Improved Selectivity and Stability in Methane Dry Reforming by Atomic Layer Deposition on Ni-CeO2–ZrO2/Al2O3 Catalysts

Ni can be used as a catalyst for dry reforming of methane (DRM), replacing more expensive and less abundant noble metal catalysts (Pt, Pd, and Rh) with little sacrifice in activity. Ni catalysts deactivate quickly under realistic DRM conditions. Rare earth oxides such as CeO2, or as CeO2–ZrO2–Al2O3 (CZA), are supports that improve both the activity and stability of Ni DRM systems due to their redox activity. However, redox-active supports can also enhance the undesired reverse water gas shift (RWGS) reaction, reducing the hydrogen selectivity. In this work, Ni on CZA was coated with an ultrathin Al2O3 overlayer using atomic layer deposition (ALD) to study the effects of the overlayer on catalyst activity, stability, and H2/CO ratio. A low-conversion screening method revealed improved DRM activity and lower coking rate upon the addition of the Al2O3 ALD overcoat, and improvements were subsequently confirmed in a high-conversion reactor at long times onstream. The overcoated samples gave an H2/CO ratio of ∼1 at high conversion, much greater than uncoated catalysts, and no evidence of deactivation. Characterization of used (but still active) catalysts using several techniques suggests that active Ni is in formal oxidation state >0, Ni–Ce–Al is most likely present as a mixed oxide at the surface, and a nominal thickness of 0.5 nm for the Al2O3 overcoat is optimal.


INTRODUCTION
Chemical transformation of CO 2 could reduce greenhouse gas emissions and lead to a more carbon-neutral petrochemical industry. 1,2Due to the high stability of the CO 2 molecule, chemical transformation is usually achieved through catalytic processes such as dry reforming of methane (DRM), hydrogenation (Sabatier reaction), formation of alkyl carbonates, etc. 3−5 In particular, DRM has the potential to transform the two major greenhouse gases, CH 4 and CO 2 , into syngas (H 2 , CO) at close to a 1:1 ratio, a mixture that would still require a water−gas shift to a higher ratio to render it usable for downstream processes such as Fischer−Tropsch. 6−9 However, at the characteristically harsh DRM conditions (T > 700 °C, P of 1−10 bar) Ni is highly susceptible to deactivation by both coking and sintering.−19 This semireducible aspect of ceria means oxygen vacancies (OVs) are formed.−22 Oxide−CeO 2 mixtures can further optimize the oxygen storage capacity (OSC) associated with such redox behavior.−28 When high concentrations of Zr are present, the fluorite structure converts to a tetragonal phase, reducing the OSC of the support. 29Due to high oxygen mobility in certain CZA mixtures, lattice oxygen can readily oxidize surface-deposited carbon. 30−35 CO 2 can adsorb and get reduced by Ce 3+ defect sites even without a nearby Ni active center, or a hydroxyl-terminated surface. 33owever, while the OVs present in CeO 2 and CZA improve DRM activity, they also enhance the activity of the RWGS reaction, thus lowering the H 2 /CO ratio. 22,36sing a shell layer to confine Ni during DRM could inhibit sintering at elevated temperatures, especially >700 °C.Atomic layer deposition (ALD) is a controllable self-limiting surface reaction method for the deposition of oxide layers.−44 Understanding the selective deposition process allows for the growth of thin layers on oxide surfaces without covering the catalytic metal sites.−48 Furthermore, Al 2 O 3 is a simple nonreducible oxide to deposit via ALD, and has been shown to limit hydrogen spillover and diffusion compared to more reducible oxides such as TiO 2 . 49The efficacy of thin shell layers, such as those produced by ALD or other core− shell techniques, is open to dispute, at least for DRM.While some core−shell catalysts (with Ni initially deposited on the core) have shown impressive turnover frequencies, such as a Ni/CeO 2 @SiO 2 catalyst (turnover frequency ∼0.39 s −1 on a total Ni atom basis), 50 it is not known if these are truly layered materials, or if oxide and/or Ni mixing into the shell is taking place.−53 There is debate about using alumina ALD overlayers due to possible formation of NiAl 2 O 4 , which is inactive for DRM. 51,54But Al 2 O 3 ALD overlayers have decreased Ni particle aggregation during DRM, 51,52,55 even though thick Al 2 O 3 -overcoated Ni−Al 2 O 3 catalysts still deactivated rapidly after reaching maximum activity, and selectivity (H 2 /CO ratio) was not explored. 56More recently, there have been conflicting reports on alumina-coated SiO 2 and CeO 2 , concluding that both thick (4 nm) 57 and thin (1 ALD cycle) 58 result in excellent catalytic performance, though long-term coking effects were not explored.This raises the question about the role of the overlayer with the support, metal, and adsorbate species.We hypothesize that ultrathin nonreducible Al 2 O 3 layers can enhance catalyst performance by limiting both Ni aggregation and adsorbate spillover due to selective deposition on the ceria surface.
In this work, we demonstrate how a nonreducible overlayer can drastically improve the selectivity and stability of a Ni-CZA catalyst.The addition of a 0.5 nm ALD Al 2 O 3 overlayer to a Ni-CZA core increased the lifetime of the uncoated catalyst and the measured coking rate by at least 10-fold.The Ni clusters of the catalyst used were also smaller and less reducible.This structure had the added benefit of decreasing, in absolute terms, the RWGS rate and increasing the DRM rate.O, Sigma-Aldrich, 99%), trimethylaluminum (Al(CH 4 ) 3 , Sigma-Aldrich, 97%), titanium(IV) isopropoxide (Ti[OCH(CH 3 ) 2 ] 4 , Sigma-Aldrich, 99%), and urea (CH 4 N 2 O, VWR, ACS grade) were used as received.The reactant gases were CH 4 (Airgas, 99%) and CO 2 (Airgas, 99%).Air (Airgas, breathing grade), N 2 (Airgas, UHP), and 5% H 2 (Airgas, certified) were also used for calcinations/reductions.

Materials.
2.2.Catalyst Syntheses.Adsorptive deposition (aka "strong electrostatic adsorption") was used to deposit 4 wt % Ni onto CZA40, as adapted from previous work. 13,59Powdered CZA40 was added along with the desired amount of Ni(NO 3 ) 2 •6H 2 O and 30 mL of 0.3 M urea per g support.The solution was stirred and reacted for 24 h under reflux at 90 °C.The product powder was washed with deionized (DI) water and dried at 100 °C overnight, then reduced as discussed subsequently.
The Al 2 O 3 -ALD method was adapted from previous work. 60he AlO x overcoats were deposited using a benchtop viscousflow ALD reactor (GEMStar-6 XT) at 150 °C, alternating exposure of trimethylaluminum (TMA) and water vapor using N 2 as both carrier and purge gas, with both TMA and water bubblers at room temperature.Each AlO x cycle consisted of 90 s of TMA exposure followed by 90 s of water exposure with 300 s of N 2 purge between each exposure (90s-300s-90s-300s).Three to 12 cycles of AlO x ALD were performed on Ni/ CZA40.The mass gain was obtained by before-and after-ALD weighing.Samples are referred to by the layer thickness increase upon ALD, as 0.3, 0.5, or 1.2 nm ALD-CZA40.
After calcination in air at 600 °C for 1 h, and reduction in 4% H 2 /Ar at 500 °C, 10% CO 2 /Ar was pulsed over the catalysts at various temperatures.The conformality of the coatings was examined by the adsorption of CO 2 (Nicolet Nexus 670 FTIR, DRIFTS mode, MCT/A detector, 4 cm −1 resolution, and 10 −6 Torr).

X-ray Diffraction (XRD).
Powder X-ray diffraction (XRD) was performed to analyze the bulk structure of the assynthesized and used catalyst samples.Diffractograms were collected using a PANalytical XRD at 45 kV and 40 mA.Spectra were recorded at 0.04°steps over the range of 5−70°, with a dwell time of 60 s.A Cu Kα radiation source was used.The average crystal size of an identified phase was calculated using the Scherrer equation = ( ) where K is the dimensionless shape factor, typically set at 0.9, λ is the X-ray wavelength for Cu Kα radiation, and β is the full width at halfmaximum of the XRD peak analyzed.
2.4.Porosimetry.The surface area, total pore volume, and pore size distributions of the fresh and used catalysts were measured by N 2 adsorption by using a Micromeritics ASAP 2020 Plus porosimeter.Samples were dried at 300 °C and degassed prior to analysis.The surface area was determined by the Brunauer−Emmett−Teller (BET) method and the pore size distribution by the BJH method.

High-Resolution Transmission Electron Microscopy (HRTEM) and Energy-Dispersive X-ray (EDX)
Spectroscopy.The morphology and size of the catalysts were analyzed by HRTEM using a 200 kV JEOL NEARM electron microscope at Oak Ridge National Laboratory equipped with double aberration correctors, a dual-energyloss spectrometer, and a cold FEG source.Before imaging, the samples were dispersed in ethanol and drop cast on a 300 mesh lacey carbon grid.EDX was performed using an FEI Quanta 3D FIB microscope equipped with an EDAX Apollo XL detector operating at an accelerating voltage of 20 kV and a current of 4 nA.Due to the strength of the TEM electron beam at high magnification, it proved impossible to obtain EDX of high magnification images due to sample deformation.Resolution was kept at a minimum of 100 nm.ImageJ (version 1.53k) was used to analyze lattice spacings of imaged catalysts.
2.6.XPS and XAS.X-ray photoelectron spectroscopy (XPS) was performed using a Scienta Omicron ESCA 2SR equipped with a monochromatic Al K α (hν = 1486.6eV) X-ray source and a hemispherical analyzer with a 128-channel detector, at 1.3 × 10 −9 Torr.The Gaussian width of the photon source was 0.5 eV with a focus voltage of 300 V.The adventitious carbon C 1s peak at 284.4 V was used to calibrate the energies.After Shirley background subtraction, all peaks were fitted by using Casa XPS (version 2.3.25) as Gaussians.
X-ray absorption spectroscopy (XAS) was performed at the Ni K-edge at the LSU Center for Advanced Microstructures and Devices (CAMD).Some spectra were collected at the HEXAS beamline using a Ge 220 double crystal monochromator, at room temperature in fluorescence mode, with a Ni foil calibration standard.Other spectra were collected at the WDCM 2.0 beamline equipped with a Si 111 channel-cut monochromator in fluorescence mode and calibrated with a Ni foil standard.Integration time was adjusted to obtain adequate counts up to a wavenumber of 12. Runs were repeated to improve the counting statistics.Both X-ray absorption nearedge (XANES) and X-ray absorption fine structure (XAFS) spectra were collected.
Background subtraction, deglitching, and merging of spectra were performed using Athena 0.9.061.Ni K-edge XAFS fitting was performed in Artemis 0.9.26.Four parameters were varied to obtain the best possible fits, S 0 2 (amplitude reduction factor), σ 2 (Debye−Waller factor), ΔE 0 (deviation in E 0 caused by structural deviations from the ideal crystal structure), and ΔR (deviation in interatomic distance).A NiO standard was fitted first to get information about S 0 2 with known coordination numbers and ΔR′s.The fitting range in R space was 1−5 Å, and all significant scattering paths were included.
2.7.Catalytic Activity Measurement.Catalytic activity was measured at differential conversions using a TA SDT Q600 differential scanning calorimeter (DSC)/Thermogravimetric Analyzer (TGA).ALD-coated catalysts were first pretreated in air (100 mL/min) at 600 °C for 1 h to remove residual water and generate pore space in the overlayer.Then the samples were reduced in 5% H 2 /95% N 2 for 3 h or as long as necessary at 600 °C.This was sufficient to fully reduce the catalysts prior to the DRM experiments, as shown in Figure S1.We varied the reduction temperature from 500−750 °C in 50 °C increments for Ni-CZA40 and 600 °C reduction gave the highest DRM activity, although over the 550−650 °C range little difference was observed.
DRM (135 mL/min total flow, 0.25 bar partial pressure CH 4 and CO 2 , 0.5 bar N 2 ) took place at 650 and 750 °C for 1.5 h at each temperature.The heat flux and change in mass were both measured.The heat flux is roughly the heat evolved by the DRM reaction, and the weight change can be related to the coking rate. 13Heat flow data were used in an Aspen HYSYS program 61 to calculate the DRM rate and methane conversion.Further details are given in the Supporting Information.
Catalysts showing promising activity at differential conditions were further tested in a fixed-bed reactor, which was a 12.5 mm quartz reactor tube with α-alumina and quartz wool as an extra packing material.Catalysts were reduced for 6 h at 600 °C in flowing 5%H 2 /95% N 2 , with a ramp rate of 10 °C/ min.Catalyst weight was varied between 0.35 and 0.25 g to vary GHSV.The feed composition was 1:1 CH 4 /CO 2 (molar) at ∼0.6 bar of each reactant.The setup here was the same as in previous work. 13An Agilent 6890N GC-MS instrument was used to analyze the outlet gas composition.The reactor tube is heated by a furnace (Teco F-5-1000, 320 W) whose temperature is controlled by a Eurotherm 818P PID controller.Figure S2 shows a schematic of the reactor setup.Information about product analysis is also in the Supporting Information (Table S1).Conversion, yield, and selectivity from these experiments were calculated from the extent of reaction.Three equations represent the main reactions (DRM, reverse water− gas shift, and coking): (1) (2) The ξ′s are molar extents of reactions in mol/min, calculated by solving the component mass balances simultaneously, using both the compositions and effluent flow rate.These results were refined using a nonlinear regression method where the objective function is the sum of the squared residuals of the CH 4 , CO 2 , H 2 O, and CO and H 2 mass balances.The terms F in,CHd 4 , and F in,COd 2 are molar flow rates of the feed components in mol/min.The yields of products on an elemental carbon basis and the conversions of CH 4 and CO 2 , and the turnover frequency (TOF), defined as the amount of CH 4 reacted per second per total Ni atoms, are calculated as follows: (5) where M Ni is the molecular weight of Ni.
2.8.Density Functional Theory (DFT) Calculations.2.8.1.Electronic Structure Methods.DFT calculations were performed using the Vienna ab initio Simulation Package (VASP). 62The electronic exchange and correlation interactions were described by the generalized gradient approximation (GGA) method with the Perdew−Burke−Ernzerhof (PBE) functional. 63The projector augmented wave method was used to represent the core electrons, and a plane wave basis set was used to represent the valence electrons 64 with an energy cutoff of 500 eV.The Monkhorst−Pack scheme 65 was used to sample the Brillouin zone, with the third vector perpendicular to the surface.Electronic SCF cycles converged with an energy difference of less than 1 × 10 −5 eV.Structural optimizations minimized forces on all atoms below 0.05 eV/Å.Spin-polarized calculations were used for systems with unpaired electrons.Slab-to-slab dipole interactions were corrected within each SCF while simulating surfaces.All isolated gas molecules were optimized with a 1 × 1 × 1 k-point grid, in a 10 × 10 × 10 Å unit cell.
DFT+U corrections were used for structures with Ce, due to the well-established difficulties that DFT faces while representing the 4f orbitals of Ce.A U value of 5 was used on the f orbitals of Ce which is consistent with our previous work on ceria-based systems. 13,66,67DFT is also known to have difficulties representing the localized d-states of transition metals; thus, U corrections were also applied for systems containing Ni.A U-value of 6.4 eV 68 was used on the d orbitals of Ni because of closer agreement between the experimental 69 and simulated band gap (4.0 vs 3.89 eV).
2.8.2.Bulk and Surface DFT Models.The bulk γ-Al 2 O 3 structure was obtained from Digne's model 70 as it has found extensive applications in DFT-based studies.The two important benefits of using this model for first-principles simulations are the availability of a smaller unit cell and the presence of fully occupied lattice sites, which enable the creation of supercell surface models. 71CeAlO 3 has two stable lowest energy structures.We used the rhombohedral perovskite-type structure, as it is similar to the NdAlO 3 structure. 72he bulk CeO 2 structure used in this study is the commonly used cubic fluorite-type structure.Geometrically optimized pure CeAlO 3 bulk and γ-Al 2 O 3 unit cells are shown in Figure S4.
To simulate surfaces of the pure oxides (γ-Al 2 O 3 and CeO 2 ), slab models were created from the respective bulk oxide unit cells with an appropriate number of layers and a sufficient vacuum space of 15 Å in the direction perpendicular to the surface.The bottom-half layers of the surfaces were frozen during the structural optimizations to represent the bulk phase.
The details of the bulk and surface structures along with their k-point grids are given in Tables S2 and S3, respectively.

Reactions and Characterization of an Uncoated CZA40
Catalyst.Uncoated CZA40 (5 wt % Ni) was the reference material for this study.Initial characterization of fresh CZA40 was performed by XRD and N 2 physisorption.The diffraction peaks (Figure 1) of the CZA40 powders were indexed to cubic CZA (ICSD 157416), 73 with no observable peaks corresponding to Ni or NiO, indicating high dispersion of Ni in the fresh catalyst.The CeO 2 crystallite size was calculated to be 5.5 nm for CZA40 based on the (111)  reflection.The N 2 physisorption (Table S4) gave a surface area of 95 m 2 /g by the BET method with a total pore volume of 0.64 cm 3 /g.Note the high surface areas maintained in the used catalysts.
CZA40 was tested in the fixed-bed reactor system at 750 °C and a gas hourly space velocity (GHSV) of 37 000 mL/(h• gcat).The catalyst exhibited initial conversions of 55% for CO 2 and 44% for CH 4 with an H 2 /CO ratio of 0.75, as shown in Figure 1.But at 16 h time onstream, an increase in reactor pressure to 1.7 bar was seen, indicating the presence of large quantities of carbon (coke) beginning to block the reactor.The final conversions were 52% for CO 2 and 39% for CH 4 , with an H 2 /CO ratio of 0.69, but upon inspection, the reactor was nearly blocked.The low H 2 /CO ratio indicates a significant RWGS rate.
XRD analysis of CZA40 (Figure 1) revealed the emergence of two additional crystalline phases.The peak at 26.5°is indicative of semigraphitic carbon formation, 13 while the peaks at 44.7 and 51.8°suggest large Ni 0 aggregates. 13The calculated CeO 2 crystallite size was 6.6 nm based on the (111) reflection; the increase in CeO 2 crystal size reflects support sintering.The presence of both Ni 0 and carbon phases confirms that the catalyst was being structurally altered, and while these changes did not deactivate the catalyst at first, they were sufficient to reduce its selectivity and lead to a situation in which the process became untenable.

Reactions on ALD-Coated CZA40
Catalysts.The TGA/DSC screening method at low CH 4 /CO 2 conversions was used to determine short-term effects on activity/stability; this method was adopted unchanged from previous work. 3,13esults for all catalysts are listed in Table 1.The neglect of the effects of RWGS on the DSC heat flux is justified both by its relative thermoneutrality, and the knowledge that RWGS rates are always less than DRM rates (by an order of magnitude at low conversion, an assertion tested later). 13For example, the calculated endothermic heat of reaction for DRM is 7.2−7.5 times that of RWGS over 650−800 °C. 61The coking rate was determined by weight change over 1.5 h, observed after the first 0.5 h of reaction.The addition of the Al 2 O 3 overlayer at a nominal 0.5 nm thickness did increase the reforming rate at 750 °C, but little to no benefit was seen at 650 °C.ALD of TiO 2 significantly decreased the activity of CZA40 and was not investigated further.
Prereducing 0.5 nm ALD-CZA40 (550 °C, 5% H 2 /N 2 , 3 h) prior to ALD was also investigated.These samples are shown as "PR" in Table 1, and the same pretreatments just prior to TGA were used.Prereduction prior to ALD greatly increased the DRM rate for 0.5 nm ALD-CZA40 at 750 °C, with a negligible coking rate at 750 °C compared to a nonpre-reduced ALD sample.Therefore, all other ALD-coated samples were also prereduced before ALD.
The DRM activity at 750 °C was comparable to uncoated CZA40 for the 0.3 nm sample, while the 1.2 nm sample had a lower activity.Both showed negligible coking rates at 650 and 750 °C.All three ALD-coated catalysts were further tested at high conversion in a bench-scale reactor system, as described below.
The 0.5 nm ALD-CZA40 was first tested at GHSV 10 900 mL/(h•gcat) at 750 °C (Figure 2).The sample was activated for more than 50 h before reaching its maximum activity and H 2 /CO ratio.After this time onstream, it exhibited ∼75% CH 4 conversion with an H 2 /CO ratio of 0.98.This H 2 /CO ratio is near the equilibrium H 2 /CO ratio calculated using an Aspen HYSYS simulation of the process, including only the DRM and RWGS reactions.The run was continued to near 140 h time onstream, with no deactivation seen.The GHSV was then increased to 37 000 mL/(h•gcat) and the reactor was run for another 4 h (Figure S5).While the CH 4 conversion and H 2 / CO ratio initially showed a slight decrease, the catalyst eventually attained nearly identical conversion and selectivity as at 10 900 GHSV.Comparing these results to uncoated CZA40 in Figure 1, the 0.5 nm ALD overcoat enhanced DRM activity and selectivity, increasing CH 4 conversion from 40 to 77%, and H 2 /CO ratio from 0.7 to 0.98.
The total amount of coke deposited during DRM was estimated by temperature-programmed oxidation with averaged coking rates shown in Table 1.For the coated samples these can be an order of magnitude lower than the uncoated CZA40 sample.The TPO plots are in the Supporting Information (Figure S6).
The 0.3 nm ALD-CZA40 catalyst was also tested in the reactor system at 37 000 GHSV (Figure 2b).The catalyst's induction period was ∼24 h, and it ran for 110 h with no deactivation.The final CH 4 conversion was 74% at 0.94 H 2 /   CO ratio, values only slightly below that of 0.5 nm ALD-CZA40.
3.3.Effect of ALD Overlayers on RWGS.The effects of Al 2 O 3 ALD overlayers on RWGS were studied using a TGA/ DSC method similar to the DRM low-conversion screening method of Table 1, at partial pressures (bar) 0.26 H 2 , 0.26 CO 2 , and 0.48 N 2 .The product of the H 2 and CO 2 partial pressures in this experiment is near the maximum possible calculated for the bench-scale reactor (0.068 bar 2 here vs 0.09 bar 2 for the bench-scale reactor at 33% fractional conversion), so this is a near-worst-case test for the undesired RWGS reaction.The heat flow values measured by DSC are highly characteristic of the RWGS reaction.Aspen HYSYS was used to simulate the equilibrium conversions and heat flows of RWGS at high CO 2 and H 2 partial pressures, along with its main competing reaction, CO 2 hydrogenation (Sabatier reaction).At 650 °C RWGS contributed 95% of the total heat flow and at 750 °C 99.5% of the total heat flow.Thus, it can be assumed that CO 2 hydrogenation to CH 4 is negligible in these TGA/DSC experiments.Using the heat flow measurements and Aspen HYSYS, we calculated the H 2 conversions for the TGA/DSC experiments were calculated.These H 2 conversions were used to compute rate constants from a plug-flow reactor mass balance for CO 2 , for an assumed second-order RWGS reaction: where Da is the Damkohler number, a ratio of reaction rate to convective mass transport rate, ε the volumetric expansion factor (here, 0), F′ the total volumetric flow rate, and X is the fraction conversion.The addition of the 0.5 nm ALD overlayer to CZA40 decreased the RWGS rate constants (Table 2) and increased the DRM rate constants that were computed from the differential rates of Table 1, assuming that the reaction is first-order in both CO 2 and CH 4 .The present results showing that the addition of the Al 2 O 3 overlayer to CZA40 decreased the RWGS rate constant are consistent with the H 2 /CO ratios near one that were observed in the high conversion reactor tests.
3.4.Characterization of ALD-Coated CZA40.Both fresh and used catalyst characterizations were preferentially performed on 0.5 nm ALD-CZA because it showed the highest activity.
3.4.1.TPR, XRD, N 2 , and CO 2 Adsorption.The CO 2 DRIFTS experiments were performed on fresh CZA40 and 0.5 nm ALD-CZA40 catalysts to explore the integrity of the overlayer (Figure S7).As expected, the fresh CZA40 showed carbonate peaks and CO/CO 2 masses (by MS) during desorption up to 400 °C.After the addition of the 0.5 nm shell and calcination, the carbonate signature was no longer detected.However, low concentrations of CO/CO 2 could still be detected in the MS data, indicating minimal adsorption of CO 2 on the coated surface.
Figure S8 shows the XRDs of fresh ALD-CZA40 catalysts and Figure 3 shows the XRDs of 0.5 nm ALD-CZA40 only.All three fresh catalysts show the characteristic peaks for cubic CZA (ICSD 157416). 73The peak at 44.4°for the 0.3 and 0.5 nm ALD-CZA40 probably arises from γ-Al 2 O 3 only, because it does not disappear or change in intensity upon use (Figure 3).The XRD of used 0.5 nm ALD-CZA40 showed no peaks around 26.5°2θ, indicating a lack of semigraphitic or graphitic carbon on the used catalyst.

TPR Was Used to Verify the Gradual Structural Change Occurring in the ALD-Prepared Catalysts during
Use. Figure 4 compares the TPR analysis of used ALD-coated   samples with the fresh CZA40 (the used sample was too coked to obtain meaningful TPR).There are four distinct peaks or shoulders.Peak D is attributed to bulk CeO 2 reduction in CZA, 74−77  °C. 93,94s the layer thickness increases, the area of peak C decreases and it appears only as a shoulder on B. The peak maximum shifts to a lower temperature, ∼480 °C for the 1.2 nm and 505 °C for the 0.5 nm ALD-CZA40.This shows that a different surface structure has been formed, which is neither entirely Al 2 O 3 nor CZA.There are no high-temperature peaks in the ALD catalysts that would indicate the reduction of NiAl 2 O 4 , 91,94 proving it is not formed during the reaction.It is notable that the 0.5 nm catalyst is the only one whose reduction is concentrated in a single broad peak.
The N 2 adsorptions (Table S4) showed a decrease in surface area and pore volume for the ALD-coated samples compared to those of CZA40, at all coating thicknesses.The average pore diameters also decreased by more than 10%.These results are indicative of the ALD coatings penetrating pores, coating some and blocking some others, as has been previously observed for ALD-coated catalysts. 51,53.4.2.Ce XPS, Ni K-Edge XAS.XPS was used to explore the oxidation state of Ce, Ce 3d spectra showing multiple bands from 880−920 eV due to the O 2p valence band − Ce 4f hybridization.The deconvolution of Ce 3d spectra results in ten bands, six of which correspond to a Ce 4+ state: v (884.8 eV), v″ (891.07 eV), v‴ (898.4 eV), u (903.2 eV), u″ (909.55 eV), u‴ (918.44), while the remaining four correspond to a Ce 3+ state: v 0 (881.4eV), v′ (887.71eV), u 0 (900.32 eV), u′ (906.27eV). 3,59,95The fraction of surface Ce 3+ was estimated as where A Ce 3+ and A Ce 4+ are the sums of the areas of the Ce 3+ and Ce 4+ peaks listed above. 96The spectra and deconvolutions applied are shown in Figure S9.The spectra of CZA40 and the fresh 0.5 nm ALD-CZA40 were similar, both with the percentage of surface Ce 3+ calculated as 35% (of total surface Ce).The used 0.5 nm ALD-CZA40 showed a reduction in the 884.7 (v) band intensity, while increasing intensities are seen at 900.8 (u 0 ) and 905.95 (u′).The surface Ce 3+ for used 0.5 nm ALD-CZA40 is calculated as 71%, a factor of 2 increase over the fresh catalyst.Ce-based catalysts used under reducing conditions typically exhibit an increase in Ce 3+ . 97,98The amount of Ce 3+ corresponds to the number of OV. 95,96,99 A high concentration of surface OV is also associated with better coke resistance in Ni/CeO 2 DRM catalysts. 3,21,34,100This is consistent with the DRM test results for both reactors.The Ce 3+ percentages should be taken as relative only, due to the potential for XPS irradiation to create additional Ce reduction, but the exact magnitude of this machine-dependent reduction is unknown. 101 1s was used to observe changes in the coordination environment of oxygen upon DRM (Figure S10).−104 There is a decrease in surface lattice oxygen from fresh to used 0.5 nm ALD-CZA40.This is consistent with the Ce 3d XPS results above showing an increase in undercoordinated surface Ce associated with Ce 4+ reduction to Ce 3+ .
The XPS spectra were also used to estimate the surface and near-surface compositions of the ALD-CZA40 samples (Table 3).The coated samples are dominated by Al, suggesting the coatings are conformal.But there is some Ce and Zr, and these atomic %'s increase both as the coating gets thinner and as the catalyst is used in DRM.For the fresh catalysts, the Ce/Zr ratio is 1/2.But after 140 h of DRM the 0.5 nm ALD-CZA40 shows enrichment of (primarily) Ce on the surface, a 2.3% increase in Ce at %.The Ce/Zr ratio increased to 3/4 in the used sample, showing that Zr does not exhibit the same mobility during DRM as Ce.
XANES was used to determine the oxidation state of Ni in the same three catalysts as used for XPS.These Ni K-edge XANES are shown in Figures 5a, S11 and S12.The used 0.5 and 0.3 nm ALD-CZA40 catalysts clearly experienced reduction during DRM (Figures 5a and S12) and their XANES spectra are characteristic of a mixture of Ni 0 /Ni 2+ .XANES spectra of the used 0.3 and 1.2 nm ALD-CZA40 (Figure S11) show a greater degree of reduction than 0.5 nm to Ni 0 .Linear combination fitting on all three catalysts based on Ni foil and NiO standards gave a 65% Ni 0 estimate for 0.5 nm ALD-CZA40 but a 90% estimate for the other two.Therefore, some Ni present in the ALD samples exists in the Ni 2+ state, but there are subtle differences between these XANES spectra and that of a NiO standard.In comparison, fresh CZA40 shows some Ni 0 , but it is mostly Ni 2+ as evident from the white line position at 8351 eV.XAFS analyses were performed to further explore the coordination environment around the Ni.While XAFS fitting of Ni-CZA systems is documented, 59,105 59 Of all the possible phases listed above, only NiO itself (Figure S14) and Ni 2+ substituted into a tetrahedral γ-Al 2 O 3 site (Figure 5b) gave reasonable fits for the fresh catalyst (the fits are given in Table S5, and the poorer fits for Ni/CeO 2 and NiAl 2 O 4 are shown in Figures S13 and S15).This conclusion was based on both visual observation and the relative standard deviations of the FT-XAFS functions (the R-values).The fit to NiO is not exact, but that is expected due to the complex nature of the coated samples.It is highly unlikely that the Ni is in bulk NiO, which would be inconsistent with the XRD results and the known reduction chemistry of NiO, but the XAFS fits do suggest a coordination environment characterized more by adjacent Ni and O atoms in the fresh sample, consistent with the XANES.
Figure 5b shows the fit for a single Ni 2+ atom substituted into a tetrahedral γ-Al 2 O 3 site for fresh 0.5 nm ALD-CZA40.The peak at 1.5 Å arises from first-shell O atom scattering.The nature of the peak at 2.5 Å cannot be so easily determined due to potential scattering from Ni, Al or Ce atoms, all of which could be located near 2.5 Å radial distance.To analyze the  XAFS was also used to estimate Ni cluster sizes in the used catalysts.XAFS analysis of used 0.5 nm ALD-CZA40 is shown in Figure 5c.The largest peak can be assigned to the Ni−Ni scattering of Ni 0 .However, this peak is broader when compared to samples with less Ni 2+ present, such as the used 0.3 and 1.2 nm ALD-CZA40 (Figures S16a,b).This suggests that additional scattering paths give rise to this peak.As previously mentioned, a weighted Ni 0 /Ni 2+ fit was applied by weighting the S 0 2 according to the mole fraction of Ni 0 and NiO.The mole fraction was a regression parameter.The coordination number for the Ni−Ni scattering path of this fit is much less than in either NiO (N = 6 for O, 12 for Ni) or Ni foil (N = 12), suggesting that Ni here exists as dispersed clusters, where much of it is coordinatively unsaturated (Table S6).This type of behavior has been observed previously, with the extent of coordinative unsaturation suggesting clusters of average size less than 3 nm. 59,105,110

.4.3. High-Angle Annular Dark Field Scanning Transmission Electron Microscopy (HAADF-STEM) and EDX.
These experiments confirmed several of the conclusions from other characterization results.Dark-field STEM and EDX images of fresh and used 0.5 nm ALD-CZA40 samples were obtained.In Figures 6a,b   ) E j is the DFT energy for species j within the reaction: where m and n are stoichiometric coefficients.Structural DFT energies of the doped oxides were minimized with respect to the volume of the unit cell, to account for the lattice distortion postdoping (details in the Supporting Information).The Ce concentration (cation %) in each bulk oxide is calculated according to eq 12. Cerium aluminate (CeAlO 3 ) is a stable bulk mixed oxide relative to its parent oxides.−114 A few nonstoichiometric cation substitutions into this mixed-oxide phase have favorable formation energies (Ce (0.44−0.53)Al (0.47−0.56) O 3 ).However, these mixed-oxide phases are not stable relative to phase separation as their formation energies do not fall on the convex hull.Together, these results indicate that the formation of mixed Ce−Al bulk oxides is unfavorable, with the exception of the stoichiometric CeAlO 3 phase, for which there is no experimental evidence.110) and ( 100) were examined (refer to Figure S22).Single/ multiple Al sites that are highlighted with yellow circles in Figure S22c and d were substituted with Ce.The oxidation state of doped Ce was varied from 3+ to 4+.Doping energy, in eV, is used to assess the thermodynamic stabilities of Ce-doped γ-Al 2 O 3 surfaces (reaction equations in Tables S7 and S8).
Figure 8 shows the doping energy as a function of the Ce 3+ concentration for γ-Al 2 O 3 ( 100) and (110).Ce doping on the (100) facet becomes highly unfavorable with increasing Ce concentration, with the lowest doping energy at ∼3% Ce concentration (0.04 eV) being positive.Doping induces surface reconstruction, which intensifies with increasing Ce concentration.This can be observed in Figure 9, where the doped Ce atoms tend to detach from the surface by breaking their bonds with near-surface O atoms (Figure 9b,c,d).However, Ce doping on the (110) facet is relatively favorable, with the doping energies being negative for a few structures, even at higher Ce concentrations.Minimal surface reconstruction is observed postdoping, as showcased in Figure 10.Owing to its corrugated surface, with ∼45% higher surface area than the (100) facet, the (110) facet is more favorable toward Ce doping.

Stability of Doped vs Adsorbed Ni-Ceria Surfaces.
The most stable (111) facet of CeO 2 (Figure S23) was chosen for this study.NiO was adsorbed on the CeO 2 surface ("adsorbed"), or a Ni atom was substituted in place of a surface Ce along with the creation of an oxygen vacancy ("doped").The doping and adsorption energies were calculated as the reaction energies given in eqs 13 and 14, respectively.A negative value of ΔΔE (in eV/Ni) = ΔE doping − ΔE adsorbed for the doped configuration (Table 4) implies that Ni prefers to be doped rather than adsorbed.Despite an equivalent average Ni−O bond distance (∼2.01 Å) in both configurations, noteworthy distinctions in the coordination environment of Ni are observed.Specifically, in the doped structure, Ni achieves relatively symmetrical coordination characterized by each Ni− O distance being around 2.00 Å.In contrast, the adsorbed structure manifests a broader range of Ni−O distances, spanning from 1.96 to 2.10 Å, indicative of a less symmetrical coordination (Figure S24).

Oxygen Vacancy (OV) Formation Energetics of Ni-Doped and Bare Ceria Surfaces. Given the observation that
Ni prefers to dope within rather than adsorb atop the ceria surface, the reducibility of Ni-doped CeO 2 surfaces was evaluated by comparing the OV formation energies (ΔE ovf ) of the doped and parent surfaces (eq 15).In both NiCe 15 O 31 and Ni 2 Ce 14 O 30 , OVs were created by removing a Ni-bonded O atom, as this yielded the most stable structures.After the creation of the OV, Ni stays in a formal Ni 2+ state, with two Ce 4+ reducing to Ce 3+ in both structures (formal charges are assigned based on the spin densities of individual atoms).
Table 5 shows ΔΔE ovf values for singly and doubly Nidoped CeO 2 surfaces.The doping of (111) CeO 2 with a single Ni 2+ atom results in a positive ΔΔE ovf , taken relative to OV formation at the same concentration in the undoped (111) CeO 2 surface.But doping of two Ni atoms into the CeO 2 −2.07 −1.67 eV/Ni (111) surface promotes OV formation relative to the undoped surface (negative ΔΔE ovf ).11).Only the DFT energy of the most stable structure under each category was employed to calculate ΔE formation and ΔΔE formation (Table 6).The ΔE formation per Ni of the doped structures was calculated with reference to bulk γ-Al 2 O 3 and NiO, as shown in eq 19.Here, ΔΔE formation the preference of sites for Ni doping relative to doping in tetrahedral−tetrahedral sites (eq 20).12 and the results in Table 7 show that the doped structure is considerably more stable.2. The k DRM for the fresh ALD-CZA40 catalysts can be used in eq 9 with other parameters charcteristics of DRM in the high conversion reactor (W = 250 mg, F′ = 150 mL/min, C A0 = 2.03 × 10 −5 mol/mL) to predict an upper bound on conversion.It is an upper bound because the reverse rate constant and product/reactant inhibition would have to be considered to determine a more realistic conversion.However, the expected CH 4 conversion in the larger reactor using k DRM from Table 2 for 0.5 nm ALD-CZA40 was only 53%, compared to an actual CH 4 conversion of 75%.This significant difference in expected vs actual conversion demonstrates the magnitude of the structural change of the ALD-coated catalyst, which is evolving to a far more active state than what was initially put into the reactor.This is evident as well from the long induction period (Figure 2).Induction periods in Al 2 O 3 ALD-coated catalysts have been observed before and were also associated with structural evolution. 51,53he analysis of the rate constants of DRM and RWGS is also useful for understanding why the ALD-coated catalysts are more selective.With uncoated CZA40, the ratio k DRM /k RWGS at 750 °C was slightly less than 1 while at 650 °C it was well below 1.For fresh 0.5 nm ALD-CZA40 the ratios at 750 and 650 °C were both greater than 1, with the ratio increasing with respect to temperature, as is generally true for good DRM catalysts.Both of these rate constants were obtained at conditions of equal partial pressures of the respective reactants for RWGS and DRM.For RWGS an equal ratio of partial pressures H 2 :CO 2 is an oversimplification of the nature of these two reactants in a reactor, where the H 2 and CO 2 concentrations vary greatly with reactor length.At the entrance of the reactor, there is a lot of CO 2 but no H 2 .Eventually, both H 2 and CO 2 are present and can react to H 2 O and CO by RWGS, although because of the nature of the DRM reaction, they are not likely to be equal in partial pressure.Further into the reactor, RWGS is less likely, because there is little CO 2 .The ALD-coated catalyst exhibited two desirable behaviors, the first being a decreased k RWGS and the second being an increased k DRM , both in absolute terms.The larger k DRM means more CO 2 consumption, and as more CO 2 is consumed the RWGS (which has an equilibrium constant of 0.48−0.76 at 650−750 °C) gradually reverses to water−gas shift, regenerating H 2 .Therefore, the change in both forward rate constants in favorable directions leads to the high H 2 /CO ratio observed for 0.5 nm ALD-CZA40.

Nature of Surface Ni−Al−Ce Interactions in DRM.
Fully understanding the operando surface structure of 0.5 nm ALD-CZA40 or any other overcoated catalyst is a difficult to impossible task.However, with the characterization and DFT data presented here, it is possible to make hypotheses about the oxidation state and coordination environment of the surface Ni, Ce, and Al.From the K-edge Ni XANES and XAFS results we concluded that some of the Ni remained electropositive, and from all the XAFS fits, and in particular, by comparing FT-XAFS fits for two vs one atoms of Ni 2+ occupying a tetrahedral alumina site (Figure 5b), we know that the most likely coordination environment for the electropositive Ni is Ni 2+ −O clustering.The Ni is not bulk NiO or surface-adsorbed NiO due to the lack of a visible NiO phase in XRD (Figure 3), the known propensity of NiO to reduce to Ni 0 at DRM conditions, 115 and the DFT calculations (Section 3.5.3.2) which reveal that a Ni 2+ -doped (110) alumina surface manifests a negative ΔΔE (in eV/Ni) relative to adsorbed NiO.As a dopant, Ni 2+ gets the benefit of attaining a coordination number closer to what it would have in bulk NiO.This validates the preference of Ni to dope rather than adopt the undercoordinated adatom configuration.Furthermore, a doping energy comparison between Ni-doped ceria and Nidoped alumina surfaces revealed a thermodynamic preference for Ni to dope γ-Al 2 O 3 , with a minimum favorability of −0.39 eV/Ni (eq 24).
In the 0.5 nm ALD-CZA40 catalyst, more Ni remained embedded in this oxide lattice (which could also be a NiO-Al 2 O 3 −CeO 2 mixture) during long-term DRM.While the tetrahedral site of γ-Al 2 O 3 was used as the model in the XAFS fitting, the fit was far from perfect (Figure 5b).There must be other atoms (Ni or Ce) around the Ni core absorber that affect its scattering.The DFT results of Section 3.5.3.1 also confirm the tendency of Ni to remain as Ni 2+ in γ-Al 2 O 3 .Among the Ni 2+ -doped structures, those with Ni doped in an octahedral− octahedral site combination have the most negative ΔE formation .While this suggests that the octahedral−octahedral combination is the more energetically favorable site arrangement (as opposed to tetrahedral−tetrahedral), both XAFS and DFT simulations predict that clustering of Ni (two or more Ni atom doping) is favorable relative to single Ni atom doping.The observed disparity in Ni location in alumina between XAFS fitting and DFT calculations can be attributed to the inherent uncertainties present in both processes and the possible presence of additional Ni clustering or Ce atom doping.
The kinetics results reinforce this interpretation of the active surface structure (Table 1), even for the catalysts prior to their final stabilized forms.In previous work we have shown that for good DRM catalysts two different ΔE groupings are evident, high (>104 kJ/mol) and low (<80 kJ/mol). 116These ranges are distinguished by the capacity to generate high OV and activate CO 2 independently of the transition-metal sites.As CeO 2 is superior in this regard, we expect catalysts with more CeO 2 mixed into the surface layer to give lower observed ΔE′s.This further suggests that the thinner ALD overcoats should exhibit lower ΔE′s.This is exactly the behavior observed in Table 1.
Ce XPS (Figure S9) was performed to analyze the electronic properties of Ce near the surface.On going from fresh to used 0.5 nm ALD-CZA40 the XPS-determined Ce 3+ concentration increased from 35 to 71%, and also showed a marked increase in total surface Ce concentration.This is consistent with CeO 2 , which has been doped with some Ni. 13,20,76,99By examining the results of Ni K-edge XANES/XAFS (especially the low Ni−Ni coordination numbers), the Ce XPS results, and the  (110) facet, and also that there can be Ce doping into Al 2 O 3 .The DFT results of Section 3.5.2confirm the enhanced favorability of Ce 4+ reduction after Ni incorporation, as also observed in the TPR and XPS studies.Furthermore, the results of Table 5 suggest that a single Ni atom incorporation does not enhance the surface reductivity, whereas Ni clustering does.This is in agreement with the TPR results of ALD-coated Ni-CZA (Figure 4), i.e., the observation that Ni incorporation into the ceria lattice during use led to the downward shift of the TPR peak, generating a higher Ce 3+ content at a lower temperature.
The intimacy of Ce−Al mixing is exemplified by both the STEM images indicating poor crystallization of both Al 2 O 3 and CeO 2 in used catalysts, by the lack of an XRD signature for the Al 2 O 3 overcoats (Figures 3 and S8), and by the TPR results (Figure 4), showing a merger of reduction peaks in the overcoated (especially the 0.5 nm) catalysts.CeO 2 and Al 2 O 3 can mix at elevated temperatures, ultimately forming a CeAlO 3 phase. 117,118There are several regions of the TEMs so indistinct that they are consistent with such a disordered mixed-oxide structure, the exact nature of which cannot be discerned from the characterization results.

Comparison of ALD-CZA40 to
Other Overcoated or Core−Shell DRM Catalysts.Core−shell architectures are now common in DRM.As previously mentioned, these architectures are designed to suppress deactivation or in some cases increase DRM activity at lower temperatures.While this work is based on the use of ALD to deposit an overlayer of one oxide on top of a mixture of Ni/second oxide, it is worth realizing that there are other ways to do this and that some of these other methods also give core−shell microstructures.Thus, it is important for the results presented here to be analyzed in context with both ALD-deposited and other core−shell DRM catalysts.Table 8 presents DRM turnover frequencies (TOF, defined here as mol CH 4 converted per second per total mol Ni), the H 2 /CO ratio, and temperature for 0.5 nm ALD-CZA40 along with other core−shell catalysts.Catalysts with a TOF > 0.25 operating in the 700−800 °C range with some proven long-range stability and with H 2 /CO > 0.7 are arbitrarily chosen as the best representative DRM catalysts; there are not many of these in the open literature.However, some other catalysts are included in Table 8 that do not exactly meet these criteria, for comparison purposes.
It can be seen from Table 8 that the TOF of 0.5 nm ALD-CZA40 is well below that of another Al 2 O 3 ALD-prepared catalyst (third row of the table).However, this catalyst deactivated very quickly with a 20% loss of activity in 60 h.By comparison, 0.5 nm ALD-CZA40 did not lose activity through 140 h time onstream.The other Al 2 O 3 ALD-prepared catalyst showed an induction period similar to ours before maximum activity was achieved, and the induction period was attributed to the reduction of NiAl 2 O 4 to separate phases of Ni 0 and Al 2 O 3 .However, this is not the case in our work because there is no evidence of bulk NiAl 2 O 4 in any of our catalysts, fresh or used.
Ahn et al. continued the work on ALD-prepared DRM catalysts by adding La to the Al 2 O 3 overlayer to fill cation vacancies in γ-Al 2 O 3 . 53This was shown to eliminate the induction period seen before.However, this catalyst lost ∼20% of its activity in ∼40 h.
Tathod et al. developed a highly active Ni/MgAl 2 O 4 /Zr catalyst for DRM. 119While the activity of this catalyst is far superior to any known Ni-based DRM catalyst, there are some drawbacks.First, it was operated at 800 °C, where it is easier to attain high DRM activity, high H 2 /CO ratio, and minimal coking based on equilibrium considerations.But even at this highly favorable temperature, the sample is far less selective (0.78 H 2 /CO ratio) than the 0.5 nm ALD-CZA40 catalyst.Finally, this Ni/MgAl 2 O 4 /Zr catalyst is less stable than ours.
The Ce-based DRM core−shell catalysts have become popular due to their ability to form an active CeO x phase to enhance Ni-site activity, attributed to the formation of OV in CeO x (from doped Ni, e.g.).These OVs are known to activate CO 2 .Das et al. synthesized a Ni/SiO 2 @CeO 2 core−shell material that was active at 700−750 °C. 50While its activity at 750 °C was greater than 0.5 nm ALD-CZA40, the latter is 9% (absolute basis) more selective.
Core−shell catalysts of Ni@oxide have been studied, as well.Han et al. coated Ni nanoparticles with a SiO 2 shell. 120This catalyst exhibited approximately the same activity as 0.5 nm ALD-CZA40 (0.28) and showed excellent stability but was again less selective by 12%.For DRM, selectivity and stability are usually considered more important than raw activity, because H 2 is a more valuable product than CO, and the catalysts would have to operate for long periods between regenerations (or replacement) to make the process economical.
4.4.Impact of the Al 2 O 3 Overcoat.The deposition of conformal ALD films, especially ultrathin films, is dependent on the reaction affinity of the precursor with available surface ligands.A cartoon of the working catalyst surface is shown in Figure 13.The preferential reaction of TMA with the hydroxyterminated oxide substrate is believed to leave the prereduced Ni particles uncoated when thin layers (∼<1 nm) are deposited.Thicker ALD layers eventually creep over smaller Ni particles, which can still be exposed to gases during final reduction and DRM.Our DFT results above suggest that the ultrathin Al 2 O 3 layers then mix with the Ce/Zr substrate when exposed to reaction conditions, forming an amorphous, thin (1−2 ML) CeAlO x /Al 2 O 3 surface layer.The thin CeAlO x layer serves to inhibit the RWGS reaction by reducing hydrogen spillover, since more Ce is stabilized in the 3+ state.However, oxygen uptake via oxidation of Ce 3+ through CO 2 adsorption and transport to the metal-oxide interface are not significantly impacted.This can be seen from the activity of the catalysts and the modest increase in activation energy (Table 1) for thin layers (43 kJ/mol) compared to the thicker layer (86 kJ/mol) compared to reactions over nonreducible substrates (>100 kJ/ mol).This dual nature of thin CeAlO x /Al 2 O 3 layers is achieved by the generation of thin, miscible coatings on reducible substrates.

CONCLUSIONS
An ALD-deposited Ni-containing catalyst, made by overcoating Al 2 O 3 on Ni/CeO 2 −ZrO 2 , compares well to other core−shell and ALD-prepared catalysts for dry reforming by all significant metrics, in particular, giving an H 2 /CO ratio near one.The mixture of a nonreducible and reducible support in a core−shell architecture, with Ni initially sandwiched between them but diffusing to the surface during both reduction and DRM, is a promising path to achieving highly active, selective, and stable DRM catalysts.An optimal Al 2 O 3 overlayer thickness of ∼0.5 nm was determined.The ALD overlayer increased the lifetime of the uncoated catalyst by at least 10-fold.The coking rate was minimized, and the Ni clusters of the used catalyst were smaller and less reducible.This structure had the added benefit of decreasing the RWGS rate and increasing the DRM rate.We speculate that the decrease in the RWGS rate reflects reduced H 2 spillover to the coated, Al 2 O 3 -containing, surface.
The induction time seen during reactor testing at high conversion is associated with the migration of CeO x from bulk CZA to the surface, generating a mixed CeO 2 −Al 2 O 3 phase.It is likely that this mixed oxide is where the Ni 2+ and small Ni clusters, which are a significant fraction of the total Ni, are located.While this active Ni would prefer association with pure Al 2 O 3 rather than pure CeO 2 , the mixing of the two oxides at the surface renders the actual surface more complex.
■ ASSOCIATED CONTENT * sı Supporting Information the National Science Foundation, and grant CNMS2021-B-00889 from the Oak Ridge National Laboratory, Center for Nanophase Materials Sciences.We also acknowledge experimental assistance from Haoran Yu and David Cullen of ORNL and Paul Toups of LSU.

d 52 a
DRM Rate in mmol/(mg cat•h).Coking Rate in mg Coke/(mg cat•h).The coking rate was measured at >0.5 h into the run, but in some cases, the weight was still decreasing, indicating an essentially zero rate of coking.b Observed activation energy (Arrhenius equation) for dry reforming.c PR is prereduced at 550 °C with 5% H 2 /N 2 for 3 h, prior to ALD. d No measurable coke over 0.5−2 h time onstream.e Computed from experiments below in the reactor system.
the addition of the overlayer and the 2+ charge on some of the Ni adds extra layers of complexity due to the unknown positions of Ni 2+ atoms in either Al 2 O 3 , CeO 2 (with ZrO 2 ), CeAlO 3 , or as a Nicontaining phase such as NiAl 2 O 4 or NiO.In particular, γ-Al 2 O 3 has two types of octahedral and one type of tetrahedral site, in any of which Ni 2+ could insert.To generate potential scattering pathway files in the Feff program, ATOMS files containing the positions of atoms and bond lengths for all of the above phases were obtained from XRD data, and then a single Ni core absorber was substituted for a cation in these files, except for NiAl 2 O 4 and NiO, where the Ni atoms were already present. 106−109 The numbers of oxygen atoms were adjusted to maintain electroneutrality by removing an O atom within the scattering volume, which consisted of a sphere of radius 5 Å from the core absorber.The XAFS analyses of used 0.3, 0.5, and 1.2 nm ALD-CZA40 proved less difficult due to the much larger concentrations of Ni 0 present in these samples.For XAFS analysis of the used 0.5 nm ALD-CZA40, the data were regressed by weighting the S 0 2 values of pure Ni and NiO phases: only the first-shell [Ni]−Ni, [NiO]−O, and [NiO]−Ni paths were included.This multiphase process is consistent with past work on first-shell fits of mixed Ni 0 /Ni 2+ in oxides.

Figure 5 .
Figure 5. (A) Ni K-edge XANES spectra for CZA40, fresh, and used 0.5 nm ALD-CZA40.(B) Fourier transform (FT) XAFS for fresh 0.5 nm ALD-CZA40 (markers) compared to simulated curves for a single-atom Ni substitution in a tetrahedral site (red) and dual ("cluster") Ni tetrahedral site substitution (blue).Green boxes are the fit range.(C) FT-XAFS for used 0.5 nm ALD-CZA40 (markers) compared to simulated curves for a Ni 0 /Ni 2+ mixture (red).Green box is the fit range.

3 . 5 .
figure).It is impossible to distinguish the (200) plane of cubic NiO or the (111) plane of cubic Ni (0.203 nm) from the (220) plane of CZA (0.191 nm CeO 2 , 0.186−0.192nm for CZA), making it difficult to assign some lattice fringes to either species, thus these fringes are marked as representing Ni, NiO or CeO 2 planes.All of these planes are common in their respective species.Planes with d spacings around ∼0.221 nm are assigned to the (100) plane of a hexagonal Ni phase.No carbon filaments, typically seen on heavily coked DRM catalysts, 111 are seen on the used 0.5 nm ALD-CZA40.But there are also regions characteristic of an Al 2 O 3 phase in Figures 6a,b and S19b.It was impossible to perform this analysis for uncoated, used CZA40 (Figure S20) given the large amount of surface coke.The DF STEM images illustrate the complex semicrystalline environment of the Ni and suggest why it is difficult to get exact fits to model structures in EXAFS.Figure 7a,b shows EDX maps of both fresh and used 0.5 nm ALD-CZA40.Both images are at lower magnification, capturing several Ni particles.The fresh sample (Figure 7a) shows dispersed Ni and no segregation of Al.The used sample (Figure 7b) clearly lost some of its initially high Ni dispersion due to crystal ripening, but there are still many regions of dispersed Ni present.An average surface Ni particle size was obtained from the EDS images for both fresh and used 0.5 nm ALD-CZA40.Upon use this average particle size increased from 15.8 to 17.3 nm, consistent with a slight loss of Ni dispersion, but given the limited resolution, this population is only characteristic of the largest Ni clusters or grouping of clusters.3.5.Insights into DFT Calculations.This section is divided into four subsections.Section 3.5.1 discusses the thermodynamic feasibility of forming ceria-Al 2 O 3 mixed oxides in the bulk and surface.The stability and reducibility of Niincorporated ceria are compared to pure ceria surfaces in Section 3.5.2,along with calculations supporting the favorability of Ce 4+ reduction post Ni incorporation, as observed in both the TPR and XPS studies.Section 3.5.3provides information on the thermodynamic stability of Ni− Al 2 O 3 bulk structures, in agreement with the XAFS fitting results.The stability of the Ni−Al 2 O 3 surfaces is analyzed to

Figure 7 .
Figure 7. (A) EDS mapping for a DF STEM image of fresh 0.5 nm ALD-CZA40.(B) EDS mapping for a DF STEM image of used 0.5 nm ALD-CZA40.

Figure
Figure S21 shows per cation formation energies for these Ce−Al mixed oxides as a function of the Ce concentration.Formation energies of all Al-doped Ce 2 O 3 structures were much greater (by >1 eV/cation) than other structures of overall similar Ce concentration and are therefore not presented for discussion.Formation energies of all Ce-doped γ-Al 2 O 3 oxides (blue squares, Figure S21) are positive (∼0.15 and 0.4 eV/cation), indicating that these are less stable than the segregated parent oxides.These data, together with an expansion of unit cell volume upon Ce doping (>2% volume increase), reveal the significance of lattice strain induced due to Ce doping; the larger ionic radius of Ce 3+ compared to that for Al 3+ (1.15 vs 0.68 Å) renders these doped structures unfavorable.Substitution of Ce into tetrahedrally coordinated positions in the γ-Al 2 O 3 lattice (Al−O ∼ 1.8 Å in parent structure) compared to octahedrally coordinated positions (Al−O ∼ 1.95 Å in parent structure) resulted in both smaller unit cell expansion (1.9 vs 2.5%) and a lower formation energy (+0.16 vs +0.20 eV/cation).Cerium aluminate (CeAlO 3 ) is a stable bulk mixed oxide relative to its parent oxides.Note here that we do not observe this bulk oxide formation in our experimental work, but it is observed in other works.112−114A few nonstoichiometric cation substitutions into this mixed-oxide phase have favorable 3.5.1.2.Surface Doping of Ce on Al 2 O 3 .Cell volume expansion can be negligible upon surface-level doping; therefore, doping of Ce on γ-Al 2 O 3 surfaces was examined to assess the stability of the experimentally observed mixed CeAlO x overlayers.The two most stable facets of γ-Al 2 O 3 , (

3 . 5 . 2 .
Stability and Reducibility of Ni-Ceria Surfaces.Next, we compare the stability of Ni in adsorbed and doped configurations on CeO 2 surfaces to further understand the reducibility of such mixed-oxide surfaces.

3 . 5 . 3 .
Stability of Ni-Alumina Systems.Results from XANES and XAFS experiments indicate the presence of Ni 2+ in γ-Al 2 O 3 and the favorability of Ni clustering in the alumina lattice.This section presents results from DFT calculations of Ni−Al 2 O 3 systems that support these experimental findings.3.5.3.1.Doping of Ni in Bulk Al 2 O 3 .Ni-doped γ-Al 2 O 3 was generated by replacing two Al sites in the γ-Al 2 O 3 bulk structure with two Ni atoms.The doped Ni atoms were forced to stay as Ni 2+ by removing either a Ni-bonded or the nearest Al-bonded O atom.The doped site combinations used for this study are classified into three broad categories: octahedral− octahedral, octahedral−tetrahedral, and tetrahedral−tetrahedral (refer to Figure

2 .
Stability of Doped vs Adsorbed Ni−Al 2 O 3 Surfaces.Understanding the doping energetics of Ni on Al 2 O 3 surfaces is crucial to understanding where Ni prefers to locate − on ceria or on Al 2 O 3 .Thus, the ΔΔE of the most stable Ni 2+ -doped (110) γ-Al 2 O 3 surface is reported here with reference to the most stable Ni 2+ -adsorbed (110) γ-Al 2 O 3 surface (eq 21).Ni 2+ -doped (110) γ-Al 2 O 3 surfaces were generated by replacing two surface Al atoms with two Ni atoms, followed by the removal of a Ni-bonded O atom.The structures are shown in Figure

4 .
Preference for Ni−Al 2 O 3 over Ni-Ceria Structures.To determine the relative stabilities of Ni-doped surfaces, we computed reaction energies for a two Ni 2+ -doped (110) γ-Al 2 O 3 surface (Ni 2 Al 30 O 47 ) and a two Ni 2+ -doped (111) CeO 2 surface (Ni 2 Ce 14 O 30 ).The Ni 2 Ce 14 O 30 surface structure was chosen because it was the most stable structure with no surface Ce 4+ reduced.DFT energies of the individual species in eq 24 were used to calculate the energy required to form Ni 2 Al 30 O 47 relative to Ni 2 Ce 14 O 30 .A negative ΔE indicates that Ni has a strong preference for doping alumina rather than ceria.

Table 2 .
Second-Order DRM and RWGS Rate Constants of Catalysts from Differential Rate Measurements, mL 2 /(min•mg cat • mmol) a 140 h time onstream DRM.
81,86,88,86,88,90appears when the ALD overlayer is present, so all bulk CZA that could be reduced was reduced before or during reaction.PeakA is usually attributed to OV formation in doped Ni/CeO 2 , 20,76−79 and note that it is largest in CZA40 and decreases substantially as the layer thickness increases, essentially disappearing for 1.2 nm ALD-CZA40.Peak C with a maximum of ∼530−535 °C is in the range (or below it) of a typical reduction peak for surface Ce with low loading Ni present, in 1:1 or near 1:1 molar CZA.77,80−89The reductions of the two elements are intertwined.It is evident that under reaction conditions, the Ni and the Ce were not completely reduced.There are also contributions from surface or nanoparticle NiO at lower temperatures, which can be identified as the broad shoulders present between 300−430 °C.20,78,84,86,88,90Ni2+reductionwithindisordered,mixed CeO 2 −Al 2 O 3 also shows a peak in this temperature range.91,92AnyNipresentas nanoparticle NiO reduces at least 50 °C lower than 530 °C.81,86,88AnyNi intercalated within Al 2 O 3 alone would reduce at higher temperatures, >580

Table 3 .
Atomic (At)% from XPS Spectra, Calculated on an Oxygen-Free Basis

Table 7 .
ΔΔE of Ni 2+ -Doped (110) γ-Al 2 O 3 Surface (the Reference is the ΔE of NiO-Al 32 O 48 )TEM results, it is hypothesized that there was gradual mixing of CeO x with the alumina overlayer, that most of the Ni is highly dispersed, some of it is in proximity to this CeO x oxide, and this gradual mixing is associated with the significant changes in DRM and RWGS kinetics observed in the first 1−2 days of reactor operation.The DFT results of Section 3.5.2(Figures 8−10) also support the thermodynamic favorability of CeO x overlayer formation on γ-Al 2 O 3 .Collectively, the DFT doping studies indicate that well-defined conformal ceria layers are not expected to form on the (100) Al 2 O 3 facet, but can form on the

Table 8 .
Turnover Frequency (Total Ni Atom Basis) and H 2 /CO of Core−Shell-type DRM Catalysts