Tautomeric mixture coordination enables efficient lead-free perovskite LEDs

Lead halide perovskite light-emitting diodes (PeLEDs) have demonstrated remarkable optoelectronic performance1–3. However, there are potential toxicity issues with lead4,5 and removing lead from the best-performing PeLEDs—without compromising their high external quantum efficiencies—remains a challenge. Here we report a tautomeric-mixture-coordination-induced electron localization strategy to stabilize the lead-free tin perovskite TEA2SnI4 (TEAI is 2-thiopheneethylammonium iodide) by incorporating cyanuric acid. We demonstrate that a crucial function of the coordination is to amplify the electronic effects, even for those Sn atoms that aren’t strongly bonded with cyanuric acid owing to the formation of hydrogen-bonded tautomeric dimer and trimer superstructures on the perovskite surface. This electron localization weakens adverse effects from Anderson localization and improves ordering in the crystal structure of TEA2SnI4. These factors result in a two-orders-of-magnitude reduction in the non-radiative recombination capture coefficient and an approximately twofold enhancement in the exciton binding energy. Our lead-free PeLED has an external quantum efficiency of up to 20.29%, representing a performance comparable to that of state-of-the-art lead-containing PeLEDs6–12. We anticipate that these findings will provide insights into the stabilization of Sn(II) perovskites and further the development of lead-free perovskite applications. Lead-free perovskite light-emitting diodes (LEDs) prepared using tautomeric mixture coordination provide improved ordering in the crystal structure, reduced recombination and enhanced exciton binding energy compared with lead-containing perovskite-based LEDs.


Article
Sn perovskite, suggesting a strong C=N bond, which arises because of the interaction of C=N and Sn perovskite.This is consistent with the electrostatic potential of the enol form of the CA molecule (Extended Data Fig. 1), indicating that the region near the nitrogen atom in the triazine ring has a high electron density.It could act as the site for donating lone-pair electrons to coordinate with Sn(II) (ref.26).The appearance of the characteristic vibration peak of C=O at 1,775 cm −1 for CA and CA-treated Sn perovskite implies the coexistence of keto form-that is, the presence of a tautomeric mixture of CA, which will be discussed later.
Subsequently, 1 H NMR measurements were carried out.Chemical shift of the resonance signal of CA (Fig. 1b) indicates the formation of hydrogen bonds in the form of N=C-OH⋯I − (ref.16).The broad resonance signal of SnI 2 -CA suggests that the CA molecule is immobilized by the N=C-OH⋯I − hydrogen bonds, which prevents the migration of I − .A similar phenomenon was also observed in 13 C NMR spectroscopy, which further confirms the strong interaction between CA and Sn(II).The FTIR and NMR spectroscopy results agreed well with the X-ray photoelectron spectroscopy (XPS) measurements (Supplementary Fig. 2), which show a chemical shift of the Sn 3d and I 3d XPS peaks after introducing CA into the Sn perovskite.AIMD simulations were used to investigate the configuration of CA on the perovskite surface, we observed the formation of CA dimers and trimers consisting of mixed enol and keto forms (Extended Data Fig. 2 and Supplementary Video 1).The possibility of a strong C=N⋯Sn 2+ coordination of the enol form leads to a partial surface-induced tautomerization of the thermodynamically more stable keto form.This tautomeric mixture is stabilized by the strong intermolecular N-H⋯O/O-H⋯N hydrogen bonds 27,28 .We also carried out a computation of the vibrational spectra of various types of surface-absorbed CA molecule (Fig. 1c).When the trimer is adsorbed on the surface, a blue shift of 25 cm −1 and 15 cm −1 (with respect to the same isolated trimer) is presented for C-N (1,650 cm −1 ) and C-O (1,775 cm −1 ) stretch vibrations, which is in good agreement with the experimental FTIR spectroscopy measurements.We note the presence of O-H vibrational modes at about 2,900-3,300 cm −1 is consistent with the formation of strong hydrogen bonds, a signature of the keto-enol hydrogen-bonded di-and trimeric forms.
We conclude that CA molecules on the Sn perovskites are composed of a mixture of keto and enol tautomers that form dimers or trimers (Supplementary Table 1).These ordered complexes are nearly commensurate with the perovskite lattice (Sn Sn second nearest distance is 1.05 nm, whereas the O-C-N⋯O-C-N distance of the extreme CA   molecules in the trimer is 1.1 nm) allowing for strong N-Sn coordination augmented by the interaction between OH-I − and H-bond (Fig. 1d-g), providing adequate protection of the perovskite surface against Sn(II) oxidation as well as Sn and I − loss.Beyond surface passivation/protection, the polyfunctional character of CA in combination with its ability to form ordered structures with a well-defined perpendicular orientation of the aromatic ring planes with respect to the surface suggests an additional potential to stabilize grain boundaries by crosslinking different grains through N/O-Sn and H-I interaction.Noticeably, this unique interaction between CA and Sn perovskite is distinctly different from previous reports (Extended Data Fig. 2) 1,16 .

Structure and stability
To understand how CA could stabilize the Sn perovskite, we performed a Bader analysis.It was found that the Sn electron density becomes more localized on coordination substantially reducing the corresponding Bader volume from around 59-62 au 3 to 24 au 3 .This effect even extends to nearby surface Sn 2+ atoms that have only weak interaction with CA (d(Sn-O) = 3.3 Å, d(Sn-N) = 3.3 Å), and still experience a clear localization of the Sn charge density (Bader volume of about 45-56 au 3 ).Both the coordinating bonds and their localizing effect contribute to the stabilization of surface Sn 2+ thus preventing Sn(II) oxidation.Furthermore, the electron localization function was calculated to provide a mapping of the electron cloud in the pristine and the CA-treated Sn perovskite (Fig. 1h).The electron cloud in the dashed region of the CA-treated Sn perovskite indicates strong electron localization of Sn 2+ .The triketone form of CA also showed similar properties (Extended Data Fig. 1).The effect of CA on electronic properties is further investigated by applying charge density difference (Fig. 1i).Electron transfer between CA and Sn perovskite is evaluated to be 0.08 e, suggesting strong interaction between CA and undercoordinated Sn 2+ , which inhibits charge transfer from Sn 2+ to O 2 molecule (Supplementary Table 2) [29][30][31] .The planar-averaged charge density difference (Extended Data Fig. 1 and Supplementary Fig. 3) also proves the charge redistribution at the interface, in combination with the negligible change of dielectric constants, which can benefit the carrier transport and improve electrical conductivity.
Scanning electron microscope images of the TEA 2 SnI 4 perovskite with 5% CA showed a smoother surface morphology than the other sample (Extended Data Fig. 3 and Supplementary Fig. 4).Atomic force microscope images further show the flatter film surface in the treated sample.Grazing incidence wide-angle X-ray scattering (GIWAXS) measurements (Fig. 2a, Extended Data Fig. 4 and Supplementary Fig. 5) suggested a highly uniform parallel orientation of the Sn perovskite films 32,33 .Compared with the pristine, the target film showed more preferred orientation, stronger diffraction intensity and narrower full width at half maximum (FWHM), which implies much more ordered stacking and better crystallinity of the perovskite crystal plates because of the introduction of CA, therefore avoiding Anderson localization 34 .Moreover, the low-n-phase components of the Sn perovskite films are suppressed effectively with the CA treatment.Such preferentially oriented crystallinity favours efficient carrier injection and prevents ion migration along the vertical direction during device operation 35 .
Kelvin probe force microscopy measurements indicated that the Sn perovskite film with CA possesses a higher surface potential (Extended Data Fig. 3 and Supplementary Fig. 6), which is mainly because of the reduced number of surface defects 36 .Calculations of the density of states (Extended Data Fig. 1) showed that iodide vacancy states in the band edge of the Sn perovskite were reduced by CA 37 , avoiding the presence of undercoordinated Sn 2+ that aggravates Sn 2+ oxidation.The suppression of I defects is vital to avoid non-radiative recombination loss in Sn perovskites.We find that the content of I − substantially increases from 82.4% in the pristine to 97.1% in the treated film (Extended Data Fig. 5 and Supplementary Table 3).The Sn:I ratios in the pristine and the target film were calculated to be 0.25 and 0.23, respectively.These results indicate that CA can prevent the migration of I − and suppress its evolution to I 2 or I 3 − (ref.38).Furthermore, we find that the content of Sn is increased by a factor of 1.09 compared with the pristine.Beyond suppressing I and Sn loss, CA prevents Sn 2+ oxidation, as demonstrated by XPS measurements, which illustrates a decrease in the content of Sn 4+ from 16.5% in the pristine to 7.6% in the target sample (Extended

Article
Data Fig. 5).To explain the effect of CA on trap states, we carried out measurements on Urbach energy (E U ) and time-resolved photoluminescence (TRPL) spectroscopy (Extended Data Fig. 6).The E U value decreased from 137 meV to 85 meV and the average carrier lifetime increased from 1.44 ns to 3.45 ns when CA was introduced.Furthermore, enhanced photoluminescence (PL) intensity is also observed in the CA-treated film.
Generally, the facile oxidation of Sn 2+ to Sn 4+ and the unstable crystal structures are responsible for undermining the stability of Sn perovskite films 29,39 .To check the shelf stability of the Sn perovskites in ambient air condition, time-dependent normalized PL measurements were performed (Fig. 2b and Supplementary Figs. 7 and 8).In contrast to the sharp reduction of PL intensity of the pristine with time, the target film retained its 90% initial PL intensity within 90 min.The non-radiative recombination capture coefficient A of the Sn perovskite sample was reduced by more than two orders of magnitude from 1.01 × 10 −9 cm 3 s −1 to 2.45 × 10 −11 cm 3 s −1 by introducing CA (Fig. 2c, Extended Data Fig. 6 and Supplementary Fig. 9).This indicates that the CA-treated Sn perovskite is potentially a remarkable luminescent material.Excitation-intensity-dependent photoluminescence quantum yield (PLQY) measurements showed that the PLQYs of the films increase with the excitation intensity to a large extent (Fig. 2d).It is well understood that, as the defect states are gradually filled, radiative bimolecular recombination begins to dominate the luminescence process during this test 40 .Afterwards, the X-ray diffraction (XRD) characterization was carried out, and the patterns of the films with CA showed stronger diffraction intensity and narrower FWHM (Extended Data Fig. 4) than the reference, which is because of the larger crystal size and better crystallinity (Extended Data Fig. 5).Furthermore, the XRD patterns of the films and nanoplates stored in dry air without encapsulation for different time are also presented for detecting the effect of CA on crystal stability of the Sn perovskites (Fig. 2e, Extended Data Fig. 5 and Supplementary Fig. 10).After 720-min storage, the diffraction intensity (002) of the films with CA decreases less than that of the pristine.The reduced non-radiative recombination coefficient and stabilized large crystal act together to improve the stability of the Sn perovskite.

Photophysical properties
As temperature increases (Fig. 3a,b), for the temperature-dependent steady-state PL spectra measurements, the PL intensity of the target film slightly decreased and the corresponding peak position of blue shifted by approximately 5 nm compared with the pristine (16 nm).These results indicate that the thermal quenching of fluorescence caused by non-radiative recombination can be essentially inhibited in the target film by CA coordination 41 .We evaluated the exciton binding energy (E b ) of the target film to be 467.6 ± 43.3 meV, which is about twofold higher than that of the pristine (263.3 ± 30.1 meV) (Extended Data Fig. 7).The enlarged E b of the target film suggests a higher energy threshold for exciton dissociation, making them promising luminescent materials.
Transient-absorption measurements showed two pronounced photo-bleaching peaks at about 615 nm and about 522 nm for both the pristine and the target sample, respectively (Fig. 3c,d and Extended Data Fig. 7), which are attributed to the 1s and 2s excitons 42 .The enhanced transient-absorption signal at the two exciton resonances for the treated sample compared with the pristine (around four times at 2s exciton and twice at 1s exciton) may be attributed to the enhanced Coulombic interaction and excitonic effect because of CA treatment.We notice that within 5 ps after photoexcitation, the exciton bleaching peaks show fast redshift, which is usually recognized as the filling process of the band tail states.The smaller shift of transient bleach further confirmed the reduced band tail states in the treated sample.Furthermore, we analysed the bleach recovery kinetics by fitting the decay curves with a tri-exponential function (Supplementary Fig. 11).The fast decay (τ 1 ) is attributed to the hot carrier relaxation, convoluted with the fast carrier trapping.The intermediate decay (τ 2 ) is responsible for the carrier trapping by defects.The treated sample shows a slower defect trapping rate, in agreement with the reduced defect traps through CA coordination-induced electron localization.As a result, the bleach signal for the treated sample displays a higher fraction of long-lived component (slow decay τ 3 > 5 ns) than that of the pristine (slow decay τ 3 > 4 ns), which is attributed to the recombination of photoexcited species.

LED performance
We then fabricated the Sn-PeLEDs with a conventional structure (Fig. 4a).The target Sn-PeLED showed electroluminescent spectra centred at about 630 nm under different voltages (Fig. 4b), and the luminance distribution is consistent with a Lambertian profile.The corresponding Commission Internationale de L'Eclairage (CIE) colour coordinates are (0.69, 0.31) (Fig. 4c), demonstrating a high colour-purity red emission.Compared with the reference device, the target Sn-PeLED had a narrower FWHM (24.9 nm) and showed stable steady-state electroluminescence output over 200 s (Supplementary Fig. 12), indicating greater potential in the field of pure-red lighting and displays.Figure 4d shows the current density-voltageluminance (J-V-L) curves of the representative Sn-PeLEDs.The target devices attained a maximum EQE of 20.29%, which is considerably higher than that of the control Sn-PeLEDs (peak EQE of 4.37%) (Fig. 4e and Extended Data Fig. 8).A statistical distribution of EQEs of the Sn-PeLEDs is shown in Fig. 4f.The average EQE of the target Sn-PeLEDs was 16.28%, which is about sevenfold higher than that of the reference devices (2.62%).Noticeably, the turn-on voltage of the target device was 0.3 V lower than that of the pristine device, which is consistent with the enhanced luminance and EQE.These results demonstrated a reduction of non-radiative recombination in the target Sn-PeLEDs.
Moreover, ultraviolet photoelectron spectroscopy measurements showed that the target Sn perovskite layer had better band alignment with both electron-and hole-transporting layers, favouring charge injection and transport, which was further validated by the space-charge-limited current (SCLC) measurements (Supplementary Figs. 13 and 14).Operational stability tests were further conducted for the Sn-PeLEDs at a constant current density (initial luminance was 30 cd m −2 in an N 2 -filled glovebox).The target device demonstrated more than threefold longer operating half-life time (T 50 ) than that of the reference (27.6 h versus 9.1 h) (Fig. 4g), which is one of the best T 50 results among lead-free PeLEDs.To highlight our work, we summarized the EQEs of lead-free PeLEDs reported so far (Fig. 4h and Extended Data Table 1).The champion EQE achieved is fully comparable to those of lead-based PeLEDs.We attribute the remarkable optoelectronic performance and operational stability of the lead-free PeLEDs to reduced defect density and halide ion migration, enhanced exciton binding energy and suppressed non-radiative recombination.Furthermore, the applicability of the CA tautomer-based mechanism is successfully extended to other two-and three-dimensional Sn-perovskites-based PeLEDs and the uniqueness of the proposed strategy is also demonstrated through structurally similar molecules with CA but not isomers (Extended Data Fig. 9, Supplementary Figs.15-19 and Supplementary Table 4).1) 16-18, 20, 32-35, 43-62 .a.u., arbitrary units.

Article
In summary, the existing performance barrier of lead-free PeLEDs was overcome by integrating CA into two-dimensional TEA 2 SnI 4 and allowed us to achieve an EQE of more than 20%.The polyfunctional character of CA led to the formation of a well-oriented and highly stable tautomeric CA complex that enables strong chemical coordination to Sn(II) with beneficial effects on crystallographic orientation and shelf stability of the Sn perovskite films yielding improved radiative recombination.Notably, the cooperative coordination-induced electron localization of Sn(II) using CA di-or trimeric surface complex is key for stabilizing Sn(II).Beyond solid-state electroluminescence, this fundamental progress could be broadly used in many lead-free perovskite semiconductors applications.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-023-06514-6.

Preparation of TEA 2 SnI 4 perovskite films
For Sn perovskite precursor solution (0.2 M), 2-thiopheneethylamine iodide and SnI 2 with a molar ratio of 2.2:1 were dissolved in anhydrous DMF:DMSO (4:1 v/v) solution, and RbI (6 mg ml −1 ) and tin powder (50 mg ml −1 ) were added.The precursor solution was stirred in an N 2 -filled glovebox (H 2 O and O 2 , <0.01 ppm) overnight at 50 °C and then filtered with a 0.22-μm polytetrafluoroethylene (PTFE) filter before use.The Sn perovskite films were spin-coated by a one-step process, 50-μl perovskite precursor (concentrations of 0%, 2%, 5% and 8% CA) was dropped onto the PEDOT:PSS substrates at 5,000 rpm for 55 s, and then 200 μl anhydrous toluene was dropped onto the substrate at 10 s after spin-coating.Finally, the obtained perovskite thin films were annealed at 100 °C for 10 min inside a glovebox.The corresponding films are denoted as pristine, with 2% CA, with 5% CA, and with 8% CA, respectively.In this text, with 5% CA is abbreviated as 'with CA'.Also, the perovskite nanoplates were prepared using a previously described method 39 .The preparation process of PEA 2 SnI 4 perovskite films is the same as that of TEA 2 SnI 4 .For CsSnBr 3 and CsSnBr 2.5 I 0.5 , stoichiometric ratios of CsBr:SnBr 2 = 1:1 and CsBr:CsI:SnBr 2 = 0.5:0.5:1were dissolved in DMSO with a concentration of 0.2 M, and CA was added.

Device fabrication
The ITO-coated glass substrates were ultrasonically cleaned for 20 min with detergent, deionized water, acetone and isopropyl alcohol.Next, the cleaned ITO-coated glass substrates were treated with ultraviolet (UV) light and ozone for 20 min.Modified PEDOT:PSS solution was prepared by mixing PEDOT:PSS solution with deionized water with a volume ratio of 7:3.Then, the PEDOT:PSS solution was spin-coated onto the treated ITO substrate at 4,000 rpm for 30 s and then the substrates were annealed in air at 150 °C for 20 min.After that the substrates were transferred to an N 2 -filled glovebox for 20-min annealing (150 °C) again.After spin-coating the Sn perovskite films as described above, the substrates were transferred to a vacuum thermal evaporator.Finally, 30 nm TPBi, 1 nm LiF and 100 nm Al electrode were deposited in sequence, under a high vacuum of less than 5 × 10 −4 Pa.The effective area was defined using masks as 0.06 cm 2 .

Characterizations
Steady-state PL (Omni-λ 300 Monochromator/Spectrometer, Zolix) and TRPL (Horiba DeltaPro fluorescence lifetime system) measurements were performed with Sn perovskite films spin-coated on quartz substrates.The TRPL curves were fitted with biexponential equations: where I is the normalized PL intensity, A, B 1 and B 2 are the decay amplitudes and A + B 1 + B 2 = 1.τ 1 and τ 2 correspond to the lifetime constants of fast and slow components, respectively.τ average is given by the equation below: For temperature-dependent PL measurements, Sn perovskite films prepared on quartz substrates were mounted in a cryostat ( Janis ST-100) and cooled by liquid nitrogen.The samples were excited by a continuous-wave laser at an excitation wavelength of 400 nm (power density of 2 μJ cm −2 and at 20 K intervals).XRD results were collected from the Rigaku Smart Lab X-ray diffractometer with Cu Kα radiation (λ = 1.54050Å).The GIWAXS was performed at the BL17B1 and BL19U2 beamline of SSRF using an X-ray energy of 10 keV.Two-dimensional patterns of Sn perovskite films on silicon-wafer substrates were acquired by a PLATUS 2M detector mounted vertically at a distance of about 240 mm from the sample with a grazing incidence angle of around 0.2-0.4°and an exposure time of 20 s.Electrochemical impedance spectroscopy experiments of the Sn-PeLED devices were carried out with an electrochemical workstation (CHI660C, CH Instruments).All Sn-PeLED devices were measured in an N 2 -filled glovebox using Keithley 2400 and Konica Minolta CS-200.Electroluminescence characteristics were recorded with a Flame spectrometer and an integrating sphere (Ocean Optic).The sweeping bias range is from 0 V to 10 V with a bias step of 100 mV and a delay time of 1 s.A schematic of the measurement system for our PeLEDs is shown in Supplementary Fig. 12h.The light distribution was measured using a Konica Minolta CS-200 detector, coordinating with various angles from 0º to 90º.Morphology and energy dispersive X-ray spectroscopy of the Sn perovskite films were characterized by scanning electron microscope (Hitachi S-4800).Atomic force microscope measurements were performed using a Bruker Nano ICON2-SYS microscope (Germany).Kelvin probe force microscopy potential and topographic mappings were obtained using a Digital Instruments Multimode atomic force microscope (Veeco Metrology Group), and the used Sn perovskite films were deposited on the ITO/ PEDOT:PSS substrate.The optical properties of the Sn perovskite films were measured using ultraviolet light-visible (UV-vis) absorption spectroscopy (Shimadzu UV-vis-NIR spectrometer (UV-2600i)).Transient-absorption spectroscopy measurements were performed using a commercial set-up (Helios, Ultrafast Systems).The wavelength and the energy density of the pump pulse were 400 nm and 8 μJ cm −2 , respectively.The perovskite films for transient-absorption characterization were fabricated on quartz substrates, spin-coating a layer of PMMA with a concentration of 20 mg ml −1 on the perovskite surface.XPS and ultraviolet photoelectron spectroscopy spectra were obtained with Hitachi Regulus 8100.The excitation-intensity-dependent PLQYs were measured by a custom instrument combining a 405-nm laser, spectrometer and integrating sphere.NMR, 1 H NMR and 13 C NMR spectra were recorded using Bruker AVIII 500MHz FT-NMR.FTIR spectroscopy was tested with Thermo Scientific Nicolet Is10 Instrument, the powder scraped from the perovskite films was mixed with dry KBr powder and ground directly, and pressure was applied to press the powder into a transparent sheet for testing.The electrostatic potentials (ESPs) were calculated using Gaussview software using a self-consistent field-density matrix.The TEM measurements were performed using a JEM-2200FSF microscope.

Defect density and carrier mobility calculations
The defect density can be deduced from the onset voltage of the trap-filled limit region (V TFL ), as shown in equation (3).The carrier mobility of the Sn perovskite films was obtained by the SCLC method using the Mott-Gurney law equation (4): where L is the thickness of perovskite film, ε r is the relative dielectric constant, ε 0 is the vacuum permittivity, q is the charge constant and μ is the carrier mobility.The hole-only device with an architecture of ITO/PEDOT:PSS/perovskite/MoO 3 (30 nm)/Ag (100 nm) was used.The electron-only device with a structure of ITO/TPBi (30 nm)/perovskite/ TPBi (30 nm)/LiF (3 nm)/Ag (100 nm) was used.

Stability measurement
The stability of the LED was measured in an N 2 -filled glovebox at room temperature without encapsulation under dark conditions.The devices were driven by a Keithley 2400 source meter at a constant current, and the luminance was measured using a Konica Minolta CS-200 with an initial luminance of 30 cd m −2 .

Arrhenius equation
where I 0 is the PL intensity at 0 K, k B is the Boltzmann constant and E b is the exciton binding energy.

System model
The structure of TEA 2 SnI 4 was initially built by combining the inorganic framework of FA 2 SnI 4 (FA is formamidinium) with the TEA cations (instead of FA).The distance between the Pb-I layers was initially set to 15.45 Å, as obtained from the XRD measurements.To model the surface, we built a slab containing two inorganic layers (Extended Data Fig. 2) and a vacuum space of 17 Å.For the AIMD simulations, this model was expanded to a 4 × 3 × 1 supercell.

Static DFT calculations
The structural and electronic properties of Sn perovskite surfaces have been investigated by first-principles calculations based on density functional theory (DFT) as implemented in the Vienna ab initio simulation package (VASP) 63,64 .The standard frozen-core projector-augmented wave method was used to describe the interactions between the ions and electrons.Exchange-correlation effects were included using the generalized gradient approximation of the Perdew-Burke-Ernzerhof (PBE) functional 65 .The kinetic energy cutoff was set at 400 eV for the plane-wave expansion.The Monkhorst-Pack k-point grid with a resolution of 0.2 Å −1 was selected to sample the reciprocal space for the calculated systems 66 .A vacuum space of 20 Å was set between the adjacent images to avoid any fictitious interactions.Dipole corrections (IDIPOL = 3) were employed to correct for the slab-to-slab dipole interactions along the surface normal direction.Grimme's dispersion correction scheme combined with Becke-Johnson damping was considered to correct for van der Waals interactions 67 .The bottom layers were fixed at the bulk crystal geometry 68 .The geometry optimizations were performed with a convergence criterion of 1.0 × 10 −6 eV in energy and 0.005 eV Å −1 in forces.Calculations of the capture coefficient calculation were based on the previous literature 69,70 .Non-radiative recombination coefficient B, which is defined as B = N def A, where N def is the defect density from being obtained single carrier devices and A is the non-radiative recombination capture coefficient.

AIMD simulations
The Born-Oppenheimer AIMD (DFT-based) simulations with a mixed Gaussian and plane wave (GPW) basis as implemented in the CP2K package [71][72][73][74] were performed to investigate the interaction between the molecules and the perovskite surface.The valence electrons were expanded in a double-ζ Gaussian basis set with polarization functions (DZVP) 75 .The energy cutoff for the electron density expansion in the GPW method was 400 Ry.The temperature was controlled by the velocity rescaling thermostat of a previous study 76 with a time constant of 0.5 fs.All AIMD simulations were performed at the Γ-point.For the three Sn-based perovskite surfaces (TEA 2 SnI 4 , PEA 2 SnI 4 and CsSnI 3 ), the lattice parameters of the triclinic simulation box were allowed to relax for 2 ps in the isothermal-isobaric ensemble (NPT) with P = 1 atm.The systems were then equilibrated over 5 ps in the NVT ensemble using the supercell volume obtained by the NPT simulation reshaped in an orthorhombic box.All the slabs were considered Sn-I terminated on both sides.The final box sizes as well as the atomic compositions of each system were reported in Supplementary Table 5.For the adsorption of the CA trimer on TEA 2 SnI 4 , 25 ps of simulation runs were used to estimate interaction patterns between the molecules and the surface.The vibrational power spectra are computed as the Fourier transform of the autocorrelation function of the atomic velocities.Three CA concentrations (CA molecules relative to Sn species, 4.16%, 8.33% and 12.5%) and two different starting scenarios (pure keto and pure enol CA isomers) were considered in the simulations (Extended Data Fig. 2).Simulations show that pure keto isomer cannot be adsorbed stably on the Sn perovskite surface at any CA concentration.The final configurations of the CA trimer on the PEA 2 SnI 4 and CsSnI 3 surfaces were considered stable after 10 ps of simulations run, when no conformational changes were further observed.Valeric acid only contains one functional group -thus if the carboxylic acid group is used to form dimers, the molecule can no longer bind to the perovskite surface, as observed with ab initio MD simulations.e, Dimer formation energies (blue) and adsorption energies (red) for CA and valeric acid (VA).f, Vibrational power spectra of the CA keto and enol monomers and dimers in gas phase.The mixed enol-keto form can also explain the observed CO stretch in FTIR.This figure shows the vibrational power spectra of the gas phase molecules of monomers and dimers of the pure keto and pure enol CA form.In the monomeric form, the CN and CO modes overlap at 1587 cm −1 and 1775 cm −1 for the enol and keto forms, respectively.Upon dimerization, these peaks split into three components with the CO contribution blue shifted compared to the CN one.From these results we can conclude that the experimental spectra contain contributions from mixed keto/enol dimers.Furthermore, Fig. 1 shows the computed vibrational power spectra of the most stable trimer configuration in vacuum and adsorbed on the perovskite surface.When the trimer is adsorbed on the surface, a blue shift of 25 cm −1 and 15 cm −1 is reported for CN and CO modes, respectively, in agreement with the experimental data.Another clear signature that CA adsorbs in the form of trimers and/or dimers is the peak around 3000~3300 cm −1 due to OH and NH groups involved in H-bond interactions (CO-HNC); indeed, this kind of peaks are not present for monomers.

Extended Data Table 1 | Summary of device performance for recently reported lead-free PeLEDs
λ EL is the peak wavelength of electroluminescent PeLEDs.

Fig. 1 |
Fig. 1 | Interaction and configuration of CA and TEA 2 SnI 4 perovskite.a, FTIR spectra of TEA 2 SnI 4 with and without CA, and pure CA. b, 1 H NMR (left) and 13 C NMR (right) spectra for pure CA and the composite CA-SnI 2 .c, Vibrational power spectra of the most stable trimer configuration in vacuum and adsorbed on the perovskite surface.d, Trienol and triketo forms of CA. e,f, The most stable configurations for the tautomeric CA trimer (e) and dimer (f) bonded to the Sn perovskite surface.The distances are in ångström (Å).Colour code: Sn, grey; I, purple; S, yellow; C, brown; H, pink; O, red; N, blue.g, Configuration and characteristic interactions of CA on the TEA 2 SnI 4 perovskite surface.Colour code: Sn, grey; I, purple; S, yellow; C, black; H, pink; O, red; N, blue.h, Electron localization function images of pristine perovskite (left) and perovskite with CA (right), which were obtained in systems with an iodide vacancy.i, Charge density difference between the Sn perovskite and CA.a.u., arbitrary units.

Fig. 2 |
Fig. 2 | Crystallization kinetics and non-radiative recombination evaluation.a, Intensity curves of GIWAXS patterns along the (002) ring for the samples with and without CA.Inset, integrated intensity of GIWAXS patterns for the samples.b, Time-dependent normalized PL intensity measurements for the pristine and the TEA 2 SnI 4 film with CA exposed to dry air (20% humidity, room temperature) for 90 min.The horizontal dashed line indicates 90% of the initial normalized PL intensity.Inset, photographs of the samples in air and in the dark with an excitation wavelength of 365 nm.c, Curves of the non-radiative capture coefficient A versus temperature of the Sn perovskites.The vertical dashed line corresponds to room temperature.d, Excitation-intensity-dependent PLQY of the Sn perovskite films with and without CA.e, The distribution of peak XRD intensity for the pristine sample and sample with CA measured in dry air for 12 h.a.u., arbitrary units.

Fig. 3 |
Fig. 3 | Photophysics and carrier dynamics of TEA 2 SnI 4 films.a,b, Temperature-dependent PL spectra mapping for the pristine film (a) and TEA 2 SnI 4 film with CA (b).c,d, Transient absorption (TA) spectra of the pristine film (c) and TEA 2 SnI 4 film with CA (d).

Fig. 4 |
Fig. 4 | Device characterization.a, Energy-level diagram of the devices used in this work.b, Electroluminescence (EL) spectra of one target device with a constant peak emission wavelength of 630 nm under different voltages.Inset, an angle-dependent EL intensity measurement.c, Commission Internationale de L'Eclairage (CIE) colour space plot of one target device.Inset, photograph of a lead-free PeLED with the pattern 'PeLED@JLU'.d, J-V-L curves of the devices with and without CA.e, EQE-J curves of the devices with and without CA.Inset, photograph of a working device (effective area: 1.5 cm × 1.5 cm).f, Histograms of peak EQEs measured from 80 devices in one batch for the Sn-PeLEDs with and without CA.g, Operational stability results of the devices with and without CA.The horizontal dashed line indicates half of the initial luminance.h, A comparison of the EQEs between our work and the previously reported results based on lead-free PeLEDs (Extended Data Table1)16-18, 20, 32-35, 43-62  .a.u., arbitrary units.

Extended Data Fig. 1 |
Density functional theory study.a, Calculated electrostatic potential distribution for enol form of the CA molecule.b, Electron localization function image for the triketo CA-treated perovskite.c, Charge density difference between the tin perovskite and triketo CA.Obtained in systems with an iodide vacancy.d, Planar-averaged charge density difference of the pristine and the Sn perovskite with CA. e, Dielectric coefficients of the Sn perovskites with and without CA along the Z-axis.f, Density of states for the pristine (blue) and the CA-treated perovskite (red) with a surface iodide vacancy.Extended Data Fig. 2 | Configuration and FTIR studies.a, b, Built slab containing two inorganic layers (a) and a vacuum space of 17 Å.For the ab-initio molecular dynamics simulations, this model was expanded to a 4×3×1 supercell (b).Color code: Sn in grey, I in purple, S in yellow, C in black, H in pink, O in red, N in blue.c, AIMD snapshoot of a stable dimer at a perovskite grain boundary.The distances are in Å. Color code: Sn in grey, I in purple, S in yellow, C in black, H in pink, O in red, N in blue.d, Valeric acid dimer on top of a pristine TEA 2 SnI 4 surface.The protons are shared between two oxygens with distances of 1.0 Å and 1.6 Å.

Fig. 3 |
Morphological characteristics.a, b, SEM images of the perovskite films without (a) and with 5% CA (b).c, d, AFM images of the pristine (c) and the 5% CA-treated sample (d).e, KPFM measurements for the pristine and the CA-treated sample.f, Linear potential profiles for the pristine and the CA-treated sample.Extended Data Fig. 4 | Crystal structure of TEA 2 SnI 4 films.a, b, GIWAXS images of TEA 2 SnI 4 perovskite films without (a) and with 5% CA (b).c, XRD patterns of the pristine and the samples with 2%, 5% and 8% CA. d, Comparison of (002) peak for the samples with and without 5% CA.Extended Data Fig. 5 | XPS and air-stability analysis of TEA 2 SnI 4 perovskites.a, XPS spectra of Sn 3d for the pristine and the CA-treated TEA 2 SnI 4 .b, XPS spectra of Sn 3d for the pristine and the CA-treated TEA 2 SnI 4 stored in dry air for 12 h.c, XPS results of I 3d spectra for the tin perovskite films without and with 5% CA. d, Ratios of Sn versus I from the XPS spectra for the tin perovskite samples with and without CA.e, XPS spectra of Sn 3d for the pristine and the CA-treated TEA 2 SnI 4 nanoplates.f, XPS spectra of Sn 3d for the pristine and the CA-treated TEA 2 SnI 4 nanoplates stored in dry air for 12 h.g, XRD patterns of TEA 2 SnI 4 nanoplates with and without CA.Time-dependent XRD patterns of TEA 2 SnI 4 nanoplates without (h) and with CA (i) stored in dry air for 24 h.

j
-m, High-resolution TEM images of perovskite films.Insets: fast Fourier transform diffractograms.Scale bar, 5 nm.To evaluate the surface effect and analysis on how the air-stability relates with crystal stability, we prepared the TEA 2 SnI 4 nanoplates with and without CA.From the XPS and XRD results (Extended Data Fig.5e-i), the TEA 2 SnI 4 nanoplates with CA showed less Sn 4+ content, thus proving that CA can enhance the air stability of both perovskite films and nanoplates.From the TEM images (Extended Data Fig.5j-m),It can be clearly seen that the perovskite with CA still show apparent crystal lattices even after 12h storage in dry air, indicating the better crystal stability with the introduction of CA.Extended Data Fig. 6 | Optical properties of the Sn perovskite films.a, PL spectra of the Sn perovskite films with various CA content.b, Logarithm curves of absorption coefficient (α) versus photon energy, the E U for the pristine and the treated sample are estimated.c, Visible absorption spectra of the Sn perovskite films with various CA content.d, TRPL spectra of the TEA 2 SnI 4 films prepared with and without 5% CA. e, Non-radiative recombination coefficients B for the single-carrier devices with and without CA treatment, the dashed line indicates the condition of room temperature.f, Excitation-intensity-dependent PLQYs of the perovskite films with different CA content.Extended Data Fig. 7 | Photophysical characterization of TEA 2 SnI 4 films.a, The fitted curves of the integrated PL intensity as a function of 1/T for the perovskites with and without CA.b, Kinetic traces at a probing wavelength of 615 nm for the Sn perovskite films with and without CA.c, d, Pseudo-color maps of femtosecond-transient absorption spectra of the pristine (c) and the sample with 5% CA (d) under an excitation wavelength of 400 nm.Extended Data Fig. 8 | Device characterization.a-d, Device performance for the Sn-PeLEDs with 2% and 8% CA: (a) J-V, (b) L-V, (c) CE-J, and (d) EQE-J curves.Extended Data Fig. 9 | The efficacy of CA on 2D PEA 2 SnI 4 perovskite.a, XRD patterns of the perovskite films with and without 5% CA. b, Absorption spectra of the perovskite films with and without CA.c, Logarithm curves of absorption coefficient (α) versus photon energy, and the Urbach energy (E U ) for the pristine and the treated sample.d, PL spectra of the films with and without CA.e, PLQYs of the perovskite films with and without CA.f, TRPL spectra of the films prepared with and without CA.g, FTIR spectra of the PEA 2 SnI 4 with and without CA, and the pure CA. h, XPS spectra of Sn 3d for the pristine and the CA-treated PEA 2 SnI 4 .Time-dependent normalized PL intensity for the pristine (i) and the film with CA ( j) exposed to dry air (20% humidity, RT) for 60 min.Time-dependent XRD patterns of the pristine (k) and the sample with CA (l) in dry air for 12 h.m, The most stable configurations for the tautomeric CA trimer bonded to the Sn perovskite surface.The distances are in Å. Color code: Sn in grey, I in purple, C in brown, H in pink, O in red, N in blue.n, Crystal orbital Hamilton population (COHP) plots of the local Sn-N bond after CA adsorption.o, Averaged integrated crystal orbital Hamilton populations (ICOHP) of the Sn-N bonds.p, EL spectra of one target device with constant peak emission wavelength of 625 nm under different voltages.J-V (q), L-V (r) and EQE-J curves (s) of the PEA 2 SnI 4 based PeLEDs with (9.22%) and without (2.87%)CA.