cond-mat0505425/qaf.tex
1: % Time-stamp: "Friday, 05.08.2005 at 13:30:50; edited by spremo"
2: % 
3: \tolerance = 10000
4: %
5: %\documentclass[twocolumn,showpacs,prl,amsmath,amssymb,floatfix]{revtex4}
6: \documentclass[twocolumn,showpacs,prb,amsmath,amssymb,floatfix,eqsecnum,subeqn]{revtex4}
7: %\documentclass[twocolumn,showpacs,superscriptaddress,prb,amsmath,amssymb,floatfix,eqsecnum]{revtex4}
8: %\documentclass[galley,showpacs,prl,amsmath,amssymb]{revtex4}
9: %
10: %\usepackage{dcolumn}
11: \usepackage{amsmath,amssymb}
12: \usepackage{bm}
13: \usepackage{epsfig}
14: \usepackage{psfrag}
15: 
16: % set \bd to \bf or \bm
17: \newcommand{\bd}{\bm}
18: \newcommand{\jonecolor}{green }
19: \newcommand{\jtwocolor}{blue }
20: 
21: 
22: \begin{document}
23: 
24: \title{Magnetic properties of a metal-organic  antiferromagnet 
25:   on  a distorted honeycomb lattice}
26: 
27: \author{Ivan Spremo, Florian Sch\"{u}tz, and Peter Kopietz}  
28: \affiliation{Institut f\"{u}r Theoretische Physik, Universit\"{a}t
29:   Frankfurt, Max-von-Laue-Strasse 1, 60438 Frankfurt, Germany}
30: 
31: 
32: \author{Volodymyr Pashchenko, Bernd Wolf, and Michael Lang}
33: \affiliation{Physikalisches Institut, Universit\"{a}t
34:   Frankfurt, Max-von-Laue-Strasse 1, 60438 Frankfurt, Germany}
35: 
36: 
37: \author{Jan W.~Bats}
38: \affiliation{Institut f\"ur Organische Chemie und Chemische Biologie, 
39:   Universit\"{a}t Frankfurt, Marie-Curie-Strasse 11, 60439 Frankfurt, Germany}
40: 
41: 
42: \author{Chunhua Hu and Martin U.~Schmidt}
43: \affiliation{Institut f\"ur Anorganische und Analytische Chemie, 
44:   Universit\"{a}t Frankfurt, Marie-Curie-Strasse 11, 60439 Frankfurt, Germany}
45: 
46: 
47: \date{\today}
48: %\date{December 22, 2004}
49: 
50: \begin{abstract}
51:   For temperatures $T$ well above the ordering temperature $T_{\ast}
52:   = 3.0 \pm 0.2 \mathrm{K}$ the magnetic properties of the metal-organic material 
53:   $\mathrm{Mn}[\mathrm{C}_{10}\mathrm{H}_{6}(\mathrm{OH})(\mathrm{COO})]_{2}
54:   \!\times\! 2\mathrm{H}_{2}\mathrm{O}$
55:   built from
56:   Mn$^{2+}$ ions and $3$-hydroxy-2-naphthoic anions can be described by a
57:   $S=5/2$ quantum antiferromagnet on a distorted honeycomb lattice
58:   with two different nearest neighbor exchange couplings $J_2 \approx
59:   2 J_1 \approx 1.8\, \mathrm{K}$.  Measurements of the magnetization
60:   $M(H,T)$ as a function of a uniform external field $H$ and of the
61:   uniform zero field susceptibility $\chi(T)$ are explained within the
62:   framework of a modified spin-wave approach which takes into account
63:   the absence of a spontaneous staggered magnetization at finite
64:   temperatures.
65: \end{abstract}
66: 
67: \pacs{75.10.Jm, 75.30.Ds, 75.50.Ee}
68: 
69: %\preprint{}
70: 
71: %\draft
72: 
73: \maketitle
74: 
75: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
76: \section{Introduction}
77: 
78: In recent years the role of fluctuations, spatial anisotropy and
79: frustration in low dimensional quantum magnets has been intensely
80: studied, both experimentally and theoretically.\cite{Schollwoeck04}
81: For a comparison of experiments with theory it is crucial to have well
82: defined crystalline materials where one or several parameters can be
83: varied externally in order to obtain quantitative  predictions for
84: physical observables.  Moreover, in order to observe interesting
85: magnetic many-body effects it is essential to have materials where the
86: magnetic moments are coupled via sufficiently strong exchange
87: interactions.  These conditions are met by transition metal oxides
88: such as cuprates, vanadates, copper-germanates, or manganites, which
89: have been the subject of many works.  However, in these materials it
90: is rather difficult to control externally microscopic parameters such
91: as the precise values of the exchange interactions or the lattice geometry by
92: changing the chemical composition in a well defined manner.  This
93: problem tends to be less severe in magnets based on metal-organic
94: materials, which offer more possibilities of modifying some
95: constituents chemically and thereby tuning the properties by a crystal 
96: engineering strategy.  
97: The challenge is then to find metal-organic
98: magnets where the magnetic moments are coupled sufficiently strongly
99: to exhibit interesting collective effects.
100: 
101: %
102: %
103: \begin{figure}[tb]
104:   \begin{center}
105:     \epsfig{file=bona_b_vp.eps,width=85mm}
106:   \end{center}
107:   \vspace{-4mm}
108:   \caption{%
109:     View along the $b$-axis of the metal-organic quantum magnet 
110:     $\mathrm{Mn}[\mathrm{C}_{10}\mathrm{H}_{6}(\mathrm{OH})(\mathrm{COO})]_{2}
111:   \!\times\! 2\mathrm{H}_{2}\mathrm{O}$.
112:    Bold lines show exchange paths 
113:    $\mathrm{Mn}-\mathrm{O}-\mathrm{C}-\mathrm{O}-\mathrm{Mn}$.
114:    The unit cell, denoted by the parallelogram, 
115:    contains four crystallographically equivalent $\mathrm{Mn}^{2+}$ ions.
116:   }
117:   \label{fig:BONS-b}
118: \end{figure}
119: %
120: %
121: 
122: These effects are of particular importance in low-dimensional magnets, 
123: e.g.~$2\mathrm{D}$ layer structures with strong magnetic couplings within the layers and 
124: weak interactions between the layers. Such layer structures can be built up chemically 
125: from spin-bearing metal ions, which are connected by short bridges, being separated 
126: by organic fragments of considerable size, see Fig.~\ref{fig:BONS-b}.
127: Motivated by these considerations we synthesized transition metal complexes 
128: of o-hydroxy-naphthoic acids. The crystal structure of 
129: $\mathrm{Mn}[\mathrm{C}_{10}\mathrm{H}_{6}(\mathrm{OH})(\mathrm{COO})]_{2}
130:   \!\times\! 2\mathrm{H}_{2}\mathrm{O}$ (systematic name: manganese(II) 
131: 3-hydroxy-2-naphthoate dihydrate, Fig.~\ref{fig:BONS-chem}) is of particular interest, 
132: because the $\mathrm{Mn}^{2+}$ ions form a distorted honeycomb lattice (Fig.~\ref{fig:BONS}).
133: For the redetermination of the crystal structure, pale brown crystals were slowly 
134: grown by diffusion of an aqueous solution of 
135: $\mathrm{Na}[\mathrm{C}_{10}\mathrm{H}_{6}(\mathrm{OH})(\mathrm{COO})]$ 
136: into an aqueous $\mathrm{MnSO}_{4}$ solution with a buffer layer of water. 
137: The single crystal X-ray analysis confirmed the previously determined 
138: structure\cite{Schmidt04} with a higher precision. 
139: The compound crystallizes in the monoclinic space group  $P2_1/c$ with the lattice 
140: parameters $a = 17.191(4)\,\mathrm{\AA}$, $b = 7.3448(10)\,\mathrm{\AA}$, 
141: $c = 15.5279(17)\,\mathrm{\AA}$, $\beta = 101.964(8)^{\circ}$, 
142: $V = 1918.1(5)\,\mathrm{\AA}^3$.\cite{CCDC} 
143: The unit cell contains four crystallographically equivalent $\mathrm{Mn}^{2+}$ ions. 
144: 
145: %
146: %
147: \begin{figure}[tb]
148:   \begin{center}
149:     \epsfig{file=BONS-chem.eps,width=85mm}
150:   \end{center}
151:   \vspace{-4mm}
152:   \caption{%
153:     Chemical formula of 
154:     $\mathrm{Mn}[\mathrm{C}_{10}\mathrm{H}_{6}(\mathrm{OH})(\mathrm{COO})]_{2}
155:   \!\times\! 2\mathrm{H}_{2}\mathrm{O}$.
156:   }
157:   \label{fig:BONS-chem}
158: \end{figure}
159: %
160: %
161: 
162: 
163: The coupling layer, parallel to the $(bc)$ plane, contains the $\mathrm{Mn}^{2+}$ 
164: ions, the $\mathrm{COO}^{-}$ and $\mathrm{OH}$ groups as well as water molecules. 
165: The isolating layer, having a thickness of about $12\,\mathrm{\AA}$ consists of 
166: the organic naphthalene moieties. 
167: These naphthalene moieties are only bound together by 
168: van der Waals contacts between C and H atoms. The relative weakness of these 
169: interactions is reflected by the morphology of the crystals: the crystals 
170: grow in ($b$) and ($c$) direction much faster than in ($a$) direction, 
171: thus forming thin plates parallel to the ($bc$) plane.
172: 
173: The magnetism is due to the $S = 5/2$ manganese ions which form a distorted 
174: honeycomb pattern parallel to the ($bc$) planes. 
175: Neighboring ions are connected by carboxylic 
176: groups, which provide an 
177: $\mathrm{Mn}-\mathrm{O}-\mathrm{C}-\mathrm{O}-\mathrm{Mn}$ 
178: magnetic exchange path. There are two 
179: different exchange paths: the first path contains a single 
180: $\mathrm{O} - \mathrm{C} - \mathrm{O}$ unit, displayed 
181: in \jonecolor in Fig.~\ref{fig:BONS}. In the second path 
182: (marked with \jtwocolor color) the $\mathrm{Mn}^{2+}$ ions are connected by two 
183: $\mathrm{O} - \mathrm{C} - \mathrm{O}$
184: moieties simultaneously. The honeycomb layers are well separated from each other; 
185: the closest distances between $\mathrm{Mn}^{2+}$ ions of different layers 
186: are as large as $16.282\,\mathrm{\AA}$.
187: %
188: %
189: \begin{figure}[tb]
190:   \begin{center}
191:     \epsfig{file=bona_bc.eps,width=85mm}
192:   \end{center}
193:   \vspace{-4mm}
194:   \caption{%
195:     View on the $(bc)$ plane of the metal-organic quantum magnet 
196:     $\mathrm{Mn}[\mathrm{C}_{10}\mathrm{H}_{6}(\mathrm{OH})(\mathrm{COO})]_{2}
197:   \!\times\! 2\mathrm{H}_{2}\mathrm{O}$.
198:   }
199:   \label{fig:BONS}
200: \end{figure}
201: %
202: %
203: 
204: The structure in Fig.~\ref{fig:BONS} suggests that the magnetic
205: properties of the material can be modeled by a spin $S=5/2$ Heisenberg
206: magnet on the distorted honeycomb lattice shown in
207: Fig.~\ref{fig:lattice}. The exchange integrals $J_{\nu} =
208: J ( {\bd{r}}_i , {\bd{r}}_i + \bd{\delta}_{\nu} )$, $\nu = 1,2,3$,
209: couple the spin at a given site ${\bd r}_i$ to its nearest neighbors at
210: ${\bd r}_i + {\bd\delta}_{\nu}$.  All exchange integrals $J_\nu$ turn
211: out to be positive, and 
212: $\left|\boldsymbol{\delta}_1\right| =
213: \left|\boldsymbol{\delta}_3\right| \equiv \delta_1 = 5.131(4)\,\mathrm{\AA}$ 
214: and $J_1 = J_3$, due to the crystal symmetry.
215: A closer look at the crystal structure in Fig.~\ref{fig:BONS} and a
216: comparison with the distorted honeycomb lattice in
217: Fig.~\ref{fig:lattice} reveals that $J_2$ acts along two exchange
218: paths while $J_1$ results from a single exchange path. 
219: Therefore we expect $J_2$ to be roughly twice as large as $J_1$.
220: Furthermore, the honeycomb lattice is bipartite, i.e., it can be
221: divided into two sublattices, labeled A and B, such that the nearest
222: neighbors of all sites belonging to sublattice A are located on
223: sublattice B.  Thus, for positive $J_\nu$ the system is not
224: frustrated, and when quantum fluctuations are neglected the ground
225: state shows classical antiferromagnetic N\'eel order. More generally, we
226: expect long-range antiferromagnetic order to persist in the quantum
227: mechanical ground state. Therefore, it should be possible to calculate
228: the magnetic properties of the system within the usual spin-wave
229: expansion, at least for temperature $T=0$.
230: Note that the actual structure shown in Fig.~\ref{fig:BONS} has an additional 
231: distortion in the $x$-direction, resulting in a primitive cell with doubled volume.
232: Due to the low symmetry of the lattice the 
233: Dzyaloshinskii-Moriya interaction might play an important role. 
234: However, we expect the corresponding energy scale to be small in
235: comparison with $J_1$ and $J_2$, so that in the first approximation
236: we can neglect this effect.
237: In the following we therefore always work with the magnetically equivalent 
238: Bravais lattice shown in Fig.~\ref{fig:lattice}.
239: 
240: %
241: %
242: \begin{figure}[t]
243:   \centering
244:   \psfrag{J1}{$J_1$}
245:   \psfrag{J2}{$J_2$}
246:   \psfrag{J3}{$J_3$}
247:   \psfrag{d1}{$\boldsymbol{\delta}_1$}
248:   \psfrag{d2}{$\boldsymbol{\delta}_2$}
249:   \psfrag{d3}{$\boldsymbol{\delta}_3$}
250:   \psfrag{a1}{$a_1$}
251:   \psfrag{a2}{$a_2$}
252:   \psfrag{a3}{$a_3$}
253:   \psfrag{f}{$\varphi$}
254:   \psfrag{x}{$x$}
255:   \psfrag{y}{$y$}
256:   \psfrag{sublattice A}{sublattice A}
257:   \psfrag{sublattice B}{sublattice B}
258:   \epsfig{file=honeycomb_lattice.eps,width=220pt}
259:   \caption{Distorted honeycomb lattice.
260:     The interactions between spins are displayed as solid lines.  The
261:     underlying magnetic sublattice is a Bravais lattice and its
262:     primitive cell can be chosen to be the dashed parallelogram.  The
263:     corresponding primitive vectors are $ {\bd a}_1 =
264:     a_{1}{\hat{\bd e}}_x$ and ${\bd a}_2 =
265:     a_{2}\cos{\varphi}\,{\hat{\bd e}}_x + 
266:     a_{2}\sin{\varphi}\,{\hat{\bd e}}_y$.  }
267:   \label{fig:lattice}
268: \end{figure}
269: %
270: %
271: 
272: Measurements of the magnetization $M(H,T)$ in a magnetic field
273: $H$ are performed at finite temperatures $T$, where long-range
274: antiferromagnetic order is ruled out by the Hohenberg-Mermin-Wagner
275: theorem.\cite{Hohenberg67}  In this case the theoretical justification
276: for the spin-wave expansion in two dimensions is more subtle.  As long
277: as there is long-range antiferromagnetic order at $T=0$, it is
278: reasonable to expect that the low-energy and long-wavelength physics
279: is still dominated by renormalized spin-waves.  The magnetic
280: properties of square lattice antiferromagnets in the absence of
281: uniform external fields have been thoroughly investigated in a
282: classical work by Chakravarty, Halperin, and Nelson.
283: \cite{Chakravarty88} Less is known about the low-energy physics of
284: two-dimensional quantum antiferromagnets subject to a uniform external
285: magnetic field. 
286: The external field breaks the rotational symmetry of the Heisenberg
287: antiferromagnet to O(2), similar to the effect of an XY anisotropy
288: in the XXZ model.\cite{Fukumoto96} However, the classical ground states
289: of the two models differ substantially: whereas the XXZ model has a 
290: collinear ground state, an uniform magnetic field in a Heisenberg 
291: antiferromagnet leads to a canted classical spin configuration shown in
292: Fig.~\ref{fig:spingroundstate}.
293: The zero-temperature magnetization curve $M(H,0)$ of
294: the square lattice antiferromagnet has been calculated a few years ago
295: by Zhitomirsky and Nikuni~\cite{Zhitomirsky98} within the spin-wave
296: expansion.  For finite temperatures, $M(H,T)$ has been extrapolated
297: from numerical diagonalizations of finite clusters.\cite{Fabricius92}
298: We are not aware of any analytical calculations in the literature
299: of $M(H,T)$ for two-dimensional quantum Heisenberg antiferromagnets at $T > 0$.
300: In this work, we calculate $M(H,T)$ using a modified spin-wave
301: approach \cite{Kollar03} which takes the absence of a spontaneous
302: staggered magnetization at finite temperatures into account.  
303: Our theoretical results for the magnetization curves as well as for the
304: zero-field susceptibility $\chi(T)$ show a satisfactory agreement with
305: our measurements for the compound
306: $\mathrm{Mn}[\mathrm{C}_{10}\mathrm{H}_{6}(\mathrm{OH})(\mathrm{COO})]_{2}
307: \!\times\! 2\mathrm{H}_{2}\mathrm{O}$.
308: 
309: The rest of the paper is organized as follows. In
310: Sec.~\ref{sec:general} we review the formalism of the spin-wave
311: expansion for non-collinear spin configurations. In
312: Sec.~\ref{sec:honeycomb} this method is applied to an antiferromagnet on a
313: bipartite lattice in the presence of a uniform magnetic field.
314: Expressions for the magnetization, the staggered magnetization and the
315: uniform susceptibility for the material of interest are obtained. We explain how a
316: self-consistently determined staggered field is used to regularize
317: divergencies at finite temperature. In Sec.~\ref{sec:results} we
318: present results and compare with experimental measurements.
319: Finally, in Sec.~\ref{sec:discussion} we present our conclusion.
320: 
321: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
322: \section{Spin waves in non-collinear spin configurations}
323: \label{sec:general}
324: 
325: In the presence of a homogeneous magnetic field an antiferromagnet on
326: a bipartite lattice has a non-collinear, canted spin configuration as
327: shown in Fig.~\ref{fig:spingroundstate}. 
328: We choose a coordinate system such that the uniform external field
329: points along the $x$-axis, and the staggered magnetization is directed
330: along the $z$-axis.  The low temperature physics is dominated by
331: spin-wave excitations.  To obtain their spectrum we should quantize
332: the spin-operators in a spatially-dependent (``co-moving'') coordinate
333: system that matches for each site the axis defined by the expectation
334: value $\langle \bd{S}_i \rangle$ of the spin operator.
335: %
336: %
337: \begin{figure}[tb]
338: % \begin{center}
339:   \centering
340:   \psfrag{x}{$x$}
341:   \psfrag{z}{$z$}
342:   \psfrag{t}{$\theta$}
343:   \psfrag{ma}{${\hat{\bd m}}_{\mathrm{A}}$}
344:   \psfrag{mb}{${\hat{\bd m}}_{\mathrm{B}}$}
345:   \psfrag{m}{$\langle {\bd S}_i \rangle $}
346:   \psfrag{B}{$ B {\hat{\bd e}}_x $}
347:   \psfrag{Bs}{$\zeta_i B_s {\hat{\bd e}}_z$}
348:   \psfrag{e2a}{$ {\hat{\bd e}}_\mathrm{A}^2$}
349:   \psfrag{e2b}{$ {\hat{\bd e}}_B^2$}
350:   \epsfig{file=magn_field_geometry.eps,width=65mm}
351: %   \end{center}
352:   \vspace{-4mm}
353:   \caption{%
354:     Spin configuration $\langle \bd{S}_i \rangle$ in the classical ground state
355:     of a two-sublattice antiferromagnet .  The dashed arrows
356:     represent a uniform magnetic field $ B \hat{\bd{e}}_x$ in the
357:     $x$-direction and a staggered magnetic field $ \zeta_i B_s
358:     \hat{\bd{e}}_z$ in the $z$-direction.  The small solid arrows
359:     represent the vectors of a ``co-moving'' basis that matches the
360:     direction defined by the local magnetization $\langle \bd{S}_i
361:     \rangle$.  Not shown are the basis vectors $\hat{\bd{e}}_\mathrm{A}^1 =
362:     \hat{\bd{e}}_\mathrm{B}^1 = \hat{\bd{e}}_y$ which point into the plane of
363:     the paper.  }
364:   \label{fig:spingroundstate}
365: \end{figure}
366: %
367: %
368: 
369: More generally, the problem of calculating the spin excitations of a
370: Heisenberg magnet subject to an arbitrary inhomogeneous magnetic field
371: ${\bd{B}}_i$ can be formulated and solved in a coordinate-free vector
372: notation.\cite{Schuetz03}  Consider the general Heisenberg hamiltonian
373: \begin{equation}
374:   \hat{H} = \frac{1}{2} \sum_{  i,j}   J_{ij} {\bd{S}}_i \cdot
375:   {\bd{S}}_j   - g \mu_{\text{B}} \sum_{i  }  {\bd{B}}_i \cdot
376:   {\bd{S}}_i
377:   \,,
378:   \label{eq:Hamiltonian}
379: \end{equation} 
380: where $J_{ij} = J ( \bd{r}_i, \bd{r}_j ) $ are some arbitrary exchange
381: couplings, the sums are over all sites ${\bd{r}}_i$ of a
382: $D$-dimensional lattice consisting of $N$ sites, and the ${\bd{S}}_i$
383: are spin-$S$ operators normalized such that ${\bd{S}}_i^2 = S ( S+1
384: )$.  The last term represents the Zeeman energy, where $g$ is the
385: gyromagnetic factor and $\mu_{\mathrm{B}}$ is the Bohr magneton.  We assume that
386: the external magnetic field is sufficiently strong to induce permanent
387: magnetic dipole moments ${\bd{m}}_i = g \mu_{\mathrm{B}} \langle {\bd{S}}_i
388: \rangle$, where $\langle \ldots \rangle$ denotes the thermal
389: equilibrium average.  It is then convenient to decompose the spin
390: operators as ${\bd{S}}_i = S^{\parallel}_i \hat{\bd{m}}_i +
391: {\bd{S}}^{\bot}_i$, where ${\bd{S}}_i^{\bot} \cdot \hat{\bd{m}}_i =
392: 0$, and $\hat{\bd{m}}_i = \bd{m}_i / | {\bd{m}}_i |$ is a unit vector
393: in the direction of $\bd{m}_i$.  Substituting this decomposition into
394: Eq.~(\ref{eq:Hamiltonian}) we obtain $\hat{H} = \hat{H}^{\parallel} +
395: \hat{H}^{\bot} + \hat{H}^{\prime}$, with
396: \begin{eqnarray}
397:   \hat{H}^{\parallel}\,\, & = & \frac{1}{2} \sum_{  i,j}  J_{ij}  
398:   \hat{\bd{m}}_i \cdot \hat{\bd{m}}_j
399:   {{S}}_i^{ \parallel} 
400:   {{S}}_j^{\parallel}   - \sum_{i  }  {\bd{h}}_i  \cdot \hat{\bd{m}}_i 
401:   {{S}}_i^{\parallel}
402:   \,,
403:   \label{eq:Hparallel}
404:   \\
405:   \hat{H}^{\bot} & = & \frac{1}{2} \sum_{  i,j}  J_{ij}  
406:   {\bd{S}}^{\bot}_i \cdot {\bd{S}}^{\bot}_j  
407:   \,,
408:   \label{eq:Hbot}
409:   \\
410:   \hat{H}^{\prime}\,\, & = &  - \sum_{  i}   {\bd{S}}^{\bot}_i
411:   \cdot \bigl(  {\bd{h}}_i -
412:   \sum_j J_{ij}  
413:   {{S}}_j^{ \parallel}   \hat{\bd{m}}_j   \bigr)
414:   \,,
415:   \label{eq:Hrest}
416: \end{eqnarray}
417: where ${\bd{h}}_i = g \mu_{\mathrm{B}} {\bd{B}}_i $.  Note that
418: $\hat{H}^{\prime}$ describes the coupling between the transverse and
419: the longitudinal spin fluctuations.  The classical ground state energy
420: ${E}_0^{\rm cl}$ is obtained by replacing $S_i^{\parallel} \rightarrow
421: S$ in Eq.~(\ref{eq:Hparallel}) and by finding the configuration $\{
422: \hat{\bd{m}}_i \}$ that minimizes the resulting classical hamiltonian
423: \begin{equation}
424:   H^{\rm cl} =
425:   \frac{S^2}{2} \sum_{  i,j}  J_{ij}  
426:   \hat{\bd{m}}_i \cdot \hat{\bd{m}}_j
427:   - S \sum_{i  }  {\bd{h}}_i  \cdot \hat{\bd{m}}_i 
428:   \label{eq:Hcl}
429:   \; .
430: \end{equation}
431: A necessary condition for an extremum of Eq.~(\ref{eq:Hcl}), taking
432: into account the constraints $\hat{\bd{m}}_i^2 =1$,
433: is~\cite{Schuetz03}
434: \begin{equation}
435:   \hat{\bd{m}}_i
436:   \times
437:   \Big(
438:   {\bd{h}}_i - S \sum_j J_{ij}  \hat{\bd{m}}_j
439:   \Big)
440:   = 0
441:   \,. 
442:   \label{eq:classical}
443: \end{equation} 
444: For given ${\bd{h}}_i $ and $J_{ij}$, this is a system of non-linear
445: equations for the spin directions $\hat{\bd{m}}_i$ in the classical
446: ground state. Using Eq.~(\ref{eq:classical}), the part
447: $\hat{H}^{\prime}$ of the hamiltonian describing the coupling between
448: transverse and longitudinal fluctuations can be written as
449: \begin{equation}
450:   \hat{H}^{\prime} = - \sum_{ i,j} J_{ij} 
451:   ( {\bd{S}}_i^{\bot} \cdot \hat{\bd{m}}_j ) ( S -S_j^{\parallel} )
452:   \; .
453:   \label{eq:Hprime}
454: \end{equation}
455: Let us expand the transverse components of $ {\bd{S}}^{\bot}_i $ in a
456: spherical basis, $ {\bd{S}}^{\bot}_i $ $=$ $\frac{1}{2} \sum_{ p = \pm
457: } S_i^{-p} {\bd{e}}^{p}_i $, with the spherical basis vectors
458: ${\bd{e}}^{p}_i $ $=$ $ \hat{\bd{e}}^1_i + i p \hat{\bd{e}}^2_i$, $p $
459: $=$ $ \pm$, where $ \left\{ \hat{\bd{e}}_i^{1}, \hat{\bd{e}}_i^2 ,
460:   \hat{\bd{m}}_i \right\}$ is a local orthogonal triad of unit
461: vectors. The transverse part of our spin hamiltonian can then be
462: written as
463: \begin{equation}
464:   \hat{H}^{\bot}  =  \frac{1}{8} \sum_{  i,j}  \sum_{ p, p^{\prime} } J_{ij}  
465:   ( {\bd{e}}^{p}_i \cdot {\bd{e}}^{p^{\prime}}_j )
466:   {{S}}^{-p}_i 
467:   {{S}}^{-p^{\prime}}_j  
468:   \,.
469:   \label{eq:Hbot2}
470: \end{equation}
471: The basis vectors $ \hat{\bd{e}}_i^{1}, \hat{\bd{e}}_i^2 $ are not
472: unique: any rotation around $\hat{\bd{m}}_i$ yields an equally
473: acceptable transverse basis.
474: 
475: So far, no approximation has been made. To obtain the spin-wave
476: spectrum, we expand the spin operators in terms of canonical boson
477: operators $b_i$ as usual \cite{Dyson56,Maleev57}, $S^{\parallel}_i$ $=$
478: $S-b_i^{\dagger}b_i^{\phantom{\dagger}}$ and $S^{+}_i$ $=$
479: $(S^{-}_i)^\dagger$ $=$
480: $\sqrt{2S}\,b_i^{\phantom{\dagger}}\,(1+O(S^{-1}))$.  Within the
481: linear spin-wave approximation the hamiltonian becomes $\hat{H}
482: \approx E_0^{\rm cl} + \hat{H}_2$, where $E_0^{\rm cl}$ is the minimum
483: of the classical hamiltonian $H^{\rm cl}$ in Eq.~(\ref{eq:Hcl}), and
484: \begin{eqnarray}
485:   \hat{H}_2 & = & \frac{S}{4} \sum_{i,j} J_{ij} [
486:   (\bd{e}^{+}_i \cdot \bd{e}^-_j ) \; b_i^{\dagger} b_j
487:   + (\bd{e}^{-}_i \cdot \bd{e}^+_j ) \; b_j^{\dagger} b_i
488:   \nonumber 
489:   \\
490:   & & \hspace{12mm} +
491:   (\bd{e}^{+}_i \cdot \bd{e}^+_j ) \; b_i^{\dagger} b_j^{\dagger}
492:   + (\bd{e}^{-}_i \cdot \bd{e}^-_j ) \; b_j b_i
493:   ]
494:   \nonumber
495:   \\
496:   & - &  \frac{S}{2} \sum_{i,j} J_{ij}  ( {\hat{\bd{m}}}_i \cdot \hat{\bd{m}}_j )
497:   [ b^{\dagger}_i b_i + b^{\dagger}_j b_j ]
498:   \nonumber
499:   \\
500:   & + &\sum_i ( \bd{h}_i \cdot \hat{\bd{m}}_i)  b^{\dagger}_i b_i
501:   \label{eq:H2}
502:   \; .
503: \end{eqnarray} 
504: Note that the contribution from $\hat{H}^{\prime}$ in
505: Eq.~(\ref{eq:Hprime}) is of order $S^{1/2}$ and hence can be neglected
506: within linear spin-wave theory.  Eq.~(\ref{eq:H2}) together with
507: Eqs.~(\ref{eq:Hcl}) and (\ref{eq:classical}) completely determine the
508: spin-wave spectrum of any Heisenberg magnet in an arbitrary
509: inhomogeneous field.
510: 
511: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
512: \section{Spin waves in the distorted honeycomb lattice}
513: \label{sec:honeycomb}
514: 
515: \subsection{Classical ground state}
516: 
517: Let us apply the general formalism of the previous section to our
518: bipartite lattice antiferromagnet in a uniform external magnetic field $B
519: \hat{\bd{e}}_x$ along the $x$ axis. We denote by $\hat{\bd{e}}_\alpha$
520: fixed unit vectors in direction $\alpha=x,y,z$.  For technical reasons 
521: we introduce an additional staggered magnetic field $\zeta_i B_s
522: \hat{\bd{e}}_z$ in the $z$-direction, where $\zeta_i = 1$ if
523: ${\bd{r}}_i$ belongs to the A-sublattice and $\zeta_i = -1$ if
524: ${\bd{r}}_i$ belongs to the B-sublattice.  
525: This auxiliary staggered field will be determined self-consistently 
526: in Sec.~\ref{subsec:magn}  
527: to insure a vanishing staggered magnetization at finite
528: temperatures. The total magnetic field is thus
529: \begin{equation}
530:   \bd{h}_i = g \mu_{\mathrm{B}} [B \hat{\bd{e}}_x
531:   +  \zeta_i B_s \hat{\bd{e}}_z]
532:   \; .
533:   \label{eq:hidef}
534: \end{equation}
535: The classical ground state configuration is then $\hat{\bd{m}}_i =
536: \zeta_i \cos \theta \hat{\bd{e}}_z + \sin \theta \hat{\bd{e}}_x $, as
537: shown in Fig.~\ref{fig:spingroundstate}.
538: 
539: For convenience we introduce the notation $n_0 = \cos \theta $ and
540: $m_0 = \sin \theta$.  Physically, $m_0$ corresponds to the classical
541: limit ($S \rightarrow \infty$) of the normalized uniform magnetization
542: \begin{equation}
543:   m  =  \frac{1}{NS} \sum_{i} \langle \hat{\bd{e}}_x \cdot \bd{S}_i \rangle
544:   \; , 
545:   \label{eq:mdef}
546: \end{equation}
547: while $n_0$ corresponds to the $S \rightarrow \infty$ limit of the
548: normalized staggered magnetization
549: \begin{equation}
550:   n  =  \frac{1}{NS} \sum_{i} \zeta_i \langle \hat{\bd{e}}_z \cdot 
551:   \bd{S}_i \rangle
552:   \; .
553:   \label{eq:ndef}
554: \end{equation}
555: By symmetry, the uniform magnetization points into the $x$-direction,
556: while the staggered magnetization points into the $z$-direction.  The
557: natural dimensionless measure for the strength of the fields is ${h} =
558: \chi_0 g \mu_{\mathrm{B}} B$ and ${h}_s = \chi_0 g \mu_{\mathrm{B}} B_s$, 
559: where $\chi_0 = (2 \tilde{J}_0 S )^{-1}$ 
560: is the classical uniform susceptibility.  Here
561: $\tilde{J}_0 = \sum_{\nu} J_{\nu}$ is the ${\bd{k}} =0$ component of
562: the Fourier transform of the exchange couplings
563: \begin{equation}
564:   \tilde{J}_{\bd{k}} = \sum_{\nu} e^{ -i {\bd{k}}
565:   \cdot \bd{\delta}_{\nu} } J_{\nu}\,.
566:   \label{eq:jdef} 
567: \end{equation}
568: For the special choice of the field $\bd{h}_i$ given in
569: Eq.~(\ref{eq:hidef}) our general Eq.~(\ref{eq:classical}) reduces 
570: to the simple relation
571: \begin{equation}
572:   {h} = m_0 [ 1 + {h}_s / n_0 ]
573:   \; ,
574:   \label{eq:hh}
575: \end{equation}
576: which together with $n_0^2 + m_0^2 =1$ determines the classical N\'eel
577: order parameter $n_0$ and the classical uniform magnetization $m_0$ as
578: functions of the fields $h$ and $h_s$.  Note that $\hat{\bd{m}}_\mathrm{A}
579: \cdot \hat{\bd{m}}_\mathrm{B} = m_0^2 - n_0^2$, and with the special transverse
580: basis shown in Fig.~\ref{fig:spingroundstate}
581: \begin{eqnarray}
582:   \bd{e}_\mathrm{A}^p \cdot \bd{e}_\mathrm{B}^{p^{\prime}} & = & 2 [\delta_{ p,p^{\prime}} n_0^2
583:   +  \delta_{p, - p^{\prime}} m_0^2 ] \; ,
584:   \label{eq:epep}
585:   \\
586:   \hat{\bd{m}}_\mathrm{A} \cdot {\bd{e}}_\mathrm{B}^p & = &
587:   2 i p n_0 m_0 =
588:   - \hat{\bd{m}}_\mathrm{B} \cdot {\bd{e}}_\mathrm{A}^p
589:   \; \; .
590:   \label{eq:medot}
591: \end{eqnarray}
592: 
593: \subsection{Spin-wave dispersion}
594: 
595: To obtain the spin-wave dispersion, we must diagonalize $\hat{H}_2$ in
596: Eq.~(\ref{eq:H2}) for the special ground-state spin configuration
597: discussed above.  Therefore, we first perform Fourier transformations
598: separately on each sublattice,
599: \begin{subequations}
600:   \begin{eqnarray}
601:     b_i & = & \sqrt{ \frac{2}{N} } \sum_{\bd{k}} e^{ i {\bd{k}} \cdot {\bd{r}}_i }
602:     a_{ \bd{k}} \; \; , \; \; \mbox{for $\bd{r}_i \in$ A }
603:     \; ,
604:     \label{eq:AFT}
605:     \\
606:     b_i & = & \sqrt{ \frac{2}{N} } \sum_{\bd{k}} e^{ i {\bd{k}} \cdot {\bd{r}}_i }
607:     b_{ \bd{k}} \; \; , \; \; \mbox{for $\bd{r}_i \in$ B }
608:     \; ,
609:     \label{eq:BFT}
610:   \end{eqnarray}
611: \end{subequations}
612: where the wave-vector sums are over the reduced (magnetic) Brillouin
613: zone of the honeycomb lattice shown in Fig.~\ref{fig:BZ}.
614: With the above definitions we obtain
615: \begin{eqnarray}
616:   \hat{H}_2 & = & \tilde{J}_0 S \sum_{ \bd{k}} \bigl[
617:   A ( a^{\dagger}_{\bd{k}} a^{}_{\bd{k}} + b^{\dagger}_{\bd{k}} b^{}_{\bd{k}} )
618:   + B^{}_{\bd{k}} b^{}_{ - \bd{k}} a^{}_{\bd{k}}  + B_{\bd{k}}^{\ast} 
619:   a^{\dagger}_{ \bd{k}} b^{\dagger}_{- \bd{k}}   
620:   \nonumber
621:   \\
622:   &  & \hspace{15mm} +
623:   C^{}_{ \bd{k}}  b^{\dagger}_{  \bd{k}} a^{}_{\bd{k}} + C_{\bd{k}}^{\ast} 
624:   a^{\dagger}_{\bd{k}} b^{}_{ \bd{k}}
625:   \bigr]
626:   \; ,
627:   \label{eq:H2a}
628: \end{eqnarray}
629: where $A =1 + 2 {h}_s / n_0$, $B_{\bd{k}} = n_0^2 \tilde{J}_{\bd{k}} /
630: \tilde{J}_0$, and $C_{\bd{k}} = m_0^2 \tilde{J}_{\bd{k}} /
631: \tilde{J}_0$.  On a honeycomb lattice $\tilde{J}_{\bd{k}} =|
632: \tilde{J}_{\bf{k}} | e^{i \phi_{\bd{k}} }$ is complex, so that
633: $B_{\bd{k}} = | B_{\bf{k}} | e^{i \phi_{\bd{k}} }$ and $C_{\bd{k}} = |
634: C_{\bf{k}} | e^{i \phi_{\bd{k}} }$.  Using $\phi_{ - \bd{k}} = -
635: \phi_{ \bd{k}}$, it is easy to see that these phase factors can be
636: removed from Eq.~(\ref{eq:H2a}) via the gauge transformation
637: $\tilde{a}_{ \bd{k}} = e^{i \phi_{\bd{k}}} a_{\bd{k}}$.  Introducing
638: then new canonical boson operators
639: \begin{equation}
640:   c_{\bd{k} \sigma } = \frac{1}{\sqrt{2}} \left[ \tilde{a}_{\bd{k}} + \sigma
641:     b_{\bd{k}} \right]
642:   \;  \; , \; \; \sigma = \pm 1
643:   \; ,
644:   \label{eq:cpmdef}
645: \end{equation}
646: the hamiltonian (\ref{eq:H2a}) assumes the block-diagonal form,
647: \begin{eqnarray}
648:   \hat{H}_2 & = & \frac{\tilde{J}_0 S}{2} \sum_{ \bd{k}  \sigma} \bigl[
649:   (A + \sigma | C^{}_{\bd{k}} | ) ( c^{\dagger}_{\bd{k} \sigma} c^{}_{\bd{k} \sigma} 
650:   + c^{\dagger}_{-\bd{k} \sigma} c^{}_{-\bd{k} \sigma} )
651:   \nonumber
652:   \\
653:   &  & \hspace{12mm} +
654:   \sigma | B^{}_{ \bd{k}} |  
655:   ( c^{\dagger}_{  \bd{k} \sigma} c^{\dagger }_{-\bd{k} \sigma} +  
656:   c^{}_{ \bd{k} \sigma} c^{}_{ -\bd{k} \sigma} )
657:   \bigr]
658:   \; .
659:   \label{eq:H2b}
660: \end{eqnarray}
661: The diagonalization is completed by means of the Bo\-go\-liu\-bov
662: transformation,
663: \begin{equation}
664:   \left( \begin{array}{c}
665:       c_{ \bd{k} \sigma } \\
666:       c^{\dagger}_{ - \bd{k} \sigma }  \end{array}
667:   \right) =
668:   \left( \begin{array}{cc}
669:       u_{ \bd{k} \sigma} & - \sigma v_{\bd{k} \sigma} \\
670:       - \sigma  v_{\bd{k} \sigma} & u_{ \bd{k} \sigma} \end{array} \right)
671:   \left( \begin{array}{c}
672:       d_{ \bd{k} \sigma } \\
673:       d^{\dagger}_{ - \bd{k} \sigma }  \end{array}
674:   \right)
675:   \; ,
676:   \label{eq:bogoliubov}
677: \end{equation}
678: where
679: \begin{subequations}
680:   \begin{eqnarray}
681:     u_{ \bd{k} \sigma } & = & 
682:     \sqrt{ \frac{ A + \sigma | C_{\bd{k}} | + \epsilon_{\bd{k} \sigma} }{ 2 \epsilon_{\bd{k} \sigma}} }
683:     \; ,
684:     \label{eq:ukdef}
685:     \\
686:     v_{ \bd{k} \sigma } & = & 
687:     \sqrt{ \frac{ A + \sigma | C_{\bd{k}} | - \epsilon_{\bd{k} \sigma} }{ 2 \epsilon_{\bd{k} \sigma}} }
688:     \; ,
689:     \label{eq:vkdef}
690:   \end{eqnarray}
691: \end{subequations}
692: with the dimensionless energy dispersion
693: \begin{equation} 
694:   \epsilon_{\bd{k} \sigma} = \sqrt{
695:     (  A + \sigma | C_{\bd{k}} | )^2 - | B_{\bd{k}} |^2 }
696:   \label{eq:epsilondef}
697:   \; .
698: \end{equation}
699: Defining $\gamma_{\bd{k}} = \tilde{J}_{\bd{k}} / \tilde{J}_0$, we may
700: write
701: \begin{equation}
702:   \epsilon_{\bd{k} \sigma} = 
703:   \Bigl[
704:   \bigl( 1 +  \frac{2 {h}_s}{n_0} + \sigma | \gamma_{\bd{k}} | \bigr)
705:   \bigl( 1 +  \frac{2 {h}_s}{n_0} - 
706:   \sigma ( n_0^2 - m_0^2 )| \gamma_{\bd{k}} | \bigr) \Bigr]^{1/2}
707:   \; .
708:   \label{eq:epsilon2}
709: \end{equation}
710: %
711: %
712: \begin{figure}[tb]
713:   \centering
714:   \psfrag{x}{$k_x$}
715:   \psfrag{y}{$k_y$}
716:   \psfrag{b1}{${\bd b}_{1}$}
717:   \psfrag{b2}{${\bd b}_{2}$}
718:   \psfrag{f}{$\varphi$}
719:   \epsfig{file=bz.eps,width=65mm}
720:   \vspace{-2mm}
721:   \caption{%
722:     Reduced Brillouin zone of the distorted honeycomb lattice.  The
723:     primitive vectors are ${\bd b}_1 =
724:     \frac{2\pi}{a_1\sin{\varphi}} (\sin{\varphi}\,{\hat{\bd e}}_x -
725:     \cos{\varphi}\,{\hat{\bd e}}_y) $ and ${\bd b}_2 =
726:     \frac{2\pi}{a_2\sin{\varphi}}{\hat{\bd e}}_y$, where $a_1$,
727:     $a_2$ and the angle $\varphi$ are defined in
728:     Fig.~\ref{fig:lattice}.  }
729:   \label{fig:BZ}
730: \end{figure}
731: % 
732: %
733: In terms of the new operators $d_{\bd{k} \sigma}$ the quadratic spin
734: wave hamiltonian $\hat{H}_2$ is diagonal,
735: \begin{equation}
736:   \hat{H}_2 = \tilde{J}_0 S \sum_{ \bd{k} \sigma} 
737:   \left\{ \epsilon^{}_{ \bd{k} \sigma} d^{\dagger}_{ \bd{k} \sigma}
738:     d^{}_{ \bd{k} \sigma} + \frac{1}{2} \left[ \epsilon_{ \bd{k} \sigma}
739:       - (A + \sigma | C_{\bd{k}} | ) \right] \right\}
740:   \; .
741: \end{equation}
742: The low temperature properties of the magnet are determined by the long-wavelength
743: behavior of the spin-wave dispersions, which follow from the expansion
744: for small ${\bd k}$,
745: \begin{equation} 
746:   |\gamma_{\bd k}| \approx 1 - \frac12 \sum_{\alpha\beta} k_{\alpha} A_{\alpha\beta}k_{\beta}\,, 
747: \end{equation} 
748: where ${\bd A}$ is a matrix with elements
749: \begin{equation} 
750:   A_{\alpha\beta} =  
751:   \sum_{\nu}  \frac{J_\nu}{\tilde{J}_0}  (  \bd{\delta}_{\nu}  \cdot \hat{\bd{e}}_{\alpha} )
752:   ( \bd{\delta}_{\nu}  \cdot \hat{\bd{e}}_{\beta} )
753:   - 
754:   \sum_{\nu,\nu'} \frac{J_\nu J_{\nu'}}{\tilde{J}_0^2}   
755:   (  \bd{\delta}_{\nu}  \cdot \hat{\bd{e}}_{\alpha} )
756:   ( \bd{\delta}_{\nu^{\prime}}  \cdot \hat{\bd{e}}_{\beta} )
757:   \,. 
758:   \label{eq:Amatrixdef}
759: \end{equation} 
760: Since ${\bd A}$ is symmetric, an orthogonal basis can always be chosen
761: such that ${\bd A}$ is diagonal, with eigenvalues $A_{\alpha}$. In this basis
762: \begin{equation} 
763:   |\gamma_{\bd k}| \approx 1 - \frac12\sum_{\alpha} A_{\alpha} k_{\alpha}^2\,. 
764: \end{equation} 
765: The matrix ${\bd A}$ is positive, since 
766: \begin{equation} 
767:   |\gamma_{\bd k}| \leq \sum_{\nu} \left|\frac{J_{\nu}}{\tilde{J}_0}\right| = 1\,, 
768: \end{equation} 
769: where the last equality assumes that all couplings have the same sign.
770: We can thus define effective length scales $\ell_\alpha$ by setting
771: $A_{\alpha} = \ell_{\alpha}^2$.  For a $D$-dimensional hypercubic
772: lattice with lattice spacing $a$ we have $\ell^2_{\alpha}={a^2}/{D}$.
773: For our honeycomb lattice shown in Fig.~\ref{fig:lattice} with $ |
774: \bd{\delta}_1 | = | \bd{\delta} _3 | $ and $J_1 = J_3$ the
775: eigenvectors of $\bd{A}$ are parallel to the $x$-axis and the
776: $y$-axis, with corresponding eigenvalues $\ell_x^2 = (J_1 / 2
777: \tilde{J}_0 ) a_1^2$ and $\ell_y^2 = (2 J_1 J_2 / \tilde{J}_0^2 )
778: a_2^2 \sin^2 \varphi$.  The spin-wave velocities $c_{\alpha} =
779: \tilde{J}_0 S \ell_{\alpha}$ along the two principal directions are
780: thus
781: \begin{eqnarray}
782:   c_x &=& S \sqrt{ \frac{J_1 \tilde{J}_0 }{2}}\,a_1\,,\\
783:   c_y &=& S  \sqrt{ 2 J_1 J_2} \,a_2\sin{\varphi}\,.
784: \end{eqnarray}
785: Note that for $ J_2 \rightarrow 0$ the velocity $c_y$ vanishes, so
786: that the system becomes one-dimensional, as is obvious from
787: Fig.~\ref{fig:lattice}.  On the other hand, for $J_1 \rightarrow 0$
788: both velocities vanish, because in this limit the system consists of
789: decoupled dimers.
790: 
791: For ${h}_s =0$ only the mode $\epsilon_{ \bd{k}-}$ is gapless for
792: ${\bd{k}} \rightarrow 0$, while the mode $\epsilon_{ \bd{k}+}$ has the
793: gap $2 m_0$.  To give a more explicit form for the long-wavelength
794: spin-wave dispersions, we further assume $h_s \ll n_0$. Then
795: \begin{eqnarray} 
796:   \epsilon_{{\bd k}-} & \approx & n_0 \left[\frac{4h_s}{n_0} 
797:     +\sum_{\alpha}(\ell_{\alpha} k_{\alpha})^2\right]^{1/2}\,,
798:   \label{eq:epsilonminuslong}
799:   \\ 
800:   \epsilon_{{\bd k}+} & \approx & \Bigl[ 4m_0^2+\frac{4h_s}{n_0}(1+m_0^2) 
801:   \nonumber
802:   \\
803:   & & 
804:   +(n_0^2-2m_0^2)\sum_{\alpha}(\ell_{\alpha}k_{\alpha})^2 \Bigr]^{1/2}\,.   
805:   \label{eq:epsilonpluslong}
806: \end{eqnarray} 
807: For $n_0\to0$ the expansion (\ref{eq:epsilonminuslong}) is not
808: appropriate any longer and for $h_s=0$ the dispersion $\epsilon_{{\bd k}-}$
809: becomes purely quadratic at $n_0=0$.  Before this happens, there is a
810: critical field $0<h^*<1$ at which the curvature of the dispersion
811: $\epsilon_{{\bd k}-}$
812: changes sign. The positive curvature for $h>h^*$ results in an
813: instability of magnons towards a spontaneous decay into two magnon
814: states.\cite{Zhitomirsky99} Furthermore, if an anisotropic exchange is
815: considered, the anisotropy gap $\Delta$ is strongly renormalized by
816: magnon interactions.\cite{Maleyev00,Syromyatnikov01} As the influence
817: of these instabilities on the thermodynamic properties is unclear at
818: the moment, they will not be further considered in this work.
819: 
820: \subsection{Uniform and staggered magnetization}
821: \label{subsec:magn}
822: 
823: We now calculate the leading spin-wave corrections to the normalized
824: uniform- and the staggered magnetization as defined in
825: Eqs.~(\ref{eq:mdef}) and (\ref{eq:ndef}).  A standard expansion in
826: powers of $1/S$ gives
827: \begin{eqnarray}
828:   m & = & \frac{m_0^2}{{h}} \left[ 1 + \frac{n {h}_s}{n_0^2}
829:     - \frac{ F (   {h} , {h}_s )   }{S}  \right]
830:   \label{eq:Fsw}
831:   \; ,
832:   \\
833:   n & = & \frac{1}{n_0} \left[ 1 - m_0 m - \frac{  I ( {h} , {h}_s )  }{S} 
834:   \right]
835:   \label{eq:Isw}
836:   \; ,
837: \end{eqnarray}
838: where
839: \begin{equation}
840:   F (   {h} , {h}_s )  = \frac{1}{N} \sum_{ \bd{k} \sigma}
841:   \frac{  n_{\bd{k} \sigma}  + \frac{1}{2} }{ \epsilon_{ \bd{k} \sigma} } 
842:   \sigma | \gamma_{\bd{k}} |   
843:   \Bigl( 1 + \frac{2 {h}_s}{n_0} + \sigma | \gamma_{\bd{k}} | \Bigr) 
844:   \; ,
845:   \label{eq:Fdef}
846: \end{equation}
847: and
848: \begin{equation}
849:   I (   {h} ,{h}_s )  = - \frac{1}{2} + \frac{1}{N} \sum_{ \bd{k} \sigma}
850:   \frac{  n_{\bd{k} \sigma}  + \frac{1}{2} }{ \epsilon_{ \bd{k} \sigma} } 
851:   \Bigl( 1 + \frac{2 {h}_s}{n_0} + \sigma m_0^2 | \gamma_{\bd{k}} | \Bigr)
852:   \; .
853:   \label{eq:Idef}
854: \end{equation}
855: Here $n_{\bd{k} \sigma} = [ e^{ \tilde{J}_0 S \epsilon_{\bd{k} \sigma}
856:   /T } -1 ]^{-1}$ is the Bose function.  The parameters $n_0$ and
857: $m_0$ on the right-hand sides of Eqs.~(\ref{eq:Fsw}--\ref{eq:Idef})
858: are determined as functions of the fields $h$ and $h_s$ by
859: Eq.~(\ref{eq:hh}) and $n_0^2 + m_0^2 =1$.  Note that for $S
860: \rightarrow \infty$ the solutions of Eqs.~(\ref{eq:Fsw}) and
861: (\ref{eq:Isw}) correctly approach $n =n_0$ and $m = m_0$: in this
862: limit Eq.~(\ref{eq:Fsw}) reduces to Eq.~(\ref{eq:hh}), while
863: Eq.~(\ref{eq:Isw}) simply becomes another way of writing $n_0^2 +
864: m_0^2 =1$.  In the thermodynamic limit, we transform Brillouin zone
865: sums to integrals according to
866: \begin{equation} 
867:   \frac{2}{N} \sum_{\bd k} \stackrel{N\to\infty}{\longrightarrow} 
868:   V_u \int_{\text{BZ}}\frac{d^2k}{(2\pi)^2}\,, 
869: \end{equation} 
870: where $V_u = a_1 a_2 \sin \varphi$ is the area of the magnetic unit
871: cell in real space and the integral is over the reduced Brillouin zone
872: shown in Fig.~\ref{fig:BZ}.
873: 
874: At $T=0$ and $h_s=0$ expressions similar to (\ref{eq:Fsw}) and
875: (\ref{eq:Isw}) have been discussed previously.\cite{Zhitomirsky98}
876: Only $m(h)$ was given explicitly and a renormalization of the canting
877: angle was found by considering spin-wave interactions. Yet, to a given
878: order in $1/S$ it is easier to calculate $m$ and $n$ directly as
879: derivatives of the free energy with respect to $h$ and $h_s$. Very
880: recently, the renormalized canting angle was also used to analyze the
881: behavior of $n(h)$ at $T=0$ for a more complicated
882: geometry.\cite{Veillette05}
883: 
884: At any finite temperature the integral $I(h,0)$ is infrared divergent
885: in two dimensions, signaling the absence of long-range
886: antiferromagnetic order, in accordance with the
887: Hohenberg-Mermin-Wagner theorem.\cite{Hohenberg67} At first sight,
888: it thus seems that the finite-temperature magnetization curve cannot
889: be calculated within our spin-wave approach.  Fortunately, there is a
890: straightforward way to obtain an approximate expression for the
891: magnetization even at finite $T$. The crucial observation is that if
892: we set $n=0$ in Eqs.~(\ref{eq:Fsw}) and (\ref{eq:Isw}), these equations can be
893: interpreted as a condition for the staggered field $h_s$ that is necessary to
894: enforce a vanishing staggered magnetization.  The solution
895: $h_s=h_s(h)$ as a function of the uniform field $h$ is not a physical
896: external staggered field, but an internal effective field that is
897: generated by strong fluctuations.  In fact, the field $h_s(h)$ is
898: nothing but the Lagrange multiplier introduced in Takahashi's modified
899: spin-wave theory.\cite{Takahashi89,Kollar03}  It is well known that
900: the internal field is related to a finite correlation length $\xi$, as
901: we will further discuss in Sec.~\ref{sec:stagg}.  Numerically, we
902: calculate the uniform magnetization $m(h,T)$ at finite temperature $T$
903: by adjusting $h_s$ for fixed external field $h$ such that the
904: condition $n=0$ is fulfilled in Eqs.~(\ref{eq:Fsw}) and
905: (\ref{eq:Isw}).  Using this $h_s(h)$ in Eq.~(\ref{eq:Fsw}) then
906: directly yields $m(h,T)$.
907: 
908: We must keep in mind that the staggered field $h_s$ does not respect
909: the rotational symmetry of the original hamiltonian, which for $h=0$
910: corresponds to a global O(3) symmetry and for $h>0$ is reduced to a
911: global O(2) symmetry around the axis of the uniform field. With the
912: parametrization that explicitly breaks this symmetry, we should
913: therefore only calculate rotationally invariant quantities.\cite{Kopietz97} 
914: Below, we
915: will find a disagreement between a rotationally invariant evaluation
916: of the zero-field uniform susceptibility and the slope of 
917: $\partial m/ \partial h$ for $h\to0$. 
918: We attribute this discrepancy to the fact that $\partial m/\partial h {\vert}_{h\to0}$
919: does not respect the O(3) symmetry in this limit. Generally, we expect
920: our approach for the finite temperature magnetization to be reasonable
921: only for $h>h_s(h,T)$. In Sec.~\ref{sec:stagg} we will see that $h_s$
922: is exponentially small at low temperatures, such that $h>h_s(h,T)$ is
923: fulfilled even for very small external fields. The condition
924: $h>h_s(h,T)$ then roughly gives a limit of validity of our approach in
925: terms of the temperature as $T\lesssim 0.5 \tilde{J}_0S$.
926: The fact that the limits $T \to 0$ and $h \to 0$ do not commute in a 
927: modified spin-wave expansion
928: was first noticed by Takahashi.\cite{Takahashi89}
929: 
930: \subsection{Uniform susceptibility}
931: 
932: In order to calculate the rotationally invariant uniform zero-field
933: susceptibility per spin
934: \begin{equation}
935:   \label{eq:chidef}
936:   \chi = \frac{1}{T N} \sum_{i,j} \langle \bd{S}_i \cdot \bd{S}_j \rangle\,,
937: \end{equation}
938: we set the uniform magnetic field $B=0$ in \mbox{Eq.(\ref{eq:hidef})}.
939: In this case $m=m_0=0$ and $n_0=1$, so that we obtain a doubly
940: degenerate mode in \mbox{Eq. (\ref{eq:epsilon2})} with dispersion
941: \begin{equation} 
942:   \epsilon_{\bd{k}\sigma}= \epsilon_{\bd{k}} = \sqrt{ (1+2h_s)^2 - |\gamma_{\bd{k}}|^2 }\,,
943:   \label{eq:epsilon0}
944: \end{equation}
945: and the expression for the staggered magnetization (\ref{eq:Isw})
946: reduces to
947: \begin{equation}
948:   \label{eq:I0SW}
949:   n = 1 + \frac{1}{2S} - 
950:   \frac{2}{NS}\sum_{\bd{k}}\frac{n_{\bd{k}}+\frac{1}{2}}{\epsilon_{\bd{k}}}
951:   \left( 1 + 2h_s \right)\,.
952: \end{equation}
953: As explained in the previous section we use a self-consistently
954: determined staggered field $h_s$ to enforce a vanishing order
955: parameter $n=0$.
956: 
957: The susceptibility (\ref{eq:chidef}) can be written as
958: \begin{equation}
959:   \label{eq:chi}
960:   \chi = \frac{1}{T}\left\langle\bd{S}_{\bd{q},+} \cdot \bd{S}_{-\bd{q},+}\right\rangle_{\bd{q}=0}\,,
961: \end{equation}
962: where we have defined the linear combinations ($\sigma=\pm$)
963: \begin{equation}
964:   \bd{S}_{\bd{q},\sigma} = \frac1{\sqrt{2}}\left(\bd{S}_{\bd{q}}^\mathrm{A}+
965:     \sigma\bd{S}_{\bd{q}}^\mathrm{B}\right)
966: \end{equation}
967: of the Fourier-transformed spin operators on each sublattice
968: \begin{equation}
969:   \label{eq:SFT}
970:   \bd{S}_{\bd{q}}^{\mathrm{A}/\mathrm{B}} = \sqrt{\frac{2}{N}}\sum_{\bd{r}_i\in \mathrm{A}/\mathrm{B}}e^{-i\bd{q}\cdot\bd{r}_i}\,\bd{S}_i\,.
971: \end{equation}
972: Next we decompose the susceptibility into a transverse and a
973: longitudinal part
974: \begin{equation}
975:   \label{eq:chitl}
976:   \chi = \chi^{+-} + \chi^{zz}\,,
977: \end{equation}
978: where 
979: \begin{eqnarray}
980:   \label{eq:chit}
981:   \chi^{+-}  & = & \frac{1}{2T} \left\langle 
982:   S^{+}_{\bd{q},+}S^{-}_{-\bd{q},+} + S^{+}_{\bd{q},+}S^{-}_{-\bd{q},+}
983:   \right\rangle_{\bd{q}=0}\,,\\
984:   \label{eq:chil}
985:   \chi^{zz}  & = & \frac{1}{T} 
986:   \left\langle S^{z}_{\bd{q},+}S^{z}_{-\bd{q},+}\right\rangle_{\bd{q}=0}\,.
987: \end{eqnarray}
988: We map the spin operators~(\ref{eq:SFT}) onto canonical boson 
989: operators via a Dyson-Maleev transformation\cite{Dyson56,Maleev57} 
990: and evaluate the thermal expectation 
991: values of the noninteracting state using the Wick theorem.
992: Then the  transverse susceptibility (\ref{eq:chit}) is
993: proportional to the right hand side of \mbox{Eq. (\ref{eq:I0SW})}, and
994: thus vanishes if we require $n=0$.
995: Therefore, in our approximation only the longitudinal
996: part contributes to the rotationally invariant uniform susceptibility,
997: \begin{equation}
998:   \label{eq:chisw}
999:   \chi =   \frac{2}{TN} \sum_{\bd{k}} n_{\bd{k}} (n_{\bd{k}}+1)\,.
1000: \end{equation}
1001: Apart from a different normalization,
1002: this result has been obtained previously in Takahashi's
1003: approach.\cite{Takahashi89}
1004: We evaluate \mbox{Eq. (\ref{eq:chisw})} numerically in the thermodynamic limit.
1005: 
1006: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1007: \section{Results}
1008: \label{sec:results}
1009: 
1010: \subsection{Zero temperature uniform and staggered magnetization}
1011: 
1012: In Figs.~\ref{fig:magn} and \ref{fig:magnst} we show results for the
1013: uniform and staggered magnetization at zero temperature. In two
1014: spatial dimensions, there are no divergent contributions to the
1015: integrals in Eqs.~(\ref{eq:Fsw}) and (\ref{eq:Isw}), indicating true
1016: long range order.  We can thus set $h_s=0$ and consequently $m_0=h$.
1017: As the deviations from the classical curves are rather small for
1018: $S=5/2$, we present the curves for the extreme quantum case $S=1/2$.
1019: 
1020: %
1021: % 
1022: \begin{figure}[tb]
1023:   \begin{center}
1024:     \psfrag{m}{$m$}
1025:     \psfrag{h}{$h$}
1026:     \psfrag{0}{\footnotesize{$0$}}
1027:     \psfrag{0.2}{\footnotesize{$0.2$}}
1028:     \psfrag{0.4}{\footnotesize{$0.4$}}
1029:     \psfrag{0.6}{\footnotesize{$0.6$}}
1030:     \psfrag{0.8}{\footnotesize{$0.8$}}
1031:     \psfrag{1}{\footnotesize{$1$}}
1032:     \psfrag{classical}{\footnotesize{classical}}
1033:     \psfrag{square lattice}{\footnotesize{square lattice}}
1034:     \psfrag{linear chain}{\footnotesize{linear chain}}
1035:     \psfrag{honeycomb lattice, J1=J2}{\footnotesize{honeycomb lattice, $J_1/J_2=1$}}
1036:     \psfrag{honeycomb lattice, J1=10J2}{\footnotesize{honeycomb lattice, $J_1/J_2=10$}}
1037:     \epsfig{file=magn-T0.eps,width=70mm,angle=270}~~~~~~
1038: 
1039:   \end{center}
1040:   \vspace{-4mm}
1041:   \caption{%
1042:     Normalized uniform magnetization $m (h)$ for $T=0$ and $h_s=0$.
1043:     The solid line is the zero-temperature magnetization curve for the
1044:     honeycomb lattice with $S=1/2$ and $J_1 = J_2$.  For comparison we
1045:     also show the corresponding curve for a square lattice and exact
1046:     results for a linear antiferromagnetic chain.\cite{Griffiths64} 
1047:     However, the $S=1/2$ chain is critical, so it is not surprising
1048:     that it is poorly described by means of the spin-wave theory.
1049:     Note that for $h_s=0$ the classical equation (\ref{eq:hh}) is simply
1050:     $m_0=h$.}
1051:   \label{fig:magn}
1052: \end{figure}
1053: %
1054: %
1055: The uniform magnetization shows a positive curvature for all $0\leq h
1056: <1$ and lies generally below the classical straight
1057: line.\cite{Zhitomirsky98} This tendency is stronger for the honeycomb
1058: lattice and is even more pronounced for anisotropic exchange couplings
1059: with $J_1\gg J_2$. The number of nearest neighbors $z=3$ for the
1060: honeycomb lattice is lower than for the square lattice ($z=4$), and in
1061: the limit $J_2\ll J_1$ the system is almost one-dimensional. The
1062: observed tendency thus simply corresponds to increased quantum
1063: fluctuations in low dimensions. Beyond the saturation field $h=1$ the
1064: ground state has full collinear ferromagnetic order. This
1065: state as well as single magnon excitations above it are easily shown
1066: to be exact eigenstates. As the single magnon states become gapless at
1067: exactly the classical value $h=1$, the saturation field is
1068: not changed by quantum fluctuations or magnon interactions.  The limit
1069: $h\to1$ is reached with infinite slope in $m(h)$. The leading behavior
1070: is given by
1071: \begin{equation}
1072:   m =  1 + \frac{V_u}{4 \ell_x \ell_y}\frac{\delta h}{\pi S}
1073:   \ln \left(4\delta h\right)\,,
1074: \end{equation} 
1075: where $\delta h=1-h$. This logarithmic asymptotics was first discussed in the
1076: language of Bose condensation of magnons below the saturation
1077: field\cite{Gluzman93} and was later found for the square lattice
1078: ($V_u/4\ell_x\ell_y$=1) within linear spin-wave
1079: theory.\cite{Zhitomirsky98} For our distorted honeycomb lattice, we
1080: have
1081: \begin{equation}
1082:   \frac{V_u}{\ell_x\ell_y}=\sqrt{\frac{(2J_1+J_2)^3}{J_1^2J_2}}\,,
1083: \end{equation}
1084: which diverges for $J_1\to0$ or $J_2\to0$ and thus exemplifies the
1085: increasing deviations from the classical curve for strongly
1086: anisotropic exchange couplings.
1087: %
1088: %
1089: \begin{figure}[tb]
1090:   \begin{center}
1091:     \psfrag{n}{$n$}
1092:     \psfrag{h}{$h$}
1093:     \psfrag{0}{\footnotesize{$0$}}
1094:     \psfrag{0.2}{\footnotesize{$0.2$}}
1095:     \psfrag{0.4}{\footnotesize{$0.4$}}
1096:     \psfrag{0.6}{\footnotesize{$0.6$}}
1097:     \psfrag{0.8}{\footnotesize{$0.8$}}
1098:     \psfrag{1}{\footnotesize{$1$}}
1099:     \psfrag{classical}{\footnotesize{classical}}
1100:     \psfrag{square lattice}{\footnotesize{square lattice}}
1101:     \psfrag{honeycomb lattice, J1=J2}{\footnotesize{honeycomb lattice, $J_1/J_2=1$}}
1102:     \psfrag{honeycomb lattice, J1=10J2}{\footnotesize{honeycomb lattice, $J_1/J_2=10$}}
1103:     \psfrag{honeycomb lattice, J1=100J2}{\footnotesize{honeycomb lattice, $J_1/J_2=100$}}
1104:     \epsfig{file=stag-T0.eps,width=70mm,angle=270}~~~~~~
1105:   \end{center}
1106:   \vspace{-4mm}
1107:   \caption{%
1108:     Normalized staggered magnetization $n(h)$ at $T=0$ for honeycomb 
1109:     (solid line) and square lattice with $S=1/2$ (dotted line). The 
1110:     classical equation $n_0=\sqrt{1-h^2}$ is plotted for comparison. We
1111:     also show the curves for the anisotropic cases $J_1/J_2=10,100$.}
1112:   \label{fig:magnst}
1113: \end{figure}
1114: %
1115: %
1116: %
1117: 
1118: The staggered magnetization in Fig.~\ref{fig:magnst} shows a
1119: non-monotonic dependence on the applied uniform field. For vanishing
1120: $h$ the staggered magnetization decreases as we lower the effective
1121: dimensionality. An external field apparently suppresses quantum
1122: fluctuations and $n(h)$ first increases with $h$ before it reaches a
1123: maximum and then vanishes for $h\to1$ with infinite slope. The
1124: asymptotic behavior is given by
1125: \begin{equation}
1126:   n = - \frac{V_u}{2\ell_x \ell_y}\frac{\sqrt{\delta h}}{\pi S}
1127:   \ln\left(4\delta h\right)\,.
1128: \end{equation}
1129: Interestingly, the quantum corrections to the staggered magnetization
1130: are positive close to the saturation field and the spin-wave result
1131: therefore intersects the classical curve. In a quasi one-dimensional
1132: situation ($J_2\ll J_1$), quantum fluctuations are strong and the
1133: leading order spin-wave theory, when pushed to the limit of validity,
1134: predicts a quantum disordered phase for small uniform fields.
1135: 
1136: 
1137: \subsection{Finite temperature magnetization and susceptibility}
1138: 
1139: Magnetic measurements were carried out on a single crystalline sample
1140: of $\mathrm{Mn}[\mathrm{C}_{10}\mathrm{H}_{6}(\mathrm{OH})(\mathrm{COO})]_{2}
1141:   \!\times\! 2\mathrm{H}_{2}\mathrm{O}$ with a
1142: mass of $m_\mathrm{BONA}=0.65\,\mathrm{mg}$ using a Quantum Design SQUID
1143: magnetometer MPMS-XL.  Isothermal magnetization runs at temperatures between
1144: $2$ and $200\,\mathrm{K}$ and fields up to $5\,\mathrm{T}$ were
1145: performed as well as measurements of the susceptibility in the
1146: temperature range $2-300\,\mathrm{K}$ for a magnetic field of
1147: $0.05-2\,\mathrm{T}$.\cite{Schmidt04}
1148: 
1149: %
1150: %
1151: %
1152: \begin{figure}[tb]
1153:   \begin{center}
1154:     \psfrag{m}{$m$}
1155:     \psfrag{h}{$h$}
1156:     \psfrag{0}{\footnotesize{$0$}}
1157:     \psfrag{0.2}{\footnotesize{$0.2$}}
1158:     \psfrag{0.4}{\footnotesize{$0.4$}}
1159:     \psfrag{0.6}{\footnotesize{$0.6$}}
1160:     \psfrag{0.8}{\footnotesize{$0.8$}}
1161:     \psfrag{1}{\footnotesize{$1$}}
1162:     \psfrag{T=0}{$T=0$}
1163:     \psfrag{T=0.5J0S}{$T=0.5\,\tilde{J}_0S$}
1164: %    \psfrag{T=2J0}{$T=2\,\tilde{J}_0$}
1165:     \epsfig{file=magn.eps,width=70mm,angle=270}~~~~~~
1166:   \end{center}
1167:   \vspace{-4mm}
1168:   \caption{%
1169:     Uniform magnetization $m (h)$ for the honeycomb lattice with
1170:     $S=5/2$ and $J_1 = J_2$ for two values of $T$.  }
1171:   \label{fig:magnT}
1172: \end{figure}
1173: %
1174: %
1175: %
1176: In Fig.~\ref{fig:magnT} we show theoretical magnetization curves
1177: $m(h)$ for the honeycomb lattice with $S=5/2$ and $J_1 = J_2$ at
1178: different temperatures $T$.  For $T\ll\tilde{J}_0S$ the magnetization
1179: is almost linear throughout the entire field range. At intermediate
1180: temperatures $m(h)$ has an S-like shape with a positive curvature at
1181: small fields $h$ that changes to a negative curvature with increasing
1182: $h$.
1183: Similar low-temperature behavior of the magnetization curve has been
1184: observed in a quantum Monte Carlo study of the two-dimensional
1185: Heisenberg antiferromagnet on a square lattice.\cite{Woodward02}
1186: 
1187: %
1188: %
1189: %
1190: \begin{figure}[tb]
1191:   \begin{center}
1192:     \psfrag{m}{$m$}
1193:     \psfrag{H[T]}{$H\,\mathrm{[T]}$}
1194:     \psfrag{0}{\footnotesize{$0$}}
1195:     \psfrag{0.1}{\footnotesize{$0.1$}}
1196:     \psfrag{0.2}{\footnotesize{$0.2$}}
1197:     \psfrag{0.3}{\footnotesize{$0.3$}}
1198:     \psfrag{1}{\footnotesize{$1$}}
1199:     \psfrag{2}{\footnotesize{$2$}}
1200:     \psfrag{3}{\footnotesize{$3$}}
1201:     \psfrag{4}{\footnotesize{$4$}}
1202:     \psfrag{5}{\footnotesize{$5$}}
1203:     \psfrag{1}{\footnotesize{$1$}}
1204:     \psfrag{2K fit}{$2\,\mathrm{K}$ calculation}
1205:     \psfrag{8K fit}{$8\,\mathrm{K}$ calculation}
1206:     \psfrag{20K fit}{$20\,\mathrm{K}$ calculation}
1207:     \psfrag{2K}{$2\,\mathrm{K}$ data}
1208:     \psfrag{8K}{$8\,\mathrm{K}$ data}
1209:     \psfrag{20K}{$20\,\mathrm{K}$ data}
1210:     \epsfig{file=magnfit.eps,width=70mm,angle=270}~~~~~~
1211:   \end{center}
1212:   \vspace{-4mm}
1213:   \caption{%
1214:     Magnetization $m (H)$ of
1215:     $\mathrm{Mn}[\mathrm{C}_{10}\mathrm{H}_{6}(\mathrm{OH})(\mathrm{COO})]_{2}
1216:   \!\times\! 2\mathrm{H}_{2}\mathrm{O}$ up to
1217:     field $H=5\,\mathrm{T}$.  Experimental data are indicated by
1218:     squares ($2\,\mathrm{K}$), circles ($8\,\mathrm{K}$) and triangles
1219:     ($20\,\mathrm{K}$).  Theoretical magnetization curves for
1220:     honeycomb lattice with $S=5/2$ and $J_2=2 J_1=1.95\,\mathrm{K}$ are denoted by
1221:     lines.  }
1222:   \label{fig:magnfit}
1223: \end{figure}
1224: %
1225: %
1226: %
1227: 
1228: It turns out that the magnetization as well as the
1229: susceptibility are not very sensitive to the ratio $J_2/J_1$ as long
1230: as $J_1$ and $J_2$ have the same order of magnitude. Thus, we cannot
1231: determine the precise value of $J_2/J_1$, but our fits are compatible
1232: with the assumption $J_2 \approx 2J_1$.
1233: 
1234: In Fig.~\ref{fig:magnfit} we show experimental data and theoretical
1235: fits for the normalized uniform magnetization $m=M/(NS)$ at different
1236: temperatures.  The magnetic field $H=2\tilde{J}_0Sh$ is given in
1237: Tesla. Surprisingly, all experimental curves are almost straight
1238: lines, whereas from Fig.~\ref{fig:magnT} we would expect an upward
1239: bend of $m(h)$ at higher temperatures. Fits for $T = 2\,
1240: \mathrm{K}$ and different ratios $J_1/J_2$ invariably give
1241: $\tilde{J}_0 \approx 4\, \mathrm{K}$.  Hence we assume $J_2 = 2 J_1$
1242: and fit the theoretical curve to the experimental data at $T = 2\,
1243: \mathrm{K}$.  Good agreement is achieved for $J_2 = 1.95\,
1244: \mathrm{K}$. For this value of the exchange couplings, we also plot
1245: theoretical magnetization curves at $T = 8\,\mathrm{K}$ and $T=20\,
1246: \mathrm{K}$ in Fig.~\ref{fig:magnfit}. These curves deviate
1247: significantly from the data, but one should be aware that $T=8K$ is
1248: already beyond the estimated limit of validity $T\lesssim 0.5
1249: \tilde{J}_0S$ of our theoretical approach. 
1250: 
1251: %
1252: %
1253: %
1254: \begin{figure}[tb]
1255:   \begin{center}
1256:     \psfrag{T[K]}{$T\,\mathrm{[K]}$}
1257:     \psfrag{chi}{\!\!\!\!\!\!\!$\chi\,\mathrm{[cm^3/mol]}$}
1258:     \psfrag{0}{\footnotesize{$0$}}
1259:     \psfrag{0.05}{\footnotesize{$0.05$}}
1260:     \psfrag{0.1}{\footnotesize{$0.1$}}
1261:     \psfrag{0.15}{\footnotesize{$0.15$}}
1262:     \psfrag{100}{\footnotesize{$100$}}
1263:     \psfrag{200}{\footnotesize{$200$}}
1264:     \psfrag{300}{\footnotesize{$300$}}
1265:     \epsfig{file=suscfit.eps,width=70mm,angle=270}~~~~~~
1266:   \end{center}
1267:   \vspace{-4mm}
1268:   \caption{%
1269:     Susceptibility $\chi (T)$ of
1270:     $\mathrm{Mn}[\mathrm{C}_{10}\mathrm{H}_{6}(\mathrm{OH})(\mathrm{COO})]_{2}
1271:   \!\times\! 2\mathrm{H}_{2}\mathrm{O}$.
1272:     Circles are experimental data in a field of $2\,\mathrm{T}$.
1273:     Theoretical fit for honeycomb lattice with $J_2=2 J_1$ (solid
1274:     line) gives $J_2=1.66K$.}
1275:   \label{fig:suscfit}
1276: \end{figure}
1277: %
1278: %
1279: %
1280: 
1281: In Fig.~\ref{fig:suscfit} the uniform susceptibility is plotted in the
1282: experimental units $\mathrm{cm^3/mol}$. When all exchange integrals
1283: have the same order of magnitude we expect a peak in the
1284: susceptibility for $T \approx \tilde{J}_0 S$.  Experimentally, the
1285: peak is at approximately $7\, \mathrm{K}$ so that we have
1286: $\tilde{J}_0\approx 3\, \mathrm{K}$, in accordance with the fits of
1287: the magnetization curves.  For a more quantitative comparison we use the
1288: following procedure. First we substract the temperature-independent
1289: contribution from the
1290: experimental susceptibility in order to get the correct paramagnetic
1291: behavior at high temperatures.  Then we fit the theoretical expression
1292: (\ref{eq:chisw}) with $J_2 = 2 J_1$ to the full set of data points.
1293: Circles in Fig.~\ref{fig:suscfit} are experimental data and the solid
1294: line is a fit with $J_2 = 1.66\, \mathrm{K}$.  The theoretical curve
1295: reproduces the behavior of the susceptibility very well and it
1296: especially gives a good estimate of the position and the form of the
1297: peak. Note that we experimentally observe an increase in the
1298: susceptibility below $T_{\ast} = 3.0 \pm 0.2 \mathrm{K}$. This coincides with an
1299: anomaly in the specific heat.
1300: The careful reader will notice at this point that the estimated value 
1301: of $T_{\ast}$ is larger than the temperature $T = 2\, \mathrm{K}$ where we obtained
1302: the best fit of our calculated magnetization curve $m (H)$ to the experimental
1303: data shown in Fig.~\ref{fig:magnfit}. Hence, at $T = 2\, \mathrm{K}$ the system 
1304: seems to have some kind of long range magnetic order, which we have ignored in our 
1305: calculation. However, the precise nature of the order and the mechanism 
1306: responsible for the ordering are not known at this point. The fact that a
1307: strictly 2D model can reasonably well explain the 
1308: magnetization curve at $T = 2\, \mathrm{K}$
1309: imposes some constraint on possible ordering mechanisms. We suspect that
1310: dipole-dipole interactions play an important role in this temperature range, 
1311: because the long-range nature of the dipole-dipole interaction can give
1312: rise to spontaneous antiferromagnetic order even in 2D.\cite{Pich93} 
1313: This point deserves further attention, both theoretically and experimentally.
1314: 
1315: 
1316: 
1317: % Possible candidates are long-range dipole-dipole
1318: % interactions or couplings between the layers 
1319: % mediated by the naphthalene moieties.
1320: 
1321: \subsection{Staggered correlation length in a magnetic field}
1322: \label{sec:stagg}
1323: 
1324: The energy gap appearing in Eq.~(\ref{eq:epsilonminuslong}) can be
1325: related to the staggered correlation length $\xi$, as discussed by
1326: Takahashi.\cite{Takahashi89} Assuming for simplicity $ |
1327: \bd{\delta}_{\nu} | = a$, we may identify
1328: \begin{equation} 
1329:   \left(\frac{a}{2 \xi}\right)^2 = \Delta^2 = \frac{4h_s}{n_0}\,.
1330:   \label{eq:xidef} 
1331: \end{equation} 
1332: In the absence of a uniform field the low temperature behavior of
1333: $\xi$ has been thoroughly studied by Chakravarty, Halperin and
1334: Nelson.\cite{Chakravarty88} Surprisingly, the effect of a uniform
1335: field $h$ on $\xi$ has so far not been investigated.  We now analyze
1336: the asymptotic behavior of $\xi$ at low temperatures.  In two spatial
1337: dimensions, the limit $T\to0$ also implies $h_s\to0$.  Our
1338: self-consistency equations (\ref{eq:Fsw}) and (\ref{eq:Isw}) can then
1339: be solved analytically by isolating divergent contributions to the
1340: integrals $I ( h, h_s)$ and $F( h, h_s)$ originating from gapless 
1341: modes in the spin-wave spectrum.  In the regular part of the integral,
1342: the limit $T\to0$ and $h_s\to0$ can be taken. For the leading behavior
1343: at small uniform fields $h\ll 1$ only the singular part of $I(h,h_s)$ 
1344: contributes, and we obtain the self-consistency condition
1345: \begin{equation}
1346:   0 = n(0) - \frac{I^{\text{sing}}(h,h_s)}{S}\,.
1347:   \label{eq:selfcon}
1348: \end{equation}
1349: Here, $I^{\text{sing}}(h,h_s)$ is the part of the integral $I (
1350: h,h_s)$ that diverges for vanishing gaps in the spin-wave dispersions,
1351: and $n(0) = n (h=0, h_s=0, T=0)$.  For $h\ll1$, we obtain
1352: \begin{eqnarray}
1353:   I^{\text{sing}}(h,h_s)&  = & \frac{T}{\tilde{J}_0 S} \frac{V_u}2 \sum_{\sigma}
1354:   \int \frac{d^2k}{(2\pi)^2}
1355:   \frac{1}{\epsilon_{{\bd k}\sigma}^2}
1356:   \nonumber
1357:   \\  
1358:   & \approx & - \frac{T}{ \tilde{J}_0 S }\frac{V_u}{8 \pi \ell_x \ell_y}
1359:   \Bigg[\ln\left(\frac{4h_s}{n_0}\right)
1360:   \nonumber\\[1mm]
1361:   &&+ \ln\left(4h^2+\frac{4h_s}{n_0}\right)
1362:   \Bigg]\,,
1363:   \label{eq:Idiv}
1364: \end{eqnarray}
1365: to leading logarithmic order. From Eqs.~(\ref{eq:xidef}) and
1366: (\ref{eq:selfcon}) we then obtain the following result for the
1367: self-consistent energy gap in a small uniform magnetic field
1368: \begin{equation} 
1369:   \Delta^2(h) = \left(\frac{a}{2\xi(h)}\right)^2
1370:   =\sqrt{\Delta_0^4+\frac{(2h)^4}4}-\frac{(2h)^2}2\,,
1371:   \label{eq:Deltah}
1372: \end{equation}
1373: where $\Delta_0=a/2\xi(0)$ is the gap for vanishing uniform field and
1374: the temperature dependence of the zero-field staggered correlation
1375: length is given by
1376: \begin{equation} 
1377:   \frac{\xi(0)}{a} \propto \exp\left(
1378:     \frac{ 2\pi \tilde{J}_0  S^2 n (0) }{T}
1379:     \frac{\ell_x \ell_y}{V_u}
1380:   \right)\,.
1381:   \label{eq:xi0}
1382: \end{equation}
1383: For a square lattice this yields with $\tilde{J}_0 = 4 J$ and $\ell_x
1384: \ell_y / V_u = 1/4$
1385: \begin{equation}
1386:   \frac{\xi(0)}{a} \propto \exp\left(
1387:     \frac{2  \pi J  S^2 n (0) }{T} 
1388:   \right)
1389:   \,,
1390:   \label{eq:xi0square}
1391: \end{equation}
1392: which is identical to Takahashi's result (see Eq.~(27a) in
1393: Ref.~\onlinecite{Takahashi89}), except that we do not include a
1394: spin-wave velocity renormalization in our approach.  To obtain this
1395: renormalization, the spin-wave interaction would have to be treated on
1396: the mean-field level in a fully self-consistent way.
1397: 
1398: The field dependence of the correlation length for fixed temperature is
1399: given by Eq.~(\ref{eq:Deltah}). For $h\ll\Delta_0(T)$, we have
1400: \begin{equation}
1401:   \xi(h)=\xi(0)\left[1+\frac12\left(\frac{h}{\Delta_0}\right)^2\right]\,,
1402: \end{equation}
1403: whereas for $h\gg\Delta_0(T)$, we obtain
1404: \begin{equation}
1405:   \frac{\xi(h)}a = 4h\,\left(\frac{\xi(0)}a\right)^2\,.
1406: \end{equation}
1407: From Eq.~(\ref{eq:Deltah}) it is clear that $\xi(h)>\xi(0)$. Thus, the
1408: correlation length is increased by a small uniform field due to reduced 
1409: quantum fluctuations. 
1410: 
1411: The temperature dependence of the correlation length for fixed uniform
1412: field $h$ can also be extracted from Eq.~(\ref{eq:Deltah}).  As long
1413: as $\Delta_0(T) \gg 2h$, this temperature dependence is still given by
1414: Eq.~(\ref{eq:xi0}). When the temperature is further reduced, 
1415: Eq.~(\ref{eq:Deltah}) predicts a crossover at $\Delta_0(T) \approx 2h$
1416: to the following temperature-dependent correlation length
1417: \begin{equation}
1418:   \frac{\xi (h)}{a} \propto \exp\left(
1419:     \frac{4  \pi \tilde{J}_0  S^2 n(0)}{T}
1420:     \frac{\ell_x \ell_y}{V_u}
1421:   \right)\,.
1422:   \label{eq:xih}
1423: \end{equation}
1424: The additional factor of two in the exponent as compared to
1425: Eq.~(\ref{eq:xi0}) is due to the fact that at very low temperatures
1426: the spin-wave mode $\epsilon_{{\bm k}-}$ yields a singular
1427: contribution, whereas the mode $\epsilon_{{\bm k}+}$ has a gap $~2h$
1428: which is fixed by the external field. In contrast, for $h=0$ both
1429: modes contribute equally, leading to Eq.~(\ref{eq:xi0}). 
1430: 
1431: The analysis in this section has been carried out for $h\ll1$. For
1432: larger fields, there are field dependent prefactors of the first
1433: logarithm in Eq.~(\ref{eq:Idiv}) leading to a field dependent
1434: renormalization factor $Z_h$ in the exponent of Eq.~(\ref{eq:xih}).
1435: The field dependence of the correlation length at fixed temperature is
1436: then no longer determined by the singular contributions to the
1437: integrals and cannot be extracted from the simple analysis presented
1438: here. Close to the critical field at $h=1$ the nature of the
1439: divergences changes, since the dispersion of the $\sigma=-$ mode
1440: becomes quadratic. As our mean-field calculation is not suitable to
1441: describe the true critical behavior in two dimensions, we do not
1442: discuss this limit in more detail.
1443: 
1444: Our approach can also describe a quasi one-dimensional anisotropic
1445: system, where the exchange coupling between chains is very weak. The
1446: dispersion is then almost flat in the transverse direction. The
1447: integrals will be quasi one-dimensional as long as the maximum
1448: variation of the dispersion in the transverse direction is smaller
1449: than the self-consistent gap $4h_s /n_0$. In this intermediate
1450: temperature regime the staggered correlation length behaves as if the
1451: system were one-dimensional.  At even lower temperatures there will be
1452: a crossover to the true asymptotic two-dimensional behavior. A rough
1453: estimate for the position of the crossover is ${a}/{\xi} \propto
1454: {\ell_{\bot}}/{a}$ where $\ell^2_{\bot}$ is the eigenvalue of the
1455: matrix $\bd{A}$ defined in Eq.~(\ref{eq:Amatrixdef}) associated with
1456: the eigenvector perpendicular to the chain direction.
1457:  
1458: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1459: \section{Conclusion}
1460: \label{sec:discussion}
1461: 
1462: In summary, we have investigated the magnetic properties of the new
1463: metal-organic quantum magnet
1464: $\mathrm{Mn}[\mathrm{C}_{10}\mathrm{H}_{6}(\mathrm{OH})(\mathrm{COO})]_{2}
1465:   \!\times\! 2\mathrm{H}_{2}\mathrm{O}$.
1466: Its layered structure contains two-dimensional arrangements of
1467: $\mathrm{Mn}^{2+}$ ions that suggest a spin $S=5/2$ Heisenberg model on a
1468: distorted honeycomb lattice as a minimal model. In order to explain
1469: measurements of the magnetization $M(H,T)$ and the
1470: susceptibility $\chi(T)$, we develop a variant of modified spin-wave
1471: theory, which can be used to describe finite temperature properties of
1472: two-dimensional magnets in a uniform external magnetic field.  A fit
1473: of the theoretical results to the experimental curves shows a
1474: satisfactory agreement for the magnetization at low temperatures
1475: where we expect our theoretical approach to be valid. The magnetic
1476: susceptibility is very well described down to temperatures of
1477: $T \gtrsim T_{\ast} \approx 3\mathrm{K}$. 
1478: Both quantities are consistently fitted by one parameter 
1479: $J_2 = 2 J_1$ to give the exchange coupling 
1480: $J_2 \approx 1.8\,
1481: \mathrm{K}$. For temperatures below $T_{\ast}$ the uniform susceptibility
1482: shows again an upturn, which together with an anomaly in the specific
1483: heat is most likely due to some ordering transition. Possible
1484: mechanisms for this transition are dipole-dipole interactions or
1485: couplings between the layers, which should be included in more refined
1486: theoretical models. From the experimental point of view  
1487: nuclear magnetic resonance or neutron scattering  measurements 
1488: could provide a more detailed insight into the nature of the 
1489: magnetic interactions.
1490: 
1491: 
1492: This work was supported by the DFG via Forschergruppe FOR 412.
1493: We thank M.~Kuli\'{c} for interesting discussions and especially for 
1494: pointing out the possible importance of
1495: dipole-dipole interactions.
1496: % , Project
1497: % No. KO 1442/5-1 and ?? (Lang group no.) and SCHM 1167/4-2.
1498: 
1499: %\vspace{-4mm}
1500: \begin{thebibliography}{99}
1501: %\vspace{-4mm}
1502:   
1503: \bibitem{Schollwoeck04}
1504: For a collection of recent reviews see U. Schollw\"{o}ck, J. Richter, D. J. J. Farnell, and R. F.
1505: Bishop (Eds.), {\it{Quantum Magnetism}}, (Springer, Berlin Heidelberg, 2004).
1506: %
1507: \bibitem{Schmidt04} 
1508: M.~U.~Schmidt, E.~Alig, L.~Fink, M.~Bolte, R.~Panisch,
1509: V.~Pashchenko, B.~Wolf, and M.~Lang, 
1510: Acta.~Cryst.~C {\bf{61}}, m361 (2005).
1511: %
1512: \bibitem{CCDC} 
1513: Details of the crystal structure determination are available from the 
1514: Cambridge Crystallographic Data Centre, 
1515: http://www.ccdc.cam.ac.uk/products/csd/request/ quoting the reference number CSD-269503.
1516: %
1517: % \bibitem{Pashchenko04}
1518: % V. Pashchenko, B. Brendel, B. Wolf, M. Lang,
1519: % M. Kollar, F. Sch\"{u}tz, P. Kopietz,
1520: % Y. Molodtsova, O. Shchegolikhina, N. Auner, and J. Bats,
1521: % J. Mag. Magn. Mat. {\bf 272-276}, e755 (2004);
1522: % V. Pashchenko, B. Brendel, B. Wolf, M. Lang, K. Lyssenko, O. Shchegolikhina, Y. Molodtsova,
1523: % L. Zherlitsina, N. Auner, F. Sch\"{u}tz, M. Kollar, P. Kopietz, and N. Harrison,
1524: % in preparation.
1525: %
1526: \bibitem{Hohenberg67}
1527: P. C. Hohenberg, Phys. Rev. {\bf{158}}, 383 (1967);
1528: N. D. Mermin and H. Wagner, Phys. Rev. Lett. {\bf{17}}, 1133 (1966).
1529: %
1530: \bibitem{Chakravarty88}
1531: S. Chakravarty, B. I. Halperin, and D. R. Nelson, 
1532: Phys. Rev. B {\bf{39}}, 2344 (1989).
1533: %
1534: \bibitem{Fukumoto96}
1535: Y.~Fukumoto, J.~Phys.~Soc.~Jpn. {\bf{65}}, 569 (1996);
1536: %
1537: \bibitem{Zhitomirsky98}
1538: M. E. Zhitomirsky and T. Nikuni, Phys. Rev. B {\bf{57}}, 5013 (1998).
1539: %
1540: \bibitem{Fabricius92}
1541: K.~Fabricius, M.~Karbach, U.~L\"{o}w, and K.-H. M\"{u}tter, 
1542: Phys. Rev. B {\bf 45}, 5315 (1992).
1543: %
1544: \bibitem{Kollar03}
1545: M. Kollar, I. Spremo, and P. Kopietz, Phys. Rev. B {\bf{67}}, 104427 (2003).
1546: %
1547: \bibitem{Schuetz03}
1548: F. Sch\"{u}tz, M. Kollar, and P. Kopietz, Phys. Rev. Lett. 
1549: {\bf{91}}, 017205 (2003); Phys. Rev. B {\bf{69}}, 035313 (2004).
1550: %
1551: %\bibitem{Holstein40}
1552: %T. Holstein and H. Primakoff, Phys. Rev. {\bf{58}}, 1098 (1940).
1553: %
1554: \bibitem{Dyson56}
1555: F. J. Dyson, Phys. Rev. {\bf{102}}, 1217 and 1230 (1956);
1556: %
1557: \bibitem{Maleev57}
1558: S. V. Maleev, Zh. Eksp. Theor. Fiz. {\bf{30}}, 1010 (1957)
1559: [Sov. Phys. JETP {\bf{64}}, 654 (1958)].
1560: %
1561: \bibitem{Zhitomirsky99}
1562: M. E. Zhitomirsky and A. L. Chernyshev, Phys. Rev. Lett. {\bf 82}, 4536 (1999).
1563: %
1564: \bibitem{Maleyev00}
1565: S. V. Maleyev, Phys. Rev. Lett. {\bf 85}, 3281 (2000).
1566: %
1567: \bibitem{Syromyatnikov01}
1568: A.~V.~Syromyatnikov and S.~V.~Maleyev, Phys. Rev. B {\bf 65}, 12401 (2001).
1569: %
1570: \bibitem{Veillette05}
1571: M. Y. Veillette, J. T. Chalker, and R. Coldea, Phys. Rev. B {\bf 71}, 
1572: 214426 (2005).
1573: % cond-mat/0501347.
1574: %
1575: \bibitem{Takahashi89}
1576: M. Takahashi, Phys. Rev. B {\bf{40}}, 2494 (1989).
1577: %
1578: \bibitem{Kopietz97}
1579: P.~Kopietz and S.~Chakravarty, Phys. Rev. B {\bf{56}}, 3338 (1997).
1580: %
1581: \bibitem{Griffiths64}
1582: R.~B.~Griffiths, Phys. Rev. {\bf{133}}, A768 (1964)
1583: %
1584: \bibitem{Gluzman93}
1585: S. Gluzman, Z. Phys. B {\bf 90}, 313 (1993).
1586: %
1587: \bibitem{Woodward02}
1588: F.~M.~Woodward, A.~S.~Albrecht, C.~M.~Wynn, C.~P.~Landee, and M.~M.~Turnbull, 
1589: Phys. Rev. B {\bf{65}}, 144412 (2002).
1590: %
1591: \bibitem{Pich93}
1592: C.~Pich and F.~Schwabl , Phys. Rev. B {\bf{47}}, 7957 (1993).
1593: %
1594: \end{thebibliography}
1595: 
1596: \end{document}
1597: 
1598: