Solvothermal Synthesis, Gas-Sensing Properties and Solar Cell-Aided Investigation of TiO2-MoOx nanocrystals

Titania anatase nanocrystals were prepared by sol-gel/solvothermal synthesis in oleic acid at 250 °C, and modified by co-reaction with Mo chloroalkoxide, aimed at investigating the effects on gas sensing properties induced by tailored nanocrystals surface modification with ultra-thin layers of MoOx species. For the lowest Mo concentration, only anatase nanocrystals were obtained, surface modified by a disordered ultra-thin layer of mainly octahedral Mo(VI) oxide species. For larger Mo concentrations, early MoO2 phase segregation occurred. Upon heat-treatment up to 500 °C, the sample with lowest Mo concentration did not feature any Mo oxide phase segregation, and the surface Mo layer was converted to dense octahedral Mo(VI) oxide. At larger Mo concentrations all segregated MoO2 was converted to MoO3. The two different materials typologies, depending on the Mo concentration, were used for processing gas-sensing devices and tested to acetone and carbon monoxide, observing a greatly enhanced response, for all Mo concentrations, to acetone (two orders of magnitude) and carbon monoxide with respect to pure TiO2. For the lowest Mo concentration, dye sensitized solar cells were also prepared for investigating the influence of anatase surface modification on the electrical transport properties, which showed that the charge transport mainly occurred in the ultra-thin MoOx surface layer.


Introduction
As the size of a nanocrystal is decreased, the overall chemical and electronic properties will be more and more affected by the chemical species generated by any reaction at the surface. Hence, if the surface is specifically activated toward a given pattern of chemical reactions, the overall system will be tuned to respond to such reactions. Such a concept is concretely illustrated by chemoresistive gas sensors, where the reaction of target gases with the surface of nanocrystalline oxide may provide huge relative changes of the electrical conductance. This result is attributed to a size effect, [1] owing to which any modulation of the surface charge depletion layer will be largely enhanced with respect to the same material in bulk form. However, if the oxide barely supports the sensing reactions by itself, preparing it in the form of nanocrystals may be not sufficient to boost its sensing 4 properties. Hence, following the initial idea, a surface modification by catalytic oxides different from the support oxide can be conceived, aimed at providing a thin surface layer capable of more favored reaction with the target gas, and generating and transporting electrical charges more rapidly than the pure oxide. In Kröger--Vink notation, and considering that only Mo VI is involved in the heattreated materials, the incorporation equations are [Eqs.^^(1)<ffr1> and (2)<ffr2>]: (2) where MoTi and Moi indicate substitutional and interstitial Mo VI ions, respectively, VTi is a Ti vacancy, and OO an oxygen ion in a regular lattice site. The effective ion size of Mo VI in the tetragonal anatase lattice is 0.59^^Å, against 0.605^^Å for Ti IV . It can be seen that the Mo-modified materials outperformed pure TiO2 by over two orders of magnitude, with evident decrease of the best operating temperature, in such a way that pure TiO2 began to give a stable and appreciable response only at very high operating temperatures (400^°C). The Mo-modified materials displayed similar responses, as also shown in the calibration curves in Figure^^8<figr8>, at the best operating temperature of 300^°C. In this case, the C0 sample response was always slightly larger than the 3^C0 sensor, and for both gases it systematically outperformed pure TiO2 by two orders of magnitude.
These data fully demonstrated the achievement of the proof of concept that the catalytic modification of anatase by Mo oxides was able of boosting the sensor performance. Although the underlying mechanisms are still unknown in detail, achieving a nanosized version of an oxidation catalyst was clearly the most important pre-condition for effectively sensing reducing gases with respect to pure TiO2. Furthermore, sample C0 deserves particular attention as: it did not display phase segregation, implying better further stability; for the same reason, it had a less complex composition, easing any further mechanistic study.

Dye-sensitized solar cells as a tool for investigating the surface properties of materials
The surface modification by MoOx layers may be expected to introduce specific states in the TiO2 band structure. The TiO2 surface atoms are no more sub-coordinated with respect to the bulk, as in pure nanocrystals, but are terminated by O<C->Mo bonds, which interrupt the periodicity of the lattice in a different way with respect to the usual surface. After heat treatment, the Mo oxidation state was always VI, as demonstrated above, which is significantly different from the typical bulk and surface states of Ti in TiO2. The presence of different states can be expected to strongly influence the conduction properties of the pure material. Although an early evaluation can be usually obtained by simple conductance measurements, in this work we introduce dye-sensitized solar cells (DSSCs) as an alternative and powerful tool to investigate the presence of additional states introduced by the surface. This application of DSSCs was suggested by the above observation that there is a large difference in the oxidation states of Mo and Ti and trapping effects can be expected. In turn, it is well known that the dynamics of solar cells can easily provide evidence for the modification of the density of surface traps. This application of DSSCs is highly attractive in the present situation (materials with large modifications of the surface composition), but to the best of our knowledge it has never been proposed for such an aim, as such devices are generally being investigated by themselves as energy conversion tools.
Pure TiO2 and TiO2--MoOx (C0 Mo concentration) materials have been used as photoanodes, that is, as the electron transport material (ETM) in solar cells (a summary of the recorded functional parameters can be found in the Supporting Information). Solar energy converting devices have been fabricated by keeping all the other components fixed (i.e., light harvesters, counter electrode materials, and electrolyte composition), thus allowing a comparative investigation between the pure and surface-modified TiO2 nanoparticles.

13
VOC decay is usually correlated to electron lifetime within the ETM (a high lifetime is highly desirable as it would imply reduced recombination processes), which here appears longer in case of surface-modified TiO2. However, this slow decay, coupled with the reduced VOC and the depressed JSC values, strongly suggested the presence of surface trap states able to capture the photogenerated charges, not allowing them to be transported through the circuit.
Photogenerated electrons were not slowly released to convert the solar light into electricity, but participated in other processes among which was, possibly, the (partial) reduction of Mo VI surface species. Photocurrent stability under pulsed light (see Figure^^S31 in the Supporting Information) highlighted these different charge dynamics between the pure anatase samples and the TiO2--MoOx photoanodes as well. The TiO2 ETM featured fast and regular dynamics of the current transients: the maximum photogenerated current value is reached almost immediately after the light had been switched on, whereas TiO2--MoOx ETM required more than 3^^s to achieve the maximum photocurrent value, this behavior being regularly observed. Upon the light being switched off, photoanodes made of pure anatase showed a sudden decrease of photocurrent to zero, whereas surface-modified TiO2 photoelectrodes showed a slow current decay where a residual current could be observed even after 5^^s.
These findings confirmed the experimental evidence that emerged in the gas-sensing analysis: the presence of surface Mo species able to capture charges is extremely beneficial in terms of detection of reducing gases (such as CO) and indeed sample C0 featured better gas-sensing performances compared with pure TiO2, whereas the same effect is, of course, detrimental in the case of solar cells. The generally supposed sensing mechanism for reducing gases is initial oxygen ionosorption, resulting in charge depletion in the sensing material, followed by charge injection upon reaction with the target gas. Hence, the surface Mo species can favor oxygen adsorption, oxidation of the target gases, and can favor the transport of electrons by delaying their injection into the underlying anatase core. Hence, the presented outcomes strongly indicated that, in gas sensing, TiO2 was playing the role of support for MoOx, which were the actual catalytic players, but which could not exist by themselves being so thin.

Conclusions
By controlling the concentration of the Mo precursor in the solvothermal processing of TiO2 nanoparticles, it is possible to induce the growth of a whole range of nanostructures, ranging from anatase nanocrystals surface modified with MoOx layers, to real nanocomposites comprising both anatase and MoO3 nanoparticles. The MoOx surface modification does not form a thick, observable shell, and is completely oxidized to Mo VI oxide after heat treatment at 500^°C, for the lowest (C0) Mo concentration. Both the typologies of materials show remarkably enhanced sensing properties toward reducing gases with respect to pure TiO2, so demonstrating that a suitably deposited catalytic layer may boost the sensing behavior of otherwise barely active oxide materials. Solar cells proved to be a useful indirect investigation tool for elucidating the origin of the improved sensing behavior. In particular, they show that the surface MoO x layers (for C 0 Mo concentration) act as trap states in the case of solar cells (so further demonstrating the presence of surface modification of pure anatase nanocrystals) and, hence, may act as a reservoir of ionosorbed oxygen and, upon reaction with the target gas, of regenerated electrons, which may travel through this layer.
If the sensing results are considered in the whole of the data obtained with other systems that have been investigated, that is, TiO2--V2O5 and TiO2--WO3, some interesting considerations can be drawn. First of all, the sensing capabilities of pure TiO2 are always enhanced by the surface modification. So the overall view of the systems confirms that if the core and surface oxide constitute a suitable reception--support couple, a real synergistic effect is obtained. Second, the sensing enhancement directly depends on the surface oxide, so for instance, TiO2--V2O5 is highly sensitive to ethanol, whereas TiO2--WO3 displayed large responses to low acetone concentrations, in agreement with the catalytic and sensing properties of V 2 O 5 and WO 3 , respectively. In the same way, TiO 2 --MoO 3 does not feature, for instance, as high an ethanol response as TiO2--V2O5 but it is able to boost the response to CO, which was negligible with the other systems. Although this is a specific advantage of this system, it also suggests the coupled use of these systems as sensor arrays where each oxide introduces a specific selectivity feature.

Experimental Section
Experimental TiO2 nanocrystals were synthesized as detailed before. [11] Briefly, Ti chloromethoxide solution (2^^mL) was dropped into n-dodecylamine (10^^mL) at room temperature, followed by heating at 100^°C for 1^^h in a closed glass vial. The white precipitate was extracted by methanol, washed two times with acetone, and then dispersed into oleic acid (10^^mL, technical grade). Then, varying volumes of molybdenum chloromethoxide, ranging from 0.5 to 2^^mL, were added, with a nominal Mo/Ti atomic concentration ranging from C0 (0.5^^mL of Mo precursor) to 4^C0 (2^^mL of Mo precursor). The Mo precursor was prepared as previously described. [12] The C0 concentration corresponded to a 23^% nominal Mo/Ti atomic concentration. The resulting suspension was poured into an open 16^^mL glass vial, inserted into a 45^^mL steel autoclave, and kept for 2^^h at 250^°C. After cooling, the bluish product was extracted with methanol, washed with acetone, and dried in air at 90^°C. Finally, the product was heat treated for 1^^h in air at various temperatures in a porcelain crucible in a muffle furnace. Pure TiO2 materials were similarly prepared by skipping the addition of the Mo precursor.
XRD data were collected in Debye--Scherrer geometry with a Rigaku RINT 2500 diffractometer, equipped with an asymmetric Johansson monochromator (Ge 111 reflection) for CuKα1 radiation (λ=1.54056^^Å) and a D/tex Ultra detector. The rotating anode source was operated at 50^^kV, 200^^mA. The powder sample was introduced in a 0.3^^mm diameter Lindemann glass capillary, set to rotation during data collection. The whole XRD profiles were fitted by the FullProf software [<url>https://www.ill.eu/sites/fullprof/</url>], by using a Rietveld approach taking into account the instrumental resolution function (IRF, that is, the instrumental broadening). A LaB6 powder sample from NIST was used as a standard to evaluate the IRF.
High-resolution transmission electron microscopy (HR-TEM) analyses of the powders were carried out by using a field emission gun FEI Tecnai F20 microscope, working at 200^^kV and with a point-to-point resolution of 0.19^^nm. Electron energy loss spectroscopy compositional characterization was performed by using a GATAN Quantum energy filter coupled to the previous TEM microscope. In this way, we combined annular dark field (ADF) scanning TEM (STEM) to the EELS spectrum imaging (SI) to obtain composition maps of our nanostructures.
Large area XPS measurements at 20^^eV pass energy were performed with an Escalab MkII spectrometer (VG Scientific Ltd., UK) equipped with a 5-channeltron detection system. The samples were pressed on the grated Au foil (99.99^%) fixed on the standard Escalab holder stubs. An unmonochromatized AlKα radiation source (1486.6^^eV) was used for the sample excitation. The binding energy (BE) scale was calibrated by measuring the reference peaks of Au^4f7/2 (84.0±0.1^^eV) from the supporting foil. The spectroscopic data were processed by using Avantage v.5 software (Thermo Fisher Scientific, UK).
Fourier transform infrared (FTIR) measurements were carried out by using a Nicolet 6700 spectrometer in diffuse reflectance setup, after dispersing the sample powders in KBr.
Gas-sensing tests were carried out by using a standard configuration described before [2b] for resistive sensor measurements, with Pt-interdigitated electrodes and a Pt-resistive-type heater deposited onto an alumina substrate. A flow-through technique was used for measuring the response to various acetone and CO concentrations, generated by mass flow controllers. The response was defined as Ggas<M->G0/G0, where G0 and Ggas are the device electrical conductance values after equilibration in pure gas carrier and in the presence of the target gas, respectively. The sensing devices selected for the gas tests had base conductance values dispersed within 10^% of the results shown in the manuscript. In this case, the measured responses were also comprised in such a range. Error bars were hence not included in the plots for the sake of clarity. Repeated experiments under the same operational conditions yielded stable and reproducible sensor responses for several months (estimated uncertainty=±10^%).
Dye-sensitized solar cells (DSSCs) were assembled as it follows. Pure TiO2 nanoparticles and the 3^C0 sample were used as electron transport materials (ETMs) after the preparation of a paste suitable for tape casting. Seven devices were fabricated for each material, the standard deviation of functional performances being lower than 5^%. The paste was prepared by mixing about 100^^mg of the oxide powders with 0.4^^g of ethylcellulose and 1^^mL of α-terpineol. Ethanol (3^^mL) and water (1^^mL) were used as solvents. The mixture was then sonicated for 15^^min and vigorously stirred overnight. Fluorine-doped tin oxide (FTO) glasses (Pilkinton) were used as conductive substrates. The paste was tape cast layer by layer (after the deposition of each layer the sample was heated at 120^°C for 6^^min) to obtain a 5^^μm thick photoanode. Finally, the electrodes were annealed in a muffle furnace at 500^°C for 1^^h under ambient atmosphere. Dye N719 (Solaronix) was used as the light harvester and the sensitization process was done by soaking the electrode in a 0.5^^mM ethanolic solution for 20^^h, followed by careful washing to remove unadsorbed molecules. Device fabrication was carried out by using as counter electrode a 5^^nm thin film of sputtered platinum on FTO glass and the iodine I <M-> /I3 <M-> couple was exploited as the electrolyte (containing 0.1^M LiI, 0.05^M I2, 0.6^M 1,2-dimethyl-3-n-propylimidazolium iodide, 0.5^M 4-tert-butylpyridine dissolved in acetonitrile). Device functional performances were investigated by using an ABET 2000 solar simulator at AM 1.5G (100^^mW^cm <M->2 ), calibrated with a silicon reference cell.

Figure^^3
Mo^3d region in the XPS spectra of the indicated C0 samples.

Figure^^4
Atomic concentrations and atomic ratios to Ti obtained from the XPS spectra of the indicated samples.

Figure^^5
FTIR spectra measured on the indicated samples. The vertical line is a guide to the eye for highlighting the b band.

Figure^^6
Dynamic response curves of the indicated sensors at an operating temperature of 300^°C towards various concentrations of acetone and CO.

Figure^^7
Response of the indicated sensors to 50^^ppm acetone and 100^^ppm CO as a function of the operating temperature.

Figure^^8
Calibration curves for acetone and CO at 300^°C for the indicated devices.