1 Introduction

Palladium and its alloys are important because of their catalytic properties in hydrogenation, oxidation and other reactions. Moreover, the application of Pd in membranes for hydrogen separation is being developed [1, 2]. Pd has been alloyed with Ag in order to (1) enhance the catalytic activity by moderating the bond strength of adsorbed species [3, 4], or (2) to suppress the transition between different hydride phases in membranes or catalysts under hydrogen exposure [5,6,7]. The structure, composition and electronic properties of the Pd–Ag alloy surface are important in this respect, but the surface and sub-surface composition will also be affected by the segregation phenomena induced by changes in the reactive environment [8]. Understanding the energetics for adsorption of relevant atomic and molecular species as well as the segregation behavior of Pd–Ag alloys, and how these mutually affect each other, is therefore important.

The interaction of CO + H2 with Pd1−xAgx surfaces is of particular relevance to membrane separation of hydrogen produced from carbon containing feedstock (natural gas, biomass or coal), as well as CO hydrogenation catalysis and the electro-oxidation of alcohols [9, 10]. We have long been involved in experimental work on thin, polycrystalline (1–10 µm) Pd alloy membranes, that may enable highly efficient separation of hydrogen from mixtures containing CO, CO2, H2O and CH4; essentially also allowing a CO2 rich retentate on the high-pressure side that is suitable for CO2 capture and sequestration (CCS) [11]. Due to the low thickness and fabrication by sputtering that allows elimination of bulk and gas phase transport limitations in the membrane separation configuration, the effect of surface phenomena becomes more apparent. Depending on the process conditions as well as the thickness, CO may inhibit H2 permeation through the Pd-based membranes [12,13,14]. We previously found, however, that the CO inhibition effect on ~ 3 µm thin PdAg membranes could be largely reduced by first exposing the membranes to air at 573 K [15]. This procedure involves the formation of a PdO layer in the surface region, and while this oxide is readily reduced upon subsequent exposure to H2, the enhanced permeation remains for the duration of the experiments (up to 85 days) [16]. Segregation of Pd to the surface has been suggested to contribute to the reduced inhibition effect towards CO [17, 18], in addition to surface roughening [18,19,20,21,22] and removal of impurities [23, 24].

Studies of hydrogen and CO adsorption on metal surfaces with controlled structure and composition has yielded information relevant to many industrial catalytic processes. For Pd and PdAg alloys, the interaction of H and CO has been investigated with a variety of different experimental techniques including low energy electron diffraction (LEED) [25,26,27,28,29,30,31], temperature desorption techniques [29,30,31,32,33,34], photoemission spectroscopy (PES) [35,36,37,38,39,40], and scanning tunneling microscopy (STM) [41,42,43]. Despite that most such investigations have been performed under ultra-high vacuum conditions in order to preserve information regarding structural, electronic, energetic, and vibrational properties, this has enabled significant insight relevant to catalyst and membrane applications. Recently, in-situ near ambient pressure photoelectron spectroscopy (NAP-PES) has opened up the possibility to investigate the reactivity and follow changes in the surface chemistry at reactant pressures more representative of the actual working conditions of a catalyst [44,45,46,47,48]. In addition, theoretical studies based on density functional theory (DFT) [17, 49,50,51,52,53,54,55,56,57,58,59] are increasingly complementing and facilitating the experimental interpretation.

H2 adsorbs dissociatively on Pd(100) and Pd(111) single crystal surfaces [60,61,62,63], and H occupies hollow sites on both surfaces forming different adsorption structures depending on the coverage [26]. Hydrogen also readily absorbs in the subsurface of Pd [42, 49, 55]; the very property that makes Pd attractive as a membrane material. Pd-Ag alloy surfaces are predicted to behave similarly to Pd with respect to hydrogen, with H preferentially occupying hollow sites independent of surface coverage [17, 50, 64]. Compared to Pd(111), however, hydrogen dissolves more easily into the subsurface of PdAg(111) and the desorption of H occurs at a higher temperature [31]. It has been observed experimentally that CO adsorbs molecularly in an upright configuration on Pd(111) where different ordered structures may stabilize depending on the CO coverage [29, 41, 65, 66]. For Pd–Ag alloy surfaces, no ordered CO overlayer structures have been observed. CO generally favors adsorption sites with high Pd coordination on Pd–Ag surfaces [31, 67], which is in agreement with DFT calculations [17, 54].

Using periodic DFT calculations, we have previously investigated the H2/CO/O2 adsorbate system over Pd3Ag(111) model systems in order to better understand how the adsorption energetics and the segregation behavior are mutually affected. Our previous theoretical investigation examined the individual adsorbates (H2, CO, O2) on different Pd3Ag(111) terminations, including effects of coverage [17]. These calculations were consistent with other experimental and theoretical studies establishing that, in absence of adsorbates (i.e. vacuum/inert atmosphere), the thermodynamically stable surface is Ag rich [68,69,70,71,72,73,74,75,76]. We further showed, however, that Pd is energetically favored in the topmost layer upon adsorption of H, O, or CO. This would make a Pd–Ag alloy surface Pd-rich at the corresponding saturation coverage if the conditions simultaneously favor the segregation kinetics. In the present work, we examine CO and H2 co-adsorption, more specifically the effect of CO adsorption on the adsorption and dissociation of H2 on Pd–Ag alloy surfaces. A bulk-terminated Pd3Ag(111) surface is compared with a Pd-terminated Pd3Ag alloy. Different compositions of Pd and Ag in the surface are considered in order to investigate the effect of segregation under reactive conditions and the dissociation of H2 is explored as a function of increasing CO coverage. Lastly, we constructed a phase diagram as a function of CO and H2 chemical potentials evaluating the coverage at relevant temperature and pressure conditions.

2 Methodology

The calculations investigating the co-adsorption of H and CO on a Pd- and a bulk-terminated Pd3Ag(111) surface, were performed using periodic DFT as implemented in DACAPO [77, 78]. The exchange-correlation energy and potential are described by the generalized gradient approximation (GGA-PW91) [79, 80], and the ionic cores were described by ultrasoft Vanderbilt pseudopotentials [81]. The Kohn–Sham wave functions were expanded in a plane-wave basis with an energy cutoff of 400 eV, and the surface Brillouin zone was sampled with 18 special k-points [82].

The Pd3Ag(111) alloy surfaces, where Pd3Ag possesses a fcc L12-like structure, are modeled using a six layer slab with a (2 × 2) surface unit cell. A vacuum layer corresponding to six atomic layers was applied between successive images of the slab. The optimized bulk lattice constant for Pd3Ag is calculated to be 4.014 Å [17], in reasonable agreement with the experimental value of 3.926 Å [83]. The four topmost layers were free to relax without any restrictions imposed during structural optimization, whereas the two bottom layers were fixed to the lattice structure of the bulk. The geometries were relaxed until the force components on each atom were less than 0.02 eV/Å, and the convergence criterion for the electronic self-consistent loop was set to 10−5 eV. The bulk-terminated surface has 25% Ag and 75% Pd in the surface and Ag evenly distributed throughout the slab. The overall Pd:Ag ratio of the slab is maintained at 3:1 for the Pd-terminated surface, for which the Ag atom in the topmost layer is exchanged with a second layer Pd atom. H and CO adsorption and co-adsorption is investigated on both surfaces at different H/CO ratios. Subsurface hydrogen is not included in this study. Adsorption is allowed on one side of the slab only, and the electrostatic potential is adjusted accordingly [84, 85]. The many possible co-adsorbed structures have been rigorously examined. The differential binding energy, \(\text{BE}_{\text{diff}}\), was defined as:

$${\text{B}\text{E}}_{\text{diff}}={E}_{{N}_{\text{Ads}}/\text{surf}}-\left({E}_{{(N}_{\text{A}\text{ds}}-1)/\text{surf}}+{E}_{\text{Ads}}\right)$$
(1)

\({E}_{{N}_{\text{Ads}}/\text{surf}}\) is the total energy of the adsorbate-substrate system with \({N}_{\text{Ads}}\) adsorbates, and \({E}_{{(N}_{\text{A}\text{ds}}-1)/\text{surf}}\) is the total energy of the adsorbate-substrate system with \(({N}_{\text{Ads}}-1)\) adsorbates. \({E}_{\text{surf}}\)\(({N}_{\text{Ads}}-1=0)\) is the total energy of the clean Pd-Ag surface and \({E}_{\text{A}\text{d}\text{s}}\) is the total energy of a gas-phase CO molecule or H atom. For the co-adsorbed system, \({E}_{\text{surf}}\) corresponds to the surface with CO pre-adsorbed. Adsorption is energetically favorable when \(\text{BE}_{\text{diff}}\) is negative. Predicted saturation coverage of H is reached when \(\text{BE}_{\text{diff}}\) is below the value of the bond energy of H2 per atom.

Climbing image nudged elastic band (CI-NEB) calculations [86] were used to calculate activation energy barriers for H2 dissociation with increasing amount of CO pre-adsorbed. 7 or 11 images were typically used for the CI-NEB calculations, and the calculations were optimized yielding forces below 0.05 eV/Å. Transition states were verified using vibrational analysis resulting in a single imaginary frequency.

Surface phase diagrams were constructed to identify the most stable CO and/or H adsorption structures as a function of reaction conditions. The surfaces were assumed to be in thermodynamic equilibrium with gas phase CO and H2 following the constrained equilibrium approach [87,88,89]. In this approach, the species are not allowed to react with each other and the Gibbs free energy for adsorption, \({\Delta G}^{\text{ads}}\), is expressed by

$${\Delta G}^{\text{ads}}= -\frac{1}{A}\left[{G}_{\text{H,CO/surf}}-{G}_{\text{surf}}-{N}_{\text{H}}{\mu}_{\text{H}}-{N}_{\text{CO}}{\mu }_{\text{CO}}\right]$$
(2)

where A is the surface area, \({G}_{\text{H,CO/surf}}\) is the Gibbs free energy for the surface with \({N}_{H}\) hydrogen and \({N}_{\text{CO}}\) CO species adsorbed. \({G}_{\text{surf}}\) is the Gibbs free energy for the surface in the absence of any adsorbates. \({\mu }_{\text{H}}\) and \({\mu }_{\text{CO}}\) represent the chemical potential of the two gas phase reservoirs of hydrogen and CO, respectively. Assuming ideal gas, they can be expressed as a function of temperature and pressure:

$${\mu }_{\text{H}}\left(T,{p}_{{\text{H}}_{\text{2}}}\right)=\frac{1}{2}\left[{E}_{{\text{H}}_{2}}+{\tilde{\mu }}_{{\text{H}}_{\text{2}}}\left(T,{p}^{0}\right)+{k}_{\text{B}}T \text{ln}\frac{{p}_{{\text{H}}_{\text{2}}}}{{p}^{0}}\right]= \frac{1}{2}\left({E}_{{\text{H}}_{\text{2}}}+\Delta {\mu }_{{\text{H}}_{\text{2}}}\right)$$
(3)

and

$${\mu }_{\text{CO}}\left(T,{p}_{\text{CO}}\right)={E}_{\text{CO}}+{\tilde{\mu }}_{\text{CO}}\left(T,{p}^{0}\right)+{k}_{\text{B}}T \text{ln}\frac{{p}_{CO}}{{p}^{0}}={E}_{\text{CO}}+\Delta {\mu }_{\text{CO}}$$
(4)

where \({p}^{0}\) is the standard pressure. The chemical potential of molecular and atomic hydrogen is related by \({\mu }_{\text{H}}=1/2\times {\mu }_{{\text{H}}_{\text{2}}}\). The chemical potentials of H2 and CO at standard pressure, \({\tilde{\mu }}_{{\text{H}}_{\text{2}}}\left(T,{p}^{0}\right)\) and \({\tilde{\mu }}_{\text{CO}}\left(T,{p}^{0}\right)\) are evaluated using the NIST-JANAF thermochemical database [90]. \({E}_{{\text{H}}_{\text{2}}}\) and \({E}_{\text{CO}}\) are the total energies of the gas-phase H2 and CO molecules, respectively. \({\Delta G}^{\text{ads}}\) can be approximated to

$${\Delta G}^{\text{ads}}\approx \frac{1}{A}\left[-{\text{BE}}_{\text{ave}}+{N}_{\text{H}}\Delta {\mu }_{\text{H}}+{N}_{\text{CO}}\Delta {\mu }_{\text{CO}}\right]$$
(5)

\(\text{BE}_{\text{ave}}\) is the average binding energy for H and CO given by

$${\text{B}\text{E}}_{\text{ave}}=\left({E}_{\text{H,CO/surf}}-{E}_{\text{surf}}-\frac{{N}_{\text{H}}}{2}{E}_{{\text{H}}_{\text{2}}}-{N}_{\text{CO}}{E}_{\text{CO}}\right)$$
(6)

where \({E}_{\text{H,CO/surf}}\) is the total energy of the co-adsorbed system. The average binding energy is corrected for zero-point energies.

3 Results and Discussion

3.1 Adsorption and Activation on Bulk-Terminated Pd3Ag(111)

For the H + CO co-adsorbed system, we consider changes in the adsorption energetics upon hydrogen adsorption on a bulk-terminated Pd3Ag(111) surface with increasing amounts of CO pre-adsorbed. The differential binding energy of hydrogen for increasing CO pre-coverage is shown in Fig. 1. We have previously published the coverage dependent interaction of H or CO alone with differently terminated Pd3Ag(111) surfaces [17]. Results from that study, representing the starting point for the current study, are included here. Table 1 summarizes the optimized adsorption geometries for different coverages of either H or CO.

Fig. 1
figure 1

Differential binding energy for H as a function of coverage on bulk-terminated Pd3Ag(111) surfaces without and with increasing amount of CO pre-adsorbed. The dotted horizontal line represents the calculated bond energy of H2 (g) per atom

Full size image
Table 1 Preferred adsorption sites and differential binding energies of CO and H adsorbed individually on the bulk-terminated Pd3Ag(111) surface
Full size table

In the absence of adsorbed CO, the predicted hydrogen saturation coverage for Pd3Ag(111) is 1.00 ML, since the differential binding energy becomes endothermic relative to H2 in the gas phase (1/2 \({E}_{{\text{H}}_{2}}\)) above this coverage (see lower curve of Fig. 1). The differential binding energy of 0.25 ML H is − 2.85 eV in a fcc site made up by Pd. The preference for the fcc site found here is consistent with predictions for Pd(111) where the fcc is generally favored over the hcp site by around 0.05 eV [50, 53, 58, 59]. With increasing coverage, H atoms populate hollow sites with the highest Pd concentration available, favoring fcc sites in agreement with previous studies [17, 50, 64]. Figure 2a illustrates the Pd3Ag(111) surface at the predicted H saturation coverage. In sites of mixed composition, e.g. an fcc site comprising 2 Pd atoms and 1 Ag atom in the top layer (here denoted fcc2PdAg), the H atom is displaced toward surface Pd atoms and away from Ag. Overall, H atoms repel each other when adsorbed on this surface and, on average, the differential binding energy of H decreases in magnitude as the H coverage increases.

Fig. 2
figure 2

Illustration of the bulk Pd3Ag(111) with the predicted saturation adsorption structures of a H (1.00 ML) and b CO (0.75 ML). No co-adsorption of H and CO is considered in this case. The (2 × 2) unit cell is outlined with dashed lines. Atoms are colored as follows: Pd—blue, Ag—grey, H—white, C—black, O—red

Full size image

In the absence of adsorbed H, CO occupies the fcc3Pd site on the bulk-terminated Pd3Ag(111) at 0.25 ML with a differential binding energy of − 2.01 eV, similar to energies obtained for other Pd-based surfaces [53, 54, 91]. Hollow sites (fcc and hcp) made up by Pd alone are preferred on this surface at 0.50 ML, while CO prefers less coordinated sites over hollow sites with mixed Pd–Ag composition above this coverage. Adsorbed CO exhibits considerable repulsive interactions as the differential binding energy increases by 0.38 eV between 0.25 and 0.50 ML. At the saturation coverage (0.75 ML) depicted in Fig. 2b, the three CO molecules in the unit cell occupy one Pd top site and two Pd–Pd bridge sites (Table 1). The CO molecules in the latter positions are shifted towards the hollow site with the O atom tilted away from the CO in the top site (towards the surface Ag atom).

As shown in Fig. 1, the adsorption energetics of hydrogen is significantly influenced by the presence of adsorbed CO. Adding 0.25 ML H to a surface containing 0.25 ML CO makes the adsorption of H 0.12 eV less stable; CO, then, occupies the Pd-coordinated fcc site while H goes to the corresponding hcp site, in a similar configuration as for 0.50 ML H or CO alone. At higher H coverage, CO persists in occupying the fcc3Pd site while H attains fcc sites of mixed Pd–Ag composition. However, with 0.25 ML CO the differential binding energy of H is more negative at 0.75 ML compared to 0.50 ML H. A similar rearrangement effect exists for hydrogen adsorption alone upon increasing the coverage from 0.75 ML to 1.00 ML H and can be ascribed to the heterogeneity of the surface. Up to a total coverage of 0.5 ML H (no CO), hollow sites consisting of only Pd (fcc and hcp) are occupied, whereas only fcc sites become occupied at 0.75 ML hydrogen. In this case, two out of three sites have mixed composition, and this leads to significant destabilization caused by surface Ag atoms. Upon increasing the coverage to 1.00 ML in total, the remaining fcc sites become populated. Nevertheless, and in agreement with experimental findings [92, 93], adsorption sites occupied by CO remain inaccessible to H and the predicted hydrogen saturation coverage is reduced to 0.75 ML when 0.25 ML CO is pre-adsorbed.

Adsorption of hydrogen is not energetically favorable upon increasing the CO pre-coverage beyond 0.25 ML. The more positive differential BE signifies a rather strong repulsive interaction between the adsorbed species resulting in weaker bonding of hydrogen to the surface. Also, as the total CO + H coverage increases, a reconfiguration of preferred sites is observed, e.g. while CO is situated in a fcc and a hcp hollow site at 0.50 ML, the hcp bonded CO gains 0.45 eV by moving to a Pd–Pd bridge site upon hydrogen addition. H persists in the fcc hollow site, but in total, this rearrangement makes the 0.50 ML CO + 0.50 ML H configuration nearly thermoneutral with respect to H2 (g) (Fig. 1).

As previously discussed, the presence of CO is also anticipated to affect the activation of H2 on Pd based surfaces. First, we have considered H2 dissociation on the clean and CO pre-covered Pd3Ag(111) surfaces up 0.75 ML CO. H2 was initially placed above the surface with the H2 bond parallel to the surface. Hydrogen dissociation relative to the most energetically stable CO adsorption structures was explored using CI-NEB calculations, and the energy profiles for dissociation relative to the gas phase are shown in Fig. 3. As expected, the dissociation of H2 is non-activated in the absence of CO. Paths without a barrier are found in Pd rich areas; between two fcc sites over a Pd–Pd bridge site or between one fcc and one hcp site. This agrees well with previous theoretical calculations for H2 dissociation on Pd(111) [52, 94] and Pd3Ag(111) [95]. For the latter, it was reported that the activation energy barrier increased for dissociation paths close to surface Ag atoms. Clean Ag is inert towards H2 dissociation [96,97,98], with barriers over Ag(111) calculated at 1.04 eV or higher [99,100,101]. Hydrogen dissociation would thus become difficult in Ag-rich areas of Pd1 − xAgx(111) surfaces.

Fig. 3
figure 3

Activation energy barrier for H2 dissociation with increasing amount of CO pre-adsorbed on the bulk-terminated Pd3Ag(111) surface

Full size image

The activation energy barrier for H2 dissociation becomes 0.04 eV when 0.25 ML CO is pre-adsorbed, with a preferred path that is well separated from the adsorbed CO molecule, i.e. over a Pd–Pd bridge site between fcc2PdAg and hcp3Pd sites. We conclude that low CO coverage has minor influence on H2 dissociation, but the final dissociated state becomes less stable relative to adsorption on the surface with no CO pre-adsorbed (Fig. 3). Increasing the CO coverage to 0.50 ML leads to an activation energy barrier for H2 dissociation of 0.94 eV, with the final state becoming endothermic with respect to H2 (g). When 0.75 ML of CO is pre-adsorbed on the surface, corresponding to the CO-saturation coverage, H2 dissociation is characterized by an activation energy barrier of 1.67 eV. In addition to the increased barrier for H2 dissociation, there is a repulsive interaction between the approaching H2 molecule and the pre-adsorbed CO molecule(s). Typically, CO is tilted away and/or moved to a different site during the dissociation process. This is more evident at higher CO coverage.

3.2 Adsorption and Activation on Pd-Terminated Pd3Ag(111)

The differential binding energy of H on the Pd-terminated Pd3Ag(111) surface as a function of hydrogen coverage with increasing amount of CO pre-adsorbed is shown in Fig. 4. The lower curve represents hydrogen adsorption on the clean surface (no CO pre-adsorbed), with a predicted saturation coverage of 1.00 ML H. The differential binding energies of H or CO adsorbed alone on Pd-terminated Pd3Ag(111) are listed in Table 2, and the preferred adsorption geometries at their predicted saturation coverages of 1.00 ML and 0.75 ML are illustrated in Fig. 5. The adsorption of H or CO on the Pd-terminated Pd3Ag(111) surface is similar to that of the bulk-terminated surface, in terms of both adsorbates displaying repulsive adsorbate interactions. Contrary to the bulk-terminated surface, however, the differential binding energy of H increases gradually since only fcc sites with the same composition are populated upon increasing coverage. The differential binding energies (Table 2) are more negative on the Pd-terminated surface compared to the bulk-terminated surface, except at 0.50 ML CO where the bonding is slightly stronger on the bulk-terminated surface. CO prefers to bind at fcc and hcp sites made up by Pd on both surfaces at 0.50 ML, with the concentration of Ag in the layer underneath causing the energy difference. The CO configuration differs between the two terminations at saturation coverage, with two hollow sites and one Pd top site occupied and an overall stronger interaction of CO on the Pd-terminated surface. This can be attributed to ensemble effects where the surface Ag atoms restrict the possibility for CO to occupy Pd hollow sites in addition to the top site in the case of the bulk-terminated Pd3Ag surface. Hydrogen, in contrast, prefers fcc sites on Pd-terminated Pd3Ag(111) for all coverages listed in Table 2.

Fig. 4
figure 4

Differential binding energy for H as a function of coverage on Pd-terminated Pd3Ag(111) surfaces without and with increasing amount of CO pre-adsorbed. The dotted horizontal line represents the calculated bond energy of H2 (g) per atom

Full size image
Fig. 5
figure 5

Illustration of the Pd-terminated Pd3Ag(111) with the predicted saturation adsorption structures of a H (1.00 ML) and b CO (0.75 ML). The (2 × 2) unit cell is indicated by the black dashed lines. Atoms are colored as follows: Pd—blue, Ag—grey, H—white, C—black, O—red

Full size image
Table 2 Preferred adsorption sites and differential binding energies of CO and H adsorbed individually on the Pd-terminated Pd3Ag(111) surface
Full size table

The energetics of hydrogen adsorption on this surface is strongly affected by the presence of pre-adsorbed CO, which destabilizes hydrogen. At low CO and H coverage (0.25 ML + 0.25 ML) the interactions of CO and H on the two surfaces are similar. However, as the coverages increase the species bind stronger to the Pd-terminated surface relative to the bulk-terminated Pd3Ag(111). As with the bulk-terminated Pd3Ag(111), in the presence of CO, the predicted hydrogen saturation coverage is also reduced on the Pd-terminated surface. In the absence of CO, the Pd-terminated surface can sustain a H saturation coverage of 1.00 ML (Fig. 4). That saturation coverage decreases to 0.75 ML H with 0.25 ML CO, as in the case of the bulk-terminated surface. But, while hydrogen adsorption is unfavorable on the bulk-terminated surface at higher CO coverage, 0.50 ML H is possible onto the 0.50 ML CO pre-covered Pd-terminated surface. H adsorption becomes endothermic only at the CO saturation coverage of 0.75 ML on the Pd-terminated surface.

The activation energy barriers for H2 dissociation on the Pd-terminated Pd3Ag(111) surface as a function of pre-adsorbed CO coverage are shown in Fig. 6. The dissociation is similar to that on the bulk-terminated surface in the absence of CO (non-activated) and at low CO coverage. With 0.25 ML CO pre-coverage there is a small activation energy barrier of 0.08 eV and the dissociation paths exhibiting the lowest barrier for both surface terminations are between two fcc sites or fcc and hcp over a Pd–Pd bridge site. The adsorbed hydrogen atoms are destabilized with CO present, but the destabilization is less than in the case of the bulk-terminated surface. Increasing the CO coverage to 0.5 ML on the Pd-terminated Pd3Ag(111) surface, leads to a H2 dissociation barrier of 0.34 eV. It is at this intermediate CO coverage that the difference between bulk- and Pd-terminated Pd3Ag(111) is most significant, since hydrogen adsorption is still exothermic and the activation barrier is more than 0.6 eV lower when the surface is terminated by palladium only. At the CO saturation coverage of 0.75 ML, the H2 dissociation barrier increases to 1.42 eV, which is 0.25 eV lower than that of the bulk-terminated surface.

Fig. 6
figure 6

Activation energy barrier for H2 dissociation with increasing amount of CO pre-adsorbed on the Pd-terminated Pd3Ag(111) surface

Full size image

Irrespective of termination, the calculations illustrate how the activation of hydrogen is affected by the presence of CO on the surface. Moreover, the two terminations studied here behave similarly for low and high CO coverage with respect to hydrogen dissociation. A transition from non-activated dissociation to a significant activation energy for increasing CO coverage is in agreement with studies of Pt(111) and Pt-terminated core-shell bimetallics by Nilekar and coworkers [102]. Our calculated activation energy barriers also agree qualitatively with studies of hydrogen dissociation over pure Pd membranes for CO coverages in the range 0–0.5 ML [103], but the semi-empirical quantification predicts somewhat lower barriers. Our findings indicate, however, that at intermediate CO coverage, a Pd-terminated Pd3Ag surface is more reactive towards hydrogen than a bulk-terminated surface, due to the lower activation barrier.

3.3 Stability and Segregation Effects

To further evaluate the effect of CO on hydrogen adsorption we have constructed phase diagrams following the constrained equilibrium approach [87,88,89]. This allows examination of pressure and temperature effects in relation to hydrogen permeation and catalytic reaction experiments at application relevant conditions. \({\Delta G}^{\text{ads}}\) (Eq. 5) was evaluated on the two surface terminations studied for the relevant co-adsorbed configurations discussed above (0.25–1.00 ML H; 0.25–0.75 ML CO).

The phase diagrams in Fig. 7 show the most stable configurations of CO and H on the two surfaces as a function of the chemical potential of H2 and CO. The secondary axes indicate the corresponding CO and H2 partial pressures for two temperatures of potential interest for membrane separation and reactor applications, 300 K and 600 K [1]. For the lowest CO and H2 chemical potentials the most stable configuration is the clean surface for both the bulk- and the Pd-terminated surface.

Fig. 7
figure 7

Phase diagrams as a function of CO and hydrogen chemical potential for the bulk-terminated Pd3Ag(111) (upper panel) and the Pd-terminated Pd3Ag(111) (lower panel) surfaces. Secondary axes give pressure in H2 and CO at two temperatures: 300 K (inner) and 600 K (outer) scales

Full size image

On the bulk-terminated Pd3Ag(111) surface, keeping the CO chemical potential low and increasing the hydrogen chemical potential, we found three stable hydrogen overlayer structures corresponding to H coverages of 0.25 ML, 0.50 ML and 1.00 ML starting at \(\Delta{\mu }_{\text{H}_2}\) of − 1.08 eV, − 0.83 eV and − 0.38 eV, respectively. 0.75 ML H is, however, found to be unstable for any \(\Delta{\mu }_{\text{H}_2}\). This correlates well with the differential binding energy for 0.75 ML H (Fig. 1) being the energetically least favorable within the saturation range and suggesting a transition directly from a 0.50 ML to 1.00 ML covered surface. A similar non-sequential increase in hydrogen coverage has also been predicted for other surfaces; e.g. configurations corresponding to 0.50 ML and 1.00 ML (but not 0.75 ML) coverage were found stable on an (unreconstructed) Ni(111) surface [104].

On the Pd-terminated Pd3Ag(111) surface, however, the hydrogen coverage at low CO chemical potential increases by 0.25 ML increments up to saturation, see Fig. 7b, including the 0.75 ML phase, consistent with predictions for hydrogen on Pd(111) [105, 106]. The clean surface is preferred for H2 chemical potentials lower than − 1.17 eV, with − 1.03 eV, − 0.88 eV and − 0.69 eV as threshold values for the next phase transitions to increased hydrogen coverage upon increasing hydrogen chemical potential. The higher coverage at lower chemical potential reflects the stronger binding of H to the Pd-terminated surface compared to the bulk-terminated Pd3Ag(111) surface. The stronger interaction of hydrogen with the Pd-terminated surface implies segregation of Pd to the surface at temperatures of appreciable segregation kinetics. Thus, our previous conclusions from Pd3Ag(111) equilibrium ground state calculations predicting Pd-termination for hydrogen coverage exceeding 0.50 ML are not significantly altered [17].

The two Pd3Ag(111) surfaces studied behave similarly at low hydrogen chemical potential, upon increasing the CO chemical potential, with stable CO phases at 0.25, 0.50 and 0.75 ML. The 0.25 ML CO configuration becomes stable for CO chemical potentials of − 1.94 and − 2.01 eV for the bulk- and Pd-terminated surface, respectively. ΔµCO of − 1.57 and − 1.40 eV is required for the two surfaces to become covered with 0.5 ML CO. The higher CO chemical potential obtained for the Pd-terminated surface relative to bulk-terminated Pd3Ag(111) can be explained by the Ag enrichment in the second topmost layer. The largest disparity is found for high CO chemical potentials, where the CO saturation coverage (0.75 ML) is reached at − 0.84 eV for the Pd-terminated surface, whereas the bulk-terminated surface maintains 0.5 ML CO up to ΔµCO of − 0.17 eV. No ordered CO induced phases have been identified on Pd–Ag alloy surfaces from experimental investigations, likely reflecting the disordered nature of the Pd–Ag surface layer [31, 33, 107]. In contrast, several ordered structures have been observed for CO on Pd(111), including (√3×√3)R30° (0.33 ML), c(4 × 2)-2CO (0.50 ML), and (2 × 2)-3CO (0.75 ML) [41, 65, 108], for which experimental ΔµCO values of − 2.18, − 1.58 and − 0.71 eV, respectively, have been reported [65]. These values, together with values from previous theoretical work of about − 2.00, − 1.52 and − 0.69 eV [109], are relatively similar to the values predicted at similar coverages for the Pd-terminated surface studied here.

The phase diagrams in Fig. 7 suggest that hydrogen and CO may co-exist on both surfaces. For the bulk-terminated Pd3Ag(111) surface, configurations consisting of 0.25 ML CO plus 0.25 ML H and 0.25 ML CO plus 0.75 ML H are stable. The 0.25 ML CO with 0.75 ML H state is stable for a very limited range of CO and H chemical potentials. The Pd-terminated surface has a larger window of chemical potentials where adsorbed CO and H co-exist and overall higher coverages as already inferred from the differential binding energy considerations; see Sect. 3.2. Stable mixed adsorbed layers are obtained corresponding to coverages of 0.25 ML CO with 0.25, 0.50 and 0.75 ML H, and 0.50 ML CO with 0.50 ML H. Interestingly, even though the Pd-terminated surface binds both adsorbates stronger than the bulk-terminated, a higher CO chemical potential is generally required in order to transition into purely CO-covered phases on this surface for modest to high H2 chemical potentials.

Hydrogen permeation applications are often targeted in the temperature range 500–700 K, at which the phase transition (α →β) in the Pd–Ag alloy is suppressed, the kinetics for hydrogen bulk diffusion is reasonably fast, and the integration into processes such as WGS and steam reforming is viable. From the analysis above it becomes apparent that, for the relevant range of CO and hydrogen partial pressures (20 ppm—20% and 10–80%, respectively, at 0.1–5 MPa), the coverages and phase changes predicted have importance. For example, the Pd-terminated surface exhibits larger variations in CO-partial pressures where hydrogen adsorption and activation are favorable than the bulk-terminated. Moreover, a higher CO partial pressure is required to shift into pure CO adsorption configurations at a specific temperature and hydrogen pressure.

Our results may help explain the performance of thin Pd–Ag alloy membranes, where the hydrogen transport may be impacted by surface phenomena. Specifically, it suggests that the increased permeation under pure hydrogen observed experimentally after oxidative treatment is correlated with Pd surface enrichment [18, 20, 21, 24, 110], since for a pure H2 environment, presence of Ag in the topmost surface layer reduces the hydrogen adsorption energy as well as the coverage. The extent of segregation between Pd and Ag in differently prepared membranes, and specifically the distribution in the outmost surface layer may hence explain the performance observed under varying conditions and treatments of the membranes [15, 18, 20, 24]. The driving force for Pd to segregate towards the surface depends on both type of adsorbate and coverage [17], but diffusion in the bulk is kinetically controlled. This introduces considerable complexity in the dynamics of reactions over alloy surfaces. Fernandes et al. [45] recently explained decreased rates of reaction during CO oxidation over PdAg(100) observed at high temperature by reverse segregation effects (i.e. increased fraction of Ag termination upon increasing the temperature from ~ 525 to ~ 725 K) occurring due to decreased adsorbates coverage and facile segregation kinetics by a combined kinetic and first principles model. Further, our results generally agree with insights derived from studies of shallow versus deep hydrogenation over Pd-based catalysts, whereby alloying has been shown to be beneficial for moderating the binding strength of key intermediates/products thereby enhancing reaction selectivity towards shallow hydrogenation products [4, 111].

4 Conclusions

Pd3Ag(111) alloy surfaces have been studied as model systems for Pd–Ag membranes or catalysts. Through periodic DFT calculations, we have addressed the effect of surface composition and segregation on co-adsorption of hydrogen and CO to understand the reactivity. At application relevant chemical potentials (p, T), we find that, for both Pd-terminated and bulk-terminated Pd3Ag(111) surfaces, adsorbed CO blocks hydrogen adsorption sites and weakens the binding of H, leading to a reduced hydrogen saturation coverage. The adsorption properties of the two surface terminations diverge at higher adsorbate coverage, where a higher coverage of H with CO present is obtained on the Pd-terminated surface. This surface also exhibits a wider range of chemical potentials for co-adsorbed mixtures of hydrogen and CO.

The surface composition of the bimetallic alloy furthermore influences the activation energy barrier for H2 dissociation in the presence of pre-adsorbed CO. For both surface terminations, 0.25 ML CO leads to a minor activation energy barrier. The presence of Ag in the surface weakens the CO binding, but it still results in an activation energy barrier for H2 dissociation on a CO saturated surface that makes H2 dissociation unlikely. The most significant effect of surface composition is found at intermediate CO coverage (0.50 ML), with a considerably lower barrier for dissociative adsorption of H2 for the Pd-terminated surface.

Under conditions relevant for hydrogen permeation and CO hydrogenation, our results imply that Pd3Ag(111) is less prone to CO poisoning if the surface becomes Pd-terminated. At the same time, the stronger binding of H and CO to pure Pd sites introduces a coverage dependent driving force for Pd segregation to the surface. This may explain some of the experimental observations related to Pd–Ag membrane performance enhancement under oxidation-reduction cycles.