1 Abstract

Single atom alloys are gaining importance as catalysts which can be extremely selective andactive towards the formation of desired products. They possess such desirable characteristicsbecause of the presence of a highly reactive single atom at the surface. In this work, we startby investigating the electronic structure of these single atom alloys to look for any unusualbehavior. Then, we tested if the single atom alloys of gold, silver and copper which havebeen doped with single atoms of platinum, palladium, iridium, rhodium and nickel adhere tothe d-band model of Hammer and Nørskov. Then, gold-palladium, copper-palladium, silver-palladium single atom alloys and the corresponding pure metal hosts are used in simpledissociation of water and nitric oxide to test their catalytic behavior. We also tested thecatalytic behavior of these metals and alloys in the selective hydrogenation of acetylene toethylene. We observed that these alloys exhibit a sharp peak in their atom projected d-banddensity of states, which we predicted could be the cause of high surface reactivity. We foundthat the d-band centers and d-band widths of these systems correlated linearly, showingadherence to the d-band model, while the energy of adsorption of a hydrogen atom on thesesurfaces could not be correlated with the d-band center, or the average reactivity of thesurface. Finally, the single atom alloys, with the exception of copper-palladium showed goodcatalytic behavior by activating the reactant molecules more strongly and showing favorablereaction pathways on the free energy diagrams for the reactions investigated.

1

2 Introduction

A catalyst is a substance that can be used to increase the rate of a reaction by providingalternate reaction pathways of lower energy. They are also known to improve the selectivityof chemical reactions. They are used in the production of essential commodities such as fuels,pharmaceuticals and speciality chemicals. Catalysts are also important in the reduction ofpollution in the atmosphere. They are used to convert NOx and other harmful pollutantsto less potent products, through catalytic converters. Transition metals are an importantclass of catalysts known for their excellent catalytic properties in chemical processes [1].They bind substances ideally, in a way that the binding is not too strong or too weak. Suchbonds are ideal in the process of catalyzing a reaction. Transition metals such as platinum,palladium, ruthenium, rhodium are known to catalyze a variety of reactions, and make theformation of products favorable [2].

However, these metals are extremely rare and precious, and efforts have been underwayfor the past decade to discover new catalysts which can be substituted in place of transitionmetal catalysts. Experimental and theoretical studies have investigated the use of metalalloys in a variety of reactions, and have had a reasonable degree of success. Metal alloysare materials whose properties can be quite different from those exhibited by its constituentelements. In these systems the electronic interactions between the different constitutents orsupports, and the change in lattice parameters due to variations in atomic sizes could giverise to a variety of properties, which could be of interest in the catalysis scheme of thinking[3]. Alloys can also be tuned to exhibit desired properties and characteristics [4]. Variousgroups of researchers have found that certain alloys of platinum like Pt3Ni show manifoldincrease in activity compared to the activity of a pure Pt catalyst for oxidation reductionreactions (ORRs) [5]. Computational studies which perform large scale screening of catalystsshow the effectiveness of alloys of late transition metals and early transition metals for ORR[6, 7, 8] hydrogen related reactions [9] and selective hydrogenation reactions [10] to name afew.

The surface of a heterogenous catalyst consists of a variety of sites, which facilitate theprocess of adsorption which is a precursor to the reaction taking place. An ideal flat (111)surface of a metal will possess at least 3 types of sites, namely the hollow, atop and bridgesites. A stepped surface will have even more sites like kinks, which can take part in areaction. A catalyst surface will never be homogenous, and will always have a huge numberof sites, which can act as active sites for a reaction [11]. A surface is said to catalyze aparticular reaction, when it makes the formation of certain products the most energeticallyfavorable. It simultaneously renders other side reactions unfavorable, which engenders theconcept of selectivity. Now, with so many sites available on a surface, chemists are trying

2

to engineer active sites so that reactivity and selectivity can be controlled. One school ofthought that has resulted out of this ’active site engineering’ process, are single atom alloycatalysts. The rationale behind this idea was to design catalysts that contain only the sitesrequried to perform or assist a target reaction [12].

Single atom alloys are alloys in which one metal constitutes the majority of the alloy, andthe other metal is doped with extremely low concentrations on the surface of the alloy. Theyare known to show enhanced reactivity and selectivity in hydrogenation reactions [13]. Itwas recently shown that alloying one percent of Pd atoms into a Cu(111) surface significantlypromotes the dissociation of H2, while not overbinding the H atoms formed [14]. Such alloysalso promote spillover the H atoms across the surface of the catalyst, as shown in the work ofSykes and coworkers [3]. A lot of research has gone into the use of alloys consisting of coinagemetals liks Au, Ag and Cu and late transition metals such as Pd, Pt etc [15]. Sykes andcoworkers have published substantial work in the sphere of single atom alloys [16], especiallyon H2 activation [17] and selective hydrogenation [13, 18].

In this work, we develop a bottom up approach to investigating the use of single atomalloy catalysts in simple reactions. To begin with, we undertook a screening process where avariety of bimetallic alloys were screened for unusual electronic structure and behavior. Thealloys were those of Cu, Au and Ag, in which one surface atom was replaced by at atom ofPt, Pd, Ir, Rh or Ni. Using the d-band density of states and other d-band characteristics,we identified 3 potential candidates for the single atom alloys, which were AuPd, AgPd andCuPd. Next, we investigated the limit of dilution from a computational standpoint, andwere able to choose a 3 × 3 surface unit cell which best represented a compromise betweencomputational expense and single atom alloy behavior. Having chosen a representativecandidate set, and their surface configuration, we set about testing their catalytic behaviorin simple dissociation reactions such as dissociation of water and nitric oxide. Owing tothe positive results from the use of single atom alloys in selective hydrogenation reactions,we also studied the selective hydrogenation of acetylene to ethylene. The results of ourinvestigations provided insight into how the unusual electronic structure of the single atomalloy surface plays a role in the reactions. We found that the adsorbates bond to the Pd atomexceptionally well when compared to bonding on pure metals, while adsorbing in a differentgeometry to that on pure surfaces of Au, Cu and Ag. We plotted reaction pathways throughpotential energy diagrams, and found that for the selective hydrogenation of acetylene, theformation of reaction intermediates is stable, and facilitates a forward reaction towards theformation of ethylene.

3

3 Methods

3.1 Computational Details

In this work, alloys chosen for investigation are those of copper, silver and gold. One atomwas replaced with a palladium atom, such that it was completely isolated from other palla-dium atoms, thus exhibiting single atom alloy behavior. All calculations were done using theVienna ab initio simulation package code (VASP) [19, 20, 21]. VASP solves the Kohn-Shamequations in an iterative manner within a planewave basis set, while employing periodicboundary conditions. The interaction between the core and valence electrons was describedby the application of the projector augmented wave (PAW) method [22]. A plane wavecutoff of 350 eV and a Gaussian smearing width of 0.1 eV was chosen as parameters forthese calculations. The exchange correlation functional used was the Perdew Burke Ernzer-hof (GGA-PBE) functional [23]. The parameters and exchange correlation functional werechosen to represent the best compromise between computational expense and depiction ofthe actual system. To begin with, bulk calculations were performed for each system in theconfiguration XY3, where X is the single atom, and Y is the host metal. Lattice constantsused were of the host metal. Lattice constants optimized for the calculation parameters usedwere obtained by fitting an equation of state and obtaining the lattice constant correspond-ing to the minimum energy. The calculated lattice constants for the metals considered arePt: 3.9671 Å, Pd: 3.9397 Å, Ir: 3.8729 Å, Rh: 3.8234 Å, Ni: 3.5099 Å, Cu: 3.6358 Å,Au: 4.1560 Å and Ag: 4.1441 Å. These values are in excellent agreement with experimentallattice constants [24]. Finally, a Monkhorst-Pack k-point grid of 12 × 12 × 12 was used for asimple unit cell consisting of one atom of the metal. The k-point grid was accordingly scaledup or down based on the dimensions of the system considered [25].

Next, calculations were performed to determine if the behavior of these single atom alloysystems can be described by the d-band model developed by Hammer and Nørskov [26]. Theelectronic structure for each system has been calculated from the atom projected densityof states (DOS). For this, a single atom of hydrogen was adsorbed at the hollow site ofthree facets fcc(111), fcc(100) and fcc(211) and relaxed to the ground state. The d-bandDOS of the single atom, which takes part in the adsorption of hydrogen was obtained. AWigner Seitz radius of 1.5 Å was used to truncate the orbitals and prevent double countingof electrons from neighboring atoms. The d-band center was calculated as the square root ofthe first moment of the d-band with respect to the Fermi level as εd =

∫ρEdE∫ρdE

. The d-bandwidth was calculated as the square root of the second moment of the d band with respect to

the Fermi level as Wd =√∫

ρE2dE∫ρdE

. These values are desciptors used in the d-band model.The d-band width was plotted against the d-band center to ascertain if these systems adhereto the model. We also plotted the peak energy in the DOS against the d-band center to

4

determine if the behavior of these systems can be determined by the shape of the d-band[27].

Finally, we proceeded to investigate the catalytic activity of single atom alloys in cer-tain reactions namely, 1) Dissociation of water, 2) Selective hydrogenation of acetylene toethylene and 3) Dissociation of nitric oxide. As a precursor to these calculations, we hadto determine the size of the surface unit cell on which adsorption would take place. Twomain considerations were kept in mind. One, multiple adsorption sites must be available foradsorption of species i.e at least four hollow sites must be available. Two, the size of theslab should ensure moderate computational expense, while describing physics of the systemin the most accurate manner possible. A larger surface unit cell would be most represen-tative of adsorbate coverages observed in these systems. These considerations ensured thatwe could carry out preliminary calculations where we determined the most favorable ad-sorption sites for all the species involved in the reactions. Surface formation energies werecalulated by subtracting the energy of a bulk system from the energy of a surface, both hav-ing identical configuration. This method calculates the energy of incorporating a vacuum inthe z-direction, hence forming a surface. The final surface energy is calculated using a peratom basis for comparison. Based on the results obtained, and the considerations mentionedabove, we chose a surface having configuration 3 × 3 × 3, which is a 3 × 3 surface unit cellhaving 3 layers in the z-direction. The bottom two layers of the slabs were fixed, while thetop layer and the adsorbates were allowed to relax.

3.2 Conventions

The binding energy for the species is calculated as the difference in the energy between anadsorbed system and the energies of the relaxed surface and the gas phase adsorbate. Thisis shown in equation 1.

∆Eadsorption = Eadsorbed system − (Esurface + Egas phase adsorbate) (1)

The reaction energy for each step in the mechanism is calculated as the difference in theenergies of the products and the reactants. Based on these energies a Born-Haber cycle isconstructed. In this cycle, the energies of every individual step in the mechanism is a sumof the energies of all preceding reactions. The reaction energies are plotted, along with theactivation barriers where applicable. The calculated energies refer only to a change in theelectronic energy of the system, and an estimate of the overall heat of the reaction. Theenergies are relative to the initial state energy, and calculation of the actual reaction energywill require calculation of zero point energies and incorporation of thermochemistry into thecalculations.

5

3.3 Dissociation of Water

For this reaction, the surfaces chosen were the fcc(111) facets of Au, AuPd, Ag, AgPd,Cu, CuPd and Pd. Calculations were performed to determine the most favorable sites foradsorption of the dissociated species and the water molecule. The species were adsorbed onthe hollow and atop sites and were allowed to relax, without any constraints. Additionally,we calculated the energies of co-adsorption of the species H and OH on adjacent hollowsites, in order to determine if the interaction energies were significant in the system. Themechanism of dissociation was that put forward by Schneider and coworkers [28]. Thestepwise dissociation mechanism is given as

H2O(g) + ∗ → H2O∗ (1)

H2O∗ → H∗ +OH∗

co-adsorbed (2)

H∗ +OH∗co-adsorbed → H∗ +OH∗

infinite separation (3)

OH∗ → O∗ +H∗co-adsorbed (4)

O∗ +H∗co-adsorbed → O∗ +H∗

infinite separation (5)

3.4 Determination of Activation Barriers

Activation barriers for step 2 and step 4 in the prescribed mechanism were calculated usinga multistep procedure. In the first step of this procedure, the dissociating H and OH groupswere separated by a specific bond length. This bond length was increased over multipleimages such that the dissociating species move towards their preferred adsorption sites. Thebond length between O and the dissociating H atom is fixed, and the groups are allowed allother degrees of freedom in the process of attaining the ground state. The images are allowedto relax using ASE’s inbuilt optimizer BFGS, to determine the geometry of the adsorbates,while constraining the bond length between O and the dissociating H. VASP was used tocalculate the energy and forces for each ionic step, which was then fed back to the BFGSoptimizer, which provided the geometry for the next iteration. This iterative process isdriven by the optimizer till a preset force convergence criteria of 0.05 eV was attained. Theimage with the highest energy provides a good initial state guess for the actual transitionstate, and its corresponding bond length. This image is then used as the initial state forobtaining the transition state using the Improved Dimer Method [29]. This method consistsof two steps. The first step involves a vibrational analysis of the system under consideration.From the step, the eigenvalues which will drive the energy of the system in an upward

6

trajectory are obtained. The second step makes use of these eigenvalues, with the dimermethod algorithm (ibrion=44), to converge at a transition state. The difference in energybetween the initial state and the transition state is the activation barrier for the step. Theseactivation barriers are incorporated in the Born-Haber cycle, providing a complete pictureof the reaction energetics.

3.5 Dissociation of Nitric Oxide

The surfaces chosen to study the dissociation of the simple diatomic molecule NO were Au,AuPd, Ag, AgPd, Cu, CuPd and Pd. For all these reactions, dissociative adsorption data hasbeen comprehensively reported in sources like in the work of Nilsson, Pettersson and Nørskov[30]. Preliminary calculations involved finding the most favorable site for adsorption of allthe species involved in the reaction. This included adsorbing the species at the hollow,atop and bridge sites and relaxing the systems to ground state. NO dissociated through asimple multistep mechanism. We referenced the work of Gajdoš, Hafner and Eichler [31, 32].The mechanism for dissociation consists of various cleavage and recombination steps. Themechanism is given below.

NO(g) + ∗ → NO∗ (1)

NO∗ → N∗ +O∗co-adsorbed (2)

N∗ +O∗co-adsorbed + ∗ → N∗ +O∗

infinite separation (3)

N∗ +N∗infinite separation → N∗ +N∗

co-adsorbed + ∗ (4)

O∗ +O∗infinite separation → O∗ +O∗

co-adsorbed + ∗ (5)

N∗ +N∗ +O∗ +O∗co-adsorbed → N2(g) +O2(g) + 2∗ (6)

Reaction energies for each step in the mechanism were calculated and plotted as part of apotential energy diagram.

3.6 Selective Hydrogenation of Acetylene to Ethylene

This reaction was studied on the fcc(111) facets of Au, AuPd, Ag, AgPd and Pd. Therationale behind choosing these metals is that Pd is most widely used as a selective hy-drogenation catalyst in the industry. Catalysts in which Pd is alloyed with Ag are alsoused extensively in the industry. It has also been reported in literature that Au is ex-tremely selective towards the formation of ethylene [33, 34, 35]. Calculations were per-

7

formed to determine the most favorable sites for adsorption of acetylene, co-adsorbed acety-lene and hydrogen and ethylene. The species were adsorbed on the hollow, atop andbridge sites and were allowed to relax into their most favorable sites, without any con-straints. The geometries of adsorption were constructed based on the work of Bos andWesterterp [36]. The mechanism of hydrogenation was that provided by the work of Neu-rock et al. [37]. The mechanism is of a Langmuir-Hinshelwood type and is given below.

C2H2(g) + ∗ → C2H∗2 (1)

H2(g) + 2∗ → 2H∗ (2)

C2H∗2 +H∗ → C2H

∗3 + ∗ (3)

C2H∗3 +H∗ → C2H

∗4 + ∗ (4)

C2H∗4 → C2H4(g) + ∗ (5)

C2H2(g) +H2(g) → C2H4(g) (overall)

The reaction energy for each step was calculated from the convention mentioned previously,and the reaction pathway diagram was plotted.

4 Results and Discussion

4.1 Density of States

The single atom alloys of Au, Ag and Cu, in which one atom was substituted by Pt, Pd, Ni,Rh, or Ir were investigated for their surface electronic structure. The d-band atom projectedDOS of the single atom for the alloys is shown in Figure 1. The grid plot is such that therows indicate the host metals Au, Ag and Cu, and the columns indicate the single atomsPt, Pd, Ni, Rh and Ir. A general observation from these plots is a sharp feature close tothe Fermi level. There appears to be a sharp increase in the number of occupied states atenergies close to the Fermi level. This peak could be a result of an ineffective mixing ofthe electron densities of the host metal and the single atom. It is also known that the hostmetals may not form homogenous alloys with the single atom metals, as they are immiscible[38]. Another observation made from the plot was the lack of sharp features in the DOSof Cu based alloys. Electron densities of the constituent metals do intermix, and result ina broadened d-band DOS. Within the space of Ag and Au alloys, the single atom is saidto possess gas phase characteristics, and may show an enhanced reactivity. The gas phase

8

characteristics can be seen in Figure 2. This could result in an enhanced catalytic behavior.There may also be an enhanced reactivity shown by the Cu alloys, but we predict that thisenhancement may be lower than that exhibited by the Au and Ag alloys.

Figure 1: d-band density of states for the single atom alloys investigated in this work. The d-band plottedin red is that of the host metal atom, while the d-band in green is that of the single atom in the alloy. Therows of the grid are the host metals, and the columns are the single atoms. For reference, the third plot inthe 2nd row is that of Au-Ir.

Figure 2: d-band density of states of gas phase single atoms.

9

4.2 Adherence to d-band model

Having predicted an enhanced reactivity due to the presence of a sharp feature in the DOS,the next step in the investigation was to see if these alloys adhered to the d-band modeldeveloped by Hammer and Nørskov. We constructed a simple 3 layer 2 × 2 surface unitcell. A hydrogen molecule was adsorbed at a hollow site. The d-band model postulates thatif the energy of the d-band increases, then there will be in upshift in the d-band center ofthe system and the d-band narrows. If there is a decrease in the energy of the d-band thenthere is a broadening of the d-band and a corresponding downshift in the d-band center. Onobserving the d-band characteristics, it can be seen that the d-band center or the averageenergy of the d-band may increase due to the gas phase characteristics of the single atom.However, on plotting the d-band center (ε) vs the d-band width (Wd) of these systems, aswell as that of pure metals in Figure 3, we see there is a general adherence to the d-bandmodel. The scatter of the data points around the linear fit is low in magnitude (~ 0.2 eV).

Figure 3: The plot of d-band width (Wd) vs the d-band center (ε). The general linear dependence is shownby the solid black line.

This result can be rationalized by the similarity in offset of the d-band center and thed-band width due to the single atom. Also, the shift in the d-band center is quite low, interms of eV’s (~ 0.1 eV). Hence, these systems are well within the purview of the d-bandmodel, and their reactivity can be predicted by the same. This is because the d-band center

10

in most cases, determines the reactivity of the surfaces due to the variations in the way thed-band electrons contribute to the bonding capability of the surface [39].

The next course of action was to see if the reactivity of the alloy agrees with that predictedby the d-band model. The dissociative energy of adsorption of a hydrogen molecule Eads wasplotted against the d-band center (ε). This plot is shown in Figure 4. There is a significantamount of scatter in the plot, which tells us that the reactivity of these alloys does notcorrespond directly to the d-band characteristics observed in the previous section. Thisobservation also lends credence to the idea that the gas phase behavior of the Pd atomis significant in determining the reactivity of the alloy. On observing the geometry of theadsorbed system, we observed that in case of the alloy, the hydrogen atom is adsorbed closerto the Pd atom, when compared to adsorption at the hollow site of the host metal surface.The lack of concurrency between figures 3 and 4 is a result of these observations. The d-band model is more inclusive of the overall surface electron density, and its correspondingd-band characteristics, while in the single atom alloy, the behavior of the single atom tendsto dominate.

Figure 4: Plot of dissociative energy of adsorption (Eads) of hydrogen on the single atom alloys and the hostmetals. The linear fit is shown by the solid line, and is for reference.

11

4.3 Choice of Surface Configuration

Having seen the unusual electronic characteristics and reactivity of single atom alloys, wemoved on to testing their applicability and effectiveness in simple dissociation reactions.The reactions studied in this work are 1) Dissociation of Water, 2) Dissociation of NitricOxide and 3) Selective Hydrogenation of Acetylene to Ethylene. It was important to choosea surface configuration which could accomodate and adsorb all species involved in the study,as well as maintain the characteristics of a single atom alloy. The surface formation energiescalculated for a number of surface unit cell configurations are shown in Table 1.

Configuration Surface Formation Energyof Unit Cell (eV)

2 × 2 0.17093 × 2 0.15493 × 3 0.14714 × 3 0.14274 × 4 0.1457

Table 1: Surface formation energies in eV for various surface unit cell configurations

We see that increasing the size of the surface unit cell increases the dilution of the Pdatom, while marginally decreasing the energy required to form the surface. These resultsare on expected lines because smaller unit cells will possess more strain on the surface dueto the mismatch in the lattice constants of Pd and host metal. These strain effects willaffect the reactivity of the surface [40]. Hence, larger surface unit cells were required forthe calculations. Next, we looked at the number of available sites for adsorption on each ofthese configurations, mainly hollow and atop sites. We find that the 3 × 3 surface unit cellhas sufficient number of sites to accomodate all the adsorbate species we encounter in ourinvestigations. It is also large enough to ensure a separation between larger adsorbates likeethylene and vinyl. This is important because it ensures that inter-adsorbate interactionsare negligibly small and do not affect the energetics of the reaction in any way. It was alsoimportant to minimize any steric hindrance that may be present in the system. The 3 ×3 unit cell also ensures that the computational expense is not too much. Hence, for allsubsequent calculations we chose the 3 × 3 surface unit cell with 3 layers in the z-direction.This configuration is shown in Figure 5.

12

Figure 5: Alloy with 3 × 3 × 3 surface unit cell, with the central Au atom replaced by Pd.

4.4 Dissociation of Water

Preliminary calculations were done to identify the site preferences for all the species presentin the dissociation process. The water molecule prefers to adsorb at the atop site for allmetals. However, the difference in adsorption energy for a hollow site and the atop site isapproximately 0.05 eV. Hence, the atop position was chosen as the preferred site for ad-sorption of the water molecule because in the eventual transition state calculation, we foundthat there were lowest site diffusion barriers associated with an atop site, when comparedto adsorption at the hollow site. We also found that OH and H preferentially adsorb atadjacent hollow sites. Grabow et al. reported that OH preferentially adsorbs in a tiltedgeometry at the atop site [41]. However co-adsorbing OH with H appears to stabilize theOH group at the hollow site. For the alloys, adsorption of the water molecule at the atopsite (on the Pd atom) was significantly more favorable than adsorption at the hollow site.OH and H relaxed at the hollow sites, and their geometry varied with alloy used. The Hatom adsorbed at the hollow site, slightly offset towards the Pd atom, while the OH groupaligned itself at a slight angle to the perpendicular. It adsorbed at the hollow site for AuPdand AgPd, and at the bridge site for CuPd. These geometries are shown in Figure 6.

13

Figure 6: Geometries indicating preferred sites for adsorption of the water molecule. Clockwise from topleft, water adsorbed atop the Pd molecule. This geometry is seen in all alloys. The next three panel imagesshow OH and H adsorbed with differing geometries.

Figure 7: Images showing geometries where the OH-H bond length was fixed, and subjected to relaxation.Images number row-wise from left to right. Plot shows the energy of each image relative to the inital state.

The next step in the dissociation process is the bond cleavage of the OH group. Preliminarycalculations determined the preferred adsorption sites for O and H to be adjacent hollowsites, on either side of the Pd atom. Again, both atoms adsorbed close to the Pd atom.For the metals, adsorption was stable at the center of the hollow sites. For the pure metalsurfaces, the preferred sites for adsorption were adjacent hollow sites.

To determine the reaction energy barriers, we relaxed the a series of adsorbate-surfacegeometries in which the OH-H bond length was steadily increased and fixed, in the processof relaxation. In every geometry, the OH and H species were positioned along a path whichwould lead them to their preferred sites, post dissociation. This increase in bond length,and subsequent relaxation is seen as a chain of images in Figure 7. The system shown isthat of AuPd. From the plot, we see that the geometry with the highest energy is the 4thimage. Using this geometry, the transition state was obtained, and consequently the barrier

14

for dissociation. The same procedure was repeated for the second dissociation step, i.e theO-H bond cleavage. The potential energy diagram for the stepwise dissociation was plotted,and is shown in Figure 8.

Figure 8: Potential energy diagram for the dissociation of water on single atom alloys. Lighter shadesrepresent dissociation on pure metal surfaces, while darker shades represent the dissociation pathway oncorresponding single atom alloys.

A quick look at Figure 8 tells us that the single atom alloys of AuPd and AgPd favordissociation by being around 0.5 eV more favorable than pure Au and Ag, for every stepof the dissociation process. CuPd performs worse than Cu, energetically, in this reaction.Transition state energies for the two dissociation steps in the reaction are also less thanthose on the pure metals. This suggests that the Brønsted-Evans-Polanyi relations do holdfor this reaction, and this has been reported in the work of Nørskov and coworkers [42].CuPd performing worse than Cu is an intriguing result. This suggests that the Pd atom isnot enhacing the reactivity of the surface. This deduction is also supported by the DOS plotfor CuPd in Figure 1, which does not show a sharp and pronounced peak. Electron densityseems to have delocalized well across the surface of the alloy. Hence, we see that adding anatom of Pd to Au and Ag surfaces does enhance the reactivity of the suface, while the samecannot be said for Cu.

15

4.5 Dissociation of Nitric Oxide

The surfaces chosen to study the dissociation of nitric oxide were pure metals of Au, Ag, Cuand Pd, and the single atom alloys of AuPd, AgPd and CuPd. Previous work by Gajdoš[32, 31] suggested that the NO molecule could adsorb on the surface in two geometries. Onegeometry is where the axis of the N=O bond is parallel to the surface, and the other is wherethe axis is perpendicular to the surface. Prelminiary calculations for the pure metal surfacesof Cu and Ag showed that NO adsorbs with a stable geometry in which the molecular axis ofNO is perpendicular to the surface, with the nitrogen atom embedded at a hollow site. Wealso found that in a co-adsorbed state oxygen and nitrogen atoms relaxed at adjacent hollowsites. This was also true for isolated nitrogen and oxygen atoms adsorbed on the surfaces.For Au, the results obtained conform perfectly with the geometry predicted by the referencework, i.e the NO molecule adsorbs with a strong tilt at the bridge sites. Geometries for theadsorption of NO on Cu and CuPd are shown in the Figure 9. For the adsorption of nitricoxide on the single atom alloys, the molecules adsorbed in a similar manner to that on puremetal surfaces, with the exception of that on CuPd, on which it adsorbs at a slight angle tothe perpendicular to the surface.

Figure 9: Nitric oxide adsorbs perpendicular to the surface on a pure Cu surface, but adsorbs with a slighttilt on the CuPd surface, at the hollow sites.

16

Figure 10: Potential energy diagram showing the dissociation pathway of NO on pure metal surfaces andsingle atom alloy surfaces.

The potential energy diagram for this reaction was constructed from the the stepwisemechanism described earlier in this work. This is shown in Figure 10. From a quick glancewe deduce that the single atom alloys AuPd and AgPd show good catalytic behavior towardsthe dissociation of NO. CuPd appears to be less active towards the dissociation of NO. Thisobservation is line with those from the dissociation of water. CuPd does not possess anenhanced reactivity at the surface, and this may be attributed to the lack of sharp featurein the DOS of of the d-band of CuPd. Another point to consider is that Pd is a larger atomembedded in a surface having smaller lattice parameters. It is possible that there are straineffects on the surface, which may lead to an apparent decrease in surface reactivity due tothe redistribution of electron density around the Pd atom [43].

4.6 Selective Hydrogenation of Acetylene to Ethylene

The final reaction studied in this work was the selective hydrogenation of acetylene to ethy-lene. This reaction was studied on Au, Ag, AuPd, AgPd and Pd surfaces. We chose thesealloys because they are all employed in the industry as hydrogenation catalysts, as puremetals and alloys. Studies have reported that alloying of Pd with noble metals enhancesselectivity towards ethylene [44]. Single atom alloys have not been reported to be used for

17

this reaction in literature. The first step was to find the preferred adsorption sites for all thespecies involved in the reaction. Bos and Westerterp [36] report that ethylene and acetyleneadsorb as π-bonded complexes or as di-σ-adsorbed complexes on catalyst surfaces. Basedon these reports, and experimental observations, acetylene was adsorbed in the fcc site onthe surface. On relaxation, we found that for Au, AuPd, Ag and AgPd, acetylene adsorbsmost favorably at the fcc site. Ethylene preferentially adsorbs atop the Pd atom for the al-loy surfaces, and also prefers to adsorb atop a pure metal molecule. Comparison of bindingenergies of acetylene at the fcc site, with that at atop sites reveals that for all surfaces exceptPd, the difference in binding energies is 0.015 eV for Au, 0.111 eV for AuPd, 0.105 eV forAg and 0.008 eV for AgPd. Adsorption on AuPd is an order of magnitude more exothermicthan adsorption over Au. Similarly, for ethylene, binding energies differ by 0.07 for Au, 0.23eV for AuPd, 0.018 eV for Ag and 0.019 eV for AgPd, all favoring the atop site. The mostsignificant takeaway from these observations is that both acetylene and ethylene prefer tobond to the metal surface through π-bonded complexes. The lack of stability at bridge sites,for both species rules out the possibility of a di-σ-adsorbed complex. The adsorbed acetylenegeometries over Au and AuPd are shown in Figure 11

Figure 11: Geometries in the first column are acetylene and ethylene adsorbed over Au. Adsorption is weak,adsorbate height is large, and both species are linear in geometry. Geometries in the second column aracetylene and ethylene adsorbed over AuPd. Adsorption is stronger, adsorbate height is smaller, and C-Hbond loses triple bond and double bond character, adsorbing with a bent geometry.

For the catalyst to be selective towards ethylene, over ethylene, the energy required todesorb ethylene from the surface should be lesser than the energy required to hydrogenate itfurther to ethane. Neurock and coworkers [37] suggested that the more strongly bound acety-lene molecule displaces the weakly bound ethylene from the surface, the instant it is formed.With adsorption on AuPd being significantly more than that on Au, the rate of adsorption,and consequently, hydrogenation is bound to be more. Ethylene prefers to adsorb atop sur-face atoms, and the smaller acetylene preferring hollow sites, It is possible for a surface to

18

be covered to a larger extent by the smaller acetylene molecule. The heat of adsorption ofethylene is more than that of ethylene. These occurrences are also bound to increase the rateof reaction, while maintaining a high degree of selectivity. Vinyl intermediates are formed inthe hydrogenation step. These intermediates are adsorbed quite strongly at the surface ofthe single atom alloys, compared the the strength of adsorption of acetylene and ethylene onthe same. The geometry of the intermediates suggest that vinyl again prefers to bond atopthe Pd atom, and exhibits a C-C bond length which is almost that of a C=C. The relaxedgeometries of vinyl, and vinyl co-adsorbed with a hydrogen atom is shown in Figure 12.

Figure 12: Geometries in the first column are vinyl adsorbed over Au. Adsorption is favorable but weak, andvinyl favors a linear geometry. Geometries in the second column ar vinyl adsorbed over AuPd. Adsorptionis stronger, adsorbate height is smaller, and favors a skewed linear geometry.

A Haber cycle was constructed to plot the potential energy diagram for this hydrogenationreaction. The energy diagram is shown in Figure 13. We see that the energetics of thehydrogenation reaction are more favorable for the single atom alloys, than their respectivepure metals.

19

Figure 13: Potential energy diagram for the hydrogenation of acetlyene to ethylene. As mentioned in thelegend accompanying the plot, darker shades represent single atom alloys, while the corresponding lightershades represent the pure metal surfaces. The pathway for Pd is represented in red.

A key point to be noted is that the single atom alloys are very favorable to the formationof the vinyl intermediate, and their reaction pathway from here on starts to mimic thatof Pd. This can be attributed to the intermediates adsorbing strongly on the Pd atom,and showing excellent selectivity towards the formation of ethylene. The co-adsorption stepafter the formation of the vinyl intermediate is indicative of the pathways being very closein energy. This suggests that the Pd atom in the alloys is singularly resposible for drivingthe reaction forward. These Pd doped alloys show excellent catalytic behavior towards theformation of ethylene.

5 Conclusion

Alloys of gold and silver which are with doped one atom of palladium on the surface showenhanced reactivity in the reactions investigated in this work. The single palladium atomshows gas phase characteristics and a correspondingly high reactivity. This is due to theineffective mixing of the electron densities of palladium and that of gold or silver. Thismanifests itself into a sharp peak near the Fermi level, in the d-band density of states for thealloys, projected on the palladium atom. The high reactivity showcased by palladium alters

20

the geometries of adsorbates by making them adsorb closer to the palladium atom, or atopthe palladium atom. This results in adsorption being a more favorable process, which in turnmakes the energetics of dissociation more favorable, in the case of water and nitric oxide. Inthe selective hydrogenation of acetlyene to ethylene, we predict an enhancement of rate ofhydrogenation, with an increased degree of selectivity, due to the more favorable energetics ofadsorption. The copper-palladium single atom alloy does not enhance reactivity, and showsless favorable energetics for the reactions studied in this work. This can be attributed to abetter mixing of surface electron densities of palladium and copper. This is seen through theabsence of the sharp peak in the d-band density of states for the alloy. We also postulate thatbecause Pd is a larger atom in a smaller copper lattice, these strain effects have an impacton the reactivity of the alloy. Finally, we can conclude by saying that these alloys exhibitunusual d-band characteristics, and this behavior manifests itself in enhanced reactivity, thusenhancing the catalytic behavior of single atom alloys.

21

References

[1] Geoffrey Colin Bond. Heterogeneous Catalysis. 1987.

[2] B. Hammer and J. K. Nørskov. Theory of Adsorption and Surface Reactions, pages 285–351. Chemisorp-tion and Reactivity on Supported Clusters and Thin Films. Springer Science + Business Media, 1997.

[3] Heather L. Tierney, Ashleigh E. Baber, John R. Kitchin, and E. Charles H. Sykes. Hydrogen Dissociationand Spillover on Individual Isolated Palladium Atoms. Phys. Rev. Lett., 103(24):nil, 2009.

[4] John R Kitchin. Tuning the electronic and chemical properties of metals: Bimetallics and transitionmetal carbides. PhD thesis, University of Delaware, 2004.

[5] V. R. Stamenkovic, B. Fowler, B. S. Mun, G. Wang, P. N. Ross, C. A. Lucas, and N. M. Markovic.Improved Oxygen Reduction Activity on Pt3Ni(111) Via Increased Surface Site Availability. Science,315(5811):493–497, 2007.

[6] J. Greeley, I. E. L. Stephens, A. S. Bondarenko, T. P. Johansson, H. A. Hansen, T. F. Jaramillo,J. Rossmeisl, I. Chorkendorff, and J. K. Nørskov. Alloys of Platinum and Early Transition Metals AsOxygen Reduction Electrocatalysts. Nature Chemistry, 1(7):552–556, 2009.

[7] Yange Suo, Lin Zhuang, and Juntao Lu. First-Principles Considerations in the Design of Pd-AlloyCatalysts for Oxygen Reduction. Angewandte Chemie, 119(16):2920–2922, 2007.

[8] Adam Holewinski, Juan-Carlos Idrobo, and Suljo Linic. High-Performance Ag-Co Alloy Catalysts forElectrochemical Oxygen Reduction. Nature Chemistry, 6(9):828–834, 2014.

[9] Jeff Greeley and Manos Mavrikakis. Alloy Catalysts Designed From First Principles. Nature Materials,3(11):810–815, 2004.

[10] F. Studt, F. Abild-Pedersen, T. Bligaard, R. Z. Sorensen, C. H. Christensen, and J. K. Norskov.Identification of Non-Precious Metal Alloy Catalysts for Selective Hydrogenation of Acetylene. Science,320(5881):1320–1322, 2008.

[11] John Meurig Thomas, Robert Raja, and Dewi W. Lewis. Single-Site Heterogeneous Catalysts. Ange-wandte Chemie International Edition, 44(40):6456–6482, 2005.

[12] John Meurig Thomas. The Concept, Reality and Utility of Single-Site Heterogeneous Catalysts(SSHCS). Phys. Chem. Chem. Phys., 16(17):7647, 2014.

[13] Matthew B. Boucher, Branko Zugic, George Cladaras, James Kammert, Matthew D. Marcinkowski,Timothy J. Lawton, E. Charles H. Sykes, and Maria Flytzani-Stephanopoulos. Single Atom AlloySurface Analogs in Pd0.18Cu15 Nanoparticles for Selective Hydrogenation Reactions. Phys. Chem.Chem. Phys., 15(29):12187, 2013.

[14] Qiang Fu and Yi Luo. Active Sites of Pd-Doped Flat and Stepped Cu(111) Surfaces for H2 Dissociationin Heterogeneous Catalytic Hydrogenation. ACS Catal., 3(6):1245–1252, 2013.

[15] O.R. Inderwildi, S.J. Jenkins, and D.A. King. When Adding an Unreactive Metal Enhances CatalyticActivity: NOx Decomposition Over Silver-rhodium Bimetallic Surfaces. Surface Science, 601(17):L103–L108, 2007.

22

[16] Felicia R. Lucci, Timothy J. Lawton, Alex Pronschinske, and E. Charles H. Sykes. Atomic Scale SurfaceStructure of Pt/Cu(111) Surface Alloys. J. Phys. Chem. C, 118(6):3015–3022, 2014.

[17] Felicia R. Lucci, Matthew D. Marcinkowski, Timothy J. Lawton, and E. Charles H. Sykes. H2 Activationand Spillover on Catalytically Relevant Pt-Cu Single Atom Alloys. J. Phys. Chem. C, 119(43):24351–24357, 2015.

[18] Felicia R. Lucci, Jilei Liu, Matthew D. Marcinkowski, Ming Yang, Lawrence F. Allard, Maria Flytzani-Stephanopoulos, and E. Charles H. Sykes. Selective Hydrogenation of 1,3-butadiene on Platinum-copperAlloys At the Single-Atom Limit. Nature Communications, 6(nil):8550, 2015.

[19] G. Kresse and J. Hafner. Ab Initio Molecular Dynamics for Liquid Metals. Phys. Rev. B, 47(1):558–561,1993.

[20] G. Kresse and J. Hafner. Ab Initio Molecular-Dynamics Simulation of the Liquid-Metal-amorphous-Semiconductor Transition in Germanium. Phys. Rev. B, 49(20):14251–14269, 1994.

[21] G. Kresse and J. Furthmüller. Efficient Iterative Schemes for Ab Initio Total-Energy Calculations Usinga Plane-Wave Basis Set. Phys. Rev. B, 54(16):11169–11186, 1996.

[22] P. E. Blöchl. Projector Augmented-Wave Method. Phys. Rev. B, 50(24):17953–17979, 1994.

[23] John P. Perdew, Kieron Burke, and Matthias Ernzerhof. Generalized Gradient Approximation MadeSimple. Phys. Rev. Lett., 77(18):3865–3868, 1996.

[24] Ralph Walter Graystone Wyckoff and Ralph WG Wyckoff. Crystal structures, volume 1. InterscienceNew York, 1960.

[25] Hendrik J. Monkhorst and James D. Pack. Special Points for Brillouin-Zone Integrations. Phys. Rev.B, 13(12):5188–5192, 1976.

[26] B. Hammer and J.K. Nørskov. Electronic Factors Determining the Reactivity of Metal Surfaces. SurfaceScience, 343(3):211–220, 1995.

[27] Hongliang Xin, Aleksandra Vojvodic, Johannes Voss, Jens K. Nørskov, and Frank Abild-Pedersen.Effects of d-band Shape on the Surface Reactivity of Transition-Metal Alloys. Phys. Rev. B, 89(11):nil,2014.

[28] Abhijit A. Phatak, W. Nicholas Delgass, Fabio H. Ribeiro, and William F. Schneider. Density FunctionalTheory Comparison of Water Dissociation Steps on Cu, Au, Ni, Pd, and Pt. J. Phys. Chem. C,113(17):7269–7276, 2009.

[29] Andreas Heyden, Alexis T. Bell, and Frerich J. Keil. Efficient Methods for Finding Transition Statesin Chemical Reactions: Comparison of Improved Dimer Method and Partitioned Rational FunctionOptimization Method. The Journal of Chemical Physics, 123(22):224101, 2005.

[30] T. Bligaard and J.K. Nørskov. Heterogeneous Catalysis, pages 255–321. Chemical Bonding at Surfacesand Interfaces. Elsevier BV, 2008.

23

[31] Marek Gajdoš, Jürgen Hafner, and Andreas Eichler. Ab Initio Density-Functional Study of NO on Close-Packed Transition and Noble Metal Surfaces: I. Molecular Adsorption. Journal of Physics: CondensedMatter, 18(1):13–40, 2005.

[32] Marek Gajdoš, Jürgen Hafner, and Andreas Eichler. Ab Initio Density-Functional Study of NO Ad-sorption on Close-Packed Transition and Noble Metal Surfaces: II. Dissociative Adsorption. Journal ofPhysics: Condensed Matter, 18(1):41–54, 2005.

[33] Andreea C. Gluhoi, Johan W. Bakker, and Bernard E. Nieuwenhuys. Gold, Still a Surprising Catalyst:Selective Hydrogenation of Acetylene To Ethylene Over Au Nanoparticles. Catalysis Today, 154(1-2):13–20, 2010.

[34] Y. Azizi, C. Petit, and V. Pitchon. Formation of Polymer-Grade Ethylene By Selective Hydrogenationof Acetylene Over Au/CeO2 Catalyst. Journal of Catalysis, 256(2):338–344, 2008.

[35] Xiaoliang Yan, James Wheeler, Ben Jang, Wen-Yuan Lin, and Binran Zhao. Stable Au Catalysts forSelective Hydrogenation of Acetylene in Ethylene. Applied Catalysis A: General, 487(nil):36–44, 2014.

[36] A.N.R. Bos and K.R. Westerterp. Mechanism and Kinetics of the Selective Hydrogenation of Ethyneand Ethene. Chemical Engineering and Processing: Process Intensification, 32(1):1–7, 1993.

[37] Priyam A. Sheth, Matthew Neurock, and C. Michael Smith. A First-Principles Analysis of AcetyleneHydrogenation Over Pd(111). The Journal of Physical Chemistry B, 107(9):2009–2017, 2003.

[38] W. Bolton. Alloying of metals, pages 39–51. Engineering Materials. Elsevier BV, 1988.

[39] B. Hammer. Special Sites At Noble and Late Transition Metal Catalysts. Topics in Catalysis, 37(1):3–16,2006.

[40] M. Mavrikakis, B. Hammer, and J. K. Nørskov. Effect of Strain on the Reactivity of Metal Surfaces.Phys. Rev. Lett., 81(13):2819–2822, 1998.

[41] L.C. Grabow, A.A. Gokhale, S.T. Evans, J.A. Dumesic, and M. Mavrikakis. Mechanism of the WaterGas Shift Reaction on Pt: First Principles, Experiments, and Microkinetic Modeling. J. Phys. Chem.C, 112(12):4608–4617, 2008.

[42] Shengguang Wang, Burcin Temel, Juan Shen, Glenn Jones, Lars C. Grabow, Felix Studt, ThomasBligaard, Frank Abild-Pedersen, Claus H. Christensen, and Jens K. Nørskov. Universal Brønsted-Evans-Polanyi Relations for C-C, C-O, C-N, N-O, N-N, and O-O Dissociation Reactions. CatalysisLetters, 141(3):370–373, 2010.

[43] J. R. Kitchin, J. K. Nørskov, M. A. Barteau, and J. G. Chen. Role of strain and ligand effects in themodification of the electronic and chemical properties of bimetallic surfaces. Phys. Rev. Lett., 93(15):nil,2004.

[44] Priyam A. Sheth, Matthew Neurock, and C. Michael Smith. First-Principles Analysis of the Effects ofAlloying Pd With Ag for the Catalytic Hydrogenation of Acetylene-ethylene Mixtures. The Journal ofPhysical Chemistry B, 109(25):12449–12466, 2005.

24