cond-mat0601231/qaf.tex
1: \documentclass{epl}
2: \usepackage{amsmath,amssymb}
3: \usepackage{bm}
4: \usepackage{epsfig}
5: 
6: 
7: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5
8: % BEGIN: key stroke saving definitions 
9: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5
10: 
11: \newcommand{\llangle}{\langle\kern-.25em\langle}
12: \newcommand{\rrangle}{\rangle\kern-.25em\rangle}
13: 
14: 
15: \newcommand{\s}{\hat {\bf S}}
16: 
17: \newcommand{\bk}{{\bf k}}
18: \newcommand{\bq}{{\bf q}}
19: 
20: \newcommand{\ve}{{\vec e}}
21: 
22: \newcommand{\bfe}{{\bf e}}
23: 
24: \newcommand{\bfk}{{\bf k}}
25: \newcommand{\bfr}{{\bf r}}
26: \newcommand{\bfQ}{{\bf Q}}
27: %\newcommand{\zit}[1]{{\sc [#1]}}
28: \newcommand{\zit}[1]{\cite{#1}}
29: 
30: 
31: 
32: \newcommand{\vn}{{\vec n}}
33: \newcommand{\vnl}{\vn}
34: \newcommand{\vnr}{\vn'}
35: 
36: \newcommand{\AAl}{a}
37: \newcommand{\AAlc}{\bar a}
38: \newcommand{\AAr}{a'}
39: \newcommand{\AArc}{\bar a'}
40: \newcommand{\Ac}{\bar a}
41: \newcommand{\DA}{\Delta a}
42: \newcommand{\DAc}{\Delta \bar a}
43: 
44: \newcommand{\BBl}{b}
45: \newcommand{\BBlc}{\bar b}
46: \newcommand{\BBr}{b'}
47: \newcommand{\BBrc}{\bar b'}
48: \newcommand{\Bc}{\bar b}
49: \newcommand{\DB}{\Delta b}
50: \newcommand{\DBc}{\Delta \bar b}
51: 
52: \newcommand{\Pl}{\phi}
53: \newcommand{\Plc}{\bar \phi}
54: \newcommand{\Prx}{\phi'}
55: \newcommand{\Prc}{\bar \phi'}
56: \newcommand{\Dt}{\Delta \tau}
57: 
58: \newcommand{\cD}{{\mathcal D}}
59: \newcommand{\cS}{{\mathcal S}}
60: 
61: \newcommand{\Gammap}{\Gamma^+}
62: \newcommand{\Gammam}{\Gamma^-}
63: \newcommand{\gammap}{\gamma^+}
64: \newcommand{\gammam}{\gamma^-}
65: 
66: \newcommand{\Jm}{J^-}
67: \newcommand{\Jp}{J^+}
68: 
69: \newcommand{\ud}{\textrm{d}}
70: 
71: \newcommand{\todo}[1]{[{\texttt {#1}}]}
72: 
73: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5
74: % END: key stroke saving definitions 
75: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5
76: 
77: 
78: \title{Frustrated Heisenberg antiferromagnets: fluctuation induced first order vs deconfined quantum criticality}
79: \author{F. Kr\"uger\inst{1} \and S. Scheidl\inst{2}}
80: \institute{
81:   \inst{1} Instituut-Lorentz, Universiteit Leiden, P.O. Box
82: 9506, 2300 RA Leiden, The Netherlands\\
83:   \inst{2} Institut f\"ur Theoretische Physik, Universit\"at zu
84: K\"oln, Z\"ulpicher Str. 77, D-50937 K\"oln, Germany
85: }
86: \pacs{75.30.Kz}{Magnetic phase boundaries}
87: \pacs{75.50.Ee}{Antiferromagnetics}
88: \pacs{05.10.Cc}{Renormalization group methods}
89: 
90: \begin{document}
91: 
92: \maketitle
93: 
94: \begin{abstract}
95: Recently it was argued that quantum phase transitions can be radically
96:   different from classical phase transitions with as a highlight the 'deconfined 
97:   critical points' exhibiting fractionalization of quantum numbers due to Berry phase 
98:   effects. Such transitions are supposed to occur in frustrated ('$J_1$-$J_2$') quantum 
99:   magnets. We have developed a novel renormalization approach for such systems which is 
100:   fully respecting the underlying lattice structure. According to our findings, another
101:   profound phenomenon is around the corner: a fluctuation  induced (order-out-of-disorder)
102:   first order transition. This has to occur for large spin and we conjecture that it is
103:   responsible for the weakly first order behavior recently observed in numerical simulations 
104:   for frustrated $S=1/2$ systems. 
105: \end{abstract}
106: 
107: \title{Frustrated Heisenberg antiferromagnets...}
108: 
109: \section{Introduction}
110: Frustrated magnets exhibit quantum phase transitions of a rich variety
111: which is subject of intense current research \zit{Sachdev99}.  Novel
112: scenarios for phase transitions beyond the Landau-Ginzburg-Wilson (LGW)
113: paradigm have been suggested \zit{Senthil+04a,Senthil+04b} joggling fundamental
114: concepts. The Heisenberg model on a square lattice with
115: antiferromagnetic couplings $J_1$ and $J_2$ between nearest and
116: next-nearest neighbors serves as a prototype for studying magnetic
117: quantum-phase transitions (see, e.g., \zit{Sushkov+01} and references
118: therein).  From the classical limit one expects that two different
119: magnetic orders can exist: the N\'eel phase with ordering wave vector
120: $(\pi,\pi)$ is favorable for $\alpha \equiv J_2/J_1 <1/2$ and columnar
121: order with $(0,\pi)$ for $\alpha > 1/2$.
122: 
123: Quantum fluctuations certainly may induce a paramagnetic (PM) phase
124: which is naturally expected near $\alpha \approx 1/2$ where both
125: orders compete \zit{Chandra+88}.  Remarkably, additional orders may
126: appear in the N\'eel phase as well as in the PM phase when translation
127: symmetry is broken by an additional spin dimerization \zit{Sushkov+01}.  
128: The existence of such enhanced order crucially depends on the spin value $S$. 
129: This becomes most apparent when the spin system is represented by a
130: nonlinear-sigma model. Topological excitations associated with Berry phases 
131: can give rise to ground-state degeneracies corresponding to a translation-symmetry 
132: breaking by dimerization and formation of valence-bond solid (VBS) phases 
133: \zit{Haldane88,Fradkin+88}.
134: 
135: The N\'eel-VBS transition has been argued to be in a novel quantum criticality
136: class that does not fit in the standard LGW paradigm. Intriguing 
137: data on this transition was obtained in simulations of the $S=1/2$ quantum
138: XY model frustrated by ring exchange \zit{Sandvik+02,Melko+04}. 
139: The transition was interpreted as a second-order one; this possibility 
140: was predicted by the theory of the deconfined critical point and suggested
141: to be generic for a variety of experimentally relevant two-dimensional 
142: antiferromagnets \zit{Senthil+04b}. However, in a more careful finite-size 
143: analysis of the XY case it was demonstrated that the N\'eel-VBS point
144: represents an anomalously weak first-order transition \zit{Kuklov+04}. 
145: 
146: Field-theoretical approaches of various kinds have been developed,
147: based on the $1/S$ \zit{Chakravarty+89} or the $1/N$ expansion
148: \zit{Read+89,Read+90}.  The latter approach is able to capture some of the
149: essence of the topological aspects on a mean-field level.  However, it
150: is the former approach, elaborated to a renormalization-group
151: analysis, which predominantly has served as basis for a comparison of
152: critical aspects between theory and experiment (see e.g.
153: \zit{Hasenfratz00}).
154: 
155: Nevertheless, this approach so far has suffered from two intrinsic
156: shortcomings: i) Spin-wave interactions are the physical mechanism
157: underlying the renormalization flow.  On one-loop level, the flow
158: equations describe corrections to the physical parameters of relative
159: order $1/S$.  These corrections have been worked out using a continuum
160: version of the nonlinear $\sigma$ model (CNL$\sigma$M) as a starting
161: point \zit{Chakravarty+89}.  However, for the original lattice model,
162: this is not systematic, since corrections of the same order are
163: dropped under the naive coarse graining of the lattice model onto the CNL$\sigma$M.
164: As a consequence, important effects such as a renormalization of the
165: spin-wave velocity and the frustration $\alpha$ are missed.  ii) 
166: Similarly, the outer large momentum region of the magnetic Brillouin zone (BZ) 
167: is only crudely treated. This entails a significant 
168: uncertainty in the computation of the phase diagram.
169: 
170: In this Letter, we lift these shortcomings by developing a
171: renormalization analysis which fully accounts for the lattice
172: structure. It combines the systematic treatment of all corrections in
173: order $1/S$ on the level of the conventional first-order spin-wave
174: theory (SWT) with the merits of a renormalization approach, which goes
175: beyond any finite order in $1/S$ by an infinite iteration of
176: differential steps, successively eliminating the spin-wave modes
177: of highest energy.  As a result, we obtain an improved description of
178: the phase transitions. In particular, the possibility of a fluctuation 
179: induced first-order transition which is not accessible on the level of 
180: the CNL$\sigma$M is included in a natural way.
181: 
182: \section{Spin-coherent state representation}
183: The key to what follows is a novel kind of path integral quantization, which 
184: makes it possible for us to treat the effects of umklapp on an equal
185: footing with the spi-wave interactions.
186: To be specific, we stick to the aforementioned $J_1$-$J_2$ Heisenberg
187: model on a square lattice.  We address the stability of the N\'eel
188: phase against quantum fluctuations controlled by $S$ and $\alpha$.
189: Fluctuations are treated in a coherent spin state path-integral
190: representation of the model, where a spin state corresponds to a unit
191: vector $\vn$.  In the absence of fluctuations, spins would assume the
192: states $|\vn_{A/B}\rangle = |\pm \ve_z\rangle = |S,\pm S\rangle$ on
193: the two sublattice A and B.  From the standard Trotter formula emerges
194: an imaginary time $\tau$ (discretized in intervals of duration $\Dt$)
195: and an action of the form \zit{Sachdev99}
196: \begin{eqnarray}
197:   \cS = - \sum_\tau \ln \langle \{\vn\} | \{\vn'\}\rangle
198:   + \sum_\tau \Dt  \frac { {\langle \{\vn\} | \hat H |\{\vn'\}\rangle}}
199:    {\langle \{\vn\} | \{\vn'\}\rangle}
200: \label{action}
201: \end{eqnarray}
202: taking $\vn$ at time $\tau$ and $\vn'$ at $\tau-\Dt$.  For weak
203: quantum fluctuations, the components of $n_{x,y}$ perpendicular to
204: the magnetization axis are small and may be considered as expansion
205: parameters (as in \zit{Chakravarty+89}).  However, attempting to
206: directly apply this expansion to the lattice model we encountered time
207: ordering difficulties in the action \zit{KS-unpub}.
208: 
209: Instead, we start form a stereographic parametrization of coherent
210: states, on sublattice A, $|\vn \rangle = (1+\AAlc\AAl/2S)^{-S}
211: \exp(\AAl \hat S_-/\sqrt{2S}) |S,S\rangle$ where $a$ is the
212: stereographic projection of $\vn$ from the unit sphere onto the
213: complex plane, $\AAl = \tan(\theta /2) \exp(i \phi)$ with the standard
214: spherical angles $\theta$ and $\phi$.  The action can be expressed in
215: terms of the stereographic coordinates using the matrix
216: elements\zit{Radcliffe71}
217: \begin{eqnarray}
218:   \frac {\langle \vnl | \hat S_z| \vnr \rangle}
219:   {\langle \vnl | \vnr \rangle}
220:   = S \frac{1-\AAlc \AAr/2S}{1+\AAlc \AAr/2S}, \quad
221:   \frac {\langle \vnl | \hat S_+| \vnr \rangle}
222:   {\langle \vnl | \vnr \rangle}
223:   = \frac{\sqrt{2S} \ \AAr}{1+\AAlc\AAr/2S}
224: \end{eqnarray}
225: Here $\bar a$ is the complex conjugate of $a$. The expressions
226: for the coordinate $b$ on sublattice B are given by the correspondences
227: $\hat S_z\to -\hat S_z$ and $\hat S_\pm\equiv\hat S_x\pm\hat S_y\to\hat S_\mp$
228: for $a\to b$.
229: 
230: The explicit expression of the action in terms of $a$ and $b$ is too
231: lengthy to be given here.  To leading order in $1/S$, the fluctuations
232: are controlled by the bilinear part $\cS_0$ of the action that
233: represents free magnons.  We also retain the quartic contribution
234: $\cS_{\rm int}$ to the action, which represents magnon interactions
235: and contain the renormalization of single-magnon parameters of
236: relative order $1/S$.  Higher order contributions to the action are
237: neglected.  Terms from the functional Jacobian are also negligible on
238: this level.  The single-magnon contribution to the action can be
239: parameterized in the form
240: \begin{eqnarray}
241:   \cS_0 &=& \sum_\tau \int_\bk \{ \frac 1{2g}
242:   [\AAlc_\bk \DA_\bk - \DAc_\bk \AAr_\bk
243:   + \BBlc_\bk \DB_\bk - \DBc_\bk \BBr_\bk]
244:   \nonumber \\ &&
245:   +\Dt S [\Gammap_\bk  (\AAlc_\bk \AAr_\bk +\BBlc_\bk \BBr_\bk)
246:   + \Gammam_\bk  (\AAlc_\bk \BBlc_\bk + \AAr_\bk \BBr_\bk)]
247:   \},\quad
248: \end{eqnarray}
249: using the Fourier transform $ \AAl(\bfr)= \int_\bk e^{i \bk \cdot \bfr
250: } \AAl_\bk$, the intgeral $\int_\bk = 2 \int \frac{d^2k}{(2\pi)^2}$
251: over the magnetic BZ $|k_x|+|k_y|\leq \pi$, and the exchange couplings
252: $\Gammap_\bk \equiv \Jp_\bk - \Jp_0 + \Jm_0$, $\Gammam_\bk \equiv
253: \Jm_\bk \equiv 2 J_1 [\cos(k_x)+\cos(k_y)]$ and
254: $\Jp_\bk=2J_2[\cos(k_x+k_y)+\cos(k_x-k_y)]$. For simplicity the
255: lattice constant is considered as unit length.  The dimensionless
256: parameter $g$ represents the strength of quantum fluctuations.  It
257: assumes the value $g=1$ in the unrenormalized model and turns out to
258: increase under renormalization.
259: 
260: Diagonalizing this bilinear action, one easily obtains the magnon
261: dispersion $E_\bk = S g |\Gammap_\bk|\\ \sqrt{1-\gamma_\bk^2}$, where
262: $\gamma_\bk=\Gammam_\bk/\Gammap_\bk$. For $\alpha \leq 1/2$, the
263: low-energy spin-wave excitations are characterized by an isotropic
264: dispersion $E(\bk) = c |\bk| + O(k^2)$ with a spin-wave velocity $c =
265: \sqrt 8 g S J_1 \sqrt{1-2\alpha}$.  Likewise, the exchange couplings
266: generate a spin stiffness $\rho = S^2 J_1 (1-2\alpha)$ for low-energy
267: modes.  One also obtains the propagators $\langle \Plc_\bk \Prx_\bk
268: \rangle = g G_\bk$ and $ \langle \Plc_\bk \Pl_\bk \rangle = \langle
269: \Prc_\bk \Pl_\bk \rangle = g (G_\bk+1)$ for fields from the same
270: sublattice ($\phi=a,b$), and $ \langle \AAl_\bk \BBl_\bk \rangle=
271: \langle \AAlc_\bk \BBlc_\bk \rangle = - g F_\bk$ for fields from
272: different sublattices. In the latter case the correlators are
273: unchanged by a replacement $\bar \phi \to \bar \phi'$.  We have
274: defined $G_\bk = (n_\bk+1/2) (1-\gamma_\bk^2)^{-1/2} - 1/2$ and $F_\bk
275: = (n_\bk+1/2) \gamma_\bk (1-\gamma_\bk^2)^{-1/2}$ where $n_\bk =
276: (e^{\beta E_\bk}-1)^{-1}$ is the bosonic occupation number.  For
277: strong frustration ($\alpha>1/2$) the stiffness becomes negative and
278: the spin-wave velocity is ill defined due to the presence of unstable
279: modes in the center of the BZ (see Fig. \ref{fig.rbz}).
280: 
281: \section{Renormalization approach}
282: Starting from this action with bilinear and quartic terms, we
283: implement a renormalization procedure as follows.  In successive
284: steps, the modes of highest energy (an infinitesimal fraction of all
285: modes) are integrated out. This decimation of modes yields an
286: effective theory for the remaining modes and gives rise to
287: differential flow equations.  As flow parameter we choose $l = \frac
288: 12 \ln (A_{\rm BZ} / A_{\rm RBZ})$, where $A_{\rm BZ}=2 \pi^2$ is the
289: area of the original BZ, and $A_{\rm RBZ}$ is the area of the residual
290: BZ (RBZ) populated by the remaining modes. On large length scales in
291: the N\'eel phase, the RBZ becomes circular and $l$ is the usual
292: logarithmic length scale.  The evolution of the RBZ and the
293: single-magnon dispersion is illustrated in Fig.~\ref{fig.rbz} for
294: various parameters.
295: 
296: \begin{figure}[htbp]
297:   \centering
298:   \epsfig{file=rbzn.eps, width = 0.6\linewidth}
299:     \caption{
300:       Evolution of modes under coarse graining. Each panel corresponds
301:       to the area $|k_x| \leq \pi$, $|k_y| \leq \pi$.  Red color
302:       represents small, yellow high positive energy. Black lines are
303:       lines of constant energy, blue areas represent unstable modes.
304:       Panels (a)-(d): In the N\'eel phase the RBZ may become
305:       disconnected first, then always shrinks to a sphere (here
306:       $S=1/2$, $\alpha=0.3$, $l=0, 0.25, 0.5, 1.0$).  Panels (e)-(h):
307:       In the N\'eel phase for strong frustration $\alpha>1/2$,
308:       initially unstable modes (blue area) are renormalized to stable
309:       modes and the RBZ eventually also shrinks to a sphere (here
310:       $S=1/2$, $\alpha=0.55$, $l=0, 0.11, 0.29, 1.30$).  Panels
311:       (i)-(k): In the columnar phase, after the elimination of all
312:       stable modes, an area of unstable modes survives (here $S=2$,
313:       $\alpha=0.6$, $l=0, 0.10, 0.20$).  }
314: \label{fig.rbz}
315: \end{figure}
316: 
317: 
318: 
319: To one-loop order, corresponding to a systematical calculation of
320: corrections in order $1/S$, the renormalization effects due to
321: spin-wave interactions can be captured by a flow of the single-magnon
322: parameters.  The resulting flow equations are given by
323: 
324: \begin{eqnarray}
325:   \ud g^{-1} &=& - \frac 1S  \ud G^0,
326:   \\
327:   \ud J_1 &=& \frac g S J_1 ( \ud F^- - 2  \ud G^0),
328:   \\
329:   \ud J_2 &=& \frac g S J_2 ( \ud G^+ - 2  \ud G^0), 
330: \end{eqnarray}
331: 
332: where the integrals over the differential fraction $\partial$ of modes
333: of highest energy are defined as $ \ud G^0 = \int_\bq^\partial G_\bq$,
334: $ \ud G^+ = \int_\bq^\partial (J^+_\bq/J^+_0) G_\bq$, and $ \ud F^-=
335: \int_\bq^\partial (J^-_\bq/J^-_0) F_\bq$.
336: 
337: Since the BZ does not remain self-similar under mode decimation, we
338: omit the usual rescaling of length and time which is necessary only
339: for the identification of fixed points under a renormalization-group
340: flow.  However, dropping this rescaling does not lead to a loss of
341: information. Then, each fixed point represents an antiferromagnetically
342: ordered state. The quantum-disordered phase and the transition into it
343: show up as run-away flow.
344: 
345: \section{Results and discussion}
346: Because of the changing geometry of the RBZ and the incorporation of
347: the full single-magnon dispersion in our renormalization approach, the
348: flow equations can be integrated  only numerically.  Here, we focus
349: on $T=0$, although the flow equations are valid also for $T >0$.  The
350: flow of parameters is characterized by the following tendencies.  Both
351: exchange couplings $J_{1,2}$ as well as $1/g$ always shrink.  These
352: fundamental parameters flow in such a way that $\alpha$ always
353: decreases, $c^2$ increases (initially it is negative for
354: $\alpha>1/2$), while $\rho$ may increase for small $l$ until it
355: decreases for sufficiently large $l$.
356: 
357: The nature of magnetic order can be identified from the flow behavior.
358: Three possibilities are observed. (i) The RBZ shrinks to a circle of
359: decreasing radius $\propto e^{-l}$ [see Fig.~\ref{fig.rbz} panels
360: (a)-(d) and panels (e)-(h)] while $J_{1,2}$ and $g$ (as well as the
361: derived quantities $c$, $\rho$, $\alpha$) converge to positive values.
362: Then N\'eel order is present, characterized by these renormalized
363: low-energy parameters. (ii) At some finite $l^*$, fluctuations become
364: so strong that $g$ diverges and $J_{1,2}$ vanish.  This indicates the
365: loss of magnetic order due to overwhelming quantum fluctuations.
366: Close to the transition to the N\'eel phase, the magnetic correlation
367: length -- which can be identified with $\xi = e^{l^*}$ -- diverges
368: algebraically like $\xi\sim (\alpha-\alpha_c)^{-1}$. For $S=1/2$ this
369: asymptotic behavior is shown in one inset of Fig.~\ref{fig.phase}.
370: (iii) For strong frustration, it is also possible that a finite RBZ of
371: unstable modes remains after decimation of {\it all} stable modes (see
372: Fig.~\ref{fig.rbz} panels (i)-(k)).  This indicates that the
373: instability towards columnar order is effective for the renormalized
374: low-energy modes.
375: 
376: 
377: \begin{figure}[htbp]
378:   \centering
379:   \epsfig{file=phase.eps, width=0.6 \linewidth}
380:   \caption{Phase diagram showing the 2nd order transition line 
381:     between the N\'eel ordered and the PM phase (dashed line) and the
382:     1st order boundary between the N\'eel and the columnar phase
383:     (solid line).  The border between the PM and the
384:     columnar phase (dotted line) is just guesswork and cannot be
385:     calculated within our approach. Insets: Nature of the phase
386:     transitions. For $S=1/2$ the renormalized spin stiffness $\rho^*$
387:     vanishes and the strength of the quantum fluctuations $Z_\phi^*$
388:     diverges at the phase boundary corresponding to a 2nd order
389:     transition. Close to the the N\'eel phase the correlation length
390:     diverges like $\xi\sim (\alpha-\alpha_c)^{-1}$. At the 1st order
391:     transition ($S=2$) $\rho^*$ and $Z_\phi^*$ remain finite. }
392:   \label{fig.phase}
393: \end{figure}
394: 
395: 
396: The region of stability of the N\'eel phase is illustrated by the
397: light grey region in Fig.~\ref{fig.phase}.  In the absence of
398: frustration, we find N\'eel order to be stable for $1/S \lesssim 5.09$
399: in remarkable agreement with conventional linear SWT \zit{Chandra+88}.
400: However, this is pure coincidence since in linear SWT the phase
401: boundary is determined by the vanishing of the local magnetization
402: calculated in order $S^0$, whereas here it is determined by the
403: divergence of $g$ due to spin-wave interactions treated in one-loop
404: order. While the phase boundary is located at academically small spin
405: values for small frustration, $S$ reaches physically meaningful values
406: at larger frustration where the discrepancies between SWT and our
407: renormalization approach become more pronounced. In linear SWT, the
408: phase boundary smoothly approaches $1/S \searrow 0$ for $\alpha \nearrow 0.5$,
409: whereas we find the N\'eel phase to reach up to $\alpha = 0.66$ for
410: $S= 0.68$. For spins smaller than this value, the N\'eel phase becomes
411: unstable towards a PM phase via a second-order transition, whereas it
412: becomes unstable for $S>0.68$ via a first order
413: transition. Since we deal with a discontinuous transition we can only 
414: speculate about the type of ordering in the adjacent phase (dark shaded 
415: region in Fig. ~\ref{fig.phase}). From the classical limit we expect columnar
416: order for very large $S$ whereas for intermediate spin also VBS order may be 
417: present.
418: 
419: In the region where the N\'eel phase reaches up to $\alpha>1/2$,
420: initially unstable modes are renormalized to stable ones by spin-wave
421: interactions. Simultaneously, $\alpha$ is renormalized to a value
422: $\alpha^*<1/2$ and the flow behavior (i) is realized.  In the columnar
423: phase, the flow of $\alpha$ saturates at a value $\alpha^*>1/2$ and
424: the flow behavior (iii) is observed.
425: 
426: Stability of N\'eel order for $\alpha>0.5$ so far has been found only
427: by a self consistently modified SWT \zit{Xu+90} and Schwinger-boson
428: mean-field theory (SBMFT) \zit{Mila+91}.  The overall shape of the
429: N\'eel phase of these approaches is consistent with our findings.
430: However, in modified SWT and SBMFT the first-order transition from
431: N\'eel to columnar order can be obtained only by a comparison of free
432: energies between the two phases.  In our theory, the transition
433: directly emerges from the analysis of spin-wave interactions in the
434: N\'eel phase.
435: 
436: The nature of the transitions out of the N\'eel phase becomes apparent
437: from the behavior of the fluctuations on large length scales ($k \to
438: 0$), where $\langle \bar \phi_\bfk \phi_\bfk' \rangle \simeq \frac
439: {Z_\phi}{\sqrt2 k }$ with $Z_\phi \equiv \frac {g}{\sqrt{1-2\alpha}}
440: \propto \frac c\rho$.  Approaching the transition into the PM phase,
441: the renormalized value $Z_\phi^*$ diverges and gives rise to a
442: divergent susceptibility (see Fig.~\ref{fig.ren}).  At the same time
443: the renormalized $\rho^*$ vanishes while $c^*$ remains finite.  The
444: continuous evolution of $Z_\phi^*$ and $\rho^*$ indicate the
445: second-order nature of the transition.  Approaching the transition
446: into the columnar phase, one observes a saturation of $Z_\phi^*$ and
447: $\rho^*$ at finite values as well as a discontinuous jump of $\alpha^*$
448: indicating a first-order transition.  Fig.~\ref{fig.ren} illustrates
449: the dependence of various renormalized quantities on $\alpha$ and
450: $1/S$ within the N\'eel phase. The insets of Fig.~2 show the evolution
451: of $\rho^*$, $Z_\phi^*$ and $\xi$ with higher resolution at the
452: transitions.
453: 
454: 
455: 
456: 
457: \begin{figure}[htbp]
458:   \centering
459:   \epsfig{file=ren2.eps, width = 0.6\linewidth}
460:    \caption{
461:      Values of the renormalized parameters within the N\'eel phase.
462:      Color corresponds to the strength of large-scale fluctuations (a),
463:      the renormalized spin stiffness (b), the frustration (c), and the
464:      spin-wave velocity (d).}
465:   \label{fig.ren}
466: \end{figure}
467: 
468: 
469: 
470: Confidence in the reliability of our findings is provided by
471: quantitative comparisons for specific parameter values.  Results for
472: $Z_c^* \equiv c^*/c$ exist from various approaches. For $S=1/2$ and
473: $\alpha=0$, first-order SWT yields a slight enhancement of spin-wave
474: velocity, $Z_c^{\rm SWT} = 1.158$.  We find an increased value $Z_c^*
475: = 1.20$, which is in agreement with Monte Carlo (MC) simulations (see
476: \cite{Manousakis91} and references therein).  As $1/S$ and/or $\alpha$
477: increases, the enhancement factor $Z_c^*$ also increases.  At the
478: phase boundary $1/S \approx 5.09$ for $\alpha=0$, the difference is
479: already more pronounced: $Z_c^{\rm SWT} = 1.40$ and $Z_c^* = 2.04$.
480: For $\alpha>0$, unfortunately, MC data for $Z_c$ are not available at
481: the transition out of the N\'eel phase, neither for $S=1/2$ nor for
482: larger physical values of $S$.
483: 
484: \section{Conclusion}
485: We have presented a novel renormalization approach for
486: frustrated quantum antiferromagnets which fully accounts for the
487: underlying lattice geometry and consistently captures the
488: renormalization of the single-magnon parameters by spin-wave
489: interactions all over the magnetic BZ. 
490: 
491: For the $J_1$-$J_2$ model 
492: we clearly demonstrated that for large spins and strong frustration, fluctuations 
493: on lattice and intermediate scales cause an instability of the N\'eel phase 
494: towards a first order transition. These effects are totally missed by any effective 
495: long-wavelength continuum theory 
496: obtained by a naive coarse graining of the lattice model.
497: 
498: We conjecture that these fluctuations which crucially depend on the underlying 
499: lattice geometry and the way of frustrating the N\'eel order are responsible 
500: for the weakly first order behavior observed recently in numerical simulations 
501: for frustrated $S=1/2$ systems \zit{Kuklov+04}.  
502: 
503: \acknowledgments
504: The authors benefited from stimulating discussions with J. Zaanen, J. Betouras and 
505: A. Sandvik and thank J. van Wezel for critical reading of the manuscript. 
506: This work was supported by Dutch Science Foundation NWO/FOM and by 
507: Deutsche Forschungsgemeinschaft SFB 608.
508: 
509: 
510: \begin{thebibliography}{0}
511: 
512: \bibitem{Sachdev99}
513:   \Name{S. Sachdev}
514:   \Book{Quantum Phase Transitions}
515:   \Editor{Cambridge University Press}
516:   \Publ{Cambridge, England}
517:   \Year{1999}.
518: 
519: \bibitem{Senthil+04a}
520:   \Name{Senthil T., Balents L., Sachdev S., Vishwanath A., \and Fisher M.P.A}
521:   \REVIEW{Phys. Rev. B}{70}{2004}{144407}.
522: 
523: \bibitem{Senthil+04b}
524:   \Name{Senthil T., Vishwanath A., Balents L., Sachdev S., \and Fisher M.P.A}
525:   \REVIEW{Science}{303}{2004}{1490}.
526: 
527: \bibitem{Sushkov+01}
528:   \Name{Sushkov O.P., Oitmaa J., \and Weihong Z.}
529:   \REVIEW{Phys. Rev. B}{63}{2001}{104420}.
530: 
531: \bibitem{Chandra+88}
532:   \Name{Chandra P. \and Doucot B.}
533:   \REVIEW{Phys. Rev. B}{38}{1988}{R9335}.
534: 
535: 
536: \bibitem{Haldane88}
537:   \Name{Haldane F.D.M.}
538:   \REVIEW{Phys. Rev. Lett}{61}{1988}{1029}.
539: 
540: \bibitem{Fradkin+88}
541:   \Name{Fradkin E. \and Stone M.}
542:   \REVIEW{Phys. Rev. B}{38}{1988}{R7215}.
543: 
544: 
545: \bibitem{Sandvik+02}
546:   \Name{Sandvik A.W., Daul S., Singh R.R.P., \and Scalapino D.J.}
547:   \REVIEW{Phys. Rev. Lett.}{89}{2002}{247201}.
548: 
549: \bibitem{Melko+04}
550:   \Name{Melko R.G., Sandvik A.W., \and Scalapino D.J.}
551:   \REVIEW{Phys. Rev. B}{38}{2004}{R9335}.
552: 
553: \bibitem{Kuklov+04}
554:   \Name{Kuklov A., Prokof'ev N., \and Svistunov B.}
555:   \REVIEW{Phys. Rev. Lett.}{93}{2004}{230402}.
556: 
557: 
558: \bibitem{Chakravarty+89}
559:   \Name{Chakravarty S., Halperin B., \and Nelson D.R.}
560:   \REVIEW{Phys. Rev. B}{39}{1989}{2344}.
561: 
562: 
563: \bibitem{Read+89}
564:   \Name{Read N. \and Sachdev S.}
565:   \REVIEW{Phys. Rev. Lett.}{62}{1989}{1694}.
566: 
567: \bibitem{Read+90}
568:   \Name{Read N. \and Sachdev S.}
569:   \REVIEW{Phys. Rev. B}{42}{1990}{4568}.
570: 
571: \bibitem{Hasenfratz00}
572:   \Name{Hasenfratz P.}
573:   \REVIEW{Eur. Phys. J. B}{13}{2000}{11}.
574: 
575: \bibitem{KS-unpub}
576:   \Name{Kr\"uger F. \and Scheidl S.}
577:   \Book{unpublished}.
578: 
579: \bibitem{Radcliffe71}
580:   \Name{Radcliffe J.M.}
581:   \REVIEW{J. Phys. A}{4}{1971}{313}.
582: 
583: \bibitem{Xu+90}
584:   \Name{Xu J.H. \and Ting C.S.}
585:   \REVIEW{Phys. Rev. B}{42}{1990}{R6861}.
586: 
587: \bibitem{Mila+91}
588:   \Name{Mila F., Poilblanc D., \and Bruder C.}
589:   \REVIEW{Phys. Rev. B}{43}{1991}{7891}.
590: 
591: \bibitem{Manousakis91}
592:   \Name{Manousakis E.}
593:   \REVIEW{Rev. Mod. Phys.}{63}{1991}{1}.
594: 
595: \end{thebibliography}
596: 
597: \end{document}
598: 
599: