Introduction

Oxygen vacancy (OV) defects at reducible metal oxide surfaces play a key role in a heterogenous catalytic oxidation process1,2,3,4. In 1954, Mars and van Krevelen reported that the oxidation of organic compounds on V2O5 includes V2O5 reduction by an organic compound and the subsequent oxidation of V2O5 by O25. This reduction and oxidation mechanism had been verified the OVs at the atomic level for the oxidation of CO on RuO2 (110) surfaces using scanning tunneling microscopy (STM) in conjunction with density-functional theory (DFT) calculations6. This has induced a boost in studies to identify OVs on metal oxide surfaces. For example, OVs have been identified on the surfaces of rutile TiO2 using high-resolution STM7. Other studies have demonstrated that many types of OVs with different catalytic reaction characteristics can exist on metal oxide surfaces. OVs have been observed on metal oxide surfaces in association with three metal (M) and oxygen (O) groups (M=O, M–O–M, and M3–O)8. Moreover, the local structures of OVs on the treated and untreated surfaces of CeO2 (110) crystal planes have been elucidated using STM in conjunction with DFT calculations9.

Recent reviews have summarized the methods that can be applied to characterize oxygen species at catalyst surfaces10 (Supplementary Table S1). An overview has focused on understanding the roles of OVs playing in the oxidation reaction at reducible metal oxide surfaces11. For example, the dissociation of O2 at OVs was found to greatly impact oxygen adsorption on TiO2 (110) surfaces, where one O atom from the dissociated O2 molecule is postulated to fill an OV and the second O atom deposited at the five-coordinate Ti4+ site12. The roles of oxygen atoms and molecules at catalyst surfaces and the properties of OVs have also been the subject of a recent review13.

The importance of OVs has led to the development of numerous strategies for increasing the concentration of OVs in metal oxide catalysts. Some success has been achieved via doping with secondary metal ions and nano structuring14,15, and the doping strategy has been expanded to develop four-layer metal oxide catalysts (CuO/VOx/Ti0.5Sn0.5O2) with layers composed of synergistic OV concentrations16. The dispersal of metal ions on the surfaces of metal oxides has also been demonstrated to increase the concentration of OVs effectively17,18. These strategies have been widely used in photocatalytic materials, electrocatalytic materials, thermal catalytic materials, and optical materials19,20,21. However, effective methods to improve the performance of metal oxide catalysts are influenced by current characterization technologies. Therefore, it is required to find an effective characterization technology to identify OVs and understand oxidation mechanisms that occur at the surfaces of metal oxides under real reaction conditions.

The operando diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) is a powerful technology that can identify surface species on a catalyst under real reaction conditions. Ye et al. found that toluene adsorption and reaction with OVs can effectively reduce the accumulation of by-products22. Li et al. investigated the structure-performance relationships of α, β, γ, and δ-MnO2 catalysts, they found that toluene adsorption is promoted by rapid dehydrogenation of methyl groups on the surface of δ-MnO223. Yao et al. used the combination of DRIFTS with a mass spectrometry (MS) to observe the functional groups on the catalyst surface and the changes in MS signals of gaseous components during the catalytic oxidation of toluene on CeO224.

Due to its multiple valence states and structural diversity (e.g., tunneling (α, β, and γ-MnO2) and layered (δ-MnO2) structures), MnO2 is an important functional metal oxide material25,26,27. β-MnO2 has a thermodynamically stable phase and high crystallinity, and become one of the hot spots in current researches28,29.

The present work addresses these issues by combining an operando DRIFTS with a temperature-programmed reaction (TPR) cell and MS to explore the behaviors of OVs and adsorbed oxygen species at β-MnO2 surfaces during H2 oxidation reaction conducted in the temperature range of 25–400 °C. The roles of OVs in H2 oxidation process are explored according to relations between OVs and oxygen species, which in turn reveal interactions between surface oxygen species with H2 at different reaction temperatures.

Results and discussion

Catalyst characterization

The crystal structure of β-MnO2 was confirmed using X-ray diffraction (XRD). β-MnO2 has good crystallization and no obvious crystal defects (Fig. 1a). High-resolution transmission electron microscope (HRTEM) image of β-MnO2 is shown in Supplementary Fig. S1. The well-identified periodic lattice fringes of 2.41 and 3.15 nm are corresponding to the interplanar distances of (101) and (110) facets of β-MnO2. Whereas severe blurring of the lattice fringes is also found (highlighted by red rectangles), suggesting the existence of OVs at β-MnO2 surfaces30.

Fig. 1: Catalyst characterization of β-MnO2.
figure 1

a XRD patterns. b Raman spectrum. c TG profile. d O2-TPD profile. e Mn2p XPS spectrum. f O1s XPS spectrum.

Figure 1b shows the Raman scattering spectrum of β-MnO2. The band at 630 cm1 corresponds to the tensile pattern of the [MnO6] octahedron, and the band at 330 cm1 is assigned to the metal-oxygen chain of Mn–O–Mn in the MnO2 octahedral lattice, indicating the presence of a well-developed rutile-shaped skeleton31.

Thermogravimetric (TG) analysis result shows that the weight loss of β-MnO2 is not obvious below 500 °C (Fig. 1c). This is due to the coordination of Mn and O in the phase structure is close to saturation, and the phase tunnel structure is stable. The weight loss at higher temperatures is attributed to the removal of lattice oxygen, resulting in the reduction of MnO2 to Mn2O3 (between 500 and 600 °C) with a weight loss of 9.63% and to Mn3O4 (between 720 and 820 °C) with a weight loss of 3.62%32.

O2 temperature-programmed desorption (O2-TPD) was used to observe the O2 desorption from β-MnO2 (Fig. 1d). There is a small desorption peak around 150 °C, which is a signal of surface oxygen desorption. When the temperature reaches about 587 and 813 °C, two obvious desorption peaks appear. The small peak at around 150 °C is due to O2 desorbed from β-MnO2, the peaks at 587 and 813 °C are related to the desorption of lattice oxygen and bulk lattice oxygen33,34.

X-ray photoelectron spectroscopy (XPS) was used to measure the valence states of Mn and the types of O at β-MnO2 surfaces (Fig. 1e, f and Table 1). The fraction ratios of Mn3+ and Mn4+ are 32.0% and 68.0%, respectively, indicating that β-MnO2 is oxidizable and reducible. O1s spectrum can be divided into lattice oxygen (Olatt) at 529.2 eV and adsorbed oxygen/surface hydroxyl groups (Oads and (OH)ads) at 531.7 and 533.2 eV (Fig. 1f)35,36. The fraction ratio of Olatt is 77.2%, indicating the presence of OVs at β-MnO2 surface.

Table 1 Mn2p, O1s binding energies, and the corresponding parameters.

H2 oxidation by surface oxygen species in the absence of O2

The DRIFTS spectra, MS signals, and normalized peak intensities during H2 oxidation by oxygen species at β-MnO2 surfaces in the absence of O2 are shown in Fig. 2. Seven kinds of oxygen species at β-MnO2 surfaces can be found, those are bridge-type (M+–O2––M+) group (750–800 cm–1)37, terminal-type (M2+–O2–) group (1300–1400 cm–1)38,39,40 (Supplementary Figs. S2S4 also prove that 1300 cm–1 belongs to M=O at β-MnO2 surfaces), M+–O group (870 cm–1)41, adsorbed molecular O2 groups including M+–O2 group (1110–1120 cm–1)42,43 and M2+–O22− group (930–960 cm–1)44, and oxidation products including δ(H2O) (1520, 1610, and 1640 cm–1) and v(OH) (3080, 3230, 3530, and 3720 cm−1)45,46 (Supplementary Table S3).

Fig. 2: Experimental results of H2 oxidation by oxygen species at β-MnO2 surfaces in H2/Ar as a function of temperature.
figure 2

a, b DRIFTS spectra. c MS signal. d Normalized intensities where the error bars are the standard deviations obtained by measuring the peak heights more than three times.

At a temperature higher than 110 °C, H2O MS signal increases obviously (Fig. 2c) and the normalized intensity of M+–O2––M+ (M–O–M) decreases (Fig. 2d), but the normalized intensity of other surface oxygen species do not change significantly below 110 °C. This finding implies that the bridge-type of oxygen atom in M+–O2––M+ (M–O–M) first reacts with H2 to form gaseous H2O and bridge-type OV (M–□–M, where, OV is represented by an empty square □) (Eq. (1)). When the temperature exceeds 150 °C, except M+–O2–M+, the normalized intensities of M+–O2, M2+–O2, M+–O, and v(OH) decrease, but the normalized intensities of δ(H2O) and M2+–O22− increase with increasing temperature. These results indicate that M+–O (M–O) and M2+–O2 (M=O) can react with H2 above 150 °C to generate H2O and terminal-type OV (bare M) (Eqs. (2) and (3)), which leads to an increase in the normalized intensity of δ(H2O)47. M2+–O2 (M=O) reacts with surface H2O to form OH (Eq. (4)), which leads to a decrease in the normalized intensity of v(OH) in H2O at β-MnO2 surfaces.

$${{{{{\rm{M}}}}}}{-}{{{{{\rm{O}}}}}}{-}{{{{{\rm{M}}}}}}+{{{{{{\rm{H}}}}}}}_{2}\to {{{{{\rm{M}}}}}}{-}{{\square }}{-}{{{{{\rm{M}}}}}}+{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}$$
(1)
$${{{{{\rm{M}}}}}}{-}{{{{{\rm{O}}}}}}+{{{{{{\rm{H}}}}}}}_{2}\to {{{{{\rm{M}}}}}}+{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}$$
(2)
$${{{{{\rm{M}}}}}}{=}{{{{{\rm{O}}}}}}+{{{{{{\rm{H}}}}}}}_{2}\to {{{{{\rm{M}}}}}}+{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}$$
(3)
$${{{{{\rm{M}}}}}}{=}{{{{{\rm{O}}}}}}+{{{{{\rm{M}}}}}}{-}{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}\to 2{{{{{\rm{M}}}}}}{-}{{{{{\rm{OH}}}}}}$$
(4)

It is interesting that the normalized intensity of M2+–O22− increases with increasing temperature even in the absence of O2 (Fig. 2d). The relation of normalized intensities of M2+–O22− and M2+–O2 is correlated (Supplementary Fig. S5). It was found that a standard deviation (R2) of the relation is 0.965, which clearly indicates that the normalized intensity of M2+–O22− is strongly correlated with that of M+–O2. Li et al. also reported similar phenomena48. The conversion reaction between M2+–O2 and M2+–O22− is shown in Eq. (5), where the valence state of the M cation in M+–O2 is kept constant via M+–O2 conversion to M2+–O22− after the formation of M–□–M.

$$\left({{{{{{\rm{O}}}}}}}_{2}^{-}\right){{{{{\rm{M}}}}}}{-}{{{{{\rm{O}}}}}}-{{{{{\rm{M}}}}}}+{{{{{{\rm{H}}}}}}}_{2}\to \left({{{{{{\rm{O}}}}}}}_{2}^{2-}\right){{{{{\rm{M}}}}}}{-}{{\square }}{-}{{{{{\rm{M}}}}}}+{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}$$
(5)

H2 oxidation by surface oxygen species in the presence of O2

The DRIFTS spectra, MS signals, and normalized intensities during H2 oxidation by oxygen species at β-MnO2 surfaces in the presence of O2 are presented in Fig. 3. The primary difference due to the presence of O2 is that the normalized intensity of v(OH) increases with temperature in the presence of O2 (Fig. 3d), but decreases in the absence of O2 (Fig. 2d). It is also noted that H2O MS signal (3.0E-09) at 400 °C in the presence of O2 (Fig. 3c) is much stronger than that (1.48E-09) in Fig. 2c in the absence of O2. These differences in the normalized intensity of v(OH) trend and H2O MS signal are evidence of O2 involvement in H2 oxidation. O2 can promote not only the release of O in M+–O2−–M+ (Eq. (1)) but also the formation rate of M–OH from M2+–O2 (Eq. (6)). With the increase in temperature, M–OH reacts with H2 (Eq. (7)) to form surface adsorption of H2O that desorbs into gaseous H2O at 250 °C (Eq. (8))49, resulting in the formation of terminal vacancies (bare Mn).

$$2{{{{{\rm{M}}}}}}{=}{{{{{\rm{O}}}}}}+{{{{{{\rm{H}}}}}}}_{2}\to 2{{{{{\rm{M}}}}}}{-}{{{{{\rm{OH}}}}}}$$
(6)
$$2{{{{{\rm{M}}}}}}{-}{{{{{\rm{OH}}}}}}+{{{{{{\rm{H}}}}}}}_{2}\to 2{{{{{\rm{M}}}}}}{-}{{{{{{\rm{OH}}}}}}}_{2}$$
(7)
$${{{{{\rm{M}}}}}}{-}{{{{{{\rm{OH}}}}}}}_{2}\to {{{{{\rm{M}}}}}}+{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}$$
(8)
Fig. 3: Experimental results of H2 oxidation by oxygen species at β-MnO2 surfaces in (H2 + O2) as a function of temperature.
figure 3

a, b DRIFTS spectra. c MS signal. d Normalized intensities where the error bars are the standard deviations obtained by measuring the peak heights more than three times.

Regeneration of H2-reduced β-MnO2 with Ar or O2

The fact that the normalized intensities of M+–O2–M+ and M2+–O2 are all negative during H2 oxidation in both H2/Ar and (H2 + O2) atmospheres (Figs. 2 and 3) indicates that OVs (M–□–M and M) can be generated even in the presence of O2. A similar result has been reported by Sun et al., where they found that M+–O2–M+ and M2+–O2 can be reduced by CO on ZnO50. The generation of the M–□–M and M may be due to either the decomposition rate of O2 at β-MnO2 surfaces is less than that of H2 oxidation or M–□–M and M cannot be regenerated. This issue was evaluated by conducting successive regeneration experiments in an Ar or O2/Ar atmosphere (Supplementary Table S2). β-MnO2 was first reduced by H2 in the TPR cell at 200 °C for 10 min, the regeneration was then carried out in an Ar or O2/Ar atmosphere by elevating the temperature from 25 °C to 400 °C.

DRIFTS spectra and normalized intensities at various temperatures during the regeneration of H2-reduced β-MnO2 in the Ar atmosphere are presented in Fig. 4. When increasing temperature from 25 °C to 300 °C, the normalized intensity of M2+–O22− decreases rapidly to zero, while those of M2+–O2 and M+–O2 increase rapidly to zero (Fig. 4b). The normalized intensities of M+–O2–M+ and M+–O asymptotically approach to –1.0 at a temperature close to 400 °C. This finding indicated that O atoms in M+–O2–M+ and O2 molecules in M2+–O22− can migrate on β-MnO2 surfaces (Eqs. (9) and (10))32.

$${{{{{\rm{M}}}}}}{-}{{{{{\rm{O}}}}}}-{{{{{\rm{M}}}}}}+{{{{{\rm{M}}}}}}\to {{{{{\rm{M}}}}}}{-}{{\square }}{-}{{{{{\rm{M}}}}}}+{{{{{\rm{M}}}}}}{=}{{{{{\rm{O}}}}}}$$
(9)
$$\left({{{{{{\rm{O}}}}}}}_{2}^{2-}\right){{{{{\rm{M}}}}}} {-}{{\square }}{-}{{{{{\rm{M}}}}}}+{{{{{\rm{M}}}}}}{-}{{{{{\rm{O}}}}}}-{{{{{\rm{M}}}}}}\to {{{{{\rm{M}}}}}}{-}{{\square }}{-}{{{{{\rm{M}}}}}}\\ +\left({{{{{{\rm{O}}}}}}}_{2}^{-}\right){{{{{\rm{M}}}}}}{-}{{{{{\rm{O}}}}}}-{{{{{\rm{M}}}}}}$$
(10)
Fig. 4: Experimental results of the regeneration of H2-reduced β-MnO2 in Ar at various temperatures.
figure 4

a DRIFTS spectra. b normalized intensities where the error bars are the standard deviations obtained by measuring the peak heights more than three times.

DRIFTS spectra and normalized intensities at various temperatures during the regeneration of H2-reduced β-MnO2 in the O2/Ar atmosphere are illustrated in Fig. 5. The normalized intensity of M+–O2–M+ increases from at a temperature higher than 100 °C, indicating that the regeneration of M+–O2–M+ from M–□–M and O2 requires a temperature higher than 100 °C51,52. Furthermore, the normalized intensity of M+–O2–M+ becomes positive at temperatures greater than 250 °C, at which all other surface oxygen species increase or decrease to 0.0, suggesting all other surface oxygen species have been completely regenerated. We may further note that M+–O2–M+ can convert to M2+–O2 (Eq. (9)). From the fact that the normalized intensity of M2+–O2 increases little at temperatures greater than 250 °C but that of M+–O2–M+ increases significantly, these findings deduce that the reaction in Eq. (9) is reversible.

Fig. 5: Experimental results of the regeneration of H2-reduced β-MnO2 in O2/Ar at various temperatures.
figure 5

a DRIFTS spectra. b normalized intensities where the error bars are the standard deviations obtained by measuring the peak heights more than three times.

The decreases in normalized intensities of δ(H2O) and v(OH) indicate that H2O can desorb from M–OH2 and M–OH at β-MnO2 surface, resulting in the formation of bare M.

Roles of OVs in H2 catalytic oxidation

The roles of OVs in H2 oxidation process at β-MnO2 surfaces in the presence of O2 can be deduced from the above discussion, and the proposed mechanism is illustrated in Fig. 6. First, when the reaction temperature is in a range of 110–150 °C, the oxygen atom in the bridge-type M+–O2––M+ can react with H2 to form H2O and OV via steps (1) and (6) in Fig. 6a. According to steps (2) and (7), the oxygen atoms in the terminal-type M2+–O2– and M–OH react with H2 to generate surface M–OH and gaseous H2O. The gaseous O2 adsorbed at the bare M site in M+–□–M+ (step (4)) yields M2+–O22−. M2+–O22− dissociates simultaneously to M+–O2––M+ and M2+–O2– (step (5), Eq. (11)). We can only find the decrease in M+–O2––M+ and increase in M–□–M in this temperature range as the step (2) is rate limited reaction.

$$\left({{{{{{\rm{O}}}}}}}_{2}^{2-}\right){{{{{\rm{M}}}}}}{-}{{\square }}{-}{{{{{\rm{M}}}}}}\to {{{{{\rm{M}}}}}}{-}{{{{{\rm{O}}}}}}{-}{{{{{\rm{M}}}}}}{=}{{{{{\rm{O}}}}}}$$
(11)
Fig. 6: Roles and mechanisms of surface oxygen species and OVs in H2 oxidation at β-MnO2 surfaces.
figure 6

a H2 oxidation between 110 and 150 °C. b H2 oxidation at a temperature higher than 150 °C.

As the oxidation process in Fig. 6b, when the temperature is above 150 °C, OH in M–OH becomes reactive enough, the O2 dissociation step (6) is slowest, resulting in the accumulation of M–□–M and M2+–O22−.

Conclusion

We explored O2 dissociation, OVs formation, and surface oxygen species conversion during H2 oxidation at β-MnO2 surface using the operando TPR-DRIFTS-MS technology. The results demonstrate that the operando TPR-DRIFTS-MS technology employed herein is a highly useful tool for identifying OVs at β-MnO2 surfaces and CeO2 and Co3O4 surfaces (Supplementary Fig. S6), and for understanding the roles of OVs and oxygen species in catalytic processes. In particular, the difference in the reaction characteristics of bridge-type (M+–O2–M+) and terminal-type (M2+–O2) oxygen species can be clearly observed using the operando TPR-DRIFTS-MS technology. Accordingly, we expect this technology could provide an important characterization method to understand the roles of surface oxygen species on metal oxide catalysts and enable the rational design of catalysts of OVs with satisfied performance.

Methods

Materials

β-MnO2 (99%) was purchased from Aladdin, Shanghai, China. Pure Ar (99.999%), pure O2 (99.999%), 5 vol% H2 standard gas (Ar balanced), and 5 vol% CO standard gas (Ar balanced) were purchased from Huayang, Changzhou, China.

Catalyst characterization

Physicochemical properties of β-MnO2 were characterized by via various techniques, such as X-ray powder diffraction (XRD), thermogravimetric (TG) analysis, Raman, X-ray photoelectron spectroscopy (XPS), oxygen temperature-programmed desorption (O2-TPD), and high-resolution transmission electron microscopy (HRTEM).

DRIFTS-MS system

A schematic diagram of the operando TPR-DRIFTS-MS system is shown in Supplementary Fig. S7. The system consisted of gas cylinders, gas flow meters (MFC, D07, Sevenstars Beijing, China), operando DRIFTS (Nicolet 50, Thermo Scientific, USA), and MS (Tilon LC-D200M, Ametek, USA). The DRIFTS was equipped with a TPR cell (HVC-DRP-5, Harrick, USA) and a narrow-band mercury cadmium telluride (MCT-A) detector with liquid nitrogen cooling for high sensitivity (0.09 cm−1) in collecting DRIFTS spectra between 4000 and 650 cm−1.

For the DRIFTS spectrum collection experiment, β-MnO2 powders were pretreated for 1 h in Ar at 450 °C (20 mL/min). Then, the β-MnO2 powders were cooled to room temperature and stabilized for 10 min, and DRIFTS background spectra were collected. The gas mixture (Supplementary Table S2) was supplied into the TPR cell for 20 min. The β-MnO2 powder temperature was elevated with programmed heating using a temperature controller. Series software was used to collect the corresponding spectra. Thirty-two scans were performed with a resolution of 4 cm−1, and the spectrum data of DRIFTS were analyzed using OMNIC software during the acquisition. The Kubelka–Munk function was used to convert the obtained spectra into absorption spectra, whose intensities were linearly related to the amount of adsorption. The gas from the TPR cell was analyzed using mass spectrometer (MS) (Tilon, LC-D200M, Ametek, USA) to obtain signals of H2 (m/z = 2), H2O (m/z = 18), O2 (m/z = 32), and Ar (m/z = 40).

Normalization of peak intensity

The collected infrared spectra at different temperatures were normalized for relatively quantitative analysis. The normalization was calculated using Eqs. (12) and (13) based on the absolute values of the highest height (Pimax) for positive peaks and lowest peak height (Pimin) for negative peaks, respectively.

$${N}_{i}=\frac{{P}_{i}}{{P}_{i\,{max }}}$$
(12)
$${N}_{i}=\frac{{P}_{i}}{\left|{P}_{i\,{min }}\right|}$$
(13)

Ni represents the normalized intensity of the absorption peak i at the corresponding temperature; Pi represents the peak height of the absorption peak i at the corresponding temperature.