Carbon Monoxide Poisoning Resistance and Structural Stability of Single Atom Alloys
Carbon Monoxide Poisoning Resistance and Structural Stability of Single Atom Alloys
Carbon Monoxide Poisoning Resistance and Structural Stability of Single Atom Alloys
https://doi.org/10.1007/s11244-017-0882-1
ORIGINAL PAPER
Abstract
Platinum group metals (PGMs) serve as highly active catalysts in a variety of heterogeneous chemical processes. Unfortu-
nately, their high activity is accompanied by a high affinity for CO and thus, PGMs are susceptible to poisoning. Alloying
PGMs with metals exhibiting lower affinity to CO could be an effective strategy toward preventing such poisoning. In this
work, we use density functional theory to demonstrate this strategy, focusing on highly dilute alloys of PGMs (Pd, Pt, Rh, Ir
and Ni) with poison resistant coinage metal hosts (Cu, Ag, Au), such that individual PGM atoms are dispersed at the atomic
limit forming single atom alloys (SAAs). We show that compared to the pure metals, CO exhibits lower binding strength on
the majority of SAAs studied, and we use kinetic Monte Carlo simulation to obtain relevant temperature programed desorp-
tion spectra, which are found to be in good agreement with experiments. Additionally, we consider the effects of CO adsorp-
tion on the structure of SAAs. We calculate segregation energies which are indicative of the stability of dopant atoms in the
bulk compared to the surface layer, as well as aggregation energies to determine the stability of isolated surface dopant atoms
compared to dimer and trimer configurations. Our calculations reveal that CO adsorption induces dopant atom segregation
into the surface layer for all SAAs considered here, whereas aggregation and island formation may be promoted or inhibited
depending on alloy constitution and CO coverage. This observation suggests the possibility of controlling ensemble effects
in novel catalyst architectures through CO-induced aggregation and kinetic trapping.
Keywords Carbon monoxide · Catalyst stability · Catalyst poisoning · Highly dilute alloy · Platinum group metals ·
Ensemble effects
13
Vol.:(0123456789)
Topics in Catalysis
that these more reactive metals disperse as isolated single We finally summarise our findings and lessons learned
atoms in the surface layer of the host material, the activity in the “Conclusions” section. Our study should provide
of the coinage metal surface can be dramatically enhanced a valuable guide for the choice of catalytically active and
whilst retaining excellent reaction selectivity [11–24]. These selective binary alloy combinations that exhibit improved
single atom alloys (SAAs) of Sykes and co-workers exhibit CO tolerance and structural stability.
tolerance to CO [18] and have been employed to catalyse
hydrogenation [16–20], dehydrogenation [22, 23], C–H acti-
vation and hydrosilylation [25] reactions with high activity
and selectivity, as extended model surfaces and/or as real 2 Computational Details
catalyst nanoparticles.
Temperature programmed desorption (TPD) of CO from 2.1 Density Functional Theory Setup
Pt/Cu(111) SAA model surfaces revealed that CO desorbs at
350 K from this SAA compared to 450 K from pure Pt(111), We perform periodic density functional theory calcu-
indicating weak binding of CO [18]. Under micro-reactor lations using the Vienna ab initio Simulation Package
conditions, it was shown that in the presence of 200 ppm (VASP) version 5.4.1 [35–37] with the projector aug-
CO (a typical industrial concentration in H2 streams) the mented wave (PAW) method to model core ionic potentials
activity of Pt/Cu SAA nanoparticle catalysts for acetylene [38, 39] and the revised Perdew–Burke–Ernzerhof (RPBE)
hydrogenation is reduced twofold, however when compared exchange–correlation functional [40, 41]. RPBE is chosen
to monometallic Pt nanoparticles there was a 15-fold activ- in this instance as it was specifically designed to overcome
ity decrease [18]. It follows that the weak binding of CO to issues of over-binding using other xc-functionals and is
single, isolated Pt atoms in Pt/Cu SAAs compared to that on proven to give CO adsorption energies that are close to
pure Pt is sufficient to give this material notable resistance those from experiment [40, 41]. We use a 3 × 3 × 5 slab
to CO poisoning, despite a relatively low number of active unit cell whereby we fully relax the top-most four layers
sites compared to monometallic Pt [18]. while we fix the bottom-most layer at the RPBE bulk FCC
It is with this in mind that we carry out a detailed theo- lattice constant of the corresponding metal (for SAAs, we
retical study of the effects of CO on an assortment of SAAs use the host material lattice parameters). A vacuum region
comprising single atoms of Ni, Pd, Pt, Rh and Ir doped with thickness of 10 Å separates periodic images in the
into Ag(111), Au(111) and Cu(111) surfaces. We perform z-direction. We model exclusively the (111) surface of all
atomistic calculations using density functional theory (DFT) materials as this is the surface with the lowest surface free
that are then used to parameterize temperature programmed energy for the each host metal in this study [42]. For binary
desorption (TPD) simulations using kinetic Monte Carlo surface alloy calculations, we replace one, two or three
(KMC). Thus, we are able to determine the strength of the surface host atoms with dopant atoms. For calculations
interaction of CO and its relation to temperature of desorp- where the dopant is in the bulk, we replace a single atom
tion from this set of candidate SAAs, with the aim of iden- in the 3rd layer of the unit cell with a dopant atom. We
tifying materials that may exhibit good resistance to CO use a 13 × 13 × 1 Monkhorst–Pack k-point mesh to sample
poisoning. the Brillouin zone and the planewave kinetic energy cutoff
Additionally, we recognize that the presence of adsorb- is set to 400 eV. The Methfessel–Paxton smearing width
ates may induce structural changes in binary alloy materials, is set to 0.1 eV. We ensure electronic self-consistency up
such as segregation of atoms from the bulk into the surface to a tolerance of 10−7 eV and during ionic relaxation, we
layer, as well as promoting aggregation and island formation perform minimization of the Hellmann–Feynman forces
[26–34]. Such changes are caused by differences in adsorp- on free atoms to within a tolerance of 10−2 eV Å−1. We
tion behaviour between an adsorbate on each metallic com- present adsorption energies Eads (mCO) , relative to m gas
ponent of the alloy; these differences can offset or increase phase CO molecules such that
the energy change upon restructuring of the material. Thus,
we perform calculations to determine the segregation and
CO
mCO+slab
Eads (mCO) = ETot slab
− ETot − m ⋅ ETot (g) , (1)
aggregation energies of PGM dopant atoms in highly dilute CO
binary alloys in the absence and presence of CO, allowing where ETot
mCO+slab
, ETot
slab
and are the DFT total ener-
ETot (g)
us to gauge the stability of SAA materials. gies of m CO molecules adsorbed on a slab, the clean slab
The rest of the paper is organised as follows: we first and gas phase CO, respectively. Thus, negative Eads (mCO)
present the setup of the DFT and KMC calculations in means exothermic adsorption. All adsorption configurations
Sect. 2, continuing with Sect. 3 where we explore the of m CO with distinct geometries are given in the support-
interactions of CO with SAAs, in the context of poison- ing information; those with comparable geometries are also
ing resistance and adsorbate-induced structural changes. noted here.
13
Topics in Catalysis
We perform simulations within the graph-theoretical 3.1 CO Adsorption on Pure Metal and SAA Surfaces
KMC framework as implemented in Zacros, version 1.02
[43–45]. We ramp the simulation temperature at a rate of Using DFT with the RPBE xc-functional, we calculate the
1 K s−1 to simulate TPD. The partial pressure of gas phase geometry of a CO molecule chemisorbed on pure metal
CO is set to zero in order to reproduce ultra-high vacuum and SAA (111) surfaces. For the pure metal (111) surfaces,
conditions. The simulation cells consist of (30 × 31) rec- our calculations are in excellent agreement with the works
tangular unit cells with sixfold symmetry. Simulations on of others [47, 48]. We determine that the CO interactions
SAAs use lattices where host metal sites have been ran- with Ag(111), Au(111) and Cu(111) are much weaker than
domly substituted with dopant metal sites giving a final for Ni(111), Pd(111), Pt(111), Rh(111) and Ir(111) (Fig. 1;
dopant atom percentage density of approximately 1%. Table 1).
We initialize the surface with only dopant sites covered For CO adsorption on a top site on Ag(111), we cal-
entirely by CO adsorbates (1:1 dopant:CO coverage), since culate the value of Eads (CO) to be 0.02 eV indicating a
in these materials CO binds significantly more strongly on slightly endothermic binding with the functional used,
the dopant, compared to the host sites. We do not account whereas on the top site of Au(111) there is a marginally
for any CO–CO lateral interactions in TPD simulations exothermic CO adsorption energy Eads(CO) of − 0.05 eV.
on SAA lattices; the high dispersion of single atom sites These values are in agreement with experimental obser-
results in CO adsorbates that reside far from each other vations that CO binds weakly to these surfaces only at
and therefore do not interact. low temperatures [49–51]. CO adsorption on Cu(111)
is exothermic and is most favourable in fcc hollow sites
though much stronger than on Au(111), with an adsorption
2.3 Rate Constants from Density Functional Theory energy of − 0.51 eV with the functional used. For Ni(111),
Pd(111) and Pt(111), the most stable site for adsorption is
In order to perform a KMC simulation, we must first cal- also the fcc hollow site with adsorption energies of − 1.50,
culate rate constants for CO desorption on each surface. − 1.67 and − 1.48 eV. CO adsorption on Rh(111) is most
According to transition state theory (TST), the rate con- favourable on the hcp hollow site with an adsorption
stant kTST of an elementary process can be calculated as energy of − 1.65 eV. Finally, for Ir(111) the most stable
k T QTS
(
ΔE
) adsorption site is the top site with the largest pure metal
kTST = B ⋅ IS exp − a (2) adsorption energy of − 1.83 eV.
h Q kB T
Though our predictions of both the adsorption energy
where kB is the Boltzmann constant, h is Planck’s constant, and preferred adsorption site for CO agree well with the
T is the temperature, QTS and QIS are the molecular partition theoretical work of others, it should be noted that the pre-
functions for the transition state and initial state, respec- diction of adsorption site preference by DFT under the
tively, and ΔEa is the activation barrier. The adsorption of generalized gradient approximation is qualitatively incor-
CO is non-activated, so ΔEa for CO desorption is taken to rect; several explanations and remedies for this phenom-
be Eads (CO) . Moreover, this implies there is no “explicit” enon have been reported, with an excellent discussion by
transition state, therefore a 2D gas phase CO transition state Kresse et al. suggesting this is due to an overestimation of
is assumed (the third translational degree of freedom is the the HOMO–LUMO gap in CO [52, 53]. Low temperature
reaction coordinate: the distance from the surface) [46]. experiments reveal that CO prefers to bind on one-fold
Thus, Eq. (2) for CO desorption becomes rather than three-fold adsorption sites on Cu(111) and
Pt(111), in disagreement with DFT.
kB T QCO(g) We determine that on each of the SAAs considered here,
( )
Eads
kTST = ⋅ CO∗ ⋅ exp . (3) the most favoured adsorption site for CO is the top site of
h Q kB T
the single dopant atom; geometry optimizations starting with
We compute the partition functions Q using the vibra- CO on shared bridge or hollow sites of SAAs typically result
tional frequency data in Table 1, under the harmonic in CO being displaced back to the dopant top site. We report
approximation [46]. The pre-exponential factor in (3) is adsorption energies of CO in the most favourable adsorption
temperature dependent, both due to the thermal factor of sites on SAA (111) surfaces in Fig. 1 and Table 1. The calcu-
kB T/h , but also because QCO(g) and QCO are functions of T
∗
lations on Pt/Cu(111) and Ni/Cu(111) are in good agreement
[46]; this is accounted for in the KMC simulation using with previous works on the adsorption of CO on Ni and Pt
fitted functions of T . impurity atoms at ¼ ML coverage in Cu(111) [54].
13
Topics in Catalysis
The vibrational modes can be described as follows; 𝜈1 C–O stretch, 𝜈2 M–C stretch, 𝜈3∕4 hindered rotations
and 𝜈5∕6 hindered translations. The imaginary frequencies on Ag and Au can be attributed to numerical
artefacts in the calculations of the soft translational modes
-1.00
Rh
Ir
-1.50 Values of 0 < 𝜑(CO) < 1 indicate that CO adsorption
on these SAAs is weaker than on pure dopant surfaces
-2.00
but stronger than on pure host materials, whereas values
where 𝜑(CO) < 0 indicate CO adsorption strength that is
-2.50 Ni SAAs Pd SAAs Pt SAAs Rh SAAs Ir SAAs
greater than on the monometallic dopant. We see for all
Pd- and Pt-doped materials, as well as Ni/Au(111) and Ni/
NiAg
NiAu
NiCu
PdAg
PdAu
PdCu
PtAg
PtAu
PtCu
RhAg
RhAu
RhCu
IrAg
IrAu
IrCu
13
Topics in Catalysis
Temperature (K)
tively. However, Ni/Ag(111) as well as all Rh- and Ir- 550
500
doped SAAs have values of 𝜑(CO) < 0 . CO binds more
450
strongly to these SAAs and thus these materials will not 400
offer any resistance to CO poisoning. However, if we use 350
NiAg
NiAu
NiCu
PdAg
PdAu
PdCu
PtAg
PtAu
PtCu
RhAg
RhAu
RhCu
IrAg
IrAu
IrCu
3.2 Temperature Programmed Desorption
Simulations
Fig. 2 Peak desorption temperatures from KMC simulated TPD of
Reductions in the adsorption strength of CO on SAAs com- CO on SAA (111) surfaces. Temperatures from pure metal simula-
pared to pure dopant surfaces will result in an increased tions are shown as full-horizontal lines whereas corresponding tem-
peratures from experiment [55–58] are shown as dotted-horizontal
tolerance to catalytic poisoning by CO. We quantify this lines [Cu (orange), Ni (pink), Pd (cyan), Pt (blue), Rh (green) and
resistance to poisoning by simulating CO TPDs from (111) Ir (purple); Au and Ag not shown due to weak or no CO binding].
surfaces of metals and alloys of interest using KMC and Experimental SAA temperatures are shown with diamonds [16, 18,
comparing desorption peak temperatures. To evaluate the 21, 55]
quality of our dataset, we compare these peak temperatures
to experimental ones for the pure metals (excluding Ag and a higher adsorption energy on Ni/Ag(111) compared to pure
Au, due to weak or no binding) and several SAAs that have Ni(111), CO desorbs with a peak temperature that is 22 K
been synthesized experimentally (Ni/Cu(111), Pd/Au(111), lower from this SAA. This qualitative deviation from the
Pd/Cu(111) and Pt/Cu(111)). expected relationship between adsorption energy and TPD
peak temperature is attributed to the soft vibrational modes
3.3 Simulated Desorption Peak Temperatures of CO bound to Ni/Ag(111) (Table 1, 𝜈5 and 𝜈6 ) giving a
lower desorption pre-exponential for this material. This dif-
We now examine the thermal desorption of CO on each pure ference offsets the stronger adsorption of CO on Ni/Ag(111)
metal and SAA (111) surface. During a TPD simulation, we [0.07 eV difference versus Ni(111)]. We also simulate TPD
record the coverage of CO* (ΘCO) on the lattice, as well as peak temperatures for Ni/Au(111) and Ni/Cu(111) that are
the number of gas molecules evolved from the surface, at 85 and 34 K below that of Ni(111), respectively.
intervals of 0.25 s. The TPD signal is obtained as a moving For Ir- and Rh-doped SAAs, we simulated CO TPD peak
average of the instantaneous desorption rate, thereby allow- temperatures that are above the corresponding temperatures
ing us to determine the time and temperature (1 K s−1 ramp for desorption from pure Rh(111) and Ir(111), in line with
rate) that the rate of CO desorption is greatest. The corre- stronger adsorption of CO on these SAAs. We calculate 88,
sponding peak desorption temperature, Tsim, is plotted for all 40 and 23 K increases in the TPD peak temperatures for CO
surfaces in Fig. 2, alongside any known experimental data desorbing from Rh/Ag(111), Rh/Au(111) and Rh/Cu(111),
[16, 18, 21, 55–58]. Comparing our simulated TPD peak respectively. For Ir-doped SAAs, the analogous tempera-
temperatures to this data, we can see that there is excellent ture differences are greater, being 204, 143 and 82 K for Ir/
agreement with a mean absolute error of 13 K, providing Ag(111), Ir/Au(111) and Ir/Cu(111), respectively.
good support for the reliability of our model and dataset. Our data suggests that there exists a strong linear correla-
The majority of SAAs show reductions in CO desorp- tion between Eads(CO) and the TPD peak temperature (both
tion temperature over their monometallic dopant analogues, for Texp and Tsim) (Fig. 3). This finding is typical for a first
including all Ni-, Pd- and Pt-doped SAAs. In particular, we order desorption process as predicted by an equation first
see that there is over a 220 K decrease in the desorption derived by Redhead [59]. Though useful, the Redhead equa-
temperature of Pd/Au(111) and Pd/Cu(111) SAAs compared tion can often produce errors as a result of poor estimation
to monometallic Pd, as well as a 185 K decrease with Pd/ of the pre-exponential; for example, arbitrarily choosing a
Ag(111). For Pt doped SAAs, the largest peak temperature pre-exponential value of kBTexp/h (as is typical) gives a mean
reduction of 102 K is simulated for Pt/Cu(111), whereas absolute error in the Redhead value of Eads(CO) of 0.20 eV
Pt/Ag(111) and Pt/Au(111) exhibit desorption temperature compared to the DFT adsorption energy. In our case, we
reductions of over 40 K compared to pure Pt(111). Despite
13
Topics in Catalysis
-1.20
-1.40 those that are as potent as CO, may induce changes in the
-1.60 surface structure of a material through effects such as seg-
-1.80 regation or formation of islands (clusters) on the surface. It
-2.00 E
ads
(CO) = -3.30 10-3 T - 5.95 10-2
sim is with this in mind that we move on to study the stability
R2 = 0.987
-2.20 of the SAA structure with respect to the aforementioned
-2.40 phenomena, both in the absence and presence of CO. This
-2.60 stability analysis will serve as a guide for the experimental
200 300 400 500 600 700 800
Temperature (K)
synthesis of SAAs, highlighting those metal combinations
with an enthalpic preference for the SAA structure over
Fig. 3 Linear correlation of TPD peak temperatures from experi-
other phases, still consistent with high dopant dilution.
ment (blue) and KMC simulation (red) with the adsorption energy
of CO calculated from DFT. Regression equations are shown with
corresponding coefficients of determination (R2). Standard errors in 3.4 Adsorbate‑Induced Structural Changes in SAAs
the slope and intercept for the experimental case are ± 1.11 × 10−4
and ± 4.51 × 10−2, respectively. The analogous standard errors in the
slope and intercept for the KMC simulated case are ± 8.79 × 10−5 and Under realistic conditions, restructuring of the surface of a
± 4.08 × 10−2, respectively catalyst can result in modifications to its function. The low
concentration of dopant atoms in a SAA means that effects
have calculated temperature dependent pre-exponentials such as dopant atom segregation into the bulk or cluster-
using TST, assuming harmonic vibrational modes. This may ing on the surface, would result in a fundamental change in
not always be an accurate approximation, especially when the surface structure of the material and transformation of
considering very soft, frustrated translations or rotations a SAA into some other class of binary alloy. In particular,
in the partition function [60], though good agreement with segregation of a single dopant atom into the bulk may result
experiment supports the use of harmonic TST in this case. in decreased catalytic activity that more closely resembles
We pose that our calibrated linear fitting of the host material. Moreover, clustering of dopant atoms in
the surface layer may result in dimer, trimer and even island
Eads (CO) = − 3.30 × 10−3 ⋅ Tsim − 5.95 × 10−2 (5) formation which will hamper the selectivity and poisoning
may be used for quick extraction of the CO adsorption resistance of the surface.
energy from future experimental work. This fitting is specific Thus, in this section we investigate the thermodynamic
for DFT using the RPBE xc-functional, though good agree- stability of SAAs under vacuum conditions and also in the
ment of our data with experiment supports the choice of this inevitable presence of CO under operating conditions. We
xc-functional in this case. We calculate the mean absolute use DFT to calculate energy changes between the SAA
error in the fitting to be 0.04 eV. structure and other highly dilute analogues where the SAA is
Our DFT results show that highly diluted alloys of sin- buried into the bulk structure of the host or aggregated into
gle atoms, out of the catalytically active group 10 metals, clusters on the surface. By comparing the values of these
dispersed into more inert hosts, out of group 11 metals, can energy changes in the absence versus presence of CO, we
reduce the adsorption strength of CO, compared to the pure quantify the effect of this species on SAA stability.
dopants. In line with these results, KMC simulations reveal
that the CO desorption temperature may be reduced by more 3.5 Surface Segregation
than 220 K in some cases, thereby dramatically reducing the
susceptibility of the surface to poisoning by CO. Decreased We perform calculations for each binary alloy in structures
TPD peak temperatures imply that it is not necessary to heat where a single dopant atom is in the surface layer (i.e. a
these SAA catalytic systems to temperatures as high as on SAA) or immersed in the bulk material. For the latter, we
pure dopant surfaces in order to circumvent CO poison- approximate the “bulk” as a single dopant atom in the 3rd
ing. Thus, one can carry out the reaction on SAAs at lower layer of the host material slab such that the dopant atom is
temperatures, thereby reducing the risk of catalyst sintering fully coordinated to host atoms and the slab is symmetric.
and hampered reaction selectivities. However, increases in The segregation energy ΔEseg is then computed relative to
the CO desorption temperature for Rh- and Ir-doped SAAs the SAA phase such that
13
Topics in Catalysis
0.00
1.20
-0.10
∆ E seg
CO
∆ Eseg (eV)
1.00
-0.20
0.80
-0.30 0.60
-0.40 0.40
0.20
-0.50
0.00
-0.60
NiAg
NiAu
NiCu
PdAg
PdAu
PdCu
PtAg
PtAu
PtCu
RhAg
RhAu
RhCu
IrAg
IrAu
IrCu
-0.70
NiAg
NiAu
NiCu
PdAg
PdAu
PdCu
PtAg
PtAu
PtCu
RhAg
RhAu
RhCu
IrAg
IrAu
IrCu
13
Topics in Catalysis
∆ Eagg(n) (eV)
0.00
addition of the dopant atoms directly into the surface layer.
-0.20
This hypothesis is evidenced to some extent thanks to the
experimental synthesis of Ni/Au, Ni/Cu, Pd/Au and Pt/Au -0.40
-0.80
3.7 Surface Aggregation and Island Formation
-1.00
NiAg
NiAu
NiCu
PdAg
PdAu
PdCu
PtAg
PtAu
PtCu
RhAg
RhAu
RhCu
IrAg
IrAu
IrCu
To evaluate the stability of single isolated dopant atoms
towards aggregation, we vary the molar fraction of CO in
the surface layer and compute the DFT energies for dimer Fig. 6 The energy of aggregation ΔEagg (n) relative to the SAA phase
and trimer configurations on the surface. The aggregation for the clustering of group 10 dopant atoms in the (111) surface of
group 11 metals into dimers (black squares) and trimers where dopant
energy for a cluster of n atoms ΔEagg (n) is given relative to
atoms surround fcc (red up-triangles) and hcp (blue, down-triangles)
a SAA such that sites
ΔEagg (n) = ETot (n) + (n − 1) ⋅ ETot (host) − n ⋅ ETot (SAA);
(8)
where ETot (n) and ETot (host) are the DFT total energies of an By performing calculations with just a single CO mol-
alloy surface with a cluster of n dopant atoms and the pure ecule adsorbed to binary alloy surfaces, we are effectively
host material, respectively. In this case, values of ΔEagg (n) considering the case where the CO partial pressure is suf-
that are negative correspond to a preference for surface clus- ficiently low that full saturation of dopant atoms in dimer
tering, whereas positive values correspond to a preference and trimer configurations is not possible. At this CO cov-
for dopant atom dispersion to the SAA structure. erage, we determine that the CO adsorption energy on an
For the majority of metal combinations we consider here, n-mer Eads
n-mer
(CO) is greater than the CO adsorption energy
our calculations show that single dopant atom isolation is on the corresponding SAA Eads SAA
(CO) for all Ni-doped and
favoured for unit cell concentrations up to 1/3 ML, with the Pd-doped alloys, as well as Pt- and Rh-doped Ag(111) and
exception of Ni-, Ir- and Rh- doped Ag(111), Ir/Au(111) and Au(111) based SAAs (Fig. 7). The most stable adsorption
Ni/Cu(111) which have negative ΔEagg (n) values (Fig. 6). In sites for a single CO molecule on these binary surfaces are
the case of Ni/Cu(111), the values of ΔEagg (2) and ΔEagg (3) the bridge sites connecting two adjacent dopant atoms in
are so small that at temperatures likely to be used in experi-
mental practice, there will be a sufficient entropic tendency
-0.60
to drive surface dopant atoms apart. In fact, configurational Ni Alloys Pd Alloys Pt Alloys Rh Alloys Ir Alloys
-0.80
contributions to the entropy term, which we do not explicitly
consider here, will always favour the SAA structure over -1.00
Au(111), as well as Ir- and Rh-doped Cu(111) and all the -2.00
Pd- or Pt-doped materials).
-2.20
NiAu
NiCu
PdAg
PdAu
PdCu
PtAg
PtAu
PtCu
RhAg
RhAu
RhCu
IrAg
IrAu
IrCu
13
Topics in Catalysis
dimers, and the hollow sites surrounded by three dopant In order for a single CO molecule to induce aggregation,
atoms in triangular trimers. Exceptions to this are Pt/ the adsorption energy of CO on an n-mer island Eads n-mer
(CO)
Cu(111), Ir/Ag(111), Ir/Au(111) and Ir/Cu(111) for which must be more negative than that on a SAA and the difference
it is more favourable for one CO molecule to adsorb on the must also offset positive values of ΔEagg (n) to make the CO
top site of the SAA, dimer or trimer analogues rather than induced aggregation energy ΔEagg m×CO
(n) negative (Fig. 8a);
on two- or three-fold sites.
m×CO SAA n-mer
{ }
ΔEagg (n) = ΔEagg (n) − m ⋅ Eads (CO) − Eads (mCO)
(9)
-0.20
-0.40
-0.60
-0.80
NiAg
NiAu
NiCu
PdAg
PdAu
PdCu
PtAg
PtAu
PtCu
RhAg
RhAu
RhCu
IrAg
IrAu
IrCu
(b) 1.20
Ni Alloys Pd Alloys Pt Alloys Rh Alloys Ir Alloys
1.00
0.80
(n) (eV)
0.60
0.40
2xCO
∆ Eagg
0.20
0.00
-0.20
-0.40
NiAg
NiAu
NiCu
PdAg
PdAu
PdCu
PtAg
PtAu
PtCu
RhAg
RhAu
RhCu
IrAg
IrAu
IrCu
(c) 1.60 Ni Alloys Pd Alloys Pt Alloys Rh Alloys Ir Alloys
1.40
1.20
(n) (eV)
1.00
0.80
3xCO
0.60
∆ Eagg
0.40
0.20
0.00
-0.20
NiAg
NiAu
NiCu
PdAg
PdAu
PdCu
PtAg
PtAu
PtCu
RhAg
RhAu
RhCu
IrAg
IrAu
IrCu
13
Topics in Catalysis
For binary alloys of Pd, Pt, Ir and Rh doped into Cu(111), of CO. For high CO coverages, the repulsive CO–CO lat-
1×CO
ΔEagg (n) is always positive indicating that these metal eral interactions are expected to promote the dispersion of
combinations will have an energetic preference for disper- dopant atoms, yielding SAA structures for all binary metals
sion into the SAA phase when exposed to low CO partial considered here. These findings may present an interesting
pressures. Pt/Ag(111) has values of ΔEagg 1×CO
(n) in dimers opportunity of controlling ensemble effects, by engineering
and trimers that are very close to 0 eV, indicating little or novel materials with primarily dimers or trimers on the sur-
no preference for the SAA phase over other aggregated face of these materials through manipulation of CO partial
phases. All other alloys including Ni/Au(111), Pd/Au(111), pressures [29, 62].
Pt/Au(111), Rh/Au(111) and Pd/Ag(111) whose values
of ΔEagg (n) were all positive (favouring dispersion in the
absence of CO) have ΔEagg 1×CO
(n) values that are negative,
indicating that CO will induce aggregation in these cases. 4 Conclusion
Under reaction conditions, it is most likely that there will
be sufficient CO present to have at least a 1:1 dopant:CO In this study, we have investigated the CO adsorption prop-
ratio and so we go on to investigate if multiple CO adsorb- erties on highly dilute binary alloys of the platinum group
ates may promote aggregation (Fig. 8b, c). The adsorption metals doped into group 11 metal hosts, with focus on the
energy of m CO molecules ( m > 1 ) on m top sites of a cluster fcc(111) surface. Using a combination of DFT with KMC,
of n dopant atoms Eadsn-mer
(mCO) is always notably more nega- we have shown that Ni-, Pd- and Pt-doped SAAs offer resist-
tive than that of a single CO molecule in its most favoured ance to catalytic CO poisoning as evidenced by reduced CO
adsorption site. We also find that Eads n-mer
(mCO) is more posi- adsorption energies and CO peak desorption temperatures
tive than the sum of m ⋅ Eads (CO) for all alloy combinations.
SAA as compared to pure Ni(111), Pd(111) and Pt(111). On the
In these cases, the CO geometries are tilted away from one other hand Rh- and Ir-doped SAAs bind CO more strongly
another and no longer in line with surface normal as is the than pure Rh(111) and Ir(111), indicating these SAAs may
case with one CO molecule, indicting the presence of repul- offer enhanced reactivity over their monometallic coun-
sive lateral interactions. The lateral interactions appear to be terparts, though also a lack of CO tolerance. Additionally,
approximately pairwise additive on trimer clusters. we have evaluated the stability of SAAs compared to other
It follows on that ΔEagg
m×CO
(n) is made more positive by the binary alloy structures and determined that the formation
presence of multiple CO for the highly dilute binary surfaces of dispersed structures (i.e. the SAA phase) is energetically
we consider here. In fact, negative values of ΔEagg (n) are favourable in a number of cases. We have considered the
offset when 2 or 3 CO molecules are co-adsorbed to clus- effect of CO on these alloys and determined that CO favours
tered islands such that ΔEagg m×CO
(n) is positive for all cluster the segregation of single dopant atoms into the surface layer
sizes for the majority of alloy combinations. Thus, for partial of the host material. Moreover, at CO dopant fractional cov-
pressures of CO giving fractional coverages of CO on dopant erage of > 0.5, our calculations suggest that CO will promote
atoms of 0.5–1, dispersion of dopant atoms into the SAA dopant atom dispersion in the surface layer, whereas lower
phase will be favoured for these alloys, rather than aggrega- CO coverages may favour aggregation leading to the forma-
tion into clustered islands. The only SAAs that do not adhere tion of dimers or trimers.
to this statement are Ir/Ag(111) and Ir/Au(111); in these
Acknowledgements M.T.D. is funded by the Engineering and Physical
cases each additional CO molecule adsorbed to the surface Sciences Research Council (EPSRC) UK as part of a Doctoral Train-
reduces the negativity of ΔEagg m×CO
(n) through repulsive inter- ing Grant (Award Ref. 1352369). The authors acknowledge the use
actions between CO molecules, until 3 CO molecules on of the UCL High Performance Computing Facilities (Legion@UCL,
Grace@UCL and Thomas@UCL), and associated support services, in
trimer configurations or 2 CO molecules on dimer configura-
the completion of the computational part of this work. We are grateful
tions is sufficient to make ΔEagg m×CO
(n) positive. to the UK Materials and Molecular Modelling Hub for computational
DFT results in the absence of CO, show that most surface resources, which is partially funded by EPSRC (EP/P020194/1). The
alloys tend to be dispersed forming SAAs at high dopant development of Zacros has been funded under the embedded Computer
Science and Engineering (eCSE) programme of the ARCHER UK
atom dilution. Notable exceptions (at least from an energetic
National Supercomputing Service (eCSE01-001, eCSE10-8). A.M.’s
point of view) are Ni-, Ir- and Rh-doped Ag(111), as well as work was supported by the European Research Council under the
Ni/Cu(111) and Ir/Au(111). Adsorption of a single CO mol- European Union’s Seventh Framework Programme (FP/2007–2013)/
ecule changes this picture for all Au-based SAAs as well as ERC Grant Agreement No. 616121 (HeteroIce project) and the Royal
Society through a Wolfson Research Merit Award. E.C.H.S. thanks
for Ir- and Pd-doped Ag(111) as in the presence of relatively
the Division of Chemical Sciences, Office of Basic Energy Sciences,
low amounts of CO on the surface, the formation of dimers/ CPIMS Program, U.S. Department of Energy, under Grant No. FG02-
trimers may be favoured. On the other hand, for Pd, Pt, Ir 10ER16170 for the financial support of this work.
and Rh doped into Cu(111) in addition to Pt/Ag(111) dis-
persion of dopant atoms is favourable despite the presence
13
Topics in Catalysis
Open Access This article is distributed under the terms of the Creative 25. Miura H, Endo K, Ogawa R, Shishido T (2017) ACS Catal
Commons Attribution 4.0 International License (http://creativecom- 7(3):1543–1553
mons.org/licenses/by/4.0/), which permits unrestricted use, distribu- 26. McCue AJ, Anderson JA (2015) J Catal 329:538–546
tion, and reproduction in any medium, provided you give appropriate 27. Han BC, Van der Ven A, Ceder G, Hwang B-J (2005) Phys Rev
credit to the original author(s) and the source, provide a link to the B 72(20):205409
Creative Commons license, and indicate if changes were made. 28. Ma Y, Balbuena PB (2009) Surf Sci 603(2):349–353
29. Zhu B, Thrimurthulu G, Delannoy L, Louis C, Mottet C, Creuze J,
Legrand B, Guesmi H (2013) J Catal 308(Supplement C):272–281
30. Dhouib A, Guesmi H (2012) Chem Phys Lett 521(Supplement
References C):98–103
31. Guesmi H, Louis C, Delannoy L (2011) Chem Phys Lett
1. Cheng X, Shi Z, Glass N, Zhang L, Zhang J, Song D, Liu Z-S, 503(1):97–100
Wang H, Shen J (2007) J Power Sources 165(2):739–756 32. Sansa M, Dhouib A, Guesmi H (2014) J Chem Phys
2. Trens P, Durand R, Coq B, Coutanceau C, Rousseau S, Lamy C 141(6):064709
(2009) Appl Catal B 92(3):280–284 33. Dhifallah M, Dhouib A, Aldulaijan S, Renzo FDI, Guesmi H
3. Vogel W, Lundquist L, Ross P, Stonehart P (1975) Electrochim (2016) J Chem Phys 145(2):024701
Acta 20(1):79–93 34. Yang K, Yang B (2017) Phys Chem Chem Phys
4. Baschuk JJ, Li X (2001) Int J Energy Res 25(8):695–713 19(27):18010–18017
5. Wang J, Liu C, Lushington A, Cheng N, Banis MN, Riese A, 35. Kresse G, Furthmüller J (1996) Comput Mater Sci 6(1):15–50
Sun X (2016) Electrochim Acta 210(Supplement C): 285–292 36. Dong W, Kresse G, Furthmüller J, Hafner J (1996) Phys Rev B
6. Groß A (2006) Top Catal 37(1):29–39 54(3):2157–2166
7. Xu X, Zhang Y, Yang Z (2015) Appl Surf Sci 357(Part 37. Kresse G, Hafner J (1993) Phys Rev B 47(1):558–561
B):1785–1791 38. Kresse G, Joubert D (1999) Phys Rev B 59(3):1758–1775
8. Wang M, Fu Z, Yang Z (2013) Phys Lett A 377(34):2189–2194 39. Blöchl PE (1994) Phys Rev B 50(24):17953–17979
9. Schmidt TJ, Jusys Z, Gasteiger HA, Behm RJ, Endruschat U, 40. Hammer B, Hansen LB, Nørskov JK (1999) Phys Rev B
Boennemann H (2001) J Electroanal Chem 501(1):132–140 59(11):7413–7421
10. Ho SF, Mendoza-Garcia A, Guo S, He K, Su D, Liu S, Metin 41. Perdew JP, Burke K, Ernzerhof M (1996) Phys Rev Lett
O, Sun S (2014) Nanoscale 6(12):6970–6973 77(18):3865–3868
11. Tierney HL, Baber AE, Kitchin JR, Sykes ECH (2009) Phys Rev 42. Tran R, Xu Z, Radhakrishnan B, Winston D, Sun W, Persson KA,
Lett 103(24):246102 Ong SP (2016) Sci Data 3:160080
12. Tierney HL, Baber AE, Sykes ECH (2009) J Phys Chem C 43. Stamatakis M, Vlachos DG (2011) J Chem Phys 134(21):214115
113(17):7246–7250 44. Stamatakis M, Chen Y, Vlachos DG (2011) J Phys Chem C
13. Baber AE, Tierney HL, Lawton TJ, Sykes ECH (2011) Chem- 115(50):24750–24762
CatChem 3(3):607–614 45. Nielsen J, d’Avezac M, Hetherington J, Stamatakis M (2013) J
14. Boucher MB, Zugic B, Cladaras G, Kammert J, Marcinkowski Chem Phys 139(22):224706
MD, Lawton TJ, Sykes ECH, Flytzani-Stephanopoulos M 46. Stamatakis M, Vlachos DG (2012) ACS Catal 2(12):2648–2663
(2013) Phys Chem Chem Phys 15(29):12187–12196 47. Liu W, Zhu YF, Lian JS, Jiang Q (2007) J Phys Chem C
15. Lucci FR, Lawton TJ, Pronschinske A, Sykes ECH (2014) J 111(2):1005–1009
Phys Chem C 118(6):3015–3022 48. Marek G, Andreas E, Jürgen H (2004) J Phys: Condens Matter
16. Marcinkowski MD, Jewell AD, Stamatakis M, Boucher MB, 16(8):1141
Lewis EA, Murphy CJ, Kyriakou G, Sykes ECH (2013) Nat 49. McElhiney G, Papp H, Pritchard J (1976) Surf Sci 54(3):617–634
Mater 12(6):523–528 50. Outka DA, Madix RJ (1987) Surf Sci 179(2):351–360
17. Kyriakou G, Boucher MB, Jewell AD, Lewis EA, Lawton TJ, 51. Kulawik M, Rust HP, Nilius N, Heyde M, Freund HJ (2005) Phys
Baber AE, Tierney HL, Flytzani-Stephanopoulos M, Sykes ECH Rev B 71(15):153405
(2012) Science 335(6073):1209–1212 52. Kresse G, Gil A, Sautet P (2003) Phys Rev B 68(7):073401
18. Liu J, Lucci FR, Yang M, Lee S, Marcinkowski MD, Ther- 53. Feibelman PJ, Hammer B, Nørskov JK, Wagner F, Schef-
rien AJ, Williams CT, Sykes ECH, Flytzani-Stephanopoulos M fler M, Stumpf R, Watwe R, Dumesic J (2001) J Phys Chem B
(2016) J Am Chem Soc 138(20):6396–6399 105(18):4018–4025
19. Lucci FR, Liu J, Marcinkowski MD, Yang M, Allard LF, 54. Ruban A, Hammer B, Stoltze P, Skriver HL, Nørskov JK (1997)
Flytzani-Stephanopoulos M, Sykes ECH (2015) Nat Commun J Mol Catal A 115(3):421–429
6:8550 55. Kirstein W, Krüger B, Thieme F (1986) Surf Sci 176(3):505–529
20. Pei GX, Liu XY, Wang A, Lee AF, Isaacs MA, Li L, Pan X, 56. Miller JB, Siddiqui HR, Gates SM, Russell JN Jr, Yates JT Jr,
Yang X, Wang X, Tai Z, Wilson K, Zhang T (2015) ACS Catal Tully JC, Cardillo MJ (1987) J Chem Phys 87(11):6725–6732
5(6):3717–3725 57. Guo X, Yates JT Jr, J Chem Phys 90(11):6761–6766
21. Lucci FR, Darby MT, Mattera MFG, Ivimey CJ, Therrien AJ, 58. Collins DM, Spicer WE (1977) Surf Sci 69(1):85–113
Michaelides A, Stamatakis M, Sykes ECH (2016) J Phys Chem 59. Redhead PA (1962) Vacuum 12(4):203–211
Lett 7(3):480–485 60. Campbell CT, Sprowl LH, Árnadóttir L (2016) J Phys Chem C
22. Shan J, Lucci FR, Liu J, El-Soda M, Marcinkowski MD, Allard 120(19):10283–10297
LF, Sykes ECH, Flytzani-Stephanopoulos M (2016) Surf Sci 61. Ruban AV, Skriver HL, Nørskov JK (1999) Phys Rev B
650:121–129 59(24):15990–16000
23. Boucher MB, Marcinkowski MD, Liriano ML, Murphy CJ, Lewis 62. Bliem R, van der Hoeven JES, Hulva J, Pavelec J, Gamba O, de
EA, Jewell AD, Mattera MFG, Kyriakou G, Flytzani-Stephanop- Jongh PE, Schmid M, Blaha P, Diebold U, Parkinson GS (2016)
oulos M, Sykes ECH (2013) ACS Nano 7(7):6181–6187 Proc Natl Acad Sci USA 113(32):8921–8926
24. Wang Z-T, Darby MT, Therrien AJ, El-Soda M, Michaelides
A, Stamatakis M, Sykes ECH (2016) J Phys Chem C
120(25):13574–13580
13