1: \documentclass[prl,aps,amssymb,showpacs,twocolumn]{revtex4}

2: 

3: \usepackage{dcolumn}

4: \usepackage{graphicx}

5: \usepackage{bm}

6: 

7: 

8: \usepackage{amsmath}

9: \usepackage{amssymb}

10: \usepackage{amsthm}

11: \usepackage{amsfonts}

12: \usepackage{enumerate}

13: \usepackage{latexsym}

14: 

15: \input{epsf}

16: 

17: 

18: \begin{document}

19: \title{Topological Mott Insulators}

20: \author{S. Raghu$^{1}$, Xiao-Liang Qi$^1$, C. Honerkamp$^2$, and Shou-Cheng Zhang$^1$}

21: \affiliation{$^1$Department of Physics, McCullough Building,

22: Stanford University, Stanford, CA 94305-4045}

23: \affiliation{$^2$Theoretical Physics, Universit\"{a}t W\"{u}rzburg, D-97074 W\"{u}rzburg, Germany}

24: \date{\today}

25: \begin{abstract}

26: We consider extended Hubbard models with repulsive interactions on

27: a Honeycomb lattice and the transitions from the semi-metal phase

28: at half-filling to Mott insulating phases.  In particular, due to

29: the frustrating nature of the second-neighbor repulsive

30: interactions, topological Mott phases displaying the quantum Hall

31: and the quantum spin Hall effects are found for spinless and

32: spinful fermion models, respectively.  We present the mean-field

33: phase diagram and consider the effects of fluctuations within the

34: random phase approximation (RPA). Functional renormalization group

35: analysis also show that these states can be favored over the

36: topologically trivial Mott insulating states. 

37: \end{abstract}

38: 

39: \pacs{71.10.-w, 71.10.Fd, 71.27.+a, 71.30.+h, 73.22.Gk, 73.43.-f}

40: 

41: \maketitle \emph{Introduction - } Partly motivated by the

42: discovery of the high $T_c$ superconductivity, Mott insulators

43: have attracted great attention in recent years. Defined in a

44: general sense, interactions drive a quantum phase transition from

45: a metallic ground state to an insulating ground state in these systems. Most Mott

46: insulators found in nature also have conventional order

47: parameters, describing, for example, the charge-density-wave (CDW)

48: or the spin-density-wave (SDW) orders. However, Mott insulators

49: with exotic ground states, such as the current carrying ground

50: states have also been proposed theoretically

51: \cite{Affleck1988,Wen1996,Varma1999,Chakravarty2001}.

52: In parallel with the study of strongly correlated systems, there

53: has recently been a growing interest in realizing topologically

54: non-trivial states of matter in band insulators. In the quantum

55: anomalous Hall (QAH) insulators\cite{Haldane1988,Qi2006}, the

56: ground state breaks time reversal symmetry but does not break the

57: lattice translational symmetry. The ground state has a bulk

58: insulating gap, but has chiral edge states. In the quantum spin

59: Hall (QSH) insulators\cite{kane2005,bernevig2006a,bernevig2006d},

60: the ground state does not break time reversal symmetry, has a

61: bulk insulating gap, but has helical edge states, where electrons

62: with the opposite spins counter-propagate. The QSH state has

63: recently been observed experimentally in $HgTe$ quantum

64: wells\cite{bernevig2006d,koenig2007}.

65: 

66: Given the tremendous interest in finding Mott insulators with

67: exotic ground states, and the recent discovery of the

68: topologically non-trivial band insulators, it is natural to ask

69: whether one can find examples of topological Mott insulators,

70: which we define as states with bulk insulating gaps driven by the

71: interaction, and inside which lie topologically protected edge

72: states. Furthermore, electronic states in the Mott insulator

73: phases are characterized by topological invariants, namely, the

74: $U(1)$ Chern number\cite{Thouless1982} in the case of the QAH

75: state, and the $Z_2$ invariant\cite{kane2005A} in the case of the

76: QSH state.  In this letter, we report on the first example of such a 

77: case by systematically studying Hubbard models with repulsive

78: interactions on a two dimensional honeycomb lattice.  In

79: particular, we demonstrate how, due to the frustrated nature of

80: second-neighbor repulsion on this lattice, topological Mott phases

81: displaying the QAH and the QSH effects are generated dynamically.

82: We present the mean-field phase diagram, proceed to consider

83: effects beyond mean-field theory via a functional renormalization

84: group (fRG) treatment, and consider the effects arising from

85: fluctuations.  The two dimensional graphene sheet, which was

86: recently realized experimentally  \cite{Novoselov2005,Zhang2005}, 

87: contains the basic degrees of freedom of

88: our model. However, presently we do not know how to tune the

89: interaction experimentally so that our proposed state can be

90: realized.

91: 

92: \emph{Spinless Fermions and the QAH state - }

93: \begin{figure}

94: \includegraphics[width=3.0in]{geom2.eps}

95: \caption{Interactions considered in our model Hamiltonian

96: (left-most hexagon), Eq. \ref{hamiltonian}. Various order

97: parameters are shown for the A-sublattice (open circles) in the

98: middle hexagon and for the B-sublattice (filled circles) in the

99: right-most hexagon.  The CDW order parameter $\rho$ is a real

100: scalar, SDW order parameter $\bm M$ a real vector and the QAH/QSH

101: order parameters $\chi_A, \chi_B$ are complex 4-vectors.  In the

102: case of spinless fermions, $\chi_A, \chi_B$ are complex scalars.  $\chi_A,\chi_B$ are both defined

103: on the directed second neighbor links defined by $\bm b_i$.    }

104: \label{geometry}

105: \end{figure}

106: The model Hamiltonian for spinless fermions with nearest-neighbor

107: and next-nearest neighbor interactions is  written

108: as\begin{eqnarray} \label{hamiltonian} H=-\sum_{\langle ij \rangle}t

109: \left(c_i^\dagger c_j+h.c. \right)+V_1 \sum_{\langle i,j \rangle }

110: (n_{i}-1)( n_{j}-1) \nonumber  \\ + V_2\sum_{\langle \langle i,j

111: \rangle \rangle}  (n_{i}-1) (n_{j}-1) -\mu\left( \sum_in_i-N \right)

112: \end{eqnarray}

113: where $V_1 $ and $V_2$ are nearest-neighbor and

114: next-nearest-neighbor interaction strengths, respectively.  Since

115: the honeycomb lattice is bipartite, consisting of two triangular

116: sublattices (referred to here as A and B), nearest-neighbor

117: repulsion will favor a charge density wave (CDW) phase with an 

118: order parameter $\rho = \frac{1}{2} \left(\langle c^{\dagger}_{i

119: A} c_{i  A} \rangle - \langle c^{\dagger}_{i  B} c_{i  B} \rangle

120: \right)$ that is consistent with overall charge conservation and

121: describes a phase with a broken discrete (inversion) symmetry.

122: However, since the second neighbor interactions within a

123: sublattice are frustrated, CDW order will be suppressed; instead

124: we consider the possibility of orbital ordering by defining the

125: following order parameter for $i,j$ next nearest neighbors:

126: $\chi_{ij}=\chi_{ji}^* = \langle{c_i^\dagger c_j} \rangle$.  Let

127: $\bm a_1, \bm a_2, \bm a_3$ be the nearest-neighbor displacements

128: from a B-site to an A-site such that $\bm z \cdot \bm a_1 \times

129: \bm a_2$ is positive.   We also define the displacements $\bm b_1

130: = \bm a_2 - \bm a_3$, $ \bm b_2 = \bm a_3 - \bm a_1$, etc, which

131: connect two neighboring sites on the same sublattice (Fig.

132: \ref{geometry}).  A translational and rotational invariant ansatz

133: of $\chi_{ij}$ is chosen as

134: \begin{eqnarray} \chi_{i,i+{\bf

135: b}_s}=\left\{\begin{array}{c c}\chi_A = \vert \chi \vert

136: e^{i\phi_A},&i\in A\\ \chi_B = \vert \chi \vert e^{i\phi_B},&i\in

137: B\end{array}\right.

138: \end{eqnarray}

139: which are \emph{complex} scalars that live along the directed second

140: neighbor links.

141: 

142: Owing to translational

143: symmetry, the mean-field free energy at $T=0$ is readily obtained: 

144: \begin{eqnarray}

145: F \left( \rho, \chi, \bar{\phi}, \phi \right) &=& -  \sum_{\bm k}

146: \sqrt{  \vert  t(\bm k) \vert ^2 + \left( V_1 \rho +2 V_2 \vert \chi

147: \vert S_{\bm k  + \bar{\phi} }   S_{\phi} \right)^2 } \nonumber

148: \\

149: &&+ 3L^2 \left( V_1  \rho^2 + 2V_2 \vert \chi \vert ^2 \right)

150: \end{eqnarray}

151: We have defined the following quantities: $t(\bm k) = \sum_{n=1}^3

152: \exp \left( i \bm k \cdot a_n \right)$, $ \bar{\phi} = \left( \phi_A

153: + \phi_B \right)/2, \phi = \left(\phi_A - \phi_B \right)/2,S_{\bm k+

154: \bar{\phi}}=\sum_{n=1}^3 \sin {\left(\bm k\cdot b_n + \bar{\phi}

155: \right)}, S_{\phi} = \sin{\phi} $.

156: Thus, the next-neighbor hopping

157: amplitudes are purely real only when both $\phi = 0$ and $\bar{\phi}

158: = 0$.

159: 

160: When both $\rho$ and $\chi= 0$, and at half-filling, the system

161: remains a semi-metal consisting of two Fermi points $\bm K_{\pm}$

162: which obey $\bm K_{\pm} \cdot \bm b_i = \pm 2 \pi/3$ and the density

163: of states vanishes linearly; the dispersion in the vicinity of these

164: so called Dirac points is governed by a 2D massless Dirac

165: Hamiltonian in $\bm k$-space . The CDW phase corresponds to an

166: ordinary insulator with a gap at the Fermi energy. As for the order

167: parameter $\chi$, which describes the second-neighbor hopping, its

168: \emph{phase} relative to the nearest neighbor hopping amplitude

169: plays an important role in determining its properties: while a

170: non-zero $Re(\chi)$ merely shifts the energy of the Dirac points, a

171: non-zero imaginary part $Im(\chi)$ \emph{opens a gap} at the Fermi

172: points.  Thus, when the system remains at half-filling, it is more

173: favorable to develop  purely imaginary next-neighbor hopping

174: amplitudes; such a configuration corresponds to a phase with

175: \emph{spontaneously broken time-reversal symmetry}.

176: 

177: To see whether such a phase can be favored, we minimize the

178: free-energy and arrive at the following self-consitent

179: equations:\begin{equation} \rho = \frac{1}{2 L^2} \sum_{\bm k}

180: \frac{V_1 \rho + 2V_2\chi S_{\bm k + \bar \phi} S_{\phi}}{\sqrt{

181: \vert t(\bm k) \vert^2

182:  +\left(V_1 \rho + 2V_2\chi S_{\bm k + \bar{\phi} } S_{\phi}\right)^2 } }

183: \end{equation}

184: \begin{equation}

185: \chi = \frac{S_{\phi}}{6L^2} \sum_{\bm k} \frac{S_{\bm k +

186: \bar{\phi} }   \left(V_1 \rho + 2V_2 \chi S_{\bm k + \bar \phi }

187: S_{\phi} \right)}{\sqrt{ \vert t(\bm k) \vert^2 +\left(V_1 \rho +

188: 2V_2 \chi S_{\bm k  + \bar{\phi} } S_{\phi}\right)^2 }

189: }

190: \end{equation} When $\chi = 0$, it is easy to see from the first

191: equation above that CDW order develops continuously at a critical

192: value $V_{1c}$ given by

193: \begin{equation}

194: \frac{1}{V_{1c}} = \frac{1}{2L^2} \sum_{\bm k} \frac{1}{\vert t(\bm k) \vert},

195: \end{equation}

196: Due to the vanishing density of states (DOS) near the Fermi points,

197: there is no instability towards CDW formation with infinitesimal

198: interactions.  Interestingly, the self-consitent equation for

199: $\chi$ shows that a non-trivial self-consistent solution can only

200: occur when $\phi \ne 0$; a detailed investigation of these

201: equations \cite{Raghu2007} show that when $V_1 = 0$, beyond a

202: critical value of $V_{2c}>0$, which satisfies\begin{equation}

203: \label{gapeqQAHE} \frac{1}{V_{2c}} = \frac{1}{3L^2} \sum_{\bm

204: k}\frac{S^2_{\bm k + \bar{\phi}} }{\vert t_{\bm k} \vert},

205: \end{equation}

206: a phase in which $\vert \chi \vert > 0$ ,$\bar{\phi} = 0$, and

207: $\phi = \pm \pi/2$ is favored.  Such a phase is also stable at

208: finite $V_1$ and is thus does not require fine-tuning (see Fig.

209: \ref{pd2}). In this phase, the system acquires purely imaginary

210: second-neighbor hoppings with a chirality which is determined by

211: the sign of $\phi$. There is a discrete symmetry breaking

212: corresponding to choosing $\phi = \pm \pi/2$, each of which breaks

213: time-reversal symmetry. The band insulator version of the CDW

214: state was considered in Ref. \cite{Semenoff1984}, while the QH

215: state on a honeycomb lattice was considered in Ref.

216: \cite{Haldane1988}. As discussed in Ref. \cite{Haldane1988} the

217: QAH phase is a topological phase in which the filled states form a

218: band which has a non-zero topological Chern number

219: \cite{Thouless1982} and is an integer quantum Hall effect phase

220: that is realized \emph{without} Landau levels. Similar states can

221: be constructed from magnetic semiconductors\cite{Qi2006}. We shall

222: generally refer to QH states without Landau levels, and with full

223: lattice translation symmetry as the \emph{Quantum Anomalous Hall}

224: (QAH) states. In our case, the topologically non-trivial gap for

225: the QAH state arises from many-body interactions rather than

226: single particle physics, and we shall refer to such states as

227: topological Mott insulators.

228: \begin{figure}

229: \includegraphics[width=2.5in]{PD2.eps}

230: \caption{Phase diagram for spinless fermions ($t=1$).  The

231: semi-metallic (SM) state that occurs at weak-coupling is separated

232: from the CDW and the topological QAH states via a continuous

233: transition (blue curve). The line separating the QAH and CDW marks a

234: first-order transition (shown in red), which terminates at a bi-critical point.}

235: \label{pd2}

236: \end{figure}

237: 

238: We have obtained the complete  phase diagram in the $V_1-V_2$

239: plane; within mean-field theory, there is a continuous transition

240: from the semi-metal to either the CDW or the QAH phase and there

241: is also a first-order transition from the CDW to the QAH phase. By

242: integrating out the fermionic fields, it is possible to construct

243: a Landau-Ginzburg (LG) theory expansion near the nodal region

244: where the order parameters are vanishingly small.  Due to the

245: linear dispersion of the Fermi points, the LG free-energy contains

246: anomalous terms of the form  $\vert \rho \vert^3$ and $\vert

247: Im(\chi) \vert^3$ arising from linear dispersion in the vicinity

248: of the Fermi points \cite{Raghu2007}.  The significance of such

249: terms is that even within mean-field theory, the CDW order

250: parameter, for instance, grows as $(V_1 - V_{1c})$ rather than the

251: usual $\left( V_1 - V_{1c} \right)^{1/2}$ \cite{Sorella1992}.

252: Furthermore, the Landau-Ginzburg theory describing the competition

253: between the CDW and QAH phases confirms the existence of the first

254: order line between these two phases that terminates in the

255: bicritical point leading into the semimetal phase

256: \cite{Raghu2007}.

257: 

258: \emph{Spinful fermions and the QSH state -} Next, we take into

259: account the spin degrees of freedom and include an onsite Hubbard

260: repulstion in our model Hamiltonian ($\mu = 0$):

261: \begin{eqnarray}

262: H=-\sum_{\langle ij \rangle  \sigma}t\left(c_{i \sigma}^\dagger c_{j

263: \sigma}+h.c. \right)+U\sum_i n_{i \uparrow} n_{i \downarrow}

264: \nonumber \\ + V_1\sum_{\langle i,j \rangle }   (n_{i}-1)( n_{j}-1)

265: + V_2\sum_{\langle \langle i,j \rangle \rangle}  (n_{i}-1)(

266: n_{j}-1)\label{hamilton}

267: \end{eqnarray}

268: where $n_i = n_{i \uparrow} + n_{i \downarrow}$.  Since the

269: Honeycomb lattice is bipartite, onsite repulsion gives rise to a

270: spin density wave phase (SDW); a standard decomposition of the Hubbard term 

271: introduces the order parameter $\bm M$ describing an

272: antiferromagnetic SDW: $\bm M = \frac{1}{2} \left( \langle \bm

273: S_{iA} \rangle-\langle \bm S_{iB} \rangle \right)$ .  As in the

274: spinless case, nearest-neighbor repulsion favors a CDW.  However,

275: there are several possible phases due to second-neighbor

276: repulsion.   Again, since the second-neighbor repulsion is

277: frustrated, we are again led to the possibility of a topological

278: phase similar to the QAH.  However, the spin degrees of freedom

279: introduce two possibilities (translation invariance along with

280: spin conservation eliminate other possibilities): 1) \emph{two

281: copies} of QAH states - i.e. the chirality of the

282: second-neighbor hopping is the \emph{same} for each spin projection, 2)

283: the QSH state, where the chiralities are \emph{opposite} for each

284: spin projection.   The latter possibility breaks a

285: \emph{continuous} global $SU(2)$ symmetry associated with choosing

286: the spin projection axis;  however, \emph{time-reversal symmetry

287: is preserved}. The QSH state on the honeycomb lattice was

288: considered in Ref. \cite{kane2005}, where the insulating gap

289: arises from the microscopic spin-orbit coupling. It was shown

290: latter that the magnitude of of the spin-orbit gap is negligibly small

291: in graphene \cite{Min2006,Yao2007}. In our case, the insulating gap is generated

292: dynamically from the many-body interaction. In this sense, our

293: effect can be viewed as an example of dynamic generation of

294: spin-orbit interaction\cite{Wu2004}.

295:   Introducing the Hubbard-Strataovich fields (sum over repeated indices implied)

296: $\chi^{\mu}_{ij} = c^{\dagger}_{i \alpha} \sigma^{\mu}_{\alpha

297: \beta} c_{j \beta}$, $\mu = 0 \ldots 3$, where $\sigma^{\mu} =

298: \left( 1, \bm \sigma \right)$, the next-neighbor interactions can

299: be recast using the identity $ (n_{i } -1)(n_{j}-1) =1  -

300: \frac{1}{2} \left( \chi^{\mu }_{ij} \right)^{\dagger}

301: \chi^{\mu}_{ij} $. Physically, if $\langle \chi^0 \rangle \neq 0$,

302: then we are in the QAH phase.  If, on the other hand, one of the

303: vector components $\langle \chi^{i} \rangle \neq 0$, then we are

304: in the QSH phase.  A translationally invariant decomposition of

305: the next-neighbor interactions via $ \langle \chi^{\mu}_{i,i+\bm

306: b_s} \rangle = \chi^{\mu}e^{i \phi^{\mu}_A} , i \in A $ (and

307: similarly for the other sublattice) gives rise to 

308: a $4 \times 4$ Hamiltonian is readily diagonalized in

309: a tensor product basis $\bm \sigma \otimes \bm \tau$, where $\bm

310: \sigma$ and $\bm \tau$ are Pauli matrices in spin and sublattice

311: space, respectively. This way, each phase corresponds to a

312: particular non-zero expectation value of a fermion bilinear

313: $\sum_{\vec{k}}\Psi_{\vec{k}}^\dagger \hat{d}(\vec{k})

314: \Psi_{\vec{k}}$, where $\hat{d}(\vec{k}) \propto \tau^3$ for the

315: CDW and QAH, $\hat{d}(\vec{k}) \propto \sigma^3\tau^3$ for SDW and

316: QSH.  A detailed study of the free-energy at $T=0$ and its saddle

317: point solutions \cite{Raghu2007} produces the phase diagram shown

318: in Fig. \ref{pd3}.  In addition to the ordinary CDW and SDW

319: insulating phases, there is a phase for $V_2 > V_{2c} \approx 1.2

320: t$ in which the 4-vector is purely imaginary (as in the spinless

321: case), collinear, and staggered from one sublattice to the next:

322: $\langle \chi^{\mu}_{i i+\bm b_n, A} \rangle = - \langle

323: \chi^{\mu}_{i i+\bm b_n, B} \rangle$, and both QAH and QSH are

324: equally favorable ground states, having identical free energies

325: within mean-field theory.  Additionally, there is never a

326: coexistence of both QAH and QSH phases; indeed, a Landau-Ginzburg

327: treatment in this region explicitly shows the absence $SO(4)$

328: symmetry of the vector $\chi^{\mu}$.  This occurs due to the

329: difference of the manner in which $\chi^0$ and $\vec{\chi}$ are

330: coupled to the fermionic fields - which favors either a phase with

331: broken $Z_2$ symmetry (QAH) or with broken $SU(2)$ symmetry, but

332: never both simultaneously \cite{Raghu2007}.

333: 

334: Quantum fluctuations, however, lift the mean-field degeneracy

335: between the QAH and QSH phases.  To quadratic order in quantum

336: fluctuations (RPA) about the QSH phase , we obtain an effective

337: action $ S_{eff} = \sum_{\vec{k}} \delta \chi^{\mu}(\vec{k},

338: \Omega) K_{\mu \nu} (\vec{k}, \Omega ) \delta \chi^{\nu}

339: (-\vec{k}, -\Omega)$ which shows the presence of six modes (2

340: longitudinal and 4 transverse modes), and 2 of the transverse

341: modes correspond to degenerate Goldstone modes whose velocity is

342: proportionality to the Fermi velocity $v \approx v_f = 3t/2\vert

343: \bm a \vert $.  Thus, the zero-point motion associated with these

344: gapless modes, lowers the free energy of the QSH state relative to

345: the QAH state.

346: 

347: \begin{figure}

348: \includegraphics[width=3.0in]{PD3.eps}

349: \caption{Complete mean-field phase diagram for the spinful model.  The transitions from the semimetal (SM) to the insulating phases are continuous, whereas transitions between any two insulating phases (red lines) are first-order.  

350: } \label{pd3}

351: \end{figure}

352: 

353: 

354: \emph{Renormalization Group Analysis - } Mean field

355: theory generally starts with a given, in a sense, biased Ansatz,

356: and investigate the self-consistency of the mean field solution.

357: Therefore, it is important to investigate the topological Mott

358: states with a method without any a priori bias. Next we go beyond

359: mean-field theory and RPA using the temperature($T$)-flow

360: functional renormalization group

361: (fRG)\cite{Honerkamp2001}\cite{Honerkamp2001A}.  

362: In this scheme, we discretize the $\vec{k}$-dependence of the

363: interaction \cite{Zanchi2000}

364: and consider all possible scattering processes between a set of initial

365: and final momenta that occur between

366: points on rings around the Dirac points (inset of Fig. \ref{rgfig}).

367: Starting with $T_0 \sim 2t$, the

368: temperature $T$ is lowered, and a flowing (renormalized) interaction

369: $V_T$ is obtained by the

370: coupled summation of the $T$-derivatives of {\em all} one-loop channels.

371: Due to this, the method is

372: unbiased and goes beyond the mean-field-level.  

373: Applying the scheme to the Hamiltonian, Eq.

374: \ref{hamilton}, we search for {\em

375:   flows to strong coupling}, where for a low temperature $T_c$ certain

376: components of $V_T $ become large. Then the approximations break

377: down, and the flow is stopped. Information on the low-$T$ state is

378: obtained from analyzing which coupling functions grow most

379: strongly and from susceptibilities for static external fields

380: coupling to the various order parameters.  In this scheme, a

381: tendency towards ordering at a finite vector $\bm Q$ can be

382: detected as a growth of the associated vertex $V_T$.  However, we

383: have found that largest couplings occur at $\bm Q = 0$, which

384: strongly supports the mean-field results presented above.

385: 

386: For onsite and nearest-neighbor repulsions $U>U_c \approx 3.8t$

387: and $V_1> V_{1c} \approx 1.2t$, the flow to strong coupling is

388: either an SDW instability for dominant $U$ or a CDW instability

389: for dominant $V_1$, in good agreement with a

390: 1/N-study\cite{Herbut2006} and

391: Quantum-Monte-Carlo\cite{Sorella1992}. For more details, see Ref.

392: \cite{Honerkamp2007}. If we include a sufficiently strong

393: second-nearest-neighbor repulsion $V_2>1.6t$, the flows change

394: qualitatively; there is a leading growth of the QSH

395: susceptibility. In Fig. \ref{rgfig} a) and b) we compare the

396: $T$-flows of various susceptibilities for $V_1>V_2$ and for

397: $V_2>V_1$. For the latter case, the QSH susceptibility grows most

398: strongly toward low $T$, followed by the QAH susceptibility, which

399: is consistent with the RPA treatment of the Goldstone modes in the

400: QSH. The QSH phase remains stable even when a moderate onsite

401: interaction of $U=t$ or $U=2t$ is introduced. Hence the global

402: structure of the mean-field phase diagram is confirmed by the fRG

403: results. Note however that the slope of the lines of critical

404: $V_1$ versus $V_2$ differs. We interpret this a competition effect

405: captured by the fRG, where $V_2$ decreases the CDW tendencies

406: induced by $V_1$.

407: 

408: 

409: \begin{figure}

410: \includegraphics[width=3.5in]{rgfig.eps}

411: \caption{{\em a)} Data for $U$=0, $V_1$=1.4$t$, $V_2$=0.  Susceptibilities of each phase vs. $T$ are shown: CDW (black);  SDW (green); QAH(red) and QSH (blue).{\em b)} Same for $U$=0, $V_1$=0, $V_2$=1.8$t$ (QSH instability).  The QSH phase has a larger susceptibility than QAH.   {\it Inset}: fRG phase diagram at $U$=0, indicating SM (blue) and insulating (red) regions (CDW dominates at large $V_1$, QSH at large $V_2$.)   The colorbar correspond to $T_c$ below which the insulating phases develop in fRG.}

412: \label{rgfig}

413: \end{figure}

414: 

415: \emph{Discussion - } We have shown that topological phases

416: displaying the QAH and QSH effects can be generated from strong

417: interactions - thus, we refer to these phases as \emph{topological

418: Mott insulators}. Both phases have associated with them

419: conventional order parameters which develop continuously at the

420: quantum critical phase transition from the semi-metallic state.

421: However, these states are also described by topological quantum

422: numbers which jump discontinuously at the transition. Although

423: the interaction strengths needed to produce these phases are

424: strong, we expect that our mean-field, RPA and fRG treatment to

425: provide strong evidence for the existence of a topological Mott

426: insulator: the Stoner criterion, which states that perturbation theory 

427: breaks down when interaction strengths are comparable to the 

428: inverse of the effective DOS, suggests that perturbation theory 

429: is more robust in our system due to the vanishing of the 

430: DOS at the Fermi points.  

431: An open issue remains to find other possible realizations

432: of the phases described here. Furthermore, the nature of the low energy

433: effective coupling of the gapless bulk Goldstone modes with the gapless

434: \emph{edge} degrees of freedom, and the possibility of

435: fractionalized excitations in these phases are interesting

436: open issues that we leave for future work.

437: 

438: \section{Acknowledgments}

439: We are grateful to D. P. Arovas and S. A. Kivelson for insightful discussions.   

440: This work is supported by the NSF under grant numbers DMR-0342832, 

441: the US Department of Energy, Office of Basic Energy Sciences

442: under contract DE-AC03-76SF00515, BaCaTec (C.H.),

443: and the Stanford Institute for Theoretical Physics (S.R.).    

444: 

445: 

446: 

447: \bibliography{tmi}

448: 

449: \end{document}

450: