Synthesis and characterizations
As illustrated in Fig. 1, six metal salt precursors (Ce, Zr, Hf, Ti, La, and Pd) are first mixed with fumed silica by ball milling. The resultant mixture is pyrolyzed at 900 °C in air to achieve the silica-templated metal oxide complex. Finally, Pd1@HEFO catalyst is obtained after etching silica with NaOH. It should be noted that the molar ratios of Ce, Zr, Hf, Ti, and La are approximately 1:1:1:1:1 confirmed by both EDS and inductively coupled plasma (ICP) results listed in Supplementary Table 1. The high surface-area HEFO is synthesized through the same steps without the addition of Pd precursor; Pd@CeO2 counterpart is prepared by the same method; the single-phase CeO2, TiO2, ZrO2, La2O3, and HfO2 are obtained from facile pyrolysis of their corresponding metal salts at 900 °C.
Powder X-ray diffraction (PXRD) (Fig. 2a) is performed to demonstrate the crystalline structure of HEFO and its corresponding single metal oxides. The cubic CeO2 (c-CeO2), monoclinic ZrO2 (m-ZrO2), monoclinic HfO2 (m-HfO2), tetragonal TiO2 (t-TiO2), and tetragonal La2O3 (t-La2O3) are observed for the single metal oxides, respectively. Interestingly, HEFO exhibits only five obvious broad peaks centered at 30.2, 34.8, 50.2, 60.1, and 62.6°, corresponding to (111), (200), (220), (311), and (222) planes of a single cubic phase. The absence of diffraction peaks indexed to m-ZrO2, m-HfO2, t-TiO2, and t-La2O3 indicates that Zr, Hf, Ti, and La are all incorporated into c-CeO2 to form a new high-entropy (CeZrHfTiLa)Ox solid solution (HEFO). Moreover, Zr4+, Hf4+, and Ti4+ except for La3+ have a smaller ion radius than Ce4+, thus resulting in an obvious shift of diffraction peaks of HEFO to a higher 2θ value compared with c-CeO2. The structural and chemical uniformity of HEFO is further evidenced by high-resolution TEM (HRTEM) and fast Fourier transfer (FFT, Fig. 2j) images, which show well-defined lattice fringes without the secondary phases. In addition, EDS-mapping results show the highly homogeneous dispersion of randomly-distributed five metal signals including Ce, Zr, Hf, Ti, and La (Fig. 2c–i), which also unambiguously suggests the formation of the high-entropy cubic phase of HEFO on nanometer scale. N2 adsorption–desorption isotherm and corresponding pore size distribution (Supplementary Figs. 1 and 2) of HEFO exhibit the emergency of rich porosity due to removal of hard-template SiO2 with a high specific surface area of 162.1 m2 g−1 (Supplementary Table 2). This porous structure of HEFO makes it a suitable candidate to be a catalyst carrier. The surface components of HEFO is mainly dominated by Zr4+, Hf4+, Ti4+, Ce4+, and La3+ shown by X-ray photoelectron spectroscopy (XPS, Supplementary Fig. 3) analysis, which indicates partial removal of oxygen in (CeZrHfTiLa)Ox (x < 2) crystal after incorporation of La compared with c-CeO2. The schematic model of cubic HEFO is then constructed (Fig. 2k) based on the above results, where the Ce atoms in c-CeO2 is randomly populated by Zr, Hf, Ti, and La atoms.
After introducing 0.5–2 wt% Pd during the synthesis of HEFO, the HEFO crystalline phase is well retained without the appearance of any additional diffraction peaks in Fig. 2b. More importantly, the absence of diffraction peaks ascribed to Pd species and the shift of diffraction peaks with the increases of Pd loading (Fig. 2b) suggests that a large portion of Pd may be incorporated into the HEFO sublattice for the formation of (PdyCeZrHfTiLa)Ox solid solution. After the introduction of Pd with different weight loading, the surface area and pore volume of Pd1@HEFO-x slightly decrease compared with pristine HEFO carrier (Supplementary Table 2). Fortunately, the surface area, pore size distribution and crystalline structure of Pd1@HEFO stay almost unchanged after both thermal and hydrothermal treatment (Supplementary Figs. 1, 2, and 4). The EDS-mapping results of Pd1@HEFO in Fig. 3a–h suggest the uniform element distribution of Pd, Ce, Zr, Hf, Ti, and La. More importantly, the agglomeration and sintering of Pd species are not observed in Pd1@HEFO (Fig. 3b), indicating the possible existence of isolated Pd sites. The HRTEM image in Fig. 3i only depicts randomly oriented lattice spacing belonging to the HEFO phase. Consistent with EDS-mapping and HRTEM results, FFT pattern in Fig. 3i (inset) again reveals the diffraction rings from (111), (200), (220), (311), and (222), attributed to the face-centered cubic (CeZrHfTiLa)Ox solid solution structure without diffraction rings ascribed to any Pd species. The above results confirm the existence of isolated Pd atoms, but the microenvironment of Pd still requires further exploration. In addition, TEM images of HEFO and Pd1@HEFO (Supplementary Fig. 5) display similar morphologies and the grains size distributions for both are centered at around 4 nm. This suggests that the morphology and particle size of HEFO will not change after the introduction of Pd. Furthermore, the chemical state of Pd in Pd1@HEFO is then investigated by XPS analysis, and the obtained binding energy (Supplementary Fig. 6) is the characteristic of electron-deficient Pd4+25,26,27. This formation of electron-deficient Pd4+ may be attributed to the electron transfer from Pd to M through Pd–O–M bonds (M = Ce, Zr, Hf, Ti, and La) in (PdyCeZrHfTiLa)Ox solid solution.
The formation of single Pd atoms is further confirmed by atomic-resolution TEM image (Fig. 4a) and the corresponding Pd EDS-mapping image (Fig. 4b). The absence of Pd atoms outside HEFO lattice (Fig. 4a) and around 6.44% surface Pd atoms determined by CO chemisorption (0.0644 μmol CO/μmol Pd, Supplementary Table 1) together indicate that the Pd atoms have been incorporated into both surface and bulk HEFO phase in Pd1@HEFO, consistent with the reported phenomena that single platinum-group metal atoms prefer to substitute cerium atoms of CeO2 rather than adsorb on its surface28. To confirm the electronic structure and coordination state of Pd in Pd1@HEFO, the X-ray absorption near-edge structure (XANES) and EXAFS measurements are performed at the Pd K-edge. XANES spectra show that the Pd K-edge absorption edge for Pd1@HEFO located between that of Pd foil and PdO (Fig. 4d), revealing the valence state of Pd is between 0 and +2, which is lower than +4 of the surface Pd from XPS. This is probably attributed to the fact that the surface Pd atoms are more likely to be contacted with oxygen and be oxidized at high temperatures. Fourier-transformed k3-weighted EXAFS spectra (Fig. 4e and Supplementary Table 3) exhibit the obvious Pd–Pd (bond length = 2.74 Å) and Pd–O–Pd (bond lengths = 3.06 and 3.44 Å) features for Pd foil and PdO references, respectively, which are both absent in Pd1@HEFO. As an alternative, the bond lengths at 3.01 and 3.26 Å corresponding to Pd–O–Zr and Pd–O–M (M = Ce/La) are identified in Pd1@HEFO (Supplementary Table 3), illustrating the existence of isolated Pdx+ (0 < x < 2) in proximity to Zr, Ce, or La atoms29,30,31. The wavelet transform plot (Fig. 4f) of Pd1@HEFO shows the wavelet transform maximum at ~10 Å−1, which corresponds to the Pd–O–Zr and Pd–O–M (M = Ce/La) bonding by comparing Pd foil and PdO counterparts and the intensity maxima of Pd–O–Ti at ca. 7 Å−1 and Pd–O–Hf at above 12 Å−132,33. Moreover, no intensity maxima corresponding to Pd−Pd and Pd–O–Pd is found, which matches well with the EXAFS fitting results in R space. Consequently, a possible schematic model of Pd1@HEFO (220) is shown in Fig. 4c.
Operando diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) upon CO adsorption in Fig. 5a is then employed to examine the coordination environment of Pd in Pd1@HEFO and Pd@CeO2. To strengthen the CO adsorption on Pd sites, HEFO, Pd1@HEFO, and Pd@CeO2 are reduced in situ at 250 °C in the DRIFTS cell before exposure to CO flow. For HEFO, the absence of FTIR bands in Fig. 4a suggests the surface of HEFO is not capable of adsorbing CO molecules. In comparison, CO frequencies between 2200 and 2000 cm−1 are clearly seen in Pd1@HEFO, which is assignable to CO molecules linearly adsorbed on single-atom Pd species34,35. The bridge and hollow-CO bands are not seen in Pd1@HEFO, indicating that Pd are atomically dispersed on the HEFO host, which agrees with HAADF-STEM and EXAFS results34,36. However, the bridge and hollow-CO peaks are obviously observed for Pd@CeO2 because of the aggregation of Pd species, thus further validating the importance of hosts for the formation of isolated Pd species37. The obvious agglomeration of Pd in Pd@CeO2 can be also evidenced by EDS-mapping results (Supplementary Fig. 7) and existence of the metallic Pd phase from PXRD pattern (Supplementary Fig. 8), in agreement with the bridge and hollow-CO peaks of DRIFTS. The chemical states of Pd in Pd@CeO2 and Pd1@HEFO are both dominated by Pd4+ based on XPS results (Supplementary Fig. 6), which suggests their similar Pd–O coordination number though totally different Pd microenvironment. As is well known, the phase stabilization is a process determined by combination of the enthalpy (H) and entropy (S) effects, which are temperature- and composition-dependent. Compared with the Pd@CeO2, Pd1@HEFO with enhanced compositional complexity provides a higher molar configurational entropy, especially for equimolar cations, which then potentially decreases the Gibbs free energy according to the equation (ΔG = ΔH−TΔS). This means that the formation of (PdyCeZrHfTiLa)Ox solid solution is an entropy-dominated process, whereas the decreased configuration entropy induces the dissociation of (PdyCe)Ox as an enthalpy-driven process. To prove this hypothesis, the PXRD patterns of Pd@ZrO2, Pd@La2O3, Pd@HfO2, Pd@TiO2, ternary Pd@CeZrTiOx, and quaternary Pd@CeZrHfTiOx samples synthesized by the same method are also collected (Supplementary Fig. 8). The diffraction peaks ascribed to metallic Pd and/or PdO can be observed, further confirming the importance of the high configurational entropy on stabilizing the Pd single atoms. We also synthesized the Pd/HEFO-p (single Pd atoms dispersed on HEFO carrier) by a post-deposition method, where the single-atom structure can be confirmed by XRD pattern and CO-DRIFTS spectra in Supplementary Fig. 9. Unfortunately, the sintering and aggregation of Pd on HEFO can be obviously observed in Pd/HEFO-p-900 after post-treatment at 900 °C, which might be ascribed to the excessive Pd density on HEFO surface.
The oxidation of CO, a key reaction in automotive emission abatement, has been extensively investigated in the past decades38,39,40,41. Therefore, the light-off curves of CO oxidation are measured to evaluate the catalytic efficiency of our as-obtained samples. As depicted in Fig. 5d, HEFO exhibits an inferior catalytic activity of CO oxidation with a high onset temperature of 230 °C. After doping 1 wt% Pd, Pd1@HEFO shows a dramatically enhanced reactivity with the onset temperature as low as ~80 °C and complete CO oxidation at 170 °C. In comparison, the onset temperature and T100 over Pd@CeO2 are 223 and 253 °C, respectively, much higher than those of Pd1@HEFO. It is generally accepted that metal on reducible carrier (CeO2) follows a Mars–van Krevelen reaction mechanism, where CO adsorbed on metals reacts with active lattice O from CeO2 to form oxygen vacancies42,43. Therefore, the reducibility of the surface lattice O from the carrier plays an important role in catalytic CO oxidation. For Pd@CeO2, the only reduction peak observed at around 250 °C is assigned to the reduction of Pd–O–Pd (PdOx) and the surface lattice oxygen in CeO244,45. In contrast, the surface lattice O of HEFO in the vicinity of Pd (Pd–O–M) is easier to be activated for Pd1@HEFO compared with Pd–O–Pd for Pd@CeO2, which offers a reduction temperature centered at 160 °C. The enhanced reducibility of the surface lattice O in Pd1@HEFO should be the main cause for its higher catalytic activity. Moreover, XPS spectra of O 1 s in Fig. 5c show that the ratio of the surface chemisorbed oxygen (Oβ, 530.4 eV) to the support’s lattice oxygen (Oα, 528.8 eV) of Pd1@HEFO is about twofold higher than that of Pd@CeO2, which might be ascribed to the existence of more under-coordinated metal cations due to the incorporation of Pd into HEFO46,47, consistent with the surface concentration of Ce3+ in Supplementary Fig. 10. These as-formed under-coordinated sites, namely oxygen vacancies, in Pd1@HEFO would render more habitation for O2 dissociation adsorption41. Combined CO-TPR with O 1s XPS results, the improved reducibility of partial surface lattice oxygen in vicinity of Pd and enhanced oxygen vacancies in Pd1@HEFO should be the main cause for its superior catalytic performance. The apparent activation energies (Ea) of Pd@CeO2 and Pd1@HEFO were calculated, as shown in Fig. 5e. Pd1@HEFO has an Ea value of 43.40 kJ/mol, which is much lower than that of Pd@CeO2 (72.21 kJ/mol), further identify the advantage of Pd1@HEFO catalyst. As shown in Fig. 5f, the cycled measurement of CO oxidation over Pd1@HEFO and time-on-stream test at 170 °C (inset) reveal the outstanding stability of Pd1@HEFO. Since water vapor is usually present in vehicle exhaust, we thereafter examine the hydrothermal stability of both Pd@CeO2 and Pd1@HEFO catalysts by treating them at 750 °C for 10 h before the activity test. After hydrothermal treatment, the textural and structural properties of Pd1@HEFO–HA remain almost unchanged (Supplementary Table 2 and Fig. 5). Correspondingly, all elements are still uniformly dispersed without agglomeration and sintering of Pd species from the EDS-mapping result in Supplementary Fig. 11. More importantly, the absence of the bridge and hollow-CO peaks further validates that single-atom Pd in Pd1@HEFO–HA is stable under hydrothermal conditions (Supplementary Fig. 12). Consequently, the complete conversion temperature of CO barely changes after the hydrothermal treatment in Fig. 5d and even 10 vol% H2O in the feed gas (Pd1@HEFO-H2O, Supplementary Fig. 13). The increase of low-temperature catalytic activity of Pd1@HEFO–HA and Pd1@HEFO-H2O (Supplementary Fig. 13) is probably ascribed to the formation of activated surface chemisorbed oxygen (Oβ) on HEFO, reported in the previous study41. However, the catalytic activity of Pd@CeO2–HA obviously decreases compared with its fresh counterpart due to the existence sintering of Pd species, evidenced by the decreased linear-CO and increased bridge and hollow-CO peaks of Pd@CeO2–HA (Supplementary Fig. 12). In addition, Pd1@HEFO not only exhibits better thermal and hydrothermal stability than Pd@CeO2 in this work, but also shows better or comparable performance relative to other reported representative catalysts of CO oxidation20,41,48,49. More importantly, Pd1@HEFO exhibits simultaneously outstanding oxidation activities of CO, C3H6, and NO at a high gas-hourly space velocity (GHSV) of 200,000 mL gcat−1 h−1 (Supplementary Fig. 14a), although the T100 of CO oxidation shifts to ca. 260 °C due to the co-presence of C3H6 and NO at a high GHSV. The catalytic performance of Pd1@HEFO is comparable to Pt/CeO2–SiAlOx regarded as a candidate of diesel oxidation catalyst (DOC)50 and Pt/CeO240. Moreover, no obvious deactivation of CO, C3H6, and NO oxidation can be observed over Pd1@HEFO even after 10 h of reaction in Supplementary Fig. 14b, implying that Pd1@HEFO shows a good DOC activity and stability. PXRD patterns of Pd1@HEFO treated in H2 at different temperatures (Supplementary Fig. 15) suggest that Pd atoms in Pd1@HEFO are not stable in reductive atmosphere. As a result, our Pd1@HEFO is more suitable for the oxidation reaction under the oxygen-rich conditions, such as catalytic destruction of pollutions emitted from diesel engines. These illustrate that our Pd1@HEFO possesses not only outstanding low-temperature CO oxidation activity but also excellent resistance to hydrothermal degradation as a candidate of DOC, thus possibly tolerating the harsh conditions during exhaust treatment of diesel engines.