Structure, Stability, Edge States and Aromaticity of Graphene Ribbons

Tobias Wassmann, Ari P. Seitsonen, A. Marco Saitta, Michele Lazzeri, and Francesco Mauri
IMPMC, Universit e Paris 6 et 7, CNRS, IPGP, 140 rue de Lourmel, 75015 Paris, France (Dated: August 17, 2008) We determine the stability, the geometry, the electronic and magnetic structure of hydrogenterminated graphene-nanoribbons edges as a function of the hydrogen content of the environment by means of density functional theory. Antiferromagnetic zigzag ribbons are stable only at extremelylow ultra-vacuum pressures. Under more standard conditions, the most stable structures are the mono- and di-hydrogenated armchair edges and a zigzag edge reconstruction with one di- and two mono-hydrogenated sites. At high hydrogen-concentration bulk graphene is not stable and spontaneously breaks to form ribbons, in analogy to the spontaneous breaking of graphene into small-width nanoribbons observed experimentally in solution. The stability and the existence of exotic edge electronic-states and/or magnetism is rationalized in terms of simple concepts from organic chemistry (Clars rule).
PACS numbers: 71.15.Mb, 71.20.Tx, 73.20.At

arXiv:0808.2283v1 [cond-mat.mtrl-sci] 17 Aug 2008

While two-dimensional graphene [1] exhibits fascinating properties such as a relativistic massless-dispersion for its charge carriers and ballistic transport on large distances, its gapless spectrum makes it unsuitable for direct application as a channel in eld-eect transistors and other semiconducting devices. This problem might be overcome, and thus open the way to a breakthrough carbon-based electronics, by designing graphene nanoribbons (GNRs) [2] in which the lateral quantum connement opens an electronic gap, function of the ribbon width. Recent works [3, 4] have shown that the use of lithographic patterning of graphene samples can yield GNRs. However, the reported ribbons have large widths, between 15 and 100 nm, small electronic gap, up to 200 meV, and are characterized by a significant egde-roughness. For practical applications, larger gaps, and correspondingly narrower and smoother-edged GNRs (width w 10 nm), are desirable. An alternative experimental route [5, 6], that makes use of chemical methods such as solution-dispersion and sonication, has shown that graphene sheets spontaneously break into ribbons of narrow width and smooth edges. Ideally, a combination of the two methods could lead to the design of GNRs having edges of controlled orientation, spatial localization, and electronic properties. In this regard, the knowledge of the structural and energetic properties of the possible edges, as well as of their thermodynamic stability, is crucial to achieve the experimental control necessary for technological applications [6]. A huge number of theoretical works has recently appeared in the literature on GNRs [7, 8, 9, 10, 11, 12, 13, 14]. In those studies, ribbons are usually chosen with the two fundamental edge geometries, zigzag and armchair, and dangling bonds are saturated with a single hydrogen per carbon atom. Zigzag GNRs have caused a great resonance in the scientic community, due to their surprising electronic and magnetic properties. In fact, zigzag GNRs feature magnetism due to spin-polarized

electronic states localized on the edge [7, 8] and they possibly turn to half-metal under high external electric elds [9, 10]. All this has red the hope to use GNRs in future spintronic-devices [9, 11], even if the robustness of the spin polarization in presence of defects has been questioned [14]. Surprisingly, no attempt has been made so far to study the energetics and thermodynamics of ribbon edges having dierent hydrogen terminations, as a function of the chemical potential typical of the experimental conditions. In this work we report density functional theory (DFT) calculations of the energetics and structure of various hydrogen-terminated graphene-nanoribbons edges as a function of the hydrogen content of the environment. We show that magnetic nanoribbons are stable only at extremely low hydrogen-concentrations, challenging to obtain experimentally. They are very reactive and thus unlikely to be stable at ambient conditions. The most stable structures at reasonable hydrogen-concentrations are non-magnetic and have a very low chemical reactivity. They are thus likely to be metastable well beyond their region of thermodynamical stability. We also show that at high hydrogen-concentrations graphene is not stable and spontaneously breaks to form hydrogenated edges. This is analogous to the spontaneous breaking of graphene into small-width nanoribbons observed experimentally in organic compounds solution [5, 6]. Finally, the stability and the existence of exotic electronic edgestates and/or magnetism are fully rationalized in terms of simple concepts from organic chemistry (Clars rule). We studied GNRs with dierent edge-terminations using periodic super-cell geometries with lattice-parameter L (periodicity) along the edge, the two edges having the same conguration. Details of the total-energy calculations are in note [15]. Zigzag edges are denoted as zn1 n2 ...nX where ni stands for the number of hydrogen atoms on a given site, and X is the number of adjacent edge-sites within the periodicity L. Armchair edges

A) z(57)00
Energy (eV)

z1 L:

a11 L:

z211 L:

a22
Energy (eV)

D)
L: a(56)2 L:

z(600)1111

L:
2 1 0 -1 -2 0 0.1 0.2 0.3 0.4 0.5 0

L:
2 1 0 -1 -2 0.1 0.2 0.3 0.4 0.5

0.1 0.2 0.3 0.4 0.5

0.1 0.2 0.3 0.4 0.5

0.1 0.2 0.3 0.4 0.5

B)
Energy (eV)
2 1 0 -1 -2 0

a00 L:

a1100 L:

C) z2111
L:

z21 L:

z(57)22
Energy (eV)

E)
=

L:
2 1 0 -1 -2 0.1 0.2 0.3 0.4 0.5

0.1 0.2 0.3 0.4 0.5 0

k (2/L)

0.1 0.2 0.3 0.4 0.5

k (2/L)

0.1 0.2 0.3 0.4 0.5 0

k (2/L)

0.1 0.2 0.3 0.4 0.5 0

k (2/L)

k (2/L)

FIG. 1: A: Scheme and electronic band-structure of the ve most stable hydrogen-terminated edges of a graphene nanoribbon. Carbon-carbon bonds are represented with the standard notation, while hydrogen atoms are the lled circles. The structures are periodic along the ribbon-edge with periodicity L. The gray area corresponds to the electronic-bands allowed in bulk graphene. The dashed line is the Fermi level. B, C: Other stable armchair and zigzag terminations. D: Further examples of possible edges. E: Standard representation of the benzenoid aromatic carbon ring.

are denoted as am1 m2 (am1 m2 m3 m4 ) for supercells containing one (two) hexagon column(s), where mi indicates the number of hydrogen atoms bonded to the ith carbon site. Some examples are in Fig. 1. Other structures which do not t this notation, z(57)ij , z(600)ijkl , and a(56)i , are also in Fig. 1. Again, the subscripts indicate the number of hydrogens on a given site. In Tab. I we report the computed zero-temperature edge formation-energy per length: EH2 = 1 2L E ribb NC E blk NH EH2 2 , (1)

where E ribb , E blk and EH2 are the total energy of the ribbon super-cell, of one atom in bulk graphene and of the isolated H2 molecule [16]. NC (NH ) are the number of carbon (hydrogen) atoms in the super-cell. EH2 can be used to determine the stability of dierent structures as a function of the experimental conditions [17]. In presence of molecular H2 gas, at a given chemical potential H2 , the relative stability is obtained by comparing GH2 = EH2 H H2 /2, where H = NH /(2L). At the absolute temperature T and for a partial H2 pressure P [18], H2 = H (T ) H (0) T S (T ) + kB T ln P P , (2)

TABLE I: Formation energy (Eq. 1) and hydrogen density (H = NH /(2L)) for all the studied edges. A) A) H ( A1 ) EH2 (eV/ H ( A1 ) EH2 (eV/ 0.000 1.4723 0.000 0.9650 a(56)0 z(57)00 0.000 1.0078 0.000 1.1452 a00 z0 0.7030 0.136 0.7854 a(56)1 0.235 z100 0.235 0.4946 0.271 0.7260 a1100 z200 0.235 0.6273 0.271 0.4306 a10 z110 0.469 0.0321 0.407 0.3337 a11 z(57)11 0.4114 0.407 0.0809 a(56)2 0.469 z1 0.704 0.2092 0.474 0.0463 a21 z211111 0.704 0.0163 0.488 0.0397 a2211 z21111 0.939 0.0710 0.508 0.0257 a22 z2111 0.542 0.0119 z211 0.0459 z(600)1111 0.542 0.610 0.0382 z21 0.678 0.1007 z221 0.813 0.2224 z2 0.813 0.2171 z(57)22 Non magnetic, metallic edges, non aromatic Magnetic, metallic edges, non aromatic Non magnetic, non-metallic edges, aromatic

where H (S ) is the enthalpy (entropy) at the pressure P = 1 bar obtained from Ref. [19]. In presence of monoatomic-hydrogen gas one should use GH = EH H H , where EH = EH2 H 2.24 eV, from DFT. The most stable structures are thus obtained by comparing GH2 or GH . G is linear in , the slope of the line being determined by H . For a given value of the stable structure is the one with the lowest value of G, thus, by increasing (i.e. going to an environment

richer in hydrogen) the favorable structures will be those with higher hydrogen-density H . This concept is visualized, as usual [17], by plotting G vs. in Fig. 2, where we report the ve most stable structures. In Fig. 3, we report the same plot for the most stable edges in the two distinct families of zigzag and armchair GNRs. Some important conclusions can be driven from Fig. 2. First, the magnetic z1 edge, widely studied as a potential component for future spintronic devices [9, 11], and the z(57)00 , proposed in [13], are stable only at extremely low hydrogen-concentrations. Second, at reasonable hydrogen pressure (at ambient conditions, the partial H2 pres-

3
100 K 300 K 600 K
1.2 10200 1070 1060 10160 1050 10120 1040 1030 1080 1020 105 1 1040 1010 100 0 100 100

1030 4

1025

1020 3

1015

1010 2

H (eV)

1.2

0.8

0.8

0.4

0.4

0.0

0.0

0.4

0.4

0.8

z1 z(57)00
4 3 10160 1060 1030 1040 2 1080 1020

z211 a11
1 0

0.8

a22
1 2 3 4 1.2

1.2 5

100 K 300 K 600 K

H2 (eV)
100 100

1020

1010

100

FIG. 2: Formation energies versus chemical potential for the ve most stable edges. Vertical lines distinguish the stability regions. The alternative bottom (top) axes show the pressure, in bar, of molecular H2 (atomic H) gas corresponding to the chemical potentials at T = 100, 300, and 600 K.
1.2 0.8

GH (eV/)

0.4 0.0 0.4 0.8 1.2 1.2 0.8

a00 a1100

a11

a22

GH (eV/)

0.4 0.0 0.4 0.8

z(57)22

z(57)00
4

z1
3 2

z2111
1

z211
0

z21
1 2 3 4

1.2 5

H2 (eV)

FIG. 3: Formation energies versus chemical potential for the most stable armchair (top) and zigzag (bottom) edges.

sure in air is 5107 bar) the most stable edges are a11 , z211 , and a22 . Finally, Fig. 2 shows that graphene is not stable at high hydrogen-concentration (G becomes negative) and spontaneously breaks to form hydrogenated edges. Similarly, it has been experimentally observed

that small-width nanoribbons spontaneously form in organic compounds solutions [5, 6]. We indicate in Tab. I, along with the energetics, whether the proposed edges are magnetic or not, i.e. whether a nite magnetization density is observed within DFT calculations. In Fig. 2, we also report the electronic bands of the ribbons and we show the bands of two-dimensional bulk graphene projected on the onedimensional Brillouin zone of the edge (gray area in Fig. 2). As currently done in surface physics, we call edge states (ESs) those electronic states localized at edges, having an energy which is not allowed in the bulk (i.e. those bands which are outside the gray area in Fig. 2). Such states are localized near the edge. Moreover, in Tab. I we indicate as metallic edges, those having partially-lled ESs, i.e. having ESs which, in absence of magnetic order, cross the Fermi level, and are thus only partially occupied. For a system with partially lled ESs, in a ribbon, it might be energetically convenient to split those bands by inducing a magnetic order between the two ribbon-edges and thus opening an energy gap, as in the well-known case of the z1 edge. Finally, we remind that the chemical reactivity of a given system can be quantitatively expressed as proportional to the electronic density of states near the Fermi level. As a consequence, all the structures having edge-states crossing the Fermi energy or in its immediate vicinity (independently on whether they are magnetic or not) are expected to be extremely reactive, since the edge-states are available to form chemical bonds. With these concepts in mind, we remark that the most stable structures at standard and high hydrogenconcentrations, namely a11 , z211 , and a22 , have no edge states, in agreement with the experimental observation of large semiconducting gaps in narrow ribbons [6]. This implies that, besides being non-magnetic, they have a very low chemical reactivity. As a consequence, once formed, those edges are likely to be metastable well beyond their respective windows of thermodynamical stability in Fig. 2. On the other hand, magnetic edges, as the well studied z1 , are stable only at extremely low hydrogen-concentrations and are also very reactive and thus unlikely to be stable at ambient conditions. We now show that the electronic structure of the edges can be interpreted in terms of aromaticity and according to Clars rule, well known in organic chemistry since the 60s [20, 21]. The fundamental stability-criterion for hydrocarbons is that each carbon atom must have four saturated bonds. In organic chemistry, the so-called benzenoid aromatic ring can be dened as a resonance (due to the delocalization of electrons over the ring [22, 23]) between 2 hexagonal rings with alternating single and double bonds as in Fig. 2-E. For such a structure, all the bonds sticking out of the hexagon are single bonds and, as a consequence, two benzenoid rings cannot be adjacent. The energetic stability of a structure increases

GH (eV/)

GH (eV/)

4 with the number of possible resonances. As a consequence, since a structure containing n benzenoid rings displays 2n possible resonances, the most stable isomer of a given hydrocarbon is the one that maximizes the number of benzenoid rings (Clars rule [20, 21]). Bulk graphene is the ideal case of aromaticity since there are three equivalent representations in which all the electrons belong to a benzenoid ring and no double-bonds are present. In such a representation 1/3 of the graphene carbon-hexagons are benzenoid rings. The edge structures of Fig. 1 are displayed in a representation that maximizes the number of benzenoid rings. However, the resulting ring-density for some of the structures (e.g. the z21 or z2111 ) is lower than the ideal graphene 1/3 density. We analyzed all the edges of Tab. I and we indicate as aromatic those structures that have a density of benzenoid rings of 1/3 (e.g. the a11 , z211 and a22 ). We note that all the aromatic structures do not have partially-lled edge-states (are non-metallic) and viceversa. This empirical rule is valid for all the structures presently studied and can be understood in terms of the Clars rule. In fact, whenever the ring-density near the edge is smaller than 1/3, there is a competition between the bulk, where aromaticity prevails with an exponential (2n ) weight, and the edge. It is precisely this competition that, imposing the ring density of the bulk, forces some edge carbon-atoms to have less or more than 4 saturated bonds and, thus, originates the electronic defects, i.e. the edge states. The z211 and the z(600)1111 ribbons are the only zigzag edges compatible with the aromaticity of the bulk. These edge congurations have in fact the same 1/3 periodicity along the zigzag edge as the benzenoid rings of bulk graphene. Moreover, monoand di-hydrogenated armchair edges are both compatible with the 1/3 aromaticity of graphene, whereas asymmetrically passivated edges, such as a21 , are not. Notice that, for a given H density, the most stable edges are aromatic, if an aromatic structure is compatible with such a concentration. The stability of fully-aromatic edges is conrmed by the fact that polycyclic aromatic hydrocarbons whose structure can be fully represented by benzenoid rings (without double bonds) show particularly high stability, high melting points, low reactivity [22], and have, whenever possible, a11 or z(600)1111 edges [23, 24]. Concluding, the knowledge of the structure and stability of the possible edges is a crucial issue to control the experimental conditions of the formation of graphene nanoribbons of desired properties. Here, we determined the structure and stability of hydrogen-terminated nanoribbons. For reasonable hydrogen concentrations the most stable structures are the a11 , a22 and z211 , which are not magnetic and do not present valence electronic states localized at the edge. These structures are, thus, expected to be non-reactive and metastable well beyond the region of their thermodynamical stability. In particular, at high hydrogen concentrations graphene spontaneously breaks into a22 nanoribbons. Our results are rationalized by means of simple concepts of organic chemistry that can be used to guide the search for the most stable edges in dierent chemical environments. We thank Th. Greber for discussions. Calculations were done at IDRIS (project n 081202).

[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15]

[16] [17] [18] [19] [20] [21] [22] [23] [24] [25]

K. S. Novoselov et al., Science 306, 666 (2004). K. Nakada et al., Phys. Rev. B 54, 17954 (1996). M. Y. Han et al., Phys. Rev. Lett. 98, 206805 (2007). Z. H. Chen et al., Physica E 40, 228 (2007). X. Li et al., Science 319, 1229 (2008). X. Wang et al., Phys. Rev. Lett. 100, 206803 (2008). A. Yamashiro et al., Phys. Rev. B 68, 193410 (2003). L. Pisani et al., Phys. Rev. B 75, 064418 (2007). Y. W. Son, M. L. Cohen, and S. G. Louie, Nature 444, 347 (2006). E. J. Kan et al., Appl. Phys. Lett. 91, 243116 (2007). O. V. Yazyev and M. I. Katsnelson, Phys. Rev. Lett. 100, 047209 (2008). S. Okada, Phys. Rev. B 77, 041408 (2008). P. Koskinen, S. Malola, and H. Hakkinen (2008), arXiv:0802.2623v1. B. Huang et al., Phys. Rev. B 77, 153411 (2008). Ab-initio calculations are done with the PWSCF code (S. Baroni et al., www.quantum-espresso.org) within planewave/ultrasoft-pseudopotential approach and exchangecorrelation from Ref. [25]. We used a kinetic energy/charge cut-o of 30/300 Ry. Edges are simulated within a super-cell geometry using a vacuum layer of 9.5 A between two edges and of 8.5 A between two graphene planes. L is xed according to the latticeconstant of graphene (2.46 A) and atomic positions are allowed to fully relax. For armchair edges, the width of the ribbons vary from 21.5 A for a(56)0 to 24.1 A for a11 . For zigzag, the widths vary from 28.81 A for z(57)00 to 30.72 A for z1 . Electronic integrations are done using grids (along the periodic direction of the ribbon) of 12 kpoints for armchair ribbons and 24, 12, 8, 6, 4, 4 k-points for the zigzag ones with 1, 2, 3, 4, 5, 6 periodicity. We checked that the inclusion of zero point motion (not considered here for sake of clarity) does not aect the relative stability of dierent edge-structures. G. X. Quian, R. M. Martin, and D. J. Chadi, Phys. Rev. B 38, 7649 (1988). C. Bailey et al. (2007), URL http://epubs.cclrc.ac.uk/work-details?w=40553 . M. W. Chase Jr. et al., JANAF Thermochamical tables Third Edition, J. Phys. Chem. Ref. Data 14 (1985). E. Clar, Polycyclic Hydrogarbons (Academic Press, London, 1964). E. Clar, The Aromatic Sextet (Wiley, New York, 1972). J. Wu, W. Pisula, and K. Mullen, Chem. Rev. 107, 718 (2007). M. Watson, A. Fechtenkotter, and K. Mullen, Chem. Rev. 101, 1267 (2001). M. M uller, C. K ubel, and K. M ullen, Chem. Eur. J. 4, 2099 (1998). J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77, 3865 (1996).