C–H oxidation enhancement on a gold nanoisland by atomic-undercoordination induced polarization

Zezhou Lin a, Hajime Hirao b, Changqing Sun c and Xi Zhang *a
aInstitute of Nanosurface Science and Engineering, Shenzhen University, Shenzhen 518060, China. E-mail: zh0005xi@szu.edu.cn
bDepartment of Chemistry, City University of Hong Kong, Hong Kong
cNOVITAS, School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore

Received 27th February 2020 , Accepted 6th May 2020

First published on 6th May 2020

C–H activation is of great significance in the chemical industry while an effective solvent-free catalyst is highly desired. This work shows that a gold nanoisland which was inert in the bulk is effective for C–H activation reactions. We investigated the C–H activation of toluene on an Au nanoisland (58 atoms) using relativistic density functional theory (DFT). We found that (i) the bonds between under-coordinated gold atoms (corner site) shrink spontaneously and become stronger; (ii) the valence charges of corner atoms are polarized to the upper edge of the valence band (near the Fermi level), indicating the electron donation ability in the catalytic process; (iii) during C–H oxidation, the indirect path (O2 dissociation and O–H bonding) and direct path (O2–H bonding) were considered. The Au–O2 complex is active enough to abstract a hydrogen atom directly from toluene, with a barrier that is 6.8 kcal mol−1 lower than that of the indirect path; and (iv) a transfer of up to ∼0.8 electrons from gold to O2 occurs. Moreover, hybridization between delocalized gold orbitals and oxygen p-orbitals leads to the stabilization of the singlet spin state of Au58O. Our results suggest that undercoordination-charge-polarization are key factors for the C–H oxidation catalyzed by an Au nanoisland.

1. Introduction

Gold had been considered to be catalytically inert1 until Haruta et al. demonstrated in 1993 that the activation energy of CO oxidation could be lowered to 3.9 kcal mol−1 by a supported gold nanoparticle.2 Since then, gold nanoparticles deposited on various supports, also called gold nanoislands, have been widely investigated as potentially “green”, stable, solvent-free and efficient heterogeneous catalysts of chemical reactions.3–8 However, the mechanisms underlying the reactions catalyzed by gold nanoislands are still under debate.9,10 So far, various mechanisms have been proposed such as the gold-only mechanism,11–13 the support-assist mechanism,14 the direct interaction mechanism15–17 and the indirect interaction mechanism.18

Under-coordinated sites can absorb small inorganic molecules, and the presence of these sites is the key to the catalytic properties of supported gold nanoparticles.10,19 Au nanoparticles were found to be catalytically active only for a diameter below 3.5 nm.8,11,16 A high activity of a gold catalyst occurs when an atomic oxygen is bound by the three-fold coordination sites on the surface of gold.20,21 However, the underlying mechanism of the under-coordination effects remain poorly understood.

C–H oxidation of hydrocarbons is particularly important in the industrial synthesis of petroleum-based chemicals.7 Of note, solid-type heterogeneous catalysts, which can be readily separated from the reaction mixture,6 are superior to the traditional homogeneous catalysts that produce environmentally undesirable waste after the reaction.7,22 In 2005, Hutchings and co-workers made a ground-breaking discovery that supported nanocrystalline gold can catalyze the gas-phase oxidation of alkenes using O2 with 98% selectivity.5 This was followed by other chemical reactions achieved by the use of gold nanoislands on a silicon surface,23,24 a TiO2 surface25 and a graphite surface,5 and also an Au–Pd alloy nanoisland on graphite.6 In comparison to the many experimental studies that have been performed so far, only a few theoretical studies of gold nanoparticle-assisted C–H activation reactions have been reported.

For the nanosize-metal cluster catalyzed CO oxidation, the Langmuir–Hinshelwood mechanism, in which both CO and O2 are absorbed onto gold before the reaction, has been proposed as a likely mechanism.17,18,26–28 In contrast, nanosize-Au-catalyzed C–H activation may follow the Eley–Rideal mechanism because a hydrocarbon molecule cannot form a chemical bond with a gold nanoisland. The O2 molecule originally has a strong double bond that may not be easily cleaved even on gold.7 In the case of CO oxidation, direct interaction of O2 with CO was shown to be possible.16–18 In a similar vein, there is the possibility that O2 “directly” abstracts a hydrogen atom from a hydroalkane on the surface of a gold island.

In this work, we use DFT to investigate two paths, i.e., the indirect and direct mechanisms of C–H oxidation of toluene on a gold nanoisland. We mainly focus on the role of the under-coordinated atoms in enhancing the catalytic ability of an Au nanoisland. Both high-spin and low-spin states are considered in the calculations.

2. Methods

Theoretical reaction path search calculations based on DFT were performed using the ADF code.29 The Perdew–Burke–Ernzerhof (PBE)30 generalized gradient approximation functional was used to describe the exchange–correlation energy. The PBE functional has been widely used for the calculation of large metal nanostructures.9,16,18,31 Triple-zeta plus polarized (TZP) numerical slater-type orbitals were used as basis sets of wave functions. A norm-conserving pseudopotential was applied to gold with a small frozen core up to the 4d level. Constrained optimization was used to find the initial guess structure of the transition state. According to the vibration analysis, the atomic positions were adjusted along with the vibration direction to eliminate the excess virtual frequency before searching for the transition state. The eigenvector-following TS optimization technique was applied to each initial TS structure. TS confirmations were followed by a vibration frequency calculation at the same DFT level. We also check whether there is only one dominant virtual frequency after TS optimization and whether the virtual frequency mode corresponds to the TS atomic vibration mode (see Fig. S1 in the ESI). The final convergence criteria for each TS and MS were set at 10−5 Hartree for the energy, 0.002 Hartree Å−1 for the forces and 0.005 Å for the displacement.

It has been experimentally shown6 that a gold nanoisland with a diameter of about 2 nm adopts fcc-arrangement structures, i.e. truncated octahedral nanocrystal structures and increasing the proportion of the (111) facet correlates with an improvement of the catalytic activity. As shown in Fig. 1, we first performed initial structural optimization of the Au116 nanoparticle, and all atoms were relaxed. The optimized Au116 nanoparticle was cut in half to obtain the geometry of an Au nanoisland (58 atoms). Fixing the underlying mimic substrate (bottom layer) of the nanoisland, further geometry optimization was performed so that the geometry was constructed precisely for investigating the realistic catalytic process. At the exposed (111) facet, there are 3 nonequivalent atomic sites. O2 can choose 5 different binding patterns on the surface for the binding. All the 5 possibilities were examined and the corner-bridge binding pattern (O2 bridges on 1–1) was found to be the most stable. Therefore, we used this pattern as the initial structure of Au58O2 for the studies of reaction mechanisms.

image file: d0cp01117g-f1.tif
Fig. 1 Illustration of a truncated octahedral Au116 nanocrystal and a half-size Au nanoisland (58 atoms). There are three nonequivalent atomic sites labeled 1, 2 and 3 on the surface of the nanoisland. There are five nonequivalent patterns for O2 absorption: 1–1, 1–2, 2–2, 1–3, and 2–3.

The principle proposed is shown in eqn (1). The core-level shift of vth energy band Ev of effective coordination number (z) over that of the bulk material (B) with respect to energy level of an isolated atom (0) can be expressed as follows:32

image file: d0cp01117g-t1.tif(1)
image file: d0cp01117g-t2.tif(2)
where Eν(0) is the energy level of an isolated atom, Vatom(r) is the intra-atomic potential and Vcry(r) is the crystal potential. |v,i〉 represents the wave function. α = 〈v,i|Vcry(r)|v,i〉 is the overlap integral and βij = 〈v,i|Vcry(r)|v,j〉 the exchange integral whose summation over the first nearest neighbor j contributes to the width of the energy band, with a periodic factor f(k) in the form of eikr, while k is the wave vector. In the localized core-levels, wave functions of core electrons between i and j overlap a little and thus βij is not comparable to the overlap integral of the ith atom, α. The core-level shift Ev(z) − Ev(0) is dominated by α. In eqn (2), m is the bond nature indicator varying with materials. For gold, m has been optimized as 1.33

According to eqn (1), as the size of the nanoparticle decreases, (i) bonds between under-coordinated atoms become shorter and stronger; (ii) as atoms becomes closer at the surface skin, the Vcry(r) to the neighbor electron will become deeper; (iii) consequently, the α and βij will be enlarged proportionally to the depth of the potential well at equilibrium; (iv) the localized core level will shift proportional to α and become trapped to a deeper energy; (v) valence band polarization (P) happens to the noble metals: valence band red shifts since nonlocal charges are polarized towards the Fermi level (EF).

3. Results and discussion

Appropriate structural optimization should be performed on an Au nanoisland to simulate a realistic C–H catalytic process. Table 1 shows the geometry optimization results on an Au nanoisland. To study the catalytic properties of the nanoisland surface, the underlying mimic substrate (bottom layer) was fixed. After geometrical optimization, the bond length of the outermost layer Au atoms is shortened with respect to that of the conventional bulk Au atoms. The spontaneous contraction of the gold bond can be explained by eqn (1). The gold bonds between corner atoms (1–1*) were found to contract most severely from 2.883 Å (bulk) to 2.723 Å (Au nanoisland). The gold bonds between the corner atom and the nearly edge atom (1–3) also contracted sharply from 2.883 Å (bulk) to 2.731 Å (Au nanoisland). The spontaneous contraction of gold bonds on the surface has been reported in recent experiments,34 which is consistent with our DFT calculations.
Table 1 Atomic Au–Au bond length di derived from the DFT geometrical optimization with respect to the bulk standard of d0
Structure Position d i (Å) d 0 (Å) 1−Ci (%) Position Mulliken charge (e)
1–1* 2.723 2.883 5.3 1 −0.032
1–2 2.780 2.883 3.6 2 −0.006
Au58 1–3 2.731 2.883 5.6 3 −0.020
2–2* 2.806 2.883 2.7
2–3 2.819 2.883 2.2

The charge transfer was demonstrated in the calculations using Mulliken charge population analysis indicating that the electrons are transferred from the inner to the outer atoms of the Au nanoisland. Fig. 2 compares the local density of states (LDOS) of three different positions of the Au nanoisland and bulk surface. Obviously, the active peak of bulk gold appears around −3.5 eV (grey), while the lower-coordinated sites move significantly toward the Fermi level (EF) and appear around −2 eV (red, orange and blue). It is demonstrated that the under-coordinated atoms will more readily interact and form bonds with others, as compared to its more “coordinated” counterparts. Moreover, the LDOS polarization illustrated that the valence charges of the corner atom (site 1) are polarized towards EF while the valence charges in the centre atom (site 2) or the edge atom (site 3) remain in deeper energy. The polarization makes the donor-type catalyst more active in losing electrons. Thus, the LDOS polarization witnesses the significance of the under-coordinated atoms moving valence band electrons to the Fermi level, indicating that the corner atom (site 1) has better activity for C–H activation. The under-coordinated corner atom plays a major role in catalytic reactions under experimental observations.35 Valence band polarization (P) happens to the noble metals: the valence band redshifts since nonlocal charges are polarized upwards to EF at the under-coordinated sites. Such polarization may turn the under-coordinated atoms into a donor-type catalyst, as that happened to Rh adatoms.36

image file: d0cp01117g-f2.tif
Fig. 2 The LDOS for the Au nanoisland and bulk surface. The electrons of the outermost shell transfer to the upper edge near the EF.

To provide atomic- and molecular-level understanding, the constrained optimization technique was used to locate the transition state in each reaction step, exploring the possible energy pathways. Vibration analysis was applied in the transition state to confirm that the final virtual frequency is single and the virtual frequency mode corresponds to the TS atomic vibration mode. As shown in Fig. 3, DFT studies have been investigated to predict whether the C–H activation reaction occurs through a direct reduction pathway or an indirect reduction pathway. In the indirect path (path 1), O2 first dissociates into two single O atoms on the gold surface and then one O atom activates the C–H bond of toluene. Path 2 is the direct path in which O2 directly abstracts an H atom from toluene to form an HO–O/Au moiety and then the O–O bond of HO–O is cleaved on the gold surface. The calculated energy profiles of path 1 and path 2 are shown in Fig. 4 and 5, respectively, in which the energies for both singlet and triplet spin states are displayed. Although O2 in the gas phase can hardly absorb abstract hydrogen, Fig. 4 and 5 show that, interestingly, path 2 has a lower total barrier (21.5 kcal mol−1) than path 1 (28.9 kcal mol−1).

image file: d0cp01117g-f3.tif
Fig. 3 Scheme of toluene oxidation in two pathways. In path 1 (indirect path: O2 dissociation and O–H bonding), O2 dissociates at first. In path 2 (direct path: O2–H bonding), C–H activates first.

image file: d0cp01117g-f4.tif
Fig. 4 Energy diagram of path 1 (indirect path: O2 dissociation and O–H bonding). The key structures of low spin states were plotted for representation. The slightly different bond lengths of low-spin/high-spin were given. T spins represent the sum of the two O atom spins in the high-spin states.

image file: d0cp01117g-f5.tif
Fig. 5 Energy diagram of path 2 (direct path: O2–H bonding). The key structures of the low spin states were plotted for representation. The slightly different bond lengths of low-spin/high-spin were given. T spins represent the sum of the two O atom spins in the high-spin states.

There are four steps in path 1 as shown in Fig. 4. The first step is the adsorption of only the O2 molecule onto gold. The toluene molecule has a nonbonding interaction directly with O2 rather than being absorbed by the Au nanoisland, as illustrated in MS1. Thus, the Eley–Rideal mechanism is really favorable in the gold-assisted C–H oxidation reaction. The second step is associated with the O–O dissociation on the gold surface. The O–O distance increases from 1.410 Å in MS1 to 2.074 Å in TS1 with a large energy barrier of 28.3 kcal mol−1 respective to MS1 in the triplet state. The oxygen atom prefers to sit at the centre of three close-contacted gold atoms in both TS1 and MS2. Two oxygen atoms from each other are separated by as large as 4.316 Å in the third step, and an O atom abstracts an H atom from toluene to induce C–H activation in TS2. Due to the high reactivity of single O, the energy barrier is only 4.8 kcal mol−1 for this step. Subsequently, there is O–C rebound at the final step (TS3: hydroxyl rebound) which has a low barrier of 0.89 kcal mol−1 to form a benzyl alcohol product.

In comparison, path 2 consists of 4 steps as shown in Fig. 5. The first step is for adsorption of the O2 molecule, the same as in path 1. The second step is direct C–H activation by the O2 molecule. This step can be further divided into two substeps.

The first substep is the approach of O2 toward the H–C bond of toluene in which the O⋯H distance decreases from 2.267 Å at MS1 to 1.933 Å at MS2, during 1.985 Å at TS1. The coordinated O2 molecule with benzene was formed at TS1 and MS2. A possible hydrogen bond is formed of C–H⋯O at MS2 which induces a slightly lower energy in MS2 and a transition state of C–H stretching mode at TS1. The second substep is the C–H activation whereby the C–H distance increases from 1.105 Å to 1.578 Å at TS2 (peroxide) while the O⋯H distance decreases from 1.985 Å to 1.150 Å. For comparison, we also performed a scan calculation of the energy profile when an O2 molecule absorbs a H atom from toluene in the absence of a catalyst. The results of the uncatalyzed reaction show that the energy keeps increasing by 32.9 kcal mol−1 until the O–H distance is 1.1 Å. Thus, the reaction is unlikely to occur without a gold-nanoisland catalyst because O2H is too unstable to hold the H atom in the final geometry optimization. In contrast, as shown in Fig. 5, the catalyzed H-abstraction has a small barrier of 10.0 kcal mol−1 and a stable Au–OOH-type intermediate MS3 is obtained in the second step. This result clearly shows that O2 on gold is reactive enough to abstract hydrogen directly. The third step is the O–OH cleavage. The barrier for the O–OH cleavage in path 2 is 21.5 kcal mol−1 (relative to triplet MS1) in the singlet spin state, lower than 28.3 kcal mol−1 of the O2 double bond in path 1. The final step of C–O rebound is similar to path 1 and has a very low barrier. Comparing Fig. 4 and 5, the total barriers of the two paths are both determined by the step of O–O dissociation. Thus, the stability of O2H/Au and the lower dissociation energy of the O–OH single bond make path 2 favourable.

It is well known that O2 prefers a spin-triplet state with two spin-parallel p-electrons, so the isolated state can hardly adopt a spin-singlet state. Unexpectedly, spin-crossover happens in both paths and singlet spin states have a lower energy barrier. To investigate the reason for spin-crossover and the functions of gold in enabling O2 direct interacting with toluene in path 2, we calculated the charge transfer and spin density distribution for all the states in path 2, as listed in Table 2.

Table 2 Mulliken charges in reaction path 2 for spin singlet (S) and triplet (T)

image file: d0cp01117g-u1.tif

S charge T charge T spin
Bn O2 Au58 Bn O2 Au58 Bn O2 Au58
Au58O2 −0.67 0.67 −0.7 0.7 1 1
MS1 0.17 −0.85 0.68 0.11 −0.77 0.66 0.01 1 1
TS1 0.11 −0.84 0.64 0.12 −0.76 0.55 0.01 0.65 1.34
TS2 −0.21 −0.84 0.68 −0.12 −0.82 0.55 0.39 0.14 1.47
TS3 −0.05 −1.1 0.82 −0.09 −1.14 0.92 −0.06 0.6 1.45
TS4 1.04 −1.62 0.26 0.07 −1.63 1.24 0.9 0.2 0.89

As shown in Mulliken charges in Table 2, in singlet spin, negative charges are transferred from gold to the O2 molecule to form an O2δ ∼ 0.67e at initial Au58O2. After toluene is absorbed by a nonbonding interaction between C–H and O in MS1, the transfer amount increases to ∼0.85e. Furthermore, δ stays at 0.84e in TS1 and TS2 and leads to a possible O2δ–H bond. Thus, the electron donation ability of a metal nano-catalyst is the key to enable the direct reaction between O2 gas and toluene.

Furthermore, the spin densities in Table 2 were used to explain the spin-crossovers as shown in Fig. 5. Initially, the total spin of O2 decreases from 2e in an isolated molecule to 1e in Au58O2. Even in the charge absence, the triplet state still precedes the singlet state until MS1. Subsequently, the strong interaction between the O2δ and Bn exacerbated the loss of oxygen spin density, which was kept lower than 0.65e until TS3. This reduced spin density results in the prevalence of spin-singlet state during TS1 to TS3. Additionally, at MS4 and TS4, Bn raises its spin more than 0.90e due to the unpaired electron of carbon, and the triplet state of the system becomes more favorable again. However, a single O atom on the Au nanoisland again adopts the singlet state in FS.

It is unexpected that a single oxygen atom on the Au nanoisland adopts a singlet spin in FS, although a single O atom and even initial Au58O2 prefer triplet states. This is due to the delocalization of the spin density (e.g. O2 has 0.14 spin and Au nanoisland has 1.47 in TS2). The strong electron donation ability of the Au nanoisland makes the low-spin favorable. To explore the reason for this, we investigated the population and frontier molecular orbitals (MOs) of the Au58O complex. The HOMO is shown in Table 3. Combining Tables 2 and 3, the donating ability of the corner gold atom and the hybridization of O and Au frontier orbitals play the crucial role of the C–H activation catalysis. Hence, it is the undercoordination-charge-polarization of the Au nanoisland that enables the donor-type catalytic ability of Au for C–H oxidation. The experimentally observed valence band polarization of the Au adatom on the TiOx/Mo(112) surface and their catalytic capability measured by the TOF also verify the polarization–catalysis correlation.12 The polarization of the valence 4d electrons associated with the under-coordinated atoms in the skin of metal nanostructures has been verified as the key for donor-type catalysts like Rh adatoms.36 The high reactivity is mainly attributed to the electron donation ability of the undercoordinated atom in the upper layer (active sites). The bottom layer atoms which provide a stronger coordination to the substrate may enhance the stability of the Au nanoisland. The intrinsic activity of the perovskite oxide in the oxygen evolution reaction also exhibits a strong dependence on the occupancy of the 3d band.37

Table 3 Comparison of energies in singlet (S) and triplet (T) spin states and the Mulliken charge and spin density of the Au58O complex
HOMO-s Singlet Triplet

image file: d0cp01117g-u2.tif

E(Au58O)/Ha −4.09211 −4.08465
E(Au58)/Ha −3.90225 −3.90331
Charge (O)/e s 0.00 0.01
p −0.80 −0.80
Spin (O)/e s 0.00
p 0.23
Charge (Au)/e s 20.15 20.13
p −28.69 −28.73
d 9.36 9.39
Spin (Au)/e s 0.76
p 0.84
d 0.17

4. Conclusions

In summary, we highlight the emergence of high C–H activation in a gold nanoisland as a widespread and universal phenomenon driven by the undercoordination of the outermost shell atoms. Through relativistic DFT calculations, we found that the spontaneous contraction of the bonds between the under-coordinated atoms (corner atoms) was stronger than the interior. LDOS polarization witnesses the significance of the under-coordinated atoms moving valence band electrons to EF, indicating that the under-coordinated Au atoms provide a major contribution for high C–H activation. The total barrier of the direct path is 6.8 kcal mol−1 lower than that of the indirect path, since the energy demanded by the cleavage of the O–OH single bond in the direct path is lower than that demanded by the O2 double bond. Moreover, the delocalization and hybridization of the gold and oxygen p-orbitals lead to the stabilization of the spin-singlet state of Au58O. The low-spin states become more favourable in both reaction paths also due to the loss of spin in oxygen. The findings determined the long-debated reaction mechanism of nano-gold-assist C–H activation as an Eley–Rideal–Direct mechanism, linked the catalytic ability of gold with the electronic structure evolution and the charge-donating ability of the gold nanoisland, and paved a way for the further design of noble metal or noble metal-alloy nanocatalysts.

Conflicts of interest

There are no conflicts to declare.


This work was supported by the National Natural Science Foundation (No. 51605306) of China, the Natural Science Foundation of Guangdong Province (No. 2016A030310060) and the Shenzhen Overseas High-level Talents Innovation and Entrepreneurship Plan (No. KQJSCX20180328094853770).


  1. B. Hammer and J. K. Norskov, Nature, 1995, 376, 238–240 CrossRef CAS.
  2. M. Haruta, S. Tsubota, T. Kobayashi, H. Kageyama, M. J. Genet and B. Delmon, J. Catal., 1993, 144, 175–192 CrossRef CAS.
  3. P. Priecel, H. Adekunle Salami, R. H. Padilla, Z. Zhong and J. A. Lopez-Sanchez, Chin. J. Catal., 2016, 37, 1619–1650 CrossRef CAS.
  4. Y. Zhang, X. J. Cui, F. Shi and Y. Q. Deng, Chem. Rev., 2011, 112, 2467–2505 CrossRef PubMed.
  5. M. D. Hughes, Y.-J. Xu, P. Jenkins, P. McMorn, P. Landon, D. I. Enache, A. F. Carley, G. A. Attard, G. J. Hutchings, F. King, E. H. Stitt, P. Johnston, K. Griffin and C. J. Kiely, Nature, 2005, 437, 1132–1135 CrossRef CAS PubMed.
  6. L. Kesavan, R. Tiruvalam, M. H. A. Rahim, M. I. bin Saiman, D. I. Enache, R. L. Jenkins, N. Dimitratos, J. A. Lopez-Sanchez, S. H. Taylor, D. W. Knight, C. J. Kiely and G. J. Hutchings, Science, 2011, 331, 195–199 CrossRef CAS PubMed.
  7. M. Haruta, Nature, 2005, 437, 1098–1099 CrossRef CAS PubMed.
  8. M. Turner, V. B. Golovko, O. P. H. Vaughan, P. Abdulkin, A. Berenguer-Murcia, M. S. Tikhov, B. F. G. Johnson and R. M. Lambert, Nature, 2008, 454, 981–983 CrossRef CAS PubMed.
  9. H. Y. Su, M. M. Yang, X. H. Bao and W. X. Li, J. Phys. Chem. C, 2008, 112, 17303–17310 CrossRef CAS.
  10. I. N. Remediakis, N. Lopez and J. K. Nørskov, Appl. Catal., A, 2005, 291, 13–20 CrossRef CAS.
  11. M. Valden, X. Lai and D. W. Goodman, Science, 1998, 281, 1647–1650 CrossRef CAS PubMed.
  12. M. Chen, Y. Cai, Z. Yan and D. W. Goodman, J. Am. Chem. Soc., 2006, 128, 6341–6346 CrossRef CAS PubMed.
  13. Y. He, J. C. Liu, L. L. Luo, Y. G. Wang, J. F. Zhu, Y. G. Du, J. Li, S. X. Mao and C. M. Wang, Proc. Natl. Acad. Sci. U. S. A., 2018, 115, 7700–7705 CrossRef CAS PubMed.
  14. M. Haruta, Catal. Today, 1997, 36, 153–166 CrossRef CAS.
  15. H. F. Wang and Z. P. Liu, J. Am. Chem. Soc., 2008, 130, 10996–11004 CrossRef CAS PubMed.
  16. W. Gao, X. F. Chen, J. C. Li and Q. Jiang, J. Phys. Chem. C, 2009, 114, 1148–1153 CrossRef.
  17. D. Y. Tang and C. W. Hu, J. Phys. Chem. Lett., 2011, 2, 2972–2977 CrossRef CAS.
  18. N. Lopez and J. K. Nørskov, J. Am. Chem. Soc., 2002, 124, 11262–11263 CrossRef CAS PubMed.
  19. J. A. Rodriguez, S. Ma, P. Liu, J. Hrbek, J. Evans and M. Pérez, Science, 2007, 318, 1757–1760 CrossRef CAS PubMed.
  20. T. A. Baker, X. Liu and C. M. Friend, Phys. Chem. Chem. Phys., 2011, 13, 34–46 RSC.
  21. T. A. Baker, C. M. Friend and E. Kaxiras, J. Chem. Theory Comput., 2009, 6, 279–287 CrossRef PubMed.
  22. Z. Y. Cai, M. Q. Zhu, J. Chen, Y. Y. Shen, J. Zhao, Y. Tang and X. Z. Chen, Catal. Commun., 2010, 12, 197–201 CrossRef CAS.
  23. F. Moreau and G. C. Bond, Catal. Commun., 2007, 8, 1403–1405 CrossRef CAS.
  24. J. J. Bravo-Suárez, K. K. Bando, J. Q. Lu, T. Fujitani and S. T. Oyama, J. Catal., 2008, 255, 114–126 CrossRef.
  25. K. Katsiev, G. Harrison, Y. Al-Salik, G. Thornton and H. Idriss, ACS Catal., 2019, 9, 8294–8305 CrossRef CAS.
  26. Q. Jiang, J. Zhang, H. Huang, Y. Wu and Z. Ao, J. Mater. Chem. A, 2020, 8, 287–295 RSC.
  27. H. Yang, J. C. Liu, L. Luo, Y. G. Wang, J. Zhu, Y. Du, J. Li, S. X. Mao and C. Wang, Proc. Natl. Acad. Sci. U. S. A., 2018, 115, 7700–7705 CrossRef PubMed.
  28. Q. G. Jiang, Z. M. Ao, S. Li and Z. Wen, RSC Adv., 2014, 4, 20290 RSC.
  29. C. Fonseca Guerra, J. G. Snijders, G. te Velde and E. J. Baerends, Theor. Chem. Acc., 1998, 99, 391–403 Search PubMed.
  30. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865–3868 CrossRef CAS PubMed.
  31. H. T. Chen, J. G. Chang, S. P. Ju and H. L. Chen, J. Comput. Chem., 2010, 31, 258–265 CAS.
  32. X. Zhang, S. Wang, Y. Liu, L. Li and C. Sun, APL Mater., 2017, 5, 053501 CrossRef.
  33. X. Zhang, C. Q. Sun and H. Hirao, Phys. Chem. Chem. Phys., 2013, 15, 19284–19292 RSC.
  34. W. J. Huang, R. Sun, J. Tao, L. D. Menard, R. G. Nuzzo and J. M. Zuo, Nat. Mater., 2008, 7, 308–313 CrossRef CAS PubMed.
  35. J. N. Crain and D. T. Pierce, Science, 2005, 307, 703–706 CrossRef CAS PubMed.
  36. C. Q. Sun, Y. Wang, Y. G. Nie, Y. Sun, J. S. Pan, L. K. Pan and Z. Sun, J. Phys. Chem. C, 2009, 113, 21889–21894 CrossRef CAS.
  37. J. Suntivich, K. J. May, H. A. Gasteiger, J. B. Goodenough and Y. Shao-Horn, Science, 2011, 334, 1383–1385 CrossRef CAS PubMed.


Electronic supplementary information (ESI) available. See DOI: 10.1039/d0cp01117g

This journal is © the Owner Societies 2020