Recrystallization of FAPbI 3 film induced by MASCN treatment. The FAPbI3 film is fabricated by a one-step antisolvent method (details are provided in the Experimental section), and the initial film without posttreatment is denoted Pristine-FAPbI3. The back-light image shows that Pristine-FAPbI3 is a deep brown film (Fig. 1a, graph ⅰ), where the corresponding scanning electron microscopy (SEM) image (Fig. 1b, graph ⅰ) indicates that the film is a compact film without holes. Then, the film is posttreated by spin-coating with a drop of MASCN/isopropanol (IPA) solution (10 mg mL− 1) and exposed to an air environment. Here, the film exposed for x minutes is denoted x min-FAPbI3 and characterized with a time gap of 10 minutes from 0 to 60 minutes (corresponding to the graph from ⅱ to ⅷ in Fig. 1a, b). The initial film (0 min-FAPbI3) is still visually deep brown (Fig. 1a, graph ⅱ); however, the SEM image shows that its morphology has become mesoporous (Fig. 1b, graph ⅱ). Over time, the film becomes rough and transparent, where the backlight turns from brown to white (Fig. 1a, the graph from ⅲ to ⅷ) and the morphology changes dramatically from a mesoporous film to large connecting crystals with holes among them (Fig. 1b, the graph from ⅲ to ⅷ). The bright-field optical microscope (BFOM) images exhibit a continuous in situ variation of the film, where small crystals gradually merge into a large one with holes emerging (Supplementary Fig. 1, the graphs from ⅰ to ⅷ correspond to the film from ⅰ to ⅷ in Fig. 1a, b). The variation spontaneously proceeds under an ambient environment, so some in situ characterizations can be employed to study this process. Ultraviolet (UV)-visible spectroscopy was employed to observe the in situ changes in film absorption from 0 min-FAPbI3 to 60 min-FAPbI3 (Fig. 1c). The absorption edges indicate that the bandgap of the film does not change, while the absorption intensity gradually decreases to a flat line, which is caused by the emerging pores on the film. However, the photoluminescence (PL) intensity of 60 min-FAPbI3 increases to several times stronger than that of 0 min-FAPbI3 (Fig. 1d), indicating that carrier recombination is prevented and that the trap states are eliminated in this process (for more discussion, see Supplementary Note 1). The X-ray diffraction (XRD) patterns support this modification effect (Fig. 1e). From 0 min-FAPbI3 to 60 min-FAPbI3, the XRD signal intensity keeps increasing, indicating the crystallization is gradually strengthened. The diffraction angle before 21° is further scanned with a tiny step of 0.002°, and the results are shown in Supplementary Fig. 2. The color map of intensity shows the normalized intensity of the (001) peak of α-FAPbI3 at 14°, and its full width at half maximum (FWHM) gradually decreases, indicating a continuous modification of crystallization (Supplementary Fig. 2a). In Supplementary Fig. 2b, the intensity ratio between the (001) facet at 14° and the (011) facet at 20° is compared to exhibit the crystal orientation. As the intensity of (001) is normalized as 1, the value of (011) gradually decreases, indicating an increasing crystal orientation along the (001) facet30,31. The characterizations of Pristine-FAPbI3 without treatment are shown in Supplementary Fig. 3a (PL) and 3b (XRD) as a reference, and there are no variations. All these results indicate that after posttreatment by MASCN, the FAPbI3 film experiences a recrystallization process, which changes its morphology and optimizes its crystallization.
Different behaviors of MA + and SCN− group. The Pristine-FAPbI3 film without treatment does not tend to vary so quickly even in a humid environment; thus, the effect of MASCN should be evaluated. The chemical components of the MASCN molecule are shown in Fig. 2a, which consists of an MA+ ion and an SCN− ion. They will be evaluated individually, where methylamine iodine (MAI) and formamidine thiocyanate (FASCN) are chosen to replace MASCN to treat the FAPbI3 film because FA+ and I− ions have tiny effects on FAPbI3. For the film posttreated by MAI, the XRD pattern shows that there is merely a difference from the initial films, and only the signal of PbI2 arises, which is caused by the decomposition of some produced MAPbI3 while annealing (Fig. 2b, MAI-FAPbI3). For the film posttreated by FASCN, interestingly, the film transforms from the α-phase to the δ-phase (Fig. 2b, FASCN-FAPbI3). Further experiments showed that posttreatment with formamidine iodide (FAI) did not cause a phase transition (Supplementary Fig. 4); thus, such a phase transition was caused by the SCN− group. Grätzel and his coworkers reported that MASCN can cause a phase transition of FAPbI3 from δ-phase to α-phase by surface interaction32, and our posttreatment method on δ-phase FAPbI3 film verifies this phenomenon (Fig. 2c). However, our reference experiment on FASCN shows that SCN− causes a reversed-phase transition from the α-phase to the δ-phase. To evaluate the difference between MASCN and FASCN, posttreatment with different concentrations was applied. For the posttreatment with MASCN, high concentrations (20 mg mL− 1 and 10 mg mL− 1, Supplementary Fig. 5a, b) optimized the crystallization of FAPbI3 in accordance with the results in Fig. 1e. However, when the concentration is low (5 mg mL− 1 and 2.5 mg mL− 1, Supplementary Fig. 5c, d), the α-phase FAPbI3 film transforms into δ-phase FAPbI3. The film treated with FASCN will cause the transition from the α-phase to the δ-phase despite the treatment concentration (Supplementary Fig. 6). Thus, it is concluded that the SCN− group will cause only an α to δ phase transition when individually, but when MA+ is involved, the effect will depend on the treatment concentration. As a result, the effect of MASCN can be divided into two parts: the SCN− group will dominate the phase transition effect on FAPbI3, and the MA+ ion will adjust this effect. Individual SCN− will just cause the α to δ transition, and when enough MA+ ions are involved, the effect of SCN− becomes reversed, causing a δ to α transition. Further discussion of the different phase transition behaviors is presented in Supplementary Note 2.
In situ time-dependent Fourier transform-transform infrared spectroscopy (FTIR) was employed to further characterize the behavior of MASCN (Fig. 2d). Three wavenumber ranges were selected to represent the three organic groups during the variation process. They are wavenumbers of 2100 − 2025 cm− 1 (stretch of C ≡ N, SCN− group, inset graph ⅰ), 1730 − 1690 cm− 1 (stretch of C = N, FA+ group, inset graph ⅱ), and 960 − 900 cm− 1 (twist of CH3-NH3+, MA+ group, inset graph ⅲ)33. The normalized variations are depicted in the color map in Figs. 2e-g. In Fig. 2e, the signal of SCN− has no change with time, neither the peak position nor FWHM. This result means that the chemical environment of SCN− did not change during this process. Considering that MASCN is spin-coated on the surface of the perovskite film, the SCN− group keeps its initial position, where it stays at the surface of the perovskite crystal. In Fig. 2f, a continuous redshift is observed for the C = N bond, indicating that the chemical environment of FA+ group has a continuous variation. In Fig. 2g, the signal of the MA+ group has an obvious redshift for the first 10 minutes and then remains in a tiny disturbance, indicating that it is not static. The behaviors of MA+ and FA+ can be understood as follows: during the recrystallization process, the MA+ ions will quickly penetrate into the crystal of the FAPbI3 film and then diffuse, resulting in the disturbance in Fig. 2g. This diffusion also causes a continuous doping process for the crystal, which can be confirmed by the continuous peak shift toward a large degree in the XRD patterns (Supplementary Fig. 7). Previous studies have reported that the doping of SCN− will not cause a degree shift34,35, so the shift must be caused by the doping of MA+ ions. The MA+ ion diffuses deeper, thus causing continuous doping of the FAPbI3 film and resulting in a continuous peak shift for the FA+ group, as shown in Fig. 2f. In conclusion, the in situ time-dependent FTIR spectra show that the SCN− group will stay at the surface of the film, while MA+ will continue to diffuse in the crystal. Further treatment is executed on the MASCN-treated film to observe the different behaviors of the MA+ and SCN− groups, where the film is washed with pure IPA. This process can wash the MA+ and SCN− ions on the surface while the MA+ ions inside the film remain, and the FTIR spectra in Supplementary Fig. 8 show that the intensity of SCN− and MA+ ions obviously decrease after IPA washing. Here, etched X-ray photoelectron spectroscopy (XPS) characterization at different depths is employed to verify the diffusion process. For the 0 min-FAPbI3 film after IPA washing, compared to the surface of the film, there is a large shift of 1.1 eV toward a high binding energy for the film etched with 10 nm for both Pb (Supplementary Fig. 9a) and I elements (Supplementary Fig. 9b). This is caused by a large MA+ content near the surface, where the MA+ group does not diffuse into the deep film because the recrystallization process has not begun (Supplementary Fig. 10a). The etched depths of 20, 30 and 40 nm showed no differences, proving that the MA+ group was not reached here. However, for the 10 min-FAPbI3 film after IPA washing, MA+ ions diffused into deeper crystals (Supplementary Fig. 10b). The shift (0.2 eV) is small at 10 nm depth (Supplementary Fig. 11a of Pb and 11b of I); meanwhile, there is also a tiny shift from a depth of 10 nm to 20 nm of 0.1 eV, exhibiting graded MA+ doping. The SCN−/MA+ intensity ratio was obviously larger in 0 min-FAPbI3 than in 10 min-FAPbI3 after IPA washing (Supplementary Fig. 12). These results further prove that there is diffusion of MA+ during the recrystallization process.
Mechanism of the recrystallization process. Combining the analysis of the different effects and behaviors of the MA+ and SCN− groups, the behaviors of MASCN are depicted in Supplementary Fig. 13. Initially, the MA+ and SCN− groups are spin-coated on the surface of the FAPbI3 film (Supplementary Fig. 13a). Then, the MA+ group diffuses into the film, while the SCN− group stays at the surface (Supplementary Fig. 13b). During the diffusion process, there will be a dynamic distribution of MA+ groups. In the MA-rich region, the SCN− group will cause a transition from δ-phase to α-phase; in the MA-poor region, the SCN− group will cause a transition from α-phase to δ-phase. With the continuous diffusion process of the MA+ group, there will be microscale phase transition cycles, which reconstructs the FAPbI3 film (Supplementary Fig. 13c). The XRD patterns prove such a coexisting situation for these two phases when the film is recrystallized with MASCN at high concentrations (Supplementary Fig. 14). The XRD patterns from 10° to 15° show that the signal of δ-phase (11.8°) also gradually emerges with the strength of α-phase (14°). Based on the aforementioned discussion, the whole recrystallization process is depicted in Fig. 3. Initially, the MASCN/IPA solution is spin-coated on the surface of the FAPbI3 film (Fig. 3a). With the evaporation of IPA, the MA+ and SCN− ions will adsorb on the crystal surface, and then MA+ ions will diffuse into the crystal while SCN− will stay at the surface (Fig. 3b). The imbalanced distribution of MA+ will cause a reverse effect of SCN− at the microscale, which results in a repeating PTC process (Fig. 3c). The repeating PTC will result in a slow recrystallization process, where the crystal will be reconstructed. The crystal system is intended to minimize the surface free energy to reach a stable construction. Compared to the cubic structure, the spherical crystal has a smaller surface/volume ratio, which is favorable for minimizing the system energy. Thus, the film will convert from a flat film with cubic crystals to a mesoporous film with continuous spherical crystals (Fig. 3d). The slow recrystallization also helps reduce the vacancies and crystal mismatches, which results in better crystallization.
Photodetectors fabricated based on PTC process. Moisture in the air will stimulate the α to δ transition, which may be employed to control the speed of the recrystallization process. The transmittance spectrum is an ideal signal to monitor the recrystallization process considering the obvious morphology change. Here, Δtran. is defined as the transmittance variation of the film between two time points. For example, Δtran.10 = transmittance10 - transmittance0, which presents the variation of transmittance from 0 min-FAPbI3 to 10 min-FAPbI3. The variation of the MASCN-treated FAPbI3 film under a sealed environment is first observed to block its contact to external stimulation, where the film is dipped into chlorobenzene (CB), which is very stable with perovskite. The variation still occurs, which indicates that this process is spontaneous even though very slow (Supplementary Fig. 15a, b). However, the film variation under an open environment is quick and fierce (Supplementary Fig. 15c, d). Therefore, the recrystallization process can proceed spontaneously, and moisture in air can exactly accelerate the process. The variation in transmittance under different RHs (0 ~ 5%, 20 ~ 25%, 50 ~ 55% and 80 ~ 85%) is shown in Supplementary Fig. 16, and the calculated Δtran. is shown in Fig. 4. From Fig. 4a-d, Δtran. is calculated with the normalized value of transmittance to compare their differences under different RHs. When the film is under a low RH of 0 ~ 5%, the variation is small and slow, similar to the film sealed in CB. However, when RH increases, the variation becomes large and harsh. Under RH values of 80 ~ 85%, a sharp increase in Δtran. occurs within 10 minutes. The absolute value of Δtran. is also depicted in Fig. 4e-h, which corresponds to the film from Fig. 4a-d. With an exposure time of 60 minutes, transmittance increases by less than 1% for RH values of 0 ~ 5% and by approximately 8% for RH values of 80 ~ 85%. In this situation, the variation of the MASCN-treated FAPbI3 film can be concluded: the MASCN-treated FAPbI3 film will recrystallize spontaneously, but the moisture will accelerate this process, where higher RH causes quicker variation (More discussions in Supplementary note 3).
The recrystallization process in air can be employed to improve the optoelectronic properties of the film, and the results in Fig. 4 prove that this process can be adjusted by optimizing the exposure time and RH. A lateral photodetector is fabricated based on the recrystallization process (Supplementary Fig. 17). In this process, IPA washing is employed to terminate the recrystallization process by eliminating the surface SCN− group, which avoids additional annealing treatment (Supplementary Fig. 18). The prepared FAPbI3 film (details in the experimental section) was treated with 10 mg mL− 1 MASCN and then exposed to air (RH ~ 25%) for different time. IPA washing is then employed to stabilize the film, and a pair of Au electrodes is evaporated on the side of the film. The device based on the film with an x-minute exposure time is denoted as Dx min, and the corresponding I-t curve is depicted in Supplementary Fig. 19. The response current increases obviously from D0 min to D10 min and then gradually decreases by a small magnitude (more discussion is provided in Supplementary Note 4). Therefore, the film treated for 10 minutes is chosen as the best parameter, and hereafter, the corresponding film is denoted MASCN-FAPbI3. Responsivity (R) is widely employed to describe the ability of a photodetector to detect incident light, and it is defined as R = \(\frac{{I}_{\text{l}\text{i}\text{g}\text{h}\text{t}}{-I}_{\text{d}\text{a}\text{r}\text{k}}}{PS}\), where Ilight is the light current, Idark is the dark current, S is the active area, and P is the intensity of incident light. The comparison between the device based on Pristine-FAPbI3 and MASCN-FAPbI3 is depicted in Fig. 5a. At an incident light of 700 nm (0.46 µW cm−2) and a low bias of 1 V, the MASCN-FAPbI3−based device shows an R of 1.44 A W−1, which is obviously improved compared to the device based on Pristine-FAPbI3 (0.96 A W− 1). The I-V curve in the dark is shown in Supplementary Fig. 20. Under a 1 V external bias, the device based on Pristine-FAPbI3 shows a dark current of 372 pA, while the device based on MASCN-FAPbI3 just shows a value of 138 pA. Under dark conditions, the current is mainly attributed to the free carriers caused by vacancies and trap states, which will cause a disturbance for detecting incident light signals. Detectivity (D*) is applied to describe the ability to detect incident light signals against the noise of photodetectors, and it is defined as D* = \(\frac{R\sqrt{S}}{\sqrt{2e{I}_{\text{d}\text{a}\text{r}\text{k}}}}\), where e is the elementary charge, as shown in Fig. 5b. The D* of the device based on MASCN-FAPbI3 can reach 6.72 × 1010 Jones, which is two times larger than the device based on Pristine-FAPbI3 of only 3.07 × 1010 Jones. For the optimized device based on MASCN-FAPbI3, the I-t and I-V curves under different wavelengths of incident light are shown in Fig. 5c, d, respectively. The device shows a good periodic response from the ultraviolet to near infrared region. Meanwhile, the device has a fast rise and decay time of 73 and 74 ms, respectively (Supplementary Fig. 21). It is also important to understand the intrinsic modification mechanism of the recrystallization process on FAPbI3 film. A temperature-dependent space charge limited current (SCLC) characterization is carried out to study the variation in the energy band. Temperature-dependent SCLC is an efficient way to extract the density of state (DOS) distribution in semiconductors, including perovskites, and the details are discussed in the Experimental section and Supplementary Note 5. The temperature-dependent I-V curves are shown in Figs. 5e-f for the Pristine-FAPbI3 and MASCN-FAPbI3 -based devices, respectively, and the extracted DOS distribution is depicted in Fig. 5g. The Pristine-FAPbI3 film exhibits a density peak at approximately 0.39 eV below the conductive band, which could be caused by PbI defects according to previous theoretical calculations36. The density peak of MASCN-FAPbI3 is at approximately 0.48 eV, with a trap density less than half that of the Pristine-FAPbI3 film, indicating that the recrystallization process helped heal defects. The stability of the film is also investigated. It is widely known that MA additives accelerate the decomposition of FA-based perovskite films because of the hydrophilicity of the MA group37. Although SCN− is reported to be able to stabilize perovskite crystals, our IPA washing process wiped most of them on the film surface. Interestingly, when the films with different treatments are stored in an environment with a high RH over 90%, the MASCN-treated film still exhibits much better stability than the Pristine-FAPbI3 film (Supplementary Fig. 22). However, the MASCN-treated film without air exposure (0 min-FAPbI3) decomposes first, which is even faster than the Pristine-FAPbI3 film. The difference in stability is attributed to the difference in system energy. When the film is exposed to air for some minutes, its morphology turns from a flat film to rounded large crystals (Fig. 1b). The rounded shape endures the film with a small surface free energy, which is more stable than the flat film (Fig. 3d). The film without air exposure still has a flat morphology; furthermore, the residue MA+ inside the film causes a faster decomposition than the Pristine-FAPbI3 film. The stability of the photodetectors is investigated in Supplementary Fig. 23. The photodetector based on the MASCN-FAPbI3 film can maintain over 80% of its initial response after storage under an RH of 50% for 30 days. As a reference, the photodetector based on Pristine-FAPbI3 falls to no response for only 3 days.
Robust modification capability for various fabricating processes. There exist some tiny process windows in the operations of the solution-based perovskite fabrication method, and the disturbances in these operations would harm the uniformity of the final device. For example, the changes at the time point of dropping antisolvent will affect the final film quality24. Here, our PTC process can correct these disturbances, and some experiments are simulated to verify it. First, the parameters during the one-step antisolvent method are regulated to simulate mechanical errors. The time point of dropping antisolvent is set at 14, 16, 18, 20 and 22 seconds before the end (the resulting films are denoted as (x s)-film, where x is 14, 16, 18, 20 or 22). The XRD patterns of these films are shown in Fig. 6a. The (14 s)-, (16 s)- and (18 s)-films show a similar peak signal, which indicates that they are all inside the process window. However, the photodetectors based on them still perform quite differently (Fig. 6b). Furthermore, the (20 s)-films and (22 s)-film show obvious signals of the δ phase, which makes their photodetectors much poorer than others. Then, each film is treated by the PTC process, and the XRD patterns of the treated films are shown in Fig. 6c. The crystal qualities are all modified and uniform as high-orientation FAPbI3 films. The morphologies of the film before and after treatment are shown in Supplementary Fig. 24. Their initial morphologies are quite different (Supplementary Fig. 24a-e), where different films have different crystal sizes. The modified films become uniform (Supplementary Fig. 24f-j), and as a result, the photodetectors based on the modified films show negligible differences (Fig. 6d). The result shows that our recrystallization process can correct the mechanical mistakes during the fabrication of initial films. Then, the PTC method is further applied to examine its effect on different fabrication methods. In addition to the one-step method, the two-step method is also widely employed for fabricating FA-based perovskite films. Here, FAPbI3 films are prepared by a two-step fabrication method (details in the experimental section). The initial film shows poor crystallization with many signals of the δ phase and PbI2, while the PTC process also modified it successfully (Fig. 6e). The morphology variation is shown in Supplementary Fig. 25, and the corresponding photodetector shows a huge improvement after modification (Fig. 6f). Finally, the PTC method is employed to save a decomposed film (Fig. 6g). Normally, when the perovskite film decomposes, it is difficult to heal the decomposition, and the materials become rubbish. For example, when α-FAPbI3 decomposes into δ-FAPbI3, heating at 150 ℃ is not able to transform it back to a useful black phase. However, the PTC method successfully transforms a decomposed film into a high-quality FAPbI3 film (Fig. 6g and Supplementary Fig. 26), and its performance is competitive with that of a normal PTC-treated device (Fig. 6h). The abovementioned results show that our PTC process can produce uniform photodetectors while ignoring the initial building blocks. A performance statistic of the Pristine- and MASCN-FAPbI3−based device is shown in Fig. 6i (for R) and Supplementary Fig. 27 (for Idark and D*) with 100 devices each. Here, the standard deviation (SD) is employed to describe the dispersion degree of the devices fabricated by different methods. For the R in Fig. 6i, the SD of the traditional method is 0.5035, and our PTC method reduces its value to 0.3969 with a decreasing percentage of 21%, which promises an attractive potential for further application.