In situ determination of the water condensation mechanisms on superhydrophobic and superhydrophilic titanium dioxide nanotubes

One-dimensional (1D) nanostructured surfaces based on high-density arrays of nanowires and nanotubes of photoactive titanium dioxide (TiO2) present a tunable wetting behavior from superhydrophobic to superhydrophilic states. These situations are depicted in a reversible way by simply irradiating with ultraviolet light (superhydrophobic to superhydrophilic) and storage in dark. In this article, we combine in-situ Environmental Scanning Electron Page 1 of 24 ACS Paragon Plus Environment Langmuir 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Espinos, 1 Jordi Fraxedas, 2 Virginia Perez-Dieste, 3 Carlos Escudero, 3 Agustin R. Gonzalez-Elipe 1  ABSTRACT. One-dimensional (1D) nanostructured surfaces based on high-density arrays of nanowires and nanotubes of photoactive titanium dioxide (TiO2) present a tunable wetting behavior from superhydrophobic to superhydrophilic states. These situations are depicted in a reversible way by simply irradiating with ultraviolet light (superhydrophobic to superhydrophilic) and storage in dark. In this article, we combine in-situ Environmental Scanning Electron Introduction Smart surfaces, equipment for power generation or lab-on-a-chip devices are examples of industrial applications where vapor condensation plays a critical role. [1][2][3][4] Implications are presented, for example, in the fabrication of water harvesting materials and heat and mass transfer performance of surfaces [3,5] controlled by either filmwise or dropwise condensation. In these and other systems, particularly when involving elements at the nanoscale, the wetting behavior of micro-and nano-structured surfaces and the condensation regimes onto them is yet not fully understood. [3,6,7] Particularly, there is a significant gap in the description of the water condensation at the nano and microscales and the eventual wetting behavior (and apparent contact angle) at meso and macroscales. [7][8] Nanostructured surfaces with a dual-scaled roughness described by a Cassie-Baxter state [9] are known to depict a dropwise condensation, where droplets with typical radius below 10 μm can be easily removed through a self-propelling mechanism. [10][11][12] Meanwhile, a detriment in water repellency is found when a high density of surface sites induces water coalescence and permeation in a partial Wenzel state. [11] In more general terms, wetting states of nanostructured systems have been also described as a modified Cassie-Baxter state, although the rich phenomenology encountered in these systems is far from being well understood. [13][14][15][16] Specially critical because of their applications in transport and architecture is the fabrication of antifogging, self-cleaning and anti-freezing surfaces [16][17][18][19][20] and the elucidation of the relationship between hydrophobicity/hydrophilicity and freezing-delaying and anti-freezing behaviors. [20][21][22] Advances in the liquid manipulation at small scales such as in microfluidic devices and for microdroplets transferring are also conditioned to the condensation and wetting mechanisms. [3,14,16,[23][24][25][26] On the other hand, the elucidation of the mechanism responsible for the superhydrophobic to superhydrophilic conversion in semiconducting oxide surfaces offers an ideal situation to advance in the understanding of the water condensation behavior and photoactivation of nanostructured surfaces. TiO2 constitutes the paradigm of photoactive oxide which, widely used for the development of photo-catalysts, biomaterials and self-cleaning surfaces, [23,24,[27][28][29][30][31][32] experiences upon UV irradiation a transformation from hydrophobic to hydrophilic without any apparent modification of surface morphology.
Within such a general context, we aim herein to unravel the factors controlling the wetting and water condensation phenomena on nanostructured surfaces at both the microscopic and molecular levels through an unprecedented "in situ" combined approach by Environmental Scanning Electron Microscopy (ESEM) and Near Ambient Pressure Photoemission (NAPP) using as one dimensional photoactive model a surface formed by hierarchical TiO2 domednanotubes (NTs) presenting porous tailor-made walls. [33][34]

Results and discussion
TiO2 domed-nanotubes shown in Figure 1 were fabricated by a full vacuum methodology providing hierarchical 1D microstructure with walls formed by the agglomeration of small nanocolumns. The system presents a singular roughness resulting from their characteristic micro-(nanotubes length: in the range between 2 and 5 um) and nano-scales (nanotube surface: formed by columns with thickness/length below 20/100 nm), as well as a high micro-and mesoporosity. [33][34][35] TiO2 NTs surface just after UV light pre-irradiation under room conditions for the indicated times and latter recovery after storage in dark for one day. Insets present photographs of the water droplets in the initial (top) state and after long UV irradiation (bottom).
In a conventional wetting test, macroscopic sessile droplets placed on the NTs surface under ambient conditions (Fig. 1 a), b)) reveal a superhydrophobic behavior, characterized by water contact angles (WCA) higher than 150º, low contact angle hysteresis (<10º) and a small sliding angle (<2º). In good agreement with previous reports on one-dimensional photoactive surfaces, [13,23,24,[35][36][37][38][39][40] a reversible change from this initial superhydrophobic situation to a superhydrophilic (WCA<10º) state occurs after UV light irradiation of the surface for a time span of several minutes. Figure 1 f) shows the time dependence of WCA for the same TiO2 NTs surface when subjected to pre-irradiation for increasing periods of time. The obtained decaying curve describes the transition with water droplets bouncing on the surface without any adhesion, to water droplets spreading onto the surface just after a few minutes of pre-irradiation. This transformation of the surface state is reversible just by keeping samples stored in the dark over one day. The initial superhydrophobic state agrees with a metastable Cassie state where the apparent contact angle (*) of a surface presenting chemical heterogeneity (in this case TiO2 and air filling the pores and gaps between NTs) can be calculated by the Caxie-Baxter [9,13,30]  for * is 171.5º very closed to the 172º measured by means of the sessile droplet method ( Figure   1 f). This behavior is typically found on dual-scaled roughness surfaces. However, for the same surface topography, UV illumination under room conditions leads to a superhydrophilic surface, typical of a wetted Wenzel state. [13,41] To reach a deeper understanding of this switchable wetting behavior of the TiO2 NTs and the relationship with the corresponding condensation regimes (dropwise or filmwise), we have carried out a thorough "in situ" characterization study by ESEM for the situations referred as "superhydrophobic" and "superhydrophilic", i.e. comparing asgrown and UV-irradiated nanotubes. Figure 2 gathers micrographs corresponding to normal and cross sectional views of the two wetting states upon progressive condensation of water at low sample temperatures, when the relative humidity of the microscope stage is increased up to induce condensation. For the as-grown NTs (Figure 2 a)-d), i)), when the water vapor pressure is raised close to the relative level of humidity saturation (860Pa at 4ºC), individual water droplets of diameters smaller than 5 m appear condensed onto the NTs surface (Figure 2 b)). These drops define a minimal pinning of contact line typical of a dropwise condensation behavior. [10][11][12] Cross section micrographs confirm the formation of isolated droplets on the superhydrophobic TiO2 NTs surface (Figure 2 c)) due to the hindrance to water spreading depicted by this surface.
For small volumes, these microdroplets present a quite spherical shape and grow contacting one or several NTs. In fact, the high resolution image in Figure 2 i) shows the nanotubes immersed within the droplets which indicates that for small droplets the wetting mechanism under condensation follows the Wenzel's law. Droplets sizes increase during the condensation process until they collapse into larger droplets that tend to cover completely the surface. This process must be accounted for by the natural decrease of the ratio between surface and bulk free energies when the nuclei size increases. [42] We have estimated the contact angle for several droplets in directly affect the applications of this type of nanostructured material as, for instance, to enhance the mass or heat transfer. [3,5] As a matter of fact, this type of surface will not allow the spontaneous ejection of droplets during condensation with the consequent detriment in the heat transfer process.
After UV light irradiation for 10 minutes, micrographs in Figure  condensation starts at the bottom of the nanowires in a filmwise way, [10][11][12] in clear contrast with the water droplet nucleation on the NTs found for the as-grown surface. Eventually, upon increasing the water pressure in the chamber, the accumulation of water within the superhydrophilic nanostructure continues up to the complete flooding of the NTs (Figure 2 h)).  Another critical issue is the elucidation of the wetting mechanism at the NT-surface, provided that the ESEM resolution under high humidity conditions is not sufficient to assess whether the observed dropwise/filmwise condensation is preceded by a water film or droplet formation at the surface of the NTs or within the porous presenting in the NTs walls (see the columnar nanostructure of the NT wall surfaces in Fig. 1). In addition, in this article we aim to contribute with new experimental insights to solve the controversy generated about the origin of the photoinduced hydrophilicity on metal oxide semiconductors such as TiO2 and ZnO. Just after the phenomenon was revealed for both anatase and rutile surfaces, Wang et al. [27][28][29] proposed that UV illumination under ambient conditions induces the formation of oxygen vacancies leading to the dissociation of water to form OH groups responsible for the hydrophilic character. A few years later, it was postulated that the UV modification of the TiO2 surface followed by water adsorption produced the changes of the OH groups from a twofold binding configuration to onefold direct binding to the Ti sites. [43] However, posterior studies carried out for pristine TiO2 surfaces under controlled atmosphere (usually in O2) indicated that UV irradiation has no influence in the uptake of OH groups. [44,45] Within this latter group, are noteworthy the results by Yates et al. on single crystal rutile combining Auger Spectroscopy and Scanning Tunnel Microscopy (STM) studies. [44] These authors concluded that poisoning with carbon residues was the main cause for the increase in surface contact angle on the TiO2 surface and that the UV driven photo-oxidation of these pollutants was the responsible for the hydrophilic conversion.
Herein, trying to get a deeper insight at nano-and molecular-scales we have carried out Near Ambient Pressure Photoemission (NAPP) analyses to in situ follow the surface chemical evolution during the initial states of water condensation. Figure 3 gathers the corresponding C1s peaks recorded under UHV conditions for the as-grown (i.e. characterized by a superhydrophobic water contact angle of sessile drops and dropwise condensation in Wenzel state) and UV treated samples (i.e. presenting superhydrophilic wetting and filmwise condensation) (see also Table S1 showing C, O and Ti atomic percentages and ratios). The different fitting bands in these C1s spectra can be attributed to -C-C/-C-H (BE ~284.5eV), C-OH/-C-O-C-(BE ~286.0eV), C=O (BE ~287.5eV) and O=C-O/CO3 = (BE ~288.5eV) species. [46] According to these spectra and surface atomic compositions both surfaces present carbonaceous species, with an approximate ratio C/Ti higher in the irradiated sample (0.27) than in the as-  [44] however, it is important to stress a significant difference: we irradiated the NT sample in air, conditions which are known to generate amphiphilic surfaces with low contact angles for both polar and non-polar species. [28] Therefore, we attribute the increment in the C/Ti ratio to the adsorption of carbonaceous compounds during the time of UV irradiation of the TiO2 surfaces.
In this regard, it is worth stressing that the original non-irradiated surface state of the TiO2 surface exhibits a higher [O=C-O/CO3 = ]/[C-C] but a lower [-C-OH]/[C-C] ratio than that after UV irradiation. This evolution can be accounted for by admitting that besides the UV activated adsorption of carbonaceous species from the air, these latter are also progressively photooxidized as proposed in the bottom panel of Figure 3.  Table S1).  Figure S2). It is noteworthy that for both samples the measured Ti/C ratio progressively decreased with the vapor pressure, suggesting the preferential coverage of the Ti sites with OH groups and/or water molecules as explained below. The initial O1s spectra can be described by the contribution of two signals attributed to OHgroups (binding energy BE~531.2eV, note also the contribution from -COx groups depicted in orange color and estimated according to the procedure described in the experimental section) and O = ions (BE ~529.6eV) bonded to Ti 4+ cations at the surface [17] (this latter used for intensity normalization). The OHsignal in the UV pre-illuminated surface has a higher intensity, a feature agreeing with the model by Wang et al. on the surface activation of TiO2 surfaces UV irradiated in air. [28] As-grown surface also presents OHgroups likely formed after exposure to ambient conditions prior its insertion in the analysis chamber. For the dropwise condensation regime state, dosing 10 Pa into the chamber (Figure 4 b)) leads to a high increase in the intensity of the OHcomponent (OH -/O = ratio increased from 0.5 to 1.3) and to the development of two new bands at 532.8 eV (H2Oads) due to adsorbed water and at 535.3 eV (H2Og) corresponding to water in the gas phase. [45] These two latter bands grow continuously in intensity when water vapor pressure increases. Unlike the behavior of this superhydrophobic state as-grown surface, in the pre-irradiated sample, the H2Oads band was smaller after dosing 10 Pa of water vapor and grew much slower with the dose at least during these initials stages of the experiment (Figure 4 f)-h)). It is also worth stressing that the OHband only increases moderately up to its saturation (OH -/O = ratio increased from 0.8 to 1.2). We tentatively link this different evolution of XPS signals in this NAPP experiment to the formation of very small water droplets on the hydrophobic as-grown NTs surface, and to the condensation of liquid water at the bottom of NTs and within the tube cavities when the sample is in the superhydrophilic state after UV irradiation. Such a condensation behavior at nanometric scale corresponds to the observations in Figure 2 when either larger water droplets or layers were observed by SEM.  According to the premises of the Tougaard analysis, [47] this hypothesis also agrees with the observation that the background slope in the Ti2p survey XPS spectra ( Figure S1 Taking together all these experimental evidences, Figure 4 bottom) shows a schematic representation of the wetting mechanism for the as-grown and irradiated NTs surfaces at the initial stages of water condensation. According to this scheme, upon water vapor exposure at 10 Pa, hydroxyl groups extensively accumulate on the surface of the as-grown sample, where they promote water molecules adsorption and condensation in the form of small droplets at the NTs tips and lateral surfaces. Following the ideas deduced from the analysis of Figure 2, these droplets are likely in a Wenzel state, where condensed water would also partially cover the gaps between the nanocolumns in the NTs shells (no shown in the schematic on the sake of simplicity). The progressive size increase of these droplets would be the responsible for the observed attenuation of the Ti2p signal and the increase in the H2Oads band. Onto the superhydrophilic surface, the concentration of hydroxyl groups is already quite significant before water exposure, and condensation of water droplets does not occur at the tip surface zones of the NTs (i.e., condensed water is then undetected by XPS) but in the interior of the tubes and in the intercolumnar space in the NTs-shells as well as in the gaps between NTs. In previous articles, we demonstrated that the equivalent columnar thin film layer (see Figure S5) presents a pore size distribution ranging from micro (pore diameter below 2 nm, within the columns) to mesoporous (diameter between 2 and 50 nm, within and between the columns). [35] Columns mean diameter in the NTs walls is smaller than that measured in the equivalent thin film, therefore, we can assume that the pore size distribution corresponding to the NTs walls microstructure is also in the range of the micro and mesopores.
Thus, the presence of condensed water in the micro and mesopores in the NTs shells and interior of the NTs at this relative low water pressure is in good agreement with the premises of the inverse Kelvin effect [48][49][50] which predicts a lower saturation water vapor pressure required for water condensation on porous hydrophilic structures with respect to plane and compact surfaces. In consequence, high water coverage of the NTs is reached for the highest pressure situation, i.e.
350Pa at low temperature, for hydrophobic and superhydrophilic states (Figure 4 d) and h)). These peak. [51] For the as grown TiO2 NTs, the OHbands after water vapor exposure (Figure 4 b) and f)) detected in our experiment are broader than those corresponding to the vacuum conditions and shifted by +1.6 eV with respect to the O = band. This fact, joined to the slightly higher width of the OH band in the superhydrophilic state (1.78eV against 1.66eV for the as-grown surface), points to a heterogeneous surface hydroxylation which, in the irradiated sample, will count with a higher contribution of acid groups (see schematics in Figure 4 Bottom). These acidic groups would favor the diffusion of water molecules that, in this way, would condense at the bottom of the NTs. The more basic OH groups in the as-grown NTs would not promote water diffusion to the interior of NTs and therefore would condense in the form of droplets at the tips surface.

Conclusions
Taking into account the results and recent precedents in the literature, we propose the processes gather in Figure 4 i) and j) as responsible for the wetting of as-grown and preirradiated nanotubes correspondently. The main mechanistic assumption in the scheme in Figure   4 i) and j) is that water adsorption is different on the hydrophobic and hydrophilic surface state not only because this last contains a higher concentration of hydroxyl groups, but also due to a different surface arrangement of O = and Ti 4+ species at the surface. [52] It is known that water dissociative adsorption onto the oxygen terminated TiO2 (001) surface plane requires an energy of 184.7 kJ/mol, which is higher than that needed for hydroxylation of other surface planes containing both Ti and O ions (e.g. lower than 70kJ/mol for TiO2 (100) and (101)). Nevertheless, the previous hydration of the surface decreases the adsorption energy by more than 45%. In this way, we think that in our NT system a hydroxylated surface due to UV preirradation is less reactive upon low vapor pressure water exposure and therefore less prone to highly increase its surface hydroxylation as it happens with the hydrophobic state (note however that almost equivalent hydroxylation levels are found at higher vapor pressures, cf. Figure 4 d) and f)). On this superhydrophilic surface, water molecules tend to strongly interact through the formation of hydrogen bonds, this leading to its diffusion to the interior of the NTs cavities and voids where, favored by capillary and inverse Kelvin effects, it starts to condense even at the lowest vapor pressures of our experiment (Figure 4j)). In this regard, it is worth mentioning that previous experiments with metals [53] and single rutile crystal surfaces [54] have confirmed the formation of one-dimensional H-bonded water chains governing by diffusion and adsorption of OH groups and water molecules onto pentacoordinated Ti surface sites, a situation which is compatible with a superhydrophilic state. Therefore, without discarding a contribution according to the suggestion of Yates et al. in the sense that carbonaceous species may contribute to increase the wetting contact angle of water on pristine, non-irradiated, TiO2, [44] our "in situ" experiments with photoactive oxide NTs support the model of Watanabe et al. and other authors [28,44,55] in the sense that the dissociation of water molecules from air is the main factor contributing to enhance hydrophilicity. This conclusion is also extensible to the similar behavior of other photoactive 1D nanostructured surfaces, like supported ZnO nanowires. [36][37] Our findings also support a revised  [56] as well as to the industrial fabrication of anti-freezing surfaces. [19] EXPERIMENTAL METHODS Sample preparation: Vertically oriented domed-TiO2 nanotubes samples, with typical individual thicknesses in the range of 200-500 nm and a length of several microns, were fabricated on Si(100) substrates by a full vacuum and plasma deposition method carried out at mild temperatures. [33,34] The samples were handled in air after their preparation and stored in a C1s spectra was carried out by using Gaussian-Lorentzian bands after subtraction of Shirley-type backgrounds. For this fitting analysis the contribution of oxygen atoms bonded to carbon (i.e., COx species) was determined from the intensity of oxidized carbon groups in the C1s spectrum. The difference in the cross section of C1s and O1s signals, as well as the atomic ratios between oxygen and carbon in the different species has been taken into account to estimate the intensity of this oxygen band. Since the position of this small intensity band was left free during fitting small shifts in BE position were observed, although these changes did not affect the main feature of the experiment related to the increase in the intensity of OHbands upon water adsorption.
Supporting Information. Additional experimental details and XPS characterization.
The following files are available free of charge. SupportingInformation.pdf

Author Contributions
The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.