cond-mat0308157/ms.tex
1: \documentclass[prb,twocolumn,aps,showpacs,eqsecnum,amsmath,
2: superscriptaddress]{revtex4}
3: \usepackage{graphicx,amsmath,amssymb,bm,times}
4: 
5: \begin{document}
6: 
7: \title{
8: Ground-state phase diagram of the one-dimensional half-filled extended Hubbard model
9: }
10: 
11: \author{M. Tsuchiizu}
12: \affiliation{Yukawa Institute for Theoretical Physics, 
13:          Kyoto University, Kyoto 606-8502, Japan}
14: \affiliation{Department of Physics, Nagoya University,
15:          Nagoya 464-8602, Japan}
16: \author{A. Furusaki}
17: \affiliation{Yukawa Institute for Theoretical Physics, 
18:          Kyoto University, Kyoto 606-8502, Japan}
19: \affiliation{Condensed-Matter Theory Laboratory,
20:     The Institute of Physical and Chemical Research (RIKEN),
21:          Saitama 351-0198, Japan}
22: \date{January 26, 2004}
23: 
24: \begin{abstract} 
25: We revisit the ground-state phase diagram of the one-dimensional
26:    half-filled extended Hubbard model with on-site ($U$) and 
27:    nearest-neighbor ($V$) repulsive interactions.
28: In the first half of the paper, using the weak-coupling
29:    renormalization-group approach ($g$-ology) 
30:    including second-order corrections to the coupling constants,
31:    we show that bond-charge-density-wave (BCDW)
32:    phase exists for $U\approx 2V$ in between
33:    charge-density-wave (CDW) and spin-density-wave (SDW) phases.
34: We find that the umklapp scattering of parallel-spin electrons
35:    disfavors the BCDW state and leads to a bicritical point
36:    where the CDW-BCDW and SDW-BCDW continuous-transition lines merge
37:    into the CDW-SDW first-order transition line.
38: In the second half of the paper, we investigate
39:    the phase diagram of the extended Hubbard model with either
40:    additional staggered site potential $\Delta$
41:    or bond alternation $\delta$.
42: Although the alternating site potential $\Delta$ strongly favors
43:    the CDW state (that is, a band insulator), the BCDW state is not
44:    destroyed completely and occupies a finite region
45:    in the phase diagram.
46: Our result is a natural generalization
47:    of the work by Fabrizio, Gogolin, and Nersesyan
48:    [Phys.\ Rev.\ Lett.\ \textbf{83}, 2014 (1999)], who predicted the
49:    existence of a spontaneously dimerized insulating state between
50:    a band insulator and a Mott insulator in the phase diagram of
51:    the ionic Hubbard model.
52: The bond alternation $\delta$ destroys the SDW state and changes it 
53:    into the BCDW state (or Peierls insulating state).
54: As a result the phase diagram of the model with $\delta$ contains
55:    only a single critical line
56:    separating the Peierls insulator phase and the CDW phase.
57: The addition of $\Delta$ or $\delta$ changes the universality class of 
58:    the CDW-BCDW transition from the Gaussian transition
59:    into the Ising transition.
60: \end{abstract}
61: 
62: \pacs{71.10.Fd, 71.10.Hf, 71.10.Pm, 71.30.+h}
63: 
64: \maketitle
65: 
66: \section{Introduction}
67: 
68: It is well known that 
69:   a one-dimensional (1D) spin system has instability to dimerization
70:   that changes the system into a nonmagnetic insulating state,
71:   the so-called spin-Peierls state. \cite{Bray}
72: Indeed the spin-Peierls state is realized in many systems including
73:    quasi-one-dimensional organic compounds \cite{Ishiguro,Kagoshima} 
74:    and the inorganic material\cite{Hase} CuGeO$_3$,
75:    and its properties have been studied extensively 
76:    both experimentally and theoretically.
77: Of particular interest is a situation in which a dimerized state
78:    appears spontaneously due to strong correlations
79:    and frustration.\cite{Sachdev}
80: A well-known example is the frustrated spin-$\frac{1}{2}$ Heisenberg
81:    chain with nearest-neighbor, $J_1$, and next-nearest-neighbor, $J_2$, 
82:    antiferromagnetic exchange interactions, where
83:    a spontaneously dimerized phase is realized
84:    for $J_2 \ge J_{2c}\simeq 0.24 J_1$.\cite{Majumdar_Ghosh}
85: Other systems of current interest are quasi-one-dimensional electron
86:    systems in organic materials, where the spin-Peierls state appears
87:    due to strong electron correlation at
88:    half filling
89:    \cite{Ukrainskii,Kivelson,Mazumdar,Hirsch83,Hara,%
90:     Sugiura2002,Malek,Sengpta2003} and 
91:    at quarter filling.\cite{Kuwabara,Sugiura2003}
92: 
93: Recently it was pointed out by Nakamura\cite{Nakamura} and co-workers that
94:    a spontaneously dimerized state occupies a finite parameter space
95:    in the ground-state phase diagram of the 1D half-filled 
96:    Hubbard model with the nearest-neighbor repulsion $V$, i.e.,
97:    the extended Hubbard model (EHM).
98: This spin-Peierls state is often called
99:    bond-charge-density-wave (BCDW) state or bond-ordered-wave state.
100: The appearance of the BCDW state in the purely electronic model is
101:    nontrivial and has attracted much attention from theoretical
102:    point of view.
103: To appreciate this surprising result, let us consider
104:    some limiting cases.
105: In the limit of weak nearest-neighbor repulsion $V$, or 
106:    in the half-filled Hubbard model with only the on-site Coulomb
107:    repulsion $U$, the ground state is in the Mott insulating state
108:    where the spin sector exhibits quasi-long-range order of
109:    spin-density wave (SDW); we call it the SDW state.
110: In the opposite limit of strong $V$,
111:    the ground state of the half-filled EHM has a long-range
112:    order of the charge-density wave (CDW);
113:    we call this state the CDW state.
114: Furthermore, in the atomic limit where the electron hopping $t$
115:   is ignored,
116:   the CDW state appears for $U<2V$ whereas the uniform
117:   state corresponding to the SDW state is stable for $U>2V$
118:   in one dimension.
119: Strong-coupling perturbation theory in $t$ has established
120:   that a first-order phase transition
121:   between the SDW state and the CDW state occurs at $U\simeq 2V$.
122:   \cite{Emery,Bari,Hirsch,Dongen}
123: As for the weak-coupling regime, perturbative renormalization-group
124:   (RG) approach or $g$-ology led to a similar conclusion that
125:    the ground state at half filling is either in the SDW state
126:    or in the CDW state with a continuous phase-transition line
127:    at $U=2V$.\cite{Emery}
128: Thus, it had been considered for a long time that
129:    the ground-state phase diagram of the EHM at half filling has
130:    only two phases, the SDW and CDW states, and that the order of
131:    the phase transition at $U\simeq2V$ changes from continuous to
132:    first order at a tricritical point which was speculated to exist
133:    in the intermediate coupling regime. 
134:    \cite{Hirsch,Cannon,Zhang,Voit}
135: This common view was revised by the Nakamura's discovery that
136:    the BCDW state exists at $U\simeq 2V$
137:    in between the SDW and CDW phases
138:    in the weak-coupling region,\cite{Nakamura}
139:    which is supported by recent large-scale Monte Carlo
140:    calculations.\cite{Sengupta,Sandvik}
141: Related studies of the dimerized state in
142:    the EHM with additional correlation effects can be found in
143:   Refs.\ [\onlinecite{Fukutome,Japaridze,Otsuka,Aligia,Nakamura2001}].
144: 
145: 
146: A related and still controversial issue of current interest is
147:    whether or not a spontaneously dimerized phase exists
148:    in the 1D Hubbard model with alternating site
149:    potential, the so-called ionic Hubbard
150:    model.\cite{Fabrizio,Tsuchiizu_JPSJ,Takada,Qin,Brune,YZZhang,Anusooya-Pati,%
151:    Wilkens,Torio,Refolio,Caprara,Gupta,Pozgajcic,Manmana}
152: This system was introduced as a simple minimal model for
153:     the neutral-ionic transitions observed in 
154:     quasi-one-dimensional organic materials\cite{MJRice,Nagaosa,Girlando}
155:    and for ferroelectric perovskites.\cite{Egami,Resta}
156: Obviously the model has two insulating phases.
157: The ground state is (i) a band insulator with the CDW order when
158:    the staggered site potential is much larger than the on-site
159:    repulsion or
160:    (ii) a Mott insulator with quasi-long-range SDW order
161:    when the staggered site potential is negligible.
162: Early exact diagonalization studies\cite{Egami,Resta,Gidopoulos}
163:    of small systems have found a transition between the two
164:    phases and also reported dramatic enhancement of the
165:    electron-lattice interaction
166:    by strong electron correlation near a boundary between
167:    the band insulating phase (the BI state) and the Mott
168:    insulating phase (the SDW state).
169: Mostly through bosonization analysis of the ionic Hubbard model,
170:    Fabrizio, Gogolin, and Nersesyan recently argued\cite{Fabrizio}
171:    that a phase of a spontaneously dimerized insulator (SDI)
172:    intervenes between the ionic insulating phase (band insulator)
173:    and the Mott insulating phase.
174: The SDI state is closely related to the BCDW state mentioned above.
175: Earlier numerical studies
176:   \cite{Takada,Qin,Brune,Anusooya-Pati,Wilkens,Gidopoulos}
177:    have drawn contradictory conclusions as to whether
178:    the SDI phase exists or not, but more recent numerical studies
179:    find two phase transitions and the SDI phase
180:    in between.\cite{Torio,YZZhang,Manmana}
181: Nevertheless there still remain unresolved issues on
182:    the critical properties near the quantum phase transitions.
183: 
184: In this paper we give supporting theoretical arguments for the existence
185:    of the spontaneously dimerized insulating states in
186:    the 1D half-filled extended Hubbard model with and without
187:    staggered potentials.
188: We adopt the standard bosonization approach and perform both perturbative
189:    RG analysis valid in the weak-coupling regime and semiclassical
190:    analysis which is expected to give a qualitatively correct picture
191:    even in the strong-coupling regime.
192: This paper is organized as follows.
193: Sections \ref{sec:model} and \ref{sec:phase_diagram} are 
194:    devoted to the analysis of the standard EHM, i.e., the system 
195:    without the staggered potential.
196: Some of the results of this part are already presented 
197:    in Ref.\ \onlinecite{Tsuchiizu_Furusaki}.
198: In Sec.\ \ref{sec:model},
199:    we introduce the model and 
200:    reformulate the weak-coupling theory, the $g$-ology,
201:    to include higher-order corrections to coupling constants.
202: We bosonize low-energy effective Hamiltonian
203:    and derive the renormalization-group equations.
204: In Sec.\ \ref{sec:phase_diagram},
205:    we determine the ground-state phase diagram. 
206: First, from the perturbative RG analysis we show that 
207:    the BCDW phase occupies a finite region near the $U=2V$ line
208:    in the weak-coupling limit.
209: Next, from the semiclassical analysis
210:    we argue that the umklapp scattering 
211:    of parallel-spin electrons
212:    destabilizes the BCDW phase and gives rise to a bicritical point
213:    where the CDW-BCDW and SDW-BCDW continuous-transition lines merge
214:    into the CDW-SDW first-order transition line.
215: Finally,
216:    combining the perturbative RG equations with the semiclassical
217:    analysis, we obtain the global phase diagram of the 1D EHM.
218: In Sec.\ \ref{sec:staggered}
219:    we study the 1D EHM with the staggered site potential.
220: We take the same strategy as in the previous sections and
221:    perform a semiclassical analysis of the bosonized Hamiltonian.
222: With the help of the perturbative RG analysis
223:    we obtain the global phase diagram that indeed has the SDI phase.
224: We find that the BCDW phase of the EHM is continuously deformed to
225:    the SDI phase
226:    upon introducing the alternating site potential.
227: In Sec.\ \ref{sec:dimer}, we study the 1D EHM with additional bond
228:    dimerization, but without the staggered potential.
229: This model exhibits a quantum phase transition between a dimerized
230:    Peierls insulator and a CDW state.
231: Section \ref{sec:conclusions} is devoted to conclusions,
232:    and details of the technical calculations are given in 
233:    Appendixes.
234: 
235: 
236: \section{Extended Hubbard Model}\label{sec:model}
237: 
238: In the first half of this paper 
239:    (Secs.\ \ref{sec:model} and \ref{sec:phase_diagram}),
240:    we consider the standard 1D EHM which has on-site, $U$, and
241:    nearest-neighbor, $V$, interactions.
242: The Hamiltonian is given by
243: %================================================================
244: \begin{eqnarray}
245: H &=&
246:  - t \sum_{j,\sigma}
247:    ( c_{j,\sigma}^\dagger c_{j+1,\sigma}^{} + \mathrm{H.c.} )
248: \nonumber \\
249: &&{}
250:  + U \sum_j n_{j,\uparrow} \, n_{j,\downarrow}
251:    + V \sum_j n_{j} \, n_{j+1} ,
252: \label{eq:H1D}
253: \end{eqnarray}
254: %================================================================
255:   where
256:   $n_{j,\sigma} \equiv c_{j,\sigma}^\dagger c_{j,\sigma}^{}
257:    -\frac{1}{2}$,
258:   $n_j \equiv n_{j,\uparrow} + n_{j,\downarrow}$, and
259:   $c_{j,\sigma}^\dagger$ denotes the creation operator of an
260:   electron with spin $\sigma$ 
261:   ($=\,\uparrow$, $\downarrow$) on the $j$th site.
262: We assume repulsive interactions, i.e.,
263:    the coupling constants $U$ and $V$ are positive.
264: Note that the Hamiltonian has global SU(2) spin symmetry.
265: Following the previous studies on models with correlated-hopping
266:   interactions,\cite{Japaridze} we consider the CDW, SDW, BCDW, and 
267:   bond-spin-density-wave (BSDW) phases as potential ordered ground
268:   states at half filling.
269: They are characterized by the order parameters
270: %================================================================
271: \begin{subequations}
272: \begin{eqnarray}
273: \mathcal{O}_\mathrm{CDW} \!\!&\equiv&\!\! (-1)^j \,
274:    (n_{j,\uparrow}+n_{j,\downarrow})
275: ,
276: \\
277: \mathcal{O}_\mathrm{SDW} \!\!&\equiv&\!\! (-1)^j \,
278:    (n_{j,\uparrow}-n_{j,\downarrow})
279: ,
280: \\
281: \mathcal{O}_\mathrm{BCDW} \!\!&\equiv&\!\! (-1)^j 
282:      (c^\dagger_{j,\uparrow}c_{j+1,\uparrow}^{}
283:       +c^\dagger_{j,\downarrow}c_{j+1,\downarrow}^{}
284:       +\mathrm{H.c.})
285: , \quad\quad
286: \\
287: \mathcal{O}_\mathrm{BSDW} \!\!&\equiv&\!\! (-1)^j
288:      (c^\dagger_{j,\uparrow}c_{j+1,\uparrow}^{}
289:       -c^\dagger_{j,\downarrow}c_{j+1,\downarrow}^{}
290:       +\mathrm{H.c.})
291: . \quad\quad
292: \end{eqnarray}
293: \label{eq:order_parameters}%
294: \end{subequations}
295: %================================================================
296: The order parameter of the BCDW state corresponds to the 
297:    Peierls dimerization operator. 
298: We note that the BCDW state can be also regarded as the
299:    $p$-density-wave state,\cite{Nayak}
300:    as the order parameter of the BCDW state can be written as
301:    $\sum_j \mathcal{O}_\mathrm{BCDW}\propto \sum_{k,\sigma} \sin(ka)\, 
302:      c_{k,\sigma}^\dagger \, c_{k+(\pi/a),\sigma}^{}$, where
303:    $c_{k,\sigma} = N^{-1/2} \sum_j e^{-ikR_j} c_{j,\sigma}$
304:    with $R_j=ja$ ($a$: the lattice spacing, $N$: the number of sites). 
305: The BSDW state describes a site-off-diagonal SDW state.\cite{Japaridze}
306: %================================================================
307: \begin{figure}[t]
308: \includegraphics[width=7cm]{fig01}
309: \caption{
310: Single-particle energy band.
311: The annihilation operator of an electron near the Fermi points with
312:    momentum
313:    $k\in [-k_F-\Lambda,-k_F+\Lambda]$
314:    ($k\in[k_F-\Lambda,k_F+\Lambda]$) is denoted $a_{k,-,\sigma}$ 
315:    ($a_{k,+,\sigma}$), and that of an electron far away from the
316:    Fermi points is denoted $b_{k,\sigma}$.
317: }
318: \label{fig:band}
319: \end{figure}
320: %================================================================
321: 
322: \subsection{$g$-ology approach}
323: 
324: The hopping $t$ generates the energy band with dispersion
325:    $\varepsilon_k=-2t \cos ka$, where
326:    the Fermi points are at $k=\pm k_F=\pm \pi/2a$ 
327:    at half filling.
328: In order to analyze the low-energy physics
329:    near the Fermi points,
330:    we introduce a momentum cutoff $\Lambda$ $(0<\Lambda<k_F)$
331:    and divide the momentum space into the three sectors 
332:    (Fig.\ \ref{fig:band})
333:    (i) $k\in R$, (ii) $k\in L$, and (iii) $k \notin (R\cup L)$, 
334:    where $R=[k_F-\Lambda,k_F+\Lambda]$ and
335:    $L=[-k_F-\Lambda,-k_F+\Lambda]$.
336: We then introduce the following fermion operators: 
337: %================================================================
338: \begin{equation}
339:  c_{k,\sigma}
340: =
341: \left\{
342: \begin{array}{cl}
343:    a_{k,+,\sigma} & \quad  \mbox{for $k\in R$}  \\
344:    a_{k,-,\sigma} & \quad  \mbox{for $k \in L$} \\
345:     b_{k,\sigma}  & \quad  \mbox{otherwise}.
346: \end{array}
347: \right.
348: \end{equation}
349: %================================================================
350: Electrons near the Fermi points are shuffled by the two-particle
351:    scattering:
352:    $H_{\mathrm{int}}=U\sum_j n_{j,\uparrow} \, n_{j,\downarrow}+
353:     V \sum_j n_{j} \, n_{j+1}$.    
354: Following the standard $g$-ology approach,\cite{Emery,Solyom}
355:    we will focus on the scattering processes 
356:    between electrons near the Fermi points,
357:    i.e.,
358:    the scattering processes which involve
359:    $a_{k,\pm,\sigma}$ only.
360: The Hamiltonian for such interaction processes is
361: %================================================================
362: \begin{eqnarray}
363: H_\mathrm{int}
364: &\!\!=\!\!& 
365: {} +
366: \frac{g_{1\parallel}}{2L}
367: \sum_{k_i,p,\sigma}
368: :\!a_{k_1,p,\sigma}^\dagger \,\,  a_{k_2,-p,\sigma}^{} \,\,
369:    a_{k_3,-p,\sigma}^\dagger \,\, a_{k_4,p,\sigma}^{} \!:
370: \nonumber \\ &&
371: {}+
372: \frac{g_{1\perp}}{2L}
373: \sum_{k_i,p,\sigma}
374: :\!a_{k_1,p,\sigma}^\dagger \,\,  a_{k_2,-p,\sigma}^{} \,\,
375:    a_{k_3,-p,\overline{\sigma}}^\dagger \,\, 
376:    a_{k_4,p,\overline{\sigma}}^{} \!:
377: \nonumber \\ &&
378: {} +
379: \frac{g_{2\parallel}}{2L}
380: \sum_{k_i,p,\sigma}
381: :\! a_{k_1,p,\sigma}^\dagger \,\, a_{k_2,p,\sigma}^{} \,\,
382:    a_{k_3,-p,\sigma}^\dagger \,\, a_{k_4,-p,\sigma}^{} \!: 
383: \nonumber \\ &&
384: {} +
385: \frac{g_{2\perp}}{2L}
386: \sum_{k_i,p,\sigma}
387: :\! a_{k_1,p,\sigma}^\dagger \,\, a_{k_2,p,\sigma}^{} \,\,
388:    a_{k_3,-p,\overline{\sigma}}^\dagger \,\, 
389:    a_{k_4,-p,\overline{\sigma}}^{} \!: 
390: \nonumber \\ &&
391: {} +
392: \frac{g_{3\parallel}}{2L}
393: \sum_{k_i,p,\sigma}
394: :\!a_{k_1,p,\sigma}^\dagger \,\, a_{k_2,-p,\sigma}^{} \,\,
395:    a_{k_3,p,\sigma}^\dagger \,\, a_{k_4,-p,\sigma}^{} \!:
396: \nonumber \\ &&
397: {} +
398: \frac{g_{3\perp}}{2L}
399: \sum_{k_i,p,\sigma}
400: :\!a_{k_1,p,\sigma}^\dagger \,\, a_{k_2,-p,\sigma}^{} \,\,
401:    a_{k_3,p,\overline{\sigma}}^\dagger \,\, 
402:    a_{k_4,-p,\overline{\sigma}}^{} \!:
403: \nonumber \\ &&
404: {} +
405: \frac{g_{4\parallel}}{2L}
406: \sum_{k_i,p,\sigma}
407: :\! a_{k_1,p,\sigma}^\dagger \,\, a_{k_2,p,\sigma}^{} \,\,
408:     a_{k_3,p,\sigma}^\dagger \,\, a_{k_4,p,\sigma}^{} \!:
409: \nonumber \\ &&
410: {} +
411: \frac{g_{4\perp}}{2L}
412: \sum_{k_i,p,\sigma}
413: :\! a_{k_1,p,\sigma}^\dagger \,\, a_{k_2,p,\sigma}^{} \,\,
414:     a_{k_3,p,\overline{\sigma}}^\dagger \,\, 
415:     a_{k_4,p,\overline{\sigma}}^{} \!:
416: ,
417: \nonumber \\
418: \label{eq:HI_fourier}
419: \end{eqnarray}
420: %================================================================
421:    where $\overline{\sigma}={} \downarrow (\uparrow)$ for
422:    $\sigma={}\uparrow (\downarrow)$, $L$ is the
423:    length of the system, and $:\ :$ denotes normal ordering.
424: The summation over the momentum $k_i$ is taken 
425:    under the condition of the total momentum being conserved
426:    (equal to $\pm2\pi/a$ for the umklapp scattering).  
427: The index $p=+/-$ denotes the right-/left-moving electron.
428: The coupling constants $g_{1\parallel}$ and $g_{1\perp}$ 
429:    ($g_{3\parallel}$ and $g_{3\perp}$) denote the matrix elements of
430:    the backward (umklapp) scattering, while
431:    $g_{2\parallel}$ and $g_{2\perp}$ ($g_{4\parallel}$ and $g_{4\perp}$)
432:    denote the matrix element of the forward scattering
433:    with the different (same) branch $p=\pm$.
434: The index $\parallel$ ($\perp$) of the coupling constants
435:    denotes the scattering of
436:    electrons with same (opposite) spins.
437: 
438: 
439: \subsection{Vertex corrections}
440: 
441: 
442: In the conventional weak-coupling approach to the
443:    1D EHM,\cite{Emery,Nakamura}
444:    one estimates 
445:    the coupling constants in Eq.\ (\ref{eq:HI_fourier})
446:    only up to the lowest order in $U$ and $V$:
447: \begin{subequations}
448: \begin{eqnarray}
449:    g_{1\perp}&=&g_{3\perp}=(U-2V)a, \\
450:    g_{2\perp}&=&g_{4\perp}=(U+2V)a, \\
451:    g_{1\parallel}&=&g_{3\parallel}=-2Va, \\
452:    g_{2\parallel}&=&g_{4\parallel}=+2Va.
453: \end{eqnarray}
454: \label{eq:g_lowest}%
455: \end{subequations}
456: In analyzing the low-energy physics of 
457:    Eq.\ (\ref{eq:HI_fourier}),
458:    one then employs the standard $g$-ology approach,\cite{Solyom}
459:    i.e., the perturbative RG method, and 
460:    obtains flow equations for the marginal terms 
461:    in Eq.\ (\ref{eq:HI_fourier}). 
462: From this RG analysis\cite{Emery,Solyom}
463:    one finds
464:    that the $g_{3\perp}$ term generates a gap in the charge excitation
465:    spectrum
466:    if $|g_{3\perp}|>-(g_{2\parallel}+g_{2\perp}-g_{1\parallel})$
467:    and $g_{3\perp}\ne0$,
468:    whereas the $g_{1\perp}$ term yields a gap in the spin excitation
469:    spectrum if
470:    $|g_{1\perp}|>-(g_{2\parallel}-g_{2\perp}-g_{1\parallel})$
471:    and $g_{1\perp}\ne0$.
472: Hence, with the lowest-order coupling constants 
473:    Eq.\ (\ref{eq:g_lowest}), one would conclude that
474:    the charge (spin) excitations become massless at $U-2V=0$ 
475:    ($U-2V \ge 0$).
476: This would mean that, as $U$ increases, 
477:    both the charge and spin sectors become critical 
478:    simultaneously at $U=2V$,
479:    where a direct and continuous CDW-SDW transition takes place.
480: This analysis is found to be insufficient from the following argument.
481: The (accidental) simultaneous vanishing
482:    of $g_{3\perp}$ and $g_{1\perp}$ results from
483:    the lowest-order estimate in $U$ and $V$
484:    and there is no symmetry principle that enforces $g_{1\perp}$ and
485:    $g_{3\perp}$ to vanish simultaneously.
486: It is possible that
487:    the higher-order corrections to $g$ lift the degeneracy of
488:    zeros and change the topology of the phase diagram.
489: Therefore, in order to analyze the phase diagram at $U\approx 2V$,
490:    we need to go beyond the lowest-order calculation
491:    of the coupling constants in the $g$-ology.
492: In this section, we compute the vertex corrections due to
493:    virtual processes involving high-energy states \cite{Penc_Mila}
494:    by integrating out $b_{k,\sigma}$.
495: This procedure allows us to 
496:    obtain the effective coupling constants $g$'s
497:    that include higher-order corrections. 
498: 
499: 
500: 
501: The second-order vertex diagrams for the coupling constants
502:    are shown in Fig.\ \ref{fig:vertex}.
503: The solid lines denote the low-energy states $a_{k,\pm,\sigma}$,  
504:    while the dashed lines denote
505:    high-energy states $b_{k,\sigma}$.
506: The nonzero contributions from the second-order virtual processes (a)-(e)
507: are
508: %================================================================
509: \begin{subequations}
510: \begin{eqnarray}
511: \delta g_{1\perp}^{(a)}
512: &=&
513: -\delta g_{3\perp}^{(b)}
514: =
515:  -  \frac{U^2}{4\pi t} D_1 a + \frac{V^2}{\pi t} D_2 a , 
516: \\
517: \delta g_{1\perp}^{(c)}
518: &=&
519: + \delta g_{3\perp}^{(c)}
520: =
521:  + \frac{V(U-2V)}{\pi t} D_1 a , 
522: \\
523: \delta g_{2\perp}^{(a)}
524: &=&
525: -\delta g_{2\perp}^{(b)}
526: =
527:  - \frac{U^2}{4\pi t} D_1 a - \frac{V^2}{\pi t} D_2 a, 
528: \\
529: \delta g_{1\parallel}^{(a)}
530: &=&
531:  +  \frac{V^2}{\pi t} D_2 a ,  
532: \\
533: \delta g_{1\parallel}^{(c)}
534: &=&
535:  - \frac{(U-2V)^2+4V^2}{4\pi t} D_1 a 
536:  - \frac{V^2}{\pi t} D_2 a, 
537: \\
538:  \delta g_{2\parallel}^{(a)}
539: &=&
540:  - \frac{V^2}{\pi t} D_2 a, 
541: \\
542: \delta g_{3\parallel}^{(c)}
543: &=&
544:  - \frac{(U-2V)^2+4V^2}{4\pi t} D_1 a 
545:  + \frac{V^2}{\pi t} D_2 a, 
546: \end{eqnarray}
547: \end{subequations}
548: %================================================================
549:   where
550: %================================================================
551: \begin{subequations}
552: \begin{eqnarray}
553: D_1(\Lambda) &\equiv& 
554: \int_{-\pi/2+a\Lambda}^{\pi/2-a\Lambda} \frac{dk}{\cos k}
555: ,
556: \\
557: D_2(\Lambda) &\equiv& 
558: \int_{-\pi/2+a\Lambda}^{\pi/2-a\Lambda} dk \,\frac{\sin^2 k}{\cos k}
559: .
560: \end{eqnarray} 
561: \end{subequations}
562: %================================================================
563: By introducing  $C_1(\Lambda)\equiv 2\ln [\cot (a\Lambda/2)]$ and
564:     $C_2(\Lambda)\equiv 2\cos a\Lambda $,
565:     $D_1(\Lambda)$ and $D_2(\Lambda)$ are rewritten as
566:   $D_1(\Lambda)=C_1(\Lambda)$ and 
567:   $D_2(\Lambda)=C_1(\Lambda)-C_2(\Lambda)$.
568: In terms of $C_1$ and $C_2$, the coupling constants 
569:    with second-order corrections are given by
570: %================================================================
571: \begin{figure}[t]
572:  \includegraphics[width=8cm]{fig02}
573: \caption{
574: Vertex diagrams with second-order corrections [(a)-(e)].
575: Solid lines denote electron states
576:   in the momentum space $k \in R$ or $k \in L$, 
577:   while the dashed lines denote electron
578:   states in the other momentum space.
579: }
580: \label{fig:vertex}
581: \end{figure}
582: %================================================================
583: %================================================================
584: \begin{subequations}
585: \begin{eqnarray}
586: g_{1\perp} \!\! &=& \!\! 
587: (U-2V)a \left[1
588: -\frac{C_1}{4\pi t}(U-2V) \right]
589: -\frac{C_2}{\pi t}V^2a ,
590: \hspace*{1cm}
591: \label{eq:g1perp}
592: \\
593: g_{1\parallel} \!\! &=& \!\! 
594: -2Va 
595: -\frac{C_1}{4\pi t}(U-2V)^2a
596: -\frac{C_2}{\pi t}V^2a ,
597: \\
598: g_{3\perp} \!\!  &=& \!\! 
599: (U-2V)a \left[1 
600: +\frac{C_1}{4\pi t}(U+6V)\right]
601: +\frac{C_2}{\pi t}V^2a ,
602: \label{eq:g3perp}
603: \\
604: g_{3\parallel} \!\!  &=& \!\! 
605: -2Va 
606: -\frac{C_1}{4\pi t}(U-2V)^2 a
607: +\frac{C_2}{\pi t}V^2a ,
608: \end{eqnarray}
609: \label{eq:g}%
610: \end{subequations}
611: %================================================================
612:    and $g_{2\parallel} = + 2Va$, $g_{2\perp} = (U+2V)a$,
613:    $g_{4\parallel} = + 2Va$, and $g_{4\perp} = (U+2V)a$.
614: Except when $a\Lambda\ll1$, the $C_i$s depend on $\Lambda$ only weakly,
615:    and we can set $\Lambda=\pi/4$ in the following analysis
616:    as we are interested in the qualitative feature of the phase
617:    diagram (different choices will only lead
618:    to small quantitative changes in phase boundaries).
619: Incidentally, the logarithmic divergence of $C_1(\Lambda)$ in the
620:    limit $\Lambda\to0$ leads to the familiar one-loop RG equations.
621: 
622: 
623: \subsection{Bosonization}
624: 
625: Having integrated out the high-energy virtual scattering processes,
626:   we now focus on the low-energy states and
627:   linearize the dispersion of $a_{k,\pm,\sigma}$ around
628:   the Fermi points. 
629: The kinetic-energy term with the linearized dispersion is given by
630: %================================================================
631: \begin{eqnarray}
632: H_{0}
633: &=&
634: \sum_{k\in R,\sigma} 
635: v_F (k-k_F) \, a_{k,+,\sigma}^\dagger \, a_{k,+,\sigma}
636: \nonumber \\ && {}
637: + \sum_{k\in L,\sigma} 
638: v_F (-k-k_F) \, a_{k,-,\sigma}^\dagger \, a_{k,-,\sigma}
639: ,
640: \label{eq:H0_linear}
641: \end{eqnarray}
642: %================================================================
643:    where $v_F=2ta$ is the Fermi velocity. 
644: The field operators of the right- and left-moving electrons 
645:    are given by
646: %================================================================
647: \begin{subequations}
648: \begin{eqnarray}
649: \psi_{+,\sigma}(x) &\equiv&
650: \frac{1}{\sqrt{L}} \sum_{k\in R}
651: e^{ikx} \, a_{k,+,\sigma}
652: ,
653: \\
654: \psi_{-,\sigma}(x) &\equiv&
655: \frac{1}{\sqrt{L}} \sum_{k\in L}
656: e^{ikx} \, a_{k,-,\sigma}
657: .
658: \end{eqnarray}
659: \label{eq:psi_fourier_trans}%
660: \end{subequations}
661: %================================================================
662: We apply the Abelian bosonization method and rewrite the
663:    kinetic-energy term $H_0=\int dx \, \mathcal{H}_0$ in terms of bosonic
664:    phase fields as (see Appendix \ref{sec:bosonization}) 
665: %================================================================
666: \begin{eqnarray}
667: \mathcal{H}_{0}
668: &=&
669: \frac{v_F}{4\pi}
670: \left[
671:  \left(2\pi \Pi_{\theta}\right)^2 + \left(\partial_x \theta\right)^2
672: \right]
673: \nonumber \\
674: && {}
675: +
676: \frac{v_F}{4\pi}
677: \left[
678:  \left(2\pi \Pi_{\phi}\right)^2 + \left(\partial_x \phi\right)^2
679: \right]
680: ,
681: \label{eq:H0}
682: \end{eqnarray}
683: %================================================================
684:    where $\theta$ ($\phi$) is the bosonic field whose spatial
685:    derivative is proportional to the charge (spin) density,
686:    $[\theta(x),\phi(y)]=0$.
687: The operators $\Pi_\theta$ and $\Pi_\phi$ are canonically conjugate
688:    variables to $\theta$ and $\phi$, respectively, and satisfy
689:    the conventional commutation relations,
690:    $[\theta(x),\Pi_\theta(x')]=[\phi(x),\Pi_\phi(x')]=i\delta(x-x')$.
691: We also introduce chiral bosonic fields
692: %================================================================
693: \begin{eqnarray}
694: \theta_\pm (x)
695: &\equiv&
696: \frac{1}{2}
697: \left[
698:   \theta(x) \mp 2\pi \int_{-\infty}^x dx' \, \Pi_\theta(x') 
699: \right]
700: ,
701: \label{eq:chiral_theta}
702: \\
703: \phi_\pm (x)
704: &\equiv&
705: \frac{1}{2}
706: \left[
707:   \phi(x) \mp 2\pi \int_{-\infty}^x dx' \, \Pi_\phi(x') 
708: \right]
709: .
710: \label{eq:chiral_phi}
711: \end{eqnarray}
712: %================================================================
713: One can easily verify that these chiral fields
714:    satisfy the commutation relations
715:   $[\theta_\pm (x),\theta_\pm (x')]=
716:    [\phi_\pm (x),\phi_\pm (x')]=\pm i (\pi/2) \, \mathrm{sgn}(x-x')$ and
717:   $[\theta_+ (x),\theta_- (x')]=
718:    [\phi_+ (x),\phi_- (x')]= i\pi/2$.
719: In terms of these fields,
720:    the kinetic-energy density reads
721: %================================================================
722: \begin{equation}
723: \mathcal{H}_{0}
724: =
725: \frac{v_F}{2\pi}
726: \sum_{p=+,-}
727: \left[
728:  \left(\partial_x \theta_p\right)^2 + \left(\partial_x \phi_p\right)^2
729: \right]
730: .
731: \end{equation}
732: %================================================================
733: 
734: To express the electron field operators $\psi_{p,\sigma}$ 
735:    with the bosonic phase fields,
736:    we introduce a new set of chiral bosonic fields 
737: %================================================================
738: \begin{equation}
739: \varphi_{p,\uparrow} = \theta_p + \phi_p, \quad
740: \varphi_{p,\downarrow} = \theta_p - \phi_p, 
741: \label{eq:varphi}
742: \end{equation}
743: %================================================================
744:    which obey the commutation relations 
745: %================================================================
746: \begin{subequations}
747: \begin{eqnarray}
748: \left[ \varphi_{\pm,\sigma}(x),\varphi_{\pm,\sigma'}(x') \right]
749:    &=&
750:  \pm i \pi \,{\rm sgn}(x-x') \, \delta_{\sigma,\sigma'}
751: , \quad\quad
752: \\
753: \left[\varphi_{+,\sigma}(x),\varphi_{-,\sigma'}(x') \right]
754: &=&
755:  i\pi \, \delta_{\sigma,\sigma'}
756: .
757: \end{eqnarray}
758: \label{eq:commutation_varphi}
759: \end{subequations}
760: %================================================================
761: In terms of the phase fields $\varphi_{p,\sigma}$,
762:    the electron field operators can be written as
763: %================================================================
764: \begin{equation}
765: \psi_{p,\sigma}(x)
766:   =
767:    \frac{\eta_\sigma}{\sqrt{2\pi a}}
768:    \exp\left[ipk_F x+ip \, \varphi_{p,\sigma}(x)\right],
769: \label{eq:field_op}
770: \end{equation}
771: %================================================================
772: where the Klein factor $\eta_\sigma$ satisfies the anticommutation
773:    relation $\{\eta_\sigma, \eta_{\sigma'}\}=2\delta_{\sigma,\sigma'}$.
774: One can verify that the operator defined in Eq.\ (\ref{eq:field_op})
775:    satisfies the same anticommutation relation as the
776:    fermion field operator. 
777: It follows from Eq.\ (\ref{eq:field_op}) that
778:    the order parameters in Eq.\ (\ref{eq:order_parameters})
779:     are rewritten as
780: %================================================================
781: \begin{subequations}
782: \begin{eqnarray}
783: \mathcal{O}_\mathrm{SDW}(x)
784: &\propto&
785: \cos\theta(x) \, \sin \phi(x),
786: \label{eq:ordersdw}
787: \\
788: \mathcal{O}_\mathrm{CDW}(x)
789: &\propto&
790: \sin\theta(x) \, \cos \phi(x),
791: \label{eq:ordercdw}
792: \\
793: \mathcal{O}_\mathrm{BCDW}(x)
794: &\propto&
795: \cos\theta(x) \, \cos \phi(x),
796: \\
797: \mathcal{O}_\mathrm{BSDW}(x)
798: &\propto&
799: \sin\theta(x) \, \sin \phi(x).
800: \label{eq:orderbsdw}
801: \end{eqnarray}
802: \label{eq:orderparam}
803: \end{subequations}
804: %================================================================
805: 
806: The interaction part of the Hamiltonian  
807:    $H_\mathrm{int}$, Eq.\ (\ref{eq:HI_fourier}), can be also
808:    expressed in terms of the boson fields $\theta_\pm$ and $\phi_\pm$.
809: It has been suggested that, besides the marginal operators,
810:    operators with  higher scaling dimensions can play an important role
811:    in the first-order CDW-SDW transition \cite{Cannon,Voit}
812:    which is known to occur in the strong-coupling region
813:    of the 1D EHM. \cite{Emery,Bari,Hirsch,Dongen}
814: We thus include all the terms of scaling dimension 4
815:    $[=2\, (\mbox{charge sector}) + 2\, (\mbox{spin sector})]$.
816: We also note that there are some complications and subtleties in
817:    bosonizing the off-site interaction term,
818:    i.e., the nearest-neighbor interaction term $V$
819:    (see Appendix \ref{sec:bosonization} for detail).
820: We obtain the bosonized Hamiltonian density,
821: %================================================================
822: \begin{eqnarray}
823: \mathcal{H} &=&
824: \frac{1}{2\pi}\sum_{p=+,-}
825:  \left[v_\rho(\partial_x\theta_p)^2
826:       +v_\sigma(\partial_x\phi_p)^2\right]
827: \nonumber \\ && {}
828: + \frac{g_\rho}{2\pi^2}
829:      \left(\partial_x \theta_+ \right)
830:      \left(\partial_x \theta_- \right)
831: - \frac{g_\sigma}{2\pi^2}
832:      \left(\partial_x \phi_+ \right) 
833:      \left(\partial_x \phi_- \right)
834: \nonumber \\ && {}
835: -\frac{g_c}{2\pi^2 a^2} \, \cos 2 \theta 
836: +\frac{g_s}{2\pi^2 a^2} \, \cos 2 \phi
837: \nonumber \\ && {}
838: -\frac{g_{cs}}{2\pi^2 a^2} \, \cos 2\theta  \,  \cos 2\phi
839: \nonumber \\ && {}
840: -\frac{g_{\rho s}}{2 \pi^2}
841:      \left(\partial_x \theta_+ \right) 
842:      \left(\partial_x \theta_- \right) \, \cos 2\phi
843: \nonumber \\ && {}
844: +\frac{g_{c\sigma}}{2 \pi^2}
845:      \left(\partial_x \phi_+\right) 
846:      \left(\partial_x \phi_- \right) \, \cos 2\theta 
847: \nonumber \\ && {}
848: +\frac{g_{\rho \sigma}}{2\pi^2}  \, a^2 
849:      \left(\partial_x \theta_+ \right) 
850:      \left(\partial_x \theta_- \right) 
851:      \left(\partial_x \phi_+ \right) 
852:      \left(\partial_x \phi_- \right) 
853: .
854: \label{eq:Hamiltonian}
855: \end{eqnarray}
856: %================================================================
857: The renormalized velocities are
858:    $v_\rho=2ta+(g_{4\parallel}+g_{4\perp}-g_{1\parallel})/2\pi$ and
859:    $v_\sigma=2ta+(g_{4\parallel}-g_{4\perp}-g_{1\parallel})/2\pi$.
860: The marginal terms with the couplings $g_{\rho}$ and $g_{c}$ 
861:    ($g_{\sigma}$ and $g_{s}$)
862:    determine low-energy properties of the charge (spin)
863:    modes, \cite{Emery,Solyom}
864:    where 
865:    $g_\rho=g_{2\perp}+g_{2\parallel}-g_{1\parallel}$,
866:    $g_c=g_{3\perp}$,
867:    $g_\sigma=g_{2\perp}-g_{2\parallel}+g_{1\parallel}$,
868:    and $g_s=g_{1\perp}$.
869: The $g_{cs}$, $g_{\rho s}$, $g_{c\sigma}$, and $g_{\rho\sigma}$ terms
870:    with scaling dimension 4
871:    couple the spin and charge degrees of freedom.
872: The $g_{cs}$ coupling comes from the 
873:    umklapp scattering $g_{3\parallel}$.
874: The $g_{\rho s}$ ($g_{\rho\sigma}$) coupling is generated 
875:    from the backward scattering of antiparallel- (parallel-) spin
876:    electrons while the $g_{c\sigma}$ coupling is generated
877:    from the umklapp scattering of electrons with antiparallel spins
878:    (see Appendix \ref{sec:bosonization}).
879: These coupling constants are given by
880:    $g_{cs}=g_{\rho s}=g_{c\sigma}=g_{\rho\sigma}=-2Va$ to lowest order
881:    in $V$.
882: Cannon and Fradkin examined the effect of the $g_{cs}$ term
883:    and argued that it plays a crucial role in 
884:    the first-order CDW-SDW transition.\cite{Cannon}
885: Voit included the $g_{\rho s}$ and $g_{c\sigma}$ terms, as well as the
886:    $g_{cs}$ term, in the perturbative RG analysis of the coupling
887:    constants, but did not consider the $g_{\rho\sigma}$ term.\cite{Voit}
888: Here we note that
889:    it is important to keep the $g_{\rho\sigma}$ term as well, 
890:    since the global SU(2) symmetry in the spin sector 
891:    is guaranteed only when $g_\sigma=g_s$,
892:    $g_{cs}=g_{c\sigma}$, and $g_{\rho s}=g_{\rho\sigma}$.
893: 
894: 
895: \subsection{Renormalization-group equations}
896: 
897: We perform a perturbative RG calculation to examine the low-energy
898:    properties of the 1D EHM in the weak-coupling regime, taking into
899:    account quantum fluctuations of the phase fields.
900: The operator product expansion (OPE) technique allows us
901:    to systematically handle the higher-order terms in the bosonized
902:    Hamiltonian (\ref{eq:Hamiltonian}).
903: The one-loop RG equations that describe changes in the coupling
904:    constants during the scaling of the short-distance cutoff
905:    ($a\to ae^{dl}$) are  given by
906:    (see Appendix \ref{sec:rg} for their derivation)
907: %================================================================
908: \begin{eqnarray}
909: \frac{d}{dl} G_\rho 
910:   &=& {}
911:     + 2 \, G_c^2 + G_{cs}^2 +  G_s \, G_{\rho s} ,
912: \label{eq:Grho}
913: \\
914: \frac{d}{dl} G_c 
915:   &=& {}
916:     + 2 \, G_\rho\, G_c - G_s \, G_{cs} - G_{cs} \, G_{\rho s} ,
917: \label{eq:Gc}
918: \\
919: \frac{d}{dl} G_s 
920:   &=& {}
921:     - 2 \, G_s^2 - G_c \, G_{cs} - G_{cs}^2 ,
922: \label{eq:Gs}
923: \\
924: \frac{d}{dl} G_{cs} 
925:   &=& {}
926:     - 2 \, G_{cs} + 2 \, G_\rho \, G_{cs} - 4 \, G_s \, G_{cs}
927: \nonumber \\ && {}
928:     - 2 \, G_c\, G_{s}
929:     - 2 \, G_c\, G_{\rho s}
930:     - 4 \, G_{cs} \, G_{\rho s} , \quad\quad
931: \label{eq:Gcs}
932: \\
933: \frac{d}{dl} G_{\rho s} 
934:   &=& {}
935:     - 2 \, G_{\rho s} + 2 \, G_\rho \, G_s
936: \nonumber \\ && {}
937:     - 4 \, G_c \, G_{cs} - 4 \, G_{cs}^2
938:     - 4 \, G_s \, G_{\rho s} ,
939: \label{eq:Grhos}
940: \end{eqnarray}
941: %================================================================
942:    where $G_\nu$ are dimensionless coupling constants with
943:    the initial values $G_\nu (0)=g_\nu /(4\pi ta)$.
944: The number of the independent coupling constants is five, 
945:    since the SU(2) spin symmetry guarantees the relations 
946:    $G_\sigma=G_s$,
947:    $G_{c\sigma}=G_{cs}$, and $G_{\rho\sigma}=G_{\rho s}$ to hold
948:    in the scaling procedure.
949: From these scaling equations, one finds that
950:    the $G_\rho$, $G_c$, and $G_s$ terms are marginal
951:    (the scaling dimension=2),\cite{Cardy,Gogolin}
952:    while the $G_{cs}$ and $G_{\rho s}$ terms are irrelevant operators
953:    of the scaling dimension 4.
954: 
955: 
956: \section{Phase Diagram of the Half-Filled Extended Hubbard Model}
957: \label{sec:phase_diagram}
958: 
959: 
960: \subsection{Bond-charge-density-wave state}\label{sec:BCDW}
961: 
962: 
963: In this section,
964:    we show that the BCDW phase exists 
965:    in between the CDW and SDW phases
966:    in the weak-coupling region of the 1D EHM.
967: 
968: First we focus on the weak-coupling limit $U,V\ll t$,
969:    where we can neglect the irrelevant terms of scaling dimension 4
970:    and restrict ourselves to the marginal terms $\propto$
971:    $g_\rho$, $g_\sigma$, $g_c$, and $g_s$.
972: Effects of the irrelevant terms
973:    are discussed later in this section.
974: Within this approximation, the Hamiltonian reduces to two decoupled
975:    sine-Gordon models, and 
976:    we can analyze the properties of the spin and 
977:    charge modes, separately.
978: The one-loop RG equations for these coupling constants
979:    are given by Eqs.\ (\ref{eq:Grho})--(\ref{eq:Gs})
980:    with $G_{cs}=G_{\rho s}=0$: 
981: %================================================================
982: \begin{eqnarray}
983: \frac{d}{dl} G_\rho(l) &=& 2 \, G_{c}^2(l),
984: \label{eq:Grho1}
985: \\ 
986: \frac{d}{dl} G_{c}(l) &=& 2 \, G_\rho(l) \, G_{c}(l),
987: \label{eq:Gc1}
988: \\
989: \frac{d}{dl} G_{s}(l)  &=& - 2 \, G_{s}^2(l).
990: \label{eq:Gs1}
991: \end{eqnarray}
992: %================================================================
993: 
994: 
995: The spin excitations are controlled by the $G_s$ coupling,
996:    which is marginally relevant (marginally irrelevant)
997:    when $G_s<0$ ($G_s>0$).
998: If $g_{s}<0$, then $|G_{s}(l)|$ increases with increasing $l$.
999: In this case the phase field $\phi$ is locked at $\phi=0$ mod $\pi$
1000:    to gain the energy [see Eq.\ (\ref{eq:Hamiltonian})],
1001:    and consequently the spin excitations have a gap.
1002: On the other hand, if $g_{s}>0$, then $|G_{s}(l)|$ decreases to zero
1003:    as $l$ increases, and the $\phi$ field becomes a free field;
1004:    the spin sector has massless excitations.
1005: The approach of $G_s$ to zero is very slow ($\sim1/l$), and the
1006:    $\phi$ field has a strong tendency to be near $\phi=\pi/2$ mod $\pi$.
1007: Although it eventually fails to lock the phase $\phi$,
1008:    the marginally irrelevant coupling still has an impact on
1009:    low-energy properties by giving rise
1010:    to logarithmic corrections to
1011:    correlation functions.\cite{Giamarchi1989} 
1012: 
1013: The charge sector is governed by the two couplings $G_c$ and $G_\rho$, 
1014:    whose RG flow diagram is of the Kosterlitz-Thouless type.
1015: Since $g_\rho=(U+6V)a >0$, $G_c$ is a relevant coupling and always
1016:    flows to strong-coupling regime, unless $g_{c}=0$.
1017: This means that $G_{c}(l)$ has two
1018:    strong-coupling fixed points,
1019:    $G_{c}(l)\to \infty$ and $G_{c}(l)\to -\infty$,
1020:    depending on its initial value $g_{c}>0$ and $g_{c}<0$.
1021: As seen from Eq.\ (\ref{eq:Hamiltonian}), the relevant $g_c$ 
1022:    with positive (negative) sign implies the 
1023:    phase locking of $\theta$ at the position 
1024:    $\theta=0$ $(\pi/2)$ mod $\pi$.
1025: 
1026: From the above standard arguments, 
1027:    the ground state can be identified by simply looking at the
1028:    initial value of the coupling constants
1029:    $g_{c}$ and $g_{s}$.
1030: The ground state is classified into four cases as summarized
1031:    in Table \ref{table:phase-locking}, and
1032:    the positions of locked phases $(\theta,\phi)$ for respective cases 
1033:    are shown in Fig.\ \ref{fig:potmin}. 
1034: 
1035: (i) $g_{s}<0$ and $g_{c}<0$: The phase fields are locked at 
1036:    $(\theta,\phi)
1037:    =\biglb( (\pi/2)+\pi I_1,\pi I_2\bigrb)$,
1038:    where $I_1$ and $I_2$ are integers.
1039: In this case, among the order parameters in Eqs.\ (\ref{eq:orderparam}),
1040:    only the CDW order parameter has a finite expectation value, and
1041:    the ground state is found to be the CDW state.
1042: Both charge and spin excitations are gapped.
1043: 
1044: (ii) $g_s<0$ and $g_c>0$: The phase fields are locked at
1045:    $(\theta,\phi)=(\pi I_1,\pi I_2)$.
1046: The nonvanishing order parameter is then $\mathcal{O}_\mathrm{BCDW}$,
1047:    and the ground state is the BCDW state.
1048: Both charge and spin excitations are gapped.
1049: 
1050: (iii) $g_s>0$ and $g_c<0$: The field $\theta$ is locked at
1051:    $\theta=(\pi/2)+\pi I_1$, and the field $\phi$ tends to be
1052:    around $\phi=(\pi/2)+\pi I_2$ although it is not locked in the
1053:    low-energy limit.
1054: In this case the dominant correlation is that of the BSDW state.
1055: The charge excitations are gapped whereas the spin excitations
1056:    are gapless.
1057: 
1058: (iv) $g_s>0$ and $g_c>0$: The field $\theta$ is locked at
1059:    $\theta=\pi I_1$, whereas the field $\phi$ tends to be near
1060:    $\phi=(\pi/2)+\pi I_2$.
1061: The dominant correlation is the SDW order.
1062: The charge excitations are gapped while the spin excitations are
1063:    gapless.
1064: 
1065: \begin{table}
1066: \caption{
1067: Possible ground-state phases and positions of (quasi)
1068:    locked phase fields
1069:    determined only from the marginal terms 
1070:    in Eq.\ (\protect{\ref{eq:Hamiltonian}}).
1071: }
1072: \label{table:phase-locking}
1073: \begin{ruledtabular}
1074: \begin{tabular}{lcc}
1075: Phase &  $(\theta,\phi)$    &  $(g_c,g_s)$   \\
1076: \hline
1077: SDW  & $(0,\pm \pi/2),
1078:         (\pi,\pm \pi/2)$
1079:      & $(+,+)$ \\
1080: CDW  & $(\pm \pi/2,0), 
1081:         (\pm \pi/2,\pi)$
1082:      & $(-,-)$  \\
1083: BCDW & $(0,0),(\pi,\pi),
1084:         (0,\pi),(\pi,0)$              
1085:      & $(+,-)$    \\
1086: BSDW & $(\pm \pi/2, 
1087:         \pm \pi/2)$ 
1088:      & $(-,+)$
1089: \end{tabular}
1090: \end{ruledtabular}
1091: \end{table}
1092: %================================================================
1093: \begin{figure}[b]
1094: \includegraphics[width=7.cm]{fig03}
1095: \caption{
1096: Positions of locked phase fields $\theta$ and $\phi$
1097:    in the SDW, CDW, BCDW, and BSDW states. 
1098: }
1099: \label{fig:potmin}
1100: \end{figure}%
1101: %================================================================
1102: 
1103: Combining the results of Table \ref{table:phase-locking} and
1104:   the coupling constants Eqs.\ (\ref{eq:g1perp}) and (\ref{eq:g3perp}),
1105:   we obtain the ground-state phase diagram
1106:   of the 1D EHM in the weak-coupling limit.
1107: For $U$ larger than $2V$ such that $g_c>0$ and $g_s>0$,
1108:    we have the SDW phase, while
1109:    for $U$ sufficiently smaller than $2V$ ($g_c<0$ and $g_s<0$)
1110:    we have the CDW phase.
1111: At $U=2V$, we see from Eqs.\ (\ref{eq:g1perp}) and (\ref{eq:g3perp})
1112:    that $g_s(=g_{1\perp})<0$ and $g_c(=g_{3\perp})>0$  due to
1113:    the $C_2$ term.
1114: This implies that a new phase 
1115:    different from the CDW and SDW states
1116:    appears for $U\approx 2V$.
1117: From Table \ref{table:phase-locking}, 
1118:    we identify the new phase with the BCDW phase.
1119: Within the approximation we employ here, the phase boundary between
1120:    the BCDW phase and the CDW (SDW) phase is located at $g_c=0$
1121:    ($g_s=0$).
1122: In this phase diagram,
1123:    the charge excitations are gapful except on the CDW-BCDW transition
1124:    line,
1125:    while the spin excitations are gapless in the SDW phase and on the
1126:    SDW-BCDW transition line.
1127: From Eqs.\ (\ref{eq:Grho1})--(\ref{eq:Gs1}), we can estimate
1128:   the charge gap $\Delta_c$ and the spin gap
1129:    $\Delta_s$ as 
1130: \begin{equation}
1131:   \Delta_c\approx t \left(\frac{|g_c|}{ta}\right)^{2\pi t a/g_\rho},
1132:  \quad
1133:   \Delta_s\approx  t\exp\left( \frac{2\pi t a}{g_s}\right)
1134: \end{equation}
1135: for $|g_c|\ll g_\rho \ll  ta$ and
1136:    $0<-g_s\ll ta$, respectively.
1137: 
1138: 
1139: Next we examine effects of the parallel-spin umklapp scattering
1140:    $g_{cs}$ on the BCDW state.
1141: We consider the situation very close to the CDW-BCDW
1142:    transition by assuming
1143:    $g_c\approx0$ and $g_s<0$, i.e.,
1144:    $U-2V = -C_2V^2/\pi t + O(V^3/t^2)$.
1145: In this case the spin gap is formed first as the energy scale is
1146:    lowered.
1147: For energies below the spin gap,
1148:    we can replace $\cos2\phi$ with its average
1149:    $\langle\cos2\phi\rangle\approx (\Delta_s/t)^2$.
1150: This means that the coupling constant $g_c$ 
1151:    is modified as
1152: %================================================================
1153: \begin{equation}
1154: g_{c}^* = g_c + g_{cs} \langle\cos2\phi\rangle.
1155: \label{eq:gc*}
1156: \end{equation}
1157: %================================================================
1158: Thus we find that the BCDW state,
1159:    which is realized for $g_c^*>0$, becomes
1160:    less favorable due to the 
1161:    $g_{cs}(<0)$ term.
1162: We note, however, that the CDW-BCDW boundary does not move across
1163:    the $U=2V$ line because
1164:    $|g_{cs}\langle\cos2\phi\rangle|\approx 2Va\exp[-c(t/V)^2]$
1165:    is much smaller than the $C_2$ term 
1166:    in Eq.\ (\ref{eq:g3perp}) for $V\ll t$,
1167:    where $c$ is a positive constant.
1168: A similar argument applies to the region near the SDW-BCDW
1169:    transition.
1170: Suppose that $U-2V=+C_2V^2/\pi t + O(V^3/t^2)$
1171:    where $g_s\approx0$ and $g_c>0$.
1172: In this case, as the energy scale is lowered, the charge gap
1173:    opens first and the $\theta$ field is pinned at $\theta=0$
1174:    mod $\pi$.
1175: Below the charge-gap energy scale, the $\phi$ field is subject to
1176:    the pinning potential $g_s^*\cos2\phi$ with
1177: %===============================================================
1178: \begin{equation}
1179: g_{s}^* = g_s - g_{cs} \langle\cos2\theta\rangle,
1180: \label{eq:gs*}
1181: \end{equation}
1182: %================================================================
1183:   where $\langle\cos2\theta\rangle\approx (\Delta_c/t)^{2(1-G_\rho)}$.
1184: Thus the BCDW phase, which is now realized 
1185:    for $g_s^*<0$, also becomes less favorable by the 
1186:    $-g_{cs}\langle\cos2\theta\rangle(>0)$ term.
1187: Again the phase boundary is not moved beyond the $U=2V$ line since
1188:    $|g_{cs}\langle\cos2\theta\rangle|\approx 2Va(c'V/t)^{\pi t/V}$
1189:    is much smaller than the $C_2$ term
1190:    in Eq.\ (\ref{eq:g1perp}), where $c'$ is a constant of order
1191:    1. 
1192: Therefore we conclude that 
1193:     the BCDW phase is robust against the $g_{cs}$ term
1194:     in the weak-coupling limit.
1195: The analysis in this section establishes the existence of the
1196:    BCDW phase near $U\approx2V$ for $0<U,V\ll t$.
1197: 
1198: 
1199: \subsection{First-order SDW-CDW transition}\label{sec:SDW-CDW_trans}
1200: 
1201: In this section, we discuss how the BCDW phase becomes unstable
1202:    at strong coupling and how the 
1203:    two continuous transitions change into the 
1204:    first-order SDW-CDW transition.
1205: 
1206: To our knowledge,
1207: Cannon and Fradkin were the first to argue that 
1208:    the $g_{3\parallel}$ term (describing the umklapp scattering of
1209:    parallel-spin electrons),
1210:    which is conventionally ignored due to its
1211:    large scaling dimension,
1212:    can become relevant at large $U$ and $V$
1213:    and cause the first-order CDW-SDW transition. \cite{Cannon}
1214: To get an insight into the effect of the 
1215:    $g_{cs}$ term in the relevant case, 
1216:    we perform a semiclassical analysis:
1217:    we neglect spatial variations
1218:    of the fields in Eq.\ (\ref{eq:Hamiltonian})
1219:    and focus on the potential,
1220: %================================================================
1221: \begin{equation}
1222:  V(\theta,\phi)=
1223:     - g_c\cos2\theta + g_s\cos 2\phi
1224:     - g_{cs} \cos2\theta \, \cos 2\phi,
1225: \label{eq:V}
1226: \end{equation}
1227: %================================================================
1228:    where $g_{cs}=g_{3\parallel}<0$.
1229: The order parameters of the SDW, CDW, BCDW, and BSDW states
1230:    take maximum amplitudes when the fields
1231:    $\theta$ and $\phi$ are pinned at 
1232:    $(\theta,\phi)
1233:    =\biglb(\pi I_1,(\pi/2)+\pi I_2\bigrb)$,
1234:    $\biglb( (\pi/2)+\pi I_1,\pi I_2\bigrb)$,
1235:    $(\pi I_1,\pi I_2)$, and 
1236:    $\biglb( (\pi/2)+\pi I_1,(\pi/2) + \pi I_2\bigrb)$,
1237:    respectively, where $I_1$ and $I_2$ are integers.
1238: The potential energy in these states is obtained
1239:    by inserting these pinned fields
1240:    into Eq.\ (\ref{eq:V}), e.g,
1241:    $V_{\mathrm{SDW}}=V\biglb(\pi I_1,(\pi/2)+\pi I_2\bigrb)$, yielding
1242: %================================================================
1243: \begin{subequations}
1244: \begin{eqnarray}
1245: V_\mathrm{SDW}  &=& - g_c-g_s-|g_{cs}|  , \\
1246: V_\mathrm{CDW}  &=& + g_c+g_s-|g_{cs}|  , \\
1247: V_\mathrm{BCDW} &=& - g_c+g_s+|g_{cs}|  , \\
1248: V_\mathrm{BSDW} &=& + g_c-g_s+|g_{cs}|  .
1249: \end{eqnarray}
1250: \end{subequations}
1251: %================================================================
1252: %================================================================
1253: \begin{figure}[t]
1254: \includegraphics[width=7cm]{fig04}
1255: \caption{
1256: Phase diagram obtained by minimizing the potential $V(\theta,\phi)$
1257:    for $g_{cs}<0$.
1258: The double line denotes the first-order transition, while the single
1259:    line denotes the second-order transition.
1260: Bicritical points are at
1261:    $(g_c,g_s)=(\pm |g_{cs}|,\mp |g_{cs}|)$.
1262: }
1263: \label{fig:classic}
1264: \end{figure}%
1265: %================================================================
1266: We find that
1267:    the $g_{cs}$ term stabilizes
1268:    the SDW and CDW states
1269:    while it works against the BCDW and BSDW states.
1270: Comparing these energies, we obtain the phase diagram in the
1271:    $g_c$-$g_s$ plane at a fixed $g_{cs}$ (Fig.\ \ref{fig:classic}).
1272: In the presence of the $g_{cs}$ term, the direct
1273:    CDW-SDW transition line appears 
1274:    in this phase diagram.
1275: 
1276: %================================================================
1277: \begin{figure}[t]
1278: \includegraphics[width=7cm]{fig05a}
1279: \includegraphics[width=7cm]{fig05b}
1280: \includegraphics[width=7cm]{fig05c}
1281: \caption{
1282: The potential $V(\theta,\phi)$ on 
1283:    the SDW-BCDW (a), BCDW-CDW (b), and CDW-SDW (c) transition lines.
1284: }
1285: \label{fig:v}
1286: \end{figure}
1287: %================================================================
1288: 
1289: We now discuss the nature of the phase transitions.
1290: The potential $V(\theta,\phi)$ on various
1291:    transition lines is shown in Fig.\ \ref{fig:v}.
1292: On the boundary between the SDW and BCDW phases, which
1293:    is located at $g_s=-|g_{cs}|$ and $g_c>|g_{cs}|$, the potential 
1294:    takes the form $V(\theta,\phi)=
1295:     -g_c \cos 2\theta + g_s \cos 2\phi (1-\cos 2\theta)$
1296:    [Fig.\ \ref{fig:v}(a)],
1297:    which pins the $\theta$ field at $\theta=\pi I_1$ and leaves
1298:    the $\phi$ field completely free.
1299: We thus find that 
1300:    the SDW-BCDW transition is continuous, i.e.,
1301:    the SDW and BCDW phases coexist
1302:    without potential barrier on the phase boundary.
1303: On the boundary between the BCDW and CDW phases,
1304:    located at $g_c=|g_{cs}|$ and $g_s<-|g_{cs}|$,
1305:    the potential now
1306:    takes the form $V(\theta,\phi)=
1307:     -g_c \cos 2\theta (1-\cos 2\phi)+ g_s \cos 2\phi $
1308:    [Fig.\ \ref{fig:v}(b)].
1309: The potential locks the $\phi$ field at $\phi=\pi I_2$, where 
1310:    it has no effect on the $\theta$ field.
1311: Thus, we find that the CDW-BCDW transition is also continuous. 
1312: From similar considerations, we find that 
1313:    the SDW-BSDW and BSDW-CDW transitions are continuous as well.
1314: In Fig.\ \ref{fig:classic}, the phase boundaries of 
1315:    continuous transitions are shown by the solid lines. 
1316: On the contrary, the phase  
1317:    boundary shown by the double line in Fig.\ \ref{fig:classic}
1318:    is of different nature from the others.
1319: The potential $V(\theta,\phi)$ on the double line is 
1320:    shown in Fig.\ \ref{fig:v}(c), where
1321:    the potential minima  
1322:    are given by the isolated points 
1323:    $(\theta,\phi)
1324:      =\biglb(\pi I_1,(\pi/2) + \pi I_2\bigrb)$
1325:    and $\biglb((\pi/2)+\pi I_1, \pi I_2\bigrb)$.
1326: These minima correspond to the SDW state
1327:    and the CDW state,
1328:    see Fig.\ \ref{fig:potmin}.
1329: The point to note is that
1330:    there is a finite potential barrier of height 
1331:    $\min(|g_{cs}|,2|g_{cs}| - 2|g_c|)$
1332:    between the corresponding minima for the SDW and CDW phases.
1333: Hence we conclude that 
1334:    the CDW-SDW transition is first order when $g_{cs}$ is relevant.
1335: 
1336: 
1337: From the above arguments, we find that
1338:    strong umklapp scattering 
1339:    of the parallel-spin electrons destabilizes the BCDW and 
1340:    BSDW states and
1341:    gives rise to bicritical points $(g_c,g_s)=\pm (g_{cs},-g_{cs})$
1342:    where the two continuous-transition lines merge
1343:    into the CDW-SDW first-order transition line.
1344: Let us take a closer look at these bicritical points.
1345: Taking into account the fact that $g_c>0$ and $g_s<0$
1346:    for $U\approx 2V$ in the original EHM,
1347:    we will focus on the bicritical point at
1348:    $(g_c,g_s)=(|g_{cs}|,-|g_{cs}|)$.
1349: The effective potential at the bicritical point takes the form
1350: %================================================================
1351: \begin{figure}[t]
1352: \includegraphics[width=7cm]{fig06}
1353: \caption{
1354: The potential $V(\theta,\phi)$ on 
1355:    the bicritical point $(g_c,g_s)=(|g_{cs}|,-|g_{cs}|)$.
1356: The potential minima are
1357:    the lines $\theta=\pi I_1$ and $\phi= \pi I_2$.
1358: }
1359: \label{fig:v4}
1360: \end{figure}%
1361: %================================================================
1362: %================================================================
1363: \begin{equation}
1364:    V(\theta,\phi)=-g(\cos2\theta+\cos2\phi-\cos2\theta\cos2\phi),
1365: \end{equation}
1366: %================================================================
1367:    which is shown in Fig.\ \ref{fig:v4}.
1368: This potential has an interesting feature that its potential minima are
1369:    not isolated points but the crossing lines $\theta=\pi m$ or
1370:    $\phi=\pi n$ ($m$, $n$: integer).
1371: On these lines either $\theta$ or $\phi$ becomes a free field;
1372:    the theory has more freedom than a single free bosonic field, but
1373:    less than two free bosonic fields.
1374: We thus expect that the theory of the bicritical point should have a
1375:    central charge larger than 1 but smaller than 2.
1376: Detailed analysis of the critical theory is left for a future study.
1377: We note that when $g_{cs}=0$ 
1378:    the first-order CDW-SDW transition line collapses into
1379:    a tetracritical point, $(g_c,g_s)=(0,0)$,
1380:    and the
1381:    phase boundaries in Fig.\ \ref{fig:classic}
1382:    reduce to the lines $g_c=0$ and $g_s=0$
1383:    where all the transitions are continuous.
1384: 
1385: 
1386: Fabrizio \textit{et al}.\ \cite{Fabrizio} and Bajnok
1387:    \textit{et al}.\ \cite{Bajnok}
1388:    discussed effects of higher-frequency terms, 
1389:    such as $\sin 3\theta$ and $\cos4\theta$, which are generated 
1390:    through the renormalization-group transformation.
1391: From the semiclassical arguments,
1392:    it can be seen that these terms can also change a second-order
1393:    transition to a first-order transition.\cite{Bajnok}
1394: In fact, it was argued that these higher-frequency terms make the
1395:    SDW-CDW transition first order in the strong-coupling regime of the 
1396:    1D EHM.\cite{Fabrizio}
1397: However,
1398:    we have shown that the SDW-CDW first-order transition can occur
1399:    simply due to the $g_{cs}$ term which is
1400:    the leading irrelevant term in this system.
1401: Since the higher-frequency terms are even less relevant than the
1402:    $g_{cs}$ term, we expect that the $g_{cs}$ term should play a
1403:    dominant role in the first-order transition in the 1D EHM.
1404: 
1405: %================================================================
1406: \begin{figure}[t]
1407: \includegraphics[width=7cm]{fig07}
1408: \caption{
1409: Phase diagram of the half-filled 1D extended Hubbard model.
1410: The double line denotes the first-order transition,
1411:   while the single lines denote the second-order transitions.
1412: The bicritical point is at $(U_c,V_c)\approx (5.0t, 2.3t)$.
1413: }
1414: \label{fig:phase2}
1415: \end{figure}
1416: %================================================================
1417: 
1418: \subsection{Global ground-state phase diagram}
1419: 
1420: To obtain the global phase diagram of the 1D EHM,
1421:    we have numerically solved the scaling equations
1422:    (\ref{eq:Grho})--(\ref{eq:Grhos}).
1423: We find out which phase is realized by looking at which one of the
1424:    couplings $G_c$, $G_s$, and $G_{cs}$ becomes relevant first,
1425:    as we have discussed 
1426:    in Secs.\ \ref{sec:BCDW} and \ref{sec:SDW-CDW_trans}.
1427: First, if $|G_c|$ grows with increasing $l$ and reaches, say, 
1428:    1 first among the three couplings,
1429:    then we stop the integration and compute
1430:    $G_s^*=G_s-G_{cs} \, \mathrm{sgn}(G_c)$.
1431: Since the charge fluctuations are suppressed below this energy scale,
1432:    we are left with Eq.\ (\ref{eq:Gs1}), 
1433:    where $G_s$ is replaced by $G_s^*$.
1434: We immediately see from Table \ref{table:phase-locking}
1435:    that a positive (negative) $G_s^*$ leads to the
1436:    SDW (BCDW) state for $G_c>0$ and the BSDW (CDW) state
1437:    for $G_c<0$. 
1438: Second, if $|G_s|$ becomes 1 first,
1439:    or more precisely, if $G_s$ reaches $-1$ first, 
1440:    then we are left with Eqs.\ (\ref{eq:Grho1}) and 
1441:    (\ref{eq:Gc1}),  where $G_\rho$  and $G_c$ are replaced by
1442:    $G_{\rho}^*=G_\rho - G_{\rho s}$ and 
1443:    $G_c^*=G_c+G_{cs}$, respectively.
1444: We see that a positive (negative) $G_c^*$    
1445:    leads to the BCDW (CDW) state.
1446: Finally, when $|G_{cs}|$ reaches 1 first, we stop the calculation and
1447:    compare $G_c$ and $G_s$.
1448: Since both charge and spin fluctuations are already suppressed by the
1449:    $G_{cs}\cos2\theta\cos2\phi$ potential, we can deduce the phase
1450:    from the semiclassical argument.
1451: From Fig.\ \ref{fig:classic} we see that we have the SDW state for
1452:    $G_s>-G_c$ and the CDW state for $G_s<-G_c$.
1453: Here we note that
1454:   in the SDW state the pinning potential to the $\phi$ field is
1455:    marginally irrelevant and 
1456:    thus the spin sector should become gapless.
1457: 
1458: 
1459: The phase diagram obtained in this manner is shown in 
1460:    Fig.\ \ref{fig:phase2}.
1461: The single lines denote continuous transitions, and
1462:    the double line denotes the first-order transition.
1463: In the weak-coupling limit, the BCDW phase appears at $U\approx 2V$ and
1464:    the successive continuous transitions between the SDW, BCDW, and
1465:    CDW states occur as $V/U$ increases.
1466: When $U$ and $V$ increase along the line $U\approx 2V$,
1467:    the BCDW phase first
1468:    expands and then shrinks up to the bicritical point
1469:    $(U_c,V_c)\approx(5.0t, 2.3t)$ where the two continuous-transition
1470:    lines meet.
1471: Beyond this point the BCDW phase disappears and we have the direct
1472:    first-order transition between the CDW and  SDW phases.
1473: The phase diagram (Fig.\ \ref{fig:phase2}) is similar to the ones
1474:    obtained by using more sophisticated numerical
1475:    methods.\cite{Nakamura,Sengupta}
1476: We note that the position of the first-order transition line 
1477:    in Fig.\ \ref{fig:phase2} is not reliable quantitatively 
1478:    as we have used the perturbative RG equations.
1479: The recent Monte Carlo calculation\cite{Sengupta} gives the most
1480:    reliable estimate for the position of the bicritical point,
1481:    $(U_c,V_c)\approx\biglb((4.7\pm0.1)t, (2.51\pm0.04)t\bigrb)$, which
1482:    agrees with our estimate in Fig.\ \ref{fig:phase2} within 10\%.
1483: The semiquantitative agreement gives us confidence that our approach,
1484:    semiclassical analysis of the low-energy effective Hamiltonian
1485:    derived with use of the perturbative RG, is reliable even in the
1486:    strong-coupling regime near the multicritical point.
1487: 
1488: 
1489: \section{Effect of staggered site potential}\label{sec:staggered}
1490: 
1491: In this section, we examine effects of
1492:    alternating on-site modulation of the chemical potential, i.e., 
1493:    the staggered site potential, in the half-filled 1D EHM.
1494: The Hamiltonian to be considered is given by $H'=H+H_\Delta$ with 
1495:    $H$ defined in Eq.\ (\ref{eq:H1D}) and
1496: %================================================================
1497: \begin{equation}
1498: H_\Delta =  \Delta \sum_{j,\sigma} (-1)^j \,  n_{j,\sigma}.
1499: \label{eq:H_Delta}
1500: \end{equation}
1501: %================================================================
1502: The model is called the ionic Hubbard model if $V=0$.
1503: When $U=V=0$, the system is a trivial band insulator, since 
1504:    the $\Delta$ term induces a gap $2|\Delta|$ at $k= \pm \pi/2$ in the 
1505:    single-particle spectrum and the lower band is fully filled.
1506: For many years effects of the \textit{on-site} repulsive interaction 
1507:    $U$ on the band insulator have been investigated intensively
1508:    \cite{MJRice,Nagaosa,Girlando,Egami,Fabrizio,%
1509:         Tsuchiizu_JPSJ,Takada,Qin,Brune,YZZhang,Anusooya-Pati,Wilkens,%
1510:         Torio,Refolio,Caprara,Gupta,Pozgajcic,%
1511:         Resta,Gidopoulos,Manmana}
1512:    from both numerical and analytical approaches.
1513: Using the standard bosonization method, 
1514:    Fabrizio, Gogolin, and Nersesyan recently argued that
1515:    the ground state of the ionic Hubbard model exhibits three phases
1516:    as $U$ increases:
1517:    the band insulator, the SDI,
1518:    and the Mott insulator.\cite{Fabrizio}
1519: The order parameter of the SDI state is 
1520:    nothing but that of the BCDW state, and we can regard the two
1521:    states as essentially identical.
1522: It was also argued that the quantum phase
1523:    transition from the band insulator to the SDI state belongs
1524:    to the Ising universality class whereas the other transition
1525:    from the SDI state to the Mott insulator is of the
1526:    Kosterlitz-Thouless type.
1527: Recent numerical studies,
1528:    \cite{Takada,Qin,Brune,YZZhang,Anusooya-Pati,Wilkens,Torio,Refolio,Manmana}
1529:    however, have reported
1530:    controversial results on the existence of the SDI phase.
1531: Some claimed to find two quantum phase transitions while others found
1532:    evidences of only one phase transition.
1533: With this issue of the SDI phase in mind,
1534:     in this section we investigate the phase diagram of
1535:     the 1D extended Hubbard model with the staggered site potential
1536:     and examine critical properties of the quantum phase transitions.
1537: 
1538: 
1539: We take into account the staggered site potential and the correlation
1540:    effects on equal footing by treating them as weak perturbations.
1541: We use Eq.\ (\ref{eq:field_op}) to rewrite $H_\Delta$ in the continuum
1542:    limit as $H_\Delta=\int dx \, \mathcal{H}_\Delta$, where
1543:    \cite{Fabrizio,Tsuchiizu_JPSJ}
1544: %================================================================
1545: \begin{equation}
1546: \mathcal{H}_\Delta =
1547: -\frac{g_\Delta}{2(\pi a)^2} \, \sin \theta \, \cos \phi
1548: \label{eq:Hw_bosonization}
1549: \end{equation}
1550: %================================================================
1551:    with $g_\Delta=4\pi \Delta a$.
1552: Note that the CDW order parameter $\mathcal{O}_\mathrm{CDW}$ is
1553:   proportional to $\mathcal{H}_\Delta$, and $g_\Delta$ can be regarded as
1554:   an external force coupled to $\mathcal{O}_\mathrm{CDW}$.
1555: This has the consequence that $\mathcal{O}_\mathrm{CDW}$ acquires
1556:   a nonvanishing expectation value for any finite $U$ and $V$,
1557:   as long as $g_\Delta\neq0$.
1558: In this section we will denote the insulating phase connected to the
1559:   free-electron band insulator ($U=V=0$ and $\Delta\ne0$) by
1560:   the BI phase, rather than the CDW phase.
1561: 
1562: The bosonized form of the Hamiltonian $H'$ can be thought of as
1563:    a generalization of the so-called double sine-Gordon (DSG) model
1564:    as $H'$ contains sine/cosine terms
1565:    with different frequencies ($\sin \theta$ and $\cos 2\theta$,
1566:    $\cos \phi$ and $\cos 2\phi$). 
1567: The DSG theory itself has been investigated intensively
1568:    \cite{Fabrizio,Bajnok,Delfino} and shown to have a critical point
1569:    belonging to the Ising universality class 
1570:    [$c=\frac{1}{2}$ conformal field theory (CFT)].
1571: To obtain a qualitative understanding of
1572:    the critical properties in our system,
1573:    we first perform a semiclassical analysis in a similar way 
1574:    to Sec.\ III B, before examining the global phase diagram of $H'$
1575:    with use of the RG method.
1576: 
1577: 
1578: \subsection{Semiclassical analysis}
1579: 
1580: 
1581: In this section, we perform a semiclassical analysis
1582:    to the Hamiltonian $\mathcal{H}'=\mathcal{H}+\mathcal{H}_\Delta$,
1583:    where $\mathcal{H}$ and $\mathcal{H}_\Delta$ are given by 
1584:    Eqs.\ (\ref{eq:Hamiltonian}) and (\ref{eq:Hw_bosonization}),
1585:    respectively.
1586: We neglect spatial variations of the field
1587:    and focus on the locking potential:
1588: %================================================================
1589: \begin{eqnarray}
1590: V_\Delta(\theta,\phi)
1591: &=&
1592: -g_c \cos2\theta + g_s \cos 2\phi
1593: -g_{cs} \cos 2\theta \cos 2\phi
1594: \nonumber \\ 
1595: && {}
1596: -g_\Delta \, \sin\theta \, \cos \phi
1597: .
1598: \label{eq:VW}
1599: \end{eqnarray}
1600: %================================================================
1601: 
1602: First, we examine the case $g_{cs}=0$, which corresponds to 
1603:    the situation where the $g_{cs}$ term becomes irrelevant in the
1604:    RG scheme.
1605: The potential to be considered is
1606: %================================================================
1607: \begin{eqnarray}
1608: V^0_\Delta(\theta,\phi)& \equiv &
1609: V_\Delta(\theta,\phi)|_{g_{cs}=0} \nonumber\\
1610: &=&
1611: -g_c \, \cos2\theta + g_s \, \cos 2\phi
1612: -g_\Delta \, \sin\theta \, \cos \phi
1613: .
1614: \nonumber\\&&
1615: \label{eq:VW0}
1616: \end{eqnarray}
1617: %================================================================
1618: Due to its double-frequency structure, possible locations
1619:    of the phase locking are different from the ones we found
1620:    in Sec.\ III B.
1621: For example, when $g_c>0$ ($g_s>0$), 
1622:    the two kinds of potentials proportional to
1623:    $\sin \theta$ and $\cos 2\theta$ ($\cos \phi$ and $\cos 2\phi$)
1624:    compete with each other. \cite{Delfino,Bajnok}
1625: The locking of the phases $\theta$ and $\phi$
1626:    are determined from the saddle-point equations: 
1627:    $\cos\theta(4 g_c \sin \theta-g_\Delta \cos \phi)=0$ and
1628:    $\sin\phi (- 4 g_s \cos \phi + g_\Delta \sin \theta)=0$.
1629: In order to simplify the notations, let us introduce
1630: %================================================================
1631: \begin{equation}
1632: \alpha_\theta^0
1633: \equiv \left| \cos^{-1} \left(\frac{g_\Delta}{4g_c}\right) \right|
1634: , \quad
1635: \alpha_\phi^0
1636: \equiv \left| \cos^{-1} \left(\frac{g_\Delta}{4g_s}\right) \right|
1637: ,
1638: \end{equation}
1639: %================================================================
1640:    where $|g_\Delta/g_c|\le 4$, $|g_\Delta/g_s|\le 4$, and
1641:    $0 \le \alpha_\theta^0,\alpha_\phi^0 \le \pi$ are assumed.
1642: %================================================================
1643: \begin{table}
1644: \caption{Possible ordered ground states  and
1645:   the position of (quasi-)locked phase fields
1646:   determined from  Eq.\ (\protect{\ref{eq:VW0}}).
1647: }
1648: \label{table:phase-locking-W}
1649: \begin{ruledtabular}
1650: \begin{tabular}{lc}
1651: Phase  &  $(\theta,\phi )$       \\
1652: \hline
1653: SDW    &  $(0,\pm \pi/2 ), (\pi,\pm \pi/2 )$  \\
1654: BI (for $g_\Delta>0$)   &  
1655:     $(+\pi/2,0 ), (-\pi/2,\pi )$  \\
1656: BI (for $g_\Delta<0$)   &
1657:     $(+\pi/2,\pi ),(-\pi/2,0 )$  \\
1658: BCDW & 
1659:     $\biglb(+(\pi/2)\pm \alpha_\theta^0,0 \bigrb), 
1660:      \biglb(-(\pi/2)\pm \alpha_\theta^0,\pi \bigrb)$  \\
1661: BSDW & 
1662:    $(+\pi/2, 0\pm \alpha_\phi^0 ),
1663:       \biglb(-\pi/2, \pm (\pi-\alpha_\phi^0) \bigrb)$
1664: \end{tabular}
1665: \end{ruledtabular}
1666: \end{table}
1667: %================================================================
1668: %================================================================
1669: \begin{figure}[b]
1670: \includegraphics[width=7.cm]{fig08}
1671: \caption{
1672: Positions of locked phase fields $\theta$ and $\phi$
1673:    in the four states when $g_\Delta>0$. 
1674: }
1675: \label{fig:potmin-W}
1676: \end{figure}%
1677: %================================================================
1678: The solutions of the saddle-point equations yield
1679:    the following four states with distinct
1680:    configurations of the locked phase fields $\theta$ and $\phi$
1681:    (modulo $2\pi$):
1682: (i) the SDW state with $\theta$ and $\phi$ locked at
1683:    $(\theta,\phi)=(0,\pm \pi/2)$ or 
1684:    $(\pi, \pm \pi/2)$;
1685: (ii) the BI state with
1686:    $(\theta,\phi) =(+\pi/2,0)$, $(-\pi/2,\pi)$
1687:    if $g_\Delta>0$ and with
1688:    $(\theta,\phi)=(+\pi/2,\pi)$, $(-\pi/2,0)$
1689:    if $g_\Delta<0$;
1690: (iii) the ``BCDW'' state where the BCDW order and the CDW order
1691:    coexist and which is realized when
1692:    $(\theta,\phi)
1693:    =( \pi/2\pm \alpha_\theta^0,0)$ or
1694:    $(-\pi/2\pm \alpha_\theta^0,\pi)$;
1695: (iv) the ``BSDW'' state where the BSDW and the CDW order coexist
1696:    and which is realized when
1697:    $(\theta,\phi)
1698:     =(\pi/2,0\pm \alpha_\phi^0)$ or
1699:     $\biglb(-\pi/2, \pm( \pi - \alpha_\phi^0) \bigrb)$.
1700: Table \ref{table:phase-locking-W} and Fig.\ \ref{fig:potmin-W}
1701:   summarize the possible ordered ground states and corresponding
1702:   positions of locked phase fields.
1703: The potential energies in these states are given by
1704: %================================================================
1705: \begin{subequations}
1706: \begin{eqnarray}
1707: V^0_{\rm SDW}
1708: &=& -g_c -g_s 
1709: ,
1710: \\
1711: V^0_{\rm BI}
1712: &=&  + g_c + g_s - \left| g_\Delta \right|
1713: ,
1714: \\
1715: V^0_{\rm BCDW}
1716: &=&
1717: -g_c +g_s -\frac{g_\Delta^2}{8g_c} 
1718: ,
1719: \label{eq:V0_BCDW+CDW}
1720: \\
1721: V^0_{\rm BSDW}
1722: &=&
1723:  + g_c -g_s 
1724:  - \frac{g_\Delta^2}{8g_s}
1725: .
1726: \label{eq:V0_BSDW+CDW}
1727: \end{eqnarray}
1728: \end{subequations}
1729: %================================================================
1730: In deriving Eqs.\ (\ref{eq:V0_BCDW+CDW}) and (\ref{eq:V0_BSDW+CDW}), 
1731:    we have assumed $|g_\Delta/g_c| \le 4$ and $|g_\Delta/g_s| \le 4$,
1732:    respectively.
1733: The CDW state is stabilized strongly by the $g_\Delta$ term
1734:    whereas
1735:    the BCDW state and the BSDW state are also stabilized by
1736:    the second-order contribution of $g_\Delta$.
1737: By comparing these energies, we arrive at the phase diagram
1738:    shown in Fig.\  \ref{fig:classical-W}.
1739: %================================================================
1740: \begin{figure}[t]
1741: \includegraphics[width=7cm]{fig09}
1742: \caption{
1743: Phase diagram obtained by minimizing the potential energy 
1744:    $V_\Delta^0(\theta,\phi)$ [Eq.\ (\ref{eq:VW0})].
1745: The phase boundaries between the SDW state and 
1746:    the BCDW state, and 
1747:    between the SDW state and the BSDW state 
1748:    are given by the curve $g_s=g_\Delta^2/(16 g_c)$ with $g_c>0$.
1749: The phase boundaries between the BI state and 
1750:    the BCDW state, and 
1751:    between the BI state and the BSDW state 
1752:    are given by the lines $g_c=\frac{1}{4}\,|g_\Delta|$
1753:    with $g_s<\frac{1}{4}\,|g_\Delta|$ and $g_s=\frac{1}{4}\,|g_\Delta|$
1754:    with $g_c<\frac{1}{4}\,|g_\Delta|$, respectively. 
1755: All the phase transitions in this figure are continuous.
1756: The tetracritical point is located at
1757:    $(g_c,g_s)=
1758:     (\frac{1}{4}\,|g_\Delta|,\frac{1}{4}\,|g_\Delta|)$.
1759: }
1760: \label{fig:classical-W}
1761: \end{figure}
1762: %================================================================
1763: As we go across the boundary ($g_c=\frac{1}{4}\, g_\Delta$) 
1764:    from the BI state to the BCDW state,
1765:    we find that each potential minimum splits into 
1766:    two minima, e.g., $( \theta, \phi )
1767:     =( \pi/2,0 ) \to 
1768:      \biglb( (\pi/2)\pm \alpha_\theta^0, 0 \bigrb)$,
1769:    and that
1770:    the potential for the $\theta$ phase field takes a
1771:    double-well structure in the BCDW state.
1772: Similarly, as we go from the BI state to the BSDW state,
1773:    each potential minimum splits into 
1774:    two minima, e.g., $( \theta, \phi )
1775:     =(\pi/2,0) \to ( \pi/2, \pm \alpha_\phi^0)$,
1776:    and now the potential for the $\phi$ phase field has a
1777:    double-well structure in the BSDW state.
1778: As long as $g_{cs}=0$, any quantum phase transition is continuous
1779:    since a potential barrier between two potential minima
1780:    corresponding to two different states vanishes at the transition.
1781: The phase diagram (Fig.\ \ref{fig:classical-W}) indicates that
1782:    a direct transition 
1783:    from the SDW state to the BI state takes place only when 
1784:    the parameters $g_c$ and $g_s$ are on the multicritical point
1785:    $(g_c,g_s)
1786:      =(\frac{1}{4}\, |g_\Delta|,\frac{1}{4}\,|g_\Delta|)$,
1787:    where the potential takes the form
1788:    $V_\Delta^0(\theta,\phi)=
1789:     \frac{1}{2}\, |g_\Delta| \{ -1 
1790:    + \left[\sin\theta - \mathrm{sgn}(g_\Delta) \cos\phi \right]^2 \}$ 
1791:    and is minimized
1792:    at $\phi=\pm [(\pi/2) -\theta]$ 
1793:    and $\phi=\pm (\frac{3}{2}\pi +\theta)$ if $g_\Delta>0$, or
1794:    at $\phi=\pm [(\pi/2) +\theta]$ 
1795:    and $\phi=\pm (\frac{3}{2}\pi -\theta)$ if $g_\Delta<0$. 
1796: 
1797: Let us take a closer look at low-energy excitations
1798:    in the BI state and the BCDW state.
1799: The massive sine-Gordon model has
1800:    topological excitations, solitons, and antisolitons.
1801: They are characterized by the
1802:    topological charges $Q$ and $S_z$
1803:     for the charge and the spin sectors,
1804: %================================================================
1805: \begin{equation}
1806: Q = \frac{1}{\pi} \int dx \, \partial_x \theta 
1807: , \quad
1808: S_z = \frac{1}{2\pi} \int dx \, \partial_x \phi
1809: .
1810: \label{eq:topological_charge}
1811: \end{equation}
1812: %================================================================
1813: In the noninteracting case ($U=V=0$) with a finite $\Delta$,
1814:    the lowest-energy excitation is a soliton 
1815:    of $\theta$ and $\phi$ connecting two neighboring minima of
1816:    the $-g_\Delta\sin\theta \, \cos \phi$, e.g., 
1817:    $(\theta,\phi)|_{x\to-\infty}=(-\pi/2,\pi)$
1818:    and $(\theta,\phi)|_{x\to\infty}=(\pi/2,0)$.
1819: Such an excitation carries the charge $Q=\pm 1$ and 
1820:    the spin $S_z=\pm \frac{1}{2}$, 
1821:    which is nothing but a single-electron excitation
1822:    in the band insulator. 
1823: It has been pointed out \cite{MJRice,Nagaosa,Fabrizio} that
1824:    in the SDI phase (i.e., in the BCDW phase),
1825:    the topological charge $Q$ of the lowest-energy excitation
1826:    becomes fractional, $Q=\pm 2\alpha_\theta^0/\pi$, 
1827:    reflecting the local double-well structure of the potential
1828:    near the potential minima, e.g., at
1829:    $(\theta,\phi)=(\pi/2\pm\alpha_\theta^0,0)$.
1830: This is a unique feature of the BCDW phase and is contrasted 
1831:    from the integer charge $Q=\pm 1$ of the lowest-energy
1832:    excitation in the pure BCDW phase where the phase fields are
1833:    locked at $(\theta,\phi)=(0,0)$. 
1834: Accordingly, the phase transition between the BCDW state
1835:    and the BI state belongs to a different universality class from 
1836:    the one between the pure BCDW state and the CDW state
1837:    discussed in Sec.\ \ref{sec:SDW-CDW_trans}.
1838: In the former case, a small potential barrier in a double-well
1839:    potential in the BCDW state vanishes at the critical point
1840:    and the effective theory for the low-energy excitations
1841:    is the ``$\varphi^4$'' theory known to describe
1842:    the Ising phase transition, rather than the 
1843:    Gaussian theory that governs the transition
1844:    between the BCDW and CDW phases.
1845: 
1846: One might expect that a similar semiclassical analysis can be
1847:    applied to the spin field $\phi$.
1848: Within the semiclassical approach the topological charge $S_z$
1849:    in the BSDW phase of Fig.\ \ref{fig:classical-W} takes a
1850:    fractional value, $\pm \alpha_\phi^0/(2\pi)$. 
1851: However, 
1852:    since the Hamiltonian has the global SU(2) spin-rotation symmetry,
1853:    the SDW state and the BSDW state cannot have a true long-range order.
1854: This implies that the phase field $\phi$ cannot be localized
1855:    except in spin-gap phases
1856:    where $\phi$ is locked at $\langle\phi\rangle=0$ mod $\pi$. 
1857: The global SU(2) symmetry thus prohibits the Ising criticality
1858:    in the spin sector.
1859: In fact, the BSDW phase in Fig.\ \ref{fig:classical-W} 
1860:    turns out to be just the BI phase.
1861: 
1862: 
1863: Let us now consider the situation in which $g_{cs}\neq 0$.
1864: In this case, the phase fields $\theta$ and $\phi$ are locked
1865:    in a similar way to the case $g_{cs}=0$, 
1866:    but $\alpha_\theta^0$ and $\alpha_\phi^0$ are 
1867:    modified into $\alpha_\theta^0\to \alpha_\theta$ and
1868:    $\alpha_\phi^0\to \alpha_\phi$, where
1869: %================================================================
1870: \begin{subequations}
1871: \begin{eqnarray}
1872: \alpha_\theta
1873: &\equiv& 
1874: \left| \cos^{-1} \left[\frac{g_\Delta}{4(g_c-|g_{cs}|)}\right] \right|
1875: , \,\,\,
1876: \\
1877: \alpha_\phi
1878: &\equiv& 
1879: \left| \cos^{-1} \left[\frac{g_\Delta}{4(g_s-|g_{cs}|)}\right] \right|.
1880: \end{eqnarray}
1881: \end{subequations}
1882: %================================================================
1883: Here we have assumed
1884:    $|g_\Delta/(g_c-|g_{cs}|)|\le 4$ and
1885:    $|g_\Delta/(g_s-|g_{cs}|)|\le 4$.
1886: The potential energies in the four states become
1887: %================================================================
1888: \begin{subequations}
1889: \begin{eqnarray}
1890: V_{\rm SDW} &=& -g_c -g_s - |g_{cs}|  ,  \\
1891: V_{\rm BI} &=& +g_c + g_s - |g_{cs}|  -|g_\Delta|  , \\
1892: V_{\rm BCDW} 
1893: &=&
1894: -g_c +g_s + |g_{cs}| 
1895: -\frac{g_\Delta^2}{8(g_c-|g_{cs}|)}
1896: , \qquad %\nonumber \\
1897: \label{eq:V_BCDW+CDW}
1898: \\
1899: V_{\rm BSDW}
1900: &=&
1901:  + g_c -g_s + |g_{cs}|  
1902: -\frac{g_\Delta^2}{8(g_s-|g_{cs}|)}
1903: . \quad %\nonumber \\
1904: \label{eq:V_BSDW+CDW}
1905: \end{eqnarray}
1906: \end{subequations}
1907: %================================================================
1908: By comparing these energies we obtain
1909:    the phase diagram (Fig.\ \ref{fig:classical-W2}).
1910: %================================================================
1911: \begin{figure}[t]
1912: \includegraphics[width=7.2cm]{fig10}
1913: \caption{
1914: Phase diagram obtained by minimizing the potential energy 
1915:    $V_\Delta(\theta,\phi)$ [Eq.\ (\ref{eq:VW})]
1916:    for $g_{cs}<0$.
1917: The phase boundaries are given by 
1918:    $g_s=-g_c+\frac{1}{2}\,|g_\Delta|$ 
1919:    between the SDW and the BI states,
1920:    $g_c=|g_{cs}|+\frac{1}{4}\,|g_\Delta|$
1921:    between the BI and the BCDW states,
1922:    $g_s=|g_{cs}|+\frac{1}{4}\,|g_\Delta|$
1923:    between the BI and the BSDW states,
1924:    $g_s=-|g_{cs}|+g_\Delta^2/[16(g_c-|g_{cs}|)]$
1925:    between the SDW and the BCDW states, and 
1926:    $g_c=-|g_{cs}|+g_\Delta^2/[16(g_s-|g_{cs}|)]$
1927:    between the SDW and the BSDW states.
1928: Multicritical points are located at 
1929:    $(g_c,g_s)
1930:    =(+|g_{cs}|+\frac{1}{4}\, |g_\Delta|,
1931:            -|g_{cs}|+\frac{1}{4}\, |g_\Delta|)$ and 
1932:    $(-|g_{cs}|+\frac{1}{4}\, |g_\Delta|,
1933:            +|g_{cs}|+\frac{1}{4}\, |g_\Delta|)$.
1934: The single lines denote second-order transitions, while 
1935:    the double lines denote first-order transitions. 
1936: }
1937: \label{fig:classical-W2}
1938: \end{figure}
1939: %================================================================
1940: In the limit $g_\Delta\to 0$ this phase diagram reduces to 
1941:    Fig.\ \ref{fig:classic}.
1942: One can easily find that the $g_{cs}$ term favors the 
1943:    SDW state and the BI state over the
1944:    BCDW state and the BSDW state. 
1945: The direct SDW-BI transition line acquires a finite length
1946:    in the phase diagram, like in Fig.\ \ref{fig:classic}.
1947: The analysis of critical properties of each quantum phase transition
1948:    is more complicated than that in Sec.\ III
1949:    due to the presence of two kinds of charge-spin coupled terms, 
1950:    the $g_\Delta$ and $g_{cs}$ terms.
1951: Along the phase boundary between the SDW state and the BI state,
1952:    the potential energy is minimized at discrete points,
1953:    $(\theta,\phi)=(-\pi/2, \pi)$, $(0,\pm \pi/2)$,
1954:    $(\pi/2,0)$, $(\pi,\pm \pi/2)$ for $g_\Delta>0$, or
1955:    at $(\theta,\phi)=(-\pi/2, 0)$, $(0,\pm \pi/2)$,
1956:    $(\pi/2,\pi)$, $(\pi,\pm \pi/2)$ for $g_\Delta<0$.
1957: These points correspond either to
1958:    the SDW state or to the BI state 
1959:    (see Table \ref{table:phase-locking-W}).
1960: Since any path connecting these potential minima has to go over
1961:    a potential barrier, the direct SDW-BI transition is first order.
1962: In addition, both the transition between the SDW state and
1963:    the BCDW state and that between the SDW state and the 
1964:    BSDW state become first order when $g_{cs}\ne0$.
1965: On the phase boundary between the SDW state and the BCDW state,
1966:    the potential has isolated minima
1967:    at $(\theta,\phi) =(0,\pm \pi/2)$,
1968:     $(\pi,\pm \pi/2)$,
1969:     $(-\pi/2\pm \alpha_\theta, \pi)$, and 
1970:     $(+\pi/2\pm \alpha_\theta, 0)$.
1971: The pinning of the phase fields at these minima corresponds either to 
1972:    the SDW state or to the BCDW state 
1973:    (see Fig.\ \ref{fig:potmin-W}).
1974: On the multicritical points at
1975:    $(g_c,g_s)
1976:    =(+|g_{cs}|+\frac{1}{4}\, |g_\Delta|,
1977:            -|g_{cs}|+\frac{1}{4}\, |g_\Delta|)$ and 
1978:    $(-|g_{cs}|+\frac{1}{4}\, |g_\Delta|,
1979:            +|g_{cs}|+\frac{1}{4}\, |g_\Delta|)$, 
1980:    the potential takes the form 
1981: %================================================================
1982: \begin{subequations}
1983: \begin{eqnarray}
1984: V_\Delta^{c1}(\theta,\phi)&=&
1985:   -  |g_{cs}| (\cos 2\theta + \cos 2\phi - \cos 2\theta \, \cos 2\phi)
1986: \nonumber \\ && {}
1987:   + \frac{1}{2}\, |g_\Delta| 
1988:     \left\{-1 + 
1989:     \left[\sin\theta - \mathrm{sgn}(g_\Delta)\cos\phi \right]^2 \right\}
1990: ,
1991: \nonumber \\
1992: \label{eq:Vc1}
1993: \\
1994: V_\Delta^{c2}(\theta,\phi)&=&
1995:   +  |g_{cs}| (\cos 2\theta + \cos 2\phi + \cos 2\theta \, \cos 2\phi)
1996: \nonumber \\ && {}
1997:   + \frac{1}{2}\, |g_\Delta| 
1998:     \left\{-1 + 
1999:     \left[\sin\theta - \mathrm{sgn}(g_\Delta)\cos\phi \right]^2 \right\}
2000: ,
2001: \nonumber \\
2002: \label{eq:Vc2}
2003: \end{eqnarray}
2004: \label{eq:Vc}%
2005: \end{subequations}
2006: %================================================================
2007:    respectively.
2008: The potential minima of $V_\Delta^{c1}(\theta,\phi)$ and
2009:    $V_\Delta^{c2}(\theta,\phi)$
2010:    are located at 
2011:    $(\theta,\phi)=(-\pi/2, \pi)$, 
2012:    $(0,\pm \pi/2)$,
2013:    $(\pi/2,0)$, and
2014:    $(\pi,\pm \pi/2)$ for $g_\Delta>0$ and
2015:    at $(\theta,\phi)=(-\pi/2, 0)$, 
2016:    $(0,\pm \pi/2)$,
2017:    $(\pi/2,\pi)$, and
2018:    $(\pi,\pm \pi/2)$ for $g_\Delta<0$.
2019: 
2020: 
2021: Finally, we note that even in the SDW state (the Mott insulator)
2022: the CDW order parameter has a nonvanishing expectation value.
2023: This is because the alternating site potential $\mathcal{H}_\Delta$
2024: has the same form as the CDW order parameter
2025: $\mathcal{O}_\mathrm{CDW}\propto\sin\theta\cos\phi$.
2026: Even though the semiclassical analysis indicates that the phase
2027: fields are pinned, say, at $(\theta,\phi)=(0,\pm\pi/2)$, quantum
2028: fluctuations of the fields around the pinning position lead to
2029: a nonvanishing $\langle\mathcal{O}_\mathrm{CDW}\rangle$.
2030: This can be easily seen in the limit of small $\Delta$, where
2031: \begin{eqnarray}
2032: \langle\mathcal{O}_\mathrm{CDW}\rangle&\propto&
2033: \mathrm{Tr}\left[
2034:   \exp\left[-\int dx(\mathcal{H}+\mathcal{H}_\Delta)\right]
2035:   \sin\theta\cos\phi
2036:            \right]
2037: \nonumber\\
2038: &\propto&
2039: g_\Delta\mathrm{Tr}\left[\exp\left(-\int dx\mathcal{H}\right)
2040: \sin^2\theta\cos^2\phi\right]
2041: \ne0.
2042: \nonumber \\
2043: \end{eqnarray}
2044: 
2045:    
2046:    
2047: 
2048: 
2049: 
2050: \subsection{Renormalization-group analysis}
2051: 
2052: We perform RG analysis to take into account quantum fluctuations
2053:    that are ignored in the semiclassical analysis.
2054: As in Sec.\ III, we obtain the RG
2055:    equations using the OPE method
2056:    (see Appendix \ref{sec:rg}): 
2057: %================================================================
2058: \begin{eqnarray}
2059: \frac{d}{dl} G_\Delta 
2060:   \!\!&=&\!\! {} + G_\Delta
2061:     + \frac{1}{2} \, G_\Delta \, G_\rho 
2062:     - G_\Delta \, G_c
2063: \nonumber \\ && {}
2064:     - \frac{3}{2} \, G_\Delta \, G_s
2065:     - \frac{3}{4} \, G_\Delta \, G_{cs}
2066:     - \frac{3}{8} \, G_\Delta \, G_{\rho s}
2067: ,\quad\quad
2068: \label{eq:RGf-b}
2069: \\
2070: \frac{d}{dl} G_\rho 
2071:   \!\!&=&\!\! {} + \frac{1}{4} \, G_\Delta^2
2072:     + 2 \, G_c^2 + G_{cs}^2 +  G_s \, G_{\rho s}
2073: ,
2074: \\
2075: \frac{d}{dl} G_c 
2076:   \!\!&=&\!\! {}
2077:   - \frac{1}{4} \, G_\Delta^2 
2078:     + 2 \, G_\rho\, G_c
2079:   - G_s \, G_{cs} - G_{cs} \, G_{\rho s}
2080: ,  \quad\quad
2081: \\
2082: \frac{d}{dl} G_s 
2083:   \!\!&=&\!\! {} - \frac{1}{4} \, G_\Delta^2 
2084:     - 2 \, G_s^2 - G_c \, G_{cs} - G_{cs}^2
2085: ,
2086: \label{eq:RGf-gs}
2087: \\
2088: \frac{d}{dl} G_{cs} 
2089:   \!\!&=&\!\! {} - \frac{1}{4}  \,G_\Delta^2 
2090:     - 2 \, G_{cs} + 2 \, G_\rho \, G_{cs} - 4 \, G_s \, G_{cs}
2091: \nonumber \\ && {}
2092:     - 2 \, G_c\, G_{s}
2093:     - 2 \, G_c\, G_{\rho s}
2094:     - 4 \, G_{cs} \, G_{\rho s}
2095: ,
2096: \\
2097: \frac{d}{dl} G_{\rho s} 
2098:   \!\!&=&\!\! {} - \frac{1}{4}  \, G_\Delta^2 
2099:     - 2 \, G_{\rho s} + 2 \, G_\rho \, G_s
2100: \nonumber \\ && {}
2101:     - 4 \, G_c \, G_{cs} - 4 \, G_{cs}^2
2102:     - 4 \, G_s \, G_{\rho s}
2103: .
2104: \label{eq:RGf-e}
2105: \end{eqnarray}
2106: %================================================================
2107: The initial value of $G_\Delta(l)$ is given by 
2108:    $G_\Delta(0)=\Delta/t$, while those of the other coupling
2109:    constants are given by $G_\nu(0) =g_\nu /(4\pi ta)$.
2110: Since the RG equations are invariant under the sign change of
2111:    $G_\Delta$ ($G_\Delta \to -G_\Delta$), we can assume 
2112:    $G_\Delta(0)\ge 0$ without losing generality
2113:    in the following arguments.
2114: 
2115: 
2116: We determine the ground-state phase diagram
2117:    in a similar way as in Sec.\ \ref{sec:phase_diagram}.
2118: That is, we integrate the scaling equations 
2119:    (\ref{eq:RGf-b})--(\ref{eq:RGf-e}) numerically
2120:    and find which one of the couplings
2121:    [$G_\Delta(l)$, $G_c(l)$, $G_s(l)$, and $G_{cs}(l)$]
2122:    becomes most relevant.
2123: By doing so, we have encountered the following four cases.
2124: 
2125: (i) The case where $G_c(l)$ grows fastest and becomes $1$
2126:    at $l=l_{\rho+}$.
2127: Below this energy scale (i.e., $l\ge l_{\rho+}$),
2128:    the charge fluctuations are suppressed and 
2129:    the phase field $\theta$ is locked at $\theta = 0$ or $\pi$.
2130: For the discussion of the ground-state properties
2131:    we may first neglect the $g_\Delta$ term 
2132:    since $\langle \sin \theta \rangle \cos \phi=0$.
2133: The Hamiltonian density $\mathcal{H}'$ then reduces to
2134: %================================================================
2135: \begin{eqnarray}
2136: \mathcal{H}^{\mathrm{eff}}_{\sigma +} &=&
2137: \frac{v_F}{2\pi} \sum_p (\partial_x \phi_p)^2
2138: - \frac{v_F}{\pi} \, G_s^* 
2139:    \left(\partial_x \phi_+\right) \left(\partial_x \phi_-\right) 
2140: \nonumber \\ && {}
2141: +\frac{v_F}{\pi a^2}  \, G_s^* \cos 2\phi
2142: ,
2143: \end{eqnarray}
2144: %================================================================
2145:    where $G_s^* = G_s(l_{\rho+}) - G_{cs}(l_{\rho+})$.
2146: We immediately see that, if $G_s^*>0$, the spin excitations are  
2147:    gapless and the ground state is the SDW state.
2148: On the other hand, if $G_s^*<0$, then 
2149:    the operators proportional to $G_s^*$ are relevant
2150:    [$G_s^*(l)\to -\infty$ under scaling]
2151:    and the phase fields are locked as 
2152:    $(\theta,\phi )=(0,0),
2153:    (0,\pi),(\pi,0),(\pi,\pi)$,   
2154:    which corresponds to the BCDW state with
2155:    $\alpha_\theta \to \pi/2$
2156:    (i.e., $g_\Delta \to 0$),
2157:    see Table \ref{table:phase-locking-W}.
2158: This would become the BCDW state with $\alpha_\theta<\pi/2$
2159:    in a more realistic treatment where the $g_\Delta$ term is not
2160:    simply ignored.
2161: 
2162: (ii) The case where $|G_c(l)|$ grows most rapidly and $G_c(l)\to -1$
2163:    at $l=l_{\rho-}$.
2164: The phase field $\theta$ is then locked at $\theta=\pm \pi/2$
2165:    for $l>l_{\rho-}$.
2166: Below this energy scale one can replace the $\sin\theta$ potential 
2167:   by its averaged value, i.e., $\sin \theta \to \langle\sin \theta
2168:   \rangle = \pm 1$.
2169: The effective Hamiltonian at $l=l_{\rho-}$ is given by
2170: %================================================================
2171: \begin{eqnarray}
2172: \mathcal{H}^{\mathrm{eff}}_{\sigma -} &=& 
2173: \frac{v_F}{2\pi} \sum_p (\partial_x \phi_p)^2
2174: - \frac{v_F}{\pi} \, G_s^* 
2175:    \left(\partial_x \phi_+\right) \left(\partial_x \phi_-\right) 
2176: \nonumber \\ && {}
2177: \mp \frac{v_F}{\pi a^2} \, G_\Delta^* \, \cos \phi
2178: + \frac{v_F}{\pi a^2} \, G_s^* \, \cos 2\phi
2179: ,
2180: \label{eq:Heff_sigma-}
2181: \end{eqnarray}
2182: %================================================================
2183:    where $G_\Delta^*=G_\Delta(l_{\rho-})$ and
2184:    $G_s^*=G_s(l_{\rho -})+G_{cs}(l_{\rho-})$, 
2185:    and the sign $-/+$ 
2186:    of the $G_\Delta$ term corresponds to the position of
2187:    the phase locking $\theta=+ (\pi/2)/- (\pi/2)$.
2188: When $G_s^*>0$, the two $G_s^*$ terms are 
2189:    marginally irrelevant,
2190:    and the only relevant operator is $\mp \cos\phi$. 
2191: Then the phase field $\phi$ is locked at $\phi=0$ or $\pi$,
2192:    depending on the position of
2193:    the charge phase locking $\theta=+ (\pi/2)$ or 
2194:    $- (\pi/2)$.
2195: On the other hand, when $G_s^*<0$,
2196:    both $G_\Delta^*$ and $G_s^*$ terms become relevant. 
2197: However, these terms do not compete with each other.
2198: The only effect of the $G_\Delta^*$ term is to lift the degeneracy
2199:    between the neighboring minima of $-\cos2\phi$,
2200:    and hence the position of the phase locking is the same as
2201:    in the case $G_s^*>0$. 
2202: Therefore, regardless of the sign of $G_s^*$,
2203:    the resultant phase is found to be the BI state
2204:    with the phase locking at
2205:    $(\theta,\phi)=(\pi/2,0)$ or $(-\pi/2,\pi)$.
2206: 
2207: (iii)
2208: The case where either $|G_{cs}(l)|$ or $|G_\Delta(l)|$ is most relevant.
2209: Then both charge and spin fluctuations are suppressed, and 
2210:    the classical treatment is sufficient at lower energy scale.
2211: In this case, we find to which phase the ground state belongs
2212:    by substituting the parameters $G_c(l)$ and $G_s(l)$ into $g_c$
2213:    and $g_s$ in Fig.\ \ref{fig:classical-W2}.
2214: 
2215: (iv) 
2216: The case where $G_{s}(l)$ is most relevant and becomes $-1$
2217:    at $l=l_\sigma$. 
2218: Below this energy scale the spin fluctuations are suppressed and 
2219:    the phase field $\phi$ is locked as $\phi \to 0$ or $\pi$ 
2220:    for $l>l_\sigma$.
2221: The effective Hamiltonian of the remaining charge sector is
2222: %================================================================
2223: \begin{eqnarray}
2224: \mathcal{H}^{\mathrm{eff}}_{\rho}
2225: &=&
2226: \frac{v_F}{2\pi} \sum_p (\partial_x \theta_p)^2
2227: + \frac{v_F}{\pi} \, G_\rho^*
2228:   \left(\partial_x \theta_+ \right)
2229:   \left(\partial_x \theta_- \right)
2230: \nonumber \\ && {}
2231: \mp \frac{v_F}{\pi a^2} \, G_\Delta^* \, \sin \theta
2232: -\frac{v_F}{\pi a^2} \, G_c^* \, \cos 2\theta, 
2233: \end{eqnarray}
2234: %================================================================
2235:    where $G_\rho^*=G_\rho(l_\sigma)-G_{\rho s}(l_\sigma)$,
2236:    $G_\Delta^*=G_\Delta(l_\sigma)$, and
2237:    $G_c^*=G_c(l_\sigma)+G_{cs}(l_\sigma)$. 
2238: The sign $-/+$ of the $G_\Delta^*$ term 
2239:    corresponds to the position of the phase locking $\phi=0/\pi$. 
2240: In this Hamiltonian, both of the nonlinear terms, $\sin \theta$
2241:    and $\cos 2\theta$, are relevant operators.
2242: If $G_c^*<0$, then the situation is the same as the case (ii):
2243:    the $G_\Delta^*$ and $G_c^*$ terms do not compete with each other and
2244:    the possible phase locking pattern is
2245:    $\theta=+\pi/2$ $(-\pi/2)$
2246:    for $\phi=0$ $(\pi)$, where the ground state is the BI state.
2247: If $G_c^*>0$, these two terms compete with each other, since
2248:    the $-(+) \sin \theta$ potential tends to lock the phase field 
2249:    $\theta$ at $\theta=+\pi/2$ $(-\pi/2)$,
2250:    while the $\cos 2\theta$ potential tends to lock it at $\theta=0$ or 
2251:    $\pi$.
2252: In this case, possible ground states are
2253:    the BI state and the BCDW state, and 
2254:    the quantum phase transition between them is
2255:    of the Ising transition type with the central charge $c=1/2$,
2256:    as discussed in the preceding section. 
2257: However, it is hard to estimate quantitatively the critical value of
2258:    the coupling constants at 
2259:    the quantum phase transition.
2260: One way to estimate it is to find a critical point separating the
2261:    basins of attraction to the two strong-coupling fixed points,
2262:    $(G_\Delta^*,G_c^*)\to (+ \infty,-\infty)$ and
2263:    $(0,+\infty)$, in the perturbative RG analysis.
2264:    \cite{Tsuchiizu_JPSJ,Tsuchiizu_JPSJ2}
2265: However, with this method where the cosine and sine terms are treated
2266:    perturbatively, we cannot see the correct picture of the DSG
2267:    theory with the double-well potential structure which leads to the
2268:    Ising transition.
2269: Instead, here we estimate the critical value for the Ising transition
2270:    from the semiclassical arguments:
2271:    The critical value is determined from the condition
2272:    $G_c^*/G_\Delta^*=1/4$. 
2273: 
2274: 
2275: %================================================================
2276: \begin{figure}[t]
2277: \includegraphics[width=7cm]{fig11}
2278: \caption{
2279: Phase diagram of the half-filled extended Hubbard model
2280:  at $\Delta/t=0.1$.
2281: The double line denotes the first-order transition,
2282:   while the single lines denote the second-order transitions.
2283: }
2284: \label{fig:UVW}
2285: \end{figure}
2286: %================================================================
2287: 
2288: We have used the above scheme to obtain the phase diagram
2289:    shown in Fig.\ \ref{fig:UVW}, for which $\Delta/t=0.1$.
2290: The phase diagram at large $U$ and $V$ is similar to
2291:    Fig.\ \ref{fig:phase2}, whereas
2292:    a qualitative charge in the phase diagram is found in
2293:    the region $U,V \lesssim t$.
2294: In agreement with Fabrizio, Gogolin, and Nersesyan,\cite{Fabrizio}
2295:    we obtain two critical points ($U_{c1}<U_{c2}$)
2296:    separating three phases on the $U$ axis:
2297:    the BI state, the BCDW state (= the SDI state\cite{Fabrizio}),
2298:    and the SDW state.
2299: From comparison of Figs.\ \ref{fig:phase2} and \ref{fig:UVW}, 
2300:    we see that the BCDW state in Fig.\ \ref{fig:phase2} has evolved
2301:    continuously into the BCDW state when the alternating site potential
2302:    $\Delta$ is switched on.
2303: The phase diagram in the $\Delta$-$V$ plane is shown in Fig.\
2304:    \ref{fig:UVW-wv}, where $U/t=1$.
2305: Both $\Delta$ and $V$ promote the BI state, while 
2306:    the SDW ground state is obtained for small $\Delta (\ll U)$
2307:    and $V(\ll U)$.
2308: We find that the region of the BCDW state obtained in the EHM
2309:    at $\Delta=0$
2310:    is connected to the region of the BCDW state in the Hubbard model
2311:    with alternating site potential at $V=0$.
2312: 
2313: %================================================================
2314: \begin{figure}[b]
2315: \includegraphics[width=7cm]{fig12}
2316: \caption{
2317: Phase diagram of the half-filled extended ionic Hubbard model 
2318:   on the plane of $\Delta/t$ and $V/t$, where $U/t=1$.
2319: }
2320: \label{fig:UVW-wv}
2321: \end{figure}
2322: %================================================================
2323: 
2324: 
2325: Let us discuss in more detail the critical regime
2326:    in the limit of small $U$, $V$, and $\Delta$.
2327: In this region we can safely neglect the irrelevant terms and set
2328:    $G_{cs}(l)=G_{\rho s}(l)=0$ in the RG equations 
2329:    (\ref{eq:RGf-b})--(\ref{eq:RGf-e}).
2330: First we consider the case $V=0$.
2331: Integrating out the RG equations 
2332:    (\ref{eq:RGf-b})--(\ref{eq:RGf-gs}) analytically 
2333: and following the criterion discussed above,
2334:    we obtain asymptotic expansion of 
2335:    the critical values for small $\Delta/t$:
2336: %================================================================
2337: \begin{eqnarray}
2338: U_{c1}^{0} \!\! &=& \!\!
2339: \frac{2\pi t}{\ln (t/\Delta)}
2340: \left[1 - \frac{C}{\ln(t/\Delta)}+\cdots\right],
2341: \\
2342: U_{c2}^{0} \!\! &=& \!\!
2343: \frac{2\pi t}{\ln (t/\Delta)}
2344: \left[1 + C'\frac{\ln\ln(t/\Delta)}{\ln(t/\Delta)} 
2345:                      + O\left(\frac{1}{\ln(t/\Delta)}\right) \right],
2346: \nonumber \\
2347: \end{eqnarray}
2348: %================================================================
2349: where $C$ and $C'$ are positive constants of order unity. 
2350: The $\Delta$ dependence of $U_{c1}^{0}$ is different from 
2351:    the result in Refs.\ \onlinecite{Fabrizio} since
2352:    the lowest correction to $2\pi t/\ln(t/\Delta)$ 
2353:    is not $O\biglb(\ln[\ln(t/\Delta)]/\ln(t/\Delta)\bigrb)$, but
2354:    $O\biglb(1/\ln(t/\Delta)\bigrb)$.
2355: Our results suggest that the ratio of $U_{c2}^{0}$ to
2356:    $U_{c1}^{0}$ becomes
2357:    $U_{c2}^{0}/U_{c1}^{0}=1+C' \ln[\ln(t/\Delta)]/\ln(t/\Delta)$.
2358: At present we do not know where this difference comes from.
2359: We extend this analysis to the case with finite $V(\ll U)$
2360:    and examine the $V$ dependence of $U_{c1}$ and $U_{c2}$.
2361: We note that $G_\rho(l) \neq G_c(l)$ in this case since
2362:    the SU(2) symmetry of the charge sector is broken.
2363: We integrate the RG equations analytically for small $V\neq 0$
2364:    and obtain the corrections to order $V$,
2365: %================================================================
2366: \begin{subequations}
2367: \label{U_c}
2368: \begin{eqnarray}
2369: U_{c1}
2370: \!\! &=& \!\!
2371: U_{c1}^0
2372: -V\left[\frac{2}{3}+O\left(\frac{1}{\ln(t/\Delta)}\right)\right]
2373: , 
2374: \\
2375: U_{c2} 
2376: \!\! &=& \!\! 
2377: U_{c2}^0
2378: -V\left[\frac{2}{3}
2379:         +O\left(\frac{\ln\ln(t/\Delta)}{\ln(t/\Delta)}\right)\right]
2380: ,
2381: \end{eqnarray}
2382: \end{subequations}
2383: %================================================================
2384: implying that the BCDW state survives upon inclusion of
2385:    the $V(\ll U)$ term. 
2386: We note that 
2387:    $U_{c1}$ and $U_{c2}$ have a similar linear dependence on $V$.
2388: From Eqs.\ (\ref{U_c}) and 
2389:    Figs.\ \ref{fig:UVW} and \ref{fig:UVW-wv}, 
2390:    we conclude that the phase diagram exhibits reentrant
2391:    behavior as $V$ increases from zero with $\Delta$ and $U$
2392:    being fixed at values near a quantum critical point.
2393: 
2394: 
2395: 
2396: %================================================================
2397: \begin{figure}[t]
2398: \includegraphics[width=8.5cm]{fig13}
2399: \caption{
2400: Schematic phase diagram of the half-filled extended Hubbard model
2401:  at (a) $V=0$, (b) $V\ll t$, and (c) $V\gg t$.
2402: The single lines represent second-order transitions, and the
2403:  double line in (c) represents a first-order transition.
2404: }
2405: \label{fig:schematicWU}
2406: \end{figure}
2407: %================================================================
2408: 
2409: 
2410: Since the Hamiltonian $H'$ has three free parameters ($U/t$, $V/t$,
2411: and $\Delta/t$) at half filling, the ground-state phase diagram becomes
2412: a three-dimensional (3D) diagram.
2413: Instead of drawing such a 3D plot, here we show two-dimensional
2414: tomographic phase diagrams.
2415: Figure \ref{fig:schematicWU} shows schematic phase diagrams in the
2416: $\Delta$-$U$ plane for three typical cases $V/t=0$, $V/t\ll1$, and
2417: $V/t\gg1$.
2418: We see that the nearest-neighbor repulsion enhances the BI phase
2419: and destroys the BCDW phase at large $V$, where the direct transition
2420: between the BI and SDW phases is first order.
2421: The recent numerical study of the ionic Hubbard model\cite{Torio}
2422: reports a similar phase diagram as Fig.\ \ref{fig:schematicWU}(a).
2423: The first-order transition line in Fig.\ \ref{fig:schematicWU}(c)
2424: asymptotically approaches the line $U=2\Delta+2V$.
2425: 
2426: Figure \ref{fig:schematicWV} shows schematic phase diagrams in
2427: the $\Delta$-$V$ plane for $U/t\ll1$ and $U/t\gg1$.
2428: At large $U$ and $V$ there appears a direct first-order transition
2429: between the BI and SDW phases in Fig.\ \ref{fig:schematicWV}(b).
2430: This first-order transition is in agreement with 
2431: the results obtained from the  strong-coupling analysis \cite{Nagaosa} 
2432: and numerical calculations. \cite{Girlando,Yonemitsu}
2433: 
2434: %================================================================
2435: \begin{figure}[t]
2436: \includegraphics[width=8.5cm]{fig14}
2437: \caption{
2438: Schematic phase diagram of the half-filled extended Hubbard model
2439:  at (a) $U\ll t$ and (b) $U\gg t$.
2440: The single lines represent second-order transitions, and the
2441:  double line represents a first-order transition.
2442: }
2443: \label{fig:schematicWV}
2444: \end{figure}
2445: %================================================================
2446: 
2447: 
2448: \subsection{Discussions on previous numerical results}
2449: 
2450: As mentioned in Introduction, many groups have already reported
2451:   on numerical studies of the ground-state phase diagram of the ionic
2452:   Hubbard model.
2453: Various numerical techniques were used in these studies, including
2454:   the density-matrix renormalization-group (DMRG) method,
2455:   \cite{Takada,Qin,Brune,YZZhang,Manmana}
2456:   the quantum Monte Carlo method, \cite{Wilkens,Refolio}
2457:   a finite-size cluster method, \cite{Anusooya-Pati} and 
2458:   a level crossing analysis. \cite{Torio}
2459: The main issue here is whether or not the SDI phase (BCDW phase) exists,
2460:   and so far these numerical studies do not seem to have reached
2461:   complete agreement yet.
2462: Although most of recent studies report that the SDI phase
2463:   appears near the boundary between the SDW phase and the BI phase,
2464:   \cite{Qin,Brune,YZZhang,Wilkens,Torio,Refolio,Manmana}
2465:   there are still some conflicting claims in the literature.
2466: A less controversial issue\cite{noteUc2} is the determination of the second
2467:   critical value $U_{c2}$ at which a spin gap closes
2468:   and which can be estimated
2469:   by computing the spin gap directly \cite{Takada,Qin}
2470:    or by examining the BCDW order parameter.\cite{YZZhang,Manmana}
2471: The determination of the critical point $U_{c1}$ and the critical
2472:   behaviors around it are more controversial issues.
2473: One way to estimate the critical value $U_{c1}$ is to use
2474:    the complex parameter introduced by Resta and Sorrela.\cite{Resta}
2475: Its diverging behavior at $U=U_{c1}$ indeed allows one to determine
2476:    the critical point.\cite{Takada,Wilkens}
2477: Another way to determine the critical point is to find a gap closing
2478:   point in excitation spectra.
2479: Since the charge sector is responsible for the quantum phase transition
2480:    at $U=U_{c1}$, one might try to look at a charge gap directly.
2481: However, numerical studies have found that a naive charge gap does not
2482:    vanish at the critical point and is always finite.
2483: Recent studies have shown\cite{Qin,Brune,Manmana} that the excitation gap
2484:    that vanishes at $U=U_{c1}$ is the gap to the first excited state
2485:    that has the same charge and spin quantum numbers as the ground state.
2486: Let us discuss this point in more detail below.
2487: 
2488: In numerical studies,\cite{Takada,Qin,Brune}
2489:    the ``charge gap'' $\Delta_c$ was \textit{defined} as
2490:     $\Delta_c = E_0(L/2+1,L/2)+E_0(L/2-1,L/2)-2E_0(L/2,L/2)$,
2491:    where $E_0(N_\uparrow,N_\downarrow)$ is the lowest energy 
2492:    of a finite-size system with an even number of sites $L$ that has
2493:    $N_\uparrow$ up-spin and $N_\downarrow$ down-spin electrons.
2494: This quantity $\Delta_c$ measures the energy of the excitation with the 
2495:    topological charge $Q=\pm 1$ and $S_z=\pm1/2$
2496:    [Eq.\ (\ref{eq:topological_charge})], and is rather a single-electron
2497:    excitation gap.
2498: According to the bosonization theory (Sec.\ IV A),
2499:    the charge transition at $U=U_{c1}$
2500:    is described by the ``$\varphi^4$'' theory and is
2501:    in the Ising universality class.
2502: The transition occurs when 
2503:    two degenerate local minima of the effective potential for the
2504:    charge fields merge into a single local minimum.
2505: As one approaches the transition point from the Ising ordered phase
2506:    (that is, the SDI phase), the topological charge 
2507:    $Q=\pm 2\alpha_\theta/\pi$ of a lowest-energy excitation is 
2508:    decreasing to zero, while excitations with $Q=1$ remain massive.
2509: Therefore the charge gap $\Delta_c$ does not vanish
2510:    at this Ising critical point, and
2511:    this quantum phase transition cannot be detected with $\Delta_c$.
2512: Qin \textit{et al}.\ and Manmana \textit{et al}.\ also used
2513:    $\Delta_e =  E_1\left(L/2,L/2\right) -E_0\left(L/2,L/2\right)$
2514:    in their numerical analysis,
2515:    where $E_1(N_\uparrow,N_\downarrow)$ is the energy of the first 
2516:    excited state.\cite{Qin,Manmana}
2517: The quantity $\Delta_e$ measures excited states with 
2518:    the same number of electrons,
2519:    whose total topological charge $Q=0$ in the sine-Gordon scheme.
2520: In the Ising ordered phase,
2521:    the first excited state with the topological charge $Q=0$
2522:    would be a bound state (or breather) of a soliton
2523:    with the topological charge $+2\alpha_\theta/\pi$ and an antisoliton
2524:    with the charge $-2\alpha_\theta/\pi$, whose energy vanishes at
2525:    the critical point.
2526: On the other hand, in the Ising disordered phase near the critical point,
2527:    the potential is almost flat and has very small curvature.
2528: The low-energy excitations would then be small oscillations around
2529:    potential minima (rather than soliton/antisoliton) whose energy
2530:    approaches zero as $U\to U_{c1}-0$.
2531: Thus the exciton gap $\Delta_e$ is a right measure to detect
2532:    the quantum phase transition at $U=U_{c1}$.
2533: 
2534: 
2535: 
2536: \section{Effect of bond dimerization}\label{sec:dimer}
2537: 
2538: In this section, we consider the 1D EHM with staggered bond 
2539:    dimerization,\cite{Su,Ortiz} i.e.,
2540:    the Peierls modulation of the hopping matrix element.
2541: The total Hamiltonian $H''$ 
2542:    is given by $H''=H+H_\delta$, where $H$ is defined in
2543:    Eq.\ (\ref{eq:H1D}) and
2544: %================================================================
2545: \begin{equation}
2546: H_{\delta} = \delta \sum_{j,\sigma} (-1)^j \, 
2547:   ( c_{j,\sigma}^\dagger c_{j+1,\sigma}+ \mathrm{H.c.} )
2548: .
2549: \end{equation}
2550: %================================================================
2551: Without loss of generality we can assume $\delta>0$.
2552: When $V=0$, the model is called ``Peierls-Hubbard model.''
2553: The one-dimensional Mott insulator, realized when
2554:    $U>0$ and $V=0$, is known to be 
2555:    unstable against the Peierls distortion, \cite{Bray,Ukrainskii} and
2556:    as a result the ground state changes from the SDW state
2557:    into the BCDW state regardless of the magnitude of the Hubbard
2558:    interaction $U$.
2559: Such an instability comes from the fact that
2560:    the bond dimerization tends to concentrate the electron density
2561:    onto bonds, without any conflict with
2562:    the Hubbard, $U$, repulsion.\cite{Brune}
2563: However, the nearest-neighbor Coulomb repulsion $V$ competes with
2564:    this $\delta$ term, since the $V$ interaction likes to localize two
2565:    electrons on a single site and promotes the CDW state.
2566: Here we investigate the instability of the BCDW state against the
2567:    intersite Coulomb repulsion $V$, and clarify the 
2568:    critical behavior near the transition between the BCDW state and
2569:    the CDW state.
2570: 
2571: The bond dimerization $H_\delta$ is bosonized as
2572:    $H_\delta=\int dx\mathcal{H}_\delta$, where
2573: %================================================================
2574: \begin{equation}
2575: \mathcal{H}_\delta =
2576:  -\frac{g_\delta}{2(\pi a)^2} \cos \theta \, \cos \phi
2577: \end{equation}
2578: %================================================================
2579:    and $g_\delta=8\pi \delta a$.
2580: One finds that the EHM with the bond dimerization
2581:    also has a two-component DSG structure.
2582: Here
2583:    the charge phase field $\theta$ is subjected to the potential
2584:    $\cos \theta$ instead of $\sin \theta$ of the $g_\Delta$ term
2585:    [Eq.\ (\ref{eq:Hw_bosonization})], while
2586:    the locking potential for the spin phase field $\phi$ has 
2587:    the same structure as that of the $g_\Delta$ term.
2588: 
2589: It is important to note that the BCDW order parameter
2590:    $\mathcal{O}_\mathrm{BCDW}$ takes a nonvanishing expectation value
2591:    for any $U$ and $V$ if $\delta \ne0$,
2592:    as $\mathcal{H}_\delta\propto\mathcal{O}_\mathrm{BCDW}$.
2593: In this section we will not use the term BCDW to characterize
2594:    phases, and, in particular, the phase containing the trivial
2595:    Peierls insulator ($U=V=0$ and $\delta\ne0$) is called the Peierls
2596:    insulating (PI) phase.
2597: 
2598: %================================================================
2599: \begin{table}[t]
2600: \caption{Possible ground states  and
2601:   the position of locked phase fields,
2602:   determined from Eq.\ (\protect{\ref{eq:VD0}}).
2603: }
2604: \label{table:phase-locking-D}
2605: \begin{ruledtabular}
2606: \begin{tabular}{lc}
2607: Phase  &  $(\theta,\phi )$       \\
2608: \hline
2609: SDW   &  $ (0,\pm\gamma_\phi ), 
2610:                 (\pi,\pm(\pi-\gamma_\phi) )$  \\
2611: CDW   &  $ (\pm \gamma_\theta, 0 ),
2612:                 (\pm(\pi - \gamma_\theta), \pi )$  \\
2613: PI (for $g_\delta>0$)   &  
2614:               $(0,0 ), (\pi,\pi )$  \\
2615: PI (for $g_\delta<0$)   &
2616:               $(0,\pi ),(\pi,0 )$  \\
2617: BSDW &        $(\pm \pi/2, \pm \pi/2 )$ 
2618: \end{tabular}
2619: \end{ruledtabular}
2620: \end{table}
2621: %================================================================
2622: \subsection{Semiclassical analysis}
2623: 
2624: We begin with semiclassical analysis of the model with the $g_\delta$
2625:    term.
2626: We neglect spatial variations of the phase fields
2627:    in $\mathcal{H}+\mathcal{H}_\delta$ and consider the potential
2628: %================================================================
2629: \begin{eqnarray}
2630: V_\delta(\theta,\phi)
2631: &=&
2632: -g_c \cos2\theta + g_s \, \cos 2\phi
2633: -g_{cs} \, \cos 2\theta \, \cos 2\phi 
2634: \nonumber \\ 
2635: && {}
2636: -g_\delta \, \cos\theta \, \cos \phi ,
2637: \label{eq:VD}
2638: \end{eqnarray}
2639: %================================================================
2640:    where $g_{cs}=g_{3\parallel}<0$.
2641: 
2642: First, we consider the simpler case where $g_{cs}=0$, which corresponds
2643:    to the situation where $g_{cs}$ is irrelevant in the RG sense.
2644: The potential in this case is
2645: %================================================================
2646: \begin{equation}
2647: V_\delta^0(\theta,\phi) =
2648: -g_c \cos2\theta + g_s \, \cos 2\phi - g_\delta \, \cos\theta \, \cos \phi.
2649: \label{eq:VD0}
2650: \end{equation}
2651: %================================================================
2652: The positions of the potential minima are determined by
2653:    the saddle-point equations,
2654:    $\partial V_\delta^0(\theta,\phi)/\partial \theta=0$ and  
2655:    $\partial V_\delta^0(\theta,\phi)/\partial \phi=0$.
2656: We find that the potential has
2657:    the double-well structure for the $\theta$ ($\phi$) phase field
2658:    when $g_c<-|g_\delta|/4$ ($g_s>|g_\delta|/4$). 
2659: Here we introduce $\gamma_\theta^0$ and $\gamma_\phi^0$
2660:    ($0 \le \gamma_\theta^0, \gamma_\phi^0 \le \pi$) defined by
2661: %================================================================
2662: \begin{equation}
2663: \gamma_\theta^0 
2664:   = \left| \cos^{-1} \left(-\frac{g_\delta}{4g_c}\right) \right|,
2665: \quad
2666: \gamma_\phi^0
2667:   = \left| \cos^{-1} \left(\frac{g_\delta}{4g_s}\right) \right|
2668: \end{equation}
2669: %================================================================
2670:    for $|g_\delta/g_c| \le 4$ and $|g_\delta/g_c|\le4$, respectively.
2671: The solutions to the saddle-point equations can be classified
2672:    into the following four classes:
2673: (i) the PI state,
2674:    $(\theta,\phi)=(0,0), (0,\pi ), (\pi,0)$, or $(\pi,\pi)$
2675: [for $g_\delta>0$, the phase fields are locked at
2676:    $(\theta,\phi)=(0,0)$ or 
2677:    $(\pi,\pi)$, while for $g_\delta<0$ the phase fields are 
2678:    locked as $(\theta,\phi)=(0,\pi)$ or $(\pi,0)$];
2679: (ii) the pure BSDW state,
2680:    $(\theta,\phi)=(\pi/2,\pm \pi/2)$ or $(-\pi/2,\pm \pi/2)$;
2681: (iii) the ``SDW'' state with both the SDW order and the BCDW order,
2682:    $(\theta,\phi)=(0,\pm \gamma_\phi^0)$ or
2683:      $\biglb(\pi, \pm (\pi - \gamma_\phi^0)\bigrb)$; and
2684: (iv) finally, the ``CDW'' state with both the CDW order and
2685:    the BCDW order,
2686:     $(\theta,\phi)=(\pm \gamma_\theta^0, 0 )$ or
2687:     $\biglb(\pm(\pi - \gamma_\theta^0), \pi \bigrb)$.
2688: %================================================================
2689: \begin{figure}[t]
2690: \includegraphics[width=7.cm]{fig15}
2691: \caption{
2692: Positions of locked phase fields $\theta$ and $\phi$
2693:    in the respective states for $g_\delta>0$. 
2694: }
2695: \label{fig:potmin-D}
2696: \end{figure}%
2697: %================================================================
2698: The possible ground states and positions of locked phase fields
2699:   are summarized in Table \ref{table:phase-locking-D} and
2700:   Fig.\ \ref{fig:potmin-D}.
2701: In these states the potential energy reads
2702: %================================================================
2703: \begin{subequations}
2704: \begin{eqnarray}
2705: V_{\rm PI}^0  &=&  - g_c + g_s  -|g_\delta| ,
2706: \\
2707: V_{\rm BSDW}^0  &=& + g_c -g_s ,
2708: \\
2709: V_{\rm SDW}^0
2710:    &=&  - g_c - g_s - \frac{g_{\delta}^2}{8g_s} ,
2711: \label{eq:V_SDW+BCDW}
2712: \\
2713: V_{\rm CDW}^0
2714:    &=&  + g_c + g_s + \frac{g_{\delta}^2}{8g_c}.
2715: \label{eq:V_CDW+BCDW}
2716: \end{eqnarray}
2717: \end{subequations}
2718: %================================================================
2719: In deriving Eqs.\ (\ref{eq:V_SDW+BCDW}) and 
2720:    (\ref{eq:V_CDW+BCDW}), we have assumed $|g_\delta/g_s| \le 4$ and 
2721:    $|g_\delta/g_c| \le 4$, respectively.
2722: The PI state is stabilized by the first-order contribution 
2723:    of the $g_\delta$ term.
2724: Furthermore, if $g_s>0$ ($g_c<0$),
2725:    the SDW state (the CDW state) is also stabilized due to 
2726:    second-order contribution of $g_\delta$.
2727: %================================================================
2728: \begin{figure}[t]
2729: \includegraphics[width=7cm]{fig16}
2730: \caption{
2731: Phase diagram obtained by minimizing the potential energy 
2732:    $V_\delta^0(\theta,\phi)$ [Eq.\ \protect{(\ref{eq:VD0})}].
2733: The phase boundary of the BSDW state is given  by
2734:   the curve $g_c g_s=-g_\delta^2/16$ with $g_c<0$.
2735: The phase boundary between the PI state and the SDW state
2736:    and that between the PI state and the CDW state are
2737:    given by the lines $g_s= |g_\delta|/4$ with $g_c > |g_\delta|/4$
2738:    and $g_c=-|g_\delta|/4$ with $g_s < |g_\delta|/4$, respectively.
2739: All the phase transitions in this figure are continuous.
2740: A multicritical point is at $(g_c,g_s)=(-|g_\delta|/4,|g_\delta|/4)$.
2741: }
2742: \label{fig:classical-D}
2743: \end{figure}
2744: %================================================================
2745: %================================================================
2746: \begin{figure}[t]
2747: \includegraphics[width=7cm]{fig17}
2748: \caption{
2749: Phase diagram obtained by minimizing the potential energy 
2750:    $V_\delta(\theta,\phi)$ [Eq.\ \protect{(\ref{eq:VD})}]
2751:    drawn for $|g_\delta|/8<|g_{cs}|<|g_\delta|/4$.
2752: Multicritical points are located at
2753:    $(g_c,g_s)=(-|g_{cs}|,|g_{cs}|)$ and
2754:    $(-\frac{1}{4}|g_\delta|+|g_{cs}|,\frac{1}{4}|g_\delta|-|g_{cs}|)$.
2755: The boundary of the BSDW phase is
2756:    $(g_c+|g_{cs}|)(g_s-|g_{cs}|)<-g_\delta^2/16$.
2757: The edges of the PI phase are defined by the lines
2758:    $g_c=-\frac{1}{4}|g_\delta|+|g_{cs}|$ and
2759:    $g_s=\frac{1}{4}|g_\delta|-|g_{cs}|$.
2760: The double line denotes a first-order transition, and the single lines
2761:    denote continuous transitions.
2762: Within the semiclassical analysis the ground state in the shaded
2763:    region has the coexisting order of the SDW, CDW, BCDW, and BSDW.
2764: }
2765: \label{fig:classical-Dcs}
2766: \end{figure}
2767: %================================================================
2768: The phase diagram obtained by comparing these energies 
2769:    is shown in Fig.\ \ref{fig:classical-D}.
2770: 
2771: From the above semiclassical analysis one might conclude that 
2772:    the topological charge $S_z$ [Eq.\ (\ref{eq:topological_charge})]
2773:    becomes fractional in the SDW phase and that
2774:    the Ising-type phase transition in the spin sector takes place
2775:    on the boundary between the PI state and the SDW state.
2776: However, as discussed in Sec.\ \ref{sec:staggered}, 
2777:    the global SU(2) symmetry prohibits 
2778:    the Ising criticality in the spin sector
2779:    and changes
2780:    the SDW phase in Fig.\ \ref{fig:classical-D}
2781:    into the PI phase.
2782: 
2783: 
2784: Next we include the $g_{cs}$ term.
2785: Table \ref{table:phase-locking-D} still stands if we replace
2786: $g_c$ and $g_s$ with $g_c-|g_{cs}|$ and $g_s+|g_{cs}|$ in
2787: $\gamma_\theta^0$ and $\gamma_\phi^0$, respectively.
2788: The phase diagram obtained by minimizing the potential energy
2789: $V_\delta(\theta,\phi)$ is shown in Fig.~\ref{fig:classical-Dcs}.
2790: New features compared with Fig.~\ref{fig:classical-D} are the appearance
2791: of a first-order transition line
2792: and of the new phase in which the ground state has the coexisting order
2793: of the SDW, CDW, BCDW, and BSDW.
2794: The new phase is shown as the shaded region in
2795: Fig.~\ref{fig:classical-Dcs}, which is surrounded by the three curves
2796: defined by
2797: \begin{subequations}
2798: \begin{eqnarray}
2799: &&(g_c+|g_{cs}|)(g_s-|g_{cs}|)=-\frac{g_\delta^2}{16},\\
2800: &&(g_c+|g_{cs}|)(g_s+|g_{cs}|)^2=-\frac{g_\delta^2}{16}(g_s-|g_{cs}|),\\
2801: &&(g_s-|g_{cs}|)(g_c-|g_{cs}|)^2=-\frac{g_\delta^2}{16}(g_c+|g_{cs}|).
2802: \qquad
2803: \end{eqnarray}
2804: \end{subequations}
2805: 
2806: 
2807: Let us focus on the phases which can be realized when
2808:    $g_s \simeq g_c$, in view of the fact that
2809:    in the extended Hubbard model both
2810:    $g_s(=g_{1\perp})$ and $g_c(=g_{3\perp})$ are given by 
2811:    $(U-2V)$ in the lowest order.
2812: Along the line $g_s\simeq g_c$ in Figs.\ \ref{fig:classical-D}
2813:    and \ref{fig:classical-Dcs}, 
2814:    there are three possible phases: the SDW state, 
2815:    the PI state, and the CDW state.
2816: Since the SDW state is prohibited by the SU(2) symmetry
2817:    and becomes the PI state, we expect to have only two phases,
2818:    the PI state and the CDW state, and a single phase
2819:    transition between them.
2820: The transition is continuous at $|g_{cs}/g_\delta|\ll1$ and changes into
2821: a discontinuous transition when $g_{cs}$ exceeds $|g_\delta|/4$.
2822: 
2823: 
2824: 
2825: \subsection{Renormalization-group analysis}
2826: 
2827: Next we perform perturbative RG analysis to take into account quantum
2828:    fluctuations.
2829: The one-loop RG equations for coupling constants
2830:    in $\mathcal{H}+\mathcal{H}_\delta$ are given by
2831: %================================================================
2832: \begin{eqnarray}
2833: \frac{d}{dl} G_\delta
2834:   &=& {} + G_\delta
2835:     + \frac{1}{2} \, G_\delta \, G_\rho 
2836:     + G_\delta \, G_c
2837: \nonumber \\ && {}
2838:     - \frac{3}{2} \, G_\delta \, G_s
2839:     + \frac{3}{4} \, G_\delta \, G_{cs}
2840:     - \frac{3}{8} \, G_\delta \, G_{\rho s}
2841: ,
2842: \label{eq:RGd-b}
2843: \\
2844: \frac{d}{dl} G_\rho 
2845:   &=& {} + \frac{1}{4} \, G_\delta^2
2846:     + 2 \, G_c^2 + G_{cs}^2 +  G_s \, G_{\rho s}
2847: ,
2848: \\
2849: \frac{d}{dl} G_c 
2850:   &=& {}
2851:   + \frac{1}{4} \, G_\delta^2 
2852:     + 2 \, G_\rho\, G_c
2853:   - G_s \, G_{cs} - G_{cs} \, G_{\rho s}
2854: , \nonumber \\
2855: \\
2856: \frac{d}{dl} G_s 
2857:   &=& {} - \frac{1}{4} \, G_\delta^2 
2858:     - 2 \, G_s^2 - G_c \, G_{cs} - G_{cs}^2
2859: ,
2860: \\
2861: \frac{d}{dl} G_{cs} 
2862:   &=& {} + \frac{1}{4}  \,G_\delta^2 
2863:     - 2 \, G_{cs} + 2 \, G_\rho \, G_{cs} - 4 \, G_s \, G_{cs}
2864: \nonumber \\ && {}
2865:     - 2 \, G_c\, G_{s}
2866:     - 2 \, G_c\, G_{\rho s}
2867:     - 4 \, G_{cs} \, G_{\rho s}
2868: ,
2869: \\
2870: \frac{d}{dl} G_{\rho s} 
2871:   &=& {} - \frac{1}{4}  \, G_\delta^2 
2872:     - 2 \, G_{\rho s} + 2 \, G_\rho \, G_s
2873: \nonumber \\ && {}
2874:     - 4 \, G_c \, G_{cs} - 4 \, G_{cs}^2
2875:     - 4 \, G_s \, G_{\rho s}
2876: .
2877: \label{eq:RGd-e}
2878: \end{eqnarray}
2879: %================================================================
2880: The initial value of $G_\delta(l)$ is given by 
2881:    $G_\delta(0)=2\delta/t$ and those of the other coupling constants
2882:    are $G_\nu(0)=g_\nu /(4\pi t)$.
2883: We note that these RG equations are invariant under 
2884:    the sign change of $G_\delta(l)$.
2885: We can thus assume $G_\delta(0) \ge 0$ without losing generality.
2886: 
2887: 
2888: To find the ground-state phase diagram of the system,
2889:    we solve the scaling equations 
2890:    (\ref{eq:RGd-b})--(\ref{eq:RGd-e}) numerically,
2891:    as in the preceding sections.
2892: We determine to which phase the ground state belongs by
2893:    looking at which one of the couplings
2894:    $G_\delta(l)$, $G_c(l)$, $G_s(l)$, and $G_{cs}(l)$ becomes most relevant. 
2895: For repulsive $U$ and $V$ there are four possibilities as listed below.
2896: 
2897: (i)
2898: If $G_c$ is most relevant and $G_c(l) \to 1 $ at $l=l_{\rho+}$,
2899:    then
2900:    the phase field $\theta$ is locked at $\theta=0$ or $\pi$, 
2901:    and 
2902:    the effective Hamiltonian for the spin sector at $l\ge l_{\rho+}$
2903:    becomes
2904: %================================================================
2905: \begin{eqnarray}
2906: \mathcal{H}^{\mathrm{eff}}_{\sigma+}
2907: &=&
2908: \frac{v_F}{2\pi} \sum_{p=\pm} (\partial_x \phi_p)^2
2909:  - \frac{v_F}{\pi} \, G_s^* (\partial_x \phi_+ )(\partial_x \phi_-) 
2910: \nonumber \\ && {}
2911:      \mp \frac{v_F}{\pi a^2} \, G_\delta^* \,  \cos \phi
2912:      +  \frac{v_F}{\pi a^2} \, G_s^* \, \cos 2\phi,
2913: \end{eqnarray}
2914: %================================================================
2915:    where $G_s^*=G_s(l_{\rho+})-G_{cs}(l_{\rho+})$ and 
2916:    $G_\delta^*=G_\delta(l_{\rho+})$,
2917:   and the sign $-/+$ of the $G_\delta^*$ term
2918:    corresponds to the location of the phase locking $\theta=0/\pi$.
2919: This effective theory is the same as Eq.\ (\ref{eq:Heff_sigma-}).
2920: As seen before, 
2921:    regardless of the sign of $G^*_s$, 
2922:    the phase field $\phi$ is locked at $\phi=0$ or $\pi$
2923:    depending on the position of the charge phase locking 
2924:    $\theta=0$ or $\pi$.
2925: Thus we have the phase locking $(\theta,\phi)
2926:     =(0,0)$ or $(\pi,\pi)$, 
2927:    i.e., the PI state as the ground state.
2928: We note that due to the SU(2) spin rotation symmetry
2929:    the SDW state cannot be realized even if $G_s^*>0$.
2930: 
2931: (ii)
2932: If $G_c$ is most relevant and $G_c(l) \to -1$ at $l=l_{\rho-}$,
2933:    then the phase field $\theta$ is locked at 
2934:    $\theta = \pm \pi/2$.
2935: The effective Hamiltonian for the spin part is
2936: %================================================================
2937: \begin{eqnarray}
2938: \mathcal{H}^{\mathrm{eff}}_{\sigma -}
2939: &=&
2940: \frac{v_F}{2\pi} \sum_p (\partial_x \phi_p)^2
2941:      -  \frac{v_F}{\pi} \, G_s^*    (\partial_x \phi_+ )(\partial_x \phi_-) 
2942: \nonumber \\ && {}
2943:      + \frac{v_F}{\pi a^2} \, G_s^* \, \cos 2\phi,
2944: \end{eqnarray}
2945: %================================================================
2946:    where $G_s^* = G_s(l_{\rho-}) + G_{cs}(l_{\rho-})$.
2947: We have verified numerically that $G_s^*$ always becomes negative
2948:    in this case.
2949: The $G_s^*$ terms are then marginally relevant
2950:    [$G_s^*(l)\to - \infty$ under scaling].
2951: The phase fields are then locked at 
2952:    $( \theta , \phi )= ( \pm \pi/2, 0), ( \pm \pi/2, \pi )$,
2953:    which corresponds to the CDW phase
2954:    with $\gamma_\theta \to \pi/2$
2955:    (i.e., $g_\delta \to 0$, see Table \ref{table:phase-locking-D}).
2956: Since $\mathcal{H}_\delta\propto\mathcal{O}_\mathrm{BCDW}$,
2957:    the order parameter of
2958:    the BCDW should have a nonvanishing expectation value.
2959: We thus conclude that the ground state is in the CDW phase.
2960: 
2961: (iii)
2962: If either $G_\delta$ or $G_{cs}$ is most relevant,
2963:    both charge and spin fluctuations are suppressed.
2964: In this case the semiclassical treatment is justified, and
2965:    we can determine to which phase the ground state belongs
2966:    by substituting $G_c$ and $G_s$ to $g_c$ and $g_s$
2967:    in Fig.\ \ref{fig:classical-Dcs}.
2968: 
2969: 
2970: (iv)
2971: If $G_s$ is most relevant and $G_{s}(l)\to -1$ at $l=l_\sigma$,
2972:    the spin fluctuations are suppressed and  
2973:    the phase field $\phi$ is locked at $\phi \to 0$ or $\pi$ 
2974:    below this energy scale.
2975: The effective Hamiltonian at $l\ge l_\sigma$ is given by
2976: %================================================================
2977: \begin{eqnarray}
2978: \mathcal{H}^{\mathrm{eff}}_{\rho}
2979: &=&
2980: \frac{v_F}{2\pi} \sum_{p=\pm} (\partial_x \theta_p)^2
2981: + \frac{v_F}{\pi} \, G_\rho^*
2982:   \left(\partial_x \theta_+ \right)
2983:   \left(\partial_x \theta_- \right)
2984: \nonumber \\ && {}
2985: \mp \frac{v_F}{\pi a^2} \, G_\delta^* \, \cos \theta
2986: -\frac{v_F}{\pi a^2} \, G_c^* \, \cos 2\theta, 
2987: \end{eqnarray}
2988: %================================================================
2989:    where $G_\rho^*=G_\rho(l_\sigma)-G_{\rho s}(l_\sigma)$,
2990:    $G_c^*=G_c(l_\sigma)+G_{cs}(l_\sigma)$, and  
2991:    $G_\delta^*=G_\delta(l_\sigma)$.
2992: The sign $-/+$ of the $G_\delta^*$ term 
2993:    corresponds to the phase locking $\phi=0/\pi$.
2994: Both of the nonlinear terms $\cos \theta$ and
2995:    $\cos 2\theta$ are relevant perturbations.
2996: If $G_c^*<0$,
2997:    these two terms compete with each other, and
2998:    this DSG model exhibits the Ising criticality.
2999: The ground state is either in the PI phase or
3000:    in the CDW phase, and there is an Ising-type quantum phase
3001:    transition between the two phases.
3002: Here we estimate the Ising critical point
3003:    from the semiclassical analysis.
3004: That is, the critical value is determined from the condition
3005:    $G_c^*/G_\delta^*=-1/4$ (see Fig.\ \ref{fig:classical-D}). 
3006: If $G_c^*>0$, these two terms do not compete and thus the
3007:    phase locking is $\theta=0$ $(\pi)$ for $\phi=0$ $(\pi)$,
3008:    where the ground state is the PI state.
3009: 
3010: %================================================================
3011: \begin{figure}[t]
3012: \includegraphics[width=7cm]{fig18}
3013: \caption{
3014: Phase diagram 
3015:    of the half-filled extended Hubbard model with
3016:    $\delta/t=0.1$.
3017: The second-order transition line (single line) turns into
3018:    the first-order transition line (double line) at the
3019:    tricritical point $(U_c,V_c)\approx(4.9t, 2.3t)$.
3020: }
3021: \label{fig:UVD}
3022: \end{figure}
3023: %================================================================
3024: 
3025: The resultant phase diagram
3026:   in the $U$-$V$ plane is shown in Fig.\ \ref{fig:UVD}.
3027: In the weak-coupling region,
3028:   the transition from the PI state to the 
3029:    CDW state is characterized by the appearance of the
3030:    double-well structure of the effective potential to the
3031:    $\theta$ field, and
3032:    thus the phase transition in Fig.\ \ref{fig:UVD} belongs to the
3033:    Ising universality class.
3034: As we increase $U$ and $V$,
3035:     there appears a tricritical point at $(U_c,V_c)\approx (4.9t,2.3t)$,
3036:     where the phase transition 
3037:     changes from second order to first order.
3038: 
3039: 
3040: 
3041: %================================================================
3042: \begin{figure}[b]
3043: \includegraphics[width=7.5cm]{fig19}
3044: \caption{
3045: Schematic phase diagram of the half-filled extended Hubbard model
3046:  at (a) $V\ll t$ and (b) $V\gg t$.
3047: The single lines represent second-order transitions, and the
3048:  double line represents a first-order transition.
3049: }
3050: \label{fig:schematicDU}
3051: \end{figure}
3052: %================================================================
3053: 
3054: %================================================================
3055: \begin{figure}[t]
3056: \includegraphics[width=8.5cm]{fig20}
3057: \caption{
3058: Schematic phase diagram of the half-filled extended Hubbard model
3059:  at (a) $U=0$, (b) $U\ll t$, and (c) $U\gg t$.
3060: The single lines represent second-order transitions, and the
3061:  double line represents a first-order transition.
3062: }
3063: \label{fig:schematicDV}
3064: \end{figure}
3065: %================================================================
3066: 
3067: 
3068: Figure \ref{fig:schematicDU} shows schematic phase diagrams in the
3069: $\delta$-$U$ plane for $V\ll t$ and $V/t\gg1$.
3070: When $\delta=0$, 
3071:   we obtain three phases (the CDW, BCDW, and SDW phases) for $V\ll t$ (a) 
3072:    and two phases (the CDW and SDW phases) for $V\gg t$ (b), 
3073:   as we discussed
3074:    in Sec.\ \ref{sec:phase_diagram} (see Fig.\ \ref{fig:phase2}).
3075: Upon turning on $\delta$, the SDW ground state changes into the PI state,
3076:    where the transition is described by the Gaussian theory.
3077: On the other hand, the BCDW state changes into the PI state without
3078:    accompanying any singularity: This change is merely lifting of the
3079:    doubly degenerate  BCDW ground states.
3080: 
3081: Figure \ref{fig:schematicDV} shows schematic phase diagrams in
3082: the $\delta$-$V$ plane for $U=0$,  $U/t\ll1$, and $U/t\gg1$.
3083: At $U=0$ we have a single critical value $V_c$ 
3084:    which has the $\delta$ dependence given by
3085:    $V_c \propto 1/\ln(t/\delta)$ for small $\delta$.  
3086: As $U$ and $V$ increase, the phase boundary approaches
3087:    the $U=2V$ line.
3088: The asymptotic form of $V_c$ for $U,V\gg \delta$
3089:    and $\delta\ll t$ 
3090: is given by $V_c=\frac{1}{2}U+ C'' \, U (\delta/t)^{2U/\pi t}$,
3091:    where $C''$ is a numerical constant of the order of unity
3092:   (see also Fig.\ \ref{fig:UVD}).
3093: 
3094: 
3095: 
3096: 
3097: \section{Conclusions}\label{sec:conclusions}
3098: 
3099: In this paper we have 
3100:    studied the ground-state phase diagram of the 
3101:    one-dimensional extended Hubbard model with 
3102:    on-site and nearest-neighbor repulsion $U$ and $V$.
3103: By including higher-order corrections to coupling constants in 
3104:    the $g$-ology,
3105:    we have given a plausible theoretical argument within the RG
3106:    approach for the mechanism of the appearance of the BCDW phase
3107:    at $U\approx 2V$ in the weak-coupling limit.
3108: Our two-step RG approach, however, is not complete in that there
3109:    remains a weak cutoff dependence in the phase boundaries.
3110: This, albeit minor, defect should be resolved with use of a more
3111:    sophisticated systematic RG procedure.
3112: Away from the weak-coupling limit the umklapp scattering between
3113:    the parallel-spin electrons $g_{3\parallel}$ tends to destabilize 
3114:    the BCDW state and eventually gives rise to a bicritical point
3115:    where the two continuous-transition lines merge
3116:    into the SDW-CDW first-order transition line 
3117:    (Fig.\ \ref{fig:phase2}).
3118: We should note, however, that there still remains a difficult
3119:    question as to whether our phase diagram is qualitatively correct
3120:    near the multicritical point (which we call bicritical).
3121: One could imagine, for example, a possibility that a continuous phase
3122:    transition between the BCDW state and the CDW state becomes first
3123:    order before reaching the multicritical point, due to higher-order
3124:    effects that are ignored in our analysis.
3125: If the correct topology of the phase diagram is indeed the same as
3126:    ours (Fig.\ \ref{fig:phase2}), then the critical properties of
3127:    the multicritical point remain to be understood.
3128: We hope that these issues will be resolved by future studies.
3129: 
3130: We have also examined effects of additional staggered site potential
3131:    and bond dimerization in the extended Hubbard model.
3132: In the presence of the staggered site potential,
3133:    we have found that the BCDW state is smoothly connected to the
3134:    SDI phase which is obtained for $V=0$ by
3135:    Fabrizio \textit{et al}. \cite{Fabrizio} 
3136: In this BCDW phase the BCDW order coexists with the CDW order, 
3137:    and the quantum phase transition
3138:    between the BI phase (or the CDW phase) and
3139:    the BCDW phase belongs to the Ising universality class
3140:    ($c=\frac{1}{2}$ CFT).
3141: For finite $V$ the BCDW phase is also destabilized by the
3142:    $g_{3\parallel}$ term, and the direct first-order quantum phase
3143:    transition between the SDW state (= Mott insulating state) and
3144:    the BI state takes place (Fig.\ \ref{fig:UVW}).
3145: In the presence of the staggered bond dimerization
3146:    the SDW phase becomes unstable and the ground state at $V=0$ turns
3147:    out to be the Peierls insulating state.
3148: For $V\ne0$ the phase diagram consists of two phases,
3149:    the PI state and the CDW state, which are
3150:    separated by a phase transition line of the Ising criticality
3151:    (Fig.\ \ref{fig:UVD}).
3152: 
3153: 
3154: \acknowledgments
3155: 
3156: One of the authors (M.T.) thanks
3157:     E.\ Orignac, M.\ Sugiura, 
3158:    Y.\ Suzumura, K.\ Yonemitsu, and H.\ Yoshioka 
3159:    for valuable discussions.
3160: The authors also thank  S.\ Qin for useful discussions.
3161: A.F.\ is grateful to Aspen Center for Physics for its hospitality,
3162:    where this paper was finished.
3163: This work was supported in part by a Grant-in-Aid for
3164:    Scientific Research on Priority Areas from the Ministry of
3165:    Education, Culture, Sports, Science and Technology, Japan
3166:    (Grant No.\ 12046238).
3167: 
3168: 
3169: \appendix 
3170: 
3171: 
3172: \section{Bosonization}\label{sec:bosonization}
3173: 
3174: In this section, we derive the phase Hamiltonian of the
3175:    1D extended Hubbard model by using
3176:    the Abelian bosonization method. \cite{Shankar}
3177: We include not only the marginal terms but the
3178:    leading irrelevant terms which play a crucial role
3179:    in the first-order SDW-CDW transition at strong coupling.
3180: 
3181: The Lagrangian for the free massless boson theory 
3182:    in a two-dimensional Euclidean
3183:    space is given by
3184: %================================================================
3185: \begin{eqnarray}
3186: L_\theta &=&
3187: \frac{1}{4\pi} \int dx 
3188: \left[
3189: v \left(\partial_x \theta \right)^2 + \frac{1}{v}
3190:    \left(\partial_\tau \theta \right)^2
3191: \right]
3192: ,
3193: \end{eqnarray}
3194: %================================================================
3195:    where $\theta$ is a bosonic field, $\tau$ is the imaginary time,
3196:    and $v$ is velocity.
3197: The variable canonically conjugate to $\theta$ is given by
3198: %================================================================
3199: \begin{eqnarray}
3200: \Pi &\equiv& \frac{\partial L}{\partial \dot{\theta}}
3201: =\frac{i}{2\pi v} \partial_\tau \theta
3202: ,
3203: \end{eqnarray}
3204: %================================================================
3205:    where $\dot{\theta}=\partial \theta/ \partial t$ and
3206:    $t$ is the real time ($\tau = it$).
3207: As usual this system is quantized by imposing
3208:    the commutation relation at equal times:
3209:   $[\theta(x), \Pi(x')] = i\delta(x-x')$.
3210: Thus the Hamiltonian for the free boson theory is given by
3211:  $H_\theta = i\int dx \, \Pi \, \partial_\tau \theta + L_\theta$, i.e.,
3212: %================================================================
3213: \begin{equation}
3214: H_\theta =
3215: \frac{v}{4\pi} \int dx 
3216: \left[
3217:   \left(2\pi \Pi\right)^2 +
3218:  \left(\frac{d\theta}{dx} \right)^2 
3219: \right]
3220: .
3221: \end{equation}
3222: %================================================================
3223: 
3224: Introducing two copies of this theory with fields
3225:    $\theta$ and $\phi$ and velocity $v=v_F$,
3226:    we arrive at $H_0$ [Eq.\ (\ref{eq:H0})], where
3227:    the fields $\theta$ and $\phi$ represent
3228:    the ``charge'' and ``spin'' degrees of freedom.
3229: The chiral bosonic fields $\theta_\pm(x,\tau)$ and
3230:    $\phi_\pm(x,\tau)$
3231:    are introduced in Eqs.\ (\ref{eq:chiral_theta}) and 
3232:    (\ref{eq:chiral_phi}), respectively, where
3233:    the right-moving (left-moving) fields 
3234:    are functions of 
3235:    $\tau-i(x/v_F)$ [$\tau+i(x/v_F)$].\cite{Shankar}  
3236: The phase field $\theta$ ($\phi$) and its dual phase field
3237:    $\tilde{\theta}$ ($\tilde{\phi}$) are written in terms of
3238:    the chiral fields as
3239: %================================================================
3240: \begin{eqnarray}
3241: &&
3242: \theta = \theta_+ + \theta_-
3243: ,
3244: \hspace*{0.2cm}
3245: \tilde{\theta} =\theta_+ - \theta_-
3246: , 
3247: \\
3248: &&
3249: \phi = \phi_+ + \phi_-
3250: ,
3251: \hspace*{0.2cm}
3252: \tilde{\phi} = \phi_+ - \phi_-
3253: .
3254: \end{eqnarray}
3255: %================================================================
3256: They satisfy the following commutation relations:
3257: %================================================================
3258: \begin{eqnarray}
3259: [ \theta(x),\tilde{\theta}(x') ] 
3260: =
3261:  [\phi (x), \tilde{\phi} (x')] 
3262: = -i \, 2\pi \, \Theta(-x+x')
3263: , 
3264: \end{eqnarray}
3265: %================================================================
3266: where $\Theta(x)$ is the Heaviside step function.
3267: 
3268: The electron field operators $\psi_{p,\sigma}(x)$ are given in
3269:    Eq.\ (\ref{eq:field_op}) in terms of
3270:    a new set of chiral bosonic fields
3271:    $\varphi_{p,\sigma}$ introduced in Eq.\ (\ref{eq:varphi}).
3272: In this bosonization scheme $\psi_{+,\sigma}$ and $\psi_{-,\sigma}$
3273:    anticommute, and we only need to introduce the Klein factor
3274:    $\kappa_\sigma$ to ensure the anticommutation relation between
3275:    fields with different spins;
3276:    cf.\ the so-called \textit{constructive} bosonization
3277:    method.\cite{vonDelft}
3278: From Eqs.\ (\ref{eq:commutation_varphi}) and (\ref{eq:field_op})
3279:    the electron-density operator becomes
3280: %================================================================
3281: \begin{equation}
3282: \rho_{p,\sigma}(x)
3283: \equiv
3284: {} : \psi_{p,\sigma}^\dagger \, \psi_{p,\sigma}\!: {}
3285: =
3286: \frac{1}{2\pi} \, \frac{d}{dx} \, \varphi_{p,\sigma}(x) 
3287: .
3288: \label{eq:density}
3289: \end{equation}
3290: %================================================================
3291: As is well known, the Hamiltonian density of free bosons (\ref{eq:H0}),
3292:   i.e., 
3293: %================================================================
3294: \begin{eqnarray}
3295: \mathcal{H}_{0}
3296: =
3297: \frac{v_F}{4\pi} \sum_{p=\pm} \sum_{\sigma}
3298:     \left( \frac{d\varphi_{p,\sigma}}{dx} \right)^2
3299: =
3300: \pi v_F \sum_{p,\sigma} 
3301: \rho^2_{p,\sigma}(x)
3302: ,
3303: \end{eqnarray}
3304: %================================================================
3305: is equivalent to the Hamiltonian density of free fermions with linear
3306: energy dispersion, Eq.\ (\ref{eq:H0_linear}).
3307: This can be shown, for example, by using the OPE method.\cite{Affleck}
3308: 
3309: 
3310: Next we bosonize the interaction term $H_\mathrm{int}$.
3311: Without the nearest-neighbor repulsion $V$, 
3312:    this can be easily done as\cite{Solyom,Emery}
3313: %================================================================
3314: \begin{eqnarray}
3315: \mathcal{H}_{\rm int}^{V=0}
3316: &=&
3317: \frac{g_{4\parallel}+g_{4\perp}}{4\pi^2} 
3318: \left[ \left(\partial_x \theta_+ \right)^2 +
3319:        \left(\partial_x \theta_- \right)^2  \right]
3320: \nonumber \\ && {}
3321: +\frac{g_{4\parallel}-g_{4\perp}}{4\pi^2} 
3322: \left[ \left(\partial_x \phi_+ \right)^2 +
3323:        \left(\partial_x \phi_- \right)^2  \right]
3324: \nonumber \\ && {}
3325: +\frac{g_{2\parallel}+g_{2\perp}-g_{1\parallel}}{2\pi^2} 
3326:    \bigl(\partial_x \theta_+ \bigr)
3327:        \bigl(\partial_x \theta_- \bigr) 
3328: \nonumber \\ && {}
3329: +\frac{g_{2\parallel}-g_{2\perp}-g_{1\parallel}}{2\pi^2} 
3330:    \bigl(\partial_x \phi_+ \bigr)
3331:        \bigl(\partial_x \phi_- \bigr) 
3332: \nonumber \\ && {}
3333: -\frac{g_{3\perp}}{2(\pi a)^2} 
3334:    \cos 2\theta
3335: +\frac{g_{1\perp}}{2(\pi a)^2} 
3336:     \cos 2\phi
3337: ,
3338: \label{eq:H_lowest}
3339: \end{eqnarray}
3340: %================================================================
3341:    where $g$'s are given in and below Eq.\ (\ref{eq:g}).
3342: In the presence of $V$, 
3343:    the matrix element of the umklapp process with parallel spins
3344:    $H_{g_{3\parallel}}$ [the $g_{3\parallel}$ process in 
3345:    Eq.\ (\ref{eq:HI_fourier})] has a finite amplitude at lowest 
3346:    order in $g$-ology.
3347: This term can be bosonized as
3348: %================================================================
3349: \begin{eqnarray}
3350: \mathcal{H}_{g_{3\parallel}}
3351: &=& {}
3352: -\frac{g_{3\parallel}}{2(\pi a)^2} \cos 2\theta \, \cos 2\phi
3353: ,
3354: \label{eq:g3para}
3355: \end{eqnarray}
3356: %================================================================
3357:    where $g_{3\parallel}=-2Va$ in the lowest order in $V$.
3358: This term, which couples the charge and spin degrees of freedom,
3359:    is often neglected since it is an irrelevant perturbation
3360:    with scaling dimension $4$,
3361:    consisting of $\mathrm{dim}[\cos2\theta]=2$ plus 
3362:    $\mathrm{dim}[\cos2\phi]=2$.
3363: Cannon and Fradkin were the first to suggest that this term should play
3364:    an important role in the first-order SDW-CDW transition in the
3365:    half-filled EHM. \cite{Cannon}
3366: Voit then derived RG equations including this term.
3367: However he did not include all the operators with scaling dimension 4
3368:    and failed to keep the spin-rotational SU(2) symmetry.\cite{Voit}
3369: We have to be careful in dealing with the $V$ interaction to
3370:    include the important terms with scaling dimension up to 4.
3371: To this end, we focus on the $V$ interaction and bosonize each
3372:    scattering process separately.
3373: 
3374: First, the $g_{1\parallel}$ term\cite{Solyom} representing the backward
3375:    scattering with parallel spins
3376:    is bosonized by using Eq.\ (\ref{eq:field_op}) as
3377: %================================================================
3378: \begin{eqnarray}
3379: && \hspace*{-.7cm}
3380: Va \sum_{p,\sigma} 
3381:   \psi^\dagger_{p,\sigma}(x) \, \psi_{-p,\sigma}(x) \,
3382:   \psi^\dagger_{-p,\sigma} (x+a) \, \psi_{p,\sigma}(x+a)
3383: \nonumber \\
3384: &=& -\frac{Va}{(2\pi a)^2} \sum_{p,s=\pm} 
3385:   e^{ip[\theta(x+a)-\theta(x)]+ips [\phi(x+a)-\phi(x)]}
3386: \nonumber \\
3387: &=& 
3388: \frac{Va}{2\pi^2} 
3389: \left[ \sum_p \left(\partial_x \theta_p \right)^2
3390:       + 2 \left(\partial_x \theta_+ \right)
3391:           \left(\partial_x \theta_- \right) 
3392: \right]
3393: \nonumber \\ && {}
3394: +\frac{Va}{2\pi^2} 
3395: \left[  \sum_p \left(\partial_x \phi_p \right)^2 
3396:       + 2 \left(\partial_x \phi_+ \right)
3397:           \left(\partial_x \phi_- \right) 
3398: \right]
3399: \nonumber \\ && {}
3400: -\frac{Va}{4\pi^2}  \, a^2 
3401: \left(\partial_x \theta\right)^2 \left(\partial_x \phi\right)^2
3402: + \cdots ,
3403: \label{eq:Hg1paraV}
3404: \end{eqnarray}
3405: %================================================================
3406:    where we have expanded the exponent in the second line up to the
3407:    order $a^4$
3408:    for the $\theta$ sector and the $\phi$ sector, separately.
3409: Since we are interested in operators that couple $\theta$ and $\phi$
3410:    as in Eq.\ (\ref{eq:g3para}),
3411:    we have discarded dimension-4 terms such as $a^4(\partial_x \theta)^4$
3412:    and $a^4(\partial_x \phi)^4$ that involve only one sector.
3413: Such terms as $(\partial_x\theta_+)(\partial_x\theta_-)$
3414:    and $(\partial_x\phi_+)(\partial_x\phi_-)$ are
3415:    already retained in Eq.\ (\ref{eq:H_lowest}), 
3416:    while the last term proportional to 
3417:    $(\partial_x\theta)^2 (\partial_x\phi)^2$ is a new term 
3418:    with scaling dimension 
3419:    $2+2$, which was missed in Ref.\ \onlinecite{Voit}.
3420: We note that the Fermi velocity is renormalized by the $g_{1\parallel}$ term
3421:    due to the presence of
3422:    $\sum_p(\partial_x \theta_p)^2$ and  
3423:    $\sum_p(\partial_x \phi_p)^2$.
3424: This is in contrast with the conventional treatment where
3425:    the velocity renormalization comes only from
3426:    the forward scattering term $g_4$.\cite{Solyom}
3427: 
3428: 
3429: In a similar way, the interaction terms of backward and umklapp 
3430:    scattering with opposite spins 
3431:    (so-called $g_{1\perp}$ and $g_{3\perp}$ terms,\cite{Solyom} 
3432:     respectively)
3433:    are bosonized as 
3434: %================================================================
3435: \begin{eqnarray}
3436: && \hspace*{-1.cm}
3437: Va \sum_{p,\sigma}
3438:   \psi^\dagger_{p,\sigma}(x) \, 
3439:   \psi_{-p,\sigma}(x) \,
3440:   \psi^\dagger_{-p,\overline{\sigma}} (x+a) \, 
3441:   \psi_{p,\overline{\sigma}}(x+a)
3442: \nonumber \\
3443: &=& -\frac{Va}{(2\pi a)^2} \sum_{p,s=\pm} 
3444:   e^{ip[\theta(x+a)-\theta(x)]-ips [\phi(x+a)+\phi(x)]}
3445: \nonumber \\
3446: &=& 
3447: -\frac{2Va}{2(\pi a)^2} \cos 2\phi
3448: + \frac{2Va}{2\pi^2} 
3449:    \left(\partial_x \theta_+ \right)
3450:    \left(\partial_x \theta_- \right) \, 
3451:    \cos 2\phi
3452: \nonumber \\ && {}
3453: + \frac{2Va}{4\pi^2} 
3454:   \left[ \sum_p \left(\partial_x \theta_p \right)^2 \right]
3455:    \cos 2\phi 
3456: + \cdots,
3457: \label{eq:Hg1perpV}
3458: \\
3459: && \hspace*{-1.cm}
3460: Va \sum_{p,\sigma} 
3461:   \psi^\dagger_{p,\sigma}(x) \, 
3462:   \psi_{-p,\sigma}(x) \,
3463:   \psi^\dagger_{p,\overline{\sigma}} (x+a) \, 
3464:   \psi_{-p,\overline{\sigma}}(x+a)
3465: \nonumber \\
3466: &=& \frac{Va}{(2\pi a)^2} \sum_{p,s=\pm} 
3467:   e^{-ip[\theta(x+a)+\theta(x)]+ips [\phi(x+a)-\phi(x)]}
3468: \nonumber \\
3469: &=& 
3470: +\frac{2Va}{2(\pi a)^2} \cos 2\theta
3471: - \frac{2Va}{2\pi^2}
3472:    \left(\partial_x \phi_+ \right)
3473:    \left(\partial_x \phi_- \right) \,
3474:    \cos 2\theta
3475: \nonumber \\ && {}
3476: - \frac{2Va}{4\pi^2} 
3477:   \left[ \sum_p \left(\partial_x \phi_p \right)^2 \right]
3478:    \cos 2\theta 
3479: + \cdots ,
3480: \label{eq:Hg3perpV}
3481: \end{eqnarray}
3482: %================================================================
3483:    where $\overline{\sigma}=\downarrow(\uparrow)$ for 
3484:    $\sigma=\uparrow(\downarrow)$.
3485: The potential $\cos 2\phi$ in Eq.\ (\ref{eq:Hg1perpV}) and 
3486:    the potential $\cos 2\theta$ in Eq.\ (\ref{eq:Hg3perpV}) are
3487:    already retained in Eq.\ (\ref{eq:H_lowest}), 
3488:    while the other terms are new and have the scaling dimension $2+2$.
3489: 
3490: The forward-scattering terms
3491:    ($g_{2\parallel}$, $g_{2\perp}$, $g_{4\parallel}$, and $g_{4\perp}$)
3492:    do not generate operators of dimension $2+2$.
3493: 
3494: 
3495: Hence the total Hamiltonian is given by
3496: %================================================================
3497: \begin{eqnarray}
3498: \mathcal{H} &\!=\!&
3499: \frac{1}{2\pi} 
3500:      \sum_p 
3501:    \left[ v_{\rho} \left(\partial_x \theta_p \right)^2
3502:         + v_{\sigma} \left(\partial_x \phi_p \right)^2 \right]
3503: \nonumber \\ && {}
3504: + \frac{g_\rho}{2\pi^2} 
3505:      \left(\partial_x \theta_+ \right)
3506:      \left(\partial_x \theta_- \right)
3507: - \frac{g_\sigma}{2\pi^2} 
3508:      \left(\partial_x \phi_+ \right) 
3509:      \left(\partial_x \phi_- \right)
3510: \nonumber \\ && {}
3511: -\frac{g_{3\perp}}{2\pi^2 a^2} 
3512:     \, \cos 2 \theta 
3513: +\frac{g_{1\perp}}{2\pi^2 a^2}  
3514:   \, \cos 2 \phi
3515: \nonumber \\ && {}
3516: +\frac{Va}{\pi^2 a^2} 
3517:    \, \cos 2\theta  \,  \cos 2\phi
3518: \nonumber \\ && {}
3519: +\frac{Va}{2 \pi^2}
3520:      \left[ \sum_{p}\left(\partial_x \theta_p \right)^2 
3521:           + 2  \left(\partial_x \theta_+ \right) 
3522:                \left(\partial_x \theta_- \right)  \right]
3523:      \cos 2\phi
3524: \nonumber \\ && {}
3525: -\frac{Va}{2 \pi^2} 
3526:      \left[ \sum_{p}\left(\partial_x \phi_p \right)^2 
3527:           + 2 \left(\partial_x \phi_+\right)
3528:               \left(\partial_x \phi_- \right) \right]
3529:      \cos 2\theta 
3530: \nonumber \\ && {}
3531: -\frac{Va^3}{4\pi^2}  
3532:      \left[ \sum_{p}\left(\partial_x \theta_p \right)^2 
3533:           + 2  \left(\partial_x \theta_+ \right) 
3534:                \left(\partial_x \theta_- \right)  \right]
3535: \nonumber \\ && {} \quad\quad \times
3536:      \left[ \sum_{p}\left(\partial_x \phi_p \right)^2 
3537:           + 2 \left(\partial_x \phi_+\right)
3538:               \left(\partial_x \phi_- \right) \right]
3539: .
3540: \label{eq:Hamiltonian_all}
3541: \end{eqnarray}
3542: %================================================================
3543: The renormalized velocities are given by
3544:   $v_\rho = 2ta + (U+6V) a/(2\pi)$ and
3545:   $v_\sigma = 2ta - (U-2V) a/(2\pi)$.
3546: The coupling constants $g_{1\perp}$ and $g_{3\perp}$ are defined in
3547:    Eq.\ (\ref{eq:g}), and
3548:    $g_\rho (\equiv g_{2\parallel}+g_{2\perp}-g_{1\parallel})$ and 
3549:    $g_\sigma (\equiv -g_{2\parallel}+g_{2\perp}+g_{1\parallel})$
3550:    are given by
3551: %================================================================
3552: \begin{subequations}
3553: \begin{eqnarray}
3554: g_\rho 
3555: &=& 
3556: (U+6V)a 
3557: + \frac{C_1}{4\pi t}(U-2V)^2a
3558: +\frac{C_2}{\pi t} V^2 a
3559: ,
3560: \\
3561: g_\sigma
3562: &=& 
3563: (U-2V)a 
3564: - \frac{C_1}{4\pi t}(U-2V)^2a
3565: -\frac{C_2}{\pi t}V^2a
3566: . \quad\quad\quad
3567: \end{eqnarray} 
3568: \end{subequations}
3569: %================================================================
3570: For the discussion of the SDW-CDW transition in the 1D EHM,
3571:    it is sufficient to have the coupling constants of
3572:    dimension 4 in lowest order in $V$.
3573: We note that due to the SU(2) spin-rotation symmetry of the theory,
3574:    the coupling constants for spin degrees of freedom must satisfy
3575:    $g_{\sigma}=g_{1\perp}$, in any order of $U$ and $V$. 
3576: To proceed further, we neglect the terms that involve
3577:    $V\sum_p (\partial_x \theta_p )^2$ or
3578:    $V\sum_p (\partial_x \phi_p )^2$ in Eq.\ (\ref{eq:Hamiltonian_all}).
3579: These terms can lead to renormalization of the velocity through the
3580:    RG transformation (see Appendix \ref{sec:rg}).
3581: This effect can be ignored if we are interested in qualitative feature
3582:    of the ground-state phase diagram of the model.
3583: The final form of the bosonized Hamiltonian is thus given by
3584:     Eq.\ (\ref{eq:Hamiltonian}).
3585: 
3586: 
3587: 
3588: \section{Derivation of renormalization-group equations}
3589: \label{sec:rg}
3590: 
3591: In this section,
3592:    we derive one-loop RG equations for the coupling constants including
3593:    those operators with higher scaling dimension.
3594: Our derivation is based on the operator product expansion (OPE) method.
3595: The interaction part of the action $S_{\rm I}$ in the presence
3596:    of the staggered site potential $G_\Delta$
3597:    is given by 
3598: %================================================================
3599: \begin{eqnarray}
3600: S_{\rm I} &=& {}
3601: \frac{G_\rho}{\pi} \int d^2 r 
3602:    \left(\partial_z \theta \right)
3603:    \left(\partial_{\bar{z}} \theta \right) 
3604: - \frac{G_\sigma}{\pi} \int d^2 r
3605:    \left(\partial_z \phi \right) 
3606:    \left(\partial_{\bar z} \phi \right) 
3607: \nonumber \\ && {}
3608: -\frac{G_\Delta}{\pi} \int \frac{d^2r}{a^2} \, 
3609:   :\! \sin \theta \!: \, :\! \cos \phi \!:
3610: \nonumber \\ && {}
3611: -\frac{G_c}{\pi} \int \frac{d^2r}{a^2} \, :\! \cos 2\theta \!:
3612: +\frac{G_s}{\pi} \int \frac{d^2r}{a^2} \, :\! \cos 2\phi \!:
3613: \nonumber \\ && {}
3614: -\frac{G_{cs}}{\pi} \int \frac{d^2r}{a^2} \, 
3615:   :\! \cos 2\theta  \!: \, :\! \cos 2\phi \!:
3616: \nonumber \\ && {}
3617: -\frac{G_{\rho s}}{\pi} \int d^2r \,
3618:    \left(\partial_z \theta \right) 
3619:    \left(\partial_{\bar{z}} \theta \right) \,
3620:    :\! \cos 2\phi \!:
3621: \nonumber \\ && {}
3622: +\frac{G_{c \sigma}}{\pi} \int d^2r \,
3623:    \left(\partial_z \phi\right) 
3624:    \left(\partial_{\bar z} \phi \right) \,
3625:    :\! \cos 2\theta \!:
3626: \nonumber \\ && {}
3627: +\frac{G_{\rho \sigma}}{\pi} \int d^2r \, a^2 \,
3628:    \left(\partial_z \theta \right) 
3629:    \left(\partial_{\bar{z}} \theta \right) 
3630:    \left(\partial_z \phi\right) 
3631:    \left(\partial_{\bar z} \phi \right) ,
3632: \end{eqnarray}
3633: %================================================================
3634:    where $z=v_F \tau+ix$, $\overline{z}=v_F \tau-ix$,
3635:    $d^2r = v_F dx \, d\tau$, and $G_i=g_i/2\pi v_F$.
3636: In this section,
3637:    the operators are explicitly normal ordered.
3638: 
3639: 
3640: In order to derive the RG equations, 
3641:    we use the following  OPE's: 
3642: %================================================================
3643: \begin{subequations}
3644: \begin{eqnarray}
3645: &&
3646: J_\rho (z) \, J_\rho (w)
3647: =
3648: \frac{1}{(z-w)^2} + \cdots ,
3649: \\
3650: &&
3651: \bar{J}_\rho (\bar{z}) \, \bar{J}_\rho (\bar{w})
3652: =
3653: \frac{1}{(\bar{z}-\bar{w})^2} + \cdots ,
3654: \\
3655: &&
3656: J_\rho (z)  :\! e^{i\alpha \theta(w,\bar{w})}\!:
3657: =
3658: \frac{\alpha}{2(z-w)} 
3659: :\! e^{i\alpha \theta(w,\bar{w})}\!:
3660: + \cdots
3661: , \,\,
3662: \\
3663: &&
3664: J_\rho (\bar{z})  :\! e^{i\alpha \theta(w,\bar{w})}\!:
3665: =
3666: \frac{-\alpha}{2(\bar{z}-\bar{w})} 
3667: :\! e^{i\alpha \theta(w,\bar{w})}\!:
3668: + \cdots
3669: , \,\,
3670: \\
3671: &&
3672: :\! e^{i\alpha\theta(z,\bar{z})}\!: \, 
3673:   :\! e^{-i\alpha\theta(0,0)}\!: \,
3674: \nonumber \\ && \hspace*{1cm}{}
3675: = 
3676: \frac{1}{|z|^{\alpha^2}}
3677:   + \frac{\alpha}{|z|^{\alpha^2}} 
3678:   \left(z J_\rho - \bar{z} \bar{J}_\rho \right)
3679: +\frac{2i}{|z|^2} 
3680:    \left(\partial_z \partial_{\bar{z}} \theta\right) 
3681: \nonumber \\ && \hspace*{1.5cm} {}
3682: +\frac{i\alpha}{2|z|^{\alpha^2}}
3683:   \left[
3684:      z^2  \left(\partial_z^2 \theta\right) 
3685:      + \bar{z}^2  \left(\partial_{\bar{z}}^2 \theta\right) 
3686:   \right]
3687: \nonumber \\ && \hspace*{1.5cm} {}
3688: +\frac{\alpha^2}{2|z|^{\alpha^2}}
3689:   \left[
3690:        z^2   : \! J_\rho ^2 \! :
3691:      + \bar{z}^2  :\! \bar{J}_\rho^2 \! :
3692:   \right]
3693: \nonumber \\ && \hspace*{1.5cm} {}
3694: -\frac{\alpha^2}{|z|^{\alpha^2-2}}
3695:      J_\rho \, \bar{J}_\rho
3696: + \cdots ,
3697: \\
3698: &&
3699: :\! e^{i\alpha\theta(z,\bar{z})}\!: \, 
3700:   :\! e^{i\beta\theta(0,0)}\!: \,
3701: = 
3702: \frac{1}{|z|^{-\alpha\beta}} :\! e^{i(\alpha+\beta) \theta}\!:
3703:  + \cdots ,
3704: \end{eqnarray}
3705: \label{eq:ope}%
3706: \end{subequations}
3707: %================================================================
3708:    where we have introduced U(1) currents:
3709:   $J_\rho(z)\equiv i \partial_z \theta(z,\bar{z})$,
3710:   $\bar{J}_\rho(\bar{z}) \equiv
3711:      - i \partial_{\bar{z}} \theta(z,\bar{z})$,
3712:   $J_\sigma(z)\equiv i \partial_z \phi(z,\bar{z})$, and
3713:   $\bar{J}_\sigma(\bar{z}) \equiv
3714:      - i \partial_{\bar{z}} \phi(z,\bar{z})$.
3715: The parameters $\alpha$ and $\beta$ ($\alpha + \beta \neq 0$) 
3716:    in the vertex operator are 
3717:    numerical constants which determine the scaling dimension.
3718: In deriving the above OPE's, we have used the Wick theorem and
3719:    the correlators: 
3720:    $\langle \theta_+(\bar{z}) \, \theta_+(\bar{\omega})\rangle
3721:       = -\frac{1}{2} \ln(\bar{z}-\bar{\omega})$,
3722:    $\langle \theta_-(z) \, \theta_-(\omega)\rangle
3723:       = -\frac{1}{2} \ln(z-\omega)$, and
3724:    $\langle \theta(z,\bar{z}) \, \theta(\omega,\bar{\omega})\rangle
3725:       = -\ln|z-\omega|$.
3726: From Eq.\ (\ref{eq:ope}), one finds
3727: %================================================================
3728: \begin{subequations}
3729: \begin{eqnarray}
3730: \quad
3731: && \hspace*{-1.5cm}
3732: \left[J_\rho(z) \, \bar{J}_\rho(\bar{z})\right] 
3733: \,
3734: \left[J_\rho(0) \, \bar{J}_\rho(0)\right] 
3735: \nonumber \\
3736: &=&
3737: \frac{1}{|z|^4}
3738: + \frac{1}{|z|^4}
3739: \left(z^2 : \!  J_\rho^2 \! : +  \bar{z}^2  :\! \bar{J}_\rho^2 \! :
3740:   \right)
3741: +\cdots, 
3742: \label{eq:ope1}
3743: \\
3744: && \hspace*{-1.5cm}
3745: \left[J_\rho(z) \, \bar{J}_\rho(\bar{z})\right] 
3746:  \, :\! \cos \alpha \theta(0,0) \!: 
3747: \nonumber \\
3748: &=&
3749:   -\frac{\alpha^2}{4 |z|^2} \, :\! \cos \alpha \theta \! : 
3750: + \cdots,
3751: \\
3752: && \hspace*{-1.5cm}
3753: \left[J_\rho(z) \, \bar{J}_\rho(\bar{z})\right] 
3754:  :\! \sin \alpha \theta(0,0) \!: 
3755: \nonumber \\
3756: &=&
3757:   -\frac{\alpha^2}{4 |z|^2} \, :\! \sin \alpha \theta \! : 
3758: + \cdots,
3759: \\
3760: && \hspace*{-1.5cm}
3761: :\! \cos \alpha\theta(z,\bar{z})\!: \, :\!\cos \alpha\theta(0,0)\! :
3762: \nonumber \\
3763: &=&
3764: \frac{1}{2|z|^{\alpha^2}}
3765: + \frac{\alpha^2}{|z|^{\alpha^2}}
3766: \left(z^2 : \! J_\rho^2 \! : +  \bar{z}^2  :\! \bar{J}_\rho^2 \! :
3767:   \right)
3768: \nonumber \\ && {}
3769: -\frac{\alpha^2}{2|z|^{\alpha^2-2}} \,
3770:     J_\rho \, \bar{J}_\rho
3771: \nonumber \\ && {}
3772: +\frac{1}{2} |z|^{\alpha^2} :\!\cos 2\alpha\theta\! :
3773: + \cdots , \quad
3774: \\
3775: && \hspace*{-1.5cm}
3776: :\! \cos \alpha\theta(z,\bar{z})\!: \, :\!\cos \beta\theta(0,0)\! :
3777: \nonumber \\
3778: &=&
3779: \frac{1}{2|z|^{\alpha\beta}} \, :\! \cos[(\alpha-\beta)\theta] \! :
3780: \nonumber \\ && {}
3781: + \frac{1}{2 |z|^{-\alpha\beta}} :\! \cos[(\alpha+\beta)\theta] \! :
3782: +\cdots .
3783: \end{eqnarray}
3784: \end{subequations}
3785: %================================================================
3786: Exchanging $\theta\to \phi$ and $\rho \to \sigma$ yields
3787:   the OPE's for spin phase fields.
3788:    
3789: 
3790: Expanding the action in powers of coupling constants and
3791:    integrating out short-distance parts, we obtain the scaling
3792:    equations,
3793: %================================================================
3794: \begin{eqnarray}
3795: \frac{d}{dl} G_\Delta
3796:  \!\!  &=& \!\! {}
3797: G_\Delta \left(1+\frac{1}{2} \, G_\rho - \frac{1}{2} \, G_\sigma
3798:     - G_c  - G_s
3799: \right. \nonumber \\ && {} \left.
3800:     - \frac{1}{2} \, G_{cs}
3801:     - \frac{1}{4} \, G_{\rho s}
3802:     - \frac{1}{4} \, G_{c\sigma}
3803:     - \frac{1}{8} \, G_{\rho\sigma}
3804: \right)
3805: , \label{eq:RG-a}
3806: \\
3807: \frac{d}{dl} G_\rho 
3808:  \!\!  &=& \!\! {}
3809:    + \frac{1}{4} \, G_\Delta^2 + 2 \, G_c^2 + G_{cs}^2 +  G_s \, G_{\rho s}
3810: ,
3811: \label{eq:RG-b}
3812: \\
3813: \frac{d}{dl} G_\sigma 
3814:  \!\!  &=& \!\! {}
3815:    - \frac{1}{4} \, G_\Delta^2  - 2 \, G_s^2 - G_{cs}^2 - G_c \, G_{c\sigma}
3816: ,
3817: \\
3818: \frac{d}{dl} G_c 
3819:  \!\!  &=& \!\! {}
3820:     -\frac{1}{4} \, G_\Delta^2
3821:     + 2 \, G_\rho\, G_c
3822:     - \left( G_s + G_{\rho s} \right) G_{cs}
3823: ,
3824: \\
3825: \frac{d}{dl} G_s 
3826:  \!\!  &=& \!\! {}
3827:    - \frac{1}{4} \, G_\Delta^2 - 2 \, G_\sigma \, G_s
3828:    -  \left( G_c + G_{c\sigma} \right)  G_{cs}
3829: ,
3830: \\
3831: \frac{d}{dl} G_{cs} 
3832:  \!\!  &=& \!\! {}
3833:      -\frac{1}{4} \, G_\Delta^2 
3834:     - 2 \left(1 - G_\rho 
3835:          +  G_\sigma  + G_{\rho \sigma}\right)    G_{cs}
3836: \nonumber \\ && {}
3837:       - 2 \, (G_c + G_{c\sigma}) (G_s + G_{\rho s})
3838: ,
3839: \\
3840: \frac{d}{dl} G_{\rho s} 
3841:  \!\!  &=& \!\! {}
3842:     - \frac{1}{4} \, G_\Delta^2
3843:     -2 \left(1 + G_\sigma \right) G_{\rho s}
3844:     + 2 \, G_{\rho} \, G_s
3845: \nonumber \\ && {}
3846:     - 4 \, ( G_c +  G_{c\sigma} ) \, G_{cs}
3847:     - 2 \, G_s \, G_{\rho \sigma}
3848: ,
3849: \\
3850: \frac{d}{dl} G_{c\sigma} 
3851:  \!\!  &=& \!\! {}
3852:     - \frac{1}{4} \, G_\Delta^2
3853:     - 2 \left(1 - G_\rho \right) G_{c \sigma}
3854:     - 2 \, G_{\sigma} \, G_c
3855: \nonumber \\ && {}
3856:     - 4 \, (G_s + G_{\rho s} ) \, G_{cs}
3857:     - 2 \, G_c \, G_{\rho \sigma}
3858: ,
3859: \\
3860: \frac{d}{dl} G_{\rho \sigma} 
3861:  \!\!  &=& \!\! {}
3862:     - \frac{1}{4} \, G_\Delta^2 
3863:     - 2 \, G_{\rho \sigma}
3864:     + 2 \, G_\rho\, G_\sigma
3865:     - 4 \, G_{cs}^2
3866: \nonumber \\ && {}
3867:     - 4 \, G_c \, G_{c\sigma} -4 \, G_s \, G_{\rho s} 
3868: .
3869: \label{eq:RG-e}
3870: \end{eqnarray}
3871: %================================================================
3872: Here we note that the number of the RG equations
3873:    can be reduced due to the spin-rotational SU(2) symmetry.
3874: To show this point more transparently,
3875:    we introduce $X(l)$, $Y(l)$, and $Z(l)$ by
3876:    $X(l)=G_\sigma(l)-G_s(l)$,
3877:    $Y(l)=G_{cs}(l)-G_{c\sigma}(l)$, and
3878:    $Z(l)=G_{\rho s}(l)-G_{\rho \sigma}(l)$.
3879: Their RG equations are obtained
3880:    from Eqs.\ (\ref{eq:RG-b})--(\ref{eq:RG-e}) as
3881: %================================================================
3882: \begin{subequations}
3883: \begin{eqnarray}
3884: \frac{d}{dl} X
3885:   &=& {}
3886:     2 \, G_s \, X  + (G_c-G_{cs}) \, Y
3887: ,
3888: \\
3889: \frac{d}{dl} Y
3890:   &=& {}
3891:    2 \, ( - 1 + G_\rho + G_s + G_{\rho s}) \, Y
3892: \nonumber \\ && {} 
3893:    + 2 \, (G_c - G_{cs})  (X-Z)
3894: , \\
3895: \frac{d}{dl} Z
3896:   &=& {}
3897:      -2 \, (1 - G_s) \, Z
3898:      -2 \, (G_\rho + G_{\rho s}) \, X
3899: \nonumber \\ && {} 
3900:      -4 \, (G_c - G_{cs}) \, Y
3901: .
3902: \end{eqnarray}
3903: \end{subequations}
3904: %================================================================
3905: One immediately finds that,
3906:   if $X(0)=Y(0)=Z(0)=0$, 
3907:   they vanish for all $l$,
3908:   i.e., $X(l)=Y(l)=Z(l)=0$.
3909: This implies that 
3910:    $G_\sigma(l)=G_s(l)$,
3911:    $G_{cs}(l)=G_{c\sigma}(l)$, and
3912:    $G_{\rho s}(l)=G_{\rho \sigma}(l)$, 
3913:    which are nothing but the constraints on the coupling constants
3914:    due to the spin-rotational SU(2) symmetry.
3915: In this case, we can set
3916:    $G_\sigma(l)=G_s(l)$, $G_{c\sigma}(l)=G_{cs}(l)$, and
3917:    $G_{\rho \sigma}(l)=G_{\rho s}(l)$
3918:    in the RG equations (\ref{eq:RG-a})--(\ref{eq:RG-e}).
3919: Then the RG equations are given by 
3920:    Eqs.\ (\ref{eq:RGf-b})--(\ref{eq:RGf-e}).
3921: The RG equations for the 1D EHM without the staggered site potential
3922:    are obtained by setting $G_\Delta(l)=0$,
3923:    Eqs.\ (\ref{eq:Grho})--(\ref{eq:Grhos}).
3924: 
3925: The RG equations can also be obtained in the presence of the
3926:    bond dimerization in a similar way.
3927: 
3928: 
3929: 
3930: 
3931: \begin{thebibliography}{}
3932: %--------------------------
3933: \bibitem{Bray}
3934: For a review, see
3935: J.W.\ Bray, L.V.\ Interrante, I.S.\ Jacobs, and J.C.\ Bonner,
3936: in 
3937: \textit{Extended Linear Chain Compounds},
3938: edited by J.S. Miller 
3939: (Plenum, New York, 1983), Vol.\ 3, p.\ 353. 
3940: %--------------------------
3941: \bibitem{Ishiguro}
3942: T.\ Ishiguro, K.\ Yamaji, and G.\ Saito, 
3943: \textit{Organic Superconductors}
3944:   (Springer-Verlag, Berlin, 1998).
3945: %--------------------------
3946: \bibitem{Kagoshima}
3947: S.\ Kagoshima, R.\ Kato, H.\ Fukuyama, H.\ Seo, and H.\ Kino, 
3948: in
3949: \textit{Advances in Synthetic Metals, Twenty years of Progress in
3950:    Science and Technology},
3951:   edited by  P.\ Bernier, S.\ Lefrant, and G.\ Bidan
3952:   (Elsevier, New York, 1999), p.\ 262.
3953: %--------------------------
3954: \bibitem{Hase}
3955: M.\ Hase, I.\ Terasaki, and K.\ Uchinokura,
3956: Phys.\ Rev.\ Lett.\ \textbf{70}, 3651 (1993).
3957: %--------------------------
3958: \bibitem{Sachdev}
3959: S.\ Sachdev, Physica A \textbf{313}, 252 (2002).
3960: %--------------------------
3961: \bibitem{Majumdar_Ghosh}
3962: C.K.\ Majumdar and D.K.\ Ghosh,
3963: J.\ Math.\ Phys.\ \textbf{10}, 1388 (1969); \textbf{10}, 1399 (1969);
3964: C.K.\ Majumdar, J.\ Phys. C \textbf{3}, 915 (1970);
3965: F.D.M.\ Haldane, Phys.\ Rev.\ B \textbf{25}, 4925 (1982);
3966: K.\ Okamoto and K.\ Nomura, Phys.\ Lett.\ A \textbf{169}, 433 (1992).
3967: %--------------------------
3968: \bibitem{Ukrainskii}
3969: I.I.\ Ukrainski\u{\i},
3970: Zh.\ Eksp.\ Teor.\ Fiz.\ \textbf{76}, 760 (1979)
3971: [Sov.\ Phys.\ JETP \textbf{49}, 381 (1979)].
3972: %--------------------------
3973: \bibitem{Kivelson}
3974: S.\ Kivelson and D.E.\ Heim, Phys.\ Rev.\ B \textbf{26}, 4278 (1982).
3975: %--------------------------
3976: \bibitem{Mazumdar}
3977: S.\ Mazumdar and S.N.\ Dixit, Phys.\ Rev.\ Lett.\ \textbf{51}, 292 (1983);
3978: S.N.\ Dixit and S.\ Mazumdar, Phys.\ Rev.\ B \textbf{29}, 1824 (1984).
3979: %--------------------------
3980: \bibitem{Hirsch83}
3981: J.E.\ Hirsch, Phys.\ Rev.\ Lett.\ \textbf{51}, 296 (1983).
3982: %--------------------------
3983: \bibitem{Hara}
3984: J.\ Hara, T.\ Nakano, and H.\ Fukuyama,
3985: J.\ Phys.\ Soc.\ Jpn.\ \textbf{51}, 341 (1982);
3986: J.\ Hara and H.\ Fukuyama,
3987: \textit{ibid.} \textbf{52}, 2128 (1983).
3988: %--------------------------
3989: \bibitem{Sugiura2002}
3990: M.\ Sugiura and Y.\ Suzumura,
3991: J.\ Phys.\ Soc.\ Jpn.\ \textbf{71}, 697 (2002).
3992: %--------------------------
3993: \bibitem{Malek}
3994: J.\ M\'alek, S.-L.\ Drechsler, S.\ Flach, E.\ Jeckelmann, and K.\ Kladko,
3995: J.\ Phys.\ Soc.\ Jpn.\ \textbf{72}, 2277 (2003).
3996: %--------------------------
3997: \bibitem{Sengpta2003}
3998: P.\ Sengupta, A.W.\ Sandvik, and D.K.\ Campbell,
3999: Phys.\ Rev.\ B \textbf{67}, 245103 (2003).
4000: %--------------------------
4001: \bibitem{Kuwabara}
4002: M.\ Kuwabara, H.\ Seo, and M.\ Ogata,
4003: J.\ Phys.\ Soc.\ Jpn.\ \textbf{72}, 225 (2003).
4004: %--------------------------
4005: \bibitem{Sugiura2003}
4006: M.\ Sugiura and Y.\ Suzumura,
4007: J.\ Phys.\ Soc.\ Jpn.\ \textbf{72}, 1458 (2003).
4008: %--------------------------
4009: \bibitem{Nakamura}
4010: M.\ Nakamura,
4011: J.\ Phys.\ Soc.\ Jpn.\ \textbf{68}, 3123 (1999);
4012: Phys.\ Rev.\ B \textbf{61}, 16377 (2000);
4013: \textbf{65}, 209902 (2002);
4014: M.\ Nakamura and J.\ Voit,
4015: \textit{ibid.} \textbf{65}, 153110 (2002).
4016: %--------------------------
4017: \bibitem{Emery}
4018: V.J.\ Emery, in
4019: \textit{Highly Conducting One-Dimensional Solids},
4020:   edited by J.\ Devreese, R.\ Evrard, and V.\ van Doren
4021:   (Plenum, New York, 1979), p.\ 247.
4022: %--------------------------
4023: \bibitem{Bari}
4024: R.A.\ Bari,
4025: Phys.\ Rev.\ B \textbf{3}, 2662 (1971).
4026: %--------------------------
4027: \bibitem{Hirsch}
4028: J.E.\ Hirsch,
4029: Phys.\ Rev.\ Lett.\ \textbf{53}, 2327 (1984).
4030: %--------------------------
4031: \bibitem{Dongen}
4032: P.G.J.\ van Dongen,
4033: Phys.\ Rev.\ B \textbf{49}, 7904 (1994).
4034: %--------------------------
4035: \bibitem{Cannon}
4036: J.W.\ Cannon and E.\ Fradkin,
4037: Phys.\ Rev.\ B \textbf{41}, 9435 (1990);
4038: J.W.\ Cannon, R.T.\ Scalettar, and E.\ Fradkin,
4039: \textit{ibid.} \textbf{44}, 5995 (1991).
4040: %--------------------------
4041: \bibitem{Zhang}
4042: G.P.\ Zhang,
4043: Phys.\ Rev.\ B \textbf{56}, 9189 (1997).
4044: %--------------------------
4045: \bibitem{Voit}
4046: J.\ Voit,
4047: Phys.\ Rev.\ B \textbf{45}, 4027 (1992).
4048: %--------------------------
4049: \bibitem{Sengupta}
4050: P.\ Sengupta, A.W.\ Sandvik, and D.K.\ Campbell,
4051: Phys.\ Rev.\ B \textbf{65}, 155113 (2002).
4052: %--------------------------
4053: \bibitem{Sandvik}
4054: See also A.W.\ Sandvik, P.\ Sengupta, and D.K.\ Campbell,
4055: Phys.\ Rev.\ Lett.\ \textbf{91}, 089701 (2003).
4056: This is a comment to
4057: E.\ Jeckelmann, Phys.\ Rev.\ Lett.\ \textbf{89}, 236401 (2002).
4058: %--------------------------
4059: \bibitem{Fukutome}
4060: N.\ Tomita and H.\ Fukutome, 
4061: Solid State Commun.\ \textbf{81}, 659 (1992); \textbf{81}, 663 (1992).
4062: %--------------------------
4063: \bibitem{Japaridze}
4064: G.I.\ Japaridze, 
4065: Phys.\ Lett.\ A \textbf{201}, 239 (1995);
4066: G.I.\ Japaridze and A.P.\ Kampf, 
4067: Phys.\ Rev.\ B \textbf{59}, 12822 (1999);
4068: G.I.\ Japaridze and S.\ Sarkar,
4069: Eur.\ Phys.\ J.\ B \textbf{27}, 139 (2002).
4070: %--------------------------
4071: \bibitem{Otsuka}
4072: H.\ Otsuka,
4073: Phys.\ Rev.\ Lett.\ \textbf{84}, 5572 (2000);
4074: Phys.\ Rev.\ B \textbf{63}, 125111 (2001).
4075: %--------------------------
4076: \bibitem{Aligia}
4077: L.\ Arrachea, E.R.\ Gagliano, and A.A.\ Aligia,  
4078: Phys.\ Rev.\ B \textbf{55}, 1173 (1997);
4079: A.A.\ Aligia,  K.\ Hallberg, C.D.\ Batista, and G.\ Ortiz,
4080: \textit{ibid.} \textbf{61}, 7883 (2000);
4081: M.E.\ Torio, A.A.\ Aligia, K.\ Hallberg, and H.A.\ Ceccatto,
4082: \textit{ibid.} \textbf{62}, 6991 (2000);
4083: M.E.\ Torio, A.A.\ Aligia, and H.A.\ Ceccatto,
4084: Phys.\ Rev.\ B \textbf{67}, 165102 (2003).
4085: %--------------------------
4086: \bibitem{Nakamura2001}
4087: M.\ Nakamura and K.\ Itoh,
4088: J.\ Phys.\ Soc.\ Jpn.\ \textbf{70}, 3606 (2001).
4089: %--------------------------
4090: \bibitem{Fabrizio}
4091: M.\ Fabrizio, A.O.\ Gogolin, and A.A.\ Nersesyan, 
4092: Phys.\ Rev.\ Lett. \textbf{83}, 2014 (1999);
4093: Nucl.\ Phys.\ B \textbf{580}, 647 (2000).
4094: %--------------------------
4095: \bibitem{Tsuchiizu_JPSJ}
4096: M.\ Tsuchiizu and Y.\ Suzumura,
4097: J.\ Phys.\ Soc.\ Jpn. \textbf{68}, 3966 (1999).
4098: %--------------------------
4099: \bibitem{Takada}
4100: Y.\ Takada and M.\ Kido,
4101: J.\ Phys.\ Soc.\ Jpn. \textbf{70}, 21 (2001).
4102: %--------------------------
4103: \bibitem{Qin}
4104: J.\ Lou, S.\ Qin, T.\ Xiang, C.\ Chen, G.S.\ Tian, and Z.\ Su, 
4105: Phys.\ Rev.\ B \textbf{68}, 045110 (2003).
4106: %--------------------------
4107: \bibitem{Brune}
4108: Ph.\ Brune, G.I.\ Japaridze, A.P.\ Kampf, and M.\ Sekania,
4109: cond-mat/0106007 (unpublished);
4110: A.P.\ Kampf, M.\ Sekania, G.I.\ Japaridze, and Ph.\ Brune,
4111: J.\ Phys.: Condens.\ Matter \textbf{15}, 5895 (2003).
4112: %--------------------------
4113: \bibitem{YZZhang} Y.Z.\ Zhang, C.Q.\ Wu, and H.Q.\ Lin,
4114: Phys.\ Rev.\ B \textbf{67}, 205109 (2003).
4115: %--------------------------
4116: \bibitem{Anusooya-Pati}
4117: Y.\ Anusooya-Pati, Z.G.\ Soos, and A.\ Painelli,
4118: Phys.\ Rev.\ B \textbf{63}, 205118 (2001).
4119: %--------------------------
4120: \bibitem{Wilkens}
4121: T.\ Wilkens and R.M.\ Martin,
4122: Phys.\ Rev.\ B \textbf{63}, 235108 (2001).
4123: %--------------------------
4124: \bibitem{Torio}
4125: M.E.\ Torio, A.A.\ Aligia, and H.A.\ Ceccatto,
4126: Phys.\ Rev.\ B \textbf{64}, 121105 (2001).
4127: %--------------------------
4128: \bibitem{Refolio}
4129: M.C.\ Refolio, J.M.\ L\'opez Sancho, and J.\ Rubio,
4130: cond-mat/0210462 (unpublished).
4131: %--------------------------
4132: \bibitem{Caprara}
4133: S.\ Caprara, M.\ Avignon, and O.\ Navarro,
4134: Phys.\ Rev.\ B \textbf{61}, 15667 (2000).
4135: %--------------------------
4136: \bibitem{Gupta}
4137: S.\ Gupta, S.\ Sil, and B.\ Bhattacharyya,
4138: Phys.\ Rev.\ B \textbf{63}, 125113 (2001).
4139: %--------------------------
4140: \bibitem{Pozgajcic}
4141: K.\ Po\v{z}gaj\v{c}i\'{c} and C.\ Gros,
4142: Phys.\ Rev.\ B \textbf{68}, 085106 (2003).
4143: %--------------------------
4144: \bibitem{Manmana}
4145: S.R.\ Manmana, V.\ Meden, R.M.\ Noack, and K.\ Sch\"onhammer,
4146: cond-mat/0307741 (unpublished).
4147: %--------------------------
4148: \bibitem{MJRice}
4149: M.J.\ Rice and E.J.\ Mele,
4150: Phys.\ Rev.\ Lett.\ \textbf{49}, 1455 (1982).
4151: %--------------------------
4152: \bibitem{Nagaosa}
4153: N.\ Nagaosa and J.\ Takimoto,
4154: J.\ Phys.\ Soc.\ Jpn.\ \textbf{55}, 2735 (1986);
4155: \textbf{55}, 2745 (1986);
4156: N.\ Nagaosa, \textit{ibid.} \textbf{55}, 2754 (1986).
4157: %--------------------------
4158: \bibitem{Girlando}
4159: A.\ Girlando and A.\ Painelli, Phys.\ Rev.\ B \textbf{34}, 2131 (1986);
4160: A.\ Painelli and A.\ Girlando, \textit{ibid.}
4161: \textbf{37}, 5748 (1988); \textbf{39}, 9663 (1989).
4162: %--------------------------
4163: \bibitem{Egami}
4164: T.\ Egami, S.\ Ishihara, and M.\ Tachiki,
4165:  Science \textbf{261}, 1307 (1993);
4166: S.\ Ishihara, T.\ Egami, and M.\ Tachiki,
4167:  Phys.\ Rev.\ B \textbf{49}, 8944 (1994).
4168: %--------------------------
4169: \bibitem{Resta}
4170: R.\ Resta and S.\ Sorella,
4171: Phys.\ Rev.\ Lett.\ \textbf{74}, 4738 (1995); 
4172: \textbf{82}, 370 (1999).
4173: %--------------------------
4174: \bibitem{Gidopoulos}
4175: N.\ Gidopoulos, S.\ Sorella, and E.\ Tosatti,
4176: Eur.\ Phys.\ J.\ B \textbf{14}, 217 (2000).
4177: %--------------------------
4178: \bibitem{Tsuchiizu_Furusaki}
4179: M.\ Tsuchiizu and A.\ Furusaki,
4180:   Phys.\ Rev.\ Lett.\ \textbf{88}, 056402 (2002).
4181: %--------------------------
4182: \bibitem{Nayak}
4183: C.\ Nayak, Phys.\ Rev.\ B \textbf{62}, 4880 (2000).
4184: %--------------------------
4185: \bibitem{Solyom}
4186: J.\ S\'olyom, 
4187: Adv.\ Phys.\ \textbf{28}, 201 (1979).
4188: %--------------------------
4189: \bibitem{Penc_Mila}
4190: K.\ Penc and F.\ Mila,
4191: Phys.\ Rev.\ B \textbf{50}, 11 429 (1994).
4192: %--------------------------
4193: \bibitem{Cardy}
4194: J.\ Cardy,
4195: {\it Scaling and Renormalization in Statistical Physics}
4196:   (Cambridge University Press, Cambridge, UK, 1996).
4197: %--------------------------
4198: \bibitem{Gogolin}
4199: A.O.\ Gogolin, A.A.\ Nersesyan, and A.M.\ Tsvelik,
4200:   \textit{Bosonization and Strongly Correlated Systems}
4201:   (Cambridge University Press, Cambridge, 1998).
4202: %--------------------------
4203: \bibitem{Giamarchi1989}
4204: T.\ Giamarchi and H.J.\ Schulz,
4205: Phys.\ Rev.\ B \textbf{39}, 4620 (1989).
4206: %--------------------------
4207: \bibitem{Bajnok}
4208: Z.\ Bajnok, L.\ Palla, G.\ Tak\'acs, and F.\ W\'agner,
4209: Nucl.\ Phys.\ B \textbf{601}, 503 (2001).
4210: %--------------------------
4211: \bibitem{Delfino}
4212: G.\ Delfino and G.\ Mussardo, 
4213: Nucl.\ Phys.\ B \textbf{516}, 675 (1998).
4214: %--------------------------
4215: \bibitem{Tsuchiizu_JPSJ2}
4216: M.\ Tsuchiizu, H.\ Yoshioka, and Y.\ Suzumura, 
4217: J.\ Phys.\ Soc.\ Jpn.\ \textbf{70}, 1460 (2001).
4218: phase to the SDI phase (see Sec.\ \ref{sec:dimer}). 
4219: %--------------------------
4220: \bibitem{Yonemitsu}
4221: K.\ Yonemitsu,
4222: Phys.\ Rev.\ B \textbf{65}, 085105 (2002); \textbf{65}, 205105 (2002).
4223: %--------------------------
4224: \bibitem{noteUc2}
4225: Brune \textit{et al.} (Ref.\ \onlinecite{Brune}) 
4226:   find very difficult to determine
4227: $U_{c2}$ in their DMRG calculations, probably because a spin gap
4228: in the SDI phase becomes too small near $U=U_{c2}$.
4229: Also, Wilkens and Martin (Ref.\ \onlinecite{Wilkens})
4230:  did not find the SDW phase.
4231: This should be because their method of adding a finite dimerization
4232: $\delta$ and then taking the $\delta\to0$ limit is delicate,
4233: as any small amount of $\delta$ is enough to change the SDW
4234: phase to the SDI phase (see Sec.\ \ref{sec:dimer}). 
4235: %--------------------------
4236: \bibitem{Su}
4237: W.P.\ Su, J.R.\ Schrieffer, and A.J.\ Heeger,
4238:  Phys.\ Rev.\ Lett.\ \textbf{42}, 1698 (1979);
4239:  Phys.\ Rev.\ B \textbf{22}, 2099 (1980).
4240: %--------------------------
4241: \bibitem{Ortiz}
4242: G.\ Ortiz, P.\ Ordej\'on, R.M.\ Martin, and G.\ Chiappe,
4243:  Phys.\ Rev.\ B \textbf{54}, 13515 (1996).
4244: %--------------------------
4245: \bibitem{Shankar}
4246: R.\ Shankar,
4247: in \textit{Current Topics in Condensed Matter and Particle Physics},
4248: edited by J.\ Pati, Q.\ Shafi, Y.\ Lu (World Scientific, Singapore, 1993),
4249: p.\ 83.
4250: %--------------------------
4251: \bibitem{vonDelft}
4252: J.\ von Delft and H. Schoeller,
4253: Ann.\ Phys.\ (Leipzig) \textbf{7}, 225 (1998).
4254: %--------------------------
4255: \bibitem{Affleck}
4256: I.\ Affleck,
4257: Nucl.\ Phys.\ B \textbf{265}, 409 (1986).
4258: %--------------------------
4259: 
4260: 
4261: 
4262: \end{thebibliography}
4263: 
4264: 
4265: \end{document}
4266: