Insight into mechanism and selectivity of propane dehydrogenation over the Pd-doped Cu(111) surface

Abstract

The catalytic activity and selectivity of the Pd-doped Cu(111) surface toward the dehydrogenation of propane have been explored by density functional theory calculations with periodic boundary conditions. Four models with different Pd ensembles are introduced to represent the Pd/Cu(111) surface, where a surface single-atom catalyst model is built by alloying Cu with Pd. Calculations reveal that the d band center of the surface Pd atom is upshifted with the increased number of Pd atoms, resulting in an enhanced adsorbate–surface bonding strength and a reduced dehydrogenation barrier. The embedded Pd atoms can significantly improve the catalytic reactivity of the pure Cu surface, whereas the presence of the relatively inactive Cu surface is beneficial for the high selectivity toward propylene dehydrogenation. In general, the Pd/Cu(111) surface with the atomically dispersed palladium catalytic centers demonstrates good balance between the activity, selectivity, thermal stability and the maximum use of the noble metal, and shows great potential in the catalytic production of light olefins.

---

1 Introduction

Light olefins, such as ethylene and propylene, are important basic raw materials in the chemical industry. With the growing demand in the global chemical market, the catalytic production of such olefins has attracted increased interest in the past few decades,¹ for which the non-oxidative and oxidative dehydrogenations of light alkanes have been widely adopted as the dominant commercial pathways. Taking the production of propylene as an example, its non-oxidative dehydrogenation process is highly endothermic and the standard enthalpy of formation for dehydrogenating propane at 298.15 K is about 29.6 kcal mol⁻¹,² which requires a considerably high energy input. More precisely, the dehydrogenation reaction of propane takes place at temperatures of 525–575 °C,³ and in order to achieve a higher conversion rate (≥50%), even higher temperatures (550–750 °C) are needed¹ if the pressure is presumed to be 1 bar. Naturally, the much increased temperature can also promote undesired side reactions, such as the deep dehydrogenation and cracking reactions. It is thus highly desirable to find a balanced way to promote the expected main dehydrogenation reactions and simultaneously suppress unwanted side reactions.

Currently, noble metal-based and metal oxide-based catalysts show high potential for the non-oxidative dehydrogenation of propane,¹ but both of them have their own drawbacks. For example, Pt-based catalysts have superior high activity for propane dehydrogenation, however, the existing large Pt aggregates are also quite active for propylene dehydrogenation and cracking reactions. The latter reactions may produce carbon deposits, leading to a decrease of the propylene yield and the deactivation of the catalyst. For metal oxide-based catalysts,^4–7 the change of the oxidation state of the metal, as well as the involvement of the oxygen atoms, can affect the catalytic process. For instance, the first detached H atom from propane may form a hydroxyl group with nearby lattice oxygen, and then the subsequently dissociated H atom could either interact with the nearby lattice oxygen to produce a new hydroxyl group or yield a water molecule with the newly-formed hydroxyl group. The desorption of the resulted water molecule can take the lattice oxygen away from the surface to form an oxygen vacancy, which may change the oxidation state of the metal to some extent and consequently affect the activity of the catalyst. Additionally, the supplement of surface lattice oxygen from the sub-layer lattice oxygen or the O₂ environment may also hinder the subsequent dehydrogenation process.

Presently, a seemingly effective way to modify the catalytic properties of metal-based catalysts is to introduce a secondary metal to form a bimetallic catalyst. Due to the changes in the electronic and geometrical structure, the bimetallic catalysts usually exhibit enhanced activity and selectivity.^8–14 Recent studies have shown that alloying the late transition metal Cu or the main-group metal Sn into Pt can improve the selectivity of the catalyst in propane dehydrogenation.^15–17 Although the addition of a promoter leads to a relatively low percentage of Cu or Sn, the host metal is still the noble metal Pt with high costs for the production. The question of how to design a low-cost catalyst with high activity and selectivity has become an urgent matter. The newly developed concept of single-atom catalyst (SAC) could be a good solution, in which the use of noble metal in the catalyst and the activity, as well as selectivity, in catalysis are well balanced.^18–23

Recently, we designed a Pd-doped Cu SAC nanoparticle, represented by a 55-atom icosahedral cluster, to examine its catalytic ability towards two types of heterogeneous catalytic reactions: H₂ dissociation²⁴ and the dehydrogenation of propane to propylene.²⁵ It was clearly demonstrated that such a SAC can not only promote the aforementioned catalytic reactions, but also reduce the unwanted side reactions effectively due to the presence of relatively inactive Cu surfaces. However, a notable distortion of the Pd-doped Cu cluster was found at the high temperature of 527 °C from ab initio molecular dynamics simulations, which may reduce the catalytic performance of the SAC in practical applications. Compared with the nanoparticle, the alloyed metal surface is much more stable under high temperatures. By now, the Pd-doped Cu(111) surface had been studied by several groups.^26–28 However, all these studies only focused on the H₂ dissociation reaction, and its catalytic performance for other reactions has not been explored. Stimulated by the novel catalytic performance of the Pd-doped Cu SAC nanoparticles, we further explored the catalytic activity and selectivity of the Pd-doped Cu surface for the dehydrogenation of propane by first-principles calculations in the present study.

On the basis of getting maximum use of the noble metal Pd, we firstly considered single Pd doping. To our best knowledge, the most favorable doping site is definitely to substitute Pd for a topmost surface Cu atom. Previous studies^24,28 have shown that the catalytic activity of the Pd-doped Cu cluster could be significantly improved if more Pd atoms are introduced into the subsurface sites to form layer–sublayer Pd ensembles. Accordingly, herein four typical doping patterns are extended to the surface and investigated thoroughly in our calculations. Such doping patterns were used by Fu et al.,²⁸ in which the Pd ensembles served as the most efficient active sites. The full dehydrogenation process from propane to propylene has been simulated by density functional theory calculations. The plausible reaction pathways and the corresponding transition states are predicted. For each Pd/Cu(111) surface, its catalytic activity and selectivity toward propane dehydrogenation have also been discussed. Additionally, we compared the catalytic performance of different Pd-doped Cu(111) surfaces with Pd-doped Cu clusters. In general, the introduction of Pd to the Cu(111) surface facilitates the initial C–H activation and accelerates the reaction rate of propane dehydrogenation. The weakened binding ability of Pd/Cu(111) lowers the propylene desorption barrier and simultaneously increases the energy barrier for propylene dehydrogenation. All of these factors ensure high selectivity toward propane dehydrogenation.

2 Methods and models

The spin-polarized DFT calculations were performed by the Vienna ab initio simulation package (VASP).^29,30 For the systems under investigation, the effects of spin polarization were found to be negligible. The interactions between the core ions and the valence electrons were described using the projector augmented wave (PAW) pseudopotentials.³¹ The exchange correlation effects were described by the GGA-PBE functional.³² A plane wave energy cutoff of 400 eV was used in the calculations. The long-range dipole correction was considered in the calculations to eliminate the spurious dipole–dipole interactions along the vertical direction.

To mimic the flat (111) surface, a four-layer slab of a (5 × 5) unit cell and a 10 Å-thick vacuum region were employed here. The periodic interaction was very small due to the separation spacing being large enough. Brillouin zone sampling was carried out using (5 × 5 × 1) Monkhorst–Pack grids. During the optimization, the lowermost two layers were fixed while the atoms in the uppermost two layers were allowed to relax until the maximum force became less than 0.02 eV Å⁻¹. The transition states of the propane dehydrogenation process were located by the climbing-image nudged elastic band (CI-NEB) method.³³ These obtained transition-state structures are characterized by frequency calculations, and they have only one imaginary frequency.

3 Results and discussion

3.1 Molecular adsorption of propane

In this study, effective Pd ensembles are taken into account in multiple Pd doping, i.e., the number of the subsurface Pd atoms that could directly connect with the surface Pd atom goes up to three. In total, there are four possible configurations for the Pd-doped Cu(111) surfaces. To better understand the impact of the doped Pd atom on the catalytic properties toward the propane dehydrogenation process, the pure Cu(111) surface is also considered for comparison. The optimized structures of the pure and the Pd-doped Cu(111) surfaces are shown in Fig. 1, and are labelled as a, b, c, d and e, respectively.

Fig. 1 Optimized structures of the pure (a) and the Pd-doped Cu(111) surfaces (b–e): (b) only one surface Pd atom; (c) one surface plus one subsurface Pd atoms; (d) one surface plus two subsurface Pd atoms; and (e) one surface plus three subsurface Pd atoms. Here, Cu atoms and Pd atoms are marked in yellow and blue, respectively.

The most stable configuration of propane adsorption on the catalyst surface generally has a parallel adsorbed state,^4,5,17,34,35 in which the intermediate methylene group is the nearest part of the molecular propane to the surface. The C–H bond cleavage could occur in the methyl or the methylene group of propane.

It should be noted that when the C–H activation starts from the terminal methyl group, the parallel adsorbed configuration may not be appropriate to serve as the initial state for the surface reaction, and it may evolve to a tilted or upright state to act as the precursor for the dehydrogenation process. In the Pd/Cu SAC nanoparticle,²⁵ the perpendicular adsorbed propane is the second most stable adsorption configuration, and the energy difference between the parallel and perpendicular states is 1 kcal mol⁻¹ by GGA-PBE calculations, and this value is even smaller in the Pd-doped Cu(111) surfaces. Such a small energy difference could make the perpendicular adsorbed state a possible reaction precursor in reality. Accordingly, we adopted these two adsorption configurations as the starting points for the two representative C–H activations to explore the dehydrogenation process on the Pd/Cu surface. The optimized structures for these configurations are given in Fig. 2, and the corresponding adsorption energies obtained by the PBE functional are listed in Table 1. Considering that the adsorption of molecular propane is usually a weak physisorption, we also employed the van der Waals density functional (vdW-DF) proposed by Dion et al.³⁶ to take into account the dispersion interaction. Previous calculations on the Pd/Cu cluster indicate that the vdW-DF single-point calculations based on the PBE-optimized configuration could predict the reasonable adsorption energies, and thus no further geometry optimization by the vdW-DF functional was carried out here. The corrected adsorption energies by vdW-DF are given in Table 1.

Fig. 2 Top and side views of two possible adsorption patterns of molecular propane on pure and Pd-doped Cu(111) surfaces. Here ∥ represents the parallel configuration while ⊥ represents the perpendicular configuration.

Table 1 Adsorption energies and equilibrium adsorption distance (H⋯Pd(Cu)) of propane on pure and Pd-doped Cu(111) surfaces (see Fig. 2). Here E_ad(propane) is defined as: E_ad = E(adsorbate/surface) − E(adsorbate) − E(surface)

Configurations	a∥	b∥	c∥	d∥	e∥	a⊥	b⊥	c⊥	d⊥	e⊥
Ead(PBE) (kcal mol^−1)	−0.7	−1.1	−1.6	−1.9	−2.1	0.2	−0.5	−0.8	−1.0	−1.2
Ead(vdW-DF) (kcal mol^−1)	−8.5	−10.0	−10.9	−11.0	−10.9	−6.8	−8.2	−7.9	−7.8	−8.0
H⋯Pd(Cu) (Å)	2.78	2.67	2.27	2.18	2.12	2.39	2.27	2.19	2.13	2.09

---

As shown in Table 1, the perpendicular adsorption states generally have slightly smaller binding energies than the parallel configurations, and all these adsorbed configurations are predicted to possibly exist by vdW-DF calculations, although the PBE-predicted adsorption energy on the pure Cu(111) surface is 0.2 kcal mol⁻¹. With the vdW correction, the PBE-underestimated adsorption energies are improved, and the calculated values range from −6.8 to −11.0 kcal mol⁻¹. We note that the adsorption energy increases with the increase of Pd concentration while the equilibrium adsorption distance decreases, indicating an enhanced binding interaction between the propane molecule and the surface. This enhanced binding strength originates from the upshift of the d band center ε_d. The famous d band theory^37–39 claimed that the higher the d band center ε_d, the stronger the adsorption. It is known that palladium has a higher ε_d while copper has a lower ε_d. Therefore, alloying copper with palladium can shift the position of ε_d upwards, which in turn causes an increased interaction between the resulting surface and the adsorbate. The calculated d band centers of the surface Pd atom in these Pd-doped Cu(111) surfaces are listed in Table 2. Clearly, the introduction of subsurface Pd atoms may modify the location of ε_d. The more the doped Pd atoms, the higher the calculated d band center. The correlation between the adsorption energy and the d band center is plotted in Fig. 5(a) and from it, one can see that the upshift of ε_d refers to an increment of the propane adsorption energy, which is in agreement with the d band rule.

Table 2 Energy barriers for propane dehydrogenation on pure and Pd-doped Cu(111) surfaces and the d band center of the Pd atom that bonds with the propane molecule (ε_d)

Surfaces	a	b	c	d	e
Step 1 (kcal mol^−1)	37.9	32.8	31.7	29.9	28.2
Step 2 (kcal mol^−1)	36.7	31.6	29.6	28.6	26.3
Step 3 (kcal mol^−1)		2.3	2.3	2.6	3.3
Step 4 (kcal mol^−1)		2.8	2.4	2.1	3.3
Step 5 (kcal mol^−1)		18.2	19.3	18.5	18.1
Step 6 (kcal mol^−1)		15.9	16.3	16.4	16.4
εd (eV)		−1.90	−1.73	−1.61	−1.55

---

3.1.1 C–H bond activation. Initial C–H bond breaking could either occur from the methyl or the methylene group of propane, giving corresponding products of 1-propyl or 2-propyl species, respectively. It is noted that the C–H bond dissociation energy (BDE) in the methyl group of molecular propane is 100.9 ± 0.5 kcal mol⁻¹, which is a little bit stronger than that in the methylene group (98.1 ± 0.7 kcal mol⁻¹);⁴⁰ therefore, the scission of C–H bond on the center carbon should be relatively more favorable than the terminal carbon from the perspective of thermodynamics. To better understand the propane dehydrogenation process, the aforementioned two C–H bond cleavage pathways are both included. In this study, we adopted a low-energy catalytic mechanism for the conversion of propane to propylene proposed in our previous work.²⁵ As Fig. 3 shows, the catalytic conversion of propane to propylene on the Pd/Cu(111) surface comprises three stages: (I) the initial C–H bond breaking to produce the propyl group; (II) the detached H atom migrating on the Cu surface; (III) and the subsequent dehydrogenation of the produced propyl group to form propylene. In stage (I), C–H activation in the methyl and methylene groups are involved, denoted as steps 1 and 2, respectively. The diffusion of the detached H atoms from steps 1 and 2 are then labeled as steps 3 and 4 in stage (II), respectively. In stage (III), we named the C–H activation in the 1-propyl and 2-propyl species as steps 5 and 6, respectively. On the basis of this scheme, two different dehydrogenation pathways have been thoroughly explored for the Pd/Cu(111) surfaces, and the corresponding transition states are shown in Fig. 4 and the related energy barriers for each step are tabulated in Table 2.

Fig. 3 Catalytic scheme for the catalytic conversion of propane to propylene.

In most cases, the first C–H activation is the rate-determining step for the whole dehydrogenation process, and it is crucial to reduce the barrier of this step here. To facilitate comparison, we also calculated the energy barrier for breaking the first C–H bond of the propane molecule catalyzed by the pure Cu(111) surface. It was well known that the pristine Cu(111) surface is relatively inert, and is hence less active in catalyzing the propane dehydrogenation reaction. From Fig. 4 (steps 1 and 2), it can be seen that the detached H atom from propane is located at the bridge site by elongating the C–H bond. The activated C–H bonds in steps 1 and 2 are stretched to 1.82 Å and 1.91 Å, respectively, which are much longer than the original C–H bond length in a propane molecule (1.10 Å). The calculated energy barriers for the C–H activation in the methyl and methylene groups are 37.9 and 36.7 kcal mol⁻¹, respectively, suggesting that the activation of the methylene group is easier than that of the methyl group of molecular propane. The following discussions will focus on this relatively feasible dehydrogenation pathway. When doping an individual Pd atom in the topmost layer of Cu(111), the energy barrier for step 2 is decreased by 5.1 kcal mol⁻¹. According to the Arrhenius equation, the change of 5 kcal mol⁻¹ in the activation energy can change the rate constant by about 23 times if we introduce an 800 K working temperature in the calculation. Clearly, the individual isolated Pd atom can effectively increase the dehydrogenation rate of propane. From Table 2, one can see that the value of the corresponding energy barrier becomes lower with the increasing number of subsurface Pd atoms, however the extent of decrease is not as remarkable as in the case of the single Pd doping on the topmost layer. As can be seen from Table 2, when three more Pd atoms are introduced into the subsurface to form a quaternary Pd ensemble, the corresponding energy barriers for steps 1 and 2 can be further reduced to 28.2 and 26.3 kcal mol⁻¹, which are ∼5 kcal mol⁻¹ lower than the single Pd doping. Obviously, the catalyst activity could be effectively improved by introducing a very small amount of the noble Pd element.

Fig. 4 Geometries of the transition states for the dehydrogenation of propane, H atom diffusion and the dehydrogenation of propyl group: (a) catalytic conversion starting from a perpendicular propane adsorption state (b) catalytic conversion starting from a parallel propane adsorption state.

Fig. 5(b–d) shows the plots of the transition state energy (E_TS) for step 1 against the d band center, the adsorption energy of parallel propane (E_ad) and the final state energy (E_FS) for this step on the Pd-doped Cu(111) surface; the reference energy is defined as the sum of the total energy of gas-phase propane and the bare Pd/Cu(111) surface. As shown in Fig. 5(b), the plot of the transition state energy against the d band center gives a straight line. A similar trend has been proposed by F. Abild-Pedersen et al.⁴¹ in a study of methane activation. This linear relationship demonstrates that the surface d band center determines the trend of the transition state energies of propane dehydrogenation catalyzed by the alloyed surface. The upshift of the d band center caused by the introduction of subsurface Pd atoms corresponds to a reduced transition state energy. In other words, higher the d band center, higher the catalytic activity. In Fig. 5(c and d), the correlations between the transition state energy and the adsorption energy of the initial and final states for step 1 also exhibit linear relationships. It is noted that the slope of Fig. 5(d) is very close to one and the obtained determination coefficient (R² = 0.982) is higher than that in Fig. 5(c), implying that the transition state in this step is more final state-dependent. As pointed out by F. Abild-Pedersen, the slope of the universal BEP relation is strongly coupled to the transition state geometry, and a more final state-like transition state leads to a slope close to one.⁴² Since all these three correlations give a linear relationship, the d band center, the adsorption energy (E_ad) and the final state energy (E_FS) could act as a general “descriptor” for the reactivity of the Pd-doped Cu(111) surface toward propane dehydrogenation.

Fig. 5 Correlation plots for the initial C–H activation: (a and b) plots of the propane adsorption energy (E_ad) and the transition state energy (E_TS) for step 1 against the d band center of the surface Pd atom (ε_d); (c and d) plots of transition state energy (E_TS) for step 1 against E_ad and final state energy (E_FS) for step 1. Here, E_TS, E_ad and E_FS are related to the initial state gas-phase energy.

As a good catalyst for the dehydrogenation reaction, accelerating the reaction rate is necessary, and also it should weakly bind intermediates such as the resulting H atom. According to the Sabatier principle, the best catalyst is a compromise which has a “just right” adsorbate–surface interaction.⁴³ An overly strong interaction would deactivate the catalyst due to the blocking of the active site by the adsorbate, while an extremely weak adsorption makes it difficult for the reactant to adsorb on the catalyst, and it cannot be activated effectively. Therefore, an intermediate strength is both beneficial for the activation and the following spillover process. Earlier works by Kyriakou et al.²⁶ and Tierney et al.²⁷ have shown that doping isolated Pd atoms into the Cu(111) surface does not alter its adsorption properties but can significantly improve its catalytic activity. The diffusion barriers for H atom migration along the single-Pd-doped Cu(111) surface were in the range of 2.5–4.6 kcal mol⁻¹ (ref. 24) from our previous theoretical studies, which also demonstrated that the H atom can readily spill over onto the Pd/Cu(111) surface. Based on these findings, the diffusion of detached H atoms from the dehydrogenation process on the host Cu(111) surface is presumed to be facile. Here the calculated results indicate that the value of the diffusion barrier for stage (II), i.e., the resulting H atom migrating across the Cu–Cu bridge along the single-Pd-doped Cu(111) surface (b), is indeed small (< 3.0 kcal mol⁻¹). When this diffusion occurs on the Pd-doped Cu cluster, it may experience a diffusion barrier of 3.8–4.3 kcal mol⁻¹.²⁴ Clearly, the movement and migration of the detached H atoms on the PdCu surface are relatively easier than those on the Pd/Cu₅₅ cluster. The optimized transition states for step 3 and step 4 are shown in Fig. 4. As Fig. 4 shows, the migrated H atom is positioned at the bridge site of the Cu–Cu bond, and the bond length between the Cu and H atoms is around 1.66 Å. It is interesting to find that the diffusion barrier for the detached H atoms doesn’t change too much with the increase of doping Pd atoms in the subsurface of Cu(111). This low diffusion barrier means that the detached H atom can easily migrate to a location away from the produced propyl species, and the forward dehydrogenation reaction could be consequently promoted. After the detached H atom is transferred to its adjacent site, the following step is the subsequent dehydrogenation of the newly-produced propyl group (stage (III)). The energy barriers for the dehydrogenation of 1-propyl and 2-propyl species on these PdCu surfaces are calculated to be 18–19 and ∼16 kcal mol⁻¹, respectively. Apparently, the activation of the second C–H bond from the propyl group is much easier than the initial C–H bond activation.

Once propylene is formed through the dehydrogenation of propyl species, two possible processes may occur. One is the direct desorption of propylene, i.e., the release of the generated propylene from this new coadsorption configuration, which is endothermic by ∼2.7 kcal mol⁻¹. The other one is H atom diffusion prior to desorption. Taking the dehydrogenation on the single-Pd-doped Cu(111) surface as an example, the second resulting H atom migrates to its adjacent site, leading to a more stable coadsorption configuration, and the new state is lower in energy than its original state by about 3 kcal mol⁻¹. The release of the generated propylene from this relatively stable configuration requires energy of ∼5.9 kcal mol⁻¹. Noted that the binding energy for the adsorption of an isolated propylene molecule on the single-Pd-doped Cu(111) surface is 7.2 kcal mol⁻¹; clearly, the coadsorbed H atoms weaken the bonding strength between the surface and the produced propylene, which in turn promotes the desorption process of propylene.

In order to better understand the dehydrogenation process of propane to propylene and the influence of consecutive Pd doping on this surface reaction, we summarized the relative energy profiles for the whole catalytic reaction on the Pd-doped Cu(111) surface in Fig. 6. From Fig. 6, one can clearly see that the rate-limiting step for the whole process is the initial C–H activation and the presence of the Pd ensemble may stabilize the transition states and lower the corresponding energy barrier. Although the introduction of subsurface Pd atoms promotes the reaction rate of propane dehydrogenation, it also increases the bonding strength between the surface and the produced propylene, which has a negative effect on the selectivity toward propylene production. Therefore, to evaluate the overall effect of the Pd ensembles on the catalytic propane dehydrogenation, the deep dehydrogenation and the cracking of propylene are also taken into account here.

Fig. 6 Low energy pathways for propane dehydrogenation to propylene on Pd-doped Cu(111) surfaces.

3.2 Selectivity toward propylene

During the propane dehydrogenation process, besides the desired product propylene, many byproducts such as ethylene, hydrocarbons and coke may be also formed due to side reactions such as deep dehydrogenation and C–C bond cracking, which can be caused by stronger propylene adsorption as well as a lower energy barrier for these aforementioned side reactions. The selectivity of this dehydrogenation reaction is determined by the competition between propylene desorption and the dehydrogenation of propylene, and therefore the energy differences between these two processes can be used as a selectivity descriptor. As mentioned before, the introduction of Pd increases the required energy for the release of propylene, which is an adverse factor for improving the selectivity toward propylene. How would the increased subsurface Pd atoms affect the possible side reactions? What is the net effect for propylene production? To address these questions, we studied propylene desorption and the possible side reactions.

It is known that the d states of the surface metal atom could be deactivated to some extent by the adsorbed H atoms,³⁸ which gives rise to weaker binding. Thus, one can expect that without the coadsorbed H atoms, the adsorption of an isolated propylene molecule on the single-Pd-doped Cu(111) surface should be stronger. Additionally, considering that the diffusion barrier for an H atom on the Pd-doped Cu(111) surface is very low, these detached H atoms could easily spill over to some locations which are far away from the C₃ intermediates and finally desorb from the surface as H₂. Accordingly, we can simply neglect these resulting H atoms and take the isolated propylene adsorption as the starting state for the possible side reactions.

3.2.1 Propylene desorption. On the Pd-doped Cu(111) surfaces, the top surface Pd is the active site. A favorable adsorption site on these surfaces for propylene is atop the Pd atom, which is also known as the π adsorption mode. On the basis of a structure with only one propylene π-adsorbed on the Pd-doped Cu(111) surface, we obtained the optimized configurations in Fig. 7, and the corresponding adsorption energies are given in Table 3. As shown in Table 3, the adsorption energy of propylene on the Pd-doped Cu(111) surface obtained by the GGA-PBE calculations increased from −7.2 to −12.4 kcal mol⁻¹ as the concentration of Pd increases, which can be explained by the upshifted d band center of the topmost Pd atom on these surfaces. The increasing binding strength hinders its desorption to a certain extent. By including the vdW correction, the adsorption energies are calculated to fall in the range of −16.6 to −19.7 kcal mol⁻¹. The C=C double bond length in these π-adsorbed propylene is around 1.38 Å which is almost identical to its gas state (1.34 Å), indicating that the C=C double bond is not activated. Generally, the energy required for the release of an adsorbed molecule from the surface is very close to its adsorption energy, and thus the desorption energy barriers for propylene release from the Pd-doped Cu(111) surface are approximately in the range of 7.2–12.4 kcal mol⁻¹. In comparison with the desorption barrier of 14 kcal mol⁻¹ for the single-Pd-doped Cu₅₅ cluster, propylene molecule desorption from the PdCu surface is much easier than that from the PdCu cluster. Therefore, the PdCu surface should be beneficial for propylene desorption in contrast with the PdCu cluster.

Fig. 7 Optimized adsorption configurations of molecular propylene on Pd-doped Cu(111) surfaces.

Table 3 Adsorption energies of propylene on Pd-doped Cu(111) surface

Surfaces	b	c	d	e
Ead(PBE) (kcal mol^−1)	−7.2	−9.5	−11.3	−12.4
Ead(vdW-DF) (kcal mol^−1)	−16.6	−18.5	−19.4	−19.7

---

3.2.2 Deep dehydrogenation and propylene cracking. In similarity to the dehydrogenation of propane, there are two possibilities for propylene dehydrogenation, and the C–H bond activation may either occur at the methylene or the methine group, which produces 1-propenyl (step 7) or 2-propenyl (step 8) species on the PdCu surface, respectively. In Table 4, we tabulate the calculated energy barriers for the C–H cleavage from propylene catalyzed by the Pd-doped Cu(111) surfaces. For these PdCu surfaces, the least energy required for the breaking of the C–H bond in propylene is 33.7 kcal mol⁻¹, which is much higher than that for propylene desorption, indicating that the desorption process would be more favorable during the two competitive reactions (Fig. 8).

Table 4 Energy barriers for C–H breaking and C=C cracking in propylene on Pd-doped Cu(111) surfaces

Surfaces	b	c	d	e
	33.8	34.2	33.9	33.0
	35.1	35.1	34.7	33.7
	60.9	60.9	60.7	61.2

---

Fig. 8 Geometries of the transition states for C–H breaking and C=C cracking in propylene on Pd-doped Cu(111) surfaces.

In propane dehydrogenation, the C–C bond cleavage could arise from propane, propyl and propylene, respectively. Among these three cracking pathways, the cleavage of the C=C bond is relatively more feasible from the point of view of energy,^17,34 which is taken as the representative for the C–C bond cracking. As shown in Table 4, the calculated energy barrier for the C=C bond cleavage of propylene (step 9) is around 61.0 kcal mol⁻¹, and this substantially high energy requirement demonstrates that this cracking is kinetically hindered and less possible, compared to the propylene desorption on the Pd-doped Cu(111) surface.

3.2.3 Selectivity. To evaluate the selectivity toward propylene on these Pd/Cu(111) surfaces, we summarize the energy differences between the possible propylene dehydrogenation/cracking and its desorption in Fig. 9. Obviously, all the obtained differences are larger than 20 kcal mol⁻¹, showing that the propylene desorption is greatly preferred among these competing processes. Although the energy difference becomes narrower with the increase of subsurface Pd atoms, the existing huge energy barrier difference will kinetically suppress the possible side reactions, which sustains a high selectivity toward propylene.

Fig. 9 Energy barrier difference (E_diff) between the possible side reactions and propylene desorption over the Pd-doped Cu(111) surfaces.

3.3 Comparison with other heterogeneous catalytic conversions of propane to propylene

It is known that the Pt-based catalysts possess high activity for many catalytic reactions, and are widely used as commercial catalysts for propane dehydrogenation in industry. The ‘true’ barrier to break the initial C–H bond cleavage in propane on Pt(111) is about 25 kcal mol⁻¹ from DFT calculations,^17,34 and this value would be even smaller if a stepped Pt(211) surface or small Pt clusters were introduced as the propane dehydrogenation catalyst. The key issue for these catalysts is their low selectivity toward propylene. To improve the selectivity of Pt-based catalysts for propane dehydrogenation, several approaches are proposed.^{7,15,17,25,44} One effective way is alloying Pt with Sn, and the addition of Sn lowers the propylene desorption barrier and meanwhile increases its energy barrier for propylene dehydrogenation.^17,44,45 Such positive effects finally improve the catalytic selectivity. A higher Sn content corresponds to a higher selectivity but a lower activity. Of these PtSn surfaces, the Pt₃Sn surface was reported to be an optimal trade-off between activity and selectivity, in which the initial C–H bond cleavage requires an energy of ∼26 kcal mol⁻¹, referenced to the adsorbed state of propane. Additionally, decreasing the ensemble of active metal centers is also an alternative method to improve the selectivity. Schweitzer et al.⁷ reported that the silica-supported single-site Zn(II) catalyst exhibits high selectivity for the dehydrogenation of propane to propylene. The rate-limiting step is the second C–H bond breaking, and the predicted barrier for this step is 49 kcal mol⁻¹, obtained by first-principles calculations based on the hybrid density functional (B3LYP). In addition, our previous studies demonstrated that doping Pd in Cu₅₅ nanoparticles can improve the selectivity and maintain a high catalytic activity.²⁵ However, its relatively lower thermal stability at high temperature is a hindrance for its practical application. By using the highly thermostable Cu(111) surface to be the host surface for the Pd doping, we constructed several PdCu surfaces with very low Pd concentrations. Compared with the PdCu nanoparticle with a high catalytic performance, the PdCu surfaces might sacrifice the activity to a lesser extent, but provide a substantially low energy barrier for both hydrogen diffusion and propylene desorption, which ensures a high selectivity toward propylene. Although the constructed PdCu surfaces show a relatively low catalytic activity compared to the PdCu nanoparticles for propane dehydrogenation, their catalytic ability, especially for four-Pd-atoms-doped Cu(111) surface, is still comparable or better than most of the reported catalysts. Meanwhile, we note that when the vdW correction is introduced to evaluate the propane dehydrogenation process, these conclusions are still valid. The corresponding energy barriers are given in Table S1.† The atomically dispersed embedded Pd doping allows a relatively high density of the palladium catalytic centers on the Cu surface, which can furnish enough active sites for propane dehydrogenation in practical applications. Considering the balance between the cost, stability, catalytic activity and selectivity, the Pd-doped Cu(111) surfaces are still promising candidates for propane dehydrogenation, although the least energy required for the initial C–H bond breaking in propane dehydrogenation catalyzed by the Pd-doped Cu(111) surfaces is around 26 kcal mol⁻¹, which is slightly higher than that required for some of the above-mentioned catalysts.

3.4 Stability

The stability of a catalyst is an important issue for its practical application, and doping processes to yield Pd/Cu(111) surfaces have been investigated theoretically. For the PdCu alloy, the Pd atom in the topmost layer has the potential to diffuse into the second layer during the annealing and equilibration process.^46,47 However, when adsorbates (especially strongly interacting adsorbates) are present, the Pd atom could be brought to the surface and the loss of Pd from the top layer to the second layer is reduced.^48,49 Additionally, the composition in the Pd–Cu(111) alloy system can be controlled by temperature, that is, a higher temperature during deposition leads to a higher Pd concentration in the subsurface region, whereas more Pd atoms would reside in the surface layer if the Cu(111) is deposited at a lower temperature.⁴⁷ When the deposition temperature is in the range of 290–475 K, surface and subsurface Pd atoms were both found in the Pd–Cu(111) alloy system experimentally,^47,50 suggesting that many factors may affect the composition of the top two layers of the Pd–Cu(111) alloy.

In general, the thermodynamic stability of a mixing or alloying bimetallic system can be evaluated by the excess energy, and a negative value of the excess energy represents a favourable mixing or alloying process.^51–53 In similarity to our previous study of Pd-doped Cu nanoparticles,²⁴ we defined the excess energy of the Pd-doped Cu surface in a vacuum (E_exc(Vac)) and under a propane atmosphere (E_exc(C₃H₈)) in eqn (1) and (2), respectively.

Here, , E_Pdsurf and E_Cusurf are the energies of the Pd-doped Cu(111), Pd(111) and Cu(111) surfaces, respectively. E(C₃H₈) is the energy of an isolated propane molecule, and is the energy of the Pd-doped Cu surface after the adsorption of one propane molecule in its most stable configuration; that is, the parallel adsorbed state. N_surf represents the number of the total metal atoms in the slab model while n is the number of the Pd atoms.

The calculated excess energies of E_exc(Vac) and E_exc(C₃H₈) are listed in Table 5. As Table 5 shows, all the predicted excess energies here are negative, indicating that these doping patterns are energetically favorable and the surrounding propane environment could further improve the stability of the Pd-doped Cu surface to a lesser extent. The increased stability makes the Pd ensemble models become possible for practical applications.

Table 5 The excess energies E_exc(Vac) and E_exc(C₃H₈) for the various Pd-doped Cu(111) surfaces

	Eexc(Vac) (eV)	Eexc(C3H8) (eV)
b	−0.69	−0.74
c	−1.12	−1.19
d	−1.46	−1.54
e	−1.74	−1.83

---

4 Conclusions

In summary, the catalytic activity and selectivity of the Pd-doped Cu(111) surfaces toward propane dehydrogenation have been systematically investigated based on first-principles calculations. The present calculations found that the dehydrogenation activity follows the famous d band rule, that an upshift of the d band center corresponds to enhanced binding between the surface and the adsorbates (intermediates and transition states), resulting in a low energy barrier for the dehydrogenation process. In addition, the introduction of a small amount of Pd not only lowers the energy barrier for propane dehydrogenation remarkably, but also the dehydrogenation barrier for propylene to a lesser extent. The obtained results reveal that the desorption barrier for the propylene being released from the Pd/Cu(111) surface increases as the Pd content increases. Considering all the aspects above, the introduction of Pd atoms into the Cu(111) surface can improve its catalytic activity remarkably, and the high selectivity is also maintained due to the predicted huge energy difference between propylene desorption and the possible side reactions. Although the predicted catalytic activity of the Pd/Cu(111) surface is a little bit lower than that of Pd-doped Cu₅₅ nanoparticles, its high thermal stability and selectivity makes it one of the most promising catalysts for propane dehydrogenation.

Acknowledgements

This work is supported by the Göran Gustafsson Foundation for Research in Natural Sciences and Medicine, the Swedish Research Council (VR), and the Fundamental Research Funds for the Central Universities of China (20720150215). The Swedish National Infrastructure for Computing (SNIC) is acknowledged for the supercomputer resources. Many thanks are due to Prof. Yi Luo from the Royal Institute of Technology for useful discussions.

References

1. J. J. H. B. Sattler, J. Ruiz-Martinez, E. Santillan-Jimenez and B. M. Weckhuysen, Chem. Rev., 2014, 114, 10613–10653.
2. D. Lide, CRC Handbook of Chemistry and Physics, Taylor and Francis, 85th edn, 2004.
3. H. You, Pet. Sci. Technol., 2005, 23, 1441–1452.
4. H. Fu, Z.-P. Liu, Z.-H. Li, W.-N. Wang and K.-N. Fan, J. Am. Chem. Soc., 2006, 128, 11114–11123.
5. J. M. H. Lo, Z. A. Premji, T. Ziegler and P. D. Clark, J. Phys. Chem. C, 2013, 117, 11258–11274.
6. Y. Liu, Z. H. Li, J. Lu and K.-N. Fan, J. Phys. Chem. C, 2008, 112, 20382–20392.
7. N. M. Schweitzer, B. Hu, U. Das, H. Kim, J. Greeley, L. A. Curtiss, P. C. Stair, J. T. Miller and A. S. Hock, ACS Catal., 2014, 1091–1098.
8. J. R. Kitchin, J. K. Nørskov, M. A. Barteau and J. G. Chen, Phys. Rev. Lett., 2004, 93, 156801.
9. M. Chen, D. Kumar, C.-W. Yi and D. W. Goodman, Science, 2005, 310, 291–293.
10. C. T. Campbell, Annu. Rev. Phys. Chem., 1990, 41, 775–837.
11. N. Toshima and T. Yonezawa, New J. Chem., 1998, 22, 1179–1201.
12. Y. Lei, B. Liu, J. Lu, R. J. Lobo-Lapidus, T. Wu, H. Feng, X. Xia, A. U. Mane, J. A. Libera, J. P. Greeley, J. T. Miller and J. W. Elam, Chem. Mater., 2012, 24, 3525–3533.
13. D. A. Hansgen, D. G. Vlachos and J. G. Chen, Nat. Chem., 2010, 2, 484–489.
14. Y.-T. Liang, S.-P. Lin, C.-W. Liu, S.-R. Chung, T.-Y. Chen, J.-H. Wang and K.-W. Wang, Chem. Commun., 2015, 51, 6605–6608.
15. Z. Han, S. Li, F. Jiang, T. Wang, X. Ma and J. Gong, Nanoscale, 2014, 6, 10000–10008.
16. Y. Zhang, Y. Zhou, J. Shi, S. Zhou, X. Sheng, Z. Zhang and S. Xiang, J. Mol. Catal. A: Chem., 2014, 381, 138–147.
17. M.-L. Yang, Y.-A. Zhu, X.-G. Zhou, Z.-J. Sui and D. Chen, ACS Catal., 2012, 2, 1247–1258.
18. X. F. Yang, A. Q. Wang, B. T. Qiao, J. Li, J. Y. Liu and T. Zhang, Acc. Chem. Res., 2013, 46, 1740–1748.
19. J. Lin, A. Wang, B. Qiao, X. Liu, X. Yang, X. Wang, J. Liang, J. Li, J. Liu and T. Zhang, J. Am. Chem. Soc., 2013, 135, 15314–15317.
20. B. T. Qiao, A. Q. Wang, X. F. Yang, L. F. Allard, Z. Jiang, Y. T. Cui, J. Y. Liu, J. Li and T. Zhang, Nat. Chem., 2011, 3, 634–641.
21. X. Wei, X.-F. Yang, A.-Q. Wang, L. Li, X.-Y. Liu, T. Zhang, C.-Y. Mou and J. Li, J. Phys. Chem. C, 2012, 116, 6222–6232.
22. J. M. Thomas, Z. Saghi and P. L. Gai, Top. Catal., 2011, 54, 588–594.
23. M. Flytzani-Stephanopoulos and B. C. Gates, Annu. Rev. Chem. Biomol. Eng., 2012, 3, 545–574.
24. X. Cao, Q. Fu and Y. Luo, Phys. Chem. Chem. Phys., 2014, 16, 8367–8375.
25. X. Cao, Y. Ji and Y. Luo, J. Phys. Chem. C, 2015, 119, 1016–1023.
26. G. Kyriakou, M. B. Boucher, A. D. Jewell, E. A. Lewis, T. J. Lawton, A. E. Baber, H. L. Tierney, M. Flytzani-Stephanopoulos and E. C. H. Sykes, Science, 2012, 335, 1209–1212.
27. H. L. Tierney, A. E. Baber, J. R. Kitchin and E. C. H. Sykes, Phys. Rev. Lett., 2009, 103, 246102.
28. Q. Fu and Y. Luo, ACS Catal., 2013, 3, 1245–1252.
29. G. Kresse and J. Furthmüller, Comput. Mater. Sci., 1996, 6, 15–50.
30. G. Kresse and J. Furthmüller, Phys. Rev. B, 1996, 54, 11169.
31. P. E. Blöchl, Phys. Rev. B, 1994, 50, 17953–17979.
32. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865–3868.
33. G. Henkelman, B. P. Uberuaga and H. Jónsson, J. Chem. Phys., 2000, 113, 9901–9904.
34. M.-L. Yang, Y.-A. Zhu, C. Fan, Z.-J. Sui, D. Chen and X.-G. Zhou, Phys. Chem. Chem. Phys., 2011, 13, 3257–3267.
35. S. Vajda, M. J. Pellin, J. P. Greeley, C. L. Marshall, L. A. Curtiss, G. A. Ballentine, J. W. Elam, S. Catillon-Mucherie, P. C. Redfern, F. Mehmood and P. Zapol, Nat. Mater., 2009, 8, 213–216.
36. M. Dion, H. Rydberg, E. Schröder, D. C. Langreth and B. I. Lundqvist, Phys. Rev. Lett., 2004, 92, 246401.
37. B. Hammer and J. K. Nørskov, Nature, 1995, 376, 238–240.
38. B. Hammer and J. K. Nørskov, Adv. Catal., 2000, 45, 71–129.
39. B. Hammer, Top. Catal., 2006, 37, 3–16.
40. Y. Luo, Comprehensive handbook of chemical bond energies, CRC Press, 2007.
41. F. Abild-Pedersen, J. Greeley and J. Nørskov, Catal. Lett., 2005, 105, 9–13.
42. A. Vojvodic, F. Calle-Vallejo, W. Guo, S. Wang, A. Toftelund, F. Studt, J. I. Martínez, J. Shen, I. C. Man, J. Rossmeisl, T. Bligaard, J. K. Nørskov and F. Abild-Pedersen, J. Chem. Phys., 2011, 134, 244509.
43. G. Ertl, H. Knözinger, F. Schüth and J. Weitkamp, Handbook of heterogeneous catalysis, VCH, Weinheim, 1997.
44. L. Nykänen and K. Honkala, J. Phys. Chem. C, 2011, 115, 9578–9586.
45. L. Nykänen and K. Honkala, ACS Catal., 2013, 3, 3026–3030.
46. M. A. Newton, S. M. Francis and M. Bowker, Surf. Sci., 1991, 259, 56–64.
47. A. E. Baber, H. L. Tierney, T. J. Lawton and E. C. H. Sykes, ChemCatChem, 2011, 3, 607–614.
48. J. B. Miller, B. D. Morreale and A. J. Gellman, Surf. Sci., 2008, 602, 1819–1825.
49. I. H. Svenum, J. A. Herron, M. Mavrikakis and H. J. Venvik, Catal. Today, 2012, 193, 111–119.
50. H. L. Tierney, A. E. Baber and E. C. H. Sykes, J. Phys. Chem. C, 2009, 113, 7246–7250.
51. A. Bruma, R. Ismail, L. O. Paz-Borbón, H. Arslan, G. Barcaro, A. Fortunelli, Z. Y. Li and R. L. Johnston, Nanoscale, 2013, 5, 646–652.
52. R. Ferrando, A. Fortunelli and G. Rossi, Phys. Rev. B, 2005, 72, 085449.
53. X. Liu, D. Tian and C. Meng, Chem. Phys., 2013, 415, 179–185.

---

Footnote

† Electronic supplementary information (ESI) available: Energy barriers for propane/propylene dehydrogenation on pure and Pd-doped Cu(111) surfaces with the van der Waals correction by vdW-DF functional. See DOI: 10.1039/c6ra15038a

---