1: \documentclass[aps,prb,twocolumn,superscriptaddress]{revtex4}
2: \usepackage{graphicx}
3: \usepackage{epsfig}
4: \usepackage{float}
5:
6: \newcommand{\bfg}[1]{\mbox {\boldmath $#1$}}
7: \newcommand{\asize}[1]{\renewcommand{\arraystretch}{#1}}
8: \newcommand{\be}{\begin{equation}}
9: \newcommand{\ee}{\end{equation}}
10: \bibliographystyle{apsrev}
11:
12:
13: \begin{document}
14:
15: \title{Sr$_2$Cu(PO$_4$)$_2$: A real material realization of the 1D nearest neighbor Heisenberg chain}
16: \author{M.D. Johannes}
17: \affiliation{Code 6390 Naval Research Laboratory, Washington, D.C. 20375}
18: \author{J. Richter}
19: \affiliation{Institute for Theoretical Physics, University of Magdeburg, Germany}
20: \author{S.-L. Drechsler}
21: \affiliation{Leibniz-Institute for Solid State and Materials Research (IFW Dresden), Dresden, Germany}
22: \author{H. Rosner}
23: \affiliation{Max-Planck-Institute for Chemical Physics of Solids, Dresden, Germany}
24:
25: \begin{abstract} We present evidence that crystalline Sr$_2$Cu(PO$_4$)$_2$ is a nearly perfect one-dimensional (1D) spin-1/2
26: anti-ferromagnetic Heisenberg model (AHM) chain compound with nearest neighbor only exchange. We undertake a broad theoretical study
27: of the magnetic properties of this compound using first principles (LDA, LDA+U calculations), exact diagonalization and Bethe-ansatz
28: methodologies to decompose the individual magnetic contributions, quantify their effect, and fit to experimental data. We calculate
29: that the conditions of one-dimensionality and short-ranged magnetic interactions are sufficiently fulfilled that Bethe's analytical
30: solution should be applicable, opening up the possibility to explore effects beyond the infinite chain limit of the AHM Hamiltonian.
31: We begin such an exploration by examining some extrinsic effects such as impurities and defects. \end{abstract}
32:
33:
34: \maketitle
35:
36:
37: \section{Introduction} Interest in magnetically low-dimensional systems began with the advent of quantum mechanics and the
38: development of spin-spin interaction models to explain magnetic behavior. Though deceptively simple, early models such as
39: Ising \cite{ising} or Heisenberg \cite{heis} cannot be solved for the general case, requiring either low spin or low
40: spatial dimensionality to obtain analytical solutions. Exact solutions for some specific cases, such as Onsager's solution
41: \cite{onsager} to the two-dimensional (2D) Ising model or Bethe's solution \cite{Bethe} to the spatially one-dimensional
42: (1D) s=1/2 Heisenberg chain, inspired general interest in the theoretical properties of magnetism in low dimensions, and
43: resulted in a variety of predictions for exotic physical behaviors. Some of the most interesting properties to arise from
44: the study of restricted-dimension magnets are due to the dramatic effect of quantum spin fluctuations \cite{qm}. These are
45: intimately involved in the emergence of novel ground states and excitation spectra, and, in recent decades, have become a
46: leading candidate as the pairing mechanism for electrons in quasi-2D high temperature superconductors \cite{JPC99}, thus
47: providing the field with a more practical aspect. Mermin and Wagner demonstrated \cite{Mermin-Wagner} that strong spin
48: fluctuations suppress magnetic long range order (LRO) until T=0 in isotropic spin systems in 1D and 2D,
49: regardless of the interaction
50: strength (exchange) between neighboring spins. Since real physical compounds are 3D by nature, the continuing effort to
51: experimentally verify such predictions and probe the nature of 1D magnets is aimed at finding materials where magnetic
52: interactions proceed predominantly along one direction. A measureable gauge of success or failure along
53: these lines can be obtained through the ratio $k_BT_N/J_1$, which compares ordering temperature of a N\'eel state to the
54: magnitude of exchange between neighboring spins. A perfectly 1D system would give $k_BT_N/J_1$ = 0. Additionally,
55: experimentally observed behaviors should conform to theoretical predictions where the conditions of the model are met.
56:
57: Recently, very low temperature measurements \cite{beliknew} on the spin-1/2 compound Sr$_2$Cu(PO$_4$)$_2$
58: identified the onset of magnetic LRO at T$_N$= 0.085 K, which, in combination with the extracted exchange
59: constant, yields a ratio $k_BT_N/J_1$ = 6x10$^{-4}$. This is nearly a full order of magnitude less than the
60: ratio \cite{moto} for prototype 1D magnet Sr$_2$CuO$_3$, which has $k_BT_N/J_1$ = 2.5x10$^{-3}$. We can
61: estimate the remaining interchain coupling, $J'$, by adopting an effective 3D chain model with $z_{\perp}$=4
62: nearest neighbor chains as in Ref. \onlinecite{irkhin00}:
63:
64: \begin{equation} J'=\frac{3.046k_BT_N}
65: {k_{AFM}z_{\perp}\sqrt{\ln \frac{5.8J_1}{k_BT_N} +0.5\ln \ln\frac{5.8J_1}{k_BT_N}}} \approx 2.9\times
66: 10^{-2}\mbox{K}, \label{qcc} \end{equation}
67:
68: where $k_{AFM}$ is the magnitude of the AFM wavevector. All signs, therefore point to an extremely high degree of
69: one-dimensionality that should justify the use of Bethe's exact analytical solution to the 1D spin-1/2 AHM in a
70: wide temperature range $J' \ll k_BT$, provided that indeed {\it only nearest-neighbor} interactions are present.
71: Previous studies \cite{helge} have shown that this condition is not satisfactorily fulfilled by the leading 1D
72: spin-1/2 AHM candidate, Sr$_2$CuO$_3$. For completeness, another candidate for a 1D-AHM system should be
73: mentioned: the linear charge transfer salt [3,3'-dimethyl-2,2'-thiazolinocyanine]-TCNQ
74: \cite{takagi96,remarktakagi}.
75:
76:
77:
78: Here, we undertake an extensive theoretical analysis of Sr$_2$Cu(PO$_4$)$_2$, employing first principles density
79: functional theory calculations, an exact diagonalization scheme, and finally, a Bethe-Ansatz derived expression
80: for the magnetic susceptibility fit to experimental data. Our results are in remarkable agreement with one
81: another and with experimental observations. We show that Sr$_2$Cu(PO$_4$)$_2$ is indeed extremely 1D and that
82: furthermore, the second-neighbor interactions are vanishingly small, eliminating any complications due to in-chain
83: frustration. We claim, therefore, that this compound is the best realization of a nearest-neighbor only
84: Heisenberg spin 1/2 chain known to date and will likely be a valuable tool for extracting intrinsic effects beyond
85: the Bethe-ansatz {\it i.e.} effects not contained in the Heisenberg Hamiltonian, such as Dzyaloshinsky-Moriya
86: interactions \cite{dzy}, ring exchange processes, or spin-lattice coupling effects. Additionally, extrinisic
87: effects due to sample imperfections such as defects, magnetic impurities or the presence of alternate phases can
88: be quantitatively explored, and we make a preliminary investigation of some of them.
89:
90:
91: \begin{figure}[tbp] \begin{center} \includegraphics[width = 0.95\linewidth]{struct.ps}
92: \end{center} \caption{(Color online) {\it
93: Upper panel} The structure of Sr$_2$Cu(PO$_4$)$_2$, showing the isolated CuO$_4$ plaquettes as
94: squares with O ions located at the corners, and the $PO_4$ units as tetrahedra with $P$ located
95: in the center. {\it Lower panel} A top view of the spin chain plane with the various hopping
96: paths labelled. Hopping to the nearest out-of-plane neighbor, t$_{\perp}$, is not shown.}
97: \label{struct} \end{figure}
98:
99:
100: \begin{figure}[b]
101: \begin{center}
102: \includegraphics[width=0.95\linewidth]{bandplot.ps}
103: \includegraphics[width=0.95\linewidth]{tbfit.ps}
104: \end{center}
105: \caption{(color online) The bandstructure of Sr$_2$Cu(PO$_4$)$_2$, showing the single metallic band
106: well separated from all others. The $X-S$ and $Y-\Gamma$ directions are along the magnetic chain, while
107: the $S-Y$ and $\Gamma-Z$ directions are
108: perpendicular to the chain. The inset shows the total and orbitally resolved DOS for the single
109: band. In the lower panel, a
110: blow-up of this band is shown with tight-binding eigenvalues superimposed to show
111: the remarkable reproduction of the dispersion.
112: } \label{BS} \end{figure}
113:
114:
115: \section{First Principles and Tight-Binding} First principles density functional theory (DFT) bandstructure
116: calculations were performed using a full-potential local orbital code, FPLO \cite{fplo}, with the following
117: valence states: Sr (4s,4p,4d,5s,5p), Cu (3s,3p,3d,4s,4p), P (2s,2p,3s,3p,3d), O (2s,2p,3d). The structure,
118: lattice constants, and atomic positions (see Fig. \ref{struct}) were taken from experiment
119: \cite{belik1,AAB+02}, $a$ = 11.515 \AA, $b$ = 5.075 \AA, $c$ = 6.5748 \AA.
120: Sr$_2$Cu(PO$_4$)$_2$ and isostructural Ba$_2$Cu(PO$_4$)$_2$, which we calculate as a
121: comparison material, are composed of isolated CuO$_4$ plaquettes surrounded by PO$_4$ tetrahedra. The
122: plaquettes are spaced evenly along the $b$ axis, forming chains that are staggered with respect to one
123: another in the $a$ crystal direction. The planar CuO$_4$ units are tilted with respect to the $a-b$ crystal
124: plane. Based on the geometry of these two systems, we enumerate five specific interactions between
125: plaquettes, and therefore between spins localized to these plaquettes, that may be necessary to describe the
126: electronic and magnetic structures. These interactions are labelled in Fig. \ref{struct} as various hopping
127: parameters to be later included in a tight-binding model.
128:
129: The paramagnetic band structure of Sr$_2$Cu(PO$_4$)$_2$ (Fig. \ref{BS}) shows a single, isolated, half-filled band,
130: derived predominantly from the Cu $3d_{x^2-y^2}$-O $2p_{\sigma}$ molecular plaquette orbital, crossing the Fermi
131: energy. In reality, the system is antiferromagnetic (AFM) and insulating, but we will follow the standard procedure
132: of importing hopping parameters from the "uncorrelated" paramagnetic system to the Hubbard model which then maps onto
133: the Heisenberg model. Magnetism and correlation effects can also be added at the DFT level with, as we will show,
134: very consistent results. The 1D character of the system is qualitatively obvious from the nearly dispersionless
135: bands in directions perpendicular to the magnetic chains (S-Y and $\Gamma$-Z) and from the characteristic logarithmic
136: divergences in the density of states (DOS) near the band edges. To quantitatively compare microscopic magnetic
137: interactions, we fit a tight-binding (TB) model to our paramagnetic band structure and calculated the individual
138: exchange constants between various CuO$_4$ plaquette spins using J$_{ij}$ = 4t$_{ij}^2$/U$_{eff}$ with U$_{eff}$=4.5
139: eV. We find this to be a reasonable choice for U$_{eff}$ because of poor screening in this geometry and because of
140: the small inter-plaquette repulsion. The hopping parameters included in the model are shown
141: schematically in Fig. \ref{struct} with the numerical values and derived superexchange constants given in Table
142: \ref{tbparam}. The resulting TB dispersion, which uses only 5 fitting parameters, is indistinguishable from the
143: full-potential calculation (Fig. \ref{BS} lower panel), indicating that further interactions can be safely ignored.
144: The ratio of the strongest in-chain coupling to the strongest interchain coupling is J$_1$/J$_1^{ic}$ $\sim$ 70 and
145: the ratio of first to second neighbor in-chain coupling is J$_1$/J$_2$ $\sim$ 700 for Sr$_2$Cu(PO$_4$)$_2$. Identical
146: calculations based on the band structure of Ba$_2$Cu(PO$_4$)$_2$ (not shown) yield similar results with somewhat more
147: inter-chain coupling but less second neighbor in-chain coupling. Both systems can therefore be considered as
148: strongly one-dimensional, with Ba$_2$Cu(PO$_4$)$_2$ slightly less so. Naturally, the choice of U$_{eff}$ is simply a
149: best estimation and results for $J$ will vary slightly with this choice, while the ratios will remain constant.
150:
151: \begin{table}[h] \caption{Tight-binding hopping parameters (in units of meV) and derived exchange constants (in units
152: of K) for A$_2$Cu(PO$_4$)$_2$, A=Sr,Ba. See Fig.
153: \ref{struct} for the
154: relationship of the hopping parameters to the structure.}
155: \begin{tabular*}{0.95 \linewidth}{@{\extracolsep{\fill}}||ccccccccccccc||} \hline \hline
156: &(meV) & \vline & t$_1$ & \vline & t$_2$ & \vline & t$_1^{ic}$ & \vline & t$_2^{ic}$ & \vline & t$_{\perp}$ & \\ \hline
157: & Sr & \vline & 135 & \vline & 5.1 & \vline & 16.3 & \vline & 3.4 & \vline & 1.4 &\\ \hline \hline
158: & Ba & \vline & 122 & \vline & 0.9 & \vline & 10.3 & \vline & 4.7 & \vline & 1.8 & \\ \hline \hline
159: \end{tabular*}
160:
161: \vspace{0.05 in}
162:
163: \begin{tabular*}{0.95 \linewidth}{@{\extracolsep{\fill}}||ccccccccccccc||} \hline \hline
164: &(K)& \vline & J$_1$ & \vline & J$_2$ & \vline & J$_1^{ic}$ & \vline & J$_2^{ic}$ &\vline & J$_{\perp}$ & \\ \hline
165: & Sr & \vline & 187 & \vline & 0.268 & \vline & 2.7 & \vline & 0.119 & \vline & 0.02 &\\ \hline \hline
166: & Ba & \vline & 154 & \vline & 0.008 & \vline & 1.09 & \vline & 0.228 & \vline & 0.03 & \\ \hline
167: \hline \end{tabular*}
168:
169: \label{tbparam}
170: \end{table}
171:
172: The energy difference between FM and AFM ordered spin configurations can be calculated using the local spin density approximation
173: (LSDA) \cite{JPP92} which allows for separate spin-up and spin-down densities. Since the LSDA is known to badly underestimate the
174: onsite Coulomb interaction in localized systems, we applied the LSDA+U methodology to better account for the correlated Cu
175: $3d$-orbitals, using the fully localized limit scheme \cite{via93} to correct for double-counting terms. We map a classical N\'eel
176: state and a classical ferromagnetic state onto the Heisenberg spin model, including only 1D nearest neighbor interactions.
177: Comparing the resulting model energy difference to the LDA energy difference between FM and AFM states (per spin), we derive an
178: effective exchange constant, J$_{eff}$, in the following way:
179:
180: \begin{equation}
181: H = \sum_{i,j} J_{ij} \mathbf{S}_i \cdot \mathbf{S}_j; \quad E_{FM} - E_{AFM} =
182: 2J_{eff}|s|^2 ,\quad s=1/2,
183: \label{heis}
184: \end{equation}
185:
186: As expected, the energy difference, and therefore $J_{eff}$, decreases as U$_d$ (not to be confused with the
187: considerably smaller one-band
188: parameter U$_{eff}$ that contains O 2$p$ contributions in addition to Cu $3d$) increasingly localizes the Cu $3d$ electrons. For a
189: range of U$_d$ between 6 eV and 9 eV, we find that $J_{eff}$ varies from 261 K to 160 K. Since U$_d$ is a local
190: quantity
191: and since the Cu-O
192: bond distance in Sr$_2$Cu(PO$_4$)$_2$ is only 1\% different than in the plaquettes of the widely studied high T$_c$ precursor systems,
193: we adopt the commonly used value of U$_d$ = \mbox{8 eV}. This corresponds to a value of 190 K for $J_{eff}$. Note that
194: since the LSDA (and LSDA+U)
195: energy differences include contributions from {\it all} exchange processes in the system, $J_{eff}$ cannot in general be considered as
196: either purely superexchange or purely 1D. However, comparison with the individual superexchange parameters derived from the TB fit
197: shows that both assumptions, in this case, are quite valid. The nearest neighbor in-chain TB exchange constant has a value $J_1$ = 187
198: K, in exceptional agreement with the $J_{eff}$ value of 190 K, indicating that the next-nearest neighbor interactions, FM exchange
199: processes and residual 2D and 3D interactions must therefore be extremely small. Of course both values can be made to vary somewhat by
200: choosing U$_d$ and U$_{eff}$ differently, thus affecting the agreement as well. We expect that our calculated value
201: of $J$ will be larger
202: than the experimental value, as it is well known that the band dispersion from which we derive $t$ and subsequently $J$ is generally
203: exaggerated by the LDA. Indeed, the experimentally derived value of $J$ is 143 K, in good but not perfect agreement with our
204: calculations. We emphasize that any renormalization of the hopping parameters stemming from effects outside the LDA will cancel in the
205: ratio ($J_1$/$J_2$) so that the precise calculated value of $J$ has, in any case, no bearing on the establishment of the compound's
206: pronounced magnetic one-dimensionality and short-ranged magnetic interactions that are the primary aim of our first principles study.
207:
208:
209: \section{Exact Diagonalization} We perform an exact diagonalization calculation using ten sites along two staggered AFM magnetic chains
210: (20 sites total) for calculating thermodynamic properties such as specific heat and magnetic susceptibility, and
211: using 36 total sites for obtaining the ground state properties.
212: We compare calculations including the three largest exchange interactions, $J_1$, $J_2$ and $J_{ic}^1$, as
213: listed in Table \ref{tbparam},
214: to calculations using only $J_1$. Since in our model each chain has only one neighboring chain, we also perform a calculation in which
215: $J_{ic}^1$ is doubled to account for the existence, in reality, of two neighboring chains. In the case of specific heat, there is no
216: discernable difference between any of the three curves using these different parameter sets. For the calculation of $\chi$, the curves
217: are identical for the majority of the temperature range explored ($ 0 < k_bT/J_1 < 3$), but a barely visible difference occurs near the
218: peak of the curve (see Fig. \ref{ed}). The maximal difference occurring between two curves is 0.9\%, at about
219: $k_BT/J_1$=0.65. The
220: ground state calculations of the spin-spin correlation, $\langle S_i \cdot S_j\rangle$ are again completely indistinguishable.
221: Obviously, more distant and weaker interactions, such as $J_{ic}^2$ and $J_{\perp}$ will have even less of an effect. From this we
222: conclude that Sr$_2$Cu(PO$_4$)$_2$ is essentially free of magnetic interactions beyond the first neighbor and is an ideal candidate for
223: Bethe-ansatz calculations, which we now discuss.
224:
225: \begin{figure}
226: \includegraphics[width=0.95\linewidth]{chi.ps}
227: \caption{A 20 site exact diagonalization calculation of the susceptibility per spin of Sr$_2$Cu(PO$_4$)$_2$. The
228: inset shows a blow-up of the
229: region where the curves have the greatest discrepancy - 0.9\%. Curves for the ground state correlation of spins and for the specific
230: heat per spin show even less deviation.}
231: \label{ed}
232: \end{figure}
233:
234: \section{Comparison with experiment}
235: The basic theory for non-periodic, open AHM
236: chains was worked out by Furusaki and Hikihara \cite{Furu} (FH) and also by Zvyagin and Makarova \cite{Zvy} (ZM). FH
237: considered a
238: half-infinite chain with $one$ free chain end applying bosonization theory whereas ZM considered finite even-membered chains on the
239: basis of a rigorous theory based on the Bethe ansatz. Both approaches result, at low temperature, in a chain length ($L=Na$)
240: dependent
241: diverging contribution to the total magnetic susceptibility
242: $\chi \propto 1/NT \ln(T_0J_1$)
243: and to the linear coefficient in the specific heat $\gamma
244: =C_p/T \propto 1/NT \ln^\beta (T_0J_1/T)$, with $\beta$=2,4 in the FH and ZM theory,
245: respectively. Since a real chain has two ends we multiplied the FH expressions
246: by a factor of 2. ZM calculated further logarithmic corrections which
247: we adopted here to be valid
248: for the FH case, too.
249: Then within both approaches we arrive finally at the same expression for the
250: chain end contributions to
251: $\chi$ (up to a factor of 3/4). For the logarithmic constant $T_0$
252: we used the same value 5.696 as proposed by Johnston {\it et al.} in
253: the fit expression (fit2) for the bulk susceptibility \cite{johnston},
254: similarly to 5.8 used in Eq.\ (1).
255: In the shown and described fits we have adopted the formalism of FH, modified as described above, for both $C/T$ and $\chi$.
256:
257: \begin{figure}[tbp]
258: \includegraphics[width= 0.95 \linewidth]{figure.ps}
259: \caption{Bethe-Ansatz fits to experimental data with corrections for broken chains, impurity phases, and domain boundaries. Note the
260: consistency of the exchange parameter vs. the wide spread in values for the chain length.}
261: \label{BA}
262: \end{figure}
263:
264: \begin{table}
265: \caption{The collected results of fits to different sets of $\chi (T)$ data for Sr$_2$Cu(PO$_4$)$_4$ and Ba$_2$Cu(PO$_4$)$_4$.}
266: \begin{tabular*}{0.95 \linewidth}{@{\extracolsep{\fill}}|c|c|c|c|} \hline \hline
267: Sr$_2$Cu(PO$_4$)$_2$ &J(K) & N & g \\ \hline \hline
268: Belik \cite{belik1} & 144.9 &114 &2.713 \\ \hline
269: Belik \cite{beliknew} & 143.9 & 337.7 & 2.154 \\ \hline
270: Nath \cite{loidl} & 152.6 & 5 & 1.952 \\ \hline
271: \end{tabular*}
272:
273: \vspace{0.05 in}
274:
275: \begin{tabular*}{0.95 \linewidth}{@{\extracolsep{\fill}}|c|c|c|c|} \hline \hline
276: Ba$_2$Cu(PO$_4$)$_2$ &J(K) & N & g \\ \hline \hline
277: Belik \cite{belik1} & 131.8 & 48 & 2.073 \\ \hline
278: Nath \cite{loidl} & 140.1 & 5 & 2.041 \\ \hline
279: \end{tabular*}
280: \end{table}
281:
282: We fit to both susceptibility and specific heat to data taken on the same sample \cite{belik1}, additionally using data from a later
283: sample \cite{loidl} for which only $\chi$ data was available. The fit quality for specific heat and susceptibility are very similar;
284: results for the latter are shown in Fig. \ref{BA}. We get very good fits throughout the entire temperature range and find a reasonably
285: consistent value for the exchange parameter, $J$, despite fitting to samples of different qualities and to two different measurements
286: ($C/T$ and $\chi$). It is worth noting that the exchange parameter derived from fitting to $C/T$ using the data of Ref.
287: \onlinecite{belik1} yields an exchange parameter of J=134.4 K, which is somewhat less than the value derived from fitting to $\chi (T)$
288: using the {\it same} sample. This is likely caused by non-magnetic impurity contributions (not accounted for in our model) that affect
289: the specific heat but not the susceptibility at low temperatures. Since the magnetic component of $C$ is $\propto 1/J$, ignoring
290: the
291: non-magnetic contribution overestimates this term and therefore underestimates the exchange. Unlike the exchange and $g$ values, the
292: chain length parameter, $N$, varies widely between samples. While this is expected for samples of different quality, the variation is
293: surprisinly high and, more importantly, the chain lengths resulting from our fits are far too small to justify the use of the open chain
294: theories that we have employed at very low temperatures. With chain lengths of this order, the low temperature region will be
295: completely dominated by broken chain physics that requires different, and as-of-yet undeveloped, formalism. Having fit throughout a
296: large temperature range, including regions where broken chain physics is inoperative, we feel that the extracted exchange constant, J
297: $\approx$ 145 K, is nonetheless relevant - a belief that is supported by its consistency between fits and its similarity to the
298: experimentally measured value.
299:
300:
301: \section{Discussion} The application of a variety of theoretical techniques to the problem of magnetism in Sr$_2$Cu(PO$_4$)$_2$
302: convincingly demonstrates that the ideal compound is highly one-dimensional. The high degree of one-dimensionality can be traced
303: back to its unusual isolated CuO$_4$ plaquette geometry. Instead of edge- or corner- shared plaquettes such as are common in
304: other quasi-1D compounds \cite{KMK+97,YMTT+98}, each Cu ion in Sr$_2$Cu(PO$_4$)$_2$ is surrounded by four O ions not shared by
305: any other Cu ion. This construction virtually eliminates the second neighbor in-chain coupling that prevents edge-shared
306: compounds such as Li$_2$CuO$_2$ from being described via a simple nearest neighbor Heisenberg model \cite{Warren,Rosner}.
307: Corner shared cuprates such as Sr$_2$CuO$_3$ have far smaller second neighbor interactions \cite{moto}, of the order J$_1$/J$_2$
308: $\sim$ 15, and yet, these must be taken into account to get good agreement between model calculations and experiment
309: \cite{helge}. The structure of Sr$_2$Cu(PO$_4$)$_2$ along the chain is that of an edge shared chain compound with every other
310: unit missing. Conceptualized in this way, one can make a correspondence between exchange constants in a edge-shared ($es$) system
311: and those in the isolated square plaquette ($sp$) geometry: J$_2^{es}$ $\rightarrow$ J$_1^{sp}$, and J$_4^{es}$ $\rightarrow$
312: J$_2^{sp}$. Since J$_4^{es}$ is known to be vanishingly small in the edge-shared geometry, it is clear that the second neighbor
313: interactions in the square plaquette geometry can be expected to be negligible. This may provide some directional guidance in the
314: search for new one-dimensional compounds: the isolated plaquette arrangement appears to be superior to the more common edge- or
315: corner-shared structures such that synthesis of new compounds with this geometry may prove to be profitable. The tilting of the
316: out-of-chain plaquettes with respect to one another further suppresses the inter-chain coupling. The staggering of plaquettes in
317: neighboring chains slightly increases the distance between spins, but more importantly, gives rise to frustration. As each chain
318: is antiferromagnetically aligned by the (relatively) strong first neighbor coupling, a given spin finds itself surrounded by four
319: interchain neighbors, two aligned in one direction and two in the other. These staggered, frustrated chains are more decoupled
320: from one another than they would be in another arrangement, {\it e.g.} a ladder configuration.
321:
322:
323: Provided that Sr$_2$Cu(PO$_4$)$_2$ is stoichiometric and largely defect-free, it is clear that this compound represents the most
324: 1D AHM chain so far investigated. These conditions are, unfortunately, not reasonably fulfilled by current samples. It is
325: interesting in this context to consider the mechanism by which the compound eventually achieves LRO (at $T_N$ = 0.085 K): is it
326: truly the result of residual third dimension interactions? Significantly, the phenomenologically estimated averaged interchain
327: interaction from Eq. \ref{qcc} is of the same order as the calculated $J_{\perp}$, rather than $J_{ic}^1$. Fluctuation induced
328: "order by disorder" coupling could be responsible for the strong reduction of two orders of magnitude within the frustrated plane.
329: On the other hand, the interchain couplings are in general phenomenally small as calculated by DFT methods and even so are likely
330: exaggerated. One alternative explanation is that in a system with many broken chains, there will be some number of chains
331: containing an odd number of spins, with each such chain carrying one uncompensated spin-1/2 electron. The relationship of the
332: uncompensated spins to one another is not defined by any of our methods and a long range ordering of these is not out of the
333: question. It would be interesting to see if the ordering temperature remains constant with sample quality. Another point to be
334: addressed in the future is the issue of spin-lattice coupling. The Heisenberg model itself assumes perfect isotropy in spin-space
335: and we have not included any relativistic (spin-orbit) interactions in our first principles calculations. The neglect of these is
336: seemingly justified by the extremely small field (H= 4mT) at which the spin-flop transition occurs \cite{beliknew}, but the
337: smallness of this field itself is unusual and a cause for further investigation. All of these facts point to the high desirability
338: of better samples that can be used to disentangle true "dimensionality" effects from behaviors due to crystal imperfections.
339: Although the investigation of true Heisenberg physics is currently limited by sample quality issues, Sr$_2$Cu(PO$_4$)$_2$ is
340: theoretically, and potentially experimentally, the best example of a magnetically 1D crystal yet studied.
341:
342: \section{Conclusion} We have shown that the isolated CuO$_4$ plaquette geometry of Sr$_2$Cu(PO$_4$)$_2$ gives
343: rise to a nearly perfect 1D spin-1/2 nearest neighbor only system. We find a ratio $k_BT_N/J_1$ = 6x10$^{-4}$,
344: in good agreement with experimental finding and show that secondary interactions (2D,3D and next-nearest
345: neighbor) are negligible in terms of calculated thermodynamic properties. Using the Bethe-ansatz solution to the
346: Heisenberg Hamiltonian along with additional terms to correct for extrinsic non-crystalline effects, we fit the
347: data over a large temperature range and derive an exchange parameter of 145 $\pm$ 5 K that is consistent between
348: samples and between fitting choices. We find that sample quality, particularly the existence of numerous broken
349: chains, currently prohibits experimental observation of true spin-1/2 AHM physics. However, Sr$_2$Cu(PO$_4$)$_2$
350: is truly intrinsically perfectly 1D with only one exchange parameter, and as better and better methods of
351: generating the compound emerge, effects beyond Bethe-ansatz can be probed experimentally.
352:
353: {\it Acknowledgements} We are grateful to A.A. Belik and A. Loidl for making data available to us and for
354: valuable input to our work. We also thank A.A. Zvyagin for fruitful discussions. We acknowledge the use of
355: J.~Schulenburg's {\it spinpack} to perform the numerical exact diagonalization. Our calculations were carried
356: out in part using the supercomputing facilities at ZIH Dresden. Funding was provided by the Emmy Noether Program
357: of Deutsche Forschungsgemeinschaft. MDJ is funded by the Office of Naval Research.
358:
359: \begin{thebibliography}{27}
360: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
361: \expandafter\ifx\csname bibnamefont\endcsname\relax
362: \def\bibnamefont#1{#1}\fi
363: \expandafter\ifx\csname bibfnamefont\endcsname\relax
364: \def\bibfnamefont#1{#1}\fi
365: \expandafter\ifx\csname citenamefont\endcsname\relax
366: \def\citenamefont#1{#1}\fi
367: \expandafter\ifx\csname url\endcsname\relax
368: \def\url#1{\texttt{#1}}\fi
369: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
370: \providecommand{\bibinfo}[2]{#2}
371: \providecommand{\eprint}[2][]{\url{#2}}
372:
373: \bibitem[{\citenamefont{{E. Ising}}(1925)}]{ising}
374: \bibinfo{author}{\bibnamefont{{E. Ising}}}, \bibinfo{journal}{Z. Physik}
375: \textbf{\bibinfo{volume}{31}}, \bibinfo{pages}{253} (\bibinfo{year}{1925}).
376:
377: \bibitem[{\citenamefont{{W. Heisenberg}}(1928)}]{heis}
378: \bibinfo{author}{\bibnamefont{{W. Heisenberg}}}, \bibinfo{journal}{Z. Physik}
379: \textbf{\bibinfo{volume}{49}}, \bibinfo{pages}{619} (\bibinfo{year}{1928}).
380:
381: \bibitem[{\citenamefont{{L. Onsager}}(1944)}]{onsager}
382: \bibinfo{author}{\bibnamefont{{L. Onsager}}}, \bibinfo{journal}{Phys. Rev.}
383: \textbf{\bibinfo{volume}{65}}, \bibinfo{pages}{117} (\bibinfo{year}{1944}).
384:
385: \bibitem[{\citenamefont{{H.A. Bethe}}(1931)}]{Bethe}
386: \bibinfo{author}{\bibnamefont{{H.A. Bethe}}}, \bibinfo{journal}{Z. Phys.}
387: \textbf{\bibinfo{volume}{71}}, \bibinfo{pages}{205} (\bibinfo{year}{1931}).
388:
389: \bibitem[{\citenamefont{{U. Schollw\"{o}ck} et~al.}(2004)\citenamefont{{U.
390: Schollw\"{o}ck}, {J. Richter}, {D. J. J. Farnell}, and {R. F. Bishop,
391: Eds.}}}]{qm}
392: \emph{\bibinfo{title}{Quantum Magnetism}},
393: \bibinfo{author}{\bibnamefont{{U. Schollw\"{o}ck}}},
394: \bibinfo{author}{\bibnamefont{{J. Richter}}},
395: \bibinfo{author}{\bibnamefont{{D. J. J. Farnell}}}, \bibnamefont{and}
396: \bibinfo{author}{\bibnamefont{{R. F. Bishop, Eds.}}},
397: \bibinfo{pages}{Lecture Notes
398: in Physics 645}
399: (\bibinfo{publisher}{Springer}, \bibinfo{address}{Berlin},
400: \bibinfo{year}{2004}).
401:
402: \bibitem[{\citenamefont{{J. P. Carbotte} et~al.}(1999)\citenamefont{{J. P.
403: Carbotte}, {E. Schachinger}, and {D. N. Basov}}}]{JPC99}
404: \bibinfo{author}{\bibnamefont{{J. P. Carbotte}}},
405: \bibinfo{author}{\bibnamefont{{E. Schachinger}}}, \bibnamefont{and}
406: \bibinfo{author}{\bibnamefont{{D. N. Basov}}}, \bibinfo{journal}{Nature}
407: \textbf{\bibinfo{volume}{401}}, \bibinfo{pages}{354} (\bibinfo{year}{1999}).
408:
409: \bibitem[{\citenamefont{{N.D. Mermin} and {H. Wagner}}(1966)}]{Mermin-Wagner}
410: \bibinfo{author}{\bibnamefont{{N.D. Mermin}}} \bibnamefont{and}
411: \bibinfo{author}{\bibnamefont{{H. Wagner}}}, \bibinfo{journal}{Phys. Rev.
412: Lett.} \textbf{\bibinfo{volume}{17}}, \bibinfo{pages}{1133}
413: (\bibinfo{year}{1966}).
414:
415: \bibitem[{\citenamefont{{A. A. Belik} et~al.}(2005)\citenamefont{{A. A. Belik},
416: {S. Uji}, {T. Terashima}, and {E. Takayama-Muromachi}}}]{beliknew}
417: \bibinfo{author}{\bibnamefont{{A. A. Belik}}},
418: \bibinfo{author}{\bibnamefont{{S. Uji}}}, \bibinfo{author}{\bibnamefont{{T.
419: Terashima}}}, \bibnamefont{and} \bibinfo{author}{\bibnamefont{{E.
420: Takayama-Muromachi}}}, \bibinfo{journal}{J. Solid State Chem.}
421: \textbf{\bibinfo{volume}{178}}, \bibinfo{pages}{3461} (\bibinfo{year}{2005}).
422:
423: \bibitem[{\citenamefont{{N. Motoyama} et~al.}(1996)\citenamefont{{N. Motoyama},
424: {E. Eisaki}, and {S. Uchida}}}]{moto}
425: \bibinfo{author}{\bibnamefont{{N. Motoyama}}},
426: \bibinfo{author}{\bibnamefont{{E. Eisaki}}}, \bibnamefont{and}
427: \bibinfo{author}{\bibnamefont{{S. Uchida}}}, \bibinfo{journal}{Phys. Rev.
428: Lett.} \textbf{\bibinfo{volume}{76}}, \bibinfo{pages}{3212}
429: (\bibinfo{year}{1996}).
430:
431: \bibitem[{\citenamefont{{V.Yu.\ Irkhin} and {A.A.\ Katanin}}(2000)}]{irkhin00}
432: \bibinfo{author}{\bibnamefont{{V.Yu.\ Irkhin}}} \bibnamefont{and}
433: \bibinfo{author}{\bibnamefont{{A.A.\ Katanin}}}, \bibinfo{journal}{Phys.\
434: Rev.\ B} \textbf{\bibinfo{volume}{61}}, \bibinfo{pages}{6757}
435: (\bibinfo{year}{2000}).
436:
437: \bibitem[{\citenamefont{{H. Rosner} et~al.}(1997)\citenamefont{{H. Rosner}, {H.
438: Eschrig}, {R. Hayn}, {S.-L. Drechsler}, and {J. Malek}}}]{helge}
439: \bibinfo{author}{\bibnamefont{{H. Rosner}}}, \bibinfo{author}{\bibnamefont{{H.
440: Eschrig}}}, \bibinfo{author}{\bibnamefont{{R. Hayn}}},
441: \bibinfo{author}{\bibnamefont{{S.-L. Drechsler}}}, \bibnamefont{and}
442: \bibinfo{author}{\bibnamefont{{J. Malek}}}, \bibinfo{journal}{Phys. Rev. B}
443: \textbf{\bibinfo{volume}{56}}, \bibinfo{pages}{3412} (\bibinfo{year}{1997}).
444:
445: \bibitem[{\citenamefont{{S.\ Takagi} et~al.}(1996)\citenamefont{{S.\ Takagi},
446: {H.\ Deguchi}, {K.\ Takeda}, {M.\ Mito}, and {M.\ Takahashi}}}]{takagi96}
447: \bibinfo{author}{\bibnamefont{{S.\ Takagi}}},
448: \bibinfo{author}{\bibnamefont{{H.\ Deguchi}}},
449: \bibinfo{author}{\bibnamefont{{K.\ Takeda}}},
450: \bibinfo{author}{\bibnamefont{{M.\ Mito}}}, \bibnamefont{and}
451: \bibinfo{author}{\bibnamefont{{M.\ Takahashi}}}, \bibinfo{journal}{J.\ Phys.\
452: Soc.\ Jpn.} \textbf{\bibinfo{volume}{65}}, \bibinfo{pages}{1934}
453: (\bibinfo{year}{1996}).
454:
455: \bibitem[{rem()}]{remarktakagi}
456: \bibinfo{note}{However, for this compound exhibiting the a steep decrease of
457: $\chi(T)$ at low $T$ the ratio $T_N/J_1$ amounts about 2.3$^.$10$^{-2}$,
458: only. In addition, the reported fit of the infinite chain Bethe-ansatz
459: solution without any subtraction of Curie-contributions shows significant
460: deviations near the maximum of $\chi (T)$ and at high $T$.}
461:
462: \bibitem[{\citenamefont{{I.E. Dzyaloshinsky}}(1957)}]{dzy}
463: \bibinfo{author}{\bibnamefont{{I.E. Dzyaloshinsky}}}, \bibinfo{journal}{JETP}
464: \textbf{\bibinfo{volume}{5}}, \bibinfo{pages}{1259} (\bibinfo{year}{1957}).
465:
466: \bibitem[{\citenamefont{{K. Koepernik} and {H. Eschrig}}(1999)}]{fplo}
467: \bibinfo{author}{\bibnamefont{{K. Koepernik}}} \bibnamefont{and}
468: \bibinfo{author}{\bibnamefont{{H. Eschrig}}}, \bibinfo{journal}{Phys. Rev.
469: B.} \textbf{\bibinfo{volume}{59}}, \bibinfo{pages}{1743}
470: (\bibinfo{year}{1999}).
471:
472: \bibitem[{\citenamefont{{A.A. Belik} et~al.}(2004)\citenamefont{{A.A. Belik},
473: {M. Azuma}, and {M. Takano}}}]{belik1}
474: \bibinfo{author}{\bibnamefont{{A.A. Belik}}}, \bibinfo{author}{\bibnamefont{{M.
475: Azuma}}}, \bibnamefont{and} \bibinfo{author}{\bibnamefont{{M. Takano}}},
476: \bibinfo{journal}{J. Solid State Chem.} \textbf{\bibinfo{volume}{177}},
477: \bibinfo{pages}{883} (\bibinfo{year}{2004}).
478:
479: \bibitem[{\citenamefont{{A. A. Belik} et~al.}(2002)\citenamefont{{A. A. Belik},
480: {A. P. Malakho}, {B. I. Lazoryak}, and {S. S. Khasanov}}}]{AAB+02}
481: \bibinfo{author}{\bibnamefont{{A. A. Belik}}},
482: \bibinfo{author}{\bibnamefont{{A. P. Malakho}}},
483: \bibinfo{author}{\bibnamefont{{B. I. Lazoryak}}}, \bibnamefont{and}
484: \bibinfo{author}{\bibnamefont{{S. S. Khasanov}}}, \bibinfo{journal}{J. Solid
485: State Chem.} \textbf{\bibinfo{volume}{163}}, \bibinfo{pages}{121}
486: (\bibinfo{year}{2002}).
487:
488: \bibitem[{\citenamefont{{J. P. Perdew} and {Y. Wang}}(1992)}]{JPP92}
489: \bibinfo{author}{\bibnamefont{{J. P. Perdew}}} \bibnamefont{and}
490: \bibinfo{author}{\bibnamefont{{Y. Wang}}}, \bibinfo{journal}{Phys. Rev. B.}
491: \textbf{\bibinfo{volume}{45}}, \bibinfo{pages}{13244} (\bibinfo{year}{1992}).
492:
493: \bibitem[{\citenamefont{{V. I. Anisimov} et~al.}(1993)\citenamefont{{V. I.
494: Anisimov}, {I. V. Solovyev}, {M. A. Korotin}, {M. T. Czyzyk}, and {G. A.
495: Sawatzky}}}]{via93}
496: \bibinfo{author}{\bibnamefont{{V. I. Anisimov}}},
497: \bibinfo{author}{\bibnamefont{{I. V. Solovyev}}},
498: \bibinfo{author}{\bibnamefont{{M. A. Korotin}}},
499: \bibinfo{author}{\bibnamefont{{M. T. Czyzyk}}}, \bibnamefont{and}
500: \bibinfo{author}{\bibnamefont{{G. A. Sawatzky}}}, \bibinfo{journal}{Phys.
501: Rev. B} \textbf{\bibinfo{volume}{48}}, \bibinfo{pages}{16929}
502: (\bibinfo{year}{1993}).
503:
504: \bibitem[{\citenamefont{{A. Furusaki} and {T. Hikihara}}(2004)}]{Furu}
505: \bibinfo{author}{\bibnamefont{{A. Furusaki}}} \bibnamefont{and}
506: \bibinfo{author}{\bibnamefont{{T. Hikihara}}}, \bibinfo{journal}{Phys. Rev.
507: B.} \textbf{\bibinfo{volume}{69}}, \bibinfo{pages}{094429}
508: (\bibinfo{year}{2004}).
509:
510: \bibitem[{\citenamefont{{A. A. Zvyagin} and {A. V. Makarova}}(2004)}]{Zvy}
511: \bibinfo{author}{\bibnamefont{{A. A. Zvyagin}}} \bibnamefont{and}
512: \bibinfo{author}{\bibnamefont{{A. V. Makarova}}}, \bibinfo{journal}{Phys.
513: Rev. B.} \textbf{\bibinfo{volume}{69}}, \bibinfo{pages}{214430}
514: (\bibinfo{year}{2004}).
515:
516: \bibitem[{\citenamefont{{D.C.\ Johnston} et~al.}(2000)\citenamefont{{D.C.\
517: Johnston}, {R.K.\ Kremer}, {M. Troyer}, , {X.\ Wang}, {A.\ Kl\"umper}, {A.L.\
518: Bud'ko}, {A.F.\ Panchula}, and {P.\ Canfield}}}]{johnston}
519: \bibinfo{author}{\bibnamefont{{D.C.\ Johnston}}},
520: \bibinfo{author}{\bibnamefont{{R.K.\ Kremer}}},
521: \bibinfo{author}{\bibnamefont{{M. Troyer}}}, ,
522: \bibinfo{author}{\bibnamefont{{X.\ Wang}}},
523: \bibinfo{author}{\bibnamefont{{A.\ Kl\"umper}}},
524: \bibinfo{author}{\bibnamefont{{A.L.\ Bud'ko}}},
525: \bibinfo{author}{\bibnamefont{{A.F.\ Panchula}}}, \bibnamefont{and}
526: \bibinfo{author}{\bibnamefont{{P.\ Canfield}}}, \bibinfo{journal}{Phys.\
527: Rev.\ B} \textbf{\bibinfo{volume}{61}}, \bibinfo{pages}{9558}
528: (\bibinfo{year}{2000}).
529:
530: \bibitem[{\citenamefont{{R. Nath} et~al.}(2005)\citenamefont{{R. Nath}, {A. V.
531: Mahajan}, {N. Buttgen}, {C. Kegler}, and {A. Loidl}}}]{loidl}
532: \bibinfo{author}{\bibnamefont{{R. Nath}}}, \bibinfo{author}{\bibnamefont{{A. V.
533: Mahajan}}}, \bibinfo{author}{\bibnamefont{{N. Buttgen}}},
534: \bibinfo{author}{\bibnamefont{{C. Kegler}}}, \bibnamefont{and}
535: \bibinfo{author}{\bibnamefont{{A. Loidl}}}, \bibinfo{journal}{Phys. Rev. B}
536: \textbf{\bibinfo{volume}{71}}, \bibinfo{pages}{174436}
537: (\bibinfo{year}{2005}).
538:
539: \bibitem[{\citenamefont{{K. M. Kojima} et~al.}(1997)\citenamefont{{K. M.
540: Kojima}, {Y. Fudamoto}, {M. Larkin}, {G. M. Luke}, {J. Merrin}, {B. Nachumi},
541: {Y. J. Uemura}, {N. Motoyama}, {H. Eisaki}, {S. Uchida} et~al.}}]{KMK+97}
542: \bibinfo{author}{\bibnamefont{{K. M. Kojima}}},
543: \bibinfo{author}{\bibnamefont{{Y. Fudamoto}}},
544: \bibinfo{author}{\bibnamefont{{M. Larkin}}},
545: \bibinfo{author}{\bibnamefont{{G. M. Luke}}},
546: \bibinfo{author}{\bibnamefont{{J. Merrin}}},
547: \bibinfo{author}{\bibnamefont{{B. Nachumi}}},
548: \bibinfo{author}{\bibnamefont{{Y. J. Uemura}}},
549: \bibinfo{author}{\bibnamefont{{N. Motoyama}}},
550: \bibinfo{author}{\bibnamefont{{H. Eisaki}}},
551: \bibinfo{author}{\bibnamefont{{S. Uchida}}}, \bibnamefont{et~al.},
552: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{78}},
553: \bibinfo{pages}{1787} (\bibinfo{year}{1997}).
554:
555: \bibitem[{\citenamefont{{Y. Mizuno} et~al.}(1998)\citenamefont{{Y. Mizuno}, {T.
556: Tohyama}, {S. Maekawa}, {T. Osafune}, {N. Motoyama}, {H. Eisaki}, and {S.
557: Uchida}}}]{YMTT+98}
558: \bibinfo{author}{\bibnamefont{{Y. Mizuno}}}, \bibinfo{author}{\bibnamefont{{T.
559: Tohyama}}}, \bibinfo{author}{\bibnamefont{{S. Maekawa}}},
560: \bibinfo{author}{\bibnamefont{{T. Osafune}}},
561: \bibinfo{author}{\bibnamefont{{N. Motoyama}}},
562: \bibinfo{author}{\bibnamefont{{H. Eisaki}}}, \bibnamefont{and}
563: \bibinfo{author}{\bibnamefont{{S. Uchida}}}, \bibinfo{journal}{Phys. Rev. B}
564: \textbf{\bibinfo{volume}{57}}, \bibinfo{pages}{5326} (\bibinfo{year}{1998}).
565:
566: \bibitem[{\citenamefont{{R. Neudert} et~al.}(1999)\citenamefont{{R. Neudert},
567: {H. Rosner}, {S.-L. Drechsler}, {M. Kielwein}, {M. Sing}, {Z. Hu}, {M.
568: Knupfer}, {M.S. Golden}, {J. Fink}, {N. Nucker} et~al.}}]{Rosner}
569: \bibinfo{author}{\bibnamefont{{R. Neudert}}}, \bibinfo{author}{\bibnamefont{{H.
570: Rosner}}}, \bibinfo{author}{\bibnamefont{{S.-L. Drechsler}}},
571: \bibinfo{author}{\bibnamefont{{M. Kielwein}}},
572: \bibinfo{author}{\bibnamefont{{M. Sing}}}, \bibinfo{author}{\bibnamefont{{Z.
573: Hu}}}, \bibinfo{author}{\bibnamefont{{M. Knupfer}}},
574: \bibinfo{author}{\bibnamefont{{M.S. Golden}}},
575: \bibinfo{author}{\bibnamefont{{J. Fink}}}, \bibinfo{author}{\bibnamefont{{N.
576: Nucker}}}, \bibnamefont{et~al.}, \bibinfo{journal}{Phys. Rev. B}
577: \textbf{\bibinfo{volume}{60}}, \bibinfo{pages}{13413} (\bibinfo{year}{1999}).
578:
579: \bibitem[{\citenamefont{{R. Weht} and {W.E. Pickett}}(1998)}]{Warren}
580: \bibinfo{author}{\bibnamefont{{R. Weht}}} \bibnamefont{and}
581: \bibinfo{author}{\bibnamefont{{W.E. Pickett}}}, \bibinfo{journal}{Phys. Rev.
582: Lett.} \textbf{\bibinfo{volume}{81}}, \bibinfo{pages}{2502}
583: (\bibinfo{year}{1998}).
584:
585: \end{thebibliography}
586: \end{document}
587: