1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3: \documentclass[twocolumn,showpacs,aps,prb,preprintnumbers,amsmath,amssymb]{revtex4}
4: % Some other (several out of many) possibilities\begin{normalsize}
5: %\documentclass[preprint,aps]{revtex4}
6: %\documentclass[preprint,aps,draft]{revtex4}
7: %\documentclass[prb]{revtex4}% Physical Review B
8:
9: \usepackage{graphicx}% Include figure files
10: \usepackage{dcolumn}% Align table columns on decimal point
11: \usepackage{bm}% bold math
12: \usepackage{epsfig}
13:
14: \begin{document}
15:
16: %\preprint{APS/123-QED}
17:
18: \title{
19: Boundary-induced violation of the Dirac fermion parity and its signatures \\
20: in local and global tunneling spectra of graphene
21: }
22:
23: \author{Grigory Tkachov and Martina Hentschel}
24: \affiliation{
25: Max Planck Institute for the Physics of Complex Systems, Dresden, Germany}
26: %\date{\today}% It is always \today, today,
27: % but any date may be explicitly specified
28:
29: \begin{abstract}
30: Extended defects in graphene, such as linear edges, break the translational invariance and
31: can also have an impact on the symmetries specific to massless Dirac-like quasiparticles in this material.
32: The paper examines the consequences of a broken Dirac fermion parity
33: in the framework of the effective boundary conditions varying from
34: the Berry-Mondragon mass confinement to a zigzag edge.
35: The parity breaking reflects the structural sublattice asymmetry of zigzag-type edges and
36: is closely related to the previously predicted time-reversal symmetric edge states.
37: We calculate the local and global densities of the edge states and show that they carry a specific polarization,
38: resembling, to some extent, that of spin-polarized materials.
39: The lack of the parity leads to a nonanalytical particle-hole asymmetry in the edge-state properties.
40: We use our findings to interpret recently observed tunneling spectra in zigzag-terminated graphene.
41: We also propose a graphene-based tunneling device where the particle-hole asymmetric edge states result in
42: a strongly nonlinear conductance-voltage characteristics,
43: which could be used to manipulate the tunneling transport.
44: \end{abstract}
45:
46: \pacs{73.20.At,73.22.Gk,73.63.Bd}
47:
48: \maketitle
49:
50: \section{Introduction}
51: \label{Intro}
52:
53: In condensed matter systems with nodal fermionic spectra,
54: quantum description of low-energy excitations can resemble that of the ultrarelativistic electron.
55: The crystal space group imposes a generic restriction on such quasiparticles known as fermion doubling:
56: they come in pairs of opposite chirality species that can be mapped to
57: the conventional "right-handed" (RH) and "left-handed" (LH) fermions of the Dirac theory.~\cite{NN81}
58: The most recently studied examples are graphene,
59: where two distinct Fermi points in the Brillouin zone
60: give rise to both chiral species,~\cite{Semenoff84,Novoselov05}
61: and 2D HgTe quantum wells where spin-orbit coupling
62: effectively results in a pair of the RH and LH fermions at low energies.~\cite{Koenig07,Koenig08,Schmidt09}
63: The fermion doubling brings the symmetry with respect to the exchange of the chiralities, RH $\leftrightarrows$ LH,
64: related to the parity symmetry of the Dirac equation.~\cite{Itzykson}
65: It is of considerable interest to investigate the consequences of the violation of such a symmetry,
66: since they could be observable in materials where quasiparticles imitate Dirac electrons.
67: Besides, the Dirac fermion parity is distinct from other discrete symmetries
68: (e.g. time-reversal invariance), and, therefore, through its violation
69: one could achieve additional control over electronic properties of the material.
70:
71: It has been noticed~\cite{McCann04,Akhmerov08} that discrete symmetries
72: of a Dirac fermion system can be broken, along with the translational invariance,
73: by the boundaries of the system.
74: In this paper we focus on the parity violation due to such a boundary effect
75: and suggest how to detect and, possibly, use it in electronic devices.
76:
77: Our main assumption is that the boundary does not cause scattering
78: between the opposite chirality quasiparticles.
79: To model this we use effective boundary conditions,~\cite{McCann04,Akhmerov08}
80: interpolating between the infinite mass confinement~\cite{Berry87}
81: and the zigzag graphene boundary.~\cite{Fujita96,Armchair}
82: The parity breaking occurs as long as the boundary deviates from the infinite mass confinement toward
83: the zigzag edge. This is due to the structural sublattice asymmetry:
84: a zigzag-type crystal face has unequal numbers of sites from
85: the two sublattices of the honeycomb structure.~\cite{Fujita96,Akhmerov08}
86: More generally, the parity in this continuum model can cannot be preserved simultaneously
87: with the time-reversal invariance near the edge of the system.
88: This is closely related to the existence of time-reversal symmetric,
89: propagating edge states~\cite{Fujita96}
90: which have become a topic of vigorous graphene-related research
91: (see, e.g., Refs.~\onlinecite{Waka00,Koba05,Niimi06,Peres06,Sasaki06,Brey06,Rycerz07,GT07},\onlinecite{Akhmerov08,Wimmer08}).
92: The time-reversal symmetry requires that the edge states from the different valleys
93: propagate in the opposite directions, forming a Kramers pair at a given energy.
94: As a result, even in the absence of the intervalley scattering
95: the problem does not reduce to a single valley, in the sense
96: that the valley contributions to observables are not identical.
97: For the low-energy states (imitating the RH and LH fermions),
98: this implies broken Dirac fermion parity.
99:
100: We intend to demonstrate several new properties of the broken-parity edge states:
101:
102: {\bf (i)} nonanalytic particle-hole asymmetry of local and global densities of states,
103:
104:
105: {\bf (ii)} time-reversal invariant pseudospin polarization
106: (which in graphene is associated with the sublattice degree of freedom), and
107:
108:
109: {\bf (iii)} asymmetric nonlinear bias-voltage dependence of the tunneling conductance.
110: \newline{
111: In view of the progress in experimental control over graphene edges,~\cite{Girit09}
112: this material is particularly suitable to test our findings.
113: Below we discuss in more detail the connection between our results
114: and the ongoing graphene-related research.
115: }
116:
117: Our finding {\bf (i)} can be tested by means of scanning tunneling spectroscopy (STS) of the density of states (DOS).
118: In fact, some STS experiments~\cite{Koba05,Niimi06} have already
119: reported a particle-hole asymmetric DOS with a peak at -20-50 meV for monoatomic zigzag graphene edges.
120: By contrast Li et al.~\cite{Li08} have observed a symmetric linear DOS in graphene bulk.
121: In our model, the crossover from an asymmetric edge DOS to a symmetric bulk one follows naturally
122: from the existence of the broken-parity Dirac fermion edge states.
123: The results of experiments~\cite{Koba05,Niimi06} could therefore provide
124: some evidence for the Dirac fermion parity violation.
125: Such an interpretation is also supported by the observation
126: that the position of the DOS peak in the experiment of Niimi et al.~\cite{Niimi06}
127: can be described very accurately by our model.
128: This is achieved by taking into account not only the structural asymmetry
129: but also a potential-energy difference between the sublattices~\cite{Akhmerov08}
130: that generates weakly dispersive edge states rather than the singular zero-energy band.~\cite{Fujita96}
131: It was shown earlier~\cite{Sasaki06} that the next-nearest-neighbor hopping on the honeycomb lattice
132: could also result in a particle-hole asymmetric edge DOS.
133: However, this symmetry-breaking mechanism results in the DOS peak at significantly larger energies,
134: of the order of next-nearest-neighbor hopping energy $\approx 300$ meV.
135:
136: Another our result {\bf (ii)} demonstrates that broken-parity edge states carry
137: a time-reversal-invariant pseudospin polarization.
138: This agrees with the general perception that
139: in graphene electronic properties and those arising from
140: the sublattice degree of freedom (pseudospin) are interrelated.
141: However, there is a great deal of uncertainty as to how such a relation can be studied.
142: The specific feature of our edge problem is that it is possible to establish
143: a one-to-one correspondence between the pseudospin polarization and the edge DOS.
144: We suggest that the pseudospin polarization can be detected via measurements
145: of the electric conductance in lateral tunnel contacts between zigzag-terminated graphene
146: and a suitably chosen metallic electrode.
147:
148: Our proposal is based on finding {\bf (iii)} that
149: the edge-state (i.e. polarization-dependent) contribution to the conductance is
150: asymmetric with respect to the bias voltage.
151: Therefore, it can be separated from the symmetric contribution of the bulk graphene states.
152: In addition, the edge-state tunnel conductance turns out to be strongly nonlinear:
153: It exhibits kink-like switching as the sign of the voltage reverses.
154: Such a behavior could serve as a prototype for the potentially useful electronic functionality.
155:
156: The outline of the paper is as follows.
157: In Sec.~\ref{Model} we formulate the boundary problem for the Green's function
158: of the Dirac equation and discuss the role of the parity symmetry.
159: In Sec.~\ref{Local} we analyze the local DOS and pseudospin polarization,
160: and compare our results for the local DOS with the experimental data of Niimi et al.~\cite{Niimi06}
161: Section~\ref{Global} describes the relation between the global edge DOS and
162: pseudospin polarization and their tunneling spectroscopy.
163: The last section~\ref{Discussion} summarizes our results and
164: discusses their validity as well as possible applications.
165:
166:
167: \section{Dirac fermions in 2D semi-space: Broken parity and edge states}
168: \label{Model}
169:
170: Edge-state spectroscopy usually deals with isolated edges in large samples
171: where finite-size effects are presumably irrelevant.~\cite{Niimi06,Koba05,Girit09}
172: We model this by considering a boundary problem
173: for a Dirac fermion retarded Green's function in a 2D semi-space
174: $-\infty < x <\infty$, $0<y<\infty$:
175: \begin{eqnarray}
176: &
177: ( \varepsilon I - v\gamma^5\mbox{\boldmath$\Sigma$}\mbox{\boldmath$p$})
178: G_{\varepsilon}( {\bf r},{\bf r}^\prime )=
179: \delta( {\bf r}-{\bf r}^\prime ),
180: &
181: \label{Eq}
182: \end{eqnarray}
183: %
184: \begin{eqnarray}
185: &
186: G_{\varepsilon}=
187: \left( \frac{ I + \gamma^5 }{ 2 }\mbox{\boldmath$\Sigma$}{\bf n}_+ +
188: \frac{ I - \gamma^5 }{ 2 }\mbox{\boldmath$\Sigma$}{\bf n}_-
189: \right)
190: G_{\varepsilon}\big|_{y=0},\,\, {\bf n}^2_\pm =1,
191: &
192: \label{BC}
193: \end{eqnarray}
194: %
195: with $G|_{y\to\infty}$ being finite.
196: In Eq.~(\ref{Eq}) $\mbox{\boldmath$p$}=-i\hbar(\partial_x, \partial_y, 0)$,
197: $\varepsilon$ and $v$ are the 2D momentum operator, energy and velocity near a Fermi point.
198: In view of the further analysis of the parity symmetry, equations~(\ref{Eq}) and (\ref{BC}) are both expressed
199: in terms of the chirality, $\gamma^5$ and effective spin, $\mbox{\boldmath$\Sigma$}$
200: through the Dirac matrices:~\cite{Itzykson}
201: \begin{eqnarray}
202: &
203: \gamma^5=i\gamma^0\gamma^1\gamma^2\gamma^3=\tau^3\otimes\sigma^0,
204: \quad
205: \mbox{\boldmath$\Sigma$}=\gamma^5\gamma^0\mbox{\boldmath$\gamma$}=\tau^0\otimes\mbox{\boldmath$\sigma$},
206: &
207: \label{Spin}\\
208: &
209: \gamma^0=-\tau^1\otimes\sigma^0,\quad
210: \mbox{\boldmath$\gamma$}=i\tau^2\otimes\mbox{\boldmath$\sigma$}.
211: &
212: \label{Gamma}
213: \end{eqnarray}
214: %
215: We introduce the two sets of Pauli matrices, $\sigma^{1,2,3}$ and $\tau^{1,2,3}$,
216: and the corresponding unit matrices, $\sigma^0,\tau^0$ and $I=\tau^0\otimes\sigma^0$.
217: In graphene, $\sigma^{1,2,3}$ represent the two sublattices of the honeycomb structure,
218: while $\tau^{1,2,3}$ act in the valley space.
219: The eigenstates and eigenvalues ($\tau=\pm 1$) of the hermitean matrix
220: $\gamma^5$ conventionally define the right-handed (RH, $+$) and
221: left-handed (LH, $-$) quasiparticles~\cite{Itzykson}.
222: In Eqs.~(\ref{Eq}) and (\ref{BC}) they are described by the projected Green's functions
223: $\frac{ 1 }{ 2 }(I \pm \gamma^5)G_{\varepsilon}$.
224:
225:
226: The boundary condition, Eq.~(\ref{BC})
227: ensures vanishing of the particle current across the edge~\cite{McCann04,Akhmerov08}
228: (i.e. no Klein tunneling~\cite{Katsnelson06}).
229: It is diagonal in chirality space with the RH and LH blocks parametrized
230: by three-dimensional unit vectors, ${\bf n}_\tau={\bf n}_{\pm}$,
231: orthogonal to the boundary normal.~\cite{McCann04,Akhmerov08}
232: In graphene, where $\tau=\pm 1$ label the valleys,
233: the case of the zigzag edge corresponds to~\cite{Akhmerov08}
234: \begin{eqnarray}
235: {\bf n}_{+}=-{\bf n}_{-}={\bf\hat z},
236: \quad
237: G_{\varepsilon}=\gamma^5 \Sigma^3G_{\varepsilon}\big|_{y=0},
238: \label{Z}
239: \end{eqnarray}
240: %
241: where ${\bf\hat z}$ is the out-of-plane unit vector.
242: This implies that one of the sublattice Green's functions must vanish at the edge,
243: reflecting the structural sublattice asymmetry of the zigzag boundary.~\cite{Fujita96}
244: The other nontrivial limit is
245: \begin{eqnarray}
246: {\bf n}_{+}={\bf n}_{-}={\bf\hat x},
247: \quad
248: G_{\varepsilon}=\Sigma^1G_{\varepsilon}\big|_{y=0}.
249: \label{MC}
250: \end{eqnarray}
251: %
252: It is the infinite mass confinement of Berry and Mondragon~\cite{Berry87}
253: (${\bf\hat x}$ is the unit vector along the edge).
254: As shown in Ref.~\onlinecite{Akhmerov08},
255: the intermediate case, when ${\bf n}_{\pm}$ interpolate between (\ref{Z}) and (\ref{MC}),
256: can be treated as a zigzag edge where the {\em structural} sublattice asymmetry
257: coexists with a {\em potential} (energy) sublattice asymmetry
258: that could result from electron-electron interactions
259: within atomic distances near the edge.~\cite{Son06}
260: This does not however exhaust the applicability of the boundary condition (\ref{BC}),
261: since it can be derived from the only requirement that the Dirac particle current is zero
262: in the normal direction at the edge.~\cite{McCann04,Akhmerov08}
263:
264: Unless restricted to the infinite mass confinement case (\ref{MC}),
265: the boundary parameters are not identical, ${\bf n}_+\not ={\bf n}_-$,
266: which is verified below on the basis of the time-reversal ($\cal T$) symmetry [see, Eq.~(\ref{K})].
267: Therefore, boundary condition~(\ref{BC})
268: (and, in particular, Eq.~(\ref{Z})) explicitly contains the chirality, $\gamma^5$.
269: This violates the symmetry under the exchange of the RH and LH quasiparticles,
270: \begin{equation}
271: G_{\varepsilon}\to \gamma^0G_{\varepsilon}\gamma^0,
272: \label{P1}
273: \end{equation}
274: %
275: since it reverses the sign of $\gamma^5$
276: (In contrast, $\mbox{\boldmath$\Sigma$}$ is even under such operation).
277: On the other hand, the RH $\leftrightarrows$ LH exchange
278: is involved in the parity transformation,~\cite{Itzykson}
279: \begin{eqnarray}
280: G_{\varepsilon}({\bf r},{\bf r}^\prime)\to
281: \gamma^0 G_{\varepsilon}(-{\bf r},-{\bf r}^\prime) \gamma^0,
282: \label{P2}
283: \end{eqnarray}
284: %
285: and in the particle-hole conjugation,
286: \begin{eqnarray}
287: G_{\varepsilon}({\bf r},{\bf r}^\prime)
288: \to -\gamma^0 G_{-\varepsilon}({\bf r},{\bf r}^\prime) \gamma^0,
289: \label{P3}
290: \end{eqnarray}
291: %
292: both leaving the Dirac equation (\ref{Eq}) invariant.
293: Therefore, if boundary condition (\ref{BC}) deviates from the infinite mass confinement (\ref{MC}),
294: our boundary problem exhibits no parity invariance and, in view of Eq.~(\ref{P3}),
295: no particle-hole symmetry. The symmetry breaking persists in the limit of the zigzag edge, Eq.~(\ref{Z}).
296: We therefore conclude that the parity breaking is due to
297: the structural sublattice asymmetry of the zigzag-type lattice termination.~\cite{Fujita96}
298: For practical calculations, we need to take into account deviations of ${\bf n}_\pm$ from ${\bf\hat z}$
299: (e.g. due to potential-energy sublattice asymmetry~\cite{Akhmerov08}),
300: because it eliminates the Green's function singularity at $\varepsilon=0$
301: characteristic to dispersionless zero-energy edge states.
302:
303: The connection between the parity breaking and
304: the existence of the edge states
305: can be established by explicit calculation of the Green's function from Eqs.~(\ref{Eq}) and (\ref{BC}).
306: The calculation details are given elsewhere.~\cite{GT07}
307: Here we present the final result:
308: \begin{eqnarray}
309: &&
310: G_{\varepsilon}({\bf r},{\bf r}^\prime)=\sum_{\tau=\pm 1, k}
311: \left( \frac{I+\tau\gamma^5}{2} \right)
312: \left( I+\frac{\tau v}{\varepsilon}\mbox{\boldmath$\Sigma$}
313: \mbox{\boldmath$p$} \right)
314: \nonumber\\
315: &&
316: \times\left( G^{(0)}_{\varepsilon\tau k}(y,y^\prime)I +
317: G^{(3)}_{\varepsilon\tau k}(y,y^\prime)\Sigma^3
318: \right)
319: \frac{{\rm e}^{ik(x-x^\prime)} }{L},
320: \label{G}
321: \end{eqnarray}
322: %
323: \begin{eqnarray}
324: &&
325: G^{(0)}_{\varepsilon\tau k}(y,y^\prime)=\frac{\varepsilon}{2\hbar^2v^2q}
326: \left(
327: {\rm e}^{-q(y+y^\prime)} - {\rm e}^{ -q|y-y^\prime| }
328: \right)
329: \nonumber\\
330: &&
331: +
332: \frac{ q + k n_{z\tau} }{ 2( \varepsilon - \hbar v \tau k n_{x\tau} ) }
333: \,{\rm e}^{-q(y+y^\prime)},
334: \label{G0}
335: \end{eqnarray}
336: %
337: \begin{eqnarray}
338: &&
339: G^{(3)}_{\varepsilon\tau k}(y,y^\prime)=
340: \frac{k + q n_{z \tau} -\tau\varepsilon n_{x \tau}/\hbar v }
341: {2( \varepsilon - \hbar v \tau k n_{x \tau} )}
342: \,{\rm e}^{-q(y+y^\prime)},
343: \label{G3}
344: \end{eqnarray}
345: %
346: where $q=\sqrt{ k^2-\varepsilon^2/\hbar^2 v^2 }$ and $k$ is the wave number.
347: In Eqs.~(\ref{G0}) and (\ref{G3}) the edge states are described by the terms
348: with the pole at $\varepsilon=\hbar v \tau k n_{x\tau}$.
349: Let us examine, for instance, Eq.~(\ref{G3}) near the pole:
350: \begin{equation}
351: G^{(3)}_{\varepsilon \tau k}(y,y^\prime)\approx
352: -\frac{n_{z\tau}\Theta( kn_{z\tau} )}{\varepsilon - \hbar v \tau k n_{x\tau}}
353: \partial_y\,{\rm e}^{-|kn_{z\tau}|(y+y^\prime)}.
354: \label{G3_1}
355: \end{equation}
356: %
357: Clearly, the pole exists only if the unit step function $\Theta(kn_{z\tau})$
358: is nonzero, which determines the spectrum as
359: \begin{equation}
360: \varepsilon_{\tau k}=\hbar v \tau k n_{x\tau}, \quad\quad kn_{z\tau}>0.
361: \label{Spectrum}
362: \end{equation}
363: %
364: These equations are not yet restricted by the ${\cal T}$ symmetry.
365: The ${\cal T}$-symmetric spectrum follows from the condition that
366: both equations (\ref{Spectrum}) are invariant under the simultaneous reversal of
367: the chirality and wave-vector, $\tau, k \to -\tau, -k$.
368: This imposes the following restrictions on ${\bf n}_\tau$:
369: \begin{eqnarray}
370: n_{x\tau}=n_x, \quad n_{z\tau}=\tau n_z,
371: \quad {\bf n}=(n_x, 0, n_z),
372: \,\, {\bf n}^2=1,
373: \label{K}
374: \end{eqnarray}
375: %
376: leaving a single free boundary parameter -
377: the direction of the unit vector ${\bf n}$.
378: The edge-state spectrum is now manifestly Kramers degenerate and
379: particle-hole asymmetric:~\cite{Akhmerov08}
380: \begin{equation}
381: \varepsilon_{\tau k}=\hbar v \tau k n_x, \quad\quad \tau kn_z>0.
382: \label{Spectrum1}
383: \end{equation}
384: %
385: The role of the parity breaking is quite apparent from the behavior of
386: the edge-state Green's function~(\ref{G3_1}):
387: it vanishes identically for the infinite mass confinement ($n_z=0$)
388: which preserves the parity symmetry (see, also Eq.~(\ref{MC})).
389:
390: The knowledge of the spectrum (\ref{Spectrum1}) is not sufficient
391: to interpret the STS measurements,~\cite{Koba05, Niimi06}
392: as they provide information on the local DOS rather than dispersion $\varepsilon_{\tau k}$.
393: In the next section we use the full Green's function (\ref{G})
394: to calculate the local DOS of the system.
395: We will see that in addition to the exponentially localized states (\ref{G3_1})
396: there is another type of edge states decaying algebraically
397: as a consequence of the lack of the energy gap in the 2D bulk.
398: This distinguishes our system from, e.g., topological insulators
399: where bulk excitations are fully gapped.~\cite{Kane05,Koenig07,Koenig08}
400:
401: \section{Local DOS and pseudospin polarization}
402: \label{Local}
403:
404: \subsection{Particle-hole symmetry and the role of parity}
405: \label{Parity}
406:
407: The spectral and pseudospin properties of the system are interrelated.
408: Let us define the local DOS
409: \begin{equation}
410: \nu_\pm(\varepsilon,{\bf r})=-\frac{1}{2\pi}{\rm Im Tr}\,
411: (I \pm \gamma^5)G_{\varepsilon}({\bf r},{\bf r}),
412: \label{nu_pm_def}
413: \end{equation}
414: %
415: and local pseudospin polarizations
416: \begin{equation}
417: p_\pm(\varepsilon,{\bf r})=-\frac{1}{2\pi}{\rm Im Tr}\,
418: (I \pm \gamma^5)\Sigma^3 G_{\varepsilon}({\bf r},{\bf r})
419: \label{p_pm_def}
420: \end{equation}
421: %
422: in terms of the RH and LH projections of the Green's function, $G_\varepsilon$.
423: Tunneling spectra are determined by the total local DOS related to $G^{(0)}$ in Eq.~(\ref{G0}):
424: \begin{eqnarray}
425: &
426: \nu(\varepsilon,{\bf r})=\nu_+ + \nu_-=
427: -\frac{2}{\pi L}\sum\limits_{\tau=\pm 1,k}
428: {\rm Im}G^{(0)}_{\varepsilon\tau k}(y,y).
429: &
430: \label{nu_def}
431: \end{eqnarray}
432: %
433: Likewise, $p_+ + p_-=-\frac{1}{\pi}{\rm Im Tr}\,\Sigma^3 G_{\varepsilon}$ is the net pseudospin polarization.
434: It vanishes by ${\cal T}$ symmetry, since Eqs.~(\ref{K}) yield $p_-= - p_+$.
435: As a $\cal T$-invariant characteristic of the pseudospin properties,
436: we use the chiral pseudospin polarization (CPP) related to $G^{(3)}$ in Eq.~(\ref{G3}):
437: \begin{eqnarray}
438: &
439: p(\varepsilon,{\bf r})=p_+-p_-=-\frac{1}{\pi}{\rm Im Tr}\,\gamma^5\Sigma^3 G_{\varepsilon}({\bf r},{\bf r})=
440: &
441: \nonumber\\
442: &
443: =-\frac{2}{\pi L}\sum\limits_{\tau=\pm 1,k}\tau {\rm Im}G^{(3)}_{\varepsilon\tau k}(y,y).
444: &
445: \label{p_def}
446: \end{eqnarray}
447: %
448: Integrating over $k$ in Eqs.~(\ref{nu_def}) and (\ref{p_def}),
449: we obtain
450: \begin{eqnarray}
451: &&
452: \nu(\varepsilon,y)=\frac{ 2|\varepsilon| }{ \pi \hbar^2 v^2 } -
453: \sum_{\tau=\pm 1} \frac{ \Theta\left( \varepsilon\, \tau n_{x\tau} n_{z\tau} \right)
454: }{h v |n_{x\tau}| }
455: \partial_y
456: {\rm e}^{
457: -\frac{ 2y }{ \hbar v } \left|\varepsilon \frac{ n_{z\tau} }{ n_{x\tau} } \right|
458: }
459: \nonumber\\
460: &&
461: -\frac{ |\varepsilon|}{ \pi^2\hbar^2v^2 }
462: \sum_{\tau=\pm 1}
463: \int_0^{ \frac{\pi}{2} } d\gamma\times
464: \label{nu}\\
465: &&
466: \times\frac{
467: n^2_{z\tau}\cos\left( \frac{2\varepsilon y}{\hbar v } \sin\gamma \right)
468: +\tau n_{x\tau}n_{z\tau}\sin\gamma\sin\left( \frac{2\varepsilon y}{\hbar v } \sin\gamma \right)
469: }
470: { n^2_{z\tau}+\tan^2\gamma },
471: \nonumber
472: \end{eqnarray}
473: %
474: \begin{eqnarray}
475: &&
476: p(\varepsilon,y)=-
477: \sum_{\tau=\pm 1}
478: \frac{ \tau n_{z\tau}\Theta\left( \varepsilon\, \tau n_{x\tau} n_{z\tau} \right)
479: }{h v |n_{x\tau}| }
480: \partial_y {\rm e}^{
481: -\frac{ 2y }{\hbar v} \left|\varepsilon \frac{ n_{z\tau} }{ n_{x\tau} } \right|
482: }
483: \nonumber\\
484: &&
485: +\frac{|\varepsilon|}{\pi^2\hbar^2v^2}
486: \sum_{\tau=\pm 1}
487: \int_0^{ \frac{\pi}{2} }d\gamma \tan^2\gamma
488: \times
489: \label{p}\\
490: &&
491: \frac{ \tau n_{z\tau}\cos\left( \frac{2\varepsilon y}{\hbar v } \sin\gamma \right)
492: +n_{x\tau}\sin\gamma\sin\left( \frac{2\varepsilon y}{\hbar v } \sin\gamma \right) }
493: { n^2_{z\tau}+\tan^2\gamma }.
494: \nonumber
495: \end{eqnarray}
496: %
497:
498: It is now easy to see that the particle-hole symmetry is controlled by the parity.
499: In the broken-parity state with ${\bf n}_+\not={\bf n}_-$ given by Eqs.~(\ref{K}),
500: the summation over chiralities $\tau=\pm 1$ in Eq.~(\ref{nu}) yields
501: an asymmetric DOS as a function of energy, $\varepsilon$:
502:
503: \begin{eqnarray}
504: &&
505: \nu(\varepsilon,y)=\frac{ 2|\varepsilon| }{ \pi \hbar^2 v^2 }-
506: \frac{ 2\Theta\left( \varepsilon\, n_{x} n_{z} \right)
507: }{h v |n_{x}| }
508: \partial_y
509: {\rm e}^{
510: -\frac{ 2y }{ \hbar v } \left|\varepsilon \frac{ n_{z} }{ n_{x} } \right|
511: }
512: \nonumber\\
513: &&
514: -\frac{ 2|\varepsilon|}{ \pi^2\hbar^2v^2 }
515: \int_0^{ \frac{\pi}{2} } d\gamma\times
516: \label{nu_1}\\
517: &&
518: \times\frac{
519: n^2_{z}\cos\left( \frac{2\varepsilon y}{\hbar v } \sin\gamma \right)
520: +n_{x}n_{z}\sin\gamma\sin\left( \frac{2\varepsilon y}{\hbar v } \sin\gamma \right)
521: }
522: { n^2_{z}+\tan^2\gamma },
523: \nonumber
524: \end{eqnarray}
525: %
526: The same is true for the CPP (\ref{p}):
527: %
528: \begin{eqnarray}
529: &&
530: p(\varepsilon,y)=-
531: \frac{ 2 n_{z}\Theta\left( \varepsilon\, n_{x} n_{z} \right)
532: }{h v |n_{x}| }
533: \partial_y {\rm e}^{
534: -\frac{ 2y }{\hbar v} \left|\varepsilon \frac{ n_{z} }{ n_{x} } \right|
535: }
536: \nonumber\\
537: &&
538: +\frac{2|\varepsilon|}{\pi^2\hbar^2v^2}
539: \int_0^{ \frac{\pi}{2} }d\gamma \tan^2\gamma
540: \times
541: \label{p_1}\\
542: &&
543: \frac{ n_{z}\cos\left( \frac{2\varepsilon y}{\hbar v } \sin\gamma \right)
544: +n_{x}\sin\gamma\sin\left( \frac{2\varepsilon y}{\hbar v } \sin\gamma \right) }
545: { n^2_{z}+\tan^2\gamma },
546: \nonumber
547: \end{eqnarray}
548: %
549: For comparison, if the parity is preserved for ${\bf n}_+={\bf n}_-$,
550: the DOS (\ref{nu}) appears to be an even function of energy:
551:
552: \begin{eqnarray}
553: &&
554: \nu(\varepsilon,y)=\frac{ 2|\varepsilon| }{ \pi \hbar^2 v^2 }-
555: \frac{ 1 }{ h v |n_{x}| }
556: \partial_y
557: {\rm e}^{
558: -\frac{ 2y }{ \hbar v } \left|\varepsilon \frac{ n_{z} }{ n_{x} } \right|
559: }
560: \nonumber\\
561: &&
562: -\frac{ 2|\varepsilon|n^2_{z}}{ \pi^2\hbar^2v^2 }
563: \int_0^{ \frac{\pi}{2} } d\gamma
564: \frac{
565: \cos\left( \frac{2\varepsilon y}{\hbar v } \sin\gamma \right)
566: }
567: { n^2_{z}+\tan^2\gamma },
568: \label{nu_2}
569: \end{eqnarray}
570: %
571: However, the requirements for the parity symmetry, ${\bf n}_+={\bf n}_-$,
572: are incompatible with conditions~(\ref{K}) for the ${\cal T}$ symmetry.
573: The only exception is the infinite mass confinement limit $n_z\to 0$.
574: In literature,~\cite{Fujita96,Son06,Wimmer08,Yazyev08}
575: ${\cal T}$-symmetry breaking on zigzag graphene edges has been discussed
576: in connection with their possible intrinsic magnetism.
577: It is still unclear whether the ${\cal T}$-symmetry breaking in the boundary condition, Eq.~(\ref{BC})
578: has anything to do with the edge magnetism.
579: We will therefore limit our analysis to the ${\cal T}$-symmetric case (\ref{K}).
580:
581: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
582: %%%%%%%%%%%%%%%%% LDOS %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
583: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
584: \begin{figure}[t]
585: \begin{center}
586: \epsfxsize=1\hsize
587: \epsffile{nu.eps}
588: \end{center}
589: \caption{(Color online) Local density of states in units of $3/4\pi$ eV$^{-1}$nm$^{-2}$:
590: {\bf (a)} vs. energy for different $n_z$ at $y=4$ nm;
591: {\bf (b)} vs. energy at different positions for $n_z=0.95$;
592: {\bf (c)} vs. position for opposite-sign energies and $n_z=0.95$.
593: Inset: local STS geometry and orientation of the unit vector ${\bf n}$
594: [Eq.~(\ref{K})] determining the boundary condition (\ref{BC}).
595: {\bf (d)} Asymmetric local DOS
596: $\Delta \nu\equiv\nu(\varepsilon)-\nu(-\varepsilon)$
597: vs. energy at $y=5$ nm for $n_z=0.95$.
598: The data are for $v=10^6$ ms$^{-1}$ and $n_x<0$.
599: }
600: \label{LDOS}
601: \end{figure}
602: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
603: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
604:
605: \subsection{Energy and position dependence of the local DOS: Analysis}
606: \label{Asymmetry}
607:
608: In Fig.~\ref{LDOS} we plot the local DOS~(\ref{nu_1})
609: as a function of energy $\varepsilon$ (in eV) and position $y$ (in nm)
610: for the Fermi velocity $v=10^6$ms$^{-1}$.
611: These units and parameters are typical for STS in graphene.
612: Panel~(a) shows an asymmetric peak, due to the edge states,
613: emerging on top of the linear DOS as the boundary condition varies from
614: the Berry-Mondragon type ($n_z=0$) to the zigzag type ($|n_z|\to 1$).
615: In the latter case, the DOS~(\ref{nu_1}) still fails
616: to recover the particle-hole symmetry because of the broken parity [see, Eq.~(\ref{Z})],
617: which is formally described by the singular energy-dependent factor $\Theta(\varepsilon n_xn_z)$.
618: The crossover between the Berry-Mondragon and zigzag cases
619: can in principle be induced by a staggered mean-field sublattice potential
620: whose strength is parametrized by
621: the angle between ${\bf n}$ and ${\bf\hat z}$.~\cite{Akhmerov08}
622:
623:
624: For $|n_z|\to 1$ the edge modes are much slower than the bulk ones,
625: and have a small characteristic energy, $\approx\hbar v/2y\times |n_x/n_z|$.
626: This can explain the observed DOS asymmetry on the scales of $20-50$ meV~\cite{Niimi06}.
627: For $n_z=0.95$, the peak position, $\varepsilon\approx 25$ meV,
628: and its overall behavior [panel~(b)] agree very well with
629: the observations (see, e.g., Fig.~5 in Ref.~\onlinecite{Niimi06}).
630: As we neglect possible level broadening,
631: the peak looks somewhat higher and narrower than in the experiment.
632: Also, in agreement with the findings of Li et al.~\cite{Li08},
633: $\nu(\varepsilon)$ approaches the symmetric Dirac DOS away from the edge.
634: The position dependence of the DOS
635: [panel~(c)] shows that at the edge
636: $\nu$ reaches either a maximum or a minimum
637: depending on the presence or absence of the exponential term in
638: Eq.~(\ref{nu_1}), which is controlled only by the sign of $\varepsilon$.
639: The Dirac waves, incident from the bulk and reflected from the edge, interfere
640: yielding an oscillatory contribution (third term) in the DOS~(\ref{nu_1}),
641: decaying as $y^{-3/2}$ on the scale of 10-20 nm.
642: Panel~(d) demonstrates similar oscillations in the energy dependence.
643:
644: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
645: %%%%%%%%%%%%%% p %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
646: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
647: \begin{figure}[t]
648: \begin{center}
649: \epsfxsize=1\hsize
650: \epsffile{p.eps}
651: \end{center}
652: \caption{(Color online)
653: Chiral pseudospin polarization
654: {\bf (a)} vs. position
655: and
656: {\bf (b)} vs. energy for $n_z=0.95$ and $n_x<0$.
657: }
658: \label{pFig}
659: \end{figure}
660: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
661: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
662:
663: \subsection{ Local chiral pseudospin polarization }
664: \label{CPP}
665:
666: To conclude the analysis of the local properties,
667: in Fig.~\ref{pFig} we plot the CPP given by Eq.~(\ref{p_1}).
668: From Fig.~\ref{pFig}(a) we see that the CPP has a purely boundary origin
669: as it decays to zero in the bulk.
670: Apart from the presence of the oscillations,
671: both position and energy dependences of the local CPP
672: differ significantly from the corresponding behaviors of the local DOS
673: [cf. Figs.~\ref{pFig} and~\ref{LDOS}(c,d)].
674: Although not obvious in the local quantities
675: $\nu(\varepsilon,y)$ and $p(\varepsilon,y)$,
676: in the next section we establish a direct relation
677: between the appropriately defined global CPP and DOS.
678:
679:
680:
681:
682: \section{ Global edge DOS and pseudospin polarization }
683: \label{Global}
684:
685: \subsection{Relation between the edge DOS and pseudospin polarization}
686: \label{Relation}
687:
688: Let us define the DOS and CPP of a finite region of space, $0\leq y \leq w$,
689: as the following dimensionless integrals,
690: \begin{eqnarray}
691: &&
692: N_e(\varepsilon,w)=hv\int_0^w dy\, \left( \nu(\varepsilon,y) - \frac{ 2|\varepsilon| }{ \pi \hbar^2 v^2 } \right),
693: \label{Ne_def}\\
694: &&
695: P(\varepsilon,w)=hv\int_0^w dy\, p(\varepsilon,y).
696: \label{P_def}
697: \end{eqnarray}
698: %
699: In the first equation we subtract the bulk Dirac DOS,
700: so that $N_e(\varepsilon,w)$ contains the contribution of the edge states only.
701: By contrast to their local counterparts, it is convenient to call
702: $N_e(\varepsilon,w)$ and $P(\varepsilon,w)$
703: the global edge DOS and global CPP, respectively.
704:
705: Inserting Eqs.~(\ref{nu_1}) and (\ref{p_1}) into Eqs.~(\ref{Ne_def}) and (\ref{P_def})
706: and integrating over position $y$, we find
707: \begin{eqnarray}
708: &&
709: N_e(\varepsilon)=
710: \frac{ 2\Theta\left( \varepsilon\, n_{x} n_{z} \right)
711: }{|n_{x}| }
712: \left[
713: 1-{\rm e}^{
714: -\frac{ 2w }{ \hbar v } \left|\varepsilon \frac{ n_{z} }{ n_{x} } \right|
715: }
716: \right]
717: -\frac{ 2 n_z{\rm sgn}\,\varepsilon }{ \pi }
718: \label{Ne}\\
719: &&
720: \times \int\limits_0^{ \frac{\pi}{2} } d\gamma
721: \left[
722: \frac{n_z}{ \sin\gamma}
723: \frac{
724: \sin\left( \frac{2\varepsilon w}{\hbar v } \sin\gamma \right)
725: }
726: { n^2_{z}+\tan^2\gamma }
727: + n_{x}
728: \frac{
729: 1-\cos\left( \frac{2\varepsilon w}{\hbar v } \sin\gamma \right)
730: }
731: { n^2_{z}+\tan^2\gamma }
732: \right],
733: \nonumber
734: \end{eqnarray}
735: %
736: \begin{eqnarray}
737: &&
738: P(\varepsilon)=
739: \frac{ 2n_z\Theta\left( \varepsilon\, n_{x} n_{z} \right)
740: }{|n_{x}| }
741: \left[
742: 1-{\rm e}^{
743: -\frac{ 2w }{ \hbar v } \left|\varepsilon \frac{ n_{z} }{ n_{x} } \right|
744: }
745: \right]
746: + \frac{ 2{\rm sgn}\,\varepsilon }{ \pi }
747: \label{P}\\
748: &&
749: \times \int\limits_0^{ \frac{\pi}{2} } d\gamma
750: \left[
751: \frac{n_z}{ \sin\gamma}
752: \frac{
753: \sin\left( \frac{2\varepsilon w}{\hbar v } \sin\gamma \right)
754: }
755: { 1 + n^2_{z}\cot^2\gamma }
756: + n_{x}
757: \frac{
758: 1-\cos\left( \frac{2\varepsilon w}{\hbar v } \sin\gamma \right)
759: }
760: { 1 + n^2_{z}\cot^2\gamma }
761: \right].
762: \nonumber
763: \end{eqnarray}
764: %
765: It is instructive to discuss first the limit $w\to\infty$, when the integrals in Eqs.~(\ref{Ne})
766: and (\ref{P}) can be evaluated analytically.
767: In this case the integrals with the rapidly oscillating cosine function vanish,
768: while those containing the sine function should be evaluated with care since
769: the ratio $\sin\left( \frac{2\varepsilon w}{\hbar v } \sin\gamma \right)/\sin\gamma $
770: becomes singular, $\pi\, {\rm sgn}(\varepsilon)\, \delta( \sin\gamma )$ as $w\to\infty$.
771: After integrating with the delta function $\delta( \sin\gamma )$, we have
772: \begin{eqnarray}
773: &&
774: N_e(\varepsilon) =
775: \frac{ 2\Theta\left( \varepsilon\, n_{x} n_{z} \right)
776: }{|n_{x}| }
777: -1
778: - \frac{ 2 n_x n_z {\rm sgn}\,\varepsilon }{ \pi }
779: \int\limits_0^{ \frac{\pi}{2} }
780: \frac{
781: d\gamma
782: }
783: { n^2_{z}+\tan^2\gamma },
784: \nonumber\\
785: &&
786: P(\varepsilon) =
787: \frac{ 2n_z\Theta\left( \varepsilon\, n_{x} n_{z} \right)
788: }{|n_{x}| }
789: + \frac{ 2 n_x {\rm sgn}\,\varepsilon }{ \pi }
790: \int\limits_0^{ \frac{\pi}{2} }
791: \frac{
792: d\gamma
793: }
794: { 1 + n^2_{z}\cot^2\gamma } .
795: \nonumber
796: \end{eqnarray}
797: %
798: The remaining integrals are easy to evaluate~\cite{Integrals}.
799: The results are
800: \begin{eqnarray}
801: N_e(\varepsilon)=\frac{ 1 + n_z {\rm sgn}(\varepsilon\, n_x) }{ |n_x| } -1,
802: \label{Ne_1}
803: \end{eqnarray}
804: %
805: \begin{eqnarray}
806: P(\varepsilon)=\frac{ {\rm sgn}(\varepsilon\, n_x) + n_z}{|n_x|},
807: |P(\varepsilon)|=\frac{ 1 + n_z {\rm sgn}(\varepsilon\, n_x) }{|n_x|}.
808: \label{P_1}
809: \end{eqnarray}
810: %
811: From Eqs.~(\ref{Ne_1}) and (\ref{P_1}) we find the relation between $N_e$ and $P$,
812: \begin{eqnarray}
813: N_e(\varepsilon)=| P(\varepsilon) | - 1.
814: \label{DOS}
815: \end{eqnarray}
816: %
817: Being a scalar, $N_e$ does not depend on the sign of $P$ which is
818: reversed under the transformation ${\bf n}\to -{\bf n}$.
819:
820: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
821: %%%%%%%%%%%%%% Corr %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
822: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
823: \begin{figure}[t]
824: \begin{center}
825: \epsfxsize=0.7\hsize
826: \epsffile{Corr.eps}
827: \end{center}
828: \caption{(Color online)
829: Typical behavior of the global edge DOS, $N_e$ and
830: chiral pseudospin polarization, $P$ as functions of the width, $w$
831: (see, Eqs.~(\ref{Ne_def}) -- (\ref{P}) ).
832: }
833: \label{Corr}
834: \end{figure}
835: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
836: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
837:
838: According to Eq.~(\ref{DOS}), for the zigzag graphene edge $|n_z|\to 1 (n_x\to 0)$
839: the absolute value of the CPP becomes equal to the edge DOS:
840: \begin{eqnarray}
841: N_e(\varepsilon)\approx | P(\varepsilon) |\approx \frac{ 2\Theta( \varepsilon n_xn_z) }{ |n_x| }.
842: \label{Ne_2}
843: \end{eqnarray}
844: %
845: This means that the CPP can, in principle, be detected through measurements of the global edge DOS.
846: The latter, in turn, can be probed by tunneling, as we discuss in the next subsection.
847: Before going to that question, we wish to point out that the correlation between $P$ and $N_e$
848: exists for finite values of $w$ as well.
849: Figure~\ref{Corr} shows the functions $N_e(w)$ [Eq.~(\ref{Ne})] and $P(w)$ [Eq.~(\ref{P})],
850: clearly approaching the relation~(\ref{DOS}) for $w\geq 100$ nm.
851: Note that the non-oscillatory components of $P$ and $N_e$ obey the relation~(\ref{DOS})
852: at much smaller $w$.
853:
854:
855:
856: \subsection{Tunneling spectroscopy}
857: \label{Non}
858:
859: Our proposal for tunneling spectroscopy of the global edge DOS exploits
860: the particle-hole asymmetric nonanalytic energy dependence of $N_e(\varepsilon)$
861: [Eqs.~(\ref{Ne_1}) and (\ref{Ne_2})].
862: It is essential that the particle-hole asymmetry persists in the case of zigzag-terminated graphene
863: ($|n_z|\to 1, n_x\to 0$) because this is an experimentally accessible system.
864:
865: It is known~\cite{Giaever60,Mahan} that a strongly energy-dependent DOS reflects in
866: the differential electric conductance, ${\rm g}(V)$ of a tunnel junction between the system of interest
867: and a metal where the DOS is almost constant near the Fermi energy.
868: Here we consider a lateral tunnel contact between a zigzag-terminated graphene sheet and
869: a metallic film, as shown in Fig.~\ref{Tunnel}.
870: It is assumed that the voltage drop, $V$, occurs predominantly across
871: the tunnel barrier in the contact area, which determines the junction resistance.
872: Under such condition, the conductance can be calculated using the tunneling Hamiltonian approach,
873: which is well described in the literature (e.g. Ref.~\onlinecite{Mahan}),
874: with the following result:
875: \begin{eqnarray}
876: &&
877: {\rm g}(V,T)={\rm g}_0 \int_{-\infty}^{\infty}
878: d\varepsilon\, N(\varepsilon) \frac{\partial f(\varepsilon -eV,T)}{\partial (eV)},
879: \label{g}
880: \end{eqnarray}
881: %
882: Here $N(\varepsilon)$ is the DOS of graphene in the contact region (dark gray area in Fig.~\ref{Tunnel}),
883: $f(\varepsilon -eV,T)$ is the Fermi-Dirac distribution of the tunneling quasiparticles
884: at voltage $V$ and temperature $T$, and the constant ${\rm g}_0$ absorbs the
885: energy-independent parameters of the metal and tunnel barrier.
886: As we are interested in the low energy regime $V,T\to 0$, in Eq.~(\ref{g}) we can neglect
887: inelastic tunneling processes (e.g. phonon emission).~\cite{Mahan}
888:
889: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
890: %%%%%%%%%%%%%% Tunnel %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
891: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
892: \begin{figure}[t]
893: \begin{center}
894: \epsfxsize=1\hsize
895: \epsffile{Tunnel.eps}
896: \end{center}
897: \caption{(Color online)
898: Suggested tunneling device for determining the density of the edge states in graphene.
899: A metallic film (M) is deposited on top of a zigzag-terminated graphene sheet (G)
900: forming a strip-like lateral contact of width, $w$.
901: The device resistance is assumed to be determined by the tunnel barrier (dark gray area)
902: so that the voltage drop $V$ predominantly occurs between the overlapping parts of M and G.
903: }
904: \label{Tunnel}
905: \end{figure}
906: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
907: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
908:
909:
910:
911:
912:
913:
914: The DOS $N(\varepsilon)$ contains the contributions of both bulk and edge states.
915: Since the bulk DOS is a symmetric function of energy ($\propto |\varepsilon|$),
916: it can be eliminated by taking the difference:
917: \begin{eqnarray}
918: &&
919: \Delta {\rm g}(V,T) = {\rm g}(V,T)-{\rm g}(-V,T)=
920: \nonumber\\
921: &&
922: ={\rm g}_0\int_{-\infty}^{\infty}
923: d\varepsilon\, [N_e(-\varepsilon)-N_e(\varepsilon)]\frac{\partial f(\varepsilon -eV,T)}{\partial\varepsilon}.
924: \label{Delta_g}
925: \end{eqnarray}
926: %
927: It contains only the particle-hole asymmetic edge DOS, $N_e(\varepsilon)$, given by Eq.~(\ref{Ne}),
928: where $w$ coincides with the width of the lateral tunnel contact [see, Fig.~\ref{Tunnel}].
929:
930: In the limit $w\to\infty$, we use Eqs.~(\ref{P_1}) and ~(\ref{DOS}) to evaluate the integral in Eq.~(\ref{Delta_g}):
931: \begin{eqnarray}
932: &&
933: \Delta {\rm g}(V,T)={\rm g}_0\, \Delta_P\,\tanh\frac{ eV }{ 2k_BT },
934: \label{Cond}\\
935: &&
936: \Delta_P=|P|_{\varepsilon>0}-|P|_{\varepsilon<0}=2\frac{n_z}{n_x}.
937: \label{Delta_P}
938: \end{eqnarray}
939: %
940: The conductance asymmetry $\Delta {\rm g}(V,T)$ reflects
941: the nonequilibrium quasiparticle accumulation that builds up
942: near the graphene edge in response to the current flow between the systems.
943: For $|eV|>2k_BT$ the conductance $\Delta {\rm g}(V,T)$ saturates at $\pm {\rm g}_0\Delta_P$,
944: where $\Delta_P$ is the difference in the absolute values of the CPP for
945: the positive- and negative-energy edge states.
946: Such a nonlinear behavior can be used to detect the edge state as well as
947: the existence of the pseudospin polarization.
948: At zero temperature $T=0$ the voltage dependence in Eq.~(\ref{Cond})
949: becomes singular ($\propto {\rm sgn}(eV)$).
950: This is specific to the $w=\infty$ limit.
951: As shown in Fig.~\ref{Dg}, the singularity is smeared due to finiteness of the contact width, $w$,
952: so that $\Delta {\rm g}(V,0)$ saturates at voltages larger than the value $\propto w^{-1}$.
953: The data in Fig.~\ref{Dg} are obtained by numerical integration of
954: Eqs.~(\ref{Delta_g}) and (\ref{Ne}) at $T\to 0$.
955:
956:
957:
958: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
959: %%%%%%%%%%%%%% Delta g %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
960: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
961: \begin{figure}[t]
962: \begin{center}
963: \epsfxsize=0.6\hsize
964: \epsffile{Dg.eps}
965: \end{center}
966: \caption{(Color online)
967: Zero-temperature conductance $\Delta {\rm g}={\rm g}(V)-{\rm g}(-V)$ vs. voltage (in volts)
968: for different widths of the tunnel contact, $w$ [see, also Fig.~\ref{Tunnel}].
969: }
970: \label{Dg}
971: \end{figure}
972: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
973: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
974:
975: \section{Summary and discussion}
976: \label{Discussion}
977:
978: We have considered the boundary problem for 2D Dirac fermions, Eq.~(\ref{Eq}) and (\ref{BC}),
979: in which the time-reversal invariance is preserved at the expense of the Dirac fermion parity.
980: Using the Green's function solution, we have shown that the broken parity
981: manifests itself in the density of the edge states and their pseudospin polarization,
982: both exhibiting a nonanalytic particle-hole asymmetry.
983: The zigzag graphene edge with its inherent structural asymmetry is an example of
984: the realization of the Dirac fermion parity breaking.
985: Taking into account additionally the potential-energy sublattice asymmetry near the zigzag edge~\cite{Akhmerov08},
986: we obtain the local DOS consistent with the experimental data of Niimi et al.~\cite{Niimi06}
987: We have also established a direct correspondence between the
988: pseudospin polarization and the density of the edge states,
989: and suggested how to detect them in a tunneling experiment.
990: The proposal relies on the broken particle-hole symmetry
991: resulting in an asymmetric nonlinear contribution to the conductance, Eq.~(\ref{Cond}),
992: in a tunnel junction between zigzag-terminated graphene and a metallic film [see, Fig.~\ref{Tunnel}].
993:
994: It is interesting to discuss possible implementations of the strong nonlinearity of the conductance, Eq.~(\ref{Cond}).
995: We suggest that it could be used for detecting weak electric signals and their polarity.
996: The operation of such a device would exploit the two different states of the tunnel junction
997: corresponding to the conductance values at positive and negative bias voltages [see, also, Fig.~\ref{Dg}].
998: Let us assume that the system is initially in one of these states.
999: Then, under externally induced change in the bias voltage
1000: the system can switch into the state with the other (lower or higher) value of the conductance.
1001: For the zigzag graphene edge ($|n_z|\to 1, n_x\to 0$),
1002: the conductance difference, Eqs.~(\ref{Cond}) and (\ref{Delta_P}), is very significant and, therefore, should be detectable.
1003: The progress in characterization of graphene edges~\cite{Girit09} may eventually lead to more understanding
1004: of how the interfaces needed to test our finding can be fabricated.
1005:
1006: We noticed that even though the conductance, Eq.~(\ref{Cond}) vanishes
1007: for the Berry-Mondragon confinement~\cite{Berry87} ($n_z\to 0$),
1008: this case is still nontrivial, because the pseudospin polarization
1009: $P(\varepsilon)$ (\ref{DOS}) is not zero:
1010: $P(\varepsilon)={\rm sgn}(\varepsilon n_x)$.
1011: This is confirmed by the numerical data for $n_z=0.05$ in Fig.~\ref{Corr}.
1012: The nonvanishing $P(\varepsilon)$ comes from the oscillatory (interference) term in Eq.~(\ref{p}),
1013: decaying as $y^{-1/2}$ on distances 50-100 nm [see, Fig.~\ref{pFig}(a)].
1014: For a given chirality (valley), the long-range polarization implies violation of
1015: the ${\cal T}$ symmetry on mesoscopic scales,
1016: which may have some connection to recent studies of the level statistics in graphene quantum dots.~\cite{Ponomarenko08,Wurm09}
1017: Also, such long-range polarization may coexist with magnetic correlations predicted for
1018: zigzag graphene edges,~\cite{Fujita96,Son06,Wimmer08,Yazyev08}
1019: since they are expected to decay on much shorter distances.~\cite{Yazyev08}
1020:
1021:
1022: We thank H. Baranger, F. Guinea, M. I. Katsnelson and A. D. Mirlin for discussions.
1023: The work was supported by the Emmy-Noether Programme (DFG).
1024:
1025:
1026:
1027:
1028: \begin{thebibliography}{00}
1029:
1030:
1031: \bibitem{NN81}
1032: We refer to the eigenstates of the Dirac chirality matrix, $\gamma^5$,
1033: see, e.g. H. B. Nielsen and M. Ninomiya, Nucl. Phys. B {\bf 185}, 20 (1981).
1034:
1035: \bibitem{Semenoff84}
1036: G. W. Semenoff, Phys. Rev. Lett. {\bf 53}, 2449 (1984).
1037:
1038: \bibitem{Novoselov05}
1039: K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, M. I. Katsnelson,
1040: I. V. Grigorieva, S. V. Dubonos, and A. A. Firsov, Nature (London)
1041: {\bf 438}, 197 (2005);
1042: Y. Zhang, Y.-W. Tan, H. L. Stormer, and P. Kim, Nature (London) {\bf
1043: 438}, 201 (2005).
1044:
1045: \bibitem{Koenig07}
1046: M. K\"onig, S. Wiedmann, C. Br\"une, A. Roth, H. Buhmann, L. W. Molenkamp, X.-L. Qi, and S.-C. Zhang,
1047: Science {\bf 318}, 766 (2007).
1048:
1049: \bibitem{Koenig08}
1050: M. K\"onig, H. Buhmann, L. W. Molenkamp, T. Hughes, C.-X. Liu, X.-L. Qi, and S.-C. Zhang,
1051: J. Phys. Soc. Jpn. {\bf 77}, 031007 (2008).
1052:
1053:
1054: \bibitem{Schmidt09}
1055: For HgTe quantum wells, the adequacy of the Dirac fermion model has recently been discussed in
1056: M. J. Schmidt, E .G. Novik, M. Kindermann, and B. Trauzettel, arXiv: 0901.0621.
1057:
1058: \bibitem{Itzykson}
1059: See, e.g., C. Itzykson and J. B. Zuber, {\em Quantum Field Theory} (McGraw-Hill Inc, 1985).
1060:
1061:
1062: \bibitem{McCann04}
1063: E. McCann and V. I. Fal'ko, J. Phys. Condens. Matter {\bf 16}, 2371 (2004).
1064:
1065: \bibitem{Akhmerov08}
1066: A. R. Akhmerov and C. W. J. Beenakker, Phys. Rev. B {\bf 77}, 085423 (2008).
1067:
1068:
1069: \bibitem{Berry87}
1070: M. V. Berry and R. J. Mondragon, Proc. R. Soc. Lond. A {\bf 412}, 53 (1987).
1071:
1072: \bibitem{Fujita96}
1073: M. Fujita, K. Wakabayashi, K. Nakada, and K. Kusakabe,
1074: J. Phys. Soc. Jpn. {\bf 65}, 1920 (1996).
1075:
1076: \bibitem{Armchair}
1077: Although this leaves out, e.g., armchair-type edges,
1078: the treatment is general enough to elucidate the origin of the parity violation.
1079:
1080: \bibitem{Waka00}
1081: K. Wakabayashi and M. Sigrist, Phys. Rev. Lett. {\bf 84}, 3390 (2000).
1082:
1083: \bibitem{Koba05}
1084: Y. Kobayashi, K. I. Fukui, T. Enoki, K. Kusakabe, and Y. Kaburagi,
1085: Phys. Rev. B {\bf 71}, 193406 (2005).
1086:
1087: \bibitem{Niimi06}
1088: Y. Niimi, T. Matsui, H. Kambara, K. Tagami, M. Tsukada, and H.
1089: Fukuyama, Phys. Rev. B {\bf 73}, 085421 (2006).
1090:
1091: \bibitem{Peres06}
1092: N. M. R. Peres, F. Guinea, and A. H. Castro Neto, Phys. Rev. B {\bf 73}, 125411 (2006).
1093:
1094: \bibitem{Sasaki06}
1095: K. Sasaki, S. Murakami, and R. Saito, Appl. Phys. Lett. {\bf 88}, 113110 (2006).
1096:
1097: \bibitem{Brey06}
1098: L. Brey and H. A. Fertig, Phys. Rev. B {\bf 73}, 235411 (2006); Phys. Rev. B {\bf 73}, 195408 (2006).
1099:
1100: \bibitem{Rycerz07}
1101: A. Rycerz, J. Tworzydlo, and C. W. J. Beenakker, Nature Phys. {\bf 3}, 172 (2007).
1102:
1103: \bibitem{GT07}
1104: G. Tkachov, Phys. Rev. B {\bf 76}, 235409 (2007); Phys. Rev. B {\bf 79}, 045429 (2009).
1105:
1106: \bibitem{Wimmer08}
1107: M. Wimmer, I. Adagideli, S. Berber, D. Tomanek, and K. Richter,
1108: Phys. Rev. Lett. {\bf 100}, 177207 (2008).
1109:
1110:
1111: \bibitem{Girit09}
1112: C. \"O. Girit, J. C. Meyer, R. Erni, M. D. Rossell, C. Kisielowski,
1113: L. Yang, C.-H. Park, M. F. Crommie, M. L. Cohen, S. G. Louie, and A. Zettl,
1114: Science {\bf 323}, 1705 (2009).
1115:
1116:
1117:
1118: \bibitem{Li08}
1119: G. Li, A. Luican and E. Y. Andrei, arXiv:0803.4016.
1120:
1121:
1122: \bibitem{Katsnelson06}
1123: For a recent review, see, e.g.
1124: M. I. Katsnelson, K. S. Novoselov, and A. K. Geim, Nature Phys. {\bf 2}, 620 (2006).
1125:
1126: \bibitem{Son06}
1127: Y.-W. Son, M. L. Cohen, and S. G. Louie, Phys. Rev. Lett. {\bf 97}, 216803 (2006).
1128:
1129: \bibitem{Kane05}
1130: C. L. Kane and E. J. Mele, Phys. Rev. Lett. {\bf 95}, 226801 (2005).
1131:
1132:
1133: \bibitem{Yazyev08}
1134: O. V. Yazyev and M. I. Katsnelson,
1135: Phys. Rev. Lett. {\bf 100}, 047209 (2008).
1136:
1137:
1138: \bibitem{Integrals}
1139: $
1140: \int_0^{ \pi/2 }
1141: \frac{
1142: d\gamma
1143: }
1144: { n^2_{z}+\tan^2\gamma }=
1145: \frac{1}{2}\oint
1146: \frac{
1147: dt
1148: }
1149: { ( 1 + t^2 )(n^2_{z} + t^2) }=
1150: \frac{\pi}{2}\frac{ 1 }{ (1 + |n_z|) |n_z| },
1151: $
1152: and
1153: $
1154: \int_0^{ \pi/2 }
1155: \frac{
1156: d\gamma
1157: }
1158: { 1 + n^2_{z}\cot^2\gamma }=
1159: \frac{1}{2}\oint
1160: \frac{
1161: t^2 dt
1162: }
1163: { ( 1 + t^2 )(n^2_{z} + t^2) }
1164: =\frac{\pi}{2}\frac{ 1 }{ 1 + |n_z| },
1165: $
1166: where the integration contour encircles either the upper (${\rm Im}\, t\geq 0$) or
1167: the lower (${\rm Im}\, t\leq 0$) halfplanes of the complex variable $t$.
1168:
1169: \bibitem{Giaever60}
1170: I. Giaever, Phys. Rev. Lett. {\bf 5}, 147 (1960).
1171:
1172: \bibitem{Mahan}
1173: G. D. Mahan, {\em Many-Particle Physics} (Plenum Press, New York, 2000).
1174:
1175: \bibitem{Ponomarenko08}
1176: L. A. Ponomarenko, F. Schedin, M. I. Katsnelson, R. Yang, E. H. Hill, K. S. Novoselov, and
1177: A. K. Geim, Science {\bf 320}, 356 (2008).
1178:
1179: \bibitem{Wurm09}
1180: J. Wurm, A. Rycerz, I. Adagideli, M. Wimmer, K. Richter, and H. U. Baranger, Phys. Rev. Lett. {\bf 102}, 056806 (2009).
1181:
1182:
1183:
1184: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1185: \end{thebibliography}
1186: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1187:
1188:
1189:
1190:
1191: \end{document}
1192:
1193:
1194:
1195:
1196:
1197:
1198:
1199: % ****** End of file apssamp.tex ******
1200:
1201:
1202:
1203:
1204:
1205:
1206:
1207:
1208:
1209:
1210:
1211:
1212:
1213:
1214:
1215:
1216:
1217:
1218:
1219: