1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2: \documentclass[12pt,floats]{article}

3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4: 

5: \setlength{\textheight}{23.0cm}

6: \setlength{\textwidth} {15.0cm}

7: \setlength{\topmargin}{-0.40cm}

8: 

9: \usepackage[]{umlaut,latexsym}

10: \usepackage[xdvi]{graphicx}

11: 

12: \usepackage{natbib}

13: \usepackage{amssymb}

14: 

15: %%%%%%%%%%%%%%%%

16: \begin{document}

17: %%%%%%%%%%%%%%%%

18: 

19: \parskip 2mm

20: 

21: \renewcommand{\refname}{\normalsize \bf \em References}

22: 

23: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

24: \title{\bf EXCHANGE INTERACTIONS AND MAGNETIC ANISOTROPY 

25:            IN THE ``Ni$_4$'' MAGNETIC MOLECULE } 

26: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%     

27: %

28: %     

29: \author{

30:       A.\ V.\ POSTNIKOV,$^{a,b\,}$\footnote{Corresponding author:

31:       Tel.: +49 541 969 2377,

32:       fax:  +49 541 969 2351,

33:       email: \mbox{apostnik@uos.de}}~

34:       M.\ BR\"UGER$^a$

35:       and J.\ SCHNACK$^a$

36: \\*[0.2cm]

37:       $^a${\small \it Universit\"at Osnabr\"uck -- Fachbereich Physik,

38:       D-49069 Osnabr\"uck, Germany;} \\

39:       $^b${\small \it Institute of Metal Physics,

40:       S. Kowalewskoj 18, Yekaterinburg 620219, Russia}

41: }

42: %\date{\small \it (Received \today)}

43: \maketitle

44: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

45: 

46: %%%%%%%%%%%%%%%%

47: \begin{abstract}

48: %%%%%%%%%%%%%%%%

49: %

50: Magnetic properties of a tetrahedral ``Ni$_4$'' molecule

51: are discussed in terms of the Heisenberg model, with magnetic anisotropy

52: terms included, and on the basis of first-principle calculations

53: within the density functional theory. It is shown that the isotropic

54: Heisenberg model does not provide an adequate description of

55: magnetization at low temperatures; an inclusion of single-site

56: anisotropy terms does not help to improve the situation either. We suppose

57: that the magnetostriction of the molecule and hence the dependence of

58: interatomic coupling parameters on the magnetization might be important

59: for an adequate description of magnetic properties.

60: The first-principle calculations confirm the system's general preference

61: for antiferromagnetic coupling, as well as the failure of the isotropic

62: Heisenberg model.

63: A conjugated-gradient search for the relaxed structure of the ``Ni$_4$'' 

64: molecule in ferromagnetic and antiferromagnetic configurations did not show

65: any clear tendency to diversification of interatomic distances.

66: These calculations however have not yet included the spin-orbit coupling,

67: which can be essential for analyzing the effects of magnetostriction.

68: %  

69: %%%%%%%%%%%%%%

70: \end{abstract}

71: %%%%%%%%%%%%%%

72: 

73: \noindent {\it Keywords:}\/ Molecular magnets; Heisenberg model; 

74: Density functional theory, \linebreak Exchange interactions

75: 

76: %%%%%%%%%%%%%%%%%%%%%%%%%%

77: \section*{1. INTRODUCTION}

78: %%%%%%%%%%%%%%%%%%%%%%%%%%

79: 

80: The system of our present interest belongs to a large family of

81: molecular magnets --  metallo-organic or -anorganic substances 

82: in which well defined

83: molecular fragments possess intrinsic magnetic structure, and can

84: crystallize forming a stable sold phase. An introduction into the subject 

85: can be found in the books by \citet{Kahn-book}, or \citet{Mol_Magnets}.%

86: \footnote{In a broader perspective,

87: one includes sometimes purely organic magnetic substances, on one side,

88: and three-dimensional connected metalloorganic systems, on the other side,

89: in the definition of molecular magnets.} 

90: The practical interest for molecular magnets is heated by high net spin moments

91: and/or large magnetic anisotropy in some of them, that seems promising 

92: for dense magnetic storage. Moreover, unique and novel properties 

93: of molecular magnets, like quantum tunneling of magnetization

94: \citep{PRL76-3830,PRL82-3903}, open perspectives of basically new

95: applications -- manipulation of magnetic states by light, work media

96: for quantum computers, etc.

97: 

98: \noindent

99: In the actual stage of ``extensive'' exploration of the field, 

100: as clear guidelines relating chemical composition to properties

101: are still missing, one witnesses a permanent quest on the side of 

102: chemists to synthesize new systems with unusual or otherwise so far unknown 

103: combination of metal or organic building blocks.

104: ``Ni$_4$'' is such a system, where the magnetic coupling between

105: four Ni ions (each carrying spin $s$=1) occurs via a long and

106: chemically not very common path.

107: 

108: \noindent

109: In the present contribution, we outline the analysis of known

110: magnetic properties in terms of the Heisenberg model in its simplest

111: form and also with some extensions. Simultaneously, we perform

112: first-principles simulations of the electronic structure and

113: try to estimate magnetic interaction parameters from them.

114: 

115: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

116: \section*{2. STRUCTURE AND BASIC MAGNETIC \newline 

117: \hspace*{0.8cm}PROPERTIES}

118: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

119: 

120: \begin{figure}[t!] 

121:   \begin{center}

122:   \includegraphics*[width=12cm]{ni4-fig1.eps}

123:   \end{center}

124: \medskip

125: \caption{\small

126: The arrangement of basic structural units Ni$_4$Mo$_{12}$ units in

127: the molecular crystal (cortesy of Paul K\"ogeler). A slightly distorted 

128: Ni$_4$ tetrahedron (dark circles) is kept together by O-bridged Mo$_{12}$ 

129: cage (gray circles). 

130: \label{fig:struc1}

131: }

132: \end{figure}

133: 

134: The full chemical formula of ``Ni$_4$'' is

135: [Mo$_{12}$O$_{30}$($\mu_2$-OH)$_{10}$H$_2$\{Ni(H$_2$O)$_3$\}$_4$]

136: $\cdot$14 H$_2$O;

137: its synthesis and characterization have been reported 

138: by \citet{InCh39-5176}.

139: The basic structural unit is a slightly distorted Ni$_4$ tetrahedron,

140: whose vertices cap the hexagonal faces of the O-bridged Mo$_{12}$ cage. 

141: The crystal structure contains 

142: two such units, related by the 180$^{\circ}$ rotation around an edge 

143: of the tetrahedron, as is shown in Fig.~\ref{fig:struc1}.

144: The nearest neighbourhood of each Ni atom is (nearly) octahedral O$_6$,

145: with three apical oxygens saturated by bonded hydrogen, and three

146: others participating in the bonding to Mo atoms.

147: The magnetic interaction path is therefore Ni--O--Mo$_2$--O--Ni,

148: with the length of 6.6--6.7 {\AA}. 

149: The nominal Ni$^{I\!I}$ state corresponds to the $3d^8$ configuration 

150: and hence $s$=1. The ground state was found to have the total spin $S$=0

151: \citep{Schnack-Ni4}.

152: 

153: \noindent

154: The quantitative description of magnetic properties (magnetization and

155: magnetic susceptibility) reported by \citet{InCh39-5176} requires

156: to choose an underlying physical model. The traditional simplest

157: choice is that of the Heisenberg model, which we introduce 

158: (including the Zeeman term) as follows:

159: %

160: %%%%%%%%%%%%%%%%

161: \begin{eqnarray}

162: %%%%%%%%%%%%%%%%

163: %

164: {\cal H} = J\!\sum_{\{\alpha,\beta\} \atop \mbox{\tiny pairs}}^6\!

165: \mathbf{s}_{\alpha}\!\cdot\!\mathbf{s}_{\beta}

166: + g\mu_{\mbox{\tiny B}}

167: \sum_{\alpha}^4\,\mathbf{B}\!\cdot\!\mathbf{s}_{\alpha}\,.

168: \label{eq:Heis}

169: %%%%%%%%%%%%%%

170: \end{eqnarray}

171: %%%%%%%%%%%%%%

172: \noindent

173: All $s_{\alpha}$=1, and $J>0$ corresponds to antiferromagnetic (AFM) 

174: coupling in this formulation.

175: In the following, we would need to extend this model over anisotropic effects.

176: The simplest case is an inclusion of the single-ion anisotropy,

177: %

178: %%%%%%%%%%%%%%%%

179: \begin{eqnarray}

180: %%%%%%%%%%%%%%%%

181: %

182: {\cal H} = J\!\sum_{\{\alpha,\beta\}\atop \mbox{\tiny pairs}}^6\!

183: \mathbf{s}_{\alpha}\cdot\mathbf{s}_{\beta}

184: + D\!\left[\sum_{\alpha}^4(\mathbf{e}_{\alpha}\!\cdot\!\mathbf{s}_{\alpha})^2

185: - \frac{8}{3}\right]

186: + g \mu_{\mbox{\tiny B}}\sum_{\alpha}^4\,\mathbf{B}\cdot\mathbf{s}_{\alpha}\,.

187: \label{eq:aniso}

188: %%%%%%%%%%%%%%

189: \end{eqnarray}

190: %%%%%%%%%%%%%%

191: \noindent

192: with $D$ being the constant of zero-field splitting, and ${\vec e}_{\alpha}$

193: the local anisotropy axis for each ion.

194: 

195: \noindent

196: With $N$=4 coupled spins $s$=1, the total dimension of problem is

197: %

198: %%%%%%%%%%%%%%%%

199: \begin{eqnarray}

200: %%%%%%%%%%%%%%%%

201: (2s+1)^N = 81 = \sum_{S=0}^4\;\sum_{M_S=-S}^S G_S\,,

202: \label{eq:H-dimen}

203: %%%%%%%%%%%%%%

204: \end{eqnarray}

205: %%%%%%%%%%%%%%

206: \noindent

207: where $G_{S=0,1,2,3,4} = 3,6,6,3,1$ are degeneracies of 

208: the $|S, M_S\rangle$ states.

209: The exact diagonalisation of this model 

210: can be straightforwardly done for any choice of parameter values. The magnetic 

211: susceptibility $\chi$ and magnetization $M$ are obtained 

212: in terms of the partition function $Z$:

213: %

214: %%%%%%%%%%%%%%%%

215: \begin{eqnarray}

216: %%%%%%%%%%%%%%%%

217: Z &=& \sum_{S=0}^4 \sum_{M=-S}^S G_S\;

218: e^{-\frac{\displaystyle E_{S,M_S}}{\displaystyle k_{\mbox{\tiny B}}T}}\,;\\

219: M &=& -\frac{g\mu_{\mbox{\tiny B}}}{Z}\;\mbox{Tr}\left[

220: e^{-\frac{\displaystyle \cal H}{\displaystyle k_{\mbox{\tiny B}}T}}

221: S_Z\right]\,,\\

222: \chi &=& \frac{\partial M}{\partial B} = 

223: \frac{(g\mu_{\mbox{\tiny B}})^2}{k_{\mbox{\tiny B}}T\,Z}\;

224: \mbox{Tr}\left[

225: e^{-\frac{\displaystyle \cal H}{\displaystyle k_{\mbox{\tiny B}}T}}

226: S_Z^{\,2}\right]\ - 

227: \frac{M^2}{k_{\mbox{\tiny B}}T}\,. 

228: %%%%%%%%%%%%%%

229: \end{eqnarray}

230: %%%%%%%%%%%%%%

231: \noindent

232: Fig.~\ref{fig:suscep} shows the fit of the measured magnetic susceptibility

233: data, as obtained by Paul K\"ogerler on powder samples and reported

234: by \citet{InCh39-5176}, to the isotropic Heisenberg model of

235: Eq.~(\ref{eq:Heis}). The Heisenberg model yields the Curie--Weiss

236: behaviour,

237: %

238: %%%%%%%%%%%%%%%%

239: \begin{eqnarray}

240: %%%%%%%%%%%%%%%%

241: %

242: \chi_{\mbox{\tiny CW}}=\frac{C}{T+T_{\mbox{\tiny C}}}

243: \quad\mbox{with}\quad

244: C=\frac{8(g\mu_{\mbox{\tiny B}})^2}{3k_{\mbox{\tiny B}}}

245: \quad\mbox{and}\quad

246: T_{\mbox{\tiny C}}=\frac{2J}{k_{\mbox{\tiny B}}}\,,

247: \label{eq:suscep}

248: %%%%%%%%%%%%%%

249: \end{eqnarray}

250: %%%%%%%%%%%%%%

251: \noindent

252: from fitting to which one can estimate the values of $J$ and $g$.

253: Fig.~\ref{fig:suscep} shows that both direct and inverse susceptibility

254: can be satisfactorily fitted through wide temperature range 

255: with only slightly varying parameter values, either 

256: $\left\{J\mbox{=6.6 K}; g\mbox{=2.22}\right\}$, or 

257: $\left\{J\mbox{=8.5 K}; g\mbox{=2.27}\right\}$

258: \citep{Brueger-diplom}.

259: 

260: \begin{figure}[t!] 

261:   \begin{center}

262:   \includegraphics*[width=13cm]{ni4-fig2.eps}

263:   \end{center}

264: \medskip

265: \caption{\small

266: Magnetic susceptibility (top panel) and its inverse (bottom panel), 

267: as reported by \citet{InCh39-5176},

268: in comparison to the fit of the Curie--Weiss law with two sets of

269: parameters \citep{Brueger-diplom}.

270: \label{fig:suscep}

271: }

272: \end{figure}

273: 

274: \begin{figure}[t!]

275:   \begin{center}

276:   \includegraphics*[width=13cm]{ni4-fig3.eps}

277:   \end{center}

278: \medskip

279: \caption{\small

280: An attempted fit \citep{Brueger-diplom} to the measured 

281: \citep[by][]{InCh39-5176} magnetization as function of the external 

282: magnetic field: within the isotropic Heisenberg model (top panel) 

283: and allowing different values of the zero-field splitting parameter $D$ 

284: (bottom panel). 

285: \label{fig:magnet}

286: }

287: \end{figure}

288: 

289: \begin{figure}[t!] 

290:   \begin{center}

291:   \includegraphics*[width=13cm]{ni4-fig4.eps}

292:   \end{center}

293: \medskip

294: \caption{

295: \small

296: A fit to the measured 

297: \citep[by][]{InCh39-5176} magnetization as function of the external 

298: magnetic field $B$, according to the Heisenberg model with

299: zero-field splitting $D$, averaged over 1000 randomly generated

300: orientations, in combination with $B$-dependent magnetic coupling (top panel);

301: the corresponding magnetic susceptibility

302: (bottom panel). 

303: }

304: \label{fig:fitJB}

305: \end{figure}

306: 

307: \noindent

308: However, the observed field dependence of the magnetization,

309: measured at sufficiently low temperature (0.44 K), cannot be

310: reasonably described by any of these sets of parameters, nor by any

311: other one, within the isotropic Heisenberg model.  

312: As is seen in Fig.~\ref{fig:magnet}, the Heisenberg model yields

313: equidistant steps in the magnetization curve, with appropriate 

314: temperature smearing.

315: In the experimental curve, on the contrary, the steps are not 

316: equidistant (in the field strength), moreover they are much stronger

317: smeared, to the extend that the first step is hardly pronounced.

318: 

319: \noindent

320: A number of attempts has been undertaken by \citet{Brueger-diplom}

321: to achieve a reasonable fit to the measured magnetization curve,

322: varying the values of magnetic coupling parameter $J$ and

323: of the zero-field splitting parameter $D$, various

324: orientations of the external magnetic field relative to the Ni$_4$

325: tetrahedron, averaging over different orientations, etc.

326: A non-equidistant steps in the magnetization, somehow similar to

327: those experimentally observed, can be simulated by allowing an increase

328: of coupling constants $J$ with magnetic field. As the physical

329: mechanisms of interatomic coupling, in the simplest picture, have to do 

330: with the overlap of orbitals centered at different sites,

331: a justification of the $J(B)$ dependency might be provided by the 

332: variation of interatomic distances in the magnetic field, i.e.,

333: the magnetostriction effects. The exponential dependence of interatomic

334: overlap integrals with distance would justify the parametrization

335: $J(B) \sim \exp(B/\gamma)$.

336: 

337: \noindent

338: Fig.~\ref{fig:fitJB} shows a fit to the experimental magnetization 

339: and magnetic susceptibility with a zero-field splitting constant included

340: in the Heisenberg model \emph{and} an empirical magnetic field dependency

341: of the magnetic interaction parameter $J$. The exact diagonalisation

342: of the corresponding model Hamiltonian yields an almost

343: acceptable description of the experimental trends, yet leaves

344: the questions open about the microscopic origins, and the specific form, 

345: of the model parameters' variation with magnetic field.

346: 

347: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

348: \section*{3. DFT CALCULATIONS SETUP}

349: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

350: 

351: The first-principles approach, on the contrary, is able to compare

352: the energies of different magnetic configurations without making

353: assumptions about relevant physical interactions in the system.

354: Moreover, the values of model Hamiltonian parameters, like exchange

355: couplings $J$ and anisotropy constants $D$, can be in principle

356: evaluated without referring to any experimental data. An abundant literature

357: exists notably to the subject of $J$ \citep[see, e.g.,]%

358: [for a recent discussion on the subject]{JCP116-2728,JCP116-3985}.

359: We emphasize the difference between multi-configurational quantum-chemical

360: approach, and that of the density functional theory (DFT).

361: The former refers to the differences in energy of, 

362: e.g., singlet and triplet multi-determinantal states.

363: In the DFT, such properties are not accessible, but one can compare

364: the expectation values of the Hamiltonian in specially prepared states,

365: e.g., a triplet (ferromagnetic) and the broken-symmetry

366: (antiferromagnetic) ones. For the comparison of approaches and a more

367: detailed discussion, see \citet{Psik-highlight}.

368: Further reference magnetic configurations can be created in the DFT

369: by using the fixed spin moment (FSM) scheme \citep{JPF14-L129}.

370: The discussion on the applicability of the Heisenberg model

371: based on the results of DFT calculations and comparison of the FSM 

372: results for another molecular magnet, hexanuclear ``ferric wheel'',

373: has been given in \citet{EMRS-Fewheel}. 

374: In the following, we make a similar analysis for the ``Ni$_4$'' system.

375: As in the last cited paper, the DFT calculations have been done with the

376: {\sc Siesta} method \citep{JPCM14-2745}, using norm-conserving 

377: pseudopotentials \citep{PRB43-1993}

378: and a compact basis set of strictly localized numerical functions.

379: Specifically, we used a triple-$\zeta$ basis for Mo$4d$ and Ni$3d$ states,

380: double-$\zeta$ with polarization orbitals for O$2s$ and $2p$, H$1s$, Mo$4p$, 

381: Ni$4p$, and single-$\zeta$ with polarization orbitals for Mo$5s$ and

382: Ni$5p$. The energy shift parameter, governing the spatial confinement of

383: basis function, was 20 mRy. The exchange-correlation was treated in

384: the generalized gradient approximation after \citet{PRL77-3865}.

385: A molecular fragment with 104 atoms, in total, was treated as an

386: isolated molecule in a cubic simulation box with the edge size 20 {\AA}. 

387: The cutoff parameter defining the real-space mesh for solving

388: the Poisson equation was set to 200 Ry, that resulted in a

389: 180$\times$180$\times$180 grid in the abovementioned cubic simulation cell.

390: In order to smoothen the variations of the total energy and other properties

391: due to the ``eggbox effect'', an averaging of calculated grid-dependent

392: properties over a finer fcc-type sub-grid has been done, as described

393: by \citet{JPCM14-2745}.

394: The calculations have been done for the ``as determined'' crystal

395: structure, i.e., corresponding to the ground state of $S$=0.

396: We allowed moreover a structure relaxation for antiferromagnetic

397: (AFM, $S$=0) and ferromagnetic (FM, $S$=1) configurations.

398: 

399: \noindent

400: Before discussing the results it should be noted that the {\sc Siesta}

401: method does not include the treatment of spin-orbit interaction,

402: hence it is not possible to discuss the magnetic anisotropy parameters

403: on the basis of calculation. In principle, the calculation of

404: zero-field splitting $D$ is not only possible, but yields the quantitative

405: results of very high accuracy. \citet{PRB59-R693,PRB60-9566} 

406: proposed a method for calculation of the second-order anisotropy energies,

407: which yielded perfect agreement with experiment for the Mn$_{12}$-acetate

408: (see the cited works) and for a number of other molecular magnets

409: \citep[see Table 3 in][]{Psik-highlight}.

410: 

411: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

412: \section*{4. FIXED SPIN MOMENT CALCULATIONS}

413: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

414: 

415: The FSM scheme fixes the total number of electrons in majority- and 

416: minority-spin channels. This amounts to imposing an effective external 

417: magnetic field, which -- at least in metal systems -- introduces

418: a difference in the chemical potential for majority- and minority-spin

419: electrons. When applied to molecular magnets, the fixed total moment

420: can be only an integer (and usually even) amount of Bohr magnetons.

421: Isolated magnetic molecules have discrete Kohn-Sham energy levels,

422: and normally, a non-zero HOMO-LUMO gap. Therefore both spin channels may

423: still possess a common chemical potential (if their respective HOMO-LUMO 

424: gaps do overlap), or they may need, similarly to the case of a metallic 

425: system, an external magnetic field for enforcing the splitting. 

426: In the former case, the DFT solutions corresponding to different

427: FSM values coexist as metastable states, whose total energies can be

428: directly compared. In the latter case, a Zeeman term must be taken

429: into account as an additional energy needed to split the chemical

430: potentials in two spin channels. These are the foundations of our

431: FSM analysis for Ni$_4$. 

432: 

433: \begin{figure}[t!] 

434:   \begin{center}

435:   \includegraphics*[width=14.5cm]{ni4-fig5.eps}

436:   \end{center}

437: \medskip

438: \caption{\small

439: Total energy relative to that of the AFM ground state (left panel)

440: and HOMO-LUMO gap values in two spin channels (right panel)

441: from the FSM calculations. The relative orientation 

442: of magnetic moments associated with Ni atoms

443: (nominally $s$=1) is shown schematically in the left panel 

444: for different magnetic configurations.

445: \label{fig:fsm}

446: }

447: \end{figure}

448: 

449: \begin{figure}[p!] 

450:   \begin{center}

451:   \includegraphics*[width=14cm]{ni4-fig6.eps}

452:   \end{center}

453: \medskip

454: \caption{\small

455: Local spin-resolved DOS 

456: of two Ni atoms in AFM configuration along with

457: bridging O and Mo atoms. The inset shows the corresponding

458: part of the ``Ni$_4$'' molecule with the numbering of atoms.

459: Local magnetic moments according to Mulliken population

460: analysis are indicated on the left. The shaded areas for Ni (Mo) atoms

461: show the contributions of the $3d$ ($4d$) states.

462: \label{fig:dos}

463: }

464: \end{figure}

465: 

466: \noindent

467: Fig.~\ref{fig:fsm} shows the total energies for a number of configurations

468: with different FSM values (left panel), along with corresponding

469: HOMO-LUMO gaps in two spin channels. For setting up an AFM configuration

470: (zero FSM value), the local magnetic moments of two (arbitrary, due to

471: nearly perfect tetrahedral symmetry) Ni atoms were inverted. This

472: configuration then always survived in the course of self-consistent

473: calculation.

474: The corresponding spin-resolved local densities of states (DOS) for

475: two oppositely magnetized Ni atoms and several of their neighbours are shown

476: in Fig.~\ref{fig:dos}. A fragment of the molecule, including the path

477: between two Ni atoms, is depicted in the inset. It is well seen

478: that the local DOS at the Ni site in the majority-spin channel

479: is dominated by a fully occupied $3d$ subband, which forms a single

480: narrow peak. In the minority-spin channel, one sees a crystal-field

481: splitting into the $t_{2g}$ (occupied) and the $e_g$ (empty) energy levels.

482: Such clear separation, without a pronounced further splitting, 

483: is due to the fact that the oxygen octahedra around the Ni atoms

484: are almost undistorted. The above electronic configuration corresponds

485: to the spin value $s$=1 per Ni atom, consistently with experiment.

486: The value of the \emph{local} magnetic moment per Ni atom, as determined

487: by the Mulliken population analysis, may however differ from

488: 2 $\mu_{\mbox{\tiny B}}$, depending on the spatial distribution of

489: the spin density and its decomposition into different basis orbitals.

490: In particular, the charge transfer from Ni to O reduces the nominal value

491: of the majority-spin Ni$3d$ occupation, at the same time inducing

492: magnetic moments at the oxygen ligands. The values of local magnetic moments

493: at several representative atoms is shown in Fig.~\ref{fig:dos}

494: on the left to each plot. Obviously the Ni atom with its six O neighbours 

495: makes a fragment which essentially carries the magnetic moment 

496: of 2 $\mu_{\mbox{\tiny B}}$; the magnetization

497: of Mo atoms is not pronounced.

498: 

499: \noindent

500: Such cumulative spins $s$=1 per NiO$_6$ fragment remain relatively intact

501: if one imposes different values of the fixed spin moment. The AFM state

502: (FSM = 0 $\mu_{\mbox{\tiny B}}$), the one with the spin of one NiO$_6$ group

503: inverted (FSM = 4 $\mu_{\mbox{\tiny B}}$) and the FM state 

504: (FSM = 8 $\mu_{\mbox{\tiny B}}$) can all easily be realized, and differ

505: only slightly in their total energies. It is noteworthy that the imposed

506: FSM value of 2 $\mu_{\mbox{\tiny B}}$ converges to a configuration where

507: three NiO$_6$ groups maintain the spins $s$=1, and the fourth Ni atom

508: becomes non-magnetic (see Fig.~\ref{fig:fsm}, left panel). This configuration

509: lies higher in energy than the three previously mentioned ones.

510: Yet in all four cases there is a common gap between occupied and

511: unoccupied states in both spin channels (see Fig.~\ref{fig:fsm}, right panel),

512: that means that all four configurations can be realized as metastable

513: magnetic solutions, without imposing an external magnetic field and hence

514: without an extra energy cost due to the Zeeman term. 

515: 

516: \noindent

517: The energy of the FSM = 2 $\mu_{\mbox{\tiny B}}$ configuration cannot be

518: used for the mapping of calculation results to the Heisenberg model,

519: because it contains a different number of isolated spins as the other three

520: cases. For FSM =0, 4, and 8 $\mu_{\mbox{\tiny B}}$ the difference 

521: in energies comes from different orientation of ``rigid'' magnetic moments,

522: and hence can serve for an attempt to extract the magnetic interaction 

523: parameters. As can be seen in Table I, which compares the energy

524: differences between three orientations of four $s$=1 spins

525: according to the Heisenberg model and to the calculation results,

526: such extraction is not straightforward. Actually it is an indication

527: that the Heisenberg model does not reasonably describe magnetic

528: interactions in the ``Ni$_4$'' molecule. This qualitative observation

529: is consistent with the analysis of experimental data discussed above.

530: The best one can do is to provide an order-of-magnitude estimate

531: of the Heisenberg exchange parameter, which is positive, i.e.

532: favouring the antiferromagnetic coupling (in the definition of

533: Eq.~\ref{eq:Heis}), and hence consistent with experiment,

534: and has the magnitude $J \approx$ 35--87 K. The earlier discussed

535: $J$ value estimated from the fit to experimental results of

536: magnetic susceptibility (Fig.~\ref{fig:suscep}) are 6--9 K. 

537: To some extent, the difference can be due to underestimating

538: the intraatomic correlation effects in a conventional DFT calculation. 

539: It has been shown by \citet{PRB65-184435} for the Mn$_{12}$-acetate molecule

540: and by \citet{PRB67-134408} for the ``V$_{15}$'' molecular magnet

541: that an artificial enhancement of the on-site Hubbard-like correlation

542: within the LDA+$U$ formalism \citep{LDA+U}

543: favours the localization of the magnetic moment at the $3d$ atoms

544: and tends to reduce the interatomic magnetic interactions. 

545: However, the selection of ``correlated'' states has to be done

546: \emph{ad hoc} in the LDA+$U$ formalism, and the magnitude of the necessary 

547: correction enters the calculation as an external tuning parameter.

548: 

549: \begin{table}[t!]

550: \caption{\small

551: An attempted mapping of the calculated total energies 

552: for three metastable FSM solutions onto the Heisenberg model.

553: The relations between total energy values are not compatible

554: with the predictions of the model. Order-of-magnitude estimates 

555: of the $J$ parameter from two energy differences to the ground state 

556: are 87 K and 35 K.

557: }

558: \begin{center}

559: \begin{tabular*}{\textwidth}{c@{\extracolsep\fill}c@{\extracolsep\fill}c}

560: \hline

561: FSM ($\mu_{\tiny B}$) \rule[-6mm]{0mm}{12mm} & 

562: $\sum\limits_{\{i,j\}\mbox{\footnotesize pairs}}\!\!\!J\,

563:  \mathbf{s}_i\,\mathbf{s}_j$ &

564:  $E_{\mbox{\footnotesize tot}}\!-\!

565:   E_{\mbox{\footnotesize tot}}^{\mbox{\footnotesize(FSM=0)}}$ (meV) \\

566:   \hline \hline

567:   0 & $-2J$ & ~0 \\

568:   4 &   0   & ~6 \\

569:   8 & $~6J$ & 60 \\

570: \hline

571: \end{tabular*}

572: \end{center}

573: \end{table}

574: 

575: \noindent

576: It should be noted that the AFM configuration, being a broken-symmetry

577: state, is \emph{not} a faithful representation of the true $S$=0

578: ground state of the system. Moreover it is obviously a frustrated state,

579: because each Ni atom interacts with one Ni magnetized in parallel 

580: and two antiparalelly magnetized counterparts, whereas all coupling

581: constants are expected to be identical, for symmetry reasons.

582: Nevertheless, the calculated total energy of this broken symmetry state 

583: (Table I) may be mapped onto the predictions of the Heisenberg model,

584: taken together with the results for other two lowest-energy configurations,

585: according to the FSM analysis of Fig.~\ref{fig:fsm}. 

586: 

587: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

588: \section*{5. STRUCTURE RELAXATION}

589: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

590: 

591: \begin{figure}[b!] 

592:   \begin{center}

593:   \includegraphics*[width=13.0cm]{ni4-fig7.eps}

594:   \end{center}

595: \medskip

596: \caption{\small

597: Total energy (top panel) and six Ni--Ni distances (two bottom panels)

598: in the course of unconstrained structure optimization, for

599: FSM values 0 $\mu_{\mbox{\tiny B}}$ (AFM state) and

600: 8 $\mu_{\mbox{\tiny B}}$ (FM state).

601: \label{fig:relax}

602: }

603: \end{figure}

604: 

605: In conclusion we discuss the structure optimization of the ``Ni$_4$''

606: molecule in AFM and FM configurations, which correspond to the FSM 

607: values of 0 and 8 $\mu_{\mbox{\tiny B}}$, correspondingly.

608: The aim of the simulation was to check whether the interatomic distances 

609: in the molecule would tend to become pronouncedly different in two 

610: magnetic configurations, that would probe the assumption of large 

611: magnetostriction in the molecule. Our preliminary results 

612: (Fig.~\ref{fig:relax}) does not seem to confirm the latter hypothesis,

613: but definite conclusion can be only be made on the basis of

614: fully relativistic calculations.

615: Our present simulation includes relativistic effects but the spin-orbit 

616: interaction; consequently, the influence of magnetic structure on total

617: energy and forces is only via the hybridisation which is slightly

618: different in FM and AFM configurations. 

619: When starting from the reported experimental geometry,

620: an initial stage of relaxation (not shown in Fig.~\ref{fig:relax}) 

621: substantially lowered the total energy, and induced strong fluctuations 

622: of interatomic distances. Eventually the molecule

623: recovers Ni--Ni distances of 6.6--6.7 {\AA}, not pronouncedly different

624: in FM and AFM cases. A subsequent smooth lowering of the total energy  

625: does not yet seem converged after 200 conjugated gradient steps.

626: 

627: \noindent

628: The fact that the total energy of the FM configuration in Fig.~\ref{fig:relax}

629: lies, over a large number of simulation steps, lower than that of the AFM,

630: does not contest our above result concerning the system's tendency

631: for the AFM coupling. The two curves in the upper panel 

632: of Fig.~\ref{fig:relax} reflect two different relaxation histories,

633: so that two solutions which accidentally correspond to the same step number

634: stay in no obvious relation to each other. 

635: A meaningful comparison of the FM and AFM total energies requires

636: either the identical structure (as it has been done in the previous

637: section), or full relaxation of structures in two magnetic configurations

638: to be compared.

639: 

640: \noindent

641: Summarizing, we attempted a fit of experimental magnetic susceptibility

642: and magnetization data for the ``Ni$_4$'' magnetic molecule to the

643: Heisenberg model, including zero-field splitting (single-ion

644: anisotropy) and allowing variations of exchange parameters $J$ 

645: with the external magnetic field.

646: It seems that fixed (magnetic field and temperature-independent)

647: values of $J$ do not allow to achieve an acceptable fit of

648: experimental dependencies. As a microscopic reason for the $J$ dependency

649: on magnetic field, we assume a magnetostriction in the ``Ni$_4$'' molecule

650: to play an important role. Without the deformation of the molecule

651: in the magnetic field taken into account, the results of first-principles

652: calculations are consistent with experimental data: they indicate

653: an antiferromagnetic interaction between Ni atoms, and confirm the

654: conclusion that the magnetic properties of the system cannot be fit

655: to the isotropic Heisenberg model. An attempt to simulate the relaxation

656: of the molecule from first principles indicated a not very pronounced

657: difference in the Ni--Ni distances between ferromagnetic and

658: antiferromagnetic configurations of the molecule. However, the spin-orbit

659: interaction has not been included in the present simulation; yet it is

660: essential for a precise treatment of magnetic anisotropy and magnetostriction,

661: so its inclusion is likely to affect the results.

662: \\*[0.5cm]

663: \noindent {\normalsize \bf \em Acknowledgements} 

664: \\*[0.5cm]

665: The authors thank the Deutsche Forschungsgemeinschaft for

666: financial support (Priority Program SPP 1137 ``Molecular Magnetism'').

667: AVP acknowledges useful discussions with Paul K\"ogerler 

668: and Stefan Bl\"ugel.

669: 

670: \begin{thebibliography}{20}

671: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi

672: \expandafter\ifx\csname url\endcsname\relax

673:   \def\url#1{{\tt #1}}\fi

674: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi

675: 

676: \bibitem[{Anisimov {\em et~al.\/}(1997)Anisimov, Aryasetiawan and

677:   Lichtenstein}]{LDA+U}

678: Anisimov, V.~I., F.~Aryasetiawan and A.~I. Lichtenstein (1997).

679: \newblock First-principles calculations of the electronic structure and spectra

680:   of strongly correlated systems: the {LDA+$U$} method.

681: \newblock {\em J.~Phys.:~Condens.~Matter\/} {\bf 9}(4), 767.

682: 

683: \bibitem[{Boukhvalov {\em et~al.\/}(2003)Boukhvalov, Kurmaev, Moewes, Zatsepin

684:   {\em et~al.\/}}]{PRB67-134408}

685: Boukhvalov, D.~W., E.~Z. Kurmaev, A.~Moewes, D.~A. Zatsepin {\em et~al.\/}

686:   (2003).

687: \newblock Electronic structure of magnetic molecules {V$_{15}$}: {LSDA+U}

688:   calculations, x-ray emissions, and photoelectron spectra.

689: \newblock {\em Phys.~Rev.~B\/} {\bf 67}(13), 134408.

690: 

691: \bibitem[{Boukhvalov {\em et~al.\/}(2002)Boukhvalov, Lichtenstein, Dobrovitski,

692:   Katsnelson {\em et~al.\/}}]{PRB65-184435}

693: Boukhvalov, D.~W., A.~I. Lichtenstein, V.~V. Dobrovitski, M.~I. Katsnelson {\em

694:   et~al.\/} (2002).

695: \newblock Effect of local {C}oulomb interactions on the electronic structure

696:   and exchange interactions in {Mn$_{12}$} magnetic molecules.

697: \newblock {\em Phys.~Rev.~B\/} {\bf 65}(18), 184435.

698: 

699: \bibitem[{Br{\"u}ger(2003)}]{Brueger-diplom}

700: Br{\"u}ger, M. (2003).

701: \newblock Einfluss der {A}nisotropie auf die {E}igenschaften magnetischer

702:   {M}olek{\"u}ule am {B}eispiel des {Ni$_4$}-{M}olek{\"u}ls.

703: \newblock Master's thesis, Universit{\"a}t {O}snabr{\"u}ck.

704: 

705: \bibitem[{Calzado {\em et~al.\/}(2002{\natexlab{a}})Calzado, Cabrero, Malrieu

706:   and Caballol}]{JCP116-2728}

707: Calzado, C.~J., J.~Cabrero, J.~P. Malrieu and R.~Caballol (2002{\natexlab{a}}).

708: \newblock Analysis of the magnetic coupling in binuclear complexes. {I}.

709:   {P}hysics of the coupling.

710: \newblock {\em J.~Chem.~Phys.\/} {\bf 116}(7), 2728.

711: 

712: \bibitem[{Calzado {\em et~al.\/}(2002{\natexlab{b}})Calzado, Cabrero, Malrieu

713:   and Caballol}]{JCP116-3985}

714: Calzado, C.~J., J.~Cabrero, J.~P. Malrieu and R.~Caballol (2002{\natexlab{b}}).

715: \newblock Analysis of the magnetic coupling in binuclear complexes. {II}.

716:   {D}erivation of valence effective {H}amiltonians from ab initio {CI} and

717:   {DFT} calculations.

718: \newblock {\em J.~Chem.~Phys.\/} {\bf 116}(10), 3985.

719: 

720: \bibitem[{Friedman {\em et~al.\/}(1996)Friedman, Sarachik, Tejada and

721:   Ziolo}]{PRL76-3830}

722: Friedman, J.~R., M.~P. Sarachik, J.~Tejada and R.~Ziolo (1996).

723: \newblock Macroscopic measurement of resonant magnetization tunneling in

724:   high-spin molecules.

725: \newblock {\em Phys.~Rev.~Lett.\/} {\bf 76}(20), 3830.

726: 

727: \bibitem[{Kahn(1993)}]{Kahn-book}

728: Kahn, O. (1993).

729: \newblock {\em Molecular Magnetism\/}.

730: \newblock John Wiley \& Sons, Singapore.

731: 

732: \bibitem[{Linert and Verdaguer(2003)}]{Mol_Magnets}

733: Linert, W. and M.~Verdaguer, editors (2003).

734: \newblock {\em Molecular Magnets\/}.

735: \newblock Springer-Verlag, Wien.

736: \newblock Special Edition of Monatshefte f{\"u}r Chemie/Chemical Monthly, Vol.

737:   134, No. 2.

738: 

739: \bibitem[{M\"uller {\em et~al.\/}(2000)M\"uller, Beugholt, K\"ogerler, B\"ogge

740:   {\em et~al.\/}}]{InCh39-5176}

741: M\"uller, A., C.~Beugholt, P.~K\"ogerler, H.~B\"ogge {\em et~al.\/} (2000).

742: \newblock

743:   {[Mo$_{12}$O$_{30}$($\mu_2$-OH)$_{10}$H$_2$\{Ni$^{II}$(H$_2$O)$_3$\}$_4$]}, a

744:   highly symmetrical e-{K}eggin unit capped with four {Ni$^{II}$} centers:

745:   Synthesis and magnetism.

746: \newblock {\em Inorg.~Chem.\/} {\bf 39}(23), 5176.

747: 

748: \bibitem[{Pederson and Khanna(1999{\natexlab{a}})}]{PRB59-R693}

749: Pederson, M.~R. and S.~N. Khanna (1999{\natexlab{a}}).

750: \newblock Electronic structure and magnetism of {Mn$_{12}$O$_{12}$} clusters.

751: \newblock {\em Phys.~Rev.~B\/} {\bf 59}(2), R693.

752: 

753: \bibitem[{Pederson and Khanna(1999{\natexlab{b}})}]{PRB60-9566}

754: Pederson, M.~R. and S.~N. Khanna (1999{\natexlab{b}}).

755: \newblock Magnetic anisotropy barrier for spin tunneling in {Mn$_{12}$O$_{12}$}

756:   molecules.

757: \newblock {\em Phys.~Rev.~B\/} {\bf 60}(13), 9566.

758: 

759: \bibitem[{Perdew {\em et~al.\/}(1996)Perdew, Burke and Ernzerhof}]{PRL77-3865}

760: Perdew, J.~P., K.~Burke and M.~Ernzerhof (1996).

761: \newblock Generalized gradient approximation made simple.

762: \newblock {\em Phys.~Rev.~Lett.\/} {\bf 77}(18), 3865.

763: 

764: \bibitem[{Postnikov {\em et~al.\/}(2004{\natexlab{a}})Postnikov,

765:   Chiuzb{\u{a}}ian, Neumann and Bl{\"u}gel}]{EMRS-Fewheel}

766: Postnikov, A.~V., S.~G. Chiuzb{\u{a}}ian, M.~Neumann and S.~Bl{\"u}gel

767:   (2004{\natexlab{a}}).

768: \newblock Electron spectroscopy and density-functional study of ``ferric

769:   wheel'' molecules.

770: \newblock {\em J. Phys. Chem. Solids\/} {\bf 65}(4), 813.

771: 

772: \bibitem[{Postnikov {\em et~al.\/}(2004{\natexlab{b}})Postnikov, Kortus and

773:   Pederson}]{Psik-highlight}

774: Postnikov, A.~V., J.~Kortus and M.~R. Pederson (2004{\natexlab{b}}).

775: \newblock Density functional studies of molecular magnets.

776: \newblock Scientific Highlight of the month February 2004, Newsletter 61 of the

777:   {$\Psi_k$}-Network.

778: \newblock \url{http://psi-k.dl.ac.uk/newsletters/News\_61/Highlight\_61.pdf}.

779: 

780: \bibitem[{Schnack {\em et~al.\/}(2004)Schnack, Br{\"u}ger, Postnikov, Luban

781:   {\em et~al.\/}}]{Schnack-Ni4}

782: Schnack, J., M.~Br{\"u}ger, A.~Postnikov, M.~Luban {\em et~al.\/} (2004).

783: \newblock Looking simple, being tricky: {Ni$_4$} -- the first magnetic molecule

784:   showing magnetostriction?

785: \newblock (in preparation).

786: 

787: \bibitem[{Schwarz and Mohn(1984)}]{JPF14-L129}

788: Schwarz, K. and P.~Mohn (1984).

789: \newblock Itinerant metamagnetism in {YCO$_2$}.

790: \newblock {\em J.~Phys.~F: Metal Phys.\/} {\bf 14}(7), L129.

791: 

792: \bibitem[{Soler {\em et~al.\/}(2002)Soler, Artacho, Gale, Garc{\'{\i}}a {\em

793:   et~al.\/}}]{JPCM14-2745}

794: Soler, J.~M., E.~Artacho, J.~D. Gale, A.~Garc{\'{\i}}a {\em et~al.\/} (2002).

795: \newblock The {SIESTA} method for \emph{ab initio} order-{$N$} materials

796:   simulation.

797: \newblock {\em J.~Phys.:~Condens.~Matter\/} {\bf 14}(11), 2745.

798: \newblock {\sc Siesta} site: \url{http://www.uam.es/siesta}.

799: 

800: \bibitem[{Troullier and Martins(1991)}]{PRB43-1993}

801: Troullier, N. and J.~L. Martins (1991).

802: \newblock Efficient pseudopotentials for plane-wave calculations.

803: \newblock {\em Phys.~Rev.~B\/} {\bf 43}(3), 1993.

804: 

805: \bibitem[{Wernsdorfer {\em et~al.\/}(1999)Wernsdorfer, Ohm, Sangregorio,

806:   Sessoli {\em et~al.\/}}]{PRL82-3903}

807: Wernsdorfer, W., T.~Ohm, C.~Sangregorio, R.~Sessoli {\em et~al.\/} (1999).

808: \newblock Observation of the distribution of molecular spin states by resonant

809:   quantum tunneling of the magnetization.

810: \newblock {\em Phys.~Rev.~Lett.\/} {\bf 82}(19), 3903.

811: 

812: \end{thebibliography}

813: 

814: %%%%%%%%%%%%%%

815: \end{document}

816: %%%%%%%%%%%%%%

817: 

818: