The design strategy integrates three components into the final catalyst: 1) a robust, high-surface-area support (e.g., SiO2, Al2O3, etc.), 2) nanoscale functional metal oxides (e.g., CeOx, TiOx, FeOx, etc.) anchored stably onto the robust support as isolated nanoglue islands, and 3) single metal atoms (M1) selectively localized to only the nanoglue islands. The nanoglue selection criteria include a) its stability in dispersed form on the support surface due to strong bonding, b) a much stronger affinity for the active metal atoms than the support, and c) interactions with the active metal that enhance activity and/or selectivity for the desired catalytic reactions. The selected nanoglue not only behaves as a “double-sided tape” but also contributes to the desired functions for the targeted catalytic reaction.
We selected SiO2, an irreducible, inexpensive support widely used in processing industries, to demonstrate our strategy because of its high-surface-area, structural stability, and availability in various forms14. Because metal atoms anchor onto reducible metal oxides (e.g., CeO2, TiO2, Fe2O3, Co3O4, etc.) at defect sites and/or via formation of strong M1-Ox bonds8,15-18, and ceria has unique redox and oxygen-storage properties19,20, we choose CeOx nanoclusters as prototype nanoglues to localize Pt atoms for CO oxidation reaction. The critical aspect of this new design strategy is to produce ultra-small, isolated CeOx nanoclusters uniformly distributed on high-surface-area SiO2 support via a scalable/practical synthesis process.
Uniform CeOx nanocluster islands were synthesized by strong electrostatic adsorption (SEA) of charged species from aqueous solution21, as schematically illustrated in Fig. 1. The point of zero charge (PZC) of the high-surface-area SiO2 (278 m2/g) is ~3.6 (ref.21), implying that its surface is negatively charged in an aqueous solution with pH > 4.0. Control of the solution OH- concentration and adsorption time yielded soluble cationic [Ce(OH)x]y+ species that quickly adsorbed onto the SiO2 surfaces, leading to (after a high-temperature calcination) formation of uniformly dispersed CeOx nanoclusters. Short adsorption time (< 3 min) usually produces uniformly coated Ce species while extended adsorption time leads to formation of large CeO2 particles (Extended Data Fig. 1a-b). High-angle annular dark-field scanning transmission electron microscope (HAADF-STEM) images showed uniform coating of the mesoporous SiO2 support with Ce species as a result of the SEA process (Extended Data Fig. 1c-e). Subsequent calcination at 600° C produced individual crystalline CeOx nanoclusters stably anchored onto SiO2 surfaces (Fig. 2a and Extended Data Fig. 1f-g). The CeOx loading was 12 wt%, determined by inductively coupled plasma mass spectrometry (ICP-MS), and the average CeOx nanocluster dimensions were 1.8 nm × 2.1 nm (Extended Data Fig. 1h-i), with ellipsoid shapes. Atomic-resolution HAADF-STEM images (Fig. 2b) show that all the as-synthesized CeOx nanocluster are well crystallized and some show visible surface steps.
Powder X-ray diffraction (XRD) patterns of the as-synthesized CeOx/SiO2 show broad diffraction peaks that represent a cubic fluorite structure (Fig. 2c). In comparison, the two control samples (12 wt% CeO2 nanoparticles (NPs) on SiO2, denoted as CeO2 NPs/SiO2, and pure CeO2 powders, made by an impregnation or precipitation method, as stated in Fig. 2c and Extended Data Fig. 2a) yielded strong diffraction peaks, demonstrating large sizes of the as-prepared CeO2 NPs. HAADF-STEM images show that the as-prepared CeO2 NPs have a wide size distribution (Extended Data Fig. 2b-c, e-f). Raman spectra show a broadened and red-shifted active mode from 462 to 448 cm-1, suggesting significant lattice distortion in the as-synthesized crystalline CeOx nanoclusters (Extended Data Fig. 2g)22. The additional Raman band near 600 cm-1 indicates presence of oxygen vacancies22. Measurements of lattice distances of the CeOx nanoclusters in numerous HAADF-STEM images showed an expansion of the CeO2{111} plane distance from 3.1 Å (bulk CeO2) to 3.3 Å in the CeOx nanoclusters (Extended Data Fig. 2h), in agreement with the slight shift of the XRD peaks to lower angles (Fig. 2c).
X-ray photoelectron spectroscopy (XPS) data (Fig. 2d and Extended Data Fig. 2d) show 28.7 %, 10.9 % and 8.4% Ce3+ species in CeOx/SiO2, CeO2 and CeO2 NPs/SiO2, respectively12,23. The significantly increased number of Ce3+ sites on the as-synthesized CeOx nanoclusters suggests more anchor sites for metal atoms8,15,24. The fact that XPS probes surfaces of nanometers in depth23 implies that the Ce3+/(Ce3+ + Ce4+) ratio (estimated from the XPS data) can be used to estimate the average composition of the CeOx nanoclusters since the heights of the as-synthesized CeOx nanoclusters are less than 2 nm. The estimated composition of the as-synthesized CeOx nanoclusters is CeO1.86. Analyses of H2 temperature-programmed reduction (H2-TPR) showed that the reduction temperature of “bulk” oxygen from the CeO1.86 nanoclusters took place at ~492 °C (290 °C lower than that of CeO2 powder) (Extended Data Fig. 2i), suggesting that full reduction of the as-synthesized CeO1.86 nanoclusters is much easier than that of CeO2 NPs, providing a route to facile formation of oxygen species. The as-synthesized ultra-small, isolated CeO1.86 nanoclusters are crystalline in nature, possess high-number density of surface defect sites, and provide labile oxygen species—well-suited to strong bonding of isolated Pt atoms.
Selective deposition of Pt atoms to the CeO1.86 nanoglue clusters (Fig. 1)—but not the SiO2 support—via a synthesis protocol was accomplished by the SEA process: opposite-charged species attract, and like-charged species repel. Because CeO2 and SiO2 possess PZCs (points of zero charge) of ~8.1 (ref.25) and 3.6, respectively, the sign of the surface charges of these two materials should be opposite when the solution pH is adjusted to between 4.0 and 6.0. After slowly titrating the PtCl62- precursor into the CeO1.86/SiO2 solution, the positively charged CeO1.86 nanoclusters strongly adsorbed the PtCl62- complexes while the negatively charged SiO2 surfaces strongly repelled these negatively charged species. The XPS measurements showed that after the preferential Pt deposition the amount of Ce3+ species in the CeO1.86/SiO2 was retained, and residual chloride was not detectable, removed by high-temperature calcination of the as-synthesized Pt/CeO1.86/SiO2 catalysts (Extended Data Fig. 3a-c). To selectively synthesize SACs, the Pt loading was controlled to be ≤ 0.4 wt% (with respect to the CeO1.86 nanoclusters) since the average number density of Pt atoms on each CeO1.86 nanocluster in the 0.4 wt% Pt/CeO1.86/SiO2 catalyst was estimated to be less than one (~0.6) Pt atom per CeO1.86 nanocluster (Extended Data Fig. 3d-f). The ICP-MS measurements showed that ~99 % of the adsorbed Pt species were deposited onto the CeO1.86 nanoglues while negligible amount of Pt was deposited onto the SiO2 control sample via the same SEA process (Extended Data Fig. 3d). These results clearly and unambiguously demonstrated selective deposition of Pt atoms onto only the CeOx nanoglues via the SEA process which is scalable and industrially practical.
The location identification of metal atoms with respect to the support surface structure is critical to understanding the catalytic properties of SACs. Even though Pt or Au atoms on well-crystallized CeO2 NPs have been reliably observed in atomic-resolution HAADF-STEM images26, Pt atoms on ultra-small CeOX nanoclusters could not be unambiguously identified (Extended Data Fig. 4a-c and Methods). After intensive investigations of HAADF-STEM imaging of CeO1.86/SiO2 supported Pt atoms/clusters, we concluded that Pt clusters with a size > 0.4 nm could be reliable detected by direct HAADF-STEM imaging (Extended Data Fig. 4d). Based on analyzing numerous atomic-resolution HAADF-STEM images of as-synthesized 0.4 wt% Pt/CeO1.86/SiO2 catalysts, we concluded that there were no Pt single atoms/clusters/particles on the SiO2 surfaces (in agreement with the ICP-MS measurement of the control sample) and all the Pt species were deposited only onto the CeO1.86 nanoglue islands either in the form of single Pt atoms or Pt clusters with sizes < 0.4 nm.
The Pt LIII-edge X-ray absorption near-edge structure (XANES) and the Fourier transform radial distribution functions of the k3-weighted extended X-ray absorption fine structure (EXAFS) spectra of the 0.4 wt% Pt/CeO1.86/SiO2 catalyst are displayed in Fig. 3a-b. The XANES spectra suggest that the oxidation state of the Pt species is very close to that of Pt in PtO2, consistent with a previous report27. The EXAFS spectrum gives no evidence in the 0.4 wt% Pt/CeO1.86/SiO2 catalyst of a Pt–Pt shell, in contrast to what is commonly observed for Pt or PtO2 NPs or clusters (Extended Data Fig. 5), bolstering the conclusion that the Pt species existed as single, isolated atoms. The coordination number of the Pt–O shell in the model was found to be 4.5 ± 0.5 with a bonding length of 1.97 ± 0.02 Å, consistent with the EXAFS analyses reported for site-isolated platinum anchored on ceria and iron oxide1,27. Diffuse-reflectance infrared Fourier-transform spectroscopy (DRIFTS) spectra on the 0.4 wt% Pt/CeO1.86/SiO2 catalyst show a well-defined peak at 2103 cm-1 (Fig. 3c), assignable to CO adsorbed on ionic single Pt atoms8,18,28.
The behavior of Pt atoms in the Pt/CeO1.86/SiO2 SAC under various treatment conditions were extensively investigated by HAADF-STEM and CO DRIFTS methods. For comparison, Pt atoms supported on SiO2 and CeO2 were synthesized and evaluated as control catalysts. Under either a reducing or oxidizing environment at temperatures >300 °C, Pt atoms sintered significantly on the SiO2 support (Extended Data Fig. 6), demonstrating extremely weak interactions between Pt atoms and the SiO2 support29. On the other hand, CeO2-supported Pt atoms were stable during calcination in air, even at high temperatures (Extended Data Fig. 4e-f)8,12. During H2 reduction at 300 °C for 1 h, however, the CeO2-supported Pt atoms sintered to form Pt NPs (Extended Data Fig. 7a-c), implying that the Pt-O-Ce bonds were broken and the Pt atoms moved on reducible CeO2 surfaces. After H2 reduction treatment, the sizes of Pt particles on CeO2 powders were smaller than those on SiO2 (Extended Data Fig. 7a-b and 6c-d), confirming that the Pt-CeO2 interaction is much stronger than the Pt-SiO2 interaction under an H2 atmosphere. Dispersed Pt atoms on CeO2 surfaces, however, became mobile at temperatures ≥ 300°C under H2 environment. In stark contrast, the Pt atoms in the 0.4 wt% Pt/CeO1.86/SiO2 catalyst did not show any sign of sintering, even after H2 reduction treatment at 300 °C for 10 h, as evidenced by DRIFTS spectra and HAADF-STEM images (Fig. 3d and Extended Data Fig. 8a-b). To further probe the stability of Pt atoms localized on the CeO1.86 nanoglue islands, catalyst samples were exposed to H2 at 400 °C to 600 °C. Even under such harsh reduction conditions, the Pt atoms remained cationic (Extended Data Fig. 8c-e). For catalysts that were reduced above 500 °C, however, significantly blue-shifted CO absorption peaks were observed, probably suggesting a major modification of the Pt-CeO1.86 interaction. Although we could not rule out the possibility that Pt atoms moved within their own CeO1.86 nanoglue islands during H2 treatment but the Pt atoms did not move onto the SiO2 surfaces—if they had, Pt clusters and particles would have been readily detected by HAADF-STEM imaging and DRIFTS.
Taking these results together, we infer that both the SiO2-supported CeO1.86 nanoglue islands and CeO2 NPs confine movement of Pt atoms during the H2 reduction treatment and that Pt clusters should form if each CeO1.86 nanocluster or CeO2 NP contains more than one Pt atom. To further test the hypothesis that Pt atoms on CeO2 sinter under H2 reduction treatment, a catalyst which contained Pt atoms dispersed on CeO2 NPs, denoted as 0.4 wt% Pt/CeO2 NPs/SiO2 SAC, was exposed to H2 at 300 °C for 1 h. All the experimental characterization data (Extended Data Fig. 7d-i) show that although the Pt atoms did not migrate onto the SiO2 surfaces to form larger particles they did sinter to form small Pt clusters on the CeO2 NPs—because some of the larger CeO2 NPs evidently contained more than one Pt atom. The CO DRIFTS spectra (Extended Data Fig. 7g-h) clearly show the representative peaks of CO adsorbed on Pt clusters/NPs. Because of the lower Pt loading (0.4 wt%), some CeO2 particles might still contain single Pt atoms after the reduction treatment. In striking contrast to the relatively sharp DRIFTS peak on the 0.4 wt% Pt/CeO1.86/SiO2 (Fig. 3d), the broad overlapping DRIFTS peaks on the reduced 0.4 wt% Pt/CeO2 NPs/SiO2 SAC clearly demonstrate the presence of highly heterogeneous dispersity of the Pt species on the CeO2 NPs.
To further verify our conclusion that during a high-temperature H2 reduction process, Pt atoms located on the same CeO1.86 cluster sinter while Pt atoms located on different CeO1.86 clusters do not move together to sinter, high amount of Pt (4 wt%) atoms were loaded onto SiO2 supported CeO1.86 clusters to produce a Pt cluster catalyst. After reducing the 4 wt% Pt/CeO1.86/SiO2 catalyst in H2 at 300 °C for 3 h, Pt clusters/NPs were clearly detected by CO DRIFTS (Extended Data Fig. 8f). Uniform Pt clusters, with an average size of ~0.9 nm when the reduction temperature was increased to 400 °C, were attached onto the CeO1.86 nanoglue islands (Extended Data Fig. 8g). After a more severe reduction, at 500 °C for 12 h, the CeO1.86 nanoglue islands became amorphous; the Pt clusters, however, retained their sizes and were still attached to the CeO1.86 nanoglue islands, unambiguously indicating strong adhesion between Pt species and the CeO1.86 nanoglue islands (Extended Data Fig. 8h). The absence of Pt particles with sizes >1 nm in diameter in these highly reduced catalysts is a clear manifestation of localization of the Pt species onto the CeO1.86 nanoglue islands even when the Pt species become mobile on their own CeO1.86 nanoglue islands—no cross-movement of Pt species among the CeOx nanoclusters. These results firmly demonstrate that our localization design applies not only to supported metal atoms but also supported subnanometer metal clusters (Extended Data Fig. 8i), significantly expand practical applications of our nanoglue localization strategy in contrast to previous stabilization approaches24,28-30.
For many catalytic reactions, the active phase is usually activated by H2 reduction prior to a desired catalytic reaction. For metal-oxide-supported SACs, H2 treatment at temperatures > 200 °C usually causes sintering of metal atoms12,13. Because of such detrimental sintering effects, many SACs were directly used without being activated by such a H2 reduction treatment, hindering the true measurement of the catalytic performance of the as-prepared SACs. Due to saturation by oxygen ligands, many SACs may not show catalytic activity after a moderate- to high-temperature calcination treatment. Since our 0.4 wt% Pt/CeO1.86/SiO2 SACs resist sintering during H2 activation processes, we can quantitatively evaluate how the H2 activation process affects CO oxidation on CeO1.86 nanoglue supported Pt1 atoms. Prior to the H2 reduction treatment, the as-synthesized 0.4 wt% Pt/CeO1.86/SiO2 SAC had relatively low activity for CO oxidation (Fig. 4), in agreement with reports on Pt1/CeO2 SACs8,12-13. The H2 activated Pt/CeO1.86/SiO2 SAC, however, achieved 50% and 90% CO conversion at 133 °C and 142 °C, respectively. For comparison, the T50 (temperatures for 50% CO conversion) for the activated Pt/CeO1.86/SiO2 (impregnation-IMP)) and activated Pt/SiO2 (IMP) catalysts were 171 and 227 °C (Fig. 4b), respectively. In particular, the H2 activation process had a much bigger effect on impregnated Pt species supported on reducible CeO1.86 nanoclusters than on Pt species supported on nonreducible SiO2, because Pt/CeO1.86/SiO2 (IMP) showed higher activity and Pt dispersion than Pt/SiO2 (IMP) after H2 reduction (Extended Data Fig. 9a-d).
The turn-over-frequency (TOF) and specific reaction rate (normalized by Pt mass) under similar CO oxidation conditions were evaluated for the various catalysts (Extended Data Fig. 9e). The TOF for CO oxidation on the as-synthesized 0.4 wt% Pt/CeO1.86/SiO2 SAC was 0.012 s-1 at 150 °C, similar to the reported value13. The H2 activation markedly increased the TOF to 1.8 s-1, 150 times higher than that of the non-activated catalyst. The specific rate at 150 °C on the activated 0.4 wt% Pt/CeO1.86/SiO2 SAC was 20 and 204 times higher than that on the activated 0.4 wt% Pt/CeO1.86/SiO2 (IMP) and Pt/SiO2 (IMP), respectively. The activated 0.4 wt% Pt/CeO1.86/SiO2 SAC possesses the lowest apparent activation energy (Ea = ~68 kJ/mol) (Extended Data Fig. 9f) and was stable during four separate light-off cycles (Fig. 4a). Time-on-stream conversion test at 140 °C verified the long-term stability of the activated 0.4 wt% Pt/CeO1.86/SiO2 SAC (Extended Data Fig. 9g-h).
In summary, we developed a scalable and practical process to localize Pt atoms on functional CeOx (x = 1.86 by XPS measurement) nanoglue islands (average size = ~ 2nm) dispersed on a robust, high-surface-area SiO2 support. A facile and scalable SEA process was used to synthesize both the ultra-small, isolated CeO1.86 nanoglue islands and the selective deposition of Pt atoms onto only those CeO1.86 nanoglue islands. The SiO2-supported CeO1.86 nanoglue islands possess abundant Ce3+ species, and strongly anchor and localize Pt atoms/clusters under both O2 or H2 environment, even at elevated temperatures. For CO oxidation reaction, activation of CeO1.86-supported Pt atoms increased the TOF by 150 times. The CeOx nanoglue islands not only localize Pt atoms to prevent sintering but also provide facile oxygen during CO oxidation reaction. Our strategy to confine the movement of metal atoms by isolated nanoglue islands extends to metals other than Pt (Extended Data Fig. 10). The use of functional nanoglues on robust high-surface-area supports to localize metal atoms can be broadly applied to creating a wide range of robust single-atom catalysts for a plethora of important catalytic transformations.