Band and bonding characteristics of N₂⁺ ion-doped graphene

Abstract

We report that the doping of energetic nitrogen cations (N₂⁺) on graphene effectively controls the local N–C bonding structures and the π-band of graphene critically depending on ion energy E_k (100 eV ≤ E_k ≤ 500 eV) by using a combined study of photoemission spectroscopy and density functional theory (DFT) calculations. With increasing E_k, we find a phase transformation of the N–C bonding structures from a graphitic phase where nitrogen substitutes carbon to a pyridinic phase where nitrogen loses one of its bonding arms, with a critical energy E^c_k = 100 eV that separates the two phases. The N₂⁺-induced changes in the π-band with varying E_k indicate an n-doping effect in the graphitic phase for E_k < E^c_k but a p-doping effect for the pyridinic graphene for E_k > E^c_k. We further show that one may control the electron charge density of graphene by two orders of magnitude by varying E_k of N₂⁺ ions within the energy range adopted. Our DFT-based band calculations reproduce the distinct doping effects observed in the π-band of the N₂⁺-doped graphene and provide an orbital origin of the different doping types. We thus demonstrate that the doping type and electron number density in the N₂⁺ ion-doped SLG can be artificially fine-controlled by adjusting the kinetic energy of incoming N₂⁺ ions.

---

1. Introduction

The surpassing properties of Dirac fermions in the π-band of a single layer graphene (SLG) over conventional materials have attracted extensive research efforts in wide branches of science and technology for future graphene-based industrial applications.^1–6 Doping of external atoms to graphene as either donors or acceptors has been adopted as a means of engineering its electronic band structure in a controllable way. This includes adsorption of charge-neutral atoms^2,3 or molecules⁴ on SLG and doping with nitrogen (N)^5–15 or boron^16,17 atoms or ions^18–21 to substitute carbon atoms of SLG or to form more complex atomic structures, as routinely done in typical semiconducting materials. In particular, doping with nitrogen in SLG appears to be quite effective in regulating electronic and chemical properties of carbon materials including sensors,^5,22–24 lithium battery,²⁵ fuel cells,²⁶ and field-effect transistors.^7,12 The comparable atomic size and five valence electrons of N best fit to form strong covalent bonds with C atoms and modify electronic properties rather strongly within a few lattice spacing from the dopants.^6,14,27–31

Here we report the characteristic changes in the π-band of SLG and the N–C local bonding configurations, as nitrogen ions are doped in SLG at 900 K in the form of N₂⁺ with kinetic energy E_k in the range of 100 eV ≤ E_k ≤ 500 eV. We have measured N₂⁺-induced changes in the valence band and core levels (N 1s and C 1s) as a function of E_k. We also have performed density functional theory (DFT) based band calculations to account for the changes in our photoemission data obtained with synchrotron photons. The behavior of the N 1s core level with increasing E_k unambiguously indicates the presence of three distinct N–C bonding configurations: (i) graphitic N where an N atom substitutes a C atom of SLG, (ii) pyridinic N where an N atom loses one of its neighboring C atoms as a vacancy defect created by the high energy N₂⁺ ions, and (iii) adsorbed molecular N₂. The relative abundance of these bonding configurations critically varies depending on E_k showing all three different forms coexisting with the graphitic N most abundant for low energy ions (E_k < 100 eV) as reported earlier,^18,20,21 but we show that only the pyridinic N survives for high energy regime for E_k ≥ 300 eV significantly reducing the charge carrier density n, which is found to vary by the two orders of magnitude as E_k changes from 100 eV to 500 eV. The measured π-bands show an n-doping effect with increased n for graphitic N and a p-doping effect for pyridinic N with remarkably reduced n. The linear Dirac feature of the π-band, however, persists upon the doping of N₂⁺ ions for all values of E_k examined in this study as confirmed by the slightly reduced Fermi velocities for these three bonding configurations from that of a clean SLG. Our DFT-based calculated bands indeed reproduce these doping effects, and reveal that the number of vacancies produced by the energetic ions rather than the number of pyridinic N atoms determines the type of doping and the resulting charge carrier density n.

2. Experimental and computational sections

2.1 Sample preparation

We have grown epitaxial SLG samples on 6H-SiC(0001) substrates by thermally annealing the SiC substrate under ultra-high vacuum (UHV) condition. The well-characterized evolution of the low energy electron diffraction (LEED) pattern upon forming SLG on SiC begins with a Si-rich 3 × 3 phase, which develops into a (√3 × √3) R30° phase when annealed at 1200 K, as measured by using an optical pyrometer. We continued to anneal the substrate further in situ at 1500 K to obtain a carbon-rich surface with the (6√3 × 6√3) R30° phase. Although there is a subtle but discernible change in the LEED pattern indicating the formation of SLG, we find the most convincing evidence from the presence of the linear π-band in our photoemission data. The N₂⁺ ions were deposited on the SLG sample by using an ion gun (PHP 3000), where E_k could be delicately controlled. To minimize surface defects and establish bonding configurations in thermal equilibrium, the SLG sample was maintained at 900 K, below desorption temperature of nitrogen on SLG, during the doping of N₂⁺ ions of 120 langmuir.

2.2 Valence band and core level data

Our angle-resolved photoemission spectroscopy (ARPES) and high-resolution core level spectroscopy (HRCLS) data were obtained at the beamline 10D of the Pohang Accelerator Laboratory (PAL) in Korea using synchrotron photons of energy 50 eV and 500 eV, respectively. Our UHV chamber kept under 1 × 10⁻¹⁰ Torr during the measurements was equipped with a PHOIBOS 150 electron analyzer. We have estimated atomic concentrations, δ_N, of the three different bonding forms of nitrogen atoms in our N₂⁺-doped graphene from their respective intensity in the N 1s core level normalized by the intensity of the C 1s. These intensities from different atoms were weighted by the respective photoionization cross sections.

2.3 DFT calculations

We have performed DFT calculations for the n/p-doped SLG treated by the high energy N₂⁺ ions. We employed the projector augmented wave method (PAW)-PBE^32,33 potential implemented in VASP³⁴ package. Every atom was relaxed in the structural optimizations with the Monkhorst–Pack k-grid 1 × 1 × 1 until force is less than 0.02 eV Å⁻¹. In order to obtain the band structures, the grids of k-points were set to the Monkhorst–Pack 2 × 2 × 1 in self-consistent field (SCF) calculations and 3 × 3 × 1 in non-self-consistent field (non-SCF) calculations. The band structures were plotted along the high symmetry k-points of the graphene supercell (13 × 13) where the k-point is equivalent to the Dirac point (DP) at K in the case of the 1 × 1 graphene. For the partial density of states (PDOS) calculations, SCF calculations with the k-point grids of 6 × 6 × 1 were performed, while SCF calculations with the k-point grids of 3 × 3 × 1 were carried out for defect-formation energy calculations. As for the DFT calculation on N 1s core-level energies, we employed the B3LYP functional³⁵ using the 6-31G* basis set^36–39 in the Gaussian (G09) suite of programs.⁴⁰

3. Results and discussion

3.1 N₂⁺ ion-induced changes in the π-band of graphene

Fig. 1 schematically illustrates a nitrogen-doped SLG, where graphitic N and pyridinic N atoms are marked as red- and blue-balls, respectively, which are sp² hybridized with neighboring C atoms of SLG. Note the presence of C-vacancies with several different forms in the vicinity of pyridinic N atoms.^20,31,41,42 Three test values for E_k were adopted (100, 300, and 500 eV) to control the electron number density n by adjusting the relative abundance of the three bonding configurations.

Fig. 1 Schematic illustration of N₂⁺-doped SLG containing two kinds of doped nitrogen, graphitic N (red) and pyridinic N (blue). Some of C atoms in graphene (black) are missing in several different forms for pyridinic N atoms, while C atoms in the buffer layer (gray circles) are not affected by the incoming N₂⁺ ions.

We present in Fig. 2 the changes of the π-band upon doping N₂⁺ ions of three different energies. These bands were measured by using synchrotron photon of energy 50 eV along the direction perpendicular to the Γ → K direction near the k-point of the Brillouin zone as indicated in Fig. 1a. The well-defined linear π-band obtained from SLG shows the DP at 0.45 eV below the Fermi energy (E_F) due to the electron charge transfer from the SiC substrate. We have estimated Fermi velocity v_F of the SLG in Fig. 2a from the slope of the linear π-band determined by connecting intensity maxima of the momentum distribution curves (MDCs) at two different energies near the DP. We then calculated electron number density n of the sample from the value of v_F.⁴³ We find v_F = 0.87 × 10⁶ m s⁻¹ and n₀ = 1.9 × 10¹³ cm⁻² of the SLG in Fig. 2a, which are quite similar to the values reported earlier.²¹ The π-bands presented in Fig. 2b–d for the N₂⁺-doped SLG with E_k = 100 eV, 300 eV, and 500 eV, respectively, appear to be much diffused and fuzzy, mainly due to the defects unavoidably created by the energetic N₂⁺ ions.²⁰ The most noteworthy observation from the π-bands of the N₂⁺-doped SLG samples is the shift of the DP depending on the ion energy E_k. As shown in Fig. 2b, the DP shifts away from E_F to E_DP = 0.98 eV for E_k = 100 eV, indicating the n-doping effect. Incidentally, for E_k = 300 eV and 500 eV, the DP shifts towards E_F with E_DP = 0.32 eV and 0.14 eV, respectively, indicating the p-doping effect. When we dope more N₂⁺ ions of 120 Langmuirs with E_k = 100 eV on the SLG in Fig. 2b, the DP shifts further down to E_DP = 1.22 eV for a more n-doped SLG. Thus, we demonstrate that the doping level of Dirac fermions can be elaborately controlled by adjusting the kinetic energy E_k of the incoming N₂⁺ ions onto the SLG even though the π-band becomes weaker and fuzzier. We also find that N₂⁺ ions of E_k = 50 eV are not proportionally doped on SLG maintained at 900 K while ions of E_k > 500 eV make the π-band of SLG almost disappeared.

Fig. 2 Change of the π-band of SLG upon doping nitrogen ions. (a) Clean SLG, N₂⁺-doped SLG with E_k = (b) 100 eV, (c) 300 eV, and (d) 500 eV, respectively.

Despite the broadening due to the random distribution of N₂⁺ ions doped and the defects created by the ions, the linear nature of the π-band still persists as seen in Fig. 2b–d. For the N₂⁺-doped SLGs of E_k = 100, 300, and 500 eV, the Fermi velocities estimated from their slopes of the linear π-bands as done for a clean SLG are v_F = 0.71, 0.70, and 0.67 × 10⁶ m s⁻¹, respectively. Note that these values are only slightly reduced from that of a clean SLG in Fig. 2a, suggesting that the Dirac nature of the π-band is preserved for all the N₂⁺-doped SLGs.

As obtained from the values of v_F, an order of magnitude increase of electron number density as compared with n₀ of the clean SLG in Fig. 2a is seen for SLG with E_k = 100 eV (n = 1.4 × 10¹⁴ cm⁻²), while a noticeable reduction from n₀ is found for E_k = 300 eV and 500 eV (n = 1.5 × 10¹³ and 3.2 × 10¹² cm⁻²), as suggested by the opposite doping effects described earlier. Such a change in the electron number density reflects the redistribution of the electron charges through charge transfers among atoms to satisfy the Bader's condition being no density gradient along the surface normal.⁴⁴

3.2 Three different phases of N₂⁺-doped SLG and their dirac points

The changes of N 1s core level shown in Fig. 3 unambiguously suggest the presence of three distinct bonding configurations of the doped N atoms in SLG. The spectrum in Fig. 3a for the N₂⁺ ion-doped SLG with E_k = 100 eV is best fitted with three components at the binding energies E_b = 398.4 eV, 401.4 eV, and 402.8 eV, representing the pyridinic N (N_p), graphitic N (N_g), and molecular nitrogen (N_m) bonding in the form of –C=N–C–, –C–N–C– and N₂ bonds, respectively as reported earlier.^20,31

Fig. 3 N 1s core level spectra obtained from N₂⁺-doped SLG with E_k = (a) 100 eV, (b) 300 eV, and (c) 500 eV, respectively. Pyridinic nitrogen (N_p), graphitic nitrogen (N_g), and molecular nitrogen (N_m) are denoted by blue, red, and violet shades, respectively. Note that N_p is 3.0 eV downshifted than N_g, which is consistent with the DFT calculations (B3LYP/6-31G*) for N 1s core level energies for which N_p is 3.7 eV downshifted than N_g, while N_g is hardly downshifted as compared with a free N atom. (d) The atomic concentrations δ_N of the three different types of nitrogen atoms in graphene doped with N₂⁺ ions of energy 100 eV, 300 eV and 500 eV (from top to bottom).

The dominant N_p component in the N₂⁺ ion-doped SLG with E_k = 100 eV increases further when E_k increases to 300 eV as seen in Fig. 3b, where N_g becomes negligibly small compared to N_p, while N_m completely disappears. It must be noted that N_g vanishes completely with N₂⁺ ions of E_k = 500 eV. These observations indicate that the defects created by the energetic ions become dominant even with ions of energy E_k = 100 eV. In Fig. 3d, we estimate the sum of three atomic concentrations δ^T_N = 1.10, 1.07 and 1.05% for E_k = 100, 300, and 500 eV, respectively. We note that the three samples of nearly the same amount of nitrogen atoms with different mixing ratios between the pyridinic N and graphitic N (Fig. 3) produce much different values for n and E_DP (Fig. 2). This unusual feature of the N₂⁺-doped SLG suggests that there must be another factor more decisive in determining the doping level than the relative abundance of pyridinic N over the graphitic N as discussed in the paragraph associated with Fig. 5. The changes in the Dirac point E_DP and the binding energies E_b of the three components of N 1s core level are summarized in Fig. 4, where the aforementioned n-doping for E_k = 100 eV and p-doping to the SLG for E_k = 300 eV and 500 eV can be found. We find a similar trend showing n-doping for E_k = 100 eV and p-doping with N₂⁺ ions of higher energies from the changes in C 1s core level (not shown). By considering one more electron in nitrogen than in carbon for graphitic N and the presence of vacancies for pyridinic N, we can ascribe the n-doping to graphitic N and the p-doping to the vacancies near pyridinic N as further supported by the trends with E_DP for N_g and N_p.²⁷

Fig. 4 Evolution of relative abundance of three nitrogen bonding configurations in N₂⁺-doped SLG with increasing ion energy. Blue, red, violet lines represent pyridinic nitrogen (N_p), graphitic nitrogen (N_g), and molecular nitrogen (N_m). The black lines denote the change of the DP of the π-band.

3.3 DFT-based band calculations and bonding characteristics of N₂⁺-doped graphene

Our DFT-based band calculations for the N₂⁺-doped SLGs presented in Fig. 5 also support our conclusion that the n-doping stems from graphitic N while the p-doping from pyridinic N atoms with vacancies. In Fig. 5a, the band structure for an optimized 13 × 13 cell freestanding SLG exhibits the π-band with the DP at the Fermi level E_F. One finds how the π-band is modified in the presence of doped nitrogen atoms and vacancy defects from Fig. 5b and c, respectively, where a band gap opens at the DP commonly to these SLGs because of the broken symmetry in the carbon sublattices of SLG.^21,45 Unfortunately, we do not directly observe such a small band from the fuzzy and weak π-bands due to the randomly distributed N₂⁺ ion with a limited instrumental energy resolution. The band gap appears to be greater in the pyridinic N than that of the graphitic-N due essentially to the enhanced asymmetry in the charge density among carbon atoms with the increased number density of the pyridinic N type defects as seen in Fig. 3.

Fig. 5 Calculated band structures for SLG with (a) no defect or dopant, (b) a graphitic N, and (c) a pyridinic N for δ_N = 0.3%. Doped nitrogen is depicted as a blue circle. Only a part of the 13 × 13 unit cell is depicted. The SLG with the graphitic-N shows n-doping with a small band gap, while the SLG with the pyridinic-N and a vacant site gives p-doping with a band gap.

The graphitic N produces an n-doping with its lone pair having one more excess electron as compared with a C atom, while the pyridinic N with a single vacancy (SV) induces the p-doping. It is interesting to note, however, the prominent n-doping effect for the SLG doped with N₂⁺ ions of E_k = 100 eV (Fig. 2b) where pyridinic N dominates (Fig. 3a). As depicted in Fig. 1, there are diverse forms of vacancies near a pyridinic N or different number of pyridinic N atoms around a single SV as reported earlier.⁴¹ Since our band calculations reveal that the number of vacancies rather than number of pyridinic N atoms determines the doping level, we thus infer that there are more graphitic N atoms than the number of vacancies although pyridinic N atoms are most abundant in the SLG doped with N₂⁺ ions of E_k = 100 eV.

The calculated partial density of states (PDOS) for SLG reveals that the orbital densities sharply increase/decrease above the Fermi level ascribed to the impurity/vacancy, reflecting the n-doping/p-doping effect as seen in Fig. 6. Interestingly, we observe in Fig. 6b and d that the N_p p_z orbital is strongly overlapping with two neighboring dangling carbon atoms p_z orbitals, forming apparently a local bonding-type molecular orbital in the positively charged triangular 3-site-ring constructed by N_p and two dangling C atoms. The DFT (PAW-PBE) predicted defect-formation energies (E_d: eV) for selected atoms are given in Fig. 7. The E_d for N_p which is bonded to only two C atoms is 11.0 eV. Though this value is smaller than the E_d for N_g (15.3 eV) which is bonded to three C atoms (namely, N_p is less stable than N_g), the value of 11.0 eV is not small in consideration of only two N–C bonds (15.3 × (2/3) = 10.2 eV). This extra binding, though small, arises from the 3-site overlap in p_z orbitals among N_p and two neighboring dangling C atoms. This substantial overlap causes transfer of a minor portion of electron density of the N 1s orbital to the N 2s and N 2p_z orbitals, and so is responsible for the significant N_p 1s core level energy downshift (by 3 eV in Fig. 3) as compared with the N_g 1s core which shows a negligible downshift as compared with a N 1s core of an isolated N atom. As for the stability of C atoms, the C atom in the graphene (A₁: 17.2 eV in Fig. 7) is much more stable, followed by the C atom bonded to N_p in the pyridinic N doped SLG (A₅: 16.4 eV), the C atom bonded to N_g in the graphitic N doped SLG (A₃: 15.2 eV), and followed by the dangling C atom in the pyridinic N doped SLG (A₆: 14.3 eV).

Fig. 6 Calculated PDOS for selected atoms (red circle) in SLG with (a) a graphitic N and (c) a pyridinic N dopant. The corresponding partial electron charge densities (which take into band below the Fermi level E_F) are shown in (b) and (d), respectively. In contrast to the N-doping effect by graphitic N in (b) and (d) shows the significant coupling between the pyridinic N p_z orbital and two neighboring dangling C p_z orbitals arising from a vacant C site.

Fig. 7 Defect-formation energy (E_d: eV) for selected atoms (red circle). (a) A₁: a normal C atom in SLG with no defect or dopant. (b) A₂/A₃: N_g/adjacent-C in SLG with graphitic N. (c) A₄/A₅/A₆: N_p/adjacent-C/dangling-C in SLG with pyridinic N. E_d(A₁) is the largest, whereas E_d(A₄) is the smallest. Note that E_d(A₂) ≈ E_d(A₃), E_d(A₂) > E_d(A₄), and E_d(A₅) > E_d(A₆).

4. Conclusions

In summary, the changes in the π-band of SLG and C–N bonding configurations when N₂⁺ ions of energy E_k (100 eV ≤ E_k ≤ 500 eV) are doped on SLG at room temperature have been investigated. The changes in N 1s core level upon varying E_k show that there are three distinct bonding configurations, graphitic N, pyridinic N, and molecular N. The pyridinic N with carbon vacancies appears to be dominant for all energies used, while the graphitic N and molecular N are seen only for SLGs doped with N₂⁺ ions of E_k = 100 eV. The changes in the π-band and the N 1s core level suggest that there is the n-doping effect for SLG with a charge density n = 1.4 × 10¹⁴ cm⁻², which is an order of magnitude increased from n = 1.9 × 10¹³ cm⁻² of the clean SLG for low energy ions of E_k = 100 eV stemming mainly from the graphitic N atoms. The prominent p-doping effects are found for the SLGs with ions of higher energies with n = 1.5 × 10¹³ cm⁻² and n = 3.2 × 10¹² cm⁻² for E_k = 300 eV and 500 eV, respectively, originating mostly from the increased number of vacancies created by the energetic N₂⁺ ions. We thus demonstrate that the doping level may be manipulated by two orders of magnitude by controlling the E_k of incident N₂⁺ ions. Our DFT-based band calculations also reproduce three different doping effects observed experimentally depending on E_k. The partial density of states calculated for the N₂⁺ ion-doped SLG reveals the main driving force for the doping level, i.e., the equally mixed C p_z and N p_z orbitals near the Fermi level for the graphitic N producing excess electrons of the substituted N atom mainly drives the n-doping while much reduced C p_z orbital seen for the pyridinic N around the vacancies accounts for the p-doping effect. The number of vacancies created by the energetic ions, however, plays a decisive role in determining the doping level in graphene for a given relative abundance between graphitic and pyridinic N atoms in the N₂⁺-doped SLGs. Our calculated bands further show the presence of a band gap at the DP due to the charge asymmetry between the two carbon sublattices of SLG. Thus, we illustrate that the doping type and electron number density in the N₂⁺ ion-doped SLG can be artificially fine-controlled by adjusting kinetic energy of incoming N₂⁺ ions. This means of doping nitrogen into SLG may play a key role in utilizing graphene in future carbon-based nano-electronic devices.

Acknowledgements

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea Government (NRF-2015R1A5A1009962), NRF (Global PhD Fellowship Program: 2011-0006466), NRF (National Honor Scientist Program: 2010-0020414), and also by the Ministry of Science, ICT and Future Planning (NRF-2013R1A1A2005598). Computation was supported by KISTI (KSC-2015-C3-059, KSC-2015-C3-061).

References

1. K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, M. I. Katsnelson, I. V. Grigorieva, S. V. Dubonos and A. A. Firsov, Nature, 2005, 438, 197–200.
2. F. Schedin, A. K. Geim, S. V. Morozov, E. W. Hill, P. Blake, M. I. Katsnelson and K. S. Novoselov, Nat. Mater., 2007, 6, 652–655.
3. G. Giovannetti, P. A. Khomyakov, G. Brocks, V. M. Karpan, J. van den Brink and P. J. Kelly, Phys. Rev. Lett., 2008, 101, 026803.
4. W. Chen, S. Chen, D. C. Qi, X. Y. Gao and A. T. S. Wee, J. Am. Chem. Soc., 2007, 129, 10418–10422.
5. Y. Wang, Y. Shao, D. W. Matson, J. Li and Y. Lin, ACS Nano, 2010, 4, 1790–1798.
6. X. Wang, X. Li, L. Zhang, Y. Yoon, P. K. Weber, H. Wang, J. Guo and H. Dai, Science, 2009, 324, 768–771.
7. B. Guo, Q. Liu, E. Chen, H. Zhu, L. Fang and J. R. Gong, Nano Lett., 2010, 10, 4975–4980.
8. Yu-P. Lin, Y. Ksari, D. Aubel, S. H. Garreau, G. Borvon, Y. Spiegel, L. Roux, L. Simon and J.-M. Themlin, Carbon, 2016, 100, 337–344.
9. H. Wang, T. Maiyalagan and X. Wang, ACS Catal., 2012, 2, 781–794.
10. F. Joucken, Y. Tison, J. Lagoute, J. Dumont, D. Cabosart, B. Zheng, V. Repain, C. Chacon, Y. Girard, A. R. B. Méndez, S. Rousset, R. Sporken, J.-C. Charlier and L. Henrard, Phys. Rev. B: Condens. Matter Mater. Phys., 2012, 85, 161408.
11. X. Li, H. Wang, J. T. Robinson, H. Sanchez, G. Diankov and H. Dai, J. Am. Chem. Soc., 2009, 131, 15939–15944.
12. D. Wei, Y. Liu, Y. Wang, H. Zhang, L. Huang and G. Yu, Nano Lett., 2009, 9, 1752–1758.
13. G. Abbas, P. Papakonstantinou, G. R. S. Iyer, I. W. Kirkman and L. C. Chen, Phys. Rev. B: Condens. Matter Mater. Phys., 2007, 75, 195429.
14. L. Zhao, R. He, K. T. Rim, T. Schiros, K. S. Kim, H. Zhou, C. Gutiérrez, S. P. Chockalingam, C. J. Arguello, L. Pálová, D. Nordlund, M. S. Hybertsen, D. R. Reichman, T. F. Heinz, P. Kim, A. Pinczuk, G. W. Flynn and A. N. Pasupathy, Science, 2011, 333, 999–1003.
15. D. Usachov, O. Vilkov, A. Grüneis, D. Haberer, A. Fedorov, V. K. Adamchuk, A. B. Preobrajenski, P. Dudin, A. Barinov, M. Oehzelt, C. Laubschat and D. V. Vyalikh, Nano Lett., 2011, 11, 5401–5407.
16. L. S. Panchakarla, K. S. Subrahmanyam, S. K. Saha, A. Govindaraj, H. R. Krishnamurthy, U. V. Waghmare and C. N. R. Rao, Adv. Mater., 2009, 21, 4726–4730.
17. A. Hanafusa, Y. Muramatsu, Y. Kaburagi, A. Yoshida, Y. Hishiyama, W. Yang, J. D. Denlinger and E. M. Gullikson, J. Appl. Phys., 2011, 110, 053504.
18. W. Zhao, O. Höfert, K. Gotterbarm, J. F. Zhu, C. Papp and H. P. Steinrück, J. Phys. Chem. C, 2012, 116, 5062–5066.
19. Z. Bai, L. Zhang and L. Liu, Nanoscale, 2016, 8, 8761–8772.
20. K.-j. Kim, S. Yang, Y. Park, M. Lee, B. Kim and H. Lee, J. Phys. Chem. C, 2013, 117, 2129–2134.
21. A. Sala, G. Zamborlini, T. O. Mentes and A. Locatelli, Small, 2015, 11, 5927–5931.
22. F. Li, H. Shu, C. Hu, Z. Shi, X. Liu, P. Liang and X. Chen, ACS Appl. Mater. Interfaces, 2015, 7, 27405–27413.
23. J. Sforzini, P. Hapala, M. Franke, G. van Straaten, A. Stöhr, S. Link, S. Soubatch, P. Jelínek, T. L. Lee, U. Starke, M. Švec, F. C. Bocquet and F. S. Tautz, Phys. Rev. Lett., 2016, 116, 126805.
24. D.-W. Wang, I. R. Gentle and G. Q. Lu, Electrochem. Commun., 2010, 12, 1423–1427.
25. A. L. M. Reddy, A. Srivastava, S. R. Gowda, H. Gullapalli, M. Dubey and P. M. Ajayan, ACS Nano, 2010, 4, 6337–6342.
26. B. Xia, Y. Yan, X. Wang and X. W. Lou, Mater. Horiz., 2014, 1, 379–399.
27. T. Schiros, D. Nordlund, L. Pálová, D. Prezzi, L. Zhao, K. S. Kim, U. Wurstbauer, C. Gutiérrez, D. Delongchamp, C. Jaye, D. Fischer, H. Ogasawara, L. G. M. Pettersson, D. R. Reichman, P. Kim, M. S. Hybertsen and A. N. Pasupathy, Nano Lett., 2012, 12, 4025–4031.
28. R. J. Koch, M. Weser, W. Zhao, F. Viñes, K. Gotterbarm, S. M. Kozlov, O. Höfert, M. Ostler, C. Papp, J. Gebhardt, H. P. Steinrück, A. Görling and T. Seyller, Phys. Rev. B: Condens. Matter Mater. Phys., 2012, 86, 075401.
29. Y. Shao, S. Zhang, M. H. Engelhard, G. Li, G. Shao, Y. Wang, J. Liu, I. A. Aksay and Y. Lin, J. Mater. Chem., 2010, 20, 7491–7496.
30. H. Liu, Y. Liu and D. Zhu, J. Mater. Chem., 2011, 21, 3335–3345.
31. K. Ki-Jeong, L. Hangil, C. Junghun, L. Hankoo, J. Min Cherl, H. J. Shin, T. H. Kang, B. Kim and K. Sehun, J. Phys.: Condens. Matter, 2010, 22, 045005.
32. P. E. Blöchl, Phys. Rev. B: Condens. Matter Mater. Phys., 1994, 50, 17953–17979.
33. G. Kresse and D. Joubert, Phys. Rev. B: Condens. Matter Mater. Phys., 1999, 59, 1758–1775.
34. G. Kresse and J. Furthmüller, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 54, 11169–11186.
35. P. J. Stephens, F. J. Devlin, C. F. Chabalowski and M. J. Frisch, J. Phys. Chem., 1994, 98, 11623–11627.
36. V. A. Rassolov, M. A. Ratner, J. A. Pople, P. C. Redfern and L. A. Curtiss, J. Comput. Chem., 2001, 22, 976–984.
37. A. D. Becke, J. Chem. Phys., 1993, 98, 5648–5652.
38. C. Lee, W. Yang and R. G. Parr, Phys. Rev. B: Condens. Matter Mater. Phys., 1988, 37, 785–789.
39. S. H. Vosko, L. Wilk and M. Nusair, Can. J. Phys., 1980, 58, 1200–1211.
40. M. J. Frisch, et al., Gaussian, Wallingford, CT, 2009.
41. Y.-C. Lin, P.-Y. Teng, C.-H. Yeh, M. Koshino, P.-W. Chiu and K. Suenaga, Nano Lett., 2015, 15, 7408–7413.
42. R. Lv, Q. Li, A. R. Botello-Méndez, T. Hayashi, B. Wang, A. Berkdemir, Q. Hao, A. L. Elías, R. Cruz-Silva, H. R. Gutiérrez, Y. A. Kim, H. Muramatsu, J. Zhu, M. Endo, H. Terrones, J.-C. Charlier, M. Pan and M. Terrones, Sci. Rep., 2012, 2, 586.
43. C. Mathieu, B. Lalmi, T. O. Menteş, E. Pallecchi, A. Locatelli, S. Latil, R. Belkhou and A. Ouerghi, Phys. Rev. B: Condens. Matter Mater. Phys., 2012, 86, 035435.
44. G. Henkelman, A. Arnaldsson and H. Jónsson, Comput. Mater. Sci., 2006, 36, 354–360.
45. P. Rani and V. K. Jindal, RSC Adv., 2013, 3, 802–812.

---