1: \documentclass[prb,preprint,groupedaddress]{revtex4}

2: %\documentclass[aps,preprint,superscriptaddress]{revtex4}

3: %\documentclass[prb,twocolumn,groupedaddress]{revtex4}

4: % \documentclass[aps,twocolumn,superscriptaddress]{revtex4} 

5: \usepackage{amsmath} 

6: \usepackage{amssymb} 

7: \usepackage{amsfonts}

8: \usepackage[dvips]{graphicx} 

9: %\usepackage[]{caption}

10: \usepackage[]{epsfig} 

11: \bibliographystyle{apsrev} 

12: \voffset 1.5cm

13: \begin{document}

14: %\bibliographystyle{apsrev} 

15: %\bibliographystyle{prsty}

16: 

17: \title{Coupling nonpolar and polar solvation free energies in implicit solvent models}

18:  \author{J. Dzubiella} \email[e-mail address:]

19: {jdzubiella@ucsd.edu} \affiliation{NSF Center for Theoretical

20: Biological Physics (CTBP),} \affiliation{Department of Chemistry and

21: Biochemistry, University of California, San Diego, La Jolla,

22: California 92093-0365} \author{J.~M.~J. Swanson} \affiliation{NSF Center for

23: Theoretical Biological Physics (CTBP),} \affiliation{Department of

24: Chemistry and Biochemistry, University of California, San Diego, La

25: Jolla, California 92093-0365} \author{J.~A. McCammon} \affiliation{NSF

26: Center for Theoretical Biological Physics (CTBP),}

27: \affiliation{Department of Chemistry and Biochemistry, University of

28: California, San Diego, La Jolla, California 92093-0365}

29: 

30: \date{\today}

31: 

32: \begin{abstract}

33: Recent studies on the solvation of atomistic and nanoscale solutes

34: indicate that a strong coupling exists between the hydrophobic,

35: dispersion, and electrostatic contributions to the solvation free

36: energy, a facet not considered in current implicit solvent models. We

37: suggest a theoretical formalism which accounts for coupling by

38: minimizing the Gibbs free energy of the solvent with respect to a

39: solvent volume exclusion function. The resulting differential equation

40: is similar to the Laplace-Young equation for the geometrical

41: description of capillary interfaces, but is extended to microscopic

42: scales by explicitly considering curvature corrections as well as

43: dispersion and electrostatic contributions.  Unlike existing implicit

44: solvent approaches, the solvent accessible surface is an output of our

45: model. The presented formalism is illustrated on spherically or

46: cylindrically symmetrical systems of neutral or charged solutes on

47: different length scales.  The results are in agreement with

48: computer simulations and, most importantly, demonstrate that our

49: method captures the strong sensitivity of solvent expulsion and

50: dewetting to the particular form of the solvent-solute interactions.

51: \end{abstract}

52: %\pacs{61.20.Ja, 68.08.Bc, 87.16.Ac, 87.16.Uv}

53: \maketitle

54: \section{Introduction}

55: Implicit solvent models are widely used in theoretical chemistry to

56: study the solvation of biomolecular systems, as well described in the

57: review of Roux and Simonson\cite{roux:biochem}. They provide a more

58: efficient, although generally less accurate, alternative to

59: atomistically-resolved explicit solvent simulations.  The solvation

60: free energy in these models is usually split into nonpolar (np) and

61: polar (p) terms,

62: \begin{eqnarray}

63: \Delta G=\Delta G_{\rm np}+\Delta G_{\rm p},

64: \end{eqnarray}

65: which are treated in separate energetic evaluations. The nonpolar term

66: includes   the   energetic   cost   of   cavity   formation,   solvent

67: rearrangement,  and solute-solvent  dispersion interactions introduced

68: when  the uncharged  solute is  brought from  vacuum into  the solvent

69: environment. The polar term describes  the free energy of charging the

70: mono- or multipolar solute in the dielectric medium.

71: 

72: The nonpolar term is commonly approximated by surface area models,

73: i.e.~$\Delta G_{\rm np}\simeq\gamma S$, where $S$ is the

74: solvent accessible surface area\cite{richards} and $\gamma$ is an

75: energy per surface area constant, which is a priori not known but

76: fit to atomistic simulations.  The deficiencies of this simple surface

77: area approach have been recognized and a further decomposition of the

78: nonpolar term into cavity (cav) and van der Waals dispersion (vdW) terms has been

79: proposed,\cite{gallicchio:jpcb,gallicchio:jcc} $\Delta G_{\rm

80: np}=\Delta G_{\rm cav}+\Delta G_{\rm vdW}$. This approach has shown

81: improved results for the solvation of

82: alkanes,\cite{gallicchio:jpcb,zacharias} the alanine

83: peptide,\cite{su:biochem} and nonpolar native and misfolded

84: proteins.\cite{levy:jacs} The electrostatic (polar) contribution of

85: the solvation free energy is often approximated by generalized

86: Born\cite{bashford} (GB) or Poisson-Boltzmann\cite{sharp:honig} (PB)

87: models. Both methods use a position-dependent dielectric

88: constant,\cite{roux:biochem} assigned on the basis of

89: the solute surface, which  can be defined in several ways,\cite{baker}

90: or defined implicitly by integration methods.

91: It has been emphasized that all three contributions, $\Delta G_{\rm

92: cav}$, $\Delta G_{\rm vdW}$, and $\Delta G_{\rm p}$ depend critically

93: on the location of the solvent-solute interface.  It has {\it also}

94: been shown that the effective location of the solvent-solute interface

95: can vary according to the local electrostatic\cite{nina}

96: and dispersion\cite{huang:jpcb:2002} potentials. This suggests

97: that interfacial, dispersion and electrostatic contributions should be

98: coupled in implicit solvent approaches. The importance of capturing

99: the right balance between nonpolar and electrostatic contributions

100: in implicit solvation models was emphasized by Ashbaugh

101: {\it et al.} in their study of amphiphiles.\cite{ashbaugh:biophys}

102: 

103: The significance of nonpolar and polar coupling becomes even

104: more evident when solvation is studied on length scales which are

105: large compared to the solvent molecule (typically $\gtrsim$ 1 nm for

106: water), where solvent dewetting ('drying') can occur. In this

107: mechanism, first envisioned by Stillinger,\cite{stilinger} the solvent

108: molecules tend to move away from the surface of a large nonpolar

109: solute forming a liquid-gas like interface parallel to the solute

110: interface. When the surfaces of two

111: large solutes come together dewetting can be amplified due to the

112: gain of interfacial free energy (by decreasing the total

113: liquid-vapor interface area) giving rise to a strong effective

114: attraction.\cite{lum:jpc,hummer:jcp:1997,chandler:review} Early evidence of 

115: confinement-induced dewetting was given only by explicit water

116: simulations for smooth plate-like solutes with a purely repulsive

117: solute-solvent interaction.\cite{wallquist:jpc} More recently,

118: however, it has been demonstrated in varying degrees in several systems with attractive

119: solute-solvent interactions including smooth parallel plate-like

120: solutes,\cite{huang:pnas} atomistically resolved

121: paraffin-plates,\cite{huang:jpcb} graphite-plates\cite{pettitt}, carbon

122: nanotubes,\cite{hummer:nature} and hydrophobic ion channels.\cite{allen:prl,beckstein:jpc,sukharev}

123: 

124: Several of these studies indicated that the magnitude of dewetting is

125: sensitive to the nature of the solute-solvent attractive dispersion

126: interactions.  \cite{huang:pnas,huang:jpcb,pettitt} A similar

127: sensitivity was found in systems where the solutes carry charges or

128: are exposed to an external electric field, e.g. electrostatic interactions have

129: been shown to strongly affect the dewetting behavior of hydrophobic

130: channels\cite{vaitheesvaran, dzubiella:channel1,dzubiella:channel2}

131: and hydrophobic spherical nanosolutes.\cite{dzubiella:jcp:2003,

132: dzubiella:jcp:2004} Furthermore, two recent simulations of proteins

133: supported the importance of solvent dewetting and its sensitivity in

134: realistic biomolecular systems. First, a simulation of the two-domain

135: BphC enzyme showed that the region between the two domains was

136: completely dewetted when vdW and electrostatic interactions were

137: turned off, but accommodated 30$\%$ of the bulk density with the

138: addition of vdW attraction (water was found mainly at the edges of the

139: considered volume, while the central region was still empty), and

140: 85-90$\%$ with the addition of electrostatic

141: interactions.\cite{zhou:science} Second, Liu {\it et al.}  observed a

142: clear dewetting transition in the simulation of the collapse of the

143: melittin tetramer, which was strongly sensitive to the type and

144: location of the hydrophobic residues around the dewetted

145: region.\cite{berne:nature}

146:  

147: Considering the aforementioned studies, we postulate that coupling of

148: the nonpolar and polar solvation contributions in implicit solvent

149: models is crucial for an accurate determination of solvation free

150: energies without too many system-dependent fit parameters. We suggest

151: a general theoretical formalism in which the particular energetic

152: contributions are coupled.  Similar to the approach of Parker {\it et

153: al.}  in their study of bubble formation at hydrophobic

154: surfaces,\cite{attard} we express the Gibbs free energy as a {\it

155: functional} of the solvent volume exclusion function,\cite{beglov:jcp}

156: and obtain the optimal solute surface via minimization.  As

157: we will show, this minimization leads to an expression which is

158: similar to the Laplace-Young equation for the description of

159: macroscopic capillary interfaces,\cite{kralchevsky} but is generalized

160: to explicitly include curvature corrections and solvent-solute

161: interactions, i.e.  short-range repulsion (excluded volume),

162: dispersion, and electrostatics.  This extension of the Laplace-Young

163: theory allows a geometric description of solvation on mesoscopic and

164: microscopic scales. Related approaches in other fields are the

165: Helfrich description of vesicle and membrane

166: surfaces,\cite{helfrich,helfrich2} wetting in colloids and granular

167: media,\cite{kralchevsky,bieker} and functional treatments of

168: electrowetting.\cite{electrowetting}

169: 

170: While most implicit solvent approaches define the solute surface with

171: a geometrical evaluation of the molecular surface, vdW surface, or

172: canonical solvent accessible surface (SAS),\cite{richards,baker} it is

173: an output of our theory.  The surface obtained by minimizing our free

174: energy functional will, in general, be very different than the

175: aforementioned established surface definitions. In particular, our

176: solvent accessible surface should not be confused with the canonical

177: SAS,\cite{richards} which is simply the envelope surrounding

178: probe-inflated spheres. Similarly, phenomenological continuum theories

179: applied to solvent dewetting always assume a certain, simplified

180: geometry for the dry region, e.g. a cylindrical volume for system like

181: hydrophobic ion channels,

182: \cite{allen:jcp,beckstein:jpc,dzubiella:channel2} plate-like

183: particles,\cite{lum:jpc,huang:pnas} or two hydrophobic spherical

184: solutes.\cite{attard2} For a few simple systems this might be a valid

185: approximation but for more complicated solute geometries the shape of

186: the dewetted volume is unknown and a different approach, as suggested

187: in this work, is necessary. We expect our formalism to be particularly

188: useful in solvation studies of large protein assemblies where the

189: hydrophobic surfaces are highly irregular and laced with hydrophilic

190: units,\cite{gerstein,rossky:nature} and for which a unified

191: description of hydration on different length scales is

192: important.\cite{chandler:review} Another potential application is the

193: solvation of superhydrophobic nanosolutes\cite{super} and

194: wetting/dewetting in near-critical colloidal mixtures.\cite{bieker}

195: 

196: A brief summary of our work has been published elsewhere.\cite{prl}

197: Here we present more challenging test cases and an expanded discussion

198: of the approximations and limitations of this model.  The rest of the

199: paper is organized as follows: In section II we present our

200: theoretical formalism and chosen approximations. In section III we

201: first verify that our method can describe solvation on molecular

202: scales with noble gases, ions, and small alkanes. We then demonstrate

203: that it captures the strong sensitivity of dewetting and hydrophobic

204: hydration to specific solute-solvent interactions on larger scales

205: with two alkane-assembled spheres. In section IV we conclude with some

206: final remarks.

207: 

208: \section{Theory}

209: 

210: \subsection{Basic formalism}

211: 

212: Let us consider an assembly of solutes

213: with arbitrary shape and composition surrounded by a dielectric

214: solvent in a volume $\cal W$.  Furthermore, we define a subvolume (or

215: cavity) $\cal V$ empty of solvent for which we can assign a volume

216: exclusion function given by

217: \begin{eqnarray}

218:  v(\vec r)= \begin{cases} 0 & {\text {for}} \,\,\vec r \in \cal V; \cr 1 & {\rm else}. \cr

219:  \end{cases}

220: \end{eqnarray}

221: We assume that the surface surrounding the volume is continuous and

222: closed, i.e.~has no boundary.  The absolute volume $V$ and surface area $S$

223: of $\cal V$ can then be expressed as functionals of $v(\vec r)$ via

224: \begin{eqnarray}

225: V[v]=\int_{\cal W}{\rm d}^3r \;[1-v(\vec r)] \nonumber \\

226: S[v]=\int_{\cal W}{\rm d}^3r \;|\nabla v(\vec r)|,

227: \end{eqnarray}

228: where $\nabla\equiv\nabla_{\vec r}$ is the usual gradient operator

229: with respect to the position vector $\vec r$ and $|\nabla v(\vec r)|$

230: gives a $\delta$-function-like contribution only at the volume

231: boundary. The expression ${\rm d}^3r \;|\nabla v(\vec r)|\equiv{\rm

232: d}S$ can thus be identified as the infinitesimal surface element. In

233: this continuum solvent model the solvent density distribution

234: is simply $\rho(\vec r)=\rho_0 v(\vec r)$, where $\rho_0$ is the

235: bulk density of the solvent at the desired temperature and

236: pressure. Local inhomogeneities of the solvent density, apart from the

237: zero to $\rho_0$ transition at the volume boundaries, are

238: neglected. The solutes' positions and conformations are fixed such that

239: the solutes can be considered as an external potential to the solvent

240: without any degrees of freedom.

241: 

242: As motivated before, we suggest expressing the Gibbs free energy

243: $G[v]$ as a functional of the volume exclusion function

244: $v(\vec r)$, and obtaining the optimal solute volume via

245: minimization

246: \begin{eqnarray}

247: \delta G[v]/\delta v(\vec r)=0,

248: \label{min}

249: \end{eqnarray}

250: where $\delta../\delta v$ denotes the functional derivative with respect to the function $v$. We adopt following

251: ansatz for the Gibbs free energy of the solvent:

252: \begin{eqnarray}

253: \label{eq:grand}

254: &G[v]&=G_{\rm pr}[v]+ G_{\rm int}[v]+ G_{\rm ne}[v]+G_{\rm es}[v]\\

255: &=& P V[v]+\int_{\cal W}{\rm d}^3r\;\gamma(\vec r;[v])|\nabla v(\vec r)|\nonumber \\

256: &+&\rho_0\int_{\cal W}{\rm d}^3r\;v(\vec r) U(\vec r)\nonumber \\ 

257: &+&\!\!\!\!\!\!\int_{\cal W}\!\!\!{\rm d}^3r\!\left\{\frac{\epsilon_0}{2}\epsilon(\vec r;[v])[\nabla\Psi(\vec r)]^2\!-\lambda(\vec r)\Psi(\vec r)\!+v(\vec r)U_{\rm mi}(\vec r)\right\}\nonumber

258: \end{eqnarray}

259: Let us discuss each of the terms in Eq.~(\ref{eq:grand}) in turn.  The

260: first term, $G_{\rm pr}[v]$, proportional to the volume $V$, is the

261: energy of creating a cavity in the solvent against the difference in

262: bulk {\it pressure} between the liquid and vapor phase, $P=P_l-P_v$.

263: For water in ambient conditions, which is close to the liquid-vapor

264: transition, this term is relatively small and can generally be

265: neglected for solutes on molecular scales. The second term $G_{\rm

266: int}[v]$ describes the energetic cost due to solvent rearrangement

267: around the cavity {\it interface} with area $S$ in terms of a free

268: energy/surface area functional $\gamma(\vec r;[v])$. This interfacial

269: energy penalty is thought to be the main driving force behind

270: hydrophobic phenomena.\cite{chandler:review} $\gamma$ is a solvent

271: specific quantity that also depends on the particular topology of the

272: cavity-solvent interface, i.e.~it varies locally in space and is a

273: functional of the volume exclusion function $\gamma=\gamma(\vec

274: r;[v])$.\cite{zwanzig} The exact form of this functional is not known.

275: 

276: The third term, $G_{\rm ne}[v]$, is the total energy of the {\it

277: non-electrostatic} solute-solvent interaction given a solvent density

278: distribution $\rho_0v(\vec r)$. The potential $U(\vec r)=\sum_{i}

279: U_i(\vec r-\vec r_i)$ is the sum of the (short-ranged) repulsive

280: exclusion and (long-ranged) attractive dispersion interaction between

281: each solute atom $i$ at position $\vec r_i$ and a solvent molecule at

282: $\vec r$. Classical solvation studies typically represent the

283: interaction $U_i$ as an isotropic Lennard-Jones (LJ) potential,

284: \begin{eqnarray}

285:  U_{\rm

286: LJ}(r)=4\epsilon\left[\left(\frac{\sigma}{r}\right)^{12}-\left(\frac{\sigma}{r}\right)^6\right],

287: \label{lj}

288: \end{eqnarray} 

289: with an energy scale $\epsilon$, length scale $\sigma$, and

290: center-to-center distance $r$. Using the form of (\ref{lj}) implies

291: that $v({\vec r})$ is defined with respect to the LJ-centers of the

292: solvent molecules.

293: 

294: The last term, $G_{\rm es}[v]$, describes the total energy of the {\it

295: electrostatic} field and the mobile ions in the system expressed by

296: the local electrostatic potential $\Psi(\vec r)$ assuming linear

297: response of the dielectric solvent.\cite{jackson} Similar to

298: $\gamma(\vec r;[v])$, the position-dependent dielectric constant

299: $\epsilon(\vec r)=\epsilon(\vec r;[v])$ depends on the geometry of

300: $v(\vec r)$ with an unknown functional form. $\lambda(\vec r)$ is the

301: fixed charge density distribution of the solutes and the local energy

302: density of the {\it mobile ions} is \cite{sharp:honig,gilson}

303: \begin{eqnarray}

304: U_{\rm mi}(\vec r) = k_BT\sum_{j}\rho_j\{\exp[-\beta q_j \Psi(\vec

305: r)]-1\}.

306: \label{pb}

307: \end{eqnarray}

308: with the thermal energy $k_BT=\beta^{-1}$.  Variation of

309: (\ref{eq:grand}) for a fixed $v(\vec r)$ with respect to $\Psi(\vec

310: r)$ yields the Poisson-Boltzmann equation\cite{sharp:honig,gilson}

311: \begin{eqnarray}

312: {\rm PB}(\vec r)=0 &=& \nabla\cdot[\epsilon_0\epsilon(\vec r;[v])\nabla\Psi(\vec

313: r)]+\lambda(\vec r)\nonumber\\&+&v(\vec r)\sum_{j}q_j\rho_j\exp[-\beta q_j

314: \Psi(\vec r)],

315: \label{pb}

316: \end{eqnarray}

317: where $q_j$ and $\rho_j$ are the charge and concentration of the

318: mobile ion species $j$. Note that the ionic charge density in

319: (\ref{pb}) is multiplied by $v(\vec r)$ to account for the fact that

320: ions usually cannot penetrate the volume empty of polar solvent due to

321: a huge free energy penalty. We remark that the treatment of the

322: electrostatics in our theory has the same limitations as other

323: implicit models using PB, for instance when describing highly charged

324: or strongly correlated electrolyte systems. In contrast to PB/SA

325: models however, the dielectric boundary is optimized such that it

326: responds to the local nonpolar and polar potential; it is not assumed

327: beforehand.

328: 

329: 

330: Let $v_{\rm min}(\vec r)$ be the exclusion function which minimizes

331: the functional (\ref{eq:grand}).  Then, the resulting Gibbs free

332: energy of the system is given by $G[v_{\rm min}]$. The solvation free

333: energy $\Delta G$ is the reversible work to solvate the solute and is

334: given by

335: \begin{eqnarray}

336: \Delta G=G[v_{\rm min}]-G_0,

337: \end{eqnarray}

338: where $G_0$ is a constant reference energy which can refer to the pure

339: solvent state and an unsolvated solute.  The potential of mean force

340: (pmf) along a given reaction coordinate $x$ (e.g. the distance

341: between two solutes' centers of mass) is given, within a constant, by

342: $G[v_{\rm min}]$, where $v_{\rm min}(\vec r)$ must be evaluated for

343: every $x$.  In order to proceed we will need valid approximations for

344: $\gamma(\vec r;[v])$ and $\epsilon(\vec r;[v])$ with which $v_{\rm

345: min}(\vec r)$ can be calculated by explicitly minimizing our free

346: energy functional (\ref{eq:grand}) according to (\ref{min}).

347: 

348: \subsection{Approximations for $\gamma(\vec r;[v])$ and $\epsilon(\vec r;[v])$}

349: 

350: Let us start with a possible description of $\gamma(\vec r;[v])$.  For

351: a planar macroscopic interface the parameter $\gamma$ is usually

352: identified by the surface tension of the solvent adjacent to the

353: second medium. This surface tension obviously depends on the

354: microscopic interactions between the medium and the solvent and is

355: generally decreased by attractive dispersion or electrostatic

356: contributions.  It seems that microscopic interactions are adequately

357: represented by a macroscopic quantity like $\gamma$ {\it if} their

358: range is much smaller than the investigated length scales, such as the

359: radii of curvature or mean particle distances. The effect of the

360: microscopic interactions are then absorbed in $\gamma$. This has been

361: exemplified with free energy estimates for the solvation of large,

362: neutral plate-like or spherical alkane-assembled

363: solutes,\cite{huang:pnas,huang:jpcb:2002} For the description of

364: solvation on smaller length scales, however, it seems important to

365: separate the free energy into a part which accounts for the formation

366: of a cavity and a part which describes the dispersion interactions

367: explicitly.\cite{gallicchio:jpcb} Furthermore, it has been shown that

368: the water liquid-vapor surface tension $\gamma_{\rm lv}$ is the

369: asymptotic value of the solvation free energy per surface area for

370: hard spherical cavities in water in the limit of large

371: radii.\cite{lum:jpc,huang:jpc} These considerations motivate our

372: choice of the second and third term in the functional (\ref{eq:grand})

373: and lead to the assumption $\gamma\equiv\gamma_{\rm lv}$ in the limit

374: of vanishing curvatures.

375: 

376: The surfaces of realistic (bio)molecules, however, display highly

377: curved shapes, so $\gamma(\vec r;[v])$ will strongly depend on the interface

378: geometry around $\vec r$ in a complicated fashion. In the following we

379: make a {\it local curvature approximation}, i.e.~we assume that

380: $\gamma(\vec r;[v])$ can be expressed solely as a function of the local mean

381: curvature

382: \begin{eqnarray}

383: H(\vec r)=\left(\kappa_1(\vec r)+\kappa_2(\vec r)\right)/2=R(\vec

384: r)^{-1},

385: \end{eqnarray}

386: where $R(\vec r)$ is the radius of mean curvature and $\kappa_1(\vec r)$ and

387: $\kappa_2(\vec r)$ are the local principal curvatures of the

388: interface.\cite{prince} The mean curvature $H$ is only defined at the

389: boundary of $v(\vec r)$. We have chosen the convention in which the

390: curvatures are positive for convex surfaces (e.g. a spherical cavity)

391: and negative for concave surfaces (e.g. a spherical droplet).

392: 

393: The curvature dependence of the liquid-vapor surface tension

394: has been a long standing subject of research and is still under steady

395: discussion.\cite{hendersonreview,bieker,evans2} For water, which is

396: close to the critical point under ambient conditions, $\gamma$ is

397: argued to be nonanalytical in curvature.\cite{evans2}

398: The first order correction term, however, is likely to be linear in

399: curvature as predicted by scaled-particle theory,\cite{spt} the

400: commonly used ansatz to study the solvation of hard spherical

401: cavities. Although this result is only strictly valid for the case of

402: spherical particles, we assume that it can be applied to local mean

403: curvatures such that $\gamma(\vec r;[v])$ reduces to the function

404: \begin{eqnarray}

405: \gamma(\vec r)=\gamma_{{\rm lv}}(1-2\delta H(\vec r)),

406: \label{H}

407: \end{eqnarray}

408: where $\delta$ is the Tolman length, which is expected to be of

409: molecular size.\cite{tolman} In our study we assume $\delta$ is

410: constant and positive, while the curvature can be positive or negative

411: as defined above.  Note that this leads to an increase of surface

412: tension for concave surfaces in agreement with the geometrical

413: arguments of Honig {\it et al.}\cite{nicholls} in their solvation

414: study of alkanes. It has been shown by computer simulations of growing

415: a hard spherical cavity in water that (\ref{H}) predicts the

416: interfacial energy rather well for radii $\gtrsim

417: 3$\AA.\cite{huang:jpc} A major drawback of approximation (\ref{H}) is

418: that it gives unphysical results if the radius of mean curvature is

419: smaller than twice the Tolman length, $|R|<2\delta$. It yields

420: negative and diverging surface tensions for convex and concave

421: surfaces, respectively.  The latter is not possible due to the finite size

422: of the solvent molecules. Thus, care has to be taken with approximation (\ref{H}) when investigating systems which can exhibit radii of

423: curvature $|R|<2\delta$.

424: %% We note that for small convex

425: %% cavities it has been shown that Gaussian fluctuations of the solvent

426: %% dominate the solvation free energy and the latter can be expressed

427: %% analytically in terms of the cavity

428: %% volume.\cite{hummer:pnas,garde:prl} An improved approximation for

429: %% $\gamma(v)$ (or in general for $G[v]$) incorporating this behavior may

430: %% be possible and needs further investigation.

431: %% Another option or additional

432: %% refinement may be the inclusion of a second order correction

433: %% proportional to the curvature squared as suggested by Helfrich, such

434: %% that

435: %% \begin{eqnarray}

436: %% \gamma(H)=\gamma_{{\rm lv}}(1+2\delta H+aH^2+bK),

437: %% \label{helf}

438: %% \end{eqnarray}

439: %% where $a$ and $b$ are a priori unknown rigidity constants, and

440: %% $K=\kappa_1\kappa_2$ is the Gaussian

441: %% curvature.\cite{helfrich,helfrich2} Using (\ref{helf}) in the

442: %% functional (\ref{eq:grand}) without excluded volume, dispersion and

443: %% electrostatic terms leads to the Helfrich

444: %% functional,\cite{helfrich,helfrich2} almost exclusively used for the

445: %% shape description of membranes and vesicles.

446: 

447: Let us now turn to electrostatics. The most common approximation for the

448: position-dependent dielectric constant is proportional to the volume

449: exclusion function $v(\vec r)$,\cite{roux:biochem} such that the functional

450: $\epsilon(\vec r;[v])$ reduces to,

451: \begin{eqnarray}

452: \epsilon(\vec r)=\epsilon_v+v(\vec r)(\epsilon_l-\epsilon_v),

453: \label{e}

454: \end{eqnarray}

455: where $\epsilon_v$ and $\epsilon_l$ are the dielectric constants

456: inside and outside the volume $\cal V$, respectively.  Eq.~(\ref{e})

457: is valid only in the limit of large solute sizes when the molecular size

458: of the solvent is negligible. For charged solutes on a molecular

459: scale, let's say mono- or polyvalent ions, two difficulties

460: arise. First the electric field close to the highly curved solutes can

461: be strong enough for the dielectric constant to be field

462: dependent. This formally affects the form of the electrostatic term in

463: the free energy functional which assumes a linear response of the

464: solvent. An improvement for continuum models along these lines has

465: been proposed by Luo et al.\cite{luo:jpc} Second, the effective

466: position of the dielectric boundary is known to depend on the sign of

467: the solute charge for asymmetric solvent molecules like water. This

468: expresses itself, for instance, in different Born radii for two

469: equally charged ions which have exact the same LJ parameters but a

470: different sign of charge. A reasonable improvement of (\ref{e}) would

471: be to shift the dielectric boundary at $\vec r$ {\it parallel} to the

472: volume boundary by a potential dependent amount $\Delta \vec r=\xi(\Psi(\vec r))\vec n(\vec r)$:

473: \begin{eqnarray}

474: \epsilon(\vec r)=\epsilon_v+v\left(\vec r-\Delta \vec r\right)(\epsilon_l-\epsilon_v),

475: \label{e2}

476: \end{eqnarray}

477: where $\vec n$ is the unit normal vector to the interface. We do not

478: attempt however, to find an approximation for the function $\xi(\Psi)$

479: in this work and postpone this investigation to later studies. For

480: further illustration of our approach we content ourselves with the

481: approximations (\ref{H}) for $\gamma(\vec r;[v])$ and (\ref{e}) for

482: $\epsilon(\vec r;[v])$.

483: 

484: \subsection{Minimization of the free energy functional}

485: For the functional derivative of the interfacial term,

486: $G_{\rm int}[v]$, we utilize

487: \begin{eqnarray}

488: \frac{\delta}{\delta v}\int_{\cal W}{\rm d}^3r \;|\nabla v(\vec r)|=\frac{\delta}{\delta v}\int_{\partial\cal W}{\rm d}S=-2H

489: \end{eqnarray}

490: and

491: \begin{eqnarray}

492: \frac{\delta}{\delta v}\int_{\cal W}{\rm d}^3r\;H \;|\nabla v(\vec r)|=\frac{\delta}{\delta v}\int_{\partial\cal W}{\rm d}S\;H=-K,

493: \end{eqnarray}

494: which has been derived in detail by Zhong-can and Helfrich by means of

495: differential geometry.\cite{helfrich} The variable

496: $K(\vec r)=\kappa_1(\vec r)\kappa_2(\vec r)$ is the Gaussian curvature of the interface, which

497: is an intrinsic geometric property of $v$. Plugging in approximations

498: (\ref{H}) and (\ref{e}) into (\ref{eq:grand}), and minimizing with

499: (\ref{min}), we obtain,

500: \begin{eqnarray}

501: 0&=&{\rm de}(\vec r)= P+2\gamma_{{\rm lv}}\left[H(\vec r)-\delta K(\vec

502: r)\right]-\rho_0 U(\vec r)\nonumber\\ &-&\frac{\epsilon_0}{2}[\nabla\Psi(\vec

503: r)\epsilon(\vec

504: r)]^2\left(\frac{1}{\epsilon_l}-\frac{1}{\epsilon_v}\right)-U_{\rm

505: mi}(\vec r).\nonumber\\

506: \label{diff}

507: \end{eqnarray}

508: Eq.~(\ref{diff}) is a partial second order differential equation (de)

509: for the optimal exclusion function $v_{\rm min}(\vec r)$ expressed in

510: terms of pressure, curvatures, short-range repulsion, dispersion, and

511: electrostatic terms, all of which have dimensions of energy density.

512: It can also be interpreted as a mechanical balance between the forces

513: per surface area generated by each of the particular contributions.

514: Thus, in our approach the surface shape and geometry, expressed by $H$

515: and $K$, are directly related to the inhomogeneous potential

516: contributions.  The constant solute charge density

517: $\lambda(\vec r)$ does not appear explicitly in (\ref{diff}) but is

518: implicitly considered in the PB equation (\ref{pb}), which must be

519: solved simultaneously. If curvature correction ($K$-term) and the last

520: three energetic terms are neglected one obtains the Laplace-Young

521: equation,

522: \begin{eqnarray}

523: P=-2\gamma_{lv}H,

524: \label{ly}

525: \end{eqnarray}

526: which is exclusively used for the shape description of macroscopic

527: capillary and interfacial phenomena in conjunction with appropriate

528: boundary conditions, e.g. prescribed liquid-solid contact angles at

529: the solid surfaces.\cite{kralchevsky} In our description the boundary

530: conditions are provided by the constraints given by the short-ranged

531: repulsive term in $U(\vec r)$, and the distribution of dispersion and

532: electrostatics, allowing an extrapolation of the Laplace-Young

533: description to mesoscopic and microscopic scales. Notice that in our

534: approach the solvent is treated as a continuum while the solute is

535: explicitly resolved.  One could use a coarse-grained treatment for the

536: solute by including the appropriate non-electrostatic and

537: electrostatic interactions in (\ref{eq:grand}).

538: 

539: The solution of (\ref{diff}) requires an appropriate

540: parametrization, i.e.~coordinate representation, for

541: the curvatures $H$ and $K$, such that the equation is expressed as a

542: function of the vector $\vec r$ and its first and second derivatives

543: in space. Analytical solutions to the much simpler Eq.~(\ref{ly}) and thus

544: to (\ref{diff}) are only available for systems with very simple geometries.\cite{kralchevsky}

545: %% Examples follow in the next section.  It seems that analytical

546: %% solutions to (\ref{diff}) are only available for very simple systems,

547: %% e.g., one or two hard and neutral spheres,\cite{kralchevsky} but

548: %% further investigation is needed. The stability of (\ref{eq:grand})

549: %% should also be checked by verifying that the second variation is

550: %% positive definite,\cite{helfrich2} $\delta^{(2)}G/\delta v^2>0$, but

551: %% seems analytically not easily tractable too.

552: Thus we use numerical solutions of (\ref{diff}) in the following to

553: further illustrate our theory.

554: 

555: \section{Applications}

556: 

557: First, we will consider the solvation of microscopic solutes such as

558: noble gases, simple alkanes, and ions which can be treated as neutral

559: or charged Lennard-Jones spheres. Then, we will investigate alkane

560: assemblies on a larger scale where interfacial and dewetting effects

561: are much more dominant.  For simplicity and a better transparency of

562: the results, mobile ions will be neglected in these illustrations.

563: 

564: \subsection{One Lennard-Jones sphere}

565: 

566: In this section we compare our approach to results from SPC and SPC/E

567: explicit solvent simulations.\cite{berendsen:jpc} We refrain from 

568: comparing to real experiments since approximations in computer

569: experiments are more easily controlled and the LJ parameters

570: of the solutes are commonly parametrized to yield accurate results in

571: classical simulations.

572: 

573: For a spherical solute with a charge $Q$ homogeneously distributed

574: over its surface, the functional (\ref{eq:grand}) with approximations

575: (\ref{H}) and (\ref{e}) and no mobile ions reduces to a function of

576: $R$, the radius of the sphere empty of solvent. The solvation free

577: energy is

578: \begin{eqnarray}

579: \Delta G(R)&=&\Delta G_{\rm pr}(R)+\Delta G_{\rm int}(R)+\Delta G_{\rm ne}(R)+\Delta G_{\rm es}(R)\nonumber\\

580: &=&\frac{4}{3}\pi R^3 P + 4\pi R^2 \gamma_{{\rm

581:  lv}}\left(1-\frac{2\delta}{R}\right)\nonumber\\&+& \int_R^{\infty} 4\pi r^2{\rm

582:  d}r\; \rho_0U_{\rm LJ}(r)\nonumber\\ &+&\frac{Q^2}{8\pi \epsilon_0

583:  R}\left(\frac{1}{\epsilon_l}-\frac{1}{\epsilon_v}\right)

584: \label{eq:sphere}. 

585: \end{eqnarray}

586: Note that the last term in (\ref{eq:sphere}) is equivalent to the Born

587: electrostatic solvation free energy.\cite{roux:biochem} Recently,

588: Manjari {\it et al.} have presented a very similar expression for the

589: solvation of a charged spherical cavity on the basis of a

590: minimization principle and have investigated the variation of $R$ with

591: thermodynamic conditions.\cite{kim} Differentiation of 

592: (\ref{eq:sphere}) with respect to $R$ and subsequent division by $4\pi

593: R^2$ yields

594:  \begin{eqnarray}

595:   0&=& P+\frac{2\gamma_{{\rm

596:   lv}}}{R}\left(1-\frac{\delta}{R}\right)-\rho_0U_{\rm

597:   LJ}(R)\nonumber\\&-&\frac{Q^2}{32\pi^2\epsilon_0

598:   R^4}\left(\frac{1}{\epsilon_l}-\frac{1}{\epsilon_v}\right)

599: \label{sphere2},

600:  \end{eqnarray}

601: which is in accord with Eq.~(\ref{diff}) given sphere-like curvatures,

602: $H=1/R$ and $K=1/R^2$.   We

603: can now calculate the solvation free energies of simple spherical

604: solutes, such as noble gases or ions.  The free parameters in

605: Eq.~(\ref{sphere2}) are the pressure $P$, Tolman length $\delta$,

606: liquid-vapor surface tension $\gamma_{\rm lv}$, and dielectric

607: constants $\epsilon_v$ and $\epsilon_l$.

608:  

609: \subsubsection{One neutral LJ sphere}

610: First, let us focus on uncharged spheres, for which the electrostatic

611: term in (\ref{eq:sphere}) can be neglected.  We compare the results

612: from our theory to those calculated by Hummer {\it et

613: al.}\cite{hummer:jpc:1996} for neutral LJ spheres in SPC water, and

614: those calculated by Paschek\cite{paschek} for noble gases in SPC and

615: SPC/E water.  The solute-water LJ parameters $\sigma$ and $\epsilon$

616: are summarized in Tab.~I.  The surface tension $\gamma_{\rm lv}$ was

617: set to that estimated for SPC and SPC/E water at 300K, $\gamma_{\rm

618: lv}=65$mJ/m$^2$ and $\gamma_{\rm lv}=72$mJ/m$^2$,

619: respectively.\cite{alejandre,huang:jpc} The pressure is fixed to 1atm.

620: Finally, the remaining free parameter $\delta$ was fit to {\it reproduce}

621: the simulation solvation free energies exactly.  The solvation

622: free energies from simulation $\Delta G_{\rm sim}$ and best fit Tolman

623: lengths $\delta_{\rm bf}$ are shown in Tables~I~and~II for the SPC and

624: SPC/E models, respectively.

625: 

626: Before we discuss the results, let us compare the particular energy

627: contributions $\Delta G_{\rm i}(R)$ with ${\rm i=pr,int,ne}$ for

628: Na$^0$ (plotted in Fig.~\ref{fig:G}). As anticipated, the pressure

629: term $\Delta G_{\rm pr}(R)$ with $P=1$atm is negligible compared to

630: the other contributions. The interfacial term $\Delta G_{\rm int}(R)$

631: increases with the cavity radius $R$. The integrated LJ-interaction

632: term $\Delta G_{\rm ne}(R)$ shows long-range attraction and a steep

633: short-ranged repulsion with a minimum at

634: $R=\sigma(Na^{0})=2.85$\AA. The total solvation free energy for the

635: Na$^0$ shows a single minimum at $R_{\rm min}=2.32$\AA~ with $\Delta

636: G=9.2$kJ/mol for a $\delta_{\rm bf}$=0.79\AA.

637: 

638: 

639: The results for the other LJ-spheres, summarized in Tab.~I and II,

640: reveal several noteworthy observations.  First, the best fit Tolman

641: lengths $\delta_{\rm bf}$ range from 0.76\AA~ to 1.00\AA~; they are

642: not only of molecular size, as expected, but are approximately half

643: the LJ-radius of a SPC or SPC/E water molecule.  Second, the

644: $\delta_{\rm bf}$ values for noble gases in SPC/E water (Tab.~II) are

645: approximately 10$\%$ larger than those in SPC water (Tab.~I).  This is

646: in qualitative agreement with the results of Huang {\it et al.} who

647: measured $\delta$=$0.76\pm0.05$\AA~ and $\delta$=$0.90\pm0.03$\AA~ for

648: SPC and SPC/E, respectively, by fitting Eq.~(\ref{H}) to the hydration

649: free energy of hard spheres with varying radii.\cite{huang:jpc}

650: 

651: Third, the quite accurate data of Paschek demonstrate a systematic increase of 

652: $\delta_{\rm bf}$ with solute size.  The inability of our theory 

653: to be fit by one fixed constant $\delta_{\rm bf}$ points to the 

654: anticipated fact that Eq.~(\ref{H}) can not capture strong curvature 

655: effects accurately and will have to be refined for small solutes. Despite this shortcoming,

656: these results show surprisingly good agreement; if we assume a fixed

657: delta, for instance $\delta=0.91$\AA~ for all noble gases in the SPC

658: data of Paschek, our theory predicts results within

659: $15\%$ of the simulation data. Finally, we

660: observe that the effective optimal sphere radius $R_{\rm min}$ is

661: always smaller than the radius of the canonical SAS with a

662: typical probe radius of 1.4\AA,\cite{connolly} $R_{\rm

663: min}<(\sigma_{ss}/2+1.4{\rm \AA})\simeq\sigma$, but larger than the

664: vdW surface, $R_{\rm min}>\sigma_{ss}/2$, where

665: $\sigma_{ss}$ is the solute-solute LJ-length.\cite{sigmanote}

666: 

667: 

668: \subsubsection{One charged LJ sphere}

669: Let us now turn to charged Lennard-Jones spheres (ions) also examined

670: in the paper by Hummer {\it et al.}  with SPC water simulations. We

671: assume $\delta$ to be the {\it fixed} by the previously obtained

672: $\delta_{\rm bf}$ values for uncharged spheres.  The dielectric

673: constants are set to $\epsilon_v=1$ and $\epsilon_l=65$, in accord

674: with SPC water.\cite{spoel} The electrostatic contribution $\Delta

675: G_{\rm es}(R)$ and the total $\Delta G(R)$ for Na$^+$ are shown in the

676: inset of Fig.~\ref{fig:G}. The electrostatic contribution decreases

677: the optimal radius to $R_{\rm min}=1.83$\AA~ giving a solvation free

678: energy of $\Delta G=-334$kJ/mol. In fact, the optimal sphere radius

679: $R_{\rm min}$ is always considerably smaller for the charged solutes

680: (Tab.~III) than for their neutral counterparts (Tab.~I). This is

681: caused by the strong compressing force of the polar solvent attempting

682: to penetrate the low dielectric cavity.  The solvation free energies

683: from theory $\Delta G$ and those from simulation $\Delta G_{\rm sim}$

684: are also shown in Tab.~III.  While our theory describes the hydration

685: free energies for positively charged ions within $15\%$, it

686: considerably underestimates those of the negative ions. This

687: qualitative disagreement between positive and negative ions was

688: expected since the Born radii for anions are always smaller than those

689: for cations, a consequence of the different solvation structure around

690: charged solutes with opposite signs. As mentioned in the previous

691: section, the position of the dielectric boundary has to be refined for

692: accurate estimates of the electrostatic contribution to the hydration

693: free energy. If we apply the correction (\ref{e2}) to the dielectric

694: boundary with a simple, potential-independent shift $\xi_+=-0.25$\AA~

695: for positive and $\xi_-=-1.05$\AA~ for negative spheres such that the

696: dielectric boundary has a radius of $R+\xi_\pm<R$, improved values

697: ($\Delta G_\xi$ in Tab.~III) are obtained which reproduce all

698: simulation values within $10\%$!

699:  

700: 

701: 

702: \subsection{Linear alkanes}

703: 

704: Let us now consider simple polyatomic molecules, such as ethane,

705: propane, or butane in a cylindrically symmetric one-dimensional (1D)

706: chain conformation.  Other conformations will be neglected. The

707: symmetry of these systems allows us to express the volume exclusion

708: function $v(\vec r)$ of the enveloping surface by a one dimensional

709: shape function $r(z)$, where $z$ is the coordinate on the symmetry

710: axis and $r$ the radial distance to it. The full surface in

711: three-dimensional space is obtained by revolving the shape function

712: $r(z)$ around the symmetry axis. Technical details are given in the

713: Appendix.

714: 

715: The LJ parameters for ethane and methane are the same as those used by

716: Ashbaugh {\it et al.}\cite{ashbaugh:biophys} in their SPC simulation

717: of linear alkanes (see Tab.~I).  The simulation solvation energy of

718: the spherical methane, $\Delta G=10.96$kJ/mol, can be reproduced with

719: $\delta_{\rm bf}=0.85$\AA. Solving the cylindrically symmetric problem

720: for ethane using the same $\delta$, we obtain a fit-parameter free

721: $\Delta G=11.40$kJ/mol, which is only 7$\%$ larger than the simulation

722: results. Alternatively, $\delta_{\rm bf}=0.87$\AA~ reproduces the

723: simulation energy exactly.  This is surprisingly good agreement

724: considering the crudeness of our curvature correction and the fact

725: that the large curvature of the system varies locally in space. The

726: curvature and shape functions are plotted in Fig.~\ref{fig:ethane}

727: together with the vdW surface and the canonical SAS obtained from

728: rolling a probe sphere with the typically chosen radius $r_{\rm

729: p}=1.4$\AA~ over the vdW surface.\cite{connolly} Away from the

730: center of mass $|z|\gtrsim 1$\AA~ the curvatures follow the expected

731: trends for the spherical surfaces: $H= 1/R$ and

732: $K=1/R^2$ with $R \simeq 3.1$\AA~.  The

733: optimal surface resulting from our theory is smaller than the

734: canonical SAS and smooth at the center of mass ($z=0$) where the

735: canonical SAS has a kink. Thus our surface has a smaller mean

736: curvature at $z=0$ and an almost zero Gaussian curvature, which is

737: typical for a cylinder geometry with one principal curvature equal to

738: zero.  These results may justify the use of smooth surfaces in

739: coarse-grained models of closely-packed hydrocarbons, a possibility we

740: will explore in the following section with solvation on larger length

741: scales where dewetting effects can occur.  If we repeat the above

742: calculation for propane and butane (three and four LJ-spheres, see

743: also Tab.~I for parameters) we need $\delta_{\rm bf}=0.94$\AA~ and

744: $\delta_{\rm bf}=0.96$\AA, respectively, to reproduce the simulation

745: results exactly. The increasing difference in $\delta_{\rm bf}$

746: compared to methane and ethane is likely due to contributions from

747: other than cylindrically symmetric conformations which were ignored in

748: our analysis.

749:  

750: 

751: \subsection{Two spherical nanosolutes}

752: 

753: \subsubsection{Model}

754: 

755: Let us now consider two spherical solutes which represent

756: homogeneously assembled CH$_2$ groups with a uniform density

757: $\rho$=0.024\AA$^{-3}$ up to a radius $R_0=15$\AA, defined by the

758: maximal distance between a CH$_2$ center and the center of the solute.

759: Integration of the CH$_2$-water LJ interaction over the entire sphere

760: yields a 9-3 like potential $U_i(r)$ for the interaction between the

761: center of the solute ($i=1,2$) and a water

762: molecule. \cite{huang:jpcb:2002} The intrinsic, nonelectrostatic

763: solute-solute interaction $U_{\rm ss}$ can be obtained in a similar

764: fashion.  The CH$_2$-water LJ parameters, $\epsilon=0.5665$kJ/mol and

765: $\sigma=3.52$\AA, are taken from the OPLSUA

766: force-field\cite{jorgensen} and are similar to those used by Huang

767: {\it et al.} in their study on dewetting between paraffin

768: plates.\cite{huang:jpcb} Minimizing Eq.~(\ref{eq:sphere}) for just one

769: sphere we obtain an optimal solvent excluded radius of $R_{\rm min}

770: \simeq 17.4$\AA, which is $R_0+2.4$\AA. Since we are also

771: interested in the effects of electrostatic interactions we place

772: opposite charges $\pm Ze$, where $e$ is the elementary charge, in the

773: center or on the edge of the two spheres. Poisson's equation is

774: simultaneously solved on a two-dimensional grid in cylindrical

775: coordinates. Numerical details are given in the Appendix.

776: 

777: The solvation of the two solutes is studied for a fixed

778: surface-to-surface distance which we define as $s_0=r_{12}-2R_0$,

779: where $r_{12}$ is the solute center-to-center distance. The effective

780: surface-to-surface distance defined by the accessibility of the

781: solvent centers is thus $s\simeq r_{\rm 12}-2R_{\rm

782: min}=s_0-4.8$\AA. In the following we focus on a separation distance

783: of $s_0=8$\AA~ to investigate the influence of different energetic

784: contributions on the shape function, $r(z)$, and the curvatures,

785: $K(z)$ and $H(z)$. For $s_0 = 8$\AA, it follows that $s\simeq 3.2

786: $\AA, such that two water molecules could fit between the solutes on

787: the $z$-axis. We systematically change the solute-solute and

788: solute-solvent interactions, as summarized in Tab.~I. We begin with

789: only the repulsive part of the nonelectrostatic interaction $U_i(r)$

790: in system I, and then add a curvature correction with

791: $\delta=0.75$\AA, vdW attractions, and sphere-centered charges $Z = 4$

792: and $Z = 5$ in systems II-V, respectively.  To study the influence of

793: charge location, we reduce the magnitude of each charge in system VI

794: to $Z=1$ and move them to the edge of the spheres on the symmetry axis

795: such that they are $8$\AA~apart (indicated by arrows in

796: Fig.~\ref{fig:nano}.  The surface tension and dielectric constants of

797: the vapor (solute) and liquid are fixed to $\gamma_{\rm

798: lv}=72$mJ/m$^2$, $\epsilon_v=1$, and $\epsilon_l=78$, respectively.

799: 

800: \subsubsection{Behavior of the shape function}

801: 

802: 

803: 

804: 

805: The results for the curvatures and shape function, defined by $r(z)$,

806: for systems I-VI are shown in Fig.~\ref{fig:nano}. Away from the

807: center of mass ($|z|\gtrsim 10$\AA), systems I-VI show very little

808: difference. The curvatures are $H=1/R$ and $K=1/R^2$ with $R \simeq

809: 17.4$\AA. Close to the center of mass ($z\simeq 0$), however, the

810: influence of changing the parameters is considerable. In system I,

811: Eq.~(\ref{diff}) reduces to the minimum surface equation $H(z)=0$ for

812: $z\simeq 0$. For two adjacent spheres the solution of this equation is

813: the catenoid $r(z)\simeq{\rm cosh}(z)$, which features zero mean

814: curvature ($\kappa_1$ and $\kappa_2$ cancel each other) and negative

815: Gaussian curvature. As a consequence, the system exhibits a vapor

816: bubble bridging the solutes, i.e. water is removed from the region

817: between the spheres even though it fits there. This dewetting is

818: driven by the interfacial term $G_{\rm int}$ which always favors

819: minimizing the liquid-vapor interface.

820: 

821: When curvature correction is applied (system II), the mean curvature

822: becomes nonzero and negative (concave) at $z\simeq 0$, while the

823: Gaussian curvature grows slightly more negative. Thus the total

824: enveloping surface area becomes larger and the solvent inaccessible

825: volume shrinks, i.e. the value of the shape function at $z \simeq 0$

826: decreases. Turning on solute-solvent dispersion attraction amplifies

827: this trend significantly as demonstrated by system III. Mean and

828: Gaussian curvatures increase fivefold, showing strongly enhanced

829: concavity, and the volume empty of water decreases considerably,

830: expressed by $r(z=0)\simeq 10.7$\AA~ dropping to $r(z=0)\simeq

831: 6.3$\AA. These trends continue with the addition of electrostatics in

832: system IV.  When the sphere charges are further increased from $Z=4$

833: to $Z=5$ (system IV$\rightarrow$V), we observe a wetting transition:

834: the bubble ruptures and the shape function jumps to the solution for

835: two isolated solutes, where $r(z\simeq 0)=0$. The same holds when

836: going from III to VI, when only one charge unit, $Z=1$, is placed at

837: each of the solutes' surfaces. Importantly, this demonstrates that the

838: present formalism captures the sensitivity of dewetting phenomena to

839: specific solvent-solute interactions as reported in previous

840: studies.\cite{huang:jpcb,pettitt,dzubiella:channel1,vaitheesvaran,dzubiella:jcp:2003,zhou:science,berne:nature}

841: Note that the optimal shape function at $|z|\simeq\pm2$\AA~ is closer

842: to the solutes in VI compared to V due to the proximity of the charge

843: to the interface. Clearly, the observed effects, particularly the

844: transition from III to VI, cannot be described by existing solvation

845: models which use the surface area (GB/SA or PB/SA)\cite{roux:biochem}

846: or effective surface tensions and macroscopic solvent-solute contact

847: angles\cite{kralchevsky,attard} as input.

848: 

849: \subsubsection{Potential of mean force}

850: The significant change of the shape function with the

851: solute-solvent interaction has a strong impact on the potential of

852: mean force (pmf) (or effective interaction) between the solutes

853: \begin{eqnarray}

854: W(s_0)= G(s_0)- G(\infty)+U_{\rm ss}(s_0).

855: \end{eqnarray} 

856: Recall that $U_{\rm ss}$ is the instrinsic dispersion interaction

857: between the two solutes. Values of $W(s_0=8{\rm\AA})$ are given in

858: Tab.~IV. From system I to VI the total attraction between the solutes

859: decreases almost two orders of magnitude. Interestingly, the curvature

860: correction (I$\rightarrow$II) lowers $W$ by a large 23.5$k_BT$, even

861: though $R_0\gg\delta$. The reason is that the mean radii of curvature

862: between the spheres can assume values $\simeq \delta$, implying that

863: curvature correction is also important for large solutes. A striking

864: effect occurs when vdW contributions are introduced

865: (II$\rightarrow$III): the inter solute attraction decreases by

866: approximately $28k_BT$ while the dispersion solute-solute potential

867: $U_{\rm ss}(s_0=8{\rm \AA})$ changes by only $-0.44 k_BT$. Similarly,

868: adding charges of $Z=5$ (III $\rightarrow$ V) at the solutes' centers

869: or $Z=1$ (III $\rightarrow$ VI) at the solutes' surfaces decreases the

870: total attraction by 1.2$k_BT$ and 5k$_B$T, respectively. Note that the

871: total attraction {\it decreases} even though electrostatic attraction

872: has been added between the solutes. The same trend has been observed

873: recently in explicit water simulations of a similar system of charged

874: hydrophobic nanosolutes.\cite{dzubiella:jcp:2003,dzubiella:jcp:2004}

875: 

876: Now we turn our attention to varying the intersolute distance.  The

877: pmfs and solute-solute mean forces $F=\partial W(s_0)/\partial s_0$

878: versus a range of $s_0$ are shown for systems I,II,III, and VI in 

879: Fig.~\ref{fig:pmf}. System I, with purely repulsive

880: solute-solvent interactions, displays a strong attraction

881: ($W\simeq-150k_BT$) at $s_0=2$\AA~ which decreases, almost linearly,

882: to zero at a distance $s_0=13.5$\AA~ where the system shows a wetting

883: transition. The corresponding force is discontinuous at this critical

884: distance.  The steep repulsion at short intersolute distances

885: ($s_0\simeq 1.5$\AA~) stems from the repulsive term of the LJ

886: interaction between the solutes. Note that the intrinsic solute-solute

887: interaction $U_{\rm ss}(s_0)$, also shown in Fig.~\ref{fig:pmf}, is almost two

888: orders of magnitude smaller than the hydrophobic attraction. Adding

889: the curvature correction in system II decreases the range and strength

890: of the pmf by approximately $20\%$, which is significant and

891: unexpected since $R_0\gg\delta$.  Adding dispersion

892: attractions in system III decreases the range and strength of the

893: hydrophobic attraction considerably, but it is still much stronger

894: than the inter solute dispersion attraction $U_{ss}$ alone. When

895: surface charges ($Z=1$) are added in system VI, the range of

896: hydrophobic attraction further decreases but the total attraction

897: increases at short intersolute distances.  This is due to the

898: increasing size of the bridging bubble ($r(z=0)$ increases) as the two

899: solutes approach each other, which decreases the high dielectric

900: screening of the solute-solute electrostatic attraction.  This again

901: underlines the importance of coupling electrostatics and dewetting

902: effects, as the electrostatic attraction (or repulsion) may be

903: magnified by more than an order of magnitude when dewetting

904: occurs. For charges with opposite sign this could be interpreted as the

905: stabilization of a salt bridge due to dehydration.\cite{fernandez}

906: Systems IV and V, not shown in Fig.~\ref{fig:pmf}, exhibit the same

907: qualitative behavior as system VI.

908: 

909: 

910: \subsubsection{Comparison of mean forces to previous MD simulations}

911: We continue considering the mean force between two nanosized solutes

912: and compare our theory now to the MD simulations of Dzubiella {\it et

913: al.}\cite{dzubiella:jcp:2003,dzubiella:jcp:2004} Their solute model

914: slightly differs from the one used in the previous section: the

915: solute-solvent interaction potential is purely repulsive and is given

916: by $U_i(r)=k_BT(r-R_0)^{-12}$, while the solute-solute interaction is

917: hard-sphere like with a hard sphere radius $R_0$. The solutes are

918: neutral or carry opposite charges $Q$ homogeneously distributed over

919: the sphere volume. The simulations were carried out with the SPC/E

920: model of water. In our theory, we fix the Tolman length to

921: $\delta=0.90$\AA~ as measured by Huang {\it et al.}\cite{huang:jpc}

922: and the dielectric constant to $\epsilon_l=71$ for SPC/E

923: water.\cite{berendsen:jpc} The mean forces are shown in

924: Fig.~\ref{fig:mf} for neutral spheres of radii $R_0=10$ and 12\AA~ and

925: oppositely charged solutes with radius $R_0=10$\AA~ and charge $Q=2$e

926: and 5e versus the solutes' surface-to-surface distance $s_0$.

927: Simulation and theory are in good, almost quantitative agreement, and

928: show that our theory captures the decreasing range of the strongly

929: hydrophobic attraction with decreasing radius and increasing

930: charge due to suppressed dewetting. We emphasize that our theory is

931: basically fit-parameter free for this system of large

932: solutes. Fig.~\ref{fig:mf} also shows the theoretical mean force for

933: the neutral $R_0=12$\AA~ solute using a smaller Tolman length

934: $\delta=0.75$\AA. The decrease of the Tolman length increases the

935: depth and range of the solvent-mediated solute-solute attractive mean

936: force by approximately 5\%, showing a nonvanishing but only slight

937: influence.

938: 

939: \section{Conclusion and final remarks}

940: 

941: In summary, we have presented a novel implicit solvent model which

942: couples polar and nonpolar solvation contributions by employing a

943: variational formalism in which the Gibbs free energy of the system is

944: expressed as a functional of the solvent volume exclusion

945: function. Minimization of the free energy leads to a Laplace-Young

946: like equation for the solvent excluded cavity around the solutes,

947: which is extended to describe solvation on mesoscopic and microscopic

948: scales.  We have shown that the theory gives a reasonable description

949: of the solvation of microscopic solutes, such as ions and

950: alkanes. Improved accuracy will require further refinement of the

951: curvature dependence of the surface tension $\gamma(\vec r;[v])$ and

952: the definition of the position-dependent dielectric constant

953: $\epsilon(\vec r;[v])$. Given the physically reasonable values of the

954: parameters $\delta$ and $\xi$ we found by fitting, we hope that

955: extensions based on physical rational, e.g. given by complementary

956: microscopic

957: approaches\cite{beglov:jcp,zwanzig,lum:jpc,hummer:pnas,garde:prl} and

958: further empirical corrections, will lead to an accurate fit-parameter

959: free implicit solvent description.

960: 

961: We have further demonstrated that on larger scales, where solvent

962: dewetting can play an important role in solvation, our formalism

963: captures the delicate balance between hydrophobic, dispersive and

964: electrostatic forces which has been observed in previous

965: systems.\cite{huang:jpcb,pettitt,dzubiella:channel1,vaitheesvaran,dzubiella:jcp:2003,zhou:science,berne:nature}

966: The dewetting in our model is driven by the interfacial term which

967: favors minimizing the solute-solvent interface. A comment must be made

968: here regarding the sensitivity of dewetting to the particular

969: solvent-solute interactions. As recently argued by

970: Chandler,\cite{chandler:review} extended fluid interfaces near phase

971: coexistence are often referred to as 'soft' because they can be

972: deformed with only little or no free-energy change.\cite{safran} Our

973: approach seems to account for this sensitivity since small changes of

974: the constraints in the differential equation (\ref{diff}) for the

975: shape function, given e.g. by the dispersion potential close to the

976: solute surface, can lead to a major deformation or even rupture

977: (wetting transition) of the inter-solute, dewetted region. As we have

978: shown, this can significantly change the pmf for the solutes. Thus we

979: anticipate that slight changes in the geometry of a system, e.g. a

980: slight concave or convex bending of two plate-like

981: solutes,\cite{huang:jpcb,pettitt} can lead to very different results

982: for the dewetting magnitude and the pmf.

983: 

984: The current illustrations utilized spherical and cylindrical

985: symmetries.  More complex molecules, such as proteins, will require

986: solving the full three dimensional problem.  Numerical algorithms for

987: the calculation of interface evolution for more complicated geometries

988: are provided by efficient level-set methods or fast marching

989: methods.\cite{level} We believe that in the full three-dimensional

990: (3D) case, our method will be much more efficient than other

991: microscopic approaches which partly resolve the water structure and

992: are able to describe dewetting effects, e.g. the Lum-Chandler-Weeks

993: theory\cite{lum:jpc} (LCW) or information theory

994: \cite{hummer:pnas,garde:prl} (IT), as only a two-dimensional surface

995: is sought rather than a 3D density distribution on a fine grid. We

996: remark that LCW and IT do not consider electrostatic interactions and

997: may benefit from our complementary approach.

998: 

999: \section*{Acknowledgment}

1000: 

1001: The authors thank Tushar Jain, John Mongan, and Cameron Mura for

1002: useful discussions.  J.D. and J.M.J.S acknowledge financial support

1003: from a DFG Forschungsstipendium and the PFC-sponsored Center for

1004: Theoretical Biological Physics (Grants No. PHY-0216576 and

1005: PHY-0225630), respectively. Work in the McCammon group is additionally

1006: supported by NIH, HHMI, NBCR, and Accelrys, Inc.

1007: 

1008: \section*{Appendix A: Curvatures in cylindrical coordinates}

1009: In our general parametrization for the shape function $r(z)$ we

1010: express the radial coordinate by $r=r(l)$ and the axial coordinate by

1011: $z=z(l)$, as functions of the parameter $l$. Depending on the geometry

1012: of the considered system, $l$ has to be conveniently chosen, for

1013: instance to be the arc length, or $r(l)=l$, or $z(l)=l$. In our

1014: illustration the most convenient choice is $z(l)=l$. The principal

1015: curvatures are generally given by\cite{frankel}

1016: \begin{eqnarray}

1017: \kappa_1(r,z)=\frac{-z'}{r\sqrt{z'^2+r'^2}} \;{\rm ,}\;\kappa_2(r,z)=\frac{z'r''-z''r'}{(z'^2+r'^2)^{3/2}},

1018: \end{eqnarray}

1019: where the primes indicate the partial derivative with respect to

1020: $l$. Additionally, the unit normal vector reads

1021: \begin{eqnarray}

1022: \vec

1023: n(r,z)=\frac{1}{\sqrt{z'^2+r'^2}}\left(\begin{array}{c}z'\\-r'\end{array}\right).

1024: \end{eqnarray}

1025: The differential equation (\ref{diff}) is then solved by a forward

1026: relaxation scheme in time $t$

1027: \begin{eqnarray}

1028: \left(\begin{array}{c}r(t+\Delta t)\\z(t+\Delta

1029: t)\end{array}\right)=\left(\begin{array}{c}r(t)\\z(t)\end{array}\right)-\Delta

1030: t\, {\vec n}(r,z){\rm de}(r,z),

1031: \end{eqnarray}

1032: where the steady-state solution $\partial(r,z)/\partial t=0$ is the

1033: solution of ${\rm de}(r,z)=0$ we are looking for.  In the numerical

1034: calculations we use a grid of 500 bins and an integration time step of

1035: $\Delta t=$0.001.  The first and second derivatives are approximated

1036: using a symmetric two and three-step finite difference equation,

1037: respectively. Convergence is usually reached after 10$^5$

1038: time steps. The result is observed to be independent of the initial

1039: choice of $r(z)$ at $t=0$.

1040: 

1041: 

1042: \section*{Appendix B: Numerical solution of the PB equation}

1043: Since we neglect mobile ions in our work, the PB equation reduces to

1044: Poisson's equation. It is solved on a two dimensional grid in

1045: cylindrical coordinates $r$ and $z$ with a finite difference method.

1046: The gradient and Laplacian are given then by $\nabla=(\partial r,

1047: \partial z)$ and $\Delta=\partial_r+\partial_r/r+\partial_z^2$,

1048: respectively. The first and second derivatives are approximated using

1049: symmetric two or three-step finite-difference equations. An explicit,

1050: forward time relaxation scheme is used to find the solution of

1051: Poisson's equation:

1052: \begin{eqnarray}

1053: \Psi(t+\Delta t;\vec r)=\Psi(t;\vec r)-\Delta t {\rm PB}(\Psi(t;\vec r)).

1054: \end{eqnarray}

1055:  In most cases we use a lattice spacing of $\Delta r=\Delta z=0.4$\AA~

1056: on a $n_r\times n_z=100\times 200$ grid, and an integration time step

1057: $\Delta t=0.05$. Convergence takes approximately $10^5$ time

1058: steps. For the charged particles which are buried in the nanosolutes

1059: we use homogeneously charged spheres with a radius of 2\AA.  Instead

1060: of a sharp transition for the dielectric boundary ~(\ref{e}), we use a

1061: smoothing function for reasons of numerical stability:

1062: \begin{eqnarray}

1063: \epsilon(\vec r)=\frac{\epsilon_l-\epsilon_v}{\exp(\kappa d(\vec r))+1}+\epsilon_v,

1064: \label{e3}

1065: \end{eqnarray}

1066: where the absolute value of the length $d(\vec r)$ is given by the nearest

1067: distance to the boundary of the volume exclusion function $v(\vec

1068: r)$. $d$ is defined to be positive when $\vec r\in \cal V$

1069: and negative elsewhere. The inverse length $\kappa$ defines the width

1070: of the boundary region and in the limit $\kappa\rightarrow \infty$ we

1071: recover the sharp transition (\ref{e}).  We choose a value $\kappa

1072: \gtrsim 3{\rm \AA}^{-1}$ for which the solution of Poisson's equation

1073: becomes basically independent of the choice of $\kappa$. An example for the

1074: dielectric boundary is shown in Fig.~\ref{eps} for two partly dewetted

1075: nanosolutes of radius $R_0=15$\AA~ at a distance $s_0=7$\AA~ carrying

1076: a charge $Q=5e$ (system V in sec. III.C). 

1077: 

1078: In order to obtain the optimal shape function $v_{\rm min}(\vec r)$

1079: the shape equation (\ref{diff}) has to be solved simultaneously with

1080: Poisson's equation when the solutes are charged. In practice, we

1081: first solve (\ref{diff}) without any electrostatic contributions. In

1082: the second step, we solve  Poisson's equation with the dielectric

1083: boundary (\ref{e3}) given by the volume exclusion function of the

1084: former solution. The result for the electric energy density is then

1085: plugged back in the shape equation in the third step. The last two

1086: steps are repeated until the solution for $v_{\rm min}(\vec r)$ is

1087: fully converged. Since the results for $r(z)$ excluding and including

1088: electrostatics are quite similar for our systems, full convergence

1089: takes usually only 6 to 7 repetitions of the described iteration

1090: steps.

1091: 

1092: \begin{thebibliography}{71}

1093: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi

1094: \expandafter\ifx\csname bibnamefont\endcsname\relax

1095:   \def\bibnamefont#1{#1}\fi

1096: \expandafter\ifx\csname bibfnamefont\endcsname\relax

1097:   \def\bibfnamefont#1{#1}\fi

1098: \expandafter\ifx\csname citenamefont\endcsname\relax

1099:   \def\citenamefont#1{#1}\fi

1100: \expandafter\ifx\csname url\endcsname\relax

1101:   \def\url#1{\texttt{#1}}\fi

1102: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi

1103: \providecommand{\bibinfo}[2]{#2}

1104: \providecommand{\eprint}[2][]{\url{#2}}

1105: 

1106: \bibitem[{\citenamefont{Roux and Simonson}(1999)}]{roux:biochem}

1107: \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Roux}} \bibnamefont{and}

1108:   \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Simonson}},

1109:   \bibinfo{journal}{Biophys. Chem.} \textbf{\bibinfo{volume}{78}},

1110:   \bibinfo{pages}{1} (\bibinfo{year}{1999}).

1111: 

1112: \bibitem[{\citenamefont{Richards}(1977)}]{richards}

1113: \bibinfo{author}{\bibfnamefont{F.~M.} \bibnamefont{Richards}},

1114:   \bibinfo{journal}{Annu. Rev. Biophys. Bioeng.} \textbf{\bibinfo{volume}{6}},

1115:   \bibinfo{pages}{151} (\bibinfo{year}{1977}).

1116: 

1117: \bibitem[{\citenamefont{Gallicchio et~al.}(2000)\citenamefont{Gallicchio, Kubo,

1118:   and Levy}}]{gallicchio:jpcb}

1119: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Gallicchio}},

1120:   \bibinfo{author}{\bibfnamefont{M.~M.} \bibnamefont{Kubo}}, \bibnamefont{and}

1121:   \bibinfo{author}{\bibfnamefont{R.~M.} \bibnamefont{Levy}},

1122:   \bibinfo{journal}{J. Phys. Chem. B} \textbf{\bibinfo{volume}{104}},

1123:   \bibinfo{pages}{6271} (\bibinfo{year}{2000}).

1124: 

1125: \bibitem[{\citenamefont{Gallicchio et~al.}(2002)\citenamefont{Gallicchio,

1126:   Zhang, and Levy}}]{gallicchio:jcc}

1127: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Gallicchio}},

1128:   \bibinfo{author}{\bibfnamefont{L.~Y.} \bibnamefont{Zhang}}, \bibnamefont{and}

1129:   \bibinfo{author}{\bibfnamefont{R.~M.} \bibnamefont{Levy}},

1130:   \bibinfo{journal}{J. Comput. Chem.} \textbf{\bibinfo{volume}{23}},

1131:   \bibinfo{pages}{517} (\bibinfo{year}{2002}).

1132: 

1133: \bibitem[{\citenamefont{Zacharias}(2003)}]{zacharias}

1134: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Zacharias}},

1135:   \bibinfo{journal}{J. Chem. Phys. A} \textbf{\bibinfo{volume}{107}},

1136:   \bibinfo{pages}{3000} (\bibinfo{year}{2003}).

1137: 

1138: \bibitem[{\citenamefont{Su and Gallicchio}(2004)}]{su:biochem}

1139: \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Su}} \bibnamefont{and}

1140:   \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Gallicchio}},

1141:   \bibinfo{journal}{Biophys. Chem.} \textbf{\bibinfo{volume}{109}},

1142:   \bibinfo{pages}{251} (\bibinfo{year}{2004}).

1143: 

1144: \bibitem[{\citenamefont{Levy et~al.}(2003)\citenamefont{Levy, Zhang,

1145:   Gallicchio, and Felts}}]{levy:jacs}

1146: \bibinfo{author}{\bibfnamefont{R.~M.} \bibnamefont{Levy}},

1147:   \bibinfo{author}{\bibfnamefont{L.~Y.} \bibnamefont{Zhang}},

1148:   \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Gallicchio}},

1149:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{A.~K.} \bibnamefont{Felts}},

1150:   \bibinfo{journal}{J. Am. Chem. Soc.} \textbf{\bibinfo{volume}{125}},

1151:   \bibinfo{pages}{9523} (\bibinfo{year}{2003}).

1152: 

1153: \bibitem[{\citenamefont{Bashford and Case}(2000)}]{bashford}

1154: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Bashford}} \bibnamefont{and}

1155:   \bibinfo{author}{\bibfnamefont{D.~A.} \bibnamefont{Case}},

1156:   \bibinfo{journal}{Ann. Rev. Phys. Chem} \textbf{\bibinfo{volume}{51}},

1157:   \bibinfo{pages}{129} (\bibinfo{year}{2000}).

1158: 

1159: \bibitem[{\citenamefont{Sharp and Honig}(1990)}]{sharp:honig}

1160: \bibinfo{author}{\bibfnamefont{K.~A.} \bibnamefont{Sharp}} \bibnamefont{and}

1161:   \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Honig}}, \bibinfo{journal}{J.

1162:   Phys. Chem.} \textbf{\bibinfo{volume}{94}}, \bibinfo{pages}{7684}

1163:   (\bibinfo{year}{1990}).

1164: 

1165: \bibitem[{\citenamefont{Baker}(2005)}]{baker}

1166: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Baker}},

1167:   \bibinfo{journal}{Curr. Opin. Struct. Bio} \textbf{\bibinfo{volume}{15}},

1168:   \bibinfo{pages}{137} (\bibinfo{year}{2005}).

1169: 

1170: \bibitem[{\citenamefont{Nina et~al.}(1997)\citenamefont{Nina, Beglov, and

1171:   Roux}}]{nina}

1172: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Nina}},

1173:   \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Beglov}}, \bibnamefont{and}

1174:   \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Roux}}, \bibinfo{journal}{J.

1175:   Phys. Chem. B} \textbf{\bibinfo{volume}{101}}, \bibinfo{pages}{5239}

1176:   (\bibinfo{year}{1997}).

1177: 

1178: \bibitem[{\citenamefont{Huang and Chandler}(2002)}]{huang:jpcb:2002}

1179: \bibinfo{author}{\bibfnamefont{D.~M.} \bibnamefont{Huang}} \bibnamefont{and}

1180:   \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Chandler}},

1181:   \bibinfo{journal}{J. Phys. Chem. B} \textbf{\bibinfo{volume}{106}},

1182:   \bibinfo{pages}{2047} (\bibinfo{year}{2002}).

1183: 

1184: \bibitem[{\citenamefont{Ashbaugh et~al.}(1998)\citenamefont{Ashbaugh, Kaler,

1185:   and Paulaitis}}]{ashbaugh:biophys}

1186: \bibinfo{author}{\bibfnamefont{H.~S.} \bibnamefont{Ashbaugh}},

1187:   \bibinfo{author}{\bibfnamefont{E.~W.} \bibnamefont{Kaler}}, \bibnamefont{and}

1188:   \bibinfo{author}{\bibfnamefont{M.~E.} \bibnamefont{Paulaitis}},

1189:   \bibinfo{journal}{Biophys. J.} \textbf{\bibinfo{volume}{75}},

1190:   \bibinfo{pages}{755} (\bibinfo{year}{1998}).

1191: 

1192: \bibitem[{\citenamefont{Stillinger}(1973)}]{stilinger}

1193: \bibinfo{author}{\bibfnamefont{F.~H.} \bibnamefont{Stillinger}},

1194:   \bibinfo{journal}{J. Solution Chem.} \textbf{\bibinfo{volume}{2}},

1195:   \bibinfo{pages}{141} (\bibinfo{year}{1973}).

1196: 

1197: \bibitem[{\citenamefont{Lum et~al.}(1999)\citenamefont{Lum, Chandler, and

1198:   Weeks}}]{lum:jpc}

1199: \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Lum}},

1200:   \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Chandler}}, \bibnamefont{and}

1201:   \bibinfo{author}{\bibfnamefont{J.~D.} \bibnamefont{Weeks}},

1202:   \bibinfo{journal}{J. Phys. Chem. B} \textbf{\bibinfo{volume}{103}},

1203:   \bibinfo{pages}{4570} (\bibinfo{year}{1999}).

1204: 

1205: \bibitem[{\citenamefont{Hummer et~al.}(1997)\citenamefont{Hummer, Pratt, and

1206:   Garcia}}]{hummer:jcp:1997}

1207: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Hummer}},

1208:   \bibinfo{author}{\bibfnamefont{L.~R.} \bibnamefont{Pratt}}, \bibnamefont{and}

1209:   \bibinfo{author}{\bibfnamefont{A.~E.} \bibnamefont{Garcia}},

1210:   \bibinfo{journal}{J. Chem. Phys} \textbf{\bibinfo{volume}{107}},

1211:   \bibinfo{pages}{9275} (\bibinfo{year}{1997}).

1212: 

1213: \bibitem[{\citenamefont{Chandler}(2005)}]{chandler:review}

1214: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Chandler}},

1215:   \bibinfo{journal}{Nature} \textbf{\bibinfo{volume}{437}},

1216:   \bibinfo{pages}{640} (\bibinfo{year}{2005}).

1217: 

1218: \bibitem[{\citenamefont{Wallquist and Berne}(1995)}]{wallquist:jpc}

1219: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Wallquist}} \bibnamefont{and}

1220:   \bibinfo{author}{\bibfnamefont{B.~J.} \bibnamefont{Berne}},

1221:   \bibinfo{journal}{J. Phys. Chem.} \textbf{\bibinfo{volume}{99}},

1222:   \bibinfo{pages}{2893} (\bibinfo{year}{1995}).

1223: 

1224: \bibitem[{\citenamefont{Huang et~al.}(2003)\citenamefont{Huang, Margulis, and

1225:   Berne}}]{huang:pnas}

1226: \bibinfo{author}{\bibfnamefont{X.}~\bibnamefont{Huang}},

1227:   \bibinfo{author}{\bibfnamefont{C.~J.} \bibnamefont{Margulis}},

1228:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{B.~J.} \bibnamefont{Berne}},

1229:   \bibinfo{journal}{PNAS} \textbf{\bibinfo{volume}{100}},

1230:   \bibinfo{pages}{11953} (\bibinfo{year}{2003}).

1231: 

1232: \bibitem[{\citenamefont{Huang et~al.}(2005)\citenamefont{Huang, Zhou, and

1233:   Berne}}]{huang:jpcb}

1234: \bibinfo{author}{\bibfnamefont{X.}~\bibnamefont{Huang}},

1235:   \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Zhou}}, \bibnamefont{and}

1236:   \bibinfo{author}{\bibfnamefont{B.~J.} \bibnamefont{Berne}},

1237:   \bibinfo{journal}{J. Phys. Chem. B} \textbf{\bibinfo{volume}{109}},

1238:   \bibinfo{pages}{3546} (\bibinfo{year}{2005}).

1239: 

1240: \bibitem[{\citenamefont{Choudhury and Pettitt}(2005)}]{pettitt}

1241: \bibinfo{author}{\bibfnamefont{N.}~\bibnamefont{Choudhury}} \bibnamefont{and}

1242:   \bibinfo{author}{\bibfnamefont{B.~M.} \bibnamefont{Pettitt}},

1243:   \bibinfo{journal}{J. Am. Chem. Soc.} \textbf{\bibinfo{volume}{127}},

1244:   \bibinfo{pages}{3556} (\bibinfo{year}{2005}).

1245: 

1246: \bibitem[{\citenamefont{Hummer et~al.}(2001)\citenamefont{Hummer, Rasaiah, and

1247:   Nowortya}}]{hummer:nature}

1248: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Hummer}},

1249:   \bibinfo{author}{\bibfnamefont{J.~C.} \bibnamefont{Rasaiah}},

1250:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{J.~P.}

1251:   \bibnamefont{Nowortya}}, \bibinfo{journal}{Nature}

1252:   \textbf{\bibinfo{volume}{414}}, \bibinfo{pages}{188} (\bibinfo{year}{2001}).

1253: 

1254: \bibitem[{\citenamefont{Allen et~al.}(2002)\citenamefont{Allen, Melchionna, and

1255:   Hansen}}]{allen:prl}

1256: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Allen}},

1257:   \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Melchionna}},

1258:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{J.-P.}

1259:   \bibnamefont{Hansen}}, \bibinfo{journal}{Phys. Rev. Lett.}

1260:   \textbf{\bibinfo{volume}{89}}, \bibinfo{pages}{175502}

1261:   (\bibinfo{year}{2002}).

1262: 

1263: \bibitem[{\citenamefont{Beckstein et~al.}(2001)\citenamefont{Beckstein, Biggin,

1264:   and Sansom}}]{beckstein:jpc}

1265: \bibinfo{author}{\bibfnamefont{O.}~\bibnamefont{Beckstein}},

1266:   \bibinfo{author}{\bibfnamefont{P.~C.} \bibnamefont{Biggin}},

1267:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{M.~S.~P.}

1268:   \bibnamefont{Sansom}}, \bibinfo{journal}{J. Phys. Chem.}

1269:   \textbf{\bibinfo{volume}{105}}, \bibinfo{pages}{12902}

1270:   (\bibinfo{year}{2001}).

1271: 

1272: \bibitem[{\citenamefont{Anishkin and Sukharev}(2004)}]{sukharev}

1273: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Anishkin}} \bibnamefont{and}

1274:   \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Sukharev}},

1275:   \bibinfo{journal}{Biophys. J.} \textbf{\bibinfo{volume}{86}},

1276:   \bibinfo{pages}{2883} (\bibinfo{year}{2004}).

1277: 

1278: \bibitem[{\citenamefont{Dzubiella and

1279:   Hansen}(2004{\natexlab{a}})}]{dzubiella:channel1}

1280: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Dzubiella}} \bibnamefont{and}

1281:   \bibinfo{author}{\bibfnamefont{J.-P.} \bibnamefont{Hansen}},

1282:   \bibinfo{journal}{J. Chem. Phys.} \textbf{\bibinfo{volume}{120}},

1283:   \bibinfo{pages}{5001} (\bibinfo{year}{2004}{\natexlab{a}}).

1284: 

1285: \bibitem[{\citenamefont{Vaitheesvaran et~al.}(2004)\citenamefont{Vaitheesvaran,

1286:   Rasaiah, and Hummer}}]{vaitheesvaran}

1287: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Vaitheesvaran}},

1288:   \bibinfo{author}{\bibfnamefont{J.~C.} \bibnamefont{Rasaiah}},

1289:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Hummer}},

1290:   \bibinfo{journal}{J. Chem. Phys.} \textbf{\bibinfo{volume}{121}},

1291:   \bibinfo{pages}{7955} (\bibinfo{year}{2004}).

1292: 

1293: \bibitem[{\citenamefont{Dzubiella and Hansen}(2005)}]{dzubiella:channel2}

1294: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Dzubiella}} \bibnamefont{and}

1295:   \bibinfo{author}{\bibfnamefont{J.-P.} \bibnamefont{Hansen}},

1296:   \bibinfo{journal}{J. Chem. Phys.} \textbf{\bibinfo{volume}{122}},

1297:   \bibinfo{pages}{234706} (\bibinfo{year}{2005}).

1298: 

1299: \bibitem[{\citenamefont{Dzubiella and Hansen}(2003)}]{dzubiella:jcp:2003}

1300: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Dzubiella}} \bibnamefont{and}

1301:   \bibinfo{author}{\bibfnamefont{J.-P.} \bibnamefont{Hansen}},

1302:   \bibinfo{journal}{J. Chem. Phys.} \textbf{\bibinfo{volume}{119}},

1303:   \bibinfo{pages}{12049} (\bibinfo{year}{2003}).

1304: 

1305: \bibitem[{\citenamefont{Dzubiella and

1306:   Hansen}(2004{\natexlab{b}})}]{dzubiella:jcp:2004}

1307: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Dzubiella}} \bibnamefont{and}

1308:   \bibinfo{author}{\bibfnamefont{J.-P.} \bibnamefont{Hansen}},

1309:   \bibinfo{journal}{J. Chem. Phys.} \textbf{\bibinfo{volume}{121}},

1310:   \bibinfo{pages}{5514} (\bibinfo{year}{2004}{\natexlab{b}}).

1311: 

1312: \bibitem[{\citenamefont{Zhou et~al.}(2004)\citenamefont{Zhou, Huang, Margulis,

1313:   and Berne}}]{zhou:science}

1314: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Zhou}},

1315:   \bibinfo{author}{\bibfnamefont{X.}~\bibnamefont{Huang}},

1316:   \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Margulis}}, \bibnamefont{and}

1317:   \bibinfo{author}{\bibfnamefont{B.~J.} \bibnamefont{Berne}},

1318:   \bibinfo{journal}{Science} \textbf{\bibinfo{volume}{305}},

1319:   \bibinfo{pages}{1605} (\bibinfo{year}{2004}).

1320: 

1321: \bibitem[{\citenamefont{Liu et~al.}(2005)\citenamefont{Liu, Huang, Zhou, and

1322:   Berne}}]{berne:nature}

1323: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Liu}},

1324:   \bibinfo{author}{\bibfnamefont{X.}~\bibnamefont{Huang}},

1325:   \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Zhou}}, \bibnamefont{and}

1326:   \bibinfo{author}{\bibfnamefont{B.~J.} \bibnamefont{Berne}},

1327:   \bibinfo{journal}{Nature} \textbf{\bibinfo{volume}{437}},

1328:   \bibinfo{pages}{159} (\bibinfo{year}{2005}).

1329: 

1330: \bibitem[{\citenamefont{Parker et~al.}(1994)\citenamefont{Parker, Claesson, and

1331:   Attard}}]{attard}

1332: \bibinfo{author}{\bibfnamefont{J.~L.} \bibnamefont{Parker}},

1333:   \bibinfo{author}{\bibfnamefont{P.~M.} \bibnamefont{Claesson}},

1334:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Attard}},

1335:   \bibinfo{journal}{J. Phys. Chem.} \textbf{\bibinfo{volume}{98}},

1336:   \bibinfo{pages}{8468} (\bibinfo{year}{1994}).

1337: 

1338: \bibitem[{\citenamefont{Beglov and Roux}(1996)}]{beglov:jcp}

1339: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Beglov}} \bibnamefont{and}

1340:   \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Roux}}, \bibinfo{journal}{J.

1341:   Chem. Phys.} \textbf{\bibinfo{volume}{104}}, \bibinfo{pages}{8678}

1342:   (\bibinfo{year}{1996}).

1343: 

1344: \bibitem[{\citenamefont{Kralchevsky and Nagayama}(Amsterdam)}]{kralchevsky}

1345: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Kralchevsky}} \bibnamefont{and}

1346:   \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Nagayama}},

1347:   \emph{\bibinfo{title}{Particles at Fluid Interfaces and Membranes}}

1348:   (\bibinfo{publisher}{Elsevier}, \bibinfo{address}{2001},

1349:   \bibinfo{year}{Amsterdam}).

1350: 

1351: \bibitem[{\citenamefont{Helfrich}(1973)}]{helfrich}

1352: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Helfrich}},

1353:   \bibinfo{journal}{Z. Naturforsch. C} \textbf{\bibinfo{volume}{28}},

1354:   \bibinfo{pages}{693} (\bibinfo{year}{1973}).

1355: 

1356: \bibitem[{\citenamefont{Zhong-can and Helfrich}(1989)}]{helfrich2}

1357: \bibinfo{author}{\bibfnamefont{O.-Y.} \bibnamefont{Zhong-can}}

1358:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Helfrich}},

1359:   \bibinfo{journal}{Phys. Rev. A} \textbf{\bibinfo{volume}{39}},

1360:   \bibinfo{pages}{5280} (\bibinfo{year}{1989}).

1361: 

1362: \bibitem[{\citenamefont{Bieker and Dietrich}(1998)}]{bieker}

1363: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Bieker}} \bibnamefont{and}

1364:   \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Dietrich}},

1365:   \bibinfo{journal}{Physica A} \textbf{\bibinfo{volume}{252}},

1366:   \bibinfo{pages}{85} (\bibinfo{year}{1998}).

1367: 

1368: \bibitem[{\citenamefont{Chou}(2001)}]{electrowetting}

1369: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Chou}}, \bibinfo{journal}{Phys.

1370:   Rev. Lett.} \textbf{\bibinfo{volume}{87}}, \bibinfo{pages}{106101}

1371:   (\bibinfo{year}{2001}).

1372: 

1373: \bibitem[{\citenamefont{Allen et~al.}(2003)\citenamefont{Allen, Melchionna, and

1374:   Hansen}}]{allen:jcp}

1375: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Allen}},

1376:   \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Melchionna}},

1377:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{J.-P.}

1378:   \bibnamefont{Hansen}}, \bibinfo{journal}{J. Chem. Phys.}

1379:   \textbf{\bibinfo{volume}{119}}, \bibinfo{pages}{3905} (\bibinfo{year}{2003}).

1380: 

1381: \bibitem[{\citenamefont{Attard}(2000)}]{attard2}

1382: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Attard}},

1383:   \bibinfo{journal}{Langmuir} \textbf{\bibinfo{volume}{16}},

1384:   \bibinfo{pages}{4455} (\bibinfo{year}{2000}).

1385: 

1386: \bibitem[{\citenamefont{Cheng and Rossky}(1998)}]{rossky:nature}

1387: \bibinfo{author}{\bibfnamefont{Y.-K.} \bibnamefont{Cheng}} \bibnamefont{and}

1388:   \bibinfo{author}{\bibfnamefont{P.~J.} \bibnamefont{Rossky}},

1389:   \bibinfo{journal}{Nature} \textbf{\bibinfo{volume}{392}},

1390:   \bibinfo{pages}{696} (\bibinfo{year}{1998}).

1391: 

1392: \bibitem[{\citenamefont{Gerstein and Chothia}(1996)}]{gerstein}

1393: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Gerstein}} \bibnamefont{and}

1394:   \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Chothia}},

1395:   \bibinfo{journal}{Proc. Natl. Acad. Sci. USA} \textbf{\bibinfo{volume}{93}},

1396:   \bibinfo{pages}{10167} (\bibinfo{year}{1996}).

1397: 

1398: \bibitem[{\citenamefont{Lau et~al.}(2003)\citenamefont{Lau, Bico, Teo,

1399:   Chhowalla, Amaratunga, Milne, McKinley, and Gleason}}]{super}

1400: \bibinfo{author}{\bibfnamefont{K.~K.~S.} \bibnamefont{Lau}},

1401:   \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Bico}},

1402:   \bibinfo{author}{\bibfnamefont{K.~B.~K.} \bibnamefont{Teo}},

1403:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Chhowalla}},

1404:   \bibinfo{author}{\bibfnamefont{G.~A.~J.} \bibnamefont{Amaratunga}},

1405:   \bibinfo{author}{\bibfnamefont{W.~I.} \bibnamefont{Milne}},

1406:   \bibinfo{author}{\bibfnamefont{G.~H.} \bibnamefont{McKinley}},

1407:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{K.~K.}

1408:   \bibnamefont{Gleason}}, \bibinfo{journal}{Nano Letters}

1409:   \textbf{\bibinfo{volume}{3}}, \bibinfo{pages}{1701} (\bibinfo{year}{2003}).

1410: 

1411: \bibitem[{\citenamefont{Dzubiella et~al.}(2005)\citenamefont{Dzubiella,

1412:   Swanson, and McCammon}}]{prl}

1413: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Dzubiella}},

1414:   \bibinfo{author}{\bibfnamefont{J.~M.~J.} \bibnamefont{Swanson}},

1415:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{J.~A.}

1416:   \bibnamefont{McCammon}} (\bibinfo{year}{2005}), \bibinfo{note}{to be

1417:   published}.

1418: 

1419: \bibitem[{\citenamefont{Triezenberg and Zwanzig}(1972)}]{zwanzig}

1420: \bibinfo{author}{\bibfnamefont{D.~G.} \bibnamefont{Triezenberg}}

1421:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Zwanzig}},

1422:   \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{28}},

1423:   \bibinfo{pages}{1183} (\bibinfo{year}{1972}).

1424: 

1425: \bibitem[{\citenamefont{Jackson}(1999)}]{jackson}

1426: \bibinfo{author}{\bibfnamefont{J.~D.} \bibnamefont{Jackson}},

1427:   \emph{\bibinfo{title}{Classical Electrodynamics}}

1428:   (\bibinfo{publisher}{Wiley}, \bibinfo{address}{New York},

1429:   \bibinfo{year}{1999}), \bibinfo{note}{3rd edn.}

1430: 

1431: \bibitem[{\citenamefont{Gilson et~al.}(1993)\citenamefont{Gilson, Davis, Luty,

1432:   and McCammon}}]{gilson}

1433: \bibinfo{author}{\bibfnamefont{M.~K.} \bibnamefont{Gilson}},

1434:   \bibinfo{author}{\bibfnamefont{M.~E.} \bibnamefont{Davis}},

1435:   \bibinfo{author}{\bibfnamefont{B.~A.} \bibnamefont{Luty}}, \bibnamefont{and}

1436:   \bibinfo{author}{\bibfnamefont{J.~A.} \bibnamefont{McCammon}},

1437:   \bibinfo{journal}{J. Phys. Chem} \textbf{\bibinfo{volume}{97}},

1438:   \bibinfo{pages}{3591} (\bibinfo{year}{1993}).

1439: 

1440: \bibitem[{\citenamefont{Huang et~al.}(2001)\citenamefont{Huang, Geissler, and

1441:   Chandler}}]{huang:jpc}

1442: \bibinfo{author}{\bibfnamefont{D.~M.} \bibnamefont{Huang}},

1443:   \bibinfo{author}{\bibfnamefont{P.~L.} \bibnamefont{Geissler}},

1444:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Chandler}},

1445:   \bibinfo{journal}{J. Phys. Chem. B} \textbf{\bibinfo{volume}{105}},

1446:   \bibinfo{pages}{6704} (\bibinfo{year}{2001}).

1447: 

1448: \bibitem[{pri()}]{prince}

1449: \bibinfo{note}{The principal curvatures are formally given by the eigenvectors

1450:   of the Hessian (or shape operator) of $v$, $\hat S$, which can be expressed

1451:   by the vector gradient of the unit normal vector field $\vec n(\vec r)$ of

1452:   the surface, $\hat S=\nabla \vec n(\vec r)={\bf \nabla} \frac{\nabla v(\vec

1453:   r)}{|\nabla v(\vec r)|}$.}

1454: 

1455: \bibitem[{\citenamefont{Henderson}(1992)}]{hendersonreview}

1456: \bibinfo{author}{\bibfnamefont{J.~R.} \bibnamefont{Henderson}},

1457:   \emph{\bibinfo{title}{Fundamentals of Inhomogeneous Fluids}}

1458:   (\bibinfo{publisher}{Dekker}, \bibinfo{address}{New York},

1459:   \bibinfo{year}{1992}), \bibinfo{note}{edited by D.~Henderson}.

1460: 

1461: \bibitem[{\citenamefont{Stewart and Evans}(2005)}]{evans2}

1462: \bibinfo{author}{\bibfnamefont{M.~C.} \bibnamefont{Stewart}} \bibnamefont{and}

1463:   \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Evans}},

1464:   \bibinfo{journal}{Phys. Rev. E} \textbf{\bibinfo{volume}{71}},

1465:   \bibinfo{pages}{011602} (\bibinfo{year}{2005}).

1466: 

1467: \bibitem[{\citenamefont{Reiss}(1965)}]{spt}

1468: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Reiss}}, \bibinfo{journal}{Adv.

1469:   Chem. Phys.} \textbf{\bibinfo{volume}{9}}, \bibinfo{pages}{1}

1470:   (\bibinfo{year}{1965}).

1471: 

1472: \bibitem[{\citenamefont{Tolman}(1949)}]{tolman}

1473: \bibinfo{author}{\bibfnamefont{R.~C.} \bibnamefont{Tolman}},

1474:   \bibinfo{journal}{J. Chem. Phys.} \textbf{\bibinfo{volume}{17}},

1475:   \bibinfo{pages}{333} (\bibinfo{year}{1949}).

1476: 

1477: \bibitem[{\citenamefont{Nicholls et~al.}(1991)\citenamefont{Nicholls, Sharp,

1478:   and Honig}}]{nicholls}

1479: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Nicholls}},

1480:   \bibinfo{author}{\bibfnamefont{K.~A.} \bibnamefont{Sharp}}, \bibnamefont{and}

1481:   \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Honig}},

1482:   \bibinfo{journal}{Proteins} \textbf{\bibinfo{volume}{11}},

1483:   \bibinfo{pages}{281} (\bibinfo{year}{1991}).

1484: 

1485: \bibitem[{\citenamefont{Luo and Tucker}(1996)}]{luo:jpc}

1486: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Luo}} \bibnamefont{and}

1487:   \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Tucker}},

1488:   \bibinfo{journal}{J. Phys. Chem.} \textbf{\bibinfo{volume}{100}},

1489:   \bibinfo{pages}{11165} (\bibinfo{year}{1996}).

1490: 

1491: \bibitem[{\citenamefont{Berendsen et~al.}(1987)\citenamefont{Berendsen,

1492:   Grigera, and Straatsma}}]{berendsen:jpc}

1493: \bibinfo{author}{\bibfnamefont{H.~J.~C.} \bibnamefont{Berendsen}},

1494:   \bibinfo{author}{\bibfnamefont{J.~R.} \bibnamefont{Grigera}},

1495:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{T.~P.}

1496:   \bibnamefont{Straatsma}}, \bibinfo{journal}{J. Phys. Chem.}

1497:   \textbf{\bibinfo{volume}{91}}, \bibinfo{pages}{6269} (\bibinfo{year}{1987}).

1498: 

1499: \bibitem[{\citenamefont{Manjari and Kim}(2005)}]{kim}

1500: \bibinfo{author}{\bibfnamefont{S.~R.} \bibnamefont{Manjari}} \bibnamefont{and}

1501:   \bibinfo{author}{\bibfnamefont{H.~J.} \bibnamefont{Kim}},

1502:   \bibinfo{journal}{J. Chem. Phys.} \textbf{\bibinfo{volume}{123}},

1503:   \bibinfo{pages}{014504} (\bibinfo{year}{2005}).

1504: 

1505: \bibitem[{\citenamefont{Hummer et~al.}(1996{\natexlab{a}})\citenamefont{Hummer,

1506:   Pratt, and Garcia}}]{hummer:jpc:1996}

1507: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Hummer}},

1508:   \bibinfo{author}{\bibfnamefont{L.~R.} \bibnamefont{Pratt}}, \bibnamefont{and}

1509:   \bibinfo{author}{\bibfnamefont{A.~E.} \bibnamefont{Garcia}},

1510:   \bibinfo{journal}{J. Phys. Chem} \textbf{\bibinfo{volume}{100}},

1511:   \bibinfo{pages}{1206} (\bibinfo{year}{1996}{\natexlab{a}}).

1512: 

1513: \bibitem[{\citenamefont{Paschek}(2004)}]{paschek}

1514: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Paschek}}, \bibinfo{journal}{J.

1515:   Chem. Phys.} \textbf{\bibinfo{volume}{120}}, \bibinfo{pages}{6674}

1516:   (\bibinfo{year}{2004}).

1517: 

1518: \bibitem[{\citenamefont{Alejandre et~al.}(1995)\citenamefont{Alejandre,

1519:   Tildesley, and Chapela}}]{alejandre}

1520: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Alejandre}},

1521:   \bibinfo{author}{\bibfnamefont{D.~J.} \bibnamefont{Tildesley}},

1522:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{G.~A.}

1523:   \bibnamefont{Chapela}}, \bibinfo{journal}{J. Chem. Phys.}

1524:   \textbf{\bibinfo{volume}{102}}, \bibinfo{pages}{4574} (\bibinfo{year}{1995}).

1525: 

1526: \bibitem[{\citenamefont{Connolly}(1993)}]{connolly}

1527: \bibinfo{author}{\bibfnamefont{M.~L.} \bibnamefont{Connolly}},

1528:   \bibinfo{journal}{J. Mol. Graphics} \textbf{\bibinfo{volume}{11}},

1529:   \bibinfo{pages}{139} (\bibinfo{year}{1993}).

1530: 

1531: \bibitem[{sig()}]{sigmanote}

1532: \bibinfo{note}{The solute-solute LJ parameters have been calculated from the

1533:   solute-water LJ parameters employing the usual combining rules. Note that the

1534:   rules used by Ashbaugh et al.,\cite{ashbaugh:biophys} are different to those

1535:   of Hummer et al.\cite{hummer:jpc:1996} and Paschek\cite{paschek}}.

1536: 

1537: \bibitem[{\citenamefont{van~der Spoel et~al.}(1998)\citenamefont{van~der Spoel,

1538:   van Maaren, and Berendsen}}]{spoel}

1539: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{van~der Spoel}},

1540:   \bibinfo{author}{\bibfnamefont{P.~J.} \bibnamefont{van Maaren}},

1541:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{H.~J.~C.}

1542:   \bibnamefont{Berendsen}}, \bibinfo{journal}{J. Chem. Phys.}

1543:   \textbf{\bibinfo{volume}{108}}, \bibinfo{pages}{10220}

1544:   (\bibinfo{year}{1998}).

1545: 

1546: \bibitem[{\citenamefont{Jorgensen et~al.}(1984)\citenamefont{Jorgensen, Madura,

1547:   and Swenson}}]{jorgensen}

1548: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Jorgensen}},

1549:   \bibinfo{author}{\bibfnamefont{J.~D.} \bibnamefont{Madura}},

1550:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{C.~J.}

1551:   \bibnamefont{Swenson}}, \bibinfo{journal}{J. Am. Chem. Soc.}

1552:   \textbf{\bibinfo{volume}{106}}, \bibinfo{pages}{6638} (\bibinfo{year}{1984}).

1553: 

1554: \bibitem[{\citenamefont{Fernandez}(2004)}]{fernandez}

1555: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Fernandez}},

1556:   \bibinfo{journal}{Nature Biotech.} \textbf{\bibinfo{volume}{22}},

1557:   \bibinfo{pages}{1081} (\bibinfo{year}{2004}).

1558: 

1559: \bibitem[{\citenamefont{Hummer et~al.}(1996{\natexlab{b}})\citenamefont{Hummer,

1560:   Garde, Garcia, Phorille, and Pratt}}]{hummer:pnas}

1561: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Hummer}},

1562:   \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Garde}},

1563:   \bibinfo{author}{\bibfnamefont{A.~E.} \bibnamefont{Garcia}},

1564:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Phorille}}, \bibnamefont{and}

1565:   \bibinfo{author}{\bibfnamefont{L.~R.} \bibnamefont{Pratt}},

1566:   \bibinfo{journal}{Proc. Natl. Acad. Sci. USA} \textbf{\bibinfo{volume}{93}},

1567:   \bibinfo{pages}{8951} (\bibinfo{year}{1996}{\natexlab{b}}).

1568: 

1569: \bibitem[{\citenamefont{Garde et~al.}(1996)\citenamefont{Garde, Hummer, Garcia,

1570:   Paulaitis, and Pratt}}]{garde:prl}

1571: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Garde}},

1572:   \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Hummer}},

1573:   \bibinfo{author}{\bibfnamefont{A.~E.} \bibnamefont{Garcia}},

1574:   \bibinfo{author}{\bibfnamefont{M.~E.} \bibnamefont{Paulaitis}},

1575:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{L.~R.} \bibnamefont{Pratt}},

1576:   \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{77}},

1577:   \bibinfo{pages}{4966} (\bibinfo{year}{1996}).

1578: 

1579: \bibitem[{\citenamefont{Safran}(1994)}]{safran}

1580: \bibinfo{author}{\bibfnamefont{S.~A.} \bibnamefont{Safran}},

1581:   \emph{\bibinfo{title}{Statistical Thermodynamics of Surfaces, Interfaces and

1582:   Membranes}} (\bibinfo{publisher}{Addison-Wesley}, \bibinfo{address}{Reading},

1583:   \bibinfo{year}{1994}), \bibinfo{note}{ch. 1-3}.

1584: 

1585: \bibitem[{\citenamefont{Sethian}(1999)}]{level}

1586: \bibinfo{author}{\bibfnamefont{J.~A.} \bibnamefont{Sethian}},

1587:   \emph{\bibinfo{title}{Level Set methods and Fast Marching Methods: Evolving

1588:   Interfaces in Geometry, Fluid Mechanics, Computer Vision, and Materials

1589:   Science}} (\bibinfo{publisher}{Cambridge University Press},

1590:   \bibinfo{address}{Cambridge}, \bibinfo{year}{1999}).

1591: 

1592: \bibitem[{\citenamefont{Frankel}(1997)}]{frankel}

1593: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Frankel}},

1594:   \emph{\bibinfo{title}{The Geometry of Physics}}

1595:   (\bibinfo{publisher}{Cambridge University Press},

1596:   \bibinfo{address}{Cambridge}, \bibinfo{year}{1997}).

1597: 

1598: \end{thebibliography}

1599: 

1600: \newpage

1601: 

1602: \begin{table}

1603: \begin{center}

1604: \begin{tabular}{l | c c c | c c | c }

1605: Solute & $\epsilon/({\rm kJ/mol})$\; & $\sigma/$\AA\; &

1606: $\Delta G_{\rm sim}/({\rm kJ/mol})$ & $\delta_{\rm bf}/$\AA &

1607: $R_{\rm min}/$\AA \\  \hline \hline SPC & 0.65 & 3.17 & -- & -- & --

1608: \\ SPC/E & 0.65 & 3.17 & -- & -- & -- \\ \hline \hline

1609: ref\cite{hummer:jpc:1996} & & & & & \\  Na$^0$ & 0.2005 & 2.85 &

1610: 9.2(1) & 0.79 & 2.32 \\ K$^0$ & 0.0061 & 4.52 & 23.7(5) & 0.76 & 2.83

1611: \\ Ca$^{0}$ & 0.6380 & 3.17 & 10.2(3) & 0.85 & 2.80 \\ F$^0$ & 0.5538

1612: & 3.05 & 9.7(2) & 0.85 & 2.68 \\ Cl$^0$ & 0.5380 & 3.75 & 21(3) & 0.80

1613: & 3.30 \\  Br$^0$ & 0.4945 & 3.83 & 24(3) & 0.77 & 3.35 \\  \hline

1614: \hline  ref\cite{paschek} & & & & & \\  Ne & 0.3156 & 3.10 &

1615: 11.41(0.05)& 0.84 & 2.61 \\  Ar & 0.8176 & 3.29 & 8.68 (0.08) & 0.90 &

1616: 2.96\\  Kr & 0.9518 & 3.42 & 8.12 (0.1) & 0.91 & 3.10 \\  Xe & 1.0710

1617: & 3.57 & 7.65 (0.15) & 0.92 & 3.27 \\ \hline  \hline

1618: ref\cite{ashbaugh:biophys} & & & & & \\ 

1619: Me & 0.8941 &3.44 & 10.96 (0.46) & 0.85 & 3.11 \\  

1620: \hline 

1621: ethane & -- & -- & 10.75 (0.50) & 0.87 & -- \\

1622: CH$_3$ & 0.7503 & 3.46&-- & -- & -- \\  

1623: \hline 

1624: propane & -- & -- & 13.81 (0.54) & 0.94 &-- \\ 

1625: butane & -- & -- & 14.69 (0.54) & 0.96 &-- \\ 

1626: CH$_2$ & 0.5665& 3.52 & -- & -- & -- \\ 

1627: CH$_3$ & 0.6900 & 3.52 & -- & -- & -- \\

1628: \end{tabular}

1629: \caption{Solute-water LJ parameters and solvation free energy $\Delta

1630:  G_{\rm sim}$ for neutral Lennard-Jones spheres from the SPC water

1631:  simulations performed by Hummer {\it et al.}\cite{hummer:jpc:1996}

1632:  and Paschek.\cite{paschek} $\delta_{\rm bf}$ is the Tolman length

1633:  best fit to $\Delta G_{\rm sim}$ (rounded to two digits after the

1634:  decimal point). $R_{\rm min}$ is the resulting optimal radius

1635:  excluded of solvent. Also shown are the values for the simple alkanes

1636:  methane (Me), ethane, propane, and butane from the study of Ashbaugh

1637:  {\it et al.}\cite{ashbaugh:biophys} Simulation errors are given in

1638:  parentheses.}

1639: \label{tab1}

1640: \end{center}

1641: \end{table} 

1642: 

1643: 

1644: \begin{table}

1645: \begin{center}

1646: \begin{tabular}{l | c c c}

1647:    Atom & $\Delta G_{\rm sim}/({\rm kJ\,mol}^{-1})$ & $\delta_{\rm

1648: bf}/$\AA & $R_{\rm min}/$\AA \\  \hline Ne & 11.65 (0.05) &0.88 & 2.60

1649: \\ Ar & 8.83 (0.08) &0.96 & 2.94 \\ Kr & 8.20 (0.1) &0.98 & 3.09 \\ Xe

1650: & 7.58 (0.15) &1.00 & 3.25 \\

1651: \end{tabular}

1652: \caption{Solvation free energies for neutral Lennard-Jones spheres in

1653: SPC/E water from the simulations of Paschek.\cite{paschek}

1654: $\delta_{\rm bf}$ and $R_{\rm min}$ are defined as in Tab.~I.}

1655: \label{tab1}

1656: \end{center}

1657: \end{table} 

1658: 

1659: \begin{table}

1660: \begin{center}

1661: \begin{tabular}{l | c c c c c}

1662:   Ion & $q$ & $\frac{\Delta G_{\rm sim}}{{\rm kJ\,mol}^{-1}}$ & $\frac{\Delta G}{{\rm kJ\,mol}^{-1}}$ & $\frac{\Delta G_{\xi}}{{\rm kJ\,mol}^{-1}}$ & $R_{\rm min}/$\AA \\ 

1663: \hline 

1664: Na$^+$ & 1 &-398 &-334     &-394 &1.83 \\ 

1665: K$^+$ & 1 & -271 &-246     &-282 &2.35 \\ 

1666: Ca$^{2+}$ & 2 &-1306 &-1181&-1364 &2.09 \\ 

1667: F$^-$ & -1 & -580 &-274    &-630 &2.25 \\ 

1668: Cl$^-$ & -1 &-371 &-198    &-342 &2.97 \\ 

1669: Br$^-$ & -1 & -358 &-192   &-328 &3.02 \\

1670: \end{tabular}

1671: \caption{Solvation free energies for charged LJ spheres in SPC water

1672: from the simulations of Hummer {\it et al.}\cite{hummer:jpc:1996}

1673: compared to the theoretical result $\Delta G$. For $\delta$ we use the

1674: best fits $\delta_{\rm bf}$ to the solvation of neutral spheres as

1675: shown in Tab.~I. $\Delta G_{\xi}$ is the result when the dielectric

1676: boundary shift $\xi$ is applied, see text.}

1677: \label{tab1}

1678: \end{center}

1679: \end{table}

1680: 

1681: \begin{table}

1682: \begin{center}

1683: \begin{tabular}{l | c c c c c}

1684:   System & $\delta/{\rm \AA}\;\;\;$ & vdW attraction & $\;\;\;Z$ & $W(s_0)/k_BT$ & dewetted\\ 

1685: \hline 

1686: I   & 0.00 & no & 0 & -57.6 &yes \\ 

1687: II  & 0.75 & no & 0 & -34.1 &yes \\ 

1688: III & 0.75 & yes & 0 & -6.3 &yes \\ 

1689: IV  & 0.75 & yes & 4 & -9.2 &yes \\ 

1690: V   & 0.75 & yes & 5 & -5.1 &no \\ 

1691: VI  & 0.75 & yes & 1 (oc)& -1.3 &no \\

1692: \end{tabular}

1693: \caption{Studied systems for two alkane-assembled spherical

1694: solutes. $W(s_0)$ is the inter-solute pmf. If $r(z=0)\neq0$ the system

1695: is 'dewetted'. In system VI the solutes' charge is located off-center

1696: (oc) at the solute surface.}

1697: \label{tab4}

1698: \end{center}

1699: \end{table}

1700: 

1701: \clearpage 

1702: 

1703:  \begin{figure}[htb]

1704:  \begin{center}

1705: %    \epsfig{file=fig1.ps, width=7cm, angle=-90}

1706:    \caption{The particular solvation energy contributions $\Delta

1707:    G_i(R)$ with $i={\rm p,int,ne}$ in Eq.~(\ref{eq:sphere}) for one LJ

1708:    sphere with Na$^0$ parameters given in Tab.~I. The pressure term

1709:    $\Delta G_{\rm pr}$ (thin solid line) with $P=1$atm is basically

1710:    zero on this scale. The interfacial term $\Delta G_{\rm int}(R)$

1711:    (dotted line) with $\gamma_{\rm lv}=65$mJ/m$^2$ increases with

1712:    radius $R$. The LJ term $\Delta G_{\rm ne}(R)$ is given by the

1713:    dashed line. The sum of the three contribution gives the total

1714:    $\Delta G(R)$ (solid line) with a minimum at $R_{\rm min}=2.32$\AA~

1715:    for the uncharged sodium Na$^0$. The inset shows the electrostatic

1716:    contribution $\Delta G_{\rm es}(R)$ (dot-dashed line) and the total

1717:    $\Delta G(R)$ for the charged Na$^+$ with a minimum at $R_{\rm

1718:    min}=1.83$\AA. The best-fit Tolman length is $\delta_{\rm

1719:    bf}=0.79$\AA.}

1720:  \label{fig:G}

1721:  \end{center}

1722:  \end{figure}

1723: 

1724:  \begin{figure}[htb]

1725:  \begin{center}

1726: %    \epsfig{file=fig2.ps, width=7cm, angle=-90}

1727:    \caption{Mean $H(z)$ and Gaussian $K(z)$ curvature and shape

1728:    function $r(z)$ (solid lines) for ethane. The canonical SAS (dashed

1729:    line) from rolling a probe sphere with radius $r_{\rm p}=1.4$\AA~ over

1730:    the vdW surface (shaded region) is also shown. The vdW surface is

1731:    defined by the solute-solute LJ-radius $\sigma_{ss}/2=1.73$\AA.\cite{sigmanote}}

1732:  \label{fig:ethane}

1733:  \end{center}

1734:  \end{figure}

1735: 

1736:  \begin{figure}[htb]

1737:  \begin{center}

1738: %    \epsfig{file=fig3.ps, width=7cm, angle=-90}

1739:    \caption{Mean $H(z)$ and Gaussian $K(z)$ curvatures and shape

1740:    function $r(z)$ for two alkane-assembled solutes of radius

1741:    $R_0=15$\AA~ (shaded region) at a distance $s_0=8$\AA~ for systems

1742:    I-VI. The position of the charges $Z=\pm 1$ in VI are indicated by

1743:    arrows. Curvatures are not shown for the 'wet' systems V and VI. }

1744:  \label{fig:nano}

1745:  \end{center}

1746:  \end{figure}

1747: 

1748:  \begin{figure}[htb]

1749:  \begin{center}

1750: %   \epsfig{file=fig4.ps, width=7cm, angle=-90} 

1751: %   \epsfig{file=fig5.ps,width=7cm, angle=-90}

1752:    \caption{Top frame: theoretical pmfs for the systems I-III, and VI

1753:    versus the solute distance $s_0$. Bottom frame: corresponding mean

1754:    forces.}

1755:  \label{fig:pmf}

1756:  \end{center}

1757:  \end{figure}

1758: 

1759: 

1760: \begin{figure}[htb]

1761:  \begin{center}

1762: %   \epsfig{file=fig6.ps,width=7cm, angle=-90}

1763:    \caption{Mean force $\beta F$\AA~ between to nanosized solutes

1764:    versus surface-to-surface distance $s_0$. The symbols denote the MD

1765:    simulation results from Dzubiella {\it et

1766:    al.}\cite{dzubiella:jcp:2003,dzubiella:jcp:2004} for neutral

1767:    spheres with radius $R_0=12$\AA~ (circles) and $R_0=10$\AA

1768:    (squares), and oppositely charged spheres with radius $R_0=10$\AA~

1769:    and charge $Q=2$e (diamonds) and $Q=5$e (triangles). The

1770:    corresponding theoretical results using $\delta=0.9$\AA~ are shown

1771:    by solid lines; the range of the strong hydrophobic attraction

1772:    decreases with decreasing radius and increasing charge. Dotted

1773:    lines through the symbols are guides for the eye. The dashed line

1774:    is the theory for $R_0=12$\AA~ and $\delta=0.75$\AA.}

1775:  \label{fig:mf}

1776:  \end{center}

1777:  \end{figure}

1778: 

1779: 

1780:  \begin{figure}[htb]

1781:  \begin{center}

1782: %   \epsfig{file=fig7.eps, width=7cm, angle=0}

1783:    \caption{Distribution of the dielectric constant in space for two

1784:    nanosolutes with $R_0=15$\AA~ at a distance $s_0=7$\AA~ carrying a

1785:    charge $Q=5e$ (system V). The region between the spheres is

1786:    dewetted. The distribution is scaled by $\epsilon_l=78$.}

1787:  \label{eps}

1788:  \end{center}

1789:  \end{figure}

1790: 

1791: \clearpage

1792: 

1793: {\Large Figure 1}

1794:  \begin{figure}[htb]

1795:  \begin{center}

1796:     \epsfig{file=fig1.ps, width=12cm, angle=-90}

1797:  \end{center}

1798:  \end{figure}

1799: 

1800: \clearpage

1801: {\Large Figure 2}

1802:  \begin{figure}[htb]

1803:  \begin{center}

1804:     \epsfig{file=fig2.ps, width=12cm, angle=-90}

1805:  \end{center}

1806:  \end{figure}

1807: 

1808: \clearpage

1809: {\Large Figure 3}

1810:  \begin{figure}[htb]

1811:  \begin{center}

1812:     \epsfig{file=fig3.ps, width=12cm, angle=-90}

1813:  \end{center}

1814:  \end{figure}

1815: 

1816: \clearpage

1817: {\Large Figure 4}

1818:  \begin{figure}[htb]

1819:  \begin{center}

1820:     \epsfig{file=fig4a.ps, width=10cm, angle=-90} 

1821:     \epsfig{file=fig4b.ps,width=10cm, angle=-90}

1822:  \end{center}

1823:  \end{figure}

1824: 

1825: \clearpage

1826: {\Large Figure 5}

1827: \begin{figure}[htb]

1828:  \begin{center}

1829:     \epsfig{file=fig5.ps,width=12cm, angle=-90}

1830:  \end{center}

1831:  \end{figure}

1832: 

1833: \clearpage

1834: {\Large Figure 6}

1835:  \begin{figure}[htb]

1836:  \begin{center}

1837:     \epsfig{file=fig6.eps, width=12cm, angle=0}

1838:  \end{center}

1839:  \end{figure}

1840: 

1841: 

1842: \end{document}

1843: 

1844: 

1845: 

1846: 

1847: 

1848: 

1849: