A density functional theory study on 3d metal/graphene for the removal of CO from H₂ feed gas in hydrogen fuel cells

Abstract

Metal/graphene has been used as a filter membrane exterior to the hydrogen fuel cell to prevent CO poisoning. The removal of CO from H₂ feed gas is important for efficient use of the anode catalyst and would increase the lifetime of the fuel cells. In this work, the adsorptions of CO and H₂ on metal/perfect-graphene (M/G_p) and metal/defect-graphene (M/G_d) (M = Sc–Zn) are investigated using density functional theory. Our results indicated that the defect sites in graphene enhance the stability of metal on the graphene surface compared to perfect graphene. For gas molecule adsorption, however, CO and H₂ adsorption is weaker on defective graphene compared with the perfect material, due to the more localized metal d electrons in the former case. For both defective and perfect graphene, Fe/G_p(d), Co/G_p(d) and Ni/G_p(d) are more effective in separating CO from H₂ feed gas, particularly for perfect graphene. Orbital analysis suggested that the d_yz and/or d_xz orbitals of metal atoms play a major role in CO and H₂ adsorption.

---

1. Introduction

Current energy productions are mainly from unsustainable resources such as fossil fuels. However, depleted resources and environmental pollution of the fossil fuels limit their use. A fuel cell is an electrochemical cell that converts chemical energy into electrical energy and is an attractive replacement for fossil fuels.^1,2 Currently, platinum is widely utilized as an anode electrocatalyst due to its high catalytic activity toward the redox reactions of H₂ and O₂.³ The hydrogen for the anode reaction is mainly produced by steam reforming of methane. This results in hydrogen feed gas containing an appreciable amount of CO (0.5–2%), and it is known that a CO level as little as 20 ppm would poison the anode catalyst by blocking hydrogen adsorption sites.⁴ Therefore, the presence of CO is an important obstacle and is particularly challenging for the development of fuel cells. In order to avoid catastrophic poisoning of the fuel cell anode, the CO concentration needs to be removed or at least reduced.

To improve the CO tolerance, many electrode materials have been investigated. It is found that the presence of a second metal, changes the CO chemisorptions by modifying the Pt electronic properties and promotes the oxidation of CO to CO₂.^5–7 Thus, many binary PtM alloys have been investigated, such as PtRu, PtRh and PtPd.^8–16 The most promising alloy is the PtRu catalyst due to its ability to more effectively promote the electro-oxidation and higher CO tolerance,^11–16 however, the high cost of Ru and Pt limit its widespread application. After that, several Pt-based electrodes alloyed with a second non-precious metal (PtFe, PtCo, PtNi, PtSn and PtW) were also found to have excellent CO tolerance and oxidation reactions.^16–24 Although the catalytic life is lengthened in these Pt alloys, CO poisoning still exists. Other techniques have been investigated to improve the CO tolerance in fuel cells, such as air bleeds (2CO + O₂ → 2CO₂) and water–gas shift reactions (WGS, CO + H₂O ↔ CO₂ + H₂).^25–31 The difficulty of air bleeds is maintaining a high reaction rate and selectivity for oxidizing CO rather than H₂,^25–27 while for WGS, the reaction shows low activities at low temperatures.^28–31 Thus, it is an ideal method for removing CO before H₂ reaches the catalyst.

The mechanism for CO poisoning has also been studied theoretically. In a practical catalytic process, the Pt catalyst is usually fully covered by H atoms and/or CO molecules and Pt catalytic activity is coverage dependent.^32,33 The dissociative energy of hydrogen on Pt(111) is between 0.70 and 0.83 eV,^34–36 while the CO adsorption energy on Pt(111) is between 1.49 and 1.8 eV.^37,38 This indicates that CO molecules will bind more strongly on Pt catalysts and block available sites for the dissociation of H atoms. Furthermore, it was found that each CO molecule could roughly exclude two H atoms on the Pt₆ cluster and the poisoning effect was partially due to the loss of Pt (5d) electrons upon CO adsorption.³⁹ It is also found that an enhancement for CO tolerance could be achieved by reducing its adsorption, facilitating CO oxidation or removing CO from H₂ feed gas, while the later seems more effective in preventing the catalyst from CO poisoning by allowing almost pure H₂ to reach the catalyst. Recently, metal deposition on carbon based materials have been widely investigated and used in fuel cells due to the high activity.^{10–14,17–19} A theoretical study indicated that several potential metal/graphene materials for the removal of CO (Ni/G, Pt/G and IrAu/G) bound CO strongly, while having minimal interaction with H₂.⁴⁰ In that work, however, only the interaction between perfect graphene and transition metals is considered.⁴⁰ It is known that defects occur frequently in graphene synthesis. Therefore, to find more applicable systems with the capability of capturing CO from H₂ feed gas, non-precious metals M (M = Sc–Zn) adsorbed on both perfect and single point defect graphene are examined. In this work, density functional theory was used to study CO and H₂ binding behaviour on different metals.

2. Computational details

2.1. Method

The calculations were performed using Vienna ab initio simulation package (VASP).^41–44 The interactions between valence electrons and ion cores were treated by Blöchl's all-electron-like projector augmented wave (PAW) method.^45,46 The exchange–correlation functional used was the generalized gradient approximation with the Perdew–Burke–Ernzerhof, known as GGA-PBE.⁴⁷ The wave functions at each k-point were expanded with a plane wave basis set and a kinetic cutoff energy of up to 400 eV. The electron occupancies were determined according to the Fermi scheme with an energy smearing of 0.1 eV. Brillouin zone integration was approximated by a sum over special selected k-points using the Monkhorst–Pack method⁴⁸ and they were set to 3 × 3 × 1. Geometries were fully optimized until the energy was converged to 1.0 × 10⁻⁶ eV per atom and the force was converged to 0.01 eV Å⁻¹. Because of the existence of the magnetic atom, spin polarization was considered in all calculations. Because H₂ adsorption would be weak physisorption, a semiempirical DFT-D2 force-field approach,^49,50 which includes the van der Waals interaction, is employed in our calculations.

2.2. Model

A 4 × 4 perfect graphene surface (G_p, including 32 C atoms) and a single point defect graphene (G_d, including 31 C atoms) are set as the planar unit cell (9.84 Å × 9.84 Å × 12 Å) for periodic calculations in this work. The structures of G_p and G_d are shown in Fig. 1a and b, respectively, as well as the possible adsorption sites of metal atom. The geometries are fully optimized after the metal atom adsorbed and the most stable structures are shown in Fig. S1 (for G_p) and S2 (for G_d) in the ESI.†

Fig. 1 Possible adsorption sites on perfect and defective graphene, respectively. B, T, Ho and D denote the bridge, top, hollow and defect sites, respectively.

The adsorption energy of the metal atoms on the graphene is calculated by eqn (1), while the adsorption energy of a gas molecule (CO or H₂) on the metal/graphene system (ΔE_g) was calculated by eqn (2). The difference between the adsorption energies of CO and H₂ is calculated by ΔE_diff in eqn (3). The larger the |ΔE_diff|, the more effective the ability to separate CO from H₂ feed gas.

ΔE_m = E_{metal/graphene} − E_graphene − E_metal (1)

ΔE_g = E_{(metal+gas)/graphene} − E_{metal/graphene} − E_gas (2)

ΔE_diff = ΔE_CO − ΔE_H2 (3)

where the E_graphene, E_{metal/graphene} and E_{(metal+gas)/graphene} denote the total energies of isolated graphene, metal/graphene and H₂/CO adsorption on metal/graphene, respectively. The E_metal and E_gas denote the total energies of a single metal atom and isolated gas, respectively, which are calculated by setting the single metal atom or gas in a box of 12 Å × 12 Å × 12 Å. Negative ΔE indicates exothermic chemisorption, and a positive value suggests endothermic chemisorption.

3. Results and discussion

3.1. Stability of 3d transition metal (from Sc to Zn) adsorption on graphene

For perfect graphene (G_p), all the possible metal adsorption sites, i.e. bridge (B), hollow (Ho) and top (T) sites are examined (Fig. 1). It is found that the Ho site is energetically the most stable site for most of the 3d metal atom adsorptions (except Cu), which is in agreement with the pervious theoretical study.⁵¹ For Cu, the most stable site is at the B site.

For perfect graphene, the adsorption energy is the largest for Ti/G_p with ΔE_m = −1.95 eV, followed by Ni/G_p with ΔE_m = −1.75 eV (Fig. 2). For Cr, Mn, Cu and Zn with either half filled d⁵ or fully filled d¹⁰ orbitals, the adsorption energy is relatively high, in particular for Zn/G_p with ΔE_m = −0.18 eV. This suggests that the interaction between M and G_p will be influenced by the number of d electrons of the metal atom.⁴⁰

Fig. 2 Adsorption energies of metal atoms (Sc–Zn) on perfect graphene (black line) and defective graphene (red line), respectively.

For defective graphene, the calculated adsorption energies show a similar trend to the perfect one. However, the absolute values of the adsorption energy of metal atoms significantly increases (Fig. 2) due to the charge transfer mechanism,^52,53 suggesting that defects on graphene surfaces can enhance the adsorption of metal atoms. This can be explained by the fact that the dangling bonds at the defect site are saturated by the metal atom and it further induces the strong interaction between the metal atom and the nearby C atoms. This is consistent with the previous observations.^52,53 The largest adsorption energy is found for Ti/G_d (ΔE_m = −8.15 eV), followed by Co/G_d (ΔE_m = −7.96 eV). Because Zn is the least stable in both perfect and defective graphene, it is excluded in the following CO/H₂ separation study.

3.2. Adsorption of CO and H₂ on M/G_p

For adsorption of H₂ and CO on M/G_p (M = Sc–Cu), the optimized geometry structures are shown in the ESI (Fig. S3 and S4†). Our calculations indicated that the average distance between the metal atom and CO molecule (Table 1) decreases from Sc (2.12 Å) to Ni (1.72 Å), with a slight increase for Cu/G_p (1.79 Å). A similar result is observed for H₂ adsorption, i.e., the average distance between M and H₂ decreases from Sc (2.05 Å) to Ni (1.54 Å), with a slight increase for Cu (1.59 Å). For CO adsorption, the energies increase from Sc (−1.17 eV) to Ni (−2.90 eV), with a dramatic decrease for Cu/G_p (−1.59 eV) compared with Ni/G_p (Fig. 3a). Thus, Ni/G_p has the strongest adsorption for CO with a value of −2.90 eV, followed by Co/G_p (−2.69 eV) and Fe/G_p (−2.56 eV). Similar to the results of CO adsorption, more stable H₂ adsorption is found for Ni/G_p, Co/G_p, and Fe/G_p with values of −1.40, −1.24 and −1.09 eV, respectively. Furthermore, the energy difference between ΔE_CO and ΔE_H2, i.e., ΔE_diff is the largest for Ni/G_p with a value of −1.50 eV, followed by Fe/G_p, Co/G_p and Mn/G_p with values of −1.47, −1.45, and −1.45 eV, respectively. Because Mn/G_p is relatively unstable (ΔE_m = −0.41 eV) compared with the other three systems (ΔE_m = −1.15, −1.54 and −1.75 eV for Fe/G_p, Co/G_p and Ni/G_p, respectively), this indicates that the M/G_p (M = Fe, Co and Ni) systems are likely to be the promising materials for the removal of CO from the H₂ feed gas, particularly for Ni/G_p.

Table 1 Average bond distances (Å) between the adsorbed molecules and metal atoms

	Perfect graphene		Defective graphene
	CO	H2	CO	H2
Sc	2.12	2.05	2.37	2.59
Ti	2.04	1.99	2.23	2.34
V	1.99	1.89	2.09	2.07
Cr	1.93	1.83	2.03	1.99
Mn	1.82	1.66	1.90	1.75
Fe	1.78	1.64	1.88	1.75
Co	1.75	1.56	1.84	1.82
Ni	1.72	1.54	1.85	1.81
Cu	1.79	1.59	1.84	1.72

---

Fig. 3 Adsorption energies of CO (ΔE_CO red line) and H₂ (ΔE_H2, black line) on (a) M/G_p and (b) M/G_d, respectively. Blue lines denote the difference between the adsorption energies of CO and H₂, i.e. ΔE_CO − ΔE_H2.

3.3. The adsorption of CO and H₂ on M/G_d

The adsorption geometries of CO and H₂ on M/G_d are shown in the ESI (Fig. S5 and S6†). For CO adsorption, the average distance between the metal atom and carbon decreases from 2.37 Å (Sc) to 1.84 Å (Cu) (Table 1). For H₂ adsorption, the average distance between the metal atom and H₂ decreases from Sc (2.59 Å) to Cu (1.72 Å) in general with a slight increase for Co (1.82 Å) and Ni (1.81 Å) compared with Mn (1.75 Å) and Fe (1.75 Å). For the adsorption energy, it is seen that Fe/G_d is the most stable system for H₂ (−0.54 eV) and CO (−1.54 eV) adsorption, followed by Co/G_d system (−0.26 and −1.34 eV for H₂ and CO, respectively) (Fig. 3b). Furthermore, the gas molecule adsorption on M/G_d (Fig. 3b) is weaker than that on M/G_p (Fig. 3a), which is consistent with the longer metal–gas molecule distances for the former. This revealed that the defect on the graphene surface will weaken the ability of the metal atom adsorbing CO and H₂. This can be explained by the change in orbital occupancy after CO and H₂ are adsorbed. The previous theoretical studies found that the d orbitals of metal atoms play a major role in gas molecule adsorption.^40,54,55 Our results indicated that the change of d orbital occupancy in M/G_d is smaller than that of M/G_p after CO and H₂ adsorption (Fig. 4). The strong adsorption of the metal atom at the defect point makes metal 3d electrons more localized. This reduces the number of electrons bonding with CO or H₂. Therefore, CO and H₂ are weakly adsorbed on M/G_d.

Fig. 4 Total change in d orbital occupancy of Fe, Co and Ni when CO and H₂ is adsorbed with reference to the isolated M/G_p(d).

Similar to perfect graphene, Fe/G_d, Co/G_d and Ni/G_d are the promising materials for the removal of CO from the H₂ feed gas, with Co/G_d slightly favoured. Furthermore, the average ΔE_diff of M/G_d is ∼0.42 eV, which is lower than that of M/G_p, suggesting that M/G_p would be more efficient than M/G_d for CO and H₂ separation.

3.4. Adsorption mechanism of CO/H₂ on M/G_p and M/G_d (M = Fe, Co and Ni)

According to the discussion above, we have seen that Fe, Co and Ni systems with large ΔE_diff are more effective in removing CO from H₂ feed gas. To further understand the mechanism of CO and H₂ adsorption on M/G_p and M/G_d (M = Fe, Co and Ni), we studied their electronic structures. The partial density of states of M/G_d (M = Fe, Co and Ni) are shown in Fig. 5–7, respectively. Because compared with Fe/G_p(d), the adsorption mechanism of CO and H₂ is the same as Co/G_p(d) and Ni/G_p(d), only the mechanism for Fe/G_p(d) is discussed. For M/G_p(d) (M = Fe, Co and Ni) without gas molecule adsorption (defined as iso. in Fig. 5–7), the five d orbitals of metal atoms split to three groups based on energy as (d_yz and d_xz), (d_xy and d_x²−y²), and d_z². Compared with the single metal atoms (Fig. 8), it is seen that all the orbitals of d band moves upwards and close to the Fermi level, particularly for d_yz and d_xz. This suggests that the graphene substrate enhances the adsorption of gas by increasing the d band, in agreement with d band center theory.^54,55

Fig. 5 Partial density of states (DOS) of CO and H₂ adsorption on Fe/G_p and Fe/G_d. Iso. represents the isolated M/G_p without CO or H₂ adsorption. The dotted line denotes the Fermi level.

Fig. 6 Partial density of states (DOS) of CO and H₂ adsorption on Co/G_p and Co/G_d. For a detailed description, see Fig. 5.

Fig. 7 Partial density of states (DOS) of CO and H₂ adsorption on Ni/G_p and Ni/G_d. For a detailed description, see Fig. 5.

Fig. 8 Partial density of states (DOS) of single Fe (a), Co (b) and Ni (c) atom.

For perfect graphene, the calculated density of states of Fe/G_p indicated that only d_yz and d_xz orbitals form bonding states (below Fermi level) and anti-bonding states at ∼3.0 eV (above Fermi level) with carbon p orbitals. For other orbitals, only the bonding state is obtained (Fig. 5a). Therefore, the main contribution to CO adsorption would come from d_yz and d_xz orbitals. For H₂ adsorption, Fig. 5b clearly shows that only d_yz orbital form clear bonding and anti-bonding states with the s orbital of H, suggesting that d_yz orbital is important in H₂ adsorption.

For defective graphene, clear bonding and anti-bonding states are observed for all d orbitals (Fig. 5), different from the case of perfect one. This demonstrated that the metal atom has a stronger interaction with defective graphene, in agreement with the high adsorption energy of M/G_d compared with M/G_p (Fig. 2). For CO adsorption (Fig. 5c), d_yz and d_xz orbitals have the largest energy gap between bonding and anti-bonding states. This indicates that these two orbitals are important for the dissociation of the metal–carbon bond and make the largest contribution to CO adsorption. Similarly, for H₂ adsorption on Fe/G_d (Fig. 5d), both d_yz/d_xz and d_xy/d_x²−y² orbitals are important for H₂ adsorption, with the former being slightly favoured.

4. Conclusions

For 3d metal/perfect-graphene (M/G_p) and metal/defect-graphene (M/G_d) (M = Sc–Zn), very different behavior is observed for the adsorption of CO and H₂. The two gas molecules interact more strongly with perfect graphene (M/G_p) compared with defect one (M/G_d). For both defective and perfect graphene, CO bonds form more strongly with M/G_d(p), compared with H₂. Fe, Co and Ni show high efficiency in separating CO from H₂ feed gas. d_yz or/and d_xz orbitals of the metal atoms are found to play a major role in CO and H₂ adsorption. We expect that the obtained results are useful for removing CO from H₂ feed gas and reducing or eliminating CO poisoning of anode catalysts.

Acknowledgements

This work is supported by the National Natural Science Foundation of China (Grant no. 21221061, 21203174), Jilin Province Youth Fund (Grant no. 20130522141JH) and Jilin Province Computing Center (Grant no. 20130101179 JC-08, 20130101179 JC-07). The authors also thank the financial support from Department of Science and Technology of Sichuan Province (Grant no. 2011GZX0077, 2012JZ0007, 2014HH0049). Part of the computational time is supported by the Performance Computing Center of Jilin University.

References

1. L. Carrette, K. A. Friedrich and U. Stimming, ChemPhysChem, 2000, 1, 162–193.
2. J. C. Davies, J. Bonde, Á. Logadóttir, J. K. Nørskov and I. Chorkendorff, Fuel Cells, 2004, 5, 429–435.
3. W. Vogel, J. Lundquist, P. Ross and P. Stonehart, Electrochim. Acta, 1975, 20, 79–93.
4. H. Igarashi, T. Fujino and M. Watanabe, J. Electroanal. Chem., 1995, 391, 119–123.
5. M. Watanabe and S. Motoo, J. Electroanal. Chem., 1975, 60, 267–273.
6. E. Christoffersen, P. Liu, A. Ruban, H. L. Skriver and J. K. Nørskov, J. Catal., 2001, 199, 123–131.
7. P. Liu, J. K. Logadottir and J. K. Nørskov, Electrochim. Acta, 2003, 48, 3731–3742.
8. P. N. Ross, K. Kinoshita, A. J. Scarpellino and P. Stonehart, J. Electroanal. Chem., 1975, 59, 177–189.
9. T. Ioroi, N. Kitazawa, K. Yasuda, Y. Yamamoto and H. Takenaka, J. Electrochem. Soc., 2000, 147, 2018–2022.
10. A. C. Garcia, V. A. Paganin and E. A. Ticianelli, Electrochim. Acta, 2008, 53, 4309–4315.
11. K. L. Ley, R. X. Liu, C. Pu, Q. B. Fan, N. Leyarovska, C. Segre and E. S. Smotkin, J. Electrochem. Soc., 1997, 144, 1543–1548.
12. F. Colmati-Júnior, W. H. Lizcano-Valbuena, G. A. Camara, E. A. Ticianelli and E. R. Gonzalez, J. Braz. Chem. Soc., 2002, 13, 474–482.
13. P. P. Lopes and E. A. Ticianelli, J. Electroanal. Chem., 2010, 644, 110–116.
14. J. Kaiser, L. Colmenares, Z. Jusys, R. Mörtel, H. Bönnemann, G. Kohl, H. Modrow, J. Hormes and R. J. Behm, Fuel Cells, 2006, 6, 190–202.
15. M. Watanabe and S. J. Motoo, J. Electroanal. Chem., 1975, 60, 275–283.
16. H. Uchida, K. Izumi and M. Watanabe, J. Phys. Chem. B, 2006, 110, 21924–21930.
17. D. H. Lim, D. H. Choi, W. D. Lee and H. I. Lee, Appl. Catal., B, 2009, 89, 484–493.
18. T. C. M. Nepel, P. P. Lopes, V. A. Paganin and E. A. Ticianelli, Electrochim. Acta, 2013, 88, 217–224.
19. S. Mukerjee, R. C. Urian, S. J. Lee, E. A. Ticianelli and J. McBeen, J. Electrochem. Soc., 2004, 151, A1094–A1103.
20. M. Watanabe, H. Igarashi and T. Fujino, Electrochemistry, 1999, 67, 1194–1196.
21. M. Watanabe, Y. Zhu and H. Uchida, J. Phys. Chem. B, 2000, 104, 1762–1768.
22. M. Watanabe, Y. Zhu, H. Igarash and H. Uchida, Electrochemistry, 2000, 68, 244–251.
23. Y. Dai, Y. Liu and S. Chen, Electrochim. Acta, 2013, 89, 744–748.
24. H. Igarashi, T. Fujino, Y. Zhu, H. Uchida and M. Watanabe, Phys. Chem. Chem. Phys., 2001, 3, 306–314.
25. S. Alayoglu, A. U. Nilekar, M. Mavrikakis and B. Eichhorn, Nat. Mater., 2008, 7, 333–338.
26. J. Zhang and R. Datta, J. Electrochem. Soc., 2005, 152, A1180–A1187.
27. E. D. Park, D. Lee and H. C. Lee, Catal. Today, 2009, 139, 280–290.
28. T. Utaka, K. Sekizawa and K. Eguchi, Appl. Catal., 2000, 194–195, 21–26.
29. J. R. Ladebeck and J. P. Wagner, in Handbook of fuel cell technology, ed. W. Vielstich, A. Lamm and H. A. Gasteiger, Wiley-VCH, New York, 2003, vol. 3, pp. 190–201.
30. E. I. Santiago, M. S. Batista, E. M. Assaf and E. A. Ticianelli, J. Electrochem. Soc., 2004, 151, A944–A949.
31. Y. Tanaka, T. Utaka, R. Kikuchi, K. Sasaki and K. Eguchi, Appl. Catal., 2003, 238, 11–18.
32. L. Chen, A. C. Cooper, G. P. Pez and H. S. Cheng, J. Phys. Chem. C, 2007, 111, 5514–5519.
33. C. Zhou, J. Wu, A. Nie, R. C. Forrey, A. Tachibana and H. S. Cheng, J. Phys. Chem. C, 2007, 111, 12773–12778.
34. G. Papoian, J. K. Nøskov and R. Hoffmann, J. Am. Chem. Soc., 2000, 122, 4129–4144.
35. K. Chrismann, Surf. Sci. Rep., 1988, 9, 1–163.
36. D. J. Godbey and G. A. Somorjai, Surf. Sci., 1988, 204, 301–313.
37. M. T. M. Koper, T. E. Shubina and R. A. van Santen, J. Phys. Chem. B, 2002, 106, 686–692.
38. S. E. Mason, I. Grinberg and A. M. Rappe, J. Phys. Chem. C, 2008, 112, 1963–1966.
39. L. Chen, B. Chen, C. G. Zhou, J. P. Wu, R. C. Forrey and H. S. Cheng, J. Phys. Chem. C, 2008, 112, 13937–13942.
40. D. J. D. Durbin and C. Malardier-Jugroot, J. Phys. Chem. C, 2011, 115, 808–815.
41. G. Kresse and J. Furthmüller, Comput. Mater. Sci., 1996, 6, 15–50.
42. G. Kresse and J. Hafner, Phys. Rev. B: Condens. Matter Mater. Phys., 1993, 47, 558–561.
43. G. Kresse and J. Hafner, Phys. Rev. B: Condens. Matter Mater. Phys., 1994, 49, 14251–14269.
44. G. Kresse and J. Furthmüller, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 54, 11169–11186.
45. P. E. Blöchl, Phys. Rev. B: Condens. Matter Mater. Phys., 1994, 50, 17953–17979.
46. G. Kresse and D. Joubert, Phys. Rev. B: Condens. Matter Mater. Phys., 1999, 59, 1758–1775.
47. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865–3868.
48. H. J. Monkhorst and J. D. Pack, Phys. Rev. B: Condens. Matter Mater. Phys., 1976, 13, 5188–5192.
49. S. Grimme, J. Comput. Chem., 2006, 27, 1787–1799.
50. S. Grimme, J. Antony, S. Ehrlich and H. Krief, J. Chem. Phys., 2010, 132, 154104.
51. X. J. Liu, C. Z. Wang, Y. X. Yao, W. C. Lu, M. Hupalo, M. C. Tringides and K. M. Ho, Phys. Rev. B: Condens. Matter Mater. Phys., 2011, 83, 235441.
52. D. Datta, J. Li and V. B. Shenoy, ACS Appl. Mater. Interfaces, 2014, 6, 1788–1795.
53. W. Song, M. G. Jiao, K. Li, Y. Wang and Z. J. Wu, Chem. Phys. Lett., 2013, 588, 203–207.
54. B. Hammer and J. K. Nørskov, Surf. Sci., 1995, 343, 211–220.
55. B. Hammer, Top. Catal., 2006, 37, 3–16.

---

Footnote

† Electronic supplementary information (ESI) available: Fig. S1 and S2 give the structures of metal atom (Sc–Zn) adsorption on G_p and G_d, respectively; Fig. S3 and S4 give the structures of CO and H₂ adsorption on M/G_p, respectively; Fig. S5 and S6 give the structures of CO and H₂ adsorption on M/G_d, respectively. See DOI: 10.1039/c4ra15937c

---