• Aucun résultat trouvé

Fluctuations in the flat and collapsed phases of polymerized membranes

N/A
N/A
Protected

Academic year: 2021

Partager "Fluctuations in the flat and collapsed phases of polymerized membranes"

Copied!
21
0
0

Texte intégral

(1)

HAL Id: jpa-00212561

https://hal.archives-ouvertes.fr/jpa-00212561

Submitted on 1 Jan 1990

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Fluctuations in the flat and collapsed phases of polymerized membranes

Farid F. Abraham, David R. Nelson

To cite this version:

Farid F. Abraham, David R. Nelson. Fluctuations in the flat and collapsed phases of polymerized membranes. Journal de Physique, 1990, 51 (23), pp.2653-2672. �10.1051/jphys:0199000510230265300�.

�jpa-00212561�

(2)

Fluctuations in the flat and collapsed phases of polymerized

membranes

Farid F. Abraham (1) and David R. Nelson (2)

(1) IBM Research Division, Almaden Research Center, 650 Harry Road, San Jose, California 95120-6099, U.S.A.

(2) Lyman Laboratory of Physics, Harvard University, Cambridge, Massachusetts 02138, U.S.A.

(Received 22 June 1990, accepted 22 August 1990)

Abstract.

2014

Fluctuations in polymerized membranes are explored via extensive molecular

dynamics simulations of simplified « tethered surface » models. The entropic rigidity associated

with repulsive second-nearest-neighbor interactions leads to a flattening of « phantom surfaces ».

An attractive interaction in the presence of distant self-avoidance leads to a collapsed membrane

with fractal dimension three at sufficiently low temperatures. When the attractive interaction is turned off, the surface returns to the flat phase found in earlier simulations. A study of density profiles and hexatic internal order allows a simple physical interpretation of results for the structure function of oriented membranes.

Classification

Physics Abstracts

68.1OC 82.70K - 87.20C

1. Introduction.

Polymerized networks appear naturally in a biological context (1), and can be made artificially by, for example, modifying traditional methods of polymer synthesis (2), or by polymerizing amphiphillic bilayers and monolayers (3). The statistical mechanics of these

« tethered surfaces » (4) has recently attracted intense theoretical interest (5), in part because, unlike conventional linear polymers, they are expected to exhibit a low temperature flat phase with an infinité persistence length. The flat phase arises because the resistance to in-

plane shear deformations leads to an anomalous stiffening of the surface in the presence of thermal fluctuations (6). Although the first simulations of such tethered surfaces were

interpreted in terms of a high temperature crumpled phase (4), simulations of much larger

surfaces with a very similar potential revealed that these objects were in fact flat (7, 8) with

very large fluctuations in the direction parallel to the average surface normal (see Fig. 1).

In this paper we present the results of extensive computer simulations of the flat phase (9).

We discuss the entropic origin of the bending rigidity introduced by distant neighbor interactions, and show explicitly that second neighbor repulsion alone is sufficient to produce

a flat phase in an initially crumpled « phantom » membrane for sufficiently short tethered

lengths. By introducing an attractive distant interaction, one can produce a compact or collapsed self-avoiding tethered surface. In contrast to flat membranes, whose fractal dimension is two, and crumpled membranes, whose fractal dimension is expected to be about

2.5 (4), the fractal dimension of this compact object is three. With the compact membrane as

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:0199000510230265300

(3)

Fig. 1.

-

Configurations of a self-avoiding tethered membrane of 4 219 particles (L

=

75). Time is

measured in molecular dynamics time-steps. Tethering bonds are drawn between neighboring

monomers, whose hard core size is not shown.

an initial condition, we turn off the attractive part of the interaction and show that the system relaxes to the same strongly fluctuating flat phase found with a « stretched » initial condition in reference [7].

We also study the internal structure of membranes in the flat phase. Density profiles perpendicular to the plane of the surface are characterized by a single exponent C, which describes the divergence of the membrane thickness as the system size tends to infinity.

Density profiles in the plane, however, are characterized by at least two different diverging length scales : an overall membrane diameter, and the width of the density distribution at the membrane edge. Both in-plane phonon fluctuations and locally transverse roughing contribute

to the apparent width of the density profile at the edge of the membrane. Although these

fluctuations destroy extended translational order in monomers with the bonding topology of a triangular lattice, we find that a small amount of long-range bond orientational order is

preserved, consistent with a prediction of Aronovitz and Lubensky (10).

This analysis of surface fluctuations in real space allows us to better understand the remarkable signature of tethered surfaces in the reciprocal space (9). A simple theoretical

approximation to the equilibrium structure function is derived in detail, and provides a good

fit to the largest tethered surfaces simulated to date.

(4)

Before proceeding further, we briefly summarize theoretical expectations (6, 10) for the

flat phase (11). In the flat phase, in-plane phonon displacements u (xl, x2 ) and an out-of-plane displacement f (xl, X2) are defined by the equation

which gives the three-dimensional position vector r (xl, x2) of an atom in the membrane as a

function of internal membrane coordinates xl and x2 attached to the monomers. These

internal parameters multiply orthogonal unit vectors êl and ê2 which span a flat zero temperature reference state (typically, a hexagonal piece of triangular lattice with lattice constant a) of characteristic linear dimension L. The unit vector ê3 is given by ê3

=

êl x ê2.

The prefactor mo is an order parameter (12) which measures the shrinkage of the surface caused by thermal fluctuations. The free energy of a nearly flat tethered membrane is a sum of

bending and stretching energies,

where K is a bending elastic constant, the elastic stretching energy has been expanded in

powers of the strain matrix, and » and À are elastic constants. The probability of a particular configuration parametrized by u (xl, x2) and f (xl, x2) is proportional to e kob T

Nonlinearities in the out-of-plane displacement enter via the strain matrix, given by

Because of such nonlinear couplings, the renormalized long-wavelength bending rigidity

and elastic constants differ considerably from their microscopic values. These renormalized elastic constants enter an effective, long-wavelength free energy for the Fourier transformed

phonon variables u (q ) and f (q ), namely

where f2 is the membrane surface area in the initial, stretched state. The probability of a particular fluctuation is now proportional to exp[-Feff/kB T]. Unlike the conventional elastic theory of thin plates (13), the renormalized wave-vector-dependent bending rigidity K R ( q ) and in-plane elastic parameters ftR(q) and ÀR(q) are singular for small q (6, 10). The bending rigidity diverges according to (6, 10)

while the elastic constants vanish as q tends to zero (10)

The singular, small q behavior of these elastic constants can be calculated via an epsilon-

expansion, for D

=

4 - e-dimensional manifolds embedded in a d-dimensional space (10), or

directly for the physically relevant case D

=

2, d

=

3 by an integral equation approach (14). A

(5)

straightforward generalization of the integral equation for K R(q ) derived in reference [6]

gives

where KR(q ) is a function of the renormalized elastic constants,

and PJ(k)

=

& ii - ki kjlk 2 is the transverse projection operator. Upon inserting the scaling

ansatzes (1.5) and (1.6) into the integral equation (1.7), we obtain an important scaling relation, first derived to all orders in an epsilon expansion by Aronovitz and Lubensky (10),

The exponent C determines the size of out-of-plane fluctuations ; using (1.4), we find

so that the membrane thickness is J"(j2) ’" L e. We introduced an upper cutoff a-1, where a

is the lattice spacing. In-plane phonon fluctuations, on the other hand, are. determined by W.

Equation (1.4) leads to

When tethered surfaces with a perfect six-fold triangular coordination topology are confined to a plane, one expects power law Bragg peaks in the in-plane scattering function (15),

The quasi-long-range translational order embodied in these peaks is destroyed when the

surface is allowed to fluctuate out of the plane, provided w z 0 (10). Long-range bond orientational order, however, is preserved. Indeed, fluctuations in the local bond angle field (15) 03B8(x) = 4 Eij 2 a 1 uj(x) are given by

The integral converges as Lu oo provided £o z 2, which is equivalent via equation (1.9) to

the inequality (clearly necessary in any flat phase), ( - 1. Thus tethered surfaces with the

topology of a triangular lattice are in fact « tethered hexatics » (10), a prediction we shall test

in section 3.

(6)

We now summarize the details of the simulation procedure. Our model for the molecular

dynamics simulations is the same as reference [7], where the tethering is enforced by a

continuous potential. The particles on the network are arranged in a triangular array and interact with their nearest neighbors through the potential

where r’ = 2 (2’/’) + f - r. The region 21/6 - r - 21/6 + f is thus force free and equivalent to

the flexible string of other models (4) of tethered membranes. In our calculations we take

f

=

0.5 unless explicitly stated otherwise. Self-avoidance is generated by the interaction between particles which are not nearest neighbors on the network. We take this interaction to be

with a a parameter. The parameter u is a measure of an « effective » hard-core radius of the

distant-neighbor particles. The case a

=

0 corresponds to the phantom membrane ;

o-

=

1 to the self-avoiding membrane in which self-intersections are impossible. We have also

considered the effects of attraction by replacing Vd(r) with the simple Lennard-Jones

potential with an attractive well,

with a

=

1, and a smoothing procedure to eliminate the small discontinuity at r

=

2.5 a. We

have considered finite systems that are hexagonal in shape and characterized by a linear

dimension L. A hexagonal sheet of size L contains (3 L 2 + 1 ) /4 particles. We have simulated

membranes up to size L

=

75 (4 219 particles) and for 106 to 107 total time steps, the longer

times corresponding to the larger clusters. The procedure is a straightforward molecular- dynamics calculation. Unless stated otherwise (see Sect. 2), the membrane is initially in a flat configuration and the particles are given random displacements and zero velocities. The clusters have zero total linear and angular momentum. The classical equations of motion are integrated forward in time, and the appropriate microcanonical ensemble averages are calculated. Time is expressed in molecular dynamics time-steps. In simulations without the attractive distant neighbor potential ( 1.16), the temperature was typically kB T -- 0.6-0.7 s.

When the attractive potential was turned on, the temperature increased to kB T .-: 1.4

After the attractive potential was turned off in the « compact » or collapsed state, the kinetic energy was rescaled upward to accelerate the retum to equilibrium. The temperature of the flat phase which was eventually recovered was kB T -. 3.5 s. The properties of the flat phase depend only weakly on temperature when the Lennard-Jones pair potential (1.16) is turned

off.

The self-avoiding membranes are characterized by long relaxation times and large fluctuations in equilibrium. Figure 1 shows snapshots of molecular-dynamics configurations

which display the fluctuations in a 4 219-atom particle membrane, an effect which contributes to long relaxation times. These snapshots show the « folding » motion, which typically has a period of = 105 time steps for this large size cluster. Between these folding configurations, the

membrane returns to a nearly « flat » hexagonal form. In order to determine if our results are

(7)

characteristic of equilibrium, we have calculated the autocorrelation functions of the

macroscopic variables of interest and have estimated the relaxation times. In all cases, the

molecular-dynamics simulation was carried out for at least 100 such relaxation times and, except for L

=

75, for a much longer period.

2. Entropic origin of the bending rigidity.

We first address the issue why triangulated tethered surfaces are flat (7). The isotropic tethering potentials of references [4] and [7] lead to very flexible membranes with no explicit microscopic bending rigidity. A priori, one might have expected such surfaces to crumple (4).

One natural explanation of the results of reference [7] is that a bending rigidity proportional

to temperature is generated for entropic reasons by excluded volume interactions, even if

there is no such term in the microscopic Hamiltonian (16). In fact, such a term is generated immediately upon introducing next nearest-neighbor excluded volume constraints into a tethered network.

To see this, note first that a repulsive next-nearest-neighbor interaction tends to align the

normals ni 1 and n2 of the two triangular plaquettes spanned by the four hard spheres in figure 2. Upon assuming for simplicity that the maximum sphere separation is just the sphere

diameter itself (i.e., the tether length is zero) we find that the average of (ni n2> is

Here, cf> 0 = ’TT - cos -1 ( 1 /3) is the largest angle .0 between normals permitted by the excluded

volume constraint. We can model this effect by adding a term 8 H = - K (n 1 . n 2 - 1 ) to the microscopic Hamiltonian, which leads to

Fig. 2. - Pictoral definition of the normals fil and n2 of the two triangular plaquettes spanned by the

four hard spheres in the figure.

(8)

where 10 (x) and I1 (x) are Bessel functions. Equations (2.1 ) and (2.2) agree provided we take

K /kB T;:- 1.13, which is larger than the ratio which produced a flat phase in the simulations of reference [12].

To test the hypothesis of an entropic bending rigidity further, we repeated the simulations of reference [7] with first and second neighbor excluded volume interactions only. By systematically shortening the tether length (f in Eqs. ( 1.14) and ( 1.15)) relative to the range

of confining potential, we were able to induce a transition from a crumpled, « phantom »

membrane (whose squared radius of gyration varies as the logarithm of the linear dimension

L) to a flat phase with long-range order in the surface normals. As shown in figure 3, a change

in behavior occurs for a tether length of about 0.6. We have plotted here the time-averaged eigenvalues Aa of the moment of inertia tensor,

Fig. 3.

-

Time-averaged eigenvalues of the moment of inertia tensor, equation (2.3), as a function of tether length f for a tethered membrane with only first and second neighbor excluded volume interactions.

where ri is the position of the i-th monomer in a membrane consisting of N monomers. Also

shown in figure 3 is the squared radius of gyration RG,

For tethering lengths of order unity, the eigenvalues A1, A2 and A3 are approximately equal

and in ratios consistent with the logarithmically crumpled phantom membrane phase

discussed in reference [12]. For shorter tethering lengths, however, two of these eigenvalues

become much larger than the remaining one, indicative of a flat phase induced by the entropic

mechanism discussed above. Although we have not investigated this crossover in detail, the

transition which occurs in figure 3 for a tether length of about 0.6 is presumably the entropic analogue of the crumpling transition driven by an energetic coupling found by Kantor and

Nelson in reference [12]. Here, the transition is driven by an entropic effective rigidity which

is increased by shortening the tethers.

(9)

Fig. 4. - Temporal sequence of configurations of L

=

75 tethered membrane often removing the

attractive part of the distant neighbor interaction and starting from a highly convoluted compact object

and time t

=

0.

We have also simulated distant neighbor interactions with an attractive potential minimum using the Lennard-Jones potential (Eq. ( 1.16)). This preserves, however, the distant neighbor repulsive excluded volume interaction of reference [7]. As shown in figure 4 (for time

=

0),

the resulting membrane collapsed into a highly convoluted compact object at a reduced

(10)

temperature of kB T kB T

=

1.4, where e is the Lennard-Jones well-depth parameter. A scaling

e

analysis of the orientationally averaged structure function,

along the lines taken by Kantor et al. in reference [4], is shown in figure 5. Surfaces with

L

=

13, 25, 49 and 75 all collapse onto a single curve when plotted as a function of qL v, with v

=

2/3, showing that RG - L L 2/3 the result characteristic of a collapsed object

with self-avoidance (4). The slope of this log-log plot in the region RG « q « a 1 is

approximately - 3, consistent with the theoretical expectation that (4) S(q) ’- Ilq d,

1

where

the fractal dimension is dF

=

2/ v. It is remarkable that the simulation can find such a compact configuration : as discussed in reference [4], folding of a surface with self-avoidance and a

finite thickness into a compact object is a nontrivial task.

Fig. 5. - The directionally average structure factors for collapsed membranes of size L

=

13, 25, 49 and

75 plotted as a function of qL ’’ where v

=

2/3.

(11)

It is interesting to take one of these collapsed compact manifolds as an initial condition for the original model, after removing the attractive part of the distant neighbor interaction. The time evolution of an L

=

75 membrane with this initial condition is shown pictorially in figure 4, and the radius of gyration and moment of inertia eigenvalues, are shown as a

function of molecular dynamics time-steps in figure 6. The molecular dynamics produces

oscillations in size at short times (Fig. 6a), but eventually leads to an equilibrated membrane

in its flat phase. For long times, these are very close to the results of reference [7], obtained

from a completely different « stretched » initial condition.

Fig. 6. - The early (a) and long (b) time evolution of the moment of inertia eigenvalues upon removal of attraction for the collapsed membrane, eventually leading to an equilibrated flat phase.

Our interpretation of this result is as follows : the attractive distant neighbor interaction leads to a collapsed surface because it drives the effective bending rigidity negative (17). When

this attraction is turned off, the positive, entropically generated bending rigidity which

remains produces a xnembrane in its flat phase, essentially indistinguishable from the equilibrated membranes in reference [7].

3. Membrane fluctuations in real space.

In the remainder of this paper we focus on membranes in the flat phase with distant self- avoidance and with no attractive interactions. These simulations were carried out on

hexagonal sheets excised from a triangular lattice containing L monomers along the diagonal.

The Fourier-transformed density correlations associated with these manifolds were analyzed

in detail in reference [9]. The basic results will be summarized in section 4. In this section, we

study monomer distribution functions in real space. Our analysis leads to a better

understanding of diffraction data, and contains information not readily accessible in

conventional diffraction measurements.

(12)

We start with a discussion of membrane density profiles, which are averages of the

microscopic density function

over the time evolution of N monomer positions {rj}. The density profile perpendicular to

the surface is defined by

where z is a coordinate aligned with the smallest eigenvalue of the moment of inertia tensor and ... ) denotes a time average. We have integrated over the in-plane coordinates x and y, and normalized p (z, L ) such that

If the monomer density distribution along î is characterized by a single length scale

B/ (T> - L e (see the discussion in Sect. 1) we expect that p (z, L ) obeys

Here, 0 (w) is a scaling function describing the density profile normal to the surface. As shown in figure 7a, L

=

49 and 75 surfaces collapse to a single universal function with C

=

0.65, the value of C determined from the diffraction analysis in reference [9] (18).

The angularly averaged in-plane density profile

Fig. 7. The normal (a) and angularly averaged in-plane (b) density profiles for L

=

49 and 75 tethered

membranes with the appropriate scaling described in the text.

(13)

normalized so that

is shown in figure 7b. Because the dominant in-plane length scale for flat membranes is clearly L, we have plotted L 2 p 1 (R, L ) vs. R/L, for different values of L. The density distribution is

approximately flat until it drops precipitously at the edge of the surface. Unlike profiles in the z-direction, at least two additional length scales come into play near this interface. If we assume that the interfacial width at the edge of the membrane is dominated by the in-plane phonon fluctuations, this width should scale like /(U2> _ L CI) / 2 so that the interfacial profile

should sharpen like 1 /L 1- s with à = w /2 when plotted as a function of R/L. The results of reference [9], obtained by an analysis of in-plane phonon fluctuations near the center of

polymerized membranes, suggest that w

=

0.66. (Note that the values C

=

0.65 and

ca

=

0.66 are in good agreement with the scaling relation (1.4).) The interfaces in figure 7b do

indeed sharpen, but with the larger exponent of 8 "’V 0.7 instead of w /2

=

0.33. In-plane phonon fluctuations alone are evidently inadequate as a model of the interface. As we show

below, the out-of-plane fluctuations (f2(X) > are most pronounced near the perimeter, suggesting that the membranes curl up significantly near the edge. We conjecture that this curling introduces an additional contribution - Le to the interfacial width, leading to 8 ===== l.

We now consider a new in-plane coordinate system which has an instantaneous orientation

aligned with the orthogonal axes with coordinates, xmax and xd, determined by the two largest eigenvalues, Amax and Amid, of the moment of inertial tensor for a membrane

configuration. Time averages of various state variables are obtained in this « dynamical »

coordinate system. The plot frms = J (j2(x) for this coordinate system is presented in figure 8 for the L

=

75 membrane and suggests that fluctuations are particularly strong near the membrane boundary. Use of the « dynamical » coordinate system leads to the two-fold

anisotropy evident in the figure. A related anisotropy is clearly demonstrated in the in-plane one-particle density functions in figure 9. While there is a very narrow interface in the direction of the largest eigenvalue, the interface is much more diffuse in the transverse direction because the predominately one-dimensional « curling » fluctuations evident in

figure 8. The width of this interface has as a contribution which scales like Le because of the

curling. The interfacial width in the direction of the largest eigenvalue, on the other hand,

should scale like L CI) / 2.

Another measure of how fluctuations vary spatially within the flat phase is the time- averaged projection of the membrane normal along the z-axis (12). In figure 10, we plot the angular average of

as a function of the internal membrane coordinates for an L

=

75 surface, where n (x) is a unit normal erected perpendicular to each triangular plaquette. The z-axis is the instantaneous direction of the smallest eigenvalue of the moment of inertia tensor. Although Q (x) is large in the interior of the surface, it decreases near the perimeter, due to the wild

fluctuations associated with the free boundary conditions. Q (x) rises to half its value at the

membrane center in about 5 or 6 interparticle spacings, which can be interpreted as a

correlation length for the flat phase. For the smaller membranes simulated in references [4]

and [8], well over half of the monomers are within a correlation length of the boundary ! In

(14)

Fig. 8. - Contour and surface plots of frms = J i2(x» as a function of the in-plane coordinate system defined by the axes of the principal moment of inertia tensor.

Fig. 9. - The one-particle density profiles for the in-plane orthogonal coordinates defined by the axes of

the principal moment of inertia tensor. The densities are normalized by replacing L with the length of

the cross-sectional cord of the membrane at each point along the respective axes.

(15)

Fig. 10.

-

A surface of Q = ([ô. i]2) - 1/3 as a function of the internal coordinates for a

L

=

75 membrane. This « order pa-rameter » tends to zero near the membrane edge, due to the curling

fluctuations discussed in the text.

our view, simulations of the larger L

=

49 and L

=

75 surfaces are essential to clearly

demonstrate the existence of a flat phase uncontaminated by these edge fluctuations.

As pointed out by Aronovitz and Lubensky (10), a special feature of membranes with a

regular triangular coordination topology (provided w 2) is that long-range bond

orientational order is preserved despite the large fluctuations in the flat phase. Although

these fluctuations are sufficient to destroy the algebraic Bragg peaks which would be present in a surface confined to a plane, there should be long-range order in the bond angle

correlation function G 6 (X) = (exp [6 i ( e (x) - 0 (0» > , where 0 (x) is the angle of near- neighbor bond projected onto the average membrane plane makes relative to some in-plane

reference axis. As shown in figure 11, our membranes do indeed appear to be « tethered hexatics » : G 6 (x) decays very slowly, with an orientational correlation length comparable to

the membrane dimensions. By identifying the asymptote of this correlation function with

1 (e6(X» 12 0.005, we see that the order parameter is, however, very small,

1 (f/1 6(X) 1 0.07.

4. Approximate form of the structure function.

In this section we present a more detailed derivation of the predictions for diffraction from

polymerized membranes summarized in reference [9]. We compare our approximate

theoretical form for the structure function with a simulation of L

=

75 membranes.

Qualitative features of our results can be understood in terms of the results for membrane fluctuations in real space tabulated in section 3.

Figure 12a shows the structure function for an oriented membrane with L

=

75 (i.e., 4 219 monomers). To calculate this structure function, we rewrite equation (2.5) as

where rz(x) is the monomer coordinate along the direction of the smallest eigenvalue of the

moment of inertia tensor, and r, (x) is the corresponding perpendicular component. The z-

(16)

Fig. 11. - Evidence for the « fixed-connectivity hexatic » from the long-range order in Gr,.

axis is thus aligned with the average normal to the surface ; the structure function is averaged

over directions perpendicular to z. Experimental realizations of oriented tethered surfaces are

possible by, e.g., confining membranes between parallel glass plates. The function

S(q,, q , L ) can then be probed directly via, e.g., light scattering or X-ray diffraction

experiments.

The structure function can be calculated theoretically using the long wavelength description

of the flat phase embodied in the Gaussian-free energy equation (1.4), in terms of the exponents ’and w defined by equations (1.5) and (1.6). The calculations are similar to those which produced equations (1.10) and (1.11). Upon inserting the decomposition (1.1) into equation (4.1) and using properties of Gaussian averages we find that

Upon using equations (1 . 5) and (1.6), we find that the exponentiated averages must take the form

and

(17)

Fig. 12. - The structure function for an oriented membrane of L = 75 from simulation (a) and theory (b).

where the coefficients A, B, and B’ depend on the coefficients in equations (1.5) and (1.6).

We shall take B = B’ for simplicity, although this assumption is easily relaxed. Upon taking

the continuum limit in (4.2) and approximating the hexagonal integration domains by disks,

we find we must evaluate

(18)

where each integration is confined to a disk of radius L/2, and

A general method for simplifying such integrals is described in the Appendix, and leads to the

result

where Jo (x) is a Bessel function. The parameters b and b2 have been introduced to normalize the behavior of the structure function for small q. We choose units of q, and q1 such that

A3 = 3 min is the smallest eigenvalue of the moment of inertia tensor, and A.L = à (7ii + A2) is

the average of the remaining two eigenvalues. We then have b 1 - lim -J A3(L)jI L2l, and

L -. ao

for C

=

0.65 and

The remaining parameter B in (12) (or, more generally B and B’, if Eq. (4.3b) is used) must

be fit to experiment.

The asymptotic large L structure function is determined once these parameters are known.

We expect equation (4.6) to be accurate for all wavelengths large compared to typical

monomer dimensions including, in particular, wavelengths large compared to either the transverse or in-plane membrane size. Although we do not expect equations (4.3a, b) to be

reliable for x - x’ [ small or x - x’ 1 . L, the factor s[cos-1 s - s -., /1 _ S2 1 deemphasizes

these regions in equation (4.6) in favor of regimes where equations (4.3a, b) are valid.

The structure function S(q,, q.l’ L ) obtained from (4.6) for L

=

75 is plotted in figure 12b,

and provides a reasonable description of the simulation data. The theory predicts a

breakdown of scaling with L for q in the transverse direction : S(o, q.l’ L ) is not a function only of the product q 1 L over a wide range of intermediate wave vectors. The physical reason

for this peculiar behavior is the large in-plane phonon fluctuations : scaling is restored in

equation (4.6) if we suppress these fluctuations by arbitrarily setting B

=

0. A related

breakdown of scaling occurs near the interface in figure 7b : For large L, the in-plane

interface sharpens like 1 /L 1- s when plotted vs. r.l / L. Sharp interfaces lead to the oscillations in S( q z, q.l ; L ) along q.l. These oscillations are also visible in figure 5 of

reference [9], but are less pronounced for small L, reflecting to a more gradual interfacial

(19)

profile in this case. In the simple theory of S( q Z’ q 1- , L ) sketched above, the damping of the

oscillations for small L is controlled by the exponent 8 = w /2. The « curling » fluctuations

near the edge which lead to 8 = C are not taken into account. It is this larger value of 8 which

makes the oscillations in the simulation less sharp than those predicted by the theory (9).

Although this hydrodynamic theory cannot describe the interesting structure in the

simulation for qd > 1, where d is an interparticle spacing, the overall shape and folds in the

structure function in figures 12 are accounted for rather well. Had we not averaged over in- plane directions, the atomic-scale oscillations for qd ± 1 would have had a six-fold modulation

reflecting the long-range hexatic order discussed in section 3. Because the expected

modulation is small (it should be of relative order 1 (i6(X» 0.07 (19», we did not search

for it.

Acknowledgements.

It is a pleasure to acknowledge helpful conversation with I. P. Batra, G. Grest, Y. Kantor, M. Kardar and M. Plischke during the course of this investigation. One of us (DRN), would

like to acknowledge the support of the National Science Foundation, through Grant DMR88- 17291 and through the Harvard Materials Research Laboratory.

Appendix.

Evaluation of an integral.

We want to simplify the expression (4.4) for S(qz, q 1-’ L), which is a constrained four- dimensional integral. Since S(q,, q,, L) clearly cannot depend on the direction of q 1

,

we can average over orientations of q, in the plane to obtain

with

We first fix the direction and magnitude of

and integrate over the center of mass coordinate

As shown in figure 13, the constraint that both x, and x2 be inside a common disk of radius

Z./2 confines the center of mass integration to a lens-shaped region. Since f(x) is

independent of the center of mass, integration over X amounts to the computation of the area

of ihis lens, which is

(20)

Fig. 13.

-

Integration over the center of mass coordinate X = 1 (X1 + x2) for a fixed direction and 2

magnitude of x

=

x, - x2. As illustrated in (a), the set of possible locations for X becomes increasingly

constrained by the requirements that 1 Xl 1 - L /2 and 1 x21 [ L /2 for large x. This leads to the lens-like shaded domain of X-integration shown in (b).

The remaining angular integral for x is trivial and so we obtain

Upon defining a new variable s

=

x/L we recover equation (4.6). The remaining one-

dimensional integral is easy to evaluate numerically. As explained in the text, the constants b 1 and b2 have been introduced into equation (4.6) to normalize the small q, and q, of the structure function in the scaling (large L) limit. The constant B must be fit to

experiment.

References

[1] ALBERTS B., BRAY D., LEWIS J., RAFF M., ROBERTS K. and WATSON J. D., The Molecular

Biology of the Cell (Garland, New York) 1983. The best example of a biological tethered

surface is probably the spectrin protein skeleton of eurythrocytes, separated from its natural

lipid environment ;

See ELGSAETER A., STOKKE B., MIKKELSEN A. and BRANTON D., Science 234 (1986) 1217.

[2] BLUMSTEIN A., BLUMSTEIN R. and VANDERSPURT T. H., J. Colloid Interface Sci. 31 (1969) 236 ;

REGEN S. L., SHIN J.-S., HAINFIELD J. F. and WALL J. S., J. Am. Chem. Soc. 106 (1984) 5756.

[3] BEREDJICK N. and BURLANT W. J., J. Polymer Sci. A 8 (1970) 2807 ;

FENDLER J. H. and TUNDO P., Acc. Chem. Res. 17 (1984) 3.

[4] KANTOR Y., KARDAR M. and NELSON D. R., Phys. Rev. Lett. 57 (1986) 791 ; Phys. Rev. A 35 (1987) 3056.

[5] Eds. D. R. Nelson, T. Piran and S. Weinberg, Statistical Mechanics of Membranes and Interfaces

(World Scientific, Singapore) 1989.

(21)

[6] NELSON D. R. and PELITI L., J. Phys. France 48 (1987) 1085.

[7] ABRAHAM F. F., RUDGE W. E. and PLISCHKE M., Phys. Rev. Lett. 62 (1989) 1757.

[8] For earlier speculations along these lines, see PLISCHKE M. and BOAL D., Phys. Rev. A 38 (1988) 4943 ;

BOAL D., LEVINSON E., LIU D. and PLISCHKE M., Phys. Rev. A 40 (1989) 3292. In our view, it is difficult to distinguish between the isotropic crumpling hypothesis of reference [4] and the hypothesis of flat, but very rough phase in these more modest simulations.

[9] For a summary which focuses on results for the structure function, see ABRAHAM F. F. and NELSON D. R., Science 249 (1990) 393.

[10] ARONOVITZ J. A. and LUBENSKY T. C., Phys. Rev. Lett. 60 (1988) 2634.

[11] Although references [6] and [10] treat the flat phase of membranes without explicit distant self- avoidance, self-avoidance is believed to be unimportant for flat surfaces, provided one

introduces a bending rigidity into the theory. See reference [5] and section 2.

[12] KANTOR Y. and NELSON D. R., Phys. Rev. A 38 (1987) 4020 ;

See also PACZUSKI M., KARDAR M. and NELSON D. R., Phys. Rev. Lett. 60 (1988) 2638.

[13] LANDAU L. D. and LIFSHITZ E. M., Theory of Elasticity (Pergamon, New York) 1970.

[14] See reference [6], and the article on the crumpling transition by NELSON D. R. in reference [5].

[15] See, e.g., NELSON D. R. and HALPERIN B. I., Phys. Rev. B 19 (1979) 2457.

[16] Similar conclusions have been reached by LEIBLER S. and MAGGS A. C., Phys. Rev. Lett. 63 (1989)

406.

[17] We are indebted to KARDAR M. for discussions on this point.

[18] In reference [9], we argued that the L

=

13 and L

=

25 membranes were too small to give

reliable results for the flat phase. This is especially true for density profiles. See also the discussion below equation (3.7). If the analysis which led to 03B6 in reference [9] is repeated with

L

=

13, 15, 49, and 75, we find 03B6

=

0.76 with a poor scaling fit.

[19] See, for example, BRUINSMA R. and NELSON D. R., Phys. Rev. B 23 (1981) 402.

Références

Documents relatifs

A genuine economic and monetary union requires establishing both a “fiscal centre” with fiscal capacity to meet the challenge of solidaristic reconstruction and transformation of

Abstract.- A model is proposed to describe the gross features of superflow in 3 He-A under conditions where it is relaxing, at a rate limited by orbital viscosity, towards a state

The results for this exponent from several simulation studies of tethered networks with free edge boundary conditions are collected in table I. The results of our

At low temperature the transition is to a glassy crumpled phase with a vanishing average tangent order parameter but with a nonvanishing spin- glass order parameter. In this phase

In particular the sound absorption into pair vibra- tion modes may be used to measure the temperature dependence of the energy gap and the quasiparticle lifetime from the

Variation with temperature of the thickness of an adsorbed polymer layer in the collapsed

it) To design the stator, we choose the following quantities. a dimensionsless factor F representing the relative saturation of ~tator and rotor teeth.. the height h~, of stator

analysis oi the existence and stability oi an ordered flat phase oi the model for sufficiently large bending rigidity.. Crystalline membranes, aise known as tethered or