Introduction

ZnO (zinc oxide), one of the most important wide-band gap semiconductors, has great potential applications in the field of solar cells1, sensors2,3, light emissions4,5,6, and catalysts7,8. For catalytic applications, ZnO is an essential component in composite catalysts, playing a key role in H2 activation and hydrogenation reactions, such as methanol and light-olefins synthesis from H2/CO/CO29,10,11,12.

For catalytic hydrogenation reaction over ZnO catalyst, the hydride species (Zn-H) formed from the dissociation of H2 on ZnO surface are usually recognized as the active species responsible for relevant hydrogenation processes13,14. Yang et al. showed that H2 can dissociate heterolytically at the ZnO surface and the generated Zn-H species hydrogenate CO/CO2 to formate15. Moreover, most studies have demonstrated that a bridging Zn-H-Zn structure originated from H2 dissociation at an oxygen vacancy exhibits reactivity toward hydrogenation of unsaturated hydrocarbon species by both computational and experimental studies. For example, previous calculation studies have shown that the Zn-H-Zn species at oxygen vacancies formed from H2 cleavage facilitate the reduction of CO to formyl in the methanol synthesis process16. Gong et al. also investigated the catalytic activity and selectivity of this bridging hydride generated via H2 dissociation at ZnO defect sites for CO2 hydrogenation using theoretical methods17. By applying various characterization methods (e.g., XPS, H2-TPD, H/D exchange, and in situ IR), Li et al. studied the hydrogenation of CO2 to methanol over composite ZnO catalysts and found that the Zn-H-Zn species formed upon H2 introduction participated in CO2 activation and enhanced catalytic performance18, which is in agreement with the findings of Polarz et al19. Despite of these advancements, the effects of different conditions on the active sites on the surfaces of ZnO require further investigation.

Solid-state NMR spectroscopy is a powerful technique that has been widely used in studying the local environments of solid materials20,21. For ZnO, previous studies have shown that surface hydrogen species, interstitial hydrogen (Hi) and oxygen-substitutional H (HO) species in ZnO nanoparticles can be distinguished by their 1H NMR shifts22,23, while the assignments of 1H resonances are still under debate24,25,26. More recently, 1H solid-state NMR has also been applied to study the H2 activation and catalytic properties of H species on ZnO27,28.

Here, we apply 1H solid-state NMR spectroscopy combined with DFT calculations to study the hydrogen species of ZnO nanorods at different conditions. In addition to H2 homolysis at oxygen vacancy, which is a widely recognized pathway for the generation of active hydride species, the migration of surface hydroxyl to nearby oxygen vacancy also leads to the formation of a bridging hydride species in the form of Zn-H-Zn, which are active toward CO2 hydrogenation. This finding shows the importance of hydroxyl species combined with oxygen vacancy for heterogeneous catalysis.

Results

Morphology of the ZnO sample

Transmission electron microscopy (TEM) and X-ray diffraction (XRD) were used to characterize the morphologies and structure. The XRD pattern confirms that the prepared ZnO nanorods have a wurtzite structure (Supplementary Fig. S1a), which are highly crystalline and predominately covered by nonpolar (10\(\bar{1}\)0) facets based on the TEM/HRTEM images (Supplementary Figs. S1b, S1c).

Different surface hydrogen species investigated with 1H NMR

The 1H NMR spectra of ZnO nanorods before and after exposing to vacuum at room temperature are given in Supplementary Fig. S2. According to our previous study27, the peaks at 4.8 and 0–3 ppm are assigned to physisorbed water, and surface hydroxyl and/or water, respectively. As obtained ZnO nanorods exposed to vacuum at room temperature and high temperatures (X °C) are denoted as “ZnO-RT-vac” and “ZnO-X-vac”, respectively, and Fig. 1a shows that the spectral intensity of ZnO nanorods decreases gradually with the increase of heating temperature under vacuum. It is necessary to point out that the center of the broad peak at 0–3 ppm gradually shifts to lower frequencies (1.5 and 0.9 ppm for ZnO-100-vac and ZnO-300-vac, respectively) upon dehydration and dehydroxylation, indicating that these species may be distinguished by their 1H NMR shifts. For ZnO-400-vac, the spectral intensity further decreases in comparison with ZnO-300-vac, while a relatively small change is observed after being heated at 500 °C (ZnO-500-vac) (Fig. 1a and Supplementary Fig. S3). The H/D exchange experiment shown in Supplementary Fig. S4 suggests that the residual resonances in ZnO-400-vac/ZnO-500-vac can be mainly ascribed to bulk hydrogen species of ZnO nanorods (e.g., H impurity), while a large fraction of surface hydrogen species is present on ZnO nanorods heated below 300 °C.

Fig. 1: 1H NMR spectra of ZnO nanorods and the structure models of ZnO used in the DFT calculations.
Fig. 1: 1H NMR spectra of ZnO nanorods and the structure models of ZnO used in the DFT calculations.
Full size image

a 1H NMR spectra of ZnO nanorods after being heated at different temperatures under vacuum. An optimized recycle delay of 5 s was used, while 64 acquisitions were collected for all spectra. b 1H-17O TRAPDOR NMR spectra of ZnO-RT-vac. 17O irradiation time: 0.2 ms; MAS rate: 5 kHz; recycle delay: 2 s; 17O irradiation radio frequency (RF) field strength: 64 kHz. The difference spectrum is obtained by subtracting the double resonance spectrum from the control spectrum. Asterisks denotes spinning sidebands. c The structure models of ZnO nanorods used for DFT calculations. Top: the optimized ZnO(10\(\bar{1}\)0) surface with a monolayer of water molecules (1 ML) in which half of the water molecules are dissociatively adsorbed while the other half molecularly adsorbed (model M1D1); bottom: ZnO(10\(\bar{1}\)0) surface with 1 ML water molecules in which all of the water molecules are molecularly adsorbed (model M2, also see Supplementary Fig. S6 for the structure). Gray and white spheres represent Zn and H atoms, red and blue spheres represent O atoms.

To confirm the spectral assignments, we further investigated the H-O proximity of ZnO-RT-vac by 1H-17O transfer of population in the Double Resonance (TRAPDOR) NMR method29. Typical TRAPDOR NMR spectra obtained with a rather short 17O irradiation time of about 0.2 ms, in which only the 17O ions in the first coordination shell (i.e., 17O directly bound to H) are close enough to have an impact, show that the intensity of the resonance at 0.9 ppm decreases significantly in the double resonance experiment (Fig. 1b). Hence, the 1H resonance centered at 0.9 ppm can be assigned to surface hydroxyl groups, while the relatively high-frequency signals centered at 1.5 ppm (in the range of 0–3 ppm) arise from adsorbed water (Fig. 1a and Supplementary Fig. S3), which is consistent with the trend that the center of the peak in the double resonance experiment gradually shifts to higher frequency from 0.9 ppm with the increase of 17O irradiation time (Supplementary Fig. S5).

DFT calculations of 1H chemical shifts were performed for the two most energetically favorable ZnO structural models identified in previous literature30,31 to help spectral assignments (Fig. 1c). In model M1D1, the calculated δH of surface OH is close to 0.9 ppm, which is the observed shift of the major peak in the spectrum of ZnO-300-vac (Fig. 1a). The calculated δH values of two H species in one water molecule are very different. However, due to rapid motion, an average value is expected in experimental observations. The average value of calculated δHs of water is 2.6, similar to the resonances at 0–3 ppm in 1H NMR spectrum of ZnO-RT-vac and ZnO-100-vac. In model M2, the average values of δHs for water molecules are in a range of 2.2–3.2 (Fig. 1c), which also agree with the frequency of the peaks at 0–3 ppm in ZnO-RT-vac and ZnO-100-vac. Therefore, the signals at 0–3 ppm (centered at 1.5 ppm) can be ascribed to surface-adsorbed water.

Notably, a high-frequency peak centered at 8.4 ppm emerges in 1H NMR spectrum of ZnO nanorods after being heated at 100–250 °C under vacuum, with the strongest signal observed at 200 °C (ZnO-200-vac) (Fig. 1a and Supplementary Fig. S7). This peak disappears in ZnO-300-vac, suggesting that the corresponding species is dependent on the temperature. Given the strong resonance at 8.4 ppm observed after vacuum treatment at 200 °C, the following investigations of this peak were mainly carried out using ZnO-200-vac samples. After exposing ZnO-200-vac to water vapor for 3 h and 3 days (Supplementary Fig. S8), only a slight decrease in intensity for the 8.4 ppm peak is observed, suggesting that the hydrogen species responsible for this signal are stable at ambient temperature. Supplementary Fig. S9 shows that hydrogen atoms contributing to this high-frequency peak can be exchanged by D2, indicating that this peak arises from hydrogen species close to the ZnO surface. A peak at 8.4 ppm has been observed in our previous work in H2 activation on ZnO nanorods27, which originates from homolytic H2 dissociation at an oxygen vacancy, leading to the formation of a bridging Zn-H-Zn hydride species. Therefore, the same hydride structure is expected in this situation. For this purpose, 1H-17O TRAPDOR NMR experiments for ZnO-200-vac were performed (Supplementary Fig. S10). The intensity of the peak at 8.4 ppm remains the same in the double resonance experiment, indicating that this hydrogen species is far away from O ions, thus the 1H resonance can be assigned to H bound to Zn ions and/or trapped in defect sites such as O vacancy.

Considering the fact that the peak at 8.4 ppm in spectrum of ZnO-200-vac is the most intense among the samples treated at different temperatures, 1H NMR experiments for ZnO-300-vac after being heated at 200 °C in a sealed glass tube under near-vacuum conditions (~50 Pa) ((ZnO-300-vac)−200, see experimental section for details) were carried out to further explore the nature of this peak. The deceased intensity of the peak at 0.9 ppm due to surface OH is observed along with the increase of the resonance at 8.4 ppm (Fig. 2a). On the whole, Table S1 shows that the increase of the high-frequency signal (8.4 ppm) in NMR spectrum during thermal treatment for ZnO-300-vac is comparable to the decrease of the signal at 0.9 ppm, implying that the peak at 8.4 ppm may arise from the conversion of surface OH sites during thermal treatment at 200 °C. At the same time, the peak attributed to oxygen vacancies (g = 2.00) becomes weaker in (ZnO-300-vac)-200 than in ZnO-300-vac, as observed in the electron paramagnetic resonance (EPR) spectra (Fig. 2b). Therefore, the high-frequency peak at 8.4 ppm can be ascribed to Zn-H-Zn formed by the migration of surface hydroxyl groups to oxygen vacancies, which is further supported by DFT calculations of chemical shifts (Fig. 2c and Supplementary Fig. S11). In addition, the peak at 8.4 ppm observed in (ZnO-300-vac)−200 in Fig. 2a is relatively weak, which may be relevant to the distributions of surface hydroxyl and oxygen vacancy on the ZnO surface. Since the concentration of oxygen vacancy is expected to be much lower than surface hydroxyls, most hydroxyls are far away from oxygen vacancy, and only the hydroxyl close to oxygen vacancy can be transformed to Zn-H-Zn successfully.

Fig. 2: Characterization and calculation for the conversion of surface OH to nearby O vacancy.
Fig. 2: Characterization and calculation for the conversion of surface OH to nearby O vacancy.
Full size image

a 1H NMR spectra of ZnO-300-vac and the ZnO-300-vac sample after being heated at 200 °C in a sealed glass tube under near-vacuum conditions (~50 Pa) (denoted as (ZnO-300-vac)−200) for 3 h. The 1H NMR spectrum of (ZnO-300-vac)-100 is also shown in Supplementary Fig. S12, together with those of ZnO-300-vac and (ZnO-300-vac)-200, in which a weaker resonance at 8.4 ppm is observed due to the lower extent of hydrogen migration at 100 °C compared to 200 °C. Note that the ZnO-300-vac was used to exclude the effects of water desorption on the origin of the resonance at 8.4 ppm during the heating process at 200 °C. Recycle delays of 5 s were used, and 64 scans were accumulated for the two spectra. b EPR spectra of ZnO-300-vac and (ZnO-300-vac)−200. The signal at g = 2.00 arises from unpaired electrons trapped in oxygen vacancies, the signal at g = 1.96 represents unpaired electrons trapped from the conductive band (CB) by shallow donors or impurities32. c The calculated models of ZnO nanorods used for the conversion of surface OH to nearby oxygen vacancy. Gray, red, and white spheres represent Zn, O, and H atoms, respectively. The dashed circles represent an oxygen vacancy.

1H NMR experiments were further carried out to explore the effects of oxygen vacancy concentrations in ZnO nanorods on the resonance at 8.4 ppm. Figure 3a shows the 1H NMR spectra of ZnO nanorods after being heated at 200 °C with (ZnO-200-vac) and without (ZnO-200) exposing to vacuum in comparison to ZnO-RT-vac (detailed experimental conditions can be found in the experimental sections), in which the high-frequency signal at 8.4 ppm emerges in the NMR spectrum of ZnO-200-vac, while this peak is absent in ZnO-200. ZnO-200-vac contains more oxygen vacancies than ZnO-200, as indicated by the stronger peak at g = 2.00 in the EPR spectrum of the former (Fig. 3b), further supporting that the resonance at 8.4 ppm is associated with oxygen vacancies.

Fig. 3: 1H NMR and EPR spectra to explore the effect of oxygen vacancy of ZnO nanorods on the 1H resonance at 8.4 ppm.
Fig. 3: 1H NMR and EPR spectra to explore the effect of oxygen vacancy of ZnO nanorods on the 1H resonance at 8.4 ppm.
Full size image

a 1H NMR spectra of ZnO nanorods with different heating conditions to study the effect of oxygen vacancy concentration on the spectral intensity at 8.4 ppm. b EPR spectra of ZnO-200-vac and ZnO-200 samples. c The corresponding EPR spectra of ZnO nanorods used in d before 200 °C thermal treatment under vacuum. d 1H NMR spectra of ZnO nanorods with different initial oxygen vacancy concentration, then being heated at 200 °C, to study the effect of different concentration of oxygen vacancy before 200 °C vacuum on the spectral intensity at 8.4 ppm. “(ZnO-300-vac)-SW-200-vac)” represents the sample of ZnO-300-vac after being exposed to a small amount of water, then being heated at 200 °C under vacuum. “(ZnO-300-vac)-LW−200-vac” corresponds to ZnO-300-vac after being exposed to saturated water, then being heated at 200 °C under vacuum. For each NMR spectrum, recycle delays of 5 s were used, and 64 scans were collected.

In addition, the relationship between the signal intensity at 8.4 ppm and the initial concentration of oxygen vacancies in ZnO nanorods was investigated. Supplementary Fig. S13 shows that the peak ascribed to oxygen vacancies increases significantly in intensities in the EPR spectrum after thermal treatment at 300 °C under vacuum, suggesting that ZnO nanorods with a high concentration of oxygen vacancies (ZnO-300-vac) can be prepared. Subsequently, ZnO-300-vac samples were exposed to varying amounts of water, as summarized in Supplementary Fig. S14. After exposing ZnO-300-vac to a small amount of water (denoted as (ZnO-300-vac)-SW), the EPR data show a decrease in the concentration of oxygen vacancies (Supplementary Fig. S15). For (ZnO-300-vac)-LW sample, obtained by exposing ZnO-300-vac to saturated water vapor, the oxygen vacancy concentration decreases further and becomes even lower than ZnO-RT-vac, as shown in the EPR data. The samples were heated at 200 °C under vacuum to generate bridging Zn-H-Zn species. With increasing initial oxygen vacancy concentration ((ZnO-300-vac)-LW <ZnO-RT-vac <(ZnO-300-vac)-SW), the intensity of the peak at 8.4 ppm increases after vacuum treatment at 200 °C (Fig. 3c, d and Supplementary Fig. S16), particularly for (ZnO-300-vac)-SW-200-vac. This suggests that oxygen vacancies can significantly promote the conversion of H in surface hydroxyl groups into bridging Zn-H-Zn species, giving rise to the high-frequency resonance at 8.4 ppm. Therefore, these results demonstrate that the concentration of bridging Zn-H-Zn species is closely related to temperature, the concentration of oxygen vacancy, and the amount of surface OH. Furthermore, DFT calculations show that these hydrogen species can desorb again at high temperatures (Supplementary Fig. S17), explaining the significant decrease in intensity of the 1H resonance at 8.4 ppm in ZnO-300-vac.

Activity of bridging Zn-H-Zn hydride species in CO2 hydrogenation to methanol investigated with 1H and 13C NMR

Based on above mentioned results, the bridging Zn-H-Zn hydride species located at O vacancy in ZnO nanorods disappear after being heated at 300 °C, thus, 1H and 13C NMR experiments were performed between 100 and 250 °C to investigate the role of this bridging hydride in CO2 hydrogenation. First, a ZnO-300-vac sample is prepared to study the interaction between surface OH groups and CO2. The amount of surface OH species decreases significantly after introducing CO2 to ZnO-300-vac at 100–200 °C (Supplementary Figs. S18a, S19a and S20a). The corresponding 13C NMR spectrum shows two sharp peaks at 171, 167 along with a broad peak at approximately 162 ppm (Supplementary Figs. S18b, S19b and S20b). According to previous reports in literatures28,33,34, these resonances at 171, 167, and 162 ppm can be assigned to the formate, carbonate, and bicarbonate species, respectively, derived from CO2 adsorption on the catalyst surface, in which formate species generate from the hydrogenation of carbonate and bicarbonate18.

(ZnO-300-vac)-SW-200-vac (denoted as ZnOZn-H-Zn), which has a high concentration of bridging hydride species derived from oxygen vacancy and nearby hydroxyl, was used to explore the impacts of this species on CO2 activation. Similar to the trend observed in Supplementary Figs. S18a, S19a and S20a, the amount of OH groups in all ZnOZn-H-Zn samples decreases after reacting with CO2 at different temperatures (Fig. 4), accompanied by the formation of formate, carbonate and bicarbonate species (Fig. 5). At 100 °C, the peak at 8.4 ppm, ascribed to bridging hydride species, remains almost unchanged (Fig. 4a), indicating that these species do not participate in CO2 activation at this temperature. At 150 °C, the resonance at 8.4 ppm becomes weaker after introducing CO2 (Fig. 4b). At the same time, 13C NMR spectrum shows a new resonance centered at 52 ppm (Fig. 5), which is assigned to methoxy species based on previous reports35,36,37. Li et al. studied CO2 hydrogenation to methanol over ZnZrOx catalysts and observed a decrease in formate species along with a simultaneous increase of methoxy species upon H2 introduction after CO2 adsorption on the catalyst surface, confirming the formation of methoxy species from formate hydrogenation38. Larmier et al. also provided evidence for the conversion of formate to methoxy species over Cu/ZrO2 catalysts during the methanol synthesis from CO2 hydrogenation using solid-state NMR spectroscopy39. Therefore, our results show that bridging Zn-H-Zn is the active hydride species for CO2 hydrogenation to methanol. In addition, since the hydrogenation of formate to methoxy occurs at a higher temperature than the formation of formate from chemisorbed CO2, our data also agree well with previous reports that the former process is kinetically more difficult40,41. After reacting with CO2 at 200 °C, the peak at 8.4 ppm disappears (Fig. 4c), while the 13C resonance at 52 ppm attributed to methoxy species remains (Fig. 5), indicating that more bridging hydride species participate in the CO2 hydrogenation reaction and that the same mechanism is involved. After introducing CO2 at a higher temperature of 250 °C, the signal at 8.4 ppm also disappears (Fig. 4d), while the peak at 52 ppm in the 13C NMR spectrum is not observed (Fig. 5). This is presumably due to desorption of methanol generated from methoxy hydrogenation, as no methanol signal was detected in in situ IR spectra in previous studies on CO2 hydrogenation to methanol18,38,42.

Fig. 4: 1H NMR spectra of ZnO nanorods before and after exposure to CO2.
Fig. 4: 1H NMR spectra of ZnO nanorods before and after exposure to CO2.
Full size image

CO2 (1000 mbar) was introduced to ZnO nanorods at 100 °C (a), 150 °C (b), 200 °C (c), and 250 °C (d) before the acquisition of the 1H NMR spectra. ‘ZnOZn-H-Zn’ represents (ZnO-300-vac)-SW-200-vac sample discussed in the above sections. Recycle delays of 5 s were used, and 64 scans were collected for each NMR spectrum.

Fig. 5: 13C CP MAS NMR spectra of ZnOZn-H-Zn samples ((ZnO-300-vac)-SW−200-vac) after exposing to CO2 (1000 mbar) at different temperatures.
Fig. 5: 13C CP MAS NMR spectra of ZnOZn-H-Zn samples ((ZnO-300-vac)-SW−200-vac) after exposing to CO2 (1000 mbar) at different temperatures.
Full size image

Recycle delays of 5 s were used, and 2000 acquisitions were accumulated for all the spectra.

The above results demonstrate that bridging hydride species are responsible for the hydrogenation of formate to methoxy species, and that methanol formation over ZnO nanorods is more likely to proceed via formate and methoxy intermediates in this case, consistent with methanol synthesis mechanism reported in the literature43,44,45,46,47,48. A possible mechanism for CO2 hydrogenation to methanol is proposed (Fig. 6).

Fig. 6: Schematic diagram for CO2 hydrogenation to methanol over ZnO nanorods in the presence of bridging hydride species.
Fig. 6: Schematic diagram for CO2 hydrogenation to methanol over ZnO nanorods in the presence of bridging hydride species.
Full size image

The red dashed circles represent oxygen vacancy.

The formation of bridging Zn-H-Zn hydride species from the migration of surface OH to nearby O vacancy under appropriate conditions in ZnO nanorods has been studied in detail using solid-state NMR spectroscopy. Within a given temperature range (100–250 °C), when abundant OH groups and O vacancies are present on the ZnO surface, OH migration into nearby vacancies produces bridging Zn-H-Zn hydrides, a process analogous to H2 homolysis at oxygen vacancies in ZnO. After introducing CO2 to ZnO nanorods containing bridging hydride species at different temperatures, 1H and 13C NMR results further show the activity of this bridging hydride species in CO2 hydrogenation toward methanol, in which hydride hydrogenates formate to methoxy species, confirming methanol production via formate and methoxy intermediates. Hydrogenation is traditionally considered to reply on gaseous H2, while our work shows an alternative route49,50. These findings reveal that surface hydroxyl and oxygen vacancy on ZnO can generate reactive Zn-H-Zn hydrides, enabling CO2 hydrogenation without direct H2 dissociation, provided that a parallel process (e.g., CO reduction or photoreduction)51,52 maintains the oxygen vacancy population. This opens new design principles for developing ZnO-based hydrogenation catalysts.

Methods

Material preparation

ZnO nanorods were synthesized hydrothermally53. First, 0.176 g Zn(Ac)2·2H2O and 0.064 g NaOH were dissolved in 40 mL of ethanol. The NaOH solution was then added to the Zn(Ac)2 solution dropwise under vigorous stirring, followed by adding 1.5 mL aqueous solution containing 0.095 g ethanediamine to the mixture. After stirring for 0.5 h, the solution was transferred to a 100 mL Teflon lined autoclave and maintained at 160 °C for 8 h. After it was cooled to room temperature, the white precipitate was centrifuged and washed with distilled water three times, before it was dried at 80 °C for 8 h to produce white ZnO nanorod powders. The prepared ZnO nanorods were subsequently heated at 300 °C under vacuum for 2 h.

Material characterization

The powder XRD analysis was performed on a Philips X’Pro X-ray diffractometer using Cu Kα irradiation (λ = 1.54184 Å) operated at 40 kV and 40 mA at 25 °C. The morphology and the exposed facets of the ZnO nanorods were characterized using a JEOL JEM-2010 type high-resolution transmission electron microscope. A Bruker EMX-10/12 spectrometer was applied to record EPR spectra of the samples at room temperature.

Measured sample preparation

ZnO-300-vac sample is obtained after exposing 150 mg ZnO nanorods to vacuum at 300 °C for 3 h. Subsequently, ZnO-300-vac is heated at 100 and 200 °C for 3 h in a sealed glass tube under near-vacuum conditions (~50 Pa) to obtain (ZnO-300-vac)−100 and (ZnO-300-vac)-200, respectively. ZnO-200 sample is achieved after 150 mg ZnO nanorods are heated at 200 °C for 3 h under atmospheric pressure. (ZnO-300-vac)-SW sample is obtained after exposing 150 mg ZnO nanorods to vacuum at 300 °C for 3 h, then adsorbed a small amount of water through the vacuum line. (ZnO-300-vac)-LW sample is obtained after exposing 150 mg ZnO nanorods to a vacuum at 300 °C for 3 h, then adsorbed saturated water. (ZnO-300-vac)-SW−200-vac and (ZnO-300-vac)-LW-200-vac samples are obtained after (ZnO-300-vac)-SW and (ZnO-300-vac)-LW samples are heated at 200 °C for 3 h under vacuum, respectively. “ZnOZn-H-Zn represents (ZnO-300-vac)-SW−200-vac sample.

Carbon dioxide activation

After the ZnOZn-H-Zn ((ZnO-300-vac)-SW-200-vac) sample is obtained, 1000 mbar CO2 is introduced to the sample in a sealed glass tube, and the mixture is heated at different temperatures from 100 to 250 °C before it is cooled to room temperature.

Solid-state NMR spectroscopy

Room temperature 1H and 13C MAS NMR spectra were measured on 9.4T Bruker Avance III spectrometers using 4.0 mm MAS probes double tuned to 1H at 400 MHz and 13C at 100.6 MHz, respectively. All samples were packed into rotors in an N2 glove box. 1H and 13C chemical shifts are referenced to adamantane at 1.79 ppm and hexamethylbenzene at 17.35 ppm (the methyl groups), respectively.

DFT calculations

The Perdew-Burke-Ernzerhof54 functional was employed to perform the spin-polarized periodic density functional theory (DFT) calculations using the Vienna ab initio Simulation Package55. The DFT + U approach56 was used. The effective U values of 7.0 and 10.0 eV were applied to the O 2p and Zn 3d orbitals, as suggested by a previous study57. The electron-core interaction was described by using the project-augmented wave method58 at a kinetic energy cutoff of 500 eV with Zn (5s, 5p, 5d, 6s), O (2s, 2p), and H (1s) electrons being treated as valence states. For geometry optimizations, the convergence criterion for electronic minimization was set to 1 × 10–5 eV, and the Hellman–Feynman force criterion of 0.02 eV/Å was applied to each relaxed ion. For chemical shift calculations, the convergence criterion for electronic minimization was set to 1 × 10–8 eV.

The calculated lattice parameters of bulk wurtzite ZnO are a = b = 3.097 Å and c = 4.995 Å; in this calculation, the 10 × 10 × 6 k-point mesh was used for the Brillouin-zone integration. ZnO(10\(\bar{1}\)0) was modeled by a slab model with a (5 × 3) surface cell and eight Zn-O layers. During the geometry optimizations, the bottom two Zn-O layers were fixed. To eliminate slab-slab interactions, a large vacuum gap of more than 10 Å along the z-direction was incorporated. The slab calculations employed a 1 × 1 × 1 k-point mesh.

The isotropic chemical shifts of 1H (δ1H) were calculated by

$${\delta }_{1{{\rm{H}}}}={\delta }_{{{\rm{cal}}}}+{\delta }_{{\rm{ref}}}$$

where δcal represents the unaligned DFT chemical shift and δref was determined by aligning the average 1H chemical shift of H in the adsorbed tetramethylsilane (−29.02 ppm) on ZnO(10\(\bar{1}\)0) to the corresponding experimental value of 0 ppm; the resulting δref is 29.02 ppm.

In calculating the desorption of hydrogen in oxygen vacancy, the ZnO(10\(\bar{1}\)0) surface was modeled using a four-layer slab with a p(3 × 2) surface cell. A large vacuum gap of more than 12 Å was introduced along the z-direction to avoid interactions between neighboring slabs. Transition states (TS) were identified using the climbing image nudged elastic band method59. The validity of the obtained TS structures was confirmed through frequency calculations.