Вы находитесь на странице: 1из 14

Theory of biomembrane phase transitions

J. F. Nagle

Citation: The Journal of Chemical Physics 58, 252 (1973); doi: 10.1063/1.1678914
View online: http://dx.doi.org/10.1063/1.1678914
View Table of Contents: http://scitation.aip.org/content/aip/journal/jcp/58/1?ver=pdfcov
Published by the AIP Publishing




















This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded
to IP: 153.132.233.26 On: Thu, 05 Dec 2013 23:44:34
THE JOURNAL OF CHEMICAL PHYSICS VOLUME 58, NUMBER 1 1 JANUARY 1973
Theory of biomembrane phase transitions
J. F. Nagle
Departments of Physics and Biology, Carnegie-Mellon University, Pittsburgh, Pennsylvania 15213
(Received 2 August 1972)
Experiments on the chain melting thermal transition in simple biological membranes are briefly reviewed
and shown to indicate that a microscopic order-disorder model may be appropriate to describe the
thermodynamics of the transition. To test this conclusion further a pair of two dimensional lattice models
(A and B) are introduced and the statistical mechanics is solved exactly using dimer techniques. The
phenomenological parameters required by the models are evaluated from experiments on other systems.
The more realistic of the two models (model A) has a second order transition at a temperature of 353K
compared to 315K for rupalmitoyl-L-o<-lecithin membranes. However, the specific heat peaks do not have
the same shape and the transition is broader for model A than for the experiment. In comparison, the less
realistic model B has a first order transition at 925K, considerably higher than the experimental
transition temperature. From these results it seems likely that the points of disagreement between model A
and experiment may be due to the simplified features of model A. It is concluded that the basic
order-disorder model is likely to be correct for the main transition. There is also a smaller transition at a
lower temperature which is also discussed in terms of an order-disorder model involving the phospholipid
head groups. In this case the calculation shows that the simple model is either wrong or that some
additional features must be added.
I. INTRODUCTION AND EXPERIMENTAL
EVIDENCE FOR AN ORDER-DISORDER
MODEL OF THE TRANSITION
The membrane of a cell provides a permeability
barrier which divides space into an inside and an out-
side of the cell. This barrier is not completely imperme-
able. In the case of nerve axons large numbers of ions
pass through the membrane in a short time when the
axon is depolarized: this dramatic permeability effect
is likely to be connected with special transport facilities
which may constitute only a small percentage of the
membrane, analogous to the windows in a house.
I
,2
Perhaps somewhat less dramatic is the fundamental
matrix of the membrane, analogous to the walls of the
house. In the hope of elucidating the structure of this
fundamental matrix many recent experimental studies
have been performed on simple micro-organisms such
as mycoplasma laidlawii.
3
-
lo
One important result of
these studies is that the fundamental membrane ma-
trix conforms to the basic bilayer model proposed by
Davson and Danielli.u This result justifies turning to
even simpler model systems such as lecithin bilayers
in order to study the structure of the phospholipids in
the bilayer state, particularly the conformational states
of the hydrocarbon chains.
12
-
22
This approach to the
membrane problem is admittedly rather specialized in
that other model membrane systems are ignored, such
as myelin which contains a great deal of cholesterol
chain lengths and degree of unsaturation depends upon
the growth temperature of the mycoplasma and the
composition of the growth medium.s,6,27 The bilayer
arrangement minimizes hydrophobic hydrocarbon-
water contact, and this provides an effectively at-
tractive interaction, via hydrogen bonds and van der
Waals interactions, which keeps the bilayer intact.
FIG. 1. One-half of the basic bilayer. The other half is a mirror
image of the first half on average. The ellipsoids represent the
head groups and the zig-zag lines represent the hydrocarbon
chains of methylene groups terminated by a methyl group. The
chains on the left are in the all-trans conformation. The chain
in the middle has several gauche rotations in the lower half.
Notice that the mean length of chain has shortened compared
to the all-trans conformation and the accumulation of density
in the lower half of the value. The two chains on the right also
have several gauche rotations in the lower half and have the tilt
proposed by McFarland and McConnell.'5 (See text.)
(although see Refs. 23 and 24 for discussion of the There may also be net attractive interactions between
incorporation of cholesterol in phospholipid systems) the charges on the head groups localized on the phos-
and retinal rod disks where the rhodopsin protein is phate group and the choline group in lecithin.
likely to be the most significant structural unit.
25
,26 At low temperatures the hydrocarbon chains are
The mycoplasma laidlawii and lecithin systems will most likely to be in the low energy all-trans conforma-
be called basic bilayers in this paper. A schematic pic- tions. As the temperature is raised there will be some
ture of the basic bilayer is shown in Fig. 1. The hydro- probability of gauche rotations about C-C bonds with
carbon chains consist mostly of CH
2
groups, with an energies typically about 0.5 kcal/mole appearing in
occasional cis double bond. The actual distribution of the chains.
28
However, the steric or hard-core repulsive
252
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded
to IP: 153.132.233.26 On: Thu, 05 Dec 2013 23:44:34
THEORY OF BIOMEMBRANE PHASE TRANSITIONS 253
interactions between chains will prevent most single
bond trans-gauche changes until relatively high tem-
peratures when a cooperative melting. of the hydro-
carbon chains can take place. Such a cooperative
"transition" is observed calorimetrically.3.4.6.12.14 Meas-
urements
6
on mycoplasma laidlawii showed the transi-
tion width to be about 20C centered at about 35C
for a growth temperature of 37C. (It has been empha-
sized by many authors that this transition takes place
at temperatures of prime importance to biological
systems. Apparently membranes need a certain amount
of chain fluidity to be viable.) Recent measurements
22
on synthetic pure lecithins show a much narrower
transition as shown in Fig. 2 with a transition enthalpy
l!.H,,-,9.5 kcal/mole for dipalmitoyl-L-a-Iecithin (DPL).
This figure also shows an anomalous specific heat at
34C which will be discussed in Sec. V. The sharpness
of the main specific heat anomaly at 41C in Fig. 2
indicates the possibility of a first or higher order phase
transition for pure, uniform lecithins and, at least, is
indicative of a strongly cooperative process.
The nature of the transition is most clearly revealed
by x-ray diffraction experiments.6.7.8.12 Below the tran-
sition the hydrocarbon chains produce a sharp diffrac-
tion ring which corresponds to a nearest neighbor chain
distance of 4.8 A. Above the transition this ring is
quite diffuse as in liquid hydrocarbons and corresponds
to distances of roughly 5.3 A.
From these two sets of experiments it seems reason-
able to propose an order-disorder model for which we
write an effective Hamiltonian
H = H att+ Hsteric+ Hrot
where H att is the sum of the attractive interactions
holding the bilayer together and Hswric prevents any
two molecules from being closer together than their
van der Waals or hard core radii. The model is called
an order-disorder model because, as in polymer theory ,29
H
rot
allows each carbon-carbon bond only the trans
conformation with energy 0 or the two gauche confor-
mations with energy E=O.5 kcal/mole. The partition
function is then a sum over discrete states and kinetic
energy plays no role in the model: motion over irans-
gauche rotational barriers is assumed in order to achieve
equilibrium but it has negligible effect on the partition
function and thermodynamic properties.
The order-disorder assumption in our model is rea-
sonable if and only if there is not much motion, i.e.,
if the correlation times To are long. Evidence supporting
this assumption has been obtained by NMR and spin-
labeling experiments above the transition. In particu-
lar, experiments
15
.
16
on multilamellar bilayers indicate
that the correlation times are of the order of 10-
6
sec
for methylene groups nearest the phosphate head groups
and for methylenes near the terminal methyl of the
hydrocarbon chains the correlation times are of the
z

:3
o
Cf)
11..50
o
::E
<
ffi40
25 50
TEMPERATURE, 'c
FIG. 2. The experimental variation with temperature of the
excess specific heat (curve A, right hand ordinates) and the
excess enthalpy (curve B, left hand ordinates) of dipalmitoyl-l-a-
lecithin (DPL) from Hinz and Sturtevant, Ref. 22. The lipid
concentration is 3.88 mg/ml.
order of 10-
8
sec indicating additional fluidity in the
center of the bilayer and suggesting that the attractive
interactions holding the bilayer together operate most
strongly on the polar head groups. These relaxation
times are characteristic of highly viscous liquids and
indicate a perturbed energy due to the motion over
the trans-gauche barrier of roughly erg,
which is negligible compared to kT m, and thereby justi-
fies using the unperturbed order-disorder model. It
may be noted that NMR performed on sonicated
vesicles indicates shorter correlation times
18
.
19
which
may be due to tumbling of the vesicles
19
or more likely
due to the very small radius of curvature (125 A)
which does not allow the hydrocarbon chains to pack
so efficiently and hence leads to greater fluidity.3 Note
that sonication for only 7 min broadens the calorimetric
transition considerably as may be seen from Fig. 9 of
Ref. 3.
The spin labeling experiments give additional infor-
mation abOl!Hhe state of the hydrocarbon chains above
the transition.
1
The end of the chains nearest the sur-
face head groups and which is mostly in the trans state
is tilted away from the perpendicular to the bilayer
surface while the free methyl end, which is more fluid,
is more ,nearly isotropically oriented as shown on the
right hand side of Fig. 1. The reason for this may be
seen by considering the possibility of no tilting. If the
relatively immobile all-trans the chains were
perpendicular to the bilayer surface, then it would
necessarily follow that the more fluid methyl end would
have a higher density than the immobile end because
the gauche bonds decrease the average membrane thick-
ness without any change in area. This state obviously
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions.
Downloaded to IP: 153.132.233.26 On: Thu, 05 Dec 2013 23:44:34
254 J. F. NAGLE
FIG. 3. A particular conformational state for model A. The
solid lines represent the hydrocarbon chain. The dotted lines
represent the edges of the underlying triangular lattice which
are not covered by a chain. The chain on the left is in the all-
trans state with Erot=O and next to it is a chain in the all-gauche
state with E
rot
= 8.. The chain on the far right has rotational
energy 3. and the remaining chain has E
rot
=2 . Several lattice
sites do not have a chain passing through them, so the density
of chain is less than the maximum.
has a higher free energy than the state shown in the
right hand side of Fig. 1 where the density may be
roughly constant or even decreasing toward the fluid
end, depending on the tilt angle which allows a change
in area per molecule at the head group end. This
tilting, due essentially to the nonuniform nature of
the attractive interaction, requires that an extra degree
of freedom be implicitly included in the theoretical
model.
It is interesting to make a rough comparison of the
order-disorder model with the observed calorimetric
transition. Since the transition is very narrow in tem-
perature we may take for the entropy of transition
cal/deg/mole. Let us assume for the
sake of definiteness that most of the entropy tJ.S goes
into disordering the chains in the sense of gauche bonds
i.e., gauche rotations. Using tJ.S""""R InW we can esti-
mate that each molecule has roughly W=e
15
=3.3X10
6
equal energy conformations available to it above T m.
If each molecule had only one gauche bond located
randomly along one of the two chains then the number
of conformations, even ignoring the expected decrease
in number due to steric interactions, is only W = 56. At
least five gauche bonds per molecule are 'needed to give
W'"'-'3 X 10
6
and this is a gross lower bound due to the
neglect of steric interactions. However, with ten gauche
bonds there are of the order of 10
10
configurations and
even more become available for larger n
g
Thus, even
with a substantial lowering of the entropy due to
steric interactions, it seems reasonable that the order-
disorder model has enough entropy to pass this experi-
mental test.
We can also use the enthalpy change tJ.H=tJ.U+
P tJ. V to estimate the number of gauche bonds ng per
molecule. First the PtJ.V term is negligible for P= 1
atm even if the volume change is as large as 25%.
Assume for the moment that the entire contribution
to tJ.U is due to tJ.Hrot Then, the number of gauche
bonds above the transition is ng =9.5 kcal/0.5 kcal= 19.
This seems very high especially since, even ignoring
steric interactions, the thermal probability of a bond
being gauche is only Since there
are only 28 possible bonds to rotate for dipalmitoyl-L-
lecithin (DPL), we would guess that 12 is an upper
bound for n
g
More likely a fair fraction of the increase
in internal energy is used to expand the membrane
against the attractive forces. Preliminary pycnometry
measurements performed by the author give a tJ. V of
0.01 ml for 0.35 g of DPL which yields a tJ.V IV of
about 5% for the hydrocarbon chains, excluding the
head groups. Engelman's x-ray measurements
7
of mem-
brane thickness and rough nearest neighbor distances
of 5.3 A above T m yields tJ.V /V"'-'5%5%. Trauble
and Haynes
30
measure tJ.V/V of only 1.4% but they
prepared their dispersion by sonication and worked
with a very high concentration; their Tm of 44C is
higher than the 41-42C found by others. Using a van
der Waals energy of 0.3 kcal/mole for methylenes at
their van der Waals separation of 4.0 A, using a 6-12
potential, assuming uniform spacings of the molecules,
and a density increase of 5% yields tJ.U
van
der
kcal/mole. Alternatively, we can make use of a pub-
lished formula
3l
for van der Waals interactions between
long chains which give tJ.H"'-'O.l1tJ.H.ublimation for a 5%
expansion from 4.8 A. Using tJ.H.ublimation= 1.84 kcal/
mole CH
2
gives tJ.H=5.7 kcal/mole for 28 CH
2
groups
per molecule. Thus, we can roughly estimate that
ng'"'-'7-8 and is probably between 5 and 12. But the
most important conclusion is that tJ.S and tJ.U are
reasonable quantities in terms of the order-disorder
model in that there are no obvious contradictions. Of
course, a more precise test of the model requires more
theoretical work as well as more refined experiments.
II. SIMPLIFIED MODELS WHICH ARE
EXACTLY SOLVABLE
The model developed in Sec. I is currently impossi-
ble to solve exactly in any reasonably complete sta-
tistical mechanical sense due to the intractable mathe-
matics. The most difficult part of the problem is the
treatment of the hard-core repulsion or H.terio which
obviously is vital to the transition. Although this
author is developing an approximation method to deal
with this, there is another more fundamental approach
which will be pursued in this paper. This approach,
which has been so successful in the statistical mechanics
of phase transitions and critical phenomena, is to sim-
plify the model until it is exactly solvable,32 but keeping
the essentials which, in this case, we consider to be
steric repulsions as well as some kind of rotational and
attractive interactions. This approach usually requires
setting up the problem in less than three dimensions
and imposing homogeneous boundary conditions. The
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded
to IP: 153.132.233.26 On: Thu, 05 Dec 2013 23:44:34
THEORY OF BIOMEMBRANE PHASE TRANSITIONS
255
advantage of this approach is that there is no uncer-
tainty in the details of the mathematical solution. Al-
though the solutions of such modified models are not
expected to agree precisely with experiment, one can
hope to answer such questions as whether or not a
sharp transition is possible in simple order-disorder
models. Also, such solutions may be valuable in guiding
the more uncertain approximate methods.
In this section two models will be described. The
first one, to be called model A, is obviously more
realistic than model B, which will, nevertheless, be
interesting as a comparison to model A.
To describe model A consider the triangular lattice
shown in Fig. 3. The mutually impenetrable liquid
hydrocarbon chains are represented by nonintersecting
solid lines in this two-dimensional lattice model. The
chain on the extreme left of the figure is defined to be
in the all-trans conformation represented in this model
by a vertical solid line. The chain second from the left
is in an all-gauche conformation. Other chain conforma-
tions are shown on the right side of the figure, where
the chains are not packed at maximum density. The
artificial restriction is made that each chain must
occupy at most one site on a given vertical level: thus
chains are prohibited from doubling back on themselves.
Notice, however, that the presence of gauche bonds
requires more cross sectional area and a shorter length
of chain per methylene, as in the real case.
For mathematical convenience the chains are allowed
to be infinitely long: this makes possible a sharp
mathematical transition. Extensive work on finite sys-
tems
33
indicates that finite chains of only 16 methylene
groups would broaden the transition somewhat but
that a considerable specific heat peak would still re-
main near the Tc found for the infinite chain limit.
With infinitely long chains on a lattice model there is
no possibility of the tilting mechanism mentioned in
Sec. I in connection with the spin labeling experiments.
A priori we expect the two-dimensional nature of
the model to raise the transition temperature relative
to comparable three-dimensional models. The reason is
that the excluded volume effect is stronger for this
two-dimensional model in the sense that a chain can
not avoid its nearest neighbors by using the third
dimension. One should specifically avoid the overly
general conclusion that it is harder to produce transi-
tions in two dimensions than in three dimensions: this
is not necessarily the case for excluded volume inter-
actions.
The rotational Hamiltonian H
rot
will be specified as
follows; for each link in the hydrocarbon chain which
is in the vertical direction assign an energy of zero and
for all other links assign an energy E= O.S kcal/mole.
This specification of H
rot
is not as close to the real case
as one might like, but more realistic choices of H
rot
are
not solvable by existing techniques as will be seen in
Sec. III. This specification of H rot does give the correct
energies for the all-trans and all-gauche conformations,
although it gives too little energy to other conforma-
tions. It also builds in a preferred orientation along the
vertical axis, even in addition to that already present
due to the restriction that chains are not allowed to
double back on themselves. However, such a preferred
orientation would be built into a more realistic model
via the boundary conditions due to the head groups
which have been eliminated by taking the chains to
be infinite, so its reappearance in this simplified model
is not unwelcome.
Finally, the attractive interactions, H
att
, will be
only the van der Waals interaction, since there are
now no head groups. Although one has no hope of
solving models with long range r
6
interactions, the
long range nature of the interactions and the softness
of the attractive potential leads one to hope that these
interactions may be taken into account adequately in
a mean field or van der Waals way.34 For example, in
a classical gas one might follow van der Waals by
adding ap2 to the pressure where p is the density.
Equivalently, one may add
_apa::.joO (-ar)r
2
dr
ro
to the internal energy per particle U which then is
used in computing the Gibbs free energy per particle
G= U-TS+PV. In our case the van der Waals energy
has a different form due to the fact that the system
does not expand in the vertical direction. Thus, the
effective form
31
of the interaction between chains is
-ar5. Integrating over all chains in three dimensions
gives an energy dependance of -aro-
3
where 1'0 is the
mean nearest neighbor distance between chains. How-
ever, now the density in three dimensions varies as 1'0-
2
so the density dependence of the van der Waals inter-
action is - ap3/2 where the constant a may be chosen
I to give the experimental energy of sublimation of bulk
. polymethylene chains, which is 1.84 kcal/mole of CH
2
groups.35
There are several alternatives to the - ap3/2 form
for H att. First, one could say that further neighbors
are unimportant and nearest neighbors only gives
_ ap5/2. Also, in more realistic models distant chains
will interact as -ar-
6
due to their finite length, and
this would give -ap2. Finally, since our model is in-
trinsically two dimensional, one may argue that one
should only perform a two dimensional integration
over all chains and further p is now proportional to
1'0-
1
so this gives Hatt= -ap4. However, if one takes
this argument to its logical conclusion, one should re-
place the rl Coulomb's law for point charges by a
logr line charge law: this adds two powers of ,. to
van der Waals forces and gives H att = - al.
There is a reason for choosing the lowest power of
p for H att. A small uniform expansion of the system
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded
to IP: 153.132.233.26 On: Thu, 05 Dec 2013 23:44:34
256
J. F. NAGLE
2n+2
3
5
./6n +2
6n+3>
. 6 ~ + 4
... ..... . ....
6n+5 /
FIG. 4. A particular conformational state for model B. The
solid lines represent the hydrocarbon chain. The dotted lines
represent the edges of the underlying square lattice which are
not covered by a chain. All chains but one are in the all-trans
conformation and the second chain from the right has rotational
energy Erot=E associated with the edge (4n+4, 6n+6) which
crosses from one channel to another.
will yield a higher energy change than many more
probable nonuniform expansions where many of the
neighbors are still near their close packing separations
and a number of larger than average gaps appear. Thus,
as p decreases from close packing, H att increases too
quickly due to the neglect of nearest neighbor correla-
tions. This may be compensated by using the lower
power of p in H att in our simplified model.
We turn next to model B. Consider the square lattice
indicated by dotted lines in Fig. 4. The lipid hydro-
carbon chains are represented by solid lines in this two
dimensional lattice model. Three of the chains in Fig. 4
are defined to be in the all-trans conformation repre-
sented in this model by ninety degree angles between
successive bonds. One of the chains has the model
equivalent of two successive gauche bonds. Notice that
because chains are not allowed to double back there
are only two alternatives for "rotations" in this model,
one trans and one gauche, instead of one trans and two
gauche as in model A.
The rotational Hamiltonian H
rot
for model B is speci-
fied as follows; links in hydrocarbon chains which
occupy the bonds whose end points in Fig. 4 are both
odd (such as 2n+S, 4n+3) or both even (such as 4,
2n+4) have energy E>O and all others have energy O.
The ground state conformations of chain will remain
the all-trans conformations positioned in the zero energy
"channels." A chain may disorder by expending E in
energy to break out of its channel into a new one,
provided that it does not run into another chain. As
in model A an orientational effect is built in, but there
is also the undesirable channeling effect in model B as
well as the shortage of gauche rotations. The attractive
Hamiltonian is included in model B in the same way
as in model A.
III. MATHEMATICAL SOLUTIONS
The models introduced in Sec. II were chosen because
their states are isomorphic to the states of close packed
lattice dimer models for which there are known methods
of solution. The states in a close packed lattice dimer
model may be pictured as follows: Every lattice site
is paired to one and only one of its nearest neighbor
lattice sites by a dimer which may be thought of as a
diatomic molecule whose atoms occupy the lattice sites
and whose bond is coincident with the edge joining the
two lattice sites.
The transformation of model A to a dimer model is
accomplished as follows. A new lattice is constructed
from the old one by first decorating each vertical edge
on the original triangular lattice with two new vertices
placed one third of the edge length from either end.
Second, each new vertex is joined by new edges to the
four closest new vertices. The new lattice is shown by
dotted and solid lines in Fig. 5 and is topologically
equivalent to a square lattice. Also shown in dashed
lines are the same hydrocarbon chains shown in Fig. 3.
Those edges in the new lattice drawn with solid lines
are meant to portray the dimers which join two vertices
or sites in the new lattice by a solid bond. Each state
of hydrocarbon chains on the original triangular lattice
transforms into a state of dimers on the new lattice
according to the rule; (a) Each nonvertical chain link
crosses one edge on the new lattice on which a dimer is
placed. (b) Each vertical chain link is coincident with
and covers one of the short vertical edges on the new
lattice on which a dimer is placed. (c) Any vertex on
the new lattice not already covered by a dimer by (a)
or (b) is now covered by a dimer placed on the long
vertical edge incident to the vertex. It may easily be
verified that part (c) of the rule may always be carried
out, that is, the other vertex on the long vertical edge
is also not covered by a dimer. Therefore, the new
lattice is filled uniquely with nonoverlapping dimers
given a chain state on the original lattice. Conversely,
FIG. 5. The dimer state corresponding to the chain state shown
in Fig. 3. The dimers are represented by solid lines joining pairs
of vertices on the new lattice. The dotted lines are edges on the
new lattice not covered by dimers. The dashed lines show the
original chains; the originallattice has been suppressed.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions.
Downloaded to IP: 153.132.233.26 On: Thu, 05 Dec 2013 23:44:34
THEORY OF BIOMEMBRANE PHASE TRANSITIONS
257
given a dimer state on the new lattice the rule may be
reversed to obtain a unique chain state, so the trans-
formation is a 1-1 correspondence between states in the
two systems.
Let the partition function for the dimer model be
Z(x, y) = 2: xnym= 2: g(n, m)xnym, (3.1)
states n,m
where n is the number of nonvertical dimers and m is
the number of dimers on the long vertical bonds and,
of course, nand m are state dependent. If N is the
number of lattice sites, then the mean density of x
dimers and the mean density of y dimers are, respec-
tively,
p.,= (x/N)(d InZ/dx) and PIl= (y/N)(d InZ/dy) ,
(3.2)
where p", and py have maximum values of 1/2. Let
x=exp(-e/kT) and y=exp(-o/kT). The free energy
of the dimer system is
where the mean energy U d= ep",+op" and the mean
entropy Sd=k lng(ii, iii), where ii=Np", and iii = Npli.
To complete the correspondence with the chain model
one obviously takes e=O.5 kcal/mole which is the rota-
tional Hamiltonian parameter. The parameter 0 in the
dimer model becomes a free parameter in the chain
model; variation of 0 allows P1l' the concentration of
"vacuum" in the chain model to be varied. Of course,
the density of hydrocarbon chain is p= pcp- p" where
Pcp=! is the close packed chain density. The internal
energy of the chain model per chain link (methylene
FIG. 6. The dimer state corresponding to the chain state
shown in Fig. 4. The dimers are represented by solid lines joining
pairs of vertices on the new lattice. The dotted lines are edges
on the new lattice. The dashed lines show the original chains;
the original lattice has been suppressed.
FIG. 7. The x-y activity
plane for the dimer model
corresponding to model A.
The lines y = 12x separate
regions of different analytic
behavior and thus locate
the phase transitions of the
dimer model.
group) is just
y
m
x
(3.4)
from the rotational and attractive Hamiltonians. The
chain model entropy per chain link is also simply
Sc= Sd/ p where Sd is calculated from (3.3). Thus, if
one can calculate the dimer model free energy F d and
the densities p", and p" as functions of x and y, one can
then calculate the Gibbs free energy per link in the
chain model
(3.5)
For fixed x (fixed T) the minimum in G
c
as a function
of y determines the thermodynamically stable state of
the chain model.
Model B can also be transformed to a dimer model
in an analogous way except that the geometry is differ-
ent. A new lattice is constructed from the old one by
expanding each vertex on the old lattice into two new
vertices joined by a vertical edge as is shown by the
dotted-and solid lines in Fig. 6. Also shown in Fig. 6 in
dashed lines are the same hydrocarbon chains shown
in Fig. 4. Each state of hydrocarbon chains on the
original lattice transforms into a state of dimers on the
new lattice according to the rule; (a) Ifa hydrocarbon
chain crosses a horizontal edge on the new lattice, then
that edge on the new lattice is covered by a dimer
represented by a solid line. (b) A vertical edge on the
new lattice is occupied by a dimer if and only if no
adjacent horizontal edges are occupied by dimers ac-
cording to (a). As with model A the correspondence is
1-1 between chain states and close packed nonover-
lapping dimer states.
Analogous to model A the dimer activities for model
B will be chosen to simulate the rotational Hamiltonian
energies. This is easily done by assigning activities
x= exp( -e/ kT) to those horizontal edges which sepa-
rate the channels in model B. (See Fig. 4.) The hori-
zontal edges in the channels have activity 1 and the
vertical edges have the free variable "vacuum" activity
y. From this point forward the procedure for solving
model B is identical to the one for solving model A. At
this point it may be worth noting that one could simu-
late a much more realistic model B which does not have
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions.
Downloaded to IP: 153.132.233.26 On: Thu, 05 Dec 2013 23:44:34
258 J. F. NAGLE
y
m
x
FIG. 8. The x-y activity phase for the dimer model correspond-
ing to model B. The lines y=xl and x+y=l separate regions
of different analytic behavior and thus locate the phase transi-
tions of the dimer model.
the channeling feature provided that one allowed addi-
tional interactions between dimers as well as the simple
single dimer activities and the hard core repulsions.
But such dimer problems are not solvable by known
methods, so models A and B have been chosen because
they are isomorphic to simple dimer problems with
single dimer activities.
The technique of solving two dimensional dimer
models is well understood, so the mathematical details
are relegated to the Appendix. For model A there are
three analytically separate regions in the x, y activity
plane as shown in Fig. 7. The solutions for these dif-
ferent regions are
Region I;
p",=O=py=lnZ(x, y),
Region II; 1 +2x,
p",=o, N-IlnZ(x, y) = (1/2) lny,
Region III; 1+2x,
p.,=t(1-1I'-I cos-
I
{ 1-[(1 +y)2/2(x2+y)]}),
(3.6a)
(3.6b)
py=t- (tp",)=F (411')-1 cos-
I
{1- [(1-y)2/2x2]}, (3.6c)
where (-) is used if 1 and (+) is used if 1 and
the argument of the inverse cosine is always in the
first or second quadrants. The function py(x, y) is
analytic on the line y= 1. The functions p", and py are
continuous across the lines y= 12x which divide the
different analytic regions but the first derivatives of
p", and PII are not only discontinuous but in Region III
they are actually divergent as one approaches these
lines in the x, y plane. The partition function in Region
III is a more difficult integral than p", or py and has not
been calculated analytically. It suffices for our pur-
poses to compute lnZ(x, y) by numerical integration
from py and the value of InZ on the line y= 1+2x. As
a check, the value at y= 1-2x for x< t was compared
to the exact value. The numerical integration interval
was chosen small enough to give an error less than
10-
6
which was quite sufficient to guarantee that no
qualitative errors and only inconsequential quantita-
tive errors resulted from this numerical integration.
For model B there are four separate analytical re-
gions which are shown in Fig. 8. The solutions for
these different regions are .
Region I;
p",=o, N-IlnZ(x, y) =t lny, (3.7a)
Region II;
py=o, p",=t, N-IlnZ(x, y) =t lnx,
Region III;
p",=O=py=N-IlnZ(x, y),
Region IV;
Pll= (1/211') cos-I[(x
2
+1-y2)/2x],
p",= (1/2) (l-Pll)
(3.7b)
(3.7c)
- (1/211') tan-I { [(x+ 1) / (x-l)] tan1l'Pll}' (3.7d)
As with model A the first derivatives of p", and Pll be-
come infinite as one approaches any of the region
boundaries from Region IV. InZ in Region IV was
also obtained by numerical integration from Pll'
IV. THERMODYNAMIC BEHAVIOR OF
MODELS A AND B
There are three parameters in our models, the rota-
tional trans-gauche energy E, the strength of the van der
0.168 2 1
Cp
0.112
keal
deg mole
0.056
350 400
2
FIG. 9. Specific heat C" vs T for model A for two cases; curve
1 for the. case a = 00 = P with T ... = 363K and curve 2 for the
case a=5.2 kcal/mole and P=4 atm for which T ... =353K.
C" is given in kcal/mole/deg where a molecule has been taken
to have 28 rotatable bonds.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded
to IP: 153.132.233.26 On: Thu, 05 Dec 2013 23:44:34
THEORY OF BIOMEMBRANE PHASE TRANSITIONS
259
Waals interaction a, and the pressure P. The param-
eter E is estimated from experiments independent of
membranes
28
to be 0.5 kcal/mole of gauche bonds. From
a(pc.p)3/2= a( 1/2)3/2= 1.84 kcal/mole CH
2
, again deter-
mined from independent experiments,35 one estimates
the strength of the attractive van der Waals forces to
be a= 5.2 kcal/mole CH
2
The value of P to be used in
the P / p term in the free energy per methylene group
is very small for pressures near atmospheric: for one
atmosphere P=0.0002S kcal/mole CH
2

Before presenting results for physical values of the
parameters let us consider the behavior of these models
if the attractive forces were infinitely strong, i.e., a
and/or P are very large, but not large enough to over-
come the hard core repulsions. Then the models are
constrained to maximum density. Model B exhibits
completely uninteresting behavior under these circum-
stances: it remains completely ordered in the all-trans
conformation with zero specific heat for all temper-
atures. In contrast model A shows a sharp second order
phase transition at a temperature kT
c
=e/ln2=0.7214
kcal or T=363K. The specific heat is zero for T<Tc 1
and diverges as (T-T
e
)-1/2 as T approaches Te from
above Te. The specific heat is plotted in Fig. 9. The
trajectory of the model in the activity plane in Fig. 7
is just the x axis and the phase transition occurs at
x= 1/2 when the model passes from Region I to
Region III.
The reason for this difference in behavior of the two
models is directly related to the extra gauche rotation
in model A as compared to model B. This can be seen
by the following simple argument which computes T ",.
The free energy, energy, and entropy are all zero for
the ordered all-trans conformation. Consider a slightly
disordered conformation with a gauche rotation of one
bond in an otherwise all-trans chain. The change in
free energy is F=e-kT Ing where g= 2 for model A
and g= 1 for model B. Of course, in models A and B at
maximum density the introduction of one gauche bond
requires an infinite sequence of additional gauche bonds.
The interested reader can show that each gauche rota-
tion in the sequence can be chosen in two ways for model
A and in only one way for model B. The argument
therefore shows that there will be a spontaneous dis-
ordering of the ground state conformation when kT=
E/lng, in agreement with the exact calculations. Al-
though this argument is no substitute for an exact
calculation, it does provide insight into the relation
between the number of gauche rotations per bond and
T m in the strong attractive binding limit.
Next, consider the opposite extreme when P=O=a.
In Fig. 10 is shown the free energy per methylene
group versus density for model A for two values of x,
i.e., at fixed T. The minimum in the free energy which
gives the stable state occurs for zero density which
simply reflects the obvious result that the membrane
decomposes if there are no attractive forces or pressure
o ~ - - - - - - - - - - - - - - - - - - - - - - -
- 0.1
S'l<iy.> - 0.3
-0.4
G
-0,5
- 0,6
- 0.7
- 0,8
- 0.9 L-__ --L ____ .L.-__ --::-'-.,.--__ ::-'-:-__ _=_'
o 0.5
FIG. 10. The free energy per methylene group versus density
for model A with no attractive forces (a=O=P) for two tem-
peratures. The upper curve is for T = 250
0
K and the lower curve
is for T = 500
o
K.
to hold it together. Notice that the free energy per
methylene in the low density limit p-?O is just the
free energy per link for an isolated chain, namely
- kT In (1+ 2x). It is also useful to calculate the slope
dG/dp of the curves in Fig. 10 at p=t,
dG(p= t) / dp= kT( t Iny-lnZ) / p2. (4.1)
When x< t this becomes
dG/dp= 2 kT In(1-2x). (4.2)
When x>t this derivative becomes infinite. For model
B Fig. 11 shows a similar plot at two temperatures.
Although the previous case is trivial, the accompany-
ing figures 10 and 11 are valuable in illustrating the
effects of adding a pressure and the attractive inter-
action which have both been taken as simple functions
of the density. Consider pressure first. For low T and
large P the negative slope of Pip at p=t is larger than
the positive slope of Gp-O vs p and it is fairly easy to
see from Figs. 10 and 11 that the thermodynamically
stable states for both models have p=t. As T is raised
the slopes eventually become equal and this defines
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions.
Downloaded to IP: 153.132.233.26 On: Thu, 05 Dec 2013 23:44:34
260 J. F. NAGLE
0
-0.02
a=5.2 kcal/mole CH
2
and P= 1 atm. For model A the
transition is second order with T m = 353K given by
...;r the modification of (4.3) proposed at the beginning of
~ \ the last paragraph. The singularities at T m are of square
J ,,,,in root type. The specific heat is shown in Fig. 9. The
~ transition temperature is a bit lower than for the infi-
nite pressure case and the specific heat at higher tem-
peratures is greater due to the continual decrease in
density. For model B the transition is first order with
a discontinuous volume change 6. V /V of about 5%,
but the transition temperature is very high, T m = 925K,
-0.04
-0.06
-0.08
G
-0.1 0
-0,12
-0,16
0.5
0,0
FIG. 11. The free energy per methylene group versus density
for model B with no attractive forces (a=O=P) for two low
temperatures, T= 125K and 175K proceeding from top to
bottom on the graph. For T= 125K the concavity, i.e., the
second derivative, changes sign near p=O.3.
the transition temperature, T m. For model A it is
It is worth mentioning some additional cases briefly.
For model A with a= 1.0 kcal/mole CH2 and P= 1 atm
the transition is first order with Tm=231K and with
a discontinuous change in density 6.V /V = 7.3%. Al-
though the transition is first order, there is an anoma-
lously large increase in C
p
as T approaches T m from
above T m just as for the previous second order transi-
tion, except that the first order transition intervenes
before C
p
actually diverges. In model B the discontinu-
ous first order effect becomes larger as a is reduced and
Tm becomes smaller. For Tm about 300
o
K, 6.V/V,.....,
300% so the transition is best described as a complete
melting transition in which the system dissipates. Also,
for model A the attractive van der Waals interaction
- abpb was varied to include b = 2 and 3: the over-all
effect was not large enough to concern us,
As can be seen by comparing Fig, 9 and Fig. 2 the
specific heat curves of model A have a considerably
different shape than the experimental curve. This is
not surprising when one recalls that (a) two dimen-
sional models usually have too much specific heat on
the high temperature side of the transition
36
and (b)
the infinite chain model specific heat would undoubtedly
be rounded by finite instead of infinite chains. It is
expected that' the shape of C
p
vs T is much more sensi-
given by
2kT
m
In(1-2x".) =4P
(4.3) tive to such things than are overall changes in enthalpy
and density as one goes through the transition. Un-
fortunately, the high temperature tails in C
p
make it
difficult to choose a temperature Tl at which the tran-
using (4.2). For higher temperatures the density de-
creases and the rotational disorder increases. Just as
was the case for model A constrained to maximum
density, the thermodynamic functions have square root
singularities as T approaches T m from above T m.
One can follow the same line of reasoning for van der
Waals forces and calculate a transition temperature
for model A given by (4.3) with 3a/21/2 added to the
right hand side. However, there is another possibility
which can negate this conclusion. This other possibility
is most conspicuous for model B at low temperatures.
From Fig. 11 we see that GP==0=4 vs p is actually con-
cave downwards at p=t and only becomes concave
upwards at much lower values of p. Since U
att
= _ ap3/2
is always concave downwards, this means that the
minimum in G must jump discontinuously from p= t
to a lower value of p as a is increased. Thus, the transi-
tion is first order,
The physical cases can now be discussed for which

II
\


FIG. 12. Each solid dot represents the projection of a hydro-
carbon chain onto a plane parallel to the membrane. The dots
form a hexagonal, i.e., triangular array. The solid lines represent
the head groups and the bonds which join pairs of hydrocarbon
chains. The assymmetry of the molecules is represented by
arrows.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions.
Downloaded to IP: 153.132.233.26 On: Thu, 05 Dec 2013 23:44:34
THEORY OF BIOMEMBRANE PHASE TRANSITIONS 261
TABLE 1. Transition quantities for model A.
a=5.2 kcal mole-
1
CH2 a= 1.0 kcal mole-
1
CH
2
a,P=co,p=i P=0.001 kcal mole-
1
CH
2
P=0.001 kcal mole-
1
CH
2 Exptl
Tm (in OK) 363 353 231 315
Tl (in OK) 372 366 267 317
C,,(T
1
) kcal deg-
1
'mole-
1
0.087 0.084 0.084 rvO.015
AVIV 0 0.004 0.31 0.05
(0.073)a
AH kcal mole-
1
2.0 2.8 8.0 9.7
(2.8)
AU rot kcal mole-
1
2.0 2.5 4.7 '"'-'4.0
(1.8)
a Quantities in parentheses for the case a= 1.0 are the discontinuities in the first order transition. Each molecule has been considered
to have 28 rotation degrees of freedom.
sition can be said to be "finished." In Table I, Tl has
somewhat arbitrarily been chosen to be that temper-
ature at which C
p
has dropped to about 0.085 kcal
deg-1mole-
1
. This Tl is also the temperature at which
roughly 90% of the enthalpy change has occurred, if
one draws in a background level to eliminate the
roughly constant very high temperature enthalpy
changes. Table I then shows the relative volume
change, IV, the change in enthalpy and the
change in rotational energy all changes occur-
ring between T m and T
1
Since the PV term is nearly
negligible, the difference between and may
be taken to be the energy required to expand the lattice
against the attractive forces.
Table I shows that the width of the experimental
c
r---------------- -----,
I Y 1 I
I I
I I
I I
I
I
I
I
I
I
I
I
I
I x
I
I
I
a
c
y
x
d
L ____
y
----------,
b
a
c
FIG. 13. The unit cell for dimer model A enclosed in dashed
lines. The choice of arrows, which satisfies Kasteleijn's clockwise
odd theorem, governs the signs of the entries in M(</>I, </>2)' The
</>1 direction is taken horizontally and the </>2 direction vertically.
tranSItIOn is narrower than any of the three cases
calculated for model A. This could be blamed on the
two dimensional nature of the model as mentioned in
(a) of the last paragraph, but one must also remember
that (b) will tend to broaden the transition. The vol-
ume change for model A is a bit small for the physical
case a= 5.2 and this is reflected in the small and
the fact that is almost as large as Clearly,
a somewhat smaller value of a would result in better
agreement with all the experimental quantities except
that the width of the transition would increase. Because
of the rough way in which the van der Waals inter-
actions are treated, there should be no binding com-
mitment to a= 5.2. However, in view of the other
simplified features of the model, there is no sense in
pursuing numerical agreement too far. Rather, we
should be satisfied that the simple but reasonably
realistic order-disorder model A produces a transition
of the experimental magnitude near the real T m, while
the less realistic model B gives much poorer agreement
with experiment.
V. LOWER TRANSITION
In this section let us consider briefly the lower transi-
tion which occurs at 34C in DPL.22 In Ref. 22 this is
called the "pretransition" in analogy to the pretransi-
tion in long chain paraffins which is a transition from
orthorhombic or lower symmetry to hexagonal symme-
try and involves the rotation of the entire hydrocarbon
chain about its long axis as well as a lattice expansion.
The enthalpy of the paraffin for 28
carbons is 8.5 kcal/mole compared to the DPL lower
transition enthalpy of 2.3 kcal/mole and only 1.1
kcal/mole for dimyristoyllecithin (DML). Certainly
rotation about the long axis of DPL is more difficult
than in the paraffins because of the bonding of two
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions.
Downloaded to IP: 153.132.233.26 On: Thu, 05 Dec 2013 23:44:34
262 J. F. NAGLE
1
c
I
d
r-- ~ - - - - - - -
I I
b
11 x b : 1
I
a
I
I
I
I
Y
I
Y
I
I
,
1:
I
c x 11
d
,
d
I
I
I I L __ _______ ...J
Y
Y
a
c
Y
FIG. 14. The unit cell
for dimer model B en-
closed in dashed lines
where the meaning of
the arrows is explained
in the caption to Fig. 13.
hydrocarbon chains by the head group, so the analogy
may not be too fruitful. Indeed, in analyzing their
data Hinz and Sturtevant
22
assume that this is not the
nature of the transition.
Steim
3
noted that the lower transition is affected by
the composition of the aqueous solution and therefore
attributed it to the head groups. The following picture
was suggested to the author by D. M. Engelmann in
private discussion. In Fig. 12 a projection onto the
plane of the membrane shows the all-trans hydrocarbon
chains in a two dimensional array which may be taken
to be the triangular or hexagonal lattice. As shown in
Fig. 12, the head groups constitute an unsymmetrical
bond between pairs of hydrocarbon chains. Suppose
that in the low temperature phase a definite order
prevails among these bonds. Such order could be in-
duced by interactions between the charged head groups
and would lower the symmetry from hexagonal. Sup-
pose that the transition involves a breaking-up of this
order with a typical state as shown in Fig. 12. Such a
transition would involve rotation of the lecithin mole-
cule about its long axis, but it would seen that such
rotations would have fairly long correlation times so
that an order-disorder model may again be appropriate,
where the potential minima are located at the six
multiples of sixty degree rotations of each molecule.
Of course, no two molecules are allowed to occupy the
same locations, so such rotations are highly cooperative.
Clearly the states of the order-disorder model are
isomorphic to the states of a dimer model on the two-
dimensional triangular lattice except that the dimers
are now unsymmetric. Of course, the dimers must be
interacting (in addition to the hard core repulsion) to
produce an ordered state and we do not know how to
solve such models. Nevertheless, it is always instruc-
tive to calculate the entropy of the model at very high
temperatures, where the nonsteric interaction can be
ignored. This gives an upper bound on the transition
entropy !:!.S or enthalpy !:!.H = T
1m
!:!.S. Fortunately, this
calculation has been performed many times.
33
The
answer for symmetric dimers in S 00 = R In (2.3565 . ) .
For unsymmetric dimers a simple factor of 2 per dimer
is included to give Soo=R In(4.713 .. ) =3.08 ... cal
deg-
1
or Hoo-Ho=0.946 kcal/mole.
The order-disorder model gives too little entropy
for DPL. It may be noted that DML and DSL have
1.1 and 1.4 kcal/mole, respectively,22 somewhat closer
to the order-disorder model but still larger. When one
considers that the actual entropy under the transition
in the order-disorder model may be only half as much
as Soo, it becomes clear that the model fails the entropy
test. It is possible that the basic picture of the transi-
tion may still be correct but that additions to the
model must be made, such as the effects of lattice
expansion and the introduction of internal degrees of
freedom of the individual head groups. 13C resonance
on the choline group 21 supports this latter possibility,
although the experiment was performed on sonicated
DPL vesicles and the lower transition could not be
distinguished from the main transition. Thus, depend-
ing on one's point of view the fact that Soo for the
order-disorder model is of the correct order of magni-
tude, though a bit small, may be considered favorable
to the basic order-disorder picture of the lower transi-
tion.
ACKNOWLEDGMENTS
The author wishes to thank D. M. Engelman, J. M.
Sturtevant, and J. Prestegard for an introduction to
the subject, for communication of results before publi-
cation and for useful and illuminating discussions. The
author wishes to thank G. R. Allen and R. B. Griffiths
for useful discussions of the theoretical models. Finally,
the financial support of the Alfred P. Sloan Foundation
during the early part of the work and the National
Science Foundation Grant GP-21093 are gratefully
acknowledged.
APPENDIX
In this appendix some of the mathematical details of the solutions of the dimer models corresponding to models
A and B are presented. The general theory of obtaining such solutions has been reviewed by Montroll.
39
Some
models similar to, but not the same as our models, were originally solved by Kasteleijn.
40
The logarithm of the
partition function per lattice site may be written
InZ 1 ( 1 )2 1
2
.. 1
2
.-
Ii = 4 211" Re 0 #1 0 d4>2lng(cp1, c/>2),
(A1)
where! g(CP1, cI>2!2=detM(CP1, c/>2) which is obtained following Kastelijn's method.
40
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions.
Downloaded to IP: 153.132.233.26 On: Thu, 05 Dec 2013 23:44:34
THEORY OF BIOMEMBRANE PHASE TRANSITIONS
263
For model A with unit cell shown in Fig. 13
0 x[1-exp( -i-Pl) J -1 +y exp(i4>2) 0
x[exp( i4>t) -1J 0 0 y-exp(i4>2)
M(4)l,4>2)=
l-y exp( -i4>2) 0 0 x[1-exp( -i4>t) J (A2)
0 x[exp( i4>1) -1J 0
and
g(4)l, 4>2) = 2r(1-cos4>l) +2y_y2 exp( (A3)
Using Eq. (3.2) we have
(A4)
and
(AS)
The -P2 integrals in (A4) and (AS) can be performed by changing variables to w= ei4>2 and using the calculus of
residues. The different analytic behavior which appears in the different regions in the x-y plane in Fig. 7 is due
to the poles of the denominators in (A4) and (AS) crossing from inside to outside of the unit circle. For example,
for y< lone pole is always inside the unit circle and the second pole is inside the unit circle iff 2r(1-cos-Pl) <
(1-y)2, This latter condition holds for all-Pl in Region I of Fig. 7 and for only some -PI in Region III. The inte-
gral over 4>2 in (A4) and (AS) then turns out to be simply trigonometric rather than the generally expected elliptic
type due to cancellations between the denominator and the numerator. The results for Po; and Pu are given in (3.6).
Similarly for model B one has for the unit cell in Fig. 14
(A6)
and
Using (3.2) we have
(A7)
and
Pu=ti r" #1 r2o' d4>2 . (A8)
2'11' J 0 J 0 y2+ (r- 2x cos4>l+ 1) e-'</>2
The -P2 integrals in (A7) and (A8) are easily performed using w= e
i
</>2 and the calculus of residues. The -PI integral
in (A8) is then trivial and the one in (A7) can be found in standard tables. Again the results are found in the
text in (3.7).
'B. Hille, Prog. Biophys. Mol. BioI. 21, 1 (1970).
2K. S. Cole, Membranes, Ions and Impulses (California, 1968),
(California, Berkeley, California, 1968), Part V.
3J. M. Stem, Adv. Chern. Ser. 84, 259 (1968).
4J. M. Stem, M. E. Tourtelotte, J.C. Reinert, R. N.
McEthaney, and R. L. Rader, Proc. Natl. Acad. Sci. USA
63, 104 (1969).
'D. L. Melchior, H. J. Morowitz, J. M. Sturtevant, and T. Y.
Tsong, Biochim. Biophys. Acta 219, 114 (1970).
6D. M. Engelman, J. Mol. BioI. 47,115 (1970).
7D. M. Engelman, J. Mol. BioI. 58, 153 (1971).
8M. H. F. Wilkins, A. E. Blaurock, and D. M. Engelman,
Nature (Lond.) 230, 72 (1971).
9W. L. Hubbell and H. M. McConnell, Proc. Natl. Acad. Sci.
USA 64, 20 (1969).
10M. E. Tourtellotte, D. Branton, and A. Keith, Proc. Natl.
Acad. Sci. USA 66, 909 (1970).
11J. F. Danielli and H. Davson, J. Cell Physiol. 5,495 (1935).
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded
to IP: 153.132.233.26 On: Thu, 05 Dec 2013 23:44:34
264
J. F. NAGLE
12 D. Chapman, R. M. Williams, and B. D. Ladbrooke, Chern.
Phys. Lipids 1,445 (1967).
13M. c. Phillips, R. M .. Williams, and D. Chapman, Chern.
Phys. Lipids 3, 234 (1969).
14B. D. Ladbrooke and D. Chapman, Chern. Phys.
Lipids 3, 304 (1969).
IsB. G. McFarland and H .. M. McConnell, Proc. Natl. Acad.
Sci. USA 68, 1274 (1971).
16S. I. Chan, G. W. Feigenson, and C. H. A. Seiter, Nature
(Lond.) 231,110 (1971).
17S. I. Chan, C. H. A. Seiter, and G. W. Feigenson, Biochem.
Biophys. Res. Commun. 46, 1488 (1972).
18A. F. Horwitz, W. J. Horsley, and M. P. Klein, Proc. Natl.
Acad. Sci. USA 69, 590 (1972).
19E. G. Finer, A. G. Flook, and H. Hauser, Biochim. Biophys.
Acta 260, 59 (1972).
20A. G. Lee, J. M. Birdsall, Y. K. Levine, and J. C. Metcalfe,
Biochim. Biophys. Acta 255, 43 (1972).
21y' K. Levine, N. J. M. Birdsall, A. G. Lee, and J. C. Metcalf,
Biochemistry (Wash., D.C.) 11, 1416 (1972).
22H. Hinz and J. M. Sturtevant, J. BioI. Chern. 247,
6071 (1972).
23H. Hinz and J. M. Sturtevant, J. BioI. Chern. 247,3697 (1972
24J. E. Rothman and D. M. Engelman, Nature (Lond.) 237, 42
(1972).
2SC. R. Worthington, Physiol. Soc. Federation Proc. 30,57
(1971).
26C. R. Worthington and A. E. Blaurock, Biophys. J. 9,970
(1969).
27S. Razin, M. E. Tourtellatte, R. N. McElhaney, and J. D.
Pollack, J. Bacteriol. 91, 609 (1966).
28A. Abe, R. L. Jernigen, and P. J. Flory, J. Am. Chern. Soc.
88, 631 (1966).
29M. Volkenstein, Configurational Statistics of Polymeric
Chains, trans. S. N. Timasheff and M. J. Timasheff
(Interscience, New York, 1963).
3OH. Trl1uble and D. H. Haynes, 7, 324 (1971).
31L. Salem, J. Chern. Phys. 37,2100 (1962).
32M. E. Fisher, Lectures in Theoretical Physics (Colorado U.
P., Boulder, Col., 1965), Vol. VIIC, p. 1.
33A. E. Ferdinand and M. E. Fisher, Phys. Rev. 185,832
(1969).
34The author has stated that the critical behavior of 1/ r6
interactions in three dimensions should be like the critical
behavior of short range interactions, rather than like the ver)
long range mean field interactions. J. Phys. C 3, 352 (1970).
However, in this paper we are only interested in the gross
features of the transition, and not the niceties of a critical
point.
3sF. W. Billmeyer, J. Appl. Phys. 28, 1114 (1957).
36c. Domb, Low Temperature Physics (Plenum, New York,
1965), Vol. LT9 (Part B), p. 637. See especially p. 644.
37A. A. Schaerer, C. J. Busso, A. E. Smith, and L. B. Skinner,
J. Am. Chern. Soc. 77, 2017 (1955).
38J. F. Nagle, Phys. Rev. 152, 190 (1966).
39E. Montroll, Applied Combinational Mathematics, edited by
E. F. Beckenbach (Wiley, New York, 1964), Chapter 4.
40p. W. Kasteleijn, J. Math. Phys. 4, 287 (1963).
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded
to IP: 153.132.233.26 On: Thu, 05 Dec 2013 23:44:34

Вам также может понравиться