Revealing the detailed path of sequential deposition for metal halide perovskite formation

See allHide authors and affiliations

Science Advances  02 Feb 2018:
Vol. 4, no. 2, e1701402
DOI: 10.1126/sciadv.1701402


Sequential deposition has been extensively used for the fabrication of perovskite solar cells. Nevertheless, fundamental aspects of the kinetics of methylammonium lead iodide perovskite formation remain obscure. We scrutinize the individual stages of the reaction and investigate the crystallization of the lead iodide film, which occurs before the intercalation of methylammonium iodide commences. Our study identifies the presence of mixed crystalline aggregates composed of perovskite and lead iodide during intercalation and structural reorganization. Furthermore, Ostwald ripening occurs in the film for reaction times beyond the point of conversion to perovskite. Using cross-sectional confocal laser scanning microscopy for the first time, we reveal that lead iodide in the over-layer and at the bottom of the mesoporous layer converts first. We identify unreacted lead iodide trapped in the mesoporous layer for samples of complete conversion. We acquire kinetic data by varying different parameters and find that the Avrami models best represent them. The model facilitates the rapid estimation of the reaction time for complete conversion for a variety of reaction conditions, thereby ascertaining a major factor previously determined by extensive experimentation. This comprehensive picture of the sequential deposition is essential for control over the perovskite film quality, which determines solar cell efficiency. Our results provide key insights to realize high-quality perovskite films for optoelectronic applications.


In the search for superior materials for use as light harvesters in mesoscopic heterojunction solar cells, organic-inorganic lead halide perovskites have emerged as strong contenders (1, 2). Numerous researchers have been investigating solar cells based on methylammonium lead iodide (CH3NH3PbI3), which has remarkable optical and electronic properties (2). Various methods for the preparation of the perovskite films have been proposed, including single-step deposition (3), sequential deposition (1), and solvent engineering (4), among others.

Sequential deposition is one of the main methods used for depositing perovskite films for various applications. It was optimized by Burschka et al. (1) for the fabrication of solar cells. In this method, the formation of a CH3NH3PbI3 film is achieved through two steps. First, a lead iodide (PbI2) solution is infiltrated into a mesoporous TiO2 film, and then the film is dried. Second, the PbI2 film is dipped in a solution of methylammonium iodide (MAI) in 2-propanol for a few seconds, resulting in the formation of the perovskite.

Research on this method has been focused on improving photovoltaic performance by controlling the perovskite morphology through optimization of the processing conditions (5, 6) or via the use of additives (7, 8). Some work focused on particular stages of the perovskite formation. For instance, Wu et al. (9) and Tu et al. (10) have looked into controlling the morphology and crystallinity of the PbI2 film to achieve better solar cell efficiency. Others have controlled the concentration of reactants, namely, PbI2 (11) and MAI (12), to optimize the perovskite cell performance. Despite the effort invested in this method, given the complexity of the reaction, it is difficult to control selected aspects of the deposition process and achieve reproducible results, while details of the remaining stages of the reaction are poorly understood. Thus, it is critical to scrutinize the various stages of perovskite formation and obtain a comprehensive picture of the path of sequential deposition.

In particular, fundamental aspects of the kinetics of CH3NH3PbI3 formation, such as the evolution of the film with the reaction coordinate (that is, with time), remain obscure. Characterization tools and techniques with the capacity to identify different chemical components need to be used to study the surface of the film during perovskite formation. In addition, there is little insight available into the directionality of the reaction in the cross section of the film during or after the dipping. Furthermore, the role of unreacted PbI2 in the film has been well studied because it affects solar cell performance (13, 14). However, the residual PbI2, likely to be trapped at some regions in the film, has not been imaged so far.

Knowledge about these aspects of perovskite film formation would allow better control over perovskite film quality, thereby enhancing the solar cell performance (15). A convenient way of determining the optimal reaction parameters that are currently determined through rigorous experimentation would aid in achieving reproducible film quality. Thus, systematic kinetic studies and a robust model that represents the kinetic data for a variety of reaction conditions are necessary.

Here, we studied the individual stages of the reaction using various spectroscopic and analytical techniques, such as x-ray diffraction (XRD), scanning electron microscopy (SEM), SEM-cathodoluminescence (CL), and confocal laser scanning fluorescence microscopy (CLSM). We demonstrate that the stages of the reaction are the unexpected crystallization of partially amorphous PbI2 followed by MAI intercalation and structural reorganization, ending with Ostwald ripening. We demonstrate the use of SEM-CL to study perovskite films, specifically looking at mixed PbI2-perovskite crystals. We show the innovative application of CLSM to look at cross sections of perovskite films at different stages of the reaction and obtain new insights into the various layers that compose the film. Last, we examine kinetic data obtained under different reaction conditions and identify that Avrami models best represent the reaction kinetics. This model allows prediction of the dipping time for a particular degree of conversion, including the point of complete conversion, as required for solar cell fabrication.


Stages of the reaction

Crystallization of PbI2. To begin with, we use XRD to examine PbI2 films dipped in MAI solution (6 mg ml−1) for different times between 2 and 400 s. PbI2 films were prepared by spin coating from a 1.3 M solution in N,N′-dimethylformamide at 70°C onto a fluorine-doped tin oxide (FTO)–coated glass substrate covered with a mesoporous Al2O3 layer, followed by heating at the same temperature for 10 min. XRD spectra of these films are presented in Fig. 1A. We find that the unreacted PbI2 film is of the 2H polytype and the (001) reflection at 12.66° is the most prominent (1, 16). The perovskite is detected at 14.01° and 14.07°, corresponding to the (002) and the (110) reflections of the tetragonal phase of the perovskite, respectively (17, 18).

Fig. 1 XRD analysis of PbI2 crystallization and perovskite formation.

(A) XRD spectra of PbI2 samples dipped in MAI for 2, 4, 8, 25, 50, 200, and 400 s, showing the (001) reflection of the 2H polytype of PbI2 and the (002) and (110) reflections of the tetragonal perovskite. (B) FWHM of the (001) reflection of PbI2 and of the (110) reflection of the perovskite versus time for the spectra shown in (A). a.u., arbitrary units.

Remarkably, the perovskite formation is delayed, with the peaks detected only after 25 s. Instead, the (001) reflection of PbI2 undergoes changes. To investigate the observations further, we analyzed the XRD data in detail. The full width at half maximum (FWHM) from the Voigt function fits of the (001) reflection of PbI2 and the (110) reflection of the perovskite is shown in Fig. 1B. We observe a narrowing of the (001) reflection of PbI2 up to 25 s, which indicates that the reaction commences with an increase in the crystallinity of PbI2. After 25 s, we observe a slight broadening in the PbI2 reflection coinciding with the formation of the perovskite, whose reflections are detected from 25 s onward, indicating that the broadening of the PbI2 reflection is due to the consumption of the crystalline PbI2 as it forms perovskite. Thus, upon dipping into MAI, the crystallinity of PbI2 increases before perovskite formation (depicted in Fig. 2A). This conclusion is consistent with suggestions reported by Harms et al. (19) and with our previous report (20) where photoluminescence (PL) imaging indicated that PbI2 crystallization is the first step in the sequential deposition, occurring before MAI intercalation commences.

Fig. 2 Schematic depicting the stages of the reaction in sequential deposition.

Dashed arrows indicate mass transfer. (A) Nucleation and growth of PbI2. (B) Intercalation of MAI and structural reorganization to form CH3NH3PbI3 perovskite. (C) Ostwald ripening where perovskite from the mesoporous layer is transported to the capping layer. (D) Further Ostwald ripening at longer dipping times where perovskite from the small crystals in the capping layer is transported to larger ones. (E) Gibbs free energy shown as a function of the reaction coordinate.

We considered the possibility that the observed changes in the XRD patterns attributed to crystallization could be due to solvent effects such as a lower solubility of PbI2 in 2-propanol that increases crystallinity upon dipping. However, this possibility has been eliminated through further XRD analysis. The analysis also ruled out the possibility that observations were due to reorientation of the PbI2 crystals in the film upon dipping (details presented in the Supplementary Materials).

Crystallization of the partly amorphous PbI2 film, rather than the direct formation of the crystalline perovskite, is explained by Ostwald’s step rule. It states that the final stable phase is formed through consecutive steps with increasing thermodynamic stability (21). In our case, the PbI2 is the phase that nucleates first, as depicted in Fig. 2. The nucleation mechanism in the PbI2 film was investigated in detail in our previous report (20).

Intercalation and structural reorganization. At this point, we look at the progress of the reaction using SEM images of PbI2 films dipped in a MAI solution for different periods. The unreacted PbI2 film is planar as seen in Fig. 3A. The growth of nuclei in the hexagonal platelet structure characteristic of the 2H polytype of PbI2 (19, 22, 23) is visible in Fig. 3B, which shows a sample dipped in MAI for 2 s. After the partial crystallization of PbI2, the reaction of the crystalline PbI2 proceeds via the intercalation of MAI between the layered PbI2 and structural reorganization to form the perovskite, as mentioned in the literature (1, 19). Ahmad et al. (24) investigated the phenomenon of intercalation and showed that controlling reactant concentrations is crucial for optimal intercalation. The report by Liu et al. (25) indicates that the intercalation of MAI starts at grain boundaries and defects in the crystalline PbI2. This onset and the subsequent structural reorganization to form the perovskite are shown schematically in Fig. 2B.

Fig. 3 Perovskite formation with time studied using SEM.

(A) Unreacted PbI2 film. (B) Sample dipped for 2 s showing the growth of nuclei as hexagonal platelets of PbI2. (C) Sample dipped for 4 s showing hexagonal platelets of PbI2 on the crystal surfaces. (D) Sample dipped for 8 s showing the textured crystals. Scale bars, 0.5 μm (A to C) and 1 μm (D).

In Fig. 3C, we observe that platelets are visible after dipping for 4 s, indicating that intercalation has not commenced. The texturing seen in Fig. 3D, on a sample dipped for 8 s, is investigated further subsequently. We see that areas in the image, which are localized around the crystals, are devoid of PbI2, and the mesoporous scaffold is exposed (indicated with arrows). This points to the transport of PbI2 during its crystallization through the MAI solution or along the sample surface, in addition to in situ crystallization. This finding is in agreement with the nucleation and growth mechanism proposed in our previous work (20).

Looking at a sample dipped for 10 s, distinct clusters are seen on the surface of the crystal in Fig. 4A. SEM-CL measurements were made on the same sample to identify the composition of these clusters that evolved from the textured crystals seen in Fig. 3D. CL spectra obtained at two different positions on the sample are shown in Fig. 4B. The emission between 480 and 550 nm, attributed to PbI2, is assigned green, and the emission between 720 and 810 nm, attributed to perovskite, is assigned red. In Fig. 4C, we see that both red and green emissions originate from the textured crystals, showing that these are mixed crystalline aggregates composed of perovskite and PbI2 formed as a result of partial MAI intercalation and structural reorganization. These mixed crystalline aggregates have, so far, never been reported. Green emission, indicated with the arrows in Fig. 4D, identifies the clusters on the crystal as PbI2, whereas the red emission from the underlying crystal shows that it is perovskite (see the Supplementary Materials for a sample of nearly complete conversion and further discussion on the CL signal strength).

Fig. 4 CL study on films of intermediate conversion.

(A) SEM image of sample dipped for 10 s with clusters visible on the crystals. Scale bar, 0.5 μm. (B) CL spectra taken at two different points on the same sample. PbI2 emission between 480 and 550 nm is assigned green, and perovskite emission between 720 and 810 nm is assigned red. (C) Pseudocolor CL image overlay on a SEM image of the same sample showing mixed crystalline aggregates composed of perovskite and PbI2. Scale bar, 0.5 μm. (D) Pseudocolor CL image overlay on a SEM image. The crystal is perovskite, whereas the clusters (indicated with arrows) on it are identified as PbI2. Scale bar, 0.5 μm.

Ostwald ripening. It is interesting to look into samples dipped in a MAI solution for extended periods of time, long after most of the PbI2 has been converted to perovskite. The presence of residual PbI2 is investigated later. Cuboids of perovskite lying over the mesoporous layer make up the capping layer (26), and SEM images in fig. S3 show the capping layer of successive samples of long dipping times of 400 and 800 s. Comparing these images, we observe an increase in the surface coverage of the capping layer, although dipping times are well beyond the point of complete PbI2 conversion. These findings support the hypothesis of Ostwald ripening taking place in the film, as previously suggested in the literature (1, 13). This thermodynamically driven phenomenon involves the transport of matter from smaller crystals to larger ones. During the initial stages of Ostwald ripening, depicted in Fig. 2C, perovskite formed in the mesoporous layer is transported to add to minute crystals on the surface, thereby increasing the surface coverage.

Upon dipping even longer (that is, for thousands of seconds), small crystals in the capping layer dissolve to add to the growth of the larger ones as shown in Fig. 2D. The disappearance of the smaller crystals would decrease the surface coverage. Cao et al. (13) investigated Ostwald ripening for very long dipping times and showed that the large crystals with low surface coverage were obtained as a result of the extensive ripening phenomenon, and this decreased the solar cell efficiency. Thus, dipping beyond the point of complete conversion of PbI2 to perovskite, which results in Ostwald ripening, is unfavorable. Further discussion on this aspect will be taken up subsequently.

Confocal laser scanning fluorescence microscopy

At this point, we delve further into the directionality of the reaction and limitations that result in the presence of residual PbI2 in the film. To spatially resolve the conversion of PbI2 within mesoporous layers in addition to the capping layer, we used CLSM innovatively to obtain cross-sectional images. Cross-sectional CLSM images allow us to study the different layers in vertical cross sections of solar cells, not just the perovskite film surface. Figure 5A shows the different parts of a sample that can be mapped using both standard surface imaging and our cross-sectional imaging technique together to obtain a complete description of the sample. This cross-sectional CLSM imaging for perovskite films will be demonstrated here for the first time.

Fig. 5 CLSM of samples with a 2.5-μm mesoporous Al2O3 layer.

Emission between 500 and 550 nm attributed to PbI2 was assigned green, and emission between 700 and 800 nm attributed to perovskite was assigned red. The color saturation scales with the emission intensity. (A) Schematic showing planes parallel to the sample surface and the cross section, showing the different parts of the sample being imaged. (B) Composite pseudocolor image of the surface of a sample of intermediate conversion dipped for 60 s. Scale bar, 5 μm. (C) Composite pseudocolor cross-sectional image of the same sample showing the perovskite capping layer over the unconverted PbI2 in the mesoporous layer. Scale bar, 1 μm. (D) Normalized emission intensity showing horizontally averaged values of PbI2 emission and perovskite emission across the depth of the sample in (C). (E) Composite pseudocolor cross-sectional image of a sample of nearly complete conversion showing trapped, unconverted PbI2 in the mesoporous layer. Scale bar, 2 μm.

Samples with a larger Al2O3 mesoporous layer, approximately 2.5 μm thick, were imaged because standard samples could not be imaged due to the resolution limit of the technique. Nevertheless, insights obtained here, such as any directionality of the conversion reaction, would be applicable to standard samples in a similar manner. The difference in the reaction for these two samples would possibly be that more time is needed to fully convert a thicker PbI2 layer as demonstrated below. Moreover, our results are particularly relevant for preparing solar cells in the architecture proposed by Mei et al. (27). Their cells have a double layer of mesoporous TiO2 and ZrO2, amounting to about 3 μm, as a scaffold infiltrated with perovskite and covered by a porous carbon film instead of a hole conductor.

Images of our thick sample, dipped for 60 s, resulting in intermediate conversion, are shown in Fig. 5 (B and C). Emission collected between 500 and 550 nm is attributed to PbI2 and represented as green, whereas the emission between 700 and 800 nm is attributed to perovskite and assigned red (20). In the surface view shown in Fig. 5B, unconverted PbI2, indicated with arrows, is visible inside the mesoporous layer, which is exposed between the perovskite crystals of the capping layer. PbI2 underneath the perovskite capping layer is not detectable due to the strong absorption of the perovskite. However, it is seen in the cross-sectional image in Fig. 5C. This is feasible because the plane chosen for cross-sectional imaging is perpendicular to that used for obtaining surface images (Fig. 5A), and the perovskite crystals of the capping layer are not in the way to hamper detection of the PbI2 emission. Thus, our novel cross-sectional CLSM imaging technique has enabled the mapping of unreacted PbI2 in the mesoporous layer for the first time (Fig. 5C). Obtaining this information was not feasible with standard surface imaging, where layers lying beneath another are often inaccessible.

The image is analyzed by horizontally averaging the PbI2 and perovskite emission, followed by normalization of the two curves plotted as a function of depth in Fig. 5D. The consumption of PbI2 in the over-layer results in the formation of the highly emissive perovskite capping layer. Strikingly, the PbI2 at the bottom of the mesoporous scaffold converts to perovskite, before the PbI2 located directly below the perovskite capping layer. These observations are consistent with the proposal in the report by Harms et al. (19) that the perovskite forms a compact layer, denying the MAI solution access to the underlying PbI2 for conversion.

As mentioned earlier, several reports (13, 28) have demonstrated that residual PbI2 is beneficial for solar cell performance. However, residual PbI2 has, so far, never been imaged or localized. Figure 5E shows a sample of long dipping time, expected to have nearly complete conversion. We see that the unconverted PbI2 is visible in green, trapped in the mesoporous layer. This finding addresses long-standing questions regarding the location of the unreacted PbI2 in the film (13). In the cross-sectional images shown in Fig. 5 (C and E), low emission from the PbI2 or perovskite in the mesoporous layer may be due to the small size of the crystals, because they are restricted by the pores of the mesoporous layer.

In situ isothermal kinetic monitoring and solid-state kinetic model

To quantify the kinetics of perovskite formation, we conducted in situ isothermal experiments at reaction temperatures of 10°, 15°, and 25°C. Standard PbI2 samples, made using a 1.3 M solution as mentioned earlier, were used. The absorbance of the perovskite at 700 nm was monitored during conversion (see Materials and Methods). The ratio of the absorbance at time t and that after complete conversion is defined as the conversion fraction [α(t); 0 ≤ α(t) ≤ 1], which represents the fraction of perovskite formed from the PbI2 at time t. The remnant PbI2 that cannot be converted, despite further dipping as discussed earlier, is neglected in the present analysis. Smoothened data, represented in terms of the conversion fraction (α) versus time, are presented in Fig. 6A.

Fig. 6 In situ kinetic data of perovskite formation.

(A) Isothermal kinetic data at 10°, 15°, and 25°C shown as the conversion fraction (α). (B) Isothermal reaction rate (dα/dt) shown as a function of conversion fraction (α) for the data in (A). (C) Kinetic data for samples of low and high PbI2 loading dipped under light and in the dark, shown as the conversion fraction (α).

The kinetic data were fitted to a generalized equation proposed by Khawam and Flanagan (29), representing various solid-state kinetic models. The generalized model equation was taken up in its integral form to determine the appropriate model for the perovskite formation. Data fitting by means of nonlinear regression showed that the Avrami models were the most suitable to describe the experimental data. Their general form under isothermal conditions isEmbedded Image(1)where n is the Avrami exponent, k is the effective reaction rate constant, α is the conversion fraction, and t is time.

We assume that the model parameters are constant throughout the reaction. The values of n for these reactions were found to be very close to 1. The value of k is expected to increase with temperature, and these values were found to be 0.8, 2.2, and 3.5 min−1 for increasing reaction temperatures of 10°, 15°, and 25°C, respectively. The change in the isothermal reaction rate (dα/dt) with the conversion fraction (α) is represented in Fig. 6B. It is seen that the reaction is accelerating in the initial stages and decelerating in the final stages, with a maxima at an intermediate conversion. This is a characteristic feature of sigmoidal models, such as the Avrami models, thereby verifying the aptness of the choice (30). The advantage of identifying the appropriate model is that upon determining the model parameters once for a temperature of choice, the reaction time can be rapidly determined using the model equation for any degree of conversion. The implications of determining the reaction time are discussed further below. The applicability of this model for a variety of reaction conditions will be taken up next.

In situ kinetic monitoring in the dark and under illumination

In our previous work (20), we demonstrated the effect of light on the perovskite formation and the resulting film morphology. To quantify the impact of illumination on the perovskite formation, we conducted in situ kinetic experiments at room temperature, in the dark, and under illumination (see Materials and Methods).

Bi et al. (11) demonstrated that increasing the concentration of the PbI2 solution would lead to better pore filling of the mesoporous scaffold, resulting in higher PbI2 loading. Using PbI2 solutions of concentrations greater than 1 M gives such high PbI2 loading that it results in the formation of an over-layer of PbI2. This over-layer of PbI2 forms a capping layer of perovskite upon dipping in a MAI solution (11). Thus, differences between perovskite formation in the mesoporous and the capping layer can be studied using samples of different PbI2 loading. Moreover, we expect thicker PbI2 layers to take longer to convert to perovskite, which indicates that the kinetic data collected for the two cases are likely to be quite different. Thus, we also include samples of two different levels of PbI2 loading in this experiment to ensure that these diverse conditions are investigated in the study. We are interested in verifying whether their kinetics can be represented well by the model identified earlier.

Samples of low PbI2 loading were prepared using a PbI2 solution with a very low concentration (0.25 M, details in Materials and Methods) to ensure that the PbI2 is present only in the mesoporous layer and in small quantities. After dipping such a sample in a MAI solution, the CH3NH3PbI3 formed remains restricted to the mesoporous layer. On the other hand, samples with high PbI2 loading, prepared using a high-concentration PbI2 solution (1.3 M, details in Materials and Methods), have an additional over-layer of PbI2. This results in the presence of perovskite within the mesoporous scaffold and as a capping layer upon dipping in a MAI solution. These samples of high PbI2 loading are the standard samples used throughout this report. The smoothened data, represented as the conversion fraction, are shown in Fig. 6C.

For the samples of high PbI2 loading, the time to achieve 99% conversion was found to be around 2.1 under light bias and 2.9 min in the dark. The analogous values for the samples of low PbI2 loading are 0.3 and 1.2 min, respectively. As expected, we found that under identical illumination conditions, samples of high PbI2 loading take longer to convert to perovskite compared to those of low PbI2 loading. Remarkably, the conversion is reached in less time upon biasing the samples with light, compared to the reaction in the dark, for both samples. Thus, light accelerates the perovskite formation within the mesoporous scaffold, as well as in the capping layer. Upon fitting the data for samples of high PbI2 loading with the Avrami models, the value of n was obtained as 1.4 and 1.1 for the dark and light bias cases, respectively. A higher value of k (1.8 min−1) was obtained for the reaction under light bias, as compared to the value of 1.0 min−1 in the dark. This is consistent with the observations in our earlier report (20), where the reaction under light was rapid compared to that in the dark.

We identified that the Avrami models are the most suitable to represent the in situ kinetic data for the reaction. They appropriately represent the reaction for various PbI2 loading levels and light intensities, in addition to our previous demonstration for different reaction temperatures, indicative of robustness.

The model facilitates the estimation of the reaction time for a given conversion at different reaction conditions, including the point of complete conversion. The loss in solar cell performance due to Ostwald ripening, as discussed earlier, can be eliminated by identifying the point of complete conversion for a given set of reaction conditions and choosing that as the dipping time. Thus, the model presents a convenient and rapid option for ascertaining the reaction time for complete conversion, a previously uncertain parameter in sequential deposition. This helps address a major source of irreproducibility in perovskite film formation.

Conclusion and broader context

Although the sequential deposition method is widely used, studies have focused on optimizing the photovoltaic performance, rather than examining the pathway of the reaction and its kinetics. Our present investigation allowed unraveling the salient features of this intriguing and widely applied conversion process. Contacting the PbI2 film deposited on a mesoscopic oxide support with a MAI solution first induces the crystallization of the PbI2 film. This is followed by the intercalation of MAI into the PbI2 crystal lattice and structural reorganization, forming mixed crystalline aggregates of perovskite and PbI2 as intermediates.

Cross-sectional CLSM imaging revealed the directionality of the conversion of the PbI2 film, showing that the PbI2 in the mesoporous scaffold, underneath the capping layer, is the last to convert and that unreacted PbI2 is trapped there even after long reaction times. Our identification of trapped, unreacted PbI2 in the film addresses open questions raised in the past (13, 14) regarding the location of residual PbI2 in perovskite solar cells. These insights obtained from our pioneering cross-sectional imaging technique were otherwise inaccessible through the traditional surface imaging technique. On a broader note, our innovative demonstration of this technique’s potential opens the door for understanding the properties of perovskites in cross sections of solar cells, not just the perovskite surface, as currently presented in the literature (31).

Dipping beyond the point of complete conversion results in Ostwald ripening, which is detrimental to solar cell performance (13). Thus, identifying the dipping time corresponding to the point of complete conversion is warranted. We found that the Avrami models are best suited to represent the kinetic data, and we verified its robustness under different reaction conditions of temperature, film thickness, and illumination. The model allows rapid estimation of the dipping time for any conversion, eliminating the need for extensive experiments. We thereby address critical problems in batch-to-batch reproducibility in perovskite film fabrication.

This comprehensive picture of the sequential deposition is essential for control over perovskite film quality, which is critical because it determines device performance (15). Our findings provide the foundation for the improvement of perovskite film fabrication through sequential deposition for various applications.



Materials used in the study were purchased from Sigma-Aldrich or Acros Organics. MAI was prepared as reported in the report by Im et al. (3). The Al2O3 paste of particles (diameter, 23 nm) was prepared in-house. Films for the in situ kinetic experiments were made using an FTO-coated glass substrate (TEC 15, Pilkington), whereas Nippon sheet glass was used for the remaining experiments.

Sample preparation

Films used for the experiments, with the exception of CLSM, had a mesoporous layer of 300-nm thickness made using the Al2O3 nanoparticles. The mesoporous layer was deposited and processed as described by Burschka et al. (1). A thicker mesoporous Al2O3 layer of approximately 2.5 μm was deposited for samples for CLSM. For standard samples, a 1.3 M solution of PbI2 in N,N′-dimethylformamide, kept at 70°C, was deposited by spin coating at 6500 rpm for 20 s. The sample was then heated at 70°C for 10 min. A low concentration of MAI solution in 2-propanol (6 mg ml−1) was used in all the kinetic experiments presented, rather than 8 or 10 mg ml−1, as used for solar cell fabrication (1), to keep the reaction rate sufficiently low so that kinetic features could be monitored and analyzed effectively. Samples of specific reaction times were prepared by dipping PbI2 films in MAI for the required period, followed by washing in 2-propanol to halt the reaction. They were then heated at 70°C for 10 min.

XRD measurements

The XRD measurements were performed on an X’Pert MPD PRO from PANalytical operated in the Bragg-Brentano geometry. It has a ceramic tube (Cu anode, λ = 1.54060 Å), a secondary graphite (002) monochromator, and a real-time multiple strip X’Celerator detector. The automatic divergence slit and beam mask were adjusted to the dimensions of the films. A step size of 0.002° and an acquisition time of up to 5 min per degree were chosen for the scan in Fig. 1A, whereas a step size of 0.008° and an acquisition time of up to 1.5 min per degree were chosen for the scan in fig. S1.


The SEM-CL measurements were conducted at room temperature using an Attolight Rosa 4634 CL microscope, with a beam probe of 5 nm, an accelerating voltage of 2 kV, and a beam current of 20 nA. The CL microscope tightly integrates a high–numerical aperture (0.72) achromatic reflective lens within the objective lens of a field emission gun SEM (FEG-SEM). The focal plane of the light lens matches the FEG-SEM optimum working distance. The CL collection efficiency is constant over a 300-μm field of view so that the CL emission can be compared quantitatively between two distinct points. The CL emission was spectrally resolved with a Czerny-Turner spectrometer (focal length, 320 mm; grating, 150 grooves/mm) and detected by an ultraviolet–visible charge-coupled device camera.

CLSM and image processing

CLSM images were captured using the Leica Application Suite X software on a confocal laser scanning microscope (Leica TCS SP8) using an HC PL APO oil objective (63×/1.40). A pulsed diode laser (440 nm) was used for excitation. The excitation power and the gain were chosen to optimize the dynamic range of the detector (photomultiplier tube, Leica HyD, or Leica HyD SMD).

Single-plane 512 × 512 images were acquired at 25°C from a unidirectional scan of 400-Hz speed. Figure 5 (B and C) was imaged using a pinhole size of 1 airy unit. A pinhole size of 0.628 airy unit was used to improve signal strength for capturing Fig. 5E. Figure 5B was acquired at a resolution of 72 nm in xy, whereas it was 46 nm for Fig. 5 (C and E). Image bit depth was 8. Fiji was used for image processing: The two components mapped to different channels were pseudocolored. Linear brightness and contrast adjustments were applied to the entire image for each channel. For display, the image in Fig. 5B was cropped to 36.9 μm × 18.45 μm, whereas the images in Fig. 5 (C and E) were cropped to show an area of 8 μm × 8 μm and 14 μm × 14 μm, respectively.


The isothermal kinetic data were recorded through absorbance spectroscopy in transmission mode in a Varian Cary 5 spectrophotometer. The amount of perovskite formed was quantified by monitoring the absorbance of the perovskite at a 700-nm wavelength.

The kinetic data of the perovskite formation, in the dark and under light bias, were recorded through absorbance spectroscopy in transmission mode (also at 700 nm) using an integrating sphere of 110-mm diameter in the same spectrophotometer. Dark conditions are as dark as achievable with attenuated probe light, whereas light biasing was achieved using a white light-emitting diode that supplemented the maximum intensity of the probe light.

A 1.3 M solution of PbI2 in N,N′-dimethylformamide was used to prepare samples of high PbI2 loading. These samples were similar to the standard samples used throughout the report. Thus, these samples were used to conduct the isothermal kinetic study as well. A 0.25 M PbI2 in N,N′-dimethylformamide solution was used to prepare samples of low PbI2 loading for some of the in situ kinetic experiments.

For both the experiments, mesoporous films infiltrated with PbI2 were placed in a cuvette. Using pipettes, MAI solution was added to start the reaction, and the reaction was stopped by rapidly replacing the MAI solution in the cuvette with 2-propanol.


Supplementary material for this article is available at

Supplementary Text

fig. S1. XRD spectra of unreacted PbI2 sample and sample dipped in MAI solution for 4 s in the dark.

fig. S2. CL study on a sample dipped for 60 s for nearly complete conversion.

fig. S3. SEM images of perovskite films with increasing dipping times showing Ostwald ripening.

table S1. FWHM of the (001) reflection of the 2H polytype of PbI2 and the corresponding height of the platelet from the Scherrer equation.

Reference (32)

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We acknowledge D. Gachet (Attolight AG) for collecting the CL images. We thank N. Pellet for experimental help and assistance with SEM. We acknowledge K. Schenk, L. Bozzo [BioImaging and Optics Platform (BIOP)–École polytechnique fédérale de Lausanne (EPFL)], and R. Guiet (BIOP-EPFL) for help with XRD analysis, CLSM, and image processing, respectively. We thank W. Tress and R. Humphry-Baker for discussions. Funding: Financial support was provided by the Swiss NSF through grant no. 200021-157135/1. Author contributions: A.U. conceptualized the study, conducted the experiments, analyzed the data, and prepared the manuscript. M.G. directed the study and commented on the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article