1: \documentclass[prb,twocolumn,floatfix,showpacs,superscriptaddress]{revtex4}
2:
3: % To create clean pdf:
4: % dvips -Ppdf -o lco8_pdf.ps lco8
5: % Then convert from postscript as usual
6:
7: \usepackage{graphics}
8:
9: \begin{document}
10:
11: \title{Oxygen Moment Formation and Canting in Li$_{2}$CuO$_{2}$}
12: \author{E.M.L Chung}
13: \affiliation{Department of Physics, University of Warwick, Coventry, CV4 7AL, UK}
14:
15: \author{G.J. McIntyre}
16: \affiliation{Institut Laue Langevin, BP 156, 38042, Grenoble Cedex 9, France}
17:
18: \author{D. M$^{c}$K. Paul}
19: \affiliation{Department of Physics, University of Warwick, Coventry, CV4 7AL, UK}
20:
21: \author{G. Balakrishnan}
22: \affiliation{Department of Physics, University of Warwick, Coventry, CV4 7AL, UK}
23:
24: \author{M.R. Lees}
25: \affiliation{Department of Physics, University of Warwick, Coventry, CV4 7AL, UK}
26: \date{\today}
27:
28: \begin{abstract}
29: The possibilities of oxygen moment formation and canting in the
30: quasi-1\emph{D} cuprate $\mathrm{Li_{2}CuO_{2}}$ are investigated
31: using single crystal neutron diffraction at 2 K. The observed
32: magnetic intensities could not be explained without the inclusion of a
33: large ordered oxygen moment of 0.11(1) $\mu_{B}$.
34: Least-squares refinement of the magnetic structure
35: of $\mathrm{Li_{2}CuO_{2}}$ in combination with a spin-density
36: Patterson analysis shows that the magnetization densities of the Cu
37: and O atoms are highly aspherical, forming quasi-1\emph{D} ribbons of
38: localised Cu and O moments. Magnetic structure refinements and
39: low-field magnetization measurements both suggest that the magnetic
40: structure of $\mathrm{Li_{2}CuO_{2}}$ at 2 K may be canted. A possible model
41: for the canted configuration is proposed.
42: \vspace{8mm}
43: \end{abstract}
44:
45: \pacs{61.12.-q, 75.25.+z, 75.30.-m, 75.40.Cx}
46:
47: \maketitle
48:
49: \section{Introduction}
50: Quasi-one-dimensional cuprates such as $\mathrm{Li_{2}CuO_{2}}$
51: exhibit a wide range of
52: interesting electronic and magnetic properties.
53: As $\mathrm{Li_{2}CuO_{2}}$ possesses
54: 1D chains of S=$\frac{1}{2}$ copper atoms it serves as a
55: useful model system for studies of low dimensional behavior and
56: quantum fluctuations in the spin $\frac{1}{2}$ quantum chain.
57: Also embedded within the chemical structure of $\mathrm{Li_{2}CuO_{2}}$
58: are sets of parallel Cu-O$_{2}$ ribbons (figure \ref{lcouc}), similar to
59: those found in the spin-Peierls compound $\mathrm{CuGeO_{3}}$ and
60: chain-ladder system $\mathrm{La_{14-x}Ca_{X}Cu_{24}O_{41}}$.
61: $\mathrm{Li_{2}CuO_{2}}$ therefore also
62: frequently serves as a model system for investigation of the
63: electronic structure and correlations of the Cu-O$_2$ ribbon
64: \cite{weht1,neudert1,mizuno1,delley1,degraaf1}.
65:
66: \begin{figure}[t]
67: \begin{center}
68: \resizebox{7.5cm}{9cm}{\includegraphics{fig1.eps}}
69: \caption{Chemical structure of $\mathrm{Li_{2}CuO_{2}}$
70: showing the CuO$_{2}$ ribbons created along $b$ by an edge-sharing arrangement of
71: CuO$_{4}$ plaquettes.}
72: \label{lcouc}
73: \end{center}
74: \end{figure}
75:
76: The Cu-O networks of $\mathrm{Li_{2}CuO_{2}}$, $\mathrm{CuGeO_{3}}$,
77: high-$\mathrm{T_{C}}$ materials and many other cuprates can be
78: thought of as being constructed from the same basic building block;
79: a $\mathrm{CuO_{4}}$ plaquette. In high-$\mathrm{T_{C}}$ compounds,
80: the plaquettes are corner sharing forming 2D planes, whilst in
81: $\mathrm{Li_{2}CuO_{2}}$ and $\mathrm{CuGeO_{3}}$ they are edge-sharing,
82: forming Cu-O ribbons. The magnetic properties of cuprates are
83: known to be crucially dependent upon the Cu-O-Cu bond angle linking
84: nearest neighboring Cu atoms. When this angle approaches 90$^{o}$
85: the strength of the superexchange
86: interaction diminishes and a change in magnetic ordering
87: from antiferromagnetic (AFM) to ferromagnetic (FM) occurs
88: \cite{goodenough1,kanamori1,anderson1}.
89: In the context of other cuprates
90: $\mathrm{Li_{2}CuO_{2}}$ is unusually placed since it possesses a
91: Cu-O-Cu exchange pathway of 94$^{o}$. This angle is close to the
92: crossover region between AFM and FM ordering and frustration effects
93: resulting from competing interactions are anticipated.
94:
95: Local spin-density approximation (LSDA) \cite{weht1,delley1} and
96: quantum chemical \cite{degraaf1} calculations have recently suggested
97: that the oxygen
98: ions may play a more novel and fundamental role in the magnetism of
99: $\mathrm{Li_{2}CuO_{2}}$ than previously thought. The LSDA
100: calculations suggest an ordered magnetic moment of 0.92 $\mu_{B}$ per
101: formula unit, 40\% of which is predicted to be associated with the
102: `non-magnetic' oxygen sites \cite{weht1}. Although covalency effects
103: (e.g. in K$_{2}$IrCl$_{6}$\cite{lynn1}) are not unknown, a moment as
104: large as 0.2 $\mu_{B}$ associated with each of the oxygen sites would
105: be unprecedented. A more conservative estimate of the oxygen moment is
106: provided by quantum chemical techniques, however, these values vary
107: dramatically (from 0.03 to 0.17 $\mu_{B}$) depending upon which
108: approximations have been employed to treat the electronic
109: correlations.
110:
111: Previous studies on polycrystalline $\mathrm{Li_{2}CuO_{2}}$ include
112: the determination of the crystal structure via X-rays \cite{hoppe1},
113: and the observation of an antiferromagnetic ordering temperature
114: of $\mathrm{T_{N}}\simeq$ 9 K
115: \cite{ohta1,shreedhar1,okuda1}. Early neutron diffraction
116: investigations performed on powder samples at 1.5 K suggested a
117: collinear arrangement of moments coupled ferromagnetically within the
118: $a$-$b$ plane, and antiferromagnetically in the $c$-direction
119: \cite{hoffmann1,sapina1}. A magnetic moment of 0.96(4) $\mu_{B}$ was
120: estimated and attributed solely to the magnetic $\mathrm{Cu^{2+}}$
121: ions \cite{sapina1}. Electron paramagnetic resonance and
122: antiferromagnetic resonance experiments showed that the easy axis for
123: the spins lies along the $a$-direction \cite{ohta1}.
124:
125: More recently, single-crystal muon spectroscopy and
126: powder susceptibility measurements have indicated that in weak fields
127: the magnetic structure of $\mathrm{Li_{2}CuO_{2}}$ below 2.6 K may be
128: canted away from the $a$-axis\cite{staub1,ortega1}. The temperature dependence of the internal fields probed by $\mu$SR also suggests that there may be magnetic moments on both the Cu and the O ions\cite{staub1}.
129:
130: An excellent test of the reported magnetic structure, and of the validity
131: of $\mathrm{Li_{2}CuO_{2}}$ LSDA predictions, can be found through
132: measurement of the magnetic moment associated with each atom of
133: $\mathrm{Li_{2}CuO_{2}}$ using single-crystal neutron diffraction.
134: By this method it is possible to survey efficiently
135: several hundred magnetic and nuclear reflections over a large
136: volume of reciprocal space, enabling the magnetic structure of
137: $\mathrm{Li_{2}CuO_{2}}$ to be determined with good statistical accuracy.
138:
139: In this paper, the growth and characterization of $\mathrm{Li_{2}CuO_{2}}$
140: single crystals, and experimental details of our neutron-scattering measurements
141: are first described. In section III the $\mathrm{Li_{2}CuO_{2}}$ magnetic form
142: factor and spin-density Patterson functions are introduced.
143: Patterson analysis of our observed $\mathrm{Li_{2}CuO_{2}}$
144: magnetic diffraction pattern reveals the existence
145: of a large moment associated with the
146: oxygen sites. Section IV contains the results of least-squares refinement of
147: the magnetic structure of $\mathrm{Li_{2}CuO_{2}}$ with respect to various
148: magnetic models. Despite complications
149: introduced by a possible slight
150: canting of the Cu and O moments, all models consistently
151: estimate that the magnitude of the oxygen moment lies
152: between 0.10 $\mu_{B}$ and 0.12 $\mu_{B}$.
153: Lastly, our magnetic model of $\mathrm{Li_{2}CuO_{2}}$ is refined further to
154: include the possibility of canting. Section V contains an additional
155: investigation of the canted phase, performed using heat capacity and
156: single-crystal magnetization measurements.
157:
158: \section{Experimental Details}
159: Single crystals of $\mathrm{Li_{2}CuO_{2}}$ were grown from phase-pure
160: polycrystalline $\mathrm{Li_{2}CuO_{2}}$\cite{patat1}
161: using a four-ellipsoid floating-zone furnace at a rate of 3-6 mm/h
162: in 1.3 atm of flowing oxygen.
163: The resulting crystalline boules were
164: faceted and possessed a pronounced
165: tendency to cleave along parallel (101) planes.
166:
167: Room-temperature powder X-ray
168: diffraction patterns obtained for finely-crushed portions of our
169: $\mathrm{Li_{2}CuO_{2}}$ single-crystals showed that the samples were
170: of high purity. All peaks were indexed in the orthorhombic
171: space group I\emph{mmm} and
172: our single crystals exhibited all of the expected bulk magnetic
173: properties (susceptibility and heat capacity) associated with
174: the onset of three-dimensional antiferromagnetic order at
175: $T_{N}=9 \pm{0.1}$ K (figure \ref{graphs1}, discussed in section V).
176:
177: Neutron diffraction measurements were performed at the high-flux reactor
178: at the Institut Laue Langevin (ILL), Grenoble, using the D10
179: diffractometer in its four-circle configuration with an offset
180: Eulerian cradle,
181: He-flow cryostat and position-sensitive $80 \times 80$ mm$^{2}$ area
182: detector. The 40 mg single crystal platelet of
183: $\mathrm{Li_{2}CuO_{2}}$ had (101) flat faces of area $3\times4$
184: mm$^{2}$ spaced 0.3 mm apart.
185: The positions and intensities of all accessible nuclear and
186: magnetic reflections with $k \geq 0$
187: were scanned at 2 K using a constant neutron wavelength of 1.257(5) \AA
188: \ to give up to four symmetry equivalents for each unique reflection.
189:
190: Background subtraction and peak integration were performed using the
191: ``\emph{minimum in $\sigma(I)/I$ method}''\cite{wilkinson1} where
192: statistically optimum results are obtained in the integration of
193: weaker magnetic reflections by using three-dimensional profiles
194: estimated from stronger neighboring reflections. The data were
195: corrected for absorption\cite{coppens1} using an calculated absorption
196: coefficient of 0.2114 mm$^{-1}$. The transmission factors ranged from
197: 41.7\% to 87.7\% depending on orientation. Examination of the
198: intensities of the stronger nuclear reflections suggested that
199: measurements of the nuclear peaks had suffered significantly from
200: extinction. The extinction could be accounted for reasonably well in
201: the structure refinements using a Becker and Coppens type I Lorentzian
202: model\cite{becker1}, with a crystal mosaicity of 0.03(2)$^{o}$.
203:
204:
205: \section{The magnetic form factor}
206:
207: \begin{figure}[t]
208: \begin{center}
209: \rotatebox{270}{\resizebox{8cm}{8cm}{\includegraphics{fig2.eps}}}
210: \caption{The magnetic form-factor of $\mathrm{Li_{2}CuO_{2}}$ (markers), calculated from the observed data by assuming a model of $a$-axis Cu moments. The lines
211: depict the $<j_{0}>$, $<j_{2}>$ and $<j_{4}>$ radial integrals
212: for the free Cu$^{2+}$ ion \cite{itc2} (scaled).}
213: \label{ff}
214: \end{center}
215: \end{figure}
216:
217: Below $\mathrm{T_{N}}$, magnetic peaks develop at $h$+$k$+$l$= 2$n$+1
218: (indexed on the basis of the nuclear unit-cell), due to the
219: antiferromagnetic ordering of Cu moments positioned at (0,0,0) and
220: ($\frac{1}{2}$,$\frac{1}{2}$,$\frac{1}{2}$). Assuming that all of the
221: magnetic scattering arises from the Cu atoms the magnetic intensity
222: can be described by,
223: %
224: \begin{equation}
225: I_{M}(\textbf{Q}) \propto T^{2} |G(\textbf{Q})f_{m}(\textbf{Q})(1 - e^{-2 \pi i (\frac{h}{2}+\frac{k}{2}+\frac{l}{2})})|^{2}
226: \label{I1}
227: \end{equation}
228: %
229: `$T$' is a temperature factor, whose variation with $Q$ at 2 K can be ignored,
230: and $f_{m}(\textbf{Q})$ is the atomic magnetic form factor of the Cu site.
231: Assuming that the moments are oriented along the $a$-axis the geometrical
232: orientation factor, $G(\textbf{Q})$, for this orthorhombic cell is given by,
233: %
234: \begin{equation}
235: G(\mathbf{Q}) = \frac{\sqrt{(\frac{k}{b})^{2}+(\frac{l}{c})^{2}}}{\sqrt{(\frac{h}{a})^{2}+(\frac{k}{b})^{2}+(\frac{l}{c})^{2}}}
236: \label{Gfact}
237: \end{equation}
238: %
239: thus, apart from some constant scale factors the magnetic form factor can be extracted using,
240: %
241: \begin{equation}
242: f_{m}(\textbf{Q}) \propto \sqrt{I_{M}(\textbf{Q})/|G(\textbf{Q})|^{2}}
243: \label{ffact}
244: \end{equation}
245:
246: The magnetic form factor of $\mathrm{Li_{2}CuO_{2}}$, calculated from the
247: measured intensities
248: by assuming that magnetic scattering arises exclusively from
249: $a$-axis Cu moments is plotted in
250: figure \ref{ff}. It can be seen that our measurements deviate radically
251: from the $<j_{0}>$ radial integral for the free Cu$^{2+}$ ion \cite{itc2}.
252: This is to
253: be expected because of the effects of crystalline anisotropy
254: on the magnetization density close to the Cu sites.
255: A further departure from spherical symmetry may also arise due to the
256: presence of a magnetic moment elsewhere in the unit cell, such as on
257: the oxygen sites.
258:
259: It should be noted that if a large oxygen moment exists,
260: or the spins in
261: $\mathrm{Li_{2}CuO_{2}}$ are
262: significantly canted, the expression for $G(\textbf{Q})$ (provided in
263: equation 2) will no longer represent a good approximation for the
264: geometrical orientation factor. In this situation, the values estimated for
265: $f_{m}(\textbf{Q})$ (using equation 3) could differ considerably from
266: those plotted.
267:
268:
269: \subsection{The Spin-density Patterson Function}
270:
271: In principle, the three-dimensional real-space magnetization density
272: of the crystal may be obtained directly through Fourier transformation
273: (F.T.) of
274: the complete magnetic structure factor, \textbf{F}$_{M}({\mathbf{Q}})$.
275: The full magnetic structure factor is a vector quantity that carries
276: information about the magnitude \emph{and} the direction of the
277: spin-density at every point in the unit cell.
278:
279: \begin{equation}
280: \mathbf{\rho}_{m}({\mathbf{r}})= {\mathrm{F.T.}}[\mathbf{F}_{M}(\mathbf{Q})]
281: \label{p1}
282: \end{equation}
283:
284: Unfortunately, analysis
285: of our neutron diffraction data by this method
286: is hindered by the lack of phase information. Since unambiguous knowledge
287: of the phase of each of the reflections was unavailable a magnetic
288: Patterson approach was adopted\cite{wilkinson2,wilkinson3,wilkinson4}.
289: Unlike maximum entropy methods, Patterson analysis requires no phase
290: information, or prior knowledge of the positions or the directions of
291: the moments. The spin-density Patterson approach treats the magnetic
292: structure as being completely unknown.
293:
294: The Patterson function was first introduced in 1934,
295: and has become an invaluable tool for the determination of
296: unknown crystal structures principally via X-rays.
297: The non-magnetic Patterson function represents the autocorrelation
298: function of the electron density in the unit cell, and is defined
299: as the Fourier transform of the squared nuclear structure factor,
300: %
301: \begin{equation}
302: P_{N}({\mathbf{u}})= {\mathrm{F.T.}}[|F_{N}(\mathbf{Q})|^{2}]
303: \label{p2}
304: \end{equation}
305: %
306: \begin{equation}
307: P_{N}({\mathbf{u}})= \rho_{n}({\mathbf{r}}) \otimes \rho_{n}({-\mathbf{r}})
308: \label{p3}
309: \end{equation}
310:
311: The nuclear Patterson function, $P_{N}(\mathbf{u})$, provides a map of the
312: vectors between \emph{pairs} of atoms in the unit cell. (i.e. in a unit cell
313: containing \emph{n} atoms, each atom will form a pair with all atoms,
314: including itself, producing $n^{2}$ Patterson peaks).
315:
316: The relationship between the magnetic Patterson function
317: and the unpaired spin-density is not as trivial as the
318: situation described for the nuclear Patterson function.
319: The dipolar nature of the electron-neutron interaction means
320: that the intensity of magnetic scattering is governed by the
321: relative orientation of the spins and the scattering vector
322: (\textbf{Q}). In terms of \textbf{F}$_{M}$(\textbf{Q}) neutrons
323: are scattered by the sample with an intensity given by\cite{halpern1},
324: %
325: \begin{equation}
326: I_{M}({\mathbf{Q}}) \propto |{\mathbf{F}}_{M}({\mathbf{Q}})|^{2}-|{\mathbf{\hat{Q}}}.{\mathbf{F}}_{M}(\mathbf{Q})|^{2}
327: \label{haleqn}
328: \end{equation}
329:
330: Where $\mathbf{\hat{Q}}$ represents a unit vector in the direction of
331: \textbf{Q}. Calculation of the spin-density Patterson map by Fourier
332: transformation of the measured magnetic intensities (equation
333: \ref{magpateqn}) generates a plot of the pair-correlation function
334: between magnetic scatterers, equation \ref{reqn},
335: %
336: \begin{equation}
337: P_{M}(\mathbf{u}) \propto {\mathrm{F.T.}}[|{\mathbf{F}}_{M}({\mathbf{Q}})|^{2}] - {\mathrm{F.T.}}[|{\mathbf{\hat{Q}}}.{\mathbf{F}}_{M}(\mathbf{Q})|^{2}]
338: \label{magpateqn}
339: \end{equation}
340: %
341: \begin{eqnarray}
342: \label{reqn}
343: P_{M}({\mathbf{u}}) & = & \sum_{i=1}^{3} \sum_{j=1}^{3}(\rho_{i}({\mathbf{r}}) \otimes \rho_{j}(-{\mathbf{r}})) \\
344: & & - \sum_{i=1}^{3} \sum_{j=1}^{3} {\mathrm{F.T.}}[\hat{Q}_{i}\hat{Q}_{j}] \otimes (\rho_{i}({\mathbf{r}) \otimes \rho_{j}(-{\mathbf{r}}))} \nonumber
345: \end{eqnarray}
346:
347: The three spatial components of spin density of the $i$'th and $j$'th atoms
348: have been written as summations over $\rho_{i}$ and $\rho_{j}$. $\hat{Q}_{i}\hat{Q}_{j}$
349: describes the product of the magnitudes of the $i$'th and $j$'th components of
350: the unit vector $\mathbf{\hat{Q}}$. The symbol $\otimes$ denotes a convolution.
351:
352: The second term in equation \ref{reqn} probes the directions of the moments,
353: causing the Patterson peaks to become elongated along the direction parallel
354: to the vector bisector between spins. Full details of this theory, and references
355: relating the uses of the Patterson function in magnetism can be found in
356: the references\cite{wilkinson2,wilkinson3,wilkinson4}.
357:
358: \begin{figure}[t]
359: \begin{center}
360: \resizebox{7.2cm}{8.6cm}{\includegraphics{fig3.eps}}
361: \caption{(a) The magnetic Patterson section generated from the observed data
362: ($|F_{M}(hkl)|^{2}$), viewed as a ($\frac{1}{2}$ $b$ $c$) slice through the
363: Li$_{2}$CuO$_{2}$ unit cell. Note that significant magnetic intensity is
364: associated with inter-atomic vectors between copper and oxygen sites.
365: This confirms the presence of unpaired spins on the oxygen sites,
366: see equation \ref{mateqn}. Filled, dotted and dashed lines represent positive, zero and negative
367: contours respectively. Contours extend from -5 to 50 in 2.5 steps.
368: (b) The corresponding positions of the atoms in the Li$_{2}$CuO$_{2}$ unit cell. }
369: \label{pat1}
370: \end{center}
371: \end{figure}
372:
373: \subsection{Li$_{2}$CuO$_{2}$ Spin-density Patterson Analysis}
374:
375: For the purposes of Patterson analysis, the $\mathrm{Li_{2}CuO_{2}}$ magnetic data
376: set of 409 observations was reduced to a complete set of 111 averaged unique reflections.
377: Including the possibility of magnetic oxygen sites there are $n$=6 magnetic atoms per
378: chemical unit cell producing $n^{2}$ = 36 Patterson peaks. The heights of the 36 peaks
379: may be represented by a symmetric matrix $\mathbf{P}$. Each matrix element corresponds
380: to the scalar product of pairs of spins at sites $i$ and $j$.
381:
382: \begin{equation}
383: \mathbf{P}=\left(
384: \begin{array}{cccc}
385: \mathbf{S}_{Cu1}.\mathbf{S}_{Cu1} & \mathbf{S}_{Cu1}.\mathbf{S}_{Cu2} & \ldots & \mathbf{S}_{Cu1}.\mathbf{S}_{O4} \\
386: \mathbf{S}_{Cu2}.\mathbf{S}_{Cu1} & \mathbf{S}_{Cu2}.\mathbf{S}_{Cu2} & & \vdots \\
387: \vdots & & \ddots & \vdots \\
388: \mathbf{S}_{O4}.\mathbf{S}_{Cu1} & \ldots & \ldots & \mathbf{S}_{O4}.\mathbf{S}_{O4}
389: \end{array}
390: \label{mateqn}
391: \right)
392: \end{equation}
393:
394: In interpreting the spin-density Patterson function it should be remembered that the
395: positions of the peaks correspond to the interatomic vectors \emph{between} scatterers
396: rather than the positions of the scatterers themselves. A large origin peak containing
397: overlapping contributions from $n$ diagonal matrix elements, ($\mathbf{S}_{n}.\mathbf{S}_{n}$),
398: occurs due to the convolution of each magnetic site with itself. With a knowledge of
399: the chemical structure, it is possible to relate the stronger peaks, dominated by
400: $\mathbf{S}_{Cu}.\mathbf{S}_{Cu}$ terms, to interatomic vectors between the centrosymmetric
401: positions of the Cu sites. Since there are only two Cu atoms per unit cell
402: (related by body centering), any peaks in addition to those at (0,0,0) and
403: ($\frac{1}{2}$,$\frac{1}{2}$,$\frac{1}{2}$) must represent additional interatomic vectors
404: between Cu sites and other magnetic scatterers.
405:
406: In the observed Patterson map (figures \ref{pat1} and \ref{pat2}) additional peaks
407: are observed at positions that do not correspond to inter-atomic Cu vectors.
408: Comparison with the positions of the atoms in the unit cell shows that a considerable
409: number of the magnetic scatterers are closely associated with the crystallographic
410: positions of the oxygen atoms. (These peaks are dominated by $\mathbf{S}_{Cu}.\mathbf{S}_{O}$ terms).
411: Asymmetry in the magnetic Patterson peaks in the immediate vicinity of the O sites
412: (as viewed along the $a$-axis) is in good agreement with the exchange
413: potential ($V_{\uparrow}-V_{\downarrow}$) predicted for the
414: ferromagnetic Cu-O chain by LSDA calculations\cite{weht1}.
415:
416: A three-dimensional Patterson plot of the observed data is
417: provided in figure \ref{pat2}. A reversal in sign between CuO$_{4}$
418: units positioned at (0,0,0) and ($\frac{1}{2}$,$\frac{1}{2}$,$\frac{1}{2}$)
419: is observed. Elongation of the peaks along the $a$-direction is consistent
420: with the direction of the bisector of the spins being parallel to the
421: $a$-axis, which does not exclude the possibility of canting.
422:
423: \begin{figure}[t]
424: \begin{center}
425: \resizebox{6.5cm}{10.5cm}{\includegraphics{fig4.eps}}
426: \caption{Three-dimensional Patterson map superimposed upon the $\mathrm{Li_{2}CuO_{2}}$ unit cell.
427: Contours at (0,0,0) and ($\frac{1}{2}$,$\frac{1}{2}$,$\frac{1}{2}$)
428: are fixed at values of +45 and -45 respectively, (illustrating the reversal in sign between
429: spins at (0,0,0) and ($\frac{1}{2}$,$\frac{1}{2}$,$\frac{1}{2}$)).}
430: \label{pat2}
431: \end{center}
432: \end{figure}
433:
434:
435: \section{Least-squares Refinement}
436:
437: Although Patterson analysis has proved useful for locating the distribution of spin-density
438: in the $\mathrm{Li_{2}CuO_{2}}$ unit cell, an estimate has not so far been obtained for the
439: magnitudes and directions of the spins.
440: A quantitative estimation of the magnitude of the moments in
441: $\mathrm{Li_{2}CuO_{2}}$ can be
442: gained through least-squares refinement of the magnetic data set with respect to an
443: appropriate magnetic model (or models).
444:
445: Before refining the magnetic structure of $\mathrm{Li_{2}CuO_{2}}$ the nuclear structure was
446: verified using the CCSL program SFLSQ \cite{ccsl}. The nuclear data set consisted of 445
447: reflections ($h$+$k$+$l$=2$n$) measured at 2 K.
448: The atomic positions and thermal parameters resulting from
449: refinement of the data are provided in Table \ref{tabnuc}. Unit cell parameters at 2 K
450: were refined to be $a$=3.661(4), $b$=2.866(7), and $c$=9.397(8) \AA. The high value
451: of $\chi^{2}$=17.9 is likely to be due to extinction effects that have not been perfectly
452: corrected by the Becker-Coppens procedure.
453:
454: \begin{table}[h]
455: \caption{$\mathrm{Li_{2}CuO_{2}}$ atomic positions and thermal parameters
456: measured at 2 K, obtained through the refinement of 9 basic variables,
457: $\chi^{2}$=17.9, $R(F)$=2.15. ($R(F)=\sum{|F_{0}-F_{C}|}/\sum{|F_{0}|}$
458: and $\chi^{2}=\frac{1}{n-p} \sum{w(F_{O}^{2}-F_{C}^{2})}$, where $n$ is the number
459: of points,
460: $p$ the number of parameters and $w$ the weighting.) Thermal parameters,
461: (assumed to be isotropic for Li and Cu, and anisotropic for O,) correspond
462: to the expression $\exp[-({h^{2}\beta_{11}+k^{2}\beta_{22}+l^{2}\beta_{33}})]$.}
463: \begin{ruledtabular}
464: \begin{tabular}{lllll}
465: Atom & X & Y & Z & Thermal parameters\\
466: \hline
467: Li & 0 & 0 & 0.7132(2) & $\beta_{11}=\beta_{22}=\beta_{33}=0.326(20)$\\
468: \hline
469: Cu & 0 & 0 & 0 & $\beta_{11}=\beta_{22}=\beta_{33}=0.095(10)$\\
470: \hline
471: & & & & $\beta_{11}=0.279(14)$\\
472: O & 0.5 & 0 & 0.6421(5) & $\beta_{22}=0.161(16)$\\
473: & & & & $\beta_{33}=0.229(15)$\\
474: \end{tabular}
475: \end{ruledtabular}
476: \label{tabnuc}
477: \end{table}
478:
479: Refinement of the magnetic structure of $\mathrm{Li_{2}CuO_{2}}$ was performed using the
480: CCSL programs MAGLSQ and MMPLSQ. The atomic co-ordinates,
481: thermal displacement parameters and overall scale factor
482: were fixed to the values found from refinement against the
483: nuclear reflections (table \ref{tabnuc}). Calculated
484: magnetic form factors were assumed for the copper\cite{itc2} and
485: oxygen\cite{itc2,Oformfactor} sites. As preliminary
486: refinements showed that extinction corrections for the magnetic peaks
487: were negligible, the magnetic reflections could be grouped and averaged.
488:
489: For completeness, refinement of the magnetic structure of $\mathrm{Li_{2}CuO_{2}}$
490: was first carried out by assuming a model based upon previous reports -
491: an antiferromagnetic arrangement
492: of $a$-axis Cu moments \cite{hoffmann1,sapina1,ohta1}. Refinements based on this model
493: (with spherical form-factors) estimated an $a$-axis Cu moment of 0.938 $\pm$ 0.024 $\mu_{B}$.
494: This value is close to the 0.96 $\mu_{B}$ estimated from powder neutron diffraction\cite{sapina1}.
495: However, considering that there are no extinction effects associated with the magnetic
496: reflections, the model provides a poor fit to the data ($\chi^{2}$ = 45), see
497: figure \ref{pat3}(a). A number of model
498: structures containing canted copper moments were also investigated but none were found that
499: enhanced the refinement. Tilting of the Cu moments away from the $a$-axis in itself was
500: not sufficient to explain the observed magnetic intensities.
501:
502:
503: \begin{figure}[t]
504: \begin{center}
505: \resizebox{6.3cm}{13cm}{\includegraphics{fig5.eps}}
506: \caption{Magnetic Patterson sections taken through the Cu-O ribbon, generated
507: from least-squares refined model data. Figure \ref{pat3}(a) spherical $a$-axis Cu moments.
508: Fig. \ref{pat3}(b) spherical $a$-axis Cu and O moments. Fig. \ref{pat3}
509: (c) non-spherical $a$-axis Cu and O moments. For comparison, the observed
510: Patterson map is provided in fig.\ref{pat1}.}
511: \label{pat3}
512: \end{center}
513: \end{figure}
514:
515: A more satisfactory fit to the data ($\chi^{2}$= 31) could be
516: achieved through the inclusion of an ordered oxygen moment,
517: see figure \ref{pat3}(b). Comparison of the Patterson map generated
518: using this model with the observed
519: Patterson map in figure \ref{pat1}(a) shows that the magnetization
520: densities of the Cu and O sites are asymmetric in the $b$-$c$ plane.
521: This strongly indicates the importance of aspherical multipolar contributions.
522:
523: Asymmetry in the magnetization densities of the Cu and O sites was modeled
524: using linear combinations of spherical harmonics. The angular dependence of the
525: magnetization density of the centro-symmetric Cu sites was modeled using an
526: approximation of the form of Equation \ref{cuff}, where the radial integrals
527: $<j_{0}>$, $<j_{2}>$ and $<j_{4}>$ (as shown in figure 2) are now modified to possess an angular dependence expressed in terms of Bessel functions $Y[l,m]$. The non-centrosymmetric O angular dependence was modeled using an approximation in the form of equation
528: \ref{off}. The coefficients $a_{l,m}$ are fitted parameters determined during the refinement.
529:
530: \begin{equation}
531: \begin{array}{rl}
532: \rho_{Cu}= & a_{0,0}<j_{0}>+a_{2,0}<j_{2}>Y[2,0]\\
533: & +<j_{4}>(a_{4,0}Y[4,0]+a_{4,2^{+}}Y[4,2^{+}])\\
534: \end{array}
535: \label{cuff}
536: \end{equation}
537:
538: \begin{equation}
539: \begin{array}{rl}
540: \rho_{O}= & a_{0,0}<j_{0}>+a_{1,0}<j_{1}>Y[1,0]\\
541: &+ <j_{2}>a_{2,2}Y[2,2^{+}]\\
542: \end{array}
543: \label{off}
544: \end{equation}
545:
546: Sequential introduction of aspherical form factors into our model,
547: refined with MMPLSQ, was found to
548: improve the fit remarkably. The Patterson section along the Cu-O chain,
549: obtained using data calculated by assuming a model containing aspherical
550: Cu and O form-factors, is depicted in figure \ref{pat3}(c). Spherical
551: harmonics that did not contribute significantly were omitted from the
552: final refinement.
553:
554: A summary of the results of analysis of the magnetic dataset using a
555: variety of starting models is
556: presented in table \ref{tabmag}. Spherical harmonics introduced to model
557: the magnetization densities of the Cu and O sites are indicated by the
558: values of $Y$[$l$,$m$]. (For example, $Y$[0,0] indicates that the
559: magnetization density was assumed to be perfectly spherical.)
560: The magnitude of the O moment, refined to be approximately 0.11 $\mu_{B}$ in size,
561: remained stable throughout the refinements. These O moments are coupled to Cu
562: moments between 0.91 $\mu_{B}$ and 0.96 $\mu_{B}$ in size.
563:
564: \begin{table}[h]
565: \caption{$\mathrm{Li_{2}CuO_{2}}$ magnetic structure refinements performed on a
566: unique set of 111 reflections. Spherical harmonics introduced to model the
567: magnetization densities of the Cu and O sites are indicated by the values of
568: $Y[l,m]$, (where $l$ and $m$ denote electronic quantum numbers). $\theta$ is the
569: angle away from $c$*; $\phi$ is the angle to $a$* in the $a$-$b$ plane.}
570: \begin{ruledtabular}
571: \begin{tabular}{llll}
572: MODEL & Cu ($\mu_{B}$) & O ($\mu_{B}$) & $\chi^{2}$ \\
573: \hline
574: $a$-axis Cu & 0.938 & -- & 44.9\\
575: Cu: $Y$[0,0] & $\pm$0.024 & & \\
576: \hline
577: $a$-axis Cu \& O, & 0.911 & 0.105 & 30.7\\
578: Cu, O: $Y$[0,0] & $\pm$0.021 & $\pm$0.012 & \\
579: \hline
580: $a$-axis Cu \& O & & & \\
581: Cu: $Y$[0,0], $Y$[2,*] \& $Y$[4,*] & 0.960 & 0.100 & 15.7 \\
582: O: $Y$[0,0] & $\pm$0.017 & $\pm$0.008& \\
583: \hline
584: $a$-axis Cu \& O, & & & \\
585: Cu: $Y$[0,0], $Y$[2,*] \& $Y$[4,*]&0.963 &0.105 & 11.1\\
586: O: $Y$[0,0], $Y$[1,0] \& $Y$[2,2$^{+}$] & $\pm$ 0.014 & $\pm$0.007 & \\
587: \hline
588: Canted($\theta$, $\phi$): & & & \\
589: Cu(76$^{o}$,0$^{o}$), O(116$^{o}$,0$^{o}$), & & & \\
590: Cu: $Y$[0,0], $Y$[2,*] \& $Y$[4,*]& 0.914 & 0.116 & 8.75 \\
591: O: $Y$[0,0], $Y$[1,0] \& $Y$[2,2$^{+}$]& $\pm$0.006 & $\pm$0.004 & \\
592: \end{tabular}
593: \end{ruledtabular}
594: \label{tabmag}
595: \end{table}
596:
597:
598:
599:
600: \subsection{Canting}
601: Although asphericity can be seen to play an important role in describing
602: the magnetism of $\mathrm{Li_{2}CuO_{2}}$, the high value of $\chi^{2}$=11.1 indicates
603: that further information can be extracted from the data. Difficulties encountered during
604: refinement of the magnetic structure of $\mathrm{Li_{2}CuO_{2}}$
605: may be explained by mounting evidence suggesting that below T = 2.6 $\pm$
606: 0.2 K the moments are slightly canted\cite{staub1,ortega1}. Other than the transition temperature,
607: little is known about the nature of the canted phase, except that the canting angle is
608: likely to be in excess of 0.7$^{o}$ \cite{staub1}.
609:
610: \begin{figure}[t]
611: \begin{center}
612: \resizebox{7cm}{8cm}{\includegraphics{fig6.eps}}
613: \caption{The antiferromagnetically canted structure found to be most consistent with the observed data.}
614: \label{cant}
615: \end{center}
616: \end{figure}
617:
618: In the present study we consider the possibility of an
619: \emph{antiferromagnetic} canted arrangement of moments. (In an
620: antiferromagnetically canted arrangement the moments are tilted so
621: that the net magnetization in all directions remains zero, and no
622: magnetic intensity is contributed to the positions of the nuclear
623: peaks.) This restriction seems reasonable considering the absence of a
624: second transition in the temperature dependence of the (012) and (001) peaks,
625: figure \ref{graphs1}(c). To the best of our knowledge no evidence
626: exists in support of a ferromagnetically canted arrangement of moments
627: in zero-field. Canting of the moments towards the $c$-axis, with the
628: oxygen moments tilted in the opposite direction to those of the
629: copper, was found to provide the best overall improvement to our
630: refinements.
631:
632: \begin{figure}
633: \begin{center}
634: \resizebox{8 cm}{18.5cm}{\includegraphics{fig7.eps}}
635: \caption{(a) $\mathrm{Li_{2}CuO_{2}}$ single-crystal susceptibility
636: curves measured in \textbf{H}=1 kOe with \textbf{H} applied parallel
637: to the $a$, $b$ and $c$-axes. The crystal was warmed to room
638: temperature and then zero-field-cooled (\emph{ZFC}) between
639: measurements. Each curve contains data obtained during warming
640: (\emph{ZFC} data) then cooling of the sample in field (\emph{FC}
641: data). In 1 kOe no low temperature magnetic transition is
642: observed. (b) $\mathrm{Li_{2}CuO_{2}}$ zero-field heat capacity
643: data. Data in the temperature range of the possible transition to the
644: canted phase are shown in the inset. No anomaly in the heat capacity
645: in the region of the transition is observed. (c) Integrated intensities
646: of the (001) and (012) magnetic Bragg peaks measured on warming.}
647: \label{graphs1}
648: \end{center}
649: \end{figure}
650:
651: A diagram of our proposed model is provided in figure
652: \ref{cant}. Adjacent Cu and O moments are coupled ferromagnetically in
653: the $a$-direction with a small antiferromagnetically coupled $c$-axis
654: component. The magnitude and direction of the moments was varied until
655: the lowest value of $\chi^{2}$ was obtained. ($\chi^{2}$=8.75 with 111
656: averaged reflections and $\chi^{2}$=6.65 with the full set of 232
657: reflections.) The resulting model consisted of 0.91 $\mu_{B}$ Cu
658: moments canted in the $a$-$c$ plane at an angle of $\theta$=76$^{o}$
659: to the $c$-axis. Oxygen moments approximately 0.12 $\mu_{B}$ in size
660: are counter-canted in the $a$-$c$ plane at an angle of
661: $\theta$=116$^{o}$. Although this result suggests that the magnetic structure of $\mathrm{Li_{2}CuO_{2}}$ is canted, it should be remembered that a small reduction in $\chi^{2}$ might also be expected as a side-effect of increasing the number of fitting parameters.
662:
663: The ordered $a$-axis moment per formula unit was 1.06 $\mu_{B}$. This
664: is consistent with the 0.96(4) $\mu_{B}$ estimated from powder neutron
665: diffraction \cite{sapina1}, and the spin-only value of the free Cu ion
666: (1 $\mu_{B}$). (For most \emph{d}-electron transition metal oxides the orbital momentum is expected to be quenched.)
667:
668: There are also several published estimates for the effective moment per Cu atom obtained from susceptibility measurements. Large discrepancies between these estimates, which vary between 1 $\mu_{B}$ and 2.3 $\mu_{B}$\cite{okuda1,ortega1,boehm1}, are probably due to the combined effects of a preferred [101] orientation of polycrystalline grains, and inappropriate fitting of the data to a Curie-Weiss law. High quality single crystal $\mathrm{Li_{2}CuO_{2}}$ susceptibility measurements clearly show that the Curie-Weiss regime (T$>>J$) does not commence until temperatures well in excess of 300 K\cite{mizuno1}.
669:
670: \section{Magnetisation data}
671:
672: In order to investigate further the nature of the possible canting in
673: $\mathrm{Li_{2}CuO_{2}}$, single-crystal magnetization measurements
674: were obtained between 1.6 and 50 K using an Oxford Instruments vibrating
675: sample magnetometer (VSM). The moderately moisture sensitive
676: $\mathrm{Li_{2}CuO_{2}}$ crystals were polished to remove surface
677: contamination and then oriented using X-ray Laue photography.
678: By this method the orientation of the crystal
679: axes relative to the applied field could be controlled to within
680: $\mathrm{\pm 5^{o}}$.
681: The magnetic properties of three separate crystal specimens weighing
682: 38 mg, 154 mg, and 86 mg were investigated. The latter crystal was
683: deliberately exposed to moisture in order
684: to identify any magnetic impurities arising from surface decomposition; none were observed.
685:
686:
687: Susceptibility curves obtained along the $a$, $b$ and $c$-axes respectively,
688: in an applied field of $\mathbf{H}$=1 kOe, are presented in figure \ref{graphs1}(a).
689: These results are in excellent agreement with previous reports
690: \cite{mizuno1,boehm1}. A sharp drop in the susceptibility associated with
691: easy $a$-axis antiferromagnetic ordering is observed at T$_{N}$ = 9 $\pm$ 0.1 K.
692: Zero-field heat capacity data is presented in figure \ref{graphs1}(b). Special
693: care was taken to follow the heat capacity in the vicinity of the expected low
694: temperature transition, \ref{graphs1}(b) (inset). The integrated intensities
695: of the (012) and (001)magnetic Bragg peaks are shown in figure \ref{graphs1}(c). None of the measurements presented
696: in figure \ref{graphs1} show any visible anomalies in the low-temperature region between 2 and 4 K.
697:
698: \begin{figure}[t]
699: \begin{center}
700: \resizebox{7.5cm}{14cm}{\includegraphics{fig8.eps}}
701: \caption{(a) $\mathrm{Li_{2}CuO_{2}}$ susceptibility curves measured in
702: $\mathbf{H}$=20 Oe along the $a$, $b$ and $c$-axes. Between each measurement
703: the sample was warmed to at least 20 K and then zero-field cooled (\emph{ZFC}).
704: Data was then collected during warming in a field of 20 Oe (\emph{ZFC}), and then on cooling in a field of 20 Oe (\emph{FC}).
705: The low-field data has been scaled to coincide with the data measured in
706: $\mathbf{H}$=1 kOe (shown as lines), see fig. \ref{graphs1}(a). (b) \textbf{M}-\textbf{H} loops
707: obtained at 1.6 K, 3 K, 7.5 K and 12.6 K with \textbf{H} applied parallel to the $a$-axis.}
708: \label{graphs2}
709: \end{center}
710: \end{figure}
711:
712: VSM data obtained between 1.6$ < $T$ < $20 K in a weak field of $\mathbf{H}$ $\simeq$ 20 Oe
713: are presented in figure \ref{graphs2}(a). We find that the response of the
714: bulk crystal to a small applied magnetic field changes abruptly at
715: approximately 2.6 K, producing a small ferromagnetic component in the
716: direction of the field. Below 2.6 K a small ferromagnetic hysteresis
717: was also observed along each of the
718: measured directions, between -100$<\mathbf{H}<$100 Oe, figure \ref{graphs2}(b).
719:
720: Although it is possible that the rise in magnetisation could be due to the ordering of a ferromagnetic impurity, the low temperature transition was consistently observed in all three of our samples, and has been reported previously in powder susceptibility\cite{ortega1} and $\mu$SR data\cite{staub1}. Assuming that the observed transition is an intrinsic property of the crystal, we find that at $\sim$ 2.6 K the Cu moments, the O moments, (or both), become free to cant toward the applied magnetic field. The observation of ferromagnetic hysteresis when $\mathbf{H}$ is parallel to the $a$-axis (figure \ref{graphs2}(b)), indicates that the moments may not be perfectly aligned along $a$, even above 2.6 K or in zero field.
721:
722:
723:
724: \section{Discussion}
725:
726: To conclude, we have performed detailed measurements of the magnetic
727: form-factor of a single crystal of $\mathrm{Li_{2}CuO_{2}}$ at 2
728: K. Patterson analysis of a complete set of unique reflections derived
729: from the observed magnetic diffraction pattern reveals the presence of
730: an effective oxygen moment, which appears to be large and
731: well-localised. Least squares refinement of the magnetic structure in
732: conjunction with the spin-density Patterson analysis reveals that the
733: magnetization densities surrounding the Cu and O sites are strongly
734: aspherical. The magnitudes of the O and Cu moments remained
735: approximately stable throughout our refinements. This allows us to
736: estimate that the magnitude of the O moment lies between 0.10
737: $\mu_{B}$ and 0.12 $\mu_{B}$. These O moments are coupled to Cu
738: moments that are approximately 0.92 $\mu_{B}$ - 0.96 $\mu_{B}$ in
739: size. Despite the magnitude of the oxygen moment being substantially
740: lower than that estimated by LSDA calculations\cite{weht1}, our
741: observations are otherwise in good qualitative agreement with LSDA
742: predictions for the exchange potential of the FM chain. The size of
743: the oxygen moment computed using the quantum chemical `difference
744: dedicated configuration interaction' technique for
745: $\mathrm{Li_{2}CuO_{2}}$ is closer to that measured in this experiment
746: \cite{degraaf1}.
747:
748: Although large covalency effects have been observed previously in
749: structures such as $\mathrm{K_{2}IrCl_{6}}$\cite{lynn1}, reports of
750: moment formation in oxides are extremely rare. The oxygen moment in
751: $\mathrm{Li_{2}CuO_{2}}$ is one of the largest so far recorded, rivaled in
752: size only by the 0.1 $\mu_{B}$ moment suggested to reside on the
753: oxygen sites in the ferromagnetic manganite
754: $\mathrm{La_{0.8}Sr_{0.2}MnO_{3}}$\cite{pierre1}. (In the
755: polarized neutron study of the manganite an oxygen moment was identified, but attempts
756: to refine its magnitude from the measured structure factors proved
757: unsuccessful.) In other examples of oxygen moment formation the
758: transfer of moment from the magnetic ion to the oxygen site has been
759: shown to be much smaller. For example, the only confirmed oxygen
760: moment in a cuprate system is of 0.02 $\mu_{B}$ in the chain-ladder
761: compound $\mathrm{La_{14-x}Ca_{x}Cu_{24}O_{41}}$\cite{matsuda1}.
762:
763: Besides the discovery of the O moment in $\mathrm{Li_{2}CuO_{2}}$, our
764: observations also seem consistent with the notion of canting. A ferromagnetic-like rise in the magnetization of our single-crystal samples was verified by low-field
765: ($\mathbf{H} =$ 20 Oe) magnetization measurements to occur at T = 2.6
766: $\pm$ 0.2 K. However, since many possible configurations of canted Cu and O
767: moments may be compatible within error with the observed magnetic diffraction
768: pattern, it is difficult to reach any reliable conclusions about the
769: nature of the canting. Having restricted our refinements to
770: antiferromagnetic canted configurations, our most successful model
771: of the observed data consists of $\sim$ 0.91
772: $\mu_{B}$ Cu moments canted in the $a$-$c$ plane at an angle of
773: $\theta$=76$^{o}$ to the $c$-axis. Nearest neighboring ~ 0.12 $\mu_{B}$ O moments are ferromagnetically coupled to the Cu along $a$, but
774: counter-canted at an angle of $\theta$=116$^{o}$ from $c$, (see also
775: figure \ref{cant}).
776:
777: Previous researchers \cite{weht1} have suggested that the Cu $d_{yz}$
778: and O $p_\sigma$ orbitals, which extend along the magnetic ribbons,
779: are energetically isolated and strongly hybridized. Covalency effects
780: serve to polarize the neighboring oxygen ions, resulting in a large
781: net transfer of the Cu moment to the oxygen sites. In
782: $\mathrm{Li_{2}CuO_{2}}$ we find that the Cu moment is large and
783: well separated from that of the Cu. Counter-canting of adjacent oxygen
784: moments seems inconsistent with the idea of a purely
785: polarization-induced oxygen magnetization, and it is possible that the
786: true magnetic structure of $\mathrm{Li_{2}CuO_{2}}$ may be
787: frustrated, ferromagnetically canted or more complicated than our proposed
788: model. (A ferromagnetically canted configuration of moments would have
789: contributed a slight intensity to the nuclear peaks and was therefore
790: difficult to refine using our existing data.) In order to determine
791: the nature configuration of any canted phase with more confidence further
792: experiments are required.
793:
794: \section{Acknowledgements}
795: The authors thank Julie Staunton, Martin Boehm, Jim Hague, Clive
796: Wilkinson and Jane Brown for useful discussions, and the Engineering and
797: Physical Sciences Research Council, UK, for financial support.
798:
799: \bibliography{lco}
800:
801: \end{document}
802:
803:
804:
805:
806:
807:
808:
809: