1: \documentclass[preprint, aps, pre, eqsecnum,amsmath, amssymb]{revtex4}

2: %\documentclass[draft,twocolumn, aps, pre, eqsecnum,amsmath, amssymb]{revtex4}

3: 

4: % \usepackage[final]{graphicx}  % to include figures

5: %\documentclass[aps,preprint,amssymb,twocolumn]{revtex4}

6: \usepackage[final]{graphicx}

7: %\usepackage{epsfig}

8: 

9:  \renewcommand{\arraystretch}{1.5}  %faktor 1.5 fuer tabellenzeilenabstand

10: 

11:  \newcommand{\Nabla}{\mbox{\bf\boldmath $\nabla$}}

12:  \newcommand{\ve}[1]{\stackrel{\rightharpoonup}{#1}}

13:  \renewcommand{\vec}[1]{{\bf #1}}

14:  \newcommand{\Vepsilon}{\mbox{\bf\boldmath $\epsilon$}}

15:  \newcommand{\Vxi}{\mbox{\bf\boldmath $\xi$}}

16:  \newcommand{\Vphi}{\mbox{\boldmath $\varphi$}}

17:  \newcommand{\Vzeta}{\mbox{\boldmath $\zeta$}}

18:  \newcommand{\VPsi}{\mbox{\boldmath $\Psi$}}

19:  \newcommand{\Overrightarrow}[1]{\stackrel{\textstyle\rightarrow}{#1}}

20: 

21:  % info und seite in kopfzeile

22: \pagestyle{myheadings}

23:  % datum in kopfzeile. falls 'twoside': {leftside}{rightside}

24: \markboth{\today}{\today}

25: 

26: \bibliographystyle{prsty}

27: 

28: \begin{document}

29: 

30: \title{Magnetization of rotating ferrofluids: the effect of polydispersity}

31: \author{A.~Leschhorn, J.~P.~Embs, M.~L\"{u}cke}

32: \affiliation{Institut f\"{u}r Theoretische Physik, Universit\"{a}t

33: des Saarlandes, D-66041~Saarbr\"{u}cken, Germany\\}

34: 

35: \date{\today}

36: 

37: \begin{abstract}

38: The influence of polydispersity on the magnetization is analyzed in a

39: nonequilibrium situation  where a cylindrical ferrofluid

40: column is enforced to rotate with constant frequency like a rigid body

41: in a homogeneous magnetic field that is applied

42: perpendicular to the cylinder axis. Then, the magnetization

43: and the internal magnetic field are not longer

44: parallel to each other and their directions differ from that of the

45: applied magnetic field. Experimental results on the transverse

46: magnetization component perpendicular to the applied field are compared

47: and analyzed as functions of rotation frequency and field strength with

48: different polydisperse Debye models that take into account the

49: polydispersity in different ways and to a varying degree.

50: \end{abstract}

51: 

52: \maketitle

53:  %\tableofcontents

54: 

55:  % skip 2/6inch

56: \vskip2pc

57: 

58:  %im gegenstz zu \widetext (z.b. bei formeln)

59:  %halbe zeitenbreite f"ur spaltentext

60:  %\narrowtext

61: 

62:  %\draft

63: %------------------------------------

64: \section{Introduction}

65: %-----------------------------------------

66: The prospect of influencing the rotational dynamics of the

67: nanoscaled magnetic particles in a ferrofluid by a macroscopic flow

68: and/or by a magnetic field in order to then observe the resulting

69: response via the magnetization and/or via changes in the flow has

70: been stimulating many research activities

71: \cite{Ro85,BlCeMa97,Od02a,Od02b} ever since McTague measured

72: \cite{McTague69} the so-called magneto-viscous effect. Of particular

73: interest are in this context flows that are shear free on the

74: macroscopic scale as in a fluid that is rotating like a rigid body

75: with a rotation frequency, say, ${\bf \Omega}=\Omega {\bf e}_z$.

76: 

77: While the colloidal magnetic particles then undergo thermally sustained

78: rotational and translational Brownian motion on the microscopic

79: scale they co-rotate in the mean with the deterministic macroscopic

80: rigid body flow. However, this mean co-rotation can be hindered by magnetic

81: torques on their moments when a magnetic field, say, ${\bf H}_0 =

82: H_0 {\bf e}_x$ is applied perpendicular to the rotation axis ${\bf

83: e}_z$ of the flow. The combination of the externally imposed forcing

84: of the particle motion by (i) the rigid body flow in which they are

85: floating and by (ii) the magnetic torques on their magnetic moments

86: drives the colloidal suspension out of equilibrium. Concerning the

87: magnetic moments, this forcing causes the mean orientation of the

88: moments, i.e., of the magnetization ${\bf M}$ of the ferrofluid to

89: be no longer parallel to the internal magnetic field ${\bf H}$.

90: Instead, ${\bf M}$ is pushed out of the direction of ${\bf H}$ as

91: well as of that of ${\bf H}_0$ thereby acquiring a nonzero

92: transverse component $M_y$. Here it should be noted that in a long

93: cylinder Maxwell's equations imply the vector relation ${\bf H} =

94: {\bf H}_0 - {\bf M}/2$ between the three fields when they are

95: stationary and homogeneous but they need not be collinear.

96: However, in equilibrium, ${\bf \Omega}=0$, the three fields are

97: indeed collinear: the equilibrium magnetization ${\bf M}_{eq}({\bf

98: H})=M_{eq}(H){\bf H}/H$ is parallel to the internal field ${\bf H}$.

99: 

100: Recently, the transverse magnetization $M_y$ of a slender cylindrical

101: column of ferrofluid that was enforced to rotate like a rigid body

102: with constant frequency $\Omega {\bf e}_z$ in an applied homogeneous

103: magnetic field $H_0 {\bf e}_x$ was measured as a function of $\Omega

104: $ and $H_0$ \cite{EmMaWaKiLeLu06}. These measurements showed among

105: others that the predictions \cite{LeLu06b} based on models for the

106: magnetization dynamics \cite{Sh72,FeKr99,Fe00,Sh01a,MuLi01} with a

107: single relaxation time overestimate the magnitude of $M_y$. One

108: reason for this discrepancy seems to be that particles with

109: different sizes and different rotational dynamics of their magnetic

110: moments contribute differently to the non-equilibrium, flow-induced

111: component $M_y$ of the magnetization. In particular only the

112: magnetic moments of the larger particles in which the magnetic

113: moments are blocked and "frozen" in the particles, i.e., those with

114: effective Brownian relaxation dynamics may be rotated by the flow

115: out of the direction of the magnetic field.

116: 

117: Here we consider poly-disperse models with single-particle Brownian

118: as well as N\'eel relaxation dynamics for the different particle

119: sizes. So we ignore the influence of any dipolar magnetic

120: interaction and of any flow induced interaction on the (rotation)

121: dynamics of the particles. Thereby collective, collision dominated

122: long-range and long-time hydrodynamic relaxation dynamics of the

123: ensemble of magnetic moments are discarded since only the individual

124: relaxation of each magnetic moment is considered --- albeit in the

125: collectively generated internal magnetic field ${\bf H}$.

126: 

127: %\clearpage

128: %------------------------------------------------

129: \section{Equilibrium magnetization}

130: %------------------------------------------------

131: In our experiments we used several ferrofluids out of the APG-series

132: of FerroTec. Their saturation magnetization was specified by the

133: manufacturer to be  $M_{sat}^{FF}=17507$ A/m ($\pm\,10\%$). This

134: corresponds to a volume concentration $\phi\approx 3.6\%$ of the

135: suspended magnetic material. We have measured the equilibrium

136: magnetization of the ferrofluids with a vibrating sample

137: magnetometer (LakeShore 7300 VSM) with a commercial PC user package.

138: In order to get information on the particle size distribution of the

139: ferrofluid under investigation, we used fits

140: \cite{EmMuKrMeNaMuWiLuHeKn01} with a lognormal form of the

141: distribution as well as with a regularization procedure

142: \cite{EmMaWaKiLeLu06} based on Tichonovs method \cite{WeSt85}.

143: Generally the equilibrium magnetization $M^{eq}(H)$ as a function of

144: the internal magnetic field $H$ can be approximated by a

145: superposition of Langevin-functions

146: \begin{equation}\label{meqvonh}

147:     M^{eq}(H)=\sum\limits_{j}w_j {\mathcal{L}}[\alpha_j(H)].

148: \end{equation}

149: Here ${\mathcal{L}}(x)=\coth(x)-1/x$ denotes the Langevin-function

150: that depends on the dimensionless Langevin-parameter

151: $\alpha_j(H)=\mu_0m_jH/k_BT$ and $w_j$ are the so-called magnetic

152: weights. $m_j$ refers to the magnetic moment of particles with

153: magnetic diameter $d_j$, i.e.,

154: $m_j=\frac{\pi}{6}d_j^3M_{sat}^{bulk}$ with $M_{sat}^{bulk}$ the

155: bulk-saturation magnetization. From Eq. (\ref{meqvonh}) we can

156: deduce the initial susceptibility $\chi_0=\frac{\pi\mu_0

157: M_{sat}^{bulk}}{18k_BT}\sum_{j} w_j d_j^3$ and the saturation

158: magnetization $M_{sat}^{FF}=\sum_j w_j$ of the ferrofluid under

159: investigation.

160: 

161: Fig.~\ref{FIG:meq} shows the experimentally determined equilibrium

162: magnetization $M_{eq}(H)$ of APG 933 versus internal field $H$

163: together with fits that were obtained with a lognormal distribution

164: \cite{EmMuKrMeNaMuWiLuHeKn01} and with the regularization method

165: \cite{EmMaWaKiLeLu06}. The saturation magnetization of the

166: ferrofluid sample was $M_{sat}^{FF}=19108.6$ A/m. From the

167: saturation magnetization the volume concentration of the magnetite

168: particles was found to be $\phi=M_{sat}^{FF}/M_{sat}^{bulk}=4.1$ \%,

169: in reasonable agreement with the manufacturer's specifications. For

170: the initial susceptibility we found the value $\chi_0=1.09$.

171: 

172: The magnetic weight distributions $w(d)$ resulting from the two fit

173: methods are shown in Fig.~\ref{FIG:vert}.

174: 

175: %\clearpage

176: %------------------ SEC III ---------------------------------------

177: \section{Experimental setup}

178: %---------------------------------------------------------

179: The experimental setup for measuring the magnetization of a rotating

180: cylindrical column of ferrofluid is sketched in Fig. \ref{FIG:sys}.

181: It is described in more detail in \cite{EmMaWaKiLeLu06}.  The

182: ferrofluid is filled into a cylindrical plexiglass sample holder

183: with inside radius $R=3.2$ mm. This radius is so small that for our

184: rotation frequencies the ferrofluid rotates as a rigid body with a

185: flow field ${\bf u}({\bf r})={\bf \Omega} \times {\bf r} = \Omega r

186: {\bf e}_{\varphi}$. Here $\Omega$ is the externally enforced

187: constant rotation rate of the sample and ${\bf e}_{\varphi}$ is the

188: unit vector in azimuthal direction. A homogeneous and temporally

189: constant magnetic field ${\bf{H}}_0=H_0{\bf{e}}_x$ is applied

190: perpendicular to the cylinder axis ${\bf e}_z$. For such a

191: combination of enforced rotation and applied field theoretical

192: models allow for a spatially and temporally constant nonequilibrium

193: magnetization ${\bf M}$ that is pushed out of the directions of

194: ${\bf H}_0$ and ${\bf H}$ by the flow.

195: 

196: According to the Maxwell equations the fields ${\bf H}$ and ${\bf

197: M}$ within the ferrofluid are related to each other via

198: \begin{eqnarray}

199: {\bf H} = {\bf H}_0 - \frac{1}{2} {\bf M} \label{EQ:maxwell}

200: \end{eqnarray}

201: for our long cylindrical sample and in particular $H_y=-M_y/2$ as

202: indicated schematically in Fig.~\ref{FIG:sys}. In addition they

203: demand that the magnetic field outside the ferrofluid cylinder

204: \begin{eqnarray}

205: {\bf H}^{out} = {\bf H}_0 + \frac{1}{2} \frac{R^{2}}{r^{2}} \left(

206: 2\frac{{\bf r}}{r} \frac{{\bf M}\cdot {\bf r}}{r} - {\bf M} \right)

207: \label{EQ:hout}

208: \end{eqnarray}

209: is a superposition of the applied field ${\bf H}_0$ and the dipolar

210: contribution from ${\bf M}$. This result yields a relation between

211: the perpendicular component of the magnetization $M_y$ resp.~of the

212: internal field $H_y=-M_y/2$  and the field $H_y^{sensor}$ measured

213: by the Hall--sensor outside the sample as indicated in

214: Fig.~\ref{FIG:sys}. Considering the finite size of the Hall--sensor,

215: $H_y^{sensor}$ is given by

216: \begin{equation}\label{EQ:hysensor}

217: H_y^{sensor} = \frac{1}{2a} \int_{-a}^a H_y^{out} dx = -

218: \frac{R^2}{a^2+b^2} H_y \, .

219: \end{equation}

220: In our experimental setup $b=4.75$ mm, $R=3.2$ mm, and $a=2$ mm;

221: here $a$ denotes the horizontal extension of the Hall sensor. So,

222: $H_y^{\text{sensor}}=-0.386\,H_y$ where $H_y=-M_y/2$ is the

223: $y$-component of the internal magnetic field in the ferrofluid.

224: 

225: %\clearpage

226: %--------------------------------------------

227: \section{Magnetization dynamics of a poly-disperse model}

228: %--------------------------------------------

229: 

230: Comparisons with experimental results showed \cite{EmMaWaKiLeLu06}

231: that theoretical predictions \cite{LeLu06b} based on models

232: \cite{Sh72,FeKr99,Fe00,Sh01a,MuLi01} with a single relaxation time

233: overestimate the magnitude of $H_y^{sensor}$. One reason is that

234: particles with different sizes and different rotational dynamics of

235: their magnetic moments contribute differently to the

236: non-equilibrium, flow-induced component $M_y$ of the magnetization

237: and that in particular only the magnetic moments of the larger

238: particles with effective Brownian relaxation dynamics may be rotated

239: by the flow out of the direction of the magnetic field.

240: 

241: Therefore, we consider here as a next step poly-disperse models with

242: single-particle Brownian and N\'eel relaxation dynamics for the

243: different particle sizes. Such models have been used \cite{LeLu06a}

244: to determine within a linear response analysis the

245: effect of polydispersity on the dynamics of a torsional ferrofluid

246: pendulum that was periodically forced close to resonance to undergo

247: small amplitude oscillations in a rigid body flow

248: \cite{EmMuWaKnLu00,EmMuLuKn00}.

249: 

250: We ignore the influence of any dipolar magnetic interaction and of

251: any flow induced interaction on the (rotation) dynamics of the

252: particles. Thereby collective, collision dominated long-range and

253: long-time hydrodynamic relaxation dynamics of the ensemble of

254: magnetic moments are discarded since only the individual relaxation

255: of each magnetic moment is considered

256: --- albeit in the collectively generated internal magnetic field

257: ${\bf H}$.

258: 

259: For numerical reasons we use a discrete partition of the particle

260: size distribution. Then, without interaction, the magnetization of

261: the resulting mixture of mono-disperse ideal paramagnetic gases is

262: given by ${\bf M}=\sum {\bf M}_j$, where ${\bf M}_j$ denotes the

263: magnetization of the particles with diameter $d_j$. We assume that

264: each magnetic moment and with it each ${\bf M}_j$  obeys a simple

265: Debye relaxation dynamics that drives them in the absence of flow

266: towards their respective equilibrium value ${\bf M}_j^{eq} ({\bf

267: H})$. Then, in the stationary situation resulting from the rigid

268: body rotation with constant ${\bf \Omega}$ the Debye relaxation

269: equation for each sub magnetization is given by

270: \begin{equation} \label{EQ:maggl}

271: {\bf \Omega} \times {\bf M}_j = \frac{1}{\tau_j} [{\bf M}_j - {\bf

272: M}_j^{eq} ({\bf H})] \, .

273: \end{equation}

274: In the absence of interactions the equilibrium magnetization of each

275: species is determined by a Langevin function

276: \begin{equation} \label{EQ:meq}

277: {\bf M}_j^{eq} ({\bf H}) = \chi_j (H) {\bf H} = w_j {\cal L}\left(

278: \frac{\mu_0 \pi M_{sat}^{bulk}}{6k_BT} d_j^3 H  \right) \,

279: \frac{{\bf H}}{H} \, .

280: \end{equation}

281: Here $M_{sat}^{bulk}$ is the bulk-saturation magnetization of the

282: magnetic material. For the magnetic weight $w_j (d_j)$ of species

283: $j$ we take the experimentally determined values in the

284: representation (\ref{meqvonh}).

285: 

286: We should like to draw the attention of the reader to the fact that

287: the magnetization equations (\ref{EQ:maggl}) for the different

288: particle sizes are coupled via the internal field ${\bf H}$:

289: according to Maxwell's equations ${\bf H}={\bf H}_0 - \frac{1}{2}

290: {\bf M}$ is given in terms of the total ${\bf M}=\sum {\bf M}_j$.

291: 

292: In the relaxation rate $1/\tau_j$ we take into account Brownian

293: and N\'eel relaxation processes by adding their rates with equal

294: weight

295: \begin{equation} \label{EQ:tau}

296: \frac{1}{\tau_j} = \frac{1}{\tau_B^j} + \frac{1}{\tau_N^j} \,.

297: \end{equation}

298: The relaxation times depend on the particle size according to

299: $\tau_B^j=\frac{\pi \eta}{2k_BT} (d_j+2s)^3$ and $\tau_N^j=f_0^{-1}

300: \exp\left(\frac{\pi Kd_j^3}{6k_BT}\right)$. Her  $\eta$ is the

301: viscosity, $s$ the thickness of the nonmagnetic particle layer, and

302: $K$ the anisotropy constant. The combined relaxation rate

303: (\ref{EQ:tau}) is dominated by the faster of the two processes. Thus,

304: large particles relax in a Brownian manner with relaxation times of

305: about some $10^{-3}s$, while small particles have the much smaller

306: N\'eel relaxation times. The boundary between N\'eel and Brownian

307: dominated relaxation as a function of particle size $d$ depends

308: sensitively on the anisotropy constant $K$. This is documented in

309: Fig.~\ref{FIG:tau} for the two values $K=15kJ/m^3$ and $K=50kJ/m^3$.

310: For these specific examples the boundaries between N\'eel and

311: Brownian dominated relaxation lie at about $d\simeq 20nm$ and

312: $d\simeq 13nm$, respectively.

313: 

314: %\clearpage

315: %----------------------------------------------------

316: \section{Comparison with experiments}

317: %----------------------------------------------------

318: For the numerical calculations we take typical values for the

319: ferrofluid APG 933 of FerroTec that is used among others in the

320: experiments described in \cite{EmMaWaKiLeLu06}: $M_{sat}^{bulk}=456

321: kA/m$, $\eta = 0.5 Pa\, s$, $s=2nm$, and $f_0=10^9 Hz$. Typical

322: values of $K$ lie between $10 kJ/m^3$ and $50 kJ/m^3$

323: \cite{FaPrCh99,Fa94}. We furthermore use as input the experimental

324: equilibrium magnetization $M_{eq}(H)$ of APG 933 shown in

325: Fig.~\ref{FIG:meq} and the magnetic weights of Fig.~\ref{FIG:vert}

326: obtained with fits to a log-normal distribution or with a

327: regularization method \cite{EmMaWaKiLeLu06}.

328: 

329: From the previous work \cite{LeLu06b,EmMaWaKiLeLu06} we know that

330: single-relaxation time (mono-disperse) models predict the maximum of

331: $M_y$ resp. of $H_y^{sensor}$ to be located roughly at $\Omega \tau

332: \sim 1$. Furthermore, in the experiments \cite{EmMaWaKiLeLu06} done

333: with poly-disperse ferrofluids for frequencies up to $\Omega \simeq

334: 3000 rad/s$ mainly the large  particles contribute to $M_y$ resp. to

335: $H_y^{sensor}$ since their magnetic moments are effectively frozen

336: in the particle's crystal lattice. Only these magnetic moments can

337: be pushed out of the direction of the magnetic field by the combined

338: action of thermally induced rotary Brownian motion and deterministic

339: macroscopic flow in the rotating cylinder. Smaller particles can

340: keep their moment aligned with the magnetic field by the N\'eel

341: process when these particles undergo rotational motion. Finally, the

342: particle diameter that separates N\'eel behavior from Brownian

343: behavior in the size distribution and that thereby determines how

344: many particles contribute to $M_y$ resp. to the experimental

345: signal $H_y^{sensor}$ depends sensitively on the anisotropy constant

346: $K$: The smaller $K$, the smaller is the number of Brownian

347: particles according to Fig.~\ref{FIG:tau}, and the smaller is $M_y$

348: resp. $H_y^{sensor}$.

349: 

350: The above sketched physical picture is corroborated by

351: Fig.~\ref{FIG:monopoly}. There we compare the experimentally

352: obtained $H_y^{sensor}(\Omega)$ (stars) as a function of $\Omega$

353: for a representative externally applied field $H_0=30 kA/m$ with

354: various model variants that take into account the polydispersity to

355: different degrees. This is done for two different anisotropy

356: constants, namely, $K=15 kJ/m^3$ and $K=50 kJ/m^3$ as representative

357: examples. However, the three uppermost curves refer to single time

358: relaxation approximations, each with $\tau = 2ms$

359: \cite{EmMaWaKiLeLu06}: the dotted line with crosses is the result of

360: a strictly monodisperse Debye model while the lines with diamonds

361: refer to polydisperse models, however, with common $\tau_j = \tau

362: =2ms$ taken in Eq.~(\ref{EQ:maggl}) but different magnetic weights

363: $w_j$ obtained either from a lognormal distribution (full line with

364: full diamonds) or from the regularization method (dashed line with

365: open diamonds). The equilibrium magnetization $M_{eq}(H)$ was taken

366: to be the experimental one, the distributions were obtained from

367: this experimental $M_{eq}(H)$ by the lognormal ansatz resp. the

368: regularization method. This, first of all, shows that models with

369: only one relaxation time show roughly the same behavior of

370: $M_y(\Omega)$ irrespective of whether the particle size and magnetic

371: moment distributions are polydisperse or not.

372: 

373: The set of curves with circles and squares in

374: Fig.~\ref{FIG:monopoly} refer to truly polydisperse models,

375: eqs.~(\ref{EQ:maggl} - \ref{EQ:tau}), but different anisotropy

376: constants of the magnetic material [$K=50 kJ/m^3$ (circles), $K=15

377: kJ/m^3$ (squares)]. Again, full lines with full symbols were

378: obtained with a lognormal distribution while dashed lines with open

379: symbols refer to a distribution resulting from the regularization

380: method. Here, one sees that these two distributions with their

381: magnetic weights displayed in Fig.~\ref{FIG:vert} yield very similar

382: results which might not be surprising in view of the fact that both

383: seem to reproduce $M_{eq}(H)$ adequately.

384: 

385: The largest and most important difference between the curves with

386: diamonds (i.e., the single-time models) and the curves with circles

387: and squares (i.e., the genuine polydisperse models) come from the

388: difference in the anisotropy constants of the magnetic material that

389: governs how many particles contribute efficiently as Brownian ones

390: to the transverse magnetization $M_y$ resp. to $H_y^{sensor}$: for

391: smaller $K$ the magnetic moments of fewer particles being Brownian

392: ones may be rotated out of the direction of the magnetic field by

393: the flow in the cylinder.

394: 

395: The curves for $K=15 kJ/m^3$ yield roughly the same maximal size

396: $H_y^{sensor}$ as the experiments --- they could be fine-tuned even

397: further. But then the location, $\Omega^{max}(H_0)$, of the maxima

398: for different $H_0$ is still off from the experimental ones as shown

399: in Fig.~\ref{FIG:hy}(a) and \ref{FIG:max}(b). However, the agreement

400: between the predictions of the polydisperse models of

401: eqs.~(\ref{EQ:maggl} - \ref{EQ:tau}) and the experiments concerning

402: the location $\Omega^{max}(H_0)$

403: %and the maximal value $H_y^{sensor}(\Omega^{max})$

404: can be improved by allowing the relaxation rates $\tau_j$ of the

405: differently sized particles to depend also on the internal field

406: $H$. To demonstate this, we use for simplicity the form

407: \cite{EmMaWaKiLeLu06}

408: \begin{equation}\label{EQ:tauhgamma}

409: \tau_j^{\gamma} (H) = \tau_j \, \frac{2{\cal L}(\gamma H)}{\gamma H

410: - {\cal L}(\gamma H)}

411: \end{equation}

412: with one additional fit parameter $\gamma$. Values of about $\gamma

413: \sim 10^{-4} m/A$ yield maximum locations $\Omega^{max}(H_0)$ that

414: agree well with the experiments as can be seen in

415: Fig.~\ref{FIG:hy}(b) and Fig.~\ref{FIG:max}(b). This generalization

416: of the model (\ref{EQ:maggl} - \ref{EQ:tau}) leaves the peak value

417: $H_y^{sensor}(\Omega^{max})$ almost unchanged, cf,

418: Fig.~\ref{FIG:max}(a).

419: 

420: However, also this augmented polydispersive model reproduces with

421: fixed values of $K$ and $\gamma$ the experimental data only in a

422: small range of $\Omega$ and $H_0$, cf, Fig.~\ref{FIG:hy} and

423: Fig.~\ref{FIG:max}.

424: 

425: %\clearpage

426: %----------------------------------------------------

427: \section{Conclusion}

428: %----------------------------------------------------

429: We have compared the predictions of polydisperse models of the

430: magnetization dynamics of ferrofluids with recent

431: experiments measuring the transverse magnetization component $M_y$

432: of a rotating ferrofluid cylinder. The models use mixtures of

433: mono-disperse ideal paramagnetic gases of differently sized

434: particles. The  magnetization dynamics of the models take into

435: account the rigid body rotation of the fluid combined with a simple

436: Debye relaxation of the magnetic moments of each particle with size

437: dependent Brownian and N\'eel magnetic relaxation times. Thus, in

438: the absence of flow, each magnetic moment and with it each

439: sub-magnetization would be driven independently of the others

440: towards its respective mean equilibrium value that, however, depends

441: on the internal magnetic field ${\bf H}$ being collectively

442: generated by all magnetic moments.

443: 

444: The comparison suggests that mainly the large  particles contribute

445: to $M_y$ since their magnetic moments are effectively frozen in the

446: particle's crystal lattice. Only these magnetic moments can be

447: pushed effectively out of the direction of the magnetic field by the

448: combined action of thermally induced rotary Brownian motion and

449: deterministic macroscopic flow in the rotating cylinder. Smaller

450: particles can keep their moment aligned with the magnetic field by

451: the N\'eel process when these particles undergo rotational motion.

452: 

453: Finally, the particle diameter that separates N\'eel behavior from

454: Brownian behavior in the size distribution and that thereby

455: determines how many particles contribute to $M_y$ resp. to the

456: experimental signal $H_y^{sensor}$ depends quite sensitively on the

457: anisotropy constant $K$ of the magnetic material. $K$ determines how

458: many magnetic moments are "frozen" or "blocked" in particles and

459: thus can be rotated by the rigid body flow: The smaller $K$, the

460: smaller is the number of Brownian particles with frozen moments, and

461: the smaller is the resulting $M_y$. Or, vice versa, a large

462: transverse magnetization can be expected in ferrofluids with large

463: anisotropy constants.

464: 

465: An analysis of the rotation rates $\Omega^{max}(H_0)$ for which

466: $M_y$ is largest indicates that the agreement between experiments

467: and model predictions can be improved by allowing the relaxation

468: rates of the differently sized particles to depend also on the

469: magnetic field $H$.

470: 

471: \begin{acknowledgments}

472: This work was supported by the DFG (SFB 277) and by INTAS(03-51-6064).

473: \end{acknowledgments}

474: 

475: \clearpage

476: 

477: %\bibliography{P_ML}

478: %\bibliographystyle{alpha}

479: %\bibliographystyle{prsty}

480: 

481: 

482: %----------------- Bibliography ---------------------

483:  \begin{thebibliography}{999}

484: %----------------------------------------------------------------

485: 

486: \bibitem{Ro85}

487: R.~E. Rosensweig, {\em Ferrohydrodynamics} (Cambridge University

488: Press,

489:   Cambridge, 1985).

490: 

491: \bibitem{BlCeMa97}

492: E. Blums, A. Cebers, and M.~M. Maiorov, {\em Magnetic Fluids}

493: (Walter de

494:   Gruyter, Berlin, 1997).

495: 

496: \bibitem{Od02a}

497: S. Odenbach, {\em Magnetoviscous Effects in Ferrofluids}, Vol.~m71

498: of {\em

499:   Lecture Notes in Physics} (Springer, Berlin, 2002).

500: 

501: \bibitem{Od02b}

502: {\em Ferrofluids –- Magnetically controllable Fluids and their

503: Applications},

504:   Vol.~594 of {\em Lecture Notes in Physics}, edited by S. Odenbach (Springer,

505:   Berlin, 2002).

506: 

507: \bibitem{McTague69}

508: J.~P. McTague, J.~Chem.~Phys. {\bf 51},  133  (1969).

509: 

510: \bibitem{EmMaWaKiLeLu06}

511: J.~P.~Embs, S.~May, C.~Wagner, A.~V.~Kityk, A.~Leschhorn, and

512: M.~L\"ucke,

513: Phys.~Rev.~E {\bf 73},  036302 (2006).

514: 

515: \bibitem{LeLu06b}

516: A. Leschhorn and M. L\"ucke, Z.~Phys.~Chem. {\bf 220},  219  (2006).

517: 

518: \bibitem{Sh72}

519: M.~I. Shliomis, Sov. Phys. JETP {\bf 34},  1291  (1972).

520: 

521: \bibitem{FeKr99}

522: B.~U. Felderhof and H.~J. Kroh, J. Chem. Phys. {\bf 110},

523: 7403(1999).

524: 

525: \bibitem{Fe00}

526: B.~U. Felderhof, Phys.~Rev.~E {\bf 62},  3848  (2000).

527: 

528: \bibitem{Sh01a}

529: M.~I. Shliomis, Phys.~Rev.~E {\bf 64},  060501  (2001).

530: 

531: \bibitem{MuLi01}

532: H.~W. M\"{u}ller and M. Liu, Phys.~Rev.~E {\bf 64},  061405  (2001).

533: 

534: \bibitem{EmMuKrMeNaMuWiLuHeKn01}

535: J.~Embs, H.~W.~M\"uller, C.~E.~Krill, F.~Meyer, H.~Natter,

536: H.~M\"uller, S.~Wiegand, M.~L\"ucke, R.~Hempelmann, and K.~Knorr,

537: Magnetohydrodynamics {\bf 37},  222 (2001).

538: 

539: \bibitem{WeSt85}

540: T. Weser and K. Stierstadt, Z.~Phys.~B~Cond.~Mat. {\bf 59},  253

541: (1985).

542: 

543: \bibitem{LeLu06a}

544: A. Leschhorn and M. L\"ucke, Z.~Phys.~Chem. {\bf 220},  89  (2006).

545: 

546: \bibitem{EmMuWaKnLu00}

547: J.~Embs, H.~W.~M\"uller, C.~Wagner, K.~Knorr, M.~L\"ucke,

548: Phys.~Rev.~E {\bf 61},  R2196 (2000).

549: 

550: \bibitem{EmMuLuKn00}

551: J. Embs, H.~W. M\"uller, M. L\"ucke, and K. Knorr,

552: Magnetohydrodynamics {\bf 36},  387  (2000).

553: 

554: \bibitem{FaPrCh99}

555: P.~C. Fannin, P.~A. Preov, and S.~W. Charles, J.~Phys. D {\bf

556: 32},1583

557:   (1999).

558: 

559: \bibitem{Fa94}

560: P.~C. Fannin, J.~Magn.~Magn.~Mater. {\bf 136},  49  (1994).

561: 

562: \end{thebibliography}

563: 

564: 

565: \clearpage

566: 

567: %-------------------------Fig. 1 ---------------------------

568: \begin{figure}[htp]

569: \centerline{\includegraphics[width=12cm,angle=0]{fig1.eps}}

570: \caption{Equilibrium magnetization versus internal magnetic field

571: for the ferrofluid AGP 933 of FerroTec. Symbols denote experimental

572: data, solid line fit with a lognormal distribution, and dashed line

573: fit with a regularization method. See text for further information}

574: \label{FIG:meq}

575: \end{figure}

576: %------------------------------------------------------

577: 

578: \clearpage

579: %-------------------------Fig. 2 ---------------------------

580: \begin{figure}[htp]

581: \centerline{\includegraphics[width=12cm,angle=0]{fig2.eps}}

582: \caption{Magnetic weights of the 30 here considered particle sizes

583: ($d_1=1nm$ to $d_{30}=30nm$) obtained from measurements of

584: $M_{eq}(H)$ in Fig.\ref{FIG:meq} by using a lognormal distribution

585: (solid line, circles) and by using a regularization method (dashed

586: line, squares).} \label{FIG:vert}

587: \end{figure}

588: %------------------------------------------------------

589: 

590: \clearpage

591: %-------------------------Fig. 3 ---------------------------

592: \begin{figure}[htp]

593: \centerline{\includegraphics[width=10cm,angle=0]{fig3.eps}}

594: \caption{Schematic plot of the system. The cylindrical sample holder

595: with inner radius $R$ rotates with angular velocity

596: $\boldsymbol{\Omega}$ in the applied static magnetic field

597: ${\bf{H}}_0$ perpendicular to $\boldsymbol{\Omega}$. The

598: magnetization $M_y$ is measured with a Hall sensor.  ${\bf M}$ and

599: ${\bf H}$ denote the magnetization and internal magnetic field of

600: the ferrofluid. Both are constant in space and time.}

601: \label{FIG:sys}

602: \end{figure}

603: %------------------------------------------------------

604: 

605: \clearpage

606: %-------------------------Fig. 4 ---------------------------

607: \begin{figure}[htp]

608: \centerline{\includegraphics[width=11cm,angle=0]{fig4.eps}}

609: \caption{Relaxation times as a function of particle diameter $d$:

610: Brownian (dot-dashed line), N\'eel (dotted line for $K=15kJ/m^3$,

611: dashed line for $K=50kJ/m^3$), and the combination (\ref{EQ:tau})

612: (solid line with crosses for $K=15kJ/m^3$, solid line with plusses

613: for $K=50kJ/m^3$).} \label{FIG:tau}

614: \end{figure}

615: %------------------------------------------------------

616: 

617: \clearpage

618: %-------------------------Fig. 5 ---------------------------

619: \begin{figure}[htp]

620: \centerline{\includegraphics[width=12cm,angle=0]{fig5.eps}}

621: \caption{$H_y^{sensor}$ as function of $\Omega$ for $H_0=30 kA/m$.

622: The three uppermost curves refer to single-time relaxation

623: approximations, each with $\tau = 2ms$: monodisperse Debye model

624: (dotted line with crosses), polydisperse models with common $\tau_j

625: = \tau$ and magnetic weights $w_j$ obtained with a lognormal

626: distribution (full line with full diamonds) and with a

627: regularization method (dashed line with open diamonds). Curves with

628: circles and squares refer to truly polydisperse models [

629: eqs.~(\ref{EQ:maggl} - \ref{EQ:tau})] with $K=50 kJ/m^3$ (circles)

630: and $K=15 kJ/m^3$ (squares). Full lines with full symbols were

631: obtained with a lognormal distribution. Dashed lines with open

632: symbols refer to a distribution resulting from the regularization

633: method. In all models the equilibrium magnetization $M_{eq}(H)$ was

634: taken to be the experimental one.} \label{FIG:monopoly}

635: \end{figure}

636: %------------------------------------------------------

637: 

638: \clearpage

639: %-------------------------Fig. 6 ---------------------------

640: \begin{figure}[htp]

641: \centerline{\includegraphics[width=12cm,angle=0]{fig6.eps}}

642: \caption{$H_y^{sensor}$ as function of $\Omega$ for $H_0=8.6 kA/m$

643: (squares), $15 kA/m$ (triangles), $30 kA/m$ (circles), and $60 kA/m$

644: (diamonds). Full symbols denote experimental data. Open symbols

645: refer to the polydisperse model [eqs.~(\ref{EQ:maggl} -

646: \ref{EQ:tau})] with lognormal distribution and $K=15 kJ/m^3$: in (a)

647: the relaxation times $\tau_j$ are independent of $H$, in (b) they

648: are replaced by $\tau_j^{\gamma}(H)$ (\ref{EQ:tauhgamma}) with

649: $\gamma = 10^{-4}m/A$. } \label{FIG:hy}

650: \end{figure}

651: %------------------------------------------------------

652: 

653: \clearpage

654: %-------------------------Fig. 7 ---------------------------

655: \begin{figure}[htp]

656: \centerline{\includegraphics[width=12cm,angle=0]{fig7.eps}}

657: \caption{Maximum magnitude $max H_y^{sensor}=

658: H_y^{sensor}(\Omega^{max})$ (a) and location of the maximum

659: $\Omega^{max}$ (b) as functions of the external field $H_0$. Stars

660: show experimental data. Lines refer to the results of polydisperse

661: models with a lognormal distribution and $K=15 kJ/m^3$:

662: $H$-independent relaxation times (solid); $H$--dependencies with

663: $\gamma = 0.8\cdot10^{-4}m/A$ (dotted), $\gamma = 10^{-4}m/A$

664: (dashed), and $\gamma = 1.1\cdot 10^{-4}m/A$ (dot-dashed). In (a)

665: the differences between the lines are too small to be seen.}

666: \label{FIG:max}

667: \end{figure}

668: %------------------------------------------------------

669: 

670: \end{document}

671: