Article pubs.acs.org/Langmuir
Tuning the Optical Properties of Mesoporous TiO2 Films by Nanoscale Engineering Birgit Schwenzer,*,† Liang Wang,† James S. Swensen,† Asanga B. Padmaperuma,† Gary Silverman,‡ Roman Korotkov,‡ and Daniel J. Gaspar† †
Pacific Northwest National Laboratory, 902 Battelle Boulevard, Richland, Washington 99352, United States Arkema Inc., 900 First Avenue, King of Prussia, Pennsylvania 19406, United States
‡
S Supporting Information *
ABSTRACT: The optical properties of spin-coated titanium dioxide films have been tuned by introducing mesoscale pores into the inorganic matrix. Differently sized pores were templated using Pluronic triblock copolymers as surfactants in the sol−gel precursor solutions and adjusted by varying the process parameters, such as the polymer concentration, annealing temperature, and time. The change in refractive index observed for different mesoporous anatase films annealed at 350, 400, or 450 °C directly correlates with changes in the pore size. Additionally, the index of refraction is influenced by the film thickness and the density of pores within the films. The band gap of these films is blue-shifted, presumably due to stress the introduction of pores exerts on the inorganic matrix. This study focused on elucidating the effect different templating materials (Pluronic F127 and P123) have on the pore size of the final mesoporous titania film and on understanding the relation of varying the polymer concentration (taking P123 as an example) in the sol−gel solution to the pore density and size in the resultant titania film. Titania thin film samples or corresponding titanium dioxide powders were characterized by X-ray diffraction, cross-section transmission electron microscopy, nitrogen adsorption, ellipsometery, UV/vis spectrometry, and other techniques to understand the interplay between mesoporosity and optical properties.
■
INTRODUCTION Titanium dioxide (titania, TiO2) is a widely used semiconducting material. It already is commercially used in solar cells,1−3 as a photocatalyst,4−6 and for optical coatings,7 just to mention a few applications. High-quality TiO2 films are prepared by high-temperature routes such as metal organic chemical vapor deposition,8 molecular beam epitaxy,9 atomic layer deposition,10 or spray pyrolysis.11 However, probably the most studied synthesis method to prepare TiO2 in either film or powder morphology is sol−gel chemistry, which normally is carried out at or around room temperature and does not require sizable upfront capital for expensive equipment.12 Because TiO2 can be prepared using economical, energy efficient, and environmentally friendly methods, its use is predicted to increase even further in coming years as efforts are made to replace toxic and/or scarce materials in devices and more emphasis is placed on developing higher throughput processes that use less energy.13 In particular, mesoporous TiO2 has received more attention lately, e.g., as a material for battery electrodes14 or in dyesensitized solar cells.15,16 Mesoporous materials in general are used in many different areas, including waste treatment17 and drug delivery.18 TiO2, however, offers the potential to combine the benefits of a mesoporous material with industrially useful optical and optoelectronic properties. Many reports describe the fabrication of mesoporous TiO2, and the low-temperature polymorph anatase 19−21 and rutile 22 TiO 2 have been © 2012 American Chemical Society
synthesized. If mesoporous TiO2 is to be used either for catalysis or in applications that require intercalation of other species, rutile is the preferred polymorph.23,24 The optoelectronic properties, such as the band gap (Eg), of anatase (Eg = 3.2 eV25) and rutile (Eg = 3.03 eV26) also differ noticeably. Nevertheless, for the most part, applications utilizing them are not tied to a specific polymorph, because these properties are tunable, e.g., by doping27 or manipulation of the crystal size.28,29 Several studies have shown that the refractive index of TiO2 can be tuned as well; it can be lowered by introducing nanometer-sized pores into TiO2 films.30,31 For dense, nonporous TiO2 films deposited on glass at 300 °C by electron beam evaporation, a refractive index >2.2 has been reported,32 and the Handbook of Optics lists a refractive index of 2.49621 for TiO2.33 Adding ionic surfactants or block copolymers to the sol−gel solution decreases the refractive index of the final product, because it leads to the formation of a porous TiO2 material after annealing.30,31 Commercially available nonionic ABA triblock copolymers, especially Pluronic triblock copolymers developed by BASF,34 are frequently used as templating agents. Pluronic polymers are amphiphilic triblock copolymers consisting of a hydrophobic poly(propylene oxide) B segment Received: April 9, 2012 Revised: May 29, 2012 Published: June 4, 2012 10072
dx.doi.org/10.1021/la301450h | Langmuir 2012, 28, 10072−10081
Langmuir
Article
believe that retention of organic residues is not a major factor in thinner films, such as the spin-coated samples on Si substrates for which we determined the refractive index and band gap values. Thin Film Preparation. Si substrates (1 cm × 1 cm, ⟨100⟩, Silicon Quest International) were cleaned with organic solvents in a sequence of hot trichloroethylene and then sonication in acetone and 1propanol. The Si substrates were then quickly blow-dried with pressurized nitrogen. Following treatment with UV ozone (UVOCleaner, Jelight Co. Inc., at 15 mW/cm2 for 20 min), the substrates were hydroxylated by dipping them into NH4OH solution (30.0% NH3 basis) for 1 min and subsequent rinsing in deionized (DI) water. After blow-drying, a thin film of TiO2 sol−gel solution was spin-coated onto the substrates from 86 μL of Ti(OEt)4/HCl/polymer/1-BuOH solution (2400 rpm, 40 s). The samples were allowed to cure at room temperature for ≥48 h and subsequently transferred into a muffle furnace. The samples were then heated to the desired temperature at a rate of 1 °C/min and held for 4 h at a maximum temperature of 350, 400, or 450 °C. At the end of the heating period, the samples were allowed to cool to 50 °C as fast as possible without opening the furnace (over the course of several hours). Analysis and Instrumentation. Powder XRD studies of finely ground bulk samples used a Rigaku Miniflex II powder diffractometer with monochromatic Cu Kα radiation (λ = 1.54059 Å) and Bragg− Brentano geometry. Small-angle XRD patterns (2θ = 0.75−10°) were obtained using a Philips Xpert X-ray diffractometer with Cu Kα radiation at λ = 1.54 Å. The BET surface area and pore size of the bulk powder samples were determined through nitrogen adsorption isotherms at −169 °C with a Quantachrome autosorb automated gas sorption system. All samples were degassed at 100 °C for 24 h before analysis. Scanning electron microscopy (SEM) was performed to study the surface topology of the films using a JEOL JSM-5900 microscope. Atomic force microscopy (AFM; DI Nanoscope IIIa Multimode) was employed to investigate the morphology of the spincoated films. AFM images were recorded at tapping mode with a silicon AFM tip. The experimental settings of all AFM images were as follows: scan rate 1 Hz, scan resolution 512 lines, amplitude set point 1 V, integral gain 0.2, and proportional gain 0.3. Transmission electron microscopy images were collected using a JEOL JEM 2010 microscope operated at an accelerating voltage of 200 kV. The point-to-point resolution of the microscope is 0.194 nm. All images were digitally recorded using a charge-coupled device (CCD) camera and were analyzed using Gatan Digital Micrograph 3.3.1. The cross-section sample preparation for the transmission electron microscopy (TEM) observation was done with an FEI Helios Nanolab dual-beam focused ion beam/scanning electron microscopy (FIB/SEM) microscope. To avoid damage of the thin film sample, an approximately 50 nm thick Pt layer (by e-beam deposition) followed by 2 μm of carbon was deposited on top of the sample before the lamella was cut from the film. The refractive index and thickness of the TiO2−P123 and TiO2− 2×P123 films were measured by ellipsometry using a spectoscopic ellipsometer alpha-SE from J. A. Woolam Co., Inc. (CompleteEASE version 3.65, model Si with transparent film). The refractive index and thickness of the TiO2−F127 films were determined using a Rudolf Auto EL-III ellipsometer (HeNe laser, 632.8 nm). Reflectance spectra of the TiO2 films on Si substrates were recorded with a UV−vis−NIR spectrophotometer (Varian, Cary 5) in the wavelength range from 200 to 800 nm at a resolution of 0.5 nm or higher. The band gap (Eg) of each sample was calculated using the Tauc relation for direct band gap materials, αhν = A(hν − Eg)2.30,35 α is the absorption coefficient, A is a scaling factor and was set to 1, and hν is the energy of the photons. The absorption coefficient was calculated from the experimental data using the relationship 2αt = ln[(Rmax − Rmin)/(R − Rmin)], with t being the thickness of the TiO2 film and Rmax and Rmin the maximum and minimum values of reflectance, directly obtained from the recorded spectra.34 For data fitting see Figure S7 in the Supporting Information. Thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC) were performed on a DSC STA 449 Jupiter Netzsch instrument equipped with an Aelos QMS 403C MS instrument. The powders were loaded into alumina crucibles, and
capped by hydrophilic poly(ethylene oxide) A segments on each end (PEOxPPOyPEOx). The molecular weight/chain length of the two components, PEO and PPO, can be modified, and as expected, the resulting Pluronic block copolymers exhibit different properties, solubility in organic solvents being one of them. In this study we used the Pluronic block copolymers F127 (PEO100PPO65PEO100, Mn = 12 600) and P123 (PEO21PPO65PEO21, Mn = 5750) as templating agents to obtain mesoporous TiO2. Ozin’s group reported several studies in which they used P123 for the preparation of mesoporous TiO2 films, and we used their experimental procedure for the sol−gel synthesis20,21 as a starting point for our investigation. A 1-butanol solution containing 0.69 g of P123 consitutes the high-concentration templating agent (2×P123) for our TiO2 materials. Because F127 is not soluble in 1-butanol at the same concentration, the other two templating solutions used here contain half the polymer, 0.345 g of P123 and 0.755 g of F127. To our knowledge, no other study has yet directly compared the effect of varying the added surfactant and also their concentration on the pore formation and thereby the refractive index of the resulting mesoporous material. Over the course of this study another interesting aspect of mesoporous TiO2 emerged: our data suggest that mesoporosity can strategically be used to widen the band gap of anatase.
■
EXPERIMENTAL SECTION
All starting materials were obtained from commercial sources (SigmaAldrich Co. or Alfa Aesar unless otherwise indicated) and were used without further purification. Unless noted otherwise, all experiments were carried out under ambient conditions in air. TiO2 Sol−Gel Synthesis. The templating polymer, F127 (755 mg, 0.06 mmol) or P123 (690 mg, 0.12 mmol, or 345 mg, 0.06 mmol) was dissolved in 7.41 mL of 1-butanol. P123 at both concentrations dissolved completely, but F127 yielded a slightly opaque colloidal solution. Separately 1.33 mL (16.1 mmol) of concentrated HCl was slowly added to 1.93 mL (9.2 mmol) of titanium tetraethoxide, Ti(OEt)4, under vigorous stirring to mitigate any temperature increase from the exothermic reaction. The respective polymer solution was then slowly pipetted into the Ti(OEt)4/HCl solution at room temperature over the course of 2−3 min while stirring. The final sol−gel solution had an overall Ti(OEt)4:HCl:polymer:1-BuOH ratio of 1:1.75:0.013:8.8 and 1:1.75:0.0065:8.8, respectively. All sol−gel solutions were clear, the originally colloidal F127 suspension dissolved upon contact with the more polar, aqueous Ti(OEt)4/HCl solution. The so-prepared sol−gel materials and the resulting TiO2 materials after annealing are denoted as follows thoughout this paper: TiO2− F127 (prepared from a sol−gel solution containing 0.06 mmol of F127), TiO2−P123 (prepared from a sol−gel solution containing 0.06 mmol of P123), and TiO2−2×P123 (prepared from a sol−gel solution containing 0.12 mmol of P123). Bulk material for Brunauer−Emmett−Teller (BET) measurements was prepared by spreading sol−gel solution onto microscope glass slides. The material was allowed to cure in air at room temperature for ≥48 h. Then the coated glass slides were transferred into a muffle furnace, heated to the desired temperature at a rate of 1 °C/min, and held for 4 h at a maximum temperature of 350, 400, or 450 °C. At the end of the heating period the samples were allowed to cool to 50 °C as fast as possible without opening the furnace (over the course of several hours). The solid material was removed from the microscope glass slides and crushed into a homogeneous powder prior to X-ray diffraction (XRD) and BET analysis. With increasing annealing temperatures the color of the bulk TiO2 material changed from dark brown to light beige, indicating that not all organic matter was burned out of these thicker films, particularly at the lower annealing temperature of 350 °C. Control experiments with films spin-coated on glass showed no visible color difference of the films, leading us to 10073
dx.doi.org/10.1021/la301450h | Langmuir 2012, 28, 10072−10081
Langmuir
Article
Figure 1. Scanning electron microscopy images of TiO2−2×P123 thin films on Si substrates after annealing at (a) 350 °C, (b) 400 °C, and (c) 450 °C. The insets show atomic force microscopy data for the same samples (1 μm × 1 μm in size, full scale for height mapping of 5 nm). the data were obtained by heating the samples under air flow (25 mL/ min) from room temperature to 550 °C at a rate of 5 °C/min.
properties of mesoporous TiO2 thin films can be tuned by adjusting the chemical composition of the initial sol−gel solution precursor. Other groups previously reported that adding surfactants or block copolymers to the sol−gel solution, which after annealing results in the formation of a porous TiO2 material, decreases the refractive index of a given material.30,31 Refractive indices determined for the different spin-coated films, as well as information on the pore size, pore volume, and surface area of the bulk TiO2 materials, are listed in Table 1. To
■
RESULTS AND DISCUSSION The acronyms TiO2−F127 (material prepared from sol−gel solution containing 0.06 mmol of F127), TiO2−P123 (material prepared from sol−gel solution containing 0.06 mmol of P123), and TiO2−2×P123 (material prepared from sol−gel solution containing 0.12 mmol of P123) are used throughout the paper to describe the TiO2 thin film or bulk materials prepared from chemically different sol−gel solutions. As much as possible, all characterization described in the following paragraphs was carried out using the spin-coated mesoporous TiO2 films on Si substrates. However, due to analysis or instrument limitations, some materials properties have been determined using similarly prepared bulk TiO2 material. It has been shown previously for other materials systems that the substrate on which materials are synthesized in film morphology, either wet-chemically or by annealing reactions, can greatly influence the material’s properties, film texture, and even chemical compositions.36,37 Nevertheless, in this study any correlation between the material’s properties of different bulk TiO2 samples can be assumed to be the same as between the analogous spin-coated TiO2 thin films. For example, the pore size of the mesoporous TiO2 materials is largely determined by reaction parameters at play before the sol−gel solution is either spin-coated onto Si substrates or drop-cast onto glass slides. Throughout the following discussion, we have clearly indicated which data have been obtained from bulk TiO2 samples and how they relate to the TiO2 thin film samples from which all optical properties have been determined. Macroscopic Thin Film Morphology. As evidenced by SEM and AFM on the micrometer scale, all TiO2 thin films seem smooth and continuous, except for the TiO2−2×P123 film annealed at 450 °C. SEM and AFM images of spin-coated films of TiO2−2×P123 annealed at 350, 400, and 450 °C, are shown in Figure 1. AFM measurements indicate that the roughness (Rms) of the mesoporous TiO2−2×P123 films annealed at 350 °C is 0.3 nm. It increases to 0.4 nm for the one annealed at 400 °C (insets in Figure 1a,b). The TiO2− 2×P123 film annealed at 450 °C exhibits cracking (Figure 1c); therefore the surface roughness of the film cannot be determined reliably. The images of TiO2−2×P123 annealed at 350 °C (Figure 1a) and TiO2−2×P123 annealed at 400 °C (Figure 1b) are representative of the microscale morphology of all TiO2−F127 and TiO2−P123 samples, including the films annealed at 450 °C (not shown). None of the other films annealed at 450 °C showed any fractures. Refractive Indices and Mesopores. Refractive index (n) measurements confirmed our hypothesis that the optical
Table 1. Refractive Index of Mesoporous TiO2 Films and BET Data of Bulk TiO2 Materials, All Prepared by Sol−Gel Synthesis Using F127 and P123 Polymer Fillers as Templating Agents sample TiO2− F127 TiO2− F127 TiO2− F127 TiO2− P123 TiO2− P123 TiO2− P123 TiO2− 2×P123 TiO2− 2×P123 TiO2− 2×P123
annealing temp (°C)
refractive index (n)
pore size (nm)
pore vol (cm3/g)
surface area (m2/g)
350
1.73
3.390
0.1445
121.962
400
1.69
3.763
0.3392
131.097
450
1.62
3.789
0.1329
133.246
350
1.73
4.283
0.0826
66.559
400
1.73
4.293
0.1643
106.256
450
1.68
4.818
0.1308
74.367
350
1.61
4.815
0.2361
192.469
400
1.59
4.824
0.1660
129.10
450
1.58
4.837
0.3877
103.439
our knowledge, this study, for the first time, directly compares what influence varying the type of templating surfactant and also the surfactant concentration has on the pore formation and thereby the refractive index of the resulting mesoporous material. For the mesoporous TiO2 films on Si substrates a decrease in refractive index was detectable with increasing annealing temperature for the samples of all three sets of sol−gel composition. This contradicts findings by Vishwas et al., who report increased refractive indices with increasing annealing temperatures for nonporous TiO2 films prepared from sol−gel solutions.30 However, their study looked mainly at amorphous materials in the annealing temperature range of 25−300 °C, and they attribute the increase in refractive indices to densification of the films. The ongoing hydrolysis and condensation reactions which affect the film density in the temperature range below ∼350 °C are not a main factor 10074
dx.doi.org/10.1021/la301450h | Langmuir 2012, 28, 10072−10081
Langmuir
Article
influencing the refractive indices of our materials, which have been annealed at higher temperatures. TGA data (Figure S1 in the Supporting Information) shows that the weight of TiO2− F127 and TiO2−P123 composite materials stabilizes around 350 °C, slightly higher for the TiO2−F127 composite material. The detectable weight loss between 350 and 400 °C was ∼0.61 wt % for TiO2−F127 and ∼0.46 wt % for TiO2−P123. For the interval between 400 and 450 °C the loss drops to ∼0.29 and ∼0.08 wt %, respectively (the depression in the TGA data in Figure S1 around 450 °C is an instrument artifact). This indicates that with the annealing temperatures used in this study (a) the polycrystalline TiO2 structure has been fully formed and (b) the polymer has been burned out of the inorganic matrix, except for possibly small amounts of trace residues in the low-temperature material (350 °C). Consequently, in the study presented here, the changes in refractive index of a given TiO2 film result from a combination of variation in the size and density of the mesopores within a condensed TiO2 matrix. To understand what governs these observed changes in refractive indices, the BET adsorption pore diameters, adsorption pore volume, and surface area were determined from N2 adsorption isotherms for the bulk mesoporous TiO2 synthesized using F127 or P123 as templating polymers (Table 1). Figure 2 shows the N2 adsorption isotherms and the pore diameter distribution (insets in Figure 2) for the TiO2−F127 bulk materials annealed at the different temperatures. For information about the other mesoporous TiO2 materials see Figure S2 in the Supporting Information. All adsoption curves display a hysteresis loop, which is indicative of mesoporous (pore opening >2 nm) or macroporous (pore opening >50 nm) materials. As expected, the pore sizes observed for the different TiO2 bulk materials (Table 1) generally become larger with increasing annealing temperatures. With increasing crystallinity of the material, smaller voids are annealed out and fuse with other pores, resulting in an increase in the observed overall pore size. According to the BET data, the TiO2−F127 bulk sample annealed at 450 °C contains pores about ∼3.8 nm in size. Cross-section TEM of a lamella cut from a spin-coated TiO2− F127 film (annealed at 450 °C) reveals that the material does not contain isolated pores, but shows a network of interpenetrating pores. On average the observable voids within the material, shown in Figure 3a, are ∼3−4 nm in size. This is in good agreement with the value obtained from the BET measurement of the corresponding bulk TiO2−F127 sample. Although we observe an interpenetrating network of pores, for simplicity reasons we will use the terms “pore”, “pore size”, and “pore density” during our discussion of refractive index and band gap analyses throughout the paper. Because the TEM lamella itself has a finite thickness, which is still larger than the average pore diameter, the sample has a highly three-dimensional nature. Consequently, not all of the TiO2 crystals that form the matrix of the film are in focus in the TEM image shown in Figure 3a. We nevertheless attempted to determine an average particle size/wall thickness of the TiO2 matrix. Measuring only the darkest features of the cross-section micrograph (Figure 3a), a crystallite size of ∼6.2 nm results. As can be seen in the high-resolution TEM image, the lattice fringes seemingly extend outside the core particle (to the left of the indicated lattice fringe direction); this implies that, due to the 3-dimensionality of the sample, measuring the core sections of the TiO2 matrix may underestimate the actual size of the individual crystallites. The selected area electron diffraction
Figure 2. N2 adsorption isotherms and pore diameter distribution (insets) = TiO2−F127 bulk materials annealed at (a) 350 °C, (b) 400 °C, and (c) 450 °C.
(SAED) pattern shown in Figure 3c, however, confirms that the matrix consists of randomly oriented smaller crystalline domains instead of a single crystalline material. The observed correlation between pore sizes (determined from bulk material) and refractive indices of the different TiO2 materials (Figure 4a) and between pore densities (determined from bulk material) and refractive indices of the different TiO2 materials (Figure 4b) is depicted schematically. Comparing the refractive indices of different materials annealed at the same maximum temperature the refractive index increases in the order TiO2−2×P123 < TiO2−F127 ≤ TiO2−P213. Figure S3 in the Supporting Information illustrates this information graphically, which is listed in Table 1. 10075
dx.doi.org/10.1021/la301450h | Langmuir 2012, 28, 10072−10081
Langmuir
Article
was increased.39 On the basis of their findings, it is reasonable to assume that the Rg or any micellar radii of the evaluated Pluronic block copolymers also decrease in the water/butanol mixtures used for our sol−gel synthesis compared to the values calculated for pure water systems. The small pore size of 3.390 nm in the TiO2−F127 bulk material annealed at 350 °C therefore indicates that the block copolymer F127 is solubilized to the molecular level in the Ti(OEt)4/HCl/F127/1-BuOH solution, during the early stages of the hydrolysis and condensation reaction. The larger pore sizes in the TiO2− P123 and TiO2−2×P213 bulk materials are more than twice the Rg calculated for P123 in aqueous solution, which implies that the pore structure was templated by oligomeric micelles of P123 rather than individual P123 molecules. The conditions of micelle formation of Pluronic block copolymers have been studied extensively.40−43 Micelle formation is an endothermic process, and for Pluronic triblock copolymers, it is a result of the hydrophobic poly(propylene oxide) section of the ABA triblock copolymer’s decreased hydrogen bonding to water with increasing temperature.40 At low polymer concentrationsa condition that is satisfied for all sol−gel solutions in this studyP123 and F127 are dissolved as individual polymer molecules in solution below a critical micellization temperature (CMT) at which the polymers start to aggregate into micelles. CMTs for Pluronic block copolymers range from below room temperature to >60 °C,40−43 generally increasing for longer hydrophilic poly(ethylene oxide) segments of the ABA triblock copolymer.41 For F127 and P123 Alexandridis et al. reported that the CMT of the former is ∼10 °C higher than for P123 at the same concentration.40 This explains why in our study at room temperature the block copolymer in the TiO2−F127 sol−gel solution is dissolved as individual molecules, whereas the polymer aggregates into micelles in the TiO2−P123 sol−gel solution, a system for which the CMT has been reported below 30 °C. Additionally, the CMT for a given polymer decreases with higher polymer concentrations in aqueous solution.40,41 Consequently, as Mortensen and Brown state, this means at a given temperature larger micelles/aggregates will form in a more concentrated polymer solution.41 Our data for TiO2− P123 and TiO2−2×P123 materials suggest this is also the case in sol−gel solutions. No clear trends are discernible for pore volume and surface area data. Our data do not support the findings reported by Gan et al., who observed a decrease in pore volume and surface area for mesoporous TiO2 thin films prepared by dip-coating with increasing annealing temperature.44 However, the values obtained for the surface area of the bulk TiO2 materials might not be representative of the actual values or trends for the surface area of the spin-coated TiO2 thin films. The surface tension of the drop-cast solutions (bulk materials) is vastly different from the centrifugal forces exerted during spin-coating (thin films). However, while the pore size in the mesoporous material certainly is the dominant parameter affecting the refractive index, the data indicate that the density of pores in a given material might be another factor influencing the optical properties of the mesoporous TiO2 films prepared from spincoated sol−gel solutions (Table 1). As discussed above, the pores within the mesoporous TiO2, both in the bulk materials and in the spin-coated thin films, are templated by either random coil assembled F127 monomers, with a defined Rg, or micelles consisting of several P123 molecules. In polymer
Figure 3. (a) Transmission electron micrograph showing the crosssection of a TiO2−F127 film spin-coated onto a silicon wafer piece and annealed at 450 °C. An interpenetrating network of pores is clearly visible. (b) High-resolution TEM image of the same material with the lattice spacing indicated. (c) SAED image obtained from the crosssection sample.
Figure 4. Schematic illustrations of the effect that changes in (a) pore size or (b) pore density have on the refractive index of mesoporous TiO2 films on Si substrates.
As discussed above, densification due to hydrolysis and condensation of the TiO2 matrix reaches its conclusion at ∼350 °C, which is approximately the same temperature at which all organic matter of the templating triblock copolymers has been removed from the spin-coated thin films. Therefore, the pore sizes determined for the different materials annealed at 350 °C should closely reflect the size of the radius of gyration (Rg) of F127 (PEO 100 PPO 65 PEO 100 , M n = 12 600) and P123 (PEO21PPO65PEO21, Mn = 5750) in solution, a templating parameter that is arrested during gelation of the sol−gel solution before spin-coating or drop-casting. Rg for F127 and P123 in water has been calculated as 3.44 and 2.12 nm, respectively,38 suggesting that the pores in P123-templated TiO2 should be smaller than in F127-templated TiO2. However, we observe that for materials annealed at 350 °C the pore size increases in the order TiO2−F127 < TiO2−P123 < TiO2−2×P213 (Table 1). With 3.390 nm the pore size of the TiO2−F127 bulk material is slightly smaller than the calculated value of 3.44 nm. TiO2−P123 and TiO2−2×P213, on the other hand, contain pores that are significantly larger (4.283 and 4.815 nm) than the 2.12 nm radius of gyration we calculated for P123. Soni et al. reported that micelles of P123 in binary water/ ethanol mixtures decreased in size as the ethanol concentration 10076
dx.doi.org/10.1021/la301450h | Langmuir 2012, 28, 10072−10081
Langmuir
Article
Table 2. Film Thickness, Calculated Volume of an Individual Pore,a and Pore Densitya [(number of pores)/g] for Mesoporous TiO2 Bulk Materials Prepared by Sol−Gel Synthesis Using F127 and P123 Polymer Fillers as Templating Agents
film thickness (nm) individual pore vol (nm3) pore density [(number of pores/ g) × 1018] a
TiO2−F127 (350 °C)
TiO2−F127 (400 °C)
TiO2−F127 (450 °C)
TiO2−P123 (350 °C)
TiO2−P123 (400 °C)
TiO2−P123 (450 °C)
TiO2−2×P123 (350 °C)
402 20.40 7.1
350 27.90 12.2
350 28.48 4.7
185 44.04 1.9
175 41.43 4.0
173 58.56 2.2
305 58.45 4.0
Values are calculated from experimental data listed in Table 1.
for TiO2−P123 materials. Most likely this explains why the introduction of additional voids between 350 and 400 °C outweighs the consequences of pore fusion, which seems to occur around or slightly below 400 °C for F127-templated materials (Table 1) and should lead to a reduction in pore density. Interestingly, even though pore fusion seems minimal between 400 and 450 °C in these materials, a decrease in the calculated number of pores per gram is observed for TiO2− F127 materials. At this point we can only speculate that the materials thickness is the underlying reason. The thin film samples are approximately twice as thick when spin-coated under the same conditions from F127-templated sol−gel solution as if P123templated solutions are used. For TiO2−2×P123 thin films on Si substrates macroscopic cracking was observed (Figure 1) and attributed to the larger film thickness. Possibly microcracking, undetectable by SEM, occurs for TiO2−F127 materials annealed at 450 °C. This could connect pores to the outside of the TiO2 matrix and decrease the overall pore volume of the material, resulting in a decrease of the number of pores per gram of TiO2 as well. Cross-section TEM observations give some indication this hypothesis is correct. Figure S5 in the Supporting Information shows that a TiO2−F127 (450 °C) thin film contains several open pores along the top surface of the film. Considering TiO2−P123 films annealed at 350 and 400 °C, respectively, which have very similar pore sizes (4.283 Å vs 4.293 Å), it appears that a thinner film (TiO2−P123 annealed at 400 °C, 175 nm) with a higher pore density (4.0 × 1018 pores/g) has the same refractive index as a thicker film (TiO2− P123 annealed at 350 °C, 185 nm) with a lower pore density (1.9 × 1018 pores/g), n = 1.73 in both cases. On the other hand, two films that have the same pore size, such as TiO2− P123 annealed at 450 °C and TiO2−2×P123 annealed at 350 °C, exhibit different refractive indices if the thicker film (305 nm for TiO2−2×P123 annealed at 350 °C vs 173 nm for TiO2−P123 annealed at 450 °C) has the higher pore density (4.0 × 1018 pores/g vs 2.2 × 1018 pores/g for TiO2−P123 annealed at 450 °C). n = 1.68 for the thinner TiO2−P123 film, and n = 1.61 for the thicker TiO2−2×P123. The data suggest the index of refraction is also a function of the pore density in addition to the correlation between the refractive index and the pore size in these materials. Films of similar thickness but with a higher pore density will exhibit a lower refractive index. This is in agreement with the understanding that more voids, such as bigger pores,30,31 in the material result in a decreased refractive index, a phenomenon we described above for our films as well. In addition, on the basis of the interplay between pore density and film thickness and how it affects the refractive index, the data indicate that for mesoporous materials the refractive index most likely is a function of the measurement angle. The observed
science random coil folding of triblock copolymers and micelles formed from triblock copolymers are treated as spherical objects. Therefore, we assume for the following calculation of the pore density for the different TiO2 bulk samples that the pores within the TiO2 matrix are similarly spherical when templated using triblock copolymers, like in this study. From Figure 3a,b we know that this assumption is not entirely correct; fusion of pores occurs during the annealing process, leading to an interpenetrating porous network within the TiO2 matrix, but for the purpose of estimating the pore density within the mesoporous material, employing a hypothetical spherical pore geometry can be considered sufficient. The induced statistical error will be the same for all calculations. The number of pores per gram of material can be taken as a measure of the pore density. It was calculated by dividing the measured pore volume per gram by the individual pore volume, which in turn was calculated from the pore size, assuming a spherical pore geometry. The measured film thickness (by spectroscopic ellipsometry), calculated individual pore volume, and number of pores per gram of TiO2 bulk material are summarized in Table 2. Because we observe macroscopic cracking for the TiO2−2×P123 film (Figure 1c), the pore volume for this filmand as a margin of error also the value for this material annealed at 400 °Cwas not regarded as accurate, and it is not included in Table 2. The bulk materials annealed at 400 °C show the highest pore densities independent of the templating agent (Table 2). Pore densities of 12.2 × 1018 and 4.0 × 1018 pores/g were calculated for TiO2−F127 and TiO2−P123 materials annealed at this temperature, respectively. As stated previously, compared to materials annealed at 350 °C, the samples annealed at 400 °C contain no residues of organic matter (TGA data, Figure S1, Supporting Information). This results in additional smaller voids within the material, which increases the porosity of the TiO2 matrix. For TiO2−P123 materials this overall increase in pore volume leads to an increase in pores per gram of material. On the basis of the difference in pore size (Table 1), pore fusion seems to occur above 400 °C, but below 450 °C for P123-templated mesoporous TiO2 films. Predictably, for this type of material an increase in the individual pore volume as well as the increasing crystalline correlation length of the TiO2 between the annealing temperatures of 400 and 450 °C is accompanied by a decrease in pore density (Table 2). Thus, the pore density peaks around 400 °C for the TiO2−P123 materials. The change in pore density, expressed as the number of pores per gram of TiO2, is more pronounced in the different TiO2−F127 samples and is mainly due to the increase in overall pore volume between the annealing temperatures of 350 and 400 °C (Table 1). TGA analysis (Figure S1, Supporting Information) indicates the weight loss due to polymer decomposition is more severe for TiO2−F127 materials than 10077
dx.doi.org/10.1021/la301450h | Langmuir 2012, 28, 10072−10081
Langmuir
Article
index of refraction is influenced by the distance the light travels within the film during the ellipsometry measurement, from the air−film interface to the film−substrate interface. The larger the angle of incident light, the thicker the mesoporous film appears and the more pores the light will travel through. Therefore, the pore density appears higher, while in reality the pore size and pore density remain unchanged. According to our observations, these perceived changes in film thickness and pore density would lead to changes in the index of refraction. Band Gaps and Crystallinity. As expected, powder XRD of the bulk TiO2 samples revealed increasing crystallinity for samples annealed at increasingly higher temperatures, 350, 400, and 450 °C, within each of the three series of materials obtained from different sol−gel solutions. All materials annealed at 350 °C were predominantly amorphous, with a very weak diffraction pattern that could possibly be attributed to more densified regions displaying a crystalline correlation length of ∼2 nm (see Figure S4 in the Supporting Information). For all other samples the observed XRD pattern could be clearly indexed to TiO2 in the anatase polymorph,45 shown in Figure 5.
small-angle X-ray scattering (SAXS) on selected bulk powder TiO2 samples from this study showed no discernible peaks (see Figure S6 in the Supporting Information). This indicates that in the bulk samples the pores are not arranged in ordered arrays within the TiO2 matrix. The cross-section TEM image of the TiO2−F127 (450 °C) thin film shown in Figure 3a similarly suggests that no ordering of the pores exists in the spin-coated TiO2 samples. Because of the thickness of the thin films, which proved to be insufficient for SAXS analysis, this could not be verified by any other means. Our findings imply that the pores are randomly distributed throughout the TiO2 matrix in the bulk as well as in the thin film samples. The crystalline correlation lengths of the different TiO2 bulk materials were calculated from the 101 reflection at ∼25.5° (by directly measuring the full width at half-maximum (fwhm) of the peak) in the XRD pattern of the different samples using the Scherrer formula (Table 3).47 As expected, the calculated crystalline correlation length increases with increasing annealing temperatures for samples derived from sol−gel solutions with the same chemical composition. This is in agreement with our observation that the size of the pores within the different TiO2 bulk materials (Table 1) generally becomes larger with increasing annealing temperatures as smaller crystallites fuse into larger crystalline domains. For TiO2−F127 the crystalline correlation length increases from 5.9 to 7.3 nm as the annealing temperature is raised from 400 to 450 °C; TiO2−P123 exhibits crystalline correlation lengths of 10.3 and 11.9 nm at these annealing temperatures. For materials prepared with twice the concentration of P123, TiO2−2×P123, the value increases from 6.5 to 10.8 nm. The smaller crystalline correlation lengths observed for TiO2−2×P123 compared to TiO2−P123 samples seems in agreement with the lower Ti(OEt)4:P123 ratio, i.e., a higher concentration of pore-forming P123, in the TiO2− 2×P123 materials. As listed in Table 3, the crystalline correlation length determined from the 100% peak in the XRD pattern of TiO2− F127 annealed at 450 °C on a Si substrate is 8.1 nm. This is in agreement with the crystallite size observed for this sample by cross-section TEM (Figure 3). Visually a crystallite size of ∼6 nm was determined, but as described above, this value might be underestimating the particle size due to the three-dimensional nature of the matrix and therefore also the TEM lamella. Overall the data reveal that the crystalline correlation length is influenced by the underlying substrate on which the mesoporous TiO2 nucleates (Table 3). Interestingly, the
Figure 5. Powder X-ray diffraction pattern of anatase TiO2 bulk materials prepared from chemically different sol−gel solutions [(a, b) TiO2−F127; (c, d) TiO2−P123; (e, f) TiO2−2×P123] annealed at 400 °C (a, c, e) and 450 °C (b, d, f).
Unlike for other mesoporous materials14,22,46 prepared by various synthesis approaches including sol−gel methods,46
Table 3. Band Gap and Crystallographic Data of Mesoporous TiO2 Films Prepared by Spin-Coating Compared to Crystallographic Data of the Corresponding Bulk Materials Prepared by Drop-Castinga spin-coated films
drop-cast bulk material
sample
annealing temp (°C)
band gap (eV)
crystalline correlation length (nm)
d101 (Å)
crystalline correlation length (nm)
d101 (Å)
TiO2−F127 TiO2−F127 TiO2−F127 TiO2−P123 TiO2−P123 TiO2−P123 TiO2−2×P123 TiO2−2×P123 TiO2−2×P123
350 400 450 350 400 450 350 400 450
3.74 3.63 3.82 3.69 3.77 3.77 3.61 3.78 3.79
amorphous 6.2 8.1 7.8 9.6 10.4 5.8 8.2 9.9
N/D 3.50587 3.49201 3.48739 3.47624 3.48109 3.45856 3.47047 3.49571
amorphous 5.9 7.3 mostly amorphous; 2.0 10.3 11.9 amorphous 6.5 10.8
N/D 3.45315 3.48842 3.53552 3.48812 3.50924 N/D 3.4828 3.49611
All materials are synthesized from sol−gel solutions using F127 and P123 polymer micelles as templating agents to tune the porosity of the final product. a
10078
dx.doi.org/10.1021/la301450h | Langmuir 2012, 28, 10072−10081
Langmuir
Article
crystalline surface of ⟨100⟩ Si seems to have an effect on mesoporous TiO2 from F127-based sol−gel solutions different from that on the same material prepared from P123-based sol− gel solutions. The crystalline correlation lengths for TiO2− F127 on Si substrates are larger than the ones observed for bulk TiO2−F127 on amorphous glass; in contrast, with one exception, P123-templated TiO2 annealed at 400 or 450 °C on Si substrates has a smaller crystalline correlation length than the analogous bulk samples. However, P123-templated TiO2 annealed at 350 °C on Si substrates is not amorphous like its counterparts drop-cast on glass. Si has a lattice parameter of a = 5.42 Å.48 The lattice parameter is an order of magnitude smaller than the observed pore sizes in the mesoporous TiO2 materials after annealing at 350 °C (Table 1). However, the relatively high crystallinity of the P123-templated TiO2 thin film on Si at 350 °C (compared to F127-templated TiO2) and the comparatively short crystalline correlation lengths at higher annealing temperatures (with regard to the bulk materials) suggest that the P123 micelles (4.282 and 4.815 nm, respectively) might form some kind of ordered layer at the substrate/sol−gel interface on Si and induce short-range order in TiO2 annealed at 350 °C, similar to creating an opal structure.49,50 Since the polymer burns out of the inorganic matrix at ≤350 °C, this could lead to disorder/shifting of the TiO2 matrix close to the substrate interface (comparable to a fragile inverse opal structure49,50) and result in the observed shorter crystalline correlation lengths at 400 and 450 °C. If, on the other hand, the inorganic matrix cast from the sol−gel solution TiO2−F127 nucleates to a larger degree on the Si surface than is the case in the TiO2−P123 systems, the influence of the underlying crystalline substrate could translate into a larger crystalline correlation length for the material at all annealing temperatures.36 As mentioned above, small-angle X-ray scattering, carried out on selected bulk TiO2 samples, revealed no additional peaks between 2θ = 0.75° and 2θ = 8°, indicating that the pores within the mesoporous material are not aligned and do not induce any short-range order within the crystal lattice (Figure S6 in the Supporting Information). The band gap, contrary to the refractive index, is an optical property that is influenced by crystallographic parameters such as order and strain on the atomic level rather than by mesoscopic features of a material. In this study, the band gaps of the spin-coated mesoporous TiO2 films were determined from reflectance data using Tauc’s relation (Table 3, Figure 6).30,35 All samples exhibited a significantly larger band gap than expected for anatase TiO2 (3.2 eV),25 even taking into consideration reported differences between epitaxially grown films (3.51 eV) and polycrystalline films (3.39 eV).26 Similar observations have been reported by Shen et al. for multilayer TiO2 films.29 In general, two effects can lead to blue-shifted band gaps in semiconducting materials: small crystallites (quantum confinement)29,51 and strain.52,53 Shen et al. attributed their findings to the former phenomenon.29 Unfortunately only the 101 reflection at ∼25.5°, which is the 100% peak of the XRD pattern for anatase, is detectable for the thin films spin-coated on Si substrates. The d101 values are listed in Table 3 for both the bulk material and thin films. For TiO2 cast from TiO2−F127 sol−gel solution d101(film) > d101(bulk), for TiO2 cast from TiO2−P123 sol−gel solutions d101(film) < d101(bulk). With just one reflection observable, the lattice parameters a and c cannot be calculated for the tetragonal system (space group I41/amd). However, compared to d101 =
Figure 6. Band gaps determined for the mesoporous TiO2 films on Si substrates vs the crystalline correlation length of the respective material: TiO2−F127 (□), TiO2−P123 (○), and TiO2−2×P123 (△).
3.520 Å reported for experimental and calculated patterns of anatase TiO2,45 d101 for all of our materials is smaller. This indicates that the crystallographic unit cell of our synthesized materials is under considerable compressive stress. Tanemura et al. calculated the dependence of the band gap of anatase for a wide range of lattice constants under compressive and tensile strain and reported that the band gap can increase up to 4 eV for smaller lattice constants.26 Without being able to determine the lattice parameters c and a, and given that the crystalline correlation lengths of some films also fall into the range where quantum confinement effects can be expected, no finite conclusions can be drawn about the extent to which either the polymer in the original solution or the pore size or density affect the band gap of TiO2. In Figure 6 the band gaps of the different films we prepared in this study are plotted against the crystalline correlation lengths of the respective materials. The plot illustrates that seemingly the pore-induced strain causes a larger blue shift in the band gap than what is caused by quantum confinement. In fact, the three films with the smallest crystalline correlation lengths (all
