Effect of fluoride-mediated transformations on electrocatalytic performance of thermally treated TiO2 nanotubular layers

The peculiarities of morphology, crystalline structure and chemical composition of TiO2 nanotubular layers (TNT) and TiO2 nanoparticulate (NP) layers obtained by fluoride-mediated transformation of TNT have been described in the present paper. The annealing of amorphous TNT in a confined space under limited air access conditions leads to TNT-to-NP transformation accompanied by fluorine doping of the titania matrix as supported by X-ray photoelectron spectroscopy investigations. The collapse of tubular structure as well as the formation of nanoparticles was confirmed using scanning electron microscopy (SEM), X-ray diffraction (XRD) and Raman spectroscopy. Electrocatalytic activity of both TNT and NP electrodes toward oxygen reduction reaction (ORR) has been examined by cyclic voltammetry (CV). The positive shift of ORR wave of the NP layers in comparison with TNT makes the NP-based electrodes AC CE PT ED M AN US CR IP T more suitable for oxygen reduction. The improved activity of the NP electrodes is attributed to the increased concentration of redox active Ti species owing to fluorine doping of TiO2 which plays a crucial role in electroreduction of oxygen molecules.


INTRODUCTION
Recently, nanostructured titanium dioxide has attracted considerable scientific interest due to combination of specific properties (unique optical properties, high catalytic activity, large surface area and unusual mechanical characteristics) in comparison to the bulk TiO2 which makes this material useful for large-scale potential applications such as solar cells, water splitting for hydrogen generation, battery cathode materials, sensors, recording materials and others [1−6]. It is generally accepted that structural and morphological features have a big impact on the performance of TiO2 nanomaterials. In nature, TiO2 occurs in three different polymorphs: anatase, rutile and brookite [7]. Rutile is widely used as a white pigment, while anatase is the most active polymorph in photocatalysis [8,9]. As for brookite, it remains to be the least studied phase, mainly owing to the difficulties with obtaining its pure phase [10]. It should be noted that not only anatase and rutile demonstrate varying photoactivity, but the different crystallographic orientations of the same TiO2 polymorph may exhibit different activities [11]. In particular, both theoretical and experimental studies have demonstrated that the surface of [001] facets of anatase exhibits a very high reactivity [12]. Moreover, incorporation of different dopants (N, C, S, F), along with Ti 3+ formation, H + uptake etc., can introduce additional energy states in the oxide band gap, thus resulting in modification of electronic and optical properties of TiO2 [2, 13−18].
Interest in nanostructured titania has extended by development of 1D structures including nanotubes, nanorods, nanowires and nanofibers [19]. Among nanostructured TiO2 materials, highly ordered TiO2 nanotubes (TNT) are the most interesting structures owing to the possibility of efficient 1D charge transport, enhanced light absorption and propagation characteristics arising from their precisely controlled and oriented porosity, and excellent performance in alkali and acidic environment [20]. Therefore, particular attention was given to the preparation of titania nanotubes and many methods were developed, including the hydro/solvothermal treatment of TiO2 nanoparticles with an alkaline solution, anodization of Ti foil, templateassisted methods and others [21−23]. Nowadays, electrochemical anodization of titanium in A C C E P T E D M A N U S C R I P T fluoride-containing electrolytes is considered to be the most simple and prevalent approach for titania nanotubular layer production [2, 24−26]. Significant efforts were made to optimize anodization conditions for preparing TNT layers of desirable morphology and thickness [6, 27−29]. Over the past few years, the third generation of TNT has attracted particular interest, where introduction of organic electrolytes such as ethylene glycol and glycerol with a small amount of water and Fions allows the formation of highly ordered, smooth, uniformly shaped and significantly long tubes [27,30]. It is well-known that as-formed TNT layers are amorphous, but via annealing they can be converted to the crystalline forms of anatase or rutile [31].
Interestingly, the annealing conditions can be crucial for reservation of tubular structure as well as for doping of TNT through thermal treatment in different gas atmosphere [32−36]. Recently, Alivov and Fan have reported on the transformation of TNT into nanoparticles (NP) under controllable annealing conditions [37,38]. Necessary parameters for such transformation are the high temperature raping rate and the close contact of TNT's open end with a supporting glass slide which needs for catalytic reaction of fluoride residues in TNT with TiO2. Such nanotube-tonanoparticle transition upon annealing in fluorine ambient can certainly alter the physicalchemical properties of the TiO2 material. For example, the formation of TiO2 nanoparticles in the presence of fluoride ions can be accompanied by the increase of surface area with a high portion of reactive [001] facets [39]. At the same time, the electron lifetime and charge collection efficiency can be decreased significantly for nanoparticulate layers as compared to nanotubular ones [40]. Thus, the monitoring of TNT-to-NP transformation is an important issue to predict the behavior of resultant material and to fabricate the nanostructured TiO2 electrodes with desired properties. Moreover, the presence of fluorine atoms could also determine the properties of the TiO2 surface in catalysis, gas sensing, etc. [41−43].
Recently, nanostructured TiO2 electrodes including nanotubes have received attention as a potential support for active metallic catalysts in fuel cells [44−48]. Besides, Sacco et al. [49] have shown that TNT layers can be considered as promising electrocatalytic material itself, since catalytic performance of the crystalline TNT in oxygen reduction reaction (ORR) is only slightly lower with respect to platinum. This fact dictates the necessity of detailed investigation of electrocatalytic activity of bare TNT as well as NP layers obtained via TNT-to-NP transformation.
In the present work we provide direct comparison of morphology, crystalline structure and chemical composition of titania nanotubes and nanoparticulate layers formed as the result of fluoride-mediated TNT-to-NP transformation during thermal treatment in a confined space under limited air access conditions. Scanning electron microscopy (SEM), powder X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS) and Raman spectroscopy were used for characterization of the obtained nanostructures. Furthermore, electrocatalytic activity of both TNT and NP electrodes in oxygen reduction reaction was studied and discussed.

EXPERIMENTAL
Self-organized highly ordered titania nanotubular layers were produced on commercially pure titanium sheets (4 cm × 1 cm; 99.7 % Ti, Alfa Aesar) by two-steps anodization in ethylene glycol electrolyte containing 0.75 wt% NH4F and 2 vol. % H2O. The Ti sheets were polished mechanically and then chemically in a HF:HNO3 (1:2 by volume) mixture to mirror finish and finally rinsed with deionized water. The anodizing cell has a two electrode configuration with a Pt sheet as the cathode and the Ti sheet as the anode. The electrochemical anodization for both steps consisted of a potential ramp from 0 to 40 V (sweep rate -200 mV s -1 ) followed by holding the potential constant for 1 h. The oxide film formed during the first step of anodization was removed by detachment in an ultrasound bath with deionized water. Before the second anodization, the electrochemical cell was filled with a fresh portion of the electrolyte. After the second step of anodization, the samples were washed with ethanol and then their surface was cleaned from debris by treatment in an ultrasound bath with distilled water during 30 s. One part of the anodized samples was used to produce crystalline TNT layers, other part was taken to convert amorphous TNT layers to the NP ones. In order to obtain crystalline TiO2, the samples were annealed at 450 °C for 3 h (heating rate -5 °C/min) using a tube furnace (TERMOLAB-Fornos Eléctricos, Lda.). To ensure a formation of well-defined tubular structure, the thermal treatment was performed in air flow (ALPHAGAZ TM 1 AIR, O2 concentration -20±1%). Under such conditions, NH4F and other fluorine-containing species, which inevitably remain in the amorphous TNT layers even after washing, are dragged by the flow without reacting with titania.
On contrary, when the thermal treatment occurs in conditions of a restricted mass transfer, the fluorides react with the amorphous TNT matrix that results in destruction of the nanotubes and formation of nanoparticles. Therefore, an annealing of the samples aimed at the TNT-to-NP transformation was carried out at limited air access in a narrow long quartz tube with a sealed end (the samples were placed near the sealed end) or by covering up them with a glass slide resulted in the transformation of nanotubes into nanoparticles.
Microstructure and surface morphology of the samples were characterized using using a Hitachi SU-70 and a Hitachi S-4800 scanning electron microscopes, as well as a Hitachi H-800 transmission electron microscope (200 kV). Phase identification was performed using a PANalytical Empyrean diffractometer (Ni-filtered Cu Kα radiation, step 0.02°, 2-s exposition per step over the angular range of 10-80°). The crystallite orientation study was carried out using a PANalytical X'Pert PRO MRD high-resolution 4-circle diffractometer in Cu Kα radiation.
Position of the detector was fixed at the reflection angle corresponding to diffraction from the (004) plane. Pole densities were plotted in stereographic projection with the plane of the projection chosen to be parallel to the sample surface. Raman spectroscopy was also applied to compare the crystallite orientations in the samples. Raman spectra were taken at room temperature using a Nanofinder HE (Lotis TII, Belarus-Japan) confocal microscope based setup.
Raman scattering was excited using the 532-nm solid-state laser (0.6 mW, 120 s). X-ray photoelectron spectroscopy (XPS) was used to analyze the influence of different annealing conditions on chemical composition and electronic state of elements in the specimens. XPS analysis was performed on a Kratos DLD Ultra spectrometer using Al Kα monochromatized radiation (E=1486.6 eV). For survey spectra, pass energy (PE) of 160 eV was used while for regions PE was 20 eV. XPS spectra were recorded before and after Ar + etching at different sputtering time (sputter rate 8 nm/min calibrated on Ta2O5). All the binding energies (BEs) were referenced to the C 1s peak at 284.8 eV of the surface adventitious carbon.
The electrocatalytic activity of both TNT and NP electrodes in oxygen reduction reaction was examined by cyclic voltammetry (CV) using an Autolab PGSTAT 302N potentiostat in a 0.1 M KOH solution saturated with oxygen during 1 h. Electrochemical experiments were performed in a single-compartment glass cell using a standard three-electrode configuration. An Hg/HgO electrode filled with 1 M KOH (Radiometer Analytical) and a Pt foil were used as the reference and counter electrodes, respectively. The potential sweep rate was 10 mV s -1 . Figure 1a shows typical SEM images of the as-grown TNT layer obtained by two-steps anodization in ethylene glycol based electrolyte. Well-defined tubular structure with a relatively narrow distribution of the inner pore diameter (60±5 nm) and the wall thickness (12±2 nm) can be observed. After the thermal treatment in the air flow, the TNT sample has grey color and preserves the ordered tubular structure (Fig. 1b). The cross-sectional view also demonstrates the well-aligned nanotubular arrays of 10±1 µm length (Fig. 1c).

Microstructure characterization
The as-grown titania layers annealed under limited air access were found to become of whitish colour and to transform into nanoparticular layers (Fig. 2a-c). The thickness of such layers was appr. 6 µm (Fig. 2c). The observed decrease in the thickness of the NP layers in comparison with the TNT ones can be explained by disturbance of the tubular architecture due to nanotube collapse and recrystallization into bigger nanoparticles. The average size of the formed nanoparticles was estimated to be ~50 nm (Fig. 2a). Meanwhile, larger particles (150−200 nm) and broken parts of the nanotubes were also observed in some areas of the sample (Fig. 2b).
TEM analysis additionally confirmed the TNT-to-NP transformation of titania matrix under limited air access. The tubular morphology was observed for the TNT sample annealed in air flow (Fig. 1d). For the NP sample, nanotubular morphology partially disappeared to form the nanoparticles (Fig. 2d).

XRD analysis and the crystallite orientation study
XRD analysis showed that the as-grown TNT layers are expectedly amorphous. The XRD

Raman spectroscopy study
The anatase phase of the TNT and NP samples was clearly identified by Raman spectroscopy. Both samples exhibit typical Raman bands (see Fig. 4) at 635 cm -1 (Eg vibration mode), 514 cm -1 (A1g mode), 396 cm -1 (B1g mode) and 198 cm -1 (Eg mode) as well as intensive band at 144 cm -1 (Eg mode) which are characteristic of anatase phase [52]. However, the relative intensities of the Raman bands are different for the TNT and NP layers. The NP sample has a Raman spectrum similar to the standard anatase polycrystalline materials. In case of the TNT sample, intensities of the bands at 396 cm -1 (B1g mode) and 514 cm -1 (A1g mode) are enhanced in comparison with the NP sample. According to the literature data [53], the percentage of anatase [001] exposed facets can be determined from the peak intensity ratio of the Eg and A1g modes.
We estimated this parameter using the Eg peak at 144 cm -1 and the A1g peak at 514 cm -1 and found that percent of the [001] facets oriented parallel to the film surface is significantly higher for TNT layers (19 %) as compared with NP ones (5 %).

XPS spectra and chemical composition
Chemical composition of the TNT and NP layers has been studied by XPS method. The layers were found to consist of titanium (Ti 2p), oxygen (O 1s), carbon (C 1s), fluorine (F 1s) and a negligible amount of nitrogen (N 1s). The Ti/O molar ratio for both samples before Ar + etching was approximately 1:3 due to the presence of surface hydroxyl groups and probably some organic contaminants. After etching with Ar + ions the Ti/O ratio becomes close to the expected stoichiometric value (1:2). Figure 5 demonstrates the high-resolution XPS spectra of the Ti 2p level for TNT and NP samples before Ar + etching. The Ti 2p (3/2) and 2p (1/2) peaks were centered at 458.8 and 464.5 eV, respectively, in accordance with literature data for octahedrally coordinated Ti 4+ in TiO2 [54−56]. Additional XPS peak with a lower binding energy of 457.4 eV (fitting curve in blue) is assigned to Ti 3+ states [57,58]. The area ratio of Ti 3+ to Ti 4+ 2p (1/2) peaks is higher for the NP sample, indicating a larger Ti 3+ state density in the NP sample in comparison with TNT one. The O 1s peaks located at 530.1 eV and 531.7 eV are assigned to lattice oxygen in TiO2 and hydroxyl groups (Ti-OH), respectively [54−56, 59, 60]. Figure 6 shows the high-resolution XPS spectra of the F 1s region. An analysis of this region for TNT and NP samples before and after Ar + etching allows identifying significant difference in fluorine content and contribution of different fluorine states. Before the etching, only one symmetrical F 1s peak at 685.3 eV was revealed for both TNT and NP samples. The F 1s XPS spectrum of the TNT layer was not markedly changed after etching. The only difference was detected in the relative area of the F 1s peak: it dropped significantly (the atomic concentration of fluorine decreases from 0.9 at.% to 0.25 at.%). In contrast, the spectrum of the NP sample after etching was clearly composed of two contributions: a main peak at 685.3 eV and a smaller peak at 687.1 eV. The main peak located at 685.3 eV can be assigned to the surface fluoride formed by ligand exchange between Fand surface hydroxyl groups on TiO2 [61−63] or to F atoms of TiOF2 [41,64]. The peak at higher binding energy (687.1 eV) can be attributed to substitutional F atoms in the TiO2 lattice [41,60,61,64]. The atomic concentration of fluorine in the NP sample before and after Ar + etching was 0.8 at.% and 2 at.%, respectively.
Additionally conducted EDS mapping revealed homogeneous distribution of fluorine on the surface of the TNT and NP layers.
It is worth mentioning that the fluorine concentration after Ar + etching was reduced by a transfer, both NH4F and the products of its decomposition (for example, HF) react with amorphous TiO2, producing complex species such as TiF6 2- [65]. Then, thermal decomposition of these complexes can give crystalline TiO2 nanoparticles.

Cyclic voltammetry measurements in Ar-saturated solutions
Electronic properties of the TNT and NP electrodes in the absence of faradaic processes were studied by cyclic voltammetry in deoxygenated alkaline solution. Figure 7 shows the cyclic voltammograms (CVs) of the electrodes in an Ar-saturated 0.1 M KOH electrolyte. At potentials less than ca. -0.9 V an exponential rise of the cathodic current is observed. This electrochemical process can be related to electron accumulation within TiO2 film coupled to proton uptake from electrolyte for charge compensation. As a result, electrochemical reductive doping of the TiO2 electrodes takes place. This reaction is reversible and an anodic current is registered when the potential is scanned in the positive direction.
Apart from the currents at E < -0.9 V, a pair of cathodic peak and related significantly broader anodic one is observed at CVs in the range from -0.4 to -0.9 V for both TNT and NP electrodes. The observed peaks are characteristic of nanostructured titania electrodes and can be attributed to filling/depopulation of deep traps located at grain boundaries [66]. Figures 8a and   8b show cyclic voltammograms of TNT and NP electrodes recorded in an Ar-saturated 0.1 M KOH solution at various scan rates in the range up to -0.9 V. The potential of the cathodic peak is slightly shifted in the negative direction when the scan rate increases. As shown in Figures 8c   and 8d, the cathodic peak current varies linearly with scan rate, indicating that there are no diffusion limitations for this process. We estimated the charge (Qtr) corresponding to the cathodic peak and found that its value varies insignificantly with the scan rate and is close to the charge of the coupled broad anodic peak. It is significant that the cathodic peak assigned to deep traps is essentially higher for NP electrodes than for TNT ones.
The value of Qtr is

Oxygen electroreduction reaction on TNT and NP electrodes
The effect of the fluoride-mediated TNT-to-NP transformation and related fluorine doping of TiO2 on its electrocatalytic activity was studied on an example of oxygen electroreduction reaction, since this process is very important in various applications. Figure 9 displays the CV curves recorded in oxygen-saturated alkaline solution for both TNT and NP electrodes. The voltammetric response of the TNT is characterized by two well defined waves at potentials more negative than -0.7 V. It is noteworthy that the second wave of cathodic current at -0.9 V (vs. Hg/HgO/1M KOH) was revealed only for highly ordered titania nanotubular layers [47,49] and is not typical for other compact polycrystalline films [67,68] and single crystals of TiO2 [69].
The origin of the second wave of oxygen electroreduction on the TNT electrodes is still not clear and requires additional investigation. Characteristically this wave almost disappears after transformation of the TNT layers to the NP ones (Fig. 9).
As seen from Figure 9, the half-wave potential of the oxygen reduction reaction on the NP measurements [62]. This additional Ti 3+ generation owing to Fdoping of the TiO2 structure can lead to enhancing ORR activity and, as a result, to a decrease of the ORR overpotential for NP electrodes.
The CV measurements at the TiO2 electrodes in deaerated electrolytes demonstrate that some surface species are involved in a reversible electron transfer in the same potential region where the irreversible electroreduction of O2 takes place (Figs. 8 and 9). The concentration of these species increases significantly after transformation of nanotubes to nanoparticles. We can suggest that these species mediate the electroreduction of oxygen on the titania surface.