1: % ****** Start of file apssamp.tex ******
2: %
3: % This file is part of the APS files in the REVTeX 4 distribution.
4: % Version 4.0 of REVTeX, August 2001
5: %
6: % Copyright (c) 2001 The American Physical Society.
7: %
8: % See the REVTeX 4 README file for restrictions and more information.
9: %
10: % TeX'ing this file requires that you have AMS-LaTeX 2.0 installed
11: % as well as the rest of the prerequisites for REVTeX 4.0
12: %
13: % See the REVTeX 4 README file
14: % It also requires running BibTeX. The commands are as follows:
15: %
16: % 1) latex apssamp.tex
17: % 2) bibtex apssamp
18: % 3) latex apssamp.tex
19: % 4) latex apssamp.tex
20: %
21: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb,prl]{revtex4}
22: %\documentclass[onecolumn,preprintnumbers,amsmath,amssymb,prb,eqsecnum]{revtex4}
23: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
24:
25: % Some other (several out of many) possibilities
26: %\documentclass[preprint,aps]{revtex4}
27: %\documentclass[preprint,aps,draft]{revtex4}
28: %\documentclass[prb]{revtex4}% Physical Review B
29:
30: \usepackage{graphicx}% Include figure files
31: \usepackage{dcolumn}% Align table columns on decimal point
32: \usepackage{bm}% bold math
33:
34: \usepackage{latexsym}
35: %\mathindent30pt
36: %---------------------------------------------------------------------
37: \newcommand{\eq}{Eq}
38: %---------------------------------------------------------------------
39: %\nofiles
40:
41: \begin{document}
42: %\mathindent10mm
43:
44: \preprint{}
45:
46: \title{Topological aspects of quantum spin Hall effect in graphene: \\
47: Z$_2$ topological order and spin Chern number}
48:
49: \author{Takahiro Fukui}
50: \affiliation{Department of Mathematical Sciences, Ibaraki University, Mito
51: 310-8512, Japan}
52: \author{Yasuhiro Hatsugai}
53: \affiliation{Department of Applied Physics, University of Tokyo, Hongo,
54: Tokyo 113-8656, Japan}
55:
56: \date{July 19, 2006}
57:
58: \begin{abstract}
59: For generic time-reversal invariant systems with spin-orbit couplings,
60: we clarify a close relationship between the Z$_2$ topological order
61: and the spin Chern number proposed by Kane and Mele and
62: by Sheng {\it et al.}, respectively, in the quantum spin Hall effect.
63: It turns out that a global gauge transformation
64: connects different spin Chern numbers (even integers) modulo 4,
65: which implies that the spin Chern number and
66: the Z$_2$ topological order yield the same classification.
67: We present a method of computing spin Chern numbers and
68: demonstrate it in single and double plane of graphene.
69: \end{abstract}
70:
71: \pacs{73.43.-f, 72.25.Hg, 73.61.Wp, 85.75.-d}
72: \maketitle
73:
74:
75: Topological orders \cite{Wen89,Hat04} play a crucial role
76: in the classification of various phases in low dimensional systems.
77: The integer quantum Hall effect (IQHE)
78: is one of the most typical examples
79: \cite{TKNN82,Koh85}, in which the quantized Hall conductance is given by
80: a topological invariant, Chern number, due to the Berry phase \cite{Ber84}
81: induced in the Brillouin zone \cite{Sim83}.
82: Such a topological feature should be more fundamental, since it has a close
83: relationship with the parity anomaly of Dirac fermions
84: \cite{Sem84,Ish84,Hal88}.
85:
86: Recently, the spin Hall effect \cite{MNZ03,Sin04,KMGA04,WKSJ04}
87: has been attracting much current interest
88: as a new device of so-called spintronics.
89: In particular,
90: Kane and Mele \cite{KanMel05a,KanMel05b}
91: have found a new class of insulator showing
92: the quantum spin Hall (QSH) effect
93: \cite{BerZha05,QWZ05,SSTH05}
94: which should be realized in
95: graphene with spin-orbit couplings.
96: They have pointed out \cite{KanMel05b} that
97: the QSH state can be specified by a Z$_2$ topological order
98: which is inherent in time-reversal (${\cal T}$) invariant systems.
99: This study is of fundamental importance, since the Z$_2$ order
100: is involved with the Z$_2$ anomaly of Majorana fermions
101: \cite{AtiSin,Wit84}.
102:
103: On the other hand, Sheng {\it et al.} \cite{SWSH06} have recently
104: computed the spin Hall conductance
105: by imposing spin-dependent twisted boundary condition,
106: generalizing the idea of Niu {\it et al.} \cite{NTW85,Hat04}.
107: They have shown that it is given by a
108: Chern number which is referred to as a spin Chern number below.
109: This seems very natural,
110: since the QSH effect is a spin-related version of IQHE.
111: The spin Chern number for graphene computed by Sheng {\it et al.}
112: indeed has a good correspondence with the
113: classification by Z$_2$.
114:
115: However, the Chern number is specified by the set of
116: integers Z, not by Z$_2$.
117: Although
118: the studies
119: by Sheng {\it et al. } \cite{SWSH06}
120: suggest a close relationship between two topological orders,
121: natural questions arise:
122: How does the concept Z$_2$ enter into the classification by Chern numbers,
123: or otherwise, does the spin Chern number carry additional information?
124:
125: In this letter, we clarify the relationship for generic ${\cal T}$
126: invariant systems. We show that while two sectors
127: in the Z$_2$ classification are separated
128: by topological changes due to {\it bulk gap-closing phenomena},
129: each of these sectors is further divided
130: into many sectors by {\it boundary-induced topological changes}
131: in the spin Chern number classification.
132: The latter is an artifact which is due to
133: calculations in finite size systems with broken translational invariance.
134: Therefore,
135: the different spin Chern numbers in each sector of Z$_2$
136: describe the same topologically ordered states
137: of the bulk.
138:
139: Consider generic electron systems on a lattice
140: with ${\cal T}$ symmetry, described by the Hamiltonian $H$.
141: Denote the electron creation operator at $j$th site as
142: $c^\dagger_j=(c^\dagger_{\uparrow j}, c^\dagger_{\downarrow j})$.
143: Then, ${\cal T}$ transformation is defined by
144: $c_j\rightarrow {\cal T}c_j$ with ${\cal T}\equiv i\sigma^2 K$,
145: where the Pauli matrix $\sigma^2$ operates the spin space
146: and $K$ stands for the
147: complex conjugation operator.
148: Let ${\cal H}(k)$ be the Fourier-transformed Hamiltonian defined by
149: $H=\sum_k c^\dagger(k){\cal H}(k)c(k)$ and let $|n(k)\rangle$ be
150: an eigenstate of ${\cal H}(k)$.
151: Assume that the ground state is composed of an $M$-dimensional multiplet of
152: degenerate single particle states
153: which is a generalized non-interacting Fermi sea \cite{Hat04}.
154: Kane and Mele \cite{KanMel05b} have found that
155: the ${\cal T}$ invariant systems have two kinds of important states
156: belonging to ``even'' subspace and ``odd'' subspace:
157: The states in the even subspace have the property that $|n(k)\rangle$
158: and ${\cal T}|n(k)\rangle$ are identical, which occurs when
159: ${\cal T}{\cal H}(k){\cal T}^{-1}={\cal H}(k)$.
160: By definition,
161: the states at $k=0$ always belong to this subspace.
162: The odd subspace has the property that the multiplet $|n(k)\rangle$
163: are orthogonal to the multiplet ${\cal T}|n(k)\rangle$.
164: These special subspaces can be detected by
165: the pfaffian
166: $p_{\rm KM}(k)\equiv \mbox{pf}~\langle n(k)|{\cal T}|m(k)\rangle$.
167: Namely, $p_{\rm KM}(k)=1$ in the even subspace and 0 in the odd subspace.
168: Kane and Mele have claimed that
169: the number of zeros of $p_{\rm KM}(k)$
170: which always appear as $\cal T$ pairs $\pm k^*$
171: with opposite vorticities
172: is a topological invariant for ${\cal T}$ invariant systems.
173: Specifically, if the number of zeros in half the
174: Brillouin zone is 1 (0) mod 2, the ground state is in the
175: QSH (insulating) phase.
176:
177:
178: We now turn to the spin Chern number
179: proposed by Sheng {\it et al.} \cite{SWSH06}.
180: According to their formulation, we impose spin-dependent
181: (-independent) twisted boundary condition along $1(2)$-direction
182: %-----------------------------------------------------------------
183: % Twisted boundary condition
184: %----------------------------------------------------------------
185: \begin{alignat}{1}
186: &
187: c_{j+L_1\hat 1}=e^{i\theta_1 \sigma^3}c_j, \quad
188: c_{j+L_2\hat 2}=e^{i\theta_2}c_j ,
189: \label{TwiBouCon}%---------------------------------------------------
190: \end{alignat}
191: where a set of integers $j\equiv(j_1,j_2)$ specifies the site and
192: $\hat 1$ and $\hat 2$ stand for the unit vectors in $1$- and
193: $2$-directions, respectively.
194: Let $ {\cal H}(\theta)$ denote the twisted Hamiltonian
195: and let $|n(\theta)\rangle$ be corresponding eigenstate.
196: It follows from Eq. (\ref{TwiBouCon}) that ${\cal T}$ transformation
197: induces
198: ${\cal T}{\cal H}(\theta_1,\theta_2){\cal T}^{-1}
199: ={\cal H}(\theta_1,-\theta_2)$
200: and therefore we can always choose
201: $|n(\theta_1,-\theta_2)\rangle={\cal T}|n(\theta_1,\theta_2)\rangle$
202: except for $\theta_2=0$. The states on the {\it line} $\theta_2=0$ belong
203: to the even subspace.
204: Below,
205: the torus spanned by $\theta$ is referred to as twist space.
206: The boundary condition (\ref{TwiBouCon}) enables us to define the spin Chern number,
207: but the cost we have to pay is the broken translational invariance.
208: In other words, we slightly break ${\cal T}$ invariance at the boundary,
209: which leads to nontrivial spin Chern numbers. The average over the twist
210: angles recovers ${\cal T}$ invariance.
211:
212:
213: Let us define a
214: pfaffian for the present twisted system as a function of the twist angles:
215: \begin{alignat}{1}
216: p(\theta)\equiv \mbox{pf}~\langle n(\theta)|{\cal T}|m(\theta)\rangle,
217: \quad n,m=1,\cdots,M ,
218: \label{Pfa}%------------------------------------------------------
219: \end{alignat}
220: where $M$ is a number of one particle states below the Fermi energy with the
221: gap opening condition.
222: Since the line $\theta_2=0$ belongs to the even subspace,
223: the zeros of the pfaffian in the $\theta_2>0$ or $<0$ twist space
224: can move only among the same half twist space keeping their vorticities,
225: and never cross the $\theta_2=0$ line.
226: Therefore, one ${\cal T}$ pair of zeros are never annihilated, like those
227: of the KM pfaffian $p_{\rm KM}(k)$.
228: Furthermore,
229: the two zeros in the same half twist space are
230: never annihilated unless they have the opposite vorticity.
231: This is a crucial difference between the KM pfaffian and the twist pfaffian.
232: The number of zeros in the twist pfaffian
233: should be classified by even integers
234: (half of them, by Z).
235:
236:
237: Are these pfaffians topologically different quantities?
238: The answer is no.
239: To show this, let us consider a global gauge transformation
240: $c_j\rightarrow g(\varphi) c_j$, where
241: %-----------------------------------------------------------------
242: % gauge transformation
243: %----------------------------------------------------------------
244: \begin{alignat}{1}
245: g(\varphi)\equiv
246: e^{i\sigma^2\varphi}=\cos\varphi + i\sigma^2 \sin\varphi .
247: \label{OrtTra}%--------------------------------------------------
248: \end{alignat}
249: This transformation replaces the the Pauli matrices
250: in the spin-orbit couplings into
251: $g^{\rm t}(\varphi){\bm \sigma}g(\varphi)=
252: (\cos2\varphi\,\sigma^1-\sin2\varphi\,\sigma^3,\sigma^2,
253: \cos2\varphi\,\sigma^3+\sin2\varphi\,\sigma^1)$.
254: On the other hand,
255: since Eq. (\ref{OrtTra}) is an orthogonal transformation,
256: one-parameter family of transformed Hamiltonian,
257: denoted by $H^\varphi$ or ${\cal H}^\varphi(\theta)$ below,
258: is equivalent. The pfaffian (\ref{Pfa}) is also invariant.
259: Therefore, when we are interested in the bulk properties,
260: we can deal with any Hamiltonian $H^\varphi$.
261:
262:
263: So far we have discussed the bulk properties.
264: However, if we consider finite periodic systems like Eq. (\ref{TwiBouCon}),
265: a family of the Hamiltonian $H^\varphi$ behaves as different models.
266: It follows from Eq. (\ref{TwiBouCon}) that
267: the gauge transformation (\ref{OrtTra}) is commutative with
268: $e^{i\theta_2}$, but {\it not} with $e^{i\theta_1\sigma^3}$.
269: This tells us that spin-dependent twisted boundary
270: condition is not invariant under the gauge transformation
271: (\ref{OrtTra}), and breaks the gauge-equivalence of the Hamiltonian
272: $H^\varphi$ which the bulk systems should have.
273:
274:
275: To understand this, the following alternative consideration
276: may be useful:
277: If we want to study bulk properties of ${\cal T}$ invariant systems,
278: we can start with any of $H^\varphi$.
279: For one $H^\varphi$ with $\varphi$ fixed, let us impose the twisted
280: boundary condition (\ref{TwiBouCon}). After that, we can make a gauge
281: transformation (\ref{OrtTra}) back to $H^{0}$.
282: Then, we can deal with the original Hamiltonian $H^{0}$, but with a
283: gauge-dependent twisted boundary condition for $x$-direction;
284: %-----------------------------------------------------------------
285: % Modified twisted boundary condition
286: %----------------------------------------------------------------
287: \begin{alignat}{1}
288: &c_{j+L_1\hat 1}=
289: e^{i\theta_1 (\cos2\varphi\sigma^3-\sin2\varphi\sigma^1)}c_j.
290: \label{ModTwiBouCon}%---------------------------------------------------
291: \end{alignat}
292: Namely, the gauge equivalence is broken only by the boundary condition
293: in $1$-direction.
294: Now, imagine a situation that at $\varphi=0$ the pfaffian (\ref{Pfa})
295: has one ${\cal T}$ pair of zeros.
296: We denote them as
297: $(\theta^*_1,\pm\theta_2^*)$ with vorticity $\pm m$.
298: Let us change $\varphi$ smoothly from 0 to $\pi/2$.
299: Then, it follows from Eq. (\ref{ModTwiBouCon}) that
300: at $\varphi=\pi/2$ the coordinate of the torus is changed from
301: $(\theta_1,\theta_2)$ into $(-\theta_1,\theta_2)$ and therefore,
302: we find the zeros at $(-\theta_1^*,\pm\theta_2^*)$ with vorticity $\mp m$.
303: We thus have a mapping
304: $(\theta_1^*,\pm\theta_2^*)\rightarrow(-\theta_1^*,\mp\theta_2^*)$
305: for $\varphi=0\rightarrow \pi/2$.
306: Namely, the zero in the $\theta_2>0$ $(<0)$ space moves into the
307: $\theta_2<0$ $(>0)$ space, and thus the zeros can move in the whole twist
308: space like those of the KM pfaffian.
309: During the mapping, there should occur a topological change,
310: but it is attributed to the boundary, i.e, an artifact of broken
311: translational invariance, and
312: half of the zeros of twist pfaffian should be also classified by Z$_2$,
313: by taking into account the gauge transformation (\ref{OrtTra}).
314:
315:
316:
317: Using this pfaffian,
318: we next show that its zeros can be
319: detectable by computing the spin Chern number which seems much easier to
320: perform than
321: searching them directly.
322: The spin Chern number \cite{SWSH06} is defined by
323: %-----------------------------------------------------------------
324: % Chern number
325: %----------------------------------------------------------------
326: %\begin{eqnarray}
327: $c_{\rm s}=\frac{1}{2\pi i}\int d^2\theta F_{12}(\theta)$,
328: %\label{CheNum}%-------------------------------------------------
329: %\end{eqnarray}
330: where
331: $F_{12}(\theta)\equiv
332: \partial_{1} A_2(\theta)-\partial_{2} A_1(\theta)$
333: is the field strength due to the U(1) part of
334: the (non-Abelian) Berry potential \cite{Hat04},
335: $A_\mu(\theta)\equiv{\rm tr}
336: \langle n(\theta)|\partial_{\mu}|m(\theta) \rangle$
337: with
338: $\partial_\mu=\partial/\partial\theta_\mu$.
339: First, we will show the relationship between
340: the zeros of the twist pfaffian
341: and the spin Chern number $c_{\rm s}$.
342: For the time being, we fix $\varphi$.
343: The degenerate ground state as the $M$-dimensional multiplet has a local U$(M)$ gauge
344: degree of freedom,
345: $|n(\theta)\rangle\rightarrow \sum_m|m(\theta)\rangle V_{mn}(\theta)$,
346: where $V(\theta)$ is a unitary matrix \cite{Hat04}.
347: Let us
348: denote
349: $V(\theta)=e^{i\alpha(\theta)/M}\tilde V(\theta)$,
350: where $\det\tilde V(k)=1$.
351: This transformation induces
352: $A_\mu(\theta)\rightarrow A_\mu(\theta)
353: +i\partial_{\mu}\alpha(\theta)$
354: to the Berry potential.
355: If one can make gauge-fixing globally over the whole twist space,
356: the Chern number is proved to be zero. Only if the global gauge-fixing
357: is impossible, the Chern number can be nonzero.
358:
359: Among various kinds of gauge-fixing, we can use the gauge that
360: {\it $p(\theta)$ is real positive}, because
361: $p(\theta)\rightarrow p(\theta) \det V(\theta)=p(\theta)e^{i\alpha(\theta)}$.
362: This rule can fix the gauge of the Berry potential except for $p(\theta)=0$.
363: Therefore, nontrivial spin Chern number is due to an obstruction to the
364: smooth gauge-fixing by the twist pfaffian.
365: This correspondence also proves that
366: {\it the spin Chern number is an even integer}, since
367: the pfaffian (\ref{Pfa}) always has ${\cal T}$ pair zeros,
368: and since $F_{12}(\theta_1,-\theta_2)=F_{12}(\theta_1,\theta_2)$.
369:
370: Let us now
371: change $\varphi$.
372: At $\varphi=0$, we obtain a certain integral $c_{\rm s}$.
373: Remember that at $\varphi=\pi/2$, the coordinate of the torus
374: is changed into $(-\theta_1,\theta_2)$.
375: Therefore, we have a mapping $c_{\rm s}\rightarrow -c_{\rm s}$
376: for $(\theta_1,\theta_2)\rightarrow (-\theta_1,\theta_2)$.
377: As in the case of the pfaffian (\ref{Pfa}), we expect
378: topological changes during the mapping.
379: However, as stressed, these changes are accompanied by
380: {\it no gap-closing in the bulk spectrum},
381: just induced by the symmetry breaking
382: {\it boundary} term which is an artifact to define the Chern numbers.
383: Therefore, the states with $\pm c_{\rm s}$ should
384: belong to an equivalent topological sector.
385: Since the minimum nonzero spin Chern number is 2, we expect
386: $c_{\rm s}$ mod 4
387: (if we define the spin Chern number in half the
388: twist space, mod 2) classifies the topological sectors.
389:
390:
391: So far we have discussed the Z$_2$ characteristics of the spin Chern
392: number $c_{\rm s}$. We next present several examples.
393: To this end, we employ an efficient method of
394: computing Chern numbers proposed in Ref. \cite{FHS05}.
395: We first discretize the twist space
396: $[0,2\pi]\otimes [0,2\pi]$ into a
397: square lattice such that $\theta_\mu=2\pi j_\mu/N_\mu$,
398: where $j_\mu=1,\cdots,N_\mu$ \cite{FooNot2}.
399: We denote the sites on this lattice as $\theta_\ell$ with
400: $\ell=1,2,\cdots, N_1N_2$.
401: We next define a U(1) link variable
402: associated with the ground state multiplet of the dimension $M$,
403: \begin{eqnarray*}
404: %$
405: U_\mu(\theta_\ell)=|\det{\bm U}_\mu(\theta_\ell)|^{-1}
406: \det {\bm U}_\mu(\theta_\ell) ,
407: %$,
408: \end{eqnarray*}
409: where
410: %\begin{eqnarray}
411: ${\bm U}_\mu(\theta_\ell)_{mn}=
412: \langle m(\theta_\ell)|n(\theta_\ell+\hat\mu)\rangle$
413: %\end{eqnarray}
414: with $n,m=1,\cdots,M$ denotes the (non-Abelian) Berry link variable.
415: Here, $\hat\mu$ is the vector in $\mu$ direction with
416: $|\hat\mu|=2\pi/N_\mu$.
417: Next define the lattice field strength,
418: \begin{eqnarray*}
419: %$
420: F_{12}(\theta_\ell)=
421: \ln U_1(\theta_\ell)U_2(\theta_\ell+\hat 1)
422: U^{-1}_1(\theta_\ell+\hat 2)U^{-1}_2(\theta_\ell) ,
423: %$,
424: \end{eqnarray*}
425: where we choose the branch of the logarithm as $|F_{12}(\theta_\ell)|<\pi$.
426: Finally, manifestly {\it gauge invariant} lattice Chern number is obtained:
427: \begin{eqnarray}
428: c_{\rm s}=\frac{1}{2\pi i}\sum_\ell F_{12}(\theta_\ell) .
429: \label{LatCheNum}%---------------------------------------------------
430: \end{eqnarray}
431: As shown in \cite{FHS05}, the spin Chern number thus defined is strictly
432: {\it integral}.
433: To see this, let us introduce a lattice gauge potential
434: %\begin{eqnarray}
435: $A_\mu(\theta_\ell)=\ln U_\mu(\theta_\ell)$
436: %\end{eqnarray}
437: which is also defined in $|A_\mu(\theta_\ell)|<\pi$.
438: Note that this field is periodic,
439: $A_\mu(\theta_\ell+N_\mu)=A_\mu(\theta_\ell)$.
440: Then, we readily find
441: \begin{eqnarray}
442: F_{12}(\theta_\ell)=\Delta_1 A_2(\theta_\ell)-\Delta_2A_1(\theta_\ell)+
443: 2\pi i n_{12}(\theta_\ell) ,
444: \label{FieStrN}%----------------------------------------------
445: \end{eqnarray}
446: where $\Delta_\mu$ stands for the difference operator and
447: $n_{12}(\theta_\ell)$ is a local {\it integral }field which is
448: referred to as $n$-field.
449: Finally, we reach
450: %\begin{eqnarray}
451: $c_{\rm s}=\sum_\ell n_{12}(\theta_\ell)$.
452: %\label{CheNumN}%---------------------------------------------------
453: %\end{eqnarray}
454: This completes the proof that the spin Chern number is integral.
455: While the $n$-field depends on a gauge we adopt, the sum is invariant.
456: For the ${\cal T}$ invariant systems, % under consideration,
457: {\it the pfaffian
458: (\ref{Pfa}) is very useful for the gauge-fixing} also for the lattice
459: computation.
460: In the continuum theory, we have stressed that the zeros of the pfaffian
461: play a central role in the Z$_2$ classification. Since such zeros occur at
462: several specific points in the twist space, it is very hard
463: to search them numerically.
464:
465:
466: Contrary to this feature in the continuum approach,
467: we can detect the zeros in the lattice approach as follows:
468: Suppose that we obtain an exact Chern
469: number using Eq. (\ref{LatCheNum}) with sufficiently large $N_\mu$.
470: Since the lattice Chern number is topological (integral),
471: which implies that even if we slightly change the lattice
472: (e.g, change the lattice size or infinitesimally shift
473: the lattice),
474: the Chern number remains unchanged.
475: Next, let us compute the $n$-field in the gauge that
476: {\it the pfaffian is real positive}.
477: If the zeros of the pfaffian happen to locate
478: on sites of the present lattice, we cannot make gauge-fixing.
479: Even in such cases, we can always avoid the zeros of the pfaffian
480: by redefining the lattice with the spin Chern number kept unchanged.
481: Thus we can always
482: compute the well-defined $n$-field configuration.
483: If the Chern number is nonzero, there exist nonzero $n$-field
484: somewhere in the twist space.
485: These nonzero $n$-field occur in general near the zeros of the
486: pfaffian: Thus, without searching zeros of pfaffian in the continuum
487: twist space, we can observe {\it
488: the trace of such zeros as nonzero $n$-fields}.
489:
490: Now let us study a graphene model with spin-orbit couplings
491: \cite{KanMel05a,KanMel05b,SWSH06};
492: %-----------------------------------------------------------------
493: % each Hamiltonian
494: %----------------------------------------------------------------
495: \begin{alignat}{1}
496: H=&-t\sum_{\langle i,j\rangle}c^\dagger_ic_j
497: +\frac{2i}{\sqrt{3}}V_{\rm so}
498: \sum_{\langle\langle i,j\rangle\rangle}c^\dagger_i{\bm \sigma}\cdot
499: \left({\bm d}_{kj}\times{\bm d}_{ik}\right)c_j
500: \nonumber\\& %-------------------------------------------------------!!!
501: +iV_{\rm R}\sum_{\langle i,j\rangle}
502: c^\dagger_i\left({\bm\sigma}\times{\bm d}_{ij}\right)^3c_j
503: +v_{\rm s}\sum_j\mbox{sgn} j~c^\dagger_jc_j,
504: \label{Ham} %---------------------------------------------------
505: \end{alignat}
506: where
507: $c^\dagger_j=(c^\dagger_{\uparrow j}, c^\dagger_{\downarrow j})$
508: is the electron creation operator at site $j$ on the honeycomb lattice,
509: $\mbox{sgn}j$ denotes $1~(-1)$ if $j$ belongs to the A (B) sublattice,
510: and ${\bm d}_{ij}$ stands for the vector from $j$ to $i$ sites.
511: The first term is the nearest neighbor hopping,
512: the second term is the $s_z$-conserving
513: next-nearest neighbor spin-orbit coupling, whereas the third term is
514: the Rashba spin-orbit coupling, and final term is a
515: on-site staggered potential.
516: Analyzing the KM pfaffian,
517: Kane and Mele \cite{KanMel05a} have derived the phase diagram:
518: It is in the QSH phase for
519: $|6\sqrt{3}V_{\rm so}-v_{\rm s}|>\sqrt{v_{\rm s}^2+9V_{\rm R}^2}$
520: and in the insulating phase otherwise.
521: Sheng {\it et. al.} \cite{SWSH06} have computed the spin Chern
522: number $c_{\rm s}=2$ in the QSH phase and 0 in the insulating phase.
523:
524:
525:
526: \begin{figure}[htb]
527: \begin{tabular}{ccc}
528: \includegraphics[width=.33\linewidth]{fig1a.eps}
529: &\includegraphics[width=.33\linewidth]{fig1b.eps}
530: &\includegraphics[width=.33\linewidth]{fig1c.eps}
531: \end{tabular}
532: \caption{Spectrum for $\theta_1=0$ as a function of $k_2$.
533: Parameters used are $V_{\rm so}=0.1t$, $v_{\rm s}=0.3t$,
534: and $V_{\rm R}=0.1t$ (left), $V_{\rm R}=0.225207t$ (middle),
535: and $V_{\rm R}=0.3t$ (right).
536: }
537: \label{f:Spectrum}
538: \end{figure}
539:
540: For numerical computations, it is convenient to use the momentum $k_2$
541: \cite{FooNot1,SWSH06}
542: instead of $\theta_2$ because of the translational invariance along this
543: direction even with the boundary condition (\ref{TwiBouCon}).
544: First, we show the spectrum in Fig. \ref{f:Spectrum} at $\theta_1=0$ as
545: a function of $k_2$.
546: The left belongs to the QSH phase with $c_{\rm s}=2$, whereas
547: the right to the insulating phase with $c_{\rm s}=0$.
548: This topological change is due to the gap-closing in the {\it bulk}
549: spectrum, as shown in the middle in Fig. \ref{f:Spectrum}.
550: Therefore, the phase with $c_{\rm s}=2$ is topologically
551: distinguishable from the phase with $c_{\rm s}=0$.
552:
553: \begin{figure}[htb]
554: \begin{tabular}{ccc}
555: \includegraphics[width=.33\linewidth]{fig2a.eps}
556: &\includegraphics[width=.33\linewidth]{fig2b.eps}
557: \hspace{-3mm}
558: &\includegraphics[width=.33\linewidth]{fig2c.eps}
559: \end{tabular}
560: \caption{Left: Spectrum
561: for nonzero $\varphi=\pi/4$ as a function of
562: $k_2$ at $\theta_1=\pi/2$.
563: Other parameters are the same as those of the left
564: in Fig. \ref{f:Spectrum}.
565: Right two:
566: the $n$-field configuration corresponding to the left in
567: Fig. \ref{f:Spectrum}. The left (right) is at $\varphi=0~(\pi/2)$.
568: The white (black) circle denotes $n=1$ $(-1)$,
569: while the blank means $n=0$.
570: We used the meshes $N_1=N_2=10$.
571: }
572: \label{f:nfield}%------------------------------------
573: \end{figure}
574:
575: Contrary to this, how about the phases with $c_{\rm s}=\pm2$?
576: As we have mentioned, the phase with $c_{\rm s}$ changes
577: into $-c_{\rm s}$ when
578: we vary $\varphi$ from 0 to $\pi/2$. In this process,
579: {\it boundary}-induced topological changes must occur.
580: We show in Fig. \ref{f:nfield} the spectrum cut at $\theta_1=\pi/2$
581: for $\varphi=\pi/4$.
582: We indeed observe a gap-closing at finite $\theta_1$,
583: and the spin Chern number $c_{\rm s}=2$ for $0\leq\varphi<\pi/4$
584: is changed into $c_{\rm s}=-2$ for $\pi/4<\varphi\leq \pi/2$.
585: As stressed,
586: this change is attributed to the boundary (edge states in Fig. \ref{f:nfield}),
587: not to the bulk, and we conclude that
588: the phase $c_{\rm s}=\pm2$ is classified as the same QSH phase.
589: These spin Chern numbers $c_{\rm s}=\pm2$ are well visible by the $n$-field.
590: In Fig. \ref{f:nfield}, we also show the $n$-field
591: for $\varphi=0$ and $\pi/2$ cases in the QSH phase. The points of nonzero
592: $n$-field is closely related with the positions of the pfaffian zeros.
593: We also note that in Eq. (\ref{LatCheNum})
594: net contributions to the
595: nonzero Chern number is just from two points.
596:
597: Next, let us study a bilayer graphene.
598: Suppose that we have two decoupled
599: sheets of graphene described by $H^{\varphi_i}$ with $i=1,2$ whose lattices
600: include $A$, $B$ sites and $\tilde A$, $\tilde B$ sites, respectively.
601: For simplicity, we take into account only the interlayer coupling $\gamma_1$
602: between $\tilde A$ and $B$ \cite{McCFal06}:
603: $V_{12}=\gamma_1\sum_{j}c_{1\tilde A,j}^\dagger c_{2B,j}+\mbox{h.c.}$,
604: where $i=1,2$ in $c_{i,j}$ indicate the $i$th sheet.
605: Now make the gauge transformation (\ref{OrtTra}) separately for
606: each sheet to obtain the same $H^0$ as Eq. (\ref{Ham}).
607: Then, we have identical bilayer system $H^0\otimes H^0$ coupled by
608: $V_{12}=\gamma_1\sum_{j}
609: c_{1\tilde A,j}^\dagger g(\varphi_1)
610: g^{\rm t}(\varphi_2)c_{2B,j+}\mbox{h.c.}$
611: with two independent boundary conditions
612: $c_{i,j+L_1\hat 1}=
613: e^{i\theta_1 (\cos2\varphi_i\sigma^3+\sin2\varphi_i\sigma^1)}c_{i,j}$.
614: \begin{figure}[htb]
615: \begin{tabular}{ccc}
616: \includegraphics[width=.33\linewidth]{fig3a.eps}
617: \hspace{-3mm}
618: &\includegraphics[width=.33\linewidth]{fig3b.eps}
619: \hspace{-3mm}
620: &\includegraphics[width=.33\linewidth]{fig3c.eps}
621: \end{tabular}
622: \caption{The $n$-field configuration for $\gamma_1=0.1t$.
623: Other parameters are same as those of the left in Fig. \ref{f:Spectrum}.
624: Left: $\varphi_1=0$ and $\varphi_2=0$.
625: Middle: $\varphi_1=\pi/2$ and $\varphi_2=0$.
626: Right: $\varphi_1=\pi/4$ and $\varphi_2=-\pi/4$.
627: In the case $\varphi_1=\pi/2$ and $\varphi_2=\pi/2$ we have the same
628: figure as the left but with black circles.
629: We used the meshes %$L_x=10$
630: $N_1=N_2=10$.
631: }
632: \label{f:NfieldDL}%------------------------------------
633: \end{figure}
634:
635: For the same parameters as those of the left in Fig. \ref{f:Spectrum},
636: the spin Chern number is, of course, $c_{\rm s}=2+2=4$ in the limit
637: $\gamma_1=0$.
638: This spin Chern number remains unchanged for small but finite interlayer
639: coupling $\gamma_1$.
640: However, taking into account the gauge transformation $g(\varphi_i)$,
641: the spin Chern number changes. % in units of 4.
642: In Fig. \ref{f:NfieldDL}, we show examples of the $n$-field for $\gamma_1=0.1t$.
643: We have the spin Chern numbers $2+2=4$, $2-2=0$, and $-2-2=-4$:
644: All of them are denoted as $c_{\rm s}=0$ mod 4, which belong to
645: the insulating phase.
646: Detail analysis of this model including the interlayer coupling $\gamma_3$
647: will be published elsewhere.
648:
649: Finally,
650: we comment that the QSH effect is understood by the edge states
651: \cite{KanMel05b}, and therefore, it is interesting to establish the
652: bulk-edge correspondence \cite{Hat93} for ${\cal T}$ invariant systems
653: with respect to Z$_2$.
654: We also mention that
655: Fu and Kane \cite{FuKan06} and Moore and Balents \cite{MooBal06}
656: have recently discussed the relationship between the Z$_2$
657: order and the spin Chern number, and reached a similar conclusion.
658:
659:
660:
661: This work was supported in part
662: by Grant-in-Aid for Scientific Research
663: (Grant No. 17540347, No. 18540365) from JSPS
664: and on Priority Areas (Grant No.18043007) from MEXT.
665:
666: \begin{thebibliography}{99} %% The number "99" means that this list has
667: %% more than nine items.
668: \bibitem{Wen89}
669: X. G. Wen,
670: Phys. Rev. B {\bf 40}, 7387 (1989).
671: %
672: \bibitem{Hat04}
673: Y. Hatsugai,
674: J. Phys. Soc. Jpn. {\bf 73}, 2604 (2004);
675: {\it ibid.} {\bf 74}, 1374 (2005).
676: %
677: \bibitem{TKNN82}
678: D. J. Thouless, M. Kohmoto, M. P. Nightingale, and M. den Nijs,
679: Phys. Rev. Lett. {\bf 49}, 405 (1982).
680: %
681: \bibitem{Koh85}
682: M. Kohmoto,
683: Ann. Phys. {\bf 160}, 355 (1985).
684: %
685: \bibitem{Ber84}
686: M. V. Berry,
687: Proc. Roy. Soc. Lond. A {\bf 392}, 45 (1984).
688: %
689: \bibitem{Sim83}
690: B. Simon,
691: Phys. Rev. Lett. {\bf 51}, 2167 (1983).
692: %
693: \bibitem{Sem84}
694: G. W. Semenoff,
695: Phys. Rev. Lett. {\bf 53}, 2449 (1984).
696: %
697: \bibitem{Ish84}
698: K. Ishikawa, Phys. Rev. Lett. {\bf 53}, 1615 (1984):
699: %
700: Phys. Rev. D {\bf 31}, 1432 (1985).
701: %
702: \bibitem{Hal88}
703: F. D. M. Haldane,
704: Phys. Rev. Lett. {\bf 61}, 2015 (1988).
705: %
706: \bibitem{MNZ03}
707: S. Murakami, N. Nagaosa, and S. C. Zhang,
708: Science {\bf 301}, 1348 (2003);
709: %
710: Phys. Rev. Lett. {\bf 93}, 156804 (2004).
711: %
712: \bibitem{Sin04}
713: J. Sinova, D. Culcer, Q. Niu, N. A. Sinitsyn, T. Jungwirth,
714: and A. H. MacDonald,
715: Phys. Rev. Lett. {\bf 92}, 126603 (2004).
716: %
717: \bibitem{KMGA04}
718: Y. K. Kato, R. C. Myers, A. C. Gossard, and D. D. Awschalom,
719: Science, {\bf 306}, 1910 (2004).
720: %
721: \bibitem{WKSJ04}
722: J. Wunderlich, B. K\"astner, J. Sinova, and T. Jungwirth,
723: Phys. Rev. Lett. {\bf 94}, 047204 (2005).
724: %
725: \bibitem{KanMel05a}
726: C. L. Kane and E. J. Mele
727: Phys. Rev. Lett. {\bf 95}, 226801 (2005).
728: %
729: \bibitem{KanMel05b}
730: C. L. Kane and E. J. Mele
731: Phys. Rev. Lett. {\bf 95}, 146802 (2005).
732: %
733: \bibitem{BerZha05}
734: B. A. Bernevig and S.-C. Zhang,
735: cond-mat/0504147.
736: %
737: \bibitem{QWZ05}
738: X.-L. Qi, Y.-S. Wu, and S.-C. Zhang,
739: cond-mat/0505308.
740: %
741: \bibitem{SSTH05}
742: L. Sheng, D. N. Sheng, C. S. Ting, and F. D. M. Haldane,
743: cond-mat/0506589.
744: %
745: \bibitem{AtiSin}
746: M. F. Atiyah and I. M. Singer, Ann. Math. {\bf 93}, 139 (1971).
747: %
748: \bibitem{Wit84}
749: E. Witten, Phys. Lett. {\bf 117B}, 324 (1982).
750: %
751: \bibitem{SWSH06}
752: D. N. Sheng, Z. Y. Weng, L. Sheng, and F. D. M. Haldane,
753: cond-mat/0603054.
754: %
755: \bibitem{NTW85}
756: Q. Niu, D. J. Thouless, and Y.-S. Wu,
757: Phys. Rev. B {\bf 31}, 3372 (1985).
758: %
759: \bibitem{FHS05}
760: T. Fukui, Y. Hatsugai, and H. Suzuki,
761: J. Phys. Soc. Jpn. {\bf 74}, 1674 (2005).
762: %
763: \bibitem{FooNot2}
764: We have shifted $\theta_\mu$, for numerical conveniences.
765: %
766: \bibitem{FooNot1}
767: In this case, we
768: use discrete $k_2$ instead of $\theta_2$.
769: Then, $L_2=N_2$
770: means the number of the meshes for discrete $k_2$.
771: %
772: \bibitem{McCFal06}
773: For details of interlayer couplings, see, e.g,
774: E. McCann and V. I. Fal'ko,
775: Phys. Rev. Lett. {\bf 96}, 086805 (2006).
776: %
777: \bibitem{Hat93}
778: Y. Hatsugai, Phys. Rev. Lett. {\bf 71}, 3697 (1993).
779: %
780: \bibitem{FuKan06}
781: L. Fu and C. L. Kane,
782: cond-mat/0606336.
783: %
784: \bibitem{MooBal06}
785: J. E. Moore and L. Balents,
786: cond-mat/0607314.
787: \end{thebibliography}
788:
789: %\input{bib_aps}
790:
791: \end{document}
792: %
793: % ****** End of file apssamp.tex ******
794:
795: