Anatase is the technologically most relevant polymorph of titanium dioxide, finding applications in dye sensitized solar cells, lithium ion batteries, as a catalyst, and in self-cleaning coatings. The drawback to using anatase in photocatalytic applications lies in its large bandgap (3.2 eV), which limits the spectrum of photons that can create electron-hole pairs to participate in oxidation or reduction reactions to the UV, and corresponds to only 4% of the incident solar energy. Hence, reducing the bandgap of TiO2 to coincide with the visible spectrum is a highly active area of research with strategies based on doping, ion implantation, metal loading, and composite semiconductors.
One of the most successful examples of improving the light harvesting capability of TiO2 is the dye-sensitized solar cell (DSSC). The ultrafast electron injection from a photo excited dye into the conduction band of an oxide semiconductor and subsequent dye regeneration by a redox electrolyte result in photocurrent generation. The overall performance of DSSCs strongly depends on the electron injection efficiency that is linked to the electronic dye-substrate coupling as well as the aggregation state of the photosensitizers. We have investigated both the adsorption geometry and electronic characteristics of single photosensitizer N3 (cis-di(thiocyanato)-bis(2,2′-bipyridyl-4,4′-dicarboxylate)ruthenium(II)) adsorbed on the technologically relevant TiO2 anatase (101) substrate . Electrospray ion beam deposition (ES-IBD) was employed for gentle deposition of the thermally fragile N3. Figure 1(a) shows a high-resolution STM image of different N3 molecules adsorbed on the TiO2 anatase (101) surface. The molecule reveals an asymmetric shape where several sub molecular lobes can be distinguished. Each N3 molecule reveals a different shape and dissimilar submolecular features whereas the corresponding apparent heights are homogeneous (≈6 Å). These observations hint at multiple binding conformations of the N3 dyes with different adsorption energies and separating barriers, indicating that in many cases the N3 molecules are trapped in local energy minimum configurations. Figure 1(b) shows STS data of the unoccupied states recorded for the individual N3 molecules marked in Fig. 1(a) (corresponding color code). The differential conductance spectrum of the pristine TiO2 anatase (101) surface is shown in black. This spectrum exhibits an onset of the TiO2 conduction band at about +0.75 eV above the Fermi level. The LUMO peak positions of the N3 molecules vary substantially from 1.6 to 2 V. Note that for some molecules no STS peak was observed in the same bias window, in which STS measurements and tip conditions are stable. Hence, the data reveal a fundamental dependence of the N3 LUMO positions on the specific N3 binding mode to its environment. We found a substantial (0.88 eV) variation of the LUMO/conduction band edge offset across the adsorption geometries considered. This variation results from the interplay of several factors, namely the intrinsic electric dipole moment of the dye molecule, the charge transfer at the interface, and the screening of the photoexcited electron through image charge effects. In addition to the above factors that have already been discussed in the literature, our calculations indicate that the energetics of the N3 LUMO can be sensitive to the distortion of the bipyridine units within the dye. Indeed, when N3 binds to the surface through two carboxylate anchor groups on the same BINA units, the strong Ti−O bonds induce an 11° tilt between the pyridine rings. This tilt is responsible for a blueshift of the LUMO by 190 meV, as we checked by subjecting an isolated N3 molecule to the same distortion. These observations can be rationalized by noting that the N3 LUMO contains sizable contributions from a good dye needs to have its LUMO located near the substrate for efficient electron injection, and its highest occupied molecular orbital far from the substrate in order to prevent electron −hole recombination. This criterion has recently been implemented in a new class of push −pull dyes based on porphyrin derivatives. Most importantly our work suggests that optimization strategies based solely on the electrochemical properties of the dye should be replaced by a more comprehensive approach where the focus is on the engineering of the chromophore−semiconductor interface at the atomic scale.
The other method which we have developed to increase the efficiency of solar harvesting is to modify the surface of anatase itself . This method is a simple and elegant method to reduce the band gap of the anatase surface through the intrinsic modification (i.e. no additional dopants, dyes, etc) of the dominant (101) facet. Figure 2(a) and (b) shows a stoichiometric oxygen-terminated surface prepared by a combination of sputtering, annealing and oxygen-annealing. By increasing the temperature of the oxygen annealing from 670 K to 920 K, a titanium-terminated anatase surface phase is created on the stoichiometric oxygen-terminated surface (Fig. 2(e) and (f)), which is overlaid with a blue color to discriminate it more clearly from the oxygen-terminated surface in red. Three observations are immediately apparent from this STM topograph: the titanium-terminated surface shows a different electronic contrast to the oxygen-terminated surface, the titanium-terminated regions reveal a high density of adsorbates while the oxygen-terminated regions are free of adsorbates and bulk defects, and terraces of the titanium-terminated surface have a different step edge configuration. Moreover, the oxygen-terminated phase can be fully recovered, when the crystal is further annealed in oxygen atmosphere at 920 K. Alternating these surface preparation procedures allows the controlled switching of a purely oxygen-terminated surface to a partially covered titanium-terminated surface, and vice-versa. The proposed and calculated structure of the Ti-terminated anatase surface phase is shown in Fig. 2(g), together with the conventional O-terminated surface phase in Fig. 2(c). The features observed in the experimental images are well reproduced by our calculations (inset of Fig. 2(b) and (f)), and can be rationalized as follows: Upon reduction of the pristine surface, adjacent rows of Ti5c and Ti6c atoms are replaced by corresponding rows of Ti4c and Ti5c atoms. The modified oxidation state of these atoms and the additional contribution of the newly-formed rows of Ti5c lead to a rounding of the protrusions and an offset of the topography.
We have mapped the local density of states in an atom-specific manner over both the oxygen- and titanium-terminated surfaces. Spectra taken on the oxygen-terminated surface show the onset of the valence band maximum (VBM) and conduction band minimum (CBM) separated by a bandgap of 4 eV – an overestimation due to tip-induced band bending common for semiconductors (CBM 0.5 V, VBM –3.5 V). Spectra obtained on the blue line reveal a drastically reduced bandgap of 2 eV (CBM 0.5 V, VBM –1.5 V) with an additional peak at –3.3 V. Comparing these values to the change of the bandgap by surface reconstruction, the reduction of the bandgap on the Ti-terminated surface is significantly larger. Our DFT calculations indicate that the additional peak arises from occupied d states of the undercoordinated Ti4c atoms at the surface.
Taken together, our combined experimental and theoretical analysis reveals the unprecedented observation of a sub-stoichiometric, titanium-terminated anatase (101) surface. This titanium-terminated surface represents a unique surface phase and cannot be considered as a simple reduced anatase surface with isolated (subsurface) oxygen vacancies. Moreover, the undercoordinated Ti4c atoms coalesce into entire domains with a high degree of structural similarity and a new electronic structure. The new proposed structure of the reactive [–111] step edge similar to the Ti-terminated surface gives an explanation for the increased reactivity. The titanium-terminated surface presented here augments the versatility of the technologically most important metal oxide, anatase TiO2. The combination of a reduced surface bandgap and enhanced reactivity of undercoordinated Ti4c sites opens new possibilities in photovoltaics and photocatalytic applications.