Research Articles
Chemie
Single-Atom Catalysis Hot Paper
Surface Density Dependent Catalytic Activity of Single Palladium
Atoms Supported on Ceria**
Yongseon Kim+, Greg Collinge+, Mal-Soon Lee, Konstantin Khivantsev, Sung June Cho,*
Vassiliki-Alexandra Glezakou, Roger Rousseau,* Janos Szanyi,* and Ja Hun Kwak*
Abstract: The analogy between single-atom catalysts (SACs)
and molecular catalysts predicts that the specific catalytic
activity of these systems is constant. We provide evidence that
this prediction is not necessarily true. As a case in point, we
show that the specific activity over ceria-supported single Pd
atoms linearly increases with metal atom density, originating
from the cumulative enhancement of CeO2 reducibility. The
long-range electrostatic footprints (  1.5 nm) around each Pd
site overlap with each other as surface Pd density increases,
resulting in an observed deviation from constant specific
activity. These cooperative effects exhaust previously active O
atoms above a certain Pd density, leading to their permanent
removal and a consequent drop in reaction rate. The findings
of our combined experimental and computational study show
that the specific catalytic activity of reducible oxide-supported
single-atom catalysts can be tuned by varying the surface
density of single metal atoms.
Introduction
Supported single-atom catalysts (SACs) have attracted
considerable attention owing to their unique activity, efficient
utilization of supported noble metals, and potential to bridge
the gap between homogeneous and heterogeneous catalysis. [1]
Increasing the surface metal density of SACs is sought after to
increase the density of active sites and the mass activity of
industrial catalysts. [2] Meanwhile, it is commonly assumed that
the active sites of SACs exhibit constant specific activity and
the activity decreases at high surface metal density due to
agglomeration of single atoms and nanoparticle formatio-
n.[1a, 2g, 3] Here using CO oxidation as a probe reaction, we
show that, contrary to this assumption, the specific activity of
Pd/CeO2 SACs instead increases with Pd density due to
[*] Y. Kim, [+] Prof. J. H. Kwak
Department of Chemical Engineering
Ulsan National Institute of Science and Technology (UNIST)
50 UNIST-gil, Ulsan 44919 (Republic of Korea)
E-mail: jhkwak@unist.ac.kr
Dr. G. Collinge,[+] Dr. M.-S. Lee, Dr. V.-A. Glezakou, Dr. R. Rousseau
Physical and Computational Sciences Directorate and Institute for
Integrated Catalysis, Pacific Northwest National Laboratory
Richland, WA 99354 (USA)
E-mail: Roger.Rousseau@pnnl.gov
Dr. G. Collinge, [+] Dr. M.-S. Lee, Dr. K. Khivantsev, Dr. V.-A. Glezakou,
Dr. R. Rousseau, Dr. J. Szanyi
Institute for Integrated Catalysis
Pacific Northwest National Laboratory
Richland, WA 99354 (USA)
Angew. Chem. Int. Ed. 2021, 60, 22769 – 22775
Angewandte
How to cite: Angew. Chem. Int. Ed. 2021, 60, 22769– 22775
15213773, 2021, 60, Downloaded from https://onlinelibrary-wiley-com.ezproxy.lib.vt.edu By Virginia Tech- on [28/10/2021]. Re-use and distribution is strictly not permitted, except for Open Access articles
International Edition: doi.org/10.1002/anie.202105750
German Edition: doi.org/10.1002/ange.202105750
synergistic nonlocal effects. This phenomenon, however, is
support-dependent as Pd SACs supported on non-reducible
g-Al2O3 exhibit constant specific activity for CO oxidation.
Numerous SACs have been characterized and reported to
exhibit high activity for the CO oxidation reaction including
the seminal Pt/FeOx catalyst.[4] The charge state of the single-
atom centers is commonly invoked to explain this activity,[5]
reportedly enhancing CO and/or O2 adsorption. Subsequent-
ly, the choice of support and its redox state has been shown to
impact CO oxidation activity greatly,[6] with irreducible
supports, such as g-Al2O3 or MgO, identified as comparatively
inactive[7] or unstable[8] without modification such as the
induction of defect sites[9] or the addition of La[10] or other
heteroatom substituents.[11] Conversely, single atoms on
reducible supports show considerable CO oxidation activi-
ty.[12] Some supports have even been reported to allow the
stabilization of metallic single atoms.[13] Spezzati et al. re-
ported high CO-oxidation activity of isolated Pd atoms
supported on CeO2 (111) and identified a PdOx species as
the active site.[14] However, despite considerable research
efforts on SACs, the effect of surface density of single metal
atoms on catalytic activity has not been elucidated.
In this study, we report the synergistic effect between
single Pd atoms supported on CeO2. H2 -temperatrue pro-
grammed reduction (TPR) and in situ Raman studies em-
ployed in this work clearly show that the reducibility of the
support is enhanced as the density of single Pd atoms
increases. Density functional theory (DFT) calculations
confirm the lowering in oxygen vacancy formation energies
as the surface Pd density increases, while CO and O2
adsorption energies remain unaffected. This suggests that
the observed specific activity trends are due to the activation
of lattice O by Pd. Charge and spin density analysis allow us to
E-mail: Janos.Szanyi@pnnl.gov
Prof. S. J. Cho
Department of Chemical Engineering
Chonnam National University
77 Yongbong-ro, Buk-gu, Gw angju 61186 (Republic of Korea)
E-mail: sjcho@chonnam.ac.kr
[+] These authors contributed equally to this work.
[**] A previous version of this manuscript has been deposited on
a preprint server (https ://doi.org/10.26434/chemrxiv.14312780.v1).
Supporting information and the ORCID identification number(s) for
the author(s) of this article can be found under :
https://doi.org/10.1002/anie.202105750.
 2021 Wiley-VCH GmbH 22769
 Angewandte
Research Articles Chemie

15213773, 2021, 60, Downloaded from https://onlinelibrary-wiley-com.ezproxy.lib.vt.edu By Virginia Tech- on [28/10/2021]. Re-use and distribution is strictly not permitted, except for Open Access articles
identify the active site as an overoxidized Pd>2+ in a [PdO4 ]
square planar complex consistent with X-ray absorption
spectroscopy (XAS). The Pd atom is overoxidized, past + 2,
and its coordinating oxygens are valence unsaturated, making
[PdO 4] an excellent oxidizer with a considerable and impor-
tantly, cumulative, range of effect. Thus, as the surface density
of [PdO4] complexes increases, their individual oxidation
power also increases. The support mediates this process,
allowing for the shuttling of excess charge to nearby [PdO4],
explaining why g-Al2 O 3, a non-reducible support, does not
exhibit the same synergy.

Results and Discussion

CO oxidation on Pd/CeO 2

The effect of surface single-atom density on the catalytic

properties of Pd/CeO2 (for characterization of CeO2 see the
Supporting Information, Figures S1–S3) and Pd/Al2O 3 cata-
lysts was investigated in the model reaction of CO oxidation
as a function of surface Pd density, displayed in Figure 1 a
(Figure S4a shows the relationship as a function of Pd weight
loading).
The specific activity of Pd/CeO2 catalysts increased
linearly with surface Pd density up to  0.8 Pd/nm2. This
behavior was observed for both initial (Figure S4b) and
steady state activities (Figure 1 a). In the contrasting case of
Pd/Al 2O3 catalysts, the specific activity was constant below
 0.034 Pd/nm 2, in agreement with previously reported re- Figure 1. a) Specific activity of CO oxidation at steady state as
sults.[1a, 3a,b] Specific activity as a function of time for a selected a function of surface Pd density of Pd/CeO2 and Pd/Al2 O3 . b) Specific
series of Pd/CeO 2 catalysts is also shown in Figure 1 b (for the activity of Pd/CeO2 as a function of time. c) Arrhenius plots of 0.1, 0.2
entire series of Pd/CeO2 and Pd/Al2O3 catalysts, see Fig- and 0.4 Pd/CeO2 catalysts for CO oxidation.
ure S4c,d). The specific activity was obtained while the CO
conversion is kept below 10% (Figure S4e,f). Pd/CeO2
catalysts with Pd surface densities in the range of 0.1– between Pd (z = 46) and Ce (z = 58), we could not observe
0.8 Pd/nm 2 deactivated slowly while catalysts with surface atomically dispersed Pd supported on CeO2 (Figures S5–
densities between 1.2–4 Pd/nm 2 underwent rapid deactiva- S8).[16] For 0.068 Pd/Al2O 3, atomically dispersed Pd were
tion. Sintering of Pd atoms/small metal clusters and, as we will observed (Figure S9). At higher Pd surface densities (e.g.,
show, the permanent removal of previously catalytically 0.20 Pd/Al2 O 3) distinctive Pd nanoparticles were observed
active lattice O atoms are primarily responsible for this (Figure S10).
activity loss. The apparent activation energies for Pd/CeO2 Since we could not directly confirm the presence of
catalysts with Pd surface densities of 0.1, 0.2 and 0.4 Pd/nm 2, atomically dispersed Pd on CeO2 with AC -HAADF-STEM,
estimated from Arrhenius plots in Figure 1 c, were 45, 35 and additional structural analysis was carried out. Pd K-edge
40 kJ mol 1, respectively. These values are consistent with X-ray absorption near-edge structure (XANES) (Figure S11)
a Mars-van Krevelen (MvK) mechanism. [15] It is important to and extended x-ray absorption fine structure (EXAFS) data
note, however, that based on these data, it is the preexpo- (Figures S12 and S13) were further used to estimate structural
nential factor that is enhanced as surface Pd density increases, parameters (Table 1). The Fourier transformed EXAFS
suggesting the observed effect is entropic in origin and not spectra show primarily a Pd-O scattering at 1.99  with
due to a lowering of the activation
barrier.
Aberration-corrected high an- Table 1: Structural parameters from Pd K-edge EXAFS analysis for 0.8 and 4 Pd/CeO2 .[a]
gle annular dark field scanning Sample Pair CN [b] r s2 DE R-factor
[c]
transmission electron microscopy [] [pm2] [d] [eV][e] [%]
(AC-HAADF-STEM) imaging 0.8 Pd/CeO 2 Pd-O 3.8  0.6 1.99  0.02 52  11 4.9  2.5 3.0
was used to visualize Pd atoms/ 4 Pd/CeO 2 Pd-O 3.7  0.5 1.99  0.01 11  18 4.3  1.2 1.3
particles in the CeO2 and Al2O 3- [a] Many-body reduction factor ; S 2 was set to 0.86. The number of the free fitting parameter for EXAFS
0
supported catalysts (Figures S5– data ; 11 was always less than the number of independent points given by the Nyquist Theorem.
S10). Because of low z-contrast [b] Coordination number. [c] Bond distance. [d] The Debye-Waller factor and [e] energy shift.

22770 www.angewandte.org  2021 Wiley-VCH GmbH Angew. Chem. Int. Ed. 2021, 60, 22769 – 22775
 Angewandte
Research Articles Chemie

15213773, 2021, 60, Downloaded from https://onlinelibrary-wiley-com.ezproxy.lib.vt.edu By Virginia Tech- on [28/10/2021]. Re-use and distribution is strictly not permitted, except for Open Access articles
coordination number of 3.8  0.6 and 3.7  0.5 for 0.8 and
4 Pd/CeO2, respectively. In bulk PdO, Pd-Pd scattering
appears around 3.06 and 3.45  with 4 and 8 coordination
number, respectively. [17] However, no Pd-Pd bond was
observed in either 0.8 or 4 Pd/CeO 2 samples. Interestingly,
the corresponding XANES is shifted to higher energies
relative to that of PdO, indicating Pd to be in an oxidation
state greater than + 2. These results suggest that Pd is highly
dispersed with Pd-O coordination number around 4 in an
overoxidized charge state.

CeO 2 Redox Properties Modified by Pd loading

Since CeO 2-based CO oxidation catalysts follow the MvK

mechanism, [15e, 18] the activity must be related to the redox
properties of the catalysts. To gain information about this, we
conducted H2-TPR (Figure 2 a for Pd/CeO2). On bare CeO 2,
surface and bulk reduction peaks at 48088C and above 700 8C
were observed, respectively.[15e] With Pd present, the TPR
profile dramatically changes. The intensity of the reduction
peak at 480 8 C decreased and a new reduction peak appeared
at 162 8C for 0.1 Pd/CeO2 . Interestingly, the reduction peak
gradually shifted toward lower temperature with increasing
Pd surface density (Figure 2 b), ultimately appearing at 93 8C
for 0.4 Pd/ CeO 2 .[19] At even higher surface Pd density, very Figure 2. a) H2 -TPR profiles of Pd/CeO2 and b) reduction temperatures
sharp and intense reduction peak arose under 50 8C. as a function of surface Pd density.
H2-TPR clearly reveals that the hydrogen consumption
(Supporting Information, Table S1) far exceeded the amount
of Pd loaded as CeO2 support is reduced. This result suggests Raman spectroscopy on the 0.2, 0.4 and 4 Pd/CeO2 catalysts at
a causal relationship between reducibility of CeO 2 support room temperature (RT), 6288C and 125 8 C. The results
and surface Pd density. To investigate the origin of initial CO obtained are displayed in Figure 3 (results of 0.1 and
oxidation activity drop shown in Figure S4c, H2 -TPR was also 1.2 Pd/CeO 2 can be found in Figure S19). All samples were
conducted on spent catalysts after CO oxidation for 2 h at pretreated at 35088C in 20 % O 2/He for 0.5 h. A Raman band
75 8C (Figure S14, Table S2). At higher Pd loading ( 0.8 representative of CeO2 at 450–470 cm 1 was observed along
Pd/nm 2), the reduction profile significantly changed, which with weak bands at 270, 600 and 1174 cm 1 (Figures S20 and
can be explained by sintering and partial reduction of Pd S21). [18a, 22] In CO + O2 , a distinct band at 830 cm1 was
during reaction (Table S3, Figures S15–S17). This may be also observed for 0.2 and 0.4 Pd/CeO 2 catalysts. On 4 Pd/CeO2
due to lattice oxygen deactivation, which we will
return to shortly. We also conducted H 2-TPR on
Pd/Al 2O3 (Figure S18). We could hardly observe
H2 consumption on 0.017–0.068 Pd/Al2O3 due to
very low Pd loading. At higher Pd loading
(0.20 Pd/Al2O 3 and 0.68 Pd/Al2O 3), a broad re-
duction peak centered at  87 8 C was observed.
At the same time, additional sharp reduction peak
at 23.5 8 C was observed on 0.68 Pd/Al 2O3, well
beyond the point at which nanoparticles form. A
negative peak at 60 8C of 0.68 Pd/Al 2O3 is attrib-
uted to decomposition of palladium hydride
species which is hardly formed on small parti-
cles.[20] The absence of reduction peak at 23.5 8 C
and feature of hydride decomposition on 0.017–
0.20 Pd/Al2 O3 indicates that Pd is highly dispersed
Figure 3. In situ Raman spectra of a) 0.2 Pd/CeO2, b) 0.4 Pd/CeO 2, c) 4 Pd/CeO2.
on 0.017–0.20 Pd/Al 2O 3.[20b, 21] Notably, Pd/Al2O 3
Conditions : (1) room temperature (RT)—He, (2) RT 1 %CO + 2.5 %O2 /He, (3) RT
catalysts do not show systematic change of
2.5 %O2/He, (4) 62 8 C/%CO + 2.5 %O2/He, (5) 62 8C2.5 %O 2/He, (6) 125 8C/
reducibility with Pd loading. %CO + 2.5 %O2/He, and (7) 125 8C2.5 %O 2 /He). Raman intensity was normalized
To investigate the observed CeO2 reduction with F2g (450–470 cm 1) band. The same scale of y-axis is used across Figure
under reaction conditions, we performed in situ panels.

Angew. Chem. Int. Ed. 2021, 60, 22769 – 22775  2021 Wiley-VCH GmbH www.angewandte.org 22771

Research Articles
catalyst, this Raman band was rather intense after pre-
treatment and He flushing, but this band decreased in the
presence of CO + O2 at room temperature. The 830 cm 1
band has been attributed to adsorbed peroxo species (O 22)
on isolated two-electron defect sites, indicating oxygen
vacancies are formed on Pd/CeO 2 under reaction condi-
tions. [22a] As CO is removed from the gas stream, the
characteristic Raman band for the peroxo species disap-
peared on the single-atom-containing catalysts (0.2 and
0.4 Pd/CeO2 ). At 62 8 C, the Raman band of peroxo species
re-appeared in the CO + O 2 stream. The most noticeable
feature is the development of new bands at 1104 and
1514 cm1 for 0.4 Pd/CeO2 . These bands are assigned to
superoxide (O 2) and weakly bound dioxygen (O 2d) species,
respectively.[23] According to Hess et al., [23] those two dioxy-
gen species can be attributed to the creation of oxygen
vacancies during CO oxidation and the transfer of electron to
adsorbed oxygen molecule. As the temperature was further
increased to 125 8C, the peroxo peak (830 cm 1) appeared in
a weak intensity under CO + O2 flow. Features for superoxide
(1104 cm 1) and weakly bound dioxygen (1514 cm 1) ap-
peared on 0.2 Pd/CeO2 , but their intensities were much lower
than in the 0.4 Pd/CeO2 catalyst. When CO was taken out of
the gas stream, the two features became very prominent on
both 0.2 and 0.4 Pd/CeO2 catalysts, which might be due to
reaction of the charged oxygen species (O2 or O 2 d) with gas
phase CO or removal of adsorbed CO on oxygen vacancies.
On the other hand, as the surface Pd density further increased
to 4, the intensity of the dioxygen species became significantly
lower than over 0.4 Pd/CeO2 . The comparison of Raman
spectra of 0.2 and 0.4 Pd/CeO2 catalysts clearly shows the
facile formation of oxygen vacancies on CeO2 surface with
higher surface density of single Pd atoms in reaction
conditions. This result manifests itself in improved oxygen
vacancy formation with surface density of atomic Pd,
consistent with results from H2-TPR.
In situ transmission FTIR spectroscopy was used to
monitor both the nature of Pd species present on the CeO2
support after oxidation at 623 K and the variation of the
reducibility of CeO 2 as a function of Pd loading using CO as
a probe molecule. The IR spectra collected during sequential
CO adsorption over Pd/CeO2 catalysts with Pd loadings of 0.2,
0.4 and 4 Pd/nm2 are displayed in Figure S22. IR bands
characteristic of Pd ions are observed exclusively for the
0.2 Pd/CeO2 sample with characteristic IR features centered
between 2098 and 2154 cm1. The 0.4 Pd/CeO 2 catalyst
exhibits similar IR features of adsorbed CO. However, the
intensity ratio of the high and low frequency bands changes
dramatically as the Pd loading increases, indicating changes in
the population of Pd ions in different environments on(in) the
CeO 2 surface. The series of IR spectra collected from the
4 Pd/CeO2 sample is fundamentally different from those of
the other two catalysts: they are dominated by IR features of
CO adsorbed on metallic Pd sites (Pd particles). IR bands at
2082–2098 and at 1917–1952 cm 1 represent CO molecules
bound linearly and in a bridging configuration to Pd0 centers
of metal particles, respectively. It is very interesting to note
that upon CO introduction onto all three samples the
formation of CO2 is clearly observed. However, while CO 2
22772 www.angewandte.org  2021 Wiley-VCH GmbH
Angewandte
Chemie
15213773, 2021, 60, Downloaded from https://onlinelibrary-wiley-com.ezproxy.lib.vt.edu By Virginia Tech- on [28/10/2021]. Re-use and distribution is strictly not permitted, except for Open Access articles
is present as an adsorbed species (sharp IR band at 2351 cm1 )
on the atomically dispersed Pd/CeO2 samples, we only
observe the formation of gas phase CO2 (2349 cm 1) over
the Pd particles. We propose that the origin of CO 2 formed
over the single-atom-containing samples is fundamentally
different from that formed on supported Pd particles. The
results of DFT calculations, which will be discussed below,
clearly demonstrate that the presence of single Pd atoms
significantly reduces the energy of oxygen vacancy formation
on CeO2 . This, in turn, means that certain oxygen species on
the CeO2 surface in the vicinity of single Pd atoms become
very reactive toward CO, and readily produce CO2. Since
these experiments are carried out at room temperature the
thus-formed CO 2 stays on the CeO2 surface interacting with
the oxygen vacancies. In contrast, when Pd particles are
present on the CeO2 support, CO 2is formed by the reaction of
CO with the surface oxide layer formed on the Pd particles
during the 623 K oxidation prior to CO adsorption. This
reaction has two consequences: 1. Metallic Pd formation upon
the removal of the surface oxide layer (the IR bands of
adsorbed CO are consistent with metallic adsorption sites)
and 2. Fo rmation of gas phase CO2 (and even some surface
carbonates via the reaction between CO2 and the ceria
support). However, no adsorbed CO2 is seen, as the concen-
tration of highly labile surface oxygen species in this catalyst
is very low due to the low number of single Pd atoms.
Calculated Properties of the Pd/CeO2 system
Density functional theory calculations were employed to
provide molecular level insights into the effect of surface Pd
density on CO oxidation activity. Since experiments use
specifically CeO2 nanocubes (Figure S3) as support, the
dominant oxygen-terminated CeO2 (100) facet was chosen as
the model system. The 4-fold hollow sites of this surface can
stabilize Pd adatoms. However, as shown in Figure S23, upon
deposition of single Pd atoms, exposing the stoichiometric
CeO2(100) surface to any partial pressure of oxygen will result
in the irreversible adsorption of one additional O2 molecule
per Pd atom to form (PdO2) 1 single-metal-centers, see Fig-
ure S23a. These (PdO2 ) 1 reside in a square planar config-
uration, with a 4-fold Pd coordination to surface-bound
oxygen atoms, labeled O1–O4 in Figure 4 a, Pd O bond
lengths of 1.94–1.98 . This is in excellent agreement with the
EXAFS results shown in Table 1 (CN of 3.8 and PdO bond
length of 1.99 ). Increasing the surface Pd density therefore
corresponds to decreasing the spatial separation of these
[PdO4 ] sites.
Through Bader analysis of the electron and spin density
around each atom in and surrounding the [PdO 4] complex
(for details, see the Supporting Information), we are able to
determine their electronic states. We find that Pd is in an
unusually overoxidized + 2.6 state, which is consistent with
our XANES results, and its surrounding O atoms are in
markedly unsaturated charge states of  1.7. The incom-
plete valence saturation of Pd 4d and O 2p orbitals can be
seen in Figure 4 b, clearly showing the presence of spin density
(i.e., unpaired charge) in these orbitals, indicative of partially
Angew. Chem. Int. Ed. 2021, 60, 22769 – 22775
 Angewandte
Research Articles Chemie

15213773, 2021, 60, Downloaded from https://onlinelibrary-wiley-com.ezproxy.lib.vt.edu By Virginia Tech- on [28/10/2021]. Re-use and distribution is strictly not permitted, except for Open Access articles
surface, nearby O atoms are made dramatically more reactive
(OVFEs of 22–43 kJmol1). The effect is diminished around
a Pd-O distance of  8  (OVFE of 70 kJ mol1) reaching
a plateau (of  88 kJ mol1) at  12 . As a difference in
reducibility between 0.1 and 0.2 Pd/nm2 is seen in our H2-TPR
experiments, and our model only allows for a minimum
surface density of  0.19 Pd/nm2, we suspect a significantly
larger distance is required to completely recover pristine
surface behavior. Using the average of 8  and 12  (i.e.,
10 ) as an approximate “radius of [PdO4] effect”, we overlay
an illustrative “zone of influence” (semi-transparent blue
circles in Figure 4 d) around each [PdO4 ] at different surface
Pd densities wherein O atoms are likely activated. Interest-
ingly, areas begin to overlap as surface Pd density increases,
encompassing neighboring [PdO4 ] by 1.13 Pd/nm2. This
suggests a cumulative effect on oxygen activity as surface
Pd density increases.
To test this presumption, we compute OVFEs for the five
most proximal (out to  8 ) O atoms to Pd (including the O
atoms in [PdO4]) as a function of surface Pd density (Fig-
ure 5 a). The results indicate that oxygen activity is progres-
sively enhanced as surface Pd density increases, with a general
downward trend in OVFEs seen for the O atoms not directly

Figure 4. a) Top-down view of the square planar [PdO4 ] active site on

the CeO 2(100) surface with atoms near Pd labeled to aid discussion
found in the Supporting Information (not to be confused with the
labels used in Figure 5). b) Spin density plot of the [PdO4 ] active site.
c) Oxygen vacancy formation energies (E OVF ) as a function of distance
from Pd. d) Approximate zone(s) of effect (semi-transparent blue
circles, shown with a 10  radius) of each [PdO4], illustrating how
overlapping zones produce a cumulative effect as Pd surface density
increases. Nominally symmetrically distinct O atoms within 8  of the
central Pd atom have been indicated with the colored and labeled
circles of Figure 5 for future reference. Atom key : Pd (brown), Ce
(green), and O (red) atoms on the CeO 2(100) surface ; O adatoms
(magenta).

charged states. This overoxidation occurs because the two

additional O atoms (O1 and O2 in Figure 4 a) require four
total electrons to reach valence saturation, while neither Pd
nor the Ce4+ of the surface can be oxidized sufficiently to
provide all four of these electrons. Instead, Pd provides  2.6
electrons (apparently the limit of its reducing power) and the
remaining electron deficiency is shared equally amongst the 4
Pd-coordinated O atoms (O1–O4 in Figure 4 a) via charge
delocalization.
This overall 2 e deficiency of the PdO4 moiety suggests
that it would be an excellent oxidizer whenever the nearby Figure 5. Quantities computed as a function of Pd loading : a) oxygen
CeO 2 surface is reduced as in CO oxidation via the MvK vacancy formation energies, b) oxygen adsorption energies, and
mechanism. We confirm this by calculating the oxygen c) average CO adsorption energies (inset : side view of the correspond-
vacancy formation energy (OVFE) of O atoms as a function ing ball-and-stick model). d) Ball-and-stick models of the four surface
Pd densities computed here. Symmetrically distinct O atoms are
of their distance from Pd. The results, shown in Figure 4 c, are
labeled 1–5 in each panel of (d). Each series in (a) and (b)
compared against the calculated OVFE of the pristine
corresponds to removal of the labeled O atoms or adsorption of an
CeO 2(100) surface (  102 kJ mol1/O). Intriguingly, while additional O atom to the labeled O atoms to form a superoxo (O 2).
the O atoms in [PdO 4] are more difficult to remove (OVFEs The series in (c) corresponds to the adsorption of CO on all single Pd
of  133 and  143 kJ mol1 /O) than those of the pristine atom centers available in each system in (d).

Angew. Chem. Int. Ed. 2021, 60, 22769 – 22775  2021 Wiley-VCH GmbH www.angewandte.org 22773

Research Articles
coordinated to Pd (labeled O3–O5 in Figure 4). Regardless of
surface Pd density, the Pd-coordinated O atoms (labeled O1
and O2 in Figure 4) remain more tightly bound than O atoms
in pristine CeO 2(100), indicating these O atoms may not be
very active in the CO oxidation reaction. In Figure 5 b,c,
respectively, we show that O2 and CO adsorption energies are
minimally affected by surface Pd density, indicating that the
primary cause of the enhanced CO oxidation activity, as seen
in experiment, is driven by the increased O reactivity. An
analysis of the charge distribution, seen in Figures S27–S29,
reveals that [PdO4] is consistently reduced after O atom
removal, regardless of which O atom is removed. As can be
seen in Figure S30a–c, for the O3 atom most proximal to
[PdO 4], the most likely participant in CO oxidation, the
OVFEs correlate well with the change in work function (DF),
with large DF corresponding to the highest OVFEs. Fig-
ure S30d,e shows that geometric relaxations do not correlate
with the OVFEs. This confirms that nonlocal electrostatic
effects are the source of the computed OVFE trends.
The OVFE and O2 adsorption energy of O3 at surface Pd
density of 1.13 Pd/nm2 (Figure 5 a,b) deserves highlighting.
We find that the OVFE is thermoneutral to within expected
errors, suggesting that this O atom should be very active.
However, the subsequent re-adsorption of O2 is lacking an
enthalpic driving force so once this O vacancy forms, it will
very seldom reform. This is consistent with experimental
results that showed decreased oxygen species adsorption on
oxygen vacancies on 4 Pd/CeO2 (Figure 3). As can be seen in
Figure S31, up to 4 such proximal O atoms (all equivalent to
O3 in Figure 5) can be removed, corresponding to two thirds
of the available [PdO4 ] active sites. This is a significant loss of
active O atoms and may explain the experimentally observed
loss in specific activity (Figure 1) in addition to simple Pd
agglomeration. This is an important observation because Pd
agglomeration does not necessarily remove the OVFE low-
ering effect, as supported metal nanoparticles have been
shown to reduce OVFEs at the nanoparticle/support interface
to a significant degree. [18d, 24]
Lastly, we wish to emphasize that the preceding results
should not be construed to suggest that O atoms more distant
than the most proximal O3 atoms are directly active in CO
oxidation. This enhancement must be entropic in nature since,
as previously stated, it is the preexponential factor and not the
activation energy that is appreciably changed as surface Pd
density increases (Figure 1 c). Since we show that the radius of
effect where O reducibility is enhanced is quite large, the
diffusivity of O vacancies is likely to be higher in these
regions, increasing the probability of O vacancies being filled
and propelling the reaction forward. This is admittedly
grounds for a much more ambitious study involving kinetic
Monte Carlo to confirm this conjecture.
Conclusion
Our study highlights a synergy between single Pd atoms
on CeO2 and enhanced catalytic activity, in contrast to the
constant-activity behavior typically assumed for single-atom
catalysts. By combining activity measurements, H2-TPR,
22774 www.angewandte.org  2021 Wiley-VCH GmbH
Angewandte
Chemie
15213773, 2021, 60, Downloaded from https://onlinelibrary-wiley-com.ezproxy.lib.vt.edu By Virginia Tech- on [28/10/2021]. Re-use and distribution is strictly not permitted, except for Open Access articles
Raman spectroscopy, FTIR, and DFT calculations, we
demonstrate that this improvement is due to increased
activity of O atoms proximal to Pd, with a surprisingly large
area of influence (at minimum 2.0 nm2 ) and O activity
accumulation as surface Pd density increases. This unique
behavior, characteristic of CeO 2 but absent in Al 2O 3, empha-
sizes the importance of the support.
XANES, EXAFS, DFT and charge analysis allowed us to
characterize the Pd/CeO2 active site as an overoxidized
square planar [PdO 4 ] complex with Pd in an approximate
+ 2.6 oxidation state and unsaturated valence O atoms in
approximately 1.7 oxidation states. The oxidative power of
[PdO4 ] is also shown to have a large lateral range of effects on
oxygen activity that accumulates as surface density of [PdO4 ]
increases. This stems from the ability of CeO2 to shuttle
charge from oxygen vacancies to the [PdO4] complex, which
reduces it to a more stable oxidation state. Both Raman
spectroscopy and DFT calculations reveal that the limits of
this cumulative effect at high surface Pd densities result in
overactivation and permanent removal of previously active O
atoms. Overall, the most interesting feature is direct corre-
lation of specific activity to CeO2 reducibility rather than Pd.
This work invites further investigation into the origin and
manifestation of nonlocal effects in single-atom catalysts that
have hitherto received little attention.
Acknowledgements
We acknowledge the financial support from the National
Research Foundation (NRF) (No. 2016R1A5A1009405,
2017R1A2B4007310). Work at the Pacific Northwest Na-
tional Laboratory (PNNL) was supported by the U.S. Depart-
ment of Energy, Office of Science, Basic Energy Sciences,
Chemical Sciences, Geosciences, and Biosciences Division.
PNNL is a multiprogram national laboratory operated for
DOE by Battelle under Contract DE-AC05-76RL01830.
Computational Resources were provided by a user proposal
at the National Energy Research Scientific Computing
Center (NERSC), a U.S. Department of Energy Office of
Science User Facility located at Lawrence Berkley National
Laboratory (LBNL).
Conflict of Interest
The authors declare no conflict of interest.
Keywords: CO oxidation · electrostatic effects · Pd/CeO 2 ·
reducible oxide supports
[1] a) X.-F. Yang, A. Wang, B. Qiao, J. Li, J. Liu, T. Zhang, Acc.
Chem. Res. 2013, 46, 1740 – 1748 ; b) S. Mitchell, E. Vorobyeva, J.
Prez-Ramrez, Angew. Chem. Int. Ed. 2018, 57, 15316 – 15329 ;
Angew. Chem. 2018, 130, 15538 – 15552; c) X. Li, X. Yang, Y.
Huang, T. Zhang, B. Liu, Adv. Mater. 2019, 31, 1902031; d) X.
Cui, W. Li, P. Ryabchuk, K. Junge, M. Beller, Nat. Catal. 2018, 1,
385 – 397.
Angew. Chem. Int. Ed. 2021, 60, 22769 – 22775

Research Articles
[2] a) J. Wang, Z. Li, Y. Wu, Y. Li, Adv. Mater. 2018, 30, 1801649 ;
b) J. Wu, L. Xiong, B. Zhao, M. Liu, L. Huang, Small Methods
2020, 4, 1900540; c) C. Tang, Y. Jiao, B. Shi, J.-N. Liu, Z. Xie, X.
Chen, Q. Zhang, S.-Z. Qiao, Angew. Chem. Int. Ed. 2020, 59,
9171 – 9176 ; Angew. Chem. 2020, 132, 9256 – 9261; d) J.-C. Liu,
H. Xiao, J. Li, J. Am. Chem. Soc. 2020, 142, 3375 – 3383 ; e) J. Li,
S. Chen, N. Yang, M. Deng, S. Ibraheem, J. Deng, J. Li, L. Li, Z.
Wei, Angew. Chem. Int. Ed. 2019, 58, 7035 – 7039 ; Angew. Chem.
2019, 131, 7109 – 7113; f) L. Zhao, Y. Zhang, L.-B. Huang, X.-Z.
Liu, Q.-H. Zhang, C. He, Z.-Y. Wu, L.-J. Zhang, J. Wu, W. Yang,
L. Gu, J.-S. Hu, L.-J. Wan, Nat. Commun. 2019, 10, 1278; g) X.
Zhang, M. Zhang, Y. Deng, M. Xu, L. Artiglia, W. Wen, R. Gao,
B. Chen, S. Yao, X. Zhang, M. Peng, J. Yan, A. Li, Z. Jiang, X.
Gao, S. Cao, C. Yang, A. J. Kropf, J. Shi, J. Xie, M. Bi, J. A.
van Bokhoven, Y.-W. Li, X. Wen, M. Flytzani-Stephanopoulos,
C. Shi, W. Zhou, D. Ma, Nature 2021, 589, 396 – 401.
[3] a) D. Kunwar, S. Zhou, A. DeLaRiva, E. J. Peterson, H. Xiong,
X. I. Pereira-Hernndez, S. C. Purdy, R. ter Veen, H. H. Bron-
gersma, J. T. Miller, H. Hashiguchi, L. Kovarik, S. Lin, H. Guo,
Y. Wang, A. K. Datye, ACS Catal. 2019, 9, 3978 – 3990 ; b) B.
Qiao, J.-X. Liang, A. Wang, C.-Q. Xu, J. Li, T. Zhang, J. J. Liu,
Nano Res. 2015, 8, 2913 – 2924 ; c) M. Yang, S. Li, Y. Wa ng, J. A.
Herron, Y. Xu, L. F. Allard, S. Lee, J. Huang, M. Mavrikakis, M.
Flytzani-Stephanopoulos, Science 2014, 346, 1498 – 1501; d) L.
Lin, W. Zhou, R. Gao, S. Yao, X. Zhang, W. Xu, S. Zheng, Z.
Jiang, Q. Yu, Y.-W. Li, C. Shi, X.-D. Wen, D. Ma, Nature 2017,
544, 80 – 83.
[4] B. Qiao, A. Wang, X. Yang, L. F. Allard, Z. Jiang, Y. Cui, J. Liu, J.
Li, T. Zhang, Nat. Chem. 2011, 3, 634 – 641.
[5] T. Yang, R. Fukuda, S. Hosokawa, T. Tanaka, S. Sakaki, M.
Ehara, ChemCatChem 2017, 9, 1222 – 1229.
[6] a) Y.-G. Wang, D. C. Cantu, M.-S. Lee, J. Li, V.-A. Glezakou, R.
Rousseau, J. Am. Chem. Soc. 2016, 138, 10467 – 10476 ; b) S. F.
Yuk, G. Collinge, M.-T. Nguyen, M.-S. Lee, V.-A. Glezakou, R.
Rousseau, J. Chem. Phys. 2020, 152, 154703.
[7] Y. Lou, J. Liu, Ind. Eng. Chem. Res. 2017, 56, 6916 – 6925.
[8] a) D. Mei, J. H. Kwak, J. Hu, S. J. Cho, J. Szanyi, L. F. Allard,
C. H. F. Peden, J. Phys. Chem. Lett. 2010, 1, 2688 – 2691; b) B. B.
Sarma, P. N. Plessow, G. Agostini, P. Concepcin, N. Pfnder, L.
Kang, F. R. Wang, F. Studt, G. Prieto, J. Am. Chem. Soc. 2020,
142, 14890 – 14902; c) S. Abbet, U. Heiz, H. Hkkinen, U.
Landman, Phys. Rev. Lett. 2001, 86, 5950.
[9] Z. Zhang, Y. Zhu, H. Asakura, B. Zhang, J. Zhang, M. Zhou, Y.
Han, T. Tanaka, A. Wang, T. Zhang, N. Yan, Nat. Commun. 2017,
8, 16100.
[10] E. J. Peterson, A. T. DeLaRiva, S. Lin, R. S. Johnson, H. Guo,
J. T. Miller, J. H. Kwak, C. H. F. Peden, B. Kiefer, L. F. Allard,
F. H. Ribeiro, A. K. Datye, Nat. Commun. 2014, 5, 4885.
[11] P. Venkataswamy, D. Jampaiah, F. Lin, I. Alxneit, B. M. Reddy,
Appl. Surf. Sci. 2015, 349, 299 – 309.
[12] a) B. Qiao, J. Lin, A. Wang, Y. Chen, T. Zhang, J. Liu, Chin. J.
Catal. 2015, 36, 1505 – 1511; b) L. Nie, D. Mei, H. Xiong, B. Peng,
Z. Ren, X. I. P. Hernandez, A. DeLaRiva, M. Wa ng, M. H.
Engelhard, L. Kovarik, A. K. Datye, Y. Wang, Science 2017, 358,
1419 – 1423 ; c) X. Zhang, J. Lei, D. Wu, X. Zhao, Y. Jing, Z.
Zhou, J. Mater. Chem. A 2016, 4, 4871 – 4876.
[13] A. J. Therrien, A. J. R. Hensley, M. D. Marcinkowski, R. Zhang,
F. R. Lucci, B. Coughlin, A. C. Schilling, J.-S. McEwen, E. C. H.
Sykes, Nat. Catal. 2018, 1, 192 – 198.
Angew. Chem. Int. Ed. 2021, 60, 22769 – 22775  2021 Wiley-VCH GmbH
Angewandte
Chemie
15213773, 2021, 60, Downloaded from https://onlinelibrary-wiley-com.ezproxy.lib.vt.edu By Virginia Tech- on [28/10/2021]. Re-use and distribution is strictly not permitted, except for Open Access articles
[14] a) G. Spezzati, Y. Su, J. P. Hofmann, A. D. Benavidez, A. T.
DeLaRiva, J. McCabe, A. K. Datye, E. J. M. Hensen, ACS Catal.
2017, 7, 6887 – 6891; b) G. Spezzati, A. Benavidez, A. T. DeLaR-
iva, Y. Su, J. P. Hofmann, S. Asahina, E. J. Olivier, J. H.
Neethling, J. T. Miller, A. K. Katye, E. J. M. Hensen, Appl.
Catal. B 2019, 243, 36 – 46.
[15] a) A. I. Boronin, E. M. Slavinskaya, I. G. Danilova, R. V.
Gulyaev, Y. I. Amosov, P. A. Kuznetsov, I. A. Polukhina, S. V.
Koscheev, V. I. Zaikovskii, A. S. Noskov, Catal. Today 2009, 144,
201 – 211; b) C. Dessal, T. Len, F. Morfin, J.-L. Rousset, M.
Aouine, P. Afanasiev, L. Piccolo, ACS Catal. 2019, 9, 5752 – 5759;
c) M. Cargnello, V. V. T. Doan-Nguyen, T. R. Gordon, R. E.
Diaz, E. A. Stach, R. J. Gorte, P. Fornasiero, C. B. Murray,
Science 2013, 341, 771 – 773 ; d) H. Jeong, J. Bae, J. W. Han, H.
Lee, ACS Catal. 2017, 7, 7097 – 7105 ; e) Z. Hu, X. Liu, D. Meng,
Y. Guo, Y. Guo, G. Lu, ACS Catal. 2016, 6, 2265 – 2279.
[16] a) H. Jeong, O. Kwon, B. -S. Kim, J. Bae, S. Shin, H.-E. Kim, J.
Kim, H. Lee, Nat. Catal. 2020, 3, 368 – 375 ; b) X. Liu, P. Ning, L.
Xu, Q. Liu, Z. Song, Q. Zhang, RSC Adv. 2016, 6, 41181 – 41188 ;
c) E. M. Slavinskaya, T. Y. Kardash, O. A. Stonkus, R. V.
Gulyaev, I. N. Lapin, V. A. Svetlichnyi, A. I. Boronin, Catal.
Sci. Technol. 2016, 6, 6650 – 6666.
[17] A. F. Lee, S. F. J. Hackett, J. S. J. Hargreaves, K. Wilson, Green
Chem. 2006, 8, 549 – 555.
[18] a) Z. Wu, M. Li, S. H. Overbury, J. Catal. 2012, 285, 61 – 73; b) J.-
X. Liu, Y. Su, I. A. W. Filot, E. J. M. Hensen, J. Am. Chem. Soc.
2018, 140, 4580 – 4587; c) H. Ha, S. Yoon, K. An, H. Y. Kim, ACS
Catal. 2018, 8, 11491 – 11501; d) B. Liu, J. Liu, T. Li, Z. Zhao, X.-
Q. Gong, Y. Chen, A. Duan, G. Jiang, Y. Wei, J. Phys. Chem. C
2015, 119, 12923 – 12934; e) M. Lohrenscheit, C. Hess, Chem-
CatChem 2016, 8, 523 – 526.
[19] a) X. Sun, J. Lin, Y. Zhou, L. Li, Y. Su, X. Wang, T. Zhang,
AIChE J. 2017, 63, 4022 – 4031; b) M.-F. Luo, Z.-Y. Hou, X.-X.
Yuan, X.-M. Zheng, Catal. Lett. 1998, 50, 205 – 209.
[20] a) S. Lin, L. Yang, X. Yang, R. Zhou, Appl. Surf. Sci. 2014, 305,
642 – 649; b) R. Strobel, J.-D. Grunwaldt, A. Camenzind, S. E.
Pratsinis, A. Baiker, Catal. Lett. 2005, 104, 9 – 16.
[21] A. A. Vedyagin, A. M. Volodin, V. O. Stoyanovskii, I. V. Mis-
hakov, D. A. Medvedev, A. S. Noskov, Appl. Catal. B 2011, 103,
397 – 403.
[22] a) Z. Wu, M. Li, J. Howe, H. M. Meyer, S. H. Overbury,
Langmuir 2010, 26, 16595 – 16606 ; b) A. Filtschew, K. Hofmann,
C. Hess, J. Phys. Chem. C 2016, 120, 6694 – 6703.
[23] C. Schilling, M. V. Ganduglia-Pirovano, C. Hess, J. Phys. Chem.
Lett. 2018, 9, 6593 – 6598.
[24] a) A. R. Puigdollers, G. Pacchioni, ChemCatChem 2017, 9,
1119 – 1127; b) S. W. Hoh, L. Thomas, G. Jones, D. J. Willock,
Res. Chem. Intermed. 2015, 41, 9587 – 9601; c) Y. Hinuma, T.
Toyao, N. Hamamoto, M. Takao, K.-i. Shimizu, T. Kamachi, J.
Phys. Chem. C 2020, 124, 27621 – 27630 ; d) P. Schlexer, D.
Widmann, R. J. Behm, G. Pacchioni, ACS Catal. 2018, 8, 6513 –
6525.
Manuscript received: April 28, 2021
Revised manuscript received: June 11, 2021
Accepted manuscript online: June 28, 2021
Version of record online: July 19, 2021
www.angewandte.org 22775