Study of Self-Sensitization of Wide-Gap Oxides Photocatalysts

3.1. The First Steps of Photo-Mass Spectrometry A pressure change above the sample is a necessary but not sufficient indicator of photoinduced adsorption/desorption effects. Such pressure shift can be caused by photochemical reactions (for example, oxidation of adsorbed contaminants) in the system under study. Therefore, to prove the photosorption effect, it is necessary to show that the system under study (for example, O₂-ZnO, O₂-TiO₂) is truly binary, i.e., it does not contain extraneous components. It is possible only if there is a reliable universal method to analyze the composition of both the gas mixture above the sample and the adsorbed layer. Replacing the integral pressure gauge with a mass spectrometer [10] allowed one to solve fundamentally new problems: besides a quantitative analysis of the gas mixture, study the gas phase composition above the adsorbent. In the preliminary experiments [11], the primary products of the photo-desorption were studied. For this purpose, the sample was placed in the chamber of a time-of-flight mass spectrometer at a distance of 30 mm from the ion source in a vacuum of 2 × 10⁻⁸ Torr and was illuminated with short 10⁻⁶ sec light pulses. Photo-desorbed particles entered the ion source “on a straight run” without collisions with other particles or the device walls. In this way, primary “hot” molecules and radicals were registered. Dispersed ZnO was studied. The illumination of the sample after oxygen adsorption led to photo-desorption of multiple products such as H₂O, OH, CO, and CO₂, but the molecular oxygen content did not exceed several percent, indicating the sample contamination with residual vacuum molecules. Attempts to obtain and maintain a clean surface for studying photo-activated oxygen reactions were unsuccessful. However, it was possible, in particular, to observe OH radicals photo-desorption after H₂O adsorption. In order to work with a thoroughly cleaned sample, another setup was designed [12]. The dispersed sample was deposited as a thin layer in water bidistillate to the quartz reactor walls. After pumping down to 10⁻⁸ Torr, the sample was heated at 800 K in a flow of pure (99.99) oxygen at 10⁻¹ Torr for tens of hours until the complete disappearance of CO and CO₂ traces in the evacuated gas. The analysis was performed with a magnetic mass spectrometer. The TDS (thermo-desorption spectroscopy) analysis of the adsorbed phase composition revealed the contaminants at the level of 10⁻⁵ monolayers. This installation enabled the study of photoadsorption and photodesorption in the O₂-ZnO system with an impurity content in the gas phase lower than 0.1%. The use of the ¹⁸O isotope revealed the dynamic nature of photosorption processes: manometrically observed pressure changes above the sample are the sum of coexisting photoadsorption from the gas phase and photo-desorption from the sample. These experiments pioneered research in photo-activated isotope exchange of oxygen over oxide photocatalysts. It is obvious that for understanding the mechanism of the photons influence on such systems, a comprehensive analysis of photo-adsorption, photodesorption, and photo-activated processes in the gas and adsorbed phases is required. Consequently, a new generation experimental setup should provide precise control over experimental conditions, allow for investigation processes involving multicomponent mixtures, and analyze the reaction kinetics for each component; using stable isotopes, separate competing terms of the photoactivated process in a binary system (for example, photo-adsorption and photo-desorption). 3.2. Experimental Technique Extensive photo-manometric studies have demonstrated that inconsistent and irreproducible experimental results were caused by insufficient sample characterization and experimental purity. Therefore, special attention was paid to address these shortcomings. The XPS, X-ray diffraction (XRD), and scanning electron microscopy (SEM) methods were used to characterize the samples before and after the experiments. However, the sensitivity of these methods does not exceed 10⁻³ monolayers, while the number of active centers in self-sensitized oxides does not exceed 10⁻⁵ monolayers. The following methods (see Figure 1): mass-spectrometry, UV (8.43 eV) photoelectron spectroscopy (PES), optical (diffusion reflectance) spectroscopy (DRS), TDS, spectra of excitation and emission of thermoactivated luminescence (TSL), and electron spin resonance spectroscopy (ESR) have been adapted for in situ investigations in three phases: gas, adsorbate and solid state [13,14,15]. The reaction block was made of quartz and stainless steel, without lubricants, taking into account the requirements for UHV installations. Dispersed powders of nominally pure oxides were deposited on the lower wall of a quartz reactor (V = 65 cm³) from the suspension in bidistilled water and preliminary purified by heating in oxygen (99.99%) flow at the pressure of 0.5 Torr for 10–20 h at 870 K in order to remove the biographic pollutants. The criterion of the sample purity was the absence of CO and CO₂ in the output oxygen flow registered by the mass spectrometer. The “reduced” samples were obtained by heating in ultrahigh vacuum (better than 10⁻⁸ Torr), and the “oxidized” samples were obtained by annealing in oxygen (99.99%). All measurements were performed at 290 K except for specially marked cases. UV (8.43 eV) PES is the photoelectron spectrometer with the Xe resonance source [14], Diff.reflect. 200 < λ < 2500 nm is the spectrophotometer for the UV-VIS-NIR range [15], Thermo Stimul.Lumin. is the spectrometer for the analysis of thermoactivated luminescence excitation and emission spectra. Thermo Desorb. Spectroscopy is realized in “flow-through” regime. Monochromator is of original construction with aperture 1:1.6 for a spectral range of 200 < λ < 800 nm [13], ESR is the electron spin resonance spectrometer [16]; Mass Spectr. is the mass-spectrometer; UHV system is the Ultra High Vacuum system. T = T₀ + βt is the system for variation of the sample temperature with a constant rate 0.1 < β < 0.5 K/s in the range of 300 < T < 1000 K for TDS and TSL measurements. L₀ is the initial gas flow, ΔL₁ is the gas flow from the sample due to photodesorption, ΔL₂ is the gas flow onto the sample due to photoadsorption. The original UV (8.43 eV) photoelectron spectrometer [13] is specialized for the in situ studies of extremely low concentrations of filled electronic states in the bandgap. Special attention was paid to obtaining and maintaining a clean surface of the sample [17]. The quality of cleaning the oxide surfaces was controlled using TDS spectra. In the quasi-stationary mode (L₀ = 0), the actual sensitivity for recording the process speed dn/dt of about (10⁻⁷−10⁻⁸) Torr/s is given by the expression $$-\frac{dn}{dt}\approx\frac{1}{A}\frac{\Delta N}{t_p}$$ where ΔN is the change in the number of molecules caused by illumination, t_p being the characteristic pump-off time of a mass-spectrometer (m-s). A typical pressure sensitivity in photo-manometric experiments in a closed volume of 20–50 cm³ was ≥10⁻⁵ Torr/s at the initial pressure of ~10⁻² Torr. 3.3. Flow-through m-s Method The next step was the development of a flow-through m-s method for studying photo processes [13]. In a quasi-stationary mode, gas sampling for mass analysis led to a drop in pressure in the reactor and thus limited the range of working pressures available for analysis. However, in a flow-through reactor, the gas sampling with a mass spectrometer is compensated for by an additional influx. Thus, the lower pressure limit, which is 10⁻³ Torr for a quasi-stationary reactor, is reduced to 10⁻⁶ Torr. In a flow-through reactor (Figure 1) containing an adsorbent with a gas phase above it, the changes in concentration of the adsorbed phase dn/dt mol·cm²·s⁻¹ lead to changes in the gas phase $$-\frac{dn}{dt}=\frac{1}{A}\left(\frac{d\Delta N}{dt}\right)+\frac{\Delta N}{t_{_p}}$$. The actual sensitivity for measuring the process rate in flow-through mode reaches 10⁻¹⁰ Torr/s or 10⁹ mol/s. This enhanced sensitivity allows one to avoid spherical and cylindrical reactors (black body-like) [14] and use flat reactors, allowing one to connect additional in situ devices. The sample thickness is optimized based on forward and backward scattering indicatrix analysis. As an example, Figure 2 shows the measurement results of the “Degussa” TiO₂ sample. Figure 2 shows Scattering diagrams of the exciting beam radiation incident normally on the “Degussa” TiO₂ photocatalyst. The curves, close in shape to circles, join the ends of the radius vectors, characterizing the scattering amplitudes of a light beam at a given angle in the horizontal plane. The differences in the diameters of the circles reflect the spectral characteristics of the scattering. In the diagram, the light is incident from the left, so the intensity of backward scattering is greater than that of forward scattering after passing the sample. Similar measurements are taken in the vertical plane. As a result, three-dimensional figures close to spheres are obtained. They give the integrated intensities of light scattered into the front and rear hemispheres. The difference between the intensity of incident and total scattered light allows one to obtain the number of light quanta absorbed by the sample. This procedure allows us to obtain quantitative characteristics of photoactivated processes without resorting to the geometry of a “black body” [14]. In a “black body” geometry, the sample is placed inside a sphere and is accessible only for manometric measurements. We used a flat reactor that allowed for the first time quantitative photo-mass spectrometric studies to be combined in situ with a range of complementary methods (Figure 1). Data processing using the Kubelka-Munk model for the layer thickness of 2.5 mg/cm² yielded the following flux values passing through the layer: 0.2 for λ = 436; 0.026 for λ = 385 and 0.0024 for λ = 365 nm, respectively. The data obtained allow one to optimize the choice of the sample thickness in photocatalytic experiments and to avoid a thick, non-illuminated sample layer. Obviously, comprehensive information about multi-stage photo processing can be obtained only by total quantitative studies of the processes in the gas phase, on the surface, and in the bulk of the sample. The experimental complex shown in Figure 1 is designed for this purpose. 3.4. Spectro-Kinetic Investigation of Photosorption Processes in Flow-through Reactor 3.4.1. Gas-Phase In the experiments, a pure (99.99%) oxygen flow was admitted over the carefully cleaned sample under the controllable dynamical pressure of 10⁻⁸–10⁻³ Torr. The sample was irradiated with a monochromatic light tuned in the spectral range of 200 < λ < 600 nm (6 > hν > 2 eV) with a power density of ~1 × 10⁻³–1 × 10⁻¹ W/cm² [18]. A high sensitivity of the “flow-through” regime (sensitivity for the photo-desorption detection from powders which reached 10⁻⁷ monolayer/s), exceeding the photo-manometric method sensitivity by 3–4 orders of magnitude, enables novel avenues for investigating photosorption processes. It allows the determination of spectral characteristics of photo-processes using monochromatic radiation for excitation. An example of determining kinetic parameters in the O₂-Al₂O₃ system is given in Figure 3. In order to determine the quantum yield of a reaction, the numbers of quanta incident on the sample, reflected from the sample and having passed through the sample, were measured using the spectro-goniometer (Figure 2). The quantum yield for photoadsorption Ф_pa or photodesorption Ф_pd is $$\Phi=\frac{dn}{dt}\frac{1}{k_\lambda I_{h\nu}}$$ where k_λ is the absorption coefficient of the sample, and I_hν is the light intensity. A system [O₂–Al₂O₃] is shown in Figure 3. The illumination shifts sharply the initial dynamical gas-surface equilibrium [19]. The sign of ΔL depends on the spectral region of the light. So, for O₂–Al₂O₃ system in the high-energy spectral region (hv > 4 eV), the irreversible photo-adsorption is observed. The opposite effect—a reversible photo-desorption—is observed in the low-energy region (hv < 3 eV). In the middle-energy region, both effects coexist. Both effects are spectrally selective and site- sensitive. For a number of studied oxides, the kinetic parameters of active centers varied in a wide range: 10⁻⁵ < τ < 10³ s, 10⁻¹⁷ < σ < 10⁻¹⁶ cm², 10⁻⁶ < Ф_pa < 10⁻⁴ molec/quantum, 10⁻⁵ < Ф_pd < 10⁻⁴. The first step of photoactivation in the sub-bandgap region is the photoexcitation of local F-type and V-type centers. The spectral maximum of photo-adsorption efficiency on Al₂O₃ is centered at ~230 nm, and has a shoulder at ~260 nm that coincides with the optical absorption of F⁺-centers [18]. The spectral efficiencies of oxygen photodesorption and CO photo-adsorption are shifted to a long-wavelength region in the absorption band of V-centers [18]. So O₂ photodesorption is activated by a photoexcitation of hole V-type centers (not of electron F-type centers). The spectral efficiency of preliminary irradiation in UHV (the “memory effect”) matches the spectral band of photo-adsorption efficiency, suggesting the same nature of short-lived and long-lived centers. The obtained results lead to the following conclusions. The quantum, nonthermic photon-driven adsorption, and desorption are proven to exist for all studied oxides. These processes are site-sensitive and spectrally selective. The kinetic model of the interaction of gas (photo-adsorption) or adsorbed (photo-desorption) molecules with photo-generated local centers is justified. The F-type and V-type centers appear to be responsible for the primary act of photon adsorption/desorption (see Table 1 and Table 2). Similar results were obtained for oxygen photo-adsorbed on wide-bandgap oxides (BeO, MgO) [20,21] and on other oxides (ZnO and TiO₂) [22,23]. The recombination of photo-generated centres, as well as the adsorption of molecules on these centres are the pathways of photo-excited surface relaxation. 3.4.2. Structure of Photoadsorbed Oxygen The structural diversity of photoactivated centers should result in different configurations of molecule adsorption. To verify this hypothesis, the oxygen adsorption configurations on illuminated Al₂O₃ were studied [19]. A combined analysis of TD spectra, thermo-programmed bleaching of optical absorption, TSL, ESR, and isotopic exchange have revealed three oxygen species A, B, C on photoactivated Al₂O₃ surface (Figure 4).



Species A: a molecular peak at 380 K,



Species B: a photoadsorbed O atom in an on-top site above a surface metal vacancy,



Species C: oxygen atoms, filling anion vacancies.

Thus, the formation of species C on the UV-irradiated surface allows the healing of the surface defects of F-type. The species B and C are typical for only UV-irradiated samples. Very similar results were obtained on the oxygen species photoadsorbed on other wide-bandgap oxides. The low-temperature peak at 380 K in the TD spectra corresponds to the molecular species A. Here, the molecule-surface bond is weak, and E_des does not exceed 1 eV. The bond is provided by the charge transfer between an oxygen molecule and surface centers of F_S and/or V_S-type, thus producing O₂ (a) and [O₂ ··· O_S] (b) complexes. It results in the emergence of the absorption band of F⁺ centers at 5.4 eV and of the ESR signal with g = 2.012 ± 0.001. Such a signal indicates a charge transfer between the adsorbed oxygen molecule and the V_S center. Another species B, has a stronger bond between the oxygen and the surface characterized by E_des = 1.2–1.8 eV. It decays in the intermediate temperature range of 450 ≤ T ≤ 600 K, just where V-type centers are decomposed. This species gives the UV absorption band at 3.1 eV and also the ESR signal characterized by two parameters: g┴ = 2.0355 ± 0.0005 and g║ = 2.0028 ± 0.0005. Such ESR signal is typical for a tetrahedral symmetry center. Its thermo-bleaching peak is close to the TSL peak at T_max = 530 K and to the thermo-bleaching peak of the optical absorption band at 2.9 eV. So, we suggest that all three parameters characterize the same center. Its decay in the temperature range of 450–600 K causes the desorption of oxygen molecules. The photo-adsorbed oxygen atom is placed in an on-top site above the surface metal vacancy, which has four surface anions on the basis of the double tetrahedron. The complex is stabilized by a hole captured by one of the basis anions. The hole delocalization results in the decay of the center. The species C (E_des reaching 4 eV) gives neither absorption band in the UV-VIS region (hν < 6 eV), nor any ESR signal. The E_des value is close to that of the thermolysis energy of the surface. This species has been attributed to oxygen atoms filling the anion vacancies. Thus, the formation of species C on the UV irradiated surface allows to heal the surface defects of F-type and, in this way, changes the surface stoichiometry and geometry even at temperatures insufficient to activate bulk atoms. A broad variety of photon-driven effects enables the oxide surface creation without altering bulk properties. Thus, oxygen photoadsorption allows to achieve extremely low oxygen vacancy concentrations below 10⁻⁶ monolayer. Our results demonstrate that oxide photoactivation within the visible light absorption region activates both electron-donating centers and hole centers essential for redox reactions. The oxygen isotope exchange model reaction enables the evaluation of the real possibility of their use. 3.4.3. POIE on ZnO Investigated with Flow-through m-s Method The actual sensitivity for measuring the process rate in flow-through mode reaches 10⁻¹⁰ Torr/s or 10⁹ mol/s. It has been previously shown that the analysis of the kinetics of thermally activated isotopic exchange of oxygen ¹⁸O of the gas phase with the catalyst can be used to predict the activity of oxide catalysts in dark oxidation reactions [15,24,25]. In 1965, the isotope ¹⁸O was used to study the elementary stages of photosorption processes in the O₂-ZnO system [2]. Having studied concentration changes in the mixture of isotopes ¹⁶O₂, ¹⁶O¹⁸O and ¹⁸O₂, the authors concluded that the manometrically observed effects of pressure changes above the illuminated sample (photo-adsorption or photo-desorption) are complex dynamic processes of coexisting adsorption and desorption in the gas-solid system. These processes lead to an isotope content change both in the gas and solid. Subsequent experiments enabled to establish correlations between the activity of photoactivated isotope exchange of oxygen and the efficiency of the oxide photocatalyst in oxidation-reduction reactions. The parameters of the isotope oxygen mixture are as follows:

(1)

$$\alpha=\frac{C_{34}}{2}+C_{36}$$, being the ¹⁸O fraction in the mixture, and

(2)

$$Y{=}2\alpha(1{-}\alpha){-}\mathrm{C}_{34}$$ giving the deviation from the equilibrium value, where C₃₂, C₃₄, C₃₆ are the fractions of ¹⁶O₂, ¹⁶O¹⁸O and ¹⁸O₂, respectively. The mixture with Y ≠ 0 is called “non-equilibrated”.

Two types of oxygen exchange in any oxygen-oxide system can occur:

1.

homoexchange (homomolecular oxygen isotope equilibration POIEq) without the participation of surface atoms; in this case, Y → 0, α_gas remains constant;

2.

heteroexchange POIEx with the participation of surface atoms; Y changes, α_gas decreases if the catalyst has a natural isotopic composition.

Thus, during the POIE, two channels are photo-activated: homo- and hetero-molecular exchanges. The first one simulates the oxidation stages, and the second one is the reduction stages of the oxidation-reduction reaction. The study of a number of wide-band oxides showed the universality of the correlation found. ZnO is the most active (along with TiO₂) photocatalyst for redox reactions due to the features of the electronic structure of ZnO, such as a large band gap (3.37 eV at 300 K), an electron affinity of ~4.0 eV, and a high carrier mobility (200 cm²/V·s at 300 K). In the UV irradiated ZnO, the e⁻/h⁺ pairs have one of the highest redox potentials among the photocatalysts and can be used in photochemical cells for water decomposition and decontamination of the environment from a wide range of organic and inorganic pollutants. However, this efficiency is small and revealed only in the intrinsic absorption region hν < 3.37 eV. The main obstacle in achieving a high photoinduced efficiency upon UV activation semiconductor is the e⁻/h⁺ pair decay when moving from bulk to surface centers of photocatalysis due to the recombination with each other and with the recombination centers. A.N. Terenin pointed out that these losses can be avoided by using excitons to transfer the excitation energy [26]. The exceptionally high exciton binding energy in ZnO (60 meV) ensures the stability of the e⁻/h⁺ pairs during their transfer at room temperature (kT = 27 meV). However, this idea required decades of experimental investigation to be put into practice. In this paper, we give the results of these studies. The activity of excitons in photocatalysis on ZnO was experimentally confirmed in 1990 [27]. The exciton excitation maximum for the POIE reaction on ZnO has been discovered. However, the observed effect was lower than that under the excitation in the interband transitions region. It was found that the main intensity losses were caused by a radiative decay of the exciton on the surface, and the internal energy dissipates in the form of the exciton photoluminescence (PL) with a nearly 100% efficiency. In Terenin’s laboratory, the problem of dissipation of exciton energy not into a photoluminescence quantum but into the generation of active centers of catalysis was solved by creating a surface layered nanostructure ZnO/ZnO_1−x/O⁻ [28]. ZnO/ZnO_1−x is a 2D structure “core-shell” and O⁻ is the adsorbed atomic oxygen capturing an electron, thus forming a locking barrier over the “core-shell” structure. A high field strength inside the well leads to a dissociative polarization of the exciton, causing its nonradiative decay (Figure 5). As a result, a “dead zone” for the exciton is created in the surface subatomic region of ZnO_1−x/O⁻. The exciton decays and a free electron and a hole are localized on F- and V-type defects of the ZnO_1−x surface layer (see Figure 6). A comprehensive description of the nanostructure and its mechanism of action is given in [28,29,30]. The energy of the absorbed light was retained for up to 8 × 10³ s, which is typical for F- and V-type centers. It is these centers that sensitize the photocatalytic activity of reduced ZnO_1−x samples [31,32]. The 2D structure ZnO/ZnO_1−x/O⁻ was formed using ZnO wurtzite structure consisting of 20–500 nm nanocrystallites [9]. The samples were characterized using optical and photo-luminescence spectroscopy (PL), TSL in the ultraviolet (UV) and visible (VIS) light, SEM, UPS (21.4, 8.43 eV), and XRD. UPS (8.43 eV) allowed one to determine in situ the φ_T of the sample and to separate the contributions of the dipole component δ from the band bending V_S. 3.4.4. Modeling of the Kinetics of POIE O₂ ⇄ ZnO/ZnO_1−x/O⁻ Figure 7 clearly illustrates the superiority of exciton excitation compared to interband excitation. The underlying process can be described mathematically using a vector-like equation, as detailed in reference [31]. The radiative decay of an exciton is suppressed in the 2D structure ZnO/ZnO_1−x/O⁻, where the exciton nonradiative decays into a pair of long-lived (up to 8 × 10³ seconds) electron and hole localized states that facilitate chemical reactions. In this context, the notations in the governing equation are as follows: the components of the vector $$\vec{N}$$ represent the number of O₂ isotope molecules in the reactor with masses 32, 34, and 36 atomic mass units. The module of this vector is defined as the sum of its components, i.e., The parameter α is defined by the ratio of isotope molecules in the gas phase $$\alpha = \frac{N_{36} + 0 . 5 \cdot N_{34}}{\left|\overset{\rightarrow}{N}\right|}$$and indicates the fraction of a specific isotope, while $$\overset{-}{\alpha} \equiv \left(1 - \alpha\right)$$ represents the fraction of ¹⁶O. The rate of homo-exchange, denoted as $$R_{x}$$, is measured in molecules per second ([molec/s]), and the rate of hetero-exchange, $$R_{q}$$, is measured in atoms per second ([atoms/s]). The outflow rate, $$\overset{\rightarrow}{F}_{o u t} = \frac{\overset{\rightarrow}{N}}{\tau_{p u m p}}$$, corresponds to the number of molecules captured by mass spectrometry during the reactor’s pump-down time constant ([molec/s]), and $$\overset{\rightarrow}{F}_{i n}$$ is the inflow rate set during the experiment ([molec/s]). All these values ($$\overset{\rightarrow}{N}$$,$$R_{x}$$, $$R_{q}$$, $$\overset{\rightarrow}{F}_{o u t}$$, $$\overset{\rightarrow}{F}_{i n}$$, and RPA) depend on time. In Equation (1), the first two terms on the right side, $$\overset{\rightarrow}{F}_{i n} - \overset{\rightarrow}{F}_{o u t}$$, describe the flow regime. The third term, $$- R_{P A} \frac{\overset{\rightarrow}{N}}{\left|\overset{\rightarrow}{N}\right|}$$, represents “photo-adsorption,” while the fourth one, $$\overset{\rightarrow}{R}_{P D}$$, corresponds to “photo-desorption.” The fifth term accounts for “hetero-molecular exchange,” and the sixth term describes “homo-molecular exchange.” Although this differential equation is nonlinear, because all variables under differentiation are known functions of time, it can be transformed into an algebraic form. In our experimental conditions, photo-desorption was not observed; therefore, it is considered to be zero in this analysis. The notation The results obtained for $$R_{q}$$ and $$R_{x}$$ are shown in Figure 8. The incident photon flux was approximately 1.8 × 10¹⁶ quanta per second (±20%), with a pressure of P = 1.2 × 10⁻⁶ Torr. It was observed that, for the POIE test reaction, irradiation within the exciton absorption band is about 5 to 8 times more effective than irradiation in the interband absorption region. This example demonstrates that the efficiency of the 2D ZnO/ZnO_1−x/O⁻ structure activated near the resonant exciton excitation is approximately 5–8 times higher than when excited in the interband absorption region [30].