1: \documentclass[prb,aps,twocolumn,showpacs]{revtex4}
2:
3: \usepackage{epsfig}
4: \usepackage{bm}
5:
6: \newcommand{\beq}{\begin{equation}}
7: \newcommand{\eeq}{\end{equation}}
8: \newcommand{\bea}{\begin{eqnarray}}
9: \newcommand{\eea}{\end{eqnarray}}
10:
11: \def\material{{Cs$_2$CuCl$_4$\ }}
12:
13: %\def\tit#1#2#3#4#5{{\em #5}, {#1}{\bf #2}, #3 (#4)}
14: \def\tit#1#2#3#4#5{{#1}{\bf #2}, #3 (#4)}
15:
16: \def\npb{Nucl.\ Phys.\ B\ }
17: \def\plb{Phys.\ Lett.\ B\ }
18: \def\prl{Phys.\ Rev.\ Lett.\ }
19: \def\prb{Phys.\ Rev.\ B\ }
20: \def\jpco{J.\ Phys.\ Cond.\ Mat.\ }
21: \def\ijmpb{Int.\ J. Mod.\ Phys.\ B\ }
22: \def\pmag{Philos.\ Mag.\ }
23: \def\sci{Science\ }
24:
25: \def\etal{{\em et al.}}
26: %\def\vecb#1{{\mathbf #1}}
27: %\def\vecb#1{{\bm{#1}}}
28: \def\vecb#1{{\bf #1}}
29:
30: \begin{document}
31:
32: \title{Ordering in \material: Is there a proximate spin liquid?}
33:
34: \author{S. V. Isakov$^1$, T. Senthil$^{2,3}$, and Yong Baek Kim,$^1$}
35:
36: \affiliation{$^1$Department of Physics, University of Toronto,
37: Toronto, Ontario, Canada M5S 1A7\\
38: $^2$ Center for Condensed Matter Theory, Indian Institute of Science, Bangalore, India 560012\\
39: $^3$ Department of Physics, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139}
40:
41: \begin{abstract}
42: The layered spiral magnet \material displays several interesting properties
43: that have been suggested as evidence of proximity to a two dimensional quantum spin liquid.
44: In this paper we study a concrete version of this proposal and suggest
45: experiments that can potentially confirm it.
46: We study universal critical properties of two-dimensional frustrated
47: quantum magnets near the quantum phase transition between a spiral magnetic state
48: and a spin liquid state with gapped bosonic spinons in the framework of the $O(4)$ invariant critical
49: theory proposed earlier\cite{chubukov}. Direct numerical
50: calculation of the anomalous exponent in spin correlations shows
51: that the critical scattering has broad
52: continua qualitatively similar to experiment. More remarkably we show that the enlarged $O(4)$ symmetry leads
53: to the same slow power law decay for the vector spin chirality and the Neel correlations. We
54: show how this may be observed through polarized neutron scattering experiments. A number of
55: other less dramatic consequences of the critical theory are outlined as well.
56:
57:
58:
59: \end{abstract}
60:
61: \pacs{}
62:
63: \maketitle
64:
65: \section{Introduction}
66:
67: Despite the great amount of progress made in our theoretical
68: understanding of spin liquid phases\cite{frustreview,fazekasanderson, anderson} of quantum magnets,
69: there are to date no convincing experimental examples of such phases in spatial dimension $d \geq 2$. A particularly
70: promising candidate - much discussed recently - is the material $Cs_2CuCl_4$. This is a layered Mott insulator
71: whose spin physics is accurately modelled by a spin-$1/2$ Heisenberg antiferromagnet on an anisotropic
72: triangular lattice. Neutron scattering measurements have revealed several unusual properties of the
73: spin excitations\cite{coldeaneu} which have motivated intensive theoretical research works
74: \cite{weihong,chungln,bocquet,chungsp,veillette}.
75: There is incommensurate spiral magnetic order at low temperatures
76: (below 0.62K). The low energy
77: excitations of this ordered state are spin waves. However spin wave theory fails to quantitatively
78: describe the spin wave line width seen in inelastic neutron scattering. In particular there are long tails
79: that extend to reasonably high frequencies in the inelastic scattering spectrum. This structure may be understood as
80: indicative of a broad continuum of excited states. The continuum survives upon heating
81: above the magnetic phase transition (though the spin wave itself does not). In addition the phase diagram in an
82: external magnetic field shows several interesting features.
83:
84:
85: The broad continuum is reminiscent of spinon excitations in one-dimensional Heisenberg antiferromagnetic
86: chain. However the continuum disperses in both spatial directions so that a two dimensional description
87: of the physics may be more appropriate.
88: A number of workers have therefore suggested interpretations of the neutron data in terms of
89: fractional spin (spinon) excitations in two dimensions
90: into which the magnons decay. In particular Coldea {\it et al.}\cite{coldeaneu} suggested that
91: the material may be viewed as being close to a quantum phase transition between a spiral Neel state and a spin liquid state.
92: The purpose of the present paper is to explore this possibility in greater detail
93: and to suggest concrete experimental tests
94: to confirm (or rule out) this proposal.
95:
96: We begin with some general considerations. It is by now established that a number of different kinds of spin
97: liquid phases are theoretically possible in two dimensional quantum magnets. We will restrict attention to a
98: particular kind of spin liquid phase whose excitations consist of {\em bosonic} spin-$1/2$ spinons with a full
99: spin gap. In addition there are gapped $Z_2$ vortices (visons) which act as sources of $\pi$ flux for the spinons.
100: As shown many years ago in Ref.~\onlinecite{chubukov} such a spin liquid phase admits a direct second order
101: transition to the spiral Neel state which is simply driven by
102: condensation of the bosonic spinons. Specifically the transition was argued to be in the universality class
103: of the classical $O(4)$ fixed point in three dimensions.
104: Here we point out several remarkable consequences of this theory which may be used
105: to test its applicability to \material or other materials. Perhaps most interestingly we show that the
106: large extra $O(4)$ symmetry that emerges at the critical fixed point unifies seemingly different
107: competing orders. The Neel vector correlations have the same slow power law decay as the vector spin chirality.
108: This non-trivial prediction can potentially be probed in polarized neutron scattering experiments.
109:
110:
111:
112:
113:
114: In passing we note that other more exotic spin liquid phases with (possibly gapless) fermionic spinons
115: could potentially exist as stable phases\cite{wen91,bfn,z2long,stableu1}
116: but the transitions to the Neel state have not been studied.
117: The corresponding critical theories are also likely to be more exotic - we will therefore defer consideration
118: of such spin liquid phases and the corresponding critical points for the future and focus here on the
119: simpler case with gapped bosonic spinons.
120:
121:
122:
123: A controlled calculation in which both the magnetically ordered state and such a
124: gapped spin liquid state appear is provided by considering
125: a large-$N$ generalization of the $S=1/2$
126: Heisenberg model \cite{chungln} on the anisotropic triangular lattice.
127: Specifically, the large-$N$ limit of a bosonic Sp($N$)
128: Heisenberg model \cite{readsachdev,sachdevfru} was used to
129: obtain the mean field phase diagram as a function
130: of the ratio of intra-plane anisotropic couplings, $J'/J$, and
131: the strength of quantum fluctuations, $\kappa=``2S"$,
132: where $\kappa$ plays the same role as $2S$ (= value of spin at each site) in the SU(2) limit.
133: It was found that, for a range of $J'/J $, the semiclassical limit
134: (large $\kappa$) leads to a spiral magnetic order.
135: As the quantum fluctuations become
136: stronger (small $\kappa$), a spin liquid state ($Z_2$ spin liquid)
137: with gapped deconfined bosonic spinons emerges.
138:
139:
140:
141:
142: The possibility that \material is proximate to the quantum critical point between the incommensurate spiral
143: and spin liquid states will affect the behavior at intermediate energy/length scales.
144: As mentioned above Chubukov \etal\ \cite{chubukov} showed quite generally that the transition was in the
145: universality class of the $O(4)$ fixed point in
146: three euclidean dimensions. This is much higher symmetry than in the
147: original microscopic model. This enlarged $O(4)$ symmetry acts naturally on spinon degrees
148: of freedom which thus emerge as the useful variables already at the transition to the
149: spin liquid. The theory predicts a large anomalous exponent $\bar{\eta}$
150: for the spin-spin correlation function since the spins
151: are composite operators in terms of spinon operators.
152: Such composite operators usually have large anomalous
153: dimensions. A qualitative physical picture is simply that the spin-$1$ magnons decay rapidly
154: into the spinons thereby leading to broad lineshapes in the inelastic
155: neutron scattering. However right at the critical point the spinons are not free particles.
156: From the large-$N$ calculation extrapolated to the physical case of $SU(2)$ spins,
157: the anomalous exponent may be estimated to be 1.54\cite{chubukov}.
158: A more accurate value of $\bar{\eta}$ can be obtained directly by
159: classical Monte Carlo simulations of the $O(4)$ nonlinear
160: sigma model in three dimensions. In the present work and
161: in Ref.~\onlinecite{ballesteros}, it is found that $\bar{\eta}=1.373$.
162: It is worth noting that the neutron experiments\cite{coldeaneu} in \material were fit
163: to functional forms that also suggest large anomalous dimension $\bar{\eta}_{\text{E}}$
164: in the range 0.7 to 1. There is however reason to question these measurements
165: of the actual numerical value of $\bar{\eta}$ as we discuss in Sec~\ref{sec:experimental}. Nevertheless the large anomalous
166: dimension predicted by the theory is qualitatively consistent with the broad lineshapes seen in experiment.
167:
168: Thus it is certainly desirable to have other qualitatively distinct
169: predictions of the critical theory. It is our hope that the enlarged $O(4)$ symmetry
170: and the consequent enhanced
171: vector spin chirality correlations will provide such a
172: sharp test. We also outline some other less dramatic consequences of the extra $O(4)$ symmetry
173: that too may be useful.
174:
175: From a theoretical point of view our considerations are rather similar to those in a recent study\cite{su4} of
176: ``algebraic spin liquid'' phases where too the low energy theory is characterized by non-trivial enlarged
177: symmetry as compared to the microscopic model. There as in the present problem this enlarged symmetry
178: acts naturally on ``spinon'' degrees of freedom and leads to non-trivial relationships between the fluctuations of
179: rather different competing orders. Perhaps such phenomena are common in correlated systems.
180:
181:
182:
183:
184:
185: The rest of this paper is organized as follows.
186: In Sec.~\ref{sec:model}, we introduce the model that we use to
187: describe Cs$_2$CuCl$_4$. We briefly review the
188: $O(4)$ invariant critical theory in Sec.~\ref{sec:critical} and
189: discuss the anomalous exponent in Sec.~\ref{sec:anomalous}.
190: In Sec~\ref{sec:chirality}, we study the vector spin-chirality and
191: scalar-chirality correlation functions.
192: In Sec~\ref{sec:experimental}, we discuss possible experiments
193: to test our predictions and the critical theory. Finally, we summarize
194: our results and conclude in Sec~\ref{sec:conclusion}.
195:
196: \section{Model}
197: \label{sec:model}
198:
199: The magnetic ions Cu$^{2+}$ in \material carry $S=1/2$ spin
200: moments that reside on a stack of triangular layers \cite{coldeaneu}.
201: The intralayer antiferromagnetic interaction (the interaction in the
202: $b$-$c$ plane; see Fig.1) is anisotropic with two coupling constants:
203: $J \approx 0.375 meV$ along the $b$ direction and $J' \approx J/3$ along the zig-zag
204: bonds \cite{coldeaneu} (see Fig.~\ref{fig:tri}).
205: The interlayer interaction (the interaction along
206: the $a$ direction) is weak\cite{coldeaneu} with a coupling
207: constant $J''=0.045J\ll J$.
208: Therefore the system is quasi two-dimensional and can be
209: modeled by a two-dimensional frustrated Heisenberg Hamiltonian
210: on the triangular lattice. The main part of the Hamiltonian reads
211: \beq
212: H = J\sum_{\langle i,j \rangle} \vecb{S}_{i} \cdot \vecb{S}_j
213: + J'\sum_{\langle \langle i,j \rangle \rangle} \vecb{S}_{i} \cdot \vecb{S}_j.
214: \label{eq:ham}
215: \eeq
216: Here $\vecb{S}_i$ are spin-$1/2$ operators at the sites $i$ of a two dimensional triangular lattice.
217: The first sum runs over bonds in the $b$ direction and the
218: second sum runs over the zig-zag bonds. There is also a weak
219: Dzyaloshinskii-Moriya interaction \cite{coldeaham}.
220: We will neglect this term when we consider the critical theory
221: and discuss later its role in experiments.
222:
223: \begin{figure}[ht]
224: {
225: \centerline{\includegraphics[angle=0, width=2.0in]{tri.eps}}
226: \caption{
227: Schematic representation of the triangular lattice in the $b$-$c$ plane.
228: The $a$ direction is perpendicular to the plane.
229: There are two coupling constants; $J$ along the horizontal chains
230: and $J'$ along the zig-zag bonds.
231: }
232: \label{fig:tri}
233: }
234: \end{figure}
235:
236: The weak interlayer interaction stabilizes magnetic long range order below
237: $T=0.62$K \cite{coldeaold}. The order is incommensurate because
238: of the frustrated anisotropic interactions in the triangular planes
239: and occurs at an incommensurate wave vector
240: $\vecb{Q}=(0.5+\epsilon_{0})\vecb{b}^*$ with
241: $\epsilon_{0}=0.030(2)$ and $\vecb{b}^*=(2\pi/b,0,0)$.
242: The corresponding classical ordering wavevector is
243: $\vecb{Q}=(0.5+\epsilon_{c})\vecb{b}^*$ with
244: $\epsilon_{c}=(1/\pi)\arcsin(J'/2J)=0.053$. The substantial difference
245: is due to large quantum fluctuations that renormalize the ordering
246: wave vector.
247: %Classically, there is another independent ordering wavevector at
248: %$\vecb{Q}^-=(0.5+\epsilon_{c})\vecb{b}^*$ in the model given by
249: %Eq.~\ref{eq:ham}. However, the wavevector $\vecb{Q}^+$ is pinned by the
250: %Dzyaloshinskii-Moriya interaction\cite{veillette}
251: %(*is that correct?).
252:
253: \subsection{Order parameter}
254: The spiral ordering pattern may be represented using two orthonormal unit
255: vectors $\vecb{n}_1$ and $\vecb{n}_2$ through:
256: \beq
257: \langle \vecb{S}(\vecb{r}_i) \rangle \propto
258: \vecb{n}_1 \cos(\vecb{Q}\cdot\vecb{r}_i)
259: + \vecb{n}_2 \sin(\vecb{Q}\cdot\vecb{r}_i),
260: \label{eq:ordpatt}
261: \eeq
262: where $\vecb{n}_1$ and $\vecb{n}_2$ satisfy
263: \beq
264: \vecb{n}_1\cdot\vecb{n}_2=0, \quad \vecb{n}_1^2=\vecb{n}_2^2=1.
265: \label{eq:constr}
266: \eeq
267:
268: The $\vecb{n}_{1,2}$ are clearly vectors under global spin rotation, and together they define the
269: order parameter for the spiral state. In addition to global spin rotations the
270: spin Hamiltonian Eq.~\ref{eq:ham} is invariant under various lattice symmetries and
271: time reversal. It is useful and straight forward to work out the transformation properties
272: of $\vecb{n}_{1,2}$ under these symmetries. We find
273: \bea
274: T_{\vecb{\hat{l}}} &:& (\vecb{n}_1+i\vecb{n}_2)\rightarrow
275: (\vecb{n}_1+i\vecb{n}_2)e^{i\vecb{Q} \cdot \vecb{\hat{l}}}, \nonumber \\
276: R_c,I &:& (\vecb{n}_1+i\vecb{n}_2)\rightarrow(\vecb{n}_1-i\vecb{n}_2),
277: \nonumber \\
278: T &:& (\vecb{n}_1+i\vecb{n}_2)\rightarrow -(\vecb{n}_1-i\vecb{n}_2).
279: \label{eq:nlatsym}
280: \eea
281: Here $T_\vecb{\hat{l}}$ is a unit translation along a lattice vector $\vecb{l}$ direction, $R_c$ is reflection about the $c$ axis,
282: $I$ is lattice inversion, and $T$ is time reversal. Other symmetry operations may be obtained as combinations
283: of the above.
284:
285:
286: \section{Theory of quantum transition}
287: \label{sec:critical}
288:
289: %Neutron scattering measurements on \material\ reveal an extended continuum
290: %of exited states that indicates possible presence of deconfined spinons. The
291: %results of Ref.~\onlinecite{chungsp} suggest that \material\ might be close to
292: %a quantum phase transition from an incommensurate spiral ordered state
293: %to a spin liquid state. Therefore some universal features of
294: %the system found in neutron scattering experiments may be described by
295: %the critical theory of a quantum phase transition between incommensurate
296: %ordered and spin-liquid phases. In this section, we review such a theory,
297: %discuss its consequences for the anomalous exponent $\bar{\eta}$, and
298: %present our Monte Carlo calculation of the exponent $\bar{\eta}$.
299:
300: %\subsection{Review of the critical theory}
301: In this section we briefly review the theory of Ref.~\onlinecite{chubukov} for the quantum transition between the Neel
302: and spin liquid states. We will also use this as an opportunity to clarify several possible confusions with other
303: superficially similar results in the literature.
304:
305: The set $\vecb{n}_1, \vecb{n}_2$ defines the order parameter for the spiral state. To study phase transitions
306: out of the spiral
307: state it is necessary to allow fluctuations where $\vecb{n}_{1,2}$ vary slowly as
308: a function of space and time. The symmetry properties in Eq.~\ref{eq:nlatsym} actually imply
309: that the resulting continuum theory has an extra $U(1)$ symmetry over and above that of spin rotations. This may be seen
310: by considering translations along the $\vecb{b}$ direction. We have
311: \begin{equation}
312: (\vecb{n}_1+i\vecb{n}_2)\rightarrow(\vecb{n}_1+i\vecb{n}_2)
313: e^{imQ_x}
314: \end{equation}
315: for a translation by $m$ lattice sites. With incommensurate $Q_x$, this effectively translates into a
316: full $U(1)$ symmetry that rotates between the $\vecb{n}_1$ and $\vecb{n}_2$ fields.
317:
318:
319:
320:
321: The order parameter manifold defined by $\vecb{n}_{1,2}$ allows
322: for topological vortex defects with a discrete $Z_2$ character\cite{z2}. These $Z_2$ vortices
323: have energy logarithmic in the system size
324: in the ordered state due to the long distance distortion of the order parameter.
325: Now consider disordering the spiral order while keeping the core energy of
326: these $Z_2$ vortices finite. Ref. \onlinecite{chubukov} argues that the resulting state is a fractionalized spin liquid
327: with bosonic spin-$1/2$ spinons. The $Z_2$ vortices survive into the spin liquid phase but now only cost
328: finite (not divergent) energy.
329:
330: A theory for this transition is obtained by writing
331: \begin{equation}
332: \vecb{n}_1 + i\vecb{n}_2 = \epsilon_{\alpha \gamma} z_{\gamma} \boldsymbol{\sigma}_{\alpha \beta} z_{\beta},
333: \label{eq:param}
334: \end{equation}
335: where $\epsilon$ is the antisymmetric tensor, $\sigma^a$ are the Pauli
336: matrices, and $z_{\alpha} = (z_\uparrow, z_\downarrow)$ a two component complex unit vector satisfying
337: \begin{equation}
338: z^{\dagger}z = 1.
339: \end{equation}
340: It is easy to check that the parametrization (\ref{eq:param}) satisfies
341: the constraints given by Eq.~\ref{eq:constr}.
342: The $z_{\alpha}$ transform as spinors under the $SU(2)$ spin rotation and describe
343: spin-$1/2$ spinons. This representation clearly has a $Z_2$ gauge redundancy associated with changing the
344: sign of $z$ at any point in space-time.
345: Thus a reformulation in terms of $z$ may be fruitfully viewed as a theory of the $z$ fields
346: coupled to a $Z_2$ gauge field. The corresponding $Z_2$ gauge flux is associated with the $Z_2$
347: vortices discussed above. These vortices stay gapped at the transition between the Neel and spin liquid states
348: and may hence be ignored for a low energy description of the critical point.
349: Thus we may obtain the critical behavior by focusing only on the spinons and ignoring their coupling to
350: the $Z_2$ gauge field. Detailed arguments show that the critical theory is in fact in the $O(4)$ universality class
351: in $D = 2+1$ dimensions where there is full rotational symmetry between the four real numbers described by
352: $(z_{\uparrow}, z_{\downarrow})$. In effect the extra $U(1)$ symmetry of rotation between
353: $\vecb{n}_1$ and $\vecb{n}_2$ has been enlarged to $SU(2)$. Combined with the $SU(2)$ spin rotations
354: the full symmetry is
355: $SU(2) \times SU(2) \sim O(4)$. Thus the critical properties may be computed using the Euclidean action
356: \beq
357: {\cal S }=\int \text{d}^2 x \, \text{d}\tau \sum_{\mu=x,\tau}
358: \frac{1}{g}
359: \partial_{\mu} z_{\alpha}^* \partial_{\mu} z_{\alpha}.
360: \eeq
361:
362:
363: The relationship between the $z$ and $\vecb{n}_{1,2}$ fields may also be
364: expressed in a different way that will be fruitful later.
365: Let us introduce an $SU(2)$ matrix $U$ built out of $z$:
366: \beq
367: U=\left(
368: \matrix{
369: z_{\uparrow} & z_{\downarrow}^* \cr
370: z_{\downarrow} & -z_{\uparrow}^* \cr
371: }
372: \right).
373: \eeq
374: Then $U$ satisfies
375: \begin{equation}
376: U^{\dagger} \sigma^a U = R^{ab} \sigma^b,
377: \end{equation}
378: where $R$ is a $3 \times 3$ rotation matrix. Clearly
379: \begin{equation}
380: R^{ab} = \frac{1}{2}tr\left(U^\dagger \sigma^a U \sigma^b \right).
381: \end{equation}
382: It is readily checked that $\vecb{n}_1$ and $\vecb{n}_2$ are the first and second column
383: respectively of this rotation matrix. The third column represents the unit vector
384: \begin{equation}
385: \vecb{n}_3 = \vecb{n}_1 \times \vecb{n}_2
386: \end{equation}
387: which is orthogonal to both $\vecb{n}_1$ and $\vecb{n}_2$. Symmetry under physical spin rotations correspond to
388: {\em left} multiplication of $U$ by an $SU(2)$ spin rotation matrix:
389: \begin{equation}
390: U \rightarrow VU
391: \end{equation}
392: with $V \in SU(2)$. The enlarged $O(4)$ symmetry implies that {\em right} multiplication
393: \begin{equation}
394: U \rightarrow UV
395: \end{equation}
396: is also a symmetry of the critical fixed point.
397:
398:
399: Several comments are in order on these results. Prior to Ref.~\onlinecite{chubukov}, analysis of a continuum
400: non-linear sigma model appropriate for non-collinear magnets had led to the suggestion of enlarged
401: $O(4)$ symmetry in $2+ \epsilon$ space-time dimensions\cite{azaria}. However numerical calculations on stacked triangular
402: lattices near their finite temperature ordering transition failed to observe the predicted $O(4)$
403: universality class\cite{kwmrrev}.
404: Rather the evidence supports a transition in the ``chiral'' universality class of Ref.~\onlinecite{kwmr,kwmrrev}.
405: How then are we to reconcile this with our claims about the quantum Neel - spin liquid transition being in the $O(4)$
406: universality class? The answer lies in the nature of the paramagnetic phase.
407: In the classical stacked triangular lattice
408: the natural paramagnetic phase is the trivial one that is smoothly connected to the high temperature limit.
409: The transition to this phase from the ordered state is obtained by proliferating the $Z_2$ vortices (which are
410: line defects in the three dimensional classical model). In principle an exotic paramagnetic phase is also possible
411: which would be the classical three dimensional analog of the spin liquid: this requires destroying the magnetic order
412: without proliferating the $Z_2$ vortices. This paramagnet will be topologically ordered and will be separated from
413: the trivial very high temperature paramagnet by a phase transition. In this classical context,
414: the arguments of Ref.~\onlinecite{chubukov} apply to the transition between the Neel state and
415: this topologically ordered
416: paramagnet which will indeed be in the $O(4)$ universality class. However neither this non-trivial paramagnet
417: nor the corresponding transition were apparently accessed in the numerical calculations\cite{kwmrrev}.
418: Equally it is hard
419: to decide which of the two possible paramagnetic phases were accessed in the $2 + \epsilon$ calculations of
420: Ref. \onlinecite{azaria}.
421: The distinction between the two phases is topological and hence sensitive to spatial dimension. This is hard to
422: disentangle in the $\epsilon$ expansion.
423:
424: Turning to the quantum problem at hand it differs from the classical stacked magnet in an important way. There are
425: extra Berry phases that are sensitive to the microscopic spin at each lattice site
426: ({\em i.e.} spin-$1/2$ or spin-$1$, etc). A close and familiar analogy is from the theory of
427: collinear quantum antiferromagnets in two dimensions where such Berry phases spoil any general direct mapping to
428: classical collinear magnets in one higher dimension. In that case in the semiclassical limit the Berry phases
429: are associated entirely with singular topological configurations known as hedgehogs.
430: A similar result also holds for the non-collinear quantum magnets of interest in this paper. The Berry phases
431: are associated entirely with the topological $Z_2$ vortex configurations. As these $Z_2$ vortices
432: are gapped across the
433: transition to the spin liquid the Berry phases play no role in the low energy universal critical physics. Indeed the
434: $O(4)$ universality class will describe the Neel - spin liquid transition for all spin-$S$ magnets
435: regardless of the value of $S$. On the other hand the phase obtained when the Neel state is
436: disordered by proliferating the
437: $Z_2$ vortices will be strongly influenced by the Berry phases. For spin-$1/2$ it is expected that the Berry phases will
438: lead to broken lattice symmetries in the resulting paramagnet\cite{sen}. The nature of the transition
439: between such a VBS ordered
440: paramagnet and the Neel state on the triangular lattice is not presently understood. Preliminary analysis suggests
441: that an interesting ``Landau-forbidden" second order transition may be possible. However in the present paper we
442: restrict ourselves to studying the transition to the spin liquid.
443:
444:
445:
446:
447:
448:
449:
450: \section{Consequences of critical theory}
451:
452: \subsection{Spin correlations}
453: \label{sec:anomalous}
454:
455: We begin by considering the spin-spin correlation function at the ordering wavevector
456: which is readily accessed in neutron scattering experiments.
457: This can be expressed in terms of the
458: correlation functions of the vectors $\vecb{n}_1$ and $\vecb{n}_2$
459: \beq
460: \langle \vecb{S}(\vecb{r},\tau) \cdot \vecb{S}(0,0) \rangle =
461: \langle \vecb{n}_1(\vecb{r},\tau) \cdot \vecb{n}_1(0,0) \rangle
462: \cos(\vecb{Q} \cdot \vecb{r}).
463: \label{eq:corr}
464: \eeq
465: We note that due to the symmetry of rotations between $\vecb{n}_1$ and $\vecb{n}_2$,
466: they will both have the same correlations.
467: Close to the critical point, the correlation function has the
468: usual power-law behavior
469: \beq
470: \langle \vecb{S}(-\vecb{q},\omega) \cdot \vecb{S}(\vecb{q},\omega)
471: \rangle \sim \frac{1}{(\omega^2-k^2)^{1-{\bar{\eta}}/2}},
472: \eeq
473: where ${\bf k}={\bf q}-{\bf Q}$.
474: Since $\vecb{n}_1$ and $\vecb{n}_2$ (and $\vecb{S}$) are composite
475: operators in terms of the $z_{\alpha}$ fields, one can expect that the
476: spin-spin correlation function has a large anomalous exponent $\bar{\eta}$.
477: Indeed, the large-$N$ calculation of $\bar{\eta}$ for an $O(2N)$ theory
478: gives \cite{chubukov}
479: \beq
480: \bar{\eta}=1+\frac{32}{3\pi^2N}.
481: \eeq
482: For the physical case $N=2$, we have $\bar{\eta}\approx1.54$.
483: This is a large number compared to anomalous exponents of
484: non-composite operators.
485:
486:
487:
488: To improve the large-$N$ estimate of the anomalous exponent, one can alternately
489: perform Monte Carlo simulations of the classical $O(4)$ nonlinear sigma model
490: in three dimensions since this model is in the same universality class as the
491: transition of interest. Ballesteros \etal\ \cite{ballesteros} measured the anomalous
492: exponent $\eta_{\text{T}}$ of the tensorial magnetization that is defined as
493: \bea
494: M_{\text{T}}&=& \left \langle \sqrt{\text{Tr}{\cal M}^2} \right \rangle, \nonumber \\
495: {\cal M}_{\alpha\beta} &=& \sum_i (\phi_{i\alpha} \phi_{i\beta}
496: -\frac{1}{4} \delta_{\alpha\beta}), \nonumber
497: \eea
498: where $\phi_{i\alpha}$ are four components of the $O(4)$ unity vector
499: at the site $i$ in three-dimensions.
500: They found $\eta_{\text{T}}=1.375(5)$. This result is directly related
501: to our case because of the following argument. Using Eq.~\ref{eq:param}
502: and writing $z_{\uparrow}=(\phi_1+i\phi_2), z_{\downarrow}=(\phi_3+i\phi_4)$,
503: where $\phi_{\alpha}$ are now real fields, we have
504: \bea
505: \vecb{n}_1&=&\left(
506: \matrix{
507: \phi_2^2+\phi_3^2-\phi_1^2-\phi_4^2 \cr
508: 2(\phi_1\phi_2+\phi_3\phi_4) \cr
509: 2(\phi_1\phi_3-\phi_2\phi_4) \cr
510: }
511: \right) \label{eq:n1} \\
512: \vecb{n}_2&=&\left(
513: \matrix{
514: 2(\phi_3\phi_4-\phi_1\phi_2) \cr
515: \phi_2^2+\phi_4^2-\phi_1^2-\phi_3^2 \cr
516: 2(\phi_1\phi_4+\phi_2\phi_3) \cr
517: }
518: \right).
519: \label{eq:n2}
520: \eea
521: We can see that the tensorial magnetization is related to $\vecb{n}_1$
522: and $\vecb{n}_2$ by symmetry and therefore the anomalous exponent of the
523: tensorial magnetization is equal to the anomalous exponent of the
524: vectors $\vecb{n}_1$ or $\vecb{n}_2$.
525: Then, from Eq.~\ref{eq:corr}, it is the same as the anomalous exponent
526: of the correlation function at wavevector $\vecb{Q}$.
527:
528: We have confirmed this conclusion by computing
529: the anomalous exponent of $\vecb{n}_1$ directly. We simulate the $O(4)$
530: nonlinear sigma model in three dimensions on the simple cubic lattice using
531: a cluster algorithm \cite{wolff}. We measure the anomalous exponent
532: $\bar{\eta}$ in the following way. We calculate the total $z$ component
533: $n_1^z=2(\phi_1\phi_3-\phi_2\phi_4)$ of the vector $\vecb{n}_1$
534: \beq
535: M=\sum_i n_{1i}^z,
536: \eeq
537: where the sum runs over all lattice sites. In our simulations, we measure
538: the ``magnetization'' and ``susceptibility``
539: \bea
540: \bar{m}&=&\frac{1}{L^3} \langle M \rangle,\nonumber \\
541: \bar{\chi}&=&\langle M^2 \rangle,\nonumber
542: \eea
543: where $L$ is the system size. We perform our simulations only at the
544: critical temperature found in Ref.~\onlinecite{ballesteros} with very high
545: accuracy, $\beta_c=1/T_c=0.935861(8)$. To find the critical exponents, we
546: use finite size scaling analysis. Close to the critical point, observables
547: satisfy the general scaling form
548: \beq
549: {\cal O}=L^{\rho/\nu} F_{\cal O}(tL^{1/\nu},L^{-\omega}),
550: \eeq
551: where ${\cal O}$ can be $\bar{m}$ or $\bar{\chi}$, $\rho$ is
552: the scaling exponent corresponding to the operator ${\cal O}$
553: ($\rho=-\bar{\beta}=-\nu(1+\bar{\eta})/2$ for $\bar{m}$ and
554: $\rho=\bar{\gamma}=\nu(2-\bar{\eta})$ for $\bar{\chi}$), $F_{\cal O}$ is a
555: universal function, $t=|T-T_c|$, and $\omega$ is a universal exponent
556: related to the leading irrelevant operator. Exactly at the critical point $t=0$,
557: one can write
558: \beq
559: {\cal O}=L^{\rho/\nu} (a+cL^{-\omega}).
560: \eeq
561: We assume that the correction to scaling is negligible for large lattice
562: sizes and fit to a simplified scaling form
563: \beq
564: {\cal O}=aL^{\rho/\nu}.
565: \eeq
566: We have performed simulations for large enough lattices (up to
567: $L_{\text{max}}=96$). Fitting from $L_{\text{min}}=20$ to $L_{\text{max}}$,
568: we obtain $\bar{\eta}=1.373(3)$.
569:
570: The large value of $\bar{\eta}$ is a reflection of the emergence of spinons as useful
571: degrees of freedom at the critical point.
572:
573:
574:
575: \subsection{Other competing orders}
576: \label{sec:chirality}
577:
578: The extra $O(4)$ symmetry actually has more striking consequences for the critical properties.
579: It implies that operators other than the natural magnetic order parameter will have enhanced
580: power law correlators. Consider the magnetic order parameters $\vecb{n}_{1,2}$. As discussed in
581: Section \ref{sec:critical}, they may be regarded as the first and second columns of a
582: rotation matrix $R$. The third column of the rotation matrix is
583: $\vecb{n}_3 =\vecb{n}_1 \times \vecb{n}_2$. The enlarged $O(4)$ symmetry at the critical point implies
584: that both left and right multiplications by an orthogonal matrix of this rotation matrix $R$
585: are symmetries of the critical theory.
586: Left multiplication is just physical spin rotation. However symmetry under right multiplication implies that
587: $\vecb{n}_{1,2,3}$ will all have the same correlations. In particular $\vecb{n}_3$ will have the same slow power law
588: decay as $\vecb{n}_{1,2}$ calculated in the previous subsection.
589: We can
590: express the vector $\vecb{n}_3$ in terms of the spin operator as follows.
591: Consider $\vecb{S}_{\vecb{r}+\vecb{\hat{l}}} \times \vecb{S}_{\vecb{r}}$, where
592: $\vecb{\hat{l}}$ is a unit lattice vector. Using (\ref{eq:ordpatt}) and expanding
593: the first spin operator around $\vecb{r}$, we have
594: \bea
595: \vecb{S}_{\vecb{r}+\vecb{\hat{l}}} \times \vecb{S}_{\vecb{r}} &\sim&
596: \vecb{n}_3 \sin(\vecb{Q}\cdot\hat{\vecb{l}}) \nonumber \\
597: &+& \text{derivative terms}.
598: \label{eq:sts}
599: \eea
600: We can ignore the derivatives terms since their contribution to the correlation
601: function is negligible. Consider the chirality vector
602: \beq
603: \vecb{C}_{\vecb{r}}=
604: \sum_{\vecb{r}'} \vecb{S}_{\vecb{r}} \times \vecb{S}_{\vecb{r}'},
605: \eeq
606: where the oriented sum is over an elementary plaquette. Using
607: Eq.~(\ref{eq:sts}), it is easy to show that
608: \beq
609: \vecb{C}=\left[\sin Q_x-2\sin \frac{Q_x}{2} \cos \frac{\sqrt{3}Q_y}{2}\right]
610: \vecb{n}_3 \propto \vecb{n}_3.
611: \eeq
612: Thus the vector spin-chirality correlation function
613: \bea
614: \langle \vecb{C}(\vecb{r},\tau) \cdot \vecb{C}(\vecb{0},0)\rangle &\propto&
615: \langle \vecb{n}_3(\vecb{r},\tau) \cdot \vecb{n}_3(\vecb{0},0)\rangle \nonumber \\
616: &=& \langle \vecb{n}_1(\vecb{r},\tau) \cdot \vecb{n}_1(\vecb{0},0)\rangle.
617: \eea
618: have the same power-law behavior at zero wavevector as those of spins
619: at the wavevector $\vecb{Q}$. This highly non-trivial statement possibly provides a
620: way to test the theory of Ref.~\onlinecite{chubukov} in
621: experiments and in model numerical calculations.
622:
623:
624:
625:
626:
627: \subsection{Conserved quantities}
628: Some less dramatic effects of the enlarged $O(4)$ symmetry are also worth noting.
629: The $O(4)$ group has six generators
630: and these will all be conserved at the critical fixed point. Three of these are just the conserved total spin
631: $\vecb{S}_{tot} = \sum _{i} \vecb{S}_i$ - they are the generators of left rotations of $U$. The remaining three
632: are conserved only in the low energy critical theory (though not in the original microscopic model). They correspond to
633: right rotations of $U$. We will denote these $K_a$, $ a = 1,2,3$.
634: The $K_a$ transform as vectors under the group
635: $SU(2)_R$ of right rotations of $U$. They generate rotations of $\vecb{n}_a$ amongst one another so that
636: \begin{equation}
637: \left[\vecb{n}_a, K_b \right] = i \epsilon_{abc} \vecb{n}_c.
638: \label{eq:comm}
639: \end{equation}
640:
641: As is well-known such conserved quantities have scaling dimension $d_{\rm scale} = 2$ with no anomalous
642: dimension\cite{footnote}.
643: %\footnote{In addition there will be twelve currents associated with these 6 conserved densities
644: %which will also have scaling dimension 2 with no anomalous dimension. Of these 6 are just the spin currents. The
645: %remaining 6 are currents of the $K^a$}.
646: Thus at the critical point at zero temperature
647: \begin{equation}
648: \langle K_a ({\bf r}, \tau)K_b({\bf 0},0) \rangle \sim \frac{\delta_{ab}}{(|{\bf r}|^2 + \tau^2)^2},
649: \end{equation}
650: where ${\bf r}$ is the spatial coordinate and $\tau$ is the (imaginary) time coordinate. A similar result also holds for
651: the conserved
652: total spin. Away from the critical point at finite temperature in the quantum critical region, the $K_a$
653: will continue to be approximately conserved. The exact conservation will be spoiled by irrelevant operators that
654: break the $O(4)$ symmetry down to $SU(2) \times lattice ~space ~group$. Thus at $T > 0$ in the quantum critical region
655: we expect that the $K_a$ will diffuse upto a long length/time scale which will be determined by
656: $T$ and the distance to the critical point. It is possible that the presence of such extra nearly diffusive modes
657: can also be looked for in experiments.
658:
659: But what do the $K_a$ correspond to in terms of the underlying spins? A useful guess for the answer
660: is provided by examining the transformation properties of $K_a$ under the microscopic symmetries of the
661: original lattice system. It is reasonable that any lattice operator that has the same transformation
662: properties will have some overlap in the continuum theory with the $K_a$. With the further assumption that
663: there are no other operators in the continuum theory with smaller scaling dimension that also have the same
664: transformation, the $K_a$ will give the dominant correlations of such lattice operators.
665: Similar considerations were also used in Ref.~\onlinecite{su4}.
666:
667:
668: The transformation properties of $K_a$ under all physical symmetries is fixed by the
669: commutation relation Eq.~\ref{eq:comm}.
670: The transformation properties of $K_a$ are related to the transformation
671: properties of the matrix $R$. First, we list the latter ones.
672: \bea
673: SU(2)_{\text{spin}} &:& R\rightarrow OR, \nonumber \\
674: T_{\vecb{\hat{l}}} &:&
675: R \rightarrow R
676: \left(
677: \matrix{
678: \cos(\vecb{Q}\cdot\vecb{\hat{l}}) &
679: \sin(\vecb{Q}\cdot\vecb{\hat{l}}) & 0 \cr
680: -\sin(\vecb{Q}\cdot\vecb{\hat{l}}) &
681: \cos(\vecb{Q}\cdot\vecb{\hat{l}}) & 0 \cr
682: 0 & 0 & 1
683: }
684: \right), \nonumber \\
685: R_c,I &:&
686: R \rightarrow R
687: \left(
688: \matrix{
689: 1 & 0 & 0 \cr
690: 0 & -1 & 0 \cr
691: 0 & 0 & -1
692: }
693: \right), \nonumber \\
694: T &:&
695: R \rightarrow R
696: \left(
697: \matrix{
698: -1 & 0 & 0 \cr
699: 0 & -1 & 0 \cr
700: 0 & 0 & 1
701: }
702: \right), \nonumber
703: \eea
704: where $O$ is an $O(3)$ matrix and $T_{\vecb{\hat{l}}}$ is the lattice
705: translation along the lattice vector $\vecb{\hat{l}}$. We can use these
706: transformations and the condition that $K_a$ are vectors under $SU(2)_R$ to
707: obtain the transformations of $K_a$
708: \bea
709: SU(2)_{\text{spin}} &:& (K_1,K_2,K_3)\rightarrow (K_1,K_2,K_3), \nonumber \\
710: T_{\vecb{\hat{l}}} &:& (K_1,K_2,K_3)\rightarrow (K_1,K_2,K_3)M, \nonumber \\
711: R_c,I &:& (K_1,K_2,K_3)\rightarrow (K_1,-K_2,-K_3), \nonumber \\
712: T &:& (K_1,K_2,K_3)\rightarrow (K_1,K_2,K_3), \nonumber
713: \eea
714: where
715: \beq
716: M=\left(\matrix{
717: \cos(\vecb{Q}\cdot\vecb{\hat{l}})
718: & -\sin(\vecb{Q}\cdot\vecb{\hat{l}}) & 0 \cr
719: \sin(\vecb{Q}\cdot\vecb{\hat{l}})
720: & \cos(\vecb{Q}\cdot\vecb{\hat{l}}) & 0 \cr
721: 0 & 0 & 1
722: }\right).
723: \eeq
724:
725: We can construct local spin operators that have the same symmetry properties
726: as $K_a$. We find that
727: \bea
728: K_1(\vecb{r}) &\sim& \cos (\vecb{Q}\cdot\vecb{r}) \vecb{S}(\vecb{r}) \cdot
729: [\vecb{S}(\vecb{r}-\vecb{\hat{x}}) + \vecb{S}(\vecb{r}+\vecb{\hat{x}})], \\
730: K_2(\vecb{r}) &\sim& \sin (\vecb{Q}\cdot\vecb{r}) \vecb{S}(\vecb{r}) \cdot
731: [\vecb{S}(\vecb{r}-\vecb{\hat{x}}) + \vecb{S}(\vecb{r}+\vecb{\hat{x}})], \\
732: K_3(\vecb{r}) &\sim& \vecb{S}(\vecb{r}-\vecb{\hat{x}}) \cdot
733: [\vecb{S}(\vecb{r}) \times \vecb{S}(\vecb{r}+\vecb{\hat{x}})].
734: \eea
735: Thus $K_1+iK_2$ transforms identically with the Fourier transform of the bond energy at wavevector
736: $\vecb{Q}$ while $K_3$ can be identified with the scalar spin-chirality
737: along the chains.
738:
739: \section{Experimental implications}
740: \label{sec:experimental}
741:
742: \subsection{Spin correlations}
743:
744: Quite generally in the quantum critical region the spin correlations at the ordering wavevector
745: must satisfy scaling - for instance as a function of $\omega/T$. \material orders at low temperature.
746: Consequently
747: data at the lowest temperatures
748: and frequencies must be excluded from the scaling analysis. Furthermore it is precisely the
749: very-low-$T$/low-$\omega$ range that will also be most sensitive to the presence of
750: weak spin anisotropies (such as Dzyaloshinski-Moriya interactions). However the data at intermediate $T$
751: and $\omega$ will be less affected by such spin anisotropies or by the low energy long range order.
752:
753:
754: The value of $\bar{\eta}$ found from Monte Carlo simulations
755: can be compared to $\bar{\eta}_E$ estimated in experiments
756: using a fitting procedure \cite{coldeaneu}. The experimental
757: values fall in a range between 0.7 and 1. Thus the Monte-Carlo
758: result is still larger than the experimentally estimated value.
759: It is important to note the following caveat on this comparison.
760: The existing measurements of $\bar{\eta}_E$ were performed in inelastic neutron scattering experiments
761: where {\em both the wavenumber and frequency transfer to the sample were simultaneously varied}.
762: In particular typical data sets (such as for instance scan G of Ref.~\onlinecite{coldeaneu}) involve
763: increasing the frequency transfer while the momentum transfer varies from points in the two-dimensional
764: Brillouin zone far away from the ordering wavevector $\vecb{Q}$ to points close to $\vecb{Q}$.
765: In the context of the ideas explored in this paper the frequency dependence at fixed $\vecb{q}$ far away from
766: $\vecb{Q}$ will not be dominated by the singular long distance critical fluctuations and will be highly
767: non-universal. In contrast the frequency dependence for $\vecb{q}$ close to $\vecb{Q}$ will be
768: determined by the long distance critical fluctuations and will be universal. It is these latter fluctuations
769: that are described by the calculated exponent $\bar{\eta}$. Thus the existing measurements of $\bar{\eta}_E$
770: are quite possibly severely contaminated by non-universal short distance effects. Hence the lack of
771: quantitative agreement between $\bar{\eta}$ and $\bar{\eta}_E$ is not surprising. However
772: the qualitative observation of broad magnon linewidths is consistent with the large $\bar{\eta}$
773: obtained in the theory. Future experiments will hopefully directly probe the fluctuations near the
774: ordering wavevector thereby allowing for quantitative comparison.
775:
776: As noted already in Ref.~\onlinecite{chubukov}, NMR experiments may be a useful way to directly
777: measure $\bar{\eta}$. For the nuclear relaxation rate we have
778: \begin{equation}
779: \frac{1}{T_1} \sim T^{\bar{\eta}}.
780: \end{equation}
781: Again this behavior will only obtain at intermediate temperatures.
782:
783:
784:
785: \subsection{Detection of vector spin-chirality fluctuations}
786: A dramatic consequence of the critical theory studied in this paper is the enhanced vector spin-chirality
787: correlations discussed in Section \ref{sec:chirality}. How can they be detected in experiments? In this section we
788: show how this may be done through
789: polarized neutron scattering. Our proposal builds on and generalizes the pioneering ideas of Maleyev\cite{Maleyev}
790: in the context of classical non-collinear magnets.
791:
792: The part of the neutron scattering rate $R_P$ that depends on the incoming neutron polarization $\vecb{P}_i$ is
793: given by
794: \begin{equation}
795: R_P(\vecb{k}_i, \vecb{P}_i \rightarrow \vecb{k}_f) \sim \left(\vecb{P}_i \cdot \hat{q} \right)
796: \left(\hat{q} \cdot \vecb{b} \right),
797: \end{equation}
798: where $\vecb{q} = \vecb{k}_f - \vecb{k}_i$ is the three-momentum transferred to the sample, and $\hat{q}$ is
799: the corresponding unit vector. The vector $\vecb{b}$ is given by
800: \begin{equation}
801: b_{\alpha} = \frac{\epsilon_{\alpha\beta\gamma}}{2i}S_{\beta \gamma},
802: \end{equation}
803: where $S_{\beta\gamma}$ is the spin structure factor. Thus the polarization dependent part
804: probes the {\em antisymmetric} part of the structure factor.
805: By the usual arguments it is clear that $b_{\alpha}$ can be obtained from a calculation of the imaginary time
806: Green function (here $0 < \tau < \beta=1/T$)
807: \begin{eqnarray}
808: \vecb{g}(\vecb{r},\tau) & = & \langle \vecb{S}(\vecb{r},\tau) \times \vecb{S}(\vecb{0},0) \rangle, \\
809: \vecb{g}(\vecb{q}, i\omega_n) & = & \int_{r\tau} e^{i\vecb{q} \cdot \vecb{r}
810: -i\omega_n \tau} \vecb{g}(\vecb{r},\tau).
811: \end{eqnarray}
812: Specifically we have
813: \begin{equation}
814: \vecb{b}(\vecb{q},\omega) = \frac{-1}{1-e^{-\beta\omega}}\left(\frac{
815: \vecb{g}(\vecb{q}, i\omega_n \rightarrow \omega+i0^+) -
816: \vecb{g}(\vecb{q}, \omega+i0^-)}{2} \right).
817: \end{equation}
818:
819: \begin{figure}[ht]
820: {
821: \centerline{\includegraphics[angle=0, width=2.0in]{dm.eps}}
822: \caption{
823: Dzyaloshinski-Moriya interaction is non-zero along oriented zigzag bonds
824: that are denoted by
825: vectors $\boldsymbol{\delta}_{1}$ and $\boldsymbol{\delta}_{2}$.
826: }
827: \label{fig:dm}
828: }
829: \end{figure}
830:
831: In the context of this paper using Eq.~\ref{eq:ordpatt} we find
832: \begin{eqnarray}
833: \vecb{g}(\vecb{r}, \tau) & = &
834: \langle \vecb{n}_1(\vecb{r}, \tau) \times \vecb{n}_1(\vecb{0},0) \rangle
835: \cos(\vecb{Q} \cdot \vecb{r}) \nonumber \\
836: & + & \langle \vecb{n}_2({\bf r}, \tau) \times \vecb{n}_1(\vecb{0},0) \rangle
837: \sin(\vecb{Q} \cdot \vecb{r}). \nonumber
838: \end{eqnarray}
839: Clearly such correlators are zero if there is full spin isotropy. In the specific case of \material
840: the presence of a weak Dzyaloshinski-Moriya(DM) interaction will lead to a non-zero result.
841: It is known that the DM interaction is non-zero along oriented zigzag bonds. For a single triangular layer
842: \begin{equation}
843: H_{DM} = -\vecb{D} \cdot \sum_{\vecb{r}}
844: \vecb{S}_{\vecb{r}} \times \left(\vecb{S}_{\vecb{r} + \boldsymbol{\delta}_1} +
845: \vecb{S}_{\vecb{r} + \boldsymbol{\delta}_2}\right).
846: \end{equation}
847: Here $\boldsymbol{\delta}_{1,2}$ are unit vectors along the oriented zigzag bonds as shown
848: in Fig.~\ref{fig:dm}.
849: In addition $H_{DM}$ is also staggered between different layers. The DM vector is oriented along the $a$ axis
850: and has magnitude $D \approx 0.02 meV \approx 0.05 J$. In the continuum theory we may again use
851: Eq.~\ref{eq:ordpatt} to write
852: \begin{eqnarray}
853: H_{DM} & \approx & -\vecb{d} \cdot \int d^2r \ \vecb{n}_1 \times \vecb{n_2}, \\
854: & = & -\vecb{d} \cdot \int d^2r \ \vecb{n}_3.
855: \end{eqnarray}
856: Here $\vecb{d} \propto \vecb{D}$. Let us now try to evaluate $\vecb{g}$ to linear order in $\vecb{d}$.
857: We get
858: \begin{eqnarray}
859: \vecb{g}(x) & = & \int d^3x' \langle \vecb{n}_1(x) \times \vecb{n}_1(0)
860: \left(\vecb{d} \cdot \vecb{n}_3(x')\right) \rangle \cos(\vecb{Q} \cdot \vecb{r}) \nonumber \\
861: & + & \int d^3x'\langle \left(\vecb{n}_2(x) \times \vecb{n}_1(0) \right)
862: \left(\vecb{d} \cdot \vecb{n}_3(x')\right) \rangle \sin(\vecb{Q} \cdot \vecb{r}). \nonumber
863: \end{eqnarray}
864: Here $x = (\vecb{r},\tau), x' = (\vecb{r'}, \tau')$ are space-time coordinates.
865: The averages are evaluated in the isotropic theory. The first average vanishes due to the symmetry
866: $\vecb{n}_3 \rightarrow -\vecb{n}_3, \vecb{n}_1 \rightarrow \vecb{n}_1$ present in the action.
867: We are therefore left with
868: \begin{equation}
869: \vecb{g}(x) = d_{\alpha}\int d^3x'\langle \vecb{n}_2(x) \times \vecb{n}_1(0)
870: n_{3\alpha}(x') \rangle \sin(\vecb{Q} \cdot \vecb{r}).
871: \end{equation}
872: Thus $\vecb{g}$ is determined by the three point correlation function of $\vecb{n}_{1,2,3}$. Using
873: spin rotation invariance we have
874: \begin{equation}
875: \vecb{g}(\vecb{r},\tau) = \frac{\vecb{d}}{3} \sin(\vecb{Q} \cdot \vecb{r})\int d^3x'
876: \langle \vecb{n}_2(x) \times \vecb{n}_1(0) \cdot \vecb{n}_{3}(x') \rangle .
877: \end{equation}
878: Such 3-point correlators of scaling fields are severely restricted by conformal invariance\cite{ginscft}.
879: As $\vecb{n}_{1,2,3}$ all have the same scaling dimension $\Delta = (1+\bar{\eta})/2$, we have (at $T = 0$)
880: \begin{equation}
881: \label{3ptcft}
882: \langle \vecb{n}_2(x) \times \vecb{n}_1(0) \cdot \vecb{n}_3(x') \rangle
883: \sim \frac{1}{x^{\Delta}x'^{\Delta}|x-x'|^{\Delta}}.
884: \end{equation}
885:
886: We now have to integrate over $x'$ which is the coordinate of
887: $\vecb{n}_3$. It is easy to see that this integral is infra-red(IR) divergent.
888: This is because at large $x'$, the integrand behaves as
889: $\frac{1}{x'^{2\Delta}}$.
890: With $2\Delta = 1 + \bar{\eta} \approx 2.37$, the integral over $x'$ will diverge in
891: the infra-red.
892: Formally this just means that perturbation theory in the DM interaction
893: diverges at $T = 0$ and the correct answer will involve some fractional
894: power of the DM vector.
895: However for our purposes it is much more meaningful and simpler to go to finite temperature
896: where this divergence
897: will be cut-off. Before doing that it is useful
898: to understand the origin of this divergence at $T = 0$ better.
899:
900: The divergence comes from large $x' \gg x$. To discuss this limit let us
901: keep $x'$ fixed and bring $x$ close to $0$. Then we need
902: the operator product expansion(OPE) of $\vecb{n}_2 \times \vecb{n}_1$. This is a vector in spin space
903: and the leading term will
904: just be $\vecb{n}_3$. Scaling requires the equation
905: \begin{equation}
906: \vecb{n}_2(x) \times \vecb{n}_1(0) \sim \frac{1}{x^{\Delta}} \vecb{n}_3 (x).
907: \end{equation}
908: There is only one power of $\Delta$ in the right hand side so that both sides will scale
909: identically.
910: Now if we calculate the correlator with $\vecb{n}_3(x')$ we will reproduce the
911: limit of the exact result Eq.~\ref{3ptcft} above when $x'$ is large.
912: It is now clear that the infra-red divergence in the integral over $x'$ actually is
913: nothing but the divergence
914: of the uniform susceptibility of the vector spin chirality at zero temperature.
915:
916: Armed with this insight let us discuss $T > 0$. Here we will assume that we
917: are interested in the scattering at an external frequency $\omega \gg T$.
918: This simplifies things because then the sole effect of $T > 0$ is to cutoff
919: the IR divergence without affecting the rest of the correlations.
920:
921: Using the OPE it is now clear that the answer is the Fourier transform of
922: \begin{equation}
923: \frac{1}{x^\Delta} \chi_{ch}(T),
924: \end{equation}
925: where $\chi_{ch}(T)$ is the uniform vector spin-chirality susceptibility at temperature $T$.
926: From scaling we have
927: \begin{equation}
928: \chi_{ch}(T) \sim \frac{1}{T^{3-2\Delta}}.
929: \end{equation}
930: Using this we may straightforwardly calculate the inelastic scattering rate.
931: For scattering right at the ordering wavevector $\vecb{Q}$ and for $\omega \gg T$, we get
932: \begin{equation}
933: R_P(\omega) \sim
934: \frac{1}{\omega^{3- \Delta}T^{3-2\Delta}}
935: \end{equation}
936: with $2\Delta = 1+ \bar{\eta} \approx 2.37$.
937:
938: This is a definite prediction that can possibly be tested.
939: In the actual experiments it will be necessary to take into account
940: the staggering of the DM interaction between the different layers. Thus it is best to choose the
941: $a$ component of the 3-vector $\vecb{q}$ to be $\pi$.
942:
943:
944:
945:
946:
947:
948:
949: \section{Summary and Conclusions}
950: \label{sec:conclusion}
951:
952: In this paper we have pursued a concrete version of the idea that Cs$_2$CuCl$_4$,
953: though magnetically ordered at
954: low temperature, may nevertheless
955: be proximate to a spin liquid phase. Such proximity suggests
956: that the intermediate length/time scale physics of \material may be governed by a
957: quantum critical point between a magnetic spiral and a genuine two dimensional quantum
958: spin liquid. A theory for such a quantum phase transition was obtained in Ref. \onlinecite{chubukov}
959: for a simple ($Z_2$) spin liquid with gapped bosonic spinons. We showed that the spin correlations at the
960: quantum critical point are characterized by a large anomalous exponent $\bar{\eta} \approx 1.37$. This is
961: qualitatively consistent with the broad power law tails observed in inelastic neutron scattering in \material.
962: Further we showed that the enlarged $O(4)$ symmetry\cite{chubukov} at the critical fixed point
963: has some remarkable consequences. The vector spin-chirality has the same slow power law decay as the
964: natural magnetic order parameter. This sharp qualitative observation should be of great use in confirming
965: (or ruling out) the applicability of the theory of Ref. \onlinecite{chubukov} to \material. Building on
966: Ref. \onlinecite{Maleyev},
967: we showed how polarized inelastic neutron scattering can be used to directly
968: detect the vector spin-chirality correlations. It is our hope that future experiments will be able to use
969: these results to clarify the physics behind the interesting properties of \material.
970:
971:
972:
973:
974:
975:
976: \section*{Acknowledgements}
977: We thank Radu Coldea, Young Lee, Diptiman Sen, and Ashvin Vishwanath for helpful discussions.
978: This work was supported by NSF Grant No. DMR-0308945 (TS), NSERC of Canada,
979: Canada Research Chair Program, and Canadian Institute for Advanced
980: Research (SVI and YBK).
981: TS also acknowledges
982: funding from the NEC Corporation, the Alfred P. Sloan Foundation, and an award from the The Research Corporation.
983:
984:
985: \begin{thebibliography}{99}
986:
987: \bibitem{chubukov}
988: A. V. Chubukov, T. Senthil, and S. Sachdev,
989: \tit{\prl}{72}{2089}{1994}
990: {Universal magnetic properties of frustrated quantum antiferromagnets in
991: two dimensions};
992: A. V. Chubukov, S. Sachdev and T. Senthil,
993: \tit{\npb}{426}{601}{1994}
994: {Quantum phase transitions in frustrated quantum antiferromagnets}.
995:
996: \bibitem{frustreview}
997: G. Misguich and C. Lhuillier, in ``Frustrated spin systems'',
998: Ed. H.T. Diep, World-Scientific (2004), and references therein.
999:
1000: \bibitem{fazekasanderson}
1001: P. Fazekas and P. W. Anderson,
1002: \tit{\pmag}{30}{23}{1974}
1003: {}.
1004:
1005: \bibitem{anderson}
1006: P. W. Anderson,
1007: \tit{\sci}{235}{1196}{1987}
1008: {The resonating-valence-bond state in La$_{2}$CuO$_{4}$ and
1009: superconductivity}.
1010:
1011: \bibitem{coldeaneu}
1012: R. Coldea, D. A. Tennant, A. M. Tsvelik, and Z. Tylczynski,
1013: \tit{\prl}{86}{1335}{2001}
1014: {Experimental Realization of a 2D Fractional Quantum Spin Liquid};
1015: R. Coldea, D. A. Tennant, and Z. Tylczynski,
1016: \tit{\prb}{68}{134424}{2003}
1017: {Extended scattering continua characteristic of spin fractionalization
1018: in the two-dimensional frustrated quantum magnet Cs$_2$CuCl$_4$ observed
1019: by neutron scattering}.
1020:
1021: \bibitem{weihong}
1022: Z. Weihong, R. H. McKenzie, and R. R. P. Singh,
1023: \tit{\prb}{59}{14367}{1999}
1024: {Phase diagram for a class of spin-$1/2$ Heisenberg models interpolating
1025: between the square-lattice, the triangular-lattice, and the linear-chain
1026: limits}.
1027:
1028: \bibitem{chungln}
1029: C. H. Chung, J. B. Marston and R. H. McKenzie,
1030: \tit{\jpco}{12}{5159}{2001}
1031: {Large-N solutions of the Heisenberg and Hubbard-Heisenberg models on the
1032: anisotropic triangular lattice: application to Cs$_2$CuCl$_4$ and to the
1033: layered organic superconductors $\kappa$-(BEDT-TTF)$_2$X
1034: (BEDT-TTF$\equiv$bis(ethylene-dithio)tetrathiafulvalene); X$\equiv$anion}.
1035:
1036: \bibitem{bocquet}
1037: M. Bocquet, F. H. L. Essler, A. M. Tsvelik, and A. O. Gogolin,
1038: \tit{\prb}{64}{094425}{2001}
1039: {Finite-temperature dynamical magnetic susceptibility of
1040: quasi-one-dimensional frustrated spin-1 / 2 Heisenberg antiferromagnets}.
1041:
1042: \bibitem{chungsp}
1043: C.-H. Chung, K. Voelker, and Y. B. Kim,
1044: \tit{\prb}{68}{094412}{2003}
1045: {Statistics of spinons in the spin-liquid phase of Cs$_2$CuCl$_4$};
1046: S. Takei, C.-H.Chung, and Y. B. Kim,
1047: \tit{\prb}{70}{104402}{2004}
1048: {}.
1049:
1050: \bibitem{veillette}
1051: M.Y. Veillette, J.T. Chalker, R. Coldea, cond-mat/0501347 (unpublished).
1052:
1053: \bibitem{wen91}X.-G. Wen, Phys. Rev. B {\bf 44}, 2664 (1991).
1054:
1055: \bibitem{bfn} L. Balents, M. P. A. Fisher, and C. Nayak,
1056: Phys. Rev. B {\bf 60}, 1654 (1999).
1057:
1058: \bibitem{z2long}
1059: T. Senthil and M. P. A. Fisher
1060: Phys. Rev. B {\bf 62}, 7850 (2000).
1061:
1062: \bibitem{stableu1}
1063: Michael Hermele et al., Phys. Rev. B {\bf 70}, 214437 (2004).
1064:
1065: \bibitem{readsachdev}
1066: N. Read and S. Sachdev,
1067: \tit{\prl}{66}{1773}{1991}
1068: {Large-$N$ expansion for frustrated quantum antiferromagnets};
1069: S. Sachdev, N. Read,
1070: \tit{\ijmpb}{5}{219}{1991}
1071: {Large $N$ expansion for frustrated and doped quantum antiferromagnets}.
1072:
1073: \bibitem{sachdevfru}
1074: S. Sachdev,
1075: \tit{\prb}{45}{12377}{1992}
1076: {Kagome and triangular-lattice Heisenberg antiferromagnets: Ordering from
1077: quantum fluctuations and quantum-disordered ground states with unconfined
1078: bosonic spinon}.
1079:
1080: \bibitem{ballesteros}
1081: H. G. Ballesteros, L. A. Fernandez, V. Martin-Mayor, and A. Munoz-Sudupe,
1082: \tit{\plb}{387}{125}{1996}
1083: {Finite size effects on measures of critical exponents in $d=3$ $O(N)$
1084: model}.
1085:
1086: \bibitem{su4} M. Hermele, T. Senthil. and Matthew P.A. Fisher, cond-mat/0502215.
1087:
1088: \bibitem{coldeaham}
1089: R. Coldea, D. A. Tennant, K. Habicht, P. Smeibidl, C. Wolters,
1090: and Z. Tylczynsk,
1091: \tit{\prl}{88}{137203}{2002}
1092: {Direct Measurement of the Spin Hamiltonian and Observation of Condensation
1093: of Magnons in the 2D Frustrated Quantum Magnet Cs$_2$CuCl$_4$}.
1094:
1095: \bibitem{coldeaold}
1096: R Coldea, D A Tennant, R A Cowley, D F McMorrow, B Dorner and Z Tylczynski,
1097: \tit{\jpco}{8}{7473}{1996}
1098: {Neutron scattering study of the magnetic structure of Cs$_2$CuCl$_4$};
1099: R. Coldea, D. A. Tennant, R. A. Cowley, D. F. McMorrow, B. Dorner,
1100: and Z. Tylczynski,
1101: \tit{\prl}{79}{151}{1997}
1102: {Quasi-1D S = 1/2 Antiferromagnet Cs$_2$CuCl$_4$ in a Magnetic Field}.
1103:
1104: \bibitem{z2}H. Kawamura and S. Miyashita, J. Phys. Soc. Jpn. {\bf 53}, 4138 (1984).
1105:
1106: \bibitem{azaria}P. Azaria, B. Delamotte and T. Jolicoeur,
1107: Phys. Rev. Lett. {\bf 64}, 3175 (1990);
1108: P. Azaria, B. Delamotte and D. Mouhanna, Phys. Rev. Lett. {\bf 68}, 1762 (1992).
1109:
1110: \bibitem{kwmrrev}For an extensive review see
1111: H. Kawamura, J. Phys. Condens. Matter {\bf 10}, 4707 (1998).
1112:
1113: \bibitem{kwmr} H. Kawamura, J. Phys. Soc. Jpn. {\bf 61}, 1299 (1992).
1114:
1115: \bibitem{sen}For a discussion of Berry phase effects in the
1116: semiclassical theory of fluctuating non-collinear order in {\em one dimensional}
1117: quantum spin chains, see S. Rao and D. Sen, J. Phys. Condens. Matter
1118: {\bf 9}, 1831 (1997).
1119:
1120: \bibitem{wolff}
1121: U. Wolff,
1122: \tit{\prl}{62}{361}{1989}
1123: {Collective Monte Carlo Updating for Spin System}.
1124:
1125: \bibitem{footnote}
1126: In addition there will be twelve currents associated with these 6 conserved densities
1127: which will also have scaling dimension 2 with no anomalous dimension. Of these 6 are just the spin currents. The
1128: remaining 6 are currents of the $K_a$.
1129:
1130: \bibitem{Maleyev}S. V. Maleyev
1131: Phys. Rev. Lett. {\bf 75}, 4682 (1995).
1132:
1133: \bibitem{ginscft}P. Ginsparg, {\em Applied Conformal Field Theory}, in
1134: "Fields, Strings, and Critical Phenomena", edited by E. Brezin and
1135: J. Zinn-Justin, Elsevier (1989).
1136:
1137:
1138: \end{thebibliography}
1139:
1140: \end{document}
1141:
1142: