14.12.2012 Views

PDF (361 K) - Laboratoire Matière et Systèmes Complexes - Paris 7

PDF (361 K) - Laboratoire Matière et Systèmes Complexes - Paris 7

PDF (361 K) - Laboratoire Matière et Systèmes Complexes - Paris 7

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

PAPER www.rsc.org/softmatter | Soft Matter<br />

Corrections to the Laplace law for vesicle aspiration in micropip<strong>et</strong>tes and<br />

other confined geom<strong>et</strong>ries<br />

J.-B. Fournier and P. Galatola*<br />

Received 18th April 2008, Accepted 26th June 2008<br />

First published as an Advance Article on the web 1st September 2008<br />

DOI: 10.1039/b806589f<br />

Based on the stress tensor associated with the Helfrich Hamiltonian, we study how the Laplace law is<br />

modified for fluid lipid membranes under small tensions or constrained within small scale devices. We<br />

derive several exact analytical results corresponding to global or local generalized Laplace relations.<br />

Analytical and numerical results based on standard energy minimization confirm these exact relations<br />

and further quantify the deviations of the standard Laplace law.<br />

PACS numbers: 87.16.D-, 68.03.Cd, 68.35.Np<br />

1 Introduction<br />

Nowadays, nanopip<strong>et</strong>tes with aperture sizes of less than 200 nm<br />

are commonly used 1,2 and the development of microfludics<br />

allows on-chip fabrication and focusing of vesicles with diam<strong>et</strong>ers<br />

in the 100 nm range. 3 At the same time, the technology of<br />

solid-state nanopores is developing fast, allowing pores with<br />

diam<strong>et</strong>ers ranging from a few nanom<strong>et</strong>res to tens of nanom<strong>et</strong>res<br />

to be made. 4 There is also an increasing interest in the use of<br />

individual artificial vesicles as containers suited for the isolation,<br />

preservation and transport of small quantities of drugs or biomolecules.<br />

5,6 It has therefore become important to understand<br />

how aspiration pressure, membrane tension and curvature effects<br />

are related at the nanoscale, i.e., how is the Laplace law modified.<br />

The case of an ordinary interface b<strong>et</strong>ween two fluids is<br />

well-known. Calling s its surface tension and dP the pressure<br />

difference b<strong>et</strong>ween two phases I and II, the Laplace law gives the<br />

local mechanical equilibrium condition,<br />

dP ¼ s(c1 + c2) (1)<br />

in which c1 and c2 are the interface’s principal curvatures. 7 The<br />

latter are the minimum and maximum curvatures of the curves<br />

obtained by cutting the surface by a plane containing the normal<br />

vector; the directions to which they belong are perpendicular. In<br />

eqn (1) the sign convention is such that if curvatures are<br />

considered positive when the curvature center is inside phase I,<br />

then dP ¼ PI PII. Lipid membranes are particular interfaces, produced by the<br />

self assembly of lipid surfactant molecules in an aqueous<br />

environment. They are highly flexible, fluid bilayers, of thickness<br />

x5 nm, forming closed vesicles of sizes ranging from a few tens<br />

of nanom<strong>et</strong>res up to hundreds of mm. Vesicles are considered as<br />

model systems for the outer walls of living cells, and they have<br />

applications as encapsulation vectors for drug delivery. 8 In<br />

addition to s and dP, membrane interfaces are characterized by<br />

the Helfrich bending modulus k, defined through the energy cost<br />

<strong>Laboratoire</strong> <strong>Matière</strong> <strong>et</strong> <strong>Systèmes</strong> <strong>Complexes</strong> (MSC), UMR 7057 CNRS<br />

& Université <strong>Paris</strong> Diderot–<strong>Paris</strong> 7, 10 rue Alice Domon <strong>et</strong> Léonie<br />

Duqu<strong>et</strong>, F-75205 <strong>Paris</strong> Cedex 13, France<br />

of curving a (symm<strong>et</strong>ric) membrane: ½k(c1 + c2) 2 per unit area. 9<br />

We recall that vesicles essentially have a constant volume V and<br />

a constant membrane area A, but that the apparent, visible area<br />

may significantly increase with s as the microscopic membrane<br />

fluctuations unfold. 10<br />

The equilibrium equation generalizing the Laplace law for a<br />

membrane interface (the so-called shape equation) is a quite<br />

complex non-linear partial differential equation. For a symm<strong>et</strong>ric<br />

membrane, it reads: 11<br />

dP ¼ 2sH 4kH(H 2 K) 2kDsH (2)<br />

where H ¼ ½(c 1 + c 2) is the mean curvature, K ¼ c 1c 2 the<br />

Gaussian curvature, and Ds the covariant surface Laplacian. 12<br />

Note that the k / 0 limit simply yields the ordinary Laplace<br />

law. Note also that other types of soft-matter materials,<br />

e.g., liquid crystals, exhibit different kinds of generalized<br />

Laplace laws. 13,14 For lipid membranes, k z 1 10 19 J, while s<br />

is an adjustable quantity that is d<strong>et</strong>ermined by the external<br />

forces exerted on the membrane. Typically, s ranges b<strong>et</strong>ween<br />

1 10 8 Jm 2 and 1 10 3 Jm 2 , the lower bound corresponding<br />

to freely floating floppy vesicles and the upper bound<br />

to the highest tensions that one can induce, by means of strong<br />

aspiration or adhesion, before the lipids tear apart.<br />

Calling R the typical curvature radius of a vesicle, or the<br />

typical length over which the curvature varies, we see that the<br />

surface tension term in eqn (2) is of order s/R, while the curvature<br />

terms (those involving k) are of order k/R 3 . Hence, the<br />

curvature corrections to the Laplace law are relevant only if R (<br />

O(k/s). With typically s x 1 10 5 Jm 2 , curvature corrections<br />

are found to be relevant when R ( 0.1 mm, i.e., at the nanoscale.<br />

Micropip<strong>et</strong>te aspiration experiment<br />

In this classical experiment (Fig. 1), used to measure k through<br />

the expansion of the visible area of the vesicle, 15,16 the standard<br />

Laplace law, eqn (1), is commonly used instead of eqn (2). This is<br />

indeed reasonable, since the typical radius of the micropip<strong>et</strong>tes is<br />

a x 3 mm, yielding a [ O(k/s) as soon as s [ 1 10 8 Jm 2<br />

(which is almost instantly attained as soon as the vesicle is<br />

pressurized).<br />

This journal is ª The Royal Soci<strong>et</strong>y of Chemistry 2008 Soft Matter, 2008, 4, 2463–2470 | 2463


Fig. 1 Geom<strong>et</strong>ry of the vesicle aspiration experiment. The pip<strong>et</strong>te radius<br />

is a, the aspiration pressure is DP ¼ P0 P1 > 0, where P0 is the pressure<br />

outside the channel and P1 the pressure inside the channel; the pressure<br />

inside the vesicle is P. The outer part of the vesicle is quasi-spherical. The<br />

regions M0, M and M1 corresponding to the outer, tubular (length L)<br />

and inner parts, respectively, are indicated. The region M is pressed onto<br />

the channel, while the other two are free.<br />

The free part of the membrane inside the micropip<strong>et</strong>te is<br />

usually described as a perfect spherical cap of radius r ¼ a. 15–17 As<br />

we shall see, this is inexact. Indeed, in the absence of adhesion, it<br />

is well-known 18 that the equilibrium curvature at the d<strong>et</strong>achment<br />

point in the plane of symm<strong>et</strong>ry must be equal to the curvature of<br />

the substrate in that direction, i.e., must vanish in the present<br />

case. Therefore, the curvature cannot be constant. Nevertheless,<br />

in the ordinary case [a [ O(k/s)], eqn (1) has to be almost<br />

perfectly satisfied—except possibly in a boundary layer. Hence<br />

the curvature radius r must be constant except very close to the<br />

micropip<strong>et</strong>te wall. As we shall see, in the absence of adhesion it<br />

turns out that this constant curvature radius is indeed r x a.<br />

Therefore the standard picture is correct in the ordinary case, i.e.,<br />

when there is no adhesion and a [ O(k/s).<br />

As for the outer free part, it is usually almost a spherical cap of<br />

radius R [ a [ O(k/s). Then, applying, far from the boundary<br />

layers, eqn (1) twice, once to the free part of the membrane<br />

inside the micropip<strong>et</strong>te and once to the outer free part, one g<strong>et</strong>s<br />

P P1 ¼ 2s/a and P P0 ¼ 2s/R, which yields the relationship<br />

b<strong>et</strong>ween s and DP ¼ P0 P1 used in the literature: 15<br />

DP ¼ 2s 1<br />

a<br />

Note that the fluctuation corrections to s are neglected here. 17,19<br />

Scope and organization of the paper<br />

In this work, we discuss several forms of the generalized Laplace<br />

law for membranes, in situations where the standard Laplace law<br />

does not hold—essentially in nanochannels or in the presence of<br />

adhesion b<strong>et</strong>ween the membrane and the channel. We consider<br />

both tubular and conical channels. We discuss global relationships<br />

of the type shown in eqn (3), that take into account the<br />

corrections coming from the curvature energy and from the<br />

adhesion energy. We also discuss local forms, simpler than eqn<br />

(2), valid on the axis of a revolution symm<strong>et</strong>ric shape. We obtain<br />

these relationships from a straightforward force equilibrium<br />

analysis, which is compl<strong>et</strong>ely equivalent to the standard Helfrich<br />

energy approach. 9 We then recover these results within a more<br />

traditional energy minimization analysis. In the last part of the<br />

paper, we present numerical results showing the departure of<br />

the free membrane part from a spherical cap and quantifying<br />

the violation of the standard Laplace law.<br />

1<br />

R<br />

(3)<br />

2 Force equilibrium analysis<br />

In the Helfrich model, 9 the free energy of a vesicle of volume V<br />

and area A within an aspiration channel can be expressed as:<br />

F ¼ sA GAc þ Wpr þ k<br />

ð<br />

dAðc1 þ c2Þ<br />

2<br />

2<br />

(4)<br />

where Ac represents the portion of the vesicle area in contact with<br />

the channel, G the adhesion energy per unit surface (possibly<br />

negligible), Wpr the work of the pressure forces, and the other<br />

quantities have been previously defined. In the present case,<br />

Wpr ¼ PV + P0Vext + P1Vint, where Vint is the volume of the<br />

vesicle portion aspirated in the channel (where the pressure is P1),<br />

Vext ¼ V Vint is the remaining volume of the vesicle outside the<br />

channel (where the pressure is P 0), and P is the pressure inside the<br />

vesicle. Note that s and P can be interpr<strong>et</strong>ed as Lagrange<br />

multipliers fixing A and V. We have not included the Gaussian<br />

rigidity k, which does not contribute in the absence of topology<br />

changes 9 and we have assumed that the membrane is symm<strong>et</strong>ric<br />

(no spontaneous curvature).<br />

2.1 Membrane stresses<br />

We have introduced the free energy in eqn (4) in order to fully<br />

specify the system under study. Until Sec. 3, however, we shall<br />

not use it . Indeed, the Helfrich model is equivalent to the<br />

following description in terms of stresses. 20,21 The force S M dlthat<br />

a piece of membrane M exerts onto an adjacent piece M 0<br />

through a cut dlhas a very simple expression in the case where<br />

the cut is parallel to a direction of principal curvature (otherwise<br />

a tensorial description must be used). L<strong>et</strong> t M denote the unit<br />

vector tangent to the membrane, perpendicular to the cut, and<br />

directed toward the interior of M (Fig. 2). L<strong>et</strong> n denote the<br />

normal to the membrane, with the convention 22 that curvatures<br />

are considered positive if the membrane is curved towards n,<br />

then: 20,21<br />

S M ¼ S M t t M + S M n n (5)<br />

S M<br />

t<br />

k 2<br />

¼ s G þ ck 2<br />

S M<br />

n<br />

¼ 2k vH<br />

vt M<br />

k 2<br />

ct 2<br />

where c || and c t are the curvatures parallel and perpendicular to<br />

the cut, respectively, H ¼ ½(c || + c t) is the mean curvature, and<br />

t M is the coordinate in the direction of t M . The term G arises<br />

when the membrane is in contact with a substrate with adhesion<br />

energy G. 21 For the free membrane parts, G ¼ 0.<br />

Fig. 2 The geom<strong>et</strong>ry for the definition of the force S M dlexerted by the<br />

region M onto the region M 0 through a cut dlin the plane tangent to the<br />

membrane.<br />

2464 | Soft Matter, 2008, 4, 2463–2470 This journal is ª The Royal Soci<strong>et</strong>y of Chemistry 2008<br />

(6)<br />

(7)


Note that eqn (5)–(7) satisfy the action–reaction principle<br />

SM0 ¼ SM , where SM0 is the force per unit length that M0 exerts<br />

onto M. Indeed, SM0 t ¼ SM t , since eqn (6) depends only on the<br />

local curvatures. On the contrary, SM0 n ¼ SM n , since tM0 ¼ tM .<br />

Then, since tM0 ¼ tM while n is the same both for M and M0 , the<br />

action–reaction principle follows.<br />

2.2 Continuity relations<br />

L<strong>et</strong> us comment on the continuity of the components of the stress<br />

along the membrane. At equilibrium, everywhere along the free<br />

part of the membrane, the continuity of S must obviously be<br />

satisfied. But what happens at the d<strong>et</strong>achment point b<strong>et</strong>ween the<br />

membrane and the pip<strong>et</strong>te? The part of the membrane in contact<br />

with the pip<strong>et</strong>te substrate is free to glide, therefore the continuity<br />

of St must be satisfied—even though the curvatures are discontinuous.<br />

However, in general, there will be a discontinuity of S n.<br />

This is due to the fact that there is actually a narrow region,<br />

intermediate b<strong>et</strong>ween the free and the bound part, where the<br />

membrane feels a potential W(z) depending on the distance z to<br />

the substrate. This intermediate region is subject to a density of<br />

normal forces dW(z)/dz. The latter results in an apparent<br />

discontinuity of Sn in a theory neglecting the width of the<br />

intermediate region.<br />

2.3 Aspiration in a cylindrical nanochannel<br />

L<strong>et</strong> us d<strong>et</strong>ermine the balance along ^z of the forces acting on<br />

region M1 (Fig. 1). They have two origins: the pressure forces<br />

acting all along the surface of M1 and the membrane elastic<br />

forces acting through the contour separating M from M1.<br />

Hence, at equilibrium, we have:<br />

Ð<br />

M1 (P P1) n$^z dA+ 2pa SM $^z ¼ 0 (8)<br />

Taking into account that Ð M 1 n$^z dA ¼ pa 2 and S M $^z ¼ S M t ,<br />

while S M t ¼ s G +½k/a 2 (within the tube c|| ¼ 1/a and ct ¼ 0),<br />

we obtain:<br />

s G k<br />

P P1 ¼ 2 þ<br />

a a3 (9)<br />

which is an exact global Laplace relation that takes into account<br />

the curvature and adhesion corrections. Note that this equation<br />

can be rewritten as:<br />

P P1 þ 2G 2s<br />

¼<br />

a a<br />

1 þ 1<br />

2<br />

l<br />

a<br />

2<br />

(10)<br />

where l ¼ O(k/s), which shows that the adhesion energy acts as<br />

a suction pressure that adds up to the pressure difference P P1,<br />

and that the curvature correction is important for a ( l, i.e.,<br />

essentially for nanopip<strong>et</strong>tes at ordinary tensions. In the latter<br />

case, the curvature corrections are important everywhere in M1,<br />

and M 1 departs significantly from a spherical cap. Nevertheless,<br />

eqn (9) holds exactly.<br />

Note now the following subtl<strong>et</strong>y that will be important in the<br />

following. In eqn (8), we have used for S M t the expression s G +<br />

½k/a 2 , which is actually valid inside the tubular region, while we<br />

should use the stress exerted immediately after the d<strong>et</strong>achment<br />

point. We are safe, however, because the tangential component of<br />

the stress is continuous. This is why eqn (9) holds independently<br />

of the fact that the shape of the cap is not a half-sphere of radius<br />

a; indeed, the size a that enters eqn (9) is the radius of the<br />

cylinder.<br />

The continuity of the tangential stress at the d<strong>et</strong>achment point<br />

gives us an important equilibrium condition. Calling c0 the<br />

boundary curvature at the d<strong>et</strong>achment point in the direction<br />

parallel to the tube axis, we obtain S M1<br />

t<br />

Then S M t ¼ S M1<br />

t yields:<br />

c0 ¼<br />

rffiffiffiffiffiffi<br />

2G<br />

k<br />

1<br />

¼ s þ<br />

2 k=a2 1<br />

2<br />

kc 2<br />

0 .<br />

(11)<br />

This corresponds to the well-known Seifert–Lipowsky adhesion<br />

boundary condition, 18 which is valid also on a curved wall<br />

when the curvature of the wall in the direction perpendicular<br />

to the contact line vanishes. 23,24 Note that when this latter<br />

condition is not satisfied, the full equilibrium condition requires<br />

also the torque balance. 25 In the absence of adhesion, eqn (11)<br />

yields:<br />

c 0 ¼ 0 (12)<br />

Therefore, as mentioned in the introduction, M 1 is never an<br />

exactly perfect spherical cap. 26<br />

Note that relation shown in eqn (9) may also be obtained by<br />

a careful minimization of the total free energy shown in eqn (4)<br />

with respect to the tube length L. Indeed, the curvature term in<br />

eqn (4), proportional to k, is equal to Ec ¼ pkL/a + Ec 1 + Ec 0 ,<br />

where the first term corresponds to the cylindrical part and the<br />

other two contributions correspond to regions M1 and M0.<br />

Thanks to the Lagrange multiplier P, V can be varied independently<br />

of the other param<strong>et</strong>ers; thus the same holds for L.<br />

Therefore, since E c1 and E c0 are independent of L, we obtain<br />

0 ¼ svA/vL GvA c/vL PvV/vL + P 0vV ext/vL + P 1vV int/vL +<br />

vE c/vL ¼ 2pas 2paG pa 2 P + pa 2 P 1 + kp/a, yielding<br />

eqn (9).<br />

Generalized global Laplace relation<br />

L<strong>et</strong> us now assume that the vesicle outside the nanochannel is<br />

much larger than O(k/s), which allows us to use the approximate<br />

Laplace relation eqn (1) instead of the exact one eqn (2). Calling<br />

R the curvature radius where the membrane me<strong>et</strong>s the revolution<br />

axis, we g<strong>et</strong> P P0 x 2s/R. Combining this equation with eqn<br />

(9), we obtain the generalized global Laplace relation:<br />

DP x 2s 1<br />

a<br />

1<br />

R<br />

G<br />

sa<br />

þ k<br />

a 3<br />

(13)<br />

which generalizes eqn (3). This is one of our central results. It<br />

holds most generally for vesicle aspiration in any cylindrical<br />

nanochannel (even when the free membrane cap M1 is far from<br />

being spherical).<br />

Macroscopic channels<br />

Note that for channels large enough such that R [ a [ O(k/s),<br />

the term involving k is negligible in eqn (9) and we obtain P P1<br />

x 2(s G)/a. Now, since in this case the ordinary Laplace law<br />

This journal is ª The Royal Soci<strong>et</strong>y of Chemistry 2008 Soft Matter, 2008, 4, 2463–2470 | 2465


(1) holds everywhere except possibly in a boundary layer (cf. the<br />

discussion in the Introduction), region M 1 must be a spherical<br />

cap almost within the whole gap. Therefore, the radius of the<br />

spherical cap M1 is not r x a, but:<br />

r x<br />

a<br />

1 G=s<br />

(14)<br />

This is our second central result. For G ¼ 0, we recover the<br />

standard approximation that M 1 is a half-sphere matching the<br />

tube radius.<br />

2.4 Aspiration in a conical nanochannel<br />

The problem is more complex for a conical channel (see Fig. 3),<br />

because of the discontinuity of the normal stress at the d<strong>et</strong>achment<br />

point. The force balance along ^z, for region M1, can be<br />

written as:<br />

pa 2 (P P1) 2pa(S M 1<br />

t cosa + S M 1<br />

n sina) ¼ 0 (15)<br />

Note that instead of using S M as previously, we have used<br />

S M 1 , which is correct, because S M 1 is indeed the force<br />

exerted on M1 by the rest of the membrane (action–reaction<br />

principle). Thanks to the continuity of the tangential stress,<br />

S M1<br />

t ¼ SM t<br />

¼ s G þ 1<br />

2 kðcos aÞ2 =a 2 , because of the conical<br />

shape. We now discuss S M 1<br />

n . The symm<strong>et</strong>ry of revolution implies<br />

H ¼ 1<br />

2 ðck þ ctÞ ¼ 1<br />

ðcos q=r þ cÞ, where r is the distance to the<br />

2<br />

revolution axis, q is the inclination of the tangent with respect to<br />

this axis, and c is the curvature of the membrane section in<br />

the plane containing the revolution axis. Thus, according to<br />

eqn (7),<br />

S M !<br />

d cos q<br />

1<br />

n ¼ k c þ ¼ k<br />

ds r 1<br />

dc cos a c1a<br />

þk<br />

ds 1 a2 sin a (16)<br />

where the index ‘‘1’’ means evaluation at the d<strong>et</strong>achment point of<br />

region M1, and where s is the curvilinear coordinate oriented in<br />

the same direction as ^z. The tangential stress continuity implies<br />

again<br />

rffiffiffiffiffiffi<br />

2G<br />

c1 ¼<br />

(17)<br />

k<br />

at the d<strong>et</strong>achment point. Hence, we obtain:<br />

Fig. 3 Geom<strong>et</strong>ry of vesicle aspiration in a conical nanochannel of<br />

aperture 2a. The regions M 0, M and M 1 corresponding to the outer,<br />

conical and inner parts, respectively, are indicated. The pressure within<br />

the channel is P 1, the pressure outside is P 0, and the pressure inside the<br />

vesicle is P. The radius of the channel at the contact point b<strong>et</strong>ween region<br />

M 1 and region M is a.<br />

s G<br />

P P1 ¼ 2 cos a þ<br />

a<br />

k<br />

a3 cos a 1 þ sin2a 2 ffiffiffiffiffiffiffiffi p<br />

2Gk<br />

a2 sin 2 a þ 2 k dc<br />

sin a<br />

a<br />

ds 1<br />

(18)<br />

This equation generalizes the global Laplace relation, eqn (9);<br />

unfortunately (dc/ds)|1 is unknown and can only be d<strong>et</strong>ermined<br />

by solving for the compl<strong>et</strong>e shape of region M1. More results<br />

concerning the conical channel will be given in Sec. 3.2<br />

2.5 Generalized local Laplace law at an umbilical point<br />

For a shape with revolution symm<strong>et</strong>ry, the point where the<br />

membrane me<strong>et</strong>s the revolution axis has its two principal<br />

curvatures equal. At such a point, called an umbilical point,<br />

the generalized Laplace law, eqn (2), takes a simpler form:<br />

indeed, the contribution proportional to H2 K vanishes,<br />

because H2 ¼ c2 ¼ K, and the covariant surface laplacian Ds<br />

equals 2v2 /vx2 , where x is a cartesian coordinate in the tangent<br />

plane. Identifying, at lowest order, the coordinate x in the<br />

tangent plane with the curvilinear coordinate s in a symm<strong>et</strong>ry<br />

plane, we obtain DsH ¼ 2 v2H/vs2 ¼ 2d2c/ds2 , where the last<br />

equality easily follows from the symm<strong>et</strong>ry of revolution and<br />

some differential geom<strong>et</strong>ry. Hence, the generalized Laplace law,<br />

eqn (2), becomes:<br />

dP ¼ 2sc 4k d2c ds2 (19)<br />

To verify this relation, we write the balance of the normal<br />

forces acting on an elementary cap of radius r / 0 centered on<br />

an umbilical point O (see Fig. 4):<br />

dPpr2 h<br />

¼ 2pr<br />

s þ k 2<br />

ck 2<br />

2k vH<br />

cos f<br />

vs<br />

k<br />

2 c2 sin f<br />

(20)<br />

where dP is the pressure difference across the membrane and one<br />

recognizes the expressions of St and Sn, as expressed in eqn (6)<br />

and (7), with ct ¼ c, G ¼ 0 (free membrane), and s the curvilinear<br />

coordinate, measured from O. Note that for r s 0, we have<br />

c|| s c. The limit r / 0 yields eqn (19), since s þ k<br />

2 c 2 k<br />

k<br />

2 c2 /s,<br />

(sinf)/r / c, cosf / 1 and (1/r)vH/vs / v 2 H/vs 2 |0.<br />

Eqn (19) is an exact generalized Laplace relation including the<br />

curvature corrections. Note that it can be applied to the tip of<br />

region M1 in the nanopip<strong>et</strong>te problem (Fig. 1), but the values of c<br />

and d 2 c/ds 2 are unknown unless we solve the whole problem. The<br />

Fig. 4 Elementary cap of radius r around an umbilical point for the<br />

d<strong>et</strong>ermination of the local generalized Laplace law at the tip of<br />

a symm<strong>et</strong>ry axis.<br />

2466 | Soft Matter, 2008, 4, 2463–2470 This journal is ª The Royal Soci<strong>et</strong>y of Chemistry 2008


m<strong>et</strong>hod we used above is therefore more efficient. Note, however,<br />

that this equation can be used at the tip of region M 0, implying<br />

that eqn (13) takes the exact form:<br />

DP ¼ 2s 1<br />

a<br />

1<br />

R<br />

G<br />

sa<br />

þ k<br />

a 3<br />

4k d2 c<br />

ds 2<br />

where d 2 c/ds 2 is evaluated at the tip of region M0.<br />

3 Energy minimization analysis<br />

(21)<br />

In this section we give the differential equations and the<br />

boundary conditions governing the shape of the free membrane<br />

inside the cylindrical or conical nanochannels of Fig. 1 and 3. As<br />

a first application, we discuss numerically the features of the<br />

solution that cannot be predicted analytically. We then redemonstrate<br />

analytically the d<strong>et</strong>achment condition eqn (17) and the<br />

generalized Laplace eqn (18) and (19) that we have obtained from<br />

the stress tensor analysis.<br />

3.1 Differential equations and boundary conditions<br />

We consider generically the region M1 in the conical nanochannel<br />

of Fig. 3. Indeed, the nanopip<strong>et</strong>te geom<strong>et</strong>ry of Fig. 1 is<br />

a particular case where a ¼ 0, while the region M 0 of the conical<br />

nanochannel of Fig. 3 corresponds to the case a < 0. The minimization<br />

of the free energy shown ineqn (4) for axisymm<strong>et</strong>ric<br />

shapes yields the s<strong>et</strong> of differential equations for the profile of the<br />

free part M1: 27,28<br />

j 00 ¼<br />

g sin j<br />

kr þ ðP1 PÞr cos j<br />

2k<br />

g 0<br />

¼ 1<br />

2 k j 0 2 sin2 j<br />

r 2<br />

þðP1 PÞr sin j<br />

j 0 cos j sin 2j<br />

þ<br />

r 2r2 þ s<br />

(22)<br />

(23)<br />

r 0 ¼ cosj (24)<br />

z 0 ¼ sinj (25)<br />

where j is the angle that the tangent to the profile forms with<br />

the radial direction r and a prime indicates derivation with<br />

respect to the arclength s. We s<strong>et</strong> s ¼ 0 at the d<strong>et</strong>achment<br />

point of M1 and s ¼ s1 at the revolution axis r ¼ 0; s1 is thus<br />

the length of the profile. The function g(s) is a Lagrange<br />

multiplier field that enforces the condition that s is everywhere<br />

the arclength of the profile. The boundary conditions are<br />

obtained by s<strong>et</strong>ting to zero the boundary terms of the free energy<br />

variation: 27,28<br />

jð0Þ ¼ p<br />

þ a; rð0Þ ¼a; zð0Þ ¼0 (26)<br />

2<br />

j(s1) ¼ p, r(s1) ¼ 0 (27)<br />

P<br />

2 gð0Þtan a<br />

P1 ¼<br />

a2 2 s k cos a<br />

þ þ<br />

a cos a a3 k<br />

a cos a j 02 ð0Þ<br />

(28)<br />

P P1 ¼<br />

2 gð0Þtan a<br />

a2 2ðs GÞ k cos a<br />

þ þ<br />

a cos a a3 (29)<br />

Eqn (26) and (27) reflect the geom<strong>et</strong>ry of the problem, eqn (28)<br />

is obtained by varying the length of the axisymm<strong>et</strong>ric contour<br />

(taking into account the existence of a conserved Hamiltonian29 ),<br />

and eqn (29) is obtained by displacing the d<strong>et</strong>achment point of<br />

region M1, taking into account the free energy FM of the<br />

adhering part M:<br />

FM ¼ 2p<br />

ð z1<br />

z0<br />

k cos a<br />

2rðzÞ<br />

ðs GÞrðzÞ<br />

þ þ<br />

cos a<br />

ðP1 PÞr2ðzÞ 2<br />

dz (30)<br />

Here, r(z) ¼ a ztana is the equation of the cone, z0 is the z<br />

coordinate of the d<strong>et</strong>achment point of M 0 and z 1 ¼ 0 is the z<br />

coordinate of the d<strong>et</strong>achment point of M 1, at which r ¼ a.<br />

Eqn (22)–(25) with the seven boundary conditions (26)–(29)<br />

fully d<strong>et</strong>ermine the problem. Indeed, we have five first-order<br />

partial differential equations, plus the two unknowns s1 and<br />

P P1.<br />

3.2 Numerical analysis<br />

We solve numerically the shape eqn (22)–(25) with the boundary<br />

conditions shown in eqn (26)–(29) by using a standard finite<br />

difference scheme with deferred correction and Newtonian iteration<br />

30 . We introduce the normalized quantities:<br />

z ¼ z r s<br />

; r ¼ ; s ¼<br />

a a a ; s1 ¼ s1<br />

a<br />

s ¼ sa2<br />

3 ðP P1Þa<br />

; p ¼<br />

k k<br />

(31)<br />

(32)<br />

G ¼ Ga2 ga<br />

; g ¼ (33)<br />

k k<br />

To cope with the unknown param<strong>et</strong>ers s1 and p, we write<br />

the differential equations in terms of the independent param<strong>et</strong>er<br />

t ¼ s/s1, with 0 # t # 1, and we treat s1 and p1 as two extra<br />

functions of the param<strong>et</strong>er t, obeying the differential equations<br />

ds1/dt ¼ dp/dt ¼ 0. Our problem is thus transformed into<br />

a standard s<strong>et</strong> of seven first-order differential equations obeying<br />

seven boundary conditions: five at the left boundary t ¼ 0(s¼ 0)<br />

and two at the right boundary t ¼ 1, (s ¼ s1).<br />

Because of the singularity of the differential equations at r ¼ 0,<br />

we replace the two right boundary conditions, eqn (27), with the<br />

following conditions at s ¼ s1 3 (with 0 < 3 s1): r (s1 3) ¼ 3 (34)<br />

jðs1 3Þ ¼p 3 1 þ 3<br />

3 gðs1 3Þ dj<br />

ds s1 3<br />

(35)<br />

that are obtained by solving the shape equations up to order 3 2<br />

around s ¼ s1.<br />

Shapes in the cylindrical nanochannel<br />

In the absence of adhesion (G ¼ 0), we find that the vesicle has<br />

a prolate shape at small tensions (see top curves of Fig. 5). As the<br />

This journal is ª The Royal Soci<strong>et</strong>y of Chemistry 2008 Soft Matter, 2008, 4, 2463–2470 | 2467


Fig. 5 Calculated shapes of the vesicle inside the cylindrical nanochannel<br />

in a symm<strong>et</strong>ry plane (x,z). Top: G ¼ 0 (no adhesion) for the<br />

reduced tensions s ¼ 0, s ¼ 10, s ¼ 1 10 2 , s ¼ 1 10 3 , s ¼ 1 10 5 (from<br />

top to bottom). Bottom: G ¼ 2 for the same reduced tensions, however<br />

from bottom to top. In both cases, for s ¼ 1 10 5 the shape is almost<br />

a half-sphere.<br />

tension increases, the vesicle tends to a sphere of radius equal to<br />

the radius of the channel—except within a small region of size<br />

xO(k/s) close the channel walls.<br />

The deviation of the shape from a spherical cap of radius<br />

equal to the radius of the channel can be characterized by the<br />

normalized height h ¼ z(r ¼ 0)/a of the top of the vesicle with<br />

respect to the d<strong>et</strong>achment point (see Fig. 6). Increasing the<br />

adhesion energy at fixed tension reduces the height of the<br />

vesicle. When the adhesion energy G equals k/(2a 2 )(G ¼ 1/2),<br />

the vesicle becomes exactly a half-sphere with radius equal to<br />

the channel radius, whatever the tension. Indeed, according to<br />

eqn (11), for G ¼ k/(2a 2 ) the curvature at the d<strong>et</strong>achment point is<br />

1/a and a perfect sphere is always a solution of the shape<br />

equations. In this case, the pressure drop is related to the<br />

tension by the ordinary Laplace law; indeed, the curvature<br />

terms in eqn (2) vanish. At adhesion energies higher than k/(2a 2 )<br />

(G > 1/2), the vesicle develops an oblate shape that tends to<br />

a sphere as the tension increases (see Fig. 6 and bottom curves<br />

of Fig. 5).<br />

Fig. 6 The normalized height h of the top of the vesicle (with respect to<br />

the d<strong>et</strong>achment point) in the cylindrical nanochannel, as a function of the<br />

reduced tension s for different normalized adhesion energies G (G ¼ 0, G<br />

¼ 0.2, G ¼ 0.5, G ¼ 2, G ¼ 10). For G ¼ 1/2, the vesicle is a perfect halfsphere<br />

of radius a whatever the tension.<br />

Fig. 7 The normalized curvature c 0 of the top of the vesicle in a cylindrical<br />

channel, as a function of the normalized tension s, for G ¼ 10.<br />

Continuous line: numerical result; dashed line: analytical approximation<br />

c0 ¼ 1 G/s according to eqn (14). Ins<strong>et</strong>, continuous line: shape of the<br />

vesicle, in a symm<strong>et</strong>ry plane (x,z), for s ¼ 40. The dashed line corresponds<br />

to a spherical cap having the same curvature as the top of the<br />

vesicle.<br />

High tension and strong adhesion in the cylindrical nanochannel<br />

As predicted above, at tensions s [ k/a 2 the vesicle should have<br />

an almost spherical shape of radius given by eqn (14). We have<br />

checked this numerically in the strong adhesion case (G x s), for<br />

which the predicted radius differs significantly from the radius<br />

a of the channel (see Fig. 7). When G > s, we find numerically<br />

that the curvature of the top of the vesicle becomes negative (the<br />

membrane invaginates), in agreement with eqn (14).<br />

Deviation from the naive prediction using the ordinary Laplace<br />

law<br />

We consider the general conical channel case of Fig. 3. The<br />

ordinary Laplace law, that neglects the curvature energy, would<br />

predict a pressure drop dPnaive ¼ (2scosa)/a, corresponding to<br />

a spherical cap of radius a/cosa matching tangentially the conical<br />

channel. In our normalized units, this yields pnaive ¼ 2scosa.<br />

Curvature and adhesion energies introduce the corrections displayed<br />

in eqn (18), and cannot be evaluated analytically for a s<br />

0. Fig. 8 shows the deviation p p naive, in the absence of adhesion,<br />

for different cone angles a. For a cylindrical channel<br />

(a ¼ 0), according to eqn (9), the deviation is p p naive ¼ 1,<br />

independent of the tension. On the other hand, for a conical<br />

nanochannel (a s 0), the deviation p pnaive increases as the<br />

tension increases. Asymptotically, for s [ 1, we find that p<br />

pnaive f s 0.5 ; therefore, the relative error (p pnaive)/pnaive x (p<br />

pnaive)/s tends to zero as 1/Os. At small tensions (s ( 1), the<br />

deviation p pnaive is almost independent of the tension; the ins<strong>et</strong><br />

of Fig. 8 shows the extrapolated deviation in the limit s / 0,<br />

which is maximum for a x 48 .<br />

3.3 Validation of the force equilibrium analysis<br />

Here, we show that the relations previously deduced from the<br />

force equilibrium analysis can be recovered directly from the<br />

shape eqn (22)–(25) and the boundary conditions shown in eqn<br />

(26)–(29).<br />

2468 | Soft Matter, 2008, 4, 2463–2470 This journal is ª The Royal Soci<strong>et</strong>y of Chemistry 2008


Fig. 8 The deviation p 2scosa of the normalized pressure difference p<br />

from the naive Laplace prediction pnaive ¼ 2scosa, in the absence of<br />

adhesion (G ¼ 0) for different cone angles a (from bottom to top, a ¼ 0 ,<br />

a ¼ 3 , a ¼ 10 , a ¼ 30 , a ¼ 45 ). Above the dashed line, the relative<br />

violation (p pnaive)/s is larger than 10%. Ins<strong>et</strong>: normalized pressure p<br />

extrapolated at zero tension as a function of the cone angle a.<br />

Curvature at the d<strong>et</strong>achment point<br />

The d<strong>et</strong>achment condition, eqn (17), can be deduced from the<br />

boundary conditions shown in eqn (28) and (29), since c1 ¼ j 0 (0).<br />

Global general Laplace relation<br />

Eqn (18) can be deduced from the shape eqn (22), using the<br />

boundary conditions shown in eqn (26) and (29), eliminating<br />

g(0).<br />

Local generalized Laplace law at an umbilical point<br />

Eqn (19) can be recovered in the following way. Eqn (22), with<br />

the help of eqn (24), can be rewritten as:<br />

k<br />

r<br />

d<br />

ds ck<br />

ðsin jÞg<br />

þ c ¼<br />

r2 þ ðP1 PÞcos j<br />

2<br />

(36)<br />

since c || ¼ (sinj)/r and c ¼ j 0 . The local generalized Laplace law<br />

(19) is then the limit for r / 0(i.e., s / s 1) of eqn (36). Indeed,<br />

using the L’Hôpital rule, we have:<br />

lim<br />

r/0<br />

1<br />

r<br />

d<br />

ds ck þ c ¼<br />

1<br />

r 0 ðs1Þ<br />

d 2 ðck þ cÞ<br />

ds 2<br />

s¼s 1<br />

¼ 2 H 00 ðs1Þ (37)<br />

sin j<br />

lim<br />

r/0 r ¼ j0 ðs1Þ cos jðs1Þ<br />

r 0 ¼ j<br />

ðs1Þ<br />

0 ðs1Þ ¼cðs1Þ (38)<br />

g<br />

lim<br />

r/0 r ¼ g0 ðs1Þ<br />

r 0 ¼ s (39)<br />

ðs1Þ<br />

where H ¼ ½(c|| + c) is the mean curvature. Here, we have taken<br />

into account that r 0 (s 1) ¼ cos[j(s 1)] ¼ 1 [see eqn (24) and (27)]<br />

and g 0 (s 1) ¼ s, according to eqn (23) with r(s 1) ¼ 0 and c(s 1) ¼<br />

c ||(s 1).<br />

4 Conclusions<br />

In micropip<strong>et</strong>te experiments, the vesicle tension is usually<br />

d<strong>et</strong>ermined by means of the ordinary Laplace relation (3). Here<br />

we have shown that the correct generalized relation, that takes<br />

into account the curvature and adhesion corrections, is eqn (13).<br />

The curvature correction is relevant when the pressure drop DP is<br />

comparable with k/a 3 , as it can occur in nanochannel. For<br />

instance, taking a ¼ 100 nm and DP ¼ 1.1 10 2 Jm 3 , with k ¼ 1<br />

10 19 J and a radius of the free part R [ a, the naive relation<br />

(3) would give s ¼ 5.5 10 6 Jm 2 , while the correct value given<br />

by eqn (13) is s ¼ 5 10 7 Jm 2 , i.e., ten times smaller.<br />

Eqn (13) also allows us to d<strong>et</strong>ermine the tension s in the<br />

presence of an adhesion energy G. Indeed, we have shown that<br />

for tensions s [ k/a 2 the vesicle profile inside the micropip<strong>et</strong>te is<br />

a spherical cap of radius depending on the adhesion energy G<br />

according to eqn (14). This relation can be used to d<strong>et</strong>ermine the<br />

ratio G/s, then s and G can be d<strong>et</strong>ermined separately with the<br />

help of eqn (13).<br />

Note that eqn (13) is an approximation holding when R [<br />

O(k/s), which is usually the case. We have also d<strong>et</strong>ermined the<br />

general relation shown in eqn (21), that is exact for cylindrical<br />

micropip<strong>et</strong>tes whatever the value of R, using an exact generalization<br />

of the Laplace law valid locally at an umbilical point [see<br />

eqn (19)].<br />

These analytical results have been confirmed by numerical<br />

calculations, that also allow us to d<strong>et</strong>ermine the exact profile of<br />

the vesicle and the relation b<strong>et</strong>ween the tension and the pressure<br />

drop in conical geom<strong>et</strong>ries (see Fig. 8). In particular, we have<br />

shown that the relative error on the Laplace pressure is significantly<br />

larger in the conical case. For instance, for a channel of<br />

aperture a ¼ 45 with size a ¼ 1 mm, with k ¼ 1 10 19 J and s ¼<br />

1 10 6 Jm 2 , the true pressure drop is dP ¼ 1.9 J m 3 , almost<br />

40% larger than the value predicted by the ordinary Laplace law.<br />

References<br />

1 F. Iwata, S. Nagami, Y. Sumiya and A. Sasaki, Nanotechnology, 2007,<br />

18, 105301.<br />

2 M. G. Schrlau, E. M. Falls, B. L Ziober and H. H. Bau,<br />

Nanotechnology, 2008, 19, 015101.<br />

3 A. Jahn, W. N. Vreeland, M. Gaitan and L. E. Locascio, J. Am.<br />

Chem. Soc., 2004, 126, 2674.<br />

4 C. Dekker, Nat. Nanotech., 2007, 2, 209.<br />

5 D. T. Chiu, C. F. Wilson, F. Ryttsén, A. Strömberg, C. Farre,<br />

A. Karlsson, S. Nordholm, A. Gaggar, B. P. Modi, A. Moscho,<br />

R. A. Garza-López, O. Orwar and R. N. Zare, Science, 1999, 283,<br />

1892.<br />

6 P. Y. Bolinger, D. Stamou and H. Vogel, J. Am. Chem. Soc., 2004,<br />

126, 8594.<br />

7 L. D. Landau and E. M. Lifshitz, Fluid Mechanics, Butterworth-<br />

Heinemann, Oxford, 2000.<br />

8 J. Darnell, H. Lodish, A. Berk, L. Zipursky, P. Matsudaira and<br />

D. Baltimore, Molecular Cell Biology, Scientific American Books,<br />

W. H. Freeman & Co. Ltd, New York, NY, 1999.<br />

9 W. Helfrich, Z. Naturforsch., 1973, 28c, 693.<br />

10 W. Helfrich and R.-M. Servuss, Nuovo Cimento Soc. Ital. Fis., D,<br />

1984, 3, 137.<br />

11 O.-Y. Zhong-Can and W. Helfrich, Phys. Rev. A, 1989, 39, 5280.<br />

12 See eqn (30) of ref. 11.<br />

13 J.-B. Fournier, Phys. Rev. L<strong>et</strong>t., 2007, 75, 854.<br />

14 P. Galatola and J.-B. Fournier, Phys. Rev. L<strong>et</strong>t., 2007, 75, 3297.<br />

15 E. Evans and W. Rawicz, Phys. Rev. L<strong>et</strong>t., 1990, 64, 2094.<br />

16 W. Rawicz, K. C. Olbrich, T. McIntosh, D. Needham and E. Evans,<br />

Biophys. J., 2000, 79, 328.<br />

17 J. R. Henriksen and J. H. Ipsen, Eur. Phys. J. E, 2004, 14, 149.<br />

18 U. Seifert and R. Lipowsky, Phys. Rev. A, 1990, 42, 4768.<br />

19 J.-B. Fournier and C. Barb<strong>et</strong>ta, Phys. Rev. L<strong>et</strong>t., 2008, 100, 078103.<br />

20 R. Capovilla and J. Guven, J. Phys. A: Math. Gen., 2002, 35, 6233.<br />

21 J.-B. Fournier, Soft Matter, 2007, 3, 883.<br />

22 The opposite convention was chosen in ref. 21.<br />

This journal is ª The Royal Soci<strong>et</strong>y of Chemistry 2008 Soft Matter, 2008, 4, 2463–2470 | 2469


23 See note 14 of ref. 18.<br />

24 See eqn (39) of R. Capovilla and J. Guven, Phys. Rev. E, 2002, 66,<br />

041604.<br />

25 M. Deserno, M. M. Müller and J. Guven, Phys. Rev. E, 2007, 76,<br />

011605.<br />

26 Unless G ¼ k/(2a 2 ), in which case the equilibrium shape is a half-sphere<br />

of radius a (see Sec. 3.2).<br />

27 U. Seifert, K. Berndl and R. Lipowsky, Phys. Rev. A, 1991, 44, 1182.<br />

28 C. Tordeux, J.-B. Fournier and P. Galatola, Phys. Rev. E, 2002, 65,<br />

041912.<br />

29 F. Jülicher and U. Seifert, Phys. Rev. E, 1994, 49, 4728.<br />

30 U. Ascher, R. M. M. Mattheij, and R. D. Russell, Numerical Solution<br />

of Boundary Value Problems for Ordinary Differential Equations,<br />

Prentice Hall, Englewood Cliffs, NJ, 1988.<br />

2470 | Soft Matter, 2008, 4, 2463–2470 This journal is ª The Royal Soci<strong>et</strong>y of Chemistry 2008

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!