Nanostructured photocatalytic titania thin-films: The effect of film morphology on photoelectrochemical properties

Rafael McDonald, Pratim Biswas

Research output: Contribution to conferencePaperpeer-review


Hydrogen has been touted as the fuel of the future 1. Since hydrogen is an element, it contains no carbon, and its use as a fuel, either in combustion or in a fuel-cell, leads to only water as a byproduct. Current techniques for hydrogen generation, however, rely upon steam reforming of hydrocarbons. This is problematic because the 1) hydrocarbons are produced on geologic timescales, which are much slower than current consumption requirements, and 2) the byproducts are carbonaceous greenhouse gasses. The photosplitting of water using solar energy is a potentially clean and renewable source of hydrogen fuel that is environmentally benign and easily distributed. Light impacting titanium dioxide electrodes immersed in water causes the water to split into oxygen and hydrogen 2. However, pure bulk TiO 2 requires UV-light (which is less than 5% of incident solar radiation) for excitation. Recent reports in the literature have indicated that some materials such as In 1-x Ni xTaO 4 3 and WO 3 doped titanium dioxides 4 are capable of harvesting visible sunlight to produce hydrogen and oxygen. The synthesis of such materials remains a challenge, since these compounds are produced by multi-step processes which are often difficult to replicate and even more difficult to scale-up. Advances in gas phase nanoparticle synthesis techniques offer hope that such materials could be synthesized that achieve breakthroughs in photocatalysis. It has been shown that for titanium dioxide semiconducting photocatalysts 5, particles involved in splitting water have optimal conversion efficiencies in the nanoparticle sizes, as quantum effects become more important. At such small sizes though, however, detrimental recombination of electrons and hole pairs occur more frequently on the surface of the particles. As such, there is an as-of-yet undetermined optimal particle size for water splitting, where quantum effects gains are balanced with surface recombination losses. Doping of titanium dioxide is a common way to shift the absorption spectrum into the visible regime. It has been shown for other photocatalytic reactions that there is an optimal dopant concentration 6, and furthermore, that this concentration is dependent upon the particle size 7. The literature also lacks information indicating how other aspects of the morphology (packing density, surface roughness, fractal dimensions, etc.) of the particles may affect efficiency. Models currently exist that predict the effect of gas-phase deposition conditions on thin film characteristics 8;9, but some of these conditions have yet to be experimentally verified. Lastly, the effect of the film morphology on system performance is not well understood. In this work, we demonstrate the effect that different process conditions have on the physical (particle size, morphology, crystal structure) and electrical characteristics of the thin film. Titania films were created via Dip-Coating (DC), Precursor-Vapor Deposition (PVD) and Flame Aerosol Deposition (FAD) onto stainless steel substrates. Process conditions were varied, and the effect on the morphology of the film was explored by various methods. SEM micrographs were analyzed for primary particle size, as well as gross morphology. AFM images were analyzed for surface fractal dimension (D a) via the Shifting Differential Box Counting (SDBC) method. 10;11 The films were examined as deposited via XRD for crystal structure and crystallite size, and relative anatase/rutile fractions were calculated via the method of Spurrand Meyers. 12 Dip coating led to the simplest films composed of unsintered primary particles (Figure 1). They had the lowest roughness (36nm) and the lowest fractal dimension (2.59). Large cracks formed in the film upon drying. As the suspension used for the DC method was made of Degussa P25, the crystallinity (77.3% Anatase) and primary particle size (d pg=30.2nm, σ g=1.26) were predetermined. figure presented. Precursor vapor deposition led to unevenly deposited films with low roughness (40nm) and a slightly more disordered surface structure (D a=2.67). High heat was used in this method to decompose the precursor on the slide, but also had the effect of sintering some of the particles that had already been deposited. As such, the primary particle size was considerably larger (d pg=80.34, σ g=1.21). Despite the high temperatures involved in this method, rutilization did not occur to a great extent (89.3% anatase). Films generated via Flame Aerosol Deposition were highly porous dendritic structures very similar to those described in Kulkarni and Biswas. 8 Primary particles showed very little sintering. Analysis of these films showed a low anatase fraction (70%), but a great enhancement in particle size (d pg=16.3nm, σ g=1.21) and surface structure (roughness of 191 nm and Da of 2.75). As such, further films were generated with this method at different process conditions. FAD deposited films were generated with various configurations and flow rates, resulting in quite different morphologies. The use of a two-port FAD reactor with low feed rates (FAD2-low) led to highly elongated aggregates (Figure 2) composed of primary particles with d pg of 43nm (σ g=1.50). These structures were highly diluted on the substrate surface, and deposition occurred mostly along the plane of the surface (xy plane). The particles most resembled diffusion-limited cluster-cluster growth 13. Higher feed rates in the same flame configuration (FAD2-high) with high dilution rates led to thicker films (Figure 3), with aggregates composed of larger primary particles (d pg=73nm, σ g=1.31). Particle density was high, and the primary direction of growth perpendicular to the surface (z direction). The resulting aggregates had the characteristic morphology of reaction-limited particle-cluster growth. Three-port Flame Aerosol Deposition (FAD 3) also led to RLPC-like aggregates, but with much smaller particle sizes. The particle size was found to be a function of both residence time (Figure 3) and precursor feedrate in the flame (Figure 4). By controlling these two parameters, the primary particle size was fairly easily controlled. Film thickness grew logarithmically, in contrast to the theoretical work of Kulkarni and Biswas who predicted a decrease in surface coverage with increasing film thickness. 8 The model employed in that study assumed a constant thermophoretic force. As the film grows, however, the temperature at the film-gas interface increase, as the deposited particles are not perfect heat conductors. This effectively decreases the thermophoretic force and, subsequently, the film growth rate. Maedler et al. (2005) accounted for this decrease in thermophoretic force, and predict the logarithmic shape that was obtained in this study. 9 As shown previously, increasing the feedrate led to an increase in primary particle size, but it also led to a decrease in roughness and fractal dimension. The pertinent details are given in Table 1. Photocurrents (current under 365nm illumination minus current in the dark) were measured under short-circuit conditions for different films. Figure 5 shows the photocurrents for the films tested. The response of the FAD film (16.3nm primary particles) is a classic photoresponse of a semiconductor. At a fixed potential, the electrode has a steady dark current. When the electrode is illuminated, electron-hole pairs in the space-charge region separate, leading to the observed anodic spike. The response is not immediate; however, as the lamp requires some time to reach maximum output (see inset). Once the maximum current is reached, the photocurrent then decays to a steady-state value, indicating that a portion of the holes that reach the surface are either recombining with electrons in the conduction band or accumulating at the surface, rather than reacting with electrons from the electrolyte 14;15. The process is mass-transfer limited. When the illumination stops, the current quickly decays back to the dark current values. The other four films (DC, PVD, and FAD 3 films with primary particles of 19.7 and 33.9nm) show a similar trend, but with important distinctions. None of these films show the same anodic spike as seen in the 16.3nm FAD 3 film. This indicates that all of the holes reaching the surface are reacting with the electrolyte. Mass transfer of electrolyte to the surface is not the limiting factor. Instead, either electron-hole production or the migration of electrons and holes to the surface is the limiting factor. Given that the size of the particles is larger in all of these cases, transport of the electrons and holes to the surface is the most likely explanation. The second important difference between the films is the actual steady-state photocurrents. All of the FAD films displayed photocurrents much higher than those produced by DC and PVD films. Between the FAD films, one sees that the smaller the primary particle, the larger the photocurrent. In this system with these primary particle sizes, the decrease in photoactivity due to quantum effects (seen in some systems 5 but not in others 16) has not become an important factor. As the particle size decreases, the time required for electron-hole transfer to the surface decreases. This decreases the likelihood of electron-hole recombination, and therefore increases efficiency. Open cell potentials were also measured vs. time.

Original languageEnglish (US)
Number of pages7
StatePublished - 2005
Externally publishedYes
Event05AIChE: 2005 AIChE Annual Meeting and Fall Showcase - Cincinnati, OH, United States
Duration: Oct 30 2005Nov 4 2005


Other05AIChE: 2005 AIChE Annual Meeting and Fall Showcase
Country/TerritoryUnited States
CityCincinnati, OH

ASJC Scopus subject areas

  • Engineering(all)


Dive into the research topics of 'Nanostructured photocatalytic titania thin-films: The effect of film morphology on photoelectrochemical properties'. Together they form a unique fingerprint.

Cite this