Using soft polymer template engineering of mesoporous TiO2 scaffolds to increase perovskite grain size and solar cell efficiency.

The mesoporous (meso)-TiO2 layer is a key component of high efficiency perovskite solar cells (PSCs). Herein, pore size controllable meso-TiO2 layers are prepared using spin coating of commercial TiO2 nanoparticle (NP) paste with added soft polymer templates (SPT) followed by removal of the SPT at 500 °C. The SPTs consist of swollen crosslinked polymer colloids (microgels, MGs) or a commercial linear polymer (denoted as LIN). The MGs and LIN were comprised of the same polymer, which was poly(N-isopropylacrylamide) (PNIPAm). Large (L-MG) and small (S-MG) MG SPTs were employed to study the effect of template size. The SPT approach enabled pore size engineering in one deposition step. The SPT/TiO2 nanoparticle films had pore sizes > 100 nm; whereas, the average pore size was 37 nm for the control meso-TiO2 scaffold. The largest pore sizes were obtained using L-MG. SPT engineering increased the perovskite grain size in the same order as the SPT sizes: LIN < S-MG < L-MG and these grain sizes were larger than obtained using the control. The power conversion efficiencies (PCEs) of the SPT/TiO2-devices were ~ 20% higher than that for the control meso-TiO2 device and the PCE of the champion S-MG device was 18.8%. This is the highest PCE reported for a PSC prepared using a pore-size engineered TiO2 layer to our knowledge and is also one of the highest for PSCs prepared using ITO/meso-TiO2. The PCE improvement is due to the increased grain size and more effective light harvesting of the SPT devices. The increased grain size was also responsible for the improved stability of the SPT/TiO2 devices. The SPT method used here is simple, scalable and versatile and should also apply to other PSCs.


■ INTRODUCTION
The power conversion efficiency (PCE) of perovskite solar cells (PSC) has increased from 3.8% in 2009 1 to 25.2%, 2 which is exceptional. Many high-efficiency PSCs use a mesoporous (meso-) TiO 2 layer, 1,3−7 which evolved from dye-sensitized solar cells. The electron transport within the meso-TiO 2 layer has been enhanced by doping Li, Ta, and Nb 8−11 into the TiO 2 layer. The PCE of PSCs and other solar cells 12 has been increased through pore size engineering of meso-TiO 2 . 13−1513−15 A seminal study by Hwang et al. used polystyrene (PS) particles as templates to include wellcontrolled pores within meso-TiO 2 layers for PSCs. 13 However, the PS particles were hard spheres in the solvent used to deposit the TiO 2 nanoparticles (NPs), which brings into question the important issue of template colloidal stability during film formation. Shao et al. added a linear polymer into TiO 2 paste to increase the pore size of the printed TiO 2 layer for PSCs. 14 Their polymer system was not particulate and should have dissolved well in the solvent used to deposit the TiO 2 NPs. Indeed, their PCE values were relatively high. Their system contained multiple components and home-made TiO 2 nanoparticles (NPs), which may render adoption of that approach challenging. In contrast, we aimed to use a polymer that was in the soft (swollen) state and as many commercial ingredients as possible to provide a simple and versatile templating method for improving the control of meso-TiO 2 pore size using pore size engineering. We hypothesized that increasing the meso-TiO 2 pore size would increase perovskite (PVK) grain size and PCE.
The regular n−i−p architecture has been widely used for fabricating PSCs 16 and consists of glass/indium tin oxide (ITO) (or fluorine-doped tin oxide, FTO)/electron-transport layer/perovskite layer/hole-transport layer/metal electrode. Compared with a planar architecture, PSCs with a meso-TiO 2 layer generally show a higher efficiency 17−22 and lower hysteresis. 17,18,22,23 This is because a larger interfacial area between TiO 2 and the perovskite in the mesoporous structure leads to faster electron transport. The extent of infiltration of the perovskite within the meso-TiO 2 layer strongly affects the PCE and this has led researchers to engineer the porosity. Xiao et al. used nanoporous TiO 2 spheres to prepare PSCs. 24 Ramos et al. constructed one-dimensional (1D) TiO 2 nanocolumnar electron-transport layers (ETLs) to give viable PSCs. 25 Chen et al. used three-dimensional (3D) rutile TiO 2 nanodendrite arrays for their PSCs. Granados et al. mixed poly-(vinylpyrrolidone) (PVP) within the TiO 2 layer and achieved pore sizes of 15−30 nm. 26 Horantner et al. used PS spheres to increase the pore size of TiO 2 . 2727 However, the above methods could not be performed in one step or only gave modest PCE values. None of these approaches used templating species that were colloidally stable when mixed with dispersed TiO 2 nanoparticles (NPs). However, colloidal stability is essential to achieve well-distributed template particles. What is missing from the literature is a simple method that enables pores greater than 100 nm in size to be prepared through onestep TiO 2 NP/template deposition. Here, we introduce such an approach to engineer pores greater than 100 nm in templated meso-TiO 2 films and use these films to provide high PCE PSCs. We denote this approach as a soft polymer template (SPT) pore size engineering because the polymers employed are in the soft (swollen) state at the point where the TiO 2 NP scaffold is deposited.
The new pore size engineering approach presented in this study is very different from additive engineering, which has been widely studied for PSCs. 28,29 The former involves using a removable template to adjust the size of the pores within the meso-TiO 2 scaffold. Additive engineering involves addition of small molecules, polymers, 30 surfactants, or quantum dots 29 to the precursor perovskite (PVK) solution 31,32 or PVK interfaces. 29 These additives usually directly interact with PbI 2 or PVK, control crystallization or provide passivation, and remain within (or at the surface of) the PVK grains. For example, Zuo et al. studied polymer-modified PVKs and used poly(4-vinylpyridine) to form coordination bonds with PVK crystals during crystallization. 30 In contrast, we use our SPT additives as pore size templates for the meso-TiO 2 scaffold and remove them before the PVK precursor solution is added. Hence, there is no interaction between the SPT and PVK.
Two of the SPTs used in this study are microgels (MGs). MGs are crosslinked polymer colloids that swell in a good solvent. 33,34 MGs have excellent colloidal stability and filmforming properties. 34 Our group has used hydrophobic PS MGs in the hole-transport matrix of PSCs 35 as well hydrophilic poly(N-vinylformamide) (PNVF) MGs as additives for the perovskite photoactive layer. 36 We postulated that the swollen nature of MGs would give a potentially useful additive for pore formation in the meso-TiO 2 layer. Unfortunately, neither PS MGs or PNVF MGs swell in ethanol (EtOH), the solvent used for TiO 2 paste, which prevented colloidal stability being achieved. However, EtOH is a thermodynamically good solvent for PNIPAm. 37 Therefore, poly(N-isopropylacrylamide) (PNIPAm) was selected to construct our SPT MG particles. It is the combination of the amide and isopropyl groups within PNIPAm that enabled the polymer to swell in EtOH. We stress that this study is the first time that (i) MGs have been used to adjust the pore size of meso-TiO 2 films and (ii) MG template meso-TiO 2 films have been used to construct PSCs. Furthermore, in contrast to our earlier studies where MGs were used as additives, 35,36 here the MGs are removed (burnt out) by heating to 500°C and do not remain in the PSCs. The method used to prepare the SPT/TiO 2 films and perovskite films (SPT/TiO 2 /PVK) is depicted in Figure 1. The PCEs of the present devices are 2−3 times higher than those obtained for the earlier PSCs that contained MGs. 35,36 Figure  S1 summarizes the previous PSC studies using MGs as well as the PCEs and confirms the novelty of the present work.
Synthesis of Small Microgels. The S-MG particles were prepared by precipitation polymerization following a reported method. 38 Briefly, NIPAm (1.50 g, 13.2 mmol), mBAAm (20.6 mg, 0.13 mmol), and SDS (40 mg) were added to water (100.0 mL). The solution was degassed with N 2 gas for 1 h in a 250 mL glass reactor equipped with a stirrer and the solution was heated to 70°C. Then, APS (65 mg) dissolved in water (0.30 mL) was quickly added to start the polymerization and stirring was maintained. The reaction mixture was stirred for 4 h under a nitrogen atmosphere and cooled to room temperature. The dispersion was then dialyzed against EtOH for 6 weeks with regular solvent changes to replace the dispersion medium with EtOH.
Synthesis of Large Microgels. The L-MG particles were prepared by precipitation polymerization based on the work of Zhan et al. 39 Briefly, NIPAm (188 mg, 1.62 mmol), mBAAm (3.0 mg, 0.019 mmol), MAA (16.2 mg, 0.18 mmol), and SDS (5.4 mg) were dissolved in water (20.0 mL). This mixture was then degassed with N 2 for 1 h at 70°C in a 50 mL glass reactor equipped with a stirrer. To begin particle growth, KPS aqueous solution (5.0 mL, 10 mM) was quickly added. The polymerization was conducted at 70°C for ∼7 h under a N 2 atmosphere. The aqueous dispersion medium was replaced using five centrifugation (at 1180 g)/resuspension cycles in EtOH.
Mesoporous TiO 2 Film Preparation. For the preparation of the control meso-TiO 2 film, 18 NRT paste (1.0 g, 19 wt %) was diluted to 9.9 mL with EtOH to give a TiO 2 concentration of 19.0 mg/mL. Then, the TAA solution (80 μL) was added to give a concentration of 6.0 mmol. Control TiO 2 films were made by dropping a portion of the TiO 2 dispersion (80 μL) onto ITO glass and then spin coating at 2000 rpm for 30 s. Finally, the substrate was transferred into a preheated furnace and sintered at 500°C for 30 min.
Linear PNIPAm/TiO 2 Film Preparation. TiO 2 NP/LIN dispersion was prepared by mixing LIN powder (21 mg) into TiO 2 paste (1.0 g, 19 wt %). Then, TAA (80 μL) and EtOH (9.92 mL) were added. The combined TiO 2 NP/LIN/TAA/EtOH dispersion was mixed with the aid of sonication. LIN/TiO 2 films were prepared by pipetting the dispersion (80 μL) onto ITO glass followed by spin coating at 2000 rpm for 30 s. The film was sintered as described above.
Microgel/TiO 2 Film Preparation. SPT/TiO 2 films prepared using S-MG or L-MG followed the same procedure. TiO 2 NP/MG dispersion was prepared by mixing MG dispersion (0.20 g/mL, 0.050 mL) into TiO 2 paste (1.0 g, 19 wt %). Then, TAA (80 μL) and EtOH (9.87 mL) were added. The combined TiO 2 NP/MG/TAA/EtOH dispersion was mixed with the aid of sonication. The spin-coating procedure and sintering conditions used to prepare the MG/TiO 2 films were the same as those described above for the LIN/TiO 2 film.
Perovskite Film Preparation. A PbI 2 solution (75 μL) containing PbI 2 (500 mg) at a concentration of 1.0 mM in a mixed solvent of DMF (950 μL) and DMSO (50 μL) was dropped onto the TiO 2 NP films on ITO and then spun at 1500 rpm for 30 s. The substrate was then placed onto a 70°C hot plate and heated for 1.0 min and cooled to room temperature. Next, 100 μL of the double cation solution FAI (60 mg, 0.349 mM), MABr (6.0 mg, 0.054 mM), and MACl (6.0 mg, 0.089 mM) was added to 1.0 mL of the IPA solution and dropped onto the PbI 2 films. The substrate was then spun at 2000 rpm for 30 s. The FA 1−x MA x Pb(I, Br, Cl) 3 film was transferred onto a 150°C hot plate and annealed for 15 min. For XRD studies, PMMA/CBZ solution (10 mg/mL) was spin-coated onto the perovskite film at 4000 rpm for 10 s for encapsulation. The thicknesses of the SPT/TiO 2 /PVK films prepared using L-MG, S-MG, LIN as SPTs and the control TiO 2 were measured (Dektak) to be 355 ± 14, 390 ± 19, 355 ± 9, and 346 ± 16 nm, respectively.
Device Fabrication. ITO glass (Ossila) was patterned using Zn powder and HCl (4.0 M), then cleaned with water and sonicated using acetone, IPA, and EtOH and washed with water, then dried under a N 2 stream, and then treated by UV-ozone plasma for 20 min. The TiO 2 blocking layer was deposited onto the cleaned ITO substrate by spin coating TIP solution (70 μL), 5000 rpm for 30 s and then sintered at 450°C for 45 min. The SPT/TiO 2 or meso-TiO 2 films were then deposited using the conditions described above. After TiO 2 layer deposition, the substrates were transferred to a 2% humidity nitrogen-filled glovebox for the next stages of fabrication. The perovskite layer was prepared via a two-step method, as described above. A Spiro solution in CBZ (90 mg/mL) was mixed with TBP (34 μL), Li-TFSI solution (17 μL of 517 mg/mL in acetonitrile), and FK209 Co(III) TFSI (10 μL of 300 mg/mL in acetonitrile). The Spiro layer was spin-coated at 4000 rpm for 15 s. Finally, a Au counter electrode (80 nm) was deposited by thermal evaporation under high vacuum. A 0.080 cm 2 metal mask was used to define the active solar cell area.
Physical Measurements. Dynamic light scattering (DLS) measurements were conducted using a Malvern Zetasizer Nano ZS instrument and provided the z-average diameter (d z ). Scanning electron microscopy (SEM) was performed using an FEI Magellan 400 XHR FEG-SEM instrument. The TiO 2 films were measured using an accelerating voltage of 5 kV. The particle and pore sizes were measured by SEM. At least 100 particles or pores were counted. Perovskite films were measured under a 3 kV accelerating voltage. The cross section of solar cells was prepared via the focused ion beam (FIB) technique using an FEI Quanta 3D FIB/SEM system with multiple current and voltage approaches to minimize spatial sample damage. High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images were captured by using FEI Talos F200A ultrahigh brightness SuperX FEG (200 kV). Chemical analysis was performed using SuperX silicon drift detectors in the Talos, and the built-in pole piece of the objective lens enabled highresolution and high-precision elemental mapping. The thickness of the films was measured using a Dektak XT profilometer. UV−visible spectra were obtained using a Hitachi U-1800 spectrophotometer. PL spectra were measured using a double monochromator FLS980 instrument (Edinburgh Instruments, U.K.). XRD patterns were obtained using a Bruker D8 Advance diffractometer (Cu Kα). TiO 2 and MGs atomic force microscopy (AFM) images were captured using a Bruker catalyst. Films were scanned using tapping mode. The Si AFM probe was purchased from Nunano. The probe was autotuned by the Nanoscope software; the drive frequency was 388 kHz. The scan rate was 0.25 Hz. Images were processed using the Nanoscope software. Perovskite film images were captured using JPK NanoWizard IV. The scanning rate was 6.5 Hz. Images were processed using the JPKSPM Data Processing software. EQE data were obtained using a Newport QuantX-300 instrument. A Micromeritics ASAP 2020 sorption analyzer was used to obtain N 2 adsorption/desorption isotherms at 77 K and to measure the Brunauer−Emmett−Teller (BET) surface area. Nitrogen adsorption/desorption isotherms were measured at 77 K using bulk samples heated at 500°C for 2 h. Helium was used for the free space determination. The pore size distribution was determined using the BJH method. Electrochemical impedance spectroscopy (EIS) measurements were performed using phase-sensitive detection of the current using a lock-in amplifier. An AC voltage of 10 mV was applied, and the frequency swept between 1.0 Hz and 1 MHz. The voltage was supplied using an Agilent 33210A function generator. Ultraviolet photoemission spectroscopy (UPS) was performed using a HIS 13 high intensity VUV source (200 mA emission, Focus GmbH) within an ESCA2SR spectrometer (Scienta Omicron GmbH) using He II (40.5 eV) photons. Thermogravimetric analysis (TGA) was performed in air using a TA Instruments Q800 instrument. The LIN sample was heated in air from room temperature to 490°C at 5°C/ min and then from 490 to 510°C at 0.3°C/min.
Device Measurements. Current−voltage measurements of the PSCs were measured under a class AAA solar simulator (Newport Oriel Sol3A) at AM1.5 and 100 mW/cm 2 illumination conditions and calibrated against a KG5 filtered silicon reference cell (Newport Oriel 91150-KG5) using a Smart PV and OLED Multiplexor Test Board. The J−V measurements were conducted in ambient air. The devices were measured following 30 s light soaking, with a 0.020 V/s scan rate and 0.1 s settling time unless otherwise stated. Both forward (short circuit to open circuit) and reverse (open circuit to short circuit) sweeps were measured.

■ RESULTS AND DISCUSSION
Soft Particle-Templated TiO 2 Nanoparticle Film Morphology. We studied three SPT species in this work. Two SPTs were MGs (L-MG and S-MG) and the third was a commercially available linear polymer (LIN). All of the SPTs are composed of PNIPAm ( Figure 1). PNIPAm was selected as the template polymer because it is soluble in EtOH, which is a key solvent used in TiO 2 pastes. 8,22,40 PNIPAm is a temperature-responsive polymer in water 41 and collapses (i.e., MG particles and LIN coils deswell) above 32°C. At such temperatures, water is a bad solvent. DLS data for L-MG, S-MG, and LIN in EtOH (a good solvent) and water at 60°C (a bad solvent) were obtained (see Figure S2A,B) and are discussed in the Supporting Information. MG particle volume swelling ratios (α p ) in EtOH for L-MG and S-MG of 25 and 15, respectively, were calculated from the DLS data. These values show that the L-MG particles swelled more in EtOH than the S-MG particles. SEM images for the SPTs deposited from water at 60°C are shown in Figure 2. The average diameters for L-MG, S-MG, and LIN obtained from SEM (D SEM ) are 264, 88, and 53 nm, respectively. The LIN particles are aggregates of collapsed polymer chains, whereas the MGs are individual particles. A summary of all of the SPT size data is presented in Table S1. All of the characterization data show that the SPT size decreases in the order L-MG > S-MG > LIN.
SPT/TiO 2 scaffolds were prepared following the method depicted in Figure 1. SEM images, pore size distributions, and average pore sizes for the nonannealed SPT/TiO 2 and meso-TiO 2 films are shown in Figure S3. It is important to note that the SPTs were removed from all of the scaffolds in this study by sintering at 500°C. This was demonstrated in three ways. First, L-MG microgel dispersion was placed on ITO-coated glass, which was heated at 500°C on a hot plate. The solvent rapidly evaporated and then the polymer turned black and was completely burnt off within 397 s (See Movie S1 and Figure  S4). Second, SEM images of L-MG before and after being heated at 500°C were measured (See Figure S5). They show that the L-MG particles completely disappeared as a result of the heat treatment. Third, TGA data for LIN ( Figure S6) show that ∼95% of the mass was lost as a consequence of heating to 500°C. The discussions that follow concern SPT/TiO 2 and meso-TiO 2 layers annealed at 500°C for 30 min unless otherwise stated.
SEM images for the MG-and LIN-templated TiO 2 films as well as the control meso-TiO 2 film are shown in Figure 3. Representative pores are highlighted by yellow arrows. While the MGs increased the pore size within the TiO 2 films when a concentration of 5 wt % was used ( Figure 3A,B), we found that a LIN concentration of 10 wt % was required to achieve comparably large pores ( Figure 3C). The pore size distributions measured from the SEM data ( Figure 3) show that relatively large pores are present for the TiO 2 scaffold prepared using L-MG ( Figure 3A), although a wide range of sizes is evident. The pore size distribution is relatively narrow for S-MG/TiO 2 ( Figure 3B) and LIN/TiO 2 ( Figure 3C). The average pore sizes measured for L-MG/TiO 2 , S-MG/TiO 2 , and LIN/TiO 2 films as well as the meso-TiO 2 film are 112 ± 86, 113 ± 49, 66 ± 33, and 37 ± 12 nm, respectively. This trend is similar to those for the films before annealing (see Figure S3) although the pore sizes had increased by a factor of about two due to annealing. The average pore sizes occupy a gap in the literature for pore size-adjusted TiO 2 films deposited in one step for PSCs. P123 polymer-adjusted meso-TiO 2 had a pore size of 16−34 nm, 14 whereas TiO 2 scaffolds prepared with a pore diameter of ∼200 nm were prepared using sacrificial PS particles. 13 The data in Figure 3 show that larger pores are produced with larger SPTs.
The final stages of the spin coating of the SPTs from EtOH will necessarily correspond to the collapsed PNIPAm state as The relationship between collapsed L-MG size and the pore size distribution ( Figure 3A) is less clear. While there are some pores similar or larger than D SEM (residing at and to the right of the red arrow), there is also a high proportion of pores residing to the left on the small size side. The latter may be due to L-MGs that deformed nonuniformly due to the relatively large shear stresses imposed during scaffold deposition via spin coating. L-MG swelled more strongly than S-MG (discussed above) and is also larger, which are two factors favoring deformation. L-MG is also lightly crosslinked. Lightly crosslinked, large, PNIPAm MGs are well known to be able to deform nonuniformly. 42 The tapping mode AFM images in Figure 3 were obtained and confirmed the evolution of the pore diameter with the SPT size discussed above. Wider-area AFM image views and line profiles for the four TiO 2 films are shown in Figure S7. The RMS roughness values for the films decrease in the order: L-MG/TiO 2 (23.7 nm) > S-MG/TiO 2 (21.3 nm) > LIN/TiO 2 (18.7 nm)> control/TiO 2 (13.6 nm). These data demonstrate the ability of the larger SPTs to create a more undulating surface. The film thicknesses for L-MG/TiO 2 , S-MG/TiO 2 , LIN/TiO 2 , and meso-TiO 2 film were 147 ± 26, 147 ± 14, 231 ± 12, and 135 ± 11 nm, respectively. The LIN/TiO 2 film was significantly thicker than the rest of the films. This may be due to the higher polymer content used for that film.
Nitrogen adsorption/desorption isotherms were measured for these systems. The minimum sample mass requirements necessitated the use of bulk scaffolds prepared by drop casting rather than spin coating. The pore size distributions appear in Figure S8. The average surface area values and pore sizes were 79−82 m 2 /g and 21−22 nm, respectively. There was no measurable difference between these values for the different scaffolds. These results do not agree with both the TEM ( Figure 3) and AFM (Figures 3 and S7) for the spin-coated scaffolds. This discrepancy is due to (a) the use of bulk samples, which meant that the scaffold formation mechanism lacked the rotational forces and large material loss of spin coating 43 and (b) the relative insensitivity of the BET method to pore sizes larger than ∼20 nm. 44 It is suggested that SPT/ TiO 2 morphology observed ( Figure 3) is favored by spin coating. We use the TEM pore sizes for the discussions below.
The ability of light to pass through the SPT/TiO 2 films is important for solar cell applications. Transmittance spectra of the films were measured ( Figure S9A) and are similar to those reported elsewhere for TiO 2 films. 45 The spectra show that the LIN/TiO 2 film had the lowest average transmittance (71.3%) over the 300−800 nm range ( Figure S9B). This is because the LIN/TiO 2 film was thicker than other SPT/TiO 2 films and meso-TiO 2 , as discussed above. The S-MG/TiO 2 and L-MG/ TiO 2 films had average transmittance values of 76.1 and 77.2%, respectively. These values are only slightly less than that measured for the meso-TiO 2 film of 78.6%.
Perovskite Films Prepared Using SPT/TiO 2 Films. We next prepared perovskite films using the two-step method depicted in Figure 1. SEM images and grain size distributions obtained from the SEM data are shown in Figure 4. Large grains (with many greater than 1 μm) can be seen in the films prepared using each of the SPT/TiO 2 films ( Figure 4A−C). In contrast, smaller grains are present for the meso-TiO 2 /PVK film ( Figure 4D). (Larger area SEM images are shown in Figure S10.) The average grain sizes for the films grown using L-MG/TiO 2 , S-MG/TiO 2 , LIN/TiO 2 , and meso-TiO 2 films are 865 ± 175, 755 ± 215, 660 ± 225, and 570 ± 170 nm, respectively. Hence, the average grain sizes decreased in the order: L-MG/TiO 2 /PVK > S-MG/TiO 2 /PVK > LIN/TiO 2 / PVK > meso-TiO 2 /PVK. Importantly, this order follows that of the D SEM values (Table S1). However, this order does not fully agree with the average pore sizes discussed above because the average pore sizes were not distinguishable for L-MG and S-MG. This result implies that crystals grown from the largest pores within the distribution of pores dominated the final crystal size.
AFM images were obtained for the films and are also shown in Figure 4. (Larger area AFM images and line profiles are shown in Figure S11.) The differences in the crystal size discussed above are confirmed by the AFM data. The RMS roughness values for the films grown using L-MG/TiO 2 , S-MG/TiO 2 , LIN/TiO 2 , and meso-TiO 2 films are 22.1, 25.3, 22.2, and 19.9 nm, respectively, based on line profiles. Hence, the SPT/TiO 2 /PVK films have greater roughness than the control meso-TiO 2 /PVK film. The RMS roughness values for the underlying SPT/TiO 2 scaffolds were also greater than that for the meso-TiO 2 scaffold (see above). Differences in the initial heights of the TiO 2 scaffold can be expected to be approximately translated to the final PVK films if capping layers recrystallize on the scaffolds. 46 In addition, the increased grain size of the SPT/TiO 2 /PVK films likely also contributes to their increased roughness.  The increased grain size for the SPT/TiO 2 /PVK films likely originates from less restriction during crystal growth as perovskite crystal domain size can be constrained by the pore size of preformed TiO 2 scaffolds. 27 Furthermore, an increase in the pore size for the meso-TiO 2 layer is known to increase PbI 2 infiltration 47 due to faster and less obstructed mass transport. Fu et al. proposed a dissolution/recrystallization model for two-step deposition 46 wherein supply of PbI 2 from within the scaffold to the surface caused MAPI crystals to grow in the capping layer. For the present TiO 2 scaffolds, as the pore size of the SPT TiO 2 scaffold increases, the rate of supply of PbI 2 should also increase, which is proposed to increase the rate of crystal growth and hence the PVK grain size. To test this proposal, we prepared a planar PVK film without any scaffold and hence without obstruction for crystal growth. The SEM images also showed many large grains (see Figure S12). The average grain size was 867 ± 438 nm. While the average grain size is indistinguishable from that for the L-MG/TiO 2 /PVK film ( Figure 4A), the grain size distribution for the planar PVK film is far more polydisperse. This result suggests a beneficial effect of our SPT/TiO 2 approach in delivering films with more uniform grain sizes compared to planar PVK films.
Devices were prepared using the SPT/TiO 2 films and the architecture is depicted in Figure 5A. Data from the champion S-MG/TiO 2 PSC cell are shown in Figure 5B, which had a PCE of 18.8%. The device performance data extracted from the J−V curves are shown in Figure 5C−F and these values are tabulated in Table S2. The highest average V oc was 1.04 V for the S-MG/TiO 2 /PVK system ( Figure 5C). Furthermore, the V oc values for the SPT/TiO 2 -based PSCs were all significantly larger than those for the meso-TiO 2 -based PSC. This is   attributed to the increased grain sizes for the SPT/TiO 2 systems noted above since V oc for PSCs is known to increase with the grain size. 48 Grain boundaries are well known to contain defects and have a high density of trap states. 49 Traps favor nonradiative recombination, which decreases V oc . 5050 Hence, because our devices prepared with the SPT/TiO 2 films have larger grains, they have fewer traps and higher V oc values compared to the system prepared using meso-TiO 2 , which had smaller grains. The fill factor (FF) and J sc values are also higher for the devices prepared using SPT/TiO 2 films than for the meso-TiO 2 films ( Figure 5D,E). This result is expected because increases in V oc are theoretically linked to increases in J sc (and EQE). 51 The devices prepared using S-MG/TiO 2 had the greatest reproducibility and highest average PCE of 17.8% ( Figure 5F). The average PCE values are clearly higher for the films prepared using SPT/TiO 2 compared to that prepared using meso-TiO 2 (14.9%). The average PCE values for the S-MG/TiO 2 /PVK and L-MG/TiO 2 /PVK (17.4%) devices were close to that for the LIN/TiO 2 /PVK devices (17.2%). Hence, it can be concluded that using SPT/TiO 2 films (either MGs or LIN) gave a similar large increase in PCE (an increase of ∼20%) compared to the control meso-TiO 2 system. The variation of J with time ( Figure 5G) and steady-state power output (SPO, Figure 5H) were measured. The SPO values for the SPT/TiO 2 /PVK devices showed much better stability compared to the control. All of these data augur well for versatility and potential widespread use of our new method as the MGs are easily prepared using a scalable preparation. Crucially, LIN is commercially available.
Hysteresis is an important topic for PSCs 52,53 because it is a source of instability that affects PSC performance. The hysteresis index (HI) for devices prepared using L-MG/ TiO 2 , S-MG/TiO 2 , LIN/TiO 2 , and meso-TiO 2 was calculated using the normalized difference between the reverse and forward PCE values. 54 We investigated the effect of the scan rate on the HI over the range of 0.020−0.20 V/s. The data (Table S3) show that the variations in the HI values with the scan rate were modest and a clear trend was not present for all samples. It follows that there are no large differences in the sweep-rate-induced ion accumulation between the samples over the sweep ranges studied. The average HI values for L-MG/TiO 2 , S-MG/TiO 2 , LIN/TiO 2 , and meso-TiO 2 are 5.8, 6.0, 4.4, and 2.4%, respectively (see Figure S13). Hence, the HI increased using SPT/TiO 2 . However, the HI was relatively low for all cases (≤6%). Our trend is opposite to that reported by Shao et al. 14 Planar PSCs are generally known to have greater HI values than meso-TiO 2 -based PSCs. 52 This is because of less perovskite/electron-transport layer contact for planar devices, which results in slower electron extraction. Fast extraction of electrons is aided by meso-TiO 2 due to the high surface area and this is known to give low HI values. 55 Because the SPT/TiO 2 films had larger pores than meso-TiO 2 ( Figure  3), the perovskite/TiO 2 contact area decreased, which would account for the relative increase of HI observed. 56 In regard to achieving a balance between PCE and HI, the LIN/TiO 2 -based device has an optimum balance of improved PCE (average of 17.2%) and low HI (average of 4.4%).
The variation of the normalized PCE for nonencapsulated devices exposed to ambient conditions is shown in Figure 5I. The variation of V oc , J sc and FF appears in Figure S14A−C. The stability of all of the devices decreased at first and then plateaued. The LIN/TiO 2 devices maintained ∼70% of their initial PCE after 1870 h (78 days) and had the best stability ( Figure 5I). The values for S-MG/TiO 2 (67%) and L-MG/ TiO 2 (56%) were higher than those of the control meso-TiO 2 system (49%). Moisture-induced degradation begins at grain boundaries and proceeds inward. 57 Larger grain sizes provide smaller surface area-to-volume ratios, which slow the rate of moisture degradation. 58 We attribute the better stability of the SPT/TiO 2 devices to the larger grain sizes compared to the control prepared using meso-TiO 2 (as shown in Figure 4).
Investigating Origin of Performance Increase for SPT/TiO 2 /PVK Devices. An intriguing question is why the devices prepared using the SPT/TiO 2 had higher PCEs than those prepared using meso-TiO 2 . We investigated the structures of the perovskite films using X-ray scattering ( Figure  S15). The scattering profiles are ascribed to the cubic αperovskite phase and are similar to those reported for related perovskites. 59,60 There are no clear differences for the perovskite phase. Residual PbI 2 is present for the films prepared using S-MG/TiO 2 and LIN/TiO 2 but mostly absent for the films prepared using L-MG/TiO 2 and meso-TiO 2 . We examined a cross section of an ITO/bl-TiO 2 /S-MG/TiO 2 / PVK/Spiro/Au device using HAADF-STEM (see Figure 6A).
The arrows highlight three regions where S-MGs had created surface pores. Such pores can also be seen in Figure 3B (arrow). The elemental mapping shows clearly the presence of I in the surface pores (arrow) as well as infiltration within the scaffold interior ( Figure 6B). The Ti mapping ( Figure 6C) shows the surface pores templated by S-MG clearly. Comparison of Figure 6B,C implies that the perovskite capping layer grew from the perovskite within the surface pores left by the S-MG templates. This suggestion is congruent with the report of Pascoe et al., who noted that crystals grow larger in unrestricted systems, such as planar films. 61 UV−visible spectra of the films are shown in Figure 7A. The S-MG/TiO 2 /PVK film had the highest absorbance. This is  because that film had the largest thickness and roughness (above). The absorbance values for L-MG/TiO 2 /PVK and LIN/TiO 2 /PVK films are also greater than those for the meso-TiO 2 /PVK film. The absorption onsets were all ∼785 nm and unaffected by the pore size. PL spectra were also obtained ( Figure 7B). Those data show quenching increases in the substrate order: meso-TiO 2 < LIN/TiO 2 < L-MG/TiO 2 < S-MG/TiO 2 . This result implies more efficient electron transport to the TiO 2 scaffold from L-MG/TiO 2 /PVK and S-MG/TiO 2 / PVK films compared to LIN/TiO 2 /PVK and meso-TiO 2 /PVK. This would be expected from the larger grain sizes for the films (above) since grain boundaries act as an impediment to charge transport. 62 We measured the EQE for the devices ( Figure 7C). The integrated current densities for the devices prepared using L-MG/TiO 2 , S-MG/TiO 2 , LIN/TiO 2 , and meso-TiO 2 devices were 23.4, 23.1, 23.1, and 21.2 mA/cm 2 , respectively. All of these values are lower than those measured from the J−V curves (Table S2) by an average of 6%. Interestingly, the EQE spectra for the SPT/TiO 2 /PVK films are very similar and there is a pronounced increase in the EQE for these systems at wavelengths greater than 500 nm compared to that for meso-TiO 2 /PVK. Ye et al. 63 attributed enhancements in EQE at such wavelengths to an optical interference effect caused by TiO 2 NP scattering. The pores for the SPT/TiO 2 films are substantially larger than those for the control film ( Figure 3) and relatively intense light scattering is expected. Furthermore, enhanced light scattering from the increased roughness of the SPT/TiO 2 /PVK film surfaces (discussed above) is also expected to have increased the EQE through increased light absorption. Indeed, higher absorption for the SPT/TiO 2 /PVK films was noted above for Figure 7A. Theoretical studies have also predicted increased EQE values for PVK films that scatter more light. 64 While increased light scattering appears to have contributed to the increased PCE of the SPT/TiO 2 -based PSCs compared to the control meso-TiO 2 system, the grain size differences apparent from SEM (and AFM) are likely the main reason for the PCE improvement. 65 EIS measurements were performed for the devices, and the real (Z′) and imaginary (−Z″) components of the impedance are shown in Figure 7D. The data were measured in the dark with a bias of V oc . A single recombination semicircle indicates that only one type of charge transport dominated 66 and has also been reported elsewhere. 67 We initially used a simple ideal semiconductor equivalent circuit model 68 to fit the data ( Figure 7D). The parameters R s , R HF , and C HF are the series resistance, high-frequency resistance, and high-frequency capacitance, respectively. The values from the fits are shown in Table S4. Because the frequencies for the maxima were greater than 300 MHz, the data are associated with the recombination of free carriers. 69 R HF increases with the semicircle size and is inversely proportional to the recombination rate. 70 The latter was lowest for the meso-TiO 2 /PVK device, which had the lowest J sc (Table S2). The electron lifetime (τ HF = R HF C HF ) was calculated 66 (Table S4). The value for meso-TiO 2 of 95 μs is close to that reported elsewhere. 67 The τ HF values for the SPT/TiO 2 /PVK devices (135−325 μs) are all much higher than those for meso-TiO 2 / PVK. This is due to the larger grain sizes of the SPT/TiO 2 / PVK films (Figure 3) because the increased grain size results in less recombination at grain boundaries 56 and higher diffusion lengths. 71 The fit for LIN/TiO 2 /PVK is improved by adding an additional RC circuit element 72 ( Figure S16), and the fitted data are shown in the last row of Table S4. The value for τ HF decreases to 265 μs but remains much greater than that for the meso-TiO 2 /PVK control. The EIS data support the view that the increased grain size is responsible for the higher PCEs of the SPT/TiO 2 /PVK systems. Figure S17 shows the UPS valence spectra for the SPT/ TiO 2 /PVK and meso-TiO 2 /PVK films. The energy difference between the valence band and the Fermi level (E F ) was 0.67 to 0.70 eV for all of the films and was not distinguishable. The spectral features and values are similar to those reported elsewhere for related systems. 73,74 The UV−visible spectra ( Figure 7A) show that the absorption onsets were similar (∼760 nm), which implies that significant bandgap differences between the systems were not present. Hence, the increased porosity for the SPT/TiO 2 /PVK did not significantly affect E F or the positions of the valence or conduction band edges for PVK in these systems.
The performance improvement of the SPT/TiO 2 /PVK devices occurs despite the decreased interfacial contact between PVK and the TiO 2 scaffold that resulted from the increased pore size. Hwang et al. noted that improvement in the grain size of their devices dominated a decreased PVK/ TiO 2 interfacial area and attributed this to a decrease in the number of PVK/interface trap sites. 13 The results of this study support the conjecture that an improvement in the PVK grain size can be more important for increasing PCE than a decrease of the PVK/TiO 2 area. Following a comment from a reviewer, we conducted a two-probe resistance test for ITO annealed at 500°C for 30 min. The resistance increased from 75 to 245 Ω due to the annealing treatment. This trend is in agreement with the literature for ITO. 75 Because FTO is stable to such annealing, 76 it follows that the PCEs reported here could be further increased by replacing ITO with FTO. This will be pursued in future work. Z′ and Z″ are the real and imaginary components of the complex impedance, respectively. R s , R HF , and C HF are the series resistance, high-frequency resistance, and highfrequency capacitance. An extended model that gives an improved fit to the LIN-based data is shown in Figure S16.

■ CONCLUSIONS
In summary, we have investigated the use of three pore size engineering additives to prepare SPT/TiO 2 scaffolds. They were MGs of two different sizes as well as linear PNIPAM. The MGs were prepared using a scaleable method and the LIN is commercially available. All three SPTs gave meso-TiO 2 with an increased pore size, which in turn increased the perovskite grain size. The PSCs for all three systems prepared from SPT/ TiO 2 had substantially higher PCE values and stabilities than those of the control prepared with the meso-TiO 2 . This improvement was due to a combination of increased perovskite grain size as well as improved charge transfer and light scattering. The study has shown that SPTs can increase the perovskite grain size via the pore size increase of the TiO 2 NP layer. The one-step SPT engineering approach to prepare porecontrolled TiO 2 films using scalable and commercially available polymer additives introduced in this study should apply to other PSCs.
Comparison of the present work with previous PSC studies involving MGs, DLS data for MGs and LIN with discussion, video and photographs and SEM of L-MG being burnt off an ITO surface, TGA of LIN being heated to 500°C, TiO 2 AFM images and line profiles, hysteresis index data, sintered TiO 2 transmittance spectra and average transmittance values, pore size distribution data, large-area SEM images for nonannealed and annealed SPT/TiO 2 as well as SPT/ TiO 2 /PVK films and a planar PVK film, device stability data, X-ray scattering patterns, electrochemical impedance spectroscopy data, UPS spectra, SPT particle size data, J−V figures of merit data (PDF) A few drops of L-MG microgel dispersion was placed on ITO-coated glass, which was heated to 500°C on a hot plate; water quickly evaporated and L-MG turned black and then is completely burnt off after only 397 s; photographs from the video are shown in Figure S4 (Movie S1) (MOV)