finkenauer B.P. Akriti Coffey A.H. Dou L. Two-dimensional halide perovskite nanomaterials and heterostructures.Chem. Soc. Rev. 2018; https://doi.org/10.1039/C7CS00886DCrossref Google Scholar Despite impressive progress, substantial challenges remain regarding stability, both intrinsic and extrinsic.13Wang Z. Shi Z. Li T. Chen Y. Huang W. Stability of perovskite solar cells: a prospective on the substitution of the A cation and X anion.Angew. Chem. Int. Ed. 2017; 56: 1190-1212Crossref PubMed Scopus (419) Google Scholar, 14Berhe T.A. Su W.-N. Chen C.-H. Pan C.-J. Cheng J.-H. Chen H.-M. Tsai M.-C. Chen L.-Y. Dubale A.A. Hwang B.-J. Organometal halide perovskite solar cells: degradation and stability.Energy. Environ. Sci. 2016; 9: 323-356Crossref Google Scholar, 15Zhao L. Gao J. Lin Y.L. Yeh Y.-W. Lee K.M. Yao N. Loo Y.-L. Rand B.P. Electrical stress influences the efficiency of CH3NH3PbI3 perovskite light emitting devices.Adv. Mater. 2017; 29: 1605317Crossref Scopus (91) Google Scholar Among various thrusts, growing attention is being paid to iodide and iodine induced degradation. For example, accelerated degradation has been observed when perovskite films are exposed to iodine vapor.16Wang S. Jiang Y. Juarez-Perez Emilio J. Ono Luis K. Qi Y. Accelerated degradation of methylammonium lead iodide perovskites induced by exposure to iodine vapour.Nat. Energy. 2016; 2: 16195Crossref Scopus (387) Google Scholar Additionally it has been reported that CH3NH3PbI3 (MAPbI3) perovskite, under long-term thermal annealing at 85°C, will decompose to PbI2 and I2.17Conings B. Drijkoningen J. Gauquelin N. Babayigit A. D'Haen J. D'Olieslaeger L. Ethirajan A. Verbeeck J. Manca J. Mosconi E. Intrinsic thermal instability of methylammonium lead trihalide perovskite.Adv. Energy Mater. 2015; 5: 1500477Crossref Scopus (1507) Google Scholar Furthermore, upon applying a large electric field, it is proposed that I− ions are lost through I2 vapor formation.18Yuan Y. Wang Q. Shao Y. Lu H. Li T. Gruverman A. Huang J. Electric-field-driven reversible conversion between methylammonium lead triiodide perovskites and lead iodide at elevated temperatures.Adv. Energy Mater. 2016; 6: 1501803Crossref Scopus (257) Google Scholar Superoxide, generated by oxygen and photogenerated electrons, can induce perovskite degradation and produce CH3NH2, PbI2, and I2.19Aristidou N. Sanchez-Molina I. Chotchuangchutchaval T. Brown M. Martinez L. Rath T. Haque S.A. The role of oxygen in the degradation of methylammonium lead trihalide perovskite photoactive layers.Angew. Chem. Int. Ed. 2015; 54: 8208-8212Crossref PubMed Scopus (665) Google Scholar Therefore, iodine/iodide loss, in particular, is an important component of the perovskite degradation process. Iodine/iodide-related degradation studies reported so far, however, have investigated polycrystalline thin films under various external environmental stresses such as moisture, heat, light, and electric field.20Kato Y. Ono L.K. Lee M.V. Wang S. Raga S.R. Qi Y. Silver iodide formation in methyl ammonium lead iodide perovskite solar cells with silver top electrodes.Adv. Mater. Interfaces. 2015; 2: 1500195Crossref Scopus (545) Google Scholar, 21Zhao L. Kerner R.A. Xiao Z. Lin Y.L. Lee K.M. Schwartz J. Rand B.P. Redox chemistry dominates the degradation and decomposition of metal halide perovskite optoelectronic devices.ACS Energy Lett. 2016; 1: 595-602Crossref Scopus (160) Google Scholar No studies, to our knowledge, have been performed to show whether iodine/iodide loss occurs spontaneously at room temperature. Furthermore, it is difficult to accurately study the intrinsic stability of perovskites using solution-processed polycrystalline thin films due to the influence, and even potentially dominant effects, of grain boundaries on the material properties.22Lee J.-W. Bae S.-H. De Marco N. Hsieh Y.-T. Dai Z. Yang Y. The role of grain boundaries in perovskite solar cells.Mater. Today Energy. 2017; 7: 149-160Crossref Scopus (168) Google Scholar For example, it has been reported that grain boundaries in polycrystalline thin films induce more defects,23Sherkar T.S. Momblona C. Gil-Escrig L. Ávila J. Sessolo M. Bolink H.J. Koster L.J.A. Recombination in perovskite solar cells: significance of grain boundaries, interface traps, and defect ions.ACS Energy Lett. 2017; 2: 1214-1222Crossref PubMed Scopus (658) Google Scholar and facilitate ion migration,24Shao Y. Fang Y. Li T. Wang Q. Dong Q. Deng Y. Yuan Y. Wei H. Wang M. Gruverman A. Grain boundary dominated ion migration in polycrystalline organic-inorganic halide perovskite films.Energy Environ. Sci. 2016; 9: 1752-1759Crossref Google Scholar compared with single crystals. In fact, if iodide/iodine loss at surfaces and/or grain boundaries occurs, the resulting vacancies are likely responsible for observed grain boundary-induced ion migration. Although perovskite single crystals have been synthesized, they are only available as bulk specimens,22Lee J.-W. Bae S.-H. De Marco N. Hsieh Y.-T. Dai Z. Yang Y. The role of grain boundaries in perovskite solar cells.Mater. Today Energy. 2017; 7: 149-160Crossref Scopus (168) Google Scholar whereas thin-film perovskites provide improved detection of surface degradation due to their larger surface-to-volume ratio.25Fan Z. Xiao H. Wang Y. Zhao Z. Lin Z. Cheng H.-C. Lee S.-J. Wang G. Feng Z. Goddard III, W.A. et al.Layer-by-layer degradation of methylammonium lead tri-iodide perovskite microplates.Joule. 2017; 1: 548-562Abstract Full Text Full Text PDF Scopus (148) Google Scholar Therefore, there is a call for studies on the intrinsic stability of metal halide perovskites based on thin single-crystal sheets. The unique layered crystal structure of 2D perovskite materials allows for the exfoliation of single crystals into ultrathin sheets,11Tian H. Zhao L. Wang X. Yeh Y.-W. Yao N. Rand B.P. Ren T.-L. Extremely low operating current resistive memory based on exfoliated 2D perovskite single crystals for neuromorphic computing.ACS Nano. 2017; 11: 12247-12256Crossref PubMed Scopus (220) Google Scholar enabling the study of intrinsic stability of perovskite single crystals with high sensitivity. In this work, we study the intrinsic stability of 2D perovskites through an exfoliated 2D perovskite single-crystal sheet/graphene heterostructure-based field-effect transistor (FET). Graphene is utilized for its high sensitivity to its surrounding environment,26Allen M.J. Tung V.C. Kaner R.B. Honeycomb carbon: a review of graphene.Chem. Rev. 2010; 110: 132-145Crossref PubMed Scopus (5810) Google Scholar an aspect that has been well established in its use in various sensors capable of detecting individual gas molecules.27Schedin F. Geim A.K. Morozov S.V. Hill E.W. Blake P. Katsnelson M.I. Novoselov K.S. Detection of individual gas molecules adsorbed on graphene.Nat. Mater. 2007; 6: 652Crossref PubMed Scopus (6780) Google Scholar The ultrahigh sensitivity of graphene to the environment is primarily due to its electronic structure:28Mao H.Y. Lu Y.H. Lin J.D. Zhong S. Wee A.T.S. Chen W. Manipulating the electronic and chemical properties of graphene via molecular functionalization.Prog. Surf. Sci. 2013; 88: 132-159Crossref Scopus (152) Google Scholar Its ambipolarity means that graphene can be doped n-type or p-type by electron-donating or -withdrawing groups, respectively, and the low density of states near the Dirac point means that a small change in doping concentration will shift the Fermi level significantly. First, single crystals of 2D perovskites are exfoliated to a thin sheet of ∼55 nm, and a 2D perovskite/graphene heterostructure-based FET is employed to monitor the 2D perovskite degradation. We show that the 2D perovskite thin crystal sheet is intrinsically unstable, leading to a continuous shift of the Fermi level with respect to the Dirac point of graphene. Moreover, the degradation is accelerated by light illumination. The mechanism underlying this effect is the spontaneous loss of iodide, which is confirmed by cross-sectional scanning transmission electron microscopy (STEM), energy-dispersive X-ray spectroscopy (EDS), UV photoelectron spectroscopy (UPS), and Kelvin probe measurements. The release of iodide generates iodide vacancies, which induce an n-type doping of the perovskite, shifting the Fermi level (Ef) of both perovskite and graphene upward. Furthermore, we show that a graphene cover on the perovskite thin crystal sheet can improve perovskite stability by preventing iodide loss. An ultrastable phototransistor based on the graphene/2D perovskite/graphene structure is demonstrated, which shows no degradation in 75 days. Millimeter-sized 2D (PEA)2PbI4 perovskite single crystals (PEA refers to phenethylammonium) were synthesized using an anti-solvent vapor crystallization method,11Tian H. Zhao L. Wang X. Yeh Y.-W. Yao N. Rand B.P. Ren T.-L. Extremely low operating current resistive memory based on exfoliated 2D perovskite single crystals for neuromorphic computing.ACS Nano. 2017; 11: 12247-12256Crossref PubMed Scopus (220) Google Scholar as shown in Figure 1A. Figure 1B shows the X-ray diffraction (XRD) patterns of the as-grown 2D perovskite single crystals, which confirm the phase purity. The scanning electron microscopy (EM) image clearly shows the 2D layered morphology (Figure 1C). The layered crystal structure and the weak van der Waals forces between the organic molecules in perovskite layers (Figure 1D) allows for the exfoliation of 2D perovskite single crystals into ultrathin sheets. Figure 1E shows an exfoliated thin crystal sheet with a thickness of ∼55 nm, enabling our study of the intrinsic stability of perovskite single crystals with high sensitivity. To study the stability of 2D perovskites, we utilized a 2D perovskite/graphene heterojunction-based FET, as shown in Figure 2A. First, graphene was mechanically exfoliated on a Si wafer terminated with a 300-nm SiO2 layer (Figure S1). Patterned source and drain electrodes were then fabricated by electron beam lithography followed by thermal evaporation of Au and lift-off. Thereafter, a 2D perovskite thin crystal sheet was exfoliated and transferred on the graphene channel. Notably, only part of the graphene channel is covered by the 2D perovskite sheet. In this way, the uncovered part of graphene maintains its initial Dirac point without any shift, thus serving as a reference. The Dirac point of the graphene channel covered by the perovskite sheet is used as a probe to monitor the Fermi-level change of the graphene/2D perovskite heterostructure. Due to extremely low carrier density of states (∼1012 cm−2) near the Dirac point, a conductance minimum is formed when the Fermi level of graphene is at the Dirac point,29Novoselov K.S. Geim A.K. Morozov S.V. Jiang D. Katsnelson M.I. Grigorieva I.V. Dubonos S.V. Firsov A.A. Two-dimensional gas of massless Dirac fermions in graphene.Nature. 2005; 438: 197Crossref PubMed Scopus (17894) Google Scholar and two conductance minima will be present if the graphene is doped differently in two regions along the graphene channel, inducing two separate coincidences of the Fermi level with the Dirac point as a function of gate voltage.30Chiu H.-Y. Perebeinos V. Lin Y.-M. Avouris P. Controllable p-n junction formation in monolayer graphene using electrostatic substrate engineering.Nano Lett. 2010; 10: 4634-4639Crossref PubMed Scopus (138) Google Scholar, 31Barraza-Lopez S. Vanević M. Kindermann M. Chou M.Y. Effects of metallic contacts on electron transport through graphene.Phys. Rev. Lett. 2010; 104: 076807Crossref PubMed Scopus (134) Google Scholar, 32Bartolomeo A.D. Giubileo F. Santandrea S. Francesco Romeo F. Citro R. Schroeder T. Lupina G. Charge transfer and partial pinning at the contacts as the origin of a double dip in the transfer characteristics of graphene-based field-effect transistors.Nanotechnology. 2011; 22: 275702Crossref PubMed Scopus (71) Google Scholar, 33Nouchi R. Tanigaki K. Competitive interfacial charge transfer to graphene from the electrode contacts and surface adsorbates.Appl. Phys. Lett. 2015; 106: 083107Crossref Scopus (9) Google Scholar, 34Sun Y. Xie D. Zhang C. Li X. Xu J. Sun M. Teng C. Li X. Zhu H. Poly (ethylene imine)-modulated transport behaviors of graphene field effect transistors with double Dirac points.J. Appl. Phys. 2017; 121: 134305Crossref Scopus (10) Google Scholar Figure 2B shows the transfer curves of the as-prepared 2D perovskite/graphene FET. Two local conductance minima are observed, and we define their corresponding gate voltages as VDirac− and VDirac+ for the left and right conductance minima, respectively. Initially, VDirac− is approximately 10 V and VDirac+ is approximately 40 V. When the 2D perovskite is transferred onto the graphene channel, the initial Fermi level/work function mismatch between the 2D perovskite and graphene induces a charge transfer between the 2D perovskite and graphene, changing the Fermi-level position with respect to the Dirac point in the graphene covered by the perovskite, while it remains unchanged in the uncovered graphene, as shown in Figure 2F. Despite this local doping of the graphene channel, the Fermi level is still below the Dirac points in both parts of the graphene, as both VDirac− and VDirac+ are positive. After storing the samples in the dark under vacuum for 4 hr, VDirac− moves to −5 V while VDirac+ remains ∼40 V, a 45-V difference (Figure 2C). Since graphene is inherently stable we can assume that the stable VDirac+ corresponds to the uncovered graphene and the VDirac− corresponds to the graphene region covered by the 2D perovskite. The shift of VDirac− corresponds to a 16-meV increase of the Fermi level, leading to 7.2 × 1011 cm−2 more free electrons in graphene (i.e., n-type doping). The origin of these extra electrons is through charge transfer from the 2D perovskite, suggesting that the 2D perovskite is self-n-doped during this period. The corresponding band diagram under this condition is shown in Figure 2G. Next, the sample is exposed to green light illumination by a laser (520-nm wavelength) with an output power of ∼1 μW for half an hour. Following this treatment, VDirac+ remains ∼40 V while VDirac− moves to −20 V (Figure 2D), which indicates that the graphene channel has been further n-doped (Figure 2H). The shifting rate of VDirac− is 30 V/hr under laser irradiation, which is significantly faster than that in darkness (3.75 V/hr). The sample is then stored in vacuum for another 24 hr in the dark, and VDirac− is further shifted to −45 V (Figure 2E), indicating that the graphene channel is further n-doped (Figure 2I), with a shifting rate of 1.04 V/hr. The continuous shift of VDirac− indicates that the 2D perovskite thin single-crystal sheet is unstable and becomes increasingly n-type, leading to continuous charge transfer between perovskite and graphene. The shift rate of VDirac− under light irradiation (30 V/hr) is about 10 times higher than that under dark conditions (3.75–1.04 V/hr), indicating that 2D perovskite degradation is accelerated by light illumination. The faster shifting rate of VDirac− is not caused by photogenerated carrier injection because the VDirac− shift is an irreversible process. If the VDirac− shift were caused by photogenerated carrier injection, we should expect VDirac− to shift back immediately after illumination ceases. The mechanism of the degradation of 2D perovskites, which induces the continuous Fermi-level shift with respect to the Dirac point of graphene, is the loss of iodide, which is confirmed by cross-sectional STEM and EDS measurements of the 2D perovskite/graphene structure, as shown in Figures 3A and 3B . The concentration of lead (Pb) is constant throughout the 2D perovskite layer, while iodine (I) concentration is lower near the top surface than at the bottom interface, in direct evidence of iodide loss from the 2D perovskite crystal. To further confirm the iodine and lead distribution, we performed EDS mapping measurements (Figure S2), which were shown to be consistent with the EDS line profile. It has been reported both theoretically and experimentally that positively charged I vacancies (VI+) result in n-type doping in perovskites,35Kim J. Lee S.-H. Lee J.H. Hong K.-H. The role of intrinsic defects in methylammonium lead iodide perovskite.J. Phys. Chem. Lett. 2014; 5: 1312-1317Crossref PubMed Scopus (666) Google Scholar, 36Yin W.-J. Shi T. Yan Y. Unusual defect physics in CH3NH3PbI3 perovskite solar cell absorber.Appl. Phys. Lett. 2014; 104: 063903Crossref Scopus (1933) Google Scholar, 37Xiao Z. Yuan Y. Shao Y. Wang Q. Dong Q. Bi C. Sharma P. Gruverman A. Huang J. Giant switchable photovoltaic effect in organometal trihalide perovskite devices.Nat. Mater. 2015; 14: 193-198Crossref PubMed Scopus (1196) Google Scholar in agreement with the shift of the Fermi level observed in this work. The loss of iodide generates VI+ and n-dopes the perovskites. The excess electrons from perovskite transfer to the graphene channel, inducing band bending in the perovskite layer at the interface with graphene (Figures 2F–2I). The accumulated positively charged VI+ in the depletion region of the perovskite enables a larger density of electrons in graphene that is covered by the perovskite and consequently shifts the Fermi level with respect to the Dirac point (Figure S3). To further confirm this mechanism, we conducted UPS measurements to study the doping effects induced by iodide loss. Due to charging effects, bulk 2D perovskite single crystals cannot be used for UPS measurements directly, while exfoliated 2D perovskite single-crystal sheets are too small in area. As an alternative, we use spin-coated polycrystalline 2D perovskite thin films on top of a conducting indium tin oxide (ITO) substrate to study the Fermi-level change (Figure S4). As shown in Figure 3C, after UV light illumination for 1 hr, UPS measurements show that the perovskite work function has decreased by 160 meV at the perovskite free surface. X-ray photoelectron spectroscopy (XPS) measurements were also conducted to study the chemical composition of the film (Figure S5). For a fresh 2D perovskite film, the initial I/Pb ratio is 4.52, slightly higher than the stoichiometry I/Pb ratio of 4 for PEA2PbI4, likely a consequence of halide termination and the polycrystalline nature of the film. After UV illumination, the I/Pb ratio decreases to 3.67, which further suggests iodide loss. A detailed comparison of the XPS I 3d core levels before and after UV exposure shows a 0.15-eV shift toward higher binding energy (Figure S6), which, along with the observed work function decrease, confirms an upward Fermi-level shift due to n-doping. The same shift is found for the Pb 4f core level (not shown here). To study whether iodide loss and n-doping of 2D perovskite is a spontaneous process that can happen at ambient pressure, we employed Kelvin probe measurements to measure the work function of 2D perovskite films in N2 at ambient pressure (Figure 3D). Initially, the work function of the 2D perovskite polycrystalline film is measured as 4.96 eV. After storage in N2 in the dark for 20 hr, the work function is minimally shifted to 4.94 eV. After 15 min of blue light (450-nm wavelength) illumination in N2 at ambient pressure, the work function is shifted to 4.83 eV. The work function shift is consistent with the iodide loss and the film becoming more n-type. The Kelvin probe measurement at ambient pressure suggests that the iodide loss is a spontaneous process. Vacuum and light may accelerate the iodide loss, but are not required conditions. Iodide loss is also observed for spin-coated polycrystalline MAPbI3, the most widely used perovskite in solar cells. As shown in the XPS spectra in Figure S7, the I/Pb ratio extracted from XPS shows a decrease from an initial 3.73 to 3.30 in vacuum following UV exposure. The slightly higher I/Pb ratio compared with the expected I/Pb ratio of 3 for MAPbI3 is likely due to halide termination and the polycrystalline nature of the film. The observation of iodide loss from MAPbI3 films triggers more attention about iodide loss as a degradation pathway for perovskite solar cells. On the other hand, the loss of iodide does not correspond to any work function change of MAPbI3 via UPS (Figure S8). This is consistent with the fact that MAPbI3 is more defect tolerant, and the amount of iodide loss in this case is insufficient to trigger any electronic property change. Figure 3E shows the schematic for the degradation mechanism revealed so far. In brief, VI+ and free electrons are generated due to the loss of iodide from the 2D perovskite, and electrons are transferred from the perovskite to the graphene channel, n-type doping the graphene channel and inducing a shift of the Fermi level with respect to the Dirac point. This process is accelerated by vacuum and light illumination. It has been reported that light is an important aspect to consider in perovskite materials processing that influences perovskite crystallization speed and film morphology,38Ummadisingu A. Steier L. Seo J.-Y. Matsui T. Abate A. Tress W. Grätzel M. The effect of illumination on the formation of metal halide perovskite films.Nature. 2017; 545: 208Crossref PubMed Scopus (210) Google Scholar and our results call for more attention on the interplay of photolysis and defect formation. Furthermore, considering that vacuum-based processing is widely used for perovskite-based device fabrication and that light is usually an unintentional factor that may be involved in the fabrication process, our results suggest that special attention should be paid to the influence of vacuum deposition and light control on device performance. In view of this degradation mechanism in 2D perovskites and the strong molecular blocking capability of conventional 2D materials such as graphene, we propose a method to improve the stability of perovskite films via a graphene encapsulation layer. Using this method, we demonstrate a stable 2D perovskite device, as shown in Figure 4A. As the perovskite layer is covered by graphene, the pathway for iodide loss is blocked, in turn significantly improving stability. As shown in Figure 4B, VDirac shifts less than 10 V in 27 days, during which period the sample is stored in nitrogen for most of the time and accumulates 1 hr of light irradiation in a vacuum chamber. The improved stability can be attributed to the inhibited iodide loss, which is confirmed by cross-sectional STEM/EDS measurements (Figures 4C and 4D). The concentration of I and Pb remains unchanged from the top surface to the bottom, showing that graphene can efficiently prevent iodide escape, and represents a promising strategy to improve the stability of perovskite devices. The strong gas-blocking capability of graphene is due to its unique structure; it has been reported that even a monolayer graphene membrane is impermeable to standard gases including helium.39Bunch J.S. Verbridge S.S. Alden J.S. van der Zande A.M. Parpia J.M. Craighead H.G. McEuen P.L. Impermeable atomic membranes from graphene sheets.Nano Lett. 2008; 8: 2458-2462Crossref PubMed Scopus (2351) Google Scholar These barrier properties can help to explain studies in which the incorporation of graphene into perovskite solar cells shows improved device stability.40Arora N. Dar M.I. Hinderhofer A. Pellet N. Schreiber F. Zakeeruddin S.M. Grätzel M. Perovskite solar cells with CuSCN hole extraction layers yield stabilized efficiencies greater than 20%.Science. 2017; 358: 768-771Crossref PubMed Scopus (1155) Google Scholar, 41Agresti A. Pescetelli S. Taheri B. Del Rio Castillo A.E. Cinà L. Bonaccorso F. Carlo A.D. Graphene-perovskite solar cells exceed 18 % efficiency: a stability study.ChemSusChem. 2016; 9: 2609-2619Crossref PubMed Scopus (148) Google Scholar, 42Kakavelakis G. Maksudov T. Konios D. Paradisanos I. Kioseoglou G. Stratakis E. Kymakis E. Efficient and highly air stable planar inverted perovskite solar cells with reduced graphene oxide doped PCBM electron transporting layer.Adv. Energy Mater. 2017; 7: 1602120Crossref Scopus (162) Google Scholar To demonstrate the practical use of the graphene/perovskite/graphene device structure, we realized a stable perovskite phototransistor.43Shao Y. Liu Y. Chen X. Chen C. Sarpkaya I. Chen Z. Fang Y. Kong J. Watanabe K. Taniguchi T. et al.Stable graphene-two-dimensional multiphase perovskite heterostructure phototransistors with high gain.Nano Lett. 2017; 17: 7330-7338Crossref PubMed Scopus (74) Google Scholar The photoresponsivity, R, is defined as R = Ilight/P, where Ilight is the difference between the current in the dark and under light, and P is the input light power. Figure 5A shows the photoresponsivity versus gate voltage (Vg) under different light power. A peak photoresponsivity of 600 A/W is reached at Vg = −8 V. The original transfer curves measured under dark and light are shown in Figure S9. Figure 5B shows a linear relation between photoresponsivity and input light power in log scale, with lower light power showing larger photoresponsivity. In assessing stability, the photocurrent shows no sign of degradation over the course of 75 days (Figure 5C). It is reasonable to infer that other conventional 2D materials, such as insulating BN and semiconducting MoS2, can also be used as iodide-blocking layers for perovskites, an aspect requiring further study, but thus widening the use of this method for different application purposes such as perovskite photovoltaics, LEDs, and memories. In summary, we reveal iodide loss as an important degradation pathway of 2D perovskite single crystals, which n-dopes the perovskite semiconductor by generating positively charged iodide vacancies. Furthermore, we show that covering perovskites with graphene can suppress the iodide loss, significantly improving perovskite stability. A perovskite phototransistor is demonstrated with a graphene/2D perovskite/graphene structure, which shows no degradation after 75 days. Our work not only provides important insights for future stable perovskite optoelectronic device development, but also demonstrates the potential of graphene as a promising sensitive diagnostic tool for device and material degradation studies.