Introduction

All-inorganic halide perovskites such as CsPbX3 (X = Cl, Br, I) are attractive materials for light emitters and photovoltaic applications due to their potential in overcoming long-term stability issues of organic–inorganic hybrid halide perovskites1,2,3,4,5. The low-temperature solution-processed phase-stable CsPbX3 perovskites are mainly based on bromide and chloride with suitable tolerance factors6. Optically active CsPbI3 black phases (α (cubic), β (tetragonal) or γ (orthorhombic)) usually require high annealing temperature (300–370 °C) to overcome the energy barrier for phase transition7,8. Moreover, the CsPbI3 black phases can readily transform to yellow-phase δ-CsPbI3 in ambient conditions due to the thermodynamic instability9,10,11, inhibiting their application in optoelectronic devices12,13. In perovskite solar cells, a general method of forming black phases CsPbI3 at low temperature is by adding hydroiodic acid (HI) in CsPbI3 precursor solution prior to spin coating13,14. It has been observed that the judicious amount of HI would decompose the solvent dimethylformamide (DMF) to form dimethylammonium iodide15, while the mechanism of how this process affects the crystallization of CsPbI3 is still under intensive debate with two arguments. One argument is that the formed DMA would sublimate and lead to a fast crystallization of CsPbI34,16,17. Another argument is that the DMA becomes the part of the crystal structure and the formed black phase is not CsPbI3 but CsxDMA1−xPbI318,19.

For CsPbI3-based light-emitting diodes (LEDs) applications, the low-temperature HI doping method is difficult to achieve high performance devices, mainly due to the high trap density and strong nonradiative recombination with those perovskite films (typical photoluminescence quantum efficiency (PLQE) < 1%)20. Alternatively, high-efficiency LEDs has been demonstrated based on CsPbI3 quantum dots (QDs)21,22. However, those colloidal QDs are synthesized ex situ in flasks by the hot-injection method, which usually requires a temperature above 170 °C and complicated processing conditions21,22,23,24. In addition, usually those perovskite QD-based LEDs only show high efficiency at low current densities with strong efficiency roll-off due to the strong nonradiative Auger process in perovskite QDs21,25,26. In this work, we report an effective approach for achieving high quality γ-CsPbI3 at low annealing temperature (~100 °C) for high performance LEDs applications. More importantly, we reveal that the low-temperature formation process of black phase CsPbI3 can be generally observed when intermediate states are formed, followed by an interionic exchange in the presence of large organic ammonium cations.

Results

Low temperature formed γ-CsPbI3 films

A DMF precursor solution of imidazolium iodide (IZI), CsI, and PbI2 with a molar ratio of 4:1.5:1 (referred as IZI-CsPbI3) is spin coated onto polyethylenimine ethoxylated (PEIE) modified ZnO substrates (referred as ZnO/PEIE). We note that ZnO/PEIE has been widely used as an electron transporting layer in perovskite LEDs27. After thermal annealing at 100 °C for 5 min, the IZI-CsPbI3 film shows X-ray diffraction (XRD) peak of 14.3 and 28.9° without any splitting (Fig. 1a), corresponding to the (110) and (220) crystal planes of γ-CsPbI3, respectively. An absorbance edge at ~1.75 eV and a photoluminescence (PL) peak at ~700 nm are also observed for this film (Fig. 1b). These results are consistent with the characteristics of γ-CsPbI3 obtained through thermal annealing above the transition temperature (around 310 °C) and rapid cooling process8,13. In addition, the scanning electron microscopic (SEM) measurement shows that the IZI-CsPbI3 film is discrete, consisting of particles with an average size of ~80 nm (Fig. 1c). The film shows good emission properties with PLQE reaching up to 38% (Fig. 1d). Time-correlated single photon counting measurement shows that the PL lifetime increases with the increasing amount of IZI (Supplementary Fig. 1a), suggesting that the nonradiative recombination of the γ-CsPbI3 is suppressed with increasing IZI. This result is consistent with the PLQE result (Supplementary Fig. 1b). More importantly, the IZI-CsPbI3 film exhibits negligible degradation after exposing for 36 days in ambient air at room temperature with 80% relative humidity (Fig. 1d). In contrast, the regular γ-CsPbI3 obtained from the high-temperature annealing process can only retain for 4 h in the same environment (Supplementary Fig. 2). The PL intensity of IZI-CsPbI3 film dropped to 50% over 8 days in the ambient air (Supplementary Fig. 3), suggesting significantly improved phase and optical stability compared to previously reported results4,19.

Fig. 1: Characteristics of IZI-CsPbI3 film on ZnO/PEIE substrate.
figure 1

a XRD pattern after exposed in the air with 80% relative humidity for various durations. Black line is the simulated powder patterns of γ-CsPbI38. b UV–vis and PL spectrum. Inset shows the image under UV light. c SEM image (scale bar, 200 nm), and d Excitation-intensity-dependent PLQE.

The mechanism behind low temperature formed γ-CsPbI3 films

To investigate the mechanism of IZI on facilitating the formation of γ-CsPbI3 perovskite at the low temperature, we monitor the crystal phase evolution of the as-spun IZI-CsPbI3 film by XRD measurements under various annealing time (Fig. 2a). At early stage of the thermal annealing process (10–15 s), CsI with the peak at 27.6° remains unchanged, while an intermediate phase with peaks at 11.3 and 25.4° is formed. This intermediate phase can be assigned to one dimensional (1D) IZPbI3, since their XRD peaks are consistent (Fig. 2a red line and Supplementary Table 1). Upon further annealing (~30–60 s), both the XRD peaks (11.3, 25.4, and 27.6°) of intermediate phase IZPbI3 and CsI disappear. In the meantime, XRD peaks (14.3 and 28.9°) of γ-CsPbI3 perovskite appear. These facts suggest that the 1D IZPbI3 perovskite transforms to the γ-CsPbI3 perovskite during the low-temperature annealing process. This transformation process requires an interionic exchange process of IZ+ embedded in face-shared PbI6 chains with external Cs+. The corresponding morphology of IZI-CsPbI3 films annealed at 100 °C for various time durations is also monitored by SEM measurement (Supplementary Fig. 4). The unannealed film displays a dense, planar morphology (Supplementary Fig. 4a). With a short-time annealing (t = 10 s), mounts of small grains of about 40 nm emerge (Supplementary Fig. 4b), corresponding to the intermediate phase. By extending the annealing time duration to 15 s, the small grains grow bigger and the layer becomes discrete (Supplementary Fig. 4c), corresponding to the mixed phase with 1D and 3D. When annealed over 30 s, the discrete γ-CsPbI3 grains with an average size of  ~80 nm form and disperse on the ZnO/PEIE substrate (Supplementary Fig. 4d–f).

Fig. 2: Formation process of γ-CsPbI3 perovskite on ZnO/PEIE substrate.
figure 2

a XRD patterns of IZI-CsPbI3 film on ZnO/PEIE substrate with different annealing time at 100 °C. XPS spectra of (b) N 1s, (c) O 1s, (d) Zn 2p1/2 and Zn 2p3/2 for IZI, IZI-CsPbI3 film on ZnO substrate and for IZI, IZ film on ITO substrate, respectively. e Schematic diagram of deprotonation of IZ+ with ZnO. f Mechanism of the γ-CsPbI3 formation process at low temperature on ZnO substrate.

We find that without the underneath ultrathin PEIE layer, the low-temperature phase transformation can be still observed by XRD measurement (Supplementary Fig. 5). It cannot be formed at 100 °C without ZnO layer, where both the 1D phase and CsI remain unchanged even after 10 min annealing (Supplementary Fig. 6). UV–vis absorption spectra measurement result is consistent with the above XRD result (Supplementary Fig. 7). These results suggest that during the formation of γ-CsPbI3 from the intermediate phase IZPbI3, the ZnO substrate plays important roles in the interionic exchange process.

We then use X-ray photoelectron spectroscopy (XPS) measurement to reveal the role of the ZnO substrate in the interionic exchange process. Particularly, the chemical interaction between the films (IZI, IZI-CsPbI3) and ZnO is investigated. Figure 2b shows the high resolution XPS spectra of N 1s, O 1s, and Zn 2p of these films. The XPS spectrum of IZI prepared on ITO substrate shows two N 1s peaks at around 401.7 eV (N-3) and 400.2 eV (N-1), respectively (Fig. 2b black and pink line). When IZI film is on top of ZnO layer, the N-3 peak disappears and a new peak positioning at a lower binding energy of 398.8 eV shows up. The two peaks of the ZnO/IZI film are in well agreement with those of imidazole (IZ), and the peak at 398.8 eV can be attributed to the N-2 of IZ (Fig. 2b blue line)28,29. The above XPS result indicates that the ZnO can deprotonate the IZ+ cation of IZI. Importantly, this deprotonation process can be also observed in the IZI-CsPbI3 film with ZnO and ZnO/PEIE substrate, respectively (Fig. 2b and Supplementary Fig. 8). Moreover, compared to the bare ZnO film on ITO, the ZnO/IZI and ZnO/IZI-CsPbI3 films show significantly suppressed XPS peak at 530 eV ascribed to O 1s of Zn–O bonding (Fig. 2c black line)30,31, and their peaks of the Zn 2p of Zn–O bonding shift to higher energies (Fig. 2d). These further XPS observations can be explained by the presence of H+ in the Zn–O–H bonding which decreases electron cloud density around Zn atoms32, consistent with the result of deprotonation of the IZ+ cation by ZnO. Figure 2e shows a scheme of this deprotonation process of IZ+ by ZnO. In addition, attenuated total reflection Fourier-transform infrared (ATR-FTIR) spectroscopy measurement result can further confirm this scenario. As shown in Supplementary Fig. 9, the N–H stretching vibration at 3260 cm−1 of IZI in the films on the ZnO/PEIE substrate is shifted to higher wavenumber (3330 cm−1). Correspondingly, the broad stretching vibration of Zn–O is red shifted from 556 to 538 cm−1,33. And the signature peak of IZ at about 3130 (CH stretching), 1543 (NH bend), 1328 (CH bend), 1263 (ring breathing), 1055 (CH bend), 841 (ring bend), 757 (CH out-of-plane bend), and 658 (torsion) cm−1 are clearly observed34. We note that the similar deprotonation process can also be observed in films with the intermediate phase of IZPbI3 (Supplementary Fig. 10) and FAPbI3 perovskites35.

On the basis of the above phase evolution and chemical elementary analysis, we can have a clear picture of the mechanism of the γ-CsPbI3 formation process at low temperature, as shown in Fig. 2f. It first forms an intermediate phase IZPbI3, followed by the formation of γ-CsPbI3 through the interionic exchange of IZ+ with external Cs+ in the process of the deprotonation of IZ+ with ZnO. The overall chemical reaction of phase formation can be presented as follows:

\({\mathrm{Stage}}\;1:\!\;{\mathrm{PbI}}_2 + {\mathrm{C}}_3{\mathrm{N}}_2{\mathrm{H}}_5{\mathrm{I}} + {\mathrm{Cs}}^ + \to ({\mathrm{C}}_3{\mathrm{N}}_2{\mathrm{H}}_5){\mathrm{PbI}}_3 + {\mathrm{Cs}}^ + {\mathrm{forming}}\;{\mathrm{intermediate}}\;{\mathrm{phase}},\)

\({\mathrm{Stage}}\;2:\!\!\;2({\mathrm{C}}_3{\mathrm{N}}_2{\mathrm{H}}_5){\mathrm{PbI}}_3 + 2{\mathrm{Cs}}^{+} + 2{\mathrm{ZnO}} \to 2{\mathrm{CsPbI}}_3 + 2{\mathrm{C}}_3{\mathrm{N}}_2{\mathrm{H}}_4 + {\mathrm{Zn}}({\mathrm{OH}})_2 + {\mathrm{Zn}}^{2 + }\; {\mathrm{ion}}\;{\mathrm{exchange}}\;{\mathrm{and}}\;{\mathrm{forming}}\;\gamma \!-\!{\mathrm{CsPbI}}_{\mathrm{3}}.\)

Since the ZnO induced deprotonation process likely mainly occurs at the ZnO interface, it is interesting to investigate how thick the perovskite can be formed by this approach. Supplementary Fig. 11 shows the absorbance of the films with various thickness fabricated from different concentration of precursor solutions. It shows the absorbance at 687 nm from the black phase CsPbI3 increases linearly with the film thickness <200 nm. Above 200 nm, the absorbance saturates and declines. This result suggests that the ZnO substrate can facilitate the γ-CsPbI3 formation within the film thickness of 200 nm.

Kinetic of phase transition

We further investigate the crystallization kinetics of the γ-CsPbI3 films with various IZI contents by using time-dependent UV–vis spectroscopy. The reaction progress of these films with different contents of IZI in precursor solutions (IZI/PbI2 molar ratio is x, x = 0, 0.5, 1, 2, 3, 4, and 5) were monitored through change in absorbance (A(t)) at ~687 nm (Fig. 1b). The value of formed γ-phase fraction χ(t) is defined as:

$${\upchi }(t) = \frac{{A(t)}}{{A(t_{\mathrm{end}})}},$$
(1)

where the A(tend) represents the maximum absorbance at the final state. As shown in Fig. 3a, the rate of γ-phase formation increases with the increasing amount of IZI at a constant annealing temperature of 90 °C. And Fig. 3b shows that the rate also increases with increasing annealing temperature when the IZI content is constant. More detailed measurement results on different mole ratios and various annealing temperature are shown in Supplementary Fig. 12. The activation energy of the γ-CsPbI3 formation for various IZI contents can be estimated by using the Mittemeijer model36, as shown in Supplementary Fig. 12g. Figure 3c shows a summary of the estimated activation energies barrier, which can be significantly decreased from 150 to 29 kJ mol−1 with increasing molar ratio of IZI to PbI2 from 0.5 to 5.

Fig. 3: Kinetic data for isothermally annealed IZI-CsPbI3 films on ZnO/PEIE substrate.
figure 3

Influence of annealing time on the γ-phase transition fraction χ(t) for (a) the IZI-CsPbI3 films with different mole ratio (0:1, 0.5:1, 1:1, 2:1, 3:1, 4:1, 5:1) at 90 °C, and for (b) IZI-CsPbI3 with mole ratio of 4:1 at different temperature. χ(t) is defined as A(t)/A(tend), A(t) represents the time-dependent absorbance at 687 nm, A(tend) represents the max absorbance. c The dependence of activation energies of the γ-CsPbI3 films on the IZI content.

Generality of the low temperature formed γ-CsPbI3 films

In order to further demonstrate the generality of the low-temperature formation of black CsPbI3 through the intermediates, we add various RNH3+-based large organic cations, such as butylammonium iodine (BAI), hexylammonium iodine (HAI), phenethylammonium iodine (PEAI), and naphthylethylammonium iodine (NMAI) into the CsPbI3 precursor and spin coated on top of ZnO substrates. As shown in Supplementary Fig. 13, all those films exhibit formation of intermediates, decomposition, and interionic exchange during the annealing process. And finally, the black CsPbI3 films form at 100 °C. Therefore, we believe that the process we plotted in Fig. 2f is general for forming black CsPbI3 at low temperature assisted by RNH3+-based large organic cations.

LED device performance

As shown in Fig. 4a, the LED devices have a structure of ITO/ZnO/PEIE/IZI-CsPbI3/poly (9,9-dioctyl-fluorene-co-N-(4-butylphenyl)diphenylamine) (TFB)/molybdenum oxide (MoOx)/gold (Au). Detailed fabrication process can be found in “Methods.” The valence bands of IZI-CsPbI3 film is obtained via ultraviolet photoelectron spectroscopy measurement (Supplementary Fig. 14), and the conduction band was estimated using the band gap derived from the absorption band edge (Fig. 1b). A cross-sectional image of the device shows the formation of the discrete CsPbI3 particles layer with a thickness of ~40 nm (Fig. 4b). We note that the residual IZI can locate between the discrete particles, which can prevent the device from short-circuit current and enhance the light-outcoupling efficiency, and similar studies have been demonstrated early37,38,39. The current density and luminance versus voltage (J–L–V) and EQE curves of the device with different contents of IZI are shown in Supplementary Fig. 15. The electroluminescence (EL) peak is located at ~698 nm and the shape remains unchanged under different bias voltages (Fig. 4c). The champion device based on the IZI-CsPbI3 film (mole ratio of 4:1.5:1 for IZI:CsI:PbI2) exhibits a peak EQE of 10.4% with luminance of 340 cd m−2, and the turn-on voltage is as low as 2 V (Fig. 4d, e). We note that our device peak EQE is a record for 3D CsPbI3 film-based red LEDs40,41. Moreover, the efficiency roll-off of the device is significantly suppressed, and the EQE remains high (~8%), under a high current density of 100 mA cm−2 (Fig. 4e). This feature is very different compared with previous QD-based CsPbI3 LED devices21,26, where high EQE can only be obtained at low excitations likely due to the strong nonradiative Auger process. In addition, the IZI-CsPbI3 devices exhibit highly reproducible with average EQE of 8.1% for 75 devices (Supplementary Fig. 15c). The best device shows a half-lifetime of 20 min at a constant current density of 100 mA cm−2, and the EL peak position remains constant over time (Fig. 4f).

Fig. 4: IZI-CsPbI3 LED structure and optoelectronic characteristics.
figure 4

a Flat-band energy level diagram and b Cross-sectional SEM image (scale bar, 200 nm). c EL spectra under various voltages. Inset shows the photograph of the deep-red device. d Current density and luminance versus voltage (J–L–V). e EQE versus current density (EQE–J) curves. f Lifetime and the maximum EL peak position measurement at 100 mA cm−2. The maximum EQE as the initial value to calculate the T50.

Discussion

In summary, we have developed a low-temperature method of forming CsPbI3 black phases for high performance CsPbI3 LED applications via synergistic effect of IZI and ZnO electron transport layer. The judicious amount of IZ+ in the precursor promotes the intermediate phase formation, followed by the formation of γ-CsPbI3 through the interionic exchange of IZ+ with external Cs+ in the process of the deprotonation of IZ+ with ZnO. The phase transition engineering can efficiently reduce the formation energy of CsPbI3 black phase, and facilitate the formation of discrete CsPbI3 particles film with high PLQE and long-term stability. The resulting CsPbI3 LED shows a peak EQE of 10.4% with suppressed efficiency roll-off. Importantly, the low-temperature formation process can be generally observed with various RNH3+-based large organic cations. So we believe that our work provides useful perspectives to tune the phase transition pathway, and offers an effective approach to fabricate low-temperature processed CsPbI3 black phase film for LEDs applications.

Methods

Synthesis of ZnO colloidal solution

ZnO were synthesized by following the previously reported method27. The dimethyl sulfoxide solution of Zn(Ac)2·2H2O (3 mmol in 30 mL) was mixed in ethanol solution of tetramethylammonium hydroxide pentahydrate (TMAH·5H2O) (5.6 mmol in 10 mL) and stirred at 30 °C for 24 h. The ZnO colloids were precipitated with ethyl acetate, and washed it three times with ethanol and ethyl acetate. Finally, the obtained ZnO colloid were dispersed in ethanol and set aside in the fridge until serve.

Synthesis of organic ammonium salt

IZI was prepared by mixing IZ (2 g) and excess hydroiodic acid (45 wt% in water) in 15 mL of ethanol at 0 °C. After the reaction mixture was stirred for 2 h, 60 mL diethyl ether was added into the mixture to obtain the precipitates. The collected precipitates were washed three times with diethyl ether and stored in an oven. BAI, HAI, PEAI, and NMAI were prepared by similar method.

Perovskite precursor solutions preparation

The CsPbI3 precursor solution were prepared by dissolving IZI, CsI, and PbI2 with a molar ratio of x in DMF at weight percent (wt%) of 6% (IZI and PbI2 molar ratio is x:1, x = 0, 1, 2, 3, 4, CsI and PbI2 molar ratio fixed at 1.5:1), and the solution with molar ratio of x = 4 is referred as IZI-CsPbI3. The IZI–PbI2 precursor solutions were prepared by dissolving IZI and PbI2 with a molar ratio of 4:1 in DMF. BAI-CsPbI3, HAI-CsPbI3, PEAI-CsPbI3, and NMAI-CsPbI3 precursor solution were prepared by dissolving BAI, HAI, PEAI, NMAI in DMF solution of CsI and PbI2 with molar ratio of 4:1.5:1, 2:1.5:1, 4:1.5:1, 2:1.5:1, respectively.

Perovskite film deposition

ZnO colloidal solution was deposited onto the ITO substrate using spin-coating technique at 4000 rpm for 45 s, followed by annealing at 150 °C for 30 min The PEIE (1.5 mg mL−1 in methoxyethanol) was spin coated onto the ZnO films at a speed of 5000 rpm and annealed at 100 °C for 10 min. Finally, the precursor solution was spin coated onto ITO/ZnO/PEIE substrate (4000 rpm, 30 s) or ITO substrate with 100 °C annealing for various time to form films, respectively.

Device fabrication

The devices were fabricated with a structure of ITO/ZnO/PEIE/perovskite/TFB/MoOx/Au. After the deposition of the ZnO, PEIE and perovskite films as mentioned above, the TFB (12 mg mL−1 in chlorobenzene) layer was spin coated by 3000 rpm for 30 s. Finally, MoOx (7 nm) and Au (80 nm) were deposited by thermal evaporation, respectively

Perovskite film characterization

XRD measurements were performed with a Rigaku Smart lab (3 kW) XRD patterns with Bragg–Brentano focusing, a diffracted beam monochromator and a conventional Cu target X-ray tube set to 40 kV and 30 mA. Time-dependent UV–vis absorption spectra were obtained on PerkinElmer Lambda 950 spectrometer. The general morphologies of the films were characterized by FEI (Quanta 200 FEG) SEM under a voltage of 5 kV. XPS tests were carried out using a Thermo ESCALAB250 Xi X-ray photoelectron spectrometer with Al Kα X-ray as the excitation source. All binding energies were referred to the C 1s peak at 284.8 eV of the surface areas of the samples. ATR-FTIR spectra of films were characterized by a Thermo-Niclet IS50 equipped with a Smart SAGA reflectance accessory in the range of 450–4000 cm−1. PL spectra were obtained using a fluorescent spectrophotometer (F-4600, HITACHI) with a 200 W Xe lamp as an excitation source. The Excitation-intensity-dependent PLQE of perovskite films was monitored by a joint control of a 450 nm continuous wave laser, 1000 μm slit width, optical fiber spectrometer, and integrating sphere42. The film thickness was determined by a surface profiler (KLA-Tencor).

Kinetic modeling of the phase transition

The dependence of rate on temperature indicates a significant activation energy barrier (Ea) for the process of the γ-CsPbI3 formation. We worked out the activation energy barrier using the Mittemeijer model:36

$${\mathrm{ln}}\left(t_{\mathrm{x}2} - {t_{\mathrm{x}1}} \right) = \frac{{E}_{\mathrm{a}}}{{RT}} - {\mathrm{ln}}\;{k}_0 + {\mathrm{ln}}\left( {\beta _{\mathrm{x}2} - \beta _{\mathrm{x}1}} \right),$$
(2)

where Ea is the effective activation energy barrier, tx1 and tx2 are the annealing time at which the transformed fraction is χ(t) = 0.2 and 0.8, R is the gas constant, T is the temperature, and k0 is a rate constant prefactor.

Device characterization

The LED was measured in glove box at room temperature, and detailed setup can be found in reference38. A Keithley 2400 source meter with a step of 0.05 V s−1 and a fiber integration sphere (FOIS-1) coupled with a QE65 Prospectrometer were used for the device measurements. The device area is 7.25 mm2. The device lifetime was measured by using the same setup under a constant current density of 100 mA cm−2.