Engineering catalyst supports to stabilize Pd/PdO 2D rafts for water-tolerant methane oxidation

The size and morphology of the active phase (metal or metal oxide) are critical for the performance of heterogeneous catalysts. Conventional approaches for catalyst synthesis involve the modication of pore size and structure, the use of ligands to anchor the metal during preparation or the use of nanostructured oxides with well-dened facets to provide suitable sites for metal nucleation and growth. However, these approaches may not yield durable catalysts for high temperature applications, such as the treatment of unburnt methane from natural gas fueled engines. Here we demonstrate an approach that relies on the trapping of metal single atoms on the support surface, in thermally stable form, to modify the nature of deposited metal/metal oxide clusters. By anchoring Pt ions on the catalyst support we can tailor the morphology of the deposited phase. In particular, two-dimensional (2-D) rafts of Pt/PtO x on the engineered catalyst support are formed by this approach, as opposed to three-dimensional (3-D) metal oxide nanoparticles on conventional supports. Adopting this approach for the synthesis of bimetallic catalysts via addition of Pd to the atom-trapped catalyst support (Pt@CeO 2 ) we found that the resulting Pd/Pt@CeO 2 catalyst provides improved thermal stability and water tolerance during methane oxidation. We attribute the improved performance to the 2-D morphology of the Pd/PdO phase present on the atom-trapped catalyst support. The results show that modifying the support by trapping single atoms could provide an important addition to the toolkit of catalyst designers to engineer catalyst supports for controlling the nucleation and growth of metal and metal oxide clusters in heterogeneous catalysts. aberration-corrected electron microscopy the 2-D with average size 0.8 as well as the presence Pt single atoms. Further methane oxidation results under the addition of 4-10 v/v% steam show that the 1Pd/2Pt@CeO 2 catalyst has improved water-tolerance in methane oxidation than conventional 1Pd/CeO 2 and (1Pd+2Pt)/CeO 2 catalysts. The critical difference is the nature of Pd/PdO phase in the 2-D rafts, which is similar a oxide on Pd metal surface. The monolayer Pd oxide is not stable, since the Pd transforms which by water vapor. The superior water tolerance 2 different form on via The performance very different The experimental observations are rationalized with model DFT calculations which show differences in the interaction of water vapor with the compared with bulk PdO. This work demonstrates that the approach of can be used to engineer a support improved catalytic performance of the deposited phase. We suggest this approach since nucleation support.


Introduction
The control over the size, morphology and oxidation state of clusters and nanoparticles is important for achieving optimum performance in heterogeneous catalysis. There are limited means available for catalyst designers to in uence the nature of the active phase, especially when the catalysts are subjected to elevated temperatures. The most widely used approach for industrial catalysts involves adsorption of nucleation and morphology of the dispersed phase (Scheme 1b). Since the trapped metal atoms are stable at high temperatures, we can generate a catalyst suitable for high temperature applications such as methane oxidation.
Methane combustion catalysis is becoming increasingly important to address the emission of unburned methane associated with the widespread use of natural gas as a clean fuel for vehicles and for power generation. Pd-based catalysts show the highest activity for methane oxidation, however they suffer from rapid deactivation due to the presence of H 2 O and sintering of the Pd phase 10 . Because of the excess O 2 present in the reactant stream, the stable phase of Pd under reaction is PdO. The sintering of Pd catalysts under steam is ascribed to the formation and migration of mobile Pd-OH entities in the presence of H 2 O molecules 11 , which are strongly adsorbed on the PdO surface. The binding of H 2 O to the PdO also causes poisoning of the surface sites for methane oxidation 12 . Such surface poisoning by steam is con rmed by the fact that the surface coverage of CH 4 in the presence of water vapor is limited to only 3% of the saturation coverage of a pure CH 4 layer on PdO (101) 13 . In our previous work, when Pd catalysts were reduced prior to reaction, we observed enhanced low-temperature reactivity 14 . Similar enhanced reactivity due to pre-reduction of Pd catalysts (also termed in-situ activation) has been reported by other workers 15,16 . We attribute this higher reactivity to a different form of surface Pd oxide formed on the surface of the reduced Pd catalyst (in contrast to bulk PdO). This is consistent with DFT calculations which show that a surface oxide is more active than bulk PdO nanoparticles 17,18 . However, when Pd based catalysts are used for methane combustion, the metal Pd nanoparticle transforms to PdO, leading to the decrease in catalytic activity 19 . A well-known approach for improving the stability and reactivity of Pd based catalysts is to add Pt, which stabilizes the Pd phase in a metallic state by forming bimetallic PtPd even under oxidizing conditions 20 , leading to enhanced resistance to adsorption of water molecules. However, the bimetallic PtPd catalysts are not as resistant to sintering as PdO catalysts, showing the formation of large alloy PtPd particles (> 20 nm) and declined catalytic reactivity 14,21 . Therefore, there is an urgent need to develop methods to form an active Pd phase which is resistant to H 2 O poisoning and to catalyst sintering 22 .
In this work, we have modi ed the catalyst support via strongly bound Pt single atoms (atom trapping 23-25 -high temperature vapor phase synthesis), to control the nature of the metal and metal oxide phase.
We show that a 2D morphology of the Pd/PdO phase synthesized with this approach is thermally-stable and resistant to water poisoning. The catalysts thus prepared overcome the disadvantages of large PdO particles which are readily poisoned by water, leading to decreased activity, or single atom Pd on the support which is less active for methane oxidation 26 .

Catalyst preparation
The initial experiments to investigate the role of atom trapped Pt were performed with the commercial ceria (Solvay-Rhodia HS 5) henceforth referred to as HSA ceria since it has a higher surface area than the ceria prepared by calcination of Ce(NO 3 ) 2 which was used for the Pd-Pt catalysts. To prepare engineered CeO 2 , tetraamine platinum nitrate solution was impregnated onto CeO 2 powder by conventional wet impregnation, followed by drying at 120 o C for 12 h. The powder was then calcined at 800 o C for 10 h in owing air. The obtained materials were denoted as 1Pt@CeO 2 and 3Pt@CeO 2 , corresponding to the Pt loadings of 1 and 3 wt.%, respectively. The high temperature treatment causes a drop in BET surface area, from 135 m 2 /g for the as-received ceria to 60 m 2 /g for the pure ceria (referred to henceforth as PS or pre-sintered ceria), 78 m 2 /g for 1Pt@CeO 2 and 100 m 2 /g for the 3Pt@CeO 2 sample. The Pt interacts strongly with the ceria surface, pinning defect sites and slowing the rate of ceria sintering 24,27 .

Catalyst characterization
Aberration corrected scanning transmission electron microscopy (AC-STEM) was carried out using a FEI Titan Themis transmission electron microscope equipped with the Gatan K2-IS camera. We also used a JEOL ARM200CF microscope for some of the samples. In each case, the sample powders were deposited on holey carbon lms after being dispersed in ethanol. The single Pt atoms on CeO 2 can be clearly seen in the AC-STEM dark eld images. Low energy ion scattering (LEIS) was used to quantify the concentration of Pt atoms on the catalyst surface. This technique selectively detects the topmost atoms and was carried out using an IONTOF Qtac100 instrument 28 . The dedicated LEIS instrument was equipped with a double toroidal analyzer for the energy analysis of the backscattered ions and the analyzer has a large solid angle of acceptance (full 360° azimuth), while the scattering angle is xed at 145°. In combination with parallel energy detection, this gives a high sensitivity while maintaining the mass resolution. He + and Ne + with ion energy 3 and 5 keV and current 5 and 2 nA, respectively, were used to analyze the surface concentration of Pt 2+ ions in the catalysts. The area scanned per sample by these two ions was 2 × 2 mm 2 and the ion ux used was 1.4 × 10 14 ions/cm 2 and 2.8 × 10 13 ions/cm 2 respectively. The spectra for the Pt reference were analyzed within a 1.5 × 1.5 mm 2 sputter crater over an area of 1 × 1 mm 2 . The analysis time was adjusted such that the surface damage was the same as for the other spectra. Assuming a sputter coe cient of 0.1 for He and 1 for Ne, this will lead to a surface damage of 1% and 2% at the end of the analysis. X-ray photoelectron spectroscopy (XPS) was performed using a Kratos Axis Ultra photoelectron spectrometer equipped with a monochromatic Al Kα source operating at 300 W. The base pressure was 2.7 × 10 -8 Pa and operating pressure was 2.7 × 10 -7 Pa. Analysis of the XPS spectra was performed using CASA XPS software. X-ray absorption spectroscopy was performed at the materials research collaborative access team (MRCAT) bending magnet and insertion device lines at the Advanced Photon Source, Argonne National Laboratory. Measurements at the Pd K edge were performed in transmission mode using a set of 3 ion chambers which allowed for concurrent measurement of a Pd foil energy reference. Measurements at the Pt L3 edge were performed in uorescence mode using a Lytle detector with a zinc foil lter used to reduce background uorescence.
At both edges, sample wafers were pressed in a stainless steel sample holder and measured at room temperature in air. Data analysis was performed using Artemis and Athena in the Demeter software suite 29 with phase and amplitude functions generated by FEFF using bulk references. At each edge, the amplitude reduction factor (S 0 2 ) was determined by tting the rst shell scattering of the respective metal foil. The obtained S 0 2 was then xed in catalyst samples. A spherical nanoparticle shape function was used in the modeling of Pd/Pt@CeO 2 at the Pd K edge 30 . Full model and t details are given in the supplemental information.

Catalytic reactivity measurements
CO oxidation was chosen as a probe reaction. The reaction rate measurements were performed using 1/4 in. diameter U-tube with 20 mg of sample. The gas ow rates for CO oxidation were the following: CO 1.5 mL/min, O 2 1 mL/min, and He 75 mL/min and temperature ramp rate was at 2 °C/min. The as-prepared, air-exposed catalyst was loaded into the reactor and temperature was increased to 300 °C under He. Once at 300 °C, a pretreatment with 10% O 2 was performed for 30 min. The gas was then switched to He and the catalyst was cooled to the reaction temperature. Subsequently, CO oxidation kinetic measurements were performed. The total pressure during CO oxidation was 83.3 kPa, the atmospheric pressure in Albuquerque. The products were analyzed by an Agilent Micro GC 490.
Catalytic methane (CH 4 ) combustion was performed using a xed bed ow reactor. 60 mg of catalyst powder was diluted by 600 mg SiC to guarantee isothermal conditions. The catalyst bed was packed between two quartz wool plugs inside a ½ inch (I.D.) quartz tube. The reaction was performed at atmospheric pressure. The reaction gas consisted of 680 ppm CH 4 , 14 vol % O 2 and 5 vol % CO 2 balanced with N 2 with a total ow of 300 mL/min. Different concentrations of steam (H 2 O) were introduced by owing the reaction gas through a temperature-controlled bubbler. The e uent gas was analyzed by an online MKS equipped with a detector. For light-off measurements, the samples were heated from room temperature to 650 o C with a ramp of 5 o C/min in reaction gas. After each light-off test, the system was cooled down in a O 2 /N 2 atmosphere. Stability under different dry/wet conditions was also evaluated by time-on-stream measurements.

Density functional theory calculations
All spin-polarized Density Functional Theory (DFT) calculations were performed using the Vienna Ab initio Simulation Package (VASP) 31,32 . The valence electrons were represented by plane waves with a cutoff energy of 400 eV, augmented with the projector-augmented wave (PAW) pseudopotentials for the core electrons 33 . A Hubbard-like term was added in the so-called DFT+U treatment with U = 4.5 eV applied to the Ce 4f states 34 . For bulk surfaces, the atoms in the top ve layers of PdO(101) and the top two layers of Pd(111) were fully relaxed. The reciprocal space was sampled with a k-point mesh of 3 × 3 × 1. For the Pd raft on the engineered Pt@CeO 2 support via atom trapping, we modeled the slab with 1 Pd(111) layer and 6 CeO 2 (111) layers, the slab was generated by a 2√2×2√2 unit CeO 2 (111) covered by a √13×√13 layer of Pd(111), and optimized distance between the Pd(111) layer and CeO 2 (111) was 2.13 Å.
The reciprocal space was sampled with a k-point mesh of 1 × 1 × 1. During geometry optimization, the atoms in the bottom three atomic layers were xed while others were fully relaxed. In these cases, the exchange-correlation potential was treated by the PW91 exchange-correlation functional 35 . A vacuum space larger than 15 Å was set between periodic slabs in order to avoid the arti cial interactions along the z-direction. Electronic energy was converged below a threshold of 10 -4 eV, while the force acting on each atom was less than 0.05 eV/Å. Climbing image nudged elastic band (CI-NEB) approach was used to determine the reaction pathway 36, 37 . 3. Results And Discussion 3.1. Engineering of the catalyst support via atom trapping In this work, we rst modi ed the ceria with atomically dispersed Pt to prepare the catalyst support (Fig.   1a). This involves deposition of a Pt precursor (1 wt% Pt) on the ceria support, then heating to 800 o C in air for 10 h. Since the Pt is present in atomically dispersed form, we will refer to this support as 1Pt@CeO 2 (Fig. 1b). Next we deposited additional 2 wt% Pt on this engineered ceria support. We found that the added Pt does not form atomically dispersed species as is typically seen on ceria supports 28,38 (see Fig. 1c-d and Supplementary Figure 1). It appears that the initial high temperature treatment has eliminated the sites on which atomically dispersed Pt would reside. Instead we see the formation of 2-D rafts as shown in  Table 1) con rm that this catalyst contains oxidized Pt but has no Pt-Pt neighbors, i.e. no Pt metal clusters. This catalyst (which is termed as 2Pt/1Pt@CeO 2 ) contains Pt rafts instead of 3-D metal or oxide clusters.
The CO oxidation reactivity for this 2Pt/1Pt@CeO 2 catalyst (containing a total of 3 wt% Pt) is greater than a 3Pt@CeO 2 prepared via atom trapping (Fig. 2). This is in agreement with a recent work showing that clusters of Pt oxide (for example Pt 8 O 14 containing Pt-O-Pt sites) are more active in CO oxidation 39 than single atom Pt. We should note that even higher CO oxidation reactivity can be achieved by reducing Pt/CeO 2 catalysts 38, 40 , but our focus here is on the oxidized state of the catalyst which is encountered during lean methane oxidation. Under the lean conditions (excess oxygen), the CO oxidation reactivity is reproducible over multiple runs. Low Energy Ion Scattering (LEIS, Fig. 2a and Supplementary Table 2), which is exquisitely surface sensitive, shows that the measured Pt concentration on the 2Pt/1Pt@CeO 2 catalyst (1.8 atoms/nm 2 ) exceeds the expected surface concentration (~ 1.2 atom/nm 2 ) calculated based on the Pt content and the BET surface area (Supplementary Table 2). In our previous work on atom-trapped Pt/CeO 2 we reported the upper limit of Pt surface concentration in single atom catalysts on this same ceria support was ~1 atom/nm 2 . 28 These results should be put in context with the surface concentration of Ce on the CeO 2 (111) facet which is 7.9 atoms/nm 2 41 and the Pt surface concentration in an oxidized Pt foil which is 9.0 atoms/nm 2 (Supplementary Table 2). Therefore, the observed surface concentration for 2-D Pt rafts of 1.8 atoms/nm 2 signi cantly exceeds that expected from atomically dispersed Pt and is consistent with the presence of Pt rafts on ceria (111) surface facets, with a preferred orientation due to prominent (111) facets leading to more prominent exposure of Pt as seen via AC-STEM ( Fig. 1c and d).
These results show that by starting with a catalyst support containing 1 wt% of strongly bound Pt single atoms, we cause the added 2 wt% Pt to form 2-D rafts. To investigate whether this 2-D raft morphology was simply a result of the high temperature treatment of the ceria, we prepared a catalyst support by heating the ceria support to 800 o C (which we term pre-sintered or PS ceria). As shown in Supplementary  Figure 3d) also leading to well de ned ceria (111) surface facets. When we deposited 3 wt% Pt on the PSceria and calcined the sample at 500 °C, we observed randomly oriented Pt in the form of 3-D nanoparticles (Supplementary Figure 3c). Therefore, the formation of 2-D rafts on 2Pt/1Pt@CeO 2 is very unusual and not simply a result of the high temperature treatment of the ceria support. Using the 1Pt@CeO 2 as a support modi es the nucleation of the deposited Pt forming 2-D rafts instead of 3-D nanoparticles which are seen on PS-ceria. The facile formation of Pt 2-D rafts and the superior reactivity of this catalyst compared to single atom Pt is consistent with the results from our DFT calculations ( Supplementary Figures 4 and 5). In summary, our results show that when the CeO 2 support is modi ed through the method of atom trapping, the strongly bound Pt single atoms act as templates, helping the Pt that is deposited on this support to form 2-D rafts, presenting superior CO reactivity compared to the Pt single atom catalyst on CeO 2 .

Methane oxidation on Pd deposited on the engineered catalyst support
Inspired by the above observations and results, we deposited Pd (1.09 wt%) on the engineered ceria support prepared by atom trapping of Pt as described in the previous section. To achieve a 1:1 Pd:Pt molar ratio, we used a catalyst support containing a Pt loading of 2 wt.% prepared via atom trapping (800°C in air for 10 h) using ceria prepared by decomposition of Ce(NO 3 ) 2 , which we refer to as polyhedral ceria (PHC). The performance of this catalyst (1Pd/2Pt@CeO 2 , Pd:Pt atomic ratio=1:1) was compared with other catalysts on the same ceria support, including the 2Pt@CeO 2 used to prepare 1Pd/2Pt@CeO 2 and 1Pd/CeO 2 prepared by impregnation. The reactivity was measured in a xed bed reactor using a gas mixture that resembles the e uent encountered during methane oxidation from vehicular exhaust (680 ppm CH 4 , 14 vol % O 2 , 5 vol % CO 2 balanced with N 2 with a total ow of 300 mL/min). For some of the experiments, we added water vapor to investigate the water tolerance of these catalysts. The light-off curves show that the 1Pd/2Pt@CeO 2 catalyst is more active than the oxidized 2Pt@CeO 2 and 1Pd/CeO 2 catalysts containing the same amount of the corresponding metal atoms under dry methane oxidation conditions (Fig. 3a). The 1Pd/2Pt@CeO 2 catalyst also shows improved low temperature reactivity compared to the reference (2Pt+Pd)/CeO 2 catalyst prepared using the same ceria support (2 times higher at 350 o C). The reference catalyst involved a conventional preparation via impregnation and calcination at 500 °C in air, but without any high temperature (800 °C) treatment.
To test the catalyst stability with & without added steam, methane oxidation was carried out at 500 o C and the conversion was kept between 90-100% by varying the space velocity ( Fig. 3b and c). The 1Pd/2Pt@CeO 2 catalyst did not suffer a signi cant loss of reactivity when switching from dry to steamcontaining methane feeds (4% loss in 5 v/v% water and 15% loss in 10 v/v% water, Fig. 3b) and the reactivity recovered after stopping water vapor ow. This shows an improvement in water tolerance for the Pd/2Pt@CeO 2 catalyst. Multiple runs of this catalyst also demonstrate the reproducibility of the data (Supplementary Figure 6a). The light off curves in the presence of varying amounts of steam (4% and 10%) are shown in the supporting information (Supplementary Figure 6b).
While methane oxidation catalysts operate under lean conditions (excess oxygen), we have seen in previous work that reduction of the catalyst forming metallic particles, or periodic pulses under reducing conditions, lead to enhanced reactivity [14][15][16] . In previous work, we have found that CO reduction at 275 °C is su cient to reduce single atom catalysts and achieve optimal CO oxidation reactivity 38 . Therefore, we tested the performance of these catalysts after CO reduction (Fig. 3c). We see de nite improvement in the light off curve for the bimetallic as well as the monometallic reduced 2Pt@CeO 2 and 1Pd/CeO 2 catalysts (Fig. 3c), but the bimetallic Pd/2Pt@CeO 2 catalyst performs better than monometallic Pt or Pd and is stable at high reaction temperatures (> 550 o C). At these high conversions, the methane is consumed so the catalyst is being subjected to an oxidizing atmosphere, which shows the improved thermal stability of this catalyst. While the Pd-only catalyst shows similar stable performance under oxidizing conditions (since it forms PdO, which is known to be active at high temperatures), the Pt only catalyst suffers a loss in reactivity. We attribute this loss of activity to the oxidation of the Pt nanoclusters and also possible re-dispersion to form Pt single atoms under oxidizing conditions at elevated temperature (Fig. 3c).
While the reduced Pd and Pt catalysts show improved reactivity after reduction (Fig. 3c), the performance of these catalysts cannot be sustained during methane oxidation and in the presence of water vapor (Fig.  3d). These catalysts suffer a major loss in reactivity due to oxidation of the metal and poisoning by water vapor. In contrast, the Pd/2Pt@CeO 2 catalysts shows only a slight drop in for the long term run at 500 o C (Fig. 3d). This drop is likely caused by the large amount of water molecules produced in the reaction at the high conversion of this test 42 . These results show that under the conditions of lean methane oxidation, the contribution from the starting single atom 2Pt@CeO 2 is negligible. Since the bimetallic catalyst Pd/2Pt@CeO 2 and the Pd/CeO 2 contain similar amounts of Pd, the enhanced performance of the former catalyst must be attributed to the morphology of the Pd/PdO phase present on the 2Pt@CeO 2 support, which is what we describe next.
To morphology of the Pd/2Pt@CeO 2 catalyst was studied via AC-STEM. Figure 4a shows that the engineered catalyst support 2Pt@CeO 2 contains atomically dispersed Pt species on the ceria support (circles in Fig. 4a), as also seen in both 1Pt@CeO 2 and 3Pt@CeO 2 reported in our previous study 24,28 .
After depositing Pd on the 2Pt@CeO 2 material, as shown in Fig. 4b, both single atoms and rafts with an average diameter of ca. 0.8 nm are present ( Fig. 4b and Supplementary Figure 7). We note that the single atoms seen in these images come from Pt and not Pd, because the contrast of single atom Pd on ceria is not su cient to make them visible in AC-STEM images (atomic number of Pd is lower than that of Ce, 46 vs 58). This is why we can only detect the clusters of Pd, and since the oxidation state cannot be inferred from this image, we will refer to them as Pd/PdO. The high magni cation STEM-EDS mapping (Fig. 4c-f) shows that both Pt and Pd are well dispersed on this catalyst and the homogeneous contrast in a domain indicates that the Pd/PdO domains are not 3D clusters, which would generally show a brighter center. A schematic view of the morphology of the Pd/PdO on 2Pt@CeO 2 is shown in the inset of Fig. 4b. The similarity in the image contrast of these Pd/PdO domains to the Pt clusters shown in Fig. 1 allows us to assign these as Pd/PdO 2-D rafts which are present on the atomically dispersed 2Pt@CeO 2 .
The environment of the Pt and Pd atoms in the Pd/PdO 2-D rafts of the 1Pd/2Pt@CeO 2 catalyst was further examined via XAS measurements performed on the calcined samples (Fig. 5). The XANES spectrum of Pt edge of 1Pd/2Pt@CeO 2 shows that the Pt is oxidized (Fig. 5a), which is similar to the 1Pt@CeO 2 catalyst reported previously 27 . Likewise, the XANES spectrum of the Pd K edge of the asprepared 1Pd/2Pt@CeO 2 resembles the bulk references PdO and Pd(OAc) 2 . The EXAFS ts for these samples are shown in the supplementary materials (Supplementary Tables 3 -8 and Supplementary  Figures 8 -13). The Pt L 3 edge EXAFS has one major peak due to Pt-O scattering, tting gave a coordination number of 6±0.4 at a bond distance of 1.99±0.008 Å. A Pt-O coordination number of 6 is consistent with platinum in the +4 oxidation state. Previous reports on atom trapped Pt on CeO 2 gave a Pt-O coordination number of 5, the extra oxygen coordination likely comes from oxygen in the Pd/PdO rafts as seen in the STEM results 28 . The interface between the raft structure and the support is not strongly ordered as evidenced by a weak broad doublet peak at 2.8-3.4 Å (phase uncorrected distance) in the Pt EXAFS. The EXAFS at the Pd K edge for 1Pd/2Pt@CeO 2 resembles that of bulk PdO with attenuated 2 nd and 3 rd shell Pd-Pd scattering. Two models were used to t the spectrum: a spherical PdO nanoparticle model, and a multi-shell PdO model without the shape function constraint. While both models gave equivalent reduced chi squared statistics, the R-factor for the multi-shell model was lower. This would be expected if the nanoparticle shape deviated from spherical. The mis t primarialy comes from under tting the 2 nd Pd-Pd coordination number. In a disk shaped particle, a spherical shape function will correctly model the data at distances lower than the shorter of the two characteristic lengths (radius and thickness) 43 . From the STEM results showing that the rafts are 1-2 atoms thick and based on the lattice parameter of PdO, the spherical model would start to fail past the rst Pd-Pd path at 3.02 Å, which is seen in the mis t of the second Pd-Pd path in the spherical model. Consistent with the Pt L3 edge EXAFS, no Pd-Ce scattering was observed, which suggests a con gurationally diverse interface between the raft and support.
We used surface-sensitive XPS to reveal differences in the Pd/PdO domains on the 1Pd/2Pt@CeO 2 compared to a reference (1Pd+2Pt)/CeO 2 sample prepared by conventional impregnation (Fig. 6). The reference sample (1Pd+2Pt)/CeO 2 sample shows signi cantly lower reactivity than the atom trapped 1Pd/2Pt@CeO 2 catalyst (Fig. 3a). The Pd 3d binding energy of Pd species on 1Pd/2Pt@CeO 2 is lower than that of the reference (1Pd+2Pt)/CeO 2 (Fig. 6b). This is suggestive of a different form of Pd oxide on the 1Pd/2Pt@CeO 2 surface, likely because it is present in the form of a raft, and different from 3D particles of bulk PdO present in the reference sample. The Pt 4f spectrum of 1Pd/2Pt@CeO 2 shows peaks at a higher binding energy in comparison to that of the reference (1Pd+2Pt)/CeO 2 (Fig. 6a), con rming that the Pt species strongly bound to ceria via atom trapping in 1Pd/2Pt@CeO 2 are different from Pt that is deposited by impregnation and calcined at 500 °C in air. The trapping of the Pt is also con rmed by CO-DRIFTS experiments wherein the 1Pd/2Pt@CeO 2 catalyst shows the intense CO band on ionic Pt seen in previous work 24 (Supplementary Figure 14). The Pd in the 2-D rafts is mobile and transforms readily from its oxidized state into a reduced state during CO oxidation at 125 °C (Supplementary Figure 14).
This mobility of Pd is consistent with previous studies of Pd single atom catalysts during CO oxidation 44,45 . What is unique to the Pd/PdO structures stabilized by the single atom Pt is the enhanced reactivity for methane oxidation and improved water tolerance.

Conclusions
In this study, we used atom-trapped Pt single atoms to modify and engineer the catalyst support to in uence the nature of the deposited metal oxide. In the case of Pt deposited on atom-trapped 1Pt@CeO 2 (2Pt/1Pt@CeO 2 ), we found that 2-D Pt rafts were formed, the support formed prominent ceria (111) surface facets and the preferred orientation caused a higher surface Pt signal in LEIS. The 2-D Pt rafts showed higher reactivity in CO oxidation because of a weaker interaction between the Pt and the engineered support. The catalyst with the strongest interaction, 3Pt@CeO 2 had the lowest reactivity for CO oxidation. Using a similar atom trapping approach, we prepared a 1Pd/2Pt@CeO 2 which showed much better reactivity than the conventional 1Pd/CeO 2 , 2Pt@CeO 2 and (1Pd+2Pt)/CeO 2 catalysts containing the same number of metal atoms when tested for methane oxidation. Low Energy Ion Scattering (LEIS) and CO oxidation reactivity of Pt 2D raft deposited on the engineered catalyst support. a, LEIS spectra of atomically dispersed 1Pt@CeO2 and 2Pt/1Pt@CeO2, the latter sample contains 2-D rafts of Pt as shown in Fig. 1. b, The CO oxidation reactivity con rms that the altered morphology of the Pt leads to enhanced CO oxidation reactivity. The 2-D Pt rafts (2Pt/1Pt@CeO2) are more active than the atom trapped catalyst 3Pt@CeO2. These two catalysts contain the same amount of Pt, the differences in reactivity can be attributed to different interaction with the support.  XAS spectra of 1Pd/2Pt@CeO2 sample. a,b, XANES spectra of Pt and Pd edges. c,d, EXAFS spectra of the Pt L and Pd K edges. The Pt and Pd standard samples were also measured.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupportinginformationforPtPdmanuscriptv4.pdf