Metastable photoinduced electrons can be generated and constantly trapped on oxide semiconductor (e.g., TiO2) nanoparticles in the presence of hole scavengers (e.g., H2O, amines and alcohols)1,2,3. These trapped electrons can be used in situ or ex situ as very effective reducing reagents to accomplish many useful and challenging chemical reactions without extra high-temperature and high-pressure conditions, such as photocatalytic CO2 reduction4, water splitting for H2 evolution5, photochemical nitrogen fixation6, and dehalogenation of halogenated organic compounds7,8. The site locations and states of trapped electrons (ecb) on the conduction band of TiO2 nanoparticles significantly influence the efficiency of the corresponding reactions9,10. Generally, photoinduced electrons in an anaerobic suspended system can be trapped at surface defect sites, such as the tetrahedral Ti4c3+ species on TiO2 nanoparticles11,12, or in the bulk, such as the Ti6c3+ species12,13. Alternatively, the electrons may also delocalize on the conduction band as free electrons14. The photoinduced electron selection of trapping locations on TiO2 nanoparticles has been considered to mainly depend on the crystalline structure of TiO212,14. In addition, a recent picosecond XAS study demonstrated that the locations of trapped electrons on TiO2 nanoparticles from dye sensitization and direct UV excitation are also different15. Although the final location of trapped electrons on TiO2 nanoparticles is still controversial, TiO2 nanoparticles with trapped electrons, no matter which hole scavenger is used, always exhibit broad visible-light absorption in the range of 400–800 nm16,17,18 and broadband IR absorption from 1100 to 3000 cm−119,20,21,22. Such characteristic UV-vis and IR absorption with a distinct blue color has long been used to recognize and confirm surface Ti3+ species as the electron-trapping states on TiO2 nanoparticles, as further evidenced by low-temperature (77 and 4 K) electron spin resonance (ESR) results13, and is denoted as the blue state22,23. These blue-state trapped electrons have been quantitatively titrated by Fe3+-1,10-phenanthroline reagent or TEMPO·/tBu3ArO· radicals16,19. However, these insights on the blue state on TiO2 nanoparticles were obtained under conditions using protic aqueous/alcohol solvents or with organic compounds as hole scavengers, and protons were available to mediate or stabilize electron trapping. We argue that the UV-vis and IR absorption characteristic of the electron-trapping features is only associated with proton participation. Until now, few other electron-trapping states have been discovered and reported on TiO2 nanoparticles that do not exhibit the above signature.

Here we report the photoinduced electron-trapping feature of TiO2 nanoparticles with iodide ions as a hole scavenger under strictly proton-free conditions. This feature is completely different from the common blue state observed under proton-participating conditions and does not exhibit the characteristic UV-vis (400–800 nm) and IR absorption (1100–3000 cm−1). The ESR (4 K) studies indicated that these proton-free trapped electrons (I hole scavenger) almost all entered the bulk of the TiO2 nanoparticles as the interstitial six-coordinated Ti6c3+ species, which is distinct from the common proton-participating electron-trapping state (alcohol hole scavenger) of the surface four-coordinated Ti4c3+ species. The synchrotron radiation ultraviolet photoelectron spectroscopy (SR-UPS) studies showed that the Ti3d binding energy of the proton-free trapped electrons (1.8 eV) is higher than that of the blue-state trapped electrons with protons (1.3 eV), together with Fe(III)-1,10-phenanthroline titration and ESR ·O2 detection, suggesting the different electron transfer abilities of the two electron-trapping states. Our work reveals that the well-known blue state exclusively occurs with the participation of available protons. To the best of our knowledge, this is the first reported scenery of trapped electrons on TiO2 nanoparticles without both characteristic UV-vis and IR absorption as diagnostic signatures.


Spectroscopic characteristics of trapped electrons

We designed two anaerobic liquid/TiO2 (anatase, 15-nm-sized nanocrystals) suspended systems, one proton and one proton-free, to observe the photoinduced electron-trapping states. In the proton system, methanol was employed as both the protic solvent and hole scavenger. In the proton-free system, I ions (LiI) dissolved in acetonitrile (MeCN, an aprotic solvent) were employed as the hole scavenger. Under UV illumination and anaerobic conditions, photoinduced holes (hvb+) were eliminated by either methanol or I ions, and the corresponding photoinduced electrons were subsequently kept on the TiO2 nanoparticles. When methanol was used, the illuminated TiO2 nanoparticles always exhibited a broad UV-vis absorption peak at 725 nm (determined by the in situ optical-fiber UV-vis spectroscopy) (Fig. 1a) and a blue color (inset of Fig. 1a), which has been universally observed and denoted as the signature of the electron-trapping state on TiO2 nanoparticles16,18,19. However, with LiI/MeCN, we unexpectedly did not observe the characteristic UV-vis absorption signal at ~700 nm. Instead, we only observed low absorption from 400 to 600 nm (Fig. 1b), which is from the I3 formed by the oxidation of I by photoinduced holes (refer to the standard UV-vis absorption spectra of I3 below in the inset of Fig. 1b and Supplementary Fig. 1). Rather than the traditional blue color, the illuminated TiO2 nanoparticle suspension in LiI/MeCN exhibited a faint yellow color (top inset of Fig. 1b). By separating preilluminated TiO2 nanoparticles from methanol and LiI/MeCN solutions, we further characterized the UV-vis diffused reflection spectra (UV-vis DRS) of different samples (Supplementary Fig. 2). Result shows that the signature absorption of electron-trapping state around 700 nm can only be observed on the preilluminated TiO2 nanoparticles from methanol solution, which is consistent with the optical-fiber UV-vis absorption spectra.

Fig. 1
figure 1

Optical-fiber UV-vis absorption spectra. a In situ optical-fiber UV-vis absorption spectra of TiO2 nanoparticles suspended in methanol under constant 365 nm illumination. The dashed line shows the peak at 725 nm. The inset shows the color change in the TiO2 suspension before and after UV illumination. b Same conditions as (a) except the TiO2 nanoparticles were suspended in a 0.1 M LiI/MeCN solution. The inset shows the color change of the TiO2 suspension before and after illumination, and the standard UV-vis absorption spectra of the I3/MeCN solution. c Spectroscopic titration of trapped electrons on TiO2 nanoparticles (0.6 g/L) in methanol with different illumination times using the Fe(III)-1,10-phenanthroline titrant. The inset shows the corresponding electron concentration as a function of the illumination time (calculated from the absorption of Fe(II)-1,10-phenanthroline at 515 nm). The arrow indicates the saturated electron concentration. d The identical titrations of thoroughly washed TiO2 nanoparticles (0.6 g/L) before and after 4 h UV illumination in a 0.1 M LiI/MeCN solution. No obvious change was found

We further quantitatively measured the saturated amount of trapped electrons on TiO2 nanoparticles under the two conditions. With methanol, the trapped electrons from the oxidation of methanol (CH3OH+2hvb+ → CH2 = O + 2H+) were quantified by titration with the Fe(III)-1,10-phenanthroline titrant (Fe3+ + ecb + 1,10-phenanthroline → Fe2+-1,10-phenanthroline) (Fig. 1c). For the 0.6 g/L TiO2 nanoparticles in the suspension, the saturated concentration of the trapped electrons was 1.7 × 10−4 M, accounting for 2.83 × 10−4 mol·g−1 on the TiO2 nanoparticles, which is similar to the value determined in our previous work19. With LiI/MeCN, the common UV-vis absorption signature at ~725 nm and the blue color were lacking, but the trapping of photoinduced electrons was also confirmed and quantified by the spectroscopic determination of I3 formation (Supplementary Fig. 3). The saturated concentration of trapped electrons in the suspension was calculated as 0.7 × 10−4 M for the same TiO2 concentration (0.6 g/L), which corresponds to 1.17 × 10−4 mol·g−1 on the TiO2 nanoparticles. The saturated concentration of trapped electrons with LiI/MeCN is less than that with methanol. Mayer’s group once observed a similar phenomenon on ZnO nanoparticles; i.e., more electrons were trapped (with CoCp*2, Cp*= pentamethylcyclopentadienyl) under acidic conditions than basic conditions, and proposed that protons may facilitate the trapping of electrons on MOx semiconductors24.

We also verified whether the trapped electrons on the TiO2 nanoparticles with LiI/MeCN can react with the Fe(III)-1,10-phenanthroline reagent. To avoid the presence of I, we separated the illuminated TiO2 nanoparticles from the LiI/MeCN solution and then thoroughly washed the TiO2 nanoparticles with MeCN. However, after the addition of the Fe(III)-1,10-phenanthroline titrant, no red Fe(II)-1,10-phenanthroline compound was generated (Fig. 1d), but the photoinduced electrons (1.17 × 10−4 mol·g−1) were present on the TiO2 nanoparticles, as evidenced by the I3 formation (Supplementary Fig. 3). This result demonstrates that the photoinduced electron-trapping states under proton and proton-free conditions have different electron transfer abilities for Fe(III)-1, 10-phenanthroline reduction. We further used ESR spectra to detect ·O2 radicals as the product of the reaction between O2 and TiO2 nanoparticles in different systems under UV illumination (O2 + ecb → ·O2). We found that the ·O2 signals can only be observed when TiO2 nanoparticles suspended in methanol (Supplementary Fig. 4a). With the TiO2/LiI/MeCN system, we only observed a background signal that is similar to the signal in the neat LiI/MeCN solution without TiO2 nanoparticles (Supplementary Fig. 4b, c). This result indicates that photoinduced electrons under the proton-free condition are also not able to react with O2 to produce ·O2 radicals, which agrees with our Fe(III)-1,10-phenanthroline titration experiments.

ATR-FTIR spectroscopy is another typical method used for in situ observation of the electron-trapping state on TiO2 nanoparticles. With methanol, we observed a very broad IR absorption band (3000–1100 cm−1) that gradually increased with the UV irradiation time, as expected (Fig. 2a), and this band has always been ascribed to photoinduced electron-trapping state of Ti3+ species on TiO2 nanoparticles19,20,21,22. However, such a diagnostic IR absorption band was not observed in the proton-free LiI/MeCN system (Fig. 2b), even with a prolonged reaction time (extended to 60 min). Moreover, when we added 10% H2O as a proton donor to the same LiI/MeCN system, the diagnostic IR absorption band of the Ti3+ electron-trapping state appeared along with negative peaks at 1640 and 3300 cm−1 that were identified as the loss of H2O by dissociation or oxidation (Fig. 2c)25,26. In this system, we also observed the UV-vis absorption peak at 725 nm, i.e., the signature of the blue-state trapped electrons, for the TiO2 nanoparticles (Supplementary Fig. 5). To further exclude the oxidation of H2O and only evaluate the effect of the proton donor in the proton-free LiI/MeCN system, we preloaded another inorganic proton donor, H3PO4, onto the TiO2 surface and dried the sample to remove residual H2O. As expected, under UV illumination, the broad IR absorption band at 3000–1100 cm−1 emerged but without the negative peaks at 1640 and 3300 cm−1 from the loss of H2O (Fig. 2d). This result clearly demonstrates that when proton donors are present in a system, even with I as the electron donor, the electron-trapping feature in the LiI/TiO2/MeCN system will change to the proton model as observed with methanol. By designing the proton-free LiI/MeCN system, we successfully revealed that the characteristic UV-vis and IR absorption signatures of the blue-state trapped electrons only occur in the presence of available protons. However, why do the proton-free trapped electrons on TiO2 nanoparticles not exhibit the characteristic UV-vis and IR absorption signals? What is the significant difference between the proton-free electron-trapping state and the common proton blue state?

Fig. 2
figure 2

In situ ATR-FTIR spectra on difference systems. a TiO2 film immersed in methanol; b TiO2 film immersed in a 0.1 M LiI/MeCN solution; c TiO2 film immersed in a 0.1 M LiI/MeCN/H2O (vMeCN:vH2O = 9:1) solution. The dashed circle and lines show the signal of water. d H3PO4-pretreated TiO2 film immersed in a 0.1 M LiI/MeCN solution. All experiments were conducted in an Ar atmosphere

ESR and SR-UPS characterizations

We used low-temperature (4 K) ESR and synchrotron radiation ultraviolet photoelectron spectroscopy (SR-UPS) to characterize the different photoinduced electron-trapping states on TiO2 nanoparticles. Trapped electrons in the form of Ti3+ can be easily recognized by their distinct ESR signals12,13,27,28. In 1985, Howe and Grätzel first observed in detail the electron-trapping locations in anaerobic, colloidal TiO2 aqueous suspensions by low-temperature (77 and 4 K) ESR13. They distinguished the surface Ti3+ species from the interstitial Ti3+ species in the bulk TiO2 nanoparticles and suggested that increasing the acidity and temperature of the TiO2 nanoparticle suspension favors the trapping of photoinduced electrons in surface sites13. However, since their observations were all carried out in aqueous solutions (H2O can be considered a common proton and electron donor under most conditions), their conclusion was that different hole scavengers (i.e., proton-containing and proton-free scavengers), such as methanol, I and polyvinyl alcohol (PVA), do not have a substantial influence on the Ti3+ species formed, but this conclusion is valid only in reaction systems with available protons. Under our experimental conditions, i.e., with methanol (Fig. 3a), we observed a rapidly increasing signal peak at g = 1.979 in the ESR spectrum with increasing UV illumination time. For illumination times over 4 h, a weak ESR signal at g = 1.930 was observed. The strong signal at g = 1.979 is attributed to surface distorted four-coordinate tetrahedral Ti4c3+ species, while the weak signal peak at g = 1.930 is attributed to anatase surface six-coordinate octahedral Ti6c3+ species12,13,27,28. We obtained the same ESR signals using ethanol and isopropanol instead of methanol as the proton-containing electron donors (Supplementary Fig. 6a, b). This result agrees with many other previous observations on the use of different organic compounds or H2O as proton and electron donors12,13,27,28, indicating that photoinduced electrons are trapped mainly as surface Ti4c3+ species with some surface Ti6c3+ species on TiO2 nanoparticles in the presence of protons. However, with LiI/MeCN (Fig. 3b), we did not observe the signals for surface Ti4c3+ and Ti6c3+ species at g = 1.979 and g = 1.930, respectively, with increasing UV illumination time. Instead, we only observed an increasing peak at g = 1.990 that corresponds to the anatase lattice octahedral Ti6c3+ state12,13, which never appeared with methanol. When using NaI instead of LiI as the proton-free electron donor, we also observed the signal of the interstitial lattice Ti6c3+ state (Supplementary Fig. 6c). This result indicates that photoinduced electrons are trapped as interior interstitial Ti6c3+ species in the lattice of the TiO2 nanoparticles in the I/MeCN system without the participation of protons. Without I in MeCN, the signal of Ti3+ species can no longer be observed (Supplementary Fig. 7). In addition, we added 10% H2O (pH = 1–13, adjusted by HClO4 and NaOH) as a controllable proton donor into identical LiI/MeCN suspension systems to observe if the electron-trapping states changed from lattice Ti6c3+ species to surface Ti4c3+ species. Indeed, with the addition of H2O, we observed that the signal of the surface Ti4c3+ species at g = 1.979 reappeared in the LiI/MeCN/H2O system, and its intensity gradually increased as the pH value of H2O decreased relative to the lattice Ti6c3+ signal at g = 1.990 (Fig. 3c). Notably, a separate ESR signal for the interior interstitial Ti6c3+ species on TiO2 nanoparticles was also obtained by Howe and Grätzel in a frozen aqueous solution when they illuminated a TiO2 sample at 77 K without hole scavengers13. In contrast to the sample illuminated at room temperature, in which the surface Ti3+ state dominates, illumination of their frozen system at ultralow temperature resulted in interior trapping of photoinduced electrons on TiO2 nanoparticles even in the presence of H2O13. Their observation is not contradictory to ours; i.e., the presence of available protons causes photoinduced electrons to be trapped on the surface of TiO2 nanoparticles as Ti4c3+ species, since proton formation and transfer are significantly inhibited at ultralow temperatures in a frozen system. Correspondingly, in our strictly proton-free I/MeCN system, we obtained the separate interior Ti6c3+ species on TiO2 nanoparticles by merely illuminating the TiO2 sample at room temperature. Thus, our work indicates that the occurrence and disappearance of the typical UV-vis and IR adsorption signatures for the electron-trapping event are essentially dependent on the location of the trapped electrons on the surface or interior lattice of the TiO2 nanoparticles, which is controlled by the presence or absence of protons. The conventional blue state on TiO2 nanoparticles requires available protons to stabilize trapped electrons as surface Ti4c3+ species.

Fig. 3
figure 3

ESR collected at 4 K and SR-UPS characterizations. a ESR spectra of TiO2 nanoparticles in methanol with different UV illumination (365 nm, 100 W) times (1–4 h). b Same conditions as (a) except TiO2 nanoparticles were suspended in a 0.1 M LiI/MeCN solution. c ESR spectra of TiO2 nanoparticles in a 0.1 M LiI/MeCN/H2O solution (vMeCN:vH2O = 9:1) after 4 h of UV illumination with the pH of the water adjusted from 13 to 1. d SR-UPS spectra (Ti3d orbital) of TiO2 nanoparticle films after UV illumination for different times (0–120 min) in methanol. e Conditions identical to those in (d) except the TiO2 nanoparticle films were illuminated in a 0.1 M LiI/MeCN solution

We further measured the Ti3d binding energy of the different electron-trapping states on TiO2 nanoparticles by SR-UPS (near the Fermi edge, 0–3 eV). Anatase TiO2 nanocrystal films were coated on a piece of Ti foil (see Methods section) and illuminated by UV light (365 nm) in methanol and LiI/MeCN solutions. The illuminated TiO2 films were subsequently washed with MeCN, dried, and transferred under anaerobic conditions for the SR-UPS characterizations. With methanol, the Ti3d binding energy of the photoinduced electron-trapping state on the TiO2 nanoparticle film was observed at 1.3 eV (Fig. 3d), which is larger than that of the well-known Ti3d defect state at 0.85 eV due to the construction of bridging oxygen vacancies (see Supplementary Fig. 8)29. However, with LiI/MeCN, the Ti3d binding energy of the electron-trapping state significantly shifted to 1.8 eV (Fig. 3e). This 0.5 eV difference in the binding energy of the Ti3d orbital directly reveals that the electrons trapped on TiO2 nanoparticles in the presence of methanol are much easier to lose than those trapped in the LiI/MeCN system, which explains why the trapped electrons in the LiI/MeCN system cannot react with the Fe(III)-1,10-phenanthroline titrant and O2.


In summary, we experimentally revealed the electron-trapping feature of TiO2 nanoparticles without the participation of protons, which does not exhibit the well-known UV-vis (400–800 nm) and IR absorption signals (1100–3000 cm−1). In the well-known blue state stabilized by protons, photoinduced electrons tend to be trapped as surface tetrahedral Ti4c3+ species on TiO2 nanoparticles. However, the electron-trapping state without available protons tends to be interior interstitial octahedral Ti6c3+ species in the lattice of the TiO2 nanoparticles. Fe(III)-1,10-phenanthroline titration experiments, ESR·O2 detection and SR-UPS spectra showed the different electron transfer abilities of the two kinds of electron-trapping states. Our results shed light on the intrinsic electron-trapping behaviors of TiO2 nanoparticles that were previously hidden under numerous proton-participating conditions. This basic insight into electron trapping in oxide semiconductors provides more information for the future design and preparation of efficient TiO2-based solar energy conversion devices.



The typical titanium dioxide (TiO2) solid sample, anatase TiO2 (15 nm nanocrystals), was purchased from Alfa Aesar (Shanghai, China). If not specifically mentioned, the TiO2 samples in all experiments were identical anatase TiO2 nanoparticles. Anhydrous FeCl2, Fe(NO3)3•9H2O, 1,10-phenanthroline, acetic acid (HAc), perchloric acid (HClO4, 70%), LiI (99.9%), methanol, and acetonitrile (MeCN) were purchased from Acros (Beijing, China) and used as received.


UV-vis absorption spectra of Fe(II)-1,10-phenanthroline and I3 ions were recorded on a Hitachi U3900 (Japan) spectrometer. The in situ optical-fiber UV-vis absorption spectroscopy measurement system was based on the Ocean Optics (Shanghai, China) HL-2000-FHSA optical-fiber analysis machine. The in situ ATR-FTIR spectra were collected using a Nicolet 8700 FT-IR instrument equipped with a DTGS detector. Low-temperature ESR spectroscopy was recorded on a JEOL (Japan) JES-FA200 ESR spectrometer. The SR-UPS experiments were performed at the Catalysis and Surface Science Endstation (BL11U beamline) in the National Synchrotron Radiation Laboratory (NSRL), Hefei, China.

Photo-illumination of TiO2 nanoparticle suspensions

In a typical procedure, a TiO2 suspension (0.2–1.0 g/L) in methanol or 0.1 M LiI/MeCN was sealed into a 50-mL quartz bottle and degassed by argon for 30 min. Then, the quartz bottle was placed into a black box for UV illumination for different times (1–12 h). A high-power (100 W) 365-nm monochromatic LED lamp was employed as the UV light source.

In situ optical-fiber UV-vis absorption spectroscopy

We designed an in situ analysis system to measure the UV-vis absorption of TiO2 suspensions (2.0 g/L). TiO2 suspensions were held in a sealed quartz cube in a black box under constant UV (365 nm, 100 W) irradiation. UV-vis absorption spectra were recorded in situ from 400 to 900 nm by an optical-fiber spectrometer.

In situ ATR-FTIR analysis

The ATR-FTIR spectroscopy experimental setup consisted of Harrick Horizon multiple internal reflection accessories coupled to a 4 mL flow-through cell containing a ZnSe crystal on the bottom plate and a quartz window on the top plate. Eleven infrared bounces were allowed using a 45° internal reflection element (50 × 10 × 2 mm3). In a typical procedure, a layer of reaction solution (methanol or I/MeCN) was dripped onto the surface of a ZnSe crystal coated with a TiO2 film. The apparatus was degassed with argon for 30 min, and the crystal was scanned to obtain the background spectrum. Time-resolved in situ FTIR data were collected under irradiation by a 365 nm LED lamp (3 W).

ESR analysis

Low-temperature (4 K) ESR spectra were collected using liquid helium as the coolant. In a typical procedure, photoreduced TiO2 was sealed in a quartz tube in an argon atmosphere. The settings for the ESR spectrometer were as follows: center field = 3500 G, sweep width = 1000 G, microwave frequency = 9.52 GHz, and field modulation frequency = 100 kHz.

SR-UPS experiments

The near-Fermi level Ti3d spectra (0–3 eV) were measured using synchrotron radiation light as the excitation source with a photon energy of 47.5 eV. In a typical procedure, anatase TiO2 films were preilluminated in different reaction systems, washed with MeCN, dried in an argon glove box, and transferred into the SR-UPS ultrahigh vacuum system under the protection of an argon flow.

Fe(III)-1,10-phenanthroline spectrometric titration

Trapped electrons on UV-illuminated TiO2 nanoparticles were quantitatively measured by the Fe(III)-1,10-phenanthroline titration spectrometric method. The 1,10-phenanthroline spectrometric titration is a simple, widely used method for the measurement of Fe(II) ions. Here, we used the Fe(III) solution to titrate the trapped electrons on TiO2 nanoparticles and quantitatively produce Fe(II) ions. Then, we used 1,10-phenanthroline to measure the concentration of Fe(II) ions produced, thereby quantifying the concentration of trapped electrons. In a typical procedure, 2.5 mL of the UV-illuminated TiO2 suspension was transferred to an argon glove box and mixed with 2.5 mL of a Fe(NO3)3 (10−3 M) solution. The resulting mixture was centrifuged. Then, 1.5 mL of the centrifuged supernatant was taken from the argon glove box and mixed with 1.5 mL of pH = 4.6 HAc-NaAc buffer solution. Then, 1 mL of a 0.2% 1,10-phenanthroline aqueous solution was added to obtain a clear red solution. The resulting solution was transferred to a quartz cuvette and measured on a Hitachi U3900 spectrometer. For the sample in the I/MeCN system, the wash treatment of the TiO2 nanoparticles was conducted in an argon glove box by MeCN, and then, the clean TiO2 nanoparticles were suspended in MeCN for the titration measurement.

ESR O2 detection

ESR detection of ·O2 radicals was conducted at room temperature on the Bruker E500 machine with DMPO as radical trapping reagent. In a typical procedure, 0.5 µg DMPO was added into 10 mL different liquid/TiO2 suspended systems (with 0.6 g/LTiO2). The resulting suspension was then collected by a capillary quartz tube for the ESR detection. During the detection, 355 nm pulse laser (1 kHz) was employed to constantly illuminate the sample. All experiments were conducted under the ambient condition.

Preparation of H3PO4-treated TiO2 nanoparticles

In a typical procedure, 100 mg TiO2 powder was dissolved in 5 mL of a diluted (5%) H3PO4 aqueous solution. Then, the suspension was dried at 70 ℃ overnight, and the resulting solid was ground into a powder to obtain H3PO4-treated TiO2 for the ATR-FTIR measurement.

Preparation of TiO2 nanoparticle films

TiO2 films for SR-UPS experiments were loaded on 0.25-mm-thick Ti metal foil. In a typical procedure, TiO2 was dissolved into n-butyl alcohol with the addition of several drops of polyethylene glycol tert-octylphenyl ether (Triton X-100) as a thickener. The solid solution ratio of the TiO2 suspension was between 10 and 15%. After ultrasonic (for 1 h) and stirring treatments (for at least 1 day), the resulting suspension was used as the precursor for the TiO2 films. The resulting precursor was coated on Ti metal foil by the traditional doctor-blade method and then calcined at 450 °C for 1 h to obtain the final product.