1: \documentclass[a4paper,english,aps,pre,showpacs,twocolumn,a4paper,floatfix]{revtex4}
2: \usepackage[T1]{fontenc}
3: \usepackage[latin1]{inputenc}
4: \usepackage{graphicx}
5: \usepackage{epsfig}
6: \bibliographystyle{apsrev}
7: \begin{document}
8: \title{Enhanced stability of tetratic phase due to clustering}
9:
10: \author{Yuri Mart\'{\i}nez-Rat\'on}
11: \email{yuri@math.uc3m.es}
12:
13: \address{Grupo Interdisciplinar de Sistemas Complejos (GISC), Departamento
14: de Matem\'aticas, Escuela
15: Polit\'ecnica Superior, Universidad Carlos III de Madrid, Avenida de la
16: Universidad 30, 28911-Legan\'es, Madrid, Spain}
17:
18: \author{Enrique Velasco}
19: \email{enrique.velasco@uam.es}
20:
21: \address{Departamento de F\'{\i}sica Te\'orica de la Materia Condensada and
22: Instituto de Ciencia
23: de Materiales Nicol\'as Cabrera, Universidad Aut\'onoma de Madrid, E-28049
24: Madrid, Spain}
25: \begin{abstract}
26: We show that the relative stability of the nematic tetratic phase with
27: respect to the usual uniaxial nematic phase can be greatly enhanced by
28: clustering effects. Two--dimensional rectangles of aspect ratio $\kappa$
29: interacting via hard interactions are considered, and the stability
30: of the two nematic phases (uniaxial and tetratic) is examined using
31: an extended scaled--particle theory applied to a polydispersed fluid
32: mixture of $n$ species. Here the $i$--th species is associated with clusters
33: of $i$ rectangles, with clusters defined as stacks of rectangles containing
34: approximately parallel rectangles, with
35: frozen internal degrees of freedom. The theory assumes an exponential cluster
36: size distribution (an assumption fully supported by Monte Carlo simulations
37: and by a simple chemical--reaction model),
38: with fixed value of the second moment. The corresponding area distribution
39: presents a shoulder, and sometimes even a well-defined peak,
40: at cluster sizes approximately corresponding to square shape (i.e. $i\simeq\kappa$),
41: meaning that square clusters have a dominant contribution to the free energy of the
42: hard--rectangle fluid. The theory predicts an enhanced region of stability of the
43: tetratic phase with respect to the standard scaled--particle theory, much closer
44: to simulation and to experimental results, demonstrating the importance of
45: clustering in this fluid.
46: \end{abstract}
47:
48:
49: \pacs{61.30.Cz, 61.30.Hn, 61.20.Gy}
50: \date{\today}
51: \maketitle
52:
53:
54: \section{Introduction}
55: The hard rectangle (HR) fluid constitutes a paradigmatic example of a two-dimensional
56: fluid exhibiting surprisingly complex phase behavior: different phase symmetries,
57: phase transitions with different order, and defect--mediated continuous transitions
58: of the Kosterlitz--Thouless type \cite{Chaikin}, all governed solely by entropy.
59: This peculiar two--dimensional system has three equilibrium fluid phases:
60: isotropic (I), where particle axes are randomly oriented,
61: uniaxial nematic (N$_{\rm{u}}$), with particles preferentially aligned along
62: a single nematic director, and tetratic nematic (N$_{\rm{t}}$), possessing
63: two equivalent
64: perpendicular nematic directors, with long particle axes oriented along one of
65: two directors with equal probability.
66:
67: In a pioneering study, Schlacken et al. \cite{Schlacken} applied scaled--particle theory
68: (SPT) on a fluid of HRs to demonstrate the stability of the N$_{\rm{t}}$ phase,
69: a phase which cannot be stabilised in a fluid of hard ellipses.
70: The intersection between the two spinodals associated with the I-N$_{\rm{t}}$ and
71: I-N$_{\rm{u}}$ transitions defines a limiting aspect ratio $\kappa=L/\sigma_0$
72: (with $L$ and $\sigma_0$ the length and width of the rectangles, respectively)
73: for the stability of the N$_{\rm t}$ phase, which is located at $\kappa\simeq 2.62$.
74: Thus, for lower values of aspect ratio, the isotropic fluid exhibits
75: a continuous transition to the N$_{\rm{t}}$ phase, whereas if $\kappa>2.62$ the I phase
76: goes directly to the uniaxial nematic phase N$_{\rm u}$.
77:
78: The study of Schlacken et al. was later supplemented by the calculation of the
79: complete phase diagram \cite{Yuri1}, also within the context of SPT. It was found,
80: in particular, that the N$_{\rm{t}}$ phase undergoes a transition to the N$_{\rm{u}}$ phase
81: at high density, the nature of which changes from second to first order at a tricritical point.
82: In addition, it was shown \cite{Yuri1} that, at the level of a particular approximation for
83: density--functional theory, the N$_{\rm{t}}$ fluid is metastable with respect to a phase
84: with (either partial or complete) spatial order; the theory was approximate in the
85: sense that it included the exact functional form of two-body correlations, but only
86: approximate higher--order correlations. On the other hand, Monte Carlo (MC) simulations
87: conducted on hard squares \cite{Frenkel} and on a HR system \cite{Donev} with $\kappa=2$
88: indicated, as expected, that the fluid exhibits quasi--long--range tetratic order, and
89: that the high--density phase consists of an aperiodic crystalline tetratic phase
90: exhibiting random tiling on a square lattice. MC simulation of a HR fluid confined
91: in a slit pore \cite{Fichthorn} show the presence of weak tetratic correlations
92: at the centre of the pore.
93:
94: In the experimental front, recent results for colloidal discs forced to stand on edge
95: by external potentials \cite{Chaikin} (and hence interacting approximately as HRs) have also
96: demonstrated that tetratic correlations play a vital role in this system. Also,
97: experiments conducted on a monolayer of vibrated granular cylinders lying on a plate
98: \cite{Narayan} have shown tetratic correlations for cylinders with aspect ratio as high
99: as $\kappa=12.6$.
100:
101: In Ref. \cite{Yuri2} strong evidence, based on MC simulations, was presented
102: for the thermodynamic stability of a N$_{\rm{t}}$ fluid when $\kappa$
103: is at least as large as $7$.
104: These simulations were supplemented by an extended SPT model that exactly incorporates
105: the second and third virial coefficients while resumming the remainder of the
106: virial series. The inclusion of the third virial coefficient increases the
107: interval in $\kappa$ where the N$_{\rm{t}}$ phase is stable, approaching the
108: simulation result. Specifically, the I-N$_{\rm{t}}$ transition line moves to lower
109: packing fractions, while the intersection point between the I-N$_{\rm{u}}$
110: and the I-N$_{\rm{t}}$ transitions shifts to $\kappa=3.23$. However, this
111: value is still lower than the value indicated by simulations.
112:
113: The properties of the HR fluid are to be contrasted with those of hard discorectangles.
114: MC simulations of this system have been conducted \cite{Frenkel2}, and the global
115: phase diagram was computed. A careful inspection of particle
116: configurations in the isotropic phase shows that there are peculiar equilibrium
117: textures, with large clusters containing particles arranged side by side, exactly
118: as in the HR fluid. These configurations are favoured by the particular shape of the
119: particles and by the reduced dimensionality. However, in contrast with the HR fluid,
120: neighbouring clusters do not exhibit strong tetratic correlations and, therefore, the
121: formation of a tetratic nematic phase is discouraged in the hard discorectangle fluid.
122:
123: In this article we address the problem of how the present theoretical understanding
124: of the HR fluid can be improved by consideration of clustering effects.
125: Our thesis is that these effects, very apparent in our own MC simulations but
126: not addressed by the theories proposed up to now,
127: are a key factor in the stabilisation of the N$_{\rm t}$ phase. Inclusion
128: of cluster formation is responsible for the enhancement of the region of N$_{\rm{t}}$
129: stability in the phase diagram. In the model proposed, clustering is approximately taken care of by
130: treating clusters as distinct species in a mixture of polydispersed rectangles.
131: The functional form for the cluster size distribution is assumed to be exponential,
132: an assumption supported by cluster statistical results based on MC simulations and
133: by a simple chemical--reaction model (see Appendix), and is introduced in the model as an input.
134: The thermodynamics of the polydispersed
135: mixture is analysed using SPT. The results indicate that clustering (assimilated
136: in the theory by means of a polydispersity parameter) stabilises the N$_{\rm{t}}$
137: phase for values of aspect ratio much higher than $\kappa =2.62$ if the
138: polydispersity is sufficiently high. Polydispersity parameters obtained from
139: simulation give support to the model.
140:
141: The article is organized as follows. In Section II we present numerical evidence
142: that the size distribution in the HR fluid is an exponentially decaying function.
143: Section III presents the main ideas of the theoretical model proposed.
144: The conclusions are drawn in Section IV.
145: Finally, details on the model and on the procedure of solution are relegated to
146: Appendices A and B, while
147: Appendix C contains a chemical--reaction model for monomer aggregation which
148: also supports the assumption of an exponential cluster size distribution.
149:
150: \section{Monte Carlo simulation of clustering}
151: \label{II}
152:
153: We started by applying standard isobaric (NPT) MC techniques on a two--dimensional fluid
154: of hard rectangles, with aspect ratios $\kappa=3$, $5$ and $7$, using $N=1400$
155: rectangles. The transition from the I phase to the N$_{\rm t}$ phase was identified
156: approximately by inspection of the tetratic order parameter
157: $q_2=\left<\cos{4\phi}\right>$, where $\phi$ is the angle between the long axis of
158: the particle and an axis fixed in space. Once the samples were equilibrated, cluster
159: statistics was applied. The criterion for pair
160: connectedness, i.e. for deciding when
161: two neighbouring rectangles can be considered to be `bonded', was based on the
162: relative angle $\phi_{12}$ between the long axes of the particles and their relative
163: centre--of--mass distance ${\bf r}_{12}$, by demanding that $\phi_{12}<\delta$ and
164: $|{\bf r}_{12}|<\epsilon$. Typical values adopted were $\delta=10^{\circ}$ and
165: $\epsilon=1.3 \sigma_0$, although the conclusions to be presented below do not seem
166: to depend qualitatively on the exact values (provided they are not too large).
167:
168: \begin{figure}
169: \epsfig{file=fig1.eps,width=3.in}
170: \caption{
171: (Colour online). Configuration of hard rectangles of aspect ratio $\kappa=7$
172: at packing fraction $\eta=0.655$, as obtained from MC simulation.
173: Clusters, defined by the pair connectedness criterion explained
174: in the text, have been coloured according to their size.}
175: \label{MC}
176: \end{figure}
177:
178: Fig. \ref{MC} presents a configuration of rectangles with $\kappa=7$ at a packing fraction
179: $\eta=0.655$. This corresponds to a tetratic phase. Identified clusters have been
180: coloured according to their size. One can see how these clusters look like large
181: ``super--rectangles'' arranged along two perpendicular directions. Therefore, the tetratic
182: structure is maintained not only for single rectangles, but also at the
183: level of clusters (``polydispersed super--rectangles''). This hierarchical feature of the
184: tetratic symmetry will give support to the theoretical model to be presented in the
185: following section. In Fig. \ref{MC1} a logarithmic histogram of the size distribution
186: (averaged over configurations) is shown; in the figure, $x_i$ is defined as the
187: fraction of clusters of size $i$ (see next section). The
188: distribution looks exponential. All the curves in the figure pertain
189: to a tetratic phase, but the same behaviour is observed also in the
190: isotropic phase (with the general trend that, for given $\kappa$,
191: the slope decreases, i.e. the size distribution function becomes
192: wider, hence the average cluster size, as density increases).
193: In the size region corresponding to square clusters
194: (i.e. clusters with aggregation number $i\sim\kappa$) it is possible to see an incipient
195: shoulder that grows as density is increased. This feature is more apparent in the
196: area distribution function, $ix_i$ (giving the fraction of area occupied by clusters of
197: a given size), which presents a shoulder and sometimes even a well--defined peak
198: (inset of Fig. \ref{MC1}), indicating that square clusters are structurally very relevant
199: and contribute very decisively to the thermodynamic properties of the fluid.
200:
201: \begin{figure}
202: \epsfig{file=fig2.eps,width=3.in}
203: \caption{
204: Size distribution function $x_i$ as obtained from MC simulation,
205: in logarithmic scale, for
206: a fluid of HRs with different values of aspect ratio and packing
207: fraction: filled circles, $\kappa=3$ and $\eta=0.635$; open
208: squares, $\kappa=7$ and $\eta=0.600$; and grey squares, $\kappa=5$ and
209: $\eta=0.657$. Straight lines are linear fits. Inset: area distribution
210: function $i x_i$ for the case $\kappa=5$ and $\eta=0.692$.}
211: \label{MC1}
212: \end{figure}
213:
214: We end this section by discussing the orientational distribution functions.
215: Let $h_m(\phi)$ be the monomer orientational distribution function,
216: giving the probability of finding a given rectangle with its long axis
217: forming an angle $\phi$ with respect to the director. Having defined
218: clusters in the fluid, we may also define an orientational distribution function
219: associated with clusters, $h_c(\phi)$, giving the probability of finding
220: an average cluster (regardless of its size) oriented with angle $\phi$.
221: In the simulations we compute $h_c(\phi)$ by averaging over all the identified
222: clusters, and then over all MC configurations. Fig. \ref{MCdist1} gives
223: the functions $h_m(\phi)$ and $h_c(\phi)$ for a state with tetratic symmetry.
224: It is interesting to note that these functions are almost identical (with
225: the same
226: symmetry and, consequently, with peaks with the same height), so that the structure
227: of tetratic ordering is maintained from the monomer level to the cluster level,
228: the functions $h_m(\phi)$ and $h_c(\phi)$ obeying a kind of ``similarity''
229: property. This property does not seem to be followed in the uniaxial nematic
230: phase.
231:
232: \begin{figure}
233: \epsfig{file=fig3.eps,width=3.in}
234: \caption{
235: Monomer $h_m(\phi)$ (solid line) and cluster $h_c(\phi)$
236: (dashed line) orientational
237: distribution functions for the case $\kappa=7$ and $\eta=0.655$, as
238: obtained from MC simulation.}
239: \label{MCdist1}
240: \end{figure}
241:
242: \section{Theoretical Model}
243: \label{III}
244:
245: The model we propose accounts for clustering effects in an approximate way.
246: The clear identification of clusters in the simulations, along with their
247: approximate rectangular shape, leads to a simple model where clusters are
248: regarded as single rectangular particles with no internal degrees of freedom. Therefore,
249: we consider a $n$--component mixture of two-dimensional hard rectangles of
250: dimensions $L$ and $\sigma_i$, with $\sigma_i=i\sigma_0$ and $i$ the
251: aggregation number. Each of these rectangles is assumed to be composed of
252: $i$ monomers in perfect contact in a side-by-side configuration.
253: The total number of monomers, $N_0$, can be written as
254: \begin{eqnarray}
255: N_0=\sum_{i=1}^n iN_i, \label{uno}
256: \end{eqnarray}
257: where $N_i$ is the number of clusters containing $i$ monomers.
258: Dividing by the volume $V$,
259: \begin{eqnarray}
260: \rho_0=\sum_{i=1}^ni\rho_i,\label{dos}
261: \end{eqnarray}
262: where $\rho_0=N_0/V$ is the total monomer density, and
263: $\rho_i=N_i/V=x_i\rho$ is the density of clusters of size $i$, with
264: $x_i=N_i/N$ their number fraction, while $N=\sum_iN_i$ is the total number
265: of clusters and $\rho=N/V$ their density. The set $\{x_i\}$, $i=1,\cdots,n$,
266: is a central quantity in our model, since it contains information about clustering
267: tendencies. We will assume $x_i$ to be an exponential with $i$, as explained
268: later. The total packing fraction of the system is
269: \begin{eqnarray}
270: \eta=\sum_{i=1}^n\rho_i a_i,
271: \end{eqnarray}
272: where $a_i=ia_0$ ($a_0=L\sigma_0$) is the area of a cluster of size $i$.
273: Using (\ref{dos}), we easily obtain $\eta=\rho_0 a_0$.
274:
275: \begin{figure}
276: \epsfig{file=fig4.eps,width=3.in}
277: \caption{
278: (Colour online).
279: Multicomponent isotropic mixture composed of different species, where
280: species, depicted in different colours, correspond to clusters of a particular size.
281: Clusters are defined as aggregates of rectangular monomers in a side-by-side configuration.
282: In this instance the monomer aspect ratio is $\kappa=5$.}
283: \label{fig1}
284: \end{figure}
285:
286: Fig. \ref{fig1} shows a schematic representation of an isotropic configuration
287: of clusters of different sizes (with monomer aspect ratio $\kappa=5$).
288: The original monomers that give rise to each cluster are indicated but note that,
289: in our model, the identity of the monomers is lost, as monomers in the same cluster
290: do not interact dynamically, always being in perfect side-by-side contact.
291:
292: The free--energy of this multicomponent mixture of hard rectangles will be modelled
293: by means of SPT applied to a fluid mixture of freely--rotating hard rectangles. The
294: orientational properties of the mixtures will be characterised by orientational
295: distribution functions $h_i(\phi)$ for each component $i$. The free--energy density
296: functional $\Phi=\beta F/V$ is written as
297: \begin{eqnarray}
298: \Phi[\{h_i\}]&=&\rho\Bigg\{\sum_{i=1}^nx_i\left[\ln
299: \left(x_i{\cal V}_i\right)+\int_0^{\pi}d\phi h_i(\phi)
300: \ln[h_i(\phi)\pi]\right]\nonumber\\
301: &-&1+\ln y+y S\left[\{h_i\}\right]\Bigg\},
302: \label{Phi}
303: \end{eqnarray}
304: where ${\cal V}_i$ is the thermal volume of $i$--sized clusters, and
305: we defined $y=\rho a_0/(1-\eta)$. Due to the head--tail symmetry of the
306: particles, the angle
307: $\phi$ can be restricted to the interval $[0,\pi]$ and the functions
308: $h_i(\phi)$ normalised accordingly. The function
309: $S\left[\{h_i\}\right]=\sum_{i,j}x_ix_j S_{ij}\left[\{h_i\}\right]$, with
310: \begin{eqnarray}
311: S_{ij}\left[\{h_i\}\right]&=&\frac{1}{2}\left(\kappa+ij\kappa^{-1}\right)
312: \langle\langle|\sin\phi_{ij}|\rangle\rangle\nonumber\\
313: &+&
314: \frac{1}{2}(i+j)\langle\langle|\cos\phi_{ij}|\rangle\rangle,
315: \label{Sij}
316: \end{eqnarray}
317: is related to $A_{ij}$,
318: the angle--averaged excluded area between clusters $i$ and $j$, as
319: $S_{ij}=(A_{ij}/a_0-i-j)/2$. The shorthand notation
320: $\langle\langle f(\phi_{ij})\rangle\rangle$ has been used
321: for the double angular average of a generic function:
322: $\langle\langle f(\phi_{ij})\rangle\rangle=\int_0^{\pi} d\phi_i\int_0^
323: {\pi} d\phi_j
324: h_i(\phi_i)h_j(\phi_j)f(\phi_{ij})$.
325: Now a bifurcation analysis of (\ref{Phi})
326: at the I-N$_{u,t}$ transition (see Appendix) allows
327: us to obtain the packing fractions of the I-N$_{u,t}$ spinodal lines as
328: \begin{eqnarray}
329: \eta^*=\left[1-\frac{4}{\pi}g_k\left(\frac{\kappa}{m_0^{(1)}}+
330: \frac{m_0^{(2)}}{\kappa m_0^{(1)}}+2(-1)^k\right)\right]^{-1},
331: \label{spino}
332: \end{eqnarray}
333: with $k=1$ for the uniaxial and $k=2$ for the tetratic nematic, while
334: $m_0^{(\alpha)}=\sum_i x_i i^{\alpha}$ ($\alpha=1,2$), are the
335: first and second moments of the cluster size distribution function.
336:
337: Based on the MC results, we adopt an exponential cluster size distribution:
338: \begin{eqnarray}
339: x_i=\frac{1-q}{1-q^n}q^{i-1},\hspace{0.4cm}i=1,\cdots,n,
340: \label{exp}
341: \end{eqnarray}
342: with $q=e^{-\lambda}$ ($\lambda>0$). The prefactor in (\ref{exp}) ensures
343: that the distribution is normalised, i.e. that $\sum_i x_i=1$.
344: The first two moments can be derived analytically:
345: \begin{eqnarray}
346: m_0^{(1)}&=&\frac{1-\left[1+n(1-q)\right]q^n}{(1-q)(1-q^n)},\label{m1}\\
347: m_0^{(2)}&=&\frac{1+q-\left[q+\left(1+n(1-q)\right)^2\right]q^n}{
348: \label{m2}
349: (1-q)^2(1-q^n)}.
350: \end{eqnarray}
351:
352: Now we use (\ref{spino}) to obtain the maximum aspect ratio which can support a
353: stable tetratic phase. This follows by imposing the condition $\eta^*_{0,t}=
354: \eta^*_{0,u}$, i.e. by searching for the intersection point of the two
355: spinodal lines I--N$_{\rm u}$ and I--N$_{\rm t}$ in the phase diagram $\eta-\kappa$.
356: Solving for the corresponding value of $\kappa$, we obtain
357: \begin{eqnarray}
358: \kappa=\frac{1}{2}\left(3m_0^{(1)}\pm\sqrt{9\left(m_0^{(1)}\right)^2-
359: m_0^{(2)}}\right).
360: \label{kap}
361: \end{eqnarray}
362: In the specific case where the number of species goes to infinity, $n\to\infty$,
363: we obtain from (\ref{m1})-(\ref{kap})
364: \begin{eqnarray}
365: \Delta=\kappa^{-1}\left[\frac{1}{2}\left(2\kappa^2-3\kappa-1\pm\sqrt{\kappa^2
366: +6\kappa+1}\right)\right]^{1/2},
367: \label{11}
368: \end{eqnarray}
369: where $\Delta=\sqrt{m_0^{(2)}/\left(m_0^{(1)}\right)^2-1}=\sqrt{q}$,
370: a measure of polydispersity, is the relative mean square deviation.
371: The two functions $\Delta(\kappa)$, corresponding to the two signs in
372: (\ref{11}), are plotted in Fig. \ref{dos0} (note that one of
373: the branches, the one with minus sign, can be only calculated for
374: $\kappa\geq \kappa^*=(3+\sqrt{5})/2$). The two lines define a region (shaded in the
375: figure) where the tetratic phase may be stabilised. Thus we see how, as the value of
376: polydispersity $\Delta$ is increased, the maximum value of $\kappa$ for which the
377: tetratic phase ceases to exist increases, which means that polydispersity
378: enhances the formation of tetratic ordering.
379:
380: \begin{figure}
381: \epsfig{file=fig5.eps,width=3.in}
382: \caption{
383: The functions $\Delta(\kappa)$ in terms of inverse aspect ratio
384: $\kappa^{-1}$. Symbols indicate the nematic phase which is
385: stable in the region in
386: question. The region of tetratic stability is shaded.
387: Open squares indicate values of polydispersity, for values of
388: aspect ratio $\kappa=3$, $5$ and $7$, at the I--N$_{\rm t}$ transition,
389: as obtained from MC simulation; error bars correspond to uncertainty
390: in $\Delta$ originating from uncertainty in location of spinodal.}
391: \label{dos0}
392: \end{figure}
393:
394: Note that the values of $\Delta$ at the upper boundary of the tetratic
395: region are very high, which means that there should be a large number of clusters
396: with their long side $i\sigma_0\gg L$, while their width is $L$. This causes
397: the effective average cluster aspect ratio $i\sigma_0/L$ to be very high,
398: which induces formation of {\it uniaxial} nematic order and thus the N$_{\rm t}$
399: phase is destabilised in favour of the N$_{\rm u}$ phase. The presence of such
400: big clusters is not observed in the simulations, which means that a more
401: sophisticated model should somehow include additional entropy terms that
402: discourage the formation of big clusters; for the sake of simplicity, we have
403: kept the ingredients of our model to a minimum and avoided any such additional
404: complications.
405:
406: Finally, in Fig. \ref{dos0} values of polydispersity $\Delta$ calculated from
407: the simulated cluster size distribution function have been indicated by symbols.
408: These values correspond to the estimated spinodal line of the I--N$_{\rm t}$
409: transition for the cases $\kappa=3$, $5$ and $7$. Since these estimations
410: are very rough, the uncertainty in packing fraction at the spinodals
411: goes over to the value of polydispersity (which naturally depends on
412: $\kappa$ and $\eta$), which is represented in each case
413: by error bars. We can see that in two of the cases the symbols are
414: well inside the tetratic stability region calculated from the theory.
415: In the case $\kappa=7$, where uncertainty is larger,
416: the MC estimate of $\Delta$ lies outside (but close to) this region.
417:
418: \section{Conclusions}
419:
420: Because of the reduced dimensionality, fluids of two--dimensional
421: hard anisotropic particles
422: exhibit strong clustering effects: particles have less freedom to orient in
423: space, which fosters configurations where neighbouring particles lie parallel
424: to each other. However, the impact of this on the onset of new macroscopic
425: symmetries depends very sensitively on the particular geometry of the
426: particles. Thus, in fluids of rectangles, neighbour clusters have a strong
427: tendency to adopt orthogonal relative configurations, since these clusters
428: are almost perfect {\it big} rectangles made of several, almost parallel,
429: monomers. These strong tetratic correlations are capable of generating full
430: macroscopic tetratic order and a thermodynamically stable tetratic phase.
431: Therefore, clustering effects are crucial to understand phase behaviour
432: in the HR fluid (and possibly also in the HDR and related fluids), but
433: simple theories at the level of two--body {\it monomer}
434: correlations (Onsager, SPT, etc.) cannot account for these effects.
435:
436: In this work we have presented a simple
437: theory that incorporates clustering in terms of cluster polydispersity, where
438: clusters are considered to be inert particles with no internal degrees of
439: freedom. This assumption may be accurate provided the cluster lifetime
440: (i.e. the average time it takes for a cluster to disappear since it was
441: formed) is longer than the typical cluster diffusion rates in the fluid.
442: Validation of this condition will have to wait for molecular dynamics simulations
443: of the hard--rectangle fluid. Once the fluid is modelled in terms of a
444: multicomponent mixture, one of the available theories for mesophase
445: formation can be used. We have used SPT and have examined the consequence
446: of polydispersity in the phase diagram. As expected, polydispersity enhances
447: the stability of the tetratic phase. Due to the limitations of our model
448: (e.g. the cluster size distribution has to be imposed from outside and
449: does not result from the theory), we cannot make any quantitative
450: comparison with available simulation and experimental results. However,
451: the model can qualitatively explain the formation of tetratic order for
452: rather high values of aspect ratio, as shown by simulation and experiment.
453:
454: \acknowledgments
455:
456: Y.M.-R. gratefully acknowledges financial support from Ministerio
457: de Educaci\'on y Ciencia (Spain) under a Ram\'on y Cajal research
458: contract and the MOSAICO grant. This work has been partly financed by
459: grants Nos. FIS2005-05243-C02-01 and FIS2007-65869-C03-01, also from
460: Ministerio de Educaci\'on y Ciencia, and S-0505/ESP-0299 from
461: Comunidad Aut\'onoma de Madrid (Spain).
462:
463: \section{Appendix}
464:
465: In this appendix we provide additional details and further information
466: on the consequences of the model. It contains two sections. In Section A,
467: details on the set of non--linear equations that have to be solved to obtain
468: the equilibrium properties of the HR fluid are provided. Also, the
469: bifurcation analysis of the I--N$_{\rm{u,t}}$ transitions is presented.
470: Section B is devoted to discussing the nature of the different phase transitions,
471: together with the behaviour of the distribution functions and to a
472: comparison with simulations. In Section C, a simple chemical model of aggregation
473: is discussed.
474:
475: \subsection{Minimisation of free energy and bifurcation analysis}
476:
477: Using Fourier series to represent the orientational distribution functions,
478: \begin{eqnarray}
479: h_i(\phi)=\frac{1}{\pi}\sum_{k\geq 0}h_k^{(i)}\cos(2k\phi),
480: \end{eqnarray}
481: with $h_0^{(i)}=1$ $\forall i$, together with Eqn. (\ref{Sij}), we find
482: \begin{eqnarray}
483: S_{ij}=\frac{1}{\pi}\sum_{k\geq 0}\left[\kappa+\frac{ij}{\kappa}+
484: (-1)^k(i+j)\right]g_kh_k^{(i)}h_k^{(j)},
485: \end{eqnarray}
486: where $g_k=-(1+\delta_{k0})/2(4k^2-1)$. Defining
487: \begin{eqnarray}
488: m_k^{(\alpha)}=\sum_i x_i i^{\alpha}h_k^{(i)},\quad \alpha=0,1,
489: \end{eqnarray}
490: we obtain
491: \begin{eqnarray}
492: \sum_{ij}x_ix_jS_{ij}=\frac{\kappa}{\pi}\sum_k
493: g_k s_k^2,\hspace{0.4cm}
494: s_k=m_k^{(0)}+(-1)^k\frac{m_k^{(1)}}{\kappa}.\nonumber \\
495: \end{eqnarray}
496: Note that $m_0^{(0)}=\sum_i x_i=1$ while $m_0^{(1)}=\sum_i x_i i$ is the first
497: moment of the discrete cluster size distribution function $\{x_i\}$.
498: Using this notation, the free-energy per particle $\varphi=\Phi/\rho_0$ can be written
499: \begin{eqnarray}
500: &\varphi&=\ln\left(\frac{y_0}{m_0^{(1)}}\right)-1+
501: \sum_{i=1}^nx_i\Bigg\{\ln\left(x_i{\cal V}_i\right)\nonumber\\
502: &+&\int_0^{\pi}
503: d\phi h_i(\phi)\ln\left[h_i(\phi)\pi\right]\Bigg\}+
504: \frac{y_0\kappa}{\pi m_0^{(1)}}\sum_kg_ks_k^2,
505: \label{varphi}
506: \end{eqnarray}
507: with $y_0=\eta/(1-\eta)$. The functional minimization of (\ref{varphi})
508: with respect to $h_i(\phi)$ gives a set of self--consistent non-linear
509: equations which, after some algebraic manipulations, can be transformed into
510: a set of equations for the new variables $s_k$:
511: \begin{eqnarray}
512: &&s_k=2\sum_{i=1}^nx_i\left[1+\frac{(-1)^k}{\kappa}i\right]
513: Q_k^{(i)},
514: \label{nolinear}\\
515: &&Q_k^{(i)}=\int_0^{\pi}d\phi\cos(2k\phi)h_i(\phi),
516: \end{eqnarray}
517: where the normalized orientational distribution functions are
518: \begin{eqnarray}
519: &&h_i(\phi)=\frac{\displaystyle e^{-\Lambda_i(\phi)}}
520: {\displaystyle\int_0^{\pi}d\phi e^{-\Lambda_i(\phi)}}, \\
521: &&\Lambda_i(\phi)=\frac{4y_0\kappa}{\pi m_0^{(1)}}\sum_{k\geq 1}
522: s_k\left[1+\frac{(-1)^k}{\kappa}i\right]g_k\cos(2k\phi).\nonumber\\
523: \label{nolinear1}
524: \end{eqnarray}
525: The linearization of (\ref{nolinear}) with respect to $s_k$ ($k\geq 1$)
526: allow us to obtain the expression (\ref{spino})
527: for the packing fractions at the I-N$_{u,t}$ spinodal lines.
528:
529: \subsection{Phase transitions and distribution functions}
530:
531: In this subsection we analyse the free energy branches of the model
532: in order to understand the nature of the different phase transitions.
533: The cluster distribution function $x_i$ of the mixture is assumed
534: to be exponential, and the width of the distribution is fixed via
535: the polydispersity parameter $q$ (or $\Delta$). Eqns.
536: (\ref{nolinear})--(\ref{nolinear1}) are solved for different values of $\eta$
537: to find all metastable and stable phases, either I, N$_{\rm{u}}$ or
538: N$_{\rm{t}}$ phases. In Fig. \ref{monton} the free-energy branches as a function
539: of $\eta^{-1}$ for different values of $q$ (and hence for different
540: polydispersities) are shown.
541: As can be seen, the N$_{\rm t}$ phase begins to be stable from
542: $q\approx 0.30$. Also, it is clear that, for $q=0.25$ and $0.35$,
543: the I-N$_{\rm{u}}$ or N$_{\rm{t}}$-N$_{\rm{u}}$ transitions are
544: of first order (free--energy branches cross with different slopes).
545: The coexistence values of $\eta$ cannot be determined from
546: the standard double--tangent construction, since the present system is
547: polydisperse. The usual procedure then is
548: to fix the distribution function $x_i^{(0)}$ and packing fraction $\eta^{(0)}$
549: for the parent phase (I or N$_{\rm{u,t}}$ phases) and find the cloud
550: and shadow curves. We have not implemented this procedure here.
551: However, for those values of $q$ for which the transitions are continuous
552: (i.e. $q=0.5$ and $0.65$), the present procedure adequately determines
553: the transition densities. A similar situation occurs for the cases
554: $\kappa=5$ and $\kappa=7$ (not shown).
555:
556: \begin{figure}
557: \epsfig{file=fig6a.eps,width=1.6in}
558: \epsfig{file=fig6b.eps,width=1.6in}
559: \epsfig{file=fig6c.eps,width=1.55in}
560: \hspace*{0.1cm}
561: \epsfig{file=fig6d.eps,width=1.55in}
562: \caption{
563: Free energy per particle $\Phi/\rho$ vs. inverse
564: packing fraction $\eta^{-1}$ from SPT results for the
565: multicomponent HR fluid of monomer
566: aspect ratio $\kappa=3$ and with different values of polydispersity:
567: (a) $q=0.25$, (b) $0.35$, (c) $0.50$ and (d) $0.65$.
568: Continuous curves: tetratic phase; dashed curves: uniaxial nematic
569: phase; dotted curves: isotropic phase. In (c) and (d) symbols indicate
570: bifurcation points from the isotropic to the tetratic (filled circles)
571: and uniaxial nematic (open circles) phases. In (a) and (b) a straight
572: line in $\eta^{-1}$ has been subtracted to better visualise the
573: curves.}
574: \label{monton}
575: \end{figure}
576:
577: It is also interesting to look at the orientational distribution functions
578: of monomers and clusters. From the corresponding functions for clusters
579: of size $i$, i.e. $h_i(\phi)$, it is easy to define a cluster
580: orientational distribution function $h_c(\phi)$ as
581: \begin{eqnarray}
582: h_c(\phi)=\sum_{i=1}^n x_i h_i(\phi).
583: \end{eqnarray}
584: From this, order parameters of the multicomponent mixture can also be
585: defined:
586: \begin{eqnarray}
587: Q^{(k)}=\sum_i x_i Q_i^{(k)},\quad k=1,2.
588: \end{eqnarray}
589: In the case of monomers the situation is a bit more complicated, since
590: in our model we have lost track of monomers as distinct entities. However,
591: we can simply count the number of monomers pointing along some angle $\phi$
592: from the set $h_i(\phi)$ and then divide by the average number of clusters.
593: Here we have to bear in mind that clusters of size $i$ (having $i$ monomers)
594: with $i<\kappa$ (and $\kappa$ an integer)
595: have a long axis in a direction perpendicular to that of
596: clusters with $i>\kappa$; therefore we write:
597: \begin{eqnarray}
598: h_m(\phi)=\frac{1}{m_0^{(1)}}\left[\sum_{i=1}^{n_0} x_i i h_i(\phi)+
599: \sum_{i=n_0+1}^n x_i i h_i(\phi+\pi/2)\right],\nonumber\\
600: \end{eqnarray}
601: with $n_0=[\kappa]$ (note that in the case where $\kappa$ is not an integer
602: this division has to be done also).
603: Thus the order parameters of the monomers can be calculated as
604: \begin{eqnarray}
605: \langle \cos(2\phi)\rangle_m&=&\frac{1}{m_0^{(1)}}\Bigg|
606: \sum_{i=1}^{n_0}x_i iQ_i^{(1)}-
607: \sum_{i=n_0+1}^nx_i iQ_i^{(1)}\Bigg|,\nonumber \\ \\
608: \langle \cos(4\phi)\rangle_m&=&
609: \frac{1}{m_0^{(1)}}\sum_{i=1}^{n_0}x_i iQ_i^{(2)}.
610: \end{eqnarray}
611: Even for values of aspect ratio for which there exists
612: a region of tetratic stability, the N$_{\rm{u}}$ is always the more
613: stable phase for high values of packing fraction.
614: It is interesting to note that, in this situation, the cluster and monomer
615: distribution functions $h_{\rm{c,m}}(\phi)$ usually have a secondary
616: peak at $\phi=\pi/2$ corresponding to tetratic ordering (this feature
617: is also present in the simple SPT for the one--component HR fluid). Fig.
618: \ref{monton4} shows these distributions for the case $\kappa=5$ and
619: $q=0.65$. Note that, according to Fig. \ref{dos0}, there exists a
620: stable tetratic phase in this case, with a value of packing fraction
621: at the N$_{\rm t}$--N$_{\rm u}$ transition of $\eta^*=0.839$. In the
622: figure, the packing fraction chosen is $\eta=0.85 > \eta^*$ and,
623: therefore, the stable phase has an orientational
624: distribution function pertaining to a N$_{\rm{u}}$.
625: However, both the cluster and the monomer orientational distribution
626: functions of N$_{\rm{u}}$ have secondary peaks.
627: We have also
628: plotted in Fig. \ref{monton4} the distribution functions of a metastable
629: N$_{\rm{t}}$ at the same value of $\eta$.
630: Finally, it is also interesting that,
631: although the distribution functions $h_m(\phi)$ and $h_c(\phi)$ in the
632: tetratic region are always tetratic--like (i.e. all maxima have the same
633: height, as it should be by construction), they do not coincide in the uniaxial
634: nematic phase and, consequently, the similarity property at work in the tetratic
635: phase is not obeyed for the uniaxial nematic (see Section \ref{II}).
636:
637: \begin{figure}
638: \epsfig{file=fig7a.eps,width=1.6in}
639: \epsfig{file=fig7b.eps,width=1.6in}
640: \caption{
641: Cluster (a) and monomer (b) orientational distribution functions
642: for $\kappa=5$ and $q=0.65$. The continuous line corresponds to
643: a stable N$_{\rm u}$ phase with $\eta=0.85$, whereas the discontinuous line is for
644: the metastable N$_{\rm t}$ phase at the same value of $\eta$.}
645: \label{monton4}
646: \end{figure}
647:
648: \subsection{Chemical reaction model}
649:
650: Aggregation phenomena in dilute fluids (e.g. micelle aggregation)
651: are very often described in terms of a chemical reaction model.
652: An exponentially decaying size distribution immediately emerges
653: from these models. Here we exploit the idea and carry it further
654: using our density--functional approximation. This model assumes
655: that the lifetime of a cluster is sufficiently long that it can
656: be defined as a distinct `chemical' species.
657:
658: One assumes a chemical reaction of the type $C_l+C_1\rightleftharpoons
659: C_{l+1}$, with $C_l$ denoting a `chemical' species (i.e. a cluster)
660: containing $l$ monomers. Chemical equilibrium between clusters and
661: monomers then implies the relation
662: \begin{eqnarray}
663: \mu_i=i\mu_1,\quad i=2,\cdots,n. \label{cuatro}
664: \end{eqnarray}
665: Now the chemical potential of the $i$--th species can be calculated from
666: our density--functional theory as
667: \begin{eqnarray}
668: \beta \mu_i=\frac{\partial\Phi}{\partial\rho_i},
669: \end{eqnarray}
670: which results in
671: \begin{eqnarray}
672: &&\beta \mu_i=\ln\left( x_i{\cal V}_i\right) +\int_0^{\pi}d\phi h_i(\phi)
673: \ln[h_i(\phi)\pi]+\ln y\nonumber\\
674: &&+y i +2y\sum_j x_j S_{ij}+y^2S i.
675: \label{chepo}
676: \end{eqnarray}
677: The $n-1$ Eqns. (\ref{cuatro}), together with the condition (\ref{dos}),
678: are a set of $n$ equations with $n$ unknowns ($x_2,\cdots, x_n$ and $\rho$)
679: which allow to find the equilibrium configuration of the fluid. Due to the
680: simplicity of the model an analytical solution can be found.
681:
682: \begin{figure}
683: \epsfig{file=fig8.eps,width=3.in}
684: \caption{
685: Packing fractions $\eta$ of the I--N$_{\rm t}$ (continuous line)
686: and I--N$_{\rm u}$ (dashed line) transitions versus aspect ratio $\kappa$.}
687: \label{figa}
688: \end{figure}
689: Since, for the isotropic phase, we have
690: \begin{eqnarray}
691: \sum_j x_j\left(iS_{1j}-S_{ij}\right)=\frac{(i-1)}{\pi}(\kappa+m_0^{(1)}),
692: \end{eqnarray}
693: Eqns. (\ref{cuatro}), together with (\ref{chepo}), give $x_i=x_1 q^{i-1}$ with
694: \begin{eqnarray}
695: q=x_1\frac{y_0}{m_0^{(1)}}\exp\left[\frac{2y_0}{\pi}\left(\frac{\kappa}{m_0^{(1)}}+1
696: \right)\right].
697: \label{final}
698: \end{eqnarray}
699: Here we assumed that ${\cal V}_i={\cal V}_1^i$ (consistent with the
700: absence of internal degrees of freedom in the clusters and also with the
701: assumption that all clusters have the same mass).
702: Now, since $\sum_i x_i=1$, we find,
703: for a fluid with an infinite number of species,
704: $1=x_1/(1-q)$ which, together with (\ref{final}), give
705: \begin{eqnarray}
706: x_1=\left\{1+\frac{y_0}{m_0^{(1)}}\exp\left[\frac{2y_0}{\pi}
707: \left(\frac{\kappa}{m_0^{(1)}}+1
708: \right)\right]\right\}^{-1}.
709: \end{eqnarray}
710: Also, the first moment can be calculated self--consistently as
711: \begin{eqnarray}
712: &&m_0^{(1)}=x_1\sum_{i=1}^{\infty}i q^{i-1}=\frac{x_1}{(1-q)^2}\nonumber\\&&=
713: 1+\frac{y_0}{m_0^{(1)}}\exp\left[\frac{2y_0}{\pi}\left(\frac{\kappa}{m_0^{(1)}}+1
714: \label{mom1}
715: \right)\right].
716: \end{eqnarray}
717: This solution means that
718: $x_i=\left(1-1/m_0^{(1)}\right)^{i-1}/m_0^{(1)}$.
719: While the first moment $m_0^{(1)}$ is the solution of
720: Eqn. (\ref{mom1}), the second moment results in
721: \begin{eqnarray}
722: m_0^{(2)}=x_1\sum_{i=1}^{\infty}i^2q^{i-1}=x_1\frac{1+q}{(1-q)^3}=
723: m_0^{(1)}(2m_0^{(1)}-1).\nonumber\\
724: \end{eqnarray}
725: The polydispersity coefficient, defined as $\Delta=\sqrt{m_0^{(2)}/
726: \left(m_0^{(1)}\right)^2-1}$, turns out to be
727: $\Delta=\sqrt{1-1/m_0^{(1)}}$.
728:
729: Now the I-N$_u$ and I-N$_t$ spinodals can be calculated by solving
730: Eqn. (\ref{mom1}), together with the value of $y_0$ obtained from
731: Eq. (\ref{spino}).
732: Fig. \ref{figa} contains the functions $\eta_{u,t}(\kappa)$
733: obtained as the solutions of (\ref{mom1}) and (\ref{spino}). As
734: can be seen the uniaxial nematic is more stable than the tetratic up
735: to $\kappa^*\approx 41.65$. This results from the peculiar behaviour of
736: the area distribution function $i x_i$: its zeroth--order moment is always
737: greater than $\kappa$, while it is larger in the tetratic phase when
738: $\kappa<\kappa^*$ [Fig. \ref{cuatro0}(a)]; this behaviour is inverted for
739: $\kappa>\kappa^*$, and the moment becomes larger for the uniaxial nematic
740: phase, as can be seen in Fig. \ref{cuatro0}(b). Again this peculiar
741: behaviour is due to the relatively
742: high proportion of very big clusters, with $i\sigma_0>L$,
743: which stabilise the N$_{\rm u}$ phase against the N$_{\rm t}$ phase. This
744: behaviour is not observed in simulations, because the formation of very big
745: clusters is penalised by fluctuations. The model has the value that
746: an exponential cluster distribution function is predicted.
747:
748: \begin{figure}
749: \epsfig{file=fig9a.eps,width=1.6in}
750: \epsfig{file=fig9b.eps,width=1.6in}
751: \caption{
752: Area distribution functions for tetratic N$_{\rm t}$ and
753: uniaxial N$_{\rm u}$ nematic phases. (a) $\kappa=10$ and (b)
754: $\kappa=100$.}
755: \label{cuatro0}
756: \end{figure}
757: \begin{thebibliography}{13}
758: \bibitem{Chaikin}K. Zhao, C. Harrison, D. Huse, W. B. Russel, and P. M.
759: Chaikin, Phys. Rev. E {\bf 76}, 040401(R) (2007).
760: \bibitem{Schlacken} H. Schlacken, H.-J. Mogel, and P. Schiller,
761: Mol. Phys. {\bf 93}, 777 (1998).
762: \bibitem{Yuri1}Y. Mart\'{\i}nez-Rat\'on, E. Velasco, and L. Mederos,
763: J. Chem. Phys. {\bf 122}, 064903 (2005).
764: \bibitem{Frenkel} K. W. Wojciechowski and D. Frenkel, Comp. Mat. Sci.
765: Tech. {\bf 10}, 235 (2004).
766: \bibitem{Donev}A. Donev, J. Burton, F. H. Stillinger, and S. Torquato,
767: Phys. Rev. B {\bf 73}, 054109 (2006).
768: \bibitem{Fichthorn} D. A. Triplett and K. A. Fichthorn, Phys. Rev. E
769: {\bf 77}, 011707 (2008).
770: \bibitem{Narayan}V. Narayan, N. Menon, and S. Ramaswamy, J. Stat. Mech.
771: P01005 (2006).
772: \bibitem{Yuri2}Y. Mart\'{\i}nez-Rat\'on, E. Velasco, and L. Mederos,
773: J. Chem. Phys. {\bf 125}, 014501 (2006).
774: \bibitem{Frenkel2} M. A. Bates and D. Frenkel, J. Chem. Phys. {\bf 112},
775: 10034 (2000).
776: \end{thebibliography}
777:
778:
779: \end{document}
780: