**Investigation of Well-Defined Pinholes in TiO2 Electron Selective Layers Used in Planar Heterojunction Perovskite Solar Cells**

**Muhammad Talha Masood 1,2, Syeda Qudsia 1, Mahboubeh Hadadian 1, Christian Weinberger 1,3, Mathias Nyman 4, Christian Ahläng 4, Sta**ff**an Dahlström 4, Maning Liu 5, Paola Vivo 5, Ronald Österbacka <sup>4</sup> and Jan-Henrik Smått 1,\***


Received: 17 December 2019; Accepted: 16 January 2020; Published: 20 January 2020

**Abstract:** The recently introduced perovskite solar cell (PSC) technology is a promising candidate for providing low-cost energy for future demands. However, one major concern with the technology can be traced back to morphological defects in the electron selective layer (ESL), which deteriorates the solar cell performance. Pinholes in the ESL may lead to an increased surface recombination rate for holes, if the perovskite absorber layer is in contact with the fluorine-doped tin oxide (FTO) substrate via the pinholes. In this work, we used sol-gel-derived mesoporous TiO2 thin films prepared by block co-polymer templating in combination with dip coating as a model system for investigating the effect of ESL pinholes on the photovoltaic performance of planar heterojunction PSCs. We studied TiO2 films with different porosities and film thicknesses, and observed that the induced pinholes only had a minor impact on the device performance. This suggests that having narrow pinholes with a diameter of about 10 nm in the ESL is in fact not detrimental for the device performance and can even, to some extent improve their performance. A probable reason for this is that the narrow pores in the ordered structure do not allow the perovskite crystals to form interconnected pathways to the underlying FTO substrate. However, for ultrathin (~20 nm) porous layers, an incomplete ESL surface coverage of the FTO layer will further deteriorate the device performance.

**Keywords:** perovskite solar cell; electron selective layer; pinhole; mesoporous TiO2; evaporation-induced self-assembly; dip coating

#### **1. Introduction**

Organic-inorganic lead halide perovskite solar cells (PSCs) have gained substantial attention in the last decade, and soon they are expected to be able to compete with conventional silicon-based solar cells due to their outstanding device performance. PSCs offer low-cost solar energy conversion due to the ease in fabrication and the possibility to make devices on top of glass or flexible substrates [1]. The first PSC was reported by Kojima et al. in 2009 [2] with 3.8% power conversion efficiency (PCE). Within the next seven years, a PCE above 20% was reached [3], while the current certified record efficiency is, amazingly, above 25% [4]. However, there are still some challenges to be addressed before the technology can reach market penetration, including device reproducibility [5], scalability [6], and stability [7].

A PSC consists of a perovskite light-absorbing layer sandwiched between an electron selective layer (ESL) and a hole selective layer (HSL). Electron-hole pairs are generated in the perovskite layer upon illumination. Electrons are selectively extracted by the ESL and transported to an external circuit via a transparent conductive substrate, such as fluorine-doped tin oxide (FTO), while holes are extracted by the HSL, most commonly lithium-doped Spiro-OMeTAD. n-Type metal oxide semiconductors, including TiO2, ZnO and SnO2 [8,9], are the most commonly used ESLs in PSCs, although conductive polymers have also been used [10–12]. The highest efficiency reported for TiO2 as the ESL reaches above 20% when using cesium containing triple-cation mixed-halide perovskite as the light absorber [3]. This involves the use of a thin mesoscopic scaffold layer consisting of TiO2 nanoparticles coated on top of a compact TiO2 layer. The perovskite is thought to be embedded in the mesoporous scaffold along with a thick capping layer on top, which isolates the TiO2 scaffold from the HSL [13]. Together with the perovskite in the mesoporous TiO2, the capping layer absorbs enough light to obtain good photocurrent. The mesoscopic scaffold plays an important role in supporting electron injection into TiO2 because it tends to establish a high TiO2/perovskite interfacial area. Without the scaffold layer, it is difficult to achieve highly efficient solar cells with TiO2, because the low contact area between the perovskite and the TiO2 results in charge build-up at the interface [8,14]. This is due to the relatively low inherent electron mobility in TiO2 and the misalignment of its work function with respect to the conduction band energy level of the perovskite. This further becomes an important source of hysteresis in planar heterojunction TiO2-based PSCs [8,9].

Morphological defects such as pinholes [15] (either in the form of larger bare patches or as porosity on the nanoscale [16]) in the ESL can result in increased surface recombination as well as charge injection barriers due to improper alignment in the energy levels between the perovskite and the ESL. Pinholes in ESLs are predominantly reported as incomplete surface coverage of the ESL film on top of the conductive FTO, which results in FTO crystals protruding through the metal oxide films creating a direct contact with the perovskite [17,18]. This degrades the device performance due to a poor hole-blocking ability or high current leakage at the ESL/perovskite interfaces [9]. These types of pinholes are formed due to poor substrate wettability or microbubble formation during the film deposition or drying stages and are most prominent in ultrathin metal oxide films. Thermally or mechanically induced cracks might also expose the underlying FTO in relatively thick ESLs [1,19]. Furthermore, the choice of deposition technique and film deposition rate are important factors that affect the surface coverage and the denseness of the produced ESL layer [16]. Atomic layer deposition (ALD) is currently considered the best deposition method for producing homogeneous and dense metal oxide thin films with well-defined morphology and crystallinity [16,20]. Whereas the atomic layer-by-layer structure buildup in ALD creates very dense metal oxide layers, the more commonly used solution-processable nanoparticle- or sol-gel-based methods typically produce layers with a lower density, as they contain some porosity at the nanoscale [16]. The increased porosity and grain boundaries in the ESLs are thought to increase the number of trap states where charge recombination can take place [21]. Nonetheless, due to expensive ALD equipment costs, sol-gel-based spray pyrolysis and spin coating are still the most popular methods for producing the compact metal oxide layers used in PSCs [22]. Moreover, some of the possible trap state defects can be alleviated by surface passivation strategies [18].

As the nature of the possible defect morphologies in ESLs can vary substantially (in the literature irregular and randomly occurring defect structures are often encountered), it is difficult to study their possible effects in the performance of PSCs systematically. In this study, we deliberately introduce well-defined morphological defects to model nanoscale pinholes evenly distributed throughout ~70 nm thick TiO2 films. The sol-gel-derived films are prepared using an inexpensive, well-controlled, and scalable dip-coating method [23] in combination with evaporation-induced self-assembly (EISA) of block co-polymers to generate an ordered pore structure in the final films with a pore size of ~10 nm in diameter [24]. Such porous films have previously been utilized in PSCs, but only in combination with an additional compact TiO2 layer underneath [25]. In our study, however, the porosity in the TiO2 films is tuned to simulate a large number of uniform pinholes in a single TiO2 layer. The TiO2 layers are characterized using grazing incidence X-ray diffraction (GI-XRD), X-ray reflectometry (XRR), field emission scanning electron microscopy (FE-SEM) and charge extraction (of injected carriers) by linearly increasing voltage in metal-insulator-semiconductor structures (MIS-CELIV) measurements, confirming that the pore system is achieved through the entire TiO2 layer and thus forms narrow pinholes down to the underlying FTO contact. Unexpectedly, devices based on TiO2 layers with high porosity still work very well. This indicates that narrow pinholes similar to the ones described by our model system do not have a significant negative impact on the device performance. This further implies that variations in the density of the TiO2 layer when using different deposition methods or deposition rates are less important as long as the ESL fully covers the FTO substrate.

#### **2. Materials and Methods**

To prepare the ordered mesoporous TiO2 films, a modified protocol of the one described by Ortel et al. [26] was used. However, instead of using triblock-based co-polymers, we used the commercially available poly(butadiene(1,4 addition)-b-ethylene oxide diblock co-polymer (P2952\_BdEO, MW = 13,000 g/mol, Polymer Source Inc., Dorval, QC, Canada) in our study. Two independent parameters were studied: (1) block co-polymer content (i.e., change in porosity), and (2) layer thickness (while keeping the porosity constant).

The TiO2 dip coating sols were prepared accordingly: Titanium (IV) chloride (TiCl4 > 99%, Fluka, Seelze, Germany) was initially diluted in ethanol (EtOH, >99.5%, ALTIA Plc, Helsinki, Finland) while stirring in an ice bath to give a 1:16 molar ratio TiCl4:EtOH stock solution. Mixtures of EtOH, Millipore water, and P2952\_BdEO were also prepared. Subsequently, the TiCl4 stock solution was added dropwise to the other solutions to produce dipping sols with the final molar ratios, as listed in Table 1.


**Table 1.** Molar compositions for the dip coating sols used in this study.

<sup>1</sup> The sample names in the porosity series (Ti-*x*) are derived according to the P2952\_BdEO/TiCl4 molar ratios, where *<sup>x</sup>* indicates the molar ratio × 106. In the sample names in the thickness series (Ti-21-*y*), the *<sup>y</sup>* parameter indicates the relative solvent amount (H2O + EtOH) in comparison to the original Ti-21 sample.

Fluorine-doped tin oxide substrates (FTO, TCO22-15, Solaronix, Aubonne, Switzerland) with dimensions 4 <sup>×</sup> 2 cm<sup>2</sup> were sonicated in water, acetone, and 2-propanol for 10 min each. The samples were subsequently dried in nitrogen flow and plasma-treated for 5 min before TiO2 film deposition. The same cleaning protocol was used when microscope glass substrates (VWR international, cut into 2.5 <sup>×</sup> 2 cm2 pieces) were used. The substrates were then dip-coated with the sols described in Table <sup>1</sup> to produce two different series (i.e., change in porosity and change in film thickness, respectively). The dip coating was performed using a withdrawal speed of 85 mm/min at a relative humidity below 20%. The substrates were then kept at the same relative humidity in the dipping chamber for at least 5–10 min before they were transferred to the oven for calcination. The films were initially kept at 80 ◦C for 4 h and then heated to 475 ◦C at a heating rate of 1 ◦C/min. The samples were held at 475 ◦C for 15 min and then allowed to naturally cool to 150 ◦C.

GI-XRD was performed on TiO2 films coated on top of FTO substrates using a Bruker AXS D8 Discover instrument. The measurements were performed between 24◦ and 40◦ using a step size of 0.04◦ and a grazing incidence angle of 0.3◦. The TOPAS P software (v. 4.2) was used to calculate the TiO2 crystallite size using the Scherrer equation [27]. The same instrument was used for XRR analysis of TiO2 films deposited on microscope glass slides instead of FTO substrates, as the high roughness of FTO would distort the interference patterns of the TiO2 layer [23]. 2θ/ω scans were performed with an increment of 0.002◦. The experimental data was fitted using the LEPTOS software (v.7.03). FE-SEM was used to determine the morphology of the TiO2 films with different porosity using a magnification of 100 kX, electron high tension of 2.70 kV, and an aperture size of 10 μm on a Zeiss Leo Gemini 1530 instrument. MIS-CELIV measurements using poly(3-hexylthiophene) (P3HT, Sigma-Aldrich, St. Louis, MO, USA) as charge injector layer were carried out to obtain information about the interconnectivity of the porous structure. The measurement setup and sample preparation are described in the Supplementary Information.

PSCs using the porous TiO2 ESLs were prepared accordingly: A 1.5 <sup>×</sup> 2 cm2 area from one side of the 4 <sup>×</sup> 2 cm2 FTO substrates was selectively etched using Zn powder (Sigma Aldrich) and 4 M HCl solution in water. The etched FTO substrates were subsequently sonicated in 2% aqueous solution of Hellmanex III detergent, water, acetone and 2-propanol for 10 min each at room temperature. The deposition of ordered porous TiO2 was performed by dip coating followed by calcination using the protocol described above. Kapton tape was used to mask the back side and the anode region of FTO substrate to avoid film deposition in those regions. Since there was a risk that the porous TiO2 films might absorb water from the moist environment, the samples were immediately transferred to a nitrogen glovebox once the samples had cooled down to 150 ◦C after calcination. The cesium-containing triple-cation mixed-halide perovskite light absorber layer was deposited by spin coating in combination with the anti-solvent method [3,28]. The mixed ion perovskite precursor solution contains a mixture of formamidium iodide (FAI, Greatcell Solar, Queanbeyan, Australia, 1 M), methylammonium bromide (MABr, Greatcell Solar, 0.2 M), lead iodide (PbI2, 99.99%, TCI Europe, Zwijndrecht, Belgium, 1.1 M), and lead bromide (PbBr2, 99.99%, TCI Europe, 0.2 M) dissolved in a 4:1 mixture of *N*,*N*-dimethylformamide (DMF, 99.8%, Sigma-Aldrich) and dimethyl sulfoxide (DMSO, 99.9%, Sigma-Aldrich). A 1.5 M solution of cesium iodide (CsI, 99.999%, ABCR, Karlsruhe, Germany) in DMSO was added to the above solution in a 1:19 volume ratio. This triple cation perovskite solution was spin-coated using a two-step program at 1000 and 6000 rpm for 10 and 20 s, respectively. During the second step, 200 μL of chlorobenzene (99.8%, TCI Europe) was pipetted onto the spinning substrate 5 s prior to the end of the program. The color of the films turned dark orange upon addition of chlorobenzene. Upon placing the samples on a hot plate at 100 ◦C, they turned dark brown within 10 s. Films were then annealed at the same temperature for 30 min. After annealing, the samples were allowed to cool down to room temperature. Subsequently the HSL solution was spin-coated on top of the perovskite layer at 4000 rpm for 30 s. The HSL solution composition was based on spiro-OMeTAD (Luminescence Technology Corporation, New Taipei, Taiwan) as the main component. Spiro-OMeTAD and 4-tert-butylpyridine (Sigma-Aldrich) were dissolved in chlorobenzene. The required amount of stock solutions of lithium bis(trifluromethylsulfonyl)imide (Li-TFSI, Sigma-Aldrich) and cobalt (III) tri[bis-(trifluromethane) sulfonamide] salt (FK209 Co (III), Greatcell Solar) in acetonitrile (99.8%, Sigma-Aldrich) were added into the first solution to obtain a molar composition of 1.0:0.5:2.5 <sup>×</sup> 10−2:3.3:131.5:7.2 (spiro-OMeTAD:Li-TFSI:FK209 Co (III):4-tert-butylpyridine:chlorobenzene:acetonitrile). Finally, an 80-nm-thick gold layer was deposited on top of the spiro-OMeTAD layer to form the back metal contact via evaporation at 2.5 <sup>×</sup> 10−<sup>5</sup> bar. An evaporation rate of 0.1 Å/s was used for the first 10 nm, after which it increased to 0.2 Å/s until 20 nm. The rate of 0.8 Å/s was then used to evaporate further up to 80 nm. The shape of the gold contacts was circular with a diameter of 0.6 cm, i.e., the contact area was 0.28 cm2.

The devices were characterized by measuring current density against voltage (J-V) scans using a 2636 Series Source Meter (Keithley Instruments, Cleveland, OH, USA) under simulated AM 1.5 sunlight close to 100 mW/cm<sup>2</sup> irradiance from an Oriel Class ABB solar simulator (150 W, 2" <sup>×</sup> 2"). The devices were masked using a black metal mask with an aperture size of 0.126 cm2. The J-V scans were performed between −0.3 V and 1.1 V at a scan speed of 10 mV/s, both in forward and reverse sweep. Time-dependent measurements of the devices were also performed for 5 min to determine the current densities close to their respective maximum power points (MPPs) under illumination at 1 Sun. Furthermore, for UV-Vis spectroscopy and photoluminescence (PL) measurements, TiO2 films deposited on glass substrates were coated with perovskite using the same deposition protocol as described above. The UV-Vis measurements were performed using a Perkin Elmer Lambda 900 UV-Vis/near infrared spectrometer. The samples were scanned in the λ range 500–900 nm and the measurements were performed in the presence of standard reflectance standards. The slit size was 2 mm. PL spectra were obtained with a FLS1000 spectrofluorometer (Edinburgh Instruments, Livingston, UK). Time-resolved photoluminescence (TR-PL) decays were determined using a time-correlated single photon counting (TCSPC) apparatus equipped with a Picoharp 300 controller and a PDL 800-B driver for excitation and a Hamamatsu R3809U-50 microchannel plate photomultiplier for detection in a 90◦ configuration. All samples were measured using a 648 nm excitation wavelength with an excitation energy intensity of 40 μJ/cm2 while exciting from the perovskite film side. The PL decays were monitored at 765 nm and well fitted with a bi-exponential function *<sup>I</sup>*(*t*) = *<sup>A</sup>*1·*e*(−*t*/τ1) + *<sup>A</sup>*2·*e*(−*t*/τ2), where *I*(*t*) is the PL intensity at time *t*, *A*<sup>x</sup> is the initial amplitude of component *x* (*x* = 1 or 2), and τ is the exponential lifetime of component *x* [29,30].

#### **3. Results and Discussion**

#### *3.1. Structural Properties of the TiO2 Thin Films*

The TiO2 thin films dip-coated on top of FTO substrates were characterized using GI-XRD to investigate the effect of block co-polymer content (i.e., film porosity) on the crystal structure. The diffractograms shown in Figure 1 reveal that all the films consist of the anatase crystal structure regardless of the amount of added block co-polymer. However, the declining peak intensity and Scherrer analysis of the full-width half maximum of the (101) reflection indicate that the anatase crystallite size decreases when higher block co-polymer amounts are used (see Table 2). The reflections at 34.0◦ and 38.1◦ 2θ originate from the underlying FTO substrate.

**Figure 1.** XRD diffractograms of TiO2 films with different block co-polymer content deposited on FTO substrates. The (101) reflection of the anatase phase is indicated in the figure. The asterisks (\*) indicate reflections from the underlying FTO substrate. The diffractograms have been normalized to the intensity of the FTO reflection at 38.1◦ 2θ as well as offset for clarity.

TiO2 films dip-coated on top of planar glass substrates were studied by X-ray reflectometry (XRR) to estimate the porosity and the film thickness. Figure S1 in the Supplementary Information shows the XRR interference patterns of TiO2 films with different porosities. The trends in the film density and thickness as a function of block co-polymer concentration are listed in Table 2. The density of the Ti-0 reference sample is lower than the literature value for a completely crystalline anatase material (3.79 g/cm3). However, it is expected that nanocrystalline thin films have lower densities. With an increase in block co-polymer concentration, the density of the films decreases, meaning that the porosity originating from the block co-polymer template increases (up to 47% for the Ti-21 sample). We aimed at keeping the film thickness constant at ~75 nm regardless of the block co-polymer content to be able to directly relate the device performance to the porosity of the films. However, initial tests simply by increasing the block co-polymer amount resulted in a rapidly increasing film thickness due to the increasing viscosity of the dip coating sol. To compensate for this, a higher dilution in water and EtOH was used for higher block co-polymer to TiCl4 ratios (see details in Table 1). Despite this adjustment, a slight variation in film thickness was observed for the samples, ranging from 75 nm for the Ti-0 sample to 50 nm for the Ti-21 sample. Thus, in order to investigate the influence of the film thickness, we made another series based on a fixed porosity. To produce the thinnest sample, 2.5 times more solvent (EtOH and water) compared to the Ti-21 sample was used in the dipping sol (sample Ti-21-2.5), while in order to produce the thickest sample, 80% of the original solvent amount was used (Ti-21-0.8). The XRR measurements shown in Figures S1 and S2 in the Supplementary Information reveal that the film thickness of the Ti-21-2.5 sample is ~20 nm, while the Ti-21-0.8 dipping sol produces a 75-nm-thick TiO2 film. However, both films have a density of ~1.7 g/cm3, which is the same as for the original 50 nm-thick film (Ti-21-1.0).


**Table 2.** Summary of the TiO2 film characteristics derived from XRD and XRR data.

<sup>1</sup> The porosity values for the block co-polymer-templated samples are calculated by relating their densities to the non-porous Ti-0 reference sample.

The films deposited on FTO substrates were further characterized using top-view SEM imaging, as shown in Figure 2. The Ti-0 sample shows a smooth granular surface with a grain size in the range of 10–20 nm, which is expected from the crystallite size obtained from XRD analysis. However, for the block co-polymer-templated films, pseudo-ordered mesoporous structures can be observed. The SEM images show a reduction in pore wall thickness from ~27 nm to ~15 nm with the increase in block co-polymer concentration, while the diameters of the pore openings remain roughly constant between 13 to 15 nm. At the bottom of some of the pores (highlighted in the dashed area in Figure 2c), spots with comparably darker contrast can be seen. These spots indicate the second row of pores deeper inside the TiO2 thin film. In the EISA process, the arrangement and interconnection of the spherical block co-polymer micellar templates determine the final pore structure [24]. In our case, it is expected that the mesopores have a body-centered-cubic (bcc) arrangement with pore shrinkage perpendicular to the substrate [26]. The ellipsoidal pores are connected via narrow channels, through which the combusted block co-polymer template escaped during the calcination process. This creates a tortuous pathway down to the underlying substrate. Based on the information obtained from XRR, XRD, and AFM, a schematic 2-D representation of the porous films can be constructed, as shown in Figure 3a.

**Figure 2.** Top-view SEM images of the (**a**) Ti-0, (**b**) Ti-6, (**c**) Ti-12, and (**d**) Ti-21 thin films made on FTO substrates. The dark spherical features indicate pore openings, while the brighter areas represent the surrounding TiO2 wall structure (see text for further details).

**Figure 3.** (**a**) A schematic 2-D representation of a porous TiO2 thin film deposited on top of a rough FTO substrate (note that the dimensions are not to scale); (**b**) suggested pore filling behavior of perovskite inside the porous TiO2 matrix: 1. Perovskite in contact with the perovskite capping layer, 2. direct perovskite pathway from the capping layer to FTO, and 3. isolated perovskite inside the porous matrix; (**c**) schematic illustration of the investigated device structures, where the TiO2 ESL is either dense or porous.

To verify whether the underlying FTO substrate is accessible via the pores from the top, the surface recombination velocity of holes at the FTO/TiO2 contact was determined in model devices where the semiconducting polymer P3HT was coated on top of the different TiO2 films and the gold contact was evaporated on top of the P3HT. Gold forms a hole-Ohmic contact to P3HT, whereas it is well known that TiO2 is hole blocking. It is expected that the P3HT can fill the porous structures more easily than the perovskite, and the more P3HT that is in direct contact with FTO the higher the surface recombination would be, since FTO does not block holes [31–33]. The surface recombination velocity at the TiO2/P3HT interfaces was determined using the MIS-CELIV technique [34]. The current transients and calculated surface recombination rates for holes, SR, for compact and porous TiO2 films are shown in Figures S3 and S4 in the Supplementary Information, respectively. For the Ti-0 sample, we obtained SR <sup>≈</sup> 10−<sup>5</sup> cm/s, which is slightly higher than previously reported values [34]. We also found that SR increases with increasing porosity. The Ti-6 sample has a SR value one order of magnitude larger than the Ti-0 sample, and when increasing the porosity to that of the Ti-21 sample, SR increases by another order of magnitude. We attribute this sharp increase to P3HT reaching all the way through the porous TiO2 films, down to the FTO, where holes will recombine much faster than at the TiO2/P3HT interface. The sharp increase in SR even at low porosity would correspond to P3HT pathways forming down to the FTO, and as the porosity increases, these pathways become more accessible and/or more numerous, seen as a further increase in SR. This leaves us with the conclusion that the porous channels in the

TiO2 films reaches all the way down to the FTO substrate. However, since these SR measurements are limited to hole-only devices and low-mobility materials, this method cannot be used to further clarify whether these porous channels are accessible to the perovskite layer. Earlier, it has been shown that the substrate is readily accessible through thinner block co-polymer-templated TiO2 porous films [35,36]. Thus, we believe that the ordered mesoporous TiO2 films can be used as a model system for TiO2 ESLs with narrow and well-defined pinholes.

#### *3.2. Device Performance*

In the next step, the non-porous reference sample (Ti-0) and the ordered mesoporous TiO2 films with different porosities were used as ESLs in PSCs. The overall device configuration is schematically illustrated in Figure 3c. In Figure 4a, representative J-V curves of devices made with TiO2 ESLs with different porosities show that there are no large deviations in device performance as a function of porosity. The corresponding dark curves are shown in Figure S5 in the Supplementary Information. Furthermore, the PCE values of the devices (measured in the reverse sweep) are plotted in Figure 4b and summarized in Table 3. The devices with a dense TiO2 layer (Ti-0) have a mean PCE of 13.1 ± 0.7%. The reason for the lower PCE compared to previously reported values for planar PSCs based on mixed perovskites [37] is mainly attributed to the relatively large active area (~0.13 cm2) and substrate size, which creates a large series resistance in the device. However, the reproducibility of the device performance is still very good, which is a prerequisite for this study.

Upon increasing the porosity in the TiO2 layer, the device efficiencies remain almost unchanged. The highest average efficiency (13.8 ± 0.7%) can actually be achieved for the samples with the most porous TiO2 layers (Ti-21). This is rather surprising, as the ordered pore structure percolates all the way through the TiO2 films and reaches down to the underlying FTO layer. If the perovskite were to be in direct contact with the FTO layer, one would expect shunt pathways and a considerable loss in device performance [9]. However, it seems like these narrow pore channels prevent the formation of detrimental perovskite pathways through the porous TiO2 to the FTO. After the high-temperature calcination of the TiO2 films, they were immediately transferred to the glove box in order to avoid contamination of the surface with volatile organics. Thus, the TiO2 pore surface is very energetic due to surface hydroxyl groups and should be readily wetted by polar solvents like the ones used in the perovskite precursor solution (DMF and DMSO). However, it is well known that the addition of the anti-solvent (chlorobenzene) brings the perovskite precursor solution into supersaturation after which a rapid perovskite crystallization commences. Our hypothesis is that the crystallization starts from the top of the porous TiO2 structure. The initially formed perovskite will then obstruct the pore entrances so that further precursor solution is not able to enter the pores upon solvent evaporation, but instead contributes to the perovskite capping layer. Nonetheless, the precursor solution that initially occupies the pores will be converted to perovskite upon annealing. However, as the resulting perovskite material is estimated to only occupy 15–20 vol.% in relation to the volume of the starting precursor solution (the solvents occupy the rest of the volume); this is scarcely enough to form percolating networks throughout the pore system. Instead, due to limited adhesion of the formed perovskite on the TiO2 surface, isolated islands of perovskite will be created inside the pore system. The suggested pore filling behavior is schematically illustrated in Figure 3b. Relating these results to the perovskite filling of nanoparticle-based mesoscopic TiO2 layers, the more accessible pores of those structures would allow for a considerably higher pore filling degree. In that case, an additional compact TiO2 layer is needed to avoid direct shunt pathways between the perovskite and FTO [23,38]. This further suggests that also the pore size could affect the pore filling degree as one would expect less pore entrance obstruction by the perovskite when the pore size is larger.

**Figure 4.** (**a**) Representative J-V curves in forward (dashed lines) and reverse (solid lines) sweep for devices with increasing porosity in the TiO2 layer. The solar cells were measured at a scan rate of 10 mV/s and AM 1.5 G illumination with light intensity of 100 mW/cm2; Changes in (**b**) device efficiencies, (**c**) *JSC*, (**d**) *VOC*, and (**e**) FF for devices measured in reverse sweep with increase in porosity in the TiO2 layer.

In the reverse sweep, it can be observed that the short circuit current (*JSC*) slightly increases with porosity (from 17.5 mA/cm2 to 18.5 mA/cm2 for the Ti-0 and Ti-21 samples, respectively). On the other hand, the open circuit voltage (*VOC*) marginally drops from 1.11 V to 1.09 V for the same set of samples. The opposing trends in *JSC* and *VOC* explain the rather constant PCE values regardless of porosity in

the TiO2 ESL layer. When comparing UV-Vis absorption data of TiO2 films coated with perovskite (see Figure S6 in the Supplementary Information), the absorption of the perovskite in the 500–750 nm wavelength range is virtually the same for all samples with a standard deviation of ~1% at λ = 700 nm. Due to the thinness of the TiO2 films, we expect optical effects like increased reflectance or optical interference to be small [39,40]. Thus, the UV-Vis results further support that the perovskite located in the pore systems is proportionally low. The slight increase in JSC with porosity could, however, be a result of more continuous perovskite pathways inside the pores at higher porosities, as illustrated in Figure 3b (case 1). Charges generated in isolated perovskite crystals inside the pores would normally recombine (case 3), but if more continuous pathways were to form, these charges could be extracted, and thus increase *JSC*. This could also explain the drop in *VOC*, as there will be a greater possibility that direct (shunt) pathways are formed between the perovskite capping layer and the FTO as the continuity in the perovskite pathways increase (case 2). As schematically illustrated in Figure 3b, we believe that such perovskite pathways can be formed in thinner regions of the TiO2 layer caused by the high surface roughness of FTO.

To evaluate the TiO2 porosity-influenced electron-injection process from the perovskite conduction band (CB) to the CB of the TiO2 layer, steady-state photoluminescence (PL) experiments on glass/perovskite and glass/TiO2/perovskite samples with different porosities of the TiO2 layer were conducted. Figure S7 shows a clear PL quenching effect for all perovskite coated TiO2 films. The calculated PL quenching efficiency (PLQE), or in other words, the electron-injection yield, is increasing with decreasing porosity, suggesting that a lower porosity of the TiO2 layer is more favorable for an efficient electron injection process. We now turn to assess the influence of the porosity of the TiO2 films on the charge transfer dynamics at the perovskite/TiO2 interface. Figure S8 shows TR-PL decays (obtained via TCSPC measurements) of the perovskite with and without coated TiO2 films with different porosities together with the extracted bi-exponential fitted data of the PL decays. All PL decays of the perovskite-coated TiO2 films show acceleration compared to that of the pristine perovskite on glass reference, suggesting that the interfacial electron injection has occurred for all cases of perovskite/TiO2 films, which is also consistent with previous PL quenching data. A clear deceleration of the decay profiles is observed with increasing porosity, although the decay lifetimes for the samples with the highest porosities (Ti-12 and Ti-21) are virtually identical. Based on the reported global analysis methods [41–43], we attribute the first component (A1, τ1) to the trap-state-mediated recombination, while assigning the second component (A2, τ2) to the nongeminate free carrier (electron and hole) recombination and electron injection process from the CB of the excited perovskite to that of the TiO2 film. It is evident that the second component dominates the overall decay process, suggesting that a low porosity is more favorable for suppressing charge recombination, which in turn results in the enhanced *VOC* [44]. Another possibility for the change in *VOC* is that the increased porosity can generate more traps on the TiO2 structure, leading to a deepening of the TiO2 CB, which could intrinsically lower the *VOC* [30,45].

**Table 3.** Mean values and standard deviations of photovoltaic parameters measured in reverse sweep of all type of devices.


Furthermore, the fill factors (FF) in reverse sweep (Figure 4e) show quite similar values for all porosities in the range of 0.67–0.70. However, when comparing the device hysteresis, larger differences are observed. As seen in Figure 4a, the device based on the dense reference TiO2 layer (Ti-0) displays an s-shaped feature in the forward sweep, which is not observed in the reverse sweep. This results in a high hysteresis index for the device (~12%). S-shapes and high hysteresis are commonly observed for planar TiO2-based devices [8,9]. The s-shape is indicative of an unstabilized power output and caused by polarization of the device (most likely due to diffusion of ionic species) [46]. Upon inducing pores in the ESL, the s-shape in the forward sweep disappears, which has also been observed when shifting from planar to mesoscopic TiO2-based devices [46]. The hysteresis between the forward and reverse sweeps remains high (in the range of 8–10%), but the FF in the forward sweep is significantly improved upon inducing porosity in TiO2 films. This is also evident from the time-dependent current density measurements performed close to the maximum power point (MPP) under illumination, shown in Figure S9 in the Supplementary Information. The device based on a dense TiO2 film requires more time to stabilize close to the MPP than the devices based on porous TiO2 films. With further increase in porosity, stabilization of the current at MPP is even faster, while the magnitude of current density at the MPP is also slightly enhanced. This correlates well with the J-V curves, as an improvement in JSC was observed in Figure 4a,c.

As mentioned earlier, the thickness of the TiO2 films decreases slightly when more block co-polymer amounts were used, i.e., from 75 nm for the Ti-0 sample to 50 nm for the Ti-21 sample. We are aware that the device performance could be affected when the thickness of the TiO2 layer is altered, due to small changes in the charge transport properties or optical effects. Nonetheless, we consider the samples in porosity series to be "thick enough", as they are all equal to or thicker than 50 nm. This is important, as it rules out the possibility that there are bare patches of FTO exposed in our devices [23]. To verify that the observed device change is an effect of the porosity rather than a change in thickness, we also made a series where the thickness of the TiO2 ESL was varied, while keeping the block co-polymer to TiCl4 ratio the same as in the Ti-21 sample. When comparing the J-V characteristics and the efficiencies in the box chart diagram in Figure 5, all thicknesses of the porous ESL display good average device performances. The devices prepared using the 75 nm-thick porous TiO2 films (Ti-21-0.8) perform slightly worse (PCE = 12.9%) than those prepared using the 50 nm porous TiO2 ESL (PCE = 14.1%). This suggests that small thickness variations can indeed also be important for optimized performance for this device structure; however, the reduced VOC trend can still be observed when comparing this Ti-21 batch to the devices with less porous TiO2 layers in Table 3. For instance, the non-porous Ti-0 sample and the highly porous Ti-21-0.8 sample have the same film thickness (75 nm) and devices made from these samples possess roughly the same efficiency (~13%). However, when comparing the VOC values for these two samples, it is clear that the VOC is substantially lower for devices based on the porous TiO2 layer (1.05 V compared to 1.11 V). This is most likely attributed to increased charge recombination as suggested by the TR-PL data. It is noteworthy that even the devices based on a ~20 nm-thick porous TiO2 layer (Ti-21-2.5) display a decent average device efficiency (PCE = 11.0%). However, as seen from the scattered PCE data points in the box diagram as well as the larger standard deviations in Table 3, the thinnest sample clearly suffers from poor reproducibility (see also Figure S10 for the statistics of the photovoltaic parameters). As the porosity is roughly the same for the different thicknesses, a similar interconnectivity of the perovskite inside the pores is to be expected. Thus, the main difference between the 75 nm (Ti-21-0.8) and 50 nm (Ti-21-1.0) samples is more likely that the thicker ESL is not able to extract electrons as well as the thinner layer, which is seen as a slightly lower PCE [9]. Furthermore, an ESL layer that is too thick can also reduce the light transmittance due to stronger light scattering and greater absorption of photons with energies higher than that of the ESL band gap, and this reduces the photon absorption by the active layer [47,48]. However, when the porous layer becomes thin enough, as in the Ti-21-2.5 sample, more direct shunt pathways are to be expected as the pore channels become shallower perpendicular to the FTO substrate. The large scattering of data points suggests that in some TiO2 layers, very few such pathways can be achieved, while others suffer severely from direct shunt pathways. In ultrathin TiO2 layers, shunt pathways can also arise from direct contacts between the perovskite and bare patches of FTO due to incomplete surface coverage of

the ESL. We previously reported that compact TiO2 layers prepared by the dip-coating method need to have a thickness of at least ~30 nm to work optimally in mesoscopic PSCs [23]. However, in the planar configuration, the TiO2 layer probably needs to be somewhat thicker (~50 nm) to avoid this kind of pinholes in the devices.

**Figure 5.** (**a**) J-V curves for representative devices based on TiO2 ESLs with the highest porosity with different thicknesses. Dashed lines indicate forward sweep and solid lines reverse sweep; (**b**) box chart for the efficiencies in reverse sweep.

#### **4. Conclusions**

We introduced a model system for pinholes in TiO2-based ESL layers using block co-polymer-templated thin films with well-defined pore structures. We believe that the investigated system predicts morphological defects, such as narrow pinholes, well. We observed that such pinholes have a very small effect on the overall device performance most likely due to the fact that very few direct pathways of perovskite reach the FTO substrate through the ordered pore system. We saw a slight improvement in *JSC* for layers at large porosities as well as a drop in *VOC* with increasing porosity. A more interconnected pathway of perovskite forming in the pores when the porosity is high can explain both effects. Larger effects on the device performance are expected if the pore size were to be larger or if the pore structure were to be more open, and we plan to investigate these parameters in a follow-up study. Furthermore, for ultrathin (~20 nm) porous layers, additional direct shunt pathways due to an incomplete ESL surface coverage of the FTO layer further deteriorates the device performance.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-4991/10/1/181/s1, Figure S1: XRR interference patterns of various TiO2 thin films, Figure S2: Film thickness and density dependence when varying the solvent amount, Figure S3: MIS-CELIV data, Figure S4: Calculated surface recombination velocities, Figure S5: J-V scans under dark conditions, Figure S6: UV-vis absorption measurements, Figure S7: Steady-state PL measurements, Figure S8: TR-PL measurements, Figure S9: Time-dependent current density measurements, Figure S10: Box charts for photovoltaic parameters of devices where the ESL thickness has been varied.

**Author Contributions:** Conceptualization, M.T.M., M.H., C.W., M.N., C.A., S.D., R.Ö. and J.-H.S.; methodology, S.Q., M.N., C.A. and S.D.; validation, M.H., C.W. and S.D.; formal analysis, M.N., C.A., S.D., M.L. and P.V.; investigation, M.T.M., S.Q. and M.L.; resources, P.V.; writing—original draft preparation, M.T.M.; writing—review and editing, all authors; visualization, M.T.M. and J.-H.S.; supervision, M.H., M.N., P.V., R.Ö. and J.-H.S.; project administration, P.V., R.Ö. and J.H.S.; funding acquisition, P.V., R.Ö. and J.-H.S. All authors have read and agree to the published version of the manuscript.

**Funding:** This research was funded by Academy of Finland, grant numbers 308307 and 326000 as well as Jane & Aatos Erkko Foundation (project "ASPIRE"). This work is also part of the Academy of Finland Flagship Programme, Photonics Research and Innovation (PREIN) (decision No 320165). C.W. thanks the Deutsche Forschungsgemeinschaft (DFG, WE 6127/1-1) for a postdoctoral fellowship.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Growth and Functionalization of Particle-Based Mesoporous Silica Films and Their Usage in Catalysis**

**Pei-Hsuan Wu 1, Peter Mäkie 1, Magnus Odén <sup>1</sup> and Emma M. Björk 1,2,\***


Received: 1 March 2019; Accepted: 2 April 2019; Published: 6 April 2019

**Abstract:** We report the formation of mesoporous films consisting of SBA-15 particles grown directly onto substrates and their usage as catalysts in esterification of acetic acid and ethanol. The film thickness was altered between 80 nm and 750 nm by adding NH4F to the synthesis solution. The salt also affects the formation rate of the particles, and substrates must be added during the formation of the siliceous network in the solution. Various substrate functionalizations were tested and hydrophobic substrates are required for a successful film growth. We show that large surfaces (> 75 cm2), as well as 3D substrates, can be homogenously coated. Further, the films were functionalized, either with acetic acid through co-condensation, or by coating the films with a thin carbon layer through exposure to furfuryl alcohol fumes followed by carbonization and sulfonation with H2SO4. The carbon-coated film was shown to be an efficient catalyst in the esterification reaction with acetic acid and ethanol. Due to the short, accessible mesopores, chemical variability, and possibility to homogenously cover large, rough surfaces. the films have a large potential for usage in various applications such as catalysis, sensing, and drug delivery.

**Keywords:** mesoporous silica; mesoporous films; direct growth; esterification; material formation

#### **1. Introduction**

Mesoporous silica films are of large interest in applications such as sensing, catalysis, and drug delivery [1–6]. Their large surface area, tunable pore characteristics, and versatile surface functionality are attractive features. For some applications, ordered cylindrical pores are preferable over spherical [7] or wormlike [8] pores. SBA-15 is a type of mesoporous silica with hexagonally ordered, cylindrical pores [9]. By alterations in the synthesis conditions, e.g. addition of swelling agents or altered reaction temperature, the pores can be increased from the regular ~8 nm to > 18 nm [10,11], and the particle morphology changed from fibers to rods or platelets [11,12]. This makes the material attractive in, e.g., catalysis [13,14], enzyme immobilization [15], and sensing [16]. However, the synthesis of mesoporous films using SBA-15 structures most often results in long pores that are aligned parallel to the substrate [17], making them inaccessible.

The most common method for synthesizing SBA-15 films is evaporation-induced self-assembly (EISA) [18], where an ethanol-containing solution is deposited onto a substrate using spin- or dip-coating. The method can be used on large substrates with various compositions, but is limited, as the dipping angle is crucial for the film thickness [19], resulting in non-homogenous coatings on non-flat substrates, and pores larger than 10 nm are rare [20]. Recently, methods have been developed for depositing thin layers of pre-synthesized mesoporous silica particles on substrates, e.g., by spinning them onto a substrate [21], or using Langmuir-Blodgett deposition [22], making also cylindrical pores easily accessible. However, these films consist of separated particles located on the substrate, which

results in poor mechanical integrity. In addition, the method requires flat substrates, which makes it less attractive for many applications. It is possible to bind the particles to a substrate through covalent linking [5], but this is a time-consuming procedure, with functionalization of both particles and substrates. A different direct growth (DiG) method, to form a monolayer of SBA-15 particles adhered to a substrate using a simple one-pot synthesis method, was recently reported [23]. This method yields densely packed, separate particles with short, accessible pores for gases and liquids, despite having the pores running parallel to the substrate surface.

In the present study, we show that mesoporous films with controlled thickness can be synthesized by growing monodispersed SBA-15 particles onto substrates. The effect of substrate functionalization, as well as the formation of the films, are investigated, and we show that it is possible to form mechanically stable films on large and rough surfaces. Functionalization of the films using co-condensation and the possibility to coat the films with a carbon layer through exposure of furfuryl alcohol vapor, forming CMK-5 [24], are investigated to study the variability of the films. To explore the accessibility of the pores, the film is tested as a catalyst for esterification of acetic acid and ethanol. The DiG-technique can be suitable for synthesizing mesoporous coatings for, e.g., catalysts, sensors, or medical implants.

#### **2. Materials and Methods**

#### *2.1. Reagents*

Hydrochloric acid (≥37%, puriss. p.a., Fluka, ACS Reagent, fuming), triblock copolymer EO20PO70EO20 (P123) (Mn ~5800, Aldrich, ammonium fluoride (≥ 98.0%, puriss. p.a., ACS reagent, Fluka), tetraethyl orthosilicate (TEOS)(reagent grade, 98%, Aldrich), heptane (99%, ReagentPlus®, Sigma-Aldrich), (3-Mercaptopropyl)trimethoxysilane (MPTMS) (Aldrich), hydrogen peroxide (≥35% at RT, purum p.a., Sigma-Aldrich), octadecyltrichlorosilane (OTS) (≥ 90%, Aldrich), chlorotrimethylsilane (TMCS) (≥ 99%, Aldrich), (3-Aminopropyl)trimethoxysilane (APTMS) (97%, Aldrich), toluene (≥ 99.5%, Sigma-Aldrich), nitric acid (≥ 64–66%, Sigma-Aldrich), sulphuric acid (95.0–98.0%, ACS Reagent, Sigma-Aldrich), glycerol (≥ 99.0%, Sigma-Aldrich), acetone (≥99.9%, Sigma-Aldrich), 1,6-diisocynanatohexane (98%, Aldrich), ethanol (95%, Kemetyl), ethanol (99.5 %, Solveco), furfuryl alcohol (98%, Aldrich), methanol (99.8%, Sigma-Aldrich), and benzene (purity ≥ 99.7%, Sigma-Aldrich)) were purchased from Sigma-Aldrich, Stockholm, Sweden, and used as received.

#### *2.2. Syntheses*

#### 2.2.1. Functionalization of Substrates

OTS-functionalized Si wafers were used as substrates for the DiG films [23]. Prior to functionalization, organic contaminants on the substrates were removed from the surface with standard Radio Corporation of America (RCA) cleaning fluid (H2O:H2O2:NH3 in a volume ratio of 5:1:1 at 85 ◦C for 10 min), washed with water, and then treated with nitric acid at room temperature for 5 min. The substrates were then washed with large amounts of water and dried with compressed N2 gas. The OTS was grafted onto the substrates by immersing them in a heptane solution with 1 mM OTS at 18 ◦C for 15 min. After the OTS treatment, the substrates were rinsed with heptane, dried at 200 ◦C for 2 h, and stored in heptane until usage.

Details of the additional functionalization methods used are presented in the Supplementary Materials.

#### 2.2.2. Film Synthesis

The films were grown following a modified protocol from Björk et al. [23]. In a typical synthesis, 2.4 g of P123 and a specific amount (0–28 mg) of NH4F were dissolved in 80 mL of 1.84 M HCl solution at 20 ◦C. The mixture was stirred in a round bottom flask until the reagents were dissolved. 5.5 mL of TEOS and 1 mL heptane were premixed and added to the micellar solution under stirring for 4 min, followed by static conditions overnight. During this static time, substrates were immersed in the

solution at specific time intervals, depending on the amount of NH4F used. The substrates were transferred to a sealed polytetrafluorethylene (PTFE) flask, while still being submersed in the solution, and placed in an oven at 100 ◦C for 24 h for hydrothermal treatment. The products were collected by filtration, rinsed with deionized water, and dried at ambient temperature overnight. For template removal, both the films and powder were calcined in air at 550 ◦C for 5 h with a 2 ◦C/min ramp rate. The materials are labelled SBA-15\_*X*, where *X* is the NH4F/P123 molar ratio.

#### 2.2.3. Direct Sulfonation (SBA-15-DS)

Films with sulfonic acid groups were synthesized by adding MPTMS and H2O2 during the particle formation, following the protocol by Margolese et al. [25]. A double batch of SBA-15\_0.0 was synthesized, and when the stirring was turned off, the solution was divided into six beakers. In five beakers, 0.17 mL of MPTMS and 0.27 mL of H2O2 were added to the synthesis solution at different times (1–20 h) after the TEOS addition. These materials are labelled SBA-15\_0.0\_*Y*DS, where *Y* corresponds to the hours between the addition of TEOS and MPTMS + H2O2. For these materials, the P123 was removed by ethanol extraction, as described in the Supplementary Materials.

#### 2.2.4. Carbon Infiltration and Sulfonation (SBA-15-CMK-5)

To get an even distribution of carbon species in the films, the carbon infiltration was performed by exposing the silica films to furfuryl alcohol vapor. Calcined SBA-15\_0.0 was exposed to alcohol vapor from furfuryl alcohol (5 mL) in a closed atmosphere at 40 ◦C overnight. The composites were then kept in a furnace at 100 ◦C overnight to ensure an even distribution of furfuryl alcohol into the mesopores, and to polymerize the alcohol. The material was then transferred to a nitrogen purged tube furnace at 800 ◦C for 1 h. Pyrolysis of the polymer during the heat treatment resulted in a thin carbon layer on the mesoporous silica walls.

The carbon infiltrated SBA-15 was sulfonated according to previous reports [13,26]. Briefly, 0.5 g of SBA-15-CMK-5 was mixed with 25 mL of H2SO4 and heated to 80 ◦C under reflux overnight. After the reaction, the solution was cooled down to room temperature, rinsed with large amounts of deionized water, and collected by filtration. The SBA-15-CMK-5-SO3H was finally dried at 80 ◦C overnight.

#### *2.3. Characterization*

The material morphology was observed by scanning electron microscopy (SEM), using a Leo 1550 Gemini Scanning Electron Microscope (Zeiss) operated at 3 kV and a working distance of 3–5 mm. The pore characteristics were determined for all powders using N2 sorption with an ASAP2020 (Micromeritics) at −196 ◦C. The specific surface area was determined with the BET method at *P*/*P*<sup>0</sup> = 0.8–0.18, and the total pore volume was calculated at *P*/*P*<sup>0</sup> = 0.98. The pore size was calculated using the KJS method at the adsorption isotherms. Small angle x-ray diffraction (SAXRD) was used to identify the pore order. Diffractograms were recorded with an Empyrean diffractometer from, in transmission mode using Cu Kα radiation (Malvern Panalytical). The pore structure was further visualized by transmission electron microscopy (TEM) performed with a Tecnai G2 TF 20 UT microscope operated at 200 kV (FEI). TEM samples were prepared by dispersing the product in acetone and depositing it on hollow carbon grids.

The contact angle of the functionalized substrates was determined by contact angle measurements using a CAM 200 Optical Contact Angle Meter (KSV Instruments). The measurements were performed using a 2 μL droplet of distilled water, which was placed in the middle of the substrate. Three independent measurements were conducted for each substrate.

Determination of sulfonic acid groups was performed using acid-base titration. For the silica-based samples, 0.10 g of the material was mixed with deionized water, followed by direct acid-base titration with a 0.005 M NaOH solution. For the SBA-15-CMK-5-SO3H, the catalyst was mixed with 30 mL 0.1 M Na2SO4 to react with the sulfonic groups, forming bisulfate, 4 h prior to the acid-base titration. The bisulfate reacted with the NaOH, giving rise to the total amount of sulfonic acid groups. Other acid groups (carboxylic, phenolic, lactonic) are less acidic than sulphate, and do not create bisulfate upon exposure to sulphate ions [27]. The number of acidic groups was calculated as

$$n\_{\rm ac} = V\_{\rm NaOH} \times \text{[NaOH]} \tag{1}$$

#### *2.4. Esterification Reaction*

The catalytic activity of an SBA-15-CMK-5-SO3H DiG film is shown in an esterification reaction with acetic acid and ethanol. In the reaction, 25 mL of ethanol and 10 mL of acetic acid was mixed and heated to 80 ◦C under stirring in a closed beaker. A 4-inch silicon wafer coated with SBA-15\_CMK-5\_SO3H cut in pieces was added to the mixture. 1 mL aliquots were removed from the solution at specific time points and mixed with deionized water to terminate the reaction. The conversion of acetic acid was determined by titration witha1M NaOH solution, and the equivalence point was found using a pH electrode. The conversion was calculated using

$$X\left(\%\right) = \left(mol\text{HAcritical} - mol\text{HAcend}\right) / mol\text{HAcinritical} \times 100\tag{2}$$

As a reference, the reaction was also performed without the presence of a catalyst.

#### **3. Results**

#### *3.1. Film Growth*

Films were synthesized with various NH4F concentrations in the solution. The particle sizes, both on the substrates and in the solution, as well as film thickness, were affected by the salt concentration, as seen in Figure 1. It was clear that densely packed films could be grown independent of the NH4F concentration. The particle size, both on the substrate and in powder form, was affected by the salt concentration. The particles became narrower, from platelets (Figure 1e,f) to rods (Figure 1g,h), with an increasing NH4F to P123 molar ratio. The particle narrowing was consistent with the results from Björk et al. [12], and was a result of the decreased solubility of the polyethylene oxide (PEO) chains of P123 when the NH4F concentration increases. The TEM micrographs show cylindrical pores, ordered in a hexagonal structure, for all NH4F concentration. The pore orientation supported that side-by-side attachment of the micelles causes the particle broadening.

**Figure 1.** SEM micrographs of films (first row) and particles (second row), and TEM micrographs of particles (third row) synthesized with a NH4F to P123 molar ratio of (**a**,**e**,**i**) 0.0, (**b**,**f**,**j**) 0.4, (**c**,**g**,**k**) 0.9, and (**d**,**h**,**l**) 1.8.

The relation between the width of the film particles and the film thickness is presented in Figure 2. It is apparent that the film thickness and particle width follow the same trend, even though the film thickness is a factor of 2–3 times smaller than the particle width. Nitrogen sorption isotherms, SAXRD diffractograms, and the corresponding physicochemical properties of the SBA-15 powders from the film syntheses are available in the Supplementary Materials (Figure S1 and Table S1). These results show that the pores were cylindrical and ~10 nm in diameter. The x-ray diffractograms show three well-resolved peaks for all materials, confirming the hexagonal ordering of the pores.

**Figure 2.** The relation between film thickness and particle width for different direct growth (DiG) films synthesized with various NH4F to P123 molar ratios.

To study the mechanism for a successful film growth, substrates were added at different times, depending on the formation rate. It is well known that the formation rate of SBA-15 is affected by addition of NH4F, where higher concentrations of salt give a faster formation rate [12,28]. The substrate addition times yielding the desired film morphology and its correlation to the formation stages of the material are presented in Figure 3. The films were evaluated by SEM, and successful film growth is here defined as a homogenous layer of densely packed particles on the substrate, see the first row in Figure 1.

**Figure 3.** The window for substrate addition for a dense film growth and its correlation for material formation stages (Adapted from [28]).

#### *3.2. Surfactant Removal*

The choice of micelle removal technique can be used to tailor the material characteristics, e.g., pore size, silanol group concentration, or survival of co-condensed functional groups. Figure 4. shows

SBA-15\_0.4 films where the surfactant was removed with various methods: calcination, ethanol extraction [9], H2O2 oxidation [11], and methanol sonication [29]. The removal techniques are presented in the Supplementary Materials. It is clear from Figure 4a,b,d that surfactant removal by calcination, ethanol extraction, and methanol sonication did not affect the film morphology. However, H2O2 oxidation (Figure 4c) removed the particles from the substrate, resulting in a nearly naked substrate surface.

**Figure 4.** SEM micrographs of DiG\_0.4 films where the surfactant was removed by (**a**) calcination, (**b**) ethanol extraction, (**c**) H2O2 oxidation, and (**d**) methanol sonication.

#### *3.3. Substrate Effects*

Several types of substrate functionalizations were used here to study the requirements for film growth. The methods for functionalization of silicon wafers with P123, sulfonic acid, thiol groups, amino groups, octadecyl groups, and methyl groups are presented in the Supplementary Materials. The contact angles for the films and SEM micrographs of the grown films are presented in Figure 5. No contact angle value is presented for the clean wafer with silanol groups, since it was so hydrophilic that no angle could be determined. As can be seen, dense DiG films (SBA-15\_0.0) could only be grown on substrates functionalized with octadecyl or methyl groups (Figure 5f,g), which are the most hydrophobic substrates. The silanol and P123 functionalized substrates also held a number of small SBA-15 particles (Figure 5a,b), while the substrate with sulfonic acid, thiol, and amino groups mainly consisted of a tissue phase and some particles (Figure 5c–e).

**Figure 5.** SEM micrographs of DiG\_0.0 films grown onto substrates functionalized with (**a**) silanol groups (**b**) P123, (**c**) sulfonic acid, (**d**) thiol groups, (**e**) amino groups, (**f**) octadecyl groups, (**g**) methyl groups, and (**h**) the contact angle for the corresponding substrates.

Figure 6 shows that it was possible to grow films with the DiG method on rough and large area substrates. An SBA-15\_0.0 film was grown on a silicon wafer that was blasted with alumina sand, creating a rough surface, prior to the OTS functionalization. Figure 6a shows a dense film coverage of the rough substrate surface, where particles are grown on all surfaces, independent of incline. Also, a full 4-inch silicon wafer was coated with SBA-15\_0.0, resulting in a homogenous film across the substrate (Figure 6b). The substrate addition time for both syntheses was 16 min after TEOS addition.

**Figure 6.** (**a**) SEM micrograph of a SBA-15\_0.0 DiG film grown on a blasted substrate, and (**b**) a photograph of (top) a SBA-15\_0.0coated and (bottom) clean 4 inch silicon wafer.

#### *3.4. Functionalization*

SBA-15\_0.0 films were functionalized with sulfonic acid using MPTMS and H2O2 during a co-condensation process during the material formation. The SBA-15\_0.0 synthesis was chosen since it has the lowest formation rate, and a mixture of MPTMS and H2O2 was added to the synthesis mixture after 1, 2, 4, and 20 h into the reaction. The morphology of the functionalized particles and films are shown in Figure 7 and it can be observed that the addition time affected the material characteristics. When the MPTMS/H2O2 was added 1 h into the synthesis, the particles were small and aggregated resulting in inhomogeneous coverage of the substrate surface. The film consists of a tissue phase with sparsely attached particles (Figure 7b). When the reagents were added after 2 h, the platelet morphology had started to form (Figure 7c), but the particles were narrower compared to the original SBA-15\_0.0, and in addition, small spherical features coexisted. The corresponding films consisted of particles, but it was apparent that these were smaller and not as developed as the unfunctionalized films in Figure 1a. When the MPTMS/H2O2 was added after 4 h, both the particle and film morphologies (Figure 7e,f) resembled the unfunctionalized materials. Finally, when the functionality was added after 20 h, i.e., directly prior to the hydrothermal treatment, the particle morphology was unaffected, but the appearance of the films was fuzzy, as if the particles had been covered with an additional layer (Figure 7g,h).

The acidity and physicochemical properties of the materials are presented in Table 1. Nitrogen sorption isotherms and small angle x-ray diffractograms are provided in the Supplementary Materials (Figure S2). The small angle x-ray diffractograms show three peaks, confirming a hexagonal order of the pores, also for the sulfonated materials. The less intense 110 and 200 peaks of SBA-15\_1DS indicate a lower degree of order compared to when functionalization was performed later in the synthesis. The highest numbers of acidic groups were found in the materials functionalized after 1 and 4 h after the TEOS addition with 0.024 mmol/g and 0.020 mmol/g, respectively, compared to 0.003 mmol/g without functionalization. The addition of the reagents after 2 h and prior to the hydrothermal treatment had a negligible effect on the acidity of the materials compared to unfunctionalized SBA-15\_0.0. It should be noted that the acidity was only measured on the powder materials, due to the small material amount on the substrates.

**Figure 7.** SEM micrographs of SBA-15\_0.0 films (first row) and particles (second row) functionalized by co-condensation (**a**,**e**) 1 h, (**b**,**f**) 2 h, (**c**,**g**) 4 h, and (**d**,**h**) 20 h after the addition of TEOS.


<sup>a</sup> Reference SBA-15\_0.0 from the same batch as the directly sulfonated materials. <sup>b</sup> SBA-15\_0.0 template for the carbon infiltration.

To study the versatility of the DiG films, an SBA-15\_0.0 film was functionalized with a thin carbon layer on the pore walls, similar to CMK-5 [24]. To form the carbon layer, the films were exposed to furfuryl alcohol vapor, instead of the commonly used induced incipient wetness impregnation, since the incipient wetness technique yielded carbon aggregates on the film surface (data not shown). As can be seen in Figure 8a, the film morphology was kept during the carbonization process. However, the unit cell was shrinking during the carbonization, most probably due to densification at 800 ◦C. Nitrogen adsorption (Figure 8b) indicated that the carbon infiltration resulted in a ~2 Å thick coating, seen as a reduced surface area and pore size compared to the parent SBA-15\_0.0, see Table 1. The silica/carbon film was further functionalized by exposure to H2SO4, which resulted in a 0.191 mmol/g of sulfonic acid sites.

**Figure 8.** SEM micrograph of (**a**) SBA-15\_0.0\_CMK-5, and (**b**) pore size distributions and physisorption isotherms of SBA-15\_0.0, SBA-15\_0.0\_CMK-5, and SBA-15\_0.0\_CMK-5\_SO3H.

#### *3.5. Catalytic Performance*

To confirm the success of carbon infiltration and functionalization, a 100 mm silicon wafer coated with an SBA-15\_CMK-5\_SO3H DiG film was prepared. This model system was then tested as a catalyst for esterification of acetic acid and ethanol. The conversion of acetic acid with and without a catalyst is presented in Figure 9. When the film was used as a catalyst, nearly 30 % of the acetic acid was converted within one hour, which was a significant increase compared to the ~5 % conversion when no catalyst was used.

**Figure 9.** Conversion of acetic acid in the esterification reaction with ethanol at 80 ◦C, with and without a DiG film catalyst.

The catalytic reaction was repeated with a new solution after cleaning the catalyst with water and ethanol. An acetic acid conversion of 11.5 % was obtained in the second cycle. The morphologies of the used films are shown in Figure 10. A majority of the film was intact, except for some areas where particles are removed.

**Figure 10.** SEM micrographs of SBA-15\_0.0\_CMK-5\_SO3H after (**a**) one cycle and (**b**) two cycles in the esterification reaction.

#### **4. Discussion**

#### *4.1. Film Formation*

This work shows that it is possible to synthesize DiG films with various thicknesses, which is of interest from an application view point, especially since the films are formed as separate particles with easily accessible pores, also close to the substrate. It has previously been shown that NH4F can be used to control the morphology of SBA-15 particles [10,12], and it is apparent from Figure 2 that the film thickness is correlated to the particle width when the salt concentration is altered. This shows that the films are affected by the solution conditions in the same ways as the particles, indicating a similar formation mechanism.

The data shows that film growth occurs when the substrates are added to the synthesis solution during the formation of the siliceous network (Figure 3). At this stage, the micelles are still spherical and the hexagonal ordering has not yet started [28]. This indicates that the films are formed by the condensation of silica species directly on the substrate, and not formed by the deposition of pre-synthesized particles. This is further corroborated by the cross-section images of the films in Figure 1.

It is clear from Figure 5 that proper substrate functionalization is needed to achieve the desired DiG film growth. The densest films are synthesized on hydrophobic substrates, functionalized with octadecyl or methyl groups, while the more hydrophilic substrates show no or poor films consisting of sparsely spaced particles in a tissue phase (Figure 5c–e). Mesoscopic simulations have shown that P123 can form hemispherical structures on hydrophobic surfaces [30]. These structures, where the hydrophobic core is in contact with the substrate, and the PEO brushes are directed towards the aqueous solution, can act as nucleation sites for growth of the film, resulting in a dense packing of particles (Figure 5f,g). Liu et al. showed that non-ionic triblock copolymers (P105) can coat hydrophobic surfaces, while micellar structures can adsorb on a hydrophilic surface [31], which is in good agreement with the simulations. The particle growth on the hydrophilic substrates with silanol groups or P123 (Figure 5a,b) can be the result of adsorbed micelles bound to the substrate through silica polymerization.

The stability of the films is shown in the surfactant removal section and also through the catalytic reaction. It is apparent that the films can sustain calcination at 550 ◦C, ethanol extraction for 24 h at 78 ◦C, or ultrasound sonication in methanol at least 5 min, confirming that the particles adhere well to the substrates. During the catalytic reaction, the films were submerged in a stirred, aqueous solution for 1 h per cycle. During the first cycle, some particles detaches from the substrate, although the vast majority of the film remains intact. The loss of particles suggests variation in adhesion among the particles, perhaps due to contaminants present at the particle/substrate interface causing incomplete attachment of the half hexagon prism to the substrate. The possibility to use various surfactant removal techniques enables functionalization through co-condensation, as the functional groups can be removed upon calcination. It also allows post-functionalization since the ethanol extraction yields more silanol groups after micelle removal [32]. H2O2 oxidation has been shown to be an efficient method for removing P123 and other organic groups from SBA-15 [11,33]. It also removes the DiG film from the substrate. The reason for this can be decomposition of the functional organic layer between the film particles and the substrate.

#### *4.2. Functionalization*

During the direct sulfonation of SBA-15, it is apparent that the addition time of the MPTMS + H2O2 strongly affects both the morphology and acidity of the final material, as seen in Figure 7 and Table 1. For SBA-15\_0.0, the optimum addition time is 4 h, which yields a material with relatively high acidity and the desired morphology of both films and particles. Adding the reagents as early as 1 h into the synthesis strongly affects the particle formation, and small aggregated particles are formed, while an addition time of 2 h yields a narrower platelet morphology compared to the non-functionalized material. This is most likely due to the fact that the particles at these time points are not completely formed [28]. During the co-condensation of MPTMS, MPTMS will condense on the silanol groups in the siliceous network [34] and hinder further silica condensation at these sites. One hour into the reaction, the siliceous network is forming, but the hexagonal ordering of the micelles has not started. At this time point, particle formation is governed by the addition of MPTMS silica species. When the MPTMS + H2O2 is added after 2 h, the hexagonal framework has formed, but the particles are still growing through side-by-side attachment of the silicated micelles. The MPTMS attaches to the surface silanols of the silica, and therefore locks the surface from further condensation, resulting in narrower platelets. When the MPTMS + H2O2 is added after 4 h or directly prior to the hydrothermal treatment, the particle morphology is already set, and no alteration is visible in the SEM micrographs

(Figure 7e,g). However, the acidity of these samples is different, which is likely related to the silica condensation also progressing after 4 h [28]. Such condensation enables larger amounts of hydrolysed MPTMS to bind to the silica particles, compared to when the condensation of particles is completed. This is in good agreement to the results by Nassor et al. [35], who showed that it is easy to wash away the MPTMS when it is added at a late stage of the synthesis.

The difference in physicochemical properties of SBA\_0.0\_no DS and SBA-15\_0.0\_carbon templates is due to the different methods for removal of the micelles, which were ethanol extraction and calcination, respectively. No external carbon features were detected when the furfuryl alcohol vapor method was used to coat the SBA-15 (Figure 8a). This method yielded a 2 Å thick carbon layer in the mesopores after carbonization (Table 1), without generating additional plugs in the pores (Figure 8b). Hence, the method of using furfuryl alcohol vapor to form CMK-5 structures in mesoporous SBA-15 films shows promising results, which also broadens the application range of DiG films.

#### *4.3. Catalytic Performance*

Introducing sulfonic acid groups in the mesoporous films result in a substantial increase in the catalytic performance of the film. A conversion of ~5 % of acetic acid during 1 hour of reaction, when no sulfonic acid groups are present, is in good agreement with other studies [13]. This value is boosted by a factor of approximately 6 times when sulfonic acid groups are present in the pores. The results confirm that the mesoporous silica DiG films are successfully coated with carbon using the evaporation technique and that the pores are accessible for both functionalization and catalytic reactions.

SEM micrographs of the used catalyst show that the film is intact after 1 h of reaction, except for a minor particle loss. It is well known that silica dissolves in water, and this reaction may result in loss of particles that have formed with defects towards the substrate. There is no observable additional particle loss after the second reaction cycle, indicating that the remaining particles are well-adhered to the substrate. The catalytic activity decreases after the first reaction cycle. Other studies of esterification reactions using sulfonated carbon catalysts show a similar trend [36,37], and the reduction can be attributed to loss of active sites of the catalyst, either by desorption of sulfonic groups [37], or formation of sulfonic esters on the catalyst surface [38].

#### **5. Conclusions**

We have shown that monodispersed SBA-15 particles with various aspect ratios can be grown onto silicon wafers using the DiG method. The film thickness follows the particle width and can be tuned between 80 nm and 750 nm by changing the NH4F concentration in the synthesis solution. The addition of salt affects the material formation rate, and therefore the substrate addition time must be adjusted so that the substrates are added during the formation of the siliceous network. It has been concluded that hydrophobic substrates are required for a dense film growth, but that substrates with surface silanols or P123 can bond smaller particles to the surface. The film growth is consistent over surfaces larger than 75 cm2, and it is possible to coat rough substrates.

We have also shown that the films can be functionalized by co-condensation of MPTMS+H2O2, but the addition time of the functional reagents must be adjusted so as to not affect the film morphology. The reagents must be added after the formation of the hexagonal order, during the final condensation, to yield a material with high acidity and accessible pores. A CMK-5 structure can also be formed in the films after exposure to furfuryl alcohol vapor. The sulfonated version of the CMK-5 film was shown to work as an efficient catalyst in the esterification reaction of acetic acid and ethanol, showing the accessibility to the pore system, even though the pores are perpendicular to the substrate, due to the separation of the grown particles. The film is stable upon the catalytic reaction, with only a minor loss of particles. Recycling experiments show, however, reduced catalytic activity after the first cycle. As an outlook, one can imagine DiG film growth on other substrates than Si-wafers, e.g. glass or titanium, and utilizing these films as catalyst hosts or drug carrying coatings for implants.

*Nanomaterials* **2019**, *9*, 562

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-4991/9/4/562/s1, Methods for substrate functionalization, methods for surfactant removal, Figure S1: Pore size distributions, physisorption isotherms, and small angle x-ray diffractograms for materials synthesized with NH4F/P123 molar ratios of 0.0–1.8, Figure S2: Pore size distributions, physisorption isotherms, and small angle x-ray diffractograms for direct sulfonated SBA\_0.0s, Table S1: Physiochemical properties and acidity of materials synthesized with NH4F/P123 molar ratios of 0.0–1.83.

**Author Contributions:** The individual contributions by the authors are: conceptualization, E.M.B. and P.M.; validation, P.-H.W.; investigation, P.-H.W.; resources, M.O.; writing—original draft preparation, P.-H.W., E.M.B.; writing—review and editing, P.M. and M.O.; supervision, E.M.B. and P.M.; funding acquisition, E.M.B. and M.O.

**Funding:** This research was funded by the Swedish research council (VR), grant number 2015-00624, the Swedish Energy Agency (grant no P42022-1), Vinnova (FunMat-II project grant no. 2016-05156), the Swedish Government Strategic Research Area in Materials Science on Functional Materials at Linköping University (Faculty Grant SFO-Mat-LiU grant no 2009-00971), and Knut and Alice Wallenberg Foundation, grant number KAW 2012.0083.

**Acknowledgments:** Rickard Melin at Ionbond Sweden AB is acknowledged for the support in blasting silicon wafers.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
