Propylene is a building block of chemical industries used to produce many chemicals such as polypropylene, propylene oxide, acrylonitrile, and acrylic acid. Propylene is conventionally obtained by steam and fluid catalytic cracking of naphtha and light diesel (Sattler et al., 2014). However, these methods are thermodynamically limited, give low yield, and have complex separation of the desired products (Che-Galicia et al., 2014). Hence, searching for an alternative method is an urgent task. The large reservoir of shale gas worldwide opens a new channel to provide abundant light alkanes such as propane (Abedin et al., 2020). Under this circumstance, propane direct dehydrogenation (PDH) becomes an alternative and profitable catalytic method of producing propene.

There are two major routes of PDH: non-oxidative and oxidative dehydrogenation. The difference between these two routes is whether an oxidant, such as oxygen, is employed in the reaction. The non-oxidative route suffers from two inherent drawbacks, i) the endothermic nature makes it economically unfavorable in that it demands high energy input, and ii) the fast deactivation of the catalyst due to coke formation (Larsson et al., 1996; Lian et al., 2021). The use of oxidants in PDH overcomes the thermodynamics restraint and lowers the reaction temperature (Atanga et al., 2018). More importantly, coke formation is inhibited under oxidizing conditions (Carrero et al., 2014). However, the strong oxidizing ability of oxygen molecules induces the deep reaction of propene, which consequently reduces selectivity. The overoxidation issue is significantly lessened by using “soft oxidants,” including CO₂, N₂O, sulfur, and halogens, in propane dehydrogenation (Jiang et al., 2021; Zhi et al., 2022). Compared to an oxygen molecule, soft oxidants possess a much-reduced oxidizing ability that is detrimental to side reactions.

Among the soft oxidants, halogens and their halides are the most promising oxidants for propane dehydrogenation. Halogens (X) can assist the PDH pathways i) by molecular halogens (X₂), ii) oxyhalogenation (X₂, O₂), iii) oxyhalogenation via halides (HX, O₂), and iv) oxyhalogenation via molten salts of metal halides (LiX, O₂). Halogens as an oxidant not only suppress the reactivity toward side reactions but also give a new active site for C-H bond activation. Therefore, the use of halogens over metal oxides and phosphate surfaces (CeO₂, EuOX, CrPO₄, and FePO₄) shows a remarkable propene selectivity of up to 95% with 70% propane conversion at 500 °C compared to conventional catalysts (Xie et al., 2018; Zichittella et al., 2019a; Zichittella et al., 2020a).

CeO₂ is extensively studied due to its activity toward HX oxidation, alkane, and alkene oxyhalogenation (Scharfe et al., 2020). Xie et al. reported the different transitions and rare metal oxides for PDH in the presence of chlorine. The study shows the instability of Fe₂O₃, CuO, and NiO, whereas the high reactivity of oxygen leads to CO_x as a major product on the RuO₂ surface. Among them, a high propylene selectivity of 55% with 38% propane conversion was observed on the CeO₂ surface in the presence of a halogen (Xie et al., 2018). The high conversion of oxygen is also found on the ceria surface, which leads to increased CO₂ selectivity of up to 20%. The formation of CO₂ can be reduced by the addition of dopants to the ceria surface. The nickel-doped ceria decreased the selectivity of CO₂ to 15%. Furthermore, the nickel-doped ceria shows the highest selectivity of propylene at 80%, while the propane conversion reached 69%. In comparison, the undoped CeO₂ gives only 55% selectivity with 38% propane conversion (Xie et al., 2018). The surface modification of CeO₂ with doped metals makes a significant change in the catalytic performance (Su et al., 2018; Yang et al., 2018). However, an understanding of this enhancement induced by dopants is still absent, even though doped ceria has attracted much attention in oxidative dehydrogenation (Yang et al., 2018). Ceria doped with transition metals is an attractive approach that gives a balanced activity and selectivity for propane dehydrogenation. Our previous work identified the role of HBr in PDH for pristine CeO₂ catalysts. This study revealed that the presence of HBr facilitated the C-H bond activation and altered the conventional reaction pathway. The apparent activation barrier is reduced with the presence of HBr (Jan et al., 2022).

In the current work, we further deepen the understanding of HCl-assisted propane dehydrogenation on metal-doped ceria. The dopant candidates include V, Mn, Fe, Co, Ni, Pd, Pt, and Cu. The geometry and electronic properties of doped ceria are carefully examined and compared. The electronic structure analysis indicated the pronounced effects induced by dopants and their implications on reactivities. The adsorption of reactants and reaction pathways are carefully examined. The thermodynamic and kinetic parameters from DFT calculations are streamed into the microkinetic simulation, and the most promising dopant candidates are proposed and verified. The current work provides a practical optimization method to further improve performance and reveal the origin of enhancement from the doping strategy.

The reported work was performed using the spin-polarized density functional theory (DFT) method with the on-site Coulomb interaction (DFT + U) using the Vienna ab initio simulation package (VASP) (Kresse and Hafner, 1993; Kresse and Furthmuller, 1996). The DFT calculations were performed on the most stable CeO₂ (111) surface according to the literature (Wang et al., 2013; Wang et al., 2015). A plane-wave basis set was used for valance electrons with a cutoff energy of 400 eV. The projector augmented-wave (PAW) method was used to describe ionic cores (Blöchl, 1994; Kresse and Joubert, 1999). The revised Perdew–Burke–Ernzerhof (RPBE) function was used as an exchange-correlation functional (Hammer et al., 1999). The Brillouin-zone integration was performed at a 1 × 2 × one Monkhorst–Pack k-point grid for transition-metal-modified M-CeO₂ (111) surfaces (M = V, Mn, Fe, Co, Ni, Pd, Pt, and Cu). The van der Waals correction was performed by the DFT-D3 method (Grimme et al., 2010). For the description of cerium 4f, localized electrons were treated according to Dudarev et al. (1998). The value of U for this calculation was selected as 5.5 eV, as previously determined (Jan et al., 2022). The calculation was converged for all optimized structures to an accuracy of 1 × 10⁻⁵ eV, and force tolerance is 0.03 eV/Å.

A nine-atomic-layer slab with a 4 × 3 supercell in which the bottom six layers are kept fixed is used to model the ceria surface. The thickness of the vacuum was taken as 15 Å in the z direction. One cerium atom is substituted by transition metal M (M = V, Mn, Fe, Co, Ni, Pd, Pt, and Cu), as shown in Figure 1. The adsorption energy E_ads was calculated as E a d s = E a d s o r b a t e / s l a b − E s l a b + E a d s o r b a t e , (1)

in which E _{adsorbate/slab} represents the total energy of interaction species with the slab, E _slab is the individual energy of the CeO₂/M-CeO₂ surface, and the adsorbate represents the isolated molecules of C₃H₈, O₂, and HCl, respectively. The climbing image nudged elastic band (CINEB) method was used to compute the reaction path (Henkelman et al., 2000). The true transition state was verified by frequency calculation on each surface.

The bonding strength is analyzed with crystal orbital Hamilton population (COHP) analysis (Dronskowski and Bloechl, 1993). The microkinetics analysis was performed by the MKMCXX simulation package (Filot et al., 2014). It is noted that microkinetic simulation is a mean-field method that discarded surface heterogeneity and dynamics evolution. Although there are some deficiencies, it is still a cost-effective way to mimic surface catalytic reactions. The Gibbs free energy with zero-point energy correction (ZPE) was calculated from DFT frequency calculations at 773 K. The ideal gas approximation was used for the entropy calculation of gas-phase species, while harmonic approximation was used for surface species. The reaction rates for adsorption/desorption species were calculated using Hertz–Knudsen kinetics. The reaction rate for elementary reactions is given by a set of ordinary differential equations as r j = k j ∏ i c i , (2)

where r _j is the rate of the elementary reaction j, k _j is the rate constant of j, and c _i represents the concentration of species i in a given reaction. The rate constant for the adsorption step is calculated as k a d s = P A 2 π m k b T . (3)

In the equation, P denotes the partial pressure of the gaseous molecule, A is the surface area on which the adsorbed molecule is present, m denotes the mass of the reactant, and T and k _b are the temperature and Boltzmann’s constant, respectively. However, for the surface reaction, the rate constant of the forward/backward reactions is calculated as k r = Q ≠ Q X   k b T h   exp ⁡ ⁡ − Δ E ≠ k b T , (4)

where X represents the reacting species, ≠ shows the transition state, k _r denotes the reaction rate for the forward/backward reaction, and Q and h represent the partition function and Planck’s constant, respectively.

Microkinetics simulation was performed based on DFT calculations to determine the influence of different reaction conditions on the reaction kinetics and evaluate the most favorable catalytic surface. The main elementary steps are divided into adsorption, desorption, and surface reactions, as listed in Supplementary Table S2. From DFT calculations, the Gibbs free energy, rate constant, and reaction order are calculated at 773 K and 1 bar. For all elementary steps, the rate constants with pre-factors are listed in Supplementary Table S2 of the supporting information file.

The relationship between the reaction rates with the partial pressure of the reactant is shown in Figure 7. The investigated catalysts exhibited different dependencies on reactant pressure. Pd-, Ni-, Pt-, and Co-doped ceria have a strong positive correlation of TOF with pressure. TOF has a sharp increase with the pressure increase, while the other catalysts have a rather mild TOF increase.

The adsorption of propane and HCl regulates the performance of investigated catalysts, as shown in Figure 8. From the heat map, the modest binding energies of propane and HCl yield the best performance, which is around −0.3 eV for both cases. It is noted that Ni, Pt, and Pd possessed a bigger TOF than the other doped and pristine CeO₂ samples. This is consistent with previous electronic structure and pathway analysis. DRC analysis was performed to identify the critical step among the elementary steps. The DRC coefficient for C₃H₇* formation was found to be unity for all investigated catalysts, as shown in Supplementary Table S3.

Therefore, it is verified that the formation of C₃H₇* is the rate-determining step. C₃H₇* is formed by the abstraction of the first hydrogen from a secondary carbon atom. Further analysis indicates that the reaction follows the first-order kinetics to C₃H₈, which further confirms that the reaction kinetics are directly linked with the propane concentration.

In conclusion, the modification of ceria with metal dopants for PDH in the presence of HCl was studied with DFT-based microkinetics simulation. The charge analysis clearly indicated that dopants caused the charge transfer and created local polarization that increased the reactivities accordingly. The surface oxygen is identified as the active site to abstract hydrogen, and its activities are tuned by different dopants. The reaction starts with the spontaneous dissociation of the HCl molecule. The propyl species spontaneously attacks the adsorbed Cl* and gives the intermediate C₃H₇Cl* that, upon dissociation, results in the formation of propene. The catalytic performance of Ni- and Pd-CeO₂ is found to be very promising for PDH. The TOF of propane conversion and propene formation is enhanced and directly related to the partial pressure of propane. DRC coefficient is found as unity for the formation of C₃H₇*, which confirmed that the first hydrogen removal is a rate-determining step on all catalytic surfaces. The reaction follows the first-order kinetics to C₃H₈. Overall, the current study provides unique insight into the reaction mechanism of HCl-assisted propane dehydrogenation and paves the way for future optimization.