Morphology and surface analyses for CH3NH3PbI3 perovskite thin films treated with versatile solvent–antisolvent vapors

Organometal halide perovskite (CH3NH3PbI3) semiconductors have been promising candidates as a photoactive layer for photovoltaics. Especially for high performance devices, the crystal structure and morphology of this perovskite layer should be optimized. In this experiment, by employing solvent–antisolvent vapor techniques during a modified sequential deposition of PbI2–CH3NH3I layers, the morphology engineering was carried out as a function of antisolvent species such as: chloroform, chlorobenzene, dichlorobenzene, toluene, and diethyl ether. Then, the optical, morphological, structural, and surface properties were characterized. When dimethyl sulfoxide (DMSO, solvent) and diethyl ether (antisolvent) vapors were employed, the CH3NH3PbI3 layer exhibited relatively desirable crystal structures and morphologies, resulting in an optical bandgap (Eg) of 1.61 eV, crystallite size (t) of 89.5 nm, and high photoluminescence (PL) intensity. Finally, the stability of perovskite films toward water was found to be dependent on the morphologies with defects such as grain boundaries, which was evaluated through contact angle measurement.


Introduction
Organometal halide perovskite solar cells (PSCs) have received tremendous interest for a next-generation photovoltaic (PV) technology. [1][2][3][4][5] Perovskite can be designated by a common formula known as ABX 3 where 'A' is a large organic cation [CH 3 NH 3 or HC (NH 2 ) 2 ], 'B' is a metal cation (Pb, Sn), and 'X' is a halide (Cl, Br, I). The perovskite material is a light-harvesting component of the PSCs, and is able to offer many desirable characteristics such as low-temperature solution processability, 6,7 high absorption coefficient, 8 long carrier diffusion length, 9 high charge carrier mobility, 10 and adjustable direct bandgap with suitable alternative metals, halogens, and organic cations. [11][12][13][14] These characteristics can be further modied by using additives, [15][16][17] compositional adjustments, 18,19 and solvent-antisolvent extraction approaches. 20 Hence, the PSCs have been a promising candidate for commercialization in the current PV industries.
In general, the PV performance of PSCs relies on the morphologies of the perovskite thin lm because the structural characteristics of a photoactive layer decide PV performances of devices. [21][22][23][24][25][26][27][28][29][30][31][32] For example, if there is a trap site (e.g., surface defect and grain boundary) in a perovskite layer, it acts as carrier recombination sites, 33 resulting in a reduced performance of devices. Thus, the morphology and crystallinity of the perovskite thin lm should be very important for fabricating high-efficiency PV devices. 34 To date, numerous approaches have been developed to obtain a high quality and defect-minimized perovskite thin lm. 35,36 For example, thermal annealing of a perovskite lm at 85-120 C has been employed. 37 Furthermore, low-temperature antisolvent assisted fabrication of devices are one of the useful techniques for obtaining a lm with desired morphologies. 18,19 Importantly, it is notable that the additive and antisolvent strategies are both signicantly promising in improving the performance of PSCs. [38][39][40][41][42][43][44][45][46][47][48][49] Moreover, the dipping time, 50 precursor's type and concentration, 51 spin-speed, 52 solvent types, 53,54 and temperature are important processing factors for optimizing a perovskite layer. In the sequential deposition of the PbI 2 and CH 3 NH 3 I (MAI) layers, the MAI's intercalation into the PbI 2 layer is critically important to obtain a high quality perovskite without any unreacted precursor material. If there is an incomplete conversion of PbI 2 -MAI into a perovskite, it may be a problem for device performances. 55 However, for improving the stability of PSCs, there are researchers who used a PbI 2 interfacial nanolayer in their device conguration. [56][57][58][59][60] In this work, we employed a modied sequential deposition method for fabricating organometal halide perovskite thin lms. For this purpose, the solvent-antisolvent vapor techniques were adopted as a method of morphology engineering. The ve anti-solvents such as chloroform (CF), chlorobenzene (CB), 1,2-dichlorobenzene (DCB), toluene (Tol), and diethyl ether (Et 2 O) were tested, which may act as an extractor of a solvent, dimethyl sulfoxide (DMSO). Then the properties of CH 3 NH 3 PbI 3 thin lms were investigated as a function of antisolvent species, which may include UV-vis light absorption, micro-/nano-structural morphologies, crystal structures, photoluminescence (PL) emission, and surface analysis through the water contact-angle measurements. In this study, it was observed that when a perovskite layer is well crystallized, the surface polarity of perovskite lms remains a longer time, i.e., an enhanced stability toward water or its vapor.

Materials and methods
In all synthesis methods, analytical grade high purity reagents were used. All solvents and antisolvents were purchased from Fine Chemicals Ltd. Indium tin oxide/uorine-doped tin oxide (ITO/FTO) coated glass substrates were purchased from TECHINSTRO Chemicals Ltd. PbI 2 precursors were purchased from Tokyo chemical industries (TCI) and synthesized using a hydrothermal method. 27 CH 3 NH 3 I (MAI) was synthesized by reacting methylamine (aqueous, 40 wt%) and hydroiodic acid (aqueous, 57 wt%) in an ice bath for 2 h with stirring. Then the solvent was evaporated using a rotary evaporator and the precipitate was collected and washed using Et 2 O three times and dried at 60 C for 24 h in a vacuum oven. The resulting product, MAI, was used without further purication. To obtain a CH 3 -NH 3 PbI 3 precursor, the synthesized PbI 2 and MAI were deposited on the top of poly (3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS)-coated substrate using a modied sequential deposition technique: (a) PbI 2 /DMSO deposition, (b) MAI/IPA deposition at a 1 : 1 mole ratio, and (c) solvent-antisolvent exposure, in which the solvent is DMSO, and the antisolvents are CF, CB, DCB, Tol, and Et 2 O.

Thin-lm preparation
To study the effect of a solvent, DMSO was used to prepare 1 M of PbI 2 (461.78 mg ml À1 of DMSO) solution and annealed at 80 C for 12 hours. ITO coated glass substrates were used to deposit the samples and sequentially washed with detergent, DI water, and ethanol in an ultrasonic bath. A hole transporting material PEDOT:PSS was deposited on the top of the ITO glass substrate. Then, the PbI 2 precursor solution was ltered by a 2 mm sterile polytetrauoroethylene (PTFE) membrane lter. PbI 2 /DMSO was spin-coated on the top of the PEDOT:PSS layer and heated at a temperature of 100 C. Then, the as-synthesized MAI in isopropyl alcohol (IPA) solution was spin-coated on the top of the PbI 2 layer. Aer deposition of the MAI, the thin lm was exposed to a DMSO vapor at 80 C for 10 minutes, and then the crystallizable perovskite layer was exposed to different antisolvent vapor at 70 C for 10 minutes. The optical, structural, morphological, and surface properties of organometal halide perovskite thin lms were then investigated, accordingly.

Characterization
UV-visible spectra measurements were taken on thin lms using Shimadzu UV-2600-Series, diffused reectance spectrophotometer double-beam source at a range of 350-800 nm tted with deuterium and halogen lamps as sources. Transmission and reection modes were recorded simultaneously. ITO glasses have been used as a background reference for thin lms. The X-ray diffraction (XRD) patterns of the as-prepared perovskite samples have been characterized by Philips X'pert PRO-240 mm diffractometer provided with an integrated germanium detector Cu-Ka, radiation source at l ¼ 1.54060Å operating at an applied voltage of 45 kV with a current intensity of 40 mA. The equatorial scans in the continuous mode were taken from 2q ¼ 4 to 80 at a step of 0.017 with a scan step time of 24.4 seconds. To probe the electron transition behavior of samples photoluminescence (PL) spectroscopy was utilized by F-7000 uorescence spectrophotometer, Hitachi, Japan, using a xenon lamp at a wavelength of 532 nm with an emission wavelength starting from 600 nm to 850 nm at a scan speed of 1200 nm min À1 . The morphologies of perovskite lms were investigated by eld-emission scanning electron microscopy (FE-SEM; Hitachi, Japan, SU8000 Series) at an accelerating voltage of 5.0 kV. Energy-dispersive X-ray spectroscopy (EDX) investigation has been performed to probe the elemental distribution variations present in the thin lm. Mapping analysis has been done to obtain elemental maps in the range of nanoscale. Contact Angle Goniometer (KRUSS GmbH, DSA25; Germany) was used to record and analyze the effect of antisolvents on the surface energies of perovskite thin lms. To determine contact angle, the edge detection of the water droplet was tted by a polynomial tting approach. Measurements were taken in time intervals of 40 ms over a period of 30 seconds.  In the text, the dimension of (cal cm À3 ) 1/2 was used for solubility parameter. top of PbI 2 /PEDOT:PSS/ITO. Finally, the solvent (DMSO) and antisolvent (CF, CB, DCB, Tol, Et 2 O) vapors are sequentially exposed to the perovskite layer. The properties of solvent and antisolvents were summarized in Table 1. Here, the solubility parameter (d) with the dimension of (cal cm À3 ) 1/2 is in the order of 14.5 (DMSO) > 10 (DCB) > 9.5 (CB) > 9. For solvent-antisolvent vapor engineering, the solvent/ antisolvent should be miscible, whereas the perovskite/ antisolvent immiscible. During the lm-formation process, if the number of nucleation cites is reduced, the crystal and grain size of perovskite may increase, resulting in a high quality lm with small grain boundaries. For this purpose, the solvent DMSO molecules should be quickly extracted from the wet DMSO/perovskite lm by help of antisolvent. 61,62 Furthermore, it is notable that although perovskite is hygroscopic and hydrophilic, the measured water-contact angle was reported to be very high (i.e., signicantly hydrophobic). This paradox was solved by recognizing that the hydrophobic PbI 2 is formed at the interface of water and CH 3 NH 3 PbI 3 . 63 In other words, the measured water contact angle is not for CH 3 NH 3 PbI 3 , but for PbI 2 (i.e., the result of a perovskite degradation). Here, of course, the contact angle data may include the effect of the morphologies of a lm including grain boundaries. Fig. 1(b) shows the chemical structure of solvent and antisolvents. Here the solubility parameter (d) is equal to a square root of cohesive energy density (CED), i.e.,

Results and discussion
=V , whereÛ vap is the molar heat of vaporization, andV is the molar volume. 64 Furthermore, two small organic molecules (here, solvent and antisolvent) are expected to be miscible because of a large entropic gain, although there is an enthalpic cost from the apparent dissimilarity in solubility parameters. Hence, DG mix ¼ DH mix À TDS mix < 0, in which DG mix , DH mix , andDS mix denote the Gibbs free energy, enthalpy, and entropy of mixing, respectively, and T is temperature. On the other hand, for the intermolecular interactions between antisolvent and perovskite, the relation should be DG mix > 0, facilitating a wet perovskite lm to undergo a drying process. Fig. 2(a) shows the UV-vis absorption spectra of perovskite lms as a function of antisolvent species. As shown in Fig. 2(a), although the overall shape of absorption is similar, the absorption edge, i.e., the optical bandgap (E g ), is a little bit different due to a non-identical ordering state of a lm. Here, the absorption data was replot using the Tauc model, 65 where a is the absorption coefficient, b is a constant (disorder parameter), h is Plank's constant and n is the frequency of light. The value 'n' is 1/2 for a direct bandgap semiconductor and 2 for an indirect bandgap. 66 Hence, n ¼ 1/2 can be used because CH 3 NH 3 PbI 3 is included in the former. As shown in Fig. 2(b), the plot (ahn) 2 vs. hn, results in the optical bandgap of $1.61-1.63 eV. As an example, the absorption edge is 770.19 nm (E g ¼ 1.61 eV) for Et 2 O vapor condition, whereas it is 760.74 nm (E g ¼ 1.63 eV) for 'None' condition, i.e., the perovskite sample was not exposed to any solvent/ antisolvent vapor. Here, the small bandgap indicates that the perovskite semiconductor has a well-organized structure, as observed in other stereoregular polymer semiconductors through red-shi in the absorption spectra. [67][68][69] Note that, if the perovskite becomes a single crystalline wafer, the bandgap was reported to be much smaller like 1.36 eV, corresponding to the light absorption onset at 910 nm. 70 This trend indicates that the allowed energy states of an electron increase with reducing defect densities in the crystalline lattice forming a periodic potential. In other words, the energy band increases and the bandgap decreases if the quality of perovskite lms is improved. Furthermore, if there are any defects in perovskite, the typical trap energies are known to be shallow because of its defecttolerance property. [71][72][73] Hence, based on the optical data, the ordering of perovskite materials is in the order of: Et 2 O > Tol > DCB > CB > CF > 'None'. Interestingly, if there is no solventantisolvent vapor treatment, the perovskite sample exhibits the smallest optical absorption, indicating that the vapor treatment is a useful technique for organizing the perovskite lms.   properties such as d ¼ 7.4 (cal cm À3 ) 1/2 and bp ¼ 34.6 C should be helpful to extract DMSO from the wet DMSO/perovskite lm. Fig. 4 shows the elemental mapping images of perovskite lms for the three representative cases, (a) 'None', (b) Tol, and (c) Et 2 O. Here, the mapping data follows the morphologies of a sample according to the SEM images (Fig. 3). Accordingly, Et 2 O-treated perovskite lm shows a uniform distribution of organic/inorganic elements, whereas Tol-treated one exhibits some voids/pinholes as shown in Fig. 4. Fig. 5(a) shows XRD patterns for the perovskite lm as a function of antisolvent species at room temperature. Importantly, CH 3 NH 3 PbI 3 is a polymorphic material, exhibiting the crystal structures of orthorhombic at T < 162.2 K, tetragonal at 162.2 K < T < 327.4 K, and cubic at T > 327.4 K. 74 Indeed, based on the data in Fig. 5(a), the calculated lattice parameters are a ¼ b ¼ 8.87Å and c ¼ 12.65Å, conrming that perovskite has a tetragonal structure at $298 K according to the literature report. 75 Interestingly, in Fig. 5(a), it is noticeable that 'None/CF/ CB/DCB' conditions display unreacted PbI 2 peak at 2q z 13 , 59 whereas Et 2 O and Tol conditions do not exhibit such a peak from unreacted PbI 2 . This observation indicates that, in a modied sequential deposition process, PbI 2 compounds would be reacted with MAI completely when DMSO-Tol or DMSO-Et 2 O was used as a solvent-antisolvent couple system. This is because Et 2 O [d ¼ 7.4 (cal cm À3 ) 1/2 and bp ¼ 35 C] and  Tol [d ¼ 8.9 (cal cm À3 ) 1/2 and bp ¼ 111 C] are relatively nonpolar and volatile, allowing the wet DMSO/perovskite lm to be dried fast (i.e., the mixed DMSO-Tol or DMSO-Et 2 O molecules are quickly evaporated from the hygroscopic perovskite). This rapid crystallization results in a complete reaction between PbI 2 and MAI. Furthermore, based on the most intense peak at (110) crystallographic planes in Fig. 5(a), the crystallite size of each perovskite lm could be estimated. The result is displayed in Fig. 5(b). Importantly, the trend of crystallite size variation is in line with the UV-vis absorption data. However, one exception was observed in "CF' condition which has volatile characteristics (bp ¼ 61 C). Recall the boiling point is in the order of 189 C (DMSO) > 180 C (DCB) > 131 C (CB) > 111 C (Tol) > 61 C (CF) > 35 C (Et 2 O). Table 2 shows the crystallite size of (110) crystallographic plane when d-spacing is 0.623 nm. Here the crystallite size (t) was calculated based on Scherrer's equation as follows, 76,77 where l (¼ 0.154 nm) is the wavelength of X-ray, and B is a full width at half maximum (FWHM) at diffraction angle, q. Here, dspacing was calculated based on the Bragg's law (l ¼ 2d sin q). Fig. 6 shows PL spectra for perovskite lm as a function of antisolvent species, in which the peak was observed at 792.6 nm ('None'), 792.0 nm (CF), 792.2 nm (CB), 792.5 nm (DCB), 792.6 nm (Tol), and 792.2 nm (Et 2 O), indicating the PL peak positions have no direct relationship with the optical bandgap (E g ) shown in Fig. 2(b). However, the PL intensity has a direct correlation with the E g in the UV-vis absorption data. For example, when E g is 1.61 eV (the most red-shi sample), the PL intensity is highest, indicating that, when crystallite size is large in a well-organized morphology, the radiative recombination process is carried out abundantly, resulting in the highest intensity of PL. In other words, when morphologies have a lot of defects like in 'None' or 'CF' conditions, the probability of nonradiative recombination is increased, resulting in a weak intensity of PL as proved in Fig. 6.
Finally, to understand the surface polarity of perovskite lms depending on solvent-antisolvent vapor exposure, the water contact angle (q c : here, subscript 'c' stands for contact angle) was measured (see Fig. 7 and 8). Here, it should be bear in mind that, when water is dropped on the surface of perovskite lm, the nanoscale PbI 2 lm is known to be immediately formed at the interface between water and perovskite through degradation of CH 3 NH 3 PbI 3 . 63 However, despite this PbI 2 formation, the stability of perovskite lm could be studied. This is because polycrystalline morphologies contain a lot of defects such as grain boundaries through which water molecules can be easily penetrated, resulting in the change of surface polarity of a lm.
The raw contact-angle data at step number 17 is displayed in Fig. 8 as an example. Fig. 7(a) shows contact angle change as a function of step number in which each measurement was taken in time intervals of 40 ms over 30 s. In Fig. 7(a), the rst striking observation is that, with increasing the step number, the contact angle decreased, indicating the polarity of a perovskite lm was changed through the water-induced degradation effect. Note that in our previous work, 27 the contact angle and surface energy for the pure PbI 2 lms (DMSO used as a processing solvent) were 130 and 6.3 mJ m À2 , respectively. However, in this work, the perovskite lm (from which PbI 2 is formed, like a water/PbI 2 /CH 3 NH 3 PbI 3 conguration) shows the water contact angle of about 120 and the average surface energy of $11.5 mJ m À2 (see Step 1 in Table 3). Hence, the water contact angle of a perovskite lm should be affected by perovskite's degradation (PbI 2 ), morphologies (including grain boundaries), and others.
The change of water contact angle with time was smaller for the cases of Et 2 O and Tol compared to the others, indicating that, when the perovskite materials were well crystallized (recall Fig. 3), the stability of lms (i.e., water-resistivity) should be signicantly improved in humid conditions. The next observation is that at steps 17 and 18, the contact angle was saturated as shown in Fig. 7(b). In this study, it is noticeable that considering the golden triangle in solar cells (that is efficiency, stability, and cost), 78 this stability-enhanced perovskite lm should be important, providing a general insight for the necessity of a single crystal 70 without any grain boundary as an ideal condition if there is a practical processibility.

Conclusion
The morphologies and surface properties of CH 3 NH 3 PbI 3 thin lms were studied by varying solvent-antisolvent vapor treatment conditions, for which the solvent was dimethyl sulfoxide (DMSO), and the antisolvents were chloroform (CF), chlorobenzene (CB), dichlorobenzene (DCB), toluene (Tol), and diethyl ether (Et 2 O). Major ndings are as follows. First, according to UV-vis absorption data, the optical bandgap of perovskite lms ranged from 1.61 eV (Et 2 O) to 1.63 eV ('None': without any solvent-antisolvent vapor treatment). Second, according to SEM images, when antisolvent was Et 2 O or Tol, the morphologies and crystal structures of perovskite lms were improved. Third, when Et 2 O or Tol was used as an antisolvent, the precursor materials (PbI 2 and CH 3 NH 3 I) were completely reacted (i.e., without any PbI 2 residue) according to the XRD patterns. Forth, according to PL emission data, when the crystallite size (t ¼ 89.5 nm for both 'Et 2 O' and 'Tol' conditions) was large, the PL intensity was higher than those of the other conditions (DCB, CB, CF, and 'None'). Fih, by measuring the water contact angle as a function of antisolvent species, the surface energy (g sv ) of each perovskite lm was estimated. Initially, the average g sv for all samples was 11.53 AE 0.64 mJ m À2 . However, when the contact angle data were saturated at step number 17, the g sv values were different depending on the antisolvent condition: g sv ¼ 25 mJ m À2 (Et 2 O), and g sv ¼ 54.3 mJ m À2 (None), indicating that the high-quality lms (exposed by Et 2 O) have more stability toward water compared to the others. Hence, the solvent-antisolvent vapor technique should be useful for the enhanced stability of perovskite layers if it is well utilized. Finally, our future works may include the device performances by extending the current study, leading to the processing-structure-property-performance relationship of perovskite solar cells.

Conflicts of interest
The authors declare no competing nancial interest. Table 3 Water contact angle ( ) and surface energy (g sv ) of organometal halide perovskite thin films at the step numbers of 1, 17, and 18 Antisolvent Contact angle ( ) Surface energy (mJ m À2 ) Step