1: % Submitted to PRB 17 July 2002, bu8239

2: % Include figure files

3: % Align table columns on decimal point

4: % bold math

5: % standard setting

6: %\topmargin= -32 truemm  % for bloody cond-mat

7: 

8: 

9: \documentclass[aps,prb,showpacs,twocolumn,floats]{revtex4}

10: \usepackage{amssymb}

11: 

12: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

13: \usepackage{graphicx,psfig}

14: \usepackage{dcolumn}

15: \usepackage{bm}

16: 

17: %TCIDATA{OutputFilter=LATEX.DLL}

18: %TCIDATA{LastRevised=Wed Feb 05 18:29:42 2003}

19: %TCIDATA{<META NAME="GraphicsSave" CONTENT="32">}

20: 

21: \newcommand{\onefourth}{\mbox{\scriptsize

22: \raisebox{1.5mm}{1}\hspace{-2.7mm}

23: \raisebox{0.5mm}{$-$}\hspace{-2.8mm}

24: \raisebox{-0.9mm}{4}\hspace{-0.7mm} \normalsize }}

25: \textheight= 243 truemm

26: \topmargin= -15 truemm

27: \newcommand{\erfc}{ {\rm erfc} }

28: \newcommand{\const}{ {\rm const} }

29: \newcommand{\arctanh}{ {\rm arctanh} }

30: \input{tcilatex}

31: 

32: \begin{document}

33: 

34: \title{Landau-Zener-Stueckelberg effect in a model of interacting tunneling

35: systems}

36: \author{D. A. Garanin}

37: \affiliation{ Institut f\"ur Physik,

38: Johannes-Gutenberg-Universit\"at,

39:  D-55099 Mainz, Germany}

40: \date{\today}

41: 

42: 

43: \begin{abstract}

44: The Landau-Zener-Stueckelberg (LZS) effect in a model system of

45: interacting tunneling particles is studied numerically and

46: analytically. Each of $N$ tunneling particles interacts with each

47: of the others with the same coupling $J.$ This problem maps onto

48: that of the LZS effect for a large spin $S=N/2.$ The mean-field

49: limit $N\rightarrow \infty $ corresponds to the classical limit

50: $S\rightarrow \infty $ for the effective spin. It is shown that

51: the ferromagnetic coupling $J>0$ tends to suppress the LZS

52: transitions. For $N\rightarrow \infty $ there is a critical value

53: of $J$ above which the staying probability $P$ does not go to zero

54: in the slow sweep limit, unlike the standard LZS effect. In the

55: same limit for $J>0$ LZS transitions are boosted and $P=0$ for a

56: set of finite values of the sweep rate. Various limiting cases

57: such as strong and weak interaction, slow and fast sweep are

58: considered analytically. It is shown that the mean-field approach

59: works well for arbitrary $N$ if the interaction $J$ is weak.

60: \end{abstract}

61: 

62: \pacs{03.65.-w, 75.10.Jm}

63: 

64: \maketitle

65: 

66: 

67: 

68: \section{Introduction}

69: 

70: The Landau-Zener-Stueckelberg (LZS)

71: problem\cite{lan32,zen32,stu32} of quantum-mechanical transitions

72: at an avoided level crossing induced by a linear-in-time energy

73: sweep is well known in many areas of physics, in particular, in

74: the physics of atomic and molecular collisions (see, e.g., Ref.\

75: \onlinecite{crohug77}). Recently the LZS\ effect was observed in

76: crystals of single-molecule magnets and it was used to extract

77: their tunneling level splittitng $\Delta

78: $.\cite{werses99,weretal00epl}

79: 

80: The specifics of the LZS effect in solid-state systems is the

81: macroscopically large number $N$ ot tunneling species such as spins of

82: magnetic molecules in the above experiments.\cite{werses99,weretal00epl}

83: These tunneling species are usually coupled with each other (mainly via the

84: dipole-dipole interaction in the case of magnetic molecules) and the

85: coupling energy $J$ can easily exceed the splitting $\Delta .$ In this

86: situation tunneling species (that can be described by pseudospins $S=1/2)$

87: do not tunnel independently, and one has to consider the Schr\"{o}dinger

88: equation for the whole system that is described by $2^{N}$ variables. The

89: latter is a new and tremendous problem since the whole Schr\"{o}dinger

90: equation becomes intractable even for a moderate number $N$ while analytical

91: treatment is difficult because the system is far from the equilibrium.

92: 

93: As a plausible first step towards the solution of the LZS problem for a

94: system of interacting particles one can consider a simplified model in which

95: each particle interacts with all $N-1$ other particles with the same

96: strength $J,$ as was done in Ref.\ \onlinecite{hamraemiysai00}. For this

97: model the Schr\"{o}dinger equation simplifies so that one has to solve only $%

98: N+1$ equations instead of the $2^{N}$ equations in the general case. In the

99: limit $N\rightarrow \infty $ the problem simplifies since the mean-filed

100: approximation (MFA) becomes exact. The latter allows one to gain insights

101: into the problem by analyzing its mean-field solution numerically as well as

102: analytically in different limiting cases. Because the \emph{molecular field}

103: exerted on a tunneling particle from the others depends nonlinearly on time

104: in the region of the level crossing, the problem cannot be linearized (c.f.

105: Ref. \onlinecite{hamraemiysai00}) and\ the results differ essentially from

106: the standard LZS solution. For instance, the ferromagnetic coupling, $J>0,$

107: makes the energy sweep faster, and thus the probability $P$ to stay on the

108: initial bare level increases. The nonlinear time dependence of the sweep

109: within the MFA has stimulated Ref.\ \onlinecite{garsch02prb}, where the

110: different kinds of sweep nonlinearities were investigated for a single

111: tunneling system. Certainly the MFA for systems of interacting tunneling

112: particles is a more complicated problem since in this case the form of the

113: sweep is unknown from the beginning and it should be found self-consistently.

114: 

115: It is very interesting to study the difference between the exact

116: quantum-mechanical and the mean-field solutions of the LZS problem for

117: finite $N.$ The original model maps of the model of a large spin $S=N/2$

118: with the uniaxial anisotropy $D\sim J$, transverse field $H_{x}\sim \Delta ,$

119: and the sweeped longitudinal field $H_{z}(t).$ That is, the mean-field limit

120: $N\rightarrow \infty $ of our model corresponds to the classical limit $%

121: S\rightarrow \infty $ for the effective large spin. In Ref.\ %

122: \onlinecite{hamraemiysai00} it was shown numerically that for the model with

123: $N=4$ and $J>0$ the exact and MFA results are qualitatively similar. It is

124: very interesting, however, to investigate the problem for larger $N$ and for

125: $J<0.$

126: 

127: Numerical and analytical study of these problems is the objective of this

128: paper, the rest of which is organized as follows. In Sec.\ \ref

129: {Sec-Hamiltonian} the Hamiltonian is written down, the Schr\"{o}dinger

130: equation is simplified making use of the symmetry of the interaction, the

131: problem is mapped onto that of a large spin and its general properties are

132: studied. In Sec.\ \ref{Sec-separated} the case of well-separated resonances,

133: $\left| J\right| \gg \Delta $ where the problem can be reduced to that of

134: successive standard LZS transitions is investigated. In Sec.\ \ref

135: {Sec-fast-sweep} the opposite case $J\ll \Delta ,$ where the problem can be

136: solved perturbatively in $J,$ is considered. Final results are worked out in

137: the cases of fast sweep and slow sweep. In Sec.\ \ref{Sec-mfa} general

138: properties of the mean-field solution of the LZS problem with interaction

139: are investigated, in particular, with the help of the mapping onto the

140: classical-spin problem. Numerical solution shows, in particular, that for $%

141: J<0$ complete LZS transition, $P=0,$ is achieved at some values of the sweep

142: rate. Analytical treatment of the slow-sweep limit within the MFA is

143: provided in Sec.\ \ref{Sec-slow-MFA}.

144: 

145: \section{The Hamiltonian}

146: 

147: \label{Sec-Hamiltonian}

148: 

149: We consider the model of $N$ double-level tunneling systems described by

150: pseudospins 1/2 and interacting each with each with an equal strength

151: \begin{equation}

152: \widehat{H}=-\frac{1}{2}\sum_{i=1}^{N}\left[ W(t)\sigma _{iz}+\Delta \sigma

153: _{ix}\right] -\frac{J}{2}\sum_{i\neq j=1}^{N}\sigma _{iz}\sigma _{jz}.

154: \label{Ham}

155: \end{equation}

156: Here $\mathbf{\sigma }_{i}$ are Pauli matrices, $W(t)$ is the energy sweep

157: that is taken to be linear in time

158: \begin{equation}

159: W(t)\equiv E_{-1}(t)-E_{1}(t)=vt,  \label{WDef}

160: \end{equation}

161: $\Delta $ is the level splitting at resonance in the absence of interaction (%

162: $t=0$ and $J=0)$, whereas $J$ is the interaction constant. The case $J>0$

163: corresponds to the ferromagnetic (FM) coupling, whereas that of $J<0$

164: corresponds to the antiferromagnetic frustrating (AFMF) coupling. In the

165: latter case, the ground state of the system with $W=\Delta =0$ is a highly

166: degenerate state with the minimal total spin. If $J=0,$ the problem

167: simplifies to the well known LZS problem for individual tunneling systems.

168: If these system at $t\rightarrow -\infty $ are in the bare ground state $%

169: \psi _{-1}\equiv |\downarrow \rangle $ before crossing the resonance, the

170: probability to stay in this state after crossing the resonance at $%

171: t\rightarrow \infty $ \ is given by \cite{zen32,stu32,akusch92,dobzve97}

172: \begin{equation}

173: P(\infty )\equiv P=e^{-\varepsilon },\qquad \varepsilon \equiv \frac{\pi

174: \Delta ^{2}}{2\hbar v}  \label{PLZ}

175: \end{equation}

176: (Probabilities without the argument are shortcuts for the final-state

177: probabilities at $t=\infty $ throughout the paper). For the problem with

178: interaction, the wave function of the system can be written as the expansion

179: over the direct-product states

180: \begin{eqnarray}

181: \Psi (t) &=&\sum_{m_{1},\ldots ,m_{N}=-1,1}C_{m_{1},\ldots ,m_{N}}(t)\psi

182: _{m_{1}}\otimes \ldots \otimes \psi _{m_{N}}  \nonumber \\

183: \psi _{-1} &=&\left(

184: \begin{array}{c}

185: 0 \\

186: 1

187: \end{array}

188: \right) =|\downarrow \rangle ,\qquad \psi _{1}=\left(

189: \begin{array}{c}

190: 1 \\

191: 0

192: \end{array}

193: \right) =|\uparrow \rangle .  \label{PsiDef}

194: \end{eqnarray}

195: The initial condition for $\Psi (t)$ is $C_{-1,\ldots ,-1}(-\infty )=1$

196: whereas all other coefficients are zero, i.e., the system starts in the

197: state with all spins down. With time the state of the system becomes a

198: superposition of all possible basis states in Eq.\ (\ref{PsiDef}). The

199: one-particle probability to remain in the initial state $-1$ for our $N$%

200: -particle system is given by

201: \begin{equation}

202: P_{N}(t)=\sum_{m_{2},\ldots ,m_{N}=-1,1}\left| C_{-1,m_{2},\ldots

203: ,m_{N}}(t)\right| ^{2}  \label{PtDef}

204: \end{equation}

205: and it starts from $P_{N}(-\infty )=1.$

206: 

207: The solution of the LZS problem for this model is simplified by the fact

208: that the coefficients $C_{m_{1},\ldots ,m_{N}}(t)$ depend only on the number

209: $k$ of spins up while they are independent on the choice of these spins.

210: Thus one can label the states by the index $k$ only that results in the

211: Schr\"{o}dinger equation

212: \begin{equation}

213: i\hbar \dot{C}_{k}=E_{k}(t)C_{k}-\frac{k\Delta }{2}C_{k-1}-\frac{(N-k)\Delta

214: }{2}C_{k+1}  \label{SchrEq}

215: \end{equation}

216: for $k=0,\ldots ,N$ and with the bare energies

217: \begin{equation}

218: E_{k}(t)=\left( \frac{N}{2}-k\right) W(t)+2Jk(N-k)+\mathrm{const}.

219: \label{Enk}

220: \end{equation}

221: Each $k$-state is realized by $N!/\left[ (N-k)!k!\right] $ different choices

222: of the indices $m_{1},\ldots ,m_{N}$ in Eq.\ (\ref{PsiDef}). For $J<0,$ the

223: ground state with $W=\Delta =0$\ corresponds to $k=N/2$ and is highly

224: degenerate. In Eq.\ (\ref{PtDef}) $m_{1}$ is fixed to $-1,$ and the

225: remaining indices $m_{2},\ldots ,m_{N}$ can also be parametrized by $k.$ In

226: the corresponding number of realizations, one should use $N-1$ instead of $N$

227: that leads to

228: \begin{equation}

229: P_{N}(t)=\sum_{k=0}^{N-1}\frac{(N-1)!}{(N-1-k)!k!}\left| C_{k}(t)\right|

230: ^{2}.  \label{PtDefk}

231: \end{equation}

232: 

233: It is convenient to introduce the coefficients

234: \begin{equation}

235: c_{k}\equiv \sqrt{\frac{N!}{(N-k)!k!}}C_{k}  \label{ckDef}

236: \end{equation}

237: that satisfy the normalization condition $1=\sum_{k=0}^{N}\left|

238: c_{k}\right| ^{2}$ and the Schr\"{o}dinger equation

239: \begin{equation}

240: i\hbar \dot{c}_{k}=E_{k}(t)c_{k}-\frac{\Delta }{2}l_{k-1,k}c_{k-1}-\frac{%

241: \Delta }{2}l_{k,k+1}c_{k+1}  \label{SchrEqc}

242: \end{equation}

243: with $l_{k,k+1}\equiv \sqrt{(N-k)(k+1)}.$ Then Eq.\ (\ref{PtDefk})

244: simplifies to

245: \begin{equation}

246: P_{N}(t)=\sum_{k=0}^{N}\left( 1-\frac{k}{N}\right) p_{k}(t),\qquad

247: p_{k}(t)\equiv \left| c_{k}(t)\right| ^{2}.  \label{PtFinal}

248: \end{equation}

249: Eq.\ (\ref{SchrEqc}) maps on the corresponding Schr\"{o}dinger equation for

250: a large pseudospin $S=N/2$ with the Hamiltonian

251: \begin{equation}

252: \widehat{H}_{S}=-H_{z}(t)S_{z}-H_{x}S_{x}-DS_{z}^{2}+\mathrm{const,}

253: \label{HamS}

254: \end{equation}

255: where

256: \begin{equation}

257: H_{z}(t)=W(t),\qquad H_{x}=\Delta ,\qquad D=2J  \label{ParIdent}

258: \end{equation}

259: (we set $g\mu _{B}=1$ so that the ``magnetic field'' is energy dimensional).

260: Note that the excitation number $k$ is related to the eigenvalue $m$ of $%

261: S_{z}$ by $k=N/2+m.$ Thus $P(t)$ maps onto

262: \begin{equation}

263: P_{N}(t)=\frac{1}{2}\left( 1-\frac{\langle S_{z}\rangle _{t}}{S}\right) .

264: \label{PtMapping}

265: \end{equation}

266: 

267: The large-spin model defined by Eq.\ (\ref{HamS}) is well known from the

268: physics of the single-molecule magnet Mn$_{12}.$ For small $\Delta $ the

269: energy levels of the system as function of $W$ are nearly straight lines

270: with avoided crossings (see Fig.\ \ref{Fig-lzn-En}). Splittings at avoided

271: crossings are or order $\Delta (\Delta /J)^{n},$ where $n$ are appropriate

272: powers.\cite{gar91jpa}

273: 

274: %TCIMACRO{

275: %\TeXButton{Fig-lzn-En}{\begin{figure}[t]

276: %\unitlength1cm

277: %\begin{picture}(11,6)

278: %\centerline{\psfig{file=Fig-lzn-En-F.eps,angle=-90,width=9cm}}

279: %\end{picture}

280: %\begin{picture}(11,6)

281: %\centerline{\psfig{file=Fig-lzn-En-AF.eps,angle=-90,width=9cm}}

282: %\end{picture}

283: %\caption{ \label{Fig-lzn-En}

284: %$a$ -- Exchange-split resonances for a ferromagnetically coupled spin cluster for $|J|\gtrsim \Delta$.

285: %Transitions shown by the dashed and dotted arrows arise in higher orders in $\Delta/J$ and are much weaker.

286: %$b$ -- Same for  an AFMF coupled spin cluster.

287: %}

288: %\end{figure}%

289: %}}%

290: %BeginExpansion

291: \begin{figure}[t]

292: \unitlength1cm

293: \begin{picture}(11,6)

294: \centerline{\psfig{file=Fig-lzn-En-F.eps,angle=-90,width=9cm}}

295: \end{picture}

296: \begin{picture}(11,6)

297: \centerline{\psfig{file=Fig-lzn-En-AF.eps,angle=-90,width=9cm}}

298: \end{picture}

299: \caption{ \label{Fig-lzn-En}

300: $a$ -- Exchange-split resonances for a ferromagnetically coupled spin cluster for $|J|\gtrsim \Delta$.

301: Transitions shown by the dashed and dotted arrows arise in higher orders in $\Delta/J$ and are much weaker.

302: $b$ -- Same for  an AFMF coupled spin cluster.

303: }

304: \end{figure}%

305: %

306: %EndExpansion

307: 

308: %TCIMACRO{

309: %\TeXButton{Fig-lzn-P3Ft}{\begin{figure}[t]

310: %\unitlength1cm

311: %\begin{picture}(11,6)

312: %\centerline{\psfig{file=Fig-lzn-P3Ft.eps,angle=-90,width=9cm}}

313: %\end{picture}

314: %\caption{ \label{Fig-lzn-P3Ft}

315: %Time dependence of the one-particle staying probability $P_N$ as function of the energy sweep $W(t)$

316: %for a system of $N=3$ ferromagnetically coupled particles.

317: %At intermediate sweep rate, $\varepsilon=1$ only the first-order LZS transition at $W=2J_0$ (the $\Delta$ transition) is seen.

318: %For rather slow sweep rates (here  $\varepsilon=10$),  higher-order transitions (here the  $\Delta^2/J$ transition at $W=0$) are switching on.

319: %}

320: %\end{figure}%

321: %}}%

322: %BeginExpansion

323: \begin{figure}[t]

324: \unitlength1cm

325: \begin{picture}(11,6)

326: \centerline{\psfig{file=Fig-lzn-P3Ft.eps,angle=-90,width=9cm}}

327: \end{picture}

328: \caption{ \label{Fig-lzn-P3Ft}

329: Time dependence of the one-particle staying probability $P_N$ as function of the energy sweep $W(t)$

330: for a system of $N=3$ ferromagnetically coupled particles.

331: At intermediate sweep rate, $\varepsilon=1$ only the first-order LZS transition at $W=2J_0$ (the $\Delta$ transition) is seen.

332: For rather slow sweep rates (here  $\varepsilon=10$),  higher-order transitions (here the  $\Delta^2/J$ transition at $W=0$) are switching on.

333: }

334: \end{figure}%

335: %

336: %EndExpansion

337: 

338: Our model described by Eqs.\ (\ref{Ham}) or (\ref{HamS}), although not

339: completely realistic, allows one to trace down the influence of the

340: interaction on the transition probabilities and to obtain a number of

341: interesting analytical results. We will see that, in particular, suppression

342: of transitions for the ferromagnetic interactions ($J>0)$ can be easily

343: understood. An advantage of this simplified model is the possibility to

344: solve the problem numerically up to much larger system sizes, compared to

345: models with realistic interactions. For the latter, the number of different

346: coefficients in Eq.\ (\ref{PsiDef}) and thus of differential equations to

347: solve is $2^{N},$ whereas in our case there are only $N+1$ equations.

348: 

349: For numerical calculations we use Wolfram Mathematica 4.0 that employs, in

350: particular, a very accurate differential equation solver needed for handling

351: strongly oscillating solutions over large time intervals. The results of

352: solving Eq.\ (\ref{SchrEqc}) are shown in Fig.\ \ref{Fig-lzn-P3Ft} for a

353: cluster of three ferromagnetically coupled particles.

354: 

355: \section{Well-separated resonances}

356: 

357: \label{Sec-separated}

358: 

359: The interaction can profoundly influence the LZS effect if it is strong

360: enough, \TEXTsymbol{\vert}$J|\gtrsim \Delta $. The general tendency can be

361: already seen from the well-known mean-field argument. If one of the

362: tunneling particles flips to another bare state during the resonance

363: crossing, then the total field (the external sweep field plus the molecular

364: field) on all other particles jumps. For the ferromagnetic coupling, $J>0,$

365: the jump of the total field is positive, other particles are brought past

366: the resonance and lose their chance to flip, and thus transition probability

367: is suppressed. For the antiferromagnetic frustrating coupling, the jump of

368: the total field is negative and other particles are being set back before

369: the resonance and are getting one more chance to cross the resonance at some

370: larger field value and flip; Thus the transition probability for the system

371: should increase.

372: 

373: The influence of the interaction can be readily illustrated within a

374: rigorous many-body quantum language for our model if one considers the

375: energy levels of the system shown in Fig.\ \ref{Fig-lzn-En}. One can see

376: that instead of a single resonance at $W=0$ for individual or noninteracting

377: particles there is an interaction-split resonance that consists of many

378: avoided line crossings. These avoided line crossings are well separated from

379: each other on the plot if \TEXTsymbol{\vert}$J|\gg \Delta .$ The question

380: whether these well-separated crossings can be considered as a succession of

381: independent LZS transitions (i.e., whether they are \emph{dynamically} well

382: separated) depends on the sweep rate. Whereas for the slow sweep, $%

383: \varepsilon \gtrsim 1$ [see Eq.\ (\ref{PLZ})] the condition \TEXTsymbol{\vert%

384: }$J|\gg \Delta $ is sufficient, for the fast sweep, $\varepsilon \lesssim 1,$

385: a more stringent condition is required that follows from Eq.\ (16) of Ref.\ %

386: \onlinecite{garsch02prb}. The resulting combined condition for the

387: dynamically well-separated resonances thus would be

388: \begin{equation}

389: |J|\gg \left\{

390: \begin{array}{cc}

391: \Delta , & \varepsilon \gtrsim 1 \\

392: \sqrt{\hbar v}\sim \Delta /\varepsilon , & \varepsilon \lesssim 1.

393: \end{array}

394: \right.  \label{SepResCond}

395: \end{equation}

396: This is, however, only an \emph{apriori} estimation considering two

397: resonances. For $N\gg 1$ there are many resonances, and the deviations from

398: the single-resonance picture can accumulate with the increase of $N.$ One

399: can do an \emph{aposteriori} estimation in the slow-sweep limit where the

400: result for the well-separated resonances $P\rightarrow 1-1/N$ for $%

401: \varepsilon \rightarrow \infty $ following from Eq.\ (\ref{PNF}) as well as

402: the mean-field result for $N\gg 1$ and non-separated resonances [the first

403: line of Eq.\ (\ref{PhxLims})] are available, in both cases $1-P\ll 1$. The

404: crossover between these results should take place at \TEXTsymbol{\vert}$%

405: J|\sim $ $\Delta N^{1/2},$ thus the first line of Eq.\ (\ref{SepResCond})

406: should be replaced by

407: \begin{equation}

408: |J|\gg \Delta N^{1/2},\qquad \varepsilon \gtrsim 1.  \label{SepResCond2}

409: \end{equation}

410: The second line of Eq.\ (\ref{SepResCond}) could be also modified by $N$ but

411: it is difficult to derive an exact condition.

412: 

413: A perturbative analysis of the stationary Schr\"{o}dinger equation for our

414: model shows that tunnel level splitting of the bare levels with different

415: values of $k$ has the form\cite{gar91jpa}

416: \begin{equation}

417: \delta E_{k_{1},k_{2}}\sim \Delta \left( \frac{\Delta }{J}\right)

418: ^{|k_{1}-k_{2}|-1}.  \label{SplittDiffOrd}

419: \end{equation}

420: Thus for \TEXTsymbol{\vert}$J|\gg \Delta $ only the first-order or direct

421: transitions with $|k_{1}-k_{2}|=1$ should be taken into account whereas the

422: higher-order transitions are weak (see Fig.\ \ref{Fig-lzn-En}).

423: 

424: Now it becomes clear that for the ferromagnetic coupling and well-separated

425: resonances only the strong transition between the initial level $k=0$ and

426: the level $k=1$ happens. That is, in another language, only one particle of $%

427: N$ has a chance to tunnel, and the tunneling probability for the whole

428: system is strongly reduced. For a quantitative analysis one can neglect all $%

429: c_{k}$ with $k>1$ in Eq.\ (\ref{SchrEqc}) that yields the system of

430: equations

431: \begin{eqnarray}

432: i\hbar \dot{c}_{0} &=&E_{0}c_{0}-\frac{\Delta }{2}\sqrt{N}c_{1}  \nonumber \\

433: i\hbar \dot{c}_{1} &=&E_{1}c_{1}-\frac{\Delta }{2}\sqrt{N}c_{0}

434: \label{SchrEqcF}

435: \end{eqnarray}

436: that maps on the standard LZS problem with $\Delta \Rightarrow \Delta

437: _{N}\equiv \Delta \sqrt{N}$ and with the resonance at $W=2(N-1)J$ instead of

438: $W=0.$ The final-state probabilities for Eq.\ (\ref{SchrEqcF}) are according

439: to Eq.\ (\ref{PLZ})

440: \begin{equation}

441: p_{0}=P^{N}=e^{-N\varepsilon },\qquad p_{1}=1-P^{N}.  \label{p0p1F}

442: \end{equation}

443: Then with the help of Eq.\ (\ref{PtFinal}) the one-particle staying

444: probability for $J>0$ can be cast into the form

445: \begin{equation}

446: P_{N}=1-\frac{1}{N}\left( 1-P^{N}\right)  \label{PNF}

447: \end{equation}

448: and it varies between $1$ in the fast-sweep limit and $1-1/N$ in the

449: slow-sweep limit. Thus suppression of transitions by the ferromagnetic

450: coupling is extremely strong for a large number of particles $N.$ Expansion

451: of the transition probability of Eq.\ (\ref{PNF}) in the fast-sweep limit

452: reads

453: \begin{equation}

454: 1-P_{N}\cong \varepsilon -\frac{N}{2}\varepsilon ^{2}+O(\varepsilon

455: ^{3}),\qquad (\varepsilon \ll 1/N).  \label{PNFfast}

456: \end{equation}

457: Note that the first term of this expansion is insensitive to the interaction

458: (in this context to the number of interacting particles) and is the same as

459: for the standard LZS problem, $1-P=1-e^{-\varepsilon }\cong \varepsilon

460: -\varepsilon ^{2}/2!+\varepsilon ^{3}/3!-\ldots $

461: 

462: %TCIMACRO{

463: %\TeXButton{Fig-lzn-separatedresonances}{\begin{figure}[t]

464: %\unitlength1cm

465: %\begin{picture}(11,6)

466: %\centerline{\psfig{file=Fig-lzn-separatedresonances.eps,angle=-90,width=9cm}}

467: %\end{picture}

468: %\caption{ \label{Fig-lzn-separatedresonances}

469: %One-particle staying probability $P_N$ vs the sweep-rate parameter $\varepsilon$ for systems with FM and AFMF interaction in the case of well separated resonances.

470: %}

471: %\end{figure}%

472: %}}%

473: %BeginExpansion

474: \begin{figure}[t]

475: \unitlength1cm

476: \begin{picture}(11,6)

477: \centerline{\psfig{file=Fig-lzn-separatedresonances.eps,angle=-90,width=9cm}}

478: \end{picture}

479: \caption{ \label{Fig-lzn-separatedresonances}

480: One-particle staying probability $P_N$ vs the sweep-rate parameter $\varepsilon$ for systems with FM and AFMF interaction in the case of well separated resonances.

481: }

482: \end{figure}%

483: %

484: %EndExpansion

485: 

486: The AFMF coupling $J<0$ favors transitions, as can be seen from Fig.\ \ref

487: {Fig-lzn-En}b. The original resonance is splitted by the interaction into $N$

488: resonances filling equidistantly the range $-2(N-1)J\leq W\leq 2(N-1)J.$ One

489: can see that in the limit of slow sweep the system remains on the lowest

490: adiabatic energy level thus $P_{N}\rightarrow 0,$ in contrast to the

491: ferromagnetic case. For well-separated resonances the problem described by

492: Eq.\ (\ref{SchrEqc}) splits into independent standard LZS problems for the

493: resonances between the levels $k$ and $k+1$ that are described by an

494: effective splitting $\Delta _{N,k}=\Delta l_{k,k+1}$ [cf. Eq.\ (\ref

495: {SchrEqcF})]. It is convenient to designate the probability to stay in state

496: $k$ after crossing with state $k-1$ but before crossing with state $k+1$

497: (see Fig.\ \ref{Fig-lzn-En}b) as $p_{k}($\textrm{mid}$)$. Then solutions of

498: the LZS problems for individual crossings along with conditions of the

499: probability conservation can be written as

500: \begin{eqnarray}

501: p_{k}(\infty ) &=&P^{(N-k)(k+1)}p_{k}(\text{\textrm{mid}})  \nonumber \\

502: p_{k}(\text{\textrm{mid}}) &=&p_{k-1}(\text{\textrm{mid}})-p_{k-1}(\infty )

503: \nonumber \\

504: p_{k-1}(\infty ) &=&P^{(N-k+1)k}p_{k-1}(\text{\textrm{mid}}),  \label{EqsAF}

505: \end{eqnarray}

506: etc. These can be combined into the recurrence relation

507: \begin{eqnarray}

508: p_{k}(\infty ) &=&P^{(N-k)(k+1)}\left( P^{-(N-k+1)k}-1\right) p_{k-1}(\infty

509: )  \nonumber \\

510: &=&P^{N-2k}\left( 1-P^{(N-k+1)k}\right) p_{k-1}(\infty )  \label{RecurrAF}

511: \end{eqnarray}

512: that has to be iterated with the initial condition $p_{0}(\infty )=P^{N}.$

513: In the slow-sweep limit $\varepsilon \gg 1/N,$ one has $P^{N}\ll 1,$ thus

514: one can drop the factor $\left( 1-P^{(N-k+1)k}\right) \cong 1$ in the second

515: line of Eq.\ (\ref{RecurrAF}) that after iteration results in $p_{k}\cong

516: P^{(N-k)(k+1)}.$ In this case, in Eq.\ (\ref{PtFinal}) $p_{0}\cong

517: p_{N-1}\cong P^{N}$ are dominant for large $N$ whereas all other summands

518: are much smaller. This yields

519: \begin{equation}

520: P_{N}\cong \left( 1+\frac{1}{N}\right) P^{N}\ll 1,\qquad \left( \varepsilon

521: \gg 1/N\right)  \label{PNAFslow}

522: \end{equation}

523: for the antiferromagnetic coupling. In the fast-sweep limit the solution of

524: Eq.\ (\ref{RecurrAF}) can be expanded in powers of $\varepsilon ,$ after

525: which Eq.\ (\ref{PtFinal}) yields for the transition probability

526: \begin{equation}

527: 1-P_{N}\cong \varepsilon +\left( \frac{3N}{2}-2\right) \varepsilon

528: ^{2}+O(\varepsilon ^{3}),\quad (\varepsilon \ll 1/N).  \label{PNAFfast}

529: \end{equation}

530: Again, the first term of this expansion is the same as in the standard LZS

531: problem. For intermediate sweep rates one can iterate Eq.\ (\ref{RecurrAF})

532: numerically and substitute the solution for $p_{k}$ into Eq.\ (\ref{PtFinal}%

533: ). The resulting curves $P_{N}$ vs $\varepsilon $ are shown in Fig.\ \ref

534: {Fig-lzn-separatedresonances} along with those for the ferromagnetic

535: coupling, Eq.\ (\ref{PNF}).

536: 

537: To conclude this section, we show the numerically computed dependences of $%

538: P_{N}$ on $J/\Delta $ for $N=3$ and different $\varepsilon $ in Fig.\ \ref

539: {Fig-lzn-PvsJ}. The case of well-separated resonances that was considered in

540: this section corresponds to the plateaus on the left and on the right sides

541: of the plot.

542: 

543: %TCIMACRO{

544: %\TeXButton{Fig-lzn-PvsJ}{\begin{figure}[t]

545: %\unitlength1cm

546: %\begin{picture}(11,6)

547: %\centerline{\psfig{file=Fig-lzn-PvsJ.eps,angle=-90,width=9cm}}

548: %\end{picture}

549: %\caption{ \label{Fig-lzn-PvsJ}

550: %Numerically computed staying probability $P_{N}$ vs $J/\Delta $ for $N=3$ and different sweep rate parameters $\varepsilon $.

551: %The horizontal dashed lines on the left side of the plot are asymptotes corresponding to well-separated resonances.

552: %}

553: %\end{figure}%

554: %}}%

555: %BeginExpansion

556: \begin{figure}[t]

557: \unitlength1cm

558: \begin{picture}(11,6)

559: \centerline{\psfig{file=Fig-lzn-PvsJ.eps,angle=-90,width=9cm}}

560: \end{picture}

561: \caption{ \label{Fig-lzn-PvsJ}

562: Numerically computed staying probability $P_{N}$ vs $J/\Delta $ for $N=3$ and different sweep rate parameters $\varepsilon $.

563: The horizontal dashed lines on the left side of the plot are asymptotes corresponding to well-separated resonances.

564: }

565: \end{figure}%

566: %

567: %EndExpansion

568: 

569: \section{Fast sweep and weak coupling}

570: 

571: \label{Sec-fast-sweep}

572: 

573: In the preceding section we have considered the strong-coupling limit

574: \TEXTsymbol{\vert}$J|\gg \max \left( \Delta N^{1/2},\Delta /\sqrt{%

575: \varepsilon }\right) $ in which individual resonances are well separated

576: from each other and one thus deals with successive standard LZS transitions.

577: The opposite limiting case is that of the weak coupling $J.$ This case can

578: be solved by a perturbative expansion in $J$ with the standard LZS solution

579: as the zeroth-order appriximation. The situation further simplifies in the

580: fast-sweep limit, where the transition probability is small and the

581: coefficients $c_{k}$ in Eq.\ (\ref{SchrEqc}) decrease with $k$ as powers of $%

582: \varepsilon \ll 1.$ In particular, $c_{0}$ contains terms of orders $%

583: \varepsilon ^{0},\varepsilon ^{1},$ and $\varepsilon ^{2},$ the coefficient $%

584: c_{1}$ contains $\varepsilon ^{1/2}$ and $\varepsilon ^{3/2},$ the

585: coefficient $c_{2}$ starts with $\varepsilon ,$ etc. Knowing these

586: contributions into $c_{0},$ $c_{1},$ and $c_{2}$ is sufficient to set up the

587: expansion of $P_{N}$ up to the first nontrivial order $\varepsilon ^{2}.$ A

588: straighforward but cumbersome calculation yields

589: \begin{eqnarray}

590: 1-P_{N} &\cong &\varepsilon -\left( \frac{1}{2}+\frac{4J_{0}}{\sqrt{2\pi

591: \hbar v}}\right) \varepsilon ^{2}+O(\varepsilon ^{3})  \nonumber \\

592: &=&\varepsilon -\frac{1}{2}\varepsilon ^{2}-\frac{4J_{0}}{\Delta }%

593: \varepsilon ^{5/2}+\ldots  \label{PNSmallJfast}

594: \end{eqnarray}

595: where we have defined

596: \begin{equation}

597: J_{0}\equiv (N-1)J.  \label{J0Def}

598: \end{equation}

599: One can see, again that ferromagnetic interaction suppresses transitions

600: whereas the AFMF interaction facilitates transitions, and that the effect of

601: interaction is increased by the number particles in the system. In fact,

602: however, in our model the interaction should scale, on physical grounds,

603: with the system's size, i.e., $J_{0}\equiv $ $(N-1)J$ should be size

604: independent. Note that in the fast-sweep limit one could do the expansion in

605: $\varepsilon $ for arbitrary $J.$ This leads, however, to cumbersome

606: multiple integrals. The strong-coupling limit for the fast sweep $%

607: \varepsilon \ll 1$ has been considered above, and the corresponding

608: counterparts of Eq.\ (\ref{PNSmallJfast}) for ferro-and antiferromagnetic

609: coupling are Eqs.\ (\ref{PNFfast}) and (\ref{PNAFfast}).

610: 

611: It should be noted that although Eq.\ (\ref{PNSmallJfast}) is valid for $%

612: \varepsilon \ll 1,$ the method of its derivation above is only justified for

613: $\varepsilon \ll 1/N,$ i.e. for much faster sweep rates, if $N\gg 1.$

614: Indeed, in the course of the derivation the terms of orders $\left(

615: \varepsilon N\right) ^{1/2},$ $\varepsilon N,$ etc., appear that are

616: required to be small. Only at the very end the leading $N$ contributions

617: cancel each other that leads to Eq.\ (\ref{PNSmallJfast}) that fortunately

618: has a larger applicability range. In fact, for large systems with weak

619: interaction the most populated final states are

620: \begin{equation}

621: k_{\max }=N(1-P)\cong N\varepsilon ,  \label{kmaxDef}

622: \end{equation}

623: so that keeping only the states with $k=0,1,2$ above was not justified for $%

624: N\varepsilon \gtrsim 1.$ Eq.\ (\ref{kmaxDef}) can be easily derived for an

625: assembly of noninteracting tunneling species. To this end, we use the

626: coefficients $C_{k}$ of Eq.\ (\ref{SchrEq}) and represent them in the

627: factorized form

628: \begin{equation}

629: C_{k}=a_{-1}^{N-k}a_{1}^{k},  \label{CFactorized}

630: \end{equation}

631: where $a_{-1}$ and $a_{1}$ are the coefficients of the wave function of the

632: one-particle problem, $\psi =a_{-1}\psi _{-1}+a_{1}\psi _{1}.$ This yields

633: \begin{equation}

634: |C_{k}|^{2}=|a_{-1}|^{2\left( N-k\right) }|a_{1}|^{2k}=P^{N-k}(1-P)^{k}

635: \label{C2Factorized}

636: \end{equation}

637: that can be plugged into Eq.\ (\ref{PtDefk}) to give

638: \begin{equation}

639: P_{N}=\sum_{k=0}^{N-1}\frac{(N-1)!}{(N-1-k)!k!}P^{N-k}(1-P)^{k}.

640: \label{PNNoninteracting}

641: \end{equation}

642: The summand in this formula has its maximum at $k=k_{\max }$ given by Eq.\ (%

643: \ref{kmaxDef}) and it becomes a narrow Gaussian packet around $k_{\max }$\

644: for $N\gg 1.$

645: 

646: A more rigorous method of handling the fast-sweep and weak coupling limits

647: uses the one-particle density matrix that is defined by

648: \begin{equation}

649: \rho _{-1,-1}=P_{N},\qquad \rho _{1,1}=1-\rho _{-1,-1}=1-P_{N}

650: \label{rhodiagDef}

651: \end{equation}

652: with $P_{N}$ of Eq.\ (\ref{PtDefk}) and

653: \begin{eqnarray}

654: \rho _{1,-1} &=&\sum_{m_{2},\ldots ,m_{N}=-1,1}C_{1,m_{2},\ldots

655: ,m_{N}}^{\ast }C_{-1,m_{2},\ldots ,m_{N}}  \nonumber \\

656: &=&\sum_{k=0}^{N-1}\frac{(N-1)!}{(N-1-k)!k!}C_{k+1}^{\ast }C_{k}  \nonumber

657: \\

658: \rho _{-1,1} &=&\rho _{1,-1}^{\ast }.  \label{rhonondiagDef}

659: \end{eqnarray}

660: The density-matrix equation (DME) can be obtained by time differentiating of

661: $\rho _{mn}$ and employing Eq.\ (\ref{SchrEq}) that yields

662: \begin{eqnarray}

663: &&\hbar \dot{\rho}_{-1,-1}=\frac{i\Delta }{2}\left( \rho _{-1,1}-\rho

664: _{1,-1}\right) =\Delta \func{Im}\rho _{1,-1}  \nonumber \\

665: &&\hbar \dot{\rho}_{1,-1}=-iW(t)\rho _{1,-1}+\frac{i\Delta }{2}\left( \rho

666: _{1,1}-\rho _{-1,-1}\right)  \nonumber \\

667: &&{}+2iJ\sum_{k=0}^{N-1}\frac{(N-1)!}{(N-1-k)!k!}\left( N-2k-1\right)

668: C_{k}C_{k+1}^{\ast }.  \label{DME}

669: \end{eqnarray}

670: In the absence of interaction, $J=0,$ the last term in $\dot{\rho}_{-1,1}$

671: disappears and one obtains a DME for one isolated particle that is

672: equivalent to the one-particle Schr\"{o}dinger equation. Note that solving

673: this one-particle equation in the fast-sweep limit requires $\varepsilon \ll

674: 1,$ in contrast to Eq.\ (\ref{SchrEq}) with $J=0$ that requires $\varepsilon

675: \ll 1/N.$

676: 

677: For $J\neq 0$ Eqs.\ (\ref{DME}) do not form a closed system of equations. In

678: this case Eqs.\ (\ref{DME}) can only be useful if an approximation for the

679: interaction term be found. In particular, one can consider the interaction

680: term perturbatively in $J$ and use

681: \begin{equation}

682: \rho _{mn}=\rho _{mn}^{(0)}+\delta \rho _{mn}  \label{rhoExpansion}

683: \end{equation}

684: where $\rho _{mn}^{(0)}$ is the density matrix without interaction that

685: satisfies

686: \begin{eqnarray}

687: &&\hbar \dot{\rho}_{-1,-1}^{(0)}=\Delta \func{Im}\rho _{1,-1}^{(0)}

688: \nonumber \\

689: &&\hbar \dot{\rho}_{1,-1}^{(\text{0)}}=-iW(t)\rho _{1,-1}^{(0)}+i\Delta

690: \left( \frac{1}{2}-\rho _{-1,-1}^{(0)}\right)  \label{DME0}

691: \end{eqnarray}

692: and $\delta \rho _{mn}$ is the correction term. The latter satisfies

693: equations similar to Eqs.\ (\ref{DME}) in which, however, $%

694: C_{k}C_{k+1}^{\ast }$ is replaced by its noninteracting-particle expression

695: following from Eq.\ (\ref{CFactorized}),

696: \begin{eqnarray}

697: &&C_{k}C_{k+1}^{\ast }=a_{-1}^{N-k}a_{1}^{k}\left( a_{-1}^{\ast }\right)

698: ^{N-k-1}\left( a_{1}^{\ast }\right) ^{k+1}  \nonumber \\

699: &&{}\qquad =|a_{-1}|^{2(N-k-1)}|a_{1}|^{2k}a_{-1}a_{1}^{\ast }  \nonumber \\

700: &&\qquad =\left( \rho _{-1,-1}^{(0)}\right) ^{N-k-1}\left( 1-\rho

701: _{-1,-1}^{(0)}\right) ^{k}\rho _{-1,1}^{(0)}.  \label{CkCkp1}

702: \end{eqnarray}

703: With the use of the latter, the sum over $k$ in Eqs.\ (\ref{DME}) can be

704: performed to yield

705: \begin{eqnarray}

706: &&\hbar \delta \dot{\rho}_{-1,-1}=\Delta \func{Im}\delta \rho _{1,-1}

707: \nonumber \\

708: &&\hbar \delta \dot{\rho}_{1,-1}=-iW(t)\delta \rho _{-1,1}-i\Delta \delta

709: \rho _{-1,-1}  \nonumber \\

710: &&{}\qquad -2iJ_{0}\left( 1-2\rho _{-1,-1}^{(0)}\right) \rho _{1,-1}^{(0)}.

711: \label{DMEdelta}

712: \end{eqnarray}

713: Solving Eqs.\ (\ref{DME0}) and (\ref{DMEdelta}) perturbatively for $%

714: \varepsilon \ll 1$ results in Eq.\ (\ref{PNSmallJfast}).

715: 

716: In accord with the remark at the beginning of this section, $J$-perturbative

717: Eqs.\ (\ref{DME0}) and (\ref{DMEdelta}) can be solved in terms of the

718: hypergeometric functions for any $\varepsilon $ (For the standard LZS

719: problem, $J=0,$ this was done by Zener. \cite{zen32}) It can be shown that

720: in the slow-sweep limit this solution simplifies to

721: \begin{equation}

722: \left. \frac{dP}{d\rho }\right| _{\rho =0}=c\varepsilon e^{-\varepsilon

723: },\qquad \varepsilon \gg 1,\qquad \rho \equiv \frac{2J_{0}}{\Delta },

724: \label{dPdrhoslowsweep}

725: \end{equation}

726: where $c$ is a number and the parameter $\rho $ should not be confused with

727: the density matrix.

728: 

729: %TCIMACRO{

730: %\TeXButton{Fig-lzn-Pvsrho}{\begin{figure}[t]

731: %\unitlength1cm

732: %\begin{picture}(11,6)

733: %\centerline{\psfig{file=Fig-lzn-Pvsrho.eps,angle=-90,width=9cm}}

734: %\end{picture}

735: %\caption{ \label{Fig-lzn-Pvsrho}

736: %Numerically computed staying probability $P_{N}$ vs $\rho\equiv 2J_0/\Delta $ for $\varepsilon=1$ and different system sizes $N$, including the mean-field result for $N\to\infty$.}

737: %\end{figure}%

738: %}}%

739: %BeginExpansion

740: \begin{figure}[t]

741: \unitlength1cm

742: \begin{picture}(11,6)

743: \centerline{\psfig{file=Fig-lzn-Pvsrho.eps,angle=-90,width=9cm}}

744: \end{picture}

745: \caption{ \label{Fig-lzn-Pvsrho}

746: Numerically computed staying probability $P_{N}$ vs $\rho\equiv 2J_0/\Delta $ for $\varepsilon=1$ and different system sizes $N$, including the mean-field result for $N\to\infty$.}

747: \end{figure}%

748: %

749: %EndExpansion

750: 

751: \section{The mean field approximation}

752: 

753: \label{Sec-mfa}

754: 

755: If each of tunneling particles interacts with all the other particles with a

756: coupling of the same sign, as is the case in our model for $N\gg 1,$ then

757: the mean-field approximation (MFA) should work well. The MFA employs a

758: one-particle density-matrix equation or a one-particle Schr\"{o}dinger

759: equation in which the interaction enters via the molecular field. For the

760: model of Eq.\ (\ref{Ham}) one has $W(t)\Rightarrow W_{\mathrm{eff}}(t)$

761: where

762: \begin{equation}

763: W_{\mathrm{eff}}(t)=W(t)+2J_{0}(1-2\rho _{-1,-1})  \label{WeffDef}

764: \end{equation}

765: in the DME of Eq.\ (\ref{DME0}), i.e.,

766: \begin{eqnarray}

767: &&\hbar \dot{\rho}_{-1,-1}=\Delta \func{Im}\rho _{1,-1}  \nonumber \\

768: &&\hbar \dot{\rho}_{1,-1}=-iW_{\mathrm{eff}}(t)\rho _{1,-1}+i\Delta \left(

769: \frac{1}{2}-\rho _{-1,-1}\right) .  \label{DMEMFA}

770: \end{eqnarray}

771: This is equivalent to the nonlinear Schr\"{o}dinger equation

772: \begin{eqnarray}

773: i\hbar \dot{C}_{-1} &=&\frac{1}{2}W_{\mathrm{eff}}(t)C_{-1}-\frac{\Delta }{2}%

774: C_{1}  \nonumber \\

775: i\hbar \dot{C}_{1} &=&-\frac{1}{2}W_{\mathrm{eff}}(t)C_{1}-\frac{\Delta }{2}%

776: C_{-1}  \label{NonlSchEq}

777: \end{eqnarray}

778: with $W_{\mathrm{eff}}(t)=W(t)+2J_{0}(1-2|C_{-1}|^{2}).$ It is interesting

779: to note that if one solves Eq.\ (\ref{DMEMFA}) perturbatively in $J$ \ using

780: Eq.\ (\ref{rhoExpansion}) one obtains Eqs.\ (\ref{DME0}) and (\ref{DMEdelta}%

781: ). This implies that the MFA works well in the weak-coupling limit in our

782: model (see Fig.\ \ref{Fig-lzn-Pvsrho}).

783: 

784: In the case $N\gg 1$ one can assume that the state of the system is

785: described by a narrow packet in the $k$-space, as is indeed the case for

786: noninteracting particles, see Eq.\ (\ref{PNNoninteracting}). Then in Eq.\ (%

787: \ref{DME}) one can replace $\left( N-2k-1\right) \Rightarrow \left(

788: N-2k_{\max }-1\right) $ \ with $k_{\max }$ of Eq.\ (\ref{kmaxDef}), after

789: which the sum over $k$ converts to $\rho _{-1,1}$ according to the

790: definition in Eq.\ (\ref{rhonondiagDef}). This leads to Eq.\ (\ref{DMEMFA}).

791: One should, however, realize that this is no more than a heuristic

792: derivation since it was not proved that the state of a system \emph{with

793: interaction} is described by a narrow packet for $N\gg 1.$ On the other

794: hand, the narrow-packet assumption is quite plausible since the limit $N\gg

795: 1 $ corresponds to the quasiclassical limit $S\gg 1$ of the equivalent model

796: of Eq.\ (\ref{HamS}) and the states of quasiclassical systems should be

797: narrow packets. It is remarkable that the density-matrix equation within the

798: MFA, Eq.\ (\ref{DMEMFA}), can be interpreted as a purely classical equation

799: of motion for a spin vector, the Landau-Lifschitz equation (LLE). Indeed,

800: rewriting Eq.\ (\ref{rhonondiagDef}) in terms of $c_{k}$ with the help of

801: Eq.\ (\ref{ckDef}) and adding Eq.\ (\ref{PtMapping}) one obtains the mapping

802: relations

803: \begin{eqnarray}

804: &&\rho _{1,-1}=\frac{1}{N}\sum_{k=0}^{N}l_{k,k+1}c_{k+1}^{\ast }c_{k}=\frac{%

805: \langle S_{+}\rangle }{2S}  \nonumber \\

806: &&\rho _{-1,-1}=\frac{1}{2}\left( 1-\frac{\langle S_{z}\rangle }{S}\right) .

807: \label{rhoSMapping}

808: \end{eqnarray}

809: Now one can see that the DME in the MFA, Eq.\ (\ref{DMEMFA}) is equivalent

810: to the LLE for the classical spin vector components

811: \begin{equation}

812: s_{z}=\frac{\langle S_{z}\rangle }{S},\qquad s_{+}=\frac{\langle

813: S_{+}\rangle }{S}  \label{sSavrIdent}

814: \end{equation}

815: that reads

816: \begin{eqnarray}

817: &&\hbar \dot{s}_{z}=-\Delta s_{y}  \nonumber \\

818: &&\hbar \dot{s}_{+}=-i\left[ W(t)+4J(S-1/2)s_{z}\right] s_{+}+i\Delta s_{z}.

819: \label{LLEzp}

820: \end{eqnarray}

821: The vector form of this LLE is

822: \begin{equation}

823: \mathbf{\dot{s}}=\gamma \left[ \mathbf{s\times H}_{\mathrm{eff}}\right]

824: ,\qquad \mathbf{H}_{\mathrm{eff}}=-\frac{\partial \mathcal{H}_{\mathrm{eff}}%

825: }{\partial \mathbf{s}},  \label{LLE}

826: \end{equation}

827: where $\gamma =1/\hbar $ and the effective classical energy is given by

828: \begin{equation}

829: \mathcal{H}_{\mathrm{eff}}=-H_{z}(t)s_{z}-H_{x}s_{x}-D_{\mathrm{cl}}s_{z}^{2}

830: \label{HamClass}

831: \end{equation}

832: with

833: \begin{eqnarray}

834: &&H_{z}(t)=W(t),\qquad H_{x}=\Delta ,\qquad  \nonumber \\

835: &&D_{\mathrm{cl}}=D(S-1/2)=J(N-1)\equiv J_{0}  \label{ParIdentClass}

836: \end{eqnarray}

837: [cf. Eqs.\ (\ref{HamS}) and (\ref{ParIdent})]. The one-particle staying

838: probability $P_{N}$ is given by the formula

839: \begin{equation}

840: P_{N}(t)=\frac{1}{2}\left[ 1-s_{z}(t)\right]  \label{PtMappingClass}

841: \end{equation}

842: that is analogous to Eq.\ (\ref{PtMapping}).

843: 

844: %TCIMACRO{

845: %\TeXButton{Fig-lzn-PMFA-AF}{\begin{figure}[t]

846: %\unitlength1cm

847: %\begin{picture}(11,6)

848: %\centerline{\psfig{file=Fig-lzn-PMFA-AF.eps,angle=-90,width=9cm}}

849: %\end{picture}

850: %\caption{ \label{Fig-lzn-PMFA-AF}

851: %The mean-field solution for the one-particle staying probability $P_N$ for the AFMF coupling, $J_0\equiv (N-1)J<0$.

852: %}

853: %\end{figure}%

854: %}}%

855: %BeginExpansion

856: \begin{figure}[t]

857: \unitlength1cm

858: \begin{picture}(11,6)

859: \centerline{\psfig{file=Fig-lzn-PMFA-AF.eps,angle=-90,width=9cm}}

860: \end{picture}

861: \caption{ \label{Fig-lzn-PMFA-AF}

862: The mean-field solution for the one-particle staying probability $P_N$ for the AFMF coupling, $J_0\equiv (N-1)J<0$.

863: }

864: \end{figure}%

865: %

866: %EndExpansion

867: 

868: %TCIMACRO{

869: %\TeXButton{Fig-lzn-PWeffvst-AF-MFA-j10m}{\begin{figure}[t]

870: %\unitlength1cm

871: %\begin{picture}(11,6)

872: %\centerline{\psfig{file=Fig-lzn-Pvst-AF-MFA-j10m.eps,angle=-90,width=9cm}}

873: %\end{picture}

874: %\begin{picture}(11,6)

875: %\centerline{\psfig{file=Fig-lzn-Weffvst-AF-MFA-j10m.eps,angle=-90,width=9cm}}

876: %\end{picture}

877: %\caption{ \label{Fig-lzn-PWeffvst-AF-MFA-j10m}

878: %$a$ -- Time dependence of the staying probability $P_N$ for the AFMF coupling

879: %$J_0\equiv (N-1)J=-10$ for the first three complete conversion cases,

880: %$P_N(\infty)=0$ (see Fig.\ \protect\ref{Fig-lzn-PMFA-AF}).

881: %$b$ -- Same for the effective sweep $W_{\rm eff}$ of Eq.\ (\protect\ref{WeffDef}).

882: %}

883: %\end{figure}%

884: %}}%

885: %BeginExpansion

886: \begin{figure}[t]

887: \unitlength1cm

888: \begin{picture}(11,6)

889: \centerline{\psfig{file=Fig-lzn-Pvst-AF-MFA-j10m.eps,angle=-90,width=9cm}}

890: \end{picture}

891: \begin{picture}(11,6)

892: \centerline{\psfig{file=Fig-lzn-Weffvst-AF-MFA-j10m.eps,angle=-90,width=9cm}}

893: \end{picture}

894: \caption{ \label{Fig-lzn-PWeffvst-AF-MFA-j10m}

895: $a$ -- Time dependence of the staying probability $P_N$ for the AFMF coupling

896: $J_0\equiv (N-1)J=-10$ for the first three complete conversion cases,

897: $P_N(\infty)=0$ (see Fig.\ \protect\ref{Fig-lzn-PMFA-AF}).

898: $b$ -- Same for the effective sweep $W_{\rm eff}$ of Eq.\ (\protect\ref{WeffDef}).

899: }

900: \end{figure}%

901: %

902: %EndExpansion

903: 

904: Let us now analyze the mean-field solution of the LZS problem based on Eqs.\

905: (\ref{DMEMFA}) or (\ref{LLE}) and compare it with the exact solution of Eq.\

906: (\ref{SchrEqc}). We start with the AFMF coupling, for convenience. Fig.\ \ref

907: {Fig-lzn-PMFA-AF} shows the dependence of the one-particle staying

908: probability $P_{N}$ on the sweep-rate parameter $\varepsilon $ for different

909: coupling strengths $J_{0}.$ The curves on the plot are qualitatively similar

910: to those in the case of well-separated resonances, Fig.\ \ref

911: {Fig-lzn-separatedresonances} and they show that AFMF interaction boosts

912: transitions, especially for slow sweep, $\varepsilon \gg 1.$ The difference

913: is that in the MFA only the combined parameter $J_{0}=(N-1)J$ enters,

914: whereas for well-separated resonances the results depend on $N$ only and not

915: on $J.$ Another difference is that for strong enough interaction the

916: mean-field solution for $P_{N}$ oscillates and vanishes at some values of $%

917: \varepsilon .$ Oscillations of this kind have been found in the solution of

918: the LZS problem with nonlinear sweep.\cite{ternak97,garsch02prb} As was

919: explained in Ref.\ \onlinecite{garsch02prb}, oscillations arise when the

920: sweep slows down in the vicinity of the resonance so that (in the extreme

921: case) the particle oscillates between the two quantum states and the

922: resulting staying probability depends on the phase of this oscillation at

923: the end of the stay near the resonance. This is also the case for the model

924: with AFMF interaction in the MFA.Tunneling of particles creates a molecular

925: field that changes in the direction opposite to the sweep so that the

926: resulting effective sweep$\ W_{\mathrm{eff}}(t)$ of Eq.\ (\ref{WeffDef}) can

927: even become nonmonotonic (in this case the analytical method of Ref.\ %

928: \onlinecite{hamraemiysai00} based on the linearization of the problem near

929: the resonance and introducing an effective sweep rate breaks down). The MFA

930: solution becomes time symmetric in the case of complete conversion, $%

931: P_{N}(\infty )=0$ that for $J_{0}/\Delta =-10$ is realized for $\varepsilon

932: =0.1137983$, 0.13207, 0.1741, 0.2233, etc. The corresponding time

933: dependences of $P_{N}(t)$ and $W_{\mathrm{eff}}(t)$ are shown in Fig.\ (\ref

934: {Fig-lzn-PWeffvst-AF-MFA-j10m}). Although they resemble the complete

935: conversion solutions found in Ref.\ \onlinecite{garsch02}, there are no

936: apparent analytical solutions for this model. The difficulty of the present

937: mean-field model is that $W_{\mathrm{eff}}(t)$ is not known from the

938: beginning but is a solution of a self-consistent nonlinear problem.

939: 

940: Comparison of the exact and the MFA solutions for the one-particle staying

941: probability $P_{N}$ for the antiferromagnetic coupling $J_{0}=-10$ is shown

942: in Fig.\ \ref{Fig-lzn-P-AF-j10}. One can see that the exact quantum solution

943: converges to the classical mean-field solution for $N\rightarrow \infty $,

944: although in the vicinity of the first complete-conversion point convergence

945: is extremely slow.

946: 

947: %TCIMACRO{

948: %\TeXButton{Fig-lzn-P-AF-j10}{\begin{figure}[t]

949: %\unitlength1cm

950: %\begin{picture}(11,6)

951: %\centerline{\psfig{file=Fig-lzn-P-AF-j10.eps,angle=-90,width=9cm}}

952: %\end{picture}

953: %\caption{ \label{Fig-lzn-P-AF-j10}

954: %Comparison of the exact quantum and the MFA solutions for the one-particle staying probability $P_N$ for the AFMF coupling

955: %showing partially slow convergence to the classical mean-field limit for $N\to\infty$.

956: %}

957: %\end{figure}%

958: %}}%

959: %BeginExpansion

960: \begin{figure}[t]

961: \unitlength1cm

962: \begin{picture}(11,6)

963: \centerline{\psfig{file=Fig-lzn-P-AF-j10.eps,angle=-90,width=9cm}}

964: \end{picture}

965: \caption{ \label{Fig-lzn-P-AF-j10}

966: Comparison of the exact quantum and the MFA solutions for the one-particle staying probability $P_N$ for the AFMF coupling

967: showing partially slow convergence to the classical mean-field limit for $N\to\infty$.

968: }

969: \end{figure}%

970: %

971: %EndExpansion

972: 

973: %TCIMACRO{

974: %\TeXButton{Fig-lzn-PMFA-F}{\begin{figure}[t]

975: %\unitlength1cm

976: %\begin{picture}(11,6)

977: %\centerline{\psfig{file=Fig-lzn-PMFA-F.eps,angle=-90,width=9cm}}

978: %\end{picture}

979: %\caption{ \label{Fig-lzn-PMFA-F}

980: %The mean-field solution for the one-particle staying probability $P_N$ for the ferromagnetic coupling, $J_0\equiv (N-1)J>0$.

981: %The dashed line is the large-$\varepsilon$ asymptote of Eq.\ (\protect\ref{PN0alSmall}).

982: %}

983: %\end{figure}%

984: %}}%

985: %BeginExpansion

986: \begin{figure}[t]

987: \unitlength1cm

988: \begin{picture}(11,6)

989: \centerline{\psfig{file=Fig-lzn-PMFA-F.eps,angle=-90,width=9cm}}

990: \end{picture}

991: \caption{ \label{Fig-lzn-PMFA-F}

992: The mean-field solution for the one-particle staying probability $P_N$ for the ferromagnetic coupling, $J_0\equiv (N-1)J>0$.

993: The dashed line is the large-$\varepsilon$ asymptote of Eq.\ (\protect\ref{PN0alSmall}).

994: }

995: \end{figure}%

996: %

997: %EndExpansion

998: 

999: %TCIMACRO{

1000: %\TeXButton{Fig-lzn-P-F-j3}{\begin{figure}[t]

1001: %\unitlength1cm

1002: %\begin{picture}(11,6)

1003: %\centerline{\psfig{file=Fig-lzn-P-F-j3.eps,angle=-90,width=9cm}}

1004: %\end{picture}

1005: %\caption{ \label{Fig-lzn-P-F-j3}

1006: %Comparison of the exact quantum and the MFA solutions for the one-particle staying probability $P_N$ for the ferromagnetic coupling.

1007: %}

1008: %\end{figure}%

1009: %}}%

1010: %BeginExpansion

1011: \begin{figure}[t]

1012: \unitlength1cm

1013: \begin{picture}(11,6)

1014: \centerline{\psfig{file=Fig-lzn-P-F-j3.eps,angle=-90,width=9cm}}

1015: \end{picture}

1016: \caption{ \label{Fig-lzn-P-F-j3}

1017: Comparison of the exact quantum and the MFA solutions for the one-particle staying probability $P_N$ for the ferromagnetic coupling.

1018: }

1019: \end{figure}%

1020: %

1021: %EndExpansion

1022: 

1023: Let us now turn to the ferromagnetic interaction. The mean-field solution

1024: for different interaction strengths is shown in Fig.\ \ref{Fig-lzn-PMFA-F}.

1025: Again, qualitatively it is similar to the solution in the limit of

1026: well-separated resonances, Eq.\ (\ref{PNF}) that is shown in Fig.\ \ref

1027: {Fig-lzn-separatedresonances}. The difference is the same as for the AFMF

1028: coupling: The MFA solution depends on $J_{0}\equiv (N-1)J$ whereas Eq.\ (\ref

1029: {PNF}) depends on $N$ only. Another difference is that in the slow-sweep

1030: limit the mean-field curve tends to a constant (for $J_{0}/\Delta >1/2)$

1031: whereas all quantum curves tend to zero. The latter are in fact combinations

1032: of several LZS exponentials corresponding to transitions of different orders

1033: shown in Fig.\ \ref{Fig-lzn-En}a by dotted arrows (see also Fig.\ \ref

1034: {Fig-lzn-P3Ft}). One can see, in particular, that the curve $N=2$ consist of

1035: two different exponentials, the curve $N=3$ consist of three different

1036: exponentials, etc. With increasing of $N,$ however, the coupling $%

1037: J=J_{0}/(N-1)$ decreases and transitions become not well separated. In this

1038: case dependence $P_{N}$ of $\varepsilon $ becomes a long-tale curve.

1039: 

1040: \section{Slow sweep in the MFA}

1041: 

1042: \label{Sec-slow-MFA}

1043: 

1044: Here we analytically consider the slow-sweep limit within the mean-field

1045: approximation, to get more insights into the mechanism that leads to

1046: suppressing transitions in the case of FM interactions (see Fig.\ \ref

1047: {Fig-lzn-PMFA-F}) as well as into that of facilitating transitions and

1048: probabillity oscillations in the case of the antiferromagnetic frustrating

1049: interactions (see Fig.\ \ref{Fig-lzn-PMFA-AF}).

1050: 

1051: \subsection{Adiabatic case: Basic equations}

1052: 

1053: For the one-particle LZS problem, a convenient method of treating the

1054: slow-sweep limit is that using the adiabatic basis.\cite{garsch02prb} Its

1055: advantage is that the probability to stay on the lowest adiabatic level is

1056: at any time close to 1, so that one can make an appropriate approximation

1057: after which the solution is obtained as a quadrature, including the case of

1058: nonlinear sweep. Direct extention of this method for the present model

1059: within the MFA is cumbersome, however, owing to the complicated

1060: self-consistent nature of the adiabatic energy levels for a \emph{nonlinear}

1061: Schr\"{o}dinger equation or \emph{nonlinear} density-matrix equation, Eq.\ (%

1062: \ref{DME}). Fortunately, an alternative description based on the LLE, Eq.\ (%

1063: \ref{LLE}), supports a physically transparent extension of the method. The

1064: adiabatic basis for a Schr\"{o}dinger equation corresponds to the adiabatic

1065: coordinate system for the classical problem of Eq.\ (\ref{LLE}). The $%

1066: z^{\prime }$ axis of this coordinate system is oriented in the direction

1067: minimizing the classical energy $\mathcal{H}_{\mathrm{eff}}$ of Eq.\ (\ref

1068: {HamClass}) at any fixed time $t.$ It belongs to the $x$-$z$ plane and it

1069: makes the angle $\theta (t)$ with the $z$ axis \ that satisfies the equation

1070: \begin{equation}

1071: h_{z}(t)\sin \theta +\nu \sin \theta \cos \theta -h_{x}\cos \theta =0

1072: \label{EMinEq}

1073: \end{equation}

1074: following from $\partial \mathcal{H}_{\mathrm{eff}}/\partial \theta =0.$

1075: Here $\nu =D_{\mathrm{cl}}/|D_{\mathrm{cl}}|=J/|J|$ and

1076: \begin{equation}

1077: h_{x}\equiv \frac{H_{x}}{2|D_{\mathrm{cl}}|}=\frac{\Delta }{2|J_{0}|},\qquad

1078: h_{z}(t)\equiv \frac{H_{z}(t)}{2|D_{\mathrm{cl}}|}=\frac{W(t)}{2|J_{0}|}.

1079: \label{hxhzDef}

1080: \end{equation}

1081: Eq.\ (\ref{EMinEq}) is well known in the theory of magnetism. In the AFMF

1082: case $\nu <0$, its solution is unique. For the ferromagnetic interaction $%

1083: \nu >0,$ Eq.\ (\ref{EMinEq}) has two solutions corresponding to the minimum

1084: and to the maximum of $\mathcal{H}_{\mathrm{eff}}$ in the range of reduced

1085: fields $h_{x}$ and $h_{z}$ outside the Stoner-Wohlfarth astroid \cite{stowoh}

1086: (see Fig.\ \ref{Fig-lzn-P-F-Astroid})

1087: \begin{equation}

1088: h_{x}^{2/3}+h_{z}^{2/3}=1.  \label{Astroid}

1089: \end{equation}

1090: For $h_{x}$ and $h_{z}$ inside the astroid, Eq.\ (\ref{EMinEq}) has four

1091: solutions corresponding to the stable and metastable minima as well as to

1092: the saddle point and to the maximum of $\mathcal{H}_{\mathrm{eff}}.$ The

1093: adiabatic solution corresponds to the minimum of the energy that changes as $%

1094: h_{z}(t)$ is sweeped from $-\infty $ to $\infty $. For $h_{x}\geq 1$ the

1095: system follows this adiabatic solution in the whole range of $h_{z}.$ In

1096: this case dependence of the effective sweep $W_{\mathrm{eff}}$ of Eq.\ (\ref

1097: {WeffDef}) or, here, $W_{\mathrm{eff}}=H_{z}+2D_{\mathrm{cl}}\cos \theta ,$

1098: on the energy sweep $W$ is shown in Fig.\ \ref{Fig-lzn-P-F-Astroid-Weff}.

1099: One can see that for $J_{0}/\Delta =0.5$ this dependence becomes

1100: nonanalytic. This corresponds to the horizontal line in Fig.\ \ref

1101: {Fig-lzn-P-F-Astroid} that touches the upper corner of the astroid. In

1102: contrast, for $h_{x}<1$ the adiabatic solution exists only until the

1103: crossing of the right branch of the astroid. After that the spin rotates

1104: away from the disappeared metastable state and its behavior becomes

1105: nonadiabatic. The latter case will be considered later while at first we

1106: concentrate on the adiabatic situation.

1107: 

1108: %TCIMACRO{

1109: %\TeXButton{Fig-lzn-P-F-Astroid}{\begin{figure}[t]

1110: %\unitlength1cm

1111: %\begin{picture}(11,4.5)

1112: %\centerline{\psfig{file=Fig-lzn-P-F-Astroid.eps,angle=-90,width=8cm}}

1113: %\end{picture}

1114: %\caption{ \label{Fig-lzn-P-F-Astroid}

1115: %Comparison of the exact quantum and the MFA solutions for the one-particle staying probability $P_N$ for the ferromagnetic coupling.

1116: %}

1117: %\end{figure}%

1118: %}}%

1119: %BeginExpansion

1120: \begin{figure}[t]

1121: \unitlength1cm

1122: \begin{picture}(11,4.5)

1123: \centerline{\psfig{file=Fig-lzn-P-F-Astroid.eps,angle=-90,width=8cm}}

1124: \end{picture}

1125: \caption{ \label{Fig-lzn-P-F-Astroid}

1126: Comparison of the exact quantum and the MFA solutions for the one-particle staying probability $P_N$ for the ferromagnetic coupling.

1127: }

1128: \end{figure}%

1129: %

1130: %EndExpansion

1131: 

1132: %TCIMACRO{

1133: %\TeXButton{Fig-lzn-P-F-Astroid-Weff}{\begin{figure}[t]

1134: %\unitlength1cm

1135: %\begin{picture}(11,6)

1136: %\centerline{\psfig{file=Fig-lzn-P-F-Astroid-Weff.eps,angle=-90,width=9cm}}

1137: %\end{picture}

1138: %\caption{ \label{Fig-lzn-P-F-Astroid-Weff}

1139: %Effective sweep in the adiabatic case for the ferromagnetic coupling of different strengths.

1140: %}

1141: %\end{figure}%

1142: %}}%

1143: %BeginExpansion

1144: \begin{figure}[t]

1145: \unitlength1cm

1146: \begin{picture}(11,6)

1147: \centerline{\psfig{file=Fig-lzn-P-F-Astroid-Weff.eps,angle=-90,width=9cm}}

1148: \end{picture}

1149: \caption{ \label{Fig-lzn-P-F-Astroid-Weff}

1150: Effective sweep in the adiabatic case for the ferromagnetic coupling of different strengths.

1151: }

1152: \end{figure}%

1153: %

1154: %EndExpansion

1155: 

1156: The Landau-Lifshitz equation, Eq.\ (\ref{LLE}) can be rewritten in the

1157: rotating adiabatic coordinate system as follows

1158: \begin{equation}

1159: \mathbf{\dot{s}}^{\prime }=\left[ \mathbf{s}^{\prime }\mathbf{\times }\left(

1160: \gamma \mathbf{H}_{\mathrm{eff}}^{\prime }+\mathbf{\Omega }\right) \right]

1161: ,\qquad \mathbf{\Omega =}\dot{\theta}\mathbf{e}_{y},  \label{LLErotating}

1162: \end{equation}

1163: where $\dot{\theta}$ is time derivative of the appropriate solution of Eq.\ (%

1164: \ref{EMinEq}). The reduced effective field

1165: \begin{equation}

1166: \mathbf{h}_{\mathrm{eff}}\equiv \frac{\mathbf{H}_{\mathrm{eff}}}{2|D_{%

1167: \mathrm{cl}}|}=\left( h_{z}+\nu s_{z}\right) \mathbf{e}_{z}+h_{x}\mathbf{e}%

1168: _{x},  \label{heff}

1169: \end{equation}

1170: in the adiabatic frame is calculated as follows

1171: \begin{eqnarray}

1172: \mathbf{h}_{\mathrm{eff}}^{\prime } &=&\left( h_{x}\cos \theta -h_{\mathrm{%

1173: eff,}z}\sin \theta \right) \mathbf{e}_{x^{\prime }}  \nonumber \\

1174: &&+\left( h_{x}\sin \theta +h_{\mathrm{eff,}z}\cos \theta \right) \mathbf{e}%

1175: _{z^{\prime }}  \nonumber \\

1176: h_{\mathrm{eff,}z} &=&h_{z}+\nu \left( -s_{x^{\prime }}\sin \theta

1177: +s_{z^{\prime }}\cos \theta \right) .  \label{heffTrans}

1178: \end{eqnarray}

1179: With the help of Eq.\ (\ref{EMinEq}) components of $\mathbf{h}_{\mathrm{eff}%

1180: }^{\prime }$ simplify to

1181: \begin{eqnarray}

1182: h_{\mathrm{eff,}x^{\prime }} &=&\left( 1-s_{z^{\prime }}+s_{x^{\prime }}\tan

1183: \theta \right) \nu \sin \theta \cos \theta  \nonumber \\

1184: h_{\mathrm{eff,}z^{\prime }} &=&h_{x}/\sin \theta -s_{x^{\prime }}\nu \sin

1185: \theta \cos \theta  \nonumber \\

1186: &&+\left( 1-s_{z^{\prime }}\right) \left( h_{z}-h_{x}\cot \theta \right)

1187: \cos \theta .  \label{heffTransFinal}

1188: \end{eqnarray}

1189: Note that if the spin is directed along the $z^{\prime }$ axis, i.e., $%

1190: s_{x^{\prime }}=s_{y}=0$ and s$_{z^{\prime }}=1,$ one has $h_{\mathrm{eff,}%

1191: x^{\prime }}=0,$ that is consistent with the choice of the adiabatic

1192: coordinate system. Introducing the dimensionless sweep variables

1193: \begin{equation}

1194: u\equiv \frac{W(t)}{\Delta }=\frac{vt}{\Delta }=\frac{h_{z}}{h_{x}},\qquad

1195: \tilde{\varepsilon}\equiv \frac{\Delta ^{2}}{\hbar v}  \label{uDef}

1196: \end{equation}

1197: [$\tilde{\varepsilon}$ should not be confused with $\varepsilon =(\pi /2)%

1198: \tilde{\varepsilon}$ of Eq.\ (\ref{PLZ})] one can rewrite Eq.\ (\ref

1199: {LLErotating}) in the form

1200: \begin{equation}

1201: \frac{d\mathbf{s}^{\prime }}{du}=\left[ \mathbf{s}^{\prime }\mathbf{\times }%

1202: \left( \tilde{\varepsilon}\frac{\mathbf{h}_{\mathrm{eff}}^{\prime }}{h_{x}}+%

1203: \frac{d\theta }{du}\mathbf{e}_{y}\right) \right] .  \label{epstilDef}

1204: \end{equation}

1205: 

1206: In the slow-sweep limit $\varepsilon \gg 1$ the solution of Eq.\ (\ref

1207: {epstilDef}) is close to the adiabatic solution $s_{x^{\prime }}=s_{y}=0$

1208: and s$_{z^{\prime }}=1.$ Hence one can linearize this equation near the

1209: adiabatic solution:

1210: \begin{eqnarray}

1211: s_{z^{\prime }} &\Rightarrow &1,\qquad h_{\mathrm{eff,}z^{\prime

1212: }}\Rightarrow h_{x}/\sin \theta  \nonumber \\

1213: h_{\mathrm{eff,}x^{\prime }} &\Rightarrow &s_{x^{\prime }}\nu \sin ^{2}\theta

1214: \label{Linearization}

1215: \end{eqnarray}

1216: that results in

1217: \begin{eqnarray}

1218: \frac{ds_{x^{\prime }}}{du} &=&\frac{\tilde{\varepsilon}}{\sin \theta }s_{y}-%

1219: \frac{d\theta }{du}  \nonumber \\

1220: \frac{ds_{y}}{du} &=&-\frac{\zeta \tilde{\varepsilon}}{\sin \theta }%

1221: s_{x^{\prime }},  \label{LLElinearized}

1222: \end{eqnarray}

1223: where

1224: \begin{equation}

1225: \zeta \equiv 1-\rho \sin ^{3}\theta ,\qquad \rho \equiv 2J_{0}/\Delta .

1226: \label{arhoDef}

1227: \end{equation}

1228: The factor $\zeta $ in the second of Eqs.\ (\ref{LLElinearized}) makes spin

1229: precession elliptic for any nonzero coupling $J_{0},$ especially in the

1230: extreme case $\rho =1$. This is an important difference from the model of

1231: one tunneling particle with a nonlinear sweep, \cite{garsch02prb} where

1232: precession remains always circular. It is convenient to introduce

1233: \begin{equation}

1234: \tilde{c}_{+}\equiv -\frac{1}{2}\left( \zeta ^{1/4}s_{x^{\prime }}+i\zeta

1235: ^{-1/4}s_{y}\right) ,\qquad \overline{\Omega }(u)\equiv \frac{\sqrt{\zeta }}{%

1236: \sin \theta }  \label{cplusOmegaDef}

1237: \end{equation}

1238: and rewrite Eqs.\ (\ref{LLElinearized}) in the form

1239: \begin{equation}

1240: \frac{d\tilde{c}_{+}}{du}=-i\tilde{\varepsilon}\overline{\Omega }(u)\tilde{c}%

1241: _{+}+\frac{1}{2}\frac{d\theta }{du}\zeta ^{1/4}.  \label{cplusEq}

1242: \end{equation}

1243: This differential equation can be easily solved with the initial condition $%

1244: \tilde{c}_{+}(-\infty )=0,$ and the staying probability $P_{N}$ can be found

1245: from Eq.\ (\ref{PtMappingClass}). Keeping in mind that at $t=\infty $ both

1246: coordinate systems coincide whereby $\theta =0$ and $a=1$, one can rewrite $%

1247: P_{N}\equiv P_{N}(\infty )$ within our linearized theory as

1248: \begin{equation}

1249: P_{N}\cong \frac{1}{4}\left[ s_{x^{\prime }}^{2}(\infty )+s_{y}^{2}(\infty )%

1250: \right] =\left| \tilde{c}_{+}(\infty )\right| ^{2}.  \label{PNLinearized}

1251: \end{equation}

1252: The final general expression for $P_{N}$ reads

1253: \begin{equation}

1254: P_{N}\cong \left| \frac{1}{2}\int_{-\infty }^{\infty }du\frac{d\theta }{du}%

1255: \zeta ^{1/4}\exp \left[ i\tilde{\varepsilon}\Phi (u)\right] \right| ^{2},

1256: \label{PNfinalgeneral}

1257: \end{equation}

1258: where

1259: \begin{equation}

1260: \Phi (u)\equiv \int_{0}^{u}du^{\prime }\,\overline{\Omega }(u^{\prime }).

1261: \label{PhiDef}

1262: \end{equation}

1263: 

1264: In the absence of interaction $J=0,$ our solution recovers that for the

1265: standard LZS problem in the slow-sweep limit up to the prefactor in front of

1266: a small exponential. Indeed, in this case the solution of Eq.\ (\ref{EMinEq}%

1267: ) is

1268: \begin{equation}

1269: \cos \theta =\frac{h_{z}}{\sqrt{h_{x}^{2}+h_{z}^{2}}}=\frac{u}{\sqrt{1+u^{2}}%

1270: },  \label{CoSiJ0}

1271: \end{equation}

1272: in Eq.\ (\ref{arhoDef}) one has $\rho =0$ and $\zeta =1,$ hence

1273: \begin{equation}

1274: \overline{\Omega }(u)=\sqrt{1+u^{2}},\qquad \frac{d\theta }{du}=-\frac{1}{%

1275: 1+u^{2}}=-\frac{1}{\overline{\Omega }^{2}(u)}.  \label{OmegaDerthetaJ0}

1276: \end{equation}

1277: One can see that in this case Eq.\ (\ref{cplusEq}) coincides with second of

1278: Eqs.\ (24) of Ref.\ \onlinecite{garsch02prb} and Eq.\ (\ref{PNfinalgeneral})

1279: coincides with Eq.\ (26) of Ref.\ \onlinecite{garsch02prb}, with $w^{\prime

1280: }(u)=1$ and $\tilde{c}_{-}\Rightarrow 1.$ The latter is exactly the

1281: approximation proposed in Ref.\ \onlinecite{garsch02prb} that for the

1282: standard LZS problem reproduces the well-known result of Eq.\ (\ref{PLZ})

1283: for $\varepsilon \gg 1$, however with a wrong prefactor: $P\approx (\pi

1284: /3)^{2}e^{-\varepsilon }.$ In Ref.\ \onlinecite{garsch02prb} we have shown

1285: how to correct the prefactor by accurately calculating $\tilde{c}_{-}.$

1286: 

1287: \subsection{Adiabatic case: The results }

1288: 

1289: In the general case $J\neq 0$ Eq.\ (\ref{EMinEq}) does not yield an

1290: analytical solution for $\theta (u).$ Fortunately, instead of $u$ one can

1291: use $x\equiv \cos \theta $ as the integration variable on the interval $%

1292: -1\leq x\leq 1$ and with the help of Eq.\ (\ref{EMinEq}) express $%

1293: u=h_{z}/h_{x}$ as a function of $x.$ Even better is then to parametrize $x$

1294: as $x=w/\sqrt{1+w^{2}},$ analogously to Eq.\ (\ref{CoSiJ0}). This brings

1295: Eq.\ (\ref{PNfinalgeneral}) into the explicit form

1296: \begin{equation}

1297: P_{N}\cong \left| \frac{1}{2}\int_{-\infty }^{\infty }\frac{dw}{1+w^{2}}%

1298: \left[ 1-\frac{\rho }{(1+w^{2})^{3/2}}\right] ^{1/4}e^{i\tilde{\varepsilon}%

1299: \Phi (w)}\right| ^{2},  \label{PNx}

1300: \end{equation}

1301: where

1302: \begin{equation}

1303: \Phi (w)=\int_{0}^{w}dw\sqrt{1+w^{2}}\left[ 1-\frac{\rho }{(1+w^{2})^{3/2}}%

1304: \right] ^{3/2}.  \label{Phix}

1305: \end{equation}

1306: In the absence of interaction, $\rho =0,$ one has $w=u$ and the formulas

1307: above describe the standard LZS effect. For $\varepsilon \gg 1,$ the value

1308: of $P_{N}$ is exponentially small and defined by the singularities of the

1309: integrand in the complex plane of $w$

1310: \begin{equation}

1311: P_{N}=P_{N0}e^{-\varepsilon \func{Im}F(\rho )},  \label{PNExp}

1312: \end{equation}

1313: where

1314: \begin{equation}

1315: F(\rho )=\frac{4}{\pi }\int_{0}^{w_{c}}dw\sqrt{1+w^{2}}\left[ 1-\frac{\rho }{%

1316: (1+w^{2})^{3/2}}\right] ^{3/2}.  \label{FrhoDef}

1317: \end{equation}

1318: For $\rho =0$ one obtains the LZS value $F(0)=i$. For the FM coupling the

1319: relevant singularity is

1320: \begin{equation}

1321: w_{c}=i\sqrt{1-\rho ^{2/3}},\qquad 0<\rho <1,  \label{wcFM}

1322: \end{equation}

1323: that corresponds to vanishing of $\overline{\Omega }(u)$ of Eq.\ (\ref

1324: {cplusOmegaDef}) due to that of the ellipticity factor $\zeta .$ For $\rho

1325: >1,$ the motion of the spin becomes nonadiabatic (see Fig.\ \ref

1326: {Fig-lzn-P-F-Astroid}), and this method does not apply any longer. For $%

1327: 0<\rho <1$ one has $\func{Re}F(\rho )=0$ and the limiting forms of Im$F(\rho

1328: )$ are given by

1329: \begin{equation}

1330: \func{Im}F(\rho )\cong \left\{

1331: \begin{array}{ll}

1332: \displaystyle1-\frac{2}{\pi }\left( \ln \frac{32}{\rho }-1\right) \rho , &

1333: \rho \ll 1 \\

1334: \displaystyle\frac{\sqrt{3}}{2\sqrt{2}}(1-\rho )^{2}, & 1-\rho \ll 1.

1335: \end{array}

1336: \right.  \label{FrhoFMLims}

1337: \end{equation}

1338: For the AFMF coupling $\rho <0,$ there is a pair of relevant singularities

1339: that also correspond to $\zeta =0$ and are given by

1340: \begin{eqnarray}

1341: w_{c\pm } &=&i\left( 1+|\rho |^{2/3}+|\rho |^{4/3}\right) ^{1/4}\exp \left(

1342: \mp i\varphi _{c}\right)  \nonumber \\

1343: \varphi _{c} &=&\frac{1}{2}\arctan \frac{\sqrt{3}|\rho |^{2/3}}{2+|\rho

1344: |^{2/3}}.  \label{wcAFMF}

1345: \end{eqnarray}

1346: Note that $\func{Im}(w_{c\pm })>1$ for $\rho <0$ and thus these

1347: singularities are \emph{further} from the real axis than the LZS singularity

1348: at $w=i.$ The latter, however, makes no contribution since, as can be easily

1349: checked, $\func{Im}F(\rho )=\infty $ for $w_{c}=i$ and $\rho <0.$ Analytical

1350: calculation of the limiting forms of $F(\rho )$ is more cumbersome for the

1351: antiferromagnetic coupling. For $|\rho |\ll 1$ one obtains

1352: \begin{equation}

1353: \func{Re}F(\rho )\cong 2|\rho |,\qquad \func{Im}F(\rho )\cong 1-\frac{2}{\pi

1354: }\left( \ln \frac{32}{|\rho |}-1\right) \rho .  \label{FrhoAFMFsmall}

1355: \end{equation}

1356: Note that $\func{Im}F(\rho )$ is given by the same formula for $|\rho |\ll 1$

1357: and both signs of $\rho .$ In the limit $|\rho |\gg 1$ the result has the

1358: form

1359: \begin{eqnarray}

1360: F(\rho ) &\cong &\frac{2}{3\sqrt{\pi }}\frac{\Gamma (1/4)}{\Gamma (3/4)}%

1361: \left( |\rho |^{3/2}-\frac{1}{4}|\rho |^{5/6}\right)  \nonumber \\

1362: &&+\frac{\sqrt{\pi }\left( \sqrt{3}+i\right) }{\Gamma (5/3)\Gamma (11/6)}%

1363: |\rho |^{2/3}  \nonumber \\

1364: &\simeq &1.1|\rho |^{3/2}-0.3\,|\rho |^{5/6}+\left( 3.6+2.1i\right) |\rho

1365: |^{2/3}.  \label{FrhoAFMFlarge}

1366: \end{eqnarray}

1367: The real and imaginary parts of $F(\rho )$ numerically calculated from Eq.\ (%

1368: \ref{FrhoDef}), as well as the analytical limiting forms, are shown in Fig.\

1369: \ref{Fig-lzn-Frho}.

1370: 

1371: %TCIMACRO{

1372: %\TeXButton{Fig-lzn-Frho}{\begin{figure}[t]

1373: %\unitlength1cm

1374: %\begin{picture}(11,6)

1375: %\centerline{\psfig{file=Fig-lzn-Frho.eps,angle=-90,width=9cm}}

1376: %\end{picture}

1377: %\caption{ \label{Fig-lzn-Frho}

1378: %Function $F(\rho) $ of Eqs.\

1379: %(\protect\ref{PNExp}) and (\protect\ref{FrhoDef}).

1380: %}

1381: %\end{figure}%

1382: %}}%

1383: %BeginExpansion

1384: \begin{figure}[t]

1385: \unitlength1cm

1386: \begin{picture}(11,6)

1387: \centerline{\psfig{file=Fig-lzn-Frho.eps,angle=-90,width=9cm}}

1388: \end{picture}

1389: \caption{ \label{Fig-lzn-Frho}

1390: Function $F(\rho) $ of Eqs.\

1391: (\protect\ref{PNExp}) and (\protect\ref{FrhoDef}).

1392: }

1393: \end{figure}%

1394: %

1395: %EndExpansion

1396: 

1397: The prefactor $P_{N0}$ in Eq.\ (\ref{PNExp}) is determined for $\varepsilon

1398: \gg 1$ by a close vicinity of the singularities $w_{c}.$ For the

1399: ferromagnetic coupling the result is

1400: \begin{equation}

1401: P_{N0}\cong \frac{\pi ^{2}}{15\varepsilon \rho \sqrt{1-\rho ^{2/3}}},\qquad

1402: 0<\rho <1.  \label{P0F}

1403: \end{equation}

1404: For the AFMF coupling $\rho <0$ interference of the two contributions from $%

1405: w_{c+}$ and $w_{c-}$ of Eq.\ (\ref{wcAFMF}) results in the oscillating

1406: prefactor

1407: \begin{equation}

1408: P_{N0}\cong \frac{\pi ^{2}}{15\varepsilon |\rho ||w_{c}|}\cos ^{2}\left[

1409: \frac{\varepsilon }{2}\func{Re}F(\rho )\right] ,\qquad \rho <0.  \label{P0AF}

1410: \end{equation}

1411: Oscillating prefactors of such a kind take place for\emph{\ one} tunneling

1412: particle if the sweep is nonlinear and decelerating in the vicinity of the

1413: resonance.\cite{garsch02prb} Here \emph{effective} nonlinearity of this type

1414: occurs because of the negative coupling between tunneling particles, even

1415: for the linear sweep.

1416: 

1417: Both Eqs.\ (\ref{P0F}) and (\ref{P0AF}) break down in the linit $\rho

1418: \rightarrow 0$ since different singularities come close to each other. One

1419: could work out the crossover from Eqs.\ (\ref{P0F}) and (\ref{P0AF}) to the

1420: value $P_{N0}=(\pi /3)^{2}$ (see Ref.\ \onlinecite{garsch02prb}) at $\rho =0$

1421: that takes place in a narrow region \TEXTsymbol{\vert}$\rho |\varepsilon

1422: \sim 1.$ This makes no sence, however, since the result $P_{N0}=\pi /3$

1423: differs from the exact LZS prefactor $P_{0}=1$ and it has to be improved by

1424: taking into account nonlinear terms dropped during the derivation of Eq.\ (%

1425: \ref{cplusEq}). The corresponding procedure for $\rho =0$ \ is described in

1426: Ref.\ \onlinecite{garsch02prb}. In the present case $\rho \neq 0$ this would

1427: be too involved and we don't try to do it. It is clear that the logarithmic

1428: singularity of the exponent of Eq.\ (\ref{FrhoAFMFsmall}) should ve

1429: compensated for by the singularity of the prefactor so that the staying

1430: probability $P_{N}$ behaves linearly in $\rho $ near $\rho =0$ [see Eq.\ (%

1431: \ref{dPdrhoslowsweep})].

1432: 

1433: Let us now consider the case $1-\rho \ll 1$ in more detail since the

1434: prefactor $P_{N0}$ of Eq.\ (\ref{P0F}) diverges at $\rho \rightarrow 1$.

1435: Here one has $\func{Im}F(\rho )\ll 1,$ and the exponential decrease of $%

1436: P_{N} $ is very slow. In this region according to Eq.\ (\ref{wcFM}) the

1437: integral in Eq.\ (\ref{PNx}) is dominated by small $w$ and it can be

1438: simplified to

1439: \begin{eqnarray}

1440: P_{N} &\cong &\frac{2}{3}\delta ^{3/2}\left| \frac{1}{2}\int_{-\infty

1441: }^{\infty }dt\left( 1+t^{2}\right) ^{1/4}e^{ia\widetilde{\Phi }(t)}\right|

1442: ^{2}  \nonumber \\

1443: \widetilde{\Phi }(t) &=&\int_{0}^{t}dt^{\prime }\left( 1+t^{\prime 2}\right)

1444: ^{3/2},  \label{PNt}

1445: \end{eqnarray}

1446: where

1447: \begin{equation}

1448: \delta \equiv 1-\rho \ll 1,\qquad t\equiv \sqrt{\frac{3}{2\delta }}w,\qquad

1449: a\equiv \sqrt{\frac{2}{3}}\delta ^{2}\tilde{\varepsilon}.  \label{atDef}

1450: \end{equation}

1451: It is convenient to compute the integral in Eq.\ (\ref{PNt}) by shifting the

1452: integration contour by $i$ to suppress oscillations of the integrand, i.e.,

1453: to parametrize $t=i+z,$ $-\infty <z<\infty .$ With $\widetilde{\Phi }%

1454: (i)=i3\pi /8$ this yields

1455: \begin{equation}

1456: P_{N}\cong P_{N0}e^{-a\func{Im}\widetilde{\Phi }(i)}=P_{N0}\exp \left( -%

1457: \frac{3\pi }{8}a\right) ,  \label{PNdelta}

1458: \end{equation}

1459: where the exponent coincides with the previously obtained $[\sqrt{3}/(2\sqrt{%

1460: 2})]\delta ^{2}\varepsilon $ [see second line of Eq.\ (\ref{FrhoFMLims})].

1461: The prefactor $P_{N0}$ can be calculated analytically for $a\gg 1$ and $a\ll

1462: 1.$ For $a\gg 1$ we need the small-$z$ expansions

1463: \begin{eqnarray}

1464: &&\delta \widetilde{\Phi }(z)\equiv \widetilde{\Phi }(i+z)-\widetilde{\Phi }%

1465: (i)\cong \frac{4}{5}\left( -1+i\right) z^{5/2}  \nonumber \\

1466: &&[1+(i+z)^{2}]^{1/4}\cong (-1)^{1/8}(2z)^{1/4}.  \label{SmallzExp}

1467: \end{eqnarray}

1468: Values of these functions for $z<0$ are obtained from $(-1)^{5/2}=-i$ and $%

1469: (-1)^{1/4}=(1-i)/\sqrt{2}.$ After that calculation in Eq.\ (\ref{PNt})

1470: yields

1471: \begin{equation}

1472: P_{N0}\cong \frac{2\pi }{15}\frac{\delta ^{3/2}}{a},\qquad a\gg 1

1473: \label{PN0alarge}

1474: \end{equation}

1475: that is a limiting form of Eq.\ (\ref{P0F}). In the opposite limit $a\ll 1,$

1476: the integral in Eq.\ (\ref{PNt}) is dominated by large $t,$ so that one can

1477: use $\widetilde{\Phi }(t)\cong t^{4}/4$ and $\left( 1+t^{2}\right)

1478: ^{1/4}\cong t^{1/2}.$ This yields

1479: \begin{equation}

1480: P_{N0}\cong \cos ^{2}\left( \frac{3\pi }{16}\right) \frac{\Gamma ^{2}(3/8)}{6%

1481: \sqrt{2}}\frac{\delta ^{3/2}}{a^{3/4}}\simeq \frac{0.747837}{\varepsilon

1482: ^{3/4}},\qquad \rho \rightarrow 1.  \label{PN0alSmall}

1483: \end{equation}

1484: It is convenient to represent $P_{N0}(a,\delta )$ in the whole range of $a$

1485: with the help of the crossover function $f(a)$ according to

1486: \begin{eqnarray}

1487: &&P_{N0}(a,\delta )=\frac{2\pi }{15}\frac{\delta ^{3/2}}{a}f(a)  \nonumber \\

1488: &&f(a)\cong \left\{

1489: \begin{array}{ll}

1490: 1, & a\gg 1 \\

1491: 1.09294a^{1/4}, & a\ll 1.

1492: \end{array}

1493: \right.  \label{PN0Scaling}

1494: \end{eqnarray}

1495: Numerically computed function $f(a)$ is shown in Fig.\ (\ref

1496: {Fig-lzn-FScaling-a}).

1497: 

1498: %TCIMACRO{

1499: %\TeXButton{Fig-lzn-FScaling-a.eps}{\begin{figure}[t]

1500: %\unitlength1cm

1501: %\begin{picture}(11,6)

1502: %\centerline{\psfig{file=Fig-lzn-FScaling-a.eps,angle=-90,width=9cm}}

1503: %\end{picture}

1504: %\caption{ \label{Fig-lzn-FScaling-a}

1505: %Scaling function $f(a)$ of Eq.\ (\protect\ref{PN0Scaling})

1506: %}

1507: %\end{figure}%

1508: %}}%

1509: %BeginExpansion

1510: \begin{figure}[t]

1511: \unitlength1cm

1512: \begin{picture}(11,6)

1513: \centerline{\psfig{file=Fig-lzn-FScaling-a.eps,angle=-90,width=9cm}}

1514: \end{picture}

1515: \caption{ \label{Fig-lzn-FScaling-a}

1516: Scaling function $f(a)$ of Eq.\ (\protect\ref{PN0Scaling})

1517: }

1518: \end{figure}%

1519: %

1520: %EndExpansion

1521: 

1522: %TCIMACRO{

1523: %\TeXButton{Fig-lzn-Phx}{\begin{figure}[t]

1524: %\unitlength1cm

1525: %\begin{picture}(11,6)

1526: %\centerline{\psfig{file=Fig-lzn-Phx.eps,angle=-90,width=9cm}}

1527: %\end{picture}

1528: %\caption{ \label{Fig-lzn-Phx}

1529: %$P_N$ in the slow-sweep limit $\varepsilon\to\infty$ vs $h_x=1/\rho=\Delta/(2J_0)$.

1530: %}

1531: %\end{figure}%

1532: %}}%

1533: %BeginExpansion

1534: \begin{figure}[t]

1535: \unitlength1cm

1536: \begin{picture}(11,6)

1537: \centerline{\psfig{file=Fig-lzn-Phx.eps,angle=-90,width=9cm}}

1538: \end{picture}

1539: \caption{ \label{Fig-lzn-Phx}

1540: $P_N$ in the slow-sweep limit $\varepsilon\to\infty$ vs $h_x=1/\rho=\Delta/(2J_0)$.

1541: }

1542: \end{figure}%

1543: %

1544: %EndExpansion

1545: 

1546: \subsection{Strong ferromagnetic interactions}

1547: 

1548: For stronger ferromagnetic interactions, $\rho >1,$ the adiabatic

1549: approximation breaks down since the spin approximately follows the initial

1550: energy minimum that becomes unstable at some $h_{z}(t)>0$ (see Fig.\ \ref

1551: {Fig-lzn-P-F-Astroid}) and than it performes a large motion that does not

1552: approach the new stable energy minimum and thus cannot be linearized around

1553: it. While the problem becomes much more complicated in this case, one can

1554: still find analytically $\stackunder{\varepsilon \rightarrow \infty }{\lim }%

1555: P_{N}$ that is nonzero$.$ To this end, one can represent the LLE for the

1556: spin, Eq.\ (\ref{LLE}), in the Hamiltonian form in terms of the canonical

1557: angle variables $\left\{ \cos \theta ,\varphi \right\} \Leftrightarrow

1558: \left\{ p,q\right\} $ (in the arbitrary frame)

1559: \begin{equation}

1560: \frac{d}{dt}\cos \theta =-\frac{\partial \mathcal{H}}{\partial \varphi }%

1561: ,\qquad \frac{d}{dt}\varphi =\frac{\partial \mathcal{H}}{\partial \cos

1562: \theta }  \label{HamEq}

1563: \end{equation}

1564: and use conservation of the action

1565: \begin{equation}

1566: \mathcal{S}=\oint pdq=\oint \cos \theta d\varphi  \label{ActionDef}

1567: \end{equation}

1568: over the period of motion for very slowly changing parameters of the system

1569: [here $h_{z}(t)]$. Since in the final state $h_{z}\rightarrow \infty $ the

1570: spin precesses around the $z$ axis with a constant value of $s_{z}(\infty

1571: )=\cos \left[ \theta (\infty )\right] $ that is related to the staying

1572: probability $P_{N}$ by Eq.\ (\ref{PtMappingClass}) and the corresponding

1573: action is simply $\mathcal{S}(\infty )=2\pi \cos \theta (\infty ),$ one

1574: obtains

1575: \begin{equation}

1576: P_{N}=\frac{1}{2}\left( 1-\frac{1}{2\pi }\mathcal{S}\right) ,\qquad

1577: \varepsilon \rightarrow \infty ,  \label{PNActionDef}

1578: \end{equation}

1579: where $\mathcal{S}$ is the action over the trajectory that starts from the

1580: metastable energy minimum that is on the verge of being unstable

1581: \begin{equation}

1582: s_{zs}=-(1-h_{x}^{2/3})^{1/2},\qquad s_{xs}=h_{x}^{1/3},\qquad s_{ys}=0.

1583: \label{sStarting}

1584: \end{equation}

1585: This trajectory can be found from the conservation of energy,

1586: \begin{equation}

1587: -s_{z}^{2}-2(1-h_{x}^{2/3})^{3/2}s_{z}-2h_{x}s_{x}=1-3h_{x}^{2/3}.

1588: \label{EnergyConservation}

1589: \end{equation}

1590: Analysis shows that in the range $0<h_{x}<3\sqrt{3}/8$ this trajectory

1591: encircles the $z$ axis, thus the $z$ axis can be used as the polar axis ($%

1592: s_{z}=\cos \theta ,$ $s_{x}=\sin \theta \cos \varphi ,$ $s_{y}=\sin \theta

1593: \sin \varphi $) to compute $\mathcal{S}.$ In the overlapping range $%

1594: 1/8<h_{x}<1$ the trajectory encircles the $x$ axis that can be used as the

1595: polar axis ($s_{x}=\cos \theta ,$ $s_{z}=-\sin \theta \cos \varphi ,$ $%

1596: s_{y}=\sin \theta \sin \varphi $). With these choices, $\cos \theta $ can be

1597: found numerically from Eq.\ (\ref{EnergyConservation}) as a function of $%

1598: \varphi $ and the action $\mathcal{S}$ can be computed from Eq.\ (\ref

1599: {ActionDef}). The result for $P_{N}$ vs $h_{x}$ is shown in Fig.\ \ref

1600: {Fig-lzn-Phx} with asymptotes

1601: \begin{equation}

1602: P_{N}\cong \left\{

1603: \begin{array}{ll}

1604: 1-(3/2)h_{x}^{2/3}, & h_{x}\ll 1 \\

1605: 0.544861\left( 1-h_{x}\right) ^{3/2}, & 1-h_{x}\ll 1.

1606: \end{array}

1607: \right.  \label{PhxLims}

1608: \end{equation}

1609: 

1610: \section{Discussion}

1611: 

1612: The simplified model of interacting tunneling particles that was considered

1613: above allows to quantify the influence of interaction on the

1614: Landau-Zener-Stueckelberg staying probability $P$. It was done here by

1615: numerically solving the problem for up to $N=200$ interacting particles as

1616: well as using a number of analytical and half-analytical approaches,

1617: including the mean-field limit $N\rightarrow \infty $.

1618: 

1619: In accord with physical expectations and considerations of the exact levels

1620: of the system in Fig.\ \ref{Fig-lzn-En}a, the ferromagnetic coupling tends

1621: to suppress LZS transitions to another bare energy level (i.e., the state on

1622: the other side of the energy barrier), in agreement with Ref.\ %

1623: \onlinecite{hamraemiysai00}. For $N\rightarrow \infty $ there is a critical

1624: value of the coupling above which the staying probability $P$ does not go to

1625: zero in the slow-sweep limit $\varepsilon \rightarrow \infty $, in contrast

1626: with the standard LZS case. For finite $N$ the dependence of $P$ on $%

1627: \varepsilon $ is a sum of many exponentials with greatly differing

1628: relaxation rates that is very slow approaching zero (see Fig.\ \ref

1629: {Fig-lzn-P-F-j3}). The same should be the case for more realistic

1630: interactions of the ferromagnetic type.

1631: 

1632: The negative coupling in our model (that corresponds to the

1633: antiferromagnetic frustrating coupling) tends to boost the LZS transitions.

1634: In the limit $N\rightarrow \infty $ the staying probability $P$ even turns

1635: to zero at \emph{finite} values of the sweep rate while the \emph{effective}

1636: sweep rate becomes an odd function of time with retardation in the

1637: resonance-crossing region (see Fig.\ \ref{Fig-lzn-PWeffvst-AF-MFA-j10m}). On

1638: the other hand, models with more realistic antiferromagnetic interactions

1639: such as the nearest-neighbor interaction have the energy-level scheme

1640: strongly differing from that shown in Fig.\ \ref{Fig-lzn-En}b. Preliminary

1641: results show that such interactions tend to hamper LZS transitions instead

1642: of boosting it, although not to such an extent as ferromagnetic interactions.

1643: 

1644: Theoretically the model considered here is interesting since it maps on the

1645: problem of a single large spin $S=N/2$ and thus the mean-field limit $%

1646: N\rightarrow \infty $ corresponds to the classical limit $S\rightarrow

1647: \infty $ for the large spin. It would be very interesting to study

1648: deviations from the mean-field solutions for large but finite $N.$ These

1649: deviations can be very large, as can be seen in Fig.\ \ref{Fig-lzn-P-AF-j10}

1650: in the region $\varepsilon \simeq 0.11.$ How quantitatively well does the

1651: MFA work for model systems with more complicated interactions and for

1652: realistic systems remains unclear, and it is an interesting topic for

1653: further work.

1654: 

1655: \section{\protect\bigskip Acknowledgments}

1656: 

1657: Many useful discussions with Rolf Schilling are greatfully acknowledged.

1658: 

1659: %\bigskip

1660: 

1661: %\bibliographystyle{prsty}

1662: %\bibliography{gar-general,gar-tunneling,gar-own,gar-oldworks}

1663: 

1664: \begin{thebibliography}{99}

1665: \bibitem{lan32}  {L. D. Landau}, Phys. Z. Sowjetunion \textbf{2}, 46 (1932).

1666: 

1667: \bibitem{zen32}  C. Zener, Proc. R. Soc. London A \textbf{137}, 696 (1932).

1668: 

1669: \bibitem{stu32}  E.~C.~G. Stueckelberg, Helv. Phys. Acta \textbf{5}, 369

1670: (1932).

1671: 

1672: \bibitem{crohug77}  {D. S. F. Crothers and J. G Huges}, J. Phys. B \textbf{10%

1673: }, L557 (1977).

1674: 

1675: \bibitem{werses99}  {W. Wernsdorfer and R. Sessoli}, Science \textbf{284},

1676: 133 (1999).

1677: 

1678: \bibitem{weretal00epl}  {W. Wernsdorfer, R. Sessoli, A. Caneshi, D.

1679: Gatteschi, and A. Cornia}, Europhys. Lett. \textbf{50}, 552 (2000).

1680: 

1681: \bibitem{hamraemiysai00}  {A. Hams, H. De Raedt, S. Miyashita, and K. Saito}%

1682: , Phys. Rev. B \textbf{62}, 13880 (2000).

1683: 

1684: \bibitem{garsch02prb}  {D. A. Garanin and R. Schilling}, Phys. Rev. B

1685: \textbf{66}, 174438 (2002).

1686: 

1687: \bibitem{akusch92}  {V. M. Akulin and W. P. Schleich}, Phys. Rev. A \textbf{%

1688: 46}, 4110 (1992).

1689: 

1690: \bibitem{dobzve97}  {V. V. Dobrovitski and A. K. Zvezdin}, Europhys. Lett.

1691: \textbf{38}, 377 (1997).

1692: 

1693: \bibitem{gar91jpa}  {D. A. Garanin}, J. Phys. A \textbf{24}, L61 (1991).

1694: 

1695: \bibitem{ternak97}  {Y. Teranishi and H. Nakamura}, J. Chem. Phys. \textbf{%

1696: 107}, 1904 (1997).

1697: 

1698: \bibitem{garsch02}  {D. A. Garanin and R. Schilling}, Europhys. Lett.

1699: \textbf{59}, 7 (2002).

1700: 

1701: \bibitem{stowoh}  {E. C. Stoner and E. P. Wohlfarth}, Philos. Trans. R. Soc.

1702: London, Ser. A \textbf{\ 240}, 599 (1948); IEEE Trans. Magn. \textbf{MAG-27}%

1703: , 3475 (1991).

1704: \end{thebibliography}

1705: 

1706: \end{document}

1707: