1: \documentclass[twocolumn,prl,superscriptaddress]{revtex4}

2: \usepackage{graphicx}

3: \usepackage{times}

4: \usepackage{amsmath}

5: 

6: \newcommand{\bra}[1]{\left\langle{#1}\right|}

7: \newcommand{\ket}[1]{\left|{#1}\right\rangle}

8: 

9: \begin{document}

10: 

11: \title{Breakdown of a topological phase: 

12: Quantum phase transition in a loop gas model with tension}

13: 

14: \author{Simon Trebst}

15: \affiliation{Microsoft Research, Station Q, University of California,

16: Santa Barbara, CA 93106}

17: \author{Philipp Werner}

18: \affiliation{Department of Physics, Columbia University, 538 West

19: 120th Street, New York, NY 10027}

20: \author{Matthias Troyer}

21: \affiliation{Theoretische Physik, Eidgen\"ossische Technische

22: Hochschule Z\"urich, CH-8093 Z\"urich, Switzerland}

23: \author{Kirill Shtengel}

24: \affiliation{Department of Physics and Astronomy, University of

25: California, Riverside, CA 92521}

26: \author{Chetan Nayak}

27: \affiliation{Microsoft Research, Station Q, University of California,

28: Santa Barbara, CA 93106}

29: \affiliation{Department of Physics and Astronomy, University of

30: California, Los Angeles, CA 90095}

31: \date{\today}

32: 

33: \begin{abstract}

34:   We study the stability of topological order against local

35:   perturbations by considering the effect of a magnetic field on a

36:   spin model -- the toric code -- which is in a topological phase.

37:   The model can be mapped onto a quantum loop gas where the

38:   perturbation introduces a bare loop tension.  When the loop tension

39:   is small, the topological order survives.  When it is large, it

40:   drives a continuous quantum phase transition into a magnetic state.

41:   The transition can be understood as the condensation of `magnetic'

42:   vortices, leading to confinement of the elementary `charge'

43:   excitations.  We also show how the topological order breaks

44:   down when the system is coupled to an Ohmic heat bath and discuss

45:   our results in the context of quantum computation applications.

46:  \end{abstract}

47: 

48: \maketitle

49: 

50: % Introduction ---------------------------------------------------------------------

51: 

52: % paragraph{Introduction --}

53: Topological phases are among the most remarkable phenomena in nature.

54: Although the underlying interactions between electrons in a solid are

55: not topologically invariant, their low-energy properties are. This

56: enhanced symmetry makes such phases an attractive platform for quantum

57: computation since it isolates the low-energy degrees of freedom from

58: local perturbations

59: -- a usual cause of errors \cite{ToricCode}. Tractable theoretical

60: models

61: with topological phases in frustrated magnets

62: \cite{ToricCode,magnets}, Josephson junction arrays

63: \cite{Ioffe02,Motrunich02}, or

64: cold atoms in traps \cite{ColdAtoms} 

65: have been proposed.  However, such phases have not, thus far, been

66: seen experimentally outside of the quantum Hall regime.

67: Is it because topological phases are very rare and

68: these models are adiabatically connected only to very small 

69: regions of the phase diagrams of real experimental systems?

70: 

71: In this paper, we take a first step towards answering this question.

72: We begin with the simplest exactly soluble model of a topological

73: phase \cite{ToricCode}, whose Hamiltonian is given below in

74: Eq.~(\ref{Eq:ToricCode}).  This model describes a frustrated magnet

75: with four-spin interactions similar to cyclic ring exchanges.  It is

76: closely related to the quantum dimer model \cite{DimerModel} for

77: frustrated magnets, which can be realized in Josephson junction arrays

78: \cite{Ioffe02}.  We consider perturbations of the soluble model that,

79: when sufficiently large, drive the system out of the topological

80: phase.

81: The question is, how large? A small answer would imply that this

82: topological phase is delicate and occupies a small portion of the

83: phase diagram.  This might explain the paucity of

84: experimentally-observed topological phases. Instead, we find that

85: `sufficiently large' is a magnetic field $h$ of order one ($h_c

86: \approx 0.6$)

87: in units of the basic four-spin plaquette interaction. 

88: Our numerical simulations demonstrate, for the first time, several key

89: signatures of the phase transition out of the topological phase,

90: including the finite-size degeneracy splitting of the topological

91: sectors, the condensation of `magnetic' excitations, and the

92: confinement of `electric' charges.

93: 

94: We also consider perturbing the system by coupling it to an 

95: Ohmic heat bath. When coupled to such a bath, a quantum mechanical

96: degree of freedom can undergo a transition

97: from coherent to incoherent behavior \cite{Leggett87}.

98: Recently, the effects of such a coupling on quantum phase transitions,

99: at which divergent numbers of quantum mechanical

100: degrees of freedom interact, have also been

101: studied \cite{Werner04}. In both contexts, the coupling to

102: the heat bath tends to make the system more

103: classical. Coherent quantum oscillations are suppressed, while

104: broken symmetry phases -- which are essentially

105: classical -- are stabilized. A topological phase is

106: quantum mechanical in nature.

107: We find that coupling the heat bath to the kinetic energy, i.e.

108: plaquette flip operator, does not destroy such a phase. 

109: However, when the dissipation is strong, the gap becomes very small,

110: and the topological phase may be too delicate to observe or use at

111: reasonable temperatures.  On the other hand, if the heat bath is

112: coupled to the classical state of each plaquette, the topological

113: phase is destroyed through a Kosterlitz-Thouless transition at a

114: dissipation strength of order one.

115: 

116: These results can be recast in quantum information language: the

117: ground states in different topological sectors are the different basis

118: states of an encoded quantum memory.  Quasiparticle excitations are

119: states outside of the code subspace.  The stability of the topological

120: phase, as measured by an energy gap $\Delta$ within a topological

121: sector (essentially the energy cost for a pair of quasiparticles),

122: translates into an error rate for topological qubits. At zero

123: temperature, errors are due to the virtual excitation of pairs of

124: quasiparticles, assuming that the system is shielded from

125: perturbations at frequencies higher than $\Delta$.  Such virtual

126: processes lead to a splitting between topological sectors

127: $\delta\! E \sim e^{-\Delta L/v}$, where $L$

128: is the system size and $v$ is a characteristic velocity. As the

129: temperature is increased, the thermal excitation of particles

130: eventually becomes more important and the error rate is $\sim

131: e^{-\beta\Delta}$ \footnote{In a special case of a topological phase

132: considered in \cite{Chamon05}, thermal activation \emph{always} dominates

133: quantum tunneling.}. (The actual concentration of excitations leading to

134: unrecoverable errors was studied in \cite{Wang03}.)

135: 

136: %- The toric code  ---------------------------------------

137: 

138: \begin{figure}[b]

139:   \includegraphics[height=25mm]{./ToricCode.eps}

140:   \hspace{3mm}

141:   \includegraphics[height=25mm]{./ToricCode2.eps}

142:   \caption{

143:     Left: Illustration of the toric code Hamiltonian (\ref{Eq:ToricCode}).

144:     Right: The elementary excitations above the loop gas ground state 

145:     are pairs of `magnetic vortices' on plaquettes connected by a string 

146:     of $\sigma^z$-operators (solid line) and `electric charges' on the 

147:     vertices connected by a string of $\sigma^x$-operators (dashed line).

148:   }

149:   \label{Fig:ToricCode}

150: \end{figure}

151: 

152: \paragraph{The model --}

153: We start with the toric code Hamiltonian \cite{ToricCode} 

154: \begin{equation}

155:   H_{\rm TC} = - A \sum_v \prod_{j\in\mbox{\scriptsize vertex}(v)}\sigma^z_j 

156:          - B  \sum_p \prod_{j\in\mbox{\scriptsize plaquette}(p)}\sigma^x_j \,,

157:   \label{Eq:ToricCode}

158: \end{equation}

159: where the $\sigma_i$ are $S=1/2$ quantum spins on the $2N$ edges of a

160: square lattice with $N$ vertices on a

161: torus. 

162: Since all terms in Hamiltonian (\ref{Eq:ToricCode}) commute with each

163: other, the model can be solved exactly \cite{ToricCode}. The

164: ground-state manifold can be described as a quantum loop gas where the

165: loops consist of chains of up-pointing $\sigma^z$-spins and the loop

166: fugacity is $d=1$.  On the torus there are four degenerate ground

167: states which can be classified by a winding number parity $P_{y/x} =

168: \prod_{i\in c_{x/y}} \sigma_i^z$ along a cut $c_{x/y}$ in the $x-$ or

169: $y-$direction.

170: 

171: % Loop tension & mapping

172: % -------------------------------------------------------------

173: 

174: Here we study the effect of perturbing the Hamiltonian

175: (\ref{Eq:ToricCode}) with a loop tension which can be introduced

176: either by a longitudinal magnetic field or local Ising interaction of

177: the form

178: \begin{equation}

179:   H = H_{\rm TC} - h \sum_i \sigma_i^z - J \sum_{\langle ij \rangle} 

180: \sigma_i^z\sigma_j^z \,,

181:   \label{Eq:LoopTension}

182: \end{equation}

183: where $h$ ($J$) is the strength of the magnetic field (Ising

184: interaction). These are the dominant perturbations expected in a

185: physical implementation, e.g. in a Josephson junction implementation

186: \cite{Ioffe02,Motrunich02} they arise from electric potential

187: perturbations or Coulomb interactions between neighboring 

188: quantum dots. 

189: We discuss this model in the limit of a large charge gap,

190: i.e. $A\gg B,h,J$, where it becomes equivalent to the `even' Ising

191: gauge theory \cite{IGT}.  The low-energy sector has no free charges

192: and any state is described by a collection of loops that can be

193: obtained from a reference state (e.g. all $\sigma_i^z={1}/{2}$) by a

194: sequence of plaquette flips.  Let us introduce a new plaquette spin

195: operator $\mu_p$ with eigenvalues $\mu_p^z = (-1)^{n_p}/2$ where $n_p$

196: is the number of times a given plaquette $p$ has been flipped,

197: counting from the reference state.  Then $\sigma_i^z = 2 \mu_p^z

198: \mu_q^z$, where $p$ and $q$ are the plaquettes separated by the edge

199: $i$. The plaquette flip term in Eq.~(\ref{Eq:ToricCode}) becomes

200: $-4B\sum_{p} \mu_p^x$. In the new variables, Hamiltonian

201: (\ref{Eq:LoopTension}) becomes equivalent to the transverse field

202: Ising model (with both nearest and next-nearest neighbor Ising

203: interactions) in a basis restricted to loop states. Independent of

204: the choice of basis states the system 

205: orders at a critical field strength $(h/B)_c = 0.65695(2)$ determined

206: by continuous-time quantum Monte Carlo simulations \cite{TIM}.  The

207: transverse field Ising model for the plaquette spins can be mapped to

208: a classical (2+1)-dimensional Ising model:

209: \begin{equation}

210:   \mathcal{H}_{\text{cl}} 

211:   = - K_{\tau} \sum_{\tau, p} S_p(\tau)

212: S_p(\tau+\Delta\tau)- K \sum_{\tau, \langle p,q \rangle}S_p(\tau) S_q(\tau), 

213:   \label{Eq:3DIsing}

214: \end{equation}

215: where the imaginary time $\tau$ is discretized in steps of

216: $\Delta\tau$ and $S_p \equiv 2 \mu_p^z = \pm 1$.  The magnetic field

217: $h$ and the local Ising interaction $J$ introduce a nearest and

218: next-nearest neighbor interaction between the classical spins

219: $S_p(\tau)$, respectively.

220: Since the exact nature of this Ising interaction does not play a role

221: in the following we do not discuss the case of next-nearest neighbor

222: couplings in detail.  The coupling along real-space coordinates is

223: then given by $K = \frac{1}{2} \Delta\tau \cdot h$ and along imaginary

224: time by $K_{\tau} = -\frac{1}{2} \ln \left[ \tanh (\Delta\tau \cdot B)

225: \right]$.  The model (\ref{Eq:3DIsing}) describes the well-known

226: continuous magnetic phase transition of the 3D Ising model. For

227: isotropic interactions, $K=K_{\tau}$, the critical coupling has been

228: determined with high precision to be $K_c = 0.221 659 5(26)$

229: \cite{Ferrenberg:91}.  Setting $B=1$ this gives a critical loop

230: tension $h_c = 0.582 24$ with isotropic couplings at $\Delta \tau =

231: 0.761 403$.

232: \begin{figure}[t]

233:   \includegraphics[height=58mm]{./Magnetization.eps}

234:   \caption{

235:     (Color online) Magnetization $M = \langle \sum_i \sigma_i^z

236:     \rangle/2N$ versus loop tension (magnetic field).  The topological

237:     phase survives for small loop tension where an almost constant

238:     susceptibility (see inset) leads to a linear increase of $M$.

239:     Above the critical loop tension (dashed line) the system

240:     approaches the fully polarized state.  }

241:   \label{Fig:Magnetization}

242: \end{figure}

243: % Phase transition for loop tension ---------------------------------------

244: \begin{figure}[b]

245:   \includegraphics[width=86mm]{./Gap_TIM.eps}

246:   \caption{

247:     (Color online) Excitation gap for magnetic vortex excitations

248:     (star symbols) versus the loop tension.

249:     At the critical loop tension the gap closes and the magnetic

250:     vortices condense.  Right ordinate: Rate $\exp(-\beta

251:     \Delta)$ for tunneling processes between topological sectors

252:     (filled circles) for $\beta=10$.  }

253:   \label{Fig:TIM}

254: \end{figure}

255: 

256: The magnetic susceptibility diverges at the transition and the

257: magnetization $\langle \sum_i \sigma_i^z \rangle/2N$ shows a corresponding

258: kink, as shown in Fig.~\ref{Fig:Magnetization}. Although

259: this is not a symmetry-breaking transition, the

260: analogous transition driven by next-nearest interaction $J$ is a

261: {\em continuous} quantum phase transition from a topologically 

262: ordered quantum state to a broken symmetry state \cite{ContinuousPhaseTransition}.

263: The magnetic transition can also be understood in terms of the condensation

264: of `magnetic vortices', plaquettes with $\prod_{j}\sigma^x_j=-1$. 

265: While for the original Hamiltonian (\ref{Eq:ToricCode}), the gap to these

266: excitations is $\Delta = 2B$, they become gapless and condense at the critical

267: loop tension, as shown in Fig.~\ref{Fig:TIM}. 

268: The gap has been estimated from measurements of the imaginary time

269: correlation length $\xi_\tau$ as $\Delta \propto 1/\xi_\tau$ which we have

270: calculated applying continuous-time quantum Monte Carlo simulations using the

271: ALPS looper code \cite{ALPS,LooperCode}). The thermal excitation of pairs

272: of magnetic vortices occurs with probability $\sim \exp(-\beta\Delta)$,

273: also shown in Fig.~\ref{Fig:TIM}.

274: \begin{figure}[t]

275:   \includegraphics[height=58mm]{./Energy.eps}

276:   \caption{

277:     (Color online)

278:     Splitting of the energy degeneracy

279:     for the four topological sectors defined by even or odd winding

280:     number parities along the $x$- and $y$-directions.  The

281:     intermediate \{even-odd / odd-even\}-parity sector was taken as a

282:     reference (dashed line).  }

283:   \label{Fig:Energy}

284: \end{figure}

285: 

286: \paragraph{Topological order --} 

287: The breakdown of topological order at

288: the phase transition can be seen from the energy

289: splitting $\delta\! E$ between the ground-states for the various

290: topological sectors. When winding 

291: parities are used as basis states for a quantum memory, this splitting

292: causes phase errors.  (The absence of `electric' charges precludes any

293: transitions between different winding parities so bit flip errors cannot

294: occur.)  In the topological phase, the virtual excitation of quasiparticles

295: leads to a small splitting $\delta\! E \propto \exp(-\Delta L/v)$ between the

296: topological sectors.  In the classically-ordered phase, on the other

297: hand, the energy splitting should scale with $L$, which corresponds to

298: the energy cost of a loop in the ordered ground state.  As the winding

299: parity is conserved by imaginary time spin-flip operations, we

300: can simulate the system in one of the topological sectors by defining

301: an initial spin configuration that corresponds to the respective limit

302: for large loop tension. 

303: Fig.~\ref{Fig:Energy} shows the calculated splitting

304: for various system sizes in the vicinity of the critical loop tension.

305: At the phase transition, the behavior qualitatively changes from

306: power-law scaling for strong loop tension to an exponential

307: suppression in the topological phase for small loop tension as

308: discussed above. A more quantitative picture arises from the

309: finite-size scaling analysis of the energy splitting $\delta\!  E(L)$

310: between the \{even-odd\}- and \{even-even\}-parity sectors shown in

311: Fig.~\ref{Fig:EnergyScaling}.  For the critical loop tension we find a

312: power-law scaling $\delta\!  E(L) \propto L^{2-z}$ with an exponent $z

313: = 1.42 \pm 0.02$.

314: Below the critical value the scaling 

315: turns into exponential scaling as expected for the topological phase.

316: 

317: \begin{figure}[t]

318:   \includegraphics[height=58mm]{./Energy_Scaling.eps}

319:   \caption{

320:     (Color online) Finite-size scaling of the energy splitting between

321:     the topological sectors around the quantum phase transition for

322:     $\beta = 10 L$.  At the critical loop tension (crosses) we find

323:     power-law scaling $\delta\! E(L) \propto L^{2-z}$ with exponent $z

324:     = 1.42 \pm 0.02$.  }

325:   \label{Fig:EnergyScaling}

326:   \vspace{-0.75mm}

327: \end{figure}

328: 

329: % Confinement transition for "electric charges" ---------------------------------------

330: 

331: \paragraph{Confinement transition --} For the loop gas Hamiltonian

332: ($\ref{Eq:ToricCode}$) the elementary electric charge excitations (end-points

333: of an open loop) are deconfined. For strong

334: loop tension, however these excitations are expected to become

335: confined, thereby eliminating all open loops. We can study this

336: confinement transition in our simulations of model (\ref{Eq:3DIsing})

337: by artificially introducing pairs of electric charge excitations for

338: the sampled loop configurations. This allows us to measure the

339: confinement length $\xi_c$ as the square root of the average second

340: moment of the distance between the two excitations, which for a torus

341: with even extent $L$ has to be normalized by a factor $6/(L^2+2)$.

342: The measured confinement length $\xi_c$ shown in

343: Fig.~\ref{Fig:Confinement} clearly demonstrates that electric charges

344: remain deconfined for the full extent of the topological phase and

345: that the confinement transitions occurs simultaneously with the

346: magnetic phase transition. At the critical loop tension the

347: confinement lengths $\xi_c(L)$ for various system sizes cross which

348: demonstrates that the confinement length diverges with the \emph{same}

349: critical exponents as the magnetic correlation length $\xi$ and there

350: is only one length scale describing the phase transition.  This can be

351: seen from the following finite-size scaling argument: We may write the

352: finite-size scaling behavior for the confinement length as

353: $\xi_c(\tau, L) \propto \tau^{-\nu} f(L \tau^{\chi})$ where $\nu$

354: ($\chi$) is the critical exponent of the correlation (confinement)

355: length respectively and $\tau = h-h_c$.  With

356: $\tilde{f}(L^{1/\chi}\tau) = (L^{1/\chi}\tau)^{-\nu} f(L \tau^\chi)$

357: we obtain for the confinement length $\xi_c(\tau, L) \propto L^{\nu /

358:   \chi} \tilde{f} (L^{1/\chi} \tau)$.  At the phase transition the

359: existence of a crossing point $\xi_c(0, L)/L \propto L^{\nu / \chi -

360:   1} = {\rm const}$ then implies $\chi = \nu$.  For our model without

361: dynamical electric charges, this measure of the confinement of test

362: charges is closely related to the calculation of a Wilson loop

363: expectation value.

364: In the presence of dynamical electric charges, this becomes trickier;

365: Polyakov loops have been used as an order parameter

366: for the {\em finite} temperature transition of the

367: 3D Ising gauge model\cite{LGT}.

368: \begin{figure}[t]

369:   \includegraphics[height=58mm]{./Confinement.eps}

370:   \caption{

371:     (Color online) Confinement length of two electric charge

372: excitations versus loop tension. The confinement transition occurs at

373: the same critical loop tension (dashed line) as the magnetic

374: transition.

375:   }

376:   \label{Fig:Confinement}

377: \end{figure}

378: 

379: 

380:  % Dissipation -------------------------------------------------------------

381: 

382: \paragraph{Dissipation --}

383: Finally, we discuss the effect of dissipation when Hamiltonian

384: (\ref{Eq:ToricCode}) is coupled to an Ohmic heat bath. Since our model

385: excludes dynamical electric charges, we do not consider coupling a

386: heat bath directly to $\sigma^x_i$.

387: Instead, we first examine coupling a heat bath to $\mu^x_p$ so that a 

388: `phonon' is created when a plaquette flips.

389: This type of dissipation could occur in a Josephson junction model

390: \cite{Schoen:90} or in a spin model through the spin-phonon coupling.

391: The standard procedure \cite{Caldeira} for a linear spectral density

392: (`Ohmic' dissipation) results in an effective action

393: for independent Ising chains with long-range couplings in an external

394: \emph{longitudinal} magnetic field (which here means parallel to

395: $\mu^x$).  We note that as a consequence of the Lee--Yang theorem

396: \cite{Lee:52}, there can be \emph{no singularities} of the respective

397: partition function at any real non-zero 

398: longitudinal field, ruling out the existence of a quantum phase

399: transition for this model. In particular, this implies that the magnetic gap

400: remains finite for any dissipation strength!

401: 

402: An entirely different behavior arises if dissipation is coupled such

403: that it stabilizes the `classical' state of the system.  Coupling the

404: bath to either $\sigma^z_i$ or ${\mu^z_p}$ stabilizes the classical

405: state of a single spin or a plaquette, respectively. We consider the

406: latter as it should be more effective at damping quantum fluctuations,

407: although we expect the former to have similar physics.

408: The same procedure as above then leads to a model for

409: decoupled Ising chains given by

410: \begin{multline}

411: \mathcal{H}_{\text{cl}} 

412: = - K_{\tau} \sum_{\tau, p} S_p(\tau)

413: S_p(\tau+\Delta\tau)

414: \\

415: -\frac{\alpha}{2}\sum_{\tau<\tau',p}

416: \Big(\frac{\pi}{N_\tau}\Big)^2\frac{S_p(\tau)S_p(\tau')}{\sin^{2}(\frac{

417: \pi}{N_\tau}|\tau-\tau'|)} \,,

418: \label{eq:dissip_action_z}

419: \end{multline}

420: where the parameter $\alpha$ measures the dissipation strength.

421: This model has been well studied \cite{Dissipation} and is known to

422: exhibit a Thouless-type phase transition into a classically

423: disordered, fluctuationless phase.  The critical value $\alpha_c$ of

424: this transition depends weakly on the cutoff in the long-range

425: interaction; in our simulations $\alpha_c\approx 0.7$.

426: At the transition the magnetic gap vanishes, in sharp contrast to the

427: previous case.

428: 

429: \begin{figure}[t]

430:   \includegraphics[height=58mm]{./Dissipation.eps}

431:   \caption{

432:     (Color online) Gap and error probability versus dissipation

433:     strength estimated from the correlation time $\xi_\tau$ of the

434:     dissipative Ising chain (\ref{eq:dissip_action_z}).

435:    }

436:   \label{Fig:Dissipation}

437:   \vspace{-0.5mm}

438: \end{figure}

439: 

440: Due to the long range interactions introduced by the dissipative

441: coupling, the spin-spin correlations will asymptotically decay as

442: $1/\tau^2$ \cite{Griffiths:67}. It is therefore a non-trivial task to

443: define a correlation-time and hence to estimate the excitation gap.

444: For $\alpha\lesssim 0.1$ one observes an exponential decay of the

445: correlation function onto the asymptotic $1/\tau^2$ behavior, and

446: subtracting the $1/\tau^2$-contribution allows one to estimate

447: $\xi_\tau$, which is found to grow linearly with $\alpha$. For

448: $\alpha>0.1$, this procedure can no longer be used. In this region we

449: estimate $\xi_\tau$ from the asymptotic decay of the correlations,

450: proportional to $(\xi_\tau/\tau)^2$. These $\xi_\tau$s grow

451: approximately exponentially in the region $0.1\lesssim \alpha

452: \lesssim \alpha_c$. Alternatively, one could define a correlation-time

453: from the crossover scale where the short-time behavior crosses over to

454: the asymptotic $1/\tau^2$ form. This definition in the spirit of a

455: `Josephson length' \cite{Chakravarty:95} yields (up to a normalization

456: factor) the same results as the $\xi_\tau$ extracted from the

457: asymptotic decay.  The gap estimated from the inverse correlation

458: function, as well as the error probability is plotted in

459: Fig.~\ref{Fig:Dissipation}. We find that the error probability remains

460: negligibly small below the crossover value $\alpha\approx 0.1$.

461: 

462: 

463: % Summary / Conclusions ----------------------------------------------------

464: 

465: \paragraph{Outlook --} 

466: We have shown that the topological phase which governs the

467: toric code model \cite{ToricCode} actually exists in an extended

468: region of phase space around the soluble point. It is stable against

469: deviations of the system Hamiltonian from the ideal one and also the

470: coupling of the system to its environment.  In general, this

471: demonstrates that a system does not necessarily have to be

472: particularly fine-tuned to reach a topological phase. The paucity of

473: their experimental observations may thus be due not to some intrinsic

474: delicateness of such phases, but rather to the experimental subtlety

475: involved in identifying them. In future work, these conclusions need

476: to be tested in other, more exotic topological phases which support universal

477: topological quantum computation \cite{Freedman01}.

478: 

479: 

480: % Acknowledgments ---------------------------------------------------------

481: 

482: We thank E. Ardonne, L. Balents, S. Chakravarty, and A. Kitaev for

483: stimulating discussions.

484: 

485: % References ---------------------------------------------------------

486: 

487: \begin{thebibliography}{99}

488: 

489: % Toric Code

490: \bibitem{ToricCode}

491: A. Yu. Kitaev, Ann. Phys. {\bf 303}, 2 (2003).

492: 

493: \bibitem{magnets}

494: R. Moessner and S. L. Sondhi, Phys. Rev. Lett {\bf 86}, 1881 (2001);

495: C. Nayak and K. Shtengel, Phys. Rev. B {\bf 64}, 064422 (2001);

496: L. Balents, M.~P.~A. Fisher, and S. Girvin, Phys. Rev. B {\bf 65}, 224412 (2002);

497: M. Freedman, {\it et al.} Ann. Phys. {\bf 310}, 428 (2004);

498: M. A. Levin and X.-G. Wen, Phys. Rev. B {\bf 71}, 045110 (2005);

499: P. Fendley and E. Fradkin, Phys. Rev. B {\bf 72}, 024412 (2005).

500: 

501: \bibitem{Ioffe02} L. B. Ioffe {\it et al.}, Nature {\bf 415}, 503 (2002).

502: 

503: \bibitem{Motrunich02}  O.~I.~Motrunich and T.~Senthil,

504:  Phys. Rev. Lett. \textbf{89}, 277004 (2002).

505: 

506: \bibitem{ColdAtoms}

507: %\bibitem{Duan03} 

508: L.-M.~Duan, E.~Demler, and M.~D.~Lukin, \prl {\bf 91}, 090402 (2003);

509: %\bibitem{Gurarie05}

510: V.~Gurarie, L.~Radzihovsky, and A.~V.~Andreev, \prl {\bf 94}, 230403 (2005).

511: 

512: \bibitem{DimerModel} 

513: D.~S.~Rokhsar and S.~A.~Kivelson, \prl {\bf 61}, 2376 (1988).

514: 

515: \bibitem{Leggett87} 

516: A.~Leggett {\it at al.}, \rmp {\bf 59}, 1 (1987).

517: 

518: \bibitem{Werner04} P.~Werner {\it et al.}, \prl {\bf 94}, 047201 (2005);

519: P.~Werner {\it et al.}, J.\ Phys.\ Soc.\ Jpn.\ Suppl.\ {\bf 74}, 67 (2005).

520: 

521: \bibitem{Chamon05} 

522: C.~Chamon, \prl \textbf{94}, 040402 (2005).

523: 

524: \bibitem{Wang03} 

525: C.~Wang, J.~Harrington, and J.~Preskill, 

526: Ann. Phys. \textbf{303}, 31 (2003).

527: 

528: \bibitem{IGT}

529: R. Moessner, S. L. Sondhi, and E. Fradkin, 

530: \prb {\bf 65}, 024504 (2002).

531: 

532: % Transverse Field Ising Model

533: \bibitem{TIM}

534: H. W. J. Bl\"ote and Y. Deng, \pre {\bf 66}, 066110 (2002).

535: 

536: % Universality 3D Ising model

537: \bibitem{Ferrenberg:91}

538: A. M. Ferrenberg and D. P. Landau, \prb {\bf 44}, 5081 (1991).

539: 

540: \bibitem{ContinuousPhaseTransition}

541: E. Ardonne, P. Fendley, and E. Fradkin, Ann. Phys. {\bf 310}, 493 (2004).

542: 

543: % ALPS

544: \bibitem{ALPS}

545: F. Alet  \emph{et al.},  J. Phys. Soc. Jpn. Suppl. {\bf 74}, 30 (2005). See also \url{http://alps.comp-phys.org}.

546: 

547: % Looper code

548: \bibitem{LooperCode}

549: S. Todo and K. Kato, \prl {\bf 87}, 047203 (2001).

550: 

551: % Lattice gauge theories, Polyakov loops, etc.

552: \bibitem{LGT}

553: M. Caselle, M. Hasenbusch, and M. Panero, JHEP {\bf 01}, 057 (2003).

554: 

555: % Dissipation

556: \bibitem{Schoen:90}

557: G. Sch\"on and A. D. Zaikin, Phys. Rep. {\bf 198}, 237 (1990).

558: 

559: \bibitem{Caldeira} 

560: A. O. Caldeira and A. J. Leggett, Ann. Phys. \textbf{149} 374 (1983).

561: 

562: \bibitem{Lee:52} 

563: T. D. Lee and C. N. Yang,  Phys. Rev. {\bf 87}, 410 (1952).

564: 

565: \bibitem{Dissipation}

566: D. J. Thouless, Phys. Rev. {\bf 187}, 732 (1969);

567: J. Fr\"ohlich and T. Spencer, Commun. Math. Phys. {\bf 84}, 87 (1982);

568: J. Bhattacharjee {\it et al.}, \prb  {\bf 24}, 3862 (1981);

569: E. Luijten and H. Messingfeld, \prl {\bf 86}, 5305 (2001).

570: 

571: \bibitem{Griffiths:67}

572: R. B. Griffiths, J. Math. Phys. {\bf 8}, 478 (1967).

573: 

574: \bibitem{Chakravarty:95}

575: S. Chakravarty and J. Rudnick, \prl {\bf 75} 501 (1995).

576: 

577: \bibitem{Freedman01}

578: M.~H. Freedman, Found. Comput. Math. {\bf 1},  183  (2001), and references therein.

579: \end{thebibliography}

580: 

581: \end{document}

582: 

583: