Introduction

Photocatalysis involves light-induced generation of electron–hole pairs and transfer of electrons and holes to adsorbed species on the semiconductor catalyst surface [1,2,3,4]. The overall quantum efficiency of a photocatalysis process is governed by the rates of charge separation, interfacial charge transfer, trapping, and recombination [5,6,7]. Photo-generated electrons and holes must reach the catalyst surface sites to initiate the charge transfer reaction for converting the adsorbed reactants to products. Many photocatalytic reactions are limited by the rapid recombination of electrons and holes.

Although the time scales of photocatalytic steps range from femtoseconds to nanoseconds [8], the photo-generated electrons produced from adsorbed species can persist in TiO2 for a long period of time at a scale of minutes to hours after terminating UV irradiation [9]. These long-life electrons, which do not get involved in either recombination with holes or transfer to the catalytic site, could stay in the conduction band and in the shallow/deep traps. The trapped electrons have been associated with Ti3+ of which optical and electrical properties are affected by its surrounding lattice [10].

Many studies on the generation and consumption of photo-generated electrons have been conducted at temperatures below 273 K and under vacuum [11,12,13], e.g., 90 K and 10−3 Pa vacuum which are not directly relevant to the working state of the photocatalyst. Thus, we have attempted to characterize the relative concentration of photo-generated electrons, conduction-band electrons, and shallow-trap electrons on TiO2 and N-TiO2 in the presence of adsorbed ethanol. Adsorbed ethanol is selected for this study because it possesses a high electron-donating potential and the structure of its adsorbed products has been well characterized [14]. N-TiO2 is worthy of the investigation because of its visible light photocatalytic activity [15].

The objective of this study is to compare the rate of the generation and consumption as well as the relative concentration of \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) on TiO2 and N-TiO2 particles under typical conditions of photocatalysis (300 K, 0.1 MPa) using in situ IR spectroscopy. Experimental IR spectra, which were obtained as a function of time during UV irradiation, were fitted into two models, to obtain simulated IR intensities of \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\). N-doping onto TiO2 increased the population of \({\text{e}}_{ST}^{ - }\) and decreased the rate of the generation of photo-generated electrons. The direct observation of the IR intensity of \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) under working state of photocatalysts will provide an in-depth insight into the role of various photo-generated charge carriers in photocatalysts.

Experimental

Materials preparation

P25 TiO2 (Degussa), ammonia (NH3, >99.98%, Praxair), and ethanol (99.5%, Fisher Scientific) were used as received. Nitrogen-doped TiO2 (N-TiO2) was prepared by treating P25 TiO2 particles with ammonia (flow rate = 60 cm3/min) at 723 K for 3 h.

Material characterization

Table 1 shows the characterization results of TiO2 and N-TiO2 catalysts. The nitrogen content of the N-TiO2 catalyst was determined to be 5.4% (atomic ratio of N/Ti) by X-ray photoelectron spectroscopy (XPS) using PHI Versaprobe II scanning XPS microphobe with SmartSoft control package. The specific surface areas of P25 TiO2 and N-TiO2, measured by Micromeritics ASAP 2020 physisorption analyzer, were in the 40–50 m2/g range. UV–Vis spectroscopic analysis (Hitachi U-3900 research grade UV–Vis spectrophotometer) gave the band-gap energy of TiO2 and N-TiO2 at 3.20 and 2.48 eV, respectively, which is consistent with literature values [16]. The last three columns show the fitting parameters used in this study, which will be discussed later.

Table 1 Properties of TiO2 and N-TiO2 photocatalysts
Full size table

In situ FTIR study

The experimental apparatus for in situ IR study is illustrated in Fig. 1 and consists of (1) a mass-flow controller (Brooks Instrument 5850), (2) a FTIR spectrometer equipped with a MCT detector (Nicolet 6700, Thermo Scientific), which accommodates a diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) cell (Harrick Scientific), and (3) a Hg lamp as UV light source (Oriel 6286) the intensity of which was set to 25 mW/cm2. For each in situ IR study, 20 mg sample of TiO2 or N-TiO2 was loaded in a sample holder with 0.7 cm diameter and 0.2 cm depth, as shown in Fig. 1. The DRIFTS cell was capped by a three-window dome. Two of the windows in the IR beam path were ZnSe which is IR-transparent, while the third one was made of CaF2 to allow the UV irradiation to pass through. TiO2 and N-TiO2 were exposed to an Ar/ethanol flow at a flow rate of 60 cm3/min which was obtained by passing Ar flow through an ethanol saturator. Excess gaseous ethanol was purged out from the DRIFTS cell by flowing Ar for 20 min before the cell was switched to batch mode. IR spectra of the catalyst with adsorbed ethanol during UV irradiation were recorded as a function of time at a resolution of 4 cm−1 with four co-added scans. The IR spectra collected were presented in two forms: (1) the single beam (SB) spectrum, corresponding to a transmission spectrum, which was obtained by the Fourier transform of the IR interferogram; (2) absorbance (A) spectrum which was obtained by A = −log (I t /I 0), where I 0 is the intensity of the background IR single beam spectrum of air and I t is the intensity of the single beam spectrum collected at a specific time during UV irradiation.

Fig. 1
figure 1

Experimental setup for in situ IR study

Full size image

Results and discussion

Simulated IR spectra of conduction-band and shallow-trap electrons produced from UV irradiation on ethanol-adsorbed TiO2

Figure 2a presents single beam spectra obtained from pristine TiO2 and TiO2 with adsorbed ethanol before, during, and after UV irradiation; Fig. 2b shows the corresponding absorbance spectra. The pristine TiO2 sample gives a broad H2O stretching at 3200–3700 cm−1 and a bending band at 1640 cm−1. The insets in Fig. 2b show the surface hydroxyl groups of pristine TiO2 sample and the adsorbed ethanol before and after the reaction, as well as the difference spectra between before and after UV irradiation. The assignments [14] of these various bands are listed in Table 2. The major IR-observable product generated after the UV irradiation is acetate at 1438 cm−1. UV irradiation of TiO2 with adsorbed ethanol resulted in an increase in the IR absorption which shows a nearly flat single beam spectrum in Fig. 2a and a featureless absorption spectrum in Fig. 2b. The featureless absorption spectrum is a result of the interactions of IR irradiation with shallow-trap electrons (\({\text{e}}_{ST}^{ - }\)) and conduction-band electrons (\({\text{e}}_{CB}^{ - }\)).

Fig. 2
figure 2

a IR single beam spectra, b absorbance spectra of pristine TiO2 sample and ethanol-adsorbed TiO2 before, with, and after UV irradiation, and c simulated \({\text{e}}_{CB}^{ - }\) (purple), \({\text{e}}_{ST}^{ - }\) (blue), and IR (black) curves of the difference spectrum A 4 − A 2 (red). Insets (i) surface hydroxyls of pristine TiO2; (ii) surface adsorbed species before and after UV irradiation; (iii) difference spectrum obtained from A 3 − A 2. (Color figure online)

Full size image
Table 2 Band assignments of adsorbed surface species on TiO2 and N-TiO2
Full size table

The specific effect of UV irradiation can be further highlighted by the difference spectrum A 4 − A 2 in Fig. 2c. To determine the contribution of \({\text{e}}_{ST}^{ - }\) and \({\text{e}}_{CB}^{ - }\) to the featureless difference spectra, we have adopted the following elaborated procedure [13] to obtain simulated \({\text{e}}_{ST}^{ - }\) and \({\text{e}}_{CB}^{ - }\) spectra using MATLAB.

  1. (1)

    Smoothing the experimental IR spectrum with the Huber function to eliminate the outlying points in a curve. For example, the spikes and minor peaks are removed in this step for the following data fitting process.

  2. (2)

    Obtaining the simulated \({\text{e}}_{CB}^{ - }\) spectrum by Eq. 1. The IR absorbance produced by the conduction-band electrons, \(\Delta A_{\text{CB}}\), can be described by the following:

    $$\Delta A_{\text{CB}} = A\tilde{v}^{ - p}$$
    (1)

    where \(\tilde{v}\) is the wavenumber, and p: is the scattering factor. The value of p for IR absorbance ranges from 1.5 to 1.7 [17]. p = 1.7 was found to give the best fit and was used to generate all simulated \({\text{e}}_{CB}^{ - }\) spectra.

  3. (3)

    The simulated IR absorbance generated by \({\text{e}}_{ST}^{ - }\), \(\Delta A_{\text{ST}}\), were obtained by Eq. 2 [13]:

    $$\Delta A_{\text{ST}} = A_{0} \sum\limits_{i = 1}^{24} {\alpha_{i} \frac{{(E - E_{\text{OP}}^{i} )^{1.5} }}{{E(E - E_{\text{OP}}^{i} + E_{\text{DI}} )^{4} }}}$$
    (2)

    where α: a weighting factor defining the contributions of the IR absorbance resulted from the shallow-trap electrons with different optical ionization energies, E iOP . E iOP : the optical ionization energy (i.e., the photon energy required to excite an electron from a trap). E DI: the donor ionization energy (i.e., the thermal energy required to generate a carrier).

A set of values for these parameters in Table 1 were selected to obtain the best fit simulated IR absorbance curve. The Eop value, which corresponds to the energy level of the trapped electron, can be obtained from the maximum point of the simulated \({\text{e}}_{ST}^{ - }\) curve.

The simulated IR spectrum which was obtained from the summation of simulated \({\text{e}}_{ST}^{ - }\) and \({\text{e}}_{CB}^{ - }\) curves (shown in Fig. 2c), is a perfect match with experimental IR spectrum in the region of 1300–3300 cm−1, which is the range of energy with most optical ionization.

Figure 3 shows the difference spectra of TiO2 with adsorbed ethanol collected as a function of time upon UV irradiation. The difference spectrum at t = 60 s corresponds to A 4 in Fig. 2c.

Fig. 3
figure 3

a Difference spectra, b \({\text{e}}_{CB}^{ - }\) and c \({\text{e}}_{ST}^{ - }\) simulated absorption curves of ethanol-adsorbed TiO2 during UV irradiation process

Full size image

Figure 3b, c show the evolution of the simulated \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) spectra resolved from difference spectra in Fig. 3a using Eqs. 1 and 2. The intensities of \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) absorption curves dramatically increased upon UV irradiation. The rate of increase slowed down after t = 3 s and then reached a plateau after around 12 s. These variations are plotted in Fig. 5 for comparison between TiO2 and N-TiO2.

The monotonic increase at the lower wavenumber region in \({\text{e}}_{CB}^{ - }\) spectra in Fig. 3b results from the Drude absorption of free electrons at the conduction band, which can be attributed to the intraband and interband transitions of the accumulated electrons [18, 19]. The peak (i.e., maxima) of the \({\text{e}}_{ST}^{ - }\) curves shifted from low to high wavenumber as their intensities grew, as shown in Fig. 3c. Higher wavenumbers of the peaks of \({\text{e}}_{ST}^{ - }\) curves correspond to higher IR photon energies which required to excite electrons from the shallow trap to conduction band.

Simulated IR spectra of conduction-band and shallow-trap electrons produced from UV irradiation on ethanol-adsorbed N-TiO2

Figure 4 shows the IR spectra obtained from ethanol-adsorbed N-TiO2 under UV irradiation. The spectra of pristine TiO2 and N-TiO2 in the region between 3550 and 3750 cm−1, shown in Figs. 2b and 4b, were attributed to surface hydroxyl (–OH) groups: linear –OH at 3693 cm−1, bridged (double-bonded) –OH at 3663 cm−1, and triple-bonded –OH at 3633 cm−1 [6]. Table 3 compares the intensities of these bands in Figs. 2b and 4b and revealed that the N-doping increased the population of double- and triple-bonded OH, which may serve as the sites where ethanol adsorbed [20]. The IR features of these OH groups vanished upon ethanol adsorption. Comparison between the IR spectra of TiO2 and N-TiO2 shows that N-TiO2 had a higher intensity of C–H bonding in the range of 2800–3000 cm−1, suggesting the presence of a higher coverage of ethanol on N-TiO2. By calculating the ratio of the C–H peak areas of the two catalysts and comparing with the calibration results from our previous work [21], the ethanol coverage at room temperature was estimated to be 820 µmol/mg on TiO2 and 3780 µmol/mg on N-TiO2.

Fig. 4
figure 4

a IR single beam spectra, b absorbance spectra of pristine N-TiO2 sample and ethanol-adsorbed N-TiO2 before, with, and after UV irradiation, and c simulated \({\text{e}}_{CB}^{ - }\) (purple), \({\text{e}}_{ST}^{ - }\) (blue), and IR (black) curves of the difference spectrum \(A_{4}^{{\prime }} - A_{2}^{{\prime }}\) (red). Insets (i) surface hydroxyls of pristine N-TiO2; (ii) surface adsorbed species before and after UV irradiation; (iii) difference spectrum obtained from \(A_{3}^{{\prime }} - A_{2}^{{\prime }}\). (Color figure online)

Full size image
Table 3 Hydroxyl groups distribution on TiO2 and N-TiO2 samples
Full size table

Adsorbed ethanol on the surface of the photocatalysts at room temperature formed both molecularly adsorbed CH3CH2OHad and dissociatively adsorbed CH3CH2Oad [22]. Although CH3CH2OHad and CH3CH2Oad exhibited similar bands of C–H stretching vibrations at 2971 and 2931 cm−1, their IR spectra in the region of 1000–1600 cm−1 exhibited different features. The most significant difference between TiO2 and N-TiO2 is that TiO2 produced CH3COO −ad while N-TiO2 produced three products, CH3COO −ad , HCOO −ad , and HCHOad. The formation of CH3COO −ad from adsorbed ethanol can be considered as a key step for the injection of electrons from adsorbed ethanol to TiO2/N-TiO2, resulting in the accumulation of \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\).

$${\text{CH}}_{3} {\text{CH}}_{2} {\text{OH}}_{\text{ad}} + {\text{nTiO}}_{2} \to {\text{CH}}_{3} {\text{COO}}^{-}_{\text{ad}} + 3{\text{H}}^{ + } + 2{\text{e}}^{-} + {\text{nTiO}}_{(2 - 1/n)}$$
(3)

The formation of proton has been verified by photo-oxidative degradation study of organics under aqueous condition [23]. CH3COO −ad may undergo further reactions to produce HCOO −ad and HCHOad as shown in our earlier studies [14].

Figure 5 shows the experimental IR absorbance, \({\text{e}}_{ST}^{ - }\) and \({\text{e}}_{CB}^{ - }\)’s intensity profiles over time and the relative distribution of \({\text{e}}_{ST}^{ - }\) and \({\text{e}}_{CB}^{ - }\) for TiO2 and N-TiO2. The experimental and simulated IR intensity profiles were taken at the IR intensity at 2000 cm−1 , which has been used as an index for the population of photo-generated electrons because they are usually not influenced by other bands related to ethanol [22]. High intensity ratio of \({\text{e}}_{ST}^{ - }\)/\({\text{e}}_{CB}^{ - }\) on N-TiO2 suggests that our N-TiO2 may contain more oxygen vacancies than TiO2 because the shallow-trap band is known to be closely related to the oxygen vacancies in the TiO2 structure [24].

Fig. 5
figure 5

The intensity of I2000, \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) as a function of UV irradiation time for ethanol-adsorbed TiO2 and N-TiO2 (insets show the normalized rate of change in the first 6 s)

Full size image

The change of \({\text{e}}_{CB}^{ - }\) intensity led that of \({\text{e}}_{ST}^{ - }\) on both TiO2 and N-TiO2, supporting the previous proposed steps [25].

  1. step (1)

    \({\text{TiO}}_{2} + hv \to {\text{e}}_{CB}^{ - } + h_{\text{VB}}^{ + }\)

  1. step (2)

    \({\text{e}}_{CB}^{ - } \to {\text{e}}_{{{\text{shallow}}\,{\text{trap}}\,{\text{with}}\,{\text{low}}\,Eop }}^{ - }\)

  1. step (3)

    \({\text{e}}_{CB}^{ - } \to {\text{e}}_{{{\text{shallow}}\,{\text{trap}}\,{\text{with}}\,{\text{high}}\,Eop }}^{-}\)

  1. step (4)

    \({\text{e}}_{trap}^{ - } + h_{ }^{ + } \to {\text{annihilation}} + {\text{heat}}\)

\({\text{e}}_{CB}^{ - }\) produced from UV irradiation was transferred to shallow trap sites, shown in step (2). Step (2) occurred at the first second after UV irradiation. After the first second, the step (3) began dominating as indicated by the shifting of the peak of the \({\text{e}}_{ST}^{ - }\) curves to high wavenumber (i.e., high optical ionization energy). The optical ionization energy for both TiO2 and N-TiO2 are plotted in Fig. 6 for comparison [25]. Termination of UV irradiation led to the decay of \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) curves, indicating the rate of charge carrier (e− and h+) recombination is higher than their generation. The decay curves levelled off after 120 s and did not return to the base line, indicating the accumulation of \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) and the long lifetime of electrons in the existence of ethanol.

Fig. 6
figure 6

The optical ionization energy of \({\text{e}}_{ST}^{ - }\) as a function of UV irradiation time

Full size image

Table 4 summarizes a number of related studies that were carried out under vacuum or aqueous conditions. These results showed accumulated \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) can persist for a long period of time. Our studies showed the accumulated \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) can co-exist with adsorbed ethanol and their products for a long period of time at 0.1 MPa and 300 K. The negative charges of accumulated \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) could be balanced by protons on TiO2 and N-TiO2 of which sites for combination of H+ and \({\text{e}}_{CB}^{ - } /{\text{e}}_{ST}^{ - }\) may be blocked by adsorbed ethanol and its products.

Table 4 Comparison of experimental techniques and conditions used in this field
Full size table

The rise and fall of the IR intensity of \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) reflect their relative generation and consumption rates. The presence of high coverage of H2O inhibited accumulation of both \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) on adsorbed ethanol on TiO2 [14]. H2O could shift the oxidation of adsorbed ethanol from a hole-initiated reaction to an electron-initiated reaction, resulting in a high rate of consumption of photo-generated electrons. These photo-generated electrons can be consumed by exposing to oxygen at a high rate, evidenced by a sharp decrease in their IR intensity [9]. The results of these studies suggest that photo-generated electrons are readily accessible to specific reactants. The key fundamental issue for the future study is identification of active sites where the transfer of \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) to adsorbed reactants occur.

Conclusions

Dynamic behaviors of photo-generated \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) on TiO2 and N-TiO2 have been studied by fitting experimental IR spectra to simulated IR curves of \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) at 300 K and 0.1 MPa. N-doping on TiO2 produced a higher population of \({\text{e}}_{ST}^{ - }\) and generated \({\text{e}}_{ST}^{ - }\) and \({\text{e}}_{CB}^{ - }\) at a lower rate than TiO2. The rise of \({\text{e}}_{CB}^{ - }\) curve leads that of \({\text{e}}_{ST}^{ - }\), indicating \({\text{e}}_{CB}^{ - }\) is produced prior to \({\text{e}}_{ST}^{ - }\). This observation further suggests that the trapped electrons were produced by the injection of \({\text{e}}_{CB}^{ - }\) to the trapped sites. Upon UV irradiation, \({\text{e}}_{CB}^{ - }\) filled in the trap sites of which energy levels are closer to the conduction band first and then fill in those farther away. Under constant UV irradiation, the average energy level of \({\text{e}}_{ST}^{ - }\) was determined to be 0.30 eV for TiO2 and 0.26 eV for N-TiO2 below conduction band. Upon terminating UV irradiation, the IR intensity of accumulated \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) at trap sites showed a gradual decay in a time scale of minutes. This study showed that the photo-generated electrons and holes can exist at a time scale of minutes in TiO2 and N-TiO2 in the presence of adsorbed ethanol under ambient conditions. To expand the practical applications of TiO2-based materials, the investigations on the specific roles of \({\text{e}}_{CB}^{ - }\) and \({\text{e}}_{ST}^{ - }\) in photocatalysis are warranted.