1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %Version from May 2003
3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4: %
5: %\documentstyle[aps,epsfig,twocolumn]{revtex}
6: %\documentclass[twocolumn,aps,prb,showpacs,preprintnumbers,amsmath,amssymb,psfrac]{revtex4}
7: \documentclass[twocolumn,aps,floats,superscriptaddress,showpacs]{revtex4}
8: \usepackage{amsmath}
9: \renewcommand{\topfraction}{0.85}
10: \renewcommand{\textfraction}{0.1}
11: \renewcommand{\floatpagefraction}{0.75}
12: \usepackage{epsfig}
13: %\usepackage{psfig}
14: \usepackage{graphicx}% Include figure files
15: \usepackage{dcolumn}% Align table columns on decimal point
16: \usepackage{bm}% bold math
17:
18: %http://www.ias.ac.in/currsci/oct252001/1011.pdf
19:
20: \renewcommand{\dag}{^{\dagger}}
21: \newcommand{\dl}{\partial_\ell}
22: \def\gapp{\lower.35em\hbox{$\stackrel{\textstyle>}{\sim}$}}
23: \def\lapp{\lower.35em\hbox{$\stackrel{\textstyle<}{\sim}$}}
24:
25: \begin{document}
26: \bibliographystyle{apsrev}
27: %
28:
29: %\draft
30: \newcommand{\beq}{\begin{equation}}
31: \newcommand{\eeq}{\end{equation}}
32: \newcommand{\beqa}{\begin{eqnarray}}
33: \newcommand{\eeqa}{\end{eqnarray}}
34: \newcommand{\da}{^\dagger}
35: \newcommand{\wh}{\widehat}
36:
37: \newcommand{\Om}{\Omega}
38: \newcommand{\om}{\omega}
39: \newcommand{\tr}{{\rm tr}}
40: \newcommand{\intf}{\int_{-\infty}^\infty}
41: \newcommand{\into}{\int_0^\infty}
42: \newcommand{\I}{{\mathcal I}}
43: \newcommand{\G}{{\mathcal G}}
44: \newcommand{\A}{{\mathcal A}}
45: \newcommand{\F}{{\mathcal F}}
46: \newcommand{\C}{{\mathcal C}}
47: \newcommand{\x}{\vec x}
48: \newcommand{\X}{\vec X}
49: \newcommand{\q}{\vec q}
50:
51: \renewcommand{\=}{\!\!=\!\!}
52:
53: \def\simleq{\; \raise0.3ex\hbox{$<$\kern-0.75em
54: \raise-1.1ex\hbox{$\sim$}}\; }
55: \def\simgeq{\; \raise0.3ex\hbox{$>$\kern-0.75em
56: \raise-1.1ex\hbox{$\sim$}}\; }
57:
58: \def\la{{\langle}}
59: \def\ra{{\rangle}} \def\vep{{\varepsilon}} \def\y{\'\i}
60: \def\half{{1\over 2}}
61: \def\an{|\Phi _N \rangle }
62: \def\bn{|\Phi ' _{N'}\rangle }
63: \def\na{ \langle \Phi _N|}
64: \def\nb{ \langle \Phi'_{N'} |}
65:
66: \def\ov{\over}
67: \def\non{\nonumber }
68: \def\beq{\begin{equation} }
69: \def\eeq{\end{equation} }
70: \def\beqa{\begin{eqnarray}}
71: \def\eeqa{\end{eqnarray}}
72: \def\del{\partial }
73: \def\D{\Delta}
74: \def\a{\alpha }
75: \def\am{\alpha^\mu }
76: \def\Xm{X^\mu}
77: \def\Xn{X^\nu}
78: \def\d{\textrm{d}}
79: \def\b{\beta}
80: \def\t{\tau}
81: \def\e{\epsilon}
82: \def\g{\gamma}
83: \def\s{\sigma}
84: \def\med{\frac{1}{2}}
85: \def\go{\vec g_1}
86: \def\gt{\vec g_2}
87: \def\eq{\!\! =\!\!}
88:
89: \title{Existence and topological stability of Fermi points in
90: multilayered graphene}
91: %
92: \author{J. L. Ma\~nes}
93: \affiliation{Departamento de F\'{\i}sica de la Materia Condensada\\
94: Universidad del Pa\'{\i}s Vasco,
95: Apdo. 644, E-48080 Bilbao, Spain}
96: \author{F. Guinea}
97: \affiliation{Unidad Asociada CSIC-UC3M,
98: Instituto de Ciencia de Materiales de Madrid,\\
99: CSIC, Cantoblanco, E-28049 Madrid, Spain.}
100: \author{ Mar\'{\i}a A. H. Vozmediano}
101: \affiliation{Unidad Asociada CSIC-UC3M,
102: Universidad Carlos III de Madrid, E-28911
103: Legan\'es, Madrid, Spain. }
104:
105:
106:
107: \date{\today}
108: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
109: \begin{abstract}
110: We study the existence and topological stability of Fermi points
111: in a graphene layer and stacks with many layers.
112: We show that the discrete symmetries
113: (spacetime inversion) stabilize the
114: Fermi points in monolayer, bilayer and multilayer
115: graphene with orthorhombic stacking. The bands near $k=0$ and $\epsilon=0$
116: in multilayers with the Bernal stacking depend on the parity of the number of
117: layers, and Fermi points are unstable when the number of layers is
118: odd. The low energy changes in the
119: electronic structure induced by commensurate perturbations which mix the two
120: Dirac points are also investigated.
121: \end{abstract}
122: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
123: %
124: \pacs{75.10.Jm, 75.10.Lp, 75.30.Ds}
125: %
126:
127: \maketitle
128: \section{Introduction.}
129:
130: The recent synthesis of monolayer graphite~\cite{Netal04,Netal05b}
131: (graphene), the experimental ability to manipulate few layer
132: samples~\cite{Betal04,Netal05,Zetal05} and the observations of
133: quasi two dimensional behavior in graphite~\cite{Ketal03}, have
134: awaken an enormous interest in these systems. The conduction band
135: of graphene is well described by a tight binding model which
136: includes the $\pi$ orbitals which are perpendicular to the plane
137: at each C atom\cite{W47,SW58}. This model describes a semimetal,
138: with zero density of states at the Fermi energy, and where the
139: Fermi surface is reduced to two inequivalent K-points located at
140: the corners of the hexagonal Brillouin Zone. The low-energy
141: excitations with momenta in the vicinity of any of the Fermi
142: points have a linear dispersion and can be described by a
143: continuum model which reduces to the Dirac equation in two
144: dimensions~\cite{GGV92,GGV93}, what has been tested by recent
145: experiments~\cite{Netal05,Zetal05,Zetal06}. Fermi points have also
146: been found in the modelling of the low energy band structure of
147: multilayer systems both theoretically~\cite{NCPG06,NCGP06} and in
148: experiments~\cite{Netal06b,Zetal06b}. A crucial issue for both
149: theory and technology is the possibility of controlling the
150: opening of a gap in the samples. From a theoretical point of view,
151: the gap is related to the chiral symmetry breaking and mass
152: generation, a classical -unresolved- problem that has been
153: explored at length in planar QED~\cite{AN88,Khv01}. For the
154: applications it is by now clear that opening a gap in monolayer
155: graphene will be a difficult task and efforts are concentrated on
156: multilayer structures~\cite{NCGP06,Oetal06,Cetal06}.
157:
158: In this paper we analyze the stability of the Fermi points under small perturbations
159: using very basic topological concepts~\cite{Naka}. We find that the Fermi points are protected
160: by the discrete symmetries (translational invariance and space and time inversion)
161: in the monolayer, bilayer AB and multilayers with $ABCA \cdots$ (rhombohedral)
162: stacking. The stability of Fermi points in stacks with the Bernal stacking,
163: $ABAB \cdots$, depends on the parity of the number of layers. We also discuss
164: the changes in the low energy and low momenta properties induced by
165: commensurate perturbations which hybridize the two $K$ points and partially
166: break translation invariance. We will not analyze here in detail the effects
167: of spin-orbit coupling, which my lead to additional changes at low temperatures\cite{KM05,HGB06,Metal06b,Yetal07}.
168:
169: The analysis reported here will be useful for the construction of
170: continuum theories
171: for long wavelength spatial perturbations, and for the study of the degeneracy
172: and spectrum of the low index Landau levels in a magnetic field.
173:
174: \section{ Electronic structure and stability in graphene.} The
175: Fermi surface (FS) is a central concept in condensed matter that
176: controls the low-energy physics of the systems. In a Landau Fermi
177: liquid at T = 0, J. Luttinger \cite{L60} defined the FS of an
178: interacting Fermi system in terms of the single-particle Green's
179: function $G({\vec k}, \omega)$, as the solution of the equation
180: $G^{-1}({\vec k}, 0) = 0 $ and showed that it encloses the same
181: volume, equal to the fermion density $n$, as in the noninteracting
182: system. The robustness of the Fermi liquid idea has been
183: understood recently in the context of the renormalization group
184: where the Fermi and Luttinger liquids are seen as infrared fixed
185: points\cite{P93,S94}. In recent works\cite{V03,V06} Volovik has
186: emphasized the idea of the topological stability of the Fermi
187: surface as the origin of the robustness of the Fermi liquid and
188: has suggested a classification of general fermionic systems in
189: universality classes dictated by momentum space topology. A more
190: recent proposal relates the stability of Fermi surfaces with
191: K-theory, a tool used to classify D-brane charges in string
192: theory\cite{H05}. The idea behind the topological stability is to
193: study the zeroes of the matrix $G^{-1}_0({\vec k}, \omega)$ (free
194: inverse propagator) that can not be lifted by small perturbations.
195: Here we will analyze the stability of the Fermi points of single
196: and multilayer graphene, where the discrete symmetries of the
197: system play a principal role. Although we will restrict ourselves
198: to perturbations that can be studied within the context of a
199: single-particle effective hamiltonian, the extension to
200: self-energy induced perturbations is rather straightforward and
201: will be reported elsewhere\cite{us07}.
202:
203: \begin{figure}[!t]
204: \begin{center}
205: \includegraphics[width=5cm]{direct.eps}
206: \caption[fig]{\label{DL}(Color online) Direct lattice and unit
207: cell for monolayer graphene. }
208: \end{center}
209: \end{figure}
210: %
211: \begin{figure}[!t]
212: \begin{center}
213: \includegraphics[width=4cm]{brillouin.eps}
214: \caption[fig]{\label{BZ} (Color online) First Brillouin zone and
215: Fermi points. The vectors $\vec K\,'_1,\vec K\,''_1$ ($\vec
216: K\,'_2,\vec K\,''_2$) are equivalent to $\vec K_1$ ($\vec K_2$).}
217: \end{center}
218: \end{figure}
219:
220:
221: As shown in Fig. \ref{DL}, monolayer graphene consists of a planar honeycomb lattice of
222: carbon atoms. Corresponding to the two atoms in the unit cell, one
223: may define two Bloch wave functions to be used in a variational
224: (tight-binding) computation of the spectrum \beq\label{bloch}
225: \Phi_i (\vec K)=\sum_{\vec t} e^{i \vec K\cdot(\vec r_i+ \vec t)}
226: \Phi (\vec r-\vec r_i-\vec t) \; \;\; , \; \;\; i=A,B \eeq where
227: the sum runs over all the points in the direct lattice, i.e.,
228: $\vec t = n_1 \vec t_1+n_2 \vec t_2$, $(\vec r_A,\vec r_B)$ are
229: the positions of the atoms in the unit cell, and $\Phi(\vec r)$
230: is a real ($\pi$-type) atomic orbital. As is well known, a
231: simple tight-binding computation~\cite{W47,SW58} yields a
232: spectrum with two Fermi points located at $\vec K_1 = -2\go/3-
233: \gt/3$ and $\vec K_2= -\vec K_1$, where $\vec t_i\cdot\vec g_j=2\pi \delta_{ij}$ (see Fig. \ref{BZ}). Near the
234: two Fermi points, the hamiltonian can be linearized, and using
235: appropriate units one finds \beq\label{lin} H(\vec K_1+\vec
236: k)\sim\left(
237: \begin{array}{cc} 0 & k^*
238: \\ k & 0 \end{array} \right)=k_x\sigma_x+k_y\sigma_y ,
239: \eeq
240: and $H(-\vec K_1+\vec k)\sim - k_x \sigma_x + k_y \sigma_y$.
241: %\beq
242: %H(-\vec K_1+\vec k)\sim\left( \begin{array}{cc} 0 &- k \\
243: %-k^* & 0 \end{array} \right)
244: %\eeq
245: where $k\equiv k_x+ik_y$ and $\sigma_i$ are the Pauli matrices.
246: Thus, the low energy electronic excitations behave like massless
247: Dirac fermions with relativistic spectrum $E=\pm |k|$.
248:
249: Under a $k$-independent, translationally invariant perturbation
250: \beqa\label{const} & H(\vec
251: K_1+\vec k)\to\left( \begin{array}{cc} a_z & k^*+a^*
252: \\ k+a & -a_z \end{array} \right)
253: %= \\
254: %& a_z\sigma_z +(k_x+a_x)\sigma_x+(k_y+a_y)\sigma_y
255: \eeqa
256: where $a\equiv a_x+i a_y$, the spectrum becomes
257: $E=\pm\sqrt{a_z^2+|k+a|^2}$ and a gap $2| a_z|$is generated.
258: This is consistent with Horava's general arguments~\cite{H05},
259: which show that Fermi points for Dirac fermions in two dimensions are unstable.
260: However, this is not the end of the story, since the
261: existence of discrete symmetries can sometimes stabilize
262: the Fermi loci. In the case of the graphene, this role
263: is played by time-reversal $T:t\to -t$ and spatial inversion
264: $I:(x,y)\to(-x,-y)$. The reality of the $\pi$~orbitals implies that
265: time reversal merely reverses $\vec K$
266: \beq\label{compl}
267: T\Phi_i(\vec K) =\Phi^*_i(\vec K)=\Phi_i (-\vec K)
268: \eeq
269: whereas the
270: spatial inversion also exchanges the two types of
271: atoms
272: \beq\label{I}
273: I\Phi_A (\vec K)=\Phi_B (-\vec K)\;\;\; , \;\;\; I\Phi_B (\vec K)=\Phi_A (-\vec K)
274: \eeq
275: Invariance under these symmetries imposes the following constraints on the hamiltonian
276: \beqa\label{trev}
277: T:\; H (\vec K)&=&H^* (-\vec K)\non\\
278: I:\;H (\vec K)&=& \sigma_x H(-\vec K) \sigma_x
279: \eeqa
280: Although these are useful properties that relate the hamiltonians at opposite values of $\vec K$,
281: what we need is a constraint on the form of $H(\vec K)$.
282: This is obtained by combining time reversal with the
283: spatial inversion
284: \beq\label{TI}
285: TI:\; H(\vec K)=\sigma_x H^*(\vec K) \sigma_x
286: \eeq
287: implying $H_{11}(\vec K)=H_{22}(\vec K)$. This enforces
288: $a_z\!=\!0$ in~(\ref{const}) and we see that no gap opens if
289: the perturbation preserves the space-time inversion $TI$.
290:
291: This has an interesting topological interpretation, which extends
292: the previous arguments to $k$-dependent ---but translationally invariant--- perturbations.
293: The low energy hamiltonian $H(\vec K_1+\vec k)$ in~(\ref{lin})
294: defines a map from the circle $k_x^2+k_y^2=R^2$ to the space
295: of $2\times 2$ hamiltonians
296: $H=\vec h \cdot\vec \sigma$:
297: \beq\label{map1}
298: k=R e^{i\theta} \to (h_x,h_y, h_z) =R (\cos\theta ,\sin\theta,0)
299: \eeq
300: Since Fermi points correspond to zeroes of the determinant
301: $ -Det(H)=h_x^2+h_y^2+h_z^2$,
302: % \beq\label{det}
303: % -Det(H)=h_x^2+h_y^2+h_z^2
304: % \eeq
305: a perturbation will be able to create a gap only if the loop
306: represented by the map~(\ref{map1}) is contractible in the space
307: hamiltonians with non-vanishing determinants, which %by~(\ref{det})
308: is just $R^3-\{ 0\}$. This is clearly the case, since
309: $ \pi_1(R^3-\{ 0\})=\pi_1(S^2)=0$.
310: % \beq\label{hom1}
311: % \pi_1(R^3-\{ 0\})=\pi_1(S^2)=0
312: % \eeq
313: On the other hand, hamiltonians invariant under $TI$ are represented
314: by points in $R^2$, and we have
315: \beq\label{hom2}
316: \pi_1(R^2-\{ 0\})=\pi_1(S^1)=Z
317: \eeq
318: This means that non-trivial maps such as the ones implied
319: by~(\ref{lin}) can only be extended to the interior of the
320: circle by going through the origin, i.e., by having at
321: least one zero. This precludes the creation of a gap.
322:
323:
324: Note that the maps defined by the low energy hamiltonian
325: in the proximity of the two Fermi points
326: \beq
327: H(\pm\vec K_1+\vec k)\; : \; k=R e^{i\theta} \to h_x+ih_y=\pm R e^{\pm i\theta} %\non\\
328: %H(-\vec K_1+\vec k)&: &\;\; k=R e^{i\theta} \to h_x+ih_y=-R e^{-i\theta}
329: \eeq
330: have opposite winding numbers $N=\pm 1$, which can be computed by
331: the formula \beq\label{wind} N={1\over 4\pi i}\int_0^{2\pi} d
332: \theta\, \mathrm{Tr} (\sigma_z H^{-1} \partial_\theta H) \eeq
333:
334: The fact that the two Dirac points carry opposite charges
335: suggests that they could annihilate mutually if brought
336: together by a perturbation. Any external potential commensurate with
337: the honeycomb lattice, which has a finite Fourier component at the
338: wavevector $\vec{G} = \vec{K}_1 - \vec{K}_2$, induces terms which hybridize
339: the two Dirac points and it will lead to the possibility of a gap.
340: We can
341: compute all the possible perturbations which are compatible with the
342: symmetries of the lattice. The most general $( 4 \times 4 )$ hamiltonian
343: including perturbations at $\vec{G} = 0$ and $\vec{G} = \vec{K}_1 -
344: \vec{K}_2\equiv -\vec K_1$ is:
345: %\begin{widetext}
346: %\begin{equation}
347: %{\cal H} \equiv \left( \begin{array}{cccc} 0& k_x - i k_y + Q_1^x - i Q_1^y &Q_2
348: % &Q_4 \\ k_x + i k_y + Q_1^x + i Q_1^y &0 &Q_4 &Q_3 \\ Q_2^* &Q_4^* &0 &-
349: % k_x - i k_y - Q_1^x - i Q_1^y \\ Q_4^* &Q_3^* &- k_x + i k_y - Q_1^x +i
350: % Q_1^y &0 \end{array} \right)
351: %\end{equation}
352: %\end{widetext}
353: \begin{equation}
354: {\cal H} \equiv \left( \begin{array}{cccc} 0& k^* + Q^*_1&Q_2
355: &Q_4 \\ k + Q_1 &0 &Q_4 &Q_3 \\ Q_2^* &Q_4^* &0 &-
356: k+Q_1 \\ Q_4^* &Q_3^* &- k^*+Q_1^*
357: &0 \end{array} \right)
358: \label{Hpert}
359: \end{equation}
360: where $Q_1=Q_1^x+i Q_1^y$ transforms according to the
361: $E_2$ representation of the $C_{6v}$ symmetry group\cite{note}
362: %\footnote{The actual symmetry group of monolayer graphene is $D_{6h}$. But,
363: % as long as we restrict ourselves to in-plane distorsions of the lattice,
364: % it is sufficient to consider the subgroup $C_{6v}$. That is what we do in
365: % this paper, since out-of-plane distorsions (phonons) can not couple
366: % linearly to electrons.}
367: at the \hbox{$\Gamma ( \vec{K}
368: = 0)$} point. $Q_2$ and $Q_3$ belong to the $E$ representation of the
369: $C_{3v}$ group at $\vec{K}_1$, and $Q_4$ to the $A_1$ representation of the
370: same group (see~\cite{BC72} for notation).
371: At this point it is worth noticing a point on notation. When
372: grouping the hamiltonians attached to the two Dirac points
373: ($K_{1,2}$) into a 4-dimensional matrix it is a common practice to
374: reverse the order of the sublattices (A,B) in one of the Fermi
375: points in such a way that the 4-dimensional wavefunctions have
376: the form
377: $$\psi=(\Phi_{K_1,A},\Phi_{K_1,B},\Phi_{K_2,B},\Phi_{K_2,A}).$$
378: If this is done the topological structure of the hamiltonian is
379: messed up and the computation of charges becomes less clear. For this reason,
380: we follow instead the convention in~\cite{KM05}, where
381: $$\psi=(\Phi_{K_1,A},\Phi_{K_1,B},\Phi_{K_2,A},\Phi_{K_2,B}).$$
382: This is also important if one tries to compare the analysis of the
383: perturbations written in eq. (\ref{Hpert}) with the ones produced
384: by the different types of disorder \cite{A06,Mcetal06,AE06}.
385:
386: The perturbation given by $Q_1^x$ and $Q_1^y$ shifts the Dirac
387: points, but does not open a gap. In fact, the only parameter which opens a gap is~$Q_4$. When only
388: $Q_4$ is different from zero, the spectrum becomes $E=\pm\sqrt{|Q_4|^2+|k|^2}$ and,
389: for $Q_4$ real, the deformation of the lattice is given in Fig.~\ref{f2},
390: where one can see that no point symmetry is broken.
391: %It is interesting that we can create a gap just by a partial breaking of translation symmetry.
392: This shows, in particular, that invariance under space-time
393: inversion is \textit{not} enough to guarantee the stability of
394: Fermi points ---translation invariace plays a crucial role: $TI$
395: by itself makes de Fermi points \textit{individually} stable, but
396: they may still annihilate against each other in the presence of a
397: perturbation that breaks translation invariance. A distortion of
398: the type of $Q_4$ can be induced by a substrate with a periodicity
399: commensurate with the lattice, or by the effect of one layer on
400: another when there is a lattice mismatch between them, as in
401: samples grown on a substrate~\cite{Betal04,Betal06}. It can be
402: responsible for the gap observed recently photoemission
403: experiments\cite{L07}. It is worth noting that the perturbation
404: denoted here $Q_4$ has been studied in a graphene ribbon
405: in~\cite{GLV06}.
406:
407: \begin{figure}[!t]
408: \begin{center}
409: \includegraphics[width=5cm]{threefold.eps}\\
410: \caption[fig]{\label{f2} (Color online)Distortion caused by the
411: condensation of $A_1$ mode with real $Q_4$ and new unit cell.}
412: \end{center}
413: \end{figure}
414:
415:
416: When only $Q_2$ or $Q_3$ are
417: different from zero, we find:
418: \begin{equation}\label{bila}
419: \epsilon_k =\pm \frac{|Q_{2,3}|}{2} \pm
420: \sqrt{\frac{|Q_{2,3}|^2}{4} + | \vec{k} |^2}.
421: \end{equation}
422: The energy bands are represented in Fig. \ref{Q3} for the
423: particular case $Q_2=0, Q_3=1$. We can see that the spectrum is
424: the same as the one obtained in a simple model for a bilayer
425: system~\cite{MF06}, which will be discussed later. A complete analysis of the most general
426: perturbation of the form~(\ref{Hpert}) will be given elsewhere~\cite{us07}.
427:
428:
429: In the absence of time reversal symmetry, other perturbations are possible, such as
430: \begin{equation}
431: \delta {\cal H} = B_1 \sigma_z \tau_z + B_2 \sigma_y \tau_y
432: \label{mag_field}
433: \end{equation}
434: where $\sigma$ and $\tau$ are Pauli matrices whose entries are the sublattice
435: and K point indices respectively, and $B_1$ and $B_2$ transform like the $z$ component of a magnetic field
436: and are odd under time inversion. Note that the
437: first term is the orbital part of the intrinsic spin orbit coupling in
438: graphene~\cite{KM05,HGB06,Metal06b,Yetal07}, and it opens a gap. The second term should appear in a
439: general spin orbit hamiltonian which takes into account the coupling between
440: the two $K$ points.
441:
442: \begin{figure}[!t]
443: \begin{center}
444: \includegraphics[width=4cm]{band3.eps}
445: \caption[fig]{\label{Q3} Energy bands along the line $(0,k_y)$
446: for $Q_3=1$, $Q_4=0$.}
447: \end{center}
448: \end{figure}
449:
450: \begin{figure}[!t]
451: \begin{center}
452: \includegraphics[width=4cm]{stacking_2.eps}
453: \caption[fig]{\label{stacking} (Color online) Skecth of the three
454: possible positions of a given
455: layer with respect to the others in a graphene stack. Bernal stacking
456: ($1,2,1,2, \cdots$, is described by two inequivalent planes, while
457: orthorhombic stacking, $1,2,3,1,2,3 \cdots$, requires the three
458: inequivalent ones.}
459: \end{center}
460: \end{figure}
461:
462: \section{ Electronic structure and stability in multilayered
463: graphene.} The case of the multilayer is more interesting. We will
464: concentrate on the ability of the $TI$ invariance to prevent the
465: creation of a gap. For the sake of definiteness, only staggered
466: ($ABA$) and rhombohedral ($ABC$) stacking will be considered. The relative
467: orientations of the $ABC$ planes are sketched in Fig.[\ref{stacking}]. The
468: two inequivalent atoms in layer $n$ will be denoted ($A_n,B_n$).
469: Our conventions are such that the couplings of an $A_n$ ($B_n$)
470: atom to the three in-plane nearest neighbors are shaped as a Y
471: (inverted Y), independently of $n$. Thus, the low energy limit of
472: the ``free'' hamiltonian obtained by neglecting inter-layer
473: couplings is block-diagonal, with $2\times 2$ blocks given
474: by~(\ref{lin}).
475:
476: The simplest model introduces interlayer hoppings $t$ only between
477: nearest neighbors. The resulting hamiltonian for bilayer graphene in the vicinity of the
478: $K_1$ Fermi point is
479: \beq\label{bil}
480: \mathcal{H}(k)=\left(
481: \begin{array}{llll}
482: 0 & k^* & 0 & t \\
483: k & 0 & 0 & 0 \\
484: 0 & 0 & 0 & k^* \\
485: t & 0 & k & 0
486: \end{array}
487: \right)
488: \eeq
489: and the energy bands are
490: given by~(\ref{bila}) with the replacement $|Q_{2,3}|\to t$. In the limit $E\ll t$, one can
491: obtain an effective hamiltonian~\cite{MF06} for the lowest energy
492: bands.
493: To this end, reorder the wavefunctions according $(A_1,B_1,A_2,B_2)\to (A_2,B_1,A_1,B_2)$,
494: so that in the new basis the hamiltonian becomes
495: \beq
496: \mathcal{H}(k)=\left(
497: \begin{array}{llll}
498: 0 & 0 & 0 & k^* \\
499: 0 & 0 & k & 0 \\
500: 0 & k^* & 0 & t \\
501: k & 0 & t & 0
502: \end{array}\right)\equiv\left(
503: \begin{array}{ll} H_{11} &H_{12}\\
504: H_{21} &H_{22}
505: \end{array}
506: \right)
507: \eeq
508: where $H_{ij}$ is a $2\times 2$ block.
509: The identity
510: %
511: \begin{eqnarray}
512: & Det(\mathcal{H}-E)\\ \nonumber =
513: &Det\Bigl(H_{11}-H_{12}(H_{22}-E)^{-1}H_{21}-E\Bigr)\,Det(H_{22}-E)
514: \end{eqnarray}
515: %
516: shows that, for $E\ll t$, the substitution $H_{22}-E\to
517: H_{22}$ reduces the computation of the lowest energy bands to the
518: diagonalization of the $2\times 2$ effective hamiltonian
519: %
520: \beq
521: \label{bieff}
522: \mathcal{H}^{eff}\equiv H_{11}-H_{12}H_{22}^{-1}H_{21}=-{1 \over t}\left(
523: \begin{array}{cc}
524: 0 & k^{*2}\\
525: k^{2} & 0
526: \end{array}
527: \right)
528: \eeq
529: %
530: This effective hamiltonian involves only the atoms $(A_2,B_1)$,
531: which are not linked by $t$ and give rise to bands with zero energy
532: at the Fermi points. Since $(A_2,B_1)$ are interchanged under spatial
533: inversion \hbox{$\vec r\to -\vec r$}, the combined $TI$-invariance imposes a constraint
534: \hbox{$\mathcal{H}^{eff}(k)=\sigma_x \mathcal{H}^{eff*}(k)\sigma_x$}
535: %\beq
536: %\mathcal{H}^{eff}(k)=\sigma_x \mathcal{H}^{eff*}(k)\sigma_x
537: %\eeq
538: identical to~(\ref{TI}). This implies
539: $\mathcal{H}_{11}^{eff}(k)=\mathcal{H}_{22}^{eff}(k)$,
540: which shows that no gap can open.
541: %This is not surprising,
542: %since by (\ref{bieff}) bilayer graphene restricted to the lowest
543: %energy bands is equivalent to monolayer graphene with the
544: %replacement $k \to -k^2/t$.
545: According to~(\ref{wind}) the topological charge for
546: the $\vec K_1$ Fermi point is $+2$ and, by time reversal
547: invariance, the charge for
548: $-\vec K_1$ is $-2$. Thus, as in the case of monolayer graphene,
549: the Fermi points are stable under perturbations that preserve $TI$ and translation invariance. For
550: instance, a
551: perturbation like trigonal warping~\cite{MF06} changes the off diagonal
552: elements in eq.(\ref{bieff}), $k^2 \rightarrow k^2 + v_3 k^*$ and splits the
553: Fermi point of charge $Q=+2$ into three Dirac points away from
554: the $K$ point, and charge $Q=+1$, and another Dirac point at the $K$ point
555: and $Q=-1$, but the total charge is conserved and no gap opens. However, a perturbation hybridizing
556: $\vec K_1$ and $-\vec K_1$ or one breaking $TI$ might lead to a gapped system with no Fermi points at all.
557: A physical example is provided by the experiment described in
558: \cite{Oetal06} where a gap is controlled by changing the carrier
559: concentration in each layer.
560:
561: This analysis can be easily generalized to multilayer graphene with
562: rhombohedral stacking. This type of staking includes the links
563: $(B_1-A_2, B_2-A_3,\ldots, B_{N-1}-A_N)$ and the effective hamiltonian,
564: which involves only the unlinked atoms $(A_1,B_N)$, is given by
565: \beq\label{muleff}
566: \mathcal{H}^{eff}\=-{1 \over t^{N-1}}\left(
567: \begin{array}{cc}
568: 0 & k^{*N}\\
569: k^{N} & 0
570: \end{array}
571: \right)
572: \eeq
573: The topological charge for the $\vec K_1$ ($-\vec K_1$) Fermi point is $+N$ ($-N$).
574: As the point group for multilayer
575: graphene with rhombohedral stacking is $D_{3d}$, which contains
576: the inversion~$I$, the system is invariant under $TI$, which interchanges $(A_1,B_N)$,
577: and the whole argument goes through
578: as before. Thus we conclude that the Fermi points for
579: multilayer graphene with rhombohedral stacking are
580: stable against perturbations that respect $TI$ and translation invariance.
581:
582: The situation is very different for $ABA$ stacking.
583: An $N$-layer graphene stack is invariant under
584: the spatial inversion $I$ only for even $N$, where the point group is
585: $D_{3d}$, while it is $D_{3h}$ for odd $N$.
586: %In the absence of a convolution such as $TI$,
587: %the stability of the Fermi points is no longer obvious.
588:
589: The $2N$ eigenstates in a stack with $N$ layers at $k=0$
590: can be divided into two sets: $N$ states at the
591: orbitals connected by the interlayer hopping $t$, and $N$ states in the other
592: sublattice sites of each layer. If we only consider the hopping $t$, the
593: first $N$ states acquire a dispersion~\cite{GCP06}, lying in the range $- 2 t
594: \le \epsilon \le 2 t$. The other $N$ states are degenerate with $\epsilon = 0$. A
595: perturbation compatible with all the symmetries of the stack is a layer
596: dependent shift of the onsite energies. This shift can be arbitrary, except
597: for the twofold degeneracy related with the equivalence between layers which
598: are symmetrically placed around the center, $\epsilon_n = \epsilon_{N-n+1}$.
599: This is illustrated in Fig.[\ref{stack_4_8}].
600:
601: \begin{figure}[!t]
602: \begin{center}
603: \includegraphics[width=6cm,angle=-90]{multilayer_gap_4_8_9.ps}\\
604: \caption[fig]{\label{stack_4_8}(Color online) Energy bands close
605: to $\epsilon = 0$ for a stack with 4 layers (left), 8 layers
606: (center), and 9 layers (right). All stacks have the Bernal
607: stacking. The Fermi velocity is $v_F = 1 , t = 0.1$ and a layer
608: dependent shift has been included: $\epsilon_n \equiv \{ 0.02 ,
609: 0.01 \}$ (left), $\epsilon_n \equiv \{ 0.02 , 0.01 , 0.005 , 0.002
610: \}$ (center), and $\epsilon_n \equiv \{ 0.02 , 0.01 , 0.005 ,
611: 0.002 , 0\}$ (right).}
612: \end{center}
613: \end{figure}
614:
615:
616: The results in Fig.[\ref{stack_4_8}] show a gap at half filling
617: in the stack with four layers,
618: and overlapping bands at all energies for the stack with eight and nine
619: layers.
620: Note that the gap in the stack with four layers does not require the
621: existence of an external electric field,
622: which will break the equivalence of the layers at
623: opposite sides of the stack. In all stacks
624: with an even number of layers, the two bands which start at the onsite energy of layers $n$ and
625: $N - n + 1$ are degenerate at $k =0$. The
626: effective $2 \times 2$ hamiltonian describing the bands near these degeneracy
627: points have off diagonal elements with a non trivial phase as in
628: eq.(\ref{bieff}). Hence, the degeneracy has topological charge $Q=2$, and it
629: cannot removed by perturbations compatible with the symmetries of the stack.
630: For even $n$, an explicit computation shows that these two bands
631: will have curvatures with the same sign near $k=0$. Thus, the corresponding
632: degeneracy points do not represent Fermi points. For odd numbered layers,
633: the two bands disperse in opposite directions away from $k=0$, and the
634: degeneracy points become stable Fermi points at the appropriate
635: doping. Note that
636: the symmetries of the system allow for a direct trigonal-like coupling
637: between the two layers, which will split this Fermi point and give rise
638: to four Fermi points showing linear dispersion, as in the bilayer. On
639: physical grounds, this coupling will be negligible, unless the two layers are
640: contiguous.
641:
642: The low energy bands in a stack with an odd number of layers also
643: contain doubly degenerate states at $k=0$, associated to the equivalence
644: between layers at opposite sides of the stack. But in this case the inversion $I$ is not part of
645: the symmetry group $D_{3h}$ of the stack, no invariance under $TI$ can be imposed and,
646: as a consequence, the first homotopy group $\pi_1$ is trivial.
647: This means that no conserved topological charge exists.
648: %Af finite values of $k$, the
649: %two bands are split by a coupling $\propto | k |^2$. Note that, unlike the
650: %case for even numbered stacks, this matrix element has no phase, indicating
651: %that the degeneracy point corresponds to zero winding number. For $N=3$, the spectrum near $k=0$
652: %contains, in addition, a band with linear dispersion.
653: %This band is associated
654: %to antisymmetric combinations of orbitals in layers adjacent to the central
655: %one.
656: %The effective effective hamiltonian is formally identical to
657: %the Dirac hamiltonian~(\ref{lin}), and has winding number one.
658: %However, the inversion $I$ is not a symmetry for $ABA$ stacking
659: %with odd $N$ and the $\pi_1$ of the
660: %effective hamiltonian is trivial.As a consequence, no conserved
661: %topological charge can be associated %to the winding number.
662: Hence, a gap
663: may open at $k=0$ when other perturbations consistent with the
664: symmetries of the stack are included. Concretely, a direct coupling between orbitals in
665: the same sublattice in layers separated by an odd number of other layers will
666: open a gap. Such a coupling, between layers which are second nearest
667: neighbors, has been proposed in graphite~\cite{DSM77}.
668:
669: %For $N=3$, the spectrum near $k=0$
670: %contains, in addition, a band with linear dispersion. This band is associated
671: %to antisymmetric combinations of orbitals in layers adjacent to the central
672: %one. The resulting effective hamiltonian is formally identical to the Dirac hamiltonian~(\ref{lin}), and
673: %has winding number one. hamiltonianpoint is unstable with respect to perturbations
674: %with break the equivalence between the two sublattices in these layers. These
675: %terms are compatible with the $D_{3h}$ symmetry group
676: %of an odd numbered stack, so that a
677: %gap will open in this Dirac like band.
678:
679: \section{ Conclusions.}
680: We have presented a classification of the
681: bands at low momenta and low energy of graphene layers and stacks
682: with many layers.
683:
684: Each Fermi point in single
685: layer graphene is stable against perturbations which preserve the discrete
686: $TI$ symmetry, and which do not mix the two Fermi points. A magnetic field,
687: for instance, induces a gap in the spectrum, see eq.~(\ref{mag_field}). This
688: term arises from the discreteness of the lattice, and it should be of higher
689: order than the minimal coupling which leads to the formation of Landau
690: levels. Combining this and dimensional arguments, we expect it to be $B_1
691: \propto v_F / l_B \times ( a / l_B )$, where $a$ is the lattice constant, and $l_B =
692: \sqrt{(e B ) / ( c \Phi_0 )}$ is the cyclotron radius. Thus, for $B \sim
693: 10$T, we have $B_1 \approx 0.1$meV.
694:
695: We have also classified the long wavelength perturbations commensurate with
696: the graphene lattice, which can hybridize the two $K$ points. Some of these
697: perturbations open a gap in the spectrum, while others shift the position of
698: the Dirac points. We expect that their strength will decay like a power law
699: with the wavelength of the distortion.
700:
701: The low energy and low momentum spectrum of stacks with many graphene layers
702: depend on the stacking order and the number of layers. For the most common
703: case of the Bernal stacking, we find that layer dependent onsite energies
704: lead to Fermi points with double degeneracy,
705: topological charge $Q= \pm 2$, and a parabolic dispersion in $k$. This
706: situation will be stable in stacks with a large (even) number of layers.
707: In stacks with an odd number of layers, there is no conserved topological charge and this
708: degeneracy will be broken by additional interactions. Stacks with
709: rhombohedral order lead to degenerate states with a large topological charge,
710: $Q=N$, which will give rise to the formation of a cascade of Fermi points
711: slightly away from $k=0$, with lower topological charges.
712:
713:
714: The most likely origin of the inequivalence between layers is the charge
715: accumulation at the layers close to the surface\cite{G06}. An induced doping of
716: $10^{10} - 10^{12}$ cm$^{-2}$ gives rise to shifts in the local potential of
717: $0.01 - 0.1$eV, so that the splittings associated to this effect can be
718: easily measurable. We find that a true gap opens, in the absence of an
719: external field which breaks spatial inversion, only in a stack with four
720: layers and Bernal stacking. Finally, stacking defects, which break
721: the equivalence between pairs of layers,
722: will also break the degeneracy of all bands at $k=0$.
723:
724:
725:
726: {\em Acknowledgments.}
727: It is a pleasure to thank J.J.~Manjar\y n and M.A.~Valle-Basagoiti
728: for useful discussions. M. A. H. V. and F. G. are thankful to the MEC (Spain) for
729: financial support through grant FIS2005-05478-C02-01
730: and the European Union Contract 12881 (NEST). J.L.M. has been
731: supported in part by the Spanish Science Ministry under Grant
732: FPA2005-04823. One of us (F. G.) also acknowledges support from the Comunidad
733: de Madrid, through the program CITECNOMIK, CM2006-S-0505-ESP-0337.
734:
735: \bibliography{PRB}
736:
737:
738:
739: \end{document}
740:
741:
742: Our topological stability arguments rely on the use of a
743: $2\times 2$ effective hamiltonian such
744: as~(\ref{bieff}) or ~(\ref{muleff}) to describe the lowest
745: energy bands degenerate at the Fermi point, but this
746: description breaks down at energies higher than the band
747: separations. As $N$ is increased we will have many
748: levels within a range of order the interlayer hopping $t$.
749: Level crossings will render the $2\times 2$ effective
750: hamiltonian useless, forcing us to include more bands.
751: But $n\times n$ hamiltonians with $n>2$ are likely to
752: have a trivial $\pi_1$ homotopy group even with $TI$
753: invariance~\cite{us06}, and stability may not be
754: guaranteed for very large values of $N$.
755:
756:
757: {\em Electronic structure and stability in graphene.}
758: The Fermi surface (FS) is a central concept in condensed matter
759: that controls the low-energy physics of the systems.
760: In a Landau Fermi liquid
761: at T = 0, J. Luttinger \cite{L60} defined the FS of an interacting Fermi
762: system in terms of
763: the single-particle Green's function $G({\bf k}, \omega)$, as the
764: solution of the equation
765: $G^{-1}(k, 0) = 0 $ and
766: showed that it encloses the same volume as in the noninteracting
767: system, equal to the fermion density $n$. The robustness of the Fermi
768: liquid idea has been understood recently in the context of
769: the renormalization group where the Fermi and Luttinger liquids are
770: seen as infrared fixed points\cite{P93,S94}.
771: In recent works\cite{V03,V06} Volovik has emphasized the idea of the
772: topological
773: stability of the Fermi surface as the origin of the robustness of the
774: Fermi liquid
775: and has suggested a classification of general fermionic systems in universality
776: classes dictated by momentum space topology. A more recent proposal relates
777: the stability of Fermi surfaces with K-theory, a tool used
778: to classify D-brane charges in string theory\cite{H05}. The idea behind
779: the topological
780: stability is
781: to study the zeroes of the matrix $G^{-1}_0({\bf k}, \omega)$ (free propagator)
782: that can not be lifted by small perturbations (self-energy induced by
783: interactions).
784: We will analyze the stability of the Fermi surface of single and multilayer
785: graphene where a principal role is played by the discrete symmetries of
786: the system.
787: