Reaction-induced dynamic evolution of PtIn/SiO₂ catalyst for propane dehydrogenation

Abstract

Understanding the dynamic evolution of heterogeneous catalysts is crucial yet challenging for elucidating the structure-performance relationships and enabling rational catalyst design. Herein, we reveal that PtIn alloy clusters gradually evolve into Pt₃In intermetallic in response to propylene, the product of propane dehydrogenation (PDH) reaction. Specifically, a PtIn_1.0/SiO₂ catalyst has been fabricated, comprising sub-nanometric PtIn alloy clusters covered by an In⁰ overlayer, with In³⁺ species locating at the metal-support interface. During the PDH reaction propylene induces the evaporation of the In⁰ overlayer, thereby exposing Pt sites. After an induction period, the evolved Pt₃In intermetallic (average size ~1.3 nm) exhibits a C₃H₆ productivity of 145 mol g_Pt⁻¹ h⁻¹. The alloyed In⁰ species effectively dilute Pt-Pt ensembles, enhancing propylene selectivity, while the interfacial In³⁺ species inhibit aggregation of Pt₃In intermetallic, ensuring excellent catalytic stability. These findings underscore the critical role of product molecules in shaping active site evolution at the atomic scale.

Understanding the dynamic evolution of heterogeneous catalysts is crucial yet challenging. This work uncovers the transformation of amorphous PtIn clusters into ultrafine Pt₃In intermetallics, driven by propylene during propane dehydrogenation, highlighting the pivotal role of product-induced changes in revealing the true active sites.

Introduction

The critical role of heterogeneous catalysts in facilitating chemical production and energy conversion processes cannot be overstated^1–3. These catalysts, typically composed of active metals dispersed on support, are subject to a continuous process of dynamic evolution during operation^4,5. Affected by the reaction environment, including temperature and atmosphere, the structural and chemical properties of catalysts are strongly altered⁶. Elevated temperatures promote the atomic rearrangement of active metals and supports, leading to phenomena such as sintering, redispersion, and facet restructuring^7–10. In addition, the coordination of reactant or product molecules on metal surfaces significantly alters the surface energy, thus changing their structure and chemical properties^11–14. For alloy catalysts, the structural diversity stemming from various distributions and proportions of components results in a richer dynamic evolution. Numerous gas molecules, including CO, NO, and O₂, have been demonstrated to induce the element segregation and even composition changes of alloy^11,13,15. This dynamic evolution of the catalyst results in dramatic changes in catalytic performance, which is an overlooked factor in studying the structure-performance relationship¹⁶. Therefore, an unambiguous understanding of the catalyst dynamics is crucial for the identification of the real active sites and the rational design of high-performance heterogeneous catalysts.

On-purpose propane dehydrogenation (PDH) technologies (ΔH_298K = 124.3 kJ mol⁻¹) have drawn intensive attention in recent years, driven by the ever-increasing market demands for propylene and the revolution of shale gas feedstocks¹⁷. The research emphasis of high-performance PDH catalysts focuses on the fabrication of highly-dispersed Pt-M alloy, where M represents promoters, including In^18–20, Sn^21–23, Cu^24,25, Ga²⁶, Zn^27–29, and so on^30–34. The addition of catalytic-inert elements is reported to dilute Pt-Pt ensembles, thus improving propylene selectivity³⁵. Meanwhile, richer structural evolution also appeared due to the introduction of promoters. In the PDH reaction, high temperature and reductive atmosphere, especially H₂, are widely recognized as key drivers of structural evolution in Pt-M alloy catalysts³⁶. However, to the best of our knowledge, the dynamic evolution of Pt-M alloys induced by propylene, the target product of PDH reaction, has not been reported. Revealing the interaction between propylene and metal sites is expected to provide a deeper understanding of this reaction and guide the rational design of PDH catalysts.

Herein, we report the dynamic evolution of the sub-nanometric PtIn clusters into ultrafine Pt₃In intermetallics, driven by the propylene molecules. Upon H₂ reduction, PtIn alloy clusters covered by In⁰ overlayer were formed, while dominant In³⁺ species were still presented on the PtIn-SiO₂ interface and surface of the support. During the PDH reaction, propylene was found to induce the evaporation of the In⁰ overlayer, markedly enhancing catalytic activity. Moreover, the remained In⁰ species effectively diluted Pt sites to improve propylene selectivity, while In³⁺ species at the Pt₃In-SiO₂ interface inhibited the aggregation of Pt₃In intermetallic clusters. As a result, the evolved Pt₃In intermetallics (average size ~1.3 nm) exhibited a C₃H₆ selectivity of 97% and a productivity of 145 mol g_Pt⁻¹ h⁻¹, which was an order of magnitude higher than previously reported PtIn-based catalysts. These results highlight the pivotal role of product molecules in governing structure-performance relationships and identifying the true active sites.

Results

Preparation and characterization of PtIn catalysts

The catalyst was fabricated using a stepwise strong electrostatic adsorption (SEA) method. First, SiO₂ was immersed in an alkaline solution to impart a negative surface charge since the pH was higher than its point of zero charge (PZC = 2.8)³⁷. To avoid the formation of In(OH)₃ precipitation in the next step, excess OH⁻ ions were removed by washing SiO₂ with deionized water. The treated SiO₂ was then re-dispersed in water, and In³⁺ cations were adsorbed on the negatively-charged SiO₂ surface. The nominal loading of In was 1.0 wt% relative to SiO₂, and the resulting powder was dried and calcined at 300 °C for 1 h, named In_1.0/SiO₂. Next, Pt(NH₃)₄²⁺ cations were adsorbed onto In_1.0/SiO₂ by controlling the pH to 9 − 10, and the resulting powder was reduced at 600 °C with H₂ for 1 h, named PtIn_1.0/SiO₂. Compared with the impregnation method, such a SEA technique ensures a uniform distribution of Pt and In on SiO₂ with low metal loading (< 1 wt%)³⁸. The actual Pt and In loadings were determined to be 0.12 wt% and 0.90 wt% by inductively coupled plasma-mass spectroscopy (ICP-MS), respectively. A series of PtIn_x/SiO₂ catalysts with fixed Pt loading but varying In loadings was further synthesized to elucidate the effect of In loading on catalytic performance. For comparison, Pt/SiO₂ was synthesized by directly adsorbing Pt(NH₃)₄²⁺ cations onto SiO₂. In addition, SiO₂-supported Pt₃In intermetallic nanoparticles (Pt₃In NPs/SiO₂) were synthesized following the procedure in literature³⁹. The nominal Pt-to-In mass ratio, along with values obtain from ICP-MS and X-ray photoelectron spectroscopy (XPS) are presented in Supplementary Table 1.

The quasi in situ X-ray absorption near-edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) spectra were measured to determine the electronic structures of Pt/SiO₂ and PtIn_1.0/SiO₂. As shown in Fig. 1a, the absorption energies of both Pt/SiO₂ and PtIn_1.0/SiO₂ were close to Pt foil, indicating a metallic state. Notably, compared with Pt/SiO₂, PtIn_1.0/SiO₂ exhibited an obvious decrease in white line intensity but an increase in the absorption energy. According to previous studies, the Pt-L_III edge represents the dipole transitions from the 2p_3/2 orbital to the unoccupied 5 d_5/2 orbital. Thus, the variations in white line intensity reflect changes in the unoccupied 5 d_5/2 orbital, such as those induced by alloy or oxidation⁴⁰. The formation of PtIn alloy is reported to result in the charge transfer from In 5p to Pt 5d orbital, thereby reducing the proportion of unoccupied Pt 5d orbital³⁹. Thus, the decreased white line intensity of the Pt-L_III edge reflects the electron redispersion between Pt and In caused by the formation of PtIn alloy. Meanwhile, the higher absorption energy of PtIn_1.0/SiO₂ was also caused by the accumulation of electrons on Pt⁴¹.

Fig. 1. Electronic and coordination structure of PtIn alloy.

a, b XANES (a) and EXAFS (b) spectra at the Pt L_III-edge of Pt/SiO₂, PtIn_1.0/SiO₂, Pt foil, and PtO₂. c, d The wavelet transformed EXAFS of Pt/SiO₂ (c) and PtIn_1.0/SiO₂ (d). e CO DRIFTS results of Pt/SiO₂, PtIn_1.0/SiO₂, and PtIn_3.0/SiO₂. a.u., arbitrary units. f Changes of Pt 4 f and In 3 d intensities in XPS spectra during the sputtering cycles.

The coordination structure of PtIn_1.0/SiO₂ was investigated by EXAFS spectra (Fig. 1b–d and Supplementary Table 2). The Pt/SiO₂ sample showed a Pt−Pt coordination number (CN) of 5.9 at a bonding distance of 2.79 Å and a low Pt−O CN (0.5) at a bonding distance of 2.02 Å. For PtIn_1.0/SiO₂, a Pt−Pt CN of 2.2 and a Pt−In CN of 2.6 were fitted, suggesting the formation of PtIn alloy⁴². Compared to that of Pt/SiO₂, the bonding distance of Pt−Pt and Pt−In was reduced to 2.55 Å and 2.56 Å, respectively. It has been reported that transitioning from bulk to small metal nanoparticles (NPs) leads to the de-hybridization of metal orbitals and a contraction in bonding distances⁴¹. Thus, the reduced bonding distance implied the small size of the PtIn alloy. In addition, XAFS measurements at the In K-edge were also measured to determine the electronic environment of In species (Supplementary Fig. 1 and Table 2). Compared to In_1.0/SiO₂, PtIn_1.0/SiO₂ exhibited a decreased absorption energy and white line intensity, indicating a lower oxidation state. EXAFS fitting results showed a CN of 5.9 for In−O in In_1.0/SiO₂, highly close to a saturated In−O CN (6.0) in bulk In₂O₃. The high CN of In−O in In_1.0/SiO₂ suggested that In³⁺ species were dominant in the reduced samples. In contrast, PtIn_1.0/SiO₂ showed a lower In−O CN of 4.4, indicating that the In³⁺ species were partially reduced upon the addition of Pt.

To further investigate the reduction degree of Pt and In, H₂ temperature-programmed reduction (TPR) experiments were performed. As illustrated in Supplementary Fig. 2, the Pt/SiO₂ sample exhibited a distinct reduction peak at 303 °C, corresponding to the reduction of Pt²⁺ species. In³⁺ species in In_1.0/SiO₂ were reduced at 550 °C. For PtIn_1.0/SiO₂, a more pronounced reduction peak of In³⁺ species was observed, along with an additional peak at 505 °C. This peak was attributed to the reduction of In³⁺ species facilitated by Pt⁰. By assuming the complete reduction of Pt²⁺, we calculated the percentage of In⁰ species based on H₂ consumption. For In_1.0/SiO₂, only 0.8% of In³⁺ was reduced to a metallic state. Upon Pt loading, the In⁰ content increased to 2.8%, indicating that Pt⁰ promoted the reduction of In³⁺ species. Nevertheless, the majority of In remained in the oxidized state as In³⁺ (97.2%). For PtIn_1.0/SiO₂, the atomic ratio of Pt⁰ to In⁰ was calculated as 1 / 0.4. EXAFS fitting results further exhibited comparable CNs for Pt-Pt and Pt-In, suggesting that Pt⁰ species were coordinated with both alloyed In⁰ species and oxidized In³⁺ species. In other words, in PtIn clusters, In³⁺ species were present at the interface between the PtIn alloy and SiO₂ support.

CO probe diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) was employed to explore the surface geometry configuration of Pt. As depicted in Fig. 1e and Supplementary Fig. 3, Pt/SiO₂ exhibited a peak at 2,066 cm⁻¹, corresponding to the stretching vibration of linearly adsorbed CO on Pt sites. For PtIn_1.0/SiO₂, the CO adsorption peak exhibited a red shift to 2042 cm⁻¹ owing to the electron accumulation on metallic Pt sites. Consistent with the XANES results, CO DRIFTS measurements indicate the electron transfer from In to Pt. Moreover, the integral intensity of the CO adsorption peaks markedly decreased for PtIn_1.0/SiO₂, and completely vanished for PtIn_3.0/SiO₂ with a higher In loading. Thus, we deduce that an In⁰ overlayer was present on the surface of the PtIn alloy, likely driven by the difference in surface energy between Pt and In. In general, elements with lower surface energy tend to segregate to the surface of an alloy to minimize the overall system energy^43,44.

The valence states of Pt and In were investigated using quasi in situ XPS. The binding energies of Pt 4 f in Pt/SiO₂ and PtIn_1.0/SiO₂ were centered at 71.6 and 71.4 eV, respectively, suggesting a metallic state of Pt¹⁹. The 0.2 eV downshift in the binding energy of PtIn_1.0/SiO₂ indicated the charge transfer from In to Pt (Supplementary Fig. 4a). The XPS results for In 3 d are shown in Supplementary Fig. 4b. Both In_1.0/SiO₂ and PtIn_1.0/SiO₂ exhibited a peak centered at 445.2 eV, attributed to dominant In³⁺ species. The minor In⁰ species were challenging to detect by XPS. To further investigate the distribution of Pt and In atoms in PtIn alloy, XPS analysis with Ar⁺ sputtering was performed. In this case, Pt₂In₁₀/SiO₂ with a Pt loading of 1.3 wt% and a In loading of 10.8 wt% was used to improve the signal-to-noise ratio of XPS spectra (Fig. 1f, Supplementary Table 1 and Fig. 5). With increasing sputtering cycles, the In signal intensity gradually decreased while the Pt signal intensity increased. Thus, we conclude that an In⁰ overlayer exists on the surface of the PtIn alloy.

The structural characteristics of Pt and In after H₂ reduction at 600 °C were investigated using electron microscopy. As shown in Supplementary Fig. 6, high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images revealed that Pt NPs in Pt/SiO₂ were randomly distributed with an average size of 2.5 nm. The presence of particles up to 6 nm indicated that Pt NPs on SiO₂ underwent sintering during H₂ reduction. In comparison, the PtIn alloy in PtIn_1.0/SiO₂ was uniformly dispersed on SiO₂ with an average size of 0.9 nm, showcasing the enhanced sintering resistance of PtIn_1.0/SiO₂ (Fig. 2a). EDX elemental mapping images indicated that the signal of Pt and In was homogeneously distributed across the surface of SiO₂, also suggesting the small and well-dispersed nature of PtIn alloy (Fig. 2b). It has been reported that these unreduced metal species at the interface between NPs and support, called oxide nano-islands, play a key role in frustrating the aggregation of metal NPs⁴⁵. On one hand, compared to In⁰ species, these unreduced In³⁺ species bonded with SiO₂ more tightly through Si-O-In bonding³⁹. On the other hand, In³⁺ species exhibited better stabilization effect on Pt⁰ species than pure SiO₂ does¹⁸. Thus, these In³⁺ species at the interface between PtIn clusters and SiO₂ support effectively improve the sintering resistance of PtIn clusters.

Fig. 2. Microscopic structure characterizations.

a HAADF-STEM image and corresponding size distribution of PtIn_1.0/SiO₂. b HAADF-STEM image and EDX elemental mapping images of PtIn_1.0/SiO₂. c, d 3D intensity distribution map and HAADF-STEM image of an individual cluster. e, f Simulated HAADF-STEM image and proposed model of the above cluster.

The atomically resolved HAADF-STEM characterizations were performed to reveal the configuration and composition of the PtIn alloy. As shown in Supplementary Fig. 7, Pt/SiO₂ exhibited well-crystallized Pt NPs with distinct lattice fringes. In contrast, PtIn_1.0/SiO₂ exhibited substantial single atoms and clusters, which were identified as atomically dispersed In³⁺ species and PtIn clusters, respectively (Supplementary Fig. 8). The structure of individual PtIn clusters was further analyzed based on the brightness of these spots. The 3D intensity distribution map derived from HAADF-STEM image defined the brightness of atomically dispersed In³⁺ species and PtIn clusters (Fig. 2c, d, and Supplementary Fig. 9). In PtIn clusters, most of the spots were identified as In atoms, while these spots with greater brightness were assumed to be a combination of Pt and In atoms. To validate this structural interpretation, we constructed an atomic-scale model of SiO₂-supported PtIn cluster and performed HAADF-STEM image simulations using QSTEM software (Fig. 2e). Consequently, the 3D intensity distribution map derived from the simulated HAADF-STEM image was in well accordance with the experimental results (Fig. 2c and Supplementary Fig. 10). Thus, the structure of the individual PtIn cluster was elucidated at an atomic scale, in which Pt atoms were covered by In atoms (Fig. 2f). This atomic-scale structural elucidation provides critical insights into the surface composition and coordination environment of the bimetallic system.

Catalytic performance towards PDH reaction

From an economic perspective in industrial production, enhancing the conversion efficiency of reactants per unit mass of catalyst is significantly advantageous. This can be strategically achieved by elevating the concentration of the reactants involved. Whereas, in PDH reaction, excessively high partial pressure of propane results in the coke formation, thus placing high demands on the stability of catalysts⁴⁶. Generally, high propane (C₃H₈) and propylene (C₃H₆) partial pressures favor coke formation, while high hydrogen (H₂) partial pressure is reported to mitigate it⁴⁷. We first evaluated the PDH catalytic performance under a relatively mild condition (550 °C, 10% C₃H₈ diluted in Ar). As shown in Supplementary Fig. 11a, the PtIn_1.0/SiO₂ catalyst exhibited poor catalytic performance. During the 30 h time on stream (TOS), C₃H₈ conversion dropped rapidly from 16.8% to 11.3%. The initial C₃H₆ productivity was calculated as 18 mol g_Pt⁻¹ h⁻¹ (Fig. 3a). To quantify deactivation of the catalyst, we calculated the mean lifetime based on an overall first-order deactivation mechanism^31,48:

$$\tau =t/\left\{\left[\ln \left[\left(1-X_{\mathrm{C}3\mathrm{H}8,\mathrm{final}}\right)/X_{\mathrm{C}3\mathrm{H}8,\mathrm{final}}\right]\right)-\mathrm{In}\left[\left(1-X_{\mathrm{C}3\mathrm{H}8,\mathrm{initial}}\right)/X_{\mathrm{C}3\mathrm{H}8,\mathrm{initial}}\right]\right\}$$ 1

where X_{C3H8, final}, X_{C3H8, initial}, and t represent the final C₃H₈ conversion, initial C₃H₈ conversion, and reaction time, respectively. PtIn_1.0/SiO₂ gave a mean lifetime of 68 h. When the C₃H₈ partial pressure was over 40%, an unexpected induction period was observed. The C₃H₈ conversion continuously increased to reach a maximum and then slightly dropped as the PDH reaction proceeded (Supplementary Fig. 11b−d). Besides, PtIn_1.0/SiO₂ exhibited higher mean lifetime values when the partial pressure of C₃H₈ exceeded 40% (Supplementary Fig. 12). Notably, the C₃H₆ productivity reached as high as 197 mol g_Pt⁻¹ h⁻¹ under pure C₃H₈, which was 11 times higher than that under 10% C₃H₈ (Fig. 3a). In short, high C₃H₈ partial pressures are beneficial for achieving high PDH catalytic activity and stability of PtIn_1.0/SiO₂ catalyst.

Fig. 3. PDH performance.

a PDH performance of PtIn_1.0/SiO₂ under different propane partial pressures at 550 °C. b PDH performance of PtIn_1.0/SiO₂ as a function of nominal Pt-to-In mass ratios. c PDH performance of PtIn_1.0/SiO₂ at different temperatures. d PDH stability test of PtIn_1.0/SiO₂, Pt/SiO₂ and Pt₃In NPs/SiO₂. Reaction condition: atmospheric pressure, WHSV = 39 h⁻¹, 550 °C, C₃H₈/H₂ = 2/1. The inset shows the bottom of the quartz tube before and after the PDH reaction. e C₃H₆ productivity versus C₃H₆ selectivity for the Pt-based catalysts described in this work and literature. The catalysts included here are only the best-performing ones from the articles considered.

To further unveil the relationship between the induction period and the In loading, a series of PtIn_x/SiO₂ catalysts with different Pt-to-In mass ratios was prepared and evaluated (Fig. 3b and Supplementary Fig. 13). We co-fed H₂ with C₃H₈ to lower the coverage of deeply dehydrogenated coke precursors, thus mitigating coke formation. The following PDH tests were operated at 550 °C with the weight hourly space velocity (WHSV) of 39 h⁻¹ (C₃H₈/H₂ = 2/1). For PtIn_0.2/SiO₂ and PtIn_0.4/SiO₂, the induction period was absent. C₃H₈ conversion and C₃H₆ selectivity decreased constantly during the 20-h TOS. The mean lifetime of PtIn_0.2/SiO₂ and PtIn_0.4/SiO₂ was calculated to be 19 h and 48 h, respectively. When In loading was further increased to 0.6 wt%, C₃H₈ conversion increased to a maximum and then gradually dropped over a 20-h TOS. PtIn_0.6/SiO₂ and PtIn_0.8/SiO₂ exhibited induction periods of 10.7 h and 15.0 h, respectively, indicating that higher In loading led to a longer induction period. Noteworthily, the lifetimes of PtIn_0.6/SiO₂ and PtIn_0.8/SiO₂ were significantly increased to 61 h and 726 h, respectively, demonstrating that increased In loading enhanced catalyst stability. For PtIn_1.0/SiO₂, C₃H₈ conversion increased constantly during 20-h TOS. Whereas further increasing the In loading to 1.5 wt% or 3.0 wt% resulted in lower overall catalytic activity, likely due to the active Pt sites being covered by excessive In⁰. At 20 h, PtIn_1.0/SiO₂ exhibited the highest C₃H₆ productivity of 121 mol g_Pt⁻¹ h⁻¹ among all PtIn_x/SiO₂ samples. Thus, to optimize C₃H₆ productivity, the PtIn_1.0/SiO₂ sample was chosen for further evaluation.

The maximum activity of PtIn_1.0/SiO₂ was investigated by increasing the WHSV after the induction period. As a result, the C₃H₆ productivity of PtIn_1.0/SiO₂ reached a peak of 317.2 mol g_Pt⁻¹ h⁻¹ under the WHSV of 118 h⁻¹ (Supplementary Fig. 14). The stability of the optimal PtIn_1.0/SiO₂ catalyst was evaluated under different temperatures after the catalytic activity reached its peak. It is worth noting that evaluating catalyst stability at equilibrium conversion should be avoided since seemingly high apparent stability can be obtained by overloading the catalyst. As shown in Fig. 3c, under the WHSV of 39 h⁻¹, the C₃H₈ conversion increased with temperature and remained steady at each stage. The C₃H₈ conversions of PtIn_1.0/SiO₂ at each stage were lower than the corresponding equilibrium conversions. After the heating and cooling processes, the C₃H₈ conversion at the final stage was completely restored to 7.9%, matching the initial stage. Therefore, the stable structure of PtIn_1.0/SiO₂ demonstrated excellent temperature tolerance, ranging from 500 to 550 °C.

To further evaluate the catalytic stability of PtIn_1.0/SiO₂, we performed the long-term PDH test. As shown in Fig. 3d, the C₃H₈ conversion of Pt/SiO₂ dropped gradually from 14.4% to 7.6% during an 18-h TOS. Concurrently, the C₃H₆ selectivity surged from 66.7% to 93.2% during the initial stage (TOS < 1 h) and remained unchanged as the PDH reaction progressed. The mean lifetime of Pt/SiO₂ was calculated as 29 h. In contrast, PtIn_1.0/SiO₂ initially exhibited a C₃H₈ conversion of 3.2%, which significantly increased to a maximum (20.1%) after 30-h TOS, corresponding to the C₃H₆ productivity of 145 mol g_Pt⁻¹ h⁻¹. As the PDH test progressed, the C₃H₈ conversion of PtIn_1.0/SiO₂ decreased slightly to 18.8%, giving a mean lifetime of 1395 h. Meanwhile, the C₃H₆ selectivity remained as high as 97% during the 150-h test.

The coke formation on Pt/SiO₂ and PtIn_1.0/SiO₂ was assessed using thermogravimetry (TG) analysis and Raman spectroscopy. As shown in Supplementary Fig. 15, the coke formation rate on Pt/SiO₂ was calculated as 1.3 mg_coke g_cat⁻¹ h⁻¹ during the 18-h PDH test. In contrast, the coke formation rate on PtIn_1.0/SiO₂ during 150-h PDH test was 0.12 mg_cokeg_cat⁻¹h⁻¹, 10.8 times lower than that of Pt/SiO₂. Furthermore, the graphitization degree of the formed coke was evaluated by Raman spectroscopy (Supplementary Fig. 16). After 18-h TOS, the I_D/I_G value for PtIn_1.0/SiO₂ was 1.88, higher than that of Pt/SiO₂ (1.78), indicating a higher degree of disorder in the formed coke. This result aligns with the TG analysis, where coke on PtIn_1.0/SiO₂ was oxidized at lower temperatures (473 °C) compared to Pt/SiO₂ (526 °C). Therefore, In⁰ species effectively improve the C₃H₆ selectivity and retard the coke formation.

According to previous studies, Pt₃In intermetallic was predicted to exhibit maximal C₃H₆ productivity among Pt-based alloy catalysts⁴⁷. We prepared SiO₂-supported Pt₃In NPs (Pt₃In NPs/SiO₂) using the colloid method to evaluate their catalytic performance³⁹. XRD and XPS characterizations demonstrated the successful preparation of Pt₃In intermetallic (Supplementary Figs. 17, 18). As displayed in Fig. 3d, Pt₃In NPs/SiO₂ catalyst exhibited an initial C₃H₆ selectivity of 95%, much higher than that of Pt/SiO₂, demonstrating the effective dilution of Pt sites by In atoms. Besides, the initial C₃H₈ conversion and C₃H₆ productivity of Pt₃In NPs/SiO₂ were as high as 18.9% and 122.3 mol g_Pt⁻¹ h⁻¹, suggesting the superior catalytic activity of Pt₃In NPs. Nevertheless, during 18-h TOS, both C₃H₈ conversion and C₃H₆ selectivity decreased constantly, resulting in a low lifetime of 42 h. We also tested the activity of In_1.0/SiO₂, which was almost inert toward PDH reaction (Supplementary Fig. 19). Thus, the high C₃H₆ productivity and C₃H₆ selectivity demonstrate the outstanding catalytic performance of PtIn_1.0/SiO₂ catalyst, which is superior to most of the state-of-the-art Pt-based catalysts (Fig. 3e and Supplementary Table S3).

Catalyst dynamics during PDH reaction

Despite numerous studies identifying distinct PtIn alloy phases as potential active sites, most PtIn-based catalysts exhibited notably inferior catalytic activity, as illustrated in Fig. 3e. The most effective PtIn alloy phase for the PDH reaction is still under debate, primarily due to an insufficient understanding of the dynamic evolution of PtIn-based catalysts during the PDH reaction. Two critical knowledge gaps persist: first, the existence of In⁰ overlayer in the reduced sample remains contentious, with no established methodology for its selective removal. Second, the characterization of PtIn-based catalysts after PDH reaction remains inadequately explored. While previous reports have noted similar induction periods in catalytic behavior^39,49, atomic-level insights into the dynamic evolution of active sites are conspicuously absent. Thus, investigation of the dynamic evolution of PtIn alloy is of crucial importance for the identification of real active sites, thus obtaining a high-performance catalyst.

The compositional change of PtIn_1.0/SiO₂ during PDH reaction was investigated using a series of spectroscopic characterizations. After a 150-h PDH test, a noticeable yellow substance was observed at the bottom of the quartz tube (insets of Fig. 3d). The Pt-to-In mass ratio of the above substance was determined as 1 / 5400, given by ICP-MS measurements. We further detected the Pt and In loading of spent catalysts. After the stability test and ICP-MS measurements twice, the mass ratio of Pt-to-In was 0.52 and 0.43, averaging 0.47, suggesting that a stable catalyst composition was achieved after the induction period. Due to its low melting point and heat of vaporization, metallic In⁰ easily melts and evaporates at high temperatures. In comparison, the evaporation of In³⁺ species is difficult due to the high melting point of In₂O₃ (~ 2000 °C)⁵⁰. Thus, given that only 2.8% of In⁰ species were present in the reduced samples, we deduced that In³⁺ species on the SiO₂ support were continually reduced and evaporated during the PDH reaction, resulting in a substantial loss of In species (Supplementary Fig. 2). As previously discussed, the PtIn alloy was covered by an In⁰ overlayer (Fig. 1f). Hence, we infer that the evaporation of In⁰ overlayer leads to the induction period of PtIn_1.0/SiO₂ catalyst.

In situ CO DRIFTS measurements were conducted to verify the above hypothesis. As shown in Supplementary Fig. 20, after catalyzing PDH reaction in the infrared cell at 550 °C for different durations, Pt/SiO₂ exhibited an unchanged adsorption peak at 2,066 cm⁻¹, the same as the reduced sample (Supplementary Fig. 3). However, the integral intensity of adsorption peaks decreased constantly, suggesting the continuous sintering of Pt sites. Figure 4a and Supplementary Fig. 21 present the CO DRIFTS results of PtIn_1.0/SiO₂. With the proceeding of the PDH reaction, the adsorption peak gradually red-shifted from 2064 to 2057 cm⁻¹. The increased electron density in the Pt 5d orbital also facilitated the desorption of C₃H₆ and thus higher C₃H₆ selectivity³⁵. Meanwhile, the increase in the integral intensity of corresponding peaks suggests that the surface Pt sites were gradually exposed, attributing to the evaporation of the In⁰ overlayer.

Fig. 4. Unveiling the origin of high PDH performance.

a In situ CO DRIFTS results of PtIn_1.0/SiO₂ after PDH reaction for different times. b Wavelet transformed EXAFS spectra of PtIn_1.0/SiO₂ after PDH reaction at the Pt L_III-edge. c HAADF-STEM image of PtIn_1.0/SiO₂ after 150-h PDH test. The scale bar of the inset was 2 nm. d Atomically-resolved HAADF-STEM images of individual PtIn NPs. e Corresponding simulated HAADF-STEM images and atomic models of Pt₃In NPs along [011] zone axis. f The extracted line profiles along directions in (d). g The dynamic evolution of PtIn_1.0/SiO₂ catalyst during the induction period of PDH reaction. The red, primrose, and yellow balls stand for Pt⁰, In³⁺, and In⁰ species, respectively.

To further investigate the characteristics of In⁰ evaporation, we studied the effect of gas components. In_1.0/SiO₂ was treated under 10% H₂/Ar, 10% C₃H₆/Ar, and 10% C₃H₈/Ar at 550 °C for 10 h in the fixed-bed reactor, respectively. As a result, under 10% C₃H₆/Ar, the amount of evaporated In⁰ species was 10 times higher than that under 10% H₂/Ar or 10% C₃H₈/Ar (Supplementary Fig. 22). The above results indicate that a C₃H₆ atmosphere is capable of inducing the evaporation of In⁰ species. Noteworthily, a high partial pressure of C₃H₆ is indispensable, since no induction period was observed for PtIn_1.0/SiO₂ during the PDH reaction under 10% C₃H₈/Ar, in which the partial pressure of C₃H₆ was calculated as only 1.5% (Supplementary Fig. 11). Thus, C₃H₆ gas with high partial pressure played a decisive role in the evaporation of In⁰ species. To further investigate the factors affecting the time required for the In⁰ evaporation, we systematically decreased the catalyst mass. PtIn_1.0/SiO₂ samples ranging from 120 to 30 mg were used, and the related WHSVs were 13 to 52 h⁻¹. A clear trend emerged that lower catalyst mass corresponded to a shorter induction period (Supplementary Figs. 23, 24). Combined with the data in Supplementary Fig. 13, these results suggest that the total amount of In⁰ present in the catalyst bed dictates the duration of the induction period.

In situ synchrotron-based vacuum ultraviolet photoionization mass spectrometry (SVUV-PIMS) was further conducted to study the evaporation of In⁰ species on SiO₂ support. A photon energy of 9.5 eV was chosen to ionize In, while the C₃H₆ molecule was not able to be ionized. When In_1.0/SiO₂ was exposed to 10% H₂/Ar atmosphere, characteristic signals of In were absent in the temperatures range of 550–800 °C, indicating that In species evaporated from In_1.0/SiO₂ was ignorable (Supplementary Fig. 25a). In comparison, under 10% C₃H₆/Ar, characteristic signals of In isotopes represented by mass/charge = 114.9 and 112.9 were observed when exposing In_1.0/SiO₂ at approximately 650 °C (Supplementary Fig. 25b). The intensity of In signals significantly increased when the temperature increased to 800 °C, suggesting that high temperatures also promote the evaporation of In species. Notably, due to the low gas pressure (2 Torr) and resultant low C₃H₆ partial pressure, the promoting effect of C₃H₆ on the evaporation of In species was weakened. Hence, the temperature of the gas phase In⁰ species detected by SVUV-PIMS was higher than that in the fixed-bed reactor. We conclude that both the high partial pressure of C₃H₆ and high temperatures are essential for inducing the evaporation of In⁰ species.

The final structure of the spent PtIn_1.0/SiO₂ catalyst (after the PDH stability test) was studied to uncover the origin of its high catalytic performance. XANES and EXAFS measurements were employed to characterize the electronic structure of the spent PtIn_1.0/SiO₂ catalyst. As depicted in Supplementary Fig. 26, the XANES spectra at the Pt L_III-edge showed that, compared to the reduced sample, the spent PtIn_1.0/SiO₂ exhibited a decrease in absorption edge and white line intensity. This indicated that Pt species were further reduced during the PDH test, consistent with the CO-DRIFTS results (Fig. 4a). The EXAFS results showed the Pt−Pt CN of 4.5 and Pt−In CN of 2.3 (Fig. 4b, Supplementary Fig. 26 and Table S2). The bonding distance of Pt−Pt and Pt−In was 2.58 Å and 2.63 Å, respectively, indicative of the small size of PtIn alloy. Noteworthily, the CN ratio of Pt−Pt and Pt−In in PtIn_1.0/SiO₂ was calculated as 2 (4.5 / 2.3), in good agreement with the theoretical value of Pt₃In (8 / 4) at a bonding distance of 2.82 Å. Hence, we deduced that the PtIn clusters evolved into Pt₃ In intermetallic during the PDH stability test.

Atomically resolved HAADF-STEM characterization was further performed to corroborate the aforementioned deduction. As shown in Fig. 4c and Supplementary Fig. 27, the spent PtIn_1.0/SiO₂ catalyst exhibited as NPs with an ordered arrangement. In addition, a large amount of residual atomically dispersed In³⁺ species were present on SiO₂ support. HAADF-STEM observations with a broad vision showed a uniform distribution with an average size of 1.3 nm. Besides, NPs larger than 2.3 nm were absent, suggesting the exceptional sintering resistance of PtIn_1.0/SiO₂. We further analyzed the structure of individual NPs by measuring the interplanar distances. For NPs in the spent Pt/SiO₂ sample, the Fast Fourier Transform (FFT) pattern showed a typical face-centered cubic crystal structure under the [011] zone axis (Supplementary Fig. 28). The d-spacing of the {111} plane was measured as 0.223 nm, closely aligning with the theoretic value. In comparison, significant lattice expansion was observed for PtIn_1.0/SiO₂. Under the [011] zone axis, the d-spacing of {111} and {200} planes were measured to be 0.234 nm and 0.203 nm, respectively (Fig. 4c and Supplementary Fig. 27). The expanded lattice spacing was in line with that of the Pt₃In intermetallic under the [011] zone axis. The structure of Pt₃In intermetallic was verified by distinguishing elements in an individual NP according to their Z-contrast (Fig. 4d–f). Along the direction of the rectangle in Fig. 4d, an oscillation of intensity was observed, attributed to the difference in Z-contrast between Pt and In. Under the [011] zone axis, the oscillation of intensity corresponded with the models and STEM simulations of Pt₃In intermetallic. Besides, the d-spacing of {111} plane was estimated to be 0.234 nm, consistent with the results in Fig. 4c. Thus, for the spent PtIn_1.0/SiO₂, we identified Pt₃In intermetallic, an ordered L1₂ structure with alternating stack of mixed Pt/In columns and Pt columns along the [011] zone axis. Combining the HAADF-STEM observations and EXAFS results, we conclude that during the PDH reaction, PtIn clusters gradually evolved into ultrafine Pt₃In intermetallic (averaging 1.3 nm), which served as the real active site in the PDH reaction (Fig. 4g).

To demonstrate the necessity of constructing ultrafine Pt₃In intermetallic via dynamic evolution, spent Pt/SiO₂ and Pt₃In NPs/SiO₂ catalysts were characterized. For Pt/SiO₂, the average size increased from 2.5 to 3.4 nm, indicating that Pt NPs underwent continuous sintering during the PDH test (Supplementary Fig. 28). For Pt₃In NPs/SiO₂, XPS results in Supplementary Fig. 18 showed that only metallic Pt and In species were present in Pt₃In NPs/SiO₂. HAADF-STEM images in Supplementary Fig. 29 indicated that Pt₃In intermetallic also underwent severe sintering, suggesting that In⁰ species could not prevent the aggregation of active Pt sites. In contrast, those In³⁺ species at the interface of Pt₃In intermetallic and SiO₂ support served as nano-islands, effectively frustrating the sintering of the evolved Pt₃In intermetallic (Supplementary Fig. 27). The quick deactivation of PtIn_0.6/SiO₂ and PtIn_0.8/SiO₂ was ascribed to In³⁺ species being not enough to prevent the sintering of PtIn clusters. To verify the above hypothesis, we characterized the reduced and spent PtIn_0.8/SiO₂. As shown in Supplementary Fig. 30, the average size of PtIn clusters increased from 1.6 nm to 2.7 nm, indicating that slight sintering occurred during PDH reaction. Thus, sufficient In³⁺ species play a crucial role in the high catalytic stability of PtIn_1.0/SiO₂. Compared with the direct synthesis of Pt₃In intermetallic with high catalytic activity, the construction of active Pt sites through dynamic evolution was demonstrated to be more effective.

Moreover, the evolved Pt₃In intermetallic exhibited an ultrafine nature (averaging 1.3 nm), smaller than reduced Pt/SiO₂. This suggests that dynamic evolution during PDH serves as an effective synthetic strategy for engineering ultrafine bimetallic alloy nanoparticles, which outperform their monometallic counterparts in catalytic performance across various reactions. To experimentally validate this hypothesis, we systematically evaluated the catalytic performance of spent PtIn_1.0/SiO₂-PDH in CO₂ hydrogenation (Supplementary Fig. 31). For Pt/SiO₂, CO₂ conversion increased from 1.8% to 28.0% as temperature rose from 300 to 500 °C. In_1.0/SiO₂ exhibited low catalytic activity and poor selectivity for CO. The catalytic activity of untreated PtIn_1.0/SiO₂ was significantly lower than that of Pt/SiO₂ due to the coverage of an In⁰ overlayer. Notably, after catalyzing PDH reaction under 67%C₃H₈/33%H₂ for 30 h, the catalytic activity of spent PtIn_1.0/SiO₂-PDH improved substantially. At 500 °C, the spent PtIn_1.0/SiO₂-PDH sample exhibited a CO₂ conversion of 41.0%, representing 3.0-fold and 1.4-fold enhancements over reduced PtIn_1.0/SiO₂ (13.5%) and Pt/SiO₂ (28.0%), respectively. These findings conclusively demonstrate that the in situ removal of In⁰ overlayer during PDH activation can be further utilized to optimize the catalytic performance in CO₂ hydrogenation. By tailoring the active metal component, this strategy can be extended to other metal-indium alloy catalysts for a broad range of reactions.

Discussion

In heterogeneous catalysis, gas-phase reaction environments drive profound dynamic evolution of catalysts that critically determine catalytic performance. In this work, we unveiled the dynamic evolution of PtIn_1.0/SiO₂ catalyst during PDH reaction at the atomic scale. Upon H₂ reduction, disordered PtIn alloy clusters with an In⁰ overlayer were formed, with large quantities of In³⁺ species still present on SiO₂ support. In³⁺ species effectively frustrated the sintering of PtIn clusters, while In⁰ species diluted Pt-Pt ensembles to improve the propylene selectivity. Propylene molecules were demonstrated to induce the evaporation of the In⁰ overlayer and the exposure of active Pt sites, leading to an observable induction period. As a result, PtIn clusters evolved into ultrafine Pt₃In intermetallic (average size ~ 1.3 nm), exhibiting a C₃H₆ productivity of 145 mol g_Pt⁻¹ h⁻¹ and mean lifetime of approximately 1400 h. These findings highlight the importance of product molecules in the dynamic evolution of active sites, offering a new perspective for the rational design of high-performance PDH catalysts.

Methods

Synthesis of In_1.0/SiO₂

500 mg of SiO₂ was dispersed in 30 mL of 10 mM NaOH aqueous solution at room temperature. The solution was stirred for 30 min. Subsequently, the mixture was centrifugated and washed with DI water twice. SiO₂ was re-dispersed in 20 mL of DI water, giving a pH of approximately 9.0. Then, 13.1 mg of In(NO₃)₃·xH₂O was added to the mixture, and the pH slightly decreased to 8.7, suggesting that most of the In³⁺ cations were adsorbed onto SiO₂ before hydrolysis. The mixture was stirred for 10 min. The sample was collected by vacuum filtration and dried at 60 °C in an oven overnight. Finally, the powder was calcined at 300 °C for 1 h. The loading of In on SiO₂ was determined as 0.9 wt% by ICP-MS measurement. For In_x/SiO₂ (x = 0.2, 0.4, 0.6, 0.8, 1.5, 3.0), the synthetic procedure was the same as that of In_1.0/SiO₂, except for varying the amount of In(NO₃)₃·xH₂O.

Synthesis of PtIn_1.0/SiO₂

The synthetic procedure of PtIn_1.0/SiO₂ follows a SEA method. In a typical synthesis, 300 mg of In_1.0/SiO₂ was immersed in 100 mL of DI water and sonicated. The pH of the solution was adjusted to 10–11 by adding 2 mL of 2 M ammonia solution. Simultaneously, 3.9 μmol of Pt(NH₃)₄(NO₃)₂ was dissolved in 25 mL of DI water to obtain the precursor solution. Under vigorous stirring, 20 mL of Pt precursor solution was slowly pumped into the aqueous solution containing In_1.0/SiO₂ in 30 min. After being stirred for another 30 min, vacuum filtration was applied to collect the sample, which was dried at 60 °C in an oven overnight. The loading of Pt on In_1.0/SiO₂ was determined as 0.12 wt% by ICP-MS measurement. The synthetic procedure PtIn_x/SiO₂ (x = 0.2, 0.4, 0.6, 0.8, 1.5) was the same as that of PtIn_1.0/SiO₂, except for changing In_1.0/SiO₂ to In_x/SiO₂.

Catalytic reaction

The PDH test was performed in a vertical, quartz fixed-bed reactor with an internal diameter of 7 mm under atmospheric pressure. Generally, 40 mg of catalysts, diluted with 960 mg quartz sand, were packed inside the quartz tube. The catalyst was reduced in 10% H₂/Ar at 600 °C for 1 h. Subsequently, the temperature was decreased to 550 °C. The WHSV was calculated based on the ratio of the mass flow rate of C₃H₈ and the mass of the catalyst. Effluent gas was analyzed by Agilent 8860 chromatography with an FID detector. For all tests, C₃H₈, C₃H₆, C₂H₆, C₂H₄, and CH₄ were detected. C₃H₈ conversion, C₃H₆ selectivity, and carbon balance were calculated via an external standard method based on Eqs. 2, 3, respectively. Carbon balance was calculated according to the following Eq. 4, giving a value between 95% and 105% for all tests.

$${\mathrm{C}}_3{\mathrm{H}}_8\mathrm{conversion}=1-\frac{{\left[{\mathrm{C}}_3{\mathrm{H}}_8\right]}_{\mathrm{out}}}{{\left[{\mathrm{C}}_3{\mathrm{H}}_8\right]}_{\mathrm{out}}+{\left[{\mathrm{C}}_3{\mathrm{H}}_6\right]}_{\mathrm{out}}+\frac{2}{3}{\left[{\mathrm{C}}_2{\mathrm{H}}_6\right]}_{\mathrm{out}}+\frac{2}{3}{\left[{\mathrm{C}}_2{\mathrm{H}}_4\right]}_{\mathrm{out}}+\frac{1}{3}{\left[\mathrm{C}{\mathrm{H}}_4\right]}_{\mathrm{out}}}\times 100\%$$ 2

$${\mathrm{C}}_3{\mathrm{H}}_6\mathrm{selectivity}=\frac{{\left[{\mathrm{C}}_3{\mathrm{H}}_6\right]}_{\mathrm{out}}}{{\left[{\mathrm{C}}_3{\mathrm{H}}_6\right]}_{\mathrm{out}}+\frac{2}{3}{\left[{\mathrm{C}}_2{\mathrm{H}}_6\right]}_{\mathrm{out}}+\frac{2}{3}{\left[{\mathrm{C}}_2{\mathrm{H}}_4\right]}_{\mathrm{out}}+\frac{1}{3}{\left[\mathrm{C}{\mathrm{H}}_4\right]}_{\mathrm{out}}}\times 100\%$$ 3

$$\mathrm{Carbon}\mathrm{balance}=\frac{{\left[{\mathrm{C}}_3{\mathrm{H}}_8\right]}_{\mathrm{out}}+{\left[{\mathrm{C}}_3{\mathrm{H}}_6\right]}_{\mathrm{out}}+\frac{2}{3}{\left[{\mathrm{C}}_2{\mathrm{H}}_6\right]}_{\mathrm{out}}+\frac{2}{3}{\left[{\mathrm{C}}_2{\mathrm{H}}_4\right]}_{\mathrm{out}}+\frac{1}{3}{\left[\mathrm{C}{\mathrm{H}}_4\right]}_{\mathrm{out}}}{{\left[{\mathrm{C}}_3{\mathrm{H}}_8\right]}_{\mathrm{in}}}\times 100\%$$ 4

SVUV-PIMS measurements

SVUV-PIMS measurements were carried out at the combustion beamline (BL03U) of the National Synchrotron Radiation Laboratory at Hefei, China. In_1.0/SiO₂ was packed in a quartz reactor with a nozzle size of approximately 0.1 mm, which was connected to an online SVUV-PIMS spectrometer. Catalysts were firstly reduced in H₂ flow (H₂:Ar = 1:9) with a total flow rate of 40 mL min⁻¹ at 600 °C for 1 h and cooled down to 550 °C. Subsequently, the gas was kept unchanged or switched to 10% C₃H₆/Ar. Data were recorded by MS from 550 °C to 800 °C. The photon energy was 9.5 eV, and the pressure was controlled at 2 Torr.

Simulations of HAADF-STEM images

The HAADF-STEM simulations were conducted with QSTEM software⁵¹. A structure model containing a PtIn cluster and SiO₂ support was established according to the experimental results. The structure model was imported into QSTEM software to generate an HAADF-STEM image. To verify the rationality of our structure model, 3D intensity distribution maps were derived and compared with the corresponding experimental ones.

Author contributions

T.Z. and H.Y. equally contributed to this work. T.Z., X.L., and J.Z. came up with the idea and designed the study. T.Z., H.Y., X.L., W.L., W.Z., and H.H. synthesized catalysts and conducted structural characterizations. S.H. conducted the HAADF-STEM simulations. R.W. processed the XAFS results. T.Z., T.X., C.L., and Y.P. conducted the SVUV-PIMS experiments. T.Z., L.L., and X.Z. conducted the XPS measurements with Ar⁺ sputtering. L.Z. and W.W. provided valuable suggestions on the research project. T.Z., H.Y., and X.L. analyzed the data and co-wrote the paper. J.Z. and X.L. supervised the work. All authors discussed the results and commented on the manuscript.

Peer review

Peer review information

Nature Communications thanks Raquel Poterla and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data supporting the findings of this study are available within the main text, figures and Supplementary Information, or from the corresponding authors upon request. Source data are provided in this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-60153-1.