cond-mat0306299/ikp.tex
1: 
2: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb,prb,superscriptaddress]{revtex4} 
3:  
4: \usepackage{epsfig}% Include figure files 
5: \usepackage{graphicx}% Include figure files 
6: \usepackage{dcolumn}% Align table columns on decimal point 
7: \usepackage{bm}% bold math 
8:  
9: \begin{document} 
10:  
11: \title{Static impurities in the kagome lattice: dimer freezing and mutual  
12: repulsion} 
13:  
14: \author{S. Dommange}
15: \affiliation{Institut Romand de Recherche Num\'erique en Physique des  
16: Materiaux (IRRMA), PPH--Ecublens, CH--1015 Lausanne, Switzerland}  
17:  
18: \author{M. Mambrini} 
19: \affiliation{Laboratoire de Physique Th\'eorique, Universit\'e Paul Sabatier,  
20: FRE 2603, F--31062 Toulouse Cedex, France} 
21:  
22: \author{B. Normand} 
23: \affiliation{D\'epartement de Physique, Universit\'e de Fribourg,  
24: CH--1700 Fribourg, Switzerland} 
25:  
26: \author{F. Mila} 
27: \affiliation{Institut de Physique Th\'eorique, Universit\'e de Lausanne,  
28: CH--1015 Lausanne, Switzerland} 
29:  
30: \date{\today} 
31:  
32: \begin{abstract} 
33:  
34: We consider the effects of doping the $S$ = 1/2 kagome lattice with  
35: static impurities. We demonstrate that impurities lower the number 
36: of low--lying singlet states, induce dimer--dimer correlations  
37: of considerable spatial extent, and do not generate free spin degrees  
38: of freedom. Most importantly, they experience a highly unconventional  
39: mutual repulsion as a direct consequence of the strong spin frustration.  
40: These properties are illustrated by exact diagonalization, and reproduced 
41: to semi--quantitative accuracy within a dimer resonating--valence--bond 
42: description which affords access to longer length scales. We calculate 
43: the local magnetization induced by doped impurities, and consider its 
44: implications for nuclear magnetic resonance measurements on known kagome 
45: systems. 
46:  
47: \end{abstract} 
48:  
49: \pacs{75.10.Jm, 75.30.Hx, 76.60.-k} 
50:  
51: \maketitle 
52:  
53: \section{Introduction} 
54:  
55: The kagome geometry (Fig.~1) presents one of the most highly frustrated 
56: quantum spin systems achievable in two dimensions (2d) with only 
57: nearest--neighbor Heisenberg interactions. The $S$ = 1/2 kagome lattice 
58: has been found to have a spin--liquid ground state,\cite{rce} whose 
59: ultra--short spin--spin correlation lengths are effectively those of a dimer 
60: liquid,\cite{rle,rze} and whose excitation spectrum\cite{rlblps} shows a 
61: manifold containing an extremely large number of low--lying singlets.\cite{rm} 
62: These properties are reproduced very well by a short--range 
63: resonating--valence--bond (RVB) description,\cite{rmm} based in fact 
64: only on nearest--neighbor dimer formation (Fig.~1).
65: 
66: On the theoretical level, a similar degree of frustration is found only 
67: in quantum dimer models,\cite{rrk,rms} which have also been considered 
68: recently in the kagome geometry,\cite{rmsp} and in models containing 
69: higher--order spin terms and multiple--spin exchange.\cite{rmblw}
70: Experimentally, a variety of materials displaying the kagome structure 
71: is known to exist, although to date none have been found which contain
72: an ideal, 2d, vacancy--free system of spins $S = 1/2$. These 
73: include the jarosites (H$_3$O)Fe$_3$(OH)$_6$(SO$_4$)$_2$,\cite{rwhrs} 
74: KFe$_3$(OH)$_6$(SO$_4$)$_2$, and KCr$_3$(OH)$_6$(SO$_4$)$_2$\cite{rkea,rlea}
75: and the volborthite Cu$_3$V$_2$O$_7$(OH)$_2$.H$_2$O.\cite{rhhkntkt}
76: The best characterized of these compounds is the magnetoplombite 
77: SrCr$_{9p}$Ga$_{12-9p}$O$_{19}$ (SCGO),\cite{rlmcmombm} in which some of 
78: the $S$ = 3/2 Cr$^{3+}$ ions form approximately 2d kagome planes. However, 
79: because in this case the ideal stoichiometry ($p = 0$) remains unachievable, 
80: discussion of the influence of static (and also, for Ga$^{3+}$, spinless) 
81: impurities represents an important issue.
82: 
83: \begin{figure}[t!] 
84: \medskip
85: \centerline{\psfig{figure=kknnc.eps,width=5.5cm,angle=0}}
86: \medskip
87: \caption{The kagome lattice. Thick black lines represent an arbitrarily 
88: chosen covering by nearest--neighbor dimers formed between $S = 1/2$ spins 
89: on each lattice site. } 
90: \label{fig:1} 
91: \end{figure} 
92:  
93: Studies of static vacancies in spin liquids have tended to focus on 
94: the question of induced free spin degrees of freedom.\cite{rlmgmdhlr} 
95: In systems such as spin ladders, and also for the 2d square 
96: lattice,\cite{rbhsl} the total energy appears to decrease as such holes 
97: in the spin background are made to approach each other, implying an 
98: effective mutual attraction. It is generally believed that in an RVB 
99: description of the ground state the minimal disruption of the wave 
100: function, and thus the most favorable energy state, occurs when holes 
101: occupy adjacent sites, thus generating a simple pairing mechanism for 
102: mobile dopants. The issue of impurites in frustrated magnets has received 
103: less attention, and appears to be more involved: it has been proposed that 
104: an absence of induced free spins corresponds to deconfined, spinon-like 
105: excitations.\cite{rsv} Consideration of the problem for the 1d frustrated 
106: chain,\cite{rnm} which provides an example with no free spins localized 
107: around the vacancies, suggests the importance of degenerate singlets in 
108: the ground--state manifold. This result indicates that the kagome geometry 
109: presents the most probable candidate for exotic behavior\cite{rs} as a
110: consequence of strong frustration in 2d.  
111:  
112: In this study we investigate the effects of static impurities in the kagome 
113: lattice. We find that, despite the very short intrinsic spin--spin 
114: correlation lengths, dimer--dimer correlations develop over considerable 
115: distances in the presence of dopants. Further, we show that there is no 
116: evidence for free local moments induced around impurity sites, and that 
117: a highly unusual effective repulsive interaction arises between such 
118: holes in the spin background. 
119: 
120: In Sec.~II we provide an outline of the numerical and anaytical methods 
121: to be employed, and discuss the fundamental effects on the excitation 
122: spectrum of doping by spinless impurities. In Sec.~III we consider the 
123: case of one hole in an odd--site cluster, demonstrating ``dimer freezing'' 
124: in the spin--spin correlation functions and dimer--dimer correlations of 
125: anomalously long range. The induced magnetizations in the spin sectors 
126: $S$ = 1 and higher are used to illustrate the consequences for nuclear 
127: magnetic resonance (NMR) measurements. Sec.~IV presents spin--spin 
128: correlation functions which characterize the situation with two impurities, 
129: supplemented by total--energy analyses quantifying the novel repulsive 
130: interaction between the holes which emerges from this study. Sec.~V 
131: contains further discussion and our conclusions.  
132:  
133: \section{Elimination of singlet states and absence of localized spins}
134:  
135: \subsection{Exact Diagonalization}
136: 
137: By ``impurity'' we refer henceforth to nonmagnetic impurities. The 
138: techniques we employ are based primarily on numerical calculations for 
139: small clusters with periodic boundary conditions chosen as symmetrically 
140: as possible. Specifically, we perform exact diagonalization (ED) studies 
141: of the Heisenberg Hamiltonian 
142: \begin{equation}
143: H = J \sum_{\langle ij \rangle} {\bf S}_i {\bf \cdot S}_j ,
144: \label{ehh}
145: \end{equation}
146: where $J$ is the antiferromagnetic superexchange interaction and $\langle 
147: ij \rangle$ denotes nearest--neighbor sites, for clusters up to 27 sites 
148: with one impurity, and up to 24 sites with two impurities for all available 
149: configurations. We analyze the total energies, the energies per doped 
150: impurity, and also the form of the energy spectra, focusing on the number 
151: of singlet states in the low--energy manifold below the first triplet for 
152: the finite systems under consideration, as well as on the number of these 
153: states removed by impurity doping. The frustrated spin interactions in the 
154: presence of impurities are characterized for the entire cluster by the bond 
155: spin--spin correlation functions and by the induced site magnetizations 
156: in every spin sector. 
157: 
158: \begin{figure}[t!] 
159: \centerline{\psfig{figure=kkded18.eps,width=7.0cm,angle=0}}
160: \medskip
161: \caption{Integrated density of states for an 18--site $S$ = 1/2 kagome 
162: cluster, computed by exact diagonalization both without impurities (solid 
163: line) and with 2 impurities (dashed line). The vertical lines mark the 
164: position of the first triplet state in the absence of impurities (solid) 
165: and in their presence (dashed). } 
166: \label{fig:2} 
167: \end{figure} 
168: 
169: The density of states integrated over the lowest--lying energy levels 
170: is shown in Fig.~2 for systems with and without impurities. Fig.~2 
171: compares ED results from systems of 18 sites, for which in the impurity 
172: configuration illustrated 8 of the 12 singlet states occurring below 
173: the first triplet, indicated by the vertical lines, are eliminated by 
174: the presence of 2 impurities. Although the separation of the triplet 
175: from the lowest remaining singlet decreases, this energy gap always 
176: remains a finite fraction of the superexchange interaction $J$. Thus the 
177: most important qualitative result emerging even for the smallest systems is 
178: that, in contrast to spin--ladder and dimerized or quadrumerized systems, 
179: no low--lying triplet state is induced by the presence of impurities. This 
180: implies directly\cite{rnm} that there are no quasi--free spin degrees of 
181: freedom, by which is meant spins effectively isolated from their local 
182: environment and interacting with each other only on an effective energy 
183: scale much smaller than $J$. The fact that no localized moments are formed 
184: in the vicinity of impurities may be ascribed to the continuum of singlets 
185: beyond the ground state, which has the same function as the two--fold 
186: degeneracy of the ground state in the frustrated chain,\cite{rnm} namely 
187: that the rearrangement of dimers incurs no energy penalty. The absence of 
188: local spins around nonmagnetic impurities represents one of the most 
189: significant differences between frustrated and unfrustrated systems, and 
190: has been further interpreted as implying a deconfinement of the spin 
191: degrees of freedom.\cite{rsv}
192: 
193: \begin{table}[b!]
194: %\begin{ruledtabular} 
195:   \begin{tabular}{|c|cccc|ccccc|} \hline 
196: $N$ & \multicolumn{4}{c|}{12} & \multicolumn{5}{c|}{18} \\
197: \hline
198: $d_{\rm M}$ & 1 & 2 & 2 & 3 & 1 & 2 & 2 & 3 & 3 \\
199: $N_s^0$ & 4 & 4 & 4 & 4 & 12 & 12 & 12 & 12 & 12 \\
200: $N_s^2$ & 2 & 2 & 3 & 1 & 3 & 4 & 4 & 3 & 4 \\
201: \hline
202: \end{tabular}
203: %\end{ruledtabular} 
204: \caption{\label{tab:deltaE} Number of low--lying singlets below the first 
205: triplet for kagome clusters without impurities ($N_s^0$, third row) and 
206: with two impurities ($N_s^2$, fourth row) as a function of the Manhattan 
207: distance $d_{\rm M}$ between impurities. Results obtained by ED are shown 
208: for clusters of $N$ = 12 and 18 sites. }
209: \end{table}
210: 
211: Table 1 provides more details of the number of states in the singlet 
212: manifold which are removed by impurity doping for clusters of 12 and 18 
213: sites, for all possible configurations of 2 impurities. The numbers vary 
214: little as a function of the proximity of the impurities, and the average 
215: number of low--lying states which are eliminated is in agreement with simple 
216: considerations of the number of singlet states in the small system (below). 
217: We have also performed analogous calculations for 24-site clusters of
218: different shapes. For all clusters the same behavior is found upon 
219: introducing two impurities, in that the number of singlets below the 
220: first triplet is reduced, but here the lowest--lying excitation is not 
221: a triplet. However, these clusters do not appear to be representative of 
222: the generic kagome physics, as the number of low--lying singlets in the 
223: absence of impurities is significantly smaller than the expected 
224: value\cite{rlblps} of $1.15^N$. This situation is presumed to be a 
225: consequence of the low symmetry of the 24--site clusters, some of which 
226: are essentially 1d systems, and as a result the densities of states are 
227: not presented here.
228: 
229: \begin{figure}[t!] 
230: \centerline{\psfig{figure=kkdrvb18.eps,width=7.0cm,angle=0}}
231: \medskip
232: \caption{Integrated density of singlet states for an 18--site $S$ = 1/2 
233: kagome cluster, calculated in the dimer--RVB basis both without impurities 
234: (solid line) and with 2 impurities (dashed). } 
235: \label{fig:3} 
236: \end{figure} 
237:  
238: \subsection{Dimer--RVB Basis}
239: 
240: We calculate in addition the ground--state properties, namely the energy 
241: and spin--spin correlation functions, of clusters doped with one and with 
242: two impurities within the dimer--RVB basis. This is the basis of states in 
243: which every spin forms a singlet dimer with one of its nearest neighbors 
244: in such a way that every spin of the lattice is involved in a dimer 
245: (Fig.~1). The use of this basis\cite{rmm} naturally implies a truncation 
246: of the total number of available states in the manifold of low--lying 
247: singlets, but we will show below that it provides a satisfactory description 
248: of the ground state of the doped system. Dimer--RVB calculations permit 
249: the consideration of somewhat larger cluster sizes, and we will present 
250: results up to and including 39 sites with one impurity and 36 sites with 
251: two impurities.
252: 
253: Calculations performed in the dimer--RVB basis yield results very similar 
254: to the exact ones for the integrated density of states (Fig.~3), giving 
255: clear evidence of a strong linear correlation (see also Fig.~9) between 
256: the ED and dimer--RVB spectra in both 0-- and 2--impurity cases. 
257: Demonstration of a quantitative correspondence between the spectra 
258: would require shift and scaling factors arising from an energetic 
259: renormalization related to the fact that calculations in the dimer--RVB 
260: basis are variational in nature, so that the true eigenstates are dressed 
261: RVB wave functions, but this procedure is not necessary for the analysis 
262: to be presented here. 
263: 
264: Given that the RVB basis represents a very significant truncation of the 
265: total Hilbert space, the qualitative similarity of the dimer--RVB spectra 
266: to those obtained from the exact calculations, both with and without 
267: impurities, provides important information about the nature of the exact 
268: wave functions of the kagome system. This observation is further confirmed 
269: by the semi--quantitative agreement with exact results for spin correlation 
270: functions and ground--state energies computed using the dimer--RVB basis. 
271: We defer this comparison to Secs.~III and IV, and comment here only that 
272: our results in the presence of impurities provide additional strong 
273: justification for the statement that the ground state of the kagome spin 
274: system is described appropriately by a maximally short--ranged RVB wave 
275: function, presumably as a consequence of the very strong frustration. 
276: 
277: The dimer--RVB basis is a natural framework for understanding the absence 
278: of localized spins around an impurity site, as all remaining spins are
279: simply incorporated within rearranged singlet dimer coverings. It also 
280: provides a qualitative reflection of the ``evaporation'' of low--lying 
281: singlets from the ground--state manifold (Figs.~2--3 and Table I) which 
282: arises on doping a cluster with two impurities, and in fact yields 
283: additional insight into the mechanism for this exclusion. It can be 
284: shown that the total number of singlet dimer coverings of a doped system 
285: is given by 
286: \begin{equation}
287: N_{\rm coverings} = 2^{N/3 - N_{\rm imp} + N_G + N_{\rm et}},
288: \label{endc}
289: \end{equation}
290: where $N$ is the number of sites of the system under consideration, 
291: $N_{\rm imp}$ the number of doped impurities, $N_G$ the number of independent 
292: clusters into which the system is divided by the dopant distribution, and 
293: $N_{\rm et}$ the number of triangles containing 3 impurities. A complete 
294: derivation of this result lies beyond the scope of the current analysis, 
295: and we present only the following heuristic explanation. By reducing each 
296: triangle to an effective $S = 1/2$ degree of freedom,\cite{rm,rmm} the 
297: dimer basis involves a truncation of the $2^N$ total states to $2^{N/3}$, 
298: multiplied by an additional factor of 2 for the overall spin direction on 
299: each independent cluster. Each added impurity removes one spin, halving 
300: the number of states in every manifold and thus justifying the dependence 
301: on $N_{\rm imp}$. The only exception to this statement occurs when the 
302: added impurity falls on a triangle which already contains two impurities, 
303: as a result of which it is already fully constrained and removal of the 
304: third site no longer affects the number of coverings. Eq.~(\ref{endc}) 
305: can be shown to remain valid for all vacancy concentrations, including 
306: most notably values of $N_{\rm imp}$ exceeding the percolation threshold, 
307: beyond which $N_G > 1$, under the condition that the impurity distribution 
308: does not cut the system into segments containing odd numbers of spins.
309: 
310: \section{Spin correlation functions and induced magnetization}
311: 
312: \subsection{Dimer ``freezing''}
313: 
314: In this section we consider the effects of a single impurity introduced  
315: in an odd--site cluster. The bond spin--spin correlation functions 
316: $\langle {\bf S}_i {\bf \cdot S}_j \rangle$ around a single impurity 
317: may be computed by ED for clusters of up to 27 sites. The ED calculations 
318: permit in addition the determination of the magnetizations induced on each 
319: site by the presence of the impurity, with results for every spin sector 
320: corresponding to the situation in all applied magnetic fields, which 
321: are presented in Subsec.~C below. 
322: 
323: \begin{figure}[t!] 
324: \centerline{\psfig{figure=kkssc27.eps,width=5.5cm,angle=0}}
325: \medskip
326: \caption{Spin--spin correlations for a single impurity in a 27--site  
327: cluster, obtained by ED. Bar width represents the strength of correlation 
328: functions on each bond on a linear scale where the strongest correlation 
329: function is $\langle {\bf S}_i {\bf \cdot S}_j \rangle = - 0.69$. Dashed 
330: lines denote bonds on which $\langle {\bf S}_i {\bf \cdot S}_j \rangle > 0$.}
331: \label{fig:4} 
332: \end{figure} 
333: 
334: The correlation functions obtained by ED are illustrated in Fig.~4 
335: for a 27--site cluster with periodic boundary conditions. The most 
336: striking feature of these results is that dimer bonds are effectively 
337: frozen in the two depleted triangles, taking a value $\langle {\bf S}_i 
338: \cdot {\bf  S}_j \rangle \simeq - 0.69$ close to the theoretical minimum 
339: of $-3/4$ for a pure singlet state. For comparison, the average value of 
340: $\langle {\bf S}_i \cdot {\bf  S}_j \rangle$ in the pure system, which 
341: retains full freedom to resonate, is a factor of 3 smaller 
342: ($- 0.22$).\cite{rle} The bonds neighboring the frozen dimers then show 
343: very weak correlations, while the next neighbors are again enhanced, and 
344: the spin correlations oscillate with distance away from the impurity 
345: [also illustrated in Fig.~5(b)]. Thus although spin--spin correlations 
346: in the kagome system are extremely short--ranged,\cite{rle} induced 
347: dimer--dimer correlations show a considerable spatial extent, reaching 
348: in fact well beyond the sizes obtainable by studying small clusters using 
349: either ED or dimer--RVB techniques. For the limited data available it is 
350: unfortunately not possible to specify the functional form of the decay to 
351: equilibrium. Further analysis of this issue, based on data obtained from 
352: mutual hole repulsion, is presented in Sec.~IV. 
353: 
354: \begin{figure}[t!] 
355: \psfig{figure=kkcfsc.eps,width=5.0cm,angle=270}
356: \centerline{(a)}
357: \centerline{\psfig{figure=kkssca39.eps,width=5.0cm,angle=270}}
358: \centerline{(b)}
359: \caption{(a) Spin--spin correlation functions $C(d_{\rm M}) = \langle 
360: {\bf S}_i \cdot {\bf  S}_j \rangle$ on nearest--neighbor bonds as a function 
361: of their Manhattan distance $d_{\rm M}$ from a single impurity, for 
362: clusters of 21 and 27 sites. (b) Spin--spin correlations as a function of 
363: (Manhattan) bond distance from a single impurity, obtained by dimer--RVB 
364: calculations for a 39--site cluster. Points represent the average over all 
365: bonds at the same distance from the impurity, the solid line the value 
366: $\langle {\bf S}_i {\bf \cdot S}_j \rangle = - 0.22$ for the impurity--free 
367: system.}
368: \label{fig:5} 
369: \end{figure} 
370:  
371: \subsection{Dimer--RVB Basis}
372: 
373: Fig.~5(a) compares the bond spin correlation values $\langle {\bf S}_i 
374: \cdot {\bf  S}_j \rangle$ obtained from ED and from the dimer--RVB basis 
375: for all bonds in clusters of 21 and 27 sites with one impurity. These 
376: clusters represent the maximal sizes accessible by ED, and for such 
377: systems it is clear that the dimer--RVB basis reproduces the exact 
378: values with quantitative accuracy. From the nature of the RVB basis it is 
379: perhaps to be expected that nearest--neighbor bond correlations are treated 
380: more accurately than any other physical property. We make use of this 
381: feature to extend the calculations of spin correlation functions to a 
382: system of 39 sites with oie impurity. Because on this cluster the boundary 
383: conditions lead to a rather low symmetry, Fig.~5(b) shows the value of 
384: $\langle {\bf S}_i {\bf \cdot S}_j \rangle$ averaged over all bonds at 
385: the same Manhattan distance from the impurity, which illustrates both 
386: the oscillatory nature of the frozen dimerization pattern and the very 
387: long range over which an impurity acts to disrupt the spin configuration.
388: 
389: \begin{figure}[t!] 
390: \centerline{\psfig{figure=kkmag27.eps,width=6.5cm,angle=0}}
391: \medskip
392: \caption{Site magnetization profile of the lowest triplet state induced 
393: by a single impurity in a 27--site cluster, obtained by ED. Dot radius 
394: represents the magnitude of the local moment on a linear scale where the 
395: largest dot corresponds to a moment of 0.13$\mu_{\rm B}$, and the grey dots 
396: represent sites with induced moments opposite to the field.} 
397: \label{fig:6} 
398: \end{figure} 
399:  
400: \subsection{Induced magnetic moments}
401: 
402: The local magnetization pattern induced in the vicinity of a doped  
403: nonmagnetic impurity may be computed by exact diagonalization in the sector 
404: of total spin $(S,S_z)$ = (1,1), and is displayed for the 27--site cluster 
405: in Fig.~6. The site magnetizations exhibit a certain anticorrelation with 
406: the strength of the dimers formed on each bond, and thus have a tendency 
407: to oscillate with increasing distance from the impurity. The induced 
408: moments are particularly enhanced on sites far from the impurity, where 
409: different paths on the periodic cluster interfere, and the interference 
410: in this regime may even result in moments oriented opposite to the 
411: effective magnetic field (Figs.~6, 7). The magnetization profile of the 
412: surrounding sites is the single most important consequence of impurities 
413: which is reflected in local--probe measurements of doped spin systems. 
414: 
415: \begin{figure}[t!] 
416: %\centerline{
417: \psfig{figure=kkks27.eps,width=7.2cm,angle=270}
418: %}
419: \medskip
420: \caption{Induced magnetization per site $\mu_i$ as a function of the spin 
421: sector $S_z$ (analogous to an applied magnetic field) for all inequivalent 
422: sites\cite{footnote1} neighboring an impurity, computed by ED for a 
423: 27--site cluster. With reference to the site labels in Fig.~6, the 
424: separate lines $i$ represent 1st (4, 5, 23, 24), 2nd (6, 9, 22, 25), 3rd 
425: (10, 12, 19, 20), 4th (2, 3), equal 5th [(7, 8, 26, 27) and (13, 15, 16, 
426: 18)], 7th (14, 17), and 8th (11, 21) neighbors of the impurity site (1).} 
427: \label{fig:7} 
428: \end{figure} 
429:   
430: Fig.~7 shows the magnetization per site induced by proximity to a 
431: nonmagnetic impurity for each spin sector obtainable by ED for a 
432: 27--site cluster. The spin sector may be considered as equivalent to an 
433: effective magnetic field which varies between zero ($S_z = 0$) and 
434: saturation ($S_z = 13$). The most pronounced variation is observed for 
435: the nearest--neighbor sites (filled circles in Fig.~7), which have very 
436: small magnetization at low fields because they are bound into the strong 
437: dimers frozen by the presence of the impurity on the same triangle. 
438: However, between $S_z = 4$ and $S_z = 6$ the moment shows an abrupt jump 
439: to saturation, implying the breaking and full polarization of these dimers 
440: at fields in excess of 1/3 of the saturation field. We speculate that this 
441: behavior may be due to the presence of a particularly stable state of the 
442: remaining spins which corresponds to a 1/3 magnetization plateau of the 
443: undoped kagome lattice.\cite{rhcs} The moments of the 
444: further--neighbor sites show a generally less systematic evolution, 
445: usually with a non--monotonic character which implies a very significant 
446: rearrangement of the induced local magnetization as a function of field. 
447: This evolution is again marked by the most significant changes occurring 
448: close to the net magnetization value of 1/3. 
449: 
450: NMR has emerged as the technique of choice for analyzing in real space  
451: the effects of doped impurities on magnetizations and spin correlations. 
452: Information concerning these quantities is extracted from Knight--shifts, 
453: linewidth alterations and, for certain cases, from relaxation times. A  
454: detailed NMR and NQR (nuclear quadrupole resonance) analysis has been 
455: conducted recently for $^{71}$Ga impurities in the kagome planes of the $S$ 
456: = 3/2 spin system SrCr$_{9p}$Ga$_{12-9p}$O$_{19}$ (SCGO).\cite{rlmcmombm} 
457: While a quantitative determination of any shifts or linewidths depends  
458: on a detailed Hamiltonian for the hyperfine interactions in a specific  
459: material, the results obtained in Fig.~7 may be considered as illustrative  
460: for the kagome geometry. 
461: 
462: The important qualitative features are the absence of a single spin--1/2 
463: degree of freedom anywhere in the vicinity of the impurity, and the rather 
464: large extent of the area of affected sites in spite of the very short spin 
465: correlation lengths. The latter effect is contained in Fig.~7 in that the 
466: induced site magnetization values on the vertical axis are directly 
467: proportional to the peak shifts measured in an NMR field sweep. Impurities 
468: in the kagome lattice therefore produce a large number of satellite peaks. 
469: 
470: Finally, with regard to the fact that most known kagome systems, including 
471: SCGO, are not $S = 1/2$, there remain rather few studies of this geometry 
472: for higher values of the spin.\cite{rhcs,rrsr} On rather general grounds 
473: based on the geometry of the lattice, one expects that both triangles and 
474: depleted triangles retain the integral or half--integral nature of the 
475: spin per site. Thus it is at least plausible that the results for the 
476: $S$ = 1/2 system are directly relevant to higher half--integral--spin 
477: systems such as SCGO.\cite{footnote2} 
478: 
479: \begin{figure}[t!] 
480: \psfig{figure=kkd2ed.eps,width=5.0cm,angle=270}
481: \medskip
482: \caption{Energy $\Delta_2$ for a pair of holes as a function of hole 
483: separation, obtained by ED for clusters of 12, 18 and 24 sites with two 
484: impurities. }
485: \label{fig:8} 
486: \end{figure} 
487: 
488: \section{Hole repulsion} 
489:  
490: In this section we present results characterizing the behavior of an 
491: even--site cluster doped with two impurities. A comparison with the 
492: undisturbed system\cite{rlblps} reveals that the average energy per 
493: remaining site not only increases in the presence of impurities, but 
494: increases further as the impurities are brought closer together. Both 
495: of these features are evident from the ED results of Fig.~8, which shows 
496: the energies of all different two--hole states available for the relevant 
497: system sizes. The higher energy per site of doped systems confirms, by 
498: considering the contributing processes, that the net loss of resonance 
499: energy for dimers constrained by the presence of impurities to occupy 
500: only the bonds on which they are essentially ``frozen'' (Sec.~III) is not 
501: compensated by the gains in dimerization energy for those bonds which 
502: are enhanced. 
503: 
504: The effective hole repulsion is most unusual in low--dimensional spin 
505: systems,\cite{rbhsl,rdrs} and in fact we are unaware of analogous results 
506: for any model other than the frustrated chain.\cite{rrnm} 
507: This repulsion of holes constitutes the central result of our analysis:
508: particularly in the context of short--range RVB wave functions, it 
509: contradicts directly the most straightforward expectation for the 
510: behavior of doped holes, namely that when mobile these would choose to 
511: minimize the net disruption of the system by occupying the same dimer, 
512: and thus would exhibit a pairing attraction.\cite{ra} The consequences 
513: for mobile impurites with repulsive interactions may be expected to 
514: include hole crystallization, phase separation and the possibility of 
515: transitions between metallic and insulating phases, and would bear 
516: considerable further investigation.
517: 
518: \begin{figure}[t!] 
519: \psfig{figure=kkd2dc.eps,width=5.0cm,angle=270}
520: \medskip
521: \caption{Comparison of energies $\Delta_{2,{\rm ED}}$ and $\Delta_{2,{\rm 
522: RVB}}$ obtained respectively from ED and from calculations in the dimer--RVB 
523: basis for a pairs of holes of different separations on clusters of 12, 
524: 18, and 24 sites with two impurities. }
525: \label{fig:9} 
526: \end{figure} 
527: 
528: \subsection{Dimer--RVB Basis}
529: 
530: We illustrate in Fig.~9 the efficacy of the dimer--RVB basis for the 
531: purposes of comparing ground--state energies. The correspondence between 
532: the dimer--RVB and ED results is shown by comparing the energies of all 
533: two--hole states in clusters of 12, 18 and 24 sites with two impurities, 
534: which represent all system sizes accessible by ED. Full details of the 
535: extraction of the energy values $\Delta_2$ in Fig.~9 are given below. 
536: Although an inspection of the abcissae reveals that the agreement between 
537: individual pairs of values is not as convincing as for the spin correlation 
538: functions (Sec.~III), the main difference between the two sets of data 
539: may be reduced to a relative upward shift of all the RVB energies. 
540: The direct linear correlation between ED and dimer--RVB results (alluded 
541: to in Sec.~II) demonstrates unambiguously that comparing the energies of 
542: different configurations of the two impurities, which is the primary goal 
543: of the energy analysis to follow, is a valid qualitative exercise when 
544: performed using the dimer--RVB results.
545: 
546: We show first the spin--spin correlation functions in the presence 
547: of a pair of impurities (Fig.~10). When considering the overlap of 
548: single--impurity dimerization patterns (Sec.~III), it is evident that 
549: these are not fully mutually compatible. In fact, one finds by inspection 
550: that not only the configuration illustrated in Fig.~10 but all possible 
551: configurations of two impurities on the kagome lattice are in some measure
552: incompatible. Comparison with other quantum spin systems suggests that 
553: this complete incompatibility is very specific to the kagome geometry, 
554: and its extent may be quantified by extracting the effective interaction 
555: energy of the holes (Fig.~11). In cluster calculations the frustration is 
556: also shown by the presence of relatively large local moments induced on 
557: the sites at the mismatch between dimerization regimes ({\it cf.}~Fig.~6). 
558: 
559: \begin{figure}[t!] 
560: \centerline{\psfig{figure=kkssc36c.eps,width=7.0cm,angle=0}}
561: \medskip
562: \caption{Spin--spin correlations for two impurities in a 36--site cluster, 
563: obtained in the dimer--RVB basis. Bar width represents the strength of 
564: correlation functions for each bond on a linear scale where the largest 
565: value is $- 0.66$. }
566: \label{fig:10} 
567: \end{figure} 
568:  
569: \subsection{Interaction between impurities}
570: 
571: Fig.~11 shows twice the energy per hole as a function of separation for all 
572: configurations of two impurities on 30-- and 36--site clusters, obtained 
573: in the dimer--RVB basis. We have analyzed the data in the following manner. 
574: $E_0$ is the ground--state energy of each cluster in the absence of 
575: impurities, and is computed for even clusters then extrapolated for 
576: odd clusters from the results for systems of sizes $N - 3$ and $N + 3$. 
577: $E_1$ is the ground--state energy in the presence of a single impurity, 
578: which is computed for odd clusters and may then be extrapolated for 
579: even clusters from the results for systems of sizes $N - 3$ and $N + 3$.
580: $E_2$ is the ground--state energy with two impurities, and again may be 
581: computed directly for even clusters and extrapolated for odd clusters.
582: We then define 
583: \begin{equation}
584: \Delta_1 = E_1 - E_0
585: \label{ed1}
586: \end{equation}
587: as the energy of one hole, 
588: \begin{equation}
589: \Delta_2 = E_2 - E_0
590: \label{ed2}
591: \end{equation}
592: as the energy of a two--hole state, and 
593: \begin{equation}
594: \epsilon = \Delta_2 - 2 \Delta_1 = E_2 - 2 E_1 + E_0
595: \label{ede}
596: \end{equation}
597: as the energy of interaction between two holes. Because of the extrapolation 
598: procedure it is important that the data sets for different system sizes be 
599: comparable. While certain problems arise for some system sizes due to the 
600: relatively large influence of the boundary conditions, we find the process 
601: to be valid for systems in excess of 27 sites. In this connection we note 
602: that the two--hole energies $\Delta_2$ shown in Fig.~11 are independent of 
603: the extrapolation between different system sizes, and it is only the direct 
604: comparison with $\Delta_1$ for the 33--site system which requires 
605: validation of the procedure.
606: 
607: \begin{figure}[t!] 
608: \psfig{figure=kkd2ls.eps,width=5.0cm,angle=270}
609: \medskip
610: \caption{Energies $\Delta_2$ for two--hole states as a function of hole
611: separation for all configurations of two impurities on 30-- and 36--site 
612: clusters, obtained by dimer--RVB calculations. The solid line marks twice 
613: the single--hole energy 2$\Delta_1$ relevant for comparison, taken from 
614: the 33--site cluster. }
615: \label{fig:11} 
616: \end{figure} 
617: 
618: It is clear that the effective mutual repulsion decays only slowly. Because 
619: of the restricted sizes of the acccessible clusters we are unable to 
620: speculate on the functional form of this decay, and in particular on the 
621: possibility that it may be long--ranged. We note only that for short 
622: distances (3--4 lattice constants) it appears to exhibit an approximate 
623: $1/d_r$ dependence, beyond which a steeper decrease is observed which may 
624: be either an intrinsic property of the interaction or a consequence of the 
625: system size. The form of the decay appears similar to that of the bond spin 
626: correlation functions towards their undisturbed value [Fig.~5(b)]. 
627: 
628: At the phenomenological level, a repulsive interaction between two 
629: impurities may be regarded as surprising in that their interference might 
630: have been expected to suppress the freezing of individual dimer patterns, 
631: leading to a restoration of resonance energy. Our findings to the contrary 
632: indicate that the dominant effect of increasing increasing impurity 
633: separation lies in the reduction in strength of the partially frozen 
634: single--impurity dimerization patterns at the boundaries where they are 
635: forced to mesh, which reduces the corresponding energetic penalty. Given 
636: the slow decay of the dimer--dimer correlations, the possibility remains 
637: that the net repulsive interaction between holes could be long--ranged. 
638: Such an unexpected result may indeed be connected with a true deconfinement 
639: of spinon--like excitations for the kagome system, which would confirm 
640: speculation to this effect in 2d.\cite{rsv} We comment further on the 
641: issue of deconfinement below. 
642: 
643: \section{Conclusion} 
644: 
645: We have illustrated a variety of novel phenomena associated with the 
646: effects of static, spinless impurities doped into the $S = 1/2$ 
647: antiferromagnetic Heisenberg model in kagome geometry. By analysis of 
648: spin--spin correlation functions and of the local magnetization patterns 
649: induced around impurities, we find that, despite the very short spin--spin 
650: correlation lengths of the highly frustrated system, dimer--dimer 
651: correlations develop over considerable distances due to a hole--induced 
652: freezing of their resonance. Furthermore, as a consequence of the 
653: continuum of states in the low--lying singlet manifold, there is no 
654: evidence for free local moments in the vicinity of impurity sites. 
655: 
656: Most importantly, we have shown from the total energies of doped systems 
657: that impurities in the kagome lattice have an effective repulsive 
658: interaction, a phenomenon unique to our knowledge and in direct 
659: contradiction to the conventional expectation for short--range RVB 
660: systems of a net pairing attraction. Furthermore, this effective 
661: repulsion is found to be rather long--ranged in nature, a most unusual 
662: result which is presumably related with the very strong frustration 
663: inherent to the kagome geometry. 
664: 
665: Our comparisons of doping effects provide additional justification for the 
666: validity of the short--range dimer--RVB framework as a suitable description 
667: of the ground--state singlet manifold in the kagome antiferromagnet. We 
668: have shown that the dimer--RVB basis provides a valuable means of probing 
669: the nature of spin and charge degrees of freedom in a further class of 
670: quantum magnets. This realization of a highly resonant RVB state, arising 
671: as a consequence of the extremely frustrated nature of the system, also 
672: presents an important contrast to other low--dimensional spin systems 
673: which are known to be described by RVB--type wave functions, particularly 
674: with regard to the localization of free spins and to the pairing of holes.
675: 
676: We summarize only briefly the consequences of our analysis for experiment. 
677: The majority of known kagome materials do not have complete filling of the 
678: lattice sites by a single value of the spin $S$, and as such the study of 
679: impurities is a significant component of their characterization. Nonmagnetic 
680: impurities in the kagome lattice have the important property that they do 
681: not generate localized spins at the vacant site, and do generate by the 
682: nature of their extended range of influence a significant number of 
683: satellites for each peak. On heuristic grounds we believe that our analysis 
684: of the qualitative features to be expected in the $S = 1/2$ system are 
685: relevant also for $S = 3/2$, and perhaps for higher half--integral spins. 
686: 
687: Finally, on the issue of deconfinement in highly frustrated systems of 
688: dimensionality higher than one, we have shown clearly from the presence 
689: of the mutual repulsive interaction between holes that charge degrees 
690: of freedom are deconfined. Further, the absence of localized $S = 1/2$ 
691: degrees of freedom around vacancies in the lattice demonstrates that 
692: spin and charge are deconfined. However, on the question of whether 
693: the spin excitations of the undoped system are appropriately described 
694: by deconfined spinon excitations, our results do not provide a proof of 
695: this hypothesis. 
696: 
697: \acknowledgments 
698:  
699: We are grateful to P. Sindzingre for helpful provision of ED spectra for 
700: undoped clusters. This work was supported by the Swiss National Science 
701: Foundation and by the French Centre National de Recherche Scientifique. 
702:  
703: \begin{thebibliography}{99} 
704:  
705: \bibitem{rce} J. T. Chalker and J. F. Eastmond, Phys. Rev. B {\bf 46}, 
706: 14 201 (1992).  
707: 
708: \bibitem{rle} P. W. Leung and V. Elser, Phys. Rev. B {\bf 47}, 5459 (1993).  
709:  
710: \bibitem{rze} C. Zeng and V. Elser, Phys. Rev. B {\bf 51}, 8318 (1995).  
711: 
712: \bibitem{rlblps} P. Lecheminant, B. Bernu, C. Lhuillier, L. Pierre, and 
713: P. Sindzingre, Phys. Rev. B {\bf 56}, 2521 (1997); C. Waldtmann, H.-U. 
714: Everts, B. Bernu, C. Lhuillier, P. Sindzingre, P. Lecheminant, and 
715: L. Pierre, Eur. Phys. J. B {\bf 2}, 501 (1998).
716: 
717: \bibitem{rm} F. Mila, Phys. Rev. Lett. {\bf 81}, 2356 (1998). 
718:  
719: \bibitem{rmm} M. Mambrini and F. Mila, Eur. Phys. J. B {\bf 17}, 651 (2000).  
720:  
721: \bibitem{rrk} D. S. Rokhsar and S. A. Kivelson, Phys. Rev. Lett. {\bf 61}, 
722: 2376 (1988).
723: 
724: \bibitem{rms} R. Moessner and S. L. Sondhi, Phys. Rev. Lett. {\bf 86}, 
725: 1881 (2001).
726: 
727: \bibitem{rmsp} G. Misguich, D. Serban and V. Pasquier, Phys. Rev. Lett. 
728: {\bf 89}, 137202 (2002); G. Misguich, D. Serban and V. Pasquier, unpublished 
729: (cond-mat/0302152).
730: 
731: \bibitem{rmblw} G. Misguich, B. Bernu, C. Lhuillier, and C. Waldtmann, 
732: Phys. Rev. Lett. {\bf 81}, 1098 (1998); W. LiMing, G. Misguich, P. Sindzingre, 
733: and C. Lhuillier, Phys. Rev. B {\bf 62}, 6372 (2000).
734: 
735: \bibitem{rwhrs} A. S. Wills, A. Harrison, C. Ritter, and  R. I. Smith, 
736: Phys. Rev. B {\bf 61}, 6156 (2000). 
737: 
738: \bibitem{rkea} A. Keren, K. Kojima, L. P. Le, G. M. Luke, W. D. Wu, 
739: Y. J. Uemura, M. Takano, H. Dabkowska, and M. J. P. Gingras, Phys. Rev. 
740: B {\bf 53}, 6451 (1996). 
741: 
742: \bibitem{rlea} S.-H. Lee, C. Broholm, M. F. Collins, L. Heller, 
743: A. P. Ramirez, Ch. Kloc, E. Bucher, R. W. Erwin, and N. Lacevic. 
744: Phys. Rev. B {\bf 56}, 8091 (1997). 
745: 
746: \bibitem{rhhkntkt} Z. Hiroi, M. Hanawa, N. Kobayashi, M. Nohara, H. Takagi, 
747: Y. Kato, and M. Takigawa, J. Phys. Soc. Jpn., {\bf 70}, 3377 (2001).
748: 
749: \bibitem{rlmcmombm} L. Limot, P. Mendels, G. Collin, C. Mondelli, B.  
750: Ouladdiaf, H. Mutka, N. Blanchard, and M. Mekata, Phys. Rev. B {\bf 65},  
751: 144447 (2002), and references therein. 
752:  
753: \bibitem{rlmgmdhlr} M. Laukamp, G. B. Martins, C. Gazza, A. L. Malvezzi, 
754: E. Dagotto, P. M. Hansen, A. C. Lopez, and J. Riera, Phys. Rev. B {\bf 57}, 
755: 10755 (1998).
756: 
757: \bibitem{rbhsl} N. Bulut, D. Hone, D. J. Scalapino, and E. Y. Loh, Phys. 
758: Rev. Lett. {\bf 62}, 2192 (1989).
759: 
760: \bibitem{rsv} S. Sachdev and M. Vojta, in Proceedings of the XIII  
761: International Congress on Mathematical Physics, edited by A. Fokas,  
762: A. Grigoryan, T. Kibble, and B. Zegarlinski, (International Press,  
763: Boston, 2001). 
764:  
765: \bibitem{rnm} B. Normand and F. Mila, Phys. Rev. B {\bf 65}, 104411 (2002). 
766: 
767: \bibitem{rs} S. Sachdev, Phys. Rev. B {\bf 45}, 12 377 (1992).  
768: 
769: \bibitem{footnote1} The behavior of the number of inequivant sites as a 
770: function of the spin sector arises from the presence of an additional 
771: permutation symmetry among the sites for the 27--site cluster with the 
772: chosen boundary conditions. This symmetry does not in general commute 
773: with the two reflection symmetries of the cluster, and so cannot be 
774: obtained simultaneously with these. However, in the majority of spin 
775: sectors it is possible to obtain all three symmetries simulataneously, 
776: making sites $i = 2$ and 3, $i = 5$ and $5^{\prime}$, and $i = 4$ and 7 
777: equivalent in these cases. 
778: 
779: \bibitem{rhcs} K. Hida, J. Phys. Soc. Jpn, 70, 3673 (2001); D. C. Cabra, 
780: M. D. Grynberg, P. C. W. Holdsworth, and P. Pujol,  Phys. Rev. B {\bf 65}, 
781: 094418 (2002); J. Schulenburg, A. Honecker, J. Schnack, J. Richter, and 
782: H.-J. Schmidt, Phys. Rev. Lett. {\bf 88}, 167207 (2002).
783: 
784: \bibitem{rrsr} I. Rudra, D. Sen, and S. Ramasesha, unpublished 
785: (cond--mat/0210122). 
786: 
787: \bibitem{footnote2} A further heuristic justification for the relevance 
788: of the current results at least to the $S = 3/2$ case may be obtained 
789: by considering that the kagome geometry supports a non--degenerate 
790: valence--bond--solid state\cite{raklt} for spins $S = 2$. In an RVB basis 
791: a system of spins $S = 3/2$ is equivalent to a set of dimer vacancies in 
792: this background which have exactly the same degeneracy as the dimers in 
793: the $S = 1/2$ system. 
794: 
795: \bibitem{raklt} I. Affleck, T. Kennedy, E. Lieb, and H. Tasaki, Phys. Rev. 
796: Lett. {\bf 59}, 799 (1987); Commun. Math. Phys.{\bf 115}, 477 (1988).
797: 
798: \bibitem{rdrs} E. Dagotto, J. Riera, and D. Scalapino, Phys. Rev. B {\bf 45}, 
799: 5744 (1992).
800: 
801: \bibitem{rrnm} Reexamination of results obtained by two of the 
802: authors\protect{\cite{rnm}} reveals that a mutual repulsion of 
803: impurities also occurs in the spin chain with frustrating next--neighbor 
804: interactions.
805: 
806: \bibitem{ra} P. W. Anderson, Science 235, 1196 (1987).
807: 
808: \end{thebibliography} 
809:  
810: \end{document} 
811:  
812: