0706.2424/an1.tex
1: \documentclass[pre,twocolumn,showpacs]{revtex4}
2: 
3: \usepackage{graphicx,amsmath}
4: 
5: %%%%%%%%%%%%%%%%%%%%
6: %% PRIVATE MACROS %%
7: %%%%%%%%%%%%%%%%%%%%
8: 
9: \newcommand{\figwidth}{0.9\columnwidth}
10: \newcommand{\eq}[1]{Eq.(\ref{#1})}
11: \newcommand{\fig}[1]{Fig.~\ref{#1}}
12: \newcommand{\avg}[1]{ {\langle #1 \rangle} }
13: \newcommand{\olcite}[1]{Ref.~\onlinecite{#1}}
14: \newcommand{\sect}[1]{Section~\ref{#1}}
15: \newcommand{\tab}[1]{Table~\ref{#1}}
16: 
17: \newcommand{\kb}{k_{\rm B}}
18: \newcommand{\et}{\epsilon^\star}
19: \newcommand{\mt}{\mu^\star}
20: \newcommand{\gt}{\gamma(\theta)}
21: 
22: %%%%%%%%%%%%%%%%%%%%%
23: %% DOC STARTS HERE %%
24: %%%%%%%%%%%%%%%%%%%%%
25: 
26: \begin{document}
27: 
28: \title{Anchoring effects at the isotropic-nematic interface in liquid 
29: crystals}
30: 
31: \author{R. L. C. Vink}
32: \affiliation{Institut f\"ur Theoretische Physik II, Heinrich Heine
33: Universit\"at D\"usseldorf, Universit\"atsstra{\ss}e 1, 40225
34: D\"usseldorf, Germany}
35: 
36: \date{\today}
37: 
38: \begin{abstract} The isotropic-to-nematic transition in liquid crystals is 
39: studied in $d=3$ spatial dimensions. A simulation method is proposed to 
40: measure the angle dependent interfacial tension $\gt$, with $\theta$ the 
41: anchoring angle of the nematic phase at the interface. In addition, an 
42: alternative liquid crystal model is introduced, defined on a lattice. The 
43: advantage of the lattice model is that accurate simulations of anchoring 
44: effects become possible. For the lattice model, $\gt$ depends sensitively 
45: on the nearest-neighbor pair interaction, and both stable and metastable 
46: anchoring angles can be detected. We also measure $\gt$ for an {\it 
47: off-lattice} fluid of soft rods. For soft rods, only one stable anchoring 
48: angle is found, corresponding to homogeneous alignment of the nematic 
49: director in the plane of the interface. This finding is in agreement with 
50: most theoretical predictions obtained for hard rods. \end{abstract}
51: 
52: %% 83.80.Xz 	Liquid crystals: nematic, cholesteric, smectic, discotic
53: %% 68.05.-n 	Liquid-liquid interfaces
54: %% 68.03.Cd 	Surface tension and related phenomena
55: %% 64.70.Md 	Transitions in liquid crystals
56: %% 61.30.Hn 	Surface phenomena: alignment, anchoring
57: 
58: \pacs{83.80.Xz, 68.05.-n, 68.03.Cd, 64.70.Md, 61.30.Hn}
59: 
60: \maketitle
61: 
62: \section{Introduction}
63: 
64: A fluid consisting of elongated molecules is more difficult to describe 
65: than one in which the molecules are simply spherical. In the case of 
66: elongated molecules, there are not only translational degrees of freedom, 
67: but also orientational ones. This additional complexity gives rise to many 
68: interesting effects, not found in spheres. For example, infinitely slender 
69: rods in three dimensions undergo a first-order phase transition from an 
70: isotropic to a nematic phase, provided the density is sufficiently high 
71: \cite{onsager:1949}. Both the isotropic and the nematic phase lack 
72: translational order, but in the nematic phase the rods have aligned, 
73: giving rise to long-range orientational order.
74: 
75: \begin{figure}
76: \begin{center}
77: \includegraphics[clip=,width=\columnwidth]{fig1}
78: \caption{\label{problem} Schematic representation of isotropic-nematic 
79: phase coexistence. The isotropic phase is on the left, the nematic on the 
80: right. Shown are (a) homogeneous anchoring, and (b) homeotropic 
81: anchoring.}
82: \end{center}
83: \end{figure}
84: 
85: The orientation of the nematic phase is an important quantity. In 
86: applications involving nematics at walls, the angle of the nematic 
87: director at the wall is often crucial. This angle is called the tilt or 
88: anchoring angle. Typically, there is a preferred tilt angle the nematic 
89: phase will assume, but the precise value depends sensitively on factors 
90: such as surface chemistry, particle shape, and temperature 
91: \cite{jerome:1991, patel.yokoyama:1993, barmes:021705}. Similarly, 
92: anchoring effects also occur at the isotropic-to-nematic (IN) transition. 
93: The first-order nature of that transition implies phase coexistence, 
94: whereby isotropic domains coexist with nematic domains, separated by 
95: interfaces. As \fig{problem} shows, the orientation of the nematic phase 
96: with respect to the interface becomes an additional parameter. In 
97: \fig{problem}(a), the nematic director points in the plane of the 
98: interface, which is called planar or homogeneous alignment. In 
99: \fig{problem}(b), the director is perpendicular to the interface, which is 
100: known as homeotropic alignment.
101: 
102: From symmetry considerations alone, it is clear that homogeneous and 
103: homeotropic anchoring are different. For homeotropic anchoring, there is 
104: still rotational symmetry around the interface normal; for homogeneous 
105: anchoring, no such symmetry is present. This difference is known to affect 
106: the spectrum of capillary waves. For homogeneous anchoring, the spectrum 
107: becomes anisotropic in the short wavelength limit 
108: \cite{akino.schmid.ea:2001, elgeti.schmid:2005, schmid.germano.ea:2007, 
109: wolfsheimer:061703}. In contrast, for homeotropic anchoring, the spectrum 
110: remains isotropic at all wavelengths. In other words, as this example 
111: shows, the anchoring angle affects the interfacial properties 
112: qualitatively. Given a set of particle interactions, it is therefore 
113: important to be able to predict the anchoring angle. This has lead to the 
114: concept of an angle dependent interfacial tension $\gt$, with $\theta$ the 
115: tilt or anchoring angle. Here, $\theta$ is defined as the angle between 
116: the nematic director and the plane of the IN interface. Homogeneous 
117: ($\theta=0$) and homeotropic ($\theta=90$) anchoring are most common, 
118: although $\theta$ could, in principle, be anywhere between 0 and 90 
119: degrees. In theoretical investigations, the anchoring angle is given by 
120: the angle which minimizes $\gt$. For hard rods, this is typically 
121: $\theta=0$, corresponding to homogeneous anchoring 
122: \cite{chen.noolandi:1992, velasco.mederos:2002, mcmullen:1988, 
123: physreva.42.6042, koch.harlen:1999, shundyak.roij:2001}, but the precise 
124: behavior is quite subtle. For example, the results of 
125: \olcite{chen.noolandi:1992} also suggest that homeotropic anchoring may be 
126: metastable. In addition, for very short rods, anchoring angles between 0 
127: and 90 degrees have also been reported \cite{physreva.42.6042}.
128: 
129: Unfortunately, it remains difficult to verify these theoretical findings 
130: in a computer simulation. On the one hand, efficient simulation 
131: methodology for problems of this kind is scarce. The state-of-the-art is 
132: to extract $\gt$ from the anisotropy of the pressure tensor 
133: \cite{allen:2000*b, mcdonald:2000, akino.schmid.ea:2001}, a technique 
134: which is somewhat prone to statistical error. On the other hand, the 
135: particle interactions used in many theoretical investigations are not 
136: convenient for simulations. The hard rod potential, for instance, often 
137: used in theory, gives rise to a very small interfacial tension. In order 
138: to stabilize the IN interface, simulations of hard rods require huge 
139: system sizes, implying long equilibration times and, consequently, data 
140: with considerable statistical uncertainty.
141: 
142: The purpose of this paper is to improve on this state of affairs. The 
143: primary aim is to present a simulation method capable of measuring the 
144: angle dependent interfacial tension accurately. The method is presented in 
145: \sect{sec:sim}. As it turns out, the method is general, and applies to 
146: lyotropic (density driven) systems, such as rods or platelets, as well as 
147: to thermotropic (temperature driven) lattice systems. The second aim is to 
148: introduce a new liquid crystal model, one which is easy to simulate, but 
149: which nevertheless features an IN transition with anchoring effects. The 
150: model we propose is defined on a {\it lattice}, and resembles the 
151: Lebwohl-Lasher (LL) model \cite{physreva.6.426}, but with two essential 
152: modifications. Since the model is easy to simulate, it lends itself 
153: perfectly for an investigation of anchoring effects. The liquid crystal 
154: model, and the subsequent determination of its $\gt$, are presented in 
155: \sect{res1}. Next, in \sect{res2}, we determine $\gt$ for a fluid of soft 
156: rods. These particles are already more complicated to simulate. 
157: Nevertheless, guided by the experience obtained for the simple lattice 
158: model, a meaningful interpretation of the simulation data is possible. We 
159: end with a summary and outlook in the last section.
160: 
161: \section{Simulation Method}
162: \label{sec:sim}
163: 
164: In this Section, we present our method to extract the angle dependent 
165: interfacial tension $\gt$ in liquid crystals. The use of 
166: so-called {\it order parameters} is crucial for our method: suitable order 
167: parameters are therefore discussed first. Next, we show how the order 
168: parameter distribution may be used to obtain phase coexistence properties, 
169: as well as interfacial tensions, which summarizes the key ingredients of 
170: previous work \cite{vink.schilling:2005, vink.wolfsheimer.ea:2005}. 
171: Finally, we show how this methodology can be modified, to also capture the 
172: angular dependence of the interfacial tension, by means of a simple 
173: constraint.
174: 
175: \subsection{Order parameters}
176: \label{op}
177: 
178: Since we are dealing with the IN transition in liquid crystals, a suitable 
179: order parameter is the nematic order parameter $S$, defined as the maximum 
180: eigenvalue of the orientational tensor $Q$, whose elements read as:
181: %%
182: \begin{equation}\label{eq:s}
183: Q_{\alpha\beta} = \frac{1}{2 N} \sum_{i=1}^N
184:    \left( 3 d_{i\alpha} d_{i\beta} - \delta_{\alpha\beta} \right).
185: \end{equation}
186: %%
187: Here, $d_{i\alpha}$ is the $\alpha$ component ($\alpha = x,y,z$) of the 
188: orientation $\vec{d}_i$ of molecule $i$ (normalized to unity), 
189: $\delta_{\alpha\beta}$ is the Kronecker delta, and $N$ the number of 
190: molecules. Note that $S$ is invariant under the inversion $\vec{d}_i \to 
191: -\vec{d}_i$ of single molecules, which is the characteristic symmetry of 
192: liquid crystals, and also that $S$ does not depend on the center of mass 
193: coordinates. In the isotropic phase, $S$ is close to zero. In the nematic 
194: phase, where the molecules have aligned, $S$ is close to unity. Another 
195: important quantity is the (normalized) eigenvector $\vec{n}=(n_x,n_y,n_z)$ 
196: associated with $S$. The vector $\vec{n}$ is called the director, and it 
197: corresponds to the overall preferred direction of the molecular 
198: orientations in the nematic phase. Again, the directions $\vec{n} \to 
199: -\vec{n}$ are equivalent: the convention in this work is to pick the 
200: vector with $n_z>0$.
201: 
202: The nematic order parameter, being zero in the isotropic phase and (close 
203: to) unity in the nematic phase, is a convenient quantity to detect the IN 
204: transition in liquid crystals. However, different quantities may be used 
205: as well. For example, in thermotropic (temperature driven) liquid 
206: crystals, there is also an energy difference between the isotropic (high 
207: energy) and nematic (low energy) phase. Therefore, in thermotropic 
208: systems, energy may also be used as order parameter. Similarly, in 
209: lyotropic (density driven) liquid crystals, such as studied by Onsager 
210: \cite{onsager:1949}, there is also a density difference between the 
211: isotropic (low density) and nematic (high density) phase. Therefore, in 
212: lyotropic systems, density is also a valid order parameter.
213: 
214: \subsection{Order parameter distributions}
215: \label{opd}
216: 
217: \begin{figure}
218: \begin{center}
219: \includegraphics[clip=,width=\figwidth]{fig2}
220: 
221: \caption{\label{bimodal} (a) Coexistence distribution $P(E)$ of a 
222: thermotropic liquid crystal, interacting via \eq{eq:ll} with $p=10$ and 
223: $\nu=0.5$, at the transition (inverse) temperature $\et \approx 
224: 1.188$. The simulation box dimensions were $L=15$ and $D=40$. (b) The 
225: logarithm of the same distribution.}
226: 
227: \end{center}
228: \end{figure}
229: 
230: Our method to obtain $\gt$ is based on the order parameter 
231: distribution $P(X)$, defined as the probability to observe the order 
232: parameter $X$ during the simulation. For liquid crystals, suitable choices 
233: for $X$ were given above. As is well known, at a first-order phase 
234: transition, the distribution $P(X)$ becomes double-peaked (bimodal). An 
235: example is provided in \fig{bimodal}(a), which shows the energy 
236: distribution $P(E)$ of a thermotropic liquid crystal (details are provided 
237: in \sect{res1}). In thermotropic systems, $P(E)$ becomes bimodal at the 
238: transition temperature. The precise value is determined using the 
239: ``equal-area'' rule \cite{binder.landau:1984}, whereby the temperature is 
240: tuned such that the area under both peaks is equal. Of course, in a 
241: lyotropic system, one would need to tune the chemical potential.
242: 
243: \begin{figure}
244: \begin{center}
245: \includegraphics[clip=,width=0.6\columnwidth]{fig3}
246: 
247: \caption{\label{snap} Schematic simulation snapshots in (a) the bulk 
248: isotropic phase, (b) the coexistence region, and (c) the bulk nematic 
249: phase.}
250: 
251: \end{center}
252: \end{figure}
253: 
254: From the bimodal energy distribution of \fig{bimodal}(a), bulk properties 
255: can readily be extracted. The peak at high energy, for example, yields the 
256: energy density of the isotropic phase; the peak at low energy of the 
257: nematic phase. Even more information is contained in the logarithm $W = 
258: \ln P(X)$, see \fig{bimodal}(b). Note that $W$ corresponds to {\it minus} 
259: the free energy of the system. We now observe a distinct flat region 
260: between the peaks. The origin of this flat region can be understood from 
261: simulation snapshots, shown schematically in \fig{snap}. When the system 
262: is in the high-energy peak, simulation snapshots reveal a homogeneous 
263: isotropic phase (a). In the low-energy peak, snapshots reveal a 
264: homogeneous nematic phase (c). At intermediate energy, coexistence between 
265: an isotropic and nematic domain is revealed, separated by an interface 
266: (b). Note that, due to periodic boundary conditions, two such interfaces 
267: are actually present. Provided the simulation box is large enough so as to 
268: accommodate two non-interacting interfaces, the order parameter can be 
269: varied (over a limited range) with no cost in the free energy at all, and 
270: hence a flat region in $W$.
271: 
272: The presence of a flat region in $W$ naturally allows for an estimate of 
273: the interfacial tension \cite{binder:1982}. In these cases, the height of 
274: the free energy barrier $\Delta F$ in \fig{bimodal}(b) may be associated 
275: with the free energy cost of having two interfaces in the system. Since 
276: the interfacial tension is defined as the excess free energy per unit 
277: area, one simply has $\gamma = \Delta F / 2A$, with $A$ the area of one 
278: interface. It was later recognized that \cite{grossmann.laursen:1993}, in 
279: an elongated $L \times L \times D$ simulation box, with $D \gg L$, the 
280: interfaces form perpendicular to the elongated direction, since this 
281: minimizes the total amount of interface in the system. This leads to 
282: $A=L^2$, and consequently
283: %%
284: \begin{equation}\label{eq:st}
285:   \gamma = \Delta F / 2L^2.
286: \end{equation}
287: %%
288: In previous work, the above ideas were successfully applied to the IN 
289: transition in fluids of rods \cite{vink.schilling:2005, 
290: vink.wolfsheimer.ea:2005} and platelets \cite{beek:087801}. Implementation 
291: details are also provided in these references. Of particular importance is 
292: the use of a biased sampling scheme \cite{virnau.muller:2004}, such that 
293: the simulation frequently traverses between the isotropic and the nematic 
294: phase.
295: 
296: \subsection{Measuring $\gt$}
297: 
298: Next, we describe how to modify the above methodology to also extract the 
299: angular dependence of the interfacial tension. We again use an elongated 
300: simulation box with periodic boundary conditions. The box is spanned by 
301: the vectors $L \hat{x}$, $L \hat{y}$, and $D \hat{z}$, with $D >> L$. As 
302: usual, $\hat{x} = (1,0,0)$, $\hat{y}=(0,1,0)$, and $\hat{z}=(0,0,1)$ 
303: denote standard Cartesian unit vectors. The key additional ingredient is 
304: to add a constraint to the Hamiltonian, such that the total energy of the 
305: system becomes $E = E_0 + E_c$. Here, $E_0$ is the energy of the 
306: unconstrained system. For example, in a thermotropic system, $E_0$ could 
307: be the LL potential. In a lyotropic system, it could be the 
308: potential of hard rods. The constraint energy $E_c$ should 
309: fulfill two criteria:
310: %%
311: \begin{enumerate}
312: \item In the bulk isotropic and nematic phase, the influence of the 
313: constraint must vanish. In other words, $E_c$ may not affect the bulk 
314: properties of the unconstrained system.
315: %%
316: \item In the coexistence region, where the system schematically resembles 
317: \fig{snap}(b), the director $\vec{n}$ of the nematic phase must point 
318: along some specified tilt angle $\theta$.
319: \end{enumerate}
320: %%
321: As it turns out, a suitable constraint can be written as:
322: %%
323: \begin{equation}\label{eq:con}
324:   E_c =
325:   \begin{cases}
326:   0 & |90 - \arccos |\vec{n} \cdot \vec{z}| - \theta_t| < \delta, \\
327:   \infty & {\rm otherwise}.
328:   \end{cases}
329: \end{equation}
330: %%
331: Here, $\vec{n}$ is the nematic director, defined in \sect{op}. The angles 
332: $\theta_t$ and $\delta$ are inputs of the method, and must be specified 
333: beforehand. By using the constraint, only states whose angle between 
334: director and $xy$-plane is within $\theta_t \pm \delta$ are retained, 
335: while all other states are rejected.
336: 
337: For large systems, \eq{eq:con} does not affect bulk properties, since bulk 
338: properties are insensitive to the overall orientation of the phase. In 
339: contrast, in the coexistence region, the constraint has a dramatic effect. 
340: In these cases, approximately half of the simulation box is filled with an 
341: isotropic domain, and the other half with a nematic domain, see 
342: \fig{snap}(b). Due to the constraint, the angle between the director of 
343: the nematic domain and the $xy$-plane is within $\theta_t \pm \delta$. At 
344: the same time, the use of an elongated simulation box forces the 
345: interfaces to form in the $xy$-plane as well. In other words, by setting 
346: $\theta_t$, the anchoring angle $\theta$ can be fixed. More precisely, one 
347: has $\theta = \theta_t$. Naturally, the threshold angle $\delta$ should be 
348: chosen as small as possible, while, at the same time, maintaining 
349: reasonable simulational efficiency. The optimal value is model dependent, 
350: and best obtained using trial-and-error.
351: 
352: The idea to obtain $\gt$ is now clear. We first choose a tilt angle 
353: $\theta$ of interest. Next, we measure the order parameter distribution 
354: $P(X)$, in an elongated simulation box, using the methodology of 
355: \sect{opd}. In addition, we incorporate the constraint of \eq{eq:con} in 
356: the simulations, using $\theta_t = \theta$. The peak positions in $P(X)$ 
357: should again yield the bulk properties of the coexisting isotropic and 
358: nematic phase. The barrier $\Delta F$, see \fig{bimodal}(b), can be 
359: plugged into \eq{eq:st} to obtain the interfacial tension at the chosen 
360: anchoring angle $\theta$. Since bulk properties should not be affected by 
361: the constraint, we expect the peak positions in $P(X)$ to coincide with 
362: those of an unconstrained simulation. In contrast to the unconstrained 
363: simulations, a dependence of the interfacial tension on the anchoring 
364: angle $\theta$ is anticipated. To what extent these expectations are met 
365: in actual simulations will be investigated next.
366: 
367: \section{Results: lattice simulations}
368: \label{res1}
369: 
370: \subsection{Lattice model and motivation}
371: 
372: As announced in the Introduction, we first test our method in a lattice 
373: model of a thermotropic liquid crystal. The aim is to measure 
374: $\gt$. The simulations are performed on a three-dimensional 
375: periodic lattice of size $V = L \times L \times D$, with $D \gg L$. To 
376: each lattice site $i$, a liquid crystal is attached with (normalized) 
377: orientation $\vec{d}_i$. The liquid crystals interact via the potential
378: %%
379: \begin{equation}\label{eq:ll}
380:  E =  - \epsilon
381:  \sum_{\langle i,j \rangle} \sigma_{ij} | \vec{d}_i \cdot \vec{d}_j |^p, 
382: \end{equation}
383: %%
384: where the summation is over nearest neighbors, coupling constant 
385: $\epsilon$, and exponent $p>0$. In what follows, factors of $\kb T$ are 
386: absorbed in the coupling constant $\epsilon$, with $T$ the temperature, 
387: and $\kb$ the Boltzmann constant. The ``anisotropy'' parameter is given by
388: %%
389: \begin{equation}\label{eq:ani}
390:   \sigma_{ij} = 1 + \nu \left[ (\vec{d}_i \cdot \vec{r}_{ij})^2
391:   + (\vec{d}_j \cdot \vec{r}_{ij})^2 \right],
392: \end{equation}
393: %%
394: with $\vec{r}_{ij}$ a unit vector pointing from site $i$ to $j$, and $\nu$ 
395: a parameter between $-0.5 \leq \nu \leq 0.5$. On a cubical lattice, each 
396: site has six nearest neighbors. Consequently, there are only three 
397: possible axes along which the vectors $\vec{r}_{ij}$ can be oriented.
398: 
399: For $p=2$ and $\nu=0$, this model reduces exactly to the LL model 
400: \cite{physreva.6.426}. In this case, the model exhibits a first-order IN 
401: transition, but it is very weak \cite{physrevlett.69.2803}. This makes the 
402: LL model rather inconvenient for our purposes. For example, to stabilize 
403: two interfaces so as to recover the coexistence of \fig{snap}(b), huge 
404: systems would be required. Such large-scale simulations are not the aim of 
405: the present work, and so we have chosen to modify the interactions 
406: appropriately. More precisely, we use a larger exponent in \eq{eq:ll}, 
407: namely $p=10$. The effect of this is a sharper pair interaction, meaning 
408: that neighboring molecules only lower their energy when they are closely 
409: aligned. It is known that, under such interactions, first-order phase 
410: transitions become enhanced \cite{physrevlett.89.285702, 
411: physrevlett.52.1535, physrevlett.88.047203, enter.romano.ea:2006, 
412: vink:2006*b} (even in two dimensions).
413: 
414: \begin{figure}
415: \begin{center}
416: \includegraphics[clip=,width=0.65\columnwidth]{fig4}
417:    
418: \caption{\label{ani} Illustration of the spatial anisotropy in the liquid 
419: crystal pair interaction. In a realistic system, the energies of the above 
420: two configurations will generally differ. The LL model, however, makes no 
421: distinction.}
422: 
423: \end{center}
424: \end{figure}
425: 
426: By using $p=10$ in \eq{eq:ll}, the model is expected to exhibit a strong
427: first-order IN transition. Nevertheless, this is not sufficient to study
428: anchoring because, for $\nu=0$, the interactions are spatially isotropic. 
429: In other words, the interactions do not depend on the relative positions
430: of the molecules, and so they cannot produce any anchoring effects at the
431: IN interface \cite{physreve.68.041709}. In realistic systems, the particle
432: interactions are typically anisotropic, see \fig{ani}. Shown are two
433: liquid crystal arrangements, labeled (a) and (b). Even though the
434: orientations of the molecules are identical in both cases, it is clear
435: that the energies need not be the same. In the LL model, however, for
436: which $\nu=0$, there is no distinction between the two arrangements. In
437: order to nevertheless study anchoring effects, we allow $\nu \neq 0$ in
438: \eq{eq:ani}, in which case the model does make the distinction. More
439: precisely, we have $\sigma_{ij} = 1$ for case (a), and $\sigma_{ij} =
440: 1+2\nu$ for case (b). By choosing $\nu<0$, side-side alignment is
441: energetically favored; choosing $\nu>0$ favors head-head alignment. For
442: $\nu=0$, the interactions are isotropic, in which case no particular
443: alignment is preferred.
444: 
445: \subsection{Bulk phase behavior}
446: 
447: We first determine the bulk behavior of \eq{eq:ll}. Recall that we keep 
448: the exponent fixed at $p=10$. The aim is to measure the variation of the 
449: bulk properties as a function of $\nu$. More precisely, we consider the 
450: transition inverse temperature $\et$, and the coexistence energy densities 
451: of the isotropic and nematic phase. To this end, we use the simulation 
452: methodology of \sect{opd} {\it without} the constraint of \eq{eq:con}. The 
453: energy distribution $P(E)$ is measured in a MC simulation, using a biased 
454: sampling scheme \cite{virnau.muller:2004}, at $\epsilon=0$. Histogram 
455: reweighting \cite{ferrenberg.swendsen:1988} is used to determine the 
456: value of $\epsilon$ for which the ``equal-area'' rule is obeyed, yielding 
457: $\et$. The energy densities are then read-off from the peak positions. An 
458: example distribution $P(E)$ is shown in \fig{bimodal}. The simulations are 
459: performed using single particle MC moves, whereby a random orientation is 
460: assigned to a randomly selected lattice site, accepted with the Metropolis 
461: criterion \cite{newman.barkema:1999}. Typical lattice sizes are $L = 
462: 10 - 20$ and $D = 20 - 40$. The CPU time required to obtain $P(E)$ 
463: accurately for a large system is around 48 hours.
464: 
465: \begin{figure}
466: \begin{center}
467: \includegraphics[clip=,width=\figwidth]{fig5}
468: \caption{\label{phase} Bulk properties of \eq{eq:ll} with $p=10$ as a 
469: function of $\nu$. Points are actual simulation data; curves serve to 
470: guide the eye. (a) Variation of $\et$ with $\nu$. (b) Binodal curves, 
471: showing the energy density $E/V$ of the isotropic phase (top curve), and 
472: of the nematic phase (lower curve), as a function of $\nu$. At energy 
473: densities between the curves, coexistence between isotropic and nematic 
474: domains occurs; simulation snapshots will then schematically resemble 
475: \fig{snap}(b).}
476: \end{center}
477: \end{figure}
478: 
479: The variation of $\et$ with $\nu$ is shown in \fig{phase}(a). The behavior 
480: is simply monotonic: by increasing $\nu$, $\et$ goes down. The energy 
481: densities, shown in \fig{phase}(b), reveal more interesting behavior. By 
482: decreasing $\nu$, the energy difference between the isotropic and the 
483: nematic phase becomes smaller. In other words, the transition becomes 
484: weaker. The simulations do not rule out that the curves meet when $\nu$ 
485: becomes sufficiently negative, possibly terminating in a critical point, 
486: but clearly additional efforts are required to resolve this. All that 
487: matters for the present work, however, is the fact that \fig{phase}(b) 
488: reveals a large coexistence region, over a substantial range of $\nu$ 
489: values. This confirms our expectation that, by using $p=10$ in \eq{eq:ll}, 
490: the first-order nature of the transition is enhanced significantly. This 
491: makes the model ideal to study anchoring effects, with which we proceed 
492: next.
493: 
494: \subsection{Anchoring effects for $\nu=0$}
495: 
496: As a benchmark, we consider \eq{eq:ll} with $\nu=0$, using the newly 
497: proposed method. Note that, for $\nu=0$, the model is spatially isotropic, 
498: and so we do not expect to see any anchoring effects. We MC simulate 
499: \eq{eq:ll} as before, with the constraint of \eq{eq:con} explicitly 
500: included. Two anchoring conditions are considered: homogeneous and 
501: homeotropic. Recall that the anchoring is set via $\theta_t$ in 
502: \eq{eq:con}. For the threshold angle, we use $\delta=0.75$~degrees. In 
503: addition to $P(E)$, we also measure $P(S)$, with $S$ the nematic order 
504: parameter defined in \sect{op}. For completeness, we mention that our 
505: simulations are performed using a bias on the nematic order parameter $S$, 
506: see details in \olcite{vink.wolfsheimer.ea:2005}.
507: 
508: \begin{figure}
509: \begin{center}
510: \includegraphics[clip=,width=\figwidth]{fig6}
511: 
512: \caption{\label{nu_zero} Order parameter distributions of \eq{eq:ll} with 
513: $p=10$ and $\nu=0$, obtained using box dimensions $L=10$ and $D=30$. Shown 
514: are the logarithm of the energy distribution (a), and of the nematic order 
515: parameter (b). In each case, three distributions are shown, corresponding 
516: to homogeneous and homeotropic enforced anchoring, as well as no enforced 
517: anchoring direction. The curves overlap almost perfectly, indicating the 
518: absence of any anchoring effects.}
519: 
520: \end{center}
521: \end{figure}
522: 
523: The resulting energy distributions are given in \fig{nu_zero}(a), which 
524: actually shows three distributions. Shown are the two distributions 
525: obtained using the new method, corresponding to homogeneous and 
526: homeotropic anchoring, as well as the distribution obtained without any 
527: enforced anchoring. The striking feature is that the curves overlap almost 
528: perfectly. This result is crucial because it demonstrates the consistency 
529: of the method. First of all, the insensitivity of the peak positions with 
530: respect to the enforced anchoring, confirms that {\it bulk} properties are 
531: {\it not} affected by the constraint. In addition, we find that the 
532: barrier $\Delta F$, defined in \fig{bimodal}(b), also does not depend on 
533: the anchoring condition. In other words, the interfacial tension is 
534: independent of the tilt angle, which is precisely what one expects for an 
535: isotropic potential. Additional confirmation of the consistency of the new 
536: method is provided in \fig{nu_zero}(b), which shows the corresponding 
537: distributions $\ln P(S)$ of the nematic order parameter. Again, the curves 
538: overlap almost perfectly. Note also that, for the interfacial tension, it 
539: does not matter whether one reads-off the barrier height in $\ln P(E)$ or 
540: $\ln P(S)$. As \fig{nu_zero} shows, the barriers are nearly equal (the 
541: slight variation gives an indication of the statistical uncertainty).
542: 
543: \begin{figure}
544: \begin{center}
545: \includegraphics[clip=,width=\figwidth]{fig7}
546: 
547: \caption{\label{et} Transition inverse temperature $\et$ versus anchoring 
548: angle $\theta$ for \eq{eq:ll} with $p=10$ and three values of $\nu$ as 
549: indicated. Closed symbols are actual simulation data; lines serve to 
550: guide the eye. The data were obtained using box dimensions $L=15$ and 
551: $D=40$. Note the much finer scale in (a) compared to (b) and (c).}
552: 
553: \end{center}
554: \end{figure}
555: 
556: \begin{figure}
557: \begin{center}
558: \includegraphics[clip=,width=\figwidth]{fig8}
559: 
560: \caption{\label{gam} Angle dependent interfacial tension $\gt$ 
561: of \eq{eq:ll} using $p=10$ and three values of $\nu$ as indicated. Closed 
562: symbols are actual simulation data; lines serve to guide the eye. The 
563: data were obtained using box dimensions $L=15$ and $D=40$.}
564: 
565: \end{center}
566: \end{figure}
567: 
568: We have repeated the above analysis using larger lattices, considering 
569: also tilt angles between 0 and 90 degrees. Shown in \fig{et}(a) is the 
570: transition inverse temperature $\et$ versus $\theta$. As expected, for the 
571: spatially isotropic case, $\et$ is insensitive to $\theta$, and we obtain 
572: $\et = 1.5860 \pm 0.0005$. Shown in \fig{gam} is the angle dependent 
573: interfacial tension $\gt$, as extracted from the barrier 
574: $\Delta F$ in $\ln P(S)$ and using \eq{eq:st}. Here, $\Delta F$ was taken 
575: to be the average height of the peaks, measured with respect to the flat 
576: region. As expected, the interfacial tension does not display any 
577: pronounced $\theta$ dependence (the variation stays below 2\%). For 
578: $\nu=0$, we thus find $\gamma = 0.108 \pm 0.002 \, \kb T$ per squared 
579: lattice spacing, independent of the tilt angle.
580: 
581: \subsection{Anchoring effects for $\nu=0.5$}
582: 
583: \begin{figure}
584: \begin{center}
585: \includegraphics[clip=,width=\figwidth]{fig9}
586: 
587: \caption{\label{nu_pos} Logarithm of the nematic order parameter 
588: distribution $P(S)$, at coexistence, of \eq{eq:ll} with $p=10$ and 
589: $\nu=0.5$. Shown is $\ln P(S)$ for various imposed anchoring angles 
590: $\theta$, with $\theta$ the angle between the nematic director and the 
591: plane of the IN interface. The distributions were obtained using box 
592: dimensions $L=15$ and $D=40$.}
593: 
594: \end{center}
595: \end{figure}
596: 
597: \begin{figure*}
598: \begin{center}
599: \includegraphics[clip=,width=2.1\columnwidth]{fig10}
600: 
601: \caption{\label{prof} Profiles $S(z)$ (dashed curves) and $\alpha(z)$ 
602: (solid curves) of \eq{eq:ll} with $p=10$ and $\nu=0.5$, for various 
603: imposed tilt angles $\theta$. The profiles were obtained at the transition 
604: inverse temperature $\et$, overall nematic order parameter $S=0.4$, and 
605: box dimensions $L=40$ and $D=100$.}
606: 
607: \end{center}
608: \end{figure*}
609: 
610: Having verified that the spatially isotropic case $\nu=0$ does not reveal 
611: any anchoring effects, we now consider \eq{eq:ll} using $\nu=0.5$. In this 
612: case, the model becomes anisotropic, and the tilt angle of the nematic 
613: phase with respect to the IN interface should become a relevant parameter. 
614: For a number of tilt angles, we have measured the order parameter 
615: distribution $\ln P(S)$ at coexistence; recall that the tilt angle is set 
616: via the constraint of \eq{eq:con}. Typical distributions are plotted in 
617: \fig{nu_pos}. Shown in (a) are distributions for tilt angles close to 
618: homogeneous and homeotropic anchoring; shown in (b) are distributions for 
619: two ``in-between'' tilt angles. Also shown in (a) is the ``unconstrained'' 
620: distribution, which one obtains without imposing the constraint of 
621: \eq{eq:con}. For tilt angles that are close to $\theta=0$ or $\theta=90$, 
622: the distributions behave as expected: they are bimodal, and also exhibit a 
623: pronounced flat region between the peaks. In addition, we observe that the 
624: barrier height, defined in \fig{bimodal}, depends profoundly on the 
625: imposed tilt angle. Since the barrier is related to the interfacial 
626: tension, via \eq{eq:st}, we can already see that anchoring effects are 
627: present. Interestingly, for the ``in-between'' tilt angles, bimodal 
628: distributions can also be identified, but the region between the peaks is 
629: not quite flat, see the arrow in \fig{nu_pos}(b). This suggests that, for 
630: these ``in-between'' angles, the constraint does not quite produce the IN 
631: coexistence scenario of \fig{snap}, but rather something else.
632: 
633: To verify what is going on, we have generated a number of snapshots, at 
634: {\it fixed} nematic order parameter $S=0.4$. For all distributions in 
635: \fig{nu_pos}, this value is well between the peak positions. The snapshots 
636: are generated at the transition inverse temperature $\et$ using MC 
637: simulation. After equilibration, we collect the profiles $S(z)$ and 
638: $\alpha(z)$. Here, $S(z)$ is the nematic order parameter in the $z$-th $L 
639: \times L$ slab perpendicular to the elongated $\hat{z}$-direction, and 
640: $\alpha(z)$ the angle between the director in that slab and the 
641: $xy$-plane. The profiles are shown in \fig{prof}, for the same tilt angles 
642: as studied in \fig{nu_pos}. For $\theta=0,10,80,90$~degrees, the profile 
643: $S(z)$ strikingly confirms IN phase coexistence. We can clearly identify 
644: one region where $S(z)$ is close to zero, corresponding to the isotropic 
645: phase, and another region where $S(z)$ is closer to unity, corresponding 
646: to the nematic phase. Moreover, in the nematic phase, $\alpha(z)$ is 
647: roughly constant, and the plateau value closely follows the imposed 
648: anchoring angle $\theta$. In other words, the constraint has the expected 
649: effect, namely to force the nematic phase to assume a specified tilt 
650: angle. Of course, in the isotropic phase, there is no preferred direction, 
651: and $\alpha(z)$ fluctuates randomly; one can show that the average should 
652: converge to $90 (\pi-2)/\pi \approx 32.7$ degrees. In contrast, for 
653: $\theta=45,50$~degrees, the scenario is completely different. Here, $S(z)$ 
654: is roughly constant at $S \approx 0.8$, implying a single nematic phase 
655: along the entire $\hat{z}$-direction. In addition, from the corresponding 
656: $\alpha(z)$, we see that the nematic is twisted: starting at $z=0$, 
657: $\alpha(z)$ rotates smoothly from 0 to 90 degrees, abruptly dropping back 
658: to 0 again as one passes through the periodic boundary at $z=100$. 
659: Clearly, this configuration does not reflect IN coexistence at all, but 
660: rather a twisted nematic phase with a surface defect.
661: 
662: In light of \fig{prof}, it is clear that the free energy barrier for 
663: ``in-between'' tilt angles does not reflect the interfacial tension, and 
664: consequently \eq{eq:st} does not apply. For angles that are close to 
665: homogeneous and homeotropic anchoring, however, the IN scenario of 
666: \fig{snap} is confirmed. Therefore, for these angles, we may use 
667: \eq{eq:st} to obtain the angle dependent interfacial tension 
668: $\gt$. The result is shown in \fig{gam}, which reveals several 
669: trends. First of all, in contrast to $\nu=0$, we now observe a profound 
670: variation of $\gt$ with the imposed tilt angle. The interfacial 
671: tension is smallest at $\theta=0$, corresponding to homogeneous anchoring. 
672: We therefore expect unconstrained simulations, whereby $\theta$ is not 
673: imposed but freely fluctuating, to mostly exhibit homogeneous anchoring. 
674: However, the data of \fig{gam} also suggest the presence of a shallow 
675: minimum at $\theta=90$, which corresponds to homeotropic anchoring. In 
676: other words, for $\nu=0.5$, homeotropic anchoring appears to be 
677: metastable. Since the difference in interfacial tension between 
678: homogeneous and homeotropic anchoring is small, it is not {\it a-priori} 
679: clear which anchoring condition will actually prevail in an unconstrained 
680: simulation. 
681: 
682: We have therefore performed a number of unconstrained simulations, 
683: i.e.~without \eq{eq:con}, and measured the coexistence distribution $\ln 
684: P(S)$. In addition, for each simulation, we also recorded the $n_z$ 
685: component of the director $\vec{n}$ as a function of $S$. In some cases, 
686: we found that the system selects $\theta=0$, in which case $n_z$ drops to 
687: zero once nematic order sets in, but quite often also $\theta=90$ is 
688: selected, in which case $n_z$ becomes close to unity. More precisely, 
689: using lattice dimensions $L=15$ and $D=40$, we performed 90~unconstrained 
690: simulations and found that metastable homeotropic anchoring ($\theta=90$) 
691: was selected 27~times, i.e.~in 30\% of the cases. This finding is 
692: important because it shows that the order parameter distribution of the 
693: unconstrained simulation actually reflects a ``weighted average'' of both 
694: stable and metastable anchoring. This feature is illustrated in 
695: \fig{nu_pos}(a), which also includes $\ln P(S)$ of the unconstrained 
696: simulation. As the figure shows, the free energy barrier of the 
697: unconstrained simulation is somewhere ``in-between'' homogeneous and 
698: homeotropic anchoring.
699: 
700: Another important finding is that the unconstrained simulations reveal 
701: only stable anchoring $(\theta=0)$, and metastable anchoring 
702: $(\theta=90)$, while no other anchoring angles were observed. This result 
703: is consistent with $\gt$ of \fig{gam}, which indeed features 
704: just two minima. In other words, all ``in-between'' tilt angles are 
705: unstable. Systems in which the anchoring is held artificially fixed at 
706: such unstable angles, for example via the constraint of \eq{eq:con}, will 
707: experience an additional strain. For highly unstable tilt angles, the 
708: strain is so strong, that it becomes favorable for the system to break-up 
709: the IN interfaces altogether, and form a twisted nematic. This is 
710: precisely the effect we observed for $\theta=45,50$ in \fig{prof}. 
711: However, also for tilt angles close to the stable and metastable angle, we 
712: noticed that the strain manifests itself. In this case, a small shift in 
713: the transition inverse temperature $\et$ can be detected. The effect is 
714: illustrated in \fig{et}(b), which shows $\et$ as a function of the imposed 
715: tilt angle $\theta$. For unstable tilt angles, $\et$ is systematically 
716: larger compared to the stable and metastable angles. Of course, for the 
717: stable and metastable angles, which are the experimentally relevant cases, 
718: one finds the same transition temperature again. Note also \fig{et}(a), 
719: which shows that the effect for the spatially isotropic potential $\nu=0$ 
720: does not occur, as expected.
721: 
722: \begin{figure}
723: \begin{center}
724: \includegraphics[clip=,width=\figwidth]{fig11}
725: \caption{\label{nu_neg} Coexistence distributions $\ln P(S)$ of 
726: \eq{eq:ll}, with $p=10$ and $\nu = -0.35$, using box dimensions $L=15$ and 
727: $D=40$. Shown are distributions for imposed tilt angles $\theta=0,90$ 
728: degrees, as well as the unconstrained distribution that one obtains when 
729: the tilt angle is allowed to freely fluctuate.}
730: \end{center}
731: \end{figure}
732: 
733: \subsection{Anchoring effects for $\nu=-0.35$}
734: 
735: For completeness, we also performed a number of simulations using a 
736: negative value of $\nu$ in \eq{eq:ll}, namely $\nu = -0.35$. Recall that 
737: for negative values, the side-side arrangement of \fig{ani} becomes 
738: energetically more favorable. Compared to $\nu=0.5$, one might intuitively 
739: expect that this reverses the stable and metastable anchoring angles. The 
740: angle dependent interfacial tension indeed confirms this, see \fig{gam}. 
741: We now observe that homeotropic anchoring yields the lowest interfacial 
742: tension, i.e.~is stable, while homogeneous anchoring appears to be 
743: metastable. In agreement with $\nu=0.5$, we again measure a shift in $\et$ 
744: when unstable anchoring angles are imposed, see \fig{et}(c). 
745: Interestingly, even though for $\nu=-0.35$ homeotropic anchoring yields 
746: the lowest interfacial tension, we observed that {\it unconstrained} 
747: simulations have difficulty ``finding'' this configuration. During a 
748: series of 95 unconstrained simulation runs, homeotropic anchoring was 
749: selected only 35~times, i.e.~in 37\% of the cases. In other words, the 
750: interfacial tension extracted from $\ln P(S)$ in the unconstrained 
751: simulation, rather reflects the metastable anchoring condition, see 
752: \fig{nu_neg}. The figure clearly shows that homeotropic anchoring 
753: ($\theta=90$) yields the lowest free energy barrier, while the barrier in 
754: the unconstrained distribution is significantly higher (and, in fact, 
755: rather closely resembles homogeneous anchoring). From a computational 
756: point of view, the result of \fig{nu_neg} is important because it shows 
757: that simulations do not generally find the optimal anchoring angle by 
758: themselves, even in a relatively simple lattice model.
759: 
760: \section{Results: soft rods}
761: \label{res2}
762: 
763: \begin{figure}
764: \begin{center}
765: \includegraphics[clip=,width=\figwidth]{fig12}
766: 
767: \caption{\label{rods} Anchoring properties of soft rods at the IN 
768: transition. Shown in (a) is the coexistence chemical potential $\mt$ 
769: versus $\theta$; in (b) the angle dependent interfacial tension $\gt$ 
770: versus $\theta$. Closed squares are raw simulation data; the curves serve 
771: to guide the eye. The horizontal line in (b) marks the interfacial tension 
772: of an {\it unconstrained} simulation, taken from previous work 
773: \cite{vink.wolfsheimer.ea:2005}. The simulations were performed using box 
774: dimensions $L=35$ and $D=105$.}
775: 
776: \end{center}
777: \end{figure}
778: 
779: Next, we investigate anchoring effects in an {\it off-lattice} fluid of 
780: soft rods. The rods are modeled as spherocylinders, of length $l$ and 
781: width $w$. In this section, we set $l/w=10$, and $w$ will be the unit of 
782: length. The rods interact via a repulsive pair potential, whereby rod 
783: overlap is penalized with an energy cost of $2 \, \kb T$. For more details 
784: about the model, the reader is referred to previous work 
785: \cite{vink.schilling:2005, vink.wolfsheimer.ea:2005}. The rods are 
786: simulated in the grand-canonical ensemble, i.e.~at constant temperature 
787: $T$, chemical potential $\mu$, and system volume $V$, while the number of 
788: rods in the system fluctuates. Again, we use an elongated simulation box 
789: $V = L \times L \times D$, with periodic boundary conditions. The 
790: simulations are performed using standard insertion/deletion moves 
791: \cite{frenkel.smit:2001}, and the distribution $\ln P(S)$ is recorded, 
792: defined as the probability to observe the nematic order parameter $S$, at 
793: the specified tilt angle $\theta$. As before, $\theta$ is imposed using 
794: the constraint of \eq{eq:con}. For soft rods, we noticed that a 
795: substantially larger threshold angle was needed to maintain efficiency. 
796: Here, we used $\delta=2.5$~degrees. Whereas in the thermotropic liquid 
797: crystal of \eq{eq:ll} phase coexistence is achieved by tuning the inverse 
798: temperature $\epsilon$, here that role is played by the chemical potential 
799: $\mu$. At the coexistence chemical potential $\mt$, $\ln P(S)$ 
800: becomes bimodal: coexistence properties and interfacial tensions may then 
801: be extracted from the peak positions and heights, as in \fig{bimodal}.
802: 
803: The results of the soft rod simulations are summarized in \fig{rods}. 
804: Compared to the lattice simulations of \eq{eq:ll}, the data reveal 
805: significant scatter. This indicates that soft rod simulations are 
806: demanding, and already close to the limit of what is currently tractable. 
807: Nevertheless, a number of trends emerge. According to \fig{rods}(b), $\gt$ 
808: increases monotonically with $\theta$, with the minimum occurring at 
809: $\theta=0$. In other words, soft rods favor homogeneous anchoring, and the 
810: presence of metastable angles is unlikely. The data also show that the 
811: anchoring angle is a remarkably ``soft'' degree of freedom: the free 
812: energy cost of tilting the nematic director away from the IN interface is 
813: small. This is apparent from the coexistence chemical potential, see 
814: \fig{rods}(a). Note that \fig{rods}(a) is the ``analogue'' of \fig{et} for 
815: the lattice model of \eq{eq:ll}. For the lattice model, the coexistence 
816: inverse temperature increases profoundly away from the stable and 
817: metastable angles. This increase is a manifestation of the strain 
818: introduced into the system when {\it unstable} anchoring angles are 
819: imposed. In contrast, for soft rods, the coexistence chemical potential 
820: remains nearly constant over a wide range; only when $\theta>30$ or so, 
821: does $\mt$ begin to exhibit a pronounced $\theta$ dependence. For soft 
822: rods, the anchoring angle can thus be varied around the stable direction 
823: over a fairly large range, without introducing excessive strain into the 
824: system. This result is important for {\it unconstrained} simulations, 
825: where the anchoring angle is allowed to fluctuate freely. It is unlikely 
826: that such simulations would always reveal homogeneous anchoring. Rather, 
827: we expect a range of anchoring angles $0 < \theta < 30$ to be present. The 
828: horizontal line in \fig{rods}(b) marks the interfacial tension obtained 
829: during an {\it unconstrained} simulation of soft rods 
830: \cite{vink.wolfsheimer.ea:2005}, and indeed confirms this expectation. 
831: Even though the lowest interfacial tension is obtained at $\theta=0$, the 
832: unconstrained simulation slightly exceeds this value. Instead, it rather 
833: reflects the average of $\gt$ over the range $0 < \theta < 30$~degrees. 
834: Additional confirmation is obtained from simulation snapshots of {\it 
835: unconstrained} simulations, which reveal substantial fluctuations of the 
836: anchoring angle around the homogeneous direction.
837: 
838: \section{Summary and outlook}
839: 
840: In this paper, an alternative simulation approach to study anchoring 
841: effects at the IN interface in liquid crystals was described. In 
842: particular, we focused on the angle dependent interfacial tension $\gt$, 
843: with $\theta$ the anchoring or tilt angle. The proposed method is based on 
844: recent innovations \cite{vink.schilling:2005, vink.wolfsheimer.ea:2005} 
845: where the order parameter distribution is used to extract interfacial 
846: properties. The new twist has been to introduce a constraint into the 
847: Hamiltonian, see \eq{eq:con}, which forces the nematic director to 
848: maintain a specified angle with respect to the $xy$-plane. The idea is 
849: that, by using a simulation box that is elongated in the $z$-direction, IN 
850: interfaces will form in the $xy$-plane as well. The constraint then allows 
851: the anchoring angle $\theta$ to be fixed to some value of interest.
852: 
853: At the same time, a new liquid crystal model was introduced. The model is 
854: defined on a lattice and exhibits a strong first-order IN transition. In 
855: addition, the preferred anchoring (homogeneous, homeotropic, or neutral) 
856: can be tuned by means of a single parameter. Compared to more elaborate 
857: {\it off-lattice} models, such as rods or platelets, the lattice variant 
858: is considerably easier to simulate. In particular, equilibration is less 
859: problematic, and high-quality data are readily generated. Precisely this 
860: property was exploited to obtain $\gt$ for the lattice model, using the 
861: new method. Indeed, when anchoring effects are ``switched-off'', by 
862: setting $\nu=0$ in \eq{eq:ani}, $\gt$ becomes constant. In contrast, when 
863: $\nu \neq 0$, a pronounced $\theta$ dependence is revealed. For these 
864: cases, only homogeneous and homeotropic anchoring were seen to be 
865: relevant. More precisely, for $\nu>0$, homogeneous anchoring is stable, 
866: and homeotropic anchoring metastable. For $\nu<0$, the trend is reversed. 
867: In other words, the preferred anchoring depends sensitively on the details 
868: of the interactions. Our results have also shown that, when unstable 
869: anchoring angles are imposed, the new method must be used with some care. 
870: In those cases, the simulations do not reveal IN coexistence, but rather a 
871: twisted nematic phase. Fortunately, when this happens, the method gives a 
872: clear warning, in the form of a shift in the coexistence temperature. A 
873: somewhat surprising finding was that, even for the simple lattice model, 
874: simulations do not generally find the ``optimal'' anchoring angle by 
875: themselves. Instead, when the nematic director is allowed to fluctuate 
876: freely, both stable and metastable anchoring are typically revealed.
877: 
878: We have also applied the new method to obtain $\gt$ for a fluid of soft 
879: rods. For soft rods, anchoring effects could also be identified, albeit 
880: that the data are significantly less accurate. The simulations reveal 
881: homogeneous anchoring to be stable, a finding which is consistent with 
882: most theoretical studies of hard rods. Interestingly, for soft rods, no 
883: metastable anchoring angle could be detected, which makes this model 
884: qualitatively very different from the lattice model of \eq{eq:ll}. It 
885: confirms, once again, that anchoring effects are extremely sensitive to 
886: the particle interactions.
887: 
888: For the future, investigations of the capillary wave spectrum for the 
889: lattice model of \eq{eq:ll} are planned. As mentioned in the Introduction, 
890: the spectrum is qualitatively affected by the anchoring condition 
891: \cite{akino.schmid.ea:2001, elgeti.schmid:2005, schmid.germano.ea:2007}. 
892: Since, in \eq{eq:ll}, the anchoring can be tuned using a single parameter, 
893: and since the model is easy to simulate anyhow, such investigations should 
894: be worthwhile. A sound understanding of the lattice model may well be a 
895: prerequisite before more complicated {\it off-lattice} simulations are 
896: attempted.
897: 
898: \acknowledgments
899: 
900: This work was supported by the {\it Deutsche Forschungsgemeinschaft} under 
901: the SFB-TR6 (project section D3).
902: 
903: \bibstyle{revtex} 
904: \bibliography{mainz}
905: 
906: \end{document}
907: