1: % INJ

2: 

3: %\documentclass[aps,pra,superscriptaddress,showpacs,twocolumn]{revtex4}

4: \documentclass[12pt]{elsart}

5: %\documentclass[aps,prb,superscriptaddress,showpacs,twocolumn]{revtex4}

6: %\documentclass[aps,prb,superscriptaddress,showpacs,preprint]{revtex4}

7: %\usepackage[draft]{graphicx}

8: \usepackage{graphicx}

9: \usepackage{amssymb}

10: \usepackage{amsmath}

11: 

12: \newcommand{\bk}{{\bf k}}

13: \newcommand{\bq}{{\bf q}}

14: \newcommand{\bQ}{{\bf Q}}

15: \newcommand{\bp}{{\bf p}}

16: \newcommand{\bx}{{\bf x}}

17: \newcommand{\br}{{\bf r}}

18: \newcommand{\Li}{{\mathop{\rm{Li}}\nolimits}}

19: \renewcommand{\Im}{{\mathop{\rm{Im}}\nolimits\,}}

20: \renewcommand{\Re}{{\mathop{\rm{Re}}\nolimits\,}}

21: \newcommand{\sgn}{{\mathop{\rm{sgn}}\nolimits\,}}

22: \newcommand{\Tr}{{\mathop{\rm{Tr}}\nolimits\,}}

23: \newcommand{\EF}{E_{\mathrm{F}}}

24: \newcommand{\kB}{k_{\mathrm{B}}}

25: \newcommand{\kF}{k_{\mathrm{F}}}

26: \newcommand{\nF}{n_{\mathrm{F}}}

27: \newcommand{\Green}{{G}}

28: \newcommand{\dSC}{{\mathrm{dSC}}}

29: \newcommand{\dPG}{{\mathrm{dPG}}}

30: \newcommand{\dDW}{{\mathrm{dDW}}}

31: \newcommand{\Ret}{{\mathrm{R}}}

32: \newcommand{\Tau}{T_\tau}

33: \newcommand{\HF}{{\mathrm{HF}}}

34: \newcommand{\TF}{{\mathrm{TF}}}

35: \newcommand{\vW}{{\mathrm{vW}}}

36: 

37: \begin{document}

38: \renewcommand{\thefootnote}{\fnsymbol{footnote}} 

39: \renewcommand{\theequation}{\arabic{section}.\arabic{equation}}

40: 

41: \journal{Phys. Lett. A}

42: 

43: \begin{frontmatter}

44: \title{Modeling vacancies and hydrogen impurities in graphene: A molecular

45: point of view}

46: 

47: \author[CTChem]{G. Forte}

48: \author[CTChem]{A. Grassi}

49: \author[CTChem]{G. M. Lombardo}

50: \author[IMM]{A. La Magna}

51: \author[CTPhys,INFN,CNISM]{G. G. N. Angilella\thanksref{corr}}

52: \author[CTPhys,CNISM]{R. Pucci}

53: \author[CTPhys]{R. Vilardi}

54: 

55: \address[CTChem]{Dipartimento di Scienze Chimiche, Facolt\`a di Farmacia,

56:    Universit\`a di Catania,\\

57: Viale A. Doria, 6, I-95126 Catania, Italy}

58: \address[IMM]{IMM, CNR, Catania, Italy}

59: 

60: \address[CTPhys]{Dipartimento di Fisica e Astronomia, Universit\`a di

61:    Catania,\\ Via S. Sofia, 64, I-95123 Catania, Italy}

62: \address[INFN]{INFN, Sez. Catania, Italy}

63: \address[CNISM]{CNISM, UdR di Catania, Italy}

64: \thanks[corr]{Corresponding author. E-mail: {\tt giuseppe.angilella@ct.infn.it}.}

65: \date{\today}

66: 

67: 

68: \begin{abstract}

69: We have followed a `molecular' approach to study impurity effects in graphene. 

70: This is thought as the limiting case of an infinitely large cluster of benzene

71: rings. Therefore, we study several carbon clusters, with increasing size, from

72: phenalene, including three benzene rings, up to coronene~61, with 61 benzene

73: rings. The impurities considered were a chemisorbed H atom, a vacancy, and a

74: substitutional proton. We performed HF and UHF calculations using the STO-3G

75: basis set. With increasing cluster size in the absence of impurities, we find a

76: decreasing energy gap, here defined as the HOMO-LUMO difference. In the case of

77: H chemisorption or a vacancy, the gap does not decrease appreciably, whereas it

78: is substantially reduced in the case of a substitutional proton. The presence of

79: an impurity invariably induces an increase of the density of states near the

80: HOMO level. We find a zero mode only in the case of a substitutional proton. In

81: agreement with experiments, we find that both the chemisorbed H, the

82: substitutional proton, and the C atom near a vacancy acquire a magnetic moment.

83: The relevance of graphene clusters for the design of novel electronic devices is

84: also discussed.

85: 

86: PACS:

87: 36.40.Cg, %Electronic and magnetic properties of clusters

88: 73.22.-f, %Electronic structure of nanoscale materials: clusters, nanoparticles, nanotubes, and nanocrystals 

89: 73.20.Hb %Impurity and defect levels; energy states of adsorbed species

90: %36.40.Mr, % Spectroscopy and geometrical structure of clusters

91: %31.25.Qm, % Electron correlation calculations for polyatomic molecules

92: %33.20.Tp % Vibrational analysis

93: \end{abstract} 

94: 

95: \end{frontmatter}

96: 

97: \section{Introduction}

98: 

99: Since the pioneering works of Wallace \cite{Wallace:47}, Coulson

100: \cite{Coulson:52a}, and Semenoff \cite{Semenoff:54}, several authors have

101: studied the peculiar electronic and structural properties of graphene, a single

102: sheet of graphite, originally only considered as a prototype of the graphite's

103: surface. A breakthrough in these studies arrived with the experimental

104: realization of graphene \cite{Novoselov:04}. The increasing interest in this

105: material stems both from the analogy of its behavior with many phenomena of

106: quantum electrodynamics (QED) \cite{Katsnelson:07a}, and because of its

107: prospective applications in the fabrication of novel electronic devices. To this

108: aim, it is important to have systems with a tunable band gap. We will stress

109: that the graphene clusters (GC) here considered, even if they are characterized

110: by a zig-zag boundary, have a semiconducting behavior, at variance with zig-zag

111: graphene nanoribbons (ZGNR), which have a metallic behavior \cite{Yan:07}. The

112: gap strongly depends on the cluster size, so that it will be highly desirable to

113: have method allowing to control the size of graphene quantum dots.

114: 

115: Graphene is a robust material. However, defects, such as ripples, adatoms,

116: vacancies and charges induced by the substrate, can modify its electronic

117: properties \cite{CastroNeto:08}. In the past, the study of vacancies and

118: impurities in carbon-based systems has been important for several aspects.

119: 

120: (i) Graphene nanostructures are believed to be attractive structures for

121: hydrogen storage. Preliminary results in this direction have been reported in

122: the case of single-walled carbon nanotubes (SWCNT) \cite{Dillon:97}.

123: 

124: (ii) It has been found \cite{Esquinazi:03} that proton irradiation of highly

125: oriented pyrolitic graphite samples triggers ferro- or ferrimagnetism. It is now

126: commonly accepted that not only vacancies, but also hydrogen chemisorbed on

127: graphene can acquire a magnetic moment.

128: 

129: (iii) From an astrophysical point of view, it would be of interest to understand

130: why H$_2$ is the most abundant molecule in the interstellar medium (ISM),

131: despite the continuous dissociation of molecular hydrogen by UV radiation and

132: cosmic rays \cite{Hornekaer:06}.

133: 

134: Of course, there can be relevant differences in the characterization of defects

135: in graphite, nanotubes, or graphene. For instance, if one considers the

136: screening of a charged impurity, Cheianov and Fal'ko \cite{Cheianov:06} have

137: shown that the electron density $\delta\rho$ in graphene decays as

138: $\delta\rho(r) \sim r^{-3}$ at long distances $r$ from the impurity. Such a

139: behavior should be contrasted with that derived by Lau and Kohn \cite{Lau:78}

140: for a two-dimensional (2D) electron gas, where $\delta\rho(r) \sim r^{-2}$. The

141: reason of such a difference is due to the peculiar chiral properties of the

142: electrons in graphene. It has been suggested \cite{Bena:07} that such a

143: difference can be used to distinguish between monolayer and bilayer graphene in

144: Fourier transformed scanning tunneling spectroscopy (FTSTS). Actually, impurity

145: induced quantum interferences have been used in scanning tunneling microscopy

146: (STM) \cite{Mallet:07} to identify experimentally mono- or bilayer graphene.

147: Previously, STM was used \cite{Ruffieux:00,Ruffieux:05} to study chemisorption

148: of hydrogen on the basal plane of graphite and atomic vacancy formation.

149: 

150: From a theoretical point of view, the effects on the electronic structure of

151: graphene due to the presence of vacancies, local impurities, and substitutional

152: impurities have been studied by Pereira \emph{et al.} \cite{Pereira:08}, by

153: using a tight-binding Hamiltonian with a local potential $U$. These authors have

154: found general trends in the low-energy spectrum of graphene connected with

155: localized zero modes, resonances and gaps induced by the above mentioned

156: defects. However, in some cases, it is important to have a more detailed

157: description of the local interaction of the defects with graphene. For this

158: reason, in the present work, we will make recourse to molecular models,

159: which in the past have been used to study the chemisorption of hydrogen on

160: graphite surface \cite{Klose:92,Fromherz:93}. According to these calculations,

161: the graphite surface is modeled as a cluster of benzene rings, whose external

162: dangling bonds are saturated by hydrogen atoms. The most frequently used model

163: clusters are polycyclic aromatic hydrocarbons (PAH), and particularly coronene

164: (C$_{24}$H$_{12}$). More recently, this model cluster has been used by

165: Patchkovskii \emph{et al.} \cite{Patchkovskii:05} to study the interaction of

166: molecular hydrogen with graphene nanostructures. These authors have performed

167: calculations at the level of second-order M\o{}ller-Plesset (MP2) perturbation

168: theory \cite{Moeller:34}.

169: 

170: In the present work, we will use such model clusters to simulate graphene, at

171: the level of \emph{ab initio} Hartree-Fock (HF) and unrestricted \emph{ab

172: initio} Hartree-Fock (UHF) theory. In some cases, also MP2 calculations will be

173: presented, but the main emphasis will be more on the trends obtained by

174: increasing the cluster size, than in the study of correlation effects. Actually,

175: Martin \emph{et al.} \cite{Martin:08} have argued that correlation effects are

176: small in graphene. A similar result was found by Siringo \cite{Siringo:84} in

177: the study of H chemisorbed on graphene \cite{note:Siringo}. Furthermore,

178: recently the Manchester group has found that the electron energy levels in

179: quantum dots in graphene follow a statistics characteristic of `Dirac

180: billiards', rather than a Poisson distribution, as is expected for an infinite

181: system \cite{Ponomarenko:08}: therefore, quantization effects are quite

182: important. It is interesting, in this context, to derive these results also

183: within a molecular approach. In the present work, we have used clusters up to 61

184: benzene rings, and we have considered the cases of a single vacancy, of a

185: chemisorbed H atom, and of a substitutional proton.

186: 

187: The paper is organized as follows. In Sec.~\ref{sec:models}, we describe the

188: models used. In Sec.~\ref{sec:results} we present our results and a comparison

189: thereof with density functional theory (DFT) and other types of calculations.

190: Finally, in Sec.~\ref{sec:conclusions} we summarize and draw our conclusions.

191: 

192: \section{The model clusters}

193: \label{sec:models}

194: 

195: \begin{figure}[t]

196: \centering

197: \includegraphics[bb=253 21 609 343,clip,width=0.9\columnwidth]{CORO_PLUS_2.ps}

198: \caption{(Color online.) Showing coronene, C$_{24}$H$_{12}$ (gray rods),

199: coronene~19, C$_{54}$H$_{18}$ (gray and blue rods), and coronene~37,

200: C$_{96}$H$_{24}$ (gray, blue, and red rods).}

201: \label{fig:coronene2plus}

202: \end{figure}

203: 

204: The molecular clusters which we have considered are: (i) phenalene (phenalenyl

205: radical); (ii) pyrene; (iii) coronene; (iv) coronene~19, \emph{viz.} coronene

206: surrounded by another series of benzene rings, for a total of 19 benzene rings;

207: (v) coronene~37, \emph{viz.} coronene surrounded by two more series of benzene

208: rings, for a total of 37 benzene rings; (vi) coronene~61. Coronene~37 is shown

209: in Fig.~\ref{fig:coronene2plus}. We believe that the larger the cluster, the

210: better it models graphene. For this reason, we have chosen to place in the

211: middle of the cluster: (a) a chemisorbed H atom; (b) a carbon vacancy; (c) a

212: proton in the substitutional position of a carbon atom.

213: 

214: For clusters (i)--(iii) we have used HF calculations with a STO-3G basis set and

215: HF calculations with a 6-311G basis set \cite{Frisch:04}. In these cases we have

216: also performed MP2 calculations. All the remaining larger clusters have been

217: treated at the HF level using a STO-3G basis set. In the cases of H

218: chemisorption, carbon vacancy and proton substitution, we have performed UHF

219: calculations, because in all these cases there is one $\pi$-electron missing.

220: The cluster geometry has been optimized in all cases.

221: 

222: \section{Results}

223: \label{sec:results}

224: 

225: \begin{figure}[t]

226: \centering

227: %\includegraphics[width=0.9\columnwidth]{ENERGY_BASI.ps}

228: \includegraphics[width=0.9\columnwidth]{NEW_FIG_2.ps}

229: \caption{(Color online.) Showing the dependence of the cluster energy as a

230: function of the cluster under consideration, for the various methods (HF or MP2)

231: and basis sets (STO-3G or 6-311G) employed in the present work. On this scale,

232: the results of different approximations are almost indistinguishable.}

233: \label{fig:levels}

234: \end{figure}

235: 

236: \begin{figure}[t]

237: \centering

238: %\includegraphics[width=0.9\columnwidth]{FIT_named.ps}

239: \includegraphics[width=0.9\columnwidth]{NEW_FIG_3.ps}

240: %\includegraphics[width=0.9\columnwidth]{GAP_BASI.ps}

241: %\caption{(Color online.) Showing the dependence of the HOMO-LUMO difference (gap

242: %energy) \emph{vs} the various clusters under consideration, for the various methods (HF or MP2)

243: %and basis sets (STO-3G or 6-311) employed in the present work.}

244: \caption{(Color online.) Showing the dependence of the HOMO-LUMO difference (gap

245: energy) \emph{vs} the number of carbon cycles for the various clusters studied

246: in the present work. The dashed line is a guide to the eye.}

247: \label{fig:HOMO-LUMO}

248: \end{figure}

249: 

250: \subsection{Pure graphene clusters}

251: \label{sec:puregraphene}

252: 

253: First of all, we note that the total energy as a function of the number of

254: benzene rings in the clusters (cf. Fig.~\ref{fig:levels}) depends weakly

255: (\emph{i.e.} to within 2~\%) on the basis set or on the specific approximation

256: employed (HF, MP2). This lends support to our conviction that relevant

257: conclusions can be drawn already for one of the largest cluster considered,

258: \emph{i.e.} coronene~37, within the framework of the HF approximation with the

259: STO-3G basis set.

260: 

261: Secondly, we note that the gap, which in these kind of calculations is described

262: by the HOMO-LUMO difference, is decreasing by increasing the size of the

263: cluster, as shown in Fig.~\ref{fig:HOMO-LUMO}. This is in agreement with the

264: consideration that our clusters resemble more closely graphene, as the number of

265: benzene rings increases. The smallest gap which we have obtained is 5.48~eV.

266: 

267: Extrapolation of our data seems to indicate that the gap would close at a

268: cluster size of $\gtrsim 1000$~benzene rings. This is in qualitative agreement

269: with the experimental finding that graphene quantum dots as large as $\sim

270: 30$~nm are characterized by a gap of $\sim 0.5$~eV \cite{Ponomarenko:08}. We

271: find a C--C nearest neighbours equilibrium distance $d_{\mathrm{C-C}} =

272: 1.42$~\AA, which is stable with respect to the different approximations and

273: cluster sizes considered here. This in striking agreement with the typical bond

274: length for $sp^2$ C--C bonds for graphene and graphite, which is 1.42~\AA.

275: 

276: Our minimization procedure yields for the clusters considered in this case a

277: planar configuration, at variance with the experimental finding for graphene,

278: showing the presence of ripples \cite{Stolyarova:07,Meyer:07}. This discrepancy

279: may be related to the specific boundary conditions used in our calculations.

280: Such a result is not in contradiction with the known fact that a nonzero amount

281: of curvature is required in order to stabilize a two-dimensional crystal, as a

282: consequence of Mermin-Wagner theorem \cite{Peierls:34,Mermin:66,Mermin:68}.

283: Indeed, the systems we are considering here are not truly infinite lattices, but

284: rather finite-size clusters. Furthermore, we note that Choi \emph{et al.}

285: \cite{Choi:08}, by using a cluster with a monovacancy and 127~C atoms, and

286: performing spin-polarized density functional theory calculations, have found an

287: almost planar configuration.

288: 

289: 

290: \begin{figure}[t]

291: \centering

292: \includegraphics[width=0.9\columnwidth]{coronene_37_DOS.ps}\\

293: \includegraphics[width=0.9\columnwidth]{coronene_37_DOS_2.ps}

294: \caption{(Color online.) Showing the DOS for pure coronene~37 (upper panel), and

295: for coronene~37 with a chemisorbed H atom (lower panel), as a function of

296: energy.}

297: \label{fig:DOS}

298: \end{figure}

299: 

300: \begin{figure}[t]

301: \centering

302: \includegraphics[width=0.8\columnwidth]{coronene_37_PS_S_15eV.ps}\\

303: \includegraphics[width=0.8\columnwidth]{coronene_37_PS_S_10eV.ps}\\

304: \includegraphics[width=0.8\columnwidth]{coronene_37_PS_S_5eV.ps}

305: \caption{(Color online.) Showing the level-spacing distribution $P(s)$ for

306: coronene~37, corresponding to different energy intervals centered around zero:

307: 15~eV (topmost panel), 10~eV (intermediate panel), 5~eV (bottom panel). It can

308: be noted that $P(s)$ looses the Poissonian behavior in favor of a Gaussian

309: behavior as the energy interval decreases.}

310: \label{fig:Ps}

311: \end{figure}

312: 

313: 

314: In Fig.~\ref{fig:DOS} we report the density of states (DOS) as a function of

315: energy (eV) for coronene~37. We note that (i) a nearly straight line joins the

316: peaks near the HOMO energy: such a trend mimics the linear dependence which is

317: found in the DOS near the Fermi level in the infinite system

318: \cite{CastroNeto:08}; (ii) particle-hole symmetry is absent, as is obtained in

319: tight-binding calculations with a nonzero next-nearest hopping parameter

320: $t^\prime$ \cite{CastroNeto:08}. However, here this asymmetry is not very

321: pronounced than in the case $t^\prime \neq 0$. This is probably due to the fact

322: that in our calculations we do not restrict the interaction to next-nearest

323: neighbors. We have also considered the level-spacing distribution $P(s)$, where

324: the dimensionless spacing $s$ is defined as the energy difference $\Delta E_i =

325: E_i - E_{i-1}$ between successive levels, divided by the average $\langle \Delta

326: E_i \rangle$ of the energy difference between successive levels

327: \cite{DeRaedt:08}. The quantity $P(s)$ yields the number of energy differences

328: for which $s-\Delta/2 < \Delta E_i / \langle \Delta E_i \rangle \leq s +

329: \Delta/2$, where $\Delta$ is the step of the histogram. We have found that

330: $P(s)$ strongly depends on the energy range considered around $E=0$ (cf.

331: Fig.~\ref{fig:Ps}). One can also recognize a strong dependence on the cluster

332: size, so that no definite conclusion can be achieved at the present status of

333: the calculations.

334: 

335: \subsection{Clusters with a chemisorbed H atom}

336: \label{sec:chemisorbed}

337: 

338: \begin{figure}[t]

339: \centering

340: \includegraphics[bb=152 20 705 341,clip,width=0.9\columnwidth]{CORO_PLUS_HF_3G_H_TOP_B_S.ps}

341: \caption{(Color online.) Showing coronene~37 with a chemisorbed H atom.}

342: \label{fig:largestHatom}

343: \end{figure}

344: 

345: 

346: 

347: The largest cluster considered in the presence of a chemisorbed H atom is

348: reported in Fig.~\ref{fig:largestHatom}, where the H adatom is depicted with a

349: yellow color. 

350: %We have indicated with A and B the carbon atoms belonging to the

351: %two different sublattices. 

352: The total energy, as a function of the cluster size

353: and the approximation used, is quite similar to that reported in

354: Fig.~\ref{fig:levels}, and will not be shown here. 

355: 

356: We find an average equilibrium distance of 1.53~\AA{} between two in-plane

357: nearest neighbor C atoms, forming an angle $\angle$HCC$=105.3^\circ$ with the H

358: adatom, while the C atom, on top of which the H atom resides, emerges from the

359: plane of 0.71~\AA. The distance of the H atom from the underneath C atom is

360: 1.10~\AA. The distance of the C atoms surrounding the adatom is of 0.30~\AA{}

361: with respect to the planar configuration. These values are in qualitative

362: agreement with recent DFT calculations \cite{Boukhvalov:08}. We find that (i)

363: the most stable position is that with H on top of a carbon atom; (ii) the

364: underlying C atom is induced out of the plane; (iii) there are small differences

365: between the $\alpha$ and $\beta$ sets of eigenvalues positions. We will

366: therefore report only the results for the $\alpha$ set.

367: 

368: An important difference found in the presence of a chemisorbed H atom, with

369: respect to the pure graphene clusters discussed in Sec.~\ref{sec:puregraphene},

370: is the trend of the gap as a function of the number of the benzene rings in the

371: cluster, which is reported in Fig.~\ref{fig:rings}. No evidence of a gap

372: decrease appears in this case, which supports the results of Pereira \emph{et

373: al.} \cite{Pereira:08} that hydrogen chemisorption on graphene induces the

374: opening of a gap. Also in agreement with such findings \cite{Pereira:08} is the

375: behavior of the DOS, which is reported a dashed blue line in Fig.~\ref{fig:DOS}.

376: One can see a clear enhancement of the DOS near the HOMO level, which Pereira

377: \emph{et al.} \cite{Pereira:08} have found as due to a resonance induced by the

378: chemisorbed hydrogen. Of course, in this case we have used UHF calculations,

379: which allowed us to estimate also the magnetic moment acquired by the H atom,

380: which for a corenene~37 cluster is $-0.33$~$\mu_{\mathrm{B}}$.

381: 

382: The above mentioned results, in the framework here considered, do not depend

383: strongly on the cluster size considered, as reported in

384: Tab.~\ref{tab:moment}, where a comparison is presented between the results

385: obtained with the coronene~19 and the coronene~37 clusters, by using a STO-3G

386: basis set.

387: 

388: \begin{figure}[t]

389: \centering

390: %\includegraphics[width=0.9\columnwidth]{GAP_BASI_H_TOP.ps}

391: \includegraphics[width=0.9\columnwidth]{NEW_FIG_7.ps}

392: \caption{(Color online.) Showing the dependence of the HOMO-LUMO difference (gap

393: energy) \emph{vs} the various clusters considered, in the presence

394: of a chemisorbed H atom, for the different methods and basis sets employed in

395: this work.}

396: \label{fig:rings}

397: \end{figure}

398: 

399: \begin{table}[t]

400: \caption{Showing values of the total energy $E$ (Hartree), HOMO energy

401: (Hartree), LUMO energy (Hartree), HOMO-LUMO gap (eV), and magnetic moment $\mu$

402: ($\mu_{\mathrm{B}}$), for the coronene~19 and coronene~37 clusters, and the same

403: clusters with a H adatom on top, with a vacancy, with substitutional proton, and

404: with a substitutional H atom, using a STO-3G basis set.}

405: \label{tab:moment}

406: \begin{tabular}{lrrrrr}

407: \hline

408:  & $E$ & HOMO & LUMO & gap & $\mu$ \\

409:  \hline

410: \multicolumn{6}{l}{coronene~19} \\

411: pure & $-2030.796$ & $-0.159$ & $0.128$ & $7.810$ & --- \\

412: with H top & $-2031.701$ & $-0.252$ & $0.243$ & $13.465$ &

413: $-0.333$\\

414: with vacancy & $-1993.461$ & $-0.262$ & $0.216$ & $13.007$ & $1.838$

415: \\

416: with proton & $-1993.834$ & $-0.335$ & $-0.120$ & $5.850$ & $0.917$

417: \\

418: with H subst. & $-1994.058$ & $-0.263$ & $0.212$ & $12.925$ & $0.849$ \\

419: \hline

420: \multicolumn{6}{l}{coronene~37} \\

421: pure & $-3605.818$ & $-0.137$ & $0.101$ & $6.476$ & --- \\

422: with H top & $-3607.059$ & $-0.251$ & $0.240$ & $13.361$ &

423: $-0.334$\\

424: with vacancy & $-3568.808$ & $-0.259$ & $0.215$ & $12.898$ & $1.838$

425: \\

426: with proton & $-3569.186$ & $-0.319$ & $-0.114$ & $5.578$ & $0.833$ \\

427: with H subst. & $-3569.401$ & $-0.261$ & $0.212$ & $12.871$ & $0.851$ \\

428: \hline

429: \end{tabular}

430: \end{table}

431: 

432: 

433: 

434: \subsection{Clusters with a vacancy}

435: \label{sec:vacancy}

436: 

437: \begin{figure}[t]

438: \centering

439: \includegraphics[width=0.9\columnwidth]{coro_plus_2_VACANCY_stick.ps}

440: \caption{(Color online.) Showing coronene~37 in the presence of a vacancy. Number

441: 68 indicates the position of the C atom around the vacancy referred to in the

442: main text.}

443: \label{fig:vacancy}

444: \end{figure}

445: 

446: 

447: \begin{figure}[t]

448: \centering

449: \includegraphics[width=0.9\columnwidth]{coronene_37_H_DOS.ps}\\

450: \includegraphics[width=0.9\columnwidth]{coronene_37_PROTON_DOS.ps}

451: \caption{Upper panel: Showing DOS of coronene~37 with a vacancy.

452: Lower panel: Showing DOS of coronene~37 with a substitutional proton.}

453: \label{fig:DOSvac}

454: \end{figure}

455: 

456: 

457: In Fig.~\ref{fig:vacancy} we report the clusters considered in the presence of a

458: vacancy. The same considerations made above for the total energy per number of

459: benzene rings in Sec.~\ref{sec:puregraphene} and \ref{sec:chemisorbed} above

460: apply to this case as well. The most stable electronic configuration is the

461: triplet state, and the gap has a similar trend and similar values as those

462: obtained for H chemisorption (cf. Tab.~\ref{tab:moment}). From

463: Fig.~\ref{fig:DOSvac}, one can see that also in this case the DOS presents an

464: increase near the HOMO level with respect to the case in the absence of the

465: vacancy. The main difference with respect to the case with H on top, is relative

466: to the magnetic moment acquired by a carbon atom near the vacancy (see atom

467: labelled C$^{68}$ for coronene~37 in Fig.~\ref{fig:vacancy}). The value obtained

468: in this case is $\approx 1.84$~$\mu_{\mathrm{B}}$, which presents just small

469: changes in going from coronene~19 to coronene~37, or by using the $\alpha$ or

470: $\beta$ sets of eigenvalues. The fact that the magnetic moment is smaller when

471: we consider H chemisorption instead of a vacancy is in qualitative agreement

472: with the results of Yaziev and Helm \cite{Yaziev:07}. These authors have found

473: $1$~$\mu_{\mathrm{B}}$ per hydrogen chemisorption defect, and

474: $1.12-1.53$~$\mu_{\mathrm{B}}$ per vacancy defect, depending on the defect

475: concentration. A value of $1.52$~$\mu_{\mathrm{B}}$ for a monovacancy is

476: reported by Choi \emph{et al.} \cite{Choi:08}. From this comparison, we can

477: deduce that the magnetic moments are sensitive to the model calculations

478: employed, but our results are in overall agreement with the most recent works.

479: 

480: Fig.~\ref{fig:vacdist} and Tab.~\ref{tab:vacdist} report the distances between

481: the various C~atoms around the vacancy.

482: 

483: 

484: \begin{figure}[t]

485: \centering

486: \includegraphics[width=0.9\columnwidth]{VACANCY_DISTANCE.ps}

487: \caption{Showing C atoms around the vacancy. The distances between the various

488: atoms are reported in Tab.~\ref{tab:vacdist}.}

489: \label{fig:vacdist}

490: \end{figure}

491: 

492: 

493: \begin{table}[t]

494: \caption{Distances (in \AA{}) between the various C~atoms $(i,j)$ around the vacancy, as

495: listed in Fig.~\ref{fig:vacdist}.}

496: \label{tab:vacdist}

497: \begin{tabular}{ccc}

498: \hline

499: $i$ & $j$ & $d_{ij}$ \\

500: \hline

501: 68 & 69 & 1.38\\

502: 68 & 70 & 1.38\\

503: 68 & 71 & 2.39\\

504: 68 & 66 & 2.39\\

505: 68 & 17 & 2.77\\

506: 68 & 6  & 2.77\\

507: 68 & 1  & 2.63\\

508: 68 & 5  & 2.63\\

509: 68 & 80 & 2.41\\

510: 68 & 82 & 2.41\\

511: \hline

512: \end{tabular}

513: \end{table}

514: 

515: 

516: \subsection{Clusters with a substitutional proton}

517: 

518: Last, we have considered the same clusters as in Fig.~\ref{fig:coronene2plus},

519: but now with a proton substituting a carbon atom at the cluster center. Such a

520: configuration has been realized by considering a de-electronated hydrogen atom.

521: The same considerations for the total energy apply also to this system, as

522: discussed in the presvious sections. The distances between the proton and the

523: nearest neighbor C atoms are 1.03~\AA{} for the C$^{68}$ atom, and 1.83~\AA{}

524: for the other two next nearest neighbor atoms. The cluster turns out to be flat.

525: The DOS presents a peak near zero energy (cf. Fig.~\ref{fig:DOSvac}), as found

526: by several authors when considering the effect of impurities in graphene

527: \cite{Pereira:08,Hu:08}. We note that, for a substitutional proton, the gap is

528: more than a half smaller with respect to the cases considered in

529: Sections~\ref{sec:chemisorbed} and \ref{sec:vacancy} (Cf.

530: Tab.~\ref{tab:moment}). It is even smaller with respect to a cluster with a

531: substitutional H, \emph{i.e.} a system presenting charge neutrality, which is

532: also reported in Tab.~\ref{tab:moment}. We argue that the gap is a sensitive

533: quantity with respect to the injections of extra charges. The magnetic moment on

534: the proton has an intermediate value with respect to the two cases considered

535: previously. It is more sensitive to the cluster size going from

536: $0.91$~$\mu_{\mathrm{B}}$ from coronene~19 to $0.83$~$\mu_{\mathrm{B}}$ for

537: coronene~37, but is similar to that obtained for the substitutional H, which is

538: $0.85$~$\mu_{\mathrm{B}}$ for both clusters (cf. Tab.~\ref{tab:moment}). The

539: present model is probably the most interesting in realtion to the experiments

540: performed by irradiating graphite with protons \cite{Esquinazi:03}.

541: 

542: \section{Conclusions}

543: \label{sec:conclusions}

544: 

545: In the present work we have followed a `molecular' approach to study impurity

546: effects in graphene. The impurities considered were a chemisorbed H atom, a

547: vacancy, and a a positively charged substitutional H. We modelled these systems

548: by considering clusters containing up to 37 benzene rings. We performed HF and

549: UHF calculations using the STO-3G basis set.

550: 

551: The main emphasis of this work is not on correlation effects, which in some

552: cases have been treated at the MP2 level, but on the size of the cluster. From

553: the results discussed in the present work, we believe that coronene~37 is

554: sufficiently large to ensure a local behavior similar to that of bulk graphene,

555: especially near the center of the cluster. This is supported by the small

556: changes in the physical properties studied here in going from coronene~19 to

557: coronene~37. Of course, some features differ considerably from the finite

558: clusters and the infinite system. For instance, we still observe quite a large

559: gap in the density of states of finite clusters, as compared to that of

560: graphene, while we have shown some important trends, which look quite promising

561: in interpreting the general features of graphene: (i) The gap decreases by

562: increasing the cluster size; (ii) The gap does not decrease if we consider H

563: chemisorption or a vacancy; (iii) the gap is drastically reduced if we consider

564: a substitutional proton. Other important features are: (i) an increase of the

565: DOS near the HOMO level for all the impurities considered here with respect to

566: the pure system; (ii) the appearance of a zero mode in the DOS if we consider a

567: substitutional proton.

568: 

569: These results are for many aspects in agreement with the theoretical work of

570: Pereira \emph{et al.} \cite{Pereira:08}, who have studied local disorder in

571: graphene by using a tight-binding approach. Such a method allowed these authors

572: to treat lattices with $4\times 10^6$ carbon atoms, thus reaching the size of

573: real samples. However, they cannot treat lattice distortions around the defect,

574: or the specific features connected with the particular impurity introduced.

575: 

576: In agreement with experiments \cite{Esquinazi:03}, we have found that

577: chemisorbed H, substitutional proton, and C atom near a vacancy acquire a

578: magnetic moment. The determination of a precise value thereof is quite delicate,

579: but we believe that the trend on going from one defect to another is correct, in

580: view of the comparison of our results with the calculations of Yazyev and Helm

581: \cite{Yaziev:07}.

582: 

583: We could not derive definite conclusions as regards the spacing of the levels

584: near the HOMO state, and further work is thus required in this direction. In the

585: future, it should be possible to treat even larger clusters, and these will be

586: able to treat more realistically especially graphene quantum dots. We find that

587: the gap strongly depends on the graphene cluster size, so that it is essential

588: to control the size of graphene quantum dots in order to obtain a particular gap

589: value. This is relevant for the design of novel electronic devices, as well as

590: it is important to control the impurity level.

591: 

592: %\cite{Schedin:07}

593: %\cite{Wehling:08}

594: %\cite{Leenaerts:08}

595: %\cite{Nakada:96}

596: %\cite{Russo:08}

597: 

598: \ack

599: 

600: The authors thank Professor N. H. March for useful discussions over the general

601: area embraced by the present paper.

602: 

603: \begin{small}

604: %\bibliographystyle{apsrev}

605: \bibliographystyle{mprsty}

606: \bibliography{a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t,u,v,w,x,y,z,zzproceedings,Angilella,notes}

607: \end{small}

608: 

609: 

610: 

611: 

612: \end{document}

613: 

614: