Surface oxygen species dictate reaction pathways in platinum-catalyzed aqueous phase reforming of methanol

The transition to low-carbon energy systems urgently requires near-zero emissions and widely applicable renewable energy sources^[1,2], such as hydrogen (H₂)^[3,4]. Due to its low volumetric energy density (~3 kW m^–3)^[5,6], H₂ remains challenging to store and transport, restricting its widespread application^[7,8]. Methanol stands out owing to its liquid nature under ambient conditions, high H₂ content (12.5 wt%), and ease of conversion^[9]. The predominant strategy for H₂ production from methanol is gas-phase steam reforming, in which methanol vapor reacts with steam to generate CO₂ and H₂^[10]. However, this process typically requires high operating temperatures (> 300 °C) and an additional vaporization unit^[11,12]. Aqueous phase reforming of methanol (APRM, CH₃OH(l) + H₂O(l) → CO₂ + 3H₂) enables the release of H₂ from the liquid phase at relatively low temperatures (150–250 °C), using compact reactors^[13]. Moreover, this method exhibits a low by-product yield, particularly for CO and CH₄, which is advantageous for H₂ utilization in proton exchange membrane fuel cells (PEMFCs). Therefore, APRM is an emerging H₂ release system, particularly attractive for distributed and mobile H₂ supply, such as transportation applications^[14].

APRM is limited by unfavorable thermodynamics at low temperatures and high kinetic constraints associated with C–H or O–H bond cleavage in methanol decomposition and water-gas shift^[15,16]. Previous studies have shown that Pt nanoparticles supported on oxide supports (e.g., Pt/Al₂O₃, Pt/CeO₂, Pt/NiAl₂O₄) are uniquely capable of promoting APRM activity via acid/alkaline sites^[17−19], oxygen vacancies^[20,21], or metal-support interaction^[22−24]. For example, K⁺-doped Pt/Al₂O₃ catalysts (PtK_x/Al₂O₃), which possess more alkaline sites, can affect the Pt d-band center and stabilize the *OH intermediates to improve the H₂ production rate^[17]. Pt single-atom-loaded oxygen-vacancy-rich CeO₂ catalysts (Pt₁/CeO₂) showed excellent activity in APRM, attributed to the role of Ce³⁺ ions and oxygen vacancies in promoting water activation^[20]. Pt/NiAl₂O₄ displayed superior activity and stability due to synergistic interaction between Pt and spinel, which facilitated the redox pathway of water-gas shift reaction (WGS, CO + H₂O → CO₂ + H₂)^[22]. Although considerable effort has been devoted to understanding metal active sites and reaction intermediates in APRM, the roles of oxygen species have not been fully understood.

This study elucidates how surface oxygen species, including hydroxyls, lattice oxygen, and inert oxygen species, govern the catalytic performance of APRM. We find that inert oxygen species induce a weak metal-support interaction, and thus the ability of Pt sites to activate methanol toward CO* and H₂ is limited. By contrast, catalysts with surface hydroxyls and reactive oxygen species promote methanol decomposition on Pt sites via strong metal-support interaction and enhance H₂ generation by supplying OH* sources for CO* conversion. Collectively, these findings demonstrate that surface oxygen species dictate catalytic performance and provide mechanistic guidance for the rational design of oxide-supported catalysts for APRM.

A series of Pt-based catalysts supported on various oxides (Pt/Al₂O₃, Pt/ZrO₂, Pt/CeO₂, Pt/TiO₂, and Pt/SiO₂) was synthesized via the wet impregnation method. The Pt precursor was H₂PtCl₆∙6H₂O (8 wt% in water), purchased from Shanghai Macklin Biochemical Co., Ltd. The information on the five supports used is as follows: Al₂O₃ (A766523, Shanghai Macklin Biochemical Co., Ltd), SiO₂ (S861577, Shanghai Macklin Biochemical Co., Ltd), ZrO₂ (J2211262, Shanghai Aladdin Biochemical Technology Co., Ltd), CeO₂ (C103981, Shanghai Aladdin Biochemical Technology Co., Ltd), and TiO₂ (T818930, Shanghai Macklin Biochemical Co., Ltd). Before the impregnation, all oxide supports were calcined at 500 °C for 4 h in air.

For the impregnation process, 1 g of pretreated supports was added to a beaker containing 0.27 g H₂PtCl₆∙6H₂O solution. Then, 50 mL of deionized water was added to the beaker to obtain a homogeneous suspension. The mixture was continuously stirred at 80 °C overnight until the water had evaporated completely. Finally, the resulting powder was calcined in a muffle or tube furnace at 500 °C for 4 h.

Catalytic performance test

Prior to evaluating catalytic performance, the Pt-based catalysts were reduced in a 10% H₂/N₂ flow at 500 °C for 2 h. After the pretreatment process, 0.1 g of the pre-reduced catalysts and 30 mL of a methanol-water mixture (methanol/water molar ratio of 1:3) were loaded into an autoclave reactor. The sealed autoclave was purged using N₂ to remove the remaining air in the reactor, after which a certain amount of N₂ was introduced as the internal standard. The reactor was subsequently heated from room temperature to the target temperatures (190, 210, 230, or 250 °C) over approximately 30 min, and then maintained at the set temperature for 30 min. After the reaction, the gaseous product (N₂, H₂, CO₂, CO, and CH₄) in the reactor was collected in a gasbag and analyzed by micro gas chromatography (Inficon, Micro GC) using a thermal conductivity detector (TCD). The amount of gas products, including H₂, CO, CO₂, and CH₄, was quantified using the ratio of their peak areas to the internal standard gas (N₂).

To evaluate the catalytic activity, we calculated the average H₂ production rate ($R_{\rm{H}_2} $, μmol g_Pt^–1 s^–1), which was defined as the amount of H₂ produced per unit mass of Pt per unit time, as shown in Eq. (1). The selectivity of methanol reforming (S_reforming) and gaseous by-products including CO and CH₄ (S_CO, $S_{ \rm{CH_4}} $ %) were calculated by Eqs (2)–(4):

In Eqs (1)–(4), m_Pt (g) is the amount of Pt loaded on the surface per gram of catalyst (Supplementary Table S1), t (s) is the constant temperature time of the reactor. In the formula for CO, CH₄, and CO₂ selectivity, n_CO (μmol), $n_{\rm{CH_4}} $ (μmol), and $n_{\rm{CO_2}} $ (μmol) are the mole quantities of CO, CH₄, and CO₂, and n_methanol corresponds to the consumed methanol, which is equal to the total molar amount of CO, CO₂, and CH₄.

Catalyst characterization

The morphology of Pt/oxides was investigated by scanning electron microscopy (SEM) and transmission electron microscopy (TEM), using JEOL JEM-2100 Plus and JSM-7900F, respectively. The Pt loading of Pt/oxide catalysts was quantified using inductively coupled plasma–optical emission spectroscopy (ICP-OES, iCAP 7400). Surface areas and pore-size distribution of supports were measured by N₂ isothermal adsorption-desorption (Micromeritics ASAP 2460) using the Brunauer-Emmett-Teller (BET) model. X-ray diffraction (XRD) patterns were collected using a Malvern Panalytical Empyrean diffractometer with Cu Kα radiation over a 2θ range of 10–90° at a step size of 0.0133°. X-ray photoelectron spectroscopy (XPS) was carried out on a Thermo Fisher Scientific ESCALAB 250Xi instrument. The H₂ temperature- programmed reduction (H₂-TPR), CO₂ or NH₃ temperature-programmed desorption (CO₂-TPD, NH₃-TPD) were conducted on Micromeritics AutoChem II 2920. For in-situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS), measurements were carried out using a Thermo Fisher Nicolet IS50 instrument with a Mercury Cadmium Telluride (MCT) detector. The pretreated catalysts were loaded into the chamber and exposed to N₂, followed by heating to 230 °C. Subsequently, the reactants (CH₃OH/H₂O = 1:3 or pure CH₃OH or H₂O) were introduced to the chamber by N₂ bubbling for 30 min, followed by N₂ purging for another 30 min. Background spectra were collected after the target temperature was reached under an N₂ atmosphere, and then the spectra were collected every 5 min while the reactants were introduced.

The catalytic activities of various 1 wt% Pt-based catalysts were evaluated across a temperature range of 190–250 °C at different pressures in an autoclave reactor (Supplementary Fig. S1). Regardless of oxide supports, the H₂ production rate increases with increasing temperature (Fig. 1a), which can be attributed to the endothermic nature of APRM^[15]. Pt-based catalysts supported on different oxides exhibit markedly different activities. The highest activity is observed for amphoteric oxides (Al₂O₃ and ZrO₂)^[25,26], followed by reducible oxides (CeO₂ and TiO₂), and inert SiO₂ displays the lowest activity. At 190 °C, the H₂ production rate over Pt/Al₂O₃ is nearly fourfold higher than that over Pt/CeO₂ and approximately 20 times higher than that over Pt/SiO₂. The activity advantage remained pronounced at 250 °C: Pt/Al₂O₃ exhibited an H₂ production rate of 846.9 μmol g_Pt^–1 s^–1, which is nearly 2.5 times higher than Pt/CeO₂ (329.8 μmol g_Pt^–1 s^–1), and 20 times higher than Pt/SiO₂ (41.4 μmol g_Pt^–1 s^–1) at 250 °C.

Figure 1. 

Catalytic performance of 1 wt% Pt/oxide catalysts. (a) H₂ production rates as a function of temperature. (b) Reforming selectivity at 190 and 250 °C. (c) Stability test over seven cycles at 250 °C. Reaction conditions: 30 mL methanol-water mixture (n[CH₃OH] : n[H₂O] = 1:3), 100 mg catalyst loading, 30 min, 0.5 MPa.

The reforming selectivity of Pt/oxide catalysts was evaluated at various temperatures, which shows that except for Pt/ZrO₂, all catalysts exhibit reforming selectivity of around 95% even at 230 °C (Fig. 1b). Regarding hydrocarbon byproducts, trace amounts of CH₄ can be detected over most Pt/oxide catalysts, while no detectable CH₄ is observed over Pt/TiO₂ and Pt/SiO₂ (Supplementary Figs S2–S6). We propose that the formation of CH₄ occurs under high pressure and H₂-rich atmosphere via CO₂ or CO methanation (CO₂ + 4H₂ → CH₄ + 2H₂O, CO + 3H₂ → CH₄ + H₂O)^[27]. CO byproducts can poison separation membranes and deactivate PEMFCs^[28], making their suppression critically important. Pt/SiO₂ exhibits the lowest CO selectivity, remaining below the detection range of GC from 190 to 250 °C, owing to its limited activity in methanol decomposition (Supplementary Fig. S7). However, as the H₂ production rate increases, the CO selectivity of Pt/Al₂O₃, Pt/CeO₂, Pt/TiO₂, and Pt/ZrO₂ reaches 0.6%, 1.9%, 0.7%, and 4.1%, respectively, at 250 °C. This variation is governed by the two-step mechanism of APRM, which includes methanol decomposition (MD, CH₃OH → CO + 2H₂) and water-gas shift (WGS, CO + H₂O → CO₂ + H₂)^[21,29]. Higher H₂ production rates typically arise from the concurrent enhancement of MD and the WGS reaction, but an overly accelerated MD pathway may result in elevated CO selectivity^[30]. The higher CO selectivity observed over Pt/ZrO₂ originates from a mismatch between the rates of MD and WGS reaction.

Cycling tests were performed to assess the stability of the Pt/oxide catalysts (Fig. 1c). The result shows that most catalysts remain stable over seven cycles with minimal activity loss. Among the tested catalysts, Pt/ZrO₂ shows the highest stability, followed by Pt/SiO₂ and Pt/TiO₂, with only minor activity losses of 2.7%, 6.8%, and 7.3%, respectively. The activity of Pt/CeO₂ decreases sharply after four cycles, from 212.5 to 130.5 μmol g_Pt^–1 s^–1, owing to the buildup of the adsorbed intermediates^[31]. The most active catalyst, Pt/Al₂O₃, shows a slight deactivation, with activity decreasing from 880.7 to 812.3 μmol g_Pt^–1 s^–1. This decline may be attributed to hydroxylation of the oxide support under reaction conditions, possibly leading to aggregation of Pt species^[32,33].

Characterization of Pt/oxide catalysts

The different oxide supports all show a spherical morphology, as observed in the SEM images (Supplementary Figs S8–S12) with varying particle sizes. SiO₂ and ZrO₂ particles show the largest particle size (~20 nm), while Al₂O₃, TiO₂, and CeO₂ are significantly smaller (nearly 10 nm). Importantly, after Pt loading, the morphology and particle size of the supports remain largely unchanged (Fig. 2a–e). TEM images further confirm the packed spherical structures of the supports, with Pt nanoparticles uniformly distributed on their surfaces. Statistical analysis reveals that the Pt particles have a narrow size distribution (1.9–3.4 nm) with no evident aggregation. The average Pt particle size increases in the order: Pt/ZrO₂, Pt/CeO₂, Pt/TiO₂, Pt/Al₂O₃, and Pt/SiO₂ (Fig. 2f, Supplementary Figs S13–S17). Notably, Pt/Al₂O₃ shows the highest activity despite having the second-largest Pt particle size, while Pt/ZrO₂ and Pt/CeO₂ with smaller Pt particles exhibit lower activity. No clear correlation is observed between Pt particle size and catalytic activity, suggesting that the dispersion of exposed active sites is not the sole determining factor in catalytic performance.

Figure 2. 

The morphology characterization of catalysts. (a) Pt/Al₂O₃, (b) Pt/ZrO₂, (c) Pt/CeO₂, (d) Pt/TiO₂, (e) Pt/SiO₂, and (f) Pt particle size distribution.

Multiple characterizations were employed to investigate the physicochemical properties of catalysts, including XRD (Supplementary Fig. S18) and BET (Supplementary Fig. S19, Supplementary Table S2), EPR, XPS, etc. Electron paramagnetic resonance (EPR) spectroscopy was performed to probe unpaired electrons associated with oxygen vacancies. No distinct EPR signal was observed for Pt/Al₂O₃ and Pt/SiO₂, suggesting a negligible concentration of oxygen vacancies (Fig. 3a). In contrast, distinct EPR signals were observed for other Pt/oxide catalysts, indicating the existence of oxygen vacancies. Pt/CeO₂ showed the strongest EPR signal, followed by Pt/ZrO₂, Pt/TiO₂, suggesting a high concentration of oxygen vacancies and active lattice oxygen on Pt/CeO₂, and a similar density of oxygen vacancies on Pt/ZrO₂, Pt/TiO₂.

Figure 3. 

The structural characterization of Pt/oxide catalysts. (a) EPR spectra, (b) O 1s XPS, (c) NH₃-TPD, (d) CO₂-TPD, (e) H₂-TPR, and (f) Pt 4f XPS.

The O 1s XPS spectra reveal differences in oxygen species among the Pt/oxide catalysts (Fig. 3b). The low binding energy peaks at 529.7–530.8 eV are assigned to lattice oxygen, while the component at 530.8 eV is associated with oxygen vacancies, and the peak at 532.1 eV is attributed to hydroxyl groups^[34]. Notably, all catalysts exhibit a lattice oxygen component, while O 1s peaks of Pt/Al₂O₃ and Pt/SiO₂ shift to higher binding energy, suggesting a modification in the local chemical environment of surface oxygen species. Hydroxyls are observed on all catalysts, with the Pt/Al₂O₃ catalyst showing the highest abundance (48.7%), followed by Pt/ZrO₂ (15.6%) (Supplementary Table S3). Oxygen vacancy-related species are detected on Pt/CeO₂, Pt/TiO₂, and Pt/ZrO₂, with abundances decreasing in the order Pt/CeO₂ > Pt/TiO₂ > Pt/ZrO₂, consistent with EPR results. Oxygen vacancies, which can serve as active sites for water dissociation, likely contribute to the improved catalytic activity.

The chemisorption techniques, such as NH₃-TPD, CO₂-TPD, and H₂-TPR experiments, are efficient methods for probing the acid, basic, or reactive properties of catalysts. In the NH₃-TPD profiles (Fig. 3c), obvious desorption peaks appear for Pt/Al₂O₃ and Pt/ZrO₂, while no significant signals are detected for the other catalysts. The peak at 118 °C for Pt/Al₂O₃ corresponds to weak acidic sites, whereas the peak at approximately 255 °C for Pt/ZrO₂ suggests the existence of medium acidic sites^[35]. These acidic sites originate from unsaturated surface Al³⁺ and Zr⁴⁺ species (Lewis acid sites) and surface hydroxyls (Brønsted acid sites). By comparison, the absence of discernible peaks for the other catalysts suggests negligible acidity. Lewis acid sites have been widely associated with enhanced Pt dispersion and redox behavior^[32,36], whereas Brønsted acid sites are known to facilitate the water adsorption in APRM.

The CO₂-TPD profiles indicate that both Pt/Al₂O₃ and Pt/ZrO₂ possess a measurable density of basic sites, while the other catalysts show negligible surface basicity (Fig. 3d). The CO₂ desorption peaks for Pt/Al₂O₃ at 154 and 370 °C are assigned to weak and medium basic sites, respectively^[37]. In comparison, Pt/ZrO₂ exhibits weaker basicity, with desorption peaks at approximately 70 and 225 °C. It is worth noting that basic sites generally serve as active sites for the WGS reaction by stabilizing OH* intermediates derived from H₂O dissociation^[38]. Accordingly, Pt/Al₂O₃ is expected to exhibit a higher WGS capability, which may account for its relatively low CO selectivity.

The redox characteristics of catalysts, which significantly affect the catalytic performance in APRM, were investigated by H₂-TPR and XPS. In the TPR graph, the H₂ consumption peak below 150 °C corresponds to Pt species with weak metal-support interaction. The peak around 200 °C is typically attributed to the reduction of interfacial oxygen species interacting with Pt particles, indicating stronger metal-support interactions^[39]. Pt/SiO₂ and Pt/TiO₂ feature the lowest reduction temperature at 80 and 101 °C, proving the existence of weak metal-support interaction (Fig. 3e). This behavior can be ascribed to the chemically inert nature of SiO₂ and the limited redox participation of TiO₂ under mild reduction conditions, which result in minimal electronic coupling between Pt and the supports. The H₂ consumption peaks of Pt/CeO₂, Pt/ZrO₂, and Pt/Al₂O₃ are higher at 201, 197, and 194 °C, owing to the stronger metal-support interaction originating from unsaturated sites. In the case of CeO₂ support, the strong metal-support interaction arises from the oxygen vacancies, whereas in ZrO₂ and Al₂O₃, it is primarily associated with coordinatively unsaturated cationic sites^[36].

The high-resolution Pt 4f XPS spectrum exhibits two peaks corresponding to the Pt 4f_7/2 and Pt 4f_5/2 orbitals (Fig. 3f). For Pt/Al₂O₃, the most intense Pt 4f photoemission signal cannot be clearly resolved due to overlap with the strong Al 2p peak^[40,41]. Among the other four catalysts, the Pt/SiO₂ shows the lowest binding energy, followed by Pt/TiO₂, indicating a more electron-rich state of Pt species on these supports. The Pt 4f_7/2 spectra of Pt/TiO₂ and Pt/SiO₂ can be deconvoluted into one component at approximately 71.1 and 70.3 eV, which are assigned to metallic Pt⁰ species (Supplementary Table S4). In contrast, the Pt 4f peaks of Pt/ZrO₂ and Pt/CeO₂ shift to higher binding energies, as evidenced by the deconvolution of the Pt 4f_7/2 signal into components attributable to Pt⁰ and Ptᵟ⁺ at approximately 72.8 and 73.6 eV, respectively. This positive binding energy shift suggests electron transfer from Pt nanoparticles to the oxide supports, thereby strengthening the metal-support interaction, consistent with H₂-TPR results. The Pt particles have been regarded as active species for methanol decomposition, especially when strong metal-support interactions are present.

Role of oxygen species on APRM

In-situ DRIFTS was employed to monitor the evolution of reaction intermediates and identify the nature of active sites during APRM. The samples pretreated under H₂ atmosphere were first exposed to a mixture of CH₃OH and H₂O vapor at 250 °C for 30 min. As shown in Fig. 4a, two bands at around 2,959 and 2,841 cm^–1 are detected, which are ascribed to chemisorbed CH₃OH and H₂O at Pt sites, respectively^[42,43]. With continued exposure, the signals of intermediates, including methoxy (CH₃O*) at 1,586, 1,033 cm^–1, CO* at 2,082 cm^–1, and formate (HCOO*) at 1,395 cm^–1 become more intense^[44]. Finally, the peaks corresponding to gaseous final product (CO₂, at 2,362 cm^–1) are observed (Fig. 4a). This phenomenon suggests that APRM on Pt/Al₂O₃ proceeded via a pathway involving methanol decomposition to CH₃O* species (first step), followed by the formation of CO* (second step), and the final conversion of CO* to CO₂ and H₂ through WGS (last step)^[24]. After purging with N₂, the signals corresponding to various intermediates became weaker, indicating their weak adsorption on the catalyst surface. Similar intermediates were detected for the other catalysts, particularly Pt/ZrO₂, suggesting analogous reaction routes (Fig. 4b). For Pt/CeO₂, abundant CH₃O* and HCOO* intermediates can be observed, and these peaks persist even after N₂ purging, indicating strong adsorption of these intermediates on the catalyst surface (Fig. 4c). This strong adsorption behavior may hinder the subsequent conversion of intermediates, thereby limiting the H₂ production. On the surface of Pt/TiO₂ and Pt/SiO₂, only adsorption features corresponding to CH₃O* signals are observed, whereas no CO* and HCOO* species are detected (Fig. 4d, e). This behavior arises from the weak interaction between the Pt sites and the inert oxygen species, which limits the activity of MD, particularly the further decomposition of CH₃O*.

Figure 4. 

The catalytic mechanism of Pt/oxide catalysts. The in-situ DRIFTS of APRM over (a) Pt/Al₂O₃, (b) Pt/ZrO₂, (c) Pt/CeO₂, (d) Pt/TiO₂, and (e) Pt/SiO₂ under CH₃OH/H₂O (1:3) at 230 °C. (f) The normalized integrated peak area of surface HCOO* (1,395 cm^–1). (g) Schematic illustration of the APRM reaction pathways over Pt/oxide catalysts.

We also integrated the HCOO* bands to enable a semi-quantitative comparison of both the formation behavior and the relative abundance of HCOO* species over the different catalysts (Fig. 4f). Notably, the peak areas of Pt/TiO₂ and Pt/SiO₂ are barely detectable, indicating that the HCOO* species cannot be formed on their surfaces. In contrast, the normalized HCOO* areas on Pt/Al₂O₃ and Pt/ZrO₂ gradually increase upon methanol introduction, but decrease during N₂ purging, with a small amount of HCOO* remaining on their surfaces. This suggests that HCOO* species are weakly adsorbed on the catalyst surface and are prone to participate in subsequent WGS reactions^[43]. On Pt/CeO₂, a large amount of HCOO* signal appears and remains even after N₂ purging. This persistence indicates the formation of strongly adsorbed HCOO* species, which likely block active sites and consequently inhibit the catalytic performance of Pt/CeO₂.

We systematically summarized the mechanism of Pt-based catalysts supported on different types of oxides (Fig. 4g). On weakly reducible and inert oxide supports (TiO₂ and SiO₂), APRM is largely restricted by CO* generation steps (first step), owing to the restriction of MD by weak metal-support interaction induced by inert oxygen species. Over strongly reducible oxides with active oxygen species (e.g., CeO₂), the strongly adsorbed HCOO* intermediates hinder the H₂ production step in WGS (second step), thereby limiting the overall reforming activity. On amphoteric oxides with hydroxyls like Al₂O₃ and ZrO₂, methanol reforming proceeds readily via the MD and WGS pathway (third step), resulting in a relatively high H₂ production rate.

We also carried out the in-situ DRIFTS by introducing CH₃OH and H₂O in sequence to probe the role of the catalysts in promoting methanol and water dissociation. On Pt/SiO₂ and Pt/TiO₂, only the limited MD reaction can happen, due to the obvious CH₃O* intermediates at 1,586, 1,033 cm^–1 and nearly undetectable CO* (2,082 cm^–1) in the DRIFTS graph (Fig. 5a, Supplementary Fig. S20). After introducing H₂O, the CH₃O* species are converted into final products (CO₂ and H₂) via WGS^[30]. We proposed that poor catalytic performance originated from the weak metal-support interaction between Pt and inert oxygen species on inert and weakly reducible oxide-based catalysts. Although they may possess water activation ability, their rate-limiting step is MD to H₂ and CO* (Fig. 5b).

Figure 5. 

Effects of surface oxygen species on APRM over Pt/oxide catalysts. (a) In-situ DRIFTS over Pt/SiO₂ catalyst upon the separate introduction of methanol and water. (b) Schematic illustration of the role of inert oxygen species on Pt/SiO₂ in methoxy formation. (c) In-situ DRIFTS over the Pt/Al₂O₃ catalyst upon separate introduction of methanol and water. (d) Schematic illustration of the role of hydroxyls on Pt/Al₂O₃ in formate formation. (e) In-situ DRIFTS over Pt/CeO₂ catalyst upon separate introduction of methanol and water. (f) Schematic illustration of the role of reactive lattice oxygen on Pt/CeO₂ in formate formation.

On Pt/Al₂O₃ and Pt/ZrO₂, the decomposition of CH₃OH over Pt/Al₂O₃ occurred, producing abundant intermediates including CH₃O* and CO* (Fig. 5c, Supplementary Fig. S21). It is worth noting that the HCOO* species (requiring an additional oxygen source) formed on Pt/Al₂O₃ via the consumption of surface hydroxyls, as evidenced by the decrease of the surface hydroxyls at ~3,700 cm^–1[38,45]. After introducing H₂O, the intensity of OH groups improved, and the previously formed CO* species rapidly disappeared, which can be attributed to the rapid H₂O activation and the occurrence of WGS. Therefore, we proposed that both CH₃OH and H₂O activation proceed rapidly on Pt/Al₂O₃. MD occurred efficiently on Pt sites, while surface OH acted as OH* intermediates to facilitate the conversion of CO*, thereby accelerating the H₂ production rate (Fig. 5d). Meanwhile, these surface OH can be continuously replenished through water dissociation, sustaining the reaction cycle. We also carried out the in-situ DRIFTS over Pt/nanorod Al₂O₃ and Pt/nanosheet Al₂O₃ to explore the applicability. Pt/Al₂O₃ with various morphologies proceeds via a hydroxyl-mediated route, as proved by the decrease of the ~3,700 cm^–1 band assigned to hydroxyls (Supplementary Figs S22 and S23). These results indicate that morphology is not the dominant factor and that surface oxygen species play a more critical role.

As for Pt/CeO₂, the decomposition of CH₃OH proceeded effectively, as evidenced by the appearance of CH₃O* and CO* (Fig. 5e). Subsequently, HCOO* species, which typically require an additional oxygen source, were formed without noticeable changes in the intermediate signals. Compared with Pt/Al₂O₃, HCOO* intermediates were not formed via consumption of surface hydroxyls, but are instead associated with reactive lattice oxygen. We propose that HCOO* formation on Pt/CeO₂ proceeds via a redox pathway involving the consumption of lattice oxygen^[46,47]. The introduction of H₂O led to the consumption of CO* intermediates and the formation of gaseous CO₂, indicating Pt/CeO₂ possessed a strong capability for WGS. However, CH₃O* and HCOO* intermediates still can be observed on Pt/CeO₂. This indicates that these species are strongly adsorbed on the catalyst surface, hindering the production of H₂. Therefore, we propose that Pt particle sites serve as active sites for MD, while the active lattice oxygen species provide OH* for CO* conversion (Fig. 5f). However, the adsorption strength of the key intermediate of WGS (HCOO*) is so strong that it limits the H₂ generation. This phenomenon is also observed on other CeO₂-based catalysts like Pt/nanorod CeO₂, where strongly adsorbed HCOO* intermediates form via consumption of lattice oxygen (Supplementary Fig. S24). It can be evidenced by the high intensity signal of HCOO* at 1,395 cm^–1 and the negligible change in the 3,700 cm^–1 band.

In this study, we reveal that oxygen species, including hydroxyls, active lattice oxygen, and inert oxygen species, govern the catalytic performance of APRM by modulating methanol activation and the evolution pathways of HCOO*. Amphoteric oxides with hydroxyls outperform other catalysts owing to the oxygen supply effects of exposed surface hydroxyls. In contrast, inert and overly reactive oxygen species lead to limited activity due to weak metal-support interactions or over-stabilization of HCOO* intermediates. These findings establish surface oxygen species as a key descriptor for APRM catalytic performance and provide mechanistic guidance for the rational design of oxide-supported catalysts. The key novelty of this work is the mechanistic classification and specific roles of oxygen species in APRM, rather than the identification of an isolated pathway for a single catalyst.