1: \documentclass[12pt]{iopart}

2: 

3: \usepackage[dvips]{graphicx}

4: \usepackage{iopams}

5: \usepackage{hyperref}

6: \begin{document}

7: 

8: \sloppy

9: 

10: \title[A new model for simulating colloidal dynamics]

11: {A new model for simulating colloidal dynamics}

12: 

13: \author{Vladimir Lobaskin\footnote[1]{

14: To whom correspondence should be addressed 

15: (lobaskin@mpip-mainz.mpg.de)} 

16: and Burkhard D\"unweg}

17: 

18: \address{Max Planck Institute for Polymer Research, 

19: Ackermannweg 10,

20: D-55128 Mainz, Germany}

21: 

22: \begin{abstract}

23: We present a new hybrid lattice-Boltzmann and Langevin molecular

24: dynamics scheme for simulating the dynamics of suspensions of

25: spherical colloidal particles. The solvent is modeled on the level of

26: the lattice-Boltzmann method while the molecular dynamics is done for

27: the solute. The coupling between the two is implemented through a

28: frictional force acting both on the solvent and on the solute, which

29: depends on the relative velocity. A spherical colloidal particle is

30: represented by interaction sites at its surface. We demonstrate that

31: this scheme quantitatively reproduces the translational and rotational

32: diffusion of a neutral spherical particle in a liquid and show

33: preliminary results for a charged spherical particle. We argue that

34: this method is especially advantageous in the case of charged

35: colloids.

36: \end{abstract}

37: 

38: %Uncomment for PACS numbers title message

39: %\pacs{00.00, 20.00, 42.10}

40: 

41: % Uncomment for Submitted to journal title message

42: \submitto{\NJP}

43: 

44: % Comment out if separate title page not required

45: \maketitle

46: 

47: \section{Introduction}

48: \label{sec:intro}

49: 

50: Understanding the dynamics of colloidal dispersions has been a

51: long-standing problem in condensed matter physics. Despite the

52: continuous need for accurate theoretical predictions concerning

53: particle mobilities or sedimentation rates, the development of the

54: field meets a nearly unpassable barrier. The many-body character of

55: the hydrodynamic interactions (HI) between the colloidal particles

56: (i.~e. their correlated motion as a result of solvent-mediated fast

57: momentum transport) poses a serious challenge for analytic theory. On

58: the other hand, the rapid growth of available computer power in the

59: last decades has made it more and more feasible to descend to a more

60: basic description of colloidal systems rather than trying to develop

61: an adequate macroscopic description. Such a step was made recently 

62: for the statics of charged colloids, where simulations of the

63: primitive electrolyte model, taking the particle character of the

64: charges and fluctuations in the charge distribution fully into

65: acount, resolved some important strong correlation issues and thus

66: have resulted in a significant refinement of the existing theories of

67: screening \cite{Hansen1}. In a similar fashion, many-body effects in

68: the HI of suspensions of only moderate density are quite

69: important. Although their analytical form is, in principle, known in

70: terms of a multipole expansion \cite{Mazur}, it is numerically very

71: cumbersome to take these higher-order many-body terms into account in

72: a Brownian or Stokesian dynamics \cite{Brady} simulation, where only

73: the positions of the colloidal particles enter. For this reason, it is

74: more convenient and for large particle numbers also more efficient,

75: to take these effects into account by simulating the solvent degrees

76: of freedom, and, in particular, the momentum transport through the solvent

77: explicitly. An additional bonus is that retardation effects, if

78: present, are automatically taken into account.

79: 

80: Standard Molecular Dynamics (MD), where just Newton's equations of

81: motion are integrated numerically, is of course able to simulate HI

82: correctly. However, it is very time-consuming to simulate the motion

83: of a huge number of particles explicitly in such detail. The time step

84: is governed by the local oscillations of the solvent particles in

85: their temporary ``cages'', which is much faster than the motion of the

86: colloidal particles. This information is however not needed for

87: correctly reproducing HI. One just needs an appropriate mechanism for

88: momentum transfer on somewhat larger time scales (however still small

89: compared to the motion of the colloidal particles). This can be done

90: in various ways: Dissipative Particle Dynamics (DPD) \cite{Warren} is

91: essentially MD, where however the particles are made very soft in

92: order to increase the time step, and where a momentum-conserving

93: Langevin thermostat is added \cite{Thoso}. Another possibility is multi-particle 

94: collision dynamics \cite{Malevanets}, where the solvent is modeled as

95: an ideal gas, and collisions are modeled as simple stochastic updating

96: rules which conserve energy and momentum. Grid-based methods can be

97: either the direct solution of the Navier-Stokes equation by a finite

98: difference method, or the Lattice Boltzmann (LB) \cite{Succi}

99: approach, where essentially a linearized Boltzmann equation is solved

100: in a fully discretized version (i.~e. space, time, and velocities are

101: discretized).

102: 

103: In our opinion, all of these approaches have their advantages and

104: disadvantages. Nevertheless, it seems that all of them are roughly

105: equivalent with respect to their computational efficiency, and with

106: respect to the level of accuracy of describing the physical

107: phenomena. Therefore, we believe that the choice of method is mainly a

108: matter of convenience.

109: 

110: For our model we have chosen the LB route, mainly because this

111: algorithm is particularly simple to implement and parallelize. Moreover, we were inspired by previous highly successful simulations of hard-sphere colloidal suspensions \cite{Ladd0,Ladd1,Ladd2,Ladd3,Ladd4,Frenkel1,Frenkel2,Frenkel3,Frenkel4}. A further advantage of a grid-based method is that thermal fluctuations can be both turned on and off, depending on the physical situation under consideration (in a particle method they are always present). This approach is essentially a hybrid method, where the solvent is modeled via LB, while the colloidal particles are run with MD.

112: 

113: The original approach by Ladd \cite{Ladd0,Ladd1,Ladd2,Ladd3,Ladd4}

114: models the colloidal particles as extended hollow spheres, while stick

115: boundary conditions at the surface are implemented (roughly spoken)

116: via bounce-back collision rules. One should note that for moving spheres

117: this involves some minor fluctuations 

118: in the fluid mass, which is included within a sphere: The moving

119: shell incorporates some new fluid at the front, while it releases

120: mass at the rear. The fluctuations are a natural effect of the

121: thermal density fluctuations of the fluid, and of the lattice

122: discretization. Using this method, the translational and the rotational

123: dynamics of colloidal spheres were accurately described.

124: 

125: This model has been recently extended to the case of charged systems

126: \cite{Horbach1}, where the colloidal particles carry a charge, while

127: the counterions and salt ions are taken into account as LB

128: populations, such that the electrostatics is essentially taken into

129: account on the level of the Poisson-Boltzmann equation. This method

130: has two disadvantages: firstly, the discrete nature of the ions and

131: correlations beyond the Poisson-Boltzmann level are not taken into

132: account; secondly, one cannot avoid a leakage of charge into (and out 

133: of) the sphere (as in the case of mass), such that it is hard

134: (if not impossible) to maintain a well-defined Debye layer of charges

135: around it \cite{Horbach2}. For these reasons, it is advisable

136: to take the counterions explicitly into account. The disadvantage,

137: however, is that only rather small size ratios (size of colloidal

138: particle vs. size of ion) are easily accessible. Nevertheless, we

139: believe it is useful to study such a system, with respect to both its

140: equilibrium and its nonequilibrium properties. The purpose of the

141: present paper is to describe first steps in the development and

142: validation of the corresponding model.

143: 

144: For the coupling of the small ions to the LB hydrodynamics, one would

145: like to simply model the former as point particles. Fortunately, such

146: a coupling has been recently developed by Ahlrichs and D\"unweg

147: \cite{Ahlrichs1,Ahlrichs2} with the purpose of studying the dynamics

148: of polymer solutions \cite{Ahlrichs3}. Each Brownian point particle is

149: assigned a phenomenological friction coefficient $\zeta$, and the

150: coupling is implemented just as a dissipative force $\vec F = - \zeta

151: (\vec V - \vec u)$ acting on the particle, where $\vec V$ is the

152: particle velocity, while $\vec u$ is the solvent flow velocity at the

153: particle's position, obtained via linear interpolation from the

154: surrounding lattice sites. Furthermore, the LB variables at these

155: sites are adjusted to ensure momentum balance, and thermal

156: fluctuations are added to both types of degrees of freedom. For more

157: details on implementation and validation of this method, see

158: Refs. \cite{Ahlrichs1,Ahlrichs2}.

159: 

160: It is then convenient to use this coupling not only for the ions, but

161: also for the colloidal particles, whose larger size is taken into

162: account by modeling them as some arrangement of interaction sites. For

163: reasons of efficiency, we take only points at the surface of the

164: colloid. Furthermore, the coupling constant $\zeta$ is chosen rather

165: large, which approximates stick boundary conditions. It should be

166: noted that the dissipative nature of the coupling does not model any

167: squeezing-out of solvent, hence the sphere is filled with the same

168: amount of solvent regardless if it is hollow or if it were filled with

169: particles. The solvent within the sphere follows the motion of the

170: surrounding shell (with some minor time lag, see below), and therefore

171: one can view the solvent inside the sphere as just belonging to the

172: colloidal particle. Adding further particles in the sphere's volume

173: would have no effect except coupling the fluid within the sphere even

174: tighter to its motion. Within our desired level of accuracy, this

175: turned out not to be necessary. The long-time motion of the sphere is

176: the same as that of a corresponding hard sphere, as we will show in

177: the present paper.

178: 

179: We analyze basic dynamic properties of such a model colloid, where we

180: restrict ourselves to the case of a single sphere. A detailed analysis

181: is done for the neutral case, while some preliminary data for the case

182: of a charged sphere are presented. We show that the (neutral) model

183: exhibits the essential dynamical features of a spherical colloidal

184: particle in a liquid. In our view, our method provides comparable

185: efficiency to Ladd's approach, while being quite straightforward to

186: implement. Furthermore, it provides substantial flexibility with

187: respect to the properties of the colloidal surface, namely, deformable,

188: permeable, and non-stick surfaces can be easily simulated.

189: 

190: The remainder of this article is organized as follows: In

191: Sec. \ref{sec:model}, we describe our simulation model, while

192: Sec. \ref{sec:results} contains the numerical results on translational

193: and rotational diffusion. Finally, Sec. \ref{sec:summary} concludes

194: with a brief summary.

195: 

196: 

197: \section{Model}

198: \label{sec:model}

199: 

200: Our hybrid simulation method involves two subsystems: the solvent that

201: is modeled via LB with fluctuating stress tensor (i.~e. we run a

202: constant-temperature version of the LB method) and a Langevin MD

203: simulation for the particles immersed in the solvent. The LB

204: simulation is performed using the 18-velocity model \cite{Ladd1},

205: using the protocol described in \cite{Ahlrichs1,Ahlrichs2}. The fluid

206: simulation consists of collision and propagation steps, the former

207: being performed with inclusion of the momentum transfer from the

208: solute particles (surface beads, and, for charged systems, ions).

209: 

210: \begin{figure}

211: \begin{center}

212: \vskip 0.1in

213: \includegraphics[clip,width=0.5\textwidth]{berry3}

214: \end{center}

215: \caption{

216: Raspberry-like model of a colloidal sphere. There is a central large

217: bead of radius $R=3$ and charge $Z=10$. The small beads of radius 1

218: are connected with their nearest neighbors on the surface via FENE

219: bonds. A repulsive soft-core potential is also operating between all

220: the monomers. The counterions are moving freely in space and interact

221: with the central bead via the Coulomb potential and with the surface

222: beads via the repulsive LJ potential. 

223: \href{http://www.mpip-mainz.mpg.de/~lobaskin/TR6/Z20.mpg}{Movie}

224: }

225: \label{fig:sphere}

226: \end{figure}

227: 

228: The colloidal particle is represented by a two-dimensional tethered

229: bead-spring network consisting of $100$ beads, which is wrapped around

230: a ball of a radius $\sigma_{cs}$ (for notation, see below), so that

231: the whole construction resembles a raspberry (see

232: Fig. \ref{fig:sphere}). The network connectivity is maintained via

233: finitely extendable nonlinear elastic (FENE) springs,

234: %

235: \begin{equation}

236: V_{FENE} (r) = - \frac{k R_0^2}{2} \ln

237: \left(1 - \left( \frac{r}{R_0}\right)^{2} \right) ,

238: \label{eq:fene}

239: \end{equation}

240: %

241: where $k$ is the spring constant, and $R_0$ the maximum bond

242: extension. Furthermore, the beads repel each other by a modified

243: Lennard-Jones (LJ) potential

244: %

245: \begin{equation}

246: V_{LJ} (r) = \left\lbrace \begin{array} {ll} 

247: 4\epsilon_{ij} \left(  \left( \frac{\sigma_{ij}}{r} \right)^{12} 

248: - \left( \frac{\sigma_{ij}}{r} \right)^{6} + 

249: \frac{1}{4} \right) & r<2^{1/6} \sigma_{ij}  \\

250: 0  & r\ge 2^{1/6} \sigma_{ij} . \end{array} \right.

251: \label{eq:lj}

252: \end{equation}

253: %

254: An additional repulsive LJ bead is introduced at the center of the

255: sphere in order to maintain its shape. In Eq. \ref{eq:lj}, $i,j$

256: denote either a central (``c'') or a surface (``s'') bead. The unit

257: system is completely defined by the surface bead parameters by setting

258: $\epsilon_{ss}$, $\sigma_{ss}$, and the surface bead mass $m_s$ to

259: unity. The interaction between the central bead and the surface beads

260: is described by $\sigma_{cs} = 3$, which is thus the sphere radius,

261: and $\epsilon_{cs}=8$. Furthermore, the FENE spring constant for the

262: surface beads is $k=300$ and the maximum bond extension is $R_0=

263: 1.25$. To simulate a charged colloidal particle, we place the charge

264: at the central bead, and add an appropriate number of counterions (LJ

265: beads with ``s'' properties) outside the sphere. The electrostatic

266: interaction is taken into account via the Coulomb potential

267: %

268: \begin{equation}

269: V_{el} (r) = \lambda_B k_B T \frac{q_i q_j}{r}

270: \label{eq:electrostatics}

271: \end{equation}

272: %

273: between the various charges, where the standard Ewald summation

274: technique \cite{Allen} is applied. In Eq. \ref{eq:electrostatics},

275: $\lambda_B=e^2/ \left(4\pi\varepsilon_0 \varepsilon_r k_B T\right)$ is

276: the Bjerrum length, $k_B$ the Boltzmann constant, $q_i$ the charge of

277: species $i$ in units of the elementary charge $e$, and $T$ the

278: temperature.

279: 

280: The LB lattice constant is chosen as one (in our LJ unit system), and

281: the fluid is simulated in a cubic box with periodic boundary

282: conditions. The force between the LB fluid and the surface beads

283: (or ions) is given by

284: %

285: \begin{equation}

286: \vec{F}= - \zeta \left( \vec{V} - \vec{u} \right) + \vec{f} .

287: \label{eq:4}

288: \end{equation}

289: %

290: Here, $\zeta$ is the ``bare'' \cite{Ahlrichs2} friction coefficient,

291: $\vec{V}$ and $\vec{u}$ are the velocities of the bead and the fluid

292: (at the position of the bead), respectively, while $\vec{f}$ is a

293: Gaussian white noise force with zero mean, whose strength is given via

294: the standard fluctuation-dissipation theorem

295: \cite{Ahlrichs1,Ahlrichs2} to keep the surface beads and ions at

296: the same temperature as the solvent. The central bead is not

297: coupled to the solvent (here $\zeta = \vec f = 0$); as discussed in the

298: Introduction, the behavior of the model would change only marginally

299: if such a coupling were included. In our simulation we used a friction

300: constant $\zeta = 20$, a temperature $k_B T = 1$, a fluid mass density

301: $\rho = 0.85$, and a kinematic viscosity $\nu=3$, resulting in a

302: dynamic viscosity $\eta=2.55$. At least 20000 MD steps were performed

303: to equilibrate the initial random bead configuration before the

304: interaction with the LB solvent was turned on. Further details on the

305: method can be found in \cite{Ahlrichs1,Ahlrichs2}.

306: 

307: 

308: \section{Results}

309: \label{sec:results}

310: 

311: We test the simulation method against basic relations for an isolated

312: sphere in solvent. First, we look at the center of mass' velocity

313: relaxation. The simplest experiment to perform is a ``kick''. The

314: sphere is placed in a LB fluid at rest (i.~e. without thermal noise),

315: and at time $t = 0$ all particles of the sphere are assigned an

316: identical velocity $\vec V = 1$ in $x$ direction. Fig. \ref{fig:vcf2}

317: monitors the time behavior of the sphere's center of mass velocity,

318: normalized by the initial value. According to linear response theory,

319: this relaxation function must be identical to the normalized

320: center-of-mass velocity autocorrelation function for Brownian motion

321: in thermal equilibrium, if the initial kick is weak enough. This is

322: indeed satisfied, as a comparison of the two curves in

323: Fig. \ref{fig:vcf2} shows. For the experiment in thermal equilibrium,

324: we performed 10 runs with different random number generator

325: initializations, in order to reduce the statistical uncertainty.

326: 

327: It is well-known that simulations of Brownian motion in a hydrodynamic

328: solvent are always strongly affected by finite size effects. The

329: diffusion constant, and therefore also the relaxation function, depend

330: on the linear system size $L$ due to hydrodynamic interactions with

331: the periodic images. The diffusion constant exhibits a finite-size

332: correction of order $R / L$ \cite{Ahlrichs2,Burkhard}. Asymptotic

333: behavior can therefore only be expected for $R / L \ll 1$, and this is

334: why we performed the experiment in a rather large box of size $L =

335: 80$. For the same reason, the equivalence between ``kick'' experiment

336: and Brownian motion will only hold if the comparison is done for the

337: same box sizes.

338: 

339: \begin{figure}

340: \vskip 0.1in

341: \begin{center}

342: \includegraphics[clip, width=0.75\textwidth]{vcf2a.eps}

343: \end{center}

344: \caption{

345: Normalized translational velocity of the center of mass of the

346: colloidal sphere in a "kick" experiment at $k_B T = 0$, and its

347: normalized velocity autocorrelation function for Brownian motion at

348: $k_B T = 1$. The dotted curve shows the exponential decay derived from

349: the Stokes law and the dashed curve the expected long-time asymptotic

350: behavior.}

351: \label{fig:vcf2}

352: \end{figure}

353: 

354: \begin{figure}

355: \begin{center}

356: \vskip 0.1in

357: \includegraphics[clip, width=0.75\textwidth]{msd}

358: \end{center}

359: \caption{

360: Mean-square displacement of the center of mass of a neutral sphere at

361: $k_B T = 1$ and different simulation box sizes. }

362: \label{fig:msd}

363: \end{figure}

364: 

365: One clearly sees that the initial decay of the relaxation function is

366: characterized by two relaxation processes, one initial fast decay

367: followed by a somewhat slower relaxation. Qualitatively, this may be

368: explained as follows: A compact sphere of radius $R$ and mass $M$

369: should exhibit a velocity relaxation which, after transient ballistic

370: motion, is initially characterized by an exponential decay $\exp

371: \left( - t / \tau \right)$, with a relaxation time $\tau = M /

372: \zeta_{tot}$. Here, $\zeta_{tot}$ is the total friction

373: coefficient, which we estimate via Stokes' law for stick boundary

374: conditions as $\zeta_{tot} \approx 6 \pi \eta R \approx 144$. However,

375: the effective mass is time-dependent: While initially only the mass of

376: the beads $M = 101$ contributes, at later times the fluid within the

377: sphere is dragged as well, such that then the mass is roughly

378: estimated as $M \approx 101 + 4 \pi \rho R^3 / 3 \approx 214$.  This

379: gives rise to the initial and final relaxation times $\tau_{in}

380: \approx 0.7$ and $\tau_{fin} \approx 1.5$. However, the initial

381: relaxation time cannot be observed, since in the extreme short-time

382: regime ballistic effects play a role. Conversely, the decay with

383: $\tau_{fin}$ is clearly visible (see Fig. \ref{fig:vcf2}).

384: 

385: \begin{figure}

386: \begin{center}

387: \vskip 0.1in

388: \includegraphics[clip, width=0.75\textwidth]{d1}

389: \end{center}

390: \caption{

391: The long-time self-diffusion coefficient and the rotational diffusion

392: coefficient of the colloidal sphere (normalized by the asymptotic

393: Stokes-Einstein and Stokes-Einstein-Debye values) as functions of the inverse size of the primary simulation cell. }

394: \label{fig:ds}

395: \end{figure}

396: 

397: After $t \approx 1$, the famous long-time tail \cite{Alder,Hansen2}

398: (normalized relaxation function $V(t) / V(0) = B t^{-3/2}$) sets

399: in. The physical mechanism of this slow relaxation is the fact that

400: the momentum is conserved and hence transported away diffusively from

401: the particle \cite{Hansen2}. For this reason, it can only be observed

402: in a sufficiently large box. The prefactor of the power law is known

403: in the colloidal limit, $B = (1 / 12) (N m / \rho) (\pi \nu)^{-3/2}$,

404: where $N$ is the number of beads, and $m$ the bead mass

405: \cite{Hinch,Cichocki}. It should be noted that the prefactor of the

406: power-law decay of the {\em unnormalized} velocity autocorrelation

407: function does not depend on the properties of the sphere at all, but

408: only on the temperature, and the hydrodynamic properties of the

409: solvent \cite{Hinch,Cichocki}. The mass dependence of $B$ results only

410: from the normalization, i.~e. the value of $\left< V^2 \right>$

411: according to the equipartition theorem, $\left< V^2 \right> = 3 k_B T

412: / (N m)$. From this consideration it is clear that the {\em

413: short-time} value of the sphere's mass ($N m$) enters (at $t = 0$

414: the shell and the inner fluid are not yet coupled). For our

415: simulation parameters, we find $B = 0.342$. As Fig. \ref{fig:vcf2}

416: shows, the data exhibit the expected behavior very nicely.

417: 

418: Figure \ref{fig:msd} illustrates the finite size effect in the diffusive

419: properties by plotting the mean square displacement of the colloidal

420: sphere as a function of time, for two different box sizes $L = 6.33 R$

421: and $L = 20 R$, for Brownian motion at $k_B T = 1$. In the long-time

422: regime, the slope is given by $6 D^S$, where $D^S$ is the

423: self-diffusion coefficient. The figure clearly shows that $D^S$

424: increases with box size. This can be explained in terms of

425: hydrodynamic interactions with the periodic images, or, equivalently,

426: in terms of suppression of long-wavelength hydrodynamic modes

427: \cite{Ahlrichs2,Burkhard}. We have therefore performed a systematic

428: finite-size analysis, and measured $D^S$ for various box sizes by

429: integrating the velocity autocorrelation function over time

430: \cite{Hansen2}. In Fig. \ref{fig:ds} we plot $D^S / D_0$ 

431: ($D_0$ denoting the asymptotic Stokes-Einstein value $D_0 = k_B T / (6

432: \pi \eta R) = 6.9 \times 10^{-3}$) as a function of $1 / L$; the

433: expected linear behavior \cite{Burkhard} is clearly seen.

434: 

435: \begin{figure}

436: \begin{center}

437: \vskip 0.1in

438: \includegraphics[clip, width=0.75\textwidth]{rotnew}

439: \end{center}

440: \caption{

441: Decay of angular velocity of the colloidal sphere. The different

442: curves are marked by the primary simulation box sizes. The solid

443: curves show the expected exponential decay according to the Debye law

444: and the long-time asymptotic behavior.}

445: \label{fig:avcf}

446: \end{figure}

447: 

448: We now look at the relaxation of the rotational motion. We performed a

449: similar kick experiment as described above, where now an initial

450: angular velocity $\omega_0 = 1$ was provided to the sphere. The data

451: for the normalized decay function $\omega(t) / \omega (0)$ are

452: presented in Fig. \ref{fig:avcf}. The sphere dynamics shows a

453: characteristic ``raw egg'' (damped) rotation pattern with a fast

454: initial decay and a subsequent slower one. For the rotational

455: relaxation of a hard sphere with moment of inertia $I$ and rotational

456: friction coefficient $8 \pi \eta R^3$, theory predicts a decay

457: according to the Debye law $\omega (t) / \omega (0) =

458: \exp \left( - t / \tau_r \right)$, where the relaxation time is given

459: by $\tau_r = I / (8 \pi \eta R^3)$ \cite{Debye}. Similarly to the

460: effective mass, the effective moment of inertia is expected to

461: increase as a function of time, due to the time-delayed dragging of

462: the fluid. Initially we expect a hollow-sphere value of roughly

463: $I_{in} \approx (2/3) M R^2 \approx 600$, where we used the mass of

464: the outer shell $M = 100$. A more accurate value is obtained by

465: direct summation over the contributing beads, yielding $I_{in} = 546$.

466: In the late-time regime the moment of

467: inertia is expected to be the compact-sphere value $I_{fin} \approx

468: (2/5) M R^2 \approx 770$, where the total mass $M \approx 214$ has

469: been used. This results in relaxation times $\tau_{r,in} = 0.34$ and

470: $\tau_{r,fin} = 0.44$. These values are roughly consistent with a fit

471: to the data in the interval $0.1 < t < 1.0$, which yields a somewhat

472: larger relaxation time $\tau_r = 0.68$ (see Fig. \ref{fig:avcf}).

473: Given the general inaccuracy of the

474: estimates, and the crossover of the Debye relaxation function both to

475: short-time ballistic behavior, and long-time hydrodynamic behavior,

476: this deviation is not too surprising.

477: 

478: Similar to the translational motion, the exponential decay is then

479: followed by a power-law long-time tail. Theory predicts $\omega (t) /

480: \omega (0) = (\pi I / \rho) (4 \pi \nu t)^{-5/2}$ \cite{Hauge}. As we

481: have seen before, the long-time tail in the translational case is

482: governed by the short-time mass as a result of normalization.

483: Similarly, the rotational tail must be controlled by the {\em

484: short-time} moment of inertia $I_{in}$, which also governs the mean

485: square fluctuations of the angular velocity via the equipartition

486: theorem $\left< \omega^2 \right> = 3 k_B T / I_{in}$. For this reason,

487: we can determine $I_{in}$ by fitting a $t^{-5/2}$ law to the data,

488: which is much more accurate than our rough geometrical estimate. The

489: fit results in $I_{in} = 533$, which is reasonably consistent.  If we

490: insert this value into the relaxation time expression, we obtain

491: $\tau_r = 0.64$, in quite good agreement with the data. It hence seems

492: that fluid dragging effects are not yet very important for the initial

493: Debye relaxation.

494: 

495: At longer times, one can again notice a significant finite-size

496: effect: the curves obtained in the smaller simulation box depart from

497: the asymptotic power-law line earlier, i.~e. the long-time diffusion

498: is hindered at the small system sizes. In fact, the power-law regime

499: is inaccessible at $L = 10$ while for $L = 100$ it extends up to $t =

500: 200$. This value characterizes the interval after which the particle

501: starts feeling its own periodic images. The rotational diffusion

502: constant $D^R$ is given by the Green-Kubo integral \cite{Berne1}

503: %

504: \begin{equation}

505: D^R = \frac{1}{3} \int_0^\infty dt\, \left< \vec \omega (t)

506: \cdot \vec \omega (0) \right> .

507: \end{equation}

508: %

509: We can evaluate this by again making use of linear response

510: theory: The correlation function $\left< \vec \omega (t)

511: \cdot \vec \omega (0) \right>$ is identical to the relaxation

512: function presented in Fig. \ref{fig:avcf}, multiplied with the initial

513: value $\left< \omega^2 \right>$, which we know from the equipartition

514: theorem (see above). Using this approach, we have calculated $D^R$ for

515: different box sizes $L$. In an infinite fluid, $D^R$ has the

516: Stokes-Einstein-Debye value $D^R = k_B T / (8 \pi \eta R^3) = 0.58 \times

517: 10^{-3}$, towards which the results indeed converge for $L \to

518: \infty$. Again a $1 / L$ behavior (but weaker than for translational

519: diffusion) is observed, as shown in Fig. \ref{fig:ds}. It should be

520: noted that the accuracy of the data is slightly hampered by the fact

521: that we had to use a somewhat arbitrary criterion for cutting off the

522: integral function, which does not arrive at a constant value but

523: rather ends up with a linear increase originating from a small nonzero

524: constant angular velocity in the long-time limit (as a manifestation

525: of the finite-size effect).

526: 

527: \begin{figure}

528: \begin{center}

529: \vskip 0.1in

530: \includegraphics[clip, width=0.75\textwidth]{vcfzz}

531: \end{center}

532: \caption{

533: Center of mass velocity autocorrelation functions for a neutral and

534: a charged colloidal sphere. The Green-Kubo integral function

535: is also shown as solid curve for $Z=0$ and dashed curve for $Z=10$.

536: }

537: 

538: \label{fig:zz}

539: \end{figure}

540: 

541: Finally, we have started to study the effect of charge on the

542: self-diffusion of the colloidal particle. We performed simulations of

543: a sphere with central charge $Z_C = 10$ in an LB box of size $L =

544: 40$. Ten counterions of charge $-1$ were also added. The Bjerrum

545: length was set to $2$. Technically, the simulation ran without

546: any problems just as well as for the neutral system. In Fig.

547: \ref{fig:zz} we compare the decay of the velocity

548: autocorrelation function for the neutral and charged spheres. The

549: difference between the two is not detectable within our error bars at

550: this charge. We cannot say yet much about the influence of the charges

551: on the dynamic properties; we do however expect that for strongly

552: charged systems the Debye layer will effectively increase the

553: hydrodynamic radius and slow down the diffusion. More work needs to be

554: done to resolve this issue.

555: 

556: \section{Summary}

557: \label{sec:summary}

558: 

559: In summary, we introduced and tested a new model for simulating

560: colloidal dynamics with inclusion of the hydrodynamic effects on the

561: level of the lattice-Boltzmann equation. The suggested

562: ``raspberry''-like colloidal object exhibits the essential features of

563: the diffusion of a spherical colloidal particle. For the first time we

564: combined a model containing explicit charges with accurate treatment

565: of the hydrodynamics. We expect this method to have some advantages

566: over the previously applied scheme \cite{Horbach1} which builds upon

567: the model of Ladd \cite{Ladd0} and has the problem of charge

568: leakage. In our method: (i) the stick boundary conditions are realized

569: not strictly, i.~e. the particle surface is softly bound to the fluid;

570: (ii) the method is easily adaptable to the multiple timestep method

571: where the ionic action on the colloid can be averaged out somewhat

572: within the integration step for the solvent. We should note that such

573: a hybrid simulation requires still some computational effort --- two

574: million hybrid steps for the colloid consisting of 100 beads in a box of

575: $20 \times 20 \times 20$ LB cells take about 12 hours on a 2 GHz Pentium

576: 4 processor, while it is expected to run a few times faster on more

577: sophisticated hardware. Here, one step is a MD step of the total system

578: of beads, plus one LB update of the whole lattice, where the latter part

579: completely dominates the CPU effort. Some improvement is possible by not updating the LB fluid at every MD step, as it was done here. For dense

580: systems, where the MD part is no longer negligible, one should also

581: optimize the interaction potentials. Finally, we would like to mention

582: that the algorithm can be fully parallelized and we are working on the

583: parallel version.

584: 

585: \ack 

586: We thank Christian Holm, J\"urgen Horbach and Kurt Kremer

587: for stimulating discussions, and the latter also for critical

588: reading of the manuscript. This work was funded by the SFB

589: TR 6 of the Deutsche Forschungsgemeinschaft.

590: 

591: 

592: \section{References}

593: \begin{thebibliography}{15}

594: \bibitem{Hansen1} 

595:   Hansen J-P and L\"owen H 2000 {\it Ann. Rev. Phys. Chem.} {\bf 51} 209

596: \bibitem{Mazur} 

597:   Mazur P and Van Saarloos W 1982 {\it Physica A} {\bf 115} 21

598: \bibitem{Brady}

599:   Brady J F and Bossis G 1998 {\it Ann. Rev. Fluid Mech.} {\bf 20} 111

600: \bibitem{Warren}

601:   Groot R and Warren P 1997 {\it J. Chem. Phys.} {\bf 107} 4423

602: \bibitem{Thoso}

603:   Soddemann T, D\"unweg B, and Kremer K 2003 {\it Phys. Rev. E} in press

604: \bibitem{Malevanets} 

605:   Malevanets A and Kapral R 1999 {\it J. Chem. Phys.} {\bf 110} 8605

606: \bibitem{Succi}

607:   Succi S 2001 {\it The Lattice Boltzmann Equation for Fluid

608:   Dynamics and Beyond} (Oxford: Oxford University Press)

609: \bibitem{Ladd0} 

610:   Ladd AJC 1993 {\it Phys. Rev. Lett.} {\bf 70} 1339

611: \bibitem{Ladd1} 

612:   Ladd AJC 1994 {\it J. Fluid Mech.} {\bf 271} 285

613: \bibitem{Ladd2} 

614:   Ladd AJC 1994 {\it J. Fluid Mech.} {\bf 271} 311

615: \bibitem{Ladd3} 

616:   Ladd AJC and Verberg R 2001 {\it J. Stat. Phys.} {\bf 104} 1191

617: \bibitem{Ladd4} 

618:   Ladd AJC, Hu Gang, Zhu JX, and Weitz DA 1995 {\it Phys. Rev. E} {\bf 52} 6550

619: \bibitem{Frenkel1}

620:  Hagen MHJ, Frenkel D, and Lowe CP 1998 {\it Physica A} {\bf 109} 275

621: \bibitem{Frenkel2}

622:  Lowe CP, Frenkel D, and Masters AJ 1995 {\it J. Chem. Phys.} {\bf 103}

623:  1582

624: \bibitem{Frenkel3}

625:  Heemels MW, Hagen MHJ, and Lowe CP 2000 {\it J. Comput. Phys} 

626:  {\bf 164} 48

627: \bibitem{Frenkel4} 

628:  Hagen MHJ, Pagonabarraga I, Lowe CP, and Frenkel D 1997 {\it Phys.

629:  Rev. Lett.} {\bf 78} 3785

630: \bibitem{Horbach1}

631:   Horbach J and Frenkel D 2001 {\it Phys. Rev. E} {\bf 64} 061507-1

632: \bibitem{Horbach2} 

633:   Horbach J {Private communication}

634: \bibitem{Ahlrichs1} 

635:   Ahlrichs P and D\"unweg B 1998 {\it Int. J. Mod. Phys. C} {\bf 9} 1429

636: \bibitem{Ahlrichs2} 

637:   Ahlrichs P and D\"unweg B 1999 {\it J. Chem. Phys.} {\bf 111} 8225

638: \bibitem{Ahlrichs3}

639:   Ahlrichs P, Everaers R, and D\"unweg B 2001 {\it Phys. Rev. E} {\bf 64}

640:   040501(R)

641: \bibitem{Allen} 

642:   Allen M and Tildesley DJ 1987 {\it Computer simulation of liquids} 

643:   (Oxford: Oxford University Press)

644: \bibitem{Burkhard} 

645:   D\"unweg B and Kremer K 1993 {\it J. Chem. Phys.} {\bf 99} 6983

646: \bibitem{Alder} 

647:   Alder B and Wainwright T 1970 {\it Phys. Rev. A} {\bf 1} 18

648: \bibitem{Hansen2} 

649:    Hansen J-P and McDonald I 1986 

650:    {\it Theory of Simple Liquids} (London: Academic Press)

651: \bibitem{Hinch}

652:   Hinch E J 1975 {\it J. Fluid Mech.} {\bf 72} 499

653: \bibitem{Cichocki}

654:   Chichocki B and Felderhof B 1995 {\it Phys. Rev. E} {\bf 51} 5549

655: \bibitem{Debye} 

656:   Debye P 1929 {\it Polar Molecules} (New York: Dover)

657: \bibitem{Hauge}

658:   Hauge EH and Martin-L\"of A 1973 {\it J. Stat. Phys.} {\bf 7} 259

659: \bibitem{Berne1}

660:   Kushick J and Berne BJ 1973 {\it J. Chem. Phys.} {\bf 59} 4486

661: \end{thebibliography}

662: 

663: \end{document}

664: