Thin films of titanium dioxide (TiO2) structured on nanometer scale have attracted considerable attention in the past few years, as nanostructured TiO2 shows outstanding properties and has a widespread application potential, for example, in photovoltaics, photocatalysis, and gas sensing.1–6 Because of its semiconducting property, TiO2 enhanced the fast growth of organic photovoltaics in terms of the so called dye-sensitized solar cell or Grätzel cell7–9 in which a mesoporous TiO2 network is filled with a hole conducting electrolyte to generate electric power from sunlight. Grätzel cells have reached rather high power conversion efficiencies, however, due to difficult processibility of the liquid electrolyte, and a necessary sealing of the cell numerous attempts have been made to replace the electrolyte by a solid organic hole conducting material. In these hybrid cells, low molecular weight molecules10–12 or conductive polymers13–15 are used. Inspired by the Grätzel cell, design one common preparation route of hybrid cells starts with the preparation of the TiO2 structure on a conducting electrode. Among the different preparation routes of the TiO2 structure, the sol–gel process turns out to be working well.16–20 In a second step, this TiO2 structure is backfilled with the organic material. A successful infiltration is crucial for an efficient working device because only a fully exploited pore structure provides the maximum interface needed for charge generation.
In the past, this has turned out to be challenging, in particular when semiconducting polymers were tried to infiltrate in porous networks, because the wetting properties and the use of a solvent to transport the organic material cause complications. Commonly applied techniques for infiltration are spin coating, solution casting, or direct submersion. Often the films are additionally annealed at a temperature close to the melting point to enhance penetration. Contradictory results about the incorporated amount of material have been reported, in particular when spin coating was applied. Some concluded on a successful infiltration of the hole conducting material in the host network,21–23 whereas others observed the formation of a solid overlying polymer layer on top of the TiO2 with only negligible incorporation.12, 24 Often the incomplete incorporation resulted in low power conversion efficiencies, necessitating a detailed understanding of the parameters being involved and determining the incorporation process.
Recently, we reported about TiO2 films with a hierarchical structure prepared by application of a sol–gel process with an amphiphilic diblock copolymer, poly(dimethyl siloxane)-block-methyl methacrylate poly(ethylene oxide) [PDMS-b-MA(PEO)], as structure directing agent and addition of poly(methyl methacrylate) (PMMA) colloidal particles as microstructure template.25 By combing these two techniques, a film structure on three size scales is obtained: (1) a mesoporous network on nanometer scale, which acts as the host network for the hole transport material and provides a large interface required for charge separation, (2) on a medium scale macropores embedded in the surface to enhance roughness and to act as penetration channels for polymer infiltration, and (3) on micrometer scale a hole structure for reduction of light reflection at the surface and improvement of absorption by additional scattering. With such hierarchical structure, these films are well-suited candidates to investigate the next step on the route to the device, the infiltration of the hole transporting material.
By use of the hierarchically structured films, TiO2:polymer composite films have been prepared as a model system for hybrid solar cells. As the polymer component poly(N-vinylcarbazole) (PVK) was chosen. Like most carbazole polymers, PVK has a low highest occupied molecular (HOMO) level and is featured by excellent chemical and thermal stability, characteristics promising a high application potential in photovoltaic devices.26 The PVK was infiltrated in the TiO2 network by solution casting, and the result compared with the application of spin coating the PVK. The structure in the bare TiO2 film and in the composite films is investigated with time-of-flight grazing incidence small-angle neutron scattering (TOF-GISANS).27 Figure 1 shows the basic principle of this technique. A scanning electron microscopy (SEM) investigation complements the GISANS analysis.
In general, the use of neutrons instead of X-rays in scattering experiments allows a fine tuning of the scattering contrast by deuteration or selection of materials with appropriate scattering length densities. In this experiment, we matched the scattering contrast of the TiO2 matrix with the infiltrated PVK, which allows conclusions on the penetration of the polymer in the TiO2 from shape and intensity of the scattering pattern. The measurements are performed in TOF mode, where a broad spectrum of wavelengths is used instead of a monochromatic beam. By subsequent slicing the neutron spectrum in narrow wavelength channels a whole set of scattering patterns is obtained, where each pattern covers a different range of the scattering vector. In other words, by application of the TOF technique, the advantages of the use of several wavelengths are accessible in one single measurement in GISANS geometry.27
Silicon wafers were used as substrates and cleaned before use in an acidic bath.28 For TiO2 film preparation, PDMS-b-MA(PEO) (0.559 g) was dissolved in a mixture of THF (8.63 g) and isopropanol (2.07 g), afterward HCl (37%, 0.093 g) as selective solvent and titanium tetraisopropoxide (TTIP, 0.278 g) as TiO2 precursor were added drop wise. After stirring for about 1 h, PMMA microspheres in an amount of 11.35 mg, corresponding to 7% PMMA weight fraction with respect to PDMS-b-MA(PEO), were added to 3 mL of the sol–gel solution, and the solution was spin coated immediately to avoid dissolution of the PMMA. Spin coating was done on a Süss MicroTec Delta6 RC spin coater under ambient conditions (temperature 23 °C, relative humidity 23%) with 2000 rpm for 60 s. To extract the PDMS-b-MA(PEO) and PMMA templates, the films were calcined for 4 h at 450 °C in air with a heating rate of 6.25 °C/min starting from room temperature, before pieces 65 × 65 mm2 in size were cut from the silicon wafers.
Poly(N-vinylcarbazole) (PVK) with a molecular weight Mw = 1100 kg mol−1, purchased from Sigma Aldrich, was used as received and dissolved in a mixture of toluene and cyclohexanone (9:1 volume ratio) in a concentration of 10 mg ml−1. For infiltration in the TiO2 network by solution casting, 2 mL PVK solution was taken and spread over the TiO2 film. The solution was dried under ambient conditions in a closed box to control the total evaporation of the solvent. An additional heating was not applied during drying. The spin coated PVK film was prepared by use of 4 mL PVK solution. Spin coating was done at 2000 rpm for 60 s. No additional annealing of the dried films was applied.
The grazing incidence small-angle neutron scattering (GISANS) experiments were performed at the REFSANS instrument29, 30 at FRM II (Garching) in time-of-flight mode (so called TOF-GISANS). Instead of a monochromatic neutron beam, a beam with a broad range of wavelengths from 2 to 19 Å was used and recorded as function of the time-of-flight. The pulsing of the beam was realized by a double chopper system. A setting with four beams, which are focused at the detector and a sample-to-detector distance of about 10 m, was chosen. The incidence angle was set to 0.5°, and the scattering signal recorded on a two-dimensional (2D) detector without movement of any motors. In front of the detector, a beam stop was installed at the position of the direct beam to avoid saturation. The counting time for each sample was 48 h. For analysis, the neutron spectrum was sliced into 22 channels with a resolution of 10% for each channel.
Scanning electron images were taken with field emission scanning electron microscopes (Zeiss LEO 1530 Gemini and Zeiss NVision 40) operated at an accelerating voltage of 3 and 5 kV, respectively, and at low working distances from 1 to 5 mm.
RESULTS AND DISCUSSION
TiO2 Film Structure
To characterize the initial structure installed in the TiO2 films, one film was investigated before the infiltration of PVK. Figure 2 (top row, a1 to a6) shows six selected two-dimensional GISANS images of this bare TiO2 film measured at different wavelengths from 3 to 15 Å. In general, the lower part of the scattering patterns contains the transmitted information and is dominated by the transmitted beam, which is partially shielded by a beam stop. In the upper part of the scattering patterns, above the sample horizon, the reflected GISANS signal is probed.31–33 It consists of an intense specular peak, and between the transmitted and specular reflected beam, a Yoneda maximum is visible. Because the Yoneda maximum of a material is located at the position of its critical angle for total reflection αc, which is dependent on the neutron wavelength, this maximum shifts to higher scattering angles with increasing wavelength. In contrast, the positions of the direct beam and the specular peak stay at a fixed position for different wavelengths.27 For neutrons with long wavelength, the critical angle for specular reflection becomes similar to the incidence angle, thus only a small portion of the neutrons can penetrate the sample and the specular reflected beam becomes the dominant feature in the scattering image. To yield also sufficient information from the depth of the sample, only the images taken at shorter wavelengths (λ < 10 Å) were used for analysis. In these images next to the Yoneda maximum a lateral structure is present, which originates from the pore structure in the TiO2 film. The lateral structure is best pronounced in the wavelength range around 5 Å, where the scattering from the TiO2 network is maximal. Around the transmitted beam no pronounced lateral structure is visible in the scattering images due to the small total film thickness.
To analyze the 2D scattering images qualitatively, line cuts were taken in vertical direction at qy = 0 and in horizontal direction at the position of the intensity maximum (Yoneda peak).34 The vertical cuts contain the information about the film structure perpendicular to the film surface, whereas the horizontal cuts provide the in-plane structural information.34 Figure 3(a) shows the vertical cuts for all images with increasing neutron wavelength from bottom to top. The transmitted beam at a scattering angle αi+αf = 0° is shielded by the beam stop, and at 0.5°, the incidence angle, the sample horizon is marked by a deep minimum in the intensity. The specular reflected beam is located at 1°, twice the value of the incident angle. The position of the specular reflected beam slightly drops to lower angles at long neutron wavelengths because of gravitational effects, but the total shift is negligible in the selected wavelength range and no correction of the data needs to be applied. Located between the horizon and the specular reflection is the Yoneda maximum, which is split in two components.34 The shift of both components with the wavelength follows the relation
with ρ being the scattering length density (SLD) of the related material. For the observed two components of the Yoneda maximum [in Fig. 3(a)], SLD values ρ1 = 0.40 × 10−6 Å−2 and ρ2 = 1.91 × 10−6 Å−2, respectively, were determined by a fit to measured Yoneda peak positions [see Fig. 4(a)]. The component at higher αI + αf, values originates from the silicon substrate (ρ2 ∼ ρSi = 2.07 × 10−6 Å−2), whereas the component at lower αi+αf, values represents the SLD of the porous TiO2 film ρ1. This determined SLD is an effective value averaged over the TiO2 network and the pore structure included in the film, which was assumed to be filled with air with a SLD ρair = 0. Thus, the SLD value of 0.40 × 10−6 Å−2 corresponds to a film with a TiO2 material content of 15%, that is, a porosity of 85%. Compared to reported porosity values for TiO2 films,35–37 this porosity is rather large, although it has to be considered that the porosity of a film depends on the morphology and values for films with different pore size are only limited available for a comparison. Moreover, the determination of porosities with different techniques than neutron scattering becomes complicated for high porosity values.38
The horizontal line cuts provide additional information about the pore structure of the TiO2 film [see Fig. 5(a)]. Because of the different wavelengths operated in TOF-GISANS, each horizontal cut covers a different qy-range. Therefore, in total, a larger range of lateral structures is probed as compared to a single-wavelength GISANS experiment.27, 31–33 In the horizontal cuts corresponding to measurements with small wavelengths [bottom curves in Fig. 5(a)], a pronounced intensity maximum is visible at large qy-values. Moreover, a broad shoulder is present in particular in the cuts corresponding to measurements at long wavelengths. Thus, all cuts were fitted with a structure factor function consisting of two lateral lengths ξ1 and ξ2. To account for deviations from these two main positions, distribution functions with Lorentzian shape were used. With an additional Lorentzian-type function, the experimental resolution was taken into account.34 From the fits, the dominant structural lengths ξ1 and ξ2 corresponding to the peak and the shoulder were determined to 52 nm and 180 nm, respectively. In comparison with an SEM image showing the hierarchical structure of the films (see Fig. 6), the structural length ξ1 is attributed to the size of the mesopores, whereas ξ2 describes the size of the macropores embedded in the surface. The micrometer-sized holes representing the third structural level are beyond the resolution limit of the REFSANS instrument in the applied experimental settings and thus cannot be identified in the horizontal cuts.
After characterization of the bare TiO2 film, a PVK solution was infiltrated in the pores and dried to the complete evaporation of the solvent (solution casting). For comparison, a second sample was spin coated with a PVK solution of the same concentration. TOF-GISANS images were taken again to probe the inner structure of these composite films (see Fig. 2, middle row b1 to b6 and bottom row c1 to c6). As compared to the scattering images of the bare TiO2 film, the images of the composite films show clear changes. The total intensity increases in the Yoneda maximum due to a strong scattering contribution of the PVK. In contrast, the lateral maxima exhibit a lower intensity in case of the spin coated film and are even vanished in case of the solution cast film. Because the SLD values of TiO2 and PVK are comparable (2.61 × 10−6 Å−2 for TiO2 and 2.29 × 10−6 Å−2 for PVK), the scattering contrast is low in case of pores completely filled with PVK. A well-filled TiO2 network shows no significant scattering in the horizontal direction. Thus, from a decreased intensity a partial filling of the pore structure in the spin coated film can be concluded, whereas the vanished lateral structure in the solution cast film suggests a well penetration of the polymer in the pores and a high degree of filling. This is also seen in the horizontal cuts [see Fig. 5(b,c)]. In the cuts of the spin coated film, the peak maximum is less pronounced, and it is not visible in the cuts of the solution cast film. However, it had to be included in the fits with a very low peak intensity to describe the cut profile, thus a weak scattering contribution from the pore structure is still present.
In the spin coated film, the lateral maximum representing the pore size is found at a position of 49 nm. As compared to the bare TiO2 film, this value is slightly decreased, which can be an indication for a wetting of only the pore walls and the formation of a only thin wetting layer instead of a complete penetration. However, the experimental error in the TOF-GISANS experiments caused by the limited statistics in the horizontal direction is on the same order, which prevents a clear statement. Because of the long counting times of 48 h per sample, improved statistics by an increased counting time are not realistic.
The findings about incomplete or complete pore filling are further quantified by the analysis of the shift in the position of the Yoneda maximum as a function of neutron wavelength in the vertical line cuts [see Fig. 4(b,c)]. In case of the spin coated film, the lower-angle component shows a higher shift compared to the bare TiO2 film, indicating an increased SLD value. From the fit to the measured peak positions results a SLD of 0.86 × 10−6 Å−2, corresponding to an TiO2 film with 64% porosity, where 25% of the primordial pores are filled. This observation of a partial pore filling is in agreement with the possible PVK wetting layer at the pore walls extracted from the horizontal cuts but can also be an effect of a compact layer on top of the TiO2, which fills only the pores close to the surface and the large macropores. For the solution cast film, no splitting of the Yoneda maximum is visible anymore, that is, the intense maximum of the TiO2:PVK composite is superimposed to the weaker Si maximum. The corresponding SLD is 1.69 × 10−6 Å−2, a value which is obtained for a film with only 28% porosity and 67% of the primordial pores filled with PVK. Hence, when solution casting is applied, the PVK can penetrate the pores and fill the TiO2 network to a higher degree.
The results of the TOF-GISANS analysis are in good agreement to observations in cross-sectional SEM images (see Fig. 7). The image of the composite film prepared by spin coating shows a compact PVK layer on top of the TiO2 film. The PVK is incorporated only in the surface part of the TiO2 network and, thereby, smoothes the surface. On the SEM image, no PVK is visible inside the pores, but it has to be considered that the resolution of the image is too low to show a thin wetting layer on the pore walls. In the film prepared by solution casting, the situation is different. A composite film consisting of polymer filled pores is visible underneath the PVK layer. Thus, in the solution cast film, the PVK has penetrated the TiO2 network substantially better than in the spin coated film. However, the image also shows some unfilled areas, in particular at the bottom interface of the film close to the substrate. This agrees with the partial pore filling derived from the GISANS data and indicates the presence of isolated pores that are not connected to the surface.
In the spin coating process, the evaporation of the solvent is fast, and the film dries rapidly in a disequilibrium state.39, 40 Under these conditions, the mobility of the polymer chains is too low to diffuse into the pores when the solution concentration increases, and only a very limited amount of polymer is incorporated in the pores. Instead, most of the material remains outside, where it forms a compact overlying layer when the film dries.12 This is different when solvent evaporation is slow, then the polymer chains keep their mobility and are able to rearrange during the drying process. The polymer concentration remains homogeneous during solvent evaporation, and the polymer chains enrich in the pores until a solid polymer film forms. Thus, under the conditions of solution casting, a filled pore structure is observed, whereas under spin coating conditions preferably an overlying layer is formed.
It is known that numerous parameters influence the penetration behavior of the polymer; the specific properties of the polymer and its solution such as molecular weight, viscosity, and concentration play a major role as well as the morphology and the surface chemistry of the host network.22, 24 It is the combination of all these parameters that determines if the network is filled or not, and variations in one parameter can change the whole process. The interplay of all parameters and their complex interrelations is not yet understood in detail, and the specific experimental conditions are different in each investigated system. Thus, it is not surprising that the results differ from complete pore filling to only negligible incorporation. As a consequence, far more work will be necessary to overcome this problem on the way to efficient devices.
TiO2:PVK composite films are prepared as a model system for a hybrid solar cell, and the inner structures are investigated with TOF-GISANS. The bare TiO2 film shows a porous network structure on two length scales. The structure is characterized by mesopores with a mean size of 52 nm and macropores with a mean size of 180 nm. When PVK is infiltrated in the TiO2 network by spin coating of the polymer solution, a weak penetration of the pores and an incomplete filling of the network structure are observed. Mostly, a solid PVK forms on top of the TiO2 and only little PVK wets the pore walls. A much higher degree of filling is obtained with solution casting of PVK. In solution casting, the polymer solution is able to penetrate the pores because the mobility is kept in the slow evaporation of the solvent. The filling of the pores results in a lower scattering contrast, which expresses in the scattering images in a disappeared lateral structure. Thus, solution casting appears to be a well-suited candidate for back filling of porous networks for photovoltaic devices and might also be successfully used for other polymers.
The application of other conducting polymers in a similar set of experiments promises to be successful as well, in particular when (partial) deuteration is used to match the scattering contrast of the host matrix material even better. In conclusion, TOF-GISANS is a well-suited tool to study the penetration of polymers in porous networks.
The authors thank M. Memesa and P. Weiser for the SEM measurements. This work was financially supported by the Deutsche Forschungsgemeinschaft in the priority program SPP 1181 “Nanomat” (MU1487/5).