Introduction
Click to copy section linkSection link copied!
Heterogeneous tandem catalysts contain two or more distinct catalytic sites that synergistically drive key reaction steps, yielding high selectivity/activity/stability often difficult to obtain using individual sites alone. (1−5) One subclass of such catalysts is metal–metal oxide catalysts, where the metal and metal oxide active sites work in tandem, while both sites remain exposed during the reaction. These catalysts have demonstrated enhanced performance in driving key complex reaction chemistries, including alkane dehydrogenation, biomass valorization, and methanol steam reforming, among others. (6−9) Advanced characterization techniques have provided insights into understanding the active site behavior of su…
Introduction
Click to copy section linkSection link copied!
Heterogeneous tandem catalysts contain two or more distinct catalytic sites that synergistically drive key reaction steps, yielding high selectivity/activity/stability often difficult to obtain using individual sites alone. (1−5) One subclass of such catalysts is metal–metal oxide catalysts, where the metal and metal oxide active sites work in tandem, while both sites remain exposed during the reaction. These catalysts have demonstrated enhanced performance in driving key complex reaction chemistries, including alkane dehydrogenation, biomass valorization, and methanol steam reforming, among others. (6−9) Advanced characterization techniques have provided insights into understanding the active site behavior of such metal–metal oxide catalysts. (6,10−13) Based on these studies, the enhanced catalytic performance is attributed to various phenomena, (14) with structural rearrangement of the metal and metal oxide phases under reaction conditions being an important factor influencing catalytic performance. (15−17) However, the chemical and geometric features of the active phases that control these structural transformations and reaction kinetics at the interface during the synthesis and reaction remain poorly understood. This lack of understanding leads to relying on trial-and-error approaches to determine optimal metal–metal oxide (M1 – M2 Ox Hy) catalysts. Novel methods that can overcome these challenges by elucidating the governing atomic features of the M1 – M2 Ox Hy active sites are hence desired to conduct a systematic physics-driven search for optimal catalysts.
First-principles methods, such as Density Functional Theory (DFT), can reveal the electronic and geometric features that govern the structure, stability, and reactivity of M1 – M2 Ox Hy catalysts. Recent computational approaches have predicted the catalyst structures at the M1 – M2 Ox Hy interface by identifying low-energy surface oxide structures on a metal at reaction conditions. (18−20) For example, Kempen and Andersen identified the stable oxide geometries of ZnyOx and InyOx on fcc(111) metal surfaces using a global optimization workflow. (20) Kumari et al. identified low-energy configurations of different ligands such as formate, oxygen, and hydroxyls on ZrOx in a ZrO2/Cu(111) catalyst under CO2 hydrogenation conditions using the grand canonical basin hopping algorithm. (19) Although significant advances have been made, incorporating the diverse atomic features that accurately represent the active site structure under in situ experimental conditions remains a key challenge, often called the Material Gap. (21) A key step in bridging this gap requires consideration of additional structural and chemical features in the atomic models. These features include the presence of potential defects on the metal, varying oxide stoichiometries in the presence of oxidizing and reducing environments, considering the variable coverage of oxide as a function of reaction conditions, and the presence of functional groups such as OH* on the oxide surface. However, considering the above-mentioned structural and chemical features to construct accurate atomic models is hindered due to the vast configuration space (sampling challenge). This work addresses this sampling challenge by adopting an algorithmic framework for systematically investigating the metal–metal oxide interface to identify M1 – M2 Ox Hy active site structures under experimental conditions. We utilize the framework to study the important case of Oxidative Propane Dehydrogenation (ODHP) reaction on the tandem overcoated In2O3-Pt/Al2O3 catalyst. (7)
ODHP is an alternative process to propane dehydrogenation (PDH) where an oxidant such as O2 (often referred to as O2 assisted-ODHP or O2-ODHP) is added to make the overall reaction exothermic and overcome its equilibrium limitations. (22) However, achieving high selectivity in O2-ODHP is a challenge due to multiple side reactions, including combustion, cracking, and overoxidation. (23) Recently, using a tandem ODHP reaction that integrates PDH and Selective Hydrogen Combustion (SHC), Yan et al. demonstrated the In2O3-Pt/Al2O3 catalyst to successfully overcome these challenges. (7) The performance of this catalyst was proposed to be intrinsically linked to the local structure of the indium oxide (InOx) overcoat at the metal–metal oxide interface. Specifically, the catalyst is shown to undergo rearrangement, exposing both metal and metal oxide active sites during the reaction, which tightly couples the reactions for direct transfer of intermediates between the catalyst domains (Pt and In2O3). Nevertheless, the nature of the underlying atomic structure and features that dictate these structural rearrangements and their subsequent effect on product selectivity remain unknown, hence hindering a physics-driven search for optimal tandem ODHP catalysts. Our proposed algorithmic framework, for the first time, elucidates the influence of different geometric, stoichiometric, and chemical features of the oxide and metal phases in selectively driving the ODHP reaction, overcoming this challenge.
The computational framework developed in this work couples first-principles-based Density Functional Theory (DFT) with a unique data-driven algorithm and successfully tackles the sampling challenge associated with modeling complex reactions on intricate M1 – M2 Ox Hy catalysts. The framework builds upon the work by Deshpande and Vlachos (24) and now incorporates a systematic consideration of Brønsted acid sites and consideration of mixed O*/OH* functional groups to study the growth and coverage of indium oxide (In2O3) on the platinum (Pt) surface in the presence of defects under ODHP conditions. Our approach utilizes a modified graph-theory-based approach (24) to identify all unique, stable configurations of indium oxide (InOxHy) on a defective Pt surface. These stable structures are used to construct a thermodynamic surface phase diagram, which for the first time elucidates the dynamic reconstruction of the surface oxide active sites from synthesis to catalytic conditions. The dominant surface oxide phase, identified as the stable structure under the reaction conditions, is used to study the thermodynamics and kinetics of the ODHP reaction. Through our analysis, we highlight the key atomic features that drive selective and stable ODHP on the Pt-InOxHy catalyst, thus establishing a structure–performance relationship between the tandem metal–metal oxide catalyst and oxidative alkane dehydrogenation chemistry. The identified relationship, along with the algorithmic framework, now becomes a stepping stone to study other important complex chemistries in the important class of M1 – M2 Ox Hy catalysts.
Computational Methods
Click to copy section linkSection link copied!
Periodic density functional theory (DFT) calculations are performed using the Vienna Ab initio Simulation Package (VASP), (25) where the Kohn–Sham equations are solved using the Perdew, Burke, and Ernzerhof functional (PBE) (26) self-consistently with the Projector Augmented Wave (PAW) method. (27) To model the Pt(111) surface, a 4 × 4 × 4 unit cell is utilized, while for Pt(322), a 4 × 5 × 4 unit cell is chosen, with an optimized lattice constant of 3.97 Å (Supporting Information, 7.8). A vacuum spacing of at least 11 Å on each side of the slab is considered in the z-direction, with the bottom layer of the slab constrained. A planewave energy cutoff of 400 eV is used with a 3 × 3 × 1 k-point set for all surfaces. The partial electronic occupancies are determined according to a Methfessel–Paxton scheme (28) with an energy smearing of 0.2 eV. The structures are relaxed until the Hellmann–Feynman forces on the atoms are less than 0.05 eV/Å. The dipole and spin corrections have a negligible effect on the energy of the relaxed system and are therefore not considered (Supporting Information, 7.9).
The indium oxide (InOxHy) structure is modeled on the Pt(322) surface, which provides realistic surface characteristics by including four layers of well-coordinated terrace sites and one layer of under-coordinated step edge sites. The binding energy of InOxHy on the Pt surface is evaluated as the formation energy of InOxHy on Pt(322) (Eform(InxOyHz)) referenced to the bulk indium oxide (In2O3) as shown in eq 1. Here, EDFT(Pt – InxOyHz) represents the potential energy of InOxHy on Pt(322), EDFT(Pt) is the potential energy of Pt(322), and EDFT(In2O3) is the potential energy of cubic In2O3 structure adopted from The Materials Project. (29)
Eform(InxOyHz)=EDFT(Pt−InxOyHz)−EDFT(Pt)−EDFT(In2O3)−(x−2)×μIn−(y−3)×μO−(z)×μH+ZPE(NOH*,NH2O*,NO*)
(1)
μIn, μO, and μH are the chemical potentials of indium, oxygen, and hydrogen, respectively. ZPE(NOH*, NH2O*, and NO*) corresponds to the sum of zero-point energy (ZPE) contributions of OH*, H2O*, and O* adsorbates on Pt-InOxHy, respectively. The zero-point energies are calculated based on harmonic vibrational states (Supporting Information, 1.1). The entropies of InOxHy and bulk In2O3 structures are assumed to be similar and will cancel out. The adsorbed oxygen and hydrogen are approximated to be in quasi-equilibrium with their respective gas-phase components. The chemical potential of indium can be estimated from bulk In2O3 as a reference, as shown in eq 2, assuming the source as bulk indium oxide (In2O3(s)).
μIn=EDFT(In2O3)−(32)μO
(2)
Under the reaction conditions, the chemical potentials of oxygen gas (μO) and hydrogen gas (μH) are evaluated using their standard chemical potentials at the reaction temperature and standard pressure (p0). A pressure correction term is then applied based on the partial pressure (pi) of the gases relevant to the experimental conditions. The chemical potential of oxygen gas (O2) is estimated at 723 K (T) using the free energy of the oxygen reduction reaction, (30) and the chemical potentials of hydrogen gas (H2) and water(H2O) based on eq 3. The computation of the chemical potential of gas-phase components at 723 K and p0 is further discussed in the Supporting Information, 1.1. The partial pressures of propane (PC3H8) and oxygen (PO2) are considered as 10 and 5 kPa, respectively, in line with the experimental conditions used in the ODHP reaction. (7) Assuming negligible gas-phase combustion during the reaction conditions, the chemical potential of oxygen gas is evaluated at a pressure of 5 kPa by adding the pressure correction to the chemical potential at 723 K and standard pressure of 1 bar, as shown in eq 4. To understand the catalyst structure under varying gas pressures of O2, a range of oxygen chemical potentials is considered. These chemical potentials are referenced to oxygen gas at 5 kPa. The chemical potential of H2 gas is computed under steady-state conditions at 1.32 kPa. The choice of the chemical potential of H2 gas is described in the Supporting Information, 1.2.
μO=12×μO2(g)=μH2O−μH2+2×1.12
(3)
Δμi(T,pi)=Δμi(T,p0)+kBTlnpip0
(4)
Stable InOxHy structures at different coverages on Pt(322) are initially estimated by calculating the formation energies using bulk In2O3 (μIn′) chemical potential, and the hydrogen (μH(gas)′) and oxygen (μO(gas)′) chemical potentials relevant to experimental conditions at 723 K (Supporting Information, 1.2). A phase diagram, as a function of the chemical potentials of hydrogen and oxygen gases, and In chemical potential referenced to the bulk In2O3, is then constructed to predict stable structures ranging from synthesis and pretreatment to reaction conditions.
Further, the formation energies of key ODHP intermediates are evaluated on Pt-InOxHy, Pt(111), and Pt(322) catalyst surfaces to understand the adsorption thermodynamics and kinetics of the ODHP reaction. These three surfaces are chosen to systematically understand the role of well-coordinated and under-coordinated (terrace vs steps) metal sites, along with InOxHy, in driving the PDH reaction. The key reaction intermediates for PDH mechanistic analysis are chosen based on a previous theoretical analysis. (31) The formation energy of an intermediate is computed relative to the gas-phase energy of propane on the Pt-InOxHy catalyst using eq 5.
Eads(C3Hx)=EDFT(C3Hx−B)−EDFT(B)−EC3H8+(8−x)×μH
(5)
Gads=Eads+ZPEads−TSads
(6)
Here, Eads(C3Hx) is the formation energy of C3Hx relative to the energy of propane denoted as EC3H8 and EDFT(C3Hx − B) is the potential energy of C3Hx adsorbed on a surface B corresponding to Pt-InOxHy, Pt(111), or Pt(322) surfaces, respectively.
The standard free energies for all adsorbates are calculated at 723 K based on eq 6. The zero-point energies are calculated based on harmonic vibrational states, and the standard state entropy corrections are calculated using the harmonic oscillator approximation. (31) For a physisorbed propane* intermediate, entropy corrections are calculated as described by Seemakurthi et al., (31) accounting for hindered translator and rotor modes. (32,33) The standard free energies of gas-phase components such as propane and propylene are calculated using ASE, (33) based on ideal gas assumption at 723 K. Activation barriers for key elementary steps are calculated using the Climbing Image Nudged Elastic Band (CI-NEB) method. (34) Eight images are generated between the initial and final states using the Image Dependent Pair Potential (IDPP), (35) and the CI-NEB calculations are conducted. The transition state (TS) for propylene gas desorption is computed using the 2D-ideal gas assumption. (31,36) TS entropies of adsorbates are approximated by the initial and final state entropies, depending on whether the TS is closer to the initial or final states. (31)
Results and Discussion
Click to copy section linkSection link copied!
To elucidate the role of the active site structure of Pt-InOxHy in driving ODHP, structural analysis is initially performed. An algorithmic approach is utilized to explore the diverse range of possible InOxHy configurations on a Pt surface. Building upon this, an ab initio thermodynamic phase diagram is constructed to understand the formation of InOxHy on Pt, mapping the evolution of the InOxHy structure on Pt from synthesis to that under reaction conditions. The latter part of the paper focuses on elucidating the role of a stable InOxHy structure in driving the SHC reaction by exploring the oxygen adsorption and reduction thermodynamics on the catalyst. Following this, a thermodynamic and kinetic analysis is presented to understand PDH on the Pt-InOxHy catalyst and discern the influence of the InOx overlayer in selectively driving the PDH chemistry.
Algorithmic Approach for InOxHy Structure Search
Predicting the intricate structure of a metal oxide overlayer on a metal is challenging due to the vast number of possible configurations arising from varying coverage, stoichiometry, oxidation states, the possible formation of Brønsted acid sites, and the presence of defects on the underlying metal. These complexities hinder our understanding of the nature of the active site structure and its subsequent role in driving catalytic reactions. Hence, to overcome these sampling challenges, a systematic and data-efficient approach is required. Herein, the structure of InOxHy on a defective Pt(322) surface is modeled by utilizing a unique workflow that integrates the modified SurfGraph algorithm with a formation energy-based evolutionary algorithm, (24) as shown in Figure 1a.
Figure 1
Figure 1. (a) Schematic of the workflow. In atoms are placed using a modified SurfGraph, followed by the addition of hydroxyl groups to form InOxHy. Structures with maximum H-bond distribution are retained, and H atoms are oriented based on H-bond donor–acceptor distances. (b) Number of structures explored at each InOx coverage on Pt(322) and the formation energy of the most stable InOxHy structure at each coverage as a function of the most stable structure at 0.39 ML. Insets (i) and (ii) show the most stable InOxHy structures on Pt(322) at 0.39 and 0.73 ML, respectively. The formation energy is calculated using μIn′ with respect to bulk In2O3, μH(gas)′ with respect to the partial pressure of H2 at the reactor output of 1.32 kPa, and μO(gas)′ with respect to partial pressure of O2 at a reactor input of 5 kPa.
The workflow initiates with populating Pt(322) with unique starting positions of indium (In) atoms on Pt using SurfGraph (37) (Step 1). In Step 2, each In atom is coordinated with three oxygen (O) atoms, as hydroxyl (OH) groups, mimicking the oxidation state of In present in the bulk. The choice of using OH* stems from a systematic investigation of the position and chemical nature of O atoms within the InOxHy framework, which revealed that they are stable in the form of OH* groups (Supporting Information, 2.2), owing to the high proton affinity of surface oxygen atoms on In2O3. (38) This stability holds within a specific range of chemical potentials of the hydrogen and oxygen gases, as discussed in subsequent sections. In Step 3, a directional graph-based approach is employed to identify stable configurations with maximized hydrogen bonding distribution. (39) In Step 4, the structures are optimized using DFT, and the stable structures with formation energies within a cutoff are screened using the formation energy-based evolutionary algorithm. (24) For any given In coverage consisting of n In atoms on the surface, the formation energy is evaluated by determining the difference in binding energies of the current structure and the most stable structure at the preceding coverage with (n – 1) In atoms. In Step 5, unique InOxHy stable configurations are segregated using the modified SurfGraph algorithm by removing OH* groups and identifying unique Pt–In relaxed structures. These unique Pt–In structures are mapped back to the InOxHy configurations to obtain a stable set of configurations at the desired coverage. The unique Pt–In relaxed structures are then used as a basis for the subsequent enumeration, where additional Indium atoms are populated based on the modified SurfGraph algorithm. This cycle is repeated until saturated InOxHy coverage on the stepped Pt(322) surface is obtained. This approach allows for the determination of the InOxHy overlayer structure as a function of coverage on the stepped Pt surface, providing insights into the key atomic interactions and geometric properties that play a role in stabilizing Pt-InOxHy catalysts.
The results from the systematic exploration of InOxHy configurations as a function of In coverage are presented in Figure 1b. With the increase in In coverage on Pt, the number of possible InOxHy configurations increases substantially due to multiple orientations of InOxHy structures and their interactions with various types of Pt sites. Importantly, by utilizing the computational workflow described in Figure 1a, the number of DFT simulations required to identify stable InOxHy structures was systematically reduced from O(104) to O(102) (Supporting Information, 2.1). Through the structural exploration, it is revealed that the InOxHy structures exhibit favorable interactions with the Pt-step sites compared to those lacking such interactions (Supporting Information, 2.3). The enhanced stability arises from strong binding between In and stepped Pt atoms, and the formation of hydrogen bonding networks through OH* groups that bridge In and Pt-step sites, as shown in insets (i) and (ii) in Figure 1b. As shown in Figure 1b inset (i), even at a modest coverage of 0.39 ML, the Pt-step edge becomes saturated with InOxHy, forming a continuous chain of hydrogen-bonded OH* groups. Utilizing these insights on Pt-InOxHy binding, we adopted the assumption that for higher coverages beyond 0.5 ML, the step edges remain saturated with InOxHy to maintain the stabilizing interactions between In and the Pt-step. This assumption allowed for efficient exploration of higher coverages beyond 0.5 ML as shown in Figure 1b. Following this approach, structures with higher In coverages revealed ring-like structures on the terrace sites coupled with chain-like arrangement preserved at the Pt-step sites (Supporting Information, 2.3). At the saturation coverage estimated at 0.73 ML, the InOxHy phase occupies most of the Pt sites including the terrace and step sites with each In atom surrounded by at least three OH* groups as shown in the right inset in Figure 1b. The resulting stable structures across different In coverages exhibit stoichiometries of Inp(OH)2p+1 and Inp(OH)2p+2 above a coverage of 0.2 ML with each In atom coordinating with three to four OH* groups where p represents the number of InOxHy units. The red line in Figure 1b represents the formation energy trends of the stable hydroxylated InOxHy structures at each coverage under experimental conditions of μH(gas)′ at 1.32 kPa of H2 and μO(gas)′ at 5 kPa of O2 with reference to the energy of the most stable InOx structure at 0.39 ML coverage on Pt. As shown in Figure 1b, the formation energy of InOxHy decreases with an increase in its coverage on Pt(322), depicting an increased binding strength. However, this holds true within a certain range of chemical potentials of hydrogen and oxygen gases, as discussed in subsequent sections. Further, by utilizing the stable hydroxylated InOx structures, we incorporated an additional framework to identify structures that might be stable under conditions away from those of the reactor. The procedure involves systematically varying both O and H coverages in the subset of InOxHy structures identified through the algorithm presented in Figure 1a (Supporting Information, 2.5), resulting in the generation of active site structures with varying concentrations of the O/OH groups. These candidate structures are then DFT relaxed and compared for stability as a function of the O2 and H2 chemical potentials. For this study, the framework is applied to the fully hydroxylated InOxHy structure at 0.39 ML coverage to generate possible partially hydroxylated structures with different stoichiometries of O and H atoms, which we now refer to as O and H concentrations in the InOxHy structure. The choice of using 0.39 ML coverage stems from the dominance of this specific InOx phase after the pretreatment conditions, as will be discussed further. Through this analysis, a library of stable atomic InOxHy structures at different In coverages on Pt is now constructed. This library now provides the basis for understanding the structure evolution of the Pt-InOxHy catalyst during synthesis, pretreatment, and reaction conditions, as discussed in the next section.
Thermodynamic Phase Diagram of the InOxHy Overlayer on the Pt(322) Surface
Elucidating the metal oxide structure during synthesis, pretreatment, and reaction conditions is crucial in understanding the role of the metal–metal oxide interface in driving the ODHP reaction. First-principles-based surface phase diagrams have been used to predict the structures of surface oxides under experimental conditions. (18,40,41) These diagrams map thermodynamically favorable structures across varying chemical environments and chemical potentials, providing a systematic approach for the prediction of the evolution of the InOxHy structure from synthesis through reaction conditions. Using the stable InOxHy structures predicted at different coverages on Pt(322), a thermodynamic surface phase diagram is constructed as a function of the chemical potentials of gas phase O2 and H2 (μO=1/2μO2, μH=1/2μH2). μO and μH are varied to mimic different synthesis and pretreatment conditions, as shown in Figure 2a. A range of chemical potentials of Δμi(gas) is considered for oxygen and hydrogen gases, where Δμi(gas) is equal to (μi(gas) – μi(gas)′). Here, μi(gas)′ refers to chemical potentials of the O2 and H2 gases corresponding to the O2 reactor input of 5 kPa and the H2 reactor output of 1.32 kPa under the steady state (Supporting Information, 1.2), respectively, at a temperature of 723 K. μi(gas) corresponds to the chemical potentials of the O2 and H2 gases inside the reactor. A decrease in Δμi(gas) corresponds to a decrease in the partial pressures of the O2 and H2 gases in the reactor under experimental conditions. Stable structures are mapped across different chemical potentials using formation energies computed using eq 1. Figure 2b–g shows the stable structures corresponding to each contour region in the surface phase diagram.
Figure 2
Figure 2. (a) Structural phase diagram of stable InOxHy structures on Pt(322) at different chemical potentials of oxygen and hydrogen. (b–g) Stable structures as depicted in the phase diagram.
During synthesis, indium oxide is overcoated on Pt nanoparticles (NPs) using atomic layer deposition (ALD) at 150 °C. (7) Given the ALD process involves an oxidant and hydrogen-containing species that form surface hydroxyl groups, (42) stable InOxHy structures under these conditions can be predicted using chemical potentials of O2 and H2 corresponding to partial pressures at and above the steady-state conditions of 5 and 1.32 kPa, respectively. Analyzing the phase diagram in Figure 2a with Δμi(gas) equal to zero reveals that In6O13H13 represents the most thermodynamically stable structure on the Pt surface in the presence of a bulk In2O3 reservoir. Sensitivity analysis at elevated partial pressures of the O2 and H2 gases above the aforementioned conditions confirms that In6O13H13 remains stable under excess O2 and H2 environments (Supporting Information, 3.1). As shown in Figure 2b, this structure predicts that the Pt surface is entirely covered by the InOxHy overlayer under synthesis conditions. This is in line with the catalyst structure reported by Yan et al., (7) wherein the Pt surface is fully covered with indium oxide after synthesis using ALD and before the pretreatment. Following synthesis, the overcoated catalyst was subjected to pretreatment in a reactor where the temperature was increased to 450 °C under a nitrogen atmosphere. (7) To understand the structure evolution under these conditions, the phase diagram is analyzed at decreasing chemical potentials of O2 and H2 gases to mimic the nitrogen environment with residual gas-phase amounts (O2 and H2) during pretreatment. Under these conditions, it is revealed that the In6O13H13 structure could transform to a set of overlayer structures with In coordination of three at ∼0.39 ML InOxHy coverage as depicted in Figure 2c–f. These structures include In3O7H7 (Figure 2c) with the maximum number of OH* at the Pt-step sites, followed by In3O6H5, In3O6H4, and In3O7H3 (Figure 2d–f), which are stable at relatively low H-chemical potentials. All of these structures shown in Figure 2c–f consist of the InOxHy phase interacting with every Pt-step site through bridging O* or OH* species. A further decrease in the chemical potential results in the formation of the InO2H2 structure with a low coverage of ∼0.1 ML, shifting the In coordination to two, as shown in Figure 2g. However, such low coverages of InOxHy appear at extremely low O2 and H2 partial pressures and are unlikely to form compared to the ∼0.39 ML InOxHy phases. This indicates that the plausible overlayer structures that result from pretreatment are in the form of ∼0.39 ML three-coordinated InOxHy phases on Pt(322). Based on our phase diagram analysis, the structures in Figure 2b–f thus reveal a structural rearrangement of synthesized InOx overcoat on Pt to preferentially decorate around the Pt-step sites during pretreatment.
We next compare the predicted structural evolution of the catalyst during pretreatment with experimental results. Experimentally, pretreatment resulted in the formation of a porous structure uncovering around half of the Pt active sites, as evidenced by an increase in the intensity of the CO diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) feature associated with Pt well-coordinated sites in the Pt-InOxHy catalyst with temperature. (7) Based on the comprehensive phase diagram analysis and the experimental observations by Yan et al., (7) we predict the three-coordinated ∼0.39 ML InOxHy phase to be the dominant overlayer structure resulting from pretreatment, as it approximately consists of 50% of free Pt sites for the ODHP reaction (Supporting Information, 2.3). The arrow in Figure 2a illustrates the shift on the phase diagram from synthesis to pretreatment, depicting the corresponding structure change from In6O13H13 to ∼0.39 ML InOxHy phase on the Pt(322) surface. In line with this, the CO-DRIFTS spectra also showed a reduced intensity of features associated with the under-coordinated Pt sites compared to uncoated Pt NPs. (7) Further, the surface diagrams at different pretreatment temperatures show that the area of the contour region representing In6O13H13 reduces with an increase in temperature, indicating a decrease in thermodynamic stability of In6O13H13 across a specific range of H2 and O2 chemical potentials (Supporting Information, 3.2). Therefore, the phase diagram analysis reveals that the experimentally observed increase in Pt sites during pretreatment could result from the destabilization of the InOxH**y structure, specifically on the Pt-terrace sites, leading to their increased availability for PDH reaction. This reconstruction of InOxHy, uncovering the Pt-terrace sites while preferentially decorating the Pt-step sites, could be a key factor in the pore formation observed experimentally in the Pt-InOxHy catalyst during pretreatment. (7) Our phase diagram further predicts that in addition to temperature, a precise control of O2 and H2 environments in the reactor during pretreatment is essential to drive this thermodynamically favorable site-selective reconstruction of InOx on the Pt surface. This analysis showcases the potential effectiveness of a combined analysis of the atomistic phase diagram with experimental characterization to elucidate the active site structure of complex metal–metal oxide interfaces. To further predict the dominant phase under reaction conditions, we assume that the H2 partial pressure is high enough (>10–4 kPa H2) to maintain the three-coordinated In3O7H7 structure on Pt(322) as the active catalytic phase during the ODHP reaction. Using this structure, we evaluate the thermodynamics and kinetics of the ODHP reaction networks to understand the role of the InOxHy overlayer in driving the catalytic reaction.
In summary, this computational investigation reports the ALD-deposited structure of InOxHy on the Pt surface at high temperatures, offering insights into the relationship between the synthesis conditions and the resulting surface morphology. To the best of our knowledge, this represents the first study reporting the atomic structure of InOxHy on a defective Pt surface and demonstrates qualitative agreement with the experiments. While this work employs thermodynamic analysis for structure prediction, it is crucial to understand that for other systems, kinetic effects can also play an important role. Nevertheless, the current approach provides an initial framework for elucidating the dynamic behavior of ALD-deposited metal oxides on a metal through atomic-level insights into metal–metal oxide interfaces.
Oxygen Activation on the Pt-InOx Catalyst
The reaction mechanistic analysis of the ODHP reaction on the Pt-InOxHy catalyst with the identified stable In3O7H7 phase is performed in two stages. First, the SHC mechanism is analyzed on the Pt-InOxHy catalyst, followed by the analysis of the PDH mechanism. Evaluating oxygen activation on the Pt-InOxHy structure is crucial for understanding the role of InOxHy in coupling the PDH reaction with SHC to maintain propylene selectivity and minimize overoxidation of PDH reaction intermediates. (23) The activation of oxygen on Pt(322)@In3O7H7 is assessed based on the binding energy trends of oxygen adsorption and water formation on the catalyst, as shown in Figure 3. The SurfGraph algorithm is then used to systematically place the oxygenated reaction intermediates (O* and OH*) at all unique possible sites on the Pt(322)@In3O7H7 surface, (37) and the binding energy is computed at 723 K using oxygen at 5 kPa and hydrogen at 1.32 kPa, respectively (Supporting Information, 1.2). Based on Figure 3, the oxygen gas molecule is dissociatively activated on the Pt-InOxHy surface with a kinetic barrier of +0.53 eV to form O*, assuming O2* adsorption from O2(gas) has no significant thermodynamic barrier on the Pt surface (Supporting Information, 4.1). The most stable configuration for O* adsorption is at a 3-fold site formed by a Pt-step atom and two Pt-terrace atoms, as shown in inset b of Figure 3. The binding energy of O* adsorption is predicted to be −0.48 eV, indicating thermodynamically favorable oxygen adsorption. While step edge sites typically enhance the reactivity of oxygen activation, (43) the presence of InOxHy prevents potential overbinding of O* on under-coordinated Pt sites and facilitates the removal of O* as H2O(g), as shown in the following analysis.
Figure 3
Figure 3. Binding energy trends to understand oxygen activation and the SHC mechanism on the Pt(322)@In3O7H7 catalyst. Here, +0.53 eV represents the kinetic barrier for oxygen molecule dissociation. Insets (a)–(e) represent the stable structures of Pt-InOxHy with their respective adsorbates mentioned in the inset text.
With H* available on the Pt surface from the dehydrogenation of hydrocarbon intermediates, O* combines with H* to form OH* at the Pt sites with a modest thermodynamic penalty of +0.3 eV (inset c of Figure 3) relative to O* on Pt(322)@In3O7H7. Further, the most stable site for OH* adsorption is identified as the bridging In and Pt-site forming In3O8H8, as shown in inset d of Figure 3. This is stable by −0.2 eV relative to OH* binding on Pt-terrace sites alone (inset c of Figure 3). Hence, our analysis predicts that OH* initially forms on Pt sites through combustion of H*, and InOxHy would then consume OH* to form a stable hydroxylated surface. With OH* bridging Pt–In sites, the formation of water occurs with thermodynamic stability of −0.17 eV, relative to In3O8H8 (regenerating the In3O7H7 active site) as shown in insets d and e of Figure 3, and by −0.3 eV relative to the formation of water from OH* on the Pt-terrace site (Supporting Information, 4). Therefore, InOxHy facilitates water formation with minimal thermodynamic and kinetic barriers, elucidating the mechanism for its high selectivity toward the SHC reaction. (44) Comparison of water abstraction at different InOxHy sites is further discussed in the Supporting Information, Section 4. Further, the current analysis reveals that InOxHy plays an important role in reducing the amount of O* on the Pt surface, thus controlling overoxidation at the Pt sites. In summary, the mechanistic analysis of SHC reveals that it is a dual-site mechanism. O* activates on Pt sites, subsequently gets adsorbed at the InOxHy active sites in the form of OH*, and the initial structure is restored when the additional OH* group leaves as water, completing the SHC catalytic cycle.
Thermodynamics and Kinetics of ODHP on the Pt-InOx Catalyst
The Pt-InOxHy catalyst facilitates a facile SHC reaction with a low activation barrier for formation of O* and subsequent reduction to H2O. We now present a thermodynamic and kinetic analysis of the PDH reaction network on the Pt-InOxHy catalyst. To understand the role of InOxHy in selectively driving the PDH reaction, a comparative reaction mechanistic analysis is performed on three different catalyst models: terrace-like Pt(111), stepped Pt(322), and Pt-InOxHy catalyst surfaces. For each system, stable adsorbate configurations are identified for key reaction intermediates, and the corresponding free energies and dehydrogenation barriers are estimated and reported in Figure 4.
Figure 4
Figure 4. (a) Free energy diagram for propane dehydrogenation on Pt(322)@In3O7H7, Pt(111), and Pt(322) catalysts. (b) Thermodynamically stable configurations of key reaction intermediates of propane dehydrogenation on Pt(322)@In3O7H7, Pt(111), and Pt(322). Here, blue, light gray, brown, red, gray, and white correspond to Pt-step, Pt-terrace, indium, oxygen, carbon, and hydrogen atoms, respectively.
Stable configurations of the intermediates, 1-propyl*, propylene*, propenyl*, and propyne* are systematically identified by enumerating the adsorbate configurations on all unique Pt sites using the SurfGraph algorithm. (37) Figure 4b presents the most stable adsorption configurations of these intermediates on the three surfaces. The preferred binding sites for 1-propyl*, propylene*, 1-propenyl*, and propyne* adsorption on all of the surfaces are consistently found to be on-top Pt, on-top Pt–Pt, bridge-top Pt–Pt, and bridge–bridge Pt–Pt sites, respectively. The adsorbates on the Pt(322) surface occupy the step sites due to their strong binding nature. (31) In contrast, on the Pt-InOxHy surface, the adsorbates occupy the Pt sites adjacent to the step sites as the step sites are saturated by the InOxHy layer. Adsorption of all intermediates is weaker on Pt-InOxHy and Pt(111) compared to Pt(322) due to the blockage of highly reactive step sites on the Pt-InOxHy surface (Supporting Information, 7.1 ). Electronic structure analysis of Pt sites occupied by the InOxHy layer shows that the d-band center shifts toward lower energy relative to the Fermi level, indicating decreased reactivity for adsorbate binding at those sites compared to Pt(322) (Supporting Information, 5).
A standard free energy diagram is then constructed and is shown in Figure 4a. It consists of the energetics of the dehydrogenation reaction, derived from the most stable adsorption sites of the PDH intermediates, and the reaction barriers for key elementary reaction steps, with propane gas as the reference. All of the intermediates on Pt(111) and Pt-InOxHy show comparable thermodynamic free energies, suggesting similar adsorption thermodynamics on terrace sites. All the intermediates on Pt(322) exhibit more negative thermodynamic free energies due to their strong binding at the defect sites. The free energy of propylene* is a key feature governing the propylene gas desorption. On Pt-InOx and Pt(111) catalysts, the thermodynamic formation of propylene gas from propylene* is downhill by 0.46 and 0.36 eV, suggesting favorable propylene desorption. However, on Pt(322), the propylene gas formation is uphill by 0.35 eV, predicting the requirement of additional energy for propylene desorption from the stepped surface. Similarly, the thermodynamic free energies of propenyl* on Pt-InOxHy and Pt(111) catalysts are slightly higher than propylene gas formation by 0.22 and 0.13 eV, respectively, while the free energy of propenyl* formation on Pt(322) is lower than propylene gas formation energy by 0.4 eV, indicating higher propensity for deep dehydrogenation to propyne* over propylene desorption on Pt(322).
The kinetic analysis of different elementary steps in the PDH reaction also shows similar activation barriers for the Pt-InOxHy and Pt(111) catalysts. The propane* activation or the first C–H bond breaking step in PDH is identified as the kinetically rate-controlling step on Pt-based catalysts. (31,45) Comparison of the first dehydrogenation barriers on Pt-InOxHy, Pt(111), and Pt(322) in Figure 4a reveals that the propane* activation barrier is approximately equal on Pt(111) and Pt-InOxHy (Pt-InOxHy lower by 0.1 eV) due to similar active sites for adsorption, with Pt(322) exhibiting a lower barrier due to adsorption at reactive step sites. The dehydrogenation barriers of 1-propyl* and propylene* are also approximately equal on Pt(111) and Pt-InOxHy due to similar potential energy barriers for C–H breaking of 1-propyl* (+0.78 eV on Pt-InOxHy and +0.8 eV on Pt(111)) and propylene* (+0.89 eV on Pt-InOxHy and +0.78 eV on Pt(111)). On Pt(322), lower dehydrogenation barriers are predicted as a result of the catalytic activity of step edge sites for C–H and C–C bond breaking. (46) The activation energy difference between propylene dehydrogenation and propylene desorption serves as a selectivity descriptor on Pt metal and its alloys. (46) Comparing the propylene* dehydrogenation barriers with the propylene* desorption barriers on Pt-InOxHy and Pt(111) catalysts reveals that propylene* desorption is kinetically favorable on both catalysts, with approximately equal desorption barriers (Pt-InOxHy lower by 0.1 eV) on Pt-InOxHy and Pt(111). In contrast, Pt(322) has a higher desorption barrier, indicating low selectivity toward propylene gas formation (Supporting Information, 7.3 ). In summary, the presence of the InOx layer modifies the adsorption thermodynamics and kinetics of PDH intermediates on Pt sites, enabling the exposed Pt sites to exhibit Pt(111)-like behavior and promoting favorable propylene formation and desorption.
Analyzing the role of indium oxide as an active site for PDH suggests that the C–H bond breaking barrier of 1-propyl* at the lattice oxygen sites is kinetically uphill by +2.5 eV, indicating unfavorable dehydrogenation on the InOxHy layer (Supporting Information, 7.6 ). This is in line with the previous analysis where In2O3 has been proven to be a better catalyst for the SHC reaction due to its preferential selectivity for hydrogen combustion over hydrocarbon conversion. (44) Therefore, this detailed analysis shows that the Pt-InOxHy enables PDH reaction at the Pt-terrace sites with improved selectivity and SHC reaction at combined Pt-InOxHy sites, controlling overoxidation of hydrocarbon intermediates at the Pt sites. The InOxHy overlayer, therefore, plays a multifaceted role in the removal of H* formed during PDH while promoting a facile pathway for H2O formation through SHC and maintaining the desired coupling between PDH and SHC reactions.
Discussion
Click to copy section linkSection link copied!
Through our results, we have demonstrated that the probable active site structure of the Pt-InOxHy catalyst is in the form of In3O7H7, which forms a continuous chain of a hydrogen-bonded network at the step sites. The proposed model predicts approximately ∼50% loss of majorly under-coordinated Pt sites, in line with the experimentally published results by Yan et al. (7) Further, our mechanistic analysis shows that the OH* network plays a critical role in the ODHP reactions and protects the catalyst from deactivation. In this section, we aim to understand the role of InOxHy in stabilizing the catalyst against deep dehydrogenation and coke deposition at the active site. Propyne* is a deep-dehydrogenated product of the PDH reaction, which has been proposed to be the starting point for the formation of coke and other undesired products since it preferentially cleaves the C–C bond over the C–H bond. (46) The C–C scission of propyne* has been proposed to be the rate-controlling step for side reactions on the Pt catalyst. (47) Therefore, understanding propyne* formation is crucial in assessing catalyst stability.
To specifically probe the effect of InOxHy in stabilizing the catalyst against coking, we analyze the formation energies of propyne* on stable Pt-InOxHy surfaces with varying O concentrations in the InOxHy overlayer, as shown in Figure 5a. The O concentration in the InOx overlayer directly correlates with the O-chemical potential as shown in the phase diagram analysis of stable InOx structures at different O and H concentrations generated through the procedure described in Section 1 (Supporting Information, 6). At each O concentration, we selected the most stable InOxHy configuration and analyzed the propyne* binding energy trends relative to the bare Pt(322) surface. The choice of the x-axis, which is the oxygen-to-indium ratio in Figure 5a, links the O2 gas concentration (O-chemical potential) in the reactor to the O/OH coverages in the Pt-InOxHy structures. The images in Figure 5b show the corresponding propyne* configurations on Pt-InOxHy structures at different O concentrations, estimated by enumerating propyne* at all unique Pt sites for each Pt-InOxHy structure using SurfGraph algorithm. (37) As shown in Figure 5a, with an increase in the O concentration in the InOxHy overlayer, the formation energy for propyne* becomes unstable compared to the bare Pt(322) surface. This results from the decreasing availability of Pt-step sites with an increase in the O concentration in the InOxHy overlayer, as shown in the images (1–6) of Figure 5b. At lower O concentration, propyne* occupies the step sites, as shown in Figure 5b images 1–2, with a binding energy of +0.25 eV relative to the Pt(322) surface. This slight decrease in binding energy is attributed to a decrease in the surface energy of the stepped Pt atoms due to the adjacent presence of the InOx overlayer. A further increase in the O concentration results in an increase in the binding energy to +0.6–0.7 eV. This results from the complete blockage of the Pt-step sites with oxygenated groups of InOx such as O*/OH* leading to the shift in propyne* binding to well-coordinated Pt-terrace sites as shown in Figure 5b images 3–6. Therefore, with step sites fully occupied by the chain of H-bonded network during the reaction, the most stable Pt(322)@In3O7H7 stabilizes the catalyst by blocking the highly reactive Pt-step sites that are conducive to C–C bo