Physics of DNA 1.0 PDF

1
PHYSICS OF DNA
Rudolf Podgornik
2 Rudolf Podgornik
3
Contents
1 PHYSICS OF DNA 1
1 Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 HELICAL CURVES . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1 Mathematical description of spatial curves . . . . . . . . . . . . 7
2.2 Material frame of reference . . . . . . . . . . . . . . . . . . . 10
2.3 Darboux vector . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 Euler angle parametrization . . . . . . . . . . . . . . . . . . . 12
2.5 Single and double helix . . . . . . . . . . . . . . . . . . . . . . 14
2.6 Călugăreanu-White-Fuller theorem . . . . . . . . . . . . . . . 20
3 MOLECULAR STRUCTURE OF DNA . . . . . . . . . . . . . . . . . 31
3.1 X-ray scattering . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2 Structure factor of a linear polymer chain . . . . . . . . . . . . 35
3.3 Structure factor of a single helix . . . . . . . . . . . . . . . . . 36
3.4 Scattering intensity of an orientationally averaged single helix . 38
3.5 Scattering intensity of a discrete double helix . . . . . . . . . . 40
3.6 Details of B-DNA structure . . . . . . . . . . . . . . . . . . . 43
3.7 Comparison of B-DNA and A-DNA structure . . . . . . . . . . 46
4 BASE-PAIR INTERACTIONS AND DNA MELTING . . . . . . . . . 50
4.1 A model for primary stabilizing interactions . . . . . . . . . . . 51
4.2 Peyrard-Bishop-Dauxois model of DNA melting . . . . . . . . 54
4.3 The DNA melting temperature . . . . . . . . . . . . . . . . . . 57
4.4 Observing DNA melting . . . . . . . . . . . . . . . . . . . . . 60
4.5 DNA denaturation bubbles . . . . . . . . . . . . . . . . . . . . 62
5 ELASTICITY OF DNA . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.1 Elastic deformation energy . . . . . . . . . . . . . . . . . . . . 67
5.2 Elastic equation of state . . . . . . . . . . . . . . . . . . . . . 68
5.3 Kirchhoff’s equation . . . . . . . . . . . . . . . . . . . . . . . 70
5.4 Kirchhoff kinematic analogy . . . . . . . . . . . . . . . . . . . 71
5.5 Elastic equilibrium with external stretching forces . . . . . . . 72
5.6 Buckling instability . . . . . . . . . . . . . . . . . . . . . . . . 82
4 Rudolf Podgornik
5.7 Euler angle of the pinched teardrop configuration . . . . . . . . 87

6 STATISTICAL MECHANICS OF DNA . . . . . . . . . . . . . . . . . 89
6.1 Kratky-Porod model . . . . . . . . . . . . . . . . . . . . . . . 89
6.2 Light scattering from a Kratky-Porod filament in solution . . . . 93
6.3 Elastic response of a Kratky-Porod filament . . . . . . . . . . . 99
6.4 Extensible semiflexible elastic filament . . . . . . . . . . . . . 102
6.5 Approximate elastic filament equation of state . . . . . . . . . . 104
6.6 Stretching DNA . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.7 Rod-like chain model of DNA . . . . . . . . . . . . . . . . . . 106
6.8 Stretching and twisting DNA . . . . . . . . . . . . . . . . . . . 110
7 ELECTROSTATICS OF DNA . . . . . . . . . . . . . . . . . . . . . . 112
7.1 Weak coupling (Poisson-Boltzmann) theory . . . . . . . . . . . 112
7.2 Cell model and Manning condensation . . . . . . . . . . . . . . 116
7.3 Forces and interactions . . . . . . . . . . . . . . . . . . . . . . 121
7.4 Strong coupling theory . . . . . . . . . . . . . . . . . . . . . . 123
7.5 DNA-DNA electrostatic interactions . . . . . . . . . . . . . . . 125
8 DNA COLLAPSE AND DNA MESOPHASES . . . . . . . . . . . . . 138
8.1 Collapse of a single DNA molecule . . . . . . . . . . . . . . . 139
8.2 DNA toroidal globule . . . . . . . . . . . . . . . . . . . . . . . 142
8.3 Nematic LC transition in a DNA solution . . . . . . . . . . . . 147
8.4 Free energy of a DNA nematic phase . . . . . . . . . . . . . . 151
8.5 Free energy of a DNA hexagonal columnar phase . . . . . . . . 155
8.6 Line Hexatic DNA Phase . . . . . . . . . . . . . . . . . . . . . 158
9 DNA ARRAYS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
9.1 Osmotic stress method . . . . . . . . . . . . . . . . . . . . . . 160
9.2 Measured DNA Equation of State . . . . . . . . . . . . . . . . 163
9.3 Cell model electrostatics . . . . . . . . . . . . . . . . . . . . . 165
9.4 Osmotic pressure of a DNA array . . . . . . . . . . . . . . . . 166
9.5 Electrostatic part of the osmotic pressure . . . . . . . . . . . . 168
9.6 Equation of state of a DNA array . . . . . . . . . . . . . . . . . 169
10 DNA ORGANIZATION IN CHROMATIN . . . . . . . . . . . . . . . 172
10.1 Nucleosomes . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
10.2 Positioning of nucleosomes along DNA . . . . . . . . . . . . . 174
11 DNA ORGANIZATION IN VIRUSES . . . . . . . . . . . . . . . . . . 178
11.1 Caspar-Klug theory end elaborations . . . . . . . . . . . . . . . 180
11.2 Continuum elasticity of viral capsids . . . . . . . . . . . . . . . 185
5
11.3 Viral capsids under mechanical stress . . . . . . . . . . . . . . 188

11.4 Osmotic encapsidation of DNA . . . . . . . . . . . . . . . . . 192
11.5 The inverse spool model . . . . . . . . . . . . . . . . . . . . . 194
11.6 DNA toroids inside viral capsids . . . . . . . . . . . . . . . . . 199
11.7 Nematic DNA Ordering in viro . . . . . . . . . . . . . . . . . 202
11.8 Orientational configurations of DNA in a nematic nanodrop . . 205
6 Rudolf Podgornik
1 Foreword
I will give a broad overview of the different aspects of the physics of Deoxyribonu-
cleic Acid (DNA) in aqueous solutions. The subject matter covered is in many respects
idiosyncratic and reflects mostly my own personal interests in the subject. I have nev-
ertheless tried to cover systematically as much ground as possible, without, hopefully,
venturing into the territory that I am completely unfamiliar with. I have tried to bring
attention to fundamental works as well as good reviews.
The book presents a personal view of the physics of DNA. Neither the list of topics
nor the list of references are exhaustive. The framework is that of soft matter physics
centered on some of the problems I worked on such as elasticity, fluctuations and in-
teractions. Many other aspects have been covered in the previous book by Alexander
Vologodskii 1 and several beautiful reviews by Maxim Franbk-Kamenetskii 2
The book is based on lectures that I gave at the Department of Physics, University
of Ljubljana, in the past 10 years. At the 37th Latin-American School of Physics 2006,
Universidade Estual Paulista, Sao Jose do Rio Preto, Brazil in 2006. At the 2007 Taiwan
International Workshop on Biological Physics and Complex Systems (BioComplex Tai-
wan), Academia Sinica, Taipei, Taiwan in 2007. At the International School of Biomem-
brane Physics, Indian Institute of Technology Madras, IIT-Madras, Chennai, India, in
2008, at the 3-eme cycle de la Physique en Suisse Romande, Ecole Polytechnique Fed-
erale de Lousanne, in 2009 and as part of the Gluckstern lectures in biological physics at
the Department of physics, University of Massachusetts, Amherst in 2011.
Thanks to Don C. Rau, V. Adrian Parsegian, M. Muthukumar, Michel Peyrard, Hel-
mut Schiessel, Anthony Maggs, Luka Curovic, Manca Podvratnik. My special thank go
to Antonio Siber who produced many of the images I used in this book.
1 A.Vologodskii, Topology and Physics of Circular DNA, CRC Press (1992).

2 ï£ijMaxim D. Frank-Kamenetskii, Biophysics of the DNA molecule, Physics Reports 288 (1997) 13-
60ï£ijï£ij.ï£ij.
7
2 HELICAL CURVES
The elucidation of the double-helical nature of the molecular structure of DNA was one
of the triumphs of science in the previous century. In fact helical structures are ubiquitous
in the world of biological macromolecules, be it nucleic acids or proteins. Not only is
the DNA a helical molecule, it is also a chiral molecule showing a right-handed helix.
Not only is DNA a double-helical molecule because it is composed of two interwound
helices, it is also an asymmetric double helix. The asymmetry in this case pertains to the
position of the two helical curves with respect to the longitudinal axis of the molecule.
In this chapter we will touch upon most fundamental definitions pertaining to the
geometry of the DNA molecule though necessarily it involves a lot of idealization since
DNA is a molecular and not a mathematical object. Nevertheless a mathematical glossary
is in order to facilitate later developments.
Apart from the mere kinematical analysis of spatial helical curves the most important
result of this chapter is the Călugăreanu-White-Fuller theorem. This theorem shows how
to decompose a topological invariant of a closed helical curve into
• twist, being a local quantity which depends on the internal configuration of the
DNA molecule. It is equal to the number of times one strand winds around the
other one in a circular DNA molecule. Contrary to twist
• writhe is a non-local quantity associated with the overall spatial shape of the DNA
helix. It is non-zero for DNA shapes that are not straight.
With certain caveats the definitions of twist and writhe can be applied also to non-closed
curves and are an essential ingredient of the statistical mechanical theories of the DNA
conformations as will become clear later on.
2.1 Mathematical description of spatial curves

A curve is a mathematical object that can be represented by an equation of the form
r = r(`), (1)
where ` is a parameter of this description. If it is defined in such a way that when it runs
from 0 to the total contour length of the curve L, the radius vector r traces the curve in
space, then it is refered to as the arclength parameter or also as the natural parameter
of the curve and the curve in Eq. 1 describes the curve in its natural parametrization. At
every point along the curve we can now define a local unit tangent vector given as
dr(`)
t= = ṙ(`) such that t · t = 1. (2)
d`
If the tangent vector along the curve is a constant, the curve is obviously a line. . The
RL
toral contour length of the curve is thus given as L = 0 d`. The derivative of the tangent
vector with respect to ` is also a vector, and can be obviously written as
ṫ(`) = (`)n(`) where ṫ(`) · t(`) = 0. (3)

8 Rudolf Podgornik
Figure 1. The unit vectors t(`), n(`) and b(`) along a spatial curve.
Vector n has to be perpendicular to the tangent vector and is usually referred to as the
normal vector. The scalar (`) is by definition the curvature of the curve. The tangent
vectro is not the only vector that changes direction along the curve. So does its normal
vector (see figure). It can either change its angle with respect to the tangent vector or can
just as well rotate around the tangent vector. In the first case the curve remains within
a single plane while in the second case the plane within which the curve lies changes as
well. Thsi plane is referred to as the osculating plane 3 . The changes in the osculating
plane are described by vector b
t(`) ⇥ ṫ(`)
b(`) = t(`) ⇥ n(`) or equivalently b(`) = , (4)
|t(`) ⇥ ṫ(`)|
the vector of the binormal to the curve. Consequently we can write the derivative of the
normal vector as
ṅ(`) = ↵(`)t(`) + ⌧ (`)b(`), (5)
where the second scalar coefficient is the torsion ⌧ (`) and the first one we will ex-
press in terms of curvature. Mathematical torsion has no connection whatsoever with
the physiacal torsion!. From t(`) · n(`) = 0 we can derive straightforwardly 0 =
ṫ(`) · n(`) + t(`) · ṅ(`) = (`) + ↵(`). Thus we realize that ↵(`)is nothing but mi-
nus the curvature fo the curve
ṅ(`) = (`)t(`) + ⌧ (`)b(`). (6)
Let us finally evaluate also the derivative of the binormal to the curve. Obviously, because
of Eq. 3 and 6
ḃ(`) = ṫ(`) ⇥ n(`) + t(`) ⇥ ṅ(`) = ⌧ (`)n(`). (7)
The three equations Eqs. 3, 6 and 7 are usually referred to as the Seret - Frenet equations.
Once again they are of the form
ṫ(`) = (`)n(`)
ṅ(`) = (`)t(`) + ⌧ (`)b(`)
ḃ(`) = ⌧ (`)n(`). (8)
3 From greek o ⌫ ! : a kiss
9
These are the basic equations of the differential geometry of a spatial curve. They tell
us that the "speed" of rotation of t around the axis b is given by (`) and that ⌧ is the
"speed" os rotation of n around the axis t. We can cast the Seret-Frenet equations Eq. 8
in the form of a matrix relation as
(ṫ, ṅ, ḃ) = K (t, n, b), (9)
where the curvature matrix is given by

2 3
0 (`) 0
6 7
K(`) = 6
4 (`) 0 ⌧ (`) 7
5. (10)
0 (`) 0
From the Seret-Frenet equations it follows that the whole curve can be reconstructed from
the local values of its curvature (`) and torsion ⌧ . If the curvature is zero the torsion of
the curve is obviously ill-defined.
Let us now find the appropriate expressions for the calculation of curvature and tor-
sion of curves. From the first Frenet-Seret equation we derive that
|t(`) ⇥ ṫ(`)| = |r̈(`)| = (`). (11)
From the third one and after evaluating ḃ(`) from Eq. 4 we get
...
|(t(`) ⇥ ẗ(`)) · ṫ(`)| |(ṙ(`) ⇥ r (`)) · r̈(`)|
|ḃ(`) · n(`)| = = = ⌧ (`). (12)
|t(`) ⇥ ṫ(`)|2 (`)2
These are the basic equations of the theory of spatial curves. Apparently the torsion of a
straight curve ( ! 0) is ill defined, as can be seen from the above equation and even in
a clearer way for a short arclength segment by assuming that both, the curvature as well
as the torsion, are arbitrary but constant. In that case the equation of the spatial curve in
the vicinity of the origin, defined as r(0), is of the form
...
r(s) = r(0) + ṙ(0)s + 12 r̈(0)s2 + 16 r (0)s3 + . . . (13)
Using the Frenet-Seret equations this leads to
1 2 3
r(s) r(0) = s 6 s t(0) + 12 s2 n(0) + 16 ⌧ s3 b(0) + . . . (14)
A straight filament with  = 0 can thus be approached with any arbitrary value of the
torsion as lim !0 ⌧ = 0! Mathematical torsion itself actually determines the rate of
deviation of the curve from a planar configuration spanned by the vectors n(0) and t(0).
This statement on the irrelevancy of mathematical torsion for a straight line has to
be acknowledged with caution since, obviously, even a straight filament can have phys-
ical torsion. The above consideration is thus applicable only to mathematical lines for
which we can not define twist about the axial line of the cross-section of a thin filament
connected with physical torsion.
10 Rudolf Podgornik
2.2 Material frame of reference

Let us assume now that three Cosserat orthogonal unit vectors d1 , d2 , d3 are frozen into
the material of the rod. By assumption d3 has the direction of the tangent to the rod and
thus
@r(`)
d3 = t(`) = ṙ(`) = .
@`
In the plane perpendicular to d3 we define two orthogonal unit vectors d2 and d1 . We
direct them in such a way that they coincide with the principal directions of the moment
tensor of the cross-section of the rod.
These two vectors are thus embedded into the material of which the rod is composed.
The coordinate system defined in such a way is said to described the material frame of
reference. Since these basis vectors are orthogonal to one another, we can set
d1 = d2 ⇥ d3 d2 = d1 ⇥ d3 . (15)
Assuming furthermore that d3 is directed along the tangent vector t and thus d1 , d2 are
in the same plane as n, b of the Frenet frame, there has to exist a rotation that can take
one pair of vectors into the other one. This rotation can be described by an angle ↵ that
depends on the position along the rodt
d1 = n cos ↵ + b sin ↵
d2 = n sin ↵ + b cos ↵, (16)
where the angle ↵ is the register angle 4 . The new unit vectors satisfy a similar equation
as t, n, b, Eq. 156 , that can be derived in the form
ḋ3 = 2 d 1 1 d 2
ḋ2 = ⌧ d 1 + 1 d 3
ḋ1 = ⌧ d2 2 d 3 , (17)
where we also remembered the definitions of the vector products between the different
basis vectors Eq. .15. The last equation can be also written in the form of a matrix
relation as
(ḋ1 , ḋ2 , ḋ3 ) = K0 (d1 , d2 , d3 ), (18)
where the curvature matrix is given by
2 3
 (`) 1 (`) 0
6 2 7
K (`) = 6
0 0
4 ⌧ (`) 0 1 (`) 7.
5 (19)
0
0 ⌧ (`) 2 (`)
The above relation plays the role of the Seret-Frenet equation for the material frame of
reference (d1 , d2 , d3 ). Here we have defined
d↵
1 =  sin ↵, 2 =  cos ↵, ⌧0 = + ⌧.
d`
4 A.E.H. Love, A treatise on the mathematical theory of elasticity, Dover, New York, 1944 p .383
11
We see that in this case the torsion of the curve is given by the previous expression
for the mathematical torsion to which one must add the twist of the curve: d↵ d` . The
complete torsion thus depends on the spatial form of the rod as well as the local twist of
the rod along its length. This last part is the fundamental difference between an idealized
mathematical line and a physical filament.
2.3 Darboux vector

Let us take the basis vectors at the beginning of the filament: (d3 (0), d2 (0), d1 (0))
and somewhere along the filament: (d3 (`), d2 (`), d1 (`)). The angle between these two
directions can be described as a vector (`) = ( 3 (`), 2 (`), 1 (`)) where
cos 3 (`) = d3 (`) · d3 (0)
cos 2 (`) = d2 (`) · d2 (0)
cos 1 (`) = d1 (`) · d1 (0). (20)
We define the "velocity" of change of these angles, or as it also referred to as the Darboux
Figure 2. The three Cosserat unit vectors, d1 , d2 and d3 along a rod represented by a
three dimensional space curve. d3 is directed along the tangent to the curve.
vector ⌦(`), as
d (`)
⌦(`) = . (21)
d`
Since all the three basis vectors are unit vectors, it is only their direction that can change
along the filament. The only change they can be subject to is thus rotation in space and
consequently we have immediately that
ḋi (`) = ⌦(`) ⇥ di (`) i = 1, 2, 3, (22)
12 Rudolf Podgornik
by simple analogy with circular motion of a point particle. It is thus obvious that di (`) ·
ḋi (`) = 0. Comparing the above equation with Eq. 19 we immediately deduce that
d 3
⌦3 = = ⌧ 0, (23)
d`
while the other two components of the Darboux vector follow straightforwardly as
d 2,1
⌦2,1 = = 2,1 . (24)
dl
The components of the Darboux vector can then be written compactly with the help of
the unit vectors di in the form
⌦(`) = ⌧ d3 (`) + 2 d2 + 1 d1 . (25)
We have thus established a connection between the components of the Darboux vector,
twist and the two curvatures of the filament directed along the two principal directions of
the moment tensor of the filament’s cross-section.
2.4 Euler angle parametrization

Let us specify the three rotation angles ( 3 (`), 2 (`), 1 (`)) more in detail. According
to Euler’s rotation theorem, any rotation in space may be described using three angles.
This is equivalent to saying that any rotation matrix can be decomposed as a product of
three elemental rotation matrices. The three angles , ✓ and are called Euler angles
and give three separate elemental rotation matrices.
There are several ways of choosing Euler angles and we will choose the s.c. z y z
convention: the first elemental rotation will be the rotation of the reference xyz frame
through an angle about the z axis, see Fig. 2.4. This rotates the first and second axis
in the xy plane. In particular, the second axis now points in the direction of y. Next we
rotate the frame through an angle ✓ about the new y axis. This moves the third axis z in
any assigned orientation in space. Finally, we rotate the frame about z through whatever
angle is needed to bring the axes into their final assigned directions. The rotation
matrix R is a product of the following three elemental rotation matrices
R( , ✓, ) = Rz ( )Ry (✓)Rz ( ),
(12) where indices z and y tell us about which axis the frame rotates.
Assume two directions in space: d(0) corresponding to the beginning of the filament
and d(`) corresponding to the point parameterized with `. We can go from one to the
other one via the rotational matrix written in terms of the Euler angles (`), ✓(`), (`)
denoting precession, nutation and self-rotation respectively and shown on Fig. 2.4.
Assuming that a triad of unit vectors (d1 (`), d2 (`), d3 (`)) can be associated with
every point along the curve, the relation between the initial and final directions can be
cast into the form
(d1 (`), d2 (`), d3 (`)) = R(`)(d1 (0), d2 (0), d3 (0)) (26)

13
Figure 3. Definition of the precession, nutation and self-rotation Euler angles , ✓,

in the zyz convention. Three successive rotations bring a reference frame into any pre-
scribed orientation in space.
where the rotational matrix R is given by

2 3
cos cos ✓ cos sin sin cos cos ✓ sin sin cos cos sin ✓
6 7
R(`) = 64 sin cos ✓ cos + cos sin sin cos ✓ sin cos cos sin sin ✓ 7
5.
sin ✓ cos sin ✓ sin cos ✓
(27)
First of all we note that after taking the derivative of Eq. 26 with respect to the arclength,
we get
(ḋ1 (`), ḋ2 (`), ḋ3 (`)) = Ṙ(`)(d1 (0), d2 (0), d3 (0)) =
= RT (`)Ṙ(`)(d1 (`), d2 (`), d3 (`)).
(28)
Obviously the frame defined by the three basis vectors rotates along the rod with an
"angular frequency" ⌦(`) so that
(ḋ1 (`), ḋ2 (`), ḋ3 (`)) = ⌦(`) ⇥ (d1 (`), d2 (`), d3 (`)) (29)
Here the cross product can be also written in terms of a matrix product of an antisym-
metric matrix with the unit vectros of the rotating frame. Introducing the components of
the "angular frequency" as ⌦(`) = RT (`)Ṙ(`) = (1 (`), 2 (`), ⌧ (`)), we get
⌦1 = 1 = ✓˙ sin ˙ sin ✓ cos

⌦2 = 2 = ✓˙ cos + ˙ sin ✓ sin
⌦3 = ⌧ 0 = ˙ cos ✓ + ˙ . (30)
It is obvious from here that the torsion is composed of the the sum of the internal twist
of the rod, ˙ , and its tortuosity, i.e. the projection of ˙ onto the local d3 (`) direction.
This geometry is displayed schematically on Fig. 2.4. As we will see later on, the helical
curve is analogous to a precessing spinning top.
14 Rudolf Podgornik
Figure 4. Euler angles and their derivatives. For precession of a spinning top or a rod
bent into a helix, the ✓ angle is constant.
2.5 Single and double helix

2.5.1 Continuous single helix
We define a helix 5 to be a spatial curve of constant angle between a spatial direction and
local tangent,
✓˙ = 0.
as well as of constant curvature and tortuosity
q
 = 21 + 22 = ˙ sin ✓ = const. and ˙ cos ✓ = const. (31)
A helical curve can thus be parameterized with only two parameters. These two param-
eters can be either curvature and tortuosity or equivalently the helical pitch, h, in the z
direction and its radius, R. If we now introduce the angle ↵ between the local tangent
and the xy cross-sectional plane of the helix, i.e. ↵ = ⇡2 ✓, then we have the Eulerian
parameterization of the helix
r( ) = (R cos , R sin , c ) (32)
with c = 2⇡h
. can thus be interpreted as an angle around the ẑ direction along the axis
of the helix and the arclength along the helix is given by
R
`= . (33)
cos ↵
From this the pitch can be derived from the fact that the length of the helix per pitch is
given by
h 2⇡R 2⇡R sin ↵
L= = and thus h=
sin ↵ cos ↵ cos ↵
The tangent vector to the helix is then obtained in the form
5 from Greek: ✏ ◆⇠ a spiral.
15
Figure 5. Helical curves can be seen everywhere. Look around and you are bound to
find them.
dr(s)
t= = ( cos ↵ sin , cos ↵ cos , sin ↵) (34)
d`
We notice that for a helix the scalar product between the tangent vector and a direction
in space, in this case the direction of the z axis, is a constant
t · ẑ = sin ↵ (35)
This is then the general definition of a helical curve in space.

Since ṫ is perpendicular to t it follows from the Frenet - Serret formulae that n· ẑ = 0.
The explicit form of the normal and binormal vectors can be obtained straightforwardly
as
n = ( cos , sin , 0)
and
b = (sin ↵ sin , sin ↵ cos , cos ↵) .
If we now also evaluate the curvature and the torsion of the helical line from the Frenet-
Serret formulae, they assume the form
cos2 ↵
(s) = |r̈| =
R
... sin ↵ cos ↵
⌧ (s) =  2 |(ṙ, r̈, r )| = (36)
R
The last two statements can be written together as

t · ẑ = cos ↵ = cot ↵. (37)
⌧
16 Rudolf Podgornik
This can be formulated as a general teorem: a helical curve must have a constant ratio of
curvature to torsion6 . Any curve with this property is bound to be a helix. A helix can
be defined equivalently by setting either (↵, R) or (, ⌧ ). There is however yet another
important property of helical curves that we need to discuss.
Figure 6. Right handed, on the left-hand side, and left handed, on the right-hand side,
helix obtained from Eq. ??. Embedded arrows show the direction in which the curve
ascends or descends along the direction of its long axis.
2.5.2 Continuous double helix
We can of course take two helical curves inscribed on the same cylinder of radius a 7 .
Parametric description of these two helical curves would then be
r1 (`) = (a cos `1 , a sin `1 , c`1 ) and r2 (`) = (a cos `2 , a sin `2 , c(`2 +

)) ,
(38)
where the angle represents the angular shift between the two helical lines if cut with
a plane perpendicular to the central axis of the helices. For two symmetrical helices,
piercing the perpendicular plane at two points across the centerline of the helices, we
have = ⇡.
Figure 2.5.1 represents six turns of a single and double continuous right-handed he-
lix. In the latter case the two strands of the double helix are shifted by an angle of 3⇡/8
making the double helix asymmetric with respect to the central axis. Take a plane per-
pendicular to the long axis of this cylinder and cut the two helices. The intersection of
6 Theorem
of Lancret, 1802.
7 K.
Olsen and J. Bohr, The generic geometry of helices and their close-packed structures, Theor Chem Acc
125 207 (2010).
17
Figure 7. Six turns of a continuous single and asymmetric double right-handed helix.
The thickness of the line is exaggerated for rendering purposes.
this plane and the two helices defines two points which do not coincide as long as the
two helices are shifted by a finite angle . Figure Fig. 2.5.1 shows two such helical
lines (yellow and blue) shifted by an angle of 3⇡/8. If this angle is different from ⇡ then
the two helical lines make an asymmetric pattern with a well defined major and minor
"groove", meaning that in the direction of the long axis of the helix with radius a, the
separation of the yellow and blue lines are uneven. This uneveness of separation has an
important meaning for the structural properties of the double helical DNA.
DNA handedness n ! h[nm] a[nm] ✓

A right 0 11 31.1 0.26 0.69 34.8
B right 0.375 10.5 31.1 0.34 0.7 37.4
Z left 0 12.0 -30.0 0.37 0.43 -58.8
Table 1. A table of different DNA structures. Adapted from K. Olsen and J. Bohr, The
generic geometry of helices and their close-packed structures, Theor Chem Acc 125 207
(2010).
It is difficult to associate continuous helical curves with DNA, though as will be clear
later, it is indeed a double helix. The problem is how to go from a detailed molecular
structure to helical lines. Most commonly, the phosphorus atoms along the phosphate
backbones are thought to correspond to idealized double helical curves though other
conventions are also possible. For the table 1 the helical lines were taken to be 0.25 nm
closer to the axis of the molecule.
18 Rudolf Podgornik
Figure 8. Two double helices with symmetric and non-symmetric positions of each helix.
LHS: symmetric arrangement, where the two helices are displaced by a finite angle not
equal to 6= ⇡. RHS: non-symmetric arrangement, = ⇡. This configuration of
the helices clearly leads to a major and a minor "groove"
2.5.3 Left and right-handed helix
Every helix posesses also handedness. Helical curves can be seen everywhere in ev-
eryday life and can be either right-handed or left-handed. The handedness of a helix is
its intrinsic property and does not depend on the way one looks at the helix. Standard
screws, nuts, and bolts are all right-handed, as are both the helices in a double-stranded
molecule of DNA. Large helical structures in animals (such as horns) usually appear in
both mirror-image forms, although the teeth of a male narwhal, usually only one which
grows into a tusk, are both left-handed.
Observing a helix in the direction of its axis its handedness is defined as follows: if
one follows the movement of an imaginary point along the helix it is said to be right-
handed if the angular velocity vector of this motion point sin the same direction as the
velocity vector along the axis. If it points in the opposite direction, then the helix is
left-handed. Both types of helices are not superimposable on their mirror-images and are
thus chiral by definition. Chirality is a property of the helix, not of the perspective: a
right-handed helix cannot be turned or flipped to look like a left-handed one unless it is
viewed through a mirror, and vice versa. Kelvin 8 defined chirality as
I call any geometrical figure, or group of points, chiral, and say it has
chirality, if its image in a plane mirror, ideally realized, cannot be brought to
coincide with itself.
8 William Thomson, Lord Kelvin, Baltimore Lectures on Molecular Dynamics and the Wave Theory of
Light, (1904).
19
Figure 9. The rotational symmetry elements of an elementary single (right front) and
double (left back) continuous helix. The lines composing the helix have a sense of direc-
tion to them, indicated by embedded arrows. In the case of the double helix (yellow-blue)
they are assumed to be antiparallel. For a single helix (red) the sense of directedness
does not matter. The angles of rotation are indicated by a green semi- and full circle
together with the axis of rotation.
In cylindrical coordinates a helical curve described by Eq. 32 has the parametric form
r( ) = (R cos , ±R sin , c ) (39)
The + sign refers to a right handed helix, where the sign of the polar angle is defined
by the right-handed screw, and the sign refers to a left handed helix. Obviously in
the first case and with it the angular velocity vector points in the direction of the axis,
i.e. z, whereas it points in the opposite direction in the second case. Two helices with
different handedness are represented in Fig. 2.5.1.
Handedness has the symmetry properties of a pseudo-scalar and can only have two
values, ±1. It thus behaves as a scalar in all symmetry transformations, expect under a
parity inversion where it changes sign. Aother example of a pseudoscalar quantity is the
magnetic flux.
2.5.4 Rotational symmetry and handedness
Assume now that the helical curve has a direction. A simple continuous model presented
above can not discern between different handedness, but we will later see that the real
DNA helix indeed has handedness. The nature of this handedness need not concern us
yet, but we will soon see that it is naturally given by the assymetry of the sugar-phosphate
bond in DNA.
20 Rudolf Podgornik
Handedness means that the direction of the helical line can either point upwards or
downwards along the helix, see Fig. 2.5.3, assuming that the helical axis is vertical.
We can now rotate the whole helix around an axis that is perpendicular to the axis of
the helix and is located right at the midpoint of the helical pitch, i.e. halfway along the
period of the helix. On rotating the helix around this axis for a single directed helical
line we have to rotate by a complete turn of 2⇡ for the helix to coincide with its position
before the rotation, see Fig. 2.5.3. For a double helix whose two single helical lines have
antiparallel directions so that one climbs along the helix and the other one descends - and
this is essential - this ceases to be the case any more. We have to rotate the antiparallel
double helix only by half that angle, i.e. by ⇡, for the two forms to coincide 9 . In this
difference of rotational symmetry elements for a single directed helix and an antiparallel
double helix Francis Crick found the essential missing piece in the elucidation of the
double-helical structure of DNA 10 that allowed him to realize that there were two chains
in DNA and that they ran in opposite directions.
2.6 Călugăreanu-White-Fuller theorem

A helical curve especially if it represents a molecule of DNA need not be on the average
straight but can trace out a curve in space of general shape, like the single helix depicted
in Fig. 2.6. These space curves can be either open or closed. In the latter case the
beginning and the end of the curve coincide so that the number of helical turns is fixed
and describes this particular configuration of the helix. It is thus a topological invariant
since one can smoothly deform the circle whichever way without changing this number
of turns.
In order first to describe these closed helical curves one needs to know what is the
equation of the average position of the helix, defined via its axial curve which passes
through the centroid of the curve. Apart from that one also needs to know how many
times the helix winds around itself when tracing out the axial line. As we will see, this
statement can be given an interesting analytical expression as the Călugăreanu-White-
Fuller theorem.
For double stranded helical DNA the number of times one phosphate strand winds
around the other one can be calculated if one knows the form of the spatial curve. For a
circular DNA molecule this number has two components: the base pairs on the two phos-
phate strands are twisted around the center line of the molecules and/or the center line
of the molecule itself can be writhed in space. Since both possibilities can in principle
coexist, the number of times one phosphate strand winds around the other is additive in
both mechanisms. This number is referred to as the linking number, Lk, of DNA and is
a topological invariant of a circular DNA loop.
2.6.1 The linking number
Calculation of the linking number can be accomplished via a simple magnetic analogy 11
- since in magnetism too, in particular in the Ampére’s law, we want to know how many
9 Herewe consider only the directionality of the two helices and not their relative spatial positions.
10 J.E. Lydon, The DNA double helix-the untold story, Liquid Crystals Today 12 1 (2003).
11 In this derivation I follow R. Kamien, The Geometry of Soft Materials: A Primer, 2002
21
Figure 10. A spatial curve traced out by a single helix that can be either open, or closed.
If closed in a circle it can have a varying number of turns of the helix.
times a current wire wraps around a certain spatial curve. Imagine a solenoid of n turns
described by a curve C with a unit current flowing through it. From Ampére’s law we
know that the number of times the solenoid passes through another closed curve C 0 is
given by a line integral of magnetic field B(r)
Z Z
1 dr ⇥ (r0 r)
n= B(r0 ) · dr0 where B(r0 ) = (40)
4⇡ C 0 C |r0 r|3
by the Biot - Savart law since the line integral of the magnetic field is 4⇡n for a unit
current. In the case of magnetic fields the curve C 0 goes along the center of the solenoid
and then closes on itself outside of the solenoid. Putting the two together we can calculate
the number of times that one curve wraps around the other when the loop is closed.
Same thinking can now be applied to a DNA double helix. We identify the two curves
C and C 0 , with the former as the axial curve, passing through the centroid of the molecule,
and the latter as one of the phosphate backbones. The linking number is now obtained
by first assigning an orientation to both curves as well as defining a sign for each of the
crossings. The linking number, Lk, is then one half the number of crossings of the two
curves and corresponds to the number of times one strand wraps around the other strand
in a 2D projection. For a closed double helix this number is an invariant and can not
change even if the spatial conformation of the molecules changes. In principle when the
helix is open it can spontaneously unravel and change its linking number.
We now use the framework of the Biot-Savart’s law for the two curves C and C 0 that
are parameterized by r(s) and r0 (s0 ). The (Gauss) linking number is then given by
Z Z
1 (ṙ(s) ⇥ ṙ0 (s0 )) · (r(s) r0 (s0 ))
Lk(C, C 0 ) = n = ds ds0 = const. (41)
4⇡ C C 0 |r(s) r0 (s0 )|3
22 Rudolf Podgornik
The above result is however valid for any two curves in particular also for the two strands
of DNA. Note that the linking number defined above is a non-local quantity and depends
on two positions along the curves C and C 0 .
Though the above expression is an interesting results giving a topological invariant
of C and C 0 it is of little use in the physics of DNA where we usually deal with the shape
of the molecule as a whole and not with the configuration of the two strands, i.e. the
two phosphate backbones of the molecule. In order to write the above invariant in terms
of the molecular configuration of DNA we assume that the axial line C and one of the
phosphate backbones C 0 are in close proximity to one another and can be written as
r0 (s) r(s) = ✏n(s), (42)
where ✏ ⌧ 1 and n(s) is assumed to be a unit vector perpendicular to the tangent vector
of the axial curve C, t(s) = ṙ(s), and pointing towards one of the phosphate backbones
C0.
2.6.2 Derivation within a magnetic analogy
The (Gauss) linking number has an interesting alternative formulation in the case of
magnetic flux filaments 12 . Assume then that the magnetic field B is zero, except in two
flux filaments centred on two unknotted
R oriented closed curves C and C 0 . The magnetic
flux in the two flux filaments, S B · dS, is the same and equals . We assume also that
B-lines within either tube on its own are unlinked closed curves. Then one can derive
the following relation for the two curves
Z Z
H= A · B dV = A · (r ⇥ A) dV = 2 Lk(C, C 0 ) 2 , (43)
(V ) (V )
where the volume integral corresponds to the helicity H of the magnetic field. Now let
us decompose the magnetic field into an axial Ba and meridional Bm components and
adopt a local cylindrical coordinate system along the flux tube. Thus
Ba = (0, 0, Bz (r)) and Bm = (0, B✓ (r), 0). (44)
The magnetic potentials are then decomposed in the same way. Since the lines of force
for the meridional magnetic field are unlinked circles one can derive the following equa-
tion for the helicity
Z Z
H= Aa · Ba dV + 2 Aa · Bm dV = Haa + 2Ham . (45)
(V ) (V )
For the axial part we can now use the Biot-Savart expression
Z
1 B(r0 ) ⇥ (r r0 ) 0
A(r) = dV with B dV ! dr = tds (46)
4⇡ (V ) |r r0 |3
12 H.K. Moffatt and R.L. Ricca, Helicity and the Caluggreanu invariant, Proc. R. Soc. Lond. A (1992) 439,
411-429.
23
where t(s) = ṙ(s) is the unit tangent vector of the axial line and thus the first term in Eq.
45 becomes
Haa = W r(C) 2 (47)
where we introduced the writhe of the curve

Z Z
1 (t(s) ⇥ t(s0 )) · (r(s) r(s0 ))
W r(C) = ds ds0 , (48)
4⇡ C C |r(s) r(s0 )|3
which a non-local quantity depending on two positions along the axial line C describing
the spatial position of the centroids of the DNA basepairs. Through writhe is similar to
the linking number for a single curve, it is nevertheless not a topological invariant of the
curve. Writhe describes the winding of the axial line in space and is equal to the number
of signed self-crossings of this line averaged over all such 2D projections. Alternatively
the writhe can be defined as the area swept by the vector n(s) on a unit sphere when
allowed to slide along C 13 .
For the send term in Eq. 45 it is simpler to calculate the change in this term as one
varies the curvature and the torsion of the curve. Assuming that the "intrinsic twist" of
the field lines within in the two flux tubes remains unchanged in any deformation of the
lines, one derives that
Ham = T w(C, C 0 ) 2 (49)
where we introduced the twist of the helix defined by

Z Z
0 1 1
T w(C, C ) = ds t(s) · (n(s) ⇥ ṅ(s)) = ds !(s), (50)
2⇡ C 2⇡ C
where n(s) was defined in Eq. 42 and is assumed to be a unit vector perpendicular to
the tangent vector of the axial curve C. !(s) is by definition the twist density and equals
the change in the local angle of the curve in the direction of the tangent. Since twist is
given by a single integral along C it is a local quantity. This decomposition of the linking
number is usually referred to as the Călugăreanu-White-Fuller theorem 14 and can be
written formally as
Lk(C, C 0 ) = T w(C, C 0 ) + W r(C). (51)
The linking number, which is a topological invariant of a given closed DNA ring, is thus
composed into two terms, the twist, T w, and the writhe of the helix.
For a spatial curve that is on the average straight, its writhe is zero since ṙ(s)⇥ṙ(s0 ) =
0, and the linking number is exactly equal to the total twist. Otherwise the linking number
is partitioned into twist and writhe according to Eq. 51 and as shown schematically for
two configurations of a DNA molecule with the same linking number in Fig. 11, where
the circular DNA has 14 crossings so that its linking number equals 7 and is partitioned
into twist (7) and writhe (0). After the twisting of the circle we then get a twist of 8 and
a writhe of -1 to satisfy the Călugăreanu-White-Fuller theorem.
13 D. Swigon, The Mathematics of DNA Structure, Mechanics, and Dynamics, IMA Volumes in Mathematics
and Its Applications, 150, 293 (2009).

14 M. R. Dennisand J. H. Hannay, Geometry of Călugăreanu’s theorem, Proc. R. Soc. A (2005) 461, 3245.
24 Rudolf Podgornik
Figure 11. Circular DNA with linking number 8 (left), corresponding to twist 8 and zero
writhe. The same segment of DNA but now with twist 24 and writhe -1. Courtesy L.
Curovic.
2.6.3 Twist and writhe of a continuous helix
In order to get a better idea about the concepts of twist and writhe let us analyze them for
two simple cases. Let us first calculate the twist of a single helix
r( ) = (a cos , ±a sin , c ) (52)
p
with a pitch given by p = 2⇡ a2 + c2 . Since the rotation of the helix per pitch is
obviously 2⇡, the rotation angle along the helical path is given by
2⇡ 1
=p .
p a + c2
2
This is not yet the twist density since we have to project it on the direction of the tangent,
therefore
cos ↵ c a
!=p = 2 2
with cos ↵ = p , (53)
2
a +c 2 a +c a + c2
2
where ↵ is thus just the angle between the local direction of the helix and the axis ẑ.
From here the value of the twist is given by
1 cL
Tw =
2⇡ a2 + c2
where L is not the (arc)length of the helix. In this case it is obviously linear in the length
of the helix.
As an example let us calculate the writhe of a circular helix, so that its linking number
is assumed to be zero, i.e. we simply glue together the two ends. Now rotate two opposite
sections of the circle yielding eventually a configuration that we term a plectoneme. Since
by definition the linking number is zero it follows from Eq. 51 that the writhe is exactly
balanced by the twist
L c
Wr = Tw = . (54)
2⇡ a2 + c2
25
Figure 12. Two curves with different twist and writhe. The upper curve has a non-zero
writhe and a twist. The lower curve has the same linking number, but has no writhe.
Thus we are able to calculate writhe exactly for the case of an ideal plectoneme where we
simply disregard the possible complications at the both ends of the plectoneme. We can
cast this expression in a different but equivalent form as follows. Recalling the definition
of the angle ↵ between the local direction of the helix and the axis ẑ, the expression for
writhe, Eq. 54, can be cast into a slightly more general form 15
L ca L sin 2↵
Wr = = , (55)
2⇡a a2 + c2 4⇡a
where we first introduced = ±1 for the chirality of the two helices in the plectoneme,
and a and ↵ now denote the superhelical radius and angle of the plectoneme. We will use
this expression later, when we will be calculating the elastic energy of a plectoneme.
2.6.4 DNA linking number
Any conformational change of DNA that involves changes in its linking number, such as
protein binding, is thus partitioned into writhe and twist. Whenever the molecule has a
high writhe we say that it is supercoiled. Supercoiling of DNA is usually tightly regulated
by enzimes topoisomerases, whose action can either change the linking number Lk =
±1 (Type I DNA topoisomerases) or by Lk = ±2 (Type II DNA topoisomerases). In
the first case one of the phosphate backbone strands is nicked, the molecule is locally
rotated around the intact phosphate backbone and then resealed. In the latter case both
phosphate backbone of a DNA chain in close proximity to another one are nicked, the
second DNA passes through the opening, which is then resealed. The same type of
enzymes can also induce knotting of DNA which however rarely occurs naturally.
DNA is often circular as in the case of various plasmids or even viruses. The changes
in the linking number, Lk, such as those performed by the topoisomerases are usually
15 S. Neukirch, Extracting DNA Twist Rigidity from Experimental Supercoiling Data, Phys. Rev. Letts. 93
198107-1 (2004). E. L. Starostin, arXiv:physics/0212095v1.

26 Rudolf Podgornik
16
expressed through the quantity defined as
Lk
= , (56)
Lk0
where Lk0 is the natural linking number associated with the unstressed molecule.
In the case of the relaxed B-DNA there is one full twist per every 3.4 nm length
comprised of 10.4 ± 0.1 base pairs. Lk0 , is then obtained by dividing the total number
of base pairs, #(bp), of the molecule by the relaxed number of base pairs per turn which
thus gives
#(bp)
Lk0 = .
10.4
This is the canonical linking number of unstressed B-DNA.
In plasmid DNA, since it is closed, Lk has to be an integer number. By convention,
Lk of a closed right-handed DNA helix is positive. The most frequent DNA conformation
for plasmids in cells is negatively supercoiled. This means that for such plasmids Lk is
less than it would be for a torsionally relaxed DNA circle - negatively supercoiled DNA is
underwound. This is a general phenomenon with important biological consequences. It
seems that free energy of negative supercoiling catalyzes processes that depend on DNA
untwisting, such as DNA replication and transcription.
2.6.5 Twist and writhe of open curves
The expression for the twist of a spatial curve is a local property and can thus be straight-
forwardly formulated also for an open curve. However, the expression for the writhe in
Eq. 48 is highly non-local, being a global property of the DNA shape. It is obviously
defined only for the cases, where the spatial curve, describing the central line of DNA
does not self-intersect. This is a consequence of the fact that our derivation was based on
Ampére’s law, which is formulated only for closed curves. The question is then, can we
use the same concepts also for open curves, such as segments of DNA of finite length?
In order to do just that we note that there is an alternative interpretation of writhe
based on the tangent indicatrix or tantrix of the curve 17 . As the tangent vector of the
curve is a unit vector its direction along an arbitrary spatial curve describes a curve on
a unit sphere where one end of the tangent vector is fixed at its origin. This curve is
referred to as its tantrix. If the original curve is closed so is its tantrix and the writhe
is proportional to the area enclosed by the tantrix, being the average number of signed
crossings one sees from all possible directions. For an open curve it can be proved that a
consistent definition of the writhe is obtained if one can close its tantrix with a geodesic
circle on a unit sphere
In the restricted case, where the shape of DNA in space satisfies some additional
constraints a second theorem due to Fuller states that the writhe of a DNA configuration
can be computed by considering an auxiliary curve and calculating the writhe around
16 A. V. Vologodskii and N.R. Cozzarelli, Conformational and thermodynamic properties of supercoiled
DNA, Ann. Rev, Biphys. Biomol. Struct. 23 609 (1994).

17 E.L. Starostin, On the writhe of non-closed curves, in: Vol. 36: Physical and Numerical Models in Knot
Theory Including Applications to the Life Sciences, ISBN 978-981-256-187-9, World Scientific 525 (2005).
27
that curve 18 . By judiciously selecting this auxiliary curve and taking due care of th
elimitations of this formula substantial simplifications can be wrought into the calculation
of writhe.
2.6.6 A local expression for writhe
Choosing the auxiliary curve to be a line in the z direction, thus its director is given by
ez = (0, 0, 1), the writhe of a curve of length L, that need not be closed, can then be
written in a local form of a single integral over the length of the curve
Z L
1 ez ⇥ t(s)
Wr = · ṫ(s)ds. (57)
2⇡ 0 1 + ez · t(s)
In order for this form of writhe to be valid, the original curve has to be smoothly trans-
formable without any self-intersections into the auxiliary curve in such a way, that the
inequality ez · t(s) 6= 1 is always satisfied, i.e. the director of the curve should never be
pointed in the direction exactly opposite the direction of the auxiliary curve. This means
that the original curve can not have any antipodal points with the auxiliary curve, which
turns out to be quite a stringent requirement and there are some fundamental issues in the
way it is or is not fulfilled for several common DNA configurations 19 . Nevertheless the
two writhe formulas are always modulo 2.
Let us evalulate the local form of the writhe in Euler parametrization. Since the unit
director of the curve t(s) is a unit vector, it can be parameterized via the Euler angles,
✓(s), (s), (s) within the intervals 0, ⇡ and 0, 2⇡, as
t(s) = (sin ✓(s) sin (s), sin ✓(s) cos (s), cos ✓(s)).
Since furthermore in this parametrization
(ez ⇥ t(s)) · ṫ(s) = sin2 ✓(s) ˙ (s) and ez · t(s) = 1 + cos ✓(s),
we obtain for the writhe of the curve

Z L
1
Wr = (1 cos ✓(s)) ˙ (s)ds. (58)
2⇡ 0
Furthermore a unit normal in the Euler parametrization, such that n(s) · t(s) = 0, can be
expressed as
n(s) = (cos (s) cos (s) cos ✓(s) sin (s) sin (s),
cos (s) sin (s) cos ✓(s) cos (s) sin (s),
sin ✓(s) sin (s)) (59)
18 S. Neukirch, E.L. Starostin, Writhe formulas and antipodal points in plectonemic DNA configurations,
Phys Rev E 78 041912 (2008).

19 Which we will not delve into here, but see J. Samuel, S. Sinha and A. Ghosh, Comment on Writhe formulas
and antipodal points in plectonemic DNA configurations, Phys Rev E 80 063901 (2009) and S. Neukirch and E.
L. Starostin, Reply to Comment on Writhe formulas and antipodal points in plectonemic DNA configurations,
Phys Rev E 80 063902 (2009).
28 Rudolf Podgornik
Figure 13. Of the three configurations, A, B and C, B has an antipodal point to the
reference curve, assumed to be the x axis. This antipodal point defined by ez · t(s) = 1
are located at the middle point of the end loop of the plectonemic structure. Adapted from
S. Neukirch and E. L. Starostin, Writhe formulas and antipodal points in plectonemic
DNA configurations, Phys. Rev. E 78 041912 (2008).
and then we obtain for the total twist, defined in Eq. 50, the result
Z Z
1 1
Tw = ds t(s) · (n(s) ⇥ ṅ(s)) = (cos ✓(s) ˙ (s) + ˙ (s))ds. (60)
2⇡ C 2⇡ C
From here it is clear that the expression
⌧ = t(s) · (n(s) ⇥ ṅ(s)) (61)
is nothing but the twist density. Combining now the expression for W r and T w we get
the Fuller theorem in the form 20
Z
1
Lk = T w + W r = ( ˙ (s) + ˙ (s))ds. (62)
2⇡ C
The linking number is thus simply expressed as a difference in the (s) and (s) angles
at the start and end of the curve
1
Lk = ( (L) + (L)), (63)
2⇡
since we can always choose the angles in such a way that (0), (0) = 0. The constraint
on the fixed linking number of a DNA molecule is then obtained by multiplying this
20 B. Fain, J. Rudnick and S. Östlund, Conformations of linear DNA, PRE 55 (1997) 7364.
29
expression by a Lagrange multiplier and inserting it in the complete Hamiltonian of the

chain which is what we will do when we delve into the statistical mechanics of a rod-like
chain.
This simple and elegant result has nevertheless its limitations. It is not correct for just
any shape of DNA, e.g. once a full plectoneme is formed there are antipodal points and
in that case the non-local variant of writhe should be used. Fig. 2.6.6 shows a plectoneme
configuration where the form of the writhe in Eq. 57 would not be valid. Though the full
non-local form can be derived for an arbitrary space curve 21 it is much more difficult
to implement in any calculation and its use is thus severely limited. On the other hand,
for a chain under a sufficiently large external traction the antipodal points can not be
approached by any chain configuration and the validity of the local form of the linking
number would remain effectively valid.
2.6.7 Writhe and magnetic monopoles
Writhe in the form derived above again leads to a very interesting physical analogy based
on the form of the vector potential of a magnetic monopole. This analogy hinges on
rewriting the definition for writhe in its local form, Eq. 58, and is thus relevant only
when this representation is reasonable.
We start with the following form of the writhe that we take over from Eq. 58
Z L Z (L)
2⇡W r = ˙
(1 cos ✓(`)) (`)d` = (1 cos ✓)d . (64)
0 (0)
Now let us recall the magnetic vector potential A(r) for a point unit magnetic monopole
described by a direction vector t. Up to an irrelevant factor of 4⇡1
its vector potential
from elementary magnetostatics equals
r⇥t 1
A(r) = = t⇥r .
r3 r
We now use a generalization of this formula where the magnetic moment is uniformly
distributed along a line (Dirac string C) in the ẑ direction with a unit strength and write
the corresponding expression for the vector potential of the string occupying the negative
z axis 22 . We can obviously write the unit vector potential with a line integral that can be
evaluated in spherical coordinates (r, ✓, ) to yield
Z
1 ẑ ⇥ r
A(r) = d` ⇥ r = (65)
C |r `| r(1 + ẑ · r)
where we took into account that ` = ẑu, with C : 1 < u  0. Introducing now
the sperical coordinate system (r, ✓, ) with appropriately defined polar and azimuthal
angles, we can rewrite Eq. 65 as
✓ ◆
1 cos ✓
A(r) = (Ar , A✓ , A ) = 0, 0, (66)
r sin ✓
21 E. Starostin, On the writhing number of a non-closed curve. In Calvo, J., Millett, K., Rawdon, E., Stasiak,
A. (Eds.). Physical and numerical models in knot theory including applications to the life sciences ( pp.525-
545). World Scientific Publishing (2005).
22 H. Kleinert, Multivalued fields in Condensed matter, Electromagnetism and Gravitation, World Scientific
(2008).
30 Rudolf Podgornik
where the integration was performed over the negative z axis. Evaluating the correspond-
ing magnetic field
r
B=r⇥A= 3
r
one can thus see that the magnetic potential Eq. 66 corresponds to a point Dirac magnetic
monopole at the origin.
The form of the Dirac monopole vector potential allows us to calculate its line integral
along a finite length segment exactly, yielding
Z L Z (L) Z (L)
A(r) · dr = A r sin ✓d = (1 cos ✓)d = 2⇡W r. (67)
0 (0) (0)
Formally this is the same as the integral derived in Eq. 64. We see that indeed the vector
potential corresponding to a unit Dirac magnetic monopole at the origin yields the same
form of interaction as the local writhe. Though maybe conceptually arcane, we will
later see that this analogy pops up directly in statistical mechanics of an Euler-Kirchhoff
filament. We should stress that this analogy hinges upon the local form of the writhe and
is therefore limited by its regime of applicability.
31
3 MOLECULAR STRUCTURE OF DNA

The breakthrough in the elucidation of the molecular structure of DNA came in 1953
with the experimental work of R. Franklin and R. Gosling as well as model building of
J.D. Watson and F.H.C. Crick. Since the story of this discovery has been often told I will
not delve into it more closely 23 .
Two structural probes, both relying on light scattering at different wavelengths, con-
tributed essentially to the discovery of the structural details of DNA. Light scattering in
the visible regime will described in a later section, while here I will delve solely with
X-ray scattering in the angstrom regime. Since the work of von Laue it is clear that X-
rays are the ideal probe for structural elucidation of crystals and since phosphates along
the DNA backbone are strong scatterers putting the two together would seem like a nat-
ural thing to do. However, long native DNA does not make crystals. It does however
make relatively ordered fibers that scatter X-rays quite well. A fiber would be an ax-
ially orientationally ordered molecular array, with variable amounts of positional order
but with ideally complete polar disorder. That means that the relative axial orientation
of one DNA molecule in the array with respect to any other one would be completely
disordered. Though this is not rigorously true 24 , it is usually assumed it in the derivation
of the scattering intensity of a discrete single and double helices.
An analysis of the X-ray scattering intensity from B-DNA fiber scattering reveals that
• DNA is composed of a periodic array of molecular residues that scatter X-rays

quite well. In the Watson-Crick model these residues are the base pairs comprised
of two phosphate groups and two nitrogen bases on opposite strands held together
by hydrogen bonds. This periodic array gives rise to a repeat of every 10 equatorial
layer lines in the scattering intensity giving 0.34 nm for the separation between the
scattering residues.
• DNA is a helical molecule. The helical pitch of the molecule gives rise to equidis-
tant equatorial layer lines. The separation between the lines leads to a value of
the helical pitch equal to 3.4 nm. There are 10 helical pitch layer lines inside a
periodical residue repeat lines. That means that there are 10 base-pair residues per
pitch of the helix.
• DNA is an asymetric double helix, where the two strands are not symmetric with
respect to the axis of the molecule but are displaced by 3/8 of the complete turn.
This leads to a destructive interference of the 4th meridional layer line which is
missing in the scattering intensity.
• Because of the helical geometry the maxima of the scattering intensity along con-
secutive meridional layer lines make a St. Andrea’s cross which is periodically
repeated within the scattering intensity pattern. From the angle of the St. Andrea’s
cross one can calculate the diameter of DNA that amounts to about 2 nm meaning
that on the average DNA has a cylindrical shape.
23 J.D. Watson, The Double Helix: A Personal Account of the Discovery of the Structure of DNA, Touchstone
(2001). B. Maddox, Rosalind Franklin: The Dark Lady of DNA, HarperCollins; 1 edition (2002).
24 A.A. Kornyshev, D.J. Lee, S. Leikin, and A. Wynveen, Structure and interactions of biological helices,
Reviews of Modern Physics 79 943 (2007).

32 Rudolf Podgornik
The detailed confirmation of the double-helical asymetric molecular model of DNA had
to wait for atomic resolution scattering from DNA crystals which was accomplished
for short fragment (dodecamere) DNAs. All the details of the model were completely
vindicated by the atomic resolution scattering experiments by Dickerson and Drew in
1981 25 . The elucidation of the iconic molecular structure of DNA amounts to one of the
most momentous discoveries in the science of 20th century if not overall.
3.1 X-ray scattering

The most commonly used probe to investigate structural properties of DNA is X-ray
scattering. The structure determinantion of crystals was initiated by the work of von Laue
in 1912 who demonstrated that the diffraction pattern of sphalerite (ZnS) on photographic
film was due to diffraction of very short wavelength electromagnetic radiation (about
1.5 Å) from a regular arrangement of atoms in the crystal. A fairly ordered sample,
preferrably a crystal, is needed in order to get detailed information on the positions of
atoms. This is true also for DNA if one wants to determine the positions of all the atoms
in the molecule. When a sample of material is exposed to a monochromatic beam of X-
rays of wavelength , one can observe the square of the molecular scattering amplitude
, given by the Fourier transform of the density as
Z
F(Q) = ⇢(r) ei Q·r d3 r, (68)
(V )
where Q is the difference between the wave-vectors of the incoming and scattered radia-
tion, Q = k k0 with |k| = |k0 |, so that the magnitude of Q is given by |Q| = 2k sin ✓2 ,
with ✓ being the scattering angle.
In general the structure factor is a complex quantity, with a magnitude and a phase.
Usually one can only detect the magnitude or the scattering intensity given by
I(Q) = F(Q)F ⇤ (Q) = |F(Q)|2 . (69)
This is true only if the scatterers are distributed regularly in space, like in a crystal. If the
scatterers are disordered, as in the case of e.g. liquids, then the scattering intensity has to
be averaged over the statistical distribution of scatterers and thus
⌦ ↵
I(Q) = |F(Q)|2 . (70)
Assume now that we have N pointlike scatterers in crisytalline arrangement located at

rn , that contribute to the scattering amplitude. In this case, the density ⇢(r) is given by a
sum of sharply peaked delta functions. If we denote by fn (Q) the scattering amplitude
of a single scatterer, we obtain for F(Q)
N
X
F(Q) = fn (Q) ei Q·rn . (71)
n=1
25 H.R. Drew et al., Structure of a B-DNA dodecamer: Conformation and dynamics, Proc.Nati.Acad.Sci.USA
78 2179 (1981).
33
Figure 14. Simulated X-ray diffraction from the mineral zoisite, which was first de-
scribed by Sigismund Zois in 1804. Chemically it is calcium aluminium hydroxy silicate
- Ca2 (Al OH) Al2 (SiO4 )3 ). The intensity fades for larger values of the Miller indices
because of the finite size of the sample and thermal motion of the atoms. The wavelength
of the X-rays is = 1.5 . The crystal thickness in this simulated scattering experiment is
50 .
Let us assume for now that we have a line with 2N + 1 identical scatterers, so that
fn (Q) = f0 (Q), separated by a along the line. In this case we can easily sum the above
series explicitly, obtaining
!
Qa 2
sin (2N + 1) h
I(Q) = lim |f0 (Q)|2 Qa
2
= |f0 (Q)|2 (Q 2⇡ ). (72)
N !1 sin 2 a
In the limit of a very large number of scatterers, N ! 1, the only term remaining in
the above expression is
Qa = 2⇡h,
where h is an integer. This analysis can be easily generalized to three dimensions. As-
sume that the scatterers are distributed periodically with distances lattice base vectors
a, b and c so that positions of the scatterers in 3D space are given as
rnmj = na + mb + jc, (73)
with n, m, j integers. By analogy we obtain that for a macroscopic sample of these
scatterers in 3D that the only terms remaining in I(Q) are those satisfying
Q·a = 2⇡ h
Q·b = 2⇡ k
Q·c = 2⇡ l, (74)
where integers h, k, l are reffered to as the Miller indices. These equations, referred
to as the von Laue conditions, define the directions at which one can observe scattered
intensity in a macroscopic sample. It furthermore follows that for the von Laue conditions
ei Q·rnmj = 1 (75)
34 Rudolf Podgornik
and thus that Q is a reciprocal lattice vector. It can be written as
Q = Qnmj = na⇤ + mb⇤ + jc⇤ , (76)
where again n, m, j are integers and a⇤ , b⇤ and c⇤ are basis vectors for the reciprocal
lattice, given by a⇤ = 2⇡(b ⇥ c)/(a · (b ⇥ c)) and analogously for b⇤ and c⇤ . It follows
from the construction that e.g. a⇤ is perpendicular to the plane defined by (b ⇥ c). The
von Laue scattering condition thus demands that the scattering vector Q should be equal
to one of the reciprocal lattice vectors Qnmj . The non-zero intensity of scattering of a
crystal can thus be only expected for very special orientations of the crystal w.r.t. the
incoming beatm as well as w.r.t. the detector. The von Laue condition is thus a kind of a
selection condition.
Reciprocal lattice vectors Qnmj are perpendicular to to a set of parallel equidistant
planes defined by the original (Bravais) lattice. These planes can be obtained by the
Miller construction. Take the original unit lattice vectors a, b and c and divide them by
Miller indices of the planes h, k and l. For different values of the Miller indices one
thus obtains three points in the original lattice that define a plane. Each of these planes is
perpendicular to Qhkl . This can be seen as follows: vectors (b/k a/h) and (c/l a/h)
are both perpendicular to Qhkl and they also lie in the plane with Miller indices h, k and
l since
b a c a
Qhkl · ( )=0 and Qhkl · ( ) = 0. (77)
k h l h
We thus know that Qhkl is perpendicular tho the Miller planes. If the separation between
these planes is denoted as dhkl then we have by definition of Qhkl that
c b a
2⇡ = Qhkl · = Qhkl · = Qhkl · = dhkl |Qhkl |. (78)
l k h
The magnitude of Qhkl is therefore a multiple of 2⇡/dhkl , where dhkl is the separation
between crystal planes with Miller indices hkl. This planes are also referred to as the
Bragg planes. Remembering the von Laue condition Qhkl = Q = k k0 and the
fact that for elastic scattering the allowed wave vectors should lie on the Ewald sphere
defined as
|k| = |k0 |,
we obtain that
⇡
Q = |Qhkl | = 2|k| sin ✓ = 2|k| cos ( ✓). (79)
2
Here ⇡2 ✓ is the angle between the incoming wave and the normal to the Bragg layers.
Putting Eq. 78 together with Eq. 79 we remain with the Bragg condition
n = 2dnmj sin ✓. (80)
This can be derived usually by considering the reflection of incoming waves on the
crystal planes and thus along completely different lines from the von Laue scattering
condition, as first shown by William H. Bragg and his son William L. Bragg in 1913.
Nevertheless the Braggs’ derivation shows that the von Laue and the Bragg formulation
are thus completely equivalent, in fact the Bragg condition presents a simplification of
the von Laue condition.
35
We have yet to determine the scattering intensity at these angles. The corresponding
scattering intensity is given as
X
F(Q) = f0 (Q) ei (hn+km+lk) = f0 (Q)FL (Q). (81)
nmk
Here we introduced also the scattering intensity of the crystalline lattice as

X
FL (Q) = ei (hn+km+lk) = N (Q · a 2⇡ h) (Q · b 2⇡ k) (Q · c 2⇡ l). (82)
nmk
The expression Eq. 81 would be exact for atomic crystals with a single atom in a unit
cell. In molecular crystals on the other hand, the unit cell is more complicated and
in general contains many atoms. In this case the atomic density is a convolution of the
electron density ⇢c (r) of a unit cell and the periodic distribution of the unit cells in space.
Assume now that the scattering intensity pertaining to the unit cell itself is given by
Z
Fc (Q) = ⇢c (r) ei Q·r d3 r. (83)
(Vc )
Since furthermore the convolution in real space is equivalent to a product in the Fourier
space the total scattering intensity of the molecular crystal is given by
X
F(Q) = Fc (Q) ei (hn+km+lk) = Fc (Q) FL (Q). (84)
nmk
To obtain the last expression, we took into account the von Laue conditions. If we know
the structure factor of the unit cell, we can thus calculate the scattering intensity of the
whole macroscopic crystal. A typical X-ray scattering intensity from a mineral crystal
zoisite is presented on Fig. 3.1
3.2 Structure factor of a linear polymer chain

For linear polymer chains the conclusions derived above assume a specific form, dictated
by the nature of the periodicity of the polymer chains. Let us assume first that we have
thin linear polymers in the direction of z-axis, where periodic repeat distance along the
chain is P , therefore the scattering residues are located at zj = jP . These can be atoms,
molecular groups or periodic segments such as the pitch of DNA. In this case Eq. 85
assumes the form
X +1 ✓ ◆
2⇡ X 2⇡`
F(Q) = Fc (Q) ei Qz jP = Fc (Q) Qz , (85)
P P
nm` `= 1
where we used the identity

+1
X +1
X
eı2⇡nx = (x m).
`= 1 `= 1
The Qz component of the wave vector can thus assume only discrete values in the form
of equatorial planes separated by 2⇡
P . The intensity of each diffraction line is given by
|Fc (Q)|2 .
36 Rudolf Podgornik
The periodicity along the Qz direction dictated by the perodicity of the linear polymer
chain is a general feature of the scattering by polymer molecules and we refer to these
equatorial planes perpendicular to the z direction as Polanyi layers, introduced first in
the study of X-ray diffractions of cellulose 26 . They are observed also in the case of long
helical molecules, among several other features that we investigate next. Because for
elastic scattering the Polanyi layers have to intersect the Ewald sphere the intersections
are not planar, but have a distorted shape, in fact they are hyperbolas.
3.3 Structure factor of a single helix

For the long DNA molecule the X-ray scattering is performed on a semicrystalline sample
as in the classical experiments of Franklin and Gosling, or on a crystalline sample as in
the case of the experiments of Arnott et al.. In the first case the ordering of DNA is
enforced by pulling of a dense DNA fiber so that all the molecules have (approximately)
the same orientation of the long axis. We will build up a theory of X-ray scattering of
DNA fibers in a stepwise manner finally converging on the complete interpretation of the
fiber scattering data.
Figure 15. Left: Six turns of a discrete single right-handed helix. The helix (yellow) and
the discrete scattering residues (blue) (base pairs for DNA) are clearly visible. Right: Six
turns of a discrete double right-handed helix. In both cases the thickness of the imaginary
helical line has again been exaggerated for rendering purposes.
Let us first deal with a single helix. The electron density distribution is given by a
convolution of a continuous helix of radius R and a one dimensional periodicity in the
direction
Saturday, of6, 2010
November the helical axis, assumed to be the z axis, of a period or pitch P and an array
of scattering residues of period h along the helix, as represented on Fig. 15. The equation
of the helix per its period is given in the cylindrical coordinates as Eq. ??
x = R cos
26 M. Polanyi, The X-ray fiber diagram. Z. Physik 7 149 (1921). M. Polanyi, My time with x-rays and
crystals (1962), 97âĂŞ104 in Marjorie Grene, ed., Knowing and being: Essays by Michael Polanyi (Chicago,
1969), 97.
37
y = R sin
P
z = , 0< < 2⇡. (86)
2⇡
We assume that the helix has a periodicity in the z direction given by the pitch P , while it
is also composed of a periodic array of scatterers (base pairs for DNA) with a neighbor-
neighbor separation h along the z axis. The position of the jth scatterer can thus be
written in the cylindrical coordinate system (r, , z) as
2⇡(zj z0 )
rj = (R, 0 + , z0 + jh), (87)
P
where (r, 0 , z0 ) is taken as an arbitrary origin. We now use Eq. 83 together with the
Jacobi-Anger expansion of the plane waves in the cylindrical coordinate basis
X ⇡
eıQ·r = eın( + 2 ) Jn (QR)e ın eıQz z
n
where Jn (x) is the standard cylindrical Bessel’s function of order n given by the follow-
ing integral representation
Z 2⇡
1 +⇡
eix cos (✓ )+in✓
d✓ = Jn (x) ein( 2) , (88)
2⇡ 0
we have written the vector Q as Q = (Q cos , Q sin , Qz ). Since the scatterers are
distributed along a thin line, with positions given by Eq. 87, the only integral that remains
in the volume integral of Eq.83 is given by d3 r ! ⇡R2 dz = ⇡R2 2⇡ P
d . Evaluating
the integral in Eq. 83 we remain with the following structure factor as first derived by
Cochran, Crick and Vand 27
1
X ⇡ 2⇡n 2⇡m
F(Q) = Fc (Q)eıQz z0 Jn (QR) ein( 0+ 2 )
(Qz ). (89)
n,m= 1
P h
The separation between the equatorial layer lines p can now be obtained from the follow-
ing selection rule
`(n, m) n m
= + .
p P h
For B-DNA with P = 3.4 nm and h = 0.34 nm this selection rule is reduced to
`(n, m) = n + 10 m and p = P.
In order to understand the consequences of the CCV structure factor one needs to recall
that the Bessel functions are oscillatory functions of their argument, Jn (x) ⇠ 0 for all
values of the argument such that x  n and J n (x) = ( 1)n Jn (x).
Let us analyze the position of the layer lines for 0 < ` < 10. The selection rule
allows the following balues of the indices n and m
27 Cochran, W.; Crick, F. H. C.; Vand, V. The structure of synthetic polypeptides. I. The transform of atoms
on a helix, Acta Crystallogr 5 581-586 (1952).

38 Rudolf Podgornik
` 0 1 2 3 4 5 6 7 8 9 10
n 0 -10 1 -9 2 -8 3 -7 4 -6 5 -5 -4 6 -3 7 -2 8 -1 9 0 10
m 01 01 01 01 01 01 01 01 01 01 01
Jn J0 J 10 J1 J 9 J2 J 8 J3 J 7 J4 J 6 J5 J 5 J 4 J6 J 3 J7 J 2 J8 J 1 J9 J0 J10
Table 2. A table of relevant Bessel function in the CCV structure factor formula for
0 < ` < 10.
As we trace the layer lines starting from ` = 0 we first note that the maximum of
the Bessel function first moves outward from the meridional line until ` = 5 and then
it moves back in. This behavior is periodic in ` with a period of 10 lines and leads to
a diamond-like first maximum pattern exhibiting St. Andrea’s cross. The angle of the
cross is obtained from the asymptotic form of the Bessel functions for large arguments:
r ⇣
2 n⇡ ⇡ ⌘
Jn (z) ' cos z + O(|z| 1 ). (90)
⇡z 2 4
The asymptotic position of the maximum of the square of the Bessel function for large n
is then at
P
' arctan .
4R
This angle can thus be used to determine the ratio between the pitch and the radius of the
molecule.
The form of the CCV structure factor tells us that we obviously still have the Polanyi
layers in the equatorial planes. They are due to the total periodicity of the helix along
the z axis, composed of the pitch as well as the periodicity of the scattering centers. On
each of the Polanyi layers we have another modulation of the scattering intensity given
by the behavior of the Bessel function. The scattering intensity is obviously symmetric
with respect to the zero order Polanyi layer (defined with n = 0). Each of the Polanyi
layers is modified by a series of scattering maxima corresponding to the maxima of the
Bessel functions for that line. By general properties of the Bessel functions the intensity
of these maxima is descreasing with the index of the line. Also for a given Polanyi line
n the first maximum is displaced towards larger values for larger n. These features in
general lead to a St. Andrea’s cross shape of the scattering intensity that is repeated in
the scattering plane in the form of a diamond (see Figure).
3.4 Scattering intensity of an orientationally averaged single helix

The structure factor of a long helical line contains the arbitrary phase corresponding to
the orientation of the origin of the helix in the plane perpendicular to its axis. But since
the observable is the scattering intensity and not the structure factor, let us now look at
the scattering intensity itself, defined as
I(Q) = <|F(Q)|2 >. (91)
The average in the above equation depends on the details fo the sample probed by X-rays.
ssuming that all the helices in the sample are parallel - fiber geometry - and that they are
orientaionally uncorrelated the average in Eq. 91 is obvipusly over the polar angle .
39
Since the square of a complex number is its product with a complex conjugate, the
total scattering intensity can be written as
1
X
+⇡ 2⇡`(n, m)
|F(Q)|2 = |Fc (Q)|2 Jn (QR) ein( 2) (Qz )
n,m= 1
p
1
X
in0 ( + ⇡ 2⇡`(n0 , m0 )
Jn0 (QR) e 2) (Qz ). (92)
p
n0 ,m0 = 1
3D graf narisan v pythonu:
Še ContourPlot narisan z Mathematico.
15
10
F(Q) Q|| 0
kz
10
15
10 5 0 5 10
k
Q⊥
Slika 1: Sipanje na enojni verigi DNA
Q⊥
Slika 2: Sipanje na enojni verigi DNA
Tu ne gre za pravi trodimenzionalni prikaz, ker sem le grafe za posamezno

vrednost kz premaknil vzdolž y - osi in zato vrednosti sipalne amplitude S(k)
niso relevantne. Ta oblika predstavitve se mi je zaradi preglednosti zdela boljša
Figure 16. The structure factor Eq. 89 as a function of the wave vector for a single
kot klasičen 3D prikaz.
helix with P = 3.4nm, h = 0.34nm and R = 1nm. Curves for different Qz have been
displaced along the y axis. (Left) Structure factor for different Qz have been displaced
alongAprilthe
Thursday, y axis. (Right) Contour plot of the scattering intensity as a function of longi-
28, 2011
tudinal and transverze components of the scattering wave vector. One can discern the
emergence of St. Andrea’s 3 cross for asymptotically large values of n in the scattering
intensity of continuous helical lines. Courtesy of L. Curovic
4
In a crystalline sample all the angles would be set and frozen. In a semi-crystalline
sample such as DNA fibers, the distribution of angles is not set and is rather disordered.
If we assume that it is completely disordered the observed scattering intensity should be
an average of I(Q) over all the angles. Dropping inessential constants we thus obtain
1
X 1
X
n0 ) ⇡
hI(Q)i = |Fc (Q)|2 Jn (QR)Jn0 (QR) ei(n 2
n,m= 1 n0 ,m0 = 1
Z 2⇡
2⇡`(n, m) 2⇡`(n0 , m0 ) 1 n0 )
(Qz ) (Qz ) ei(n d .
p p 2⇡ 0
(93)
The above integral is obviously zero, unless n = m which furthermore implies that there
40 Rudolf Podgornik
is no difference between m and m0 either. In this case one then obtains

1
X 2⇡n 2⇡m
hI(Q)i = |Fc (Q)|2 Jn2 (QR) (Qz ). (94)
n,m= 1
P h
Note that in this context the prodict of the two delta functions is the same is one single
delta function, i.e. it simply sets the periodicity along the Qz direction. This is what is
actually plotted on Figs. 3.4 and 3.4. Analogous expression could be abtained also for a
discrete, commensurate or incommensurate, helix.
The above analysis apllies to integer as well as non-integer helices, emaning that a
helix can have either an integer number of residues per pitch or not. First, let us take a
helix with an integer number, k, of discrete residues per pitch, thus Ph = k. In this case
the scattering condition on Qz becomes
2⇡n 2⇡m 2⇡
Qz = + = (n + km). (95)
P h P
Let us assume that k = 3. Then the Qz = 0 line will contain the sum over all n = 3m.
The same is true if now Qz is an integral multiple of 3, i.e. Qz = 3p, where p is also an
integer. In this case n = 3(p m), which again goes through the same values as 3m.
That of course signifies that the scattering amplitude has a repeat period of 3 along Qz .
The spacings between layer lines are in all cases given invariably by 1/P . The scattering
amplitudes for other values of Qz can be obtained analogously. Everything said so far
can be discerned from the Figure.
If on the other hand the helix does not have an integer number of residues per pitch,
then the situation is a bit more complicated. Let us assume that there are k residues per
V pitches. The discreet helical pattern thus has a repeat period of V P and the spacing
between adjacent residues is obviously h = V P/k. In this case the equation for Qz
assumes the form
2⇡n 2⇡m 2⇡
Qz = + = (V n + km). (96)
P h VP
For V = 1 this of course reduces to Eq. 95. Since V, n, k, m are all integers, V n +
km must also be an integer. For any given k, V one can find values of n, m such that
V n + km = 0, ±1, ±2 . . .. This signifies that the layer spacing in this case has to be
V P . The periodicity in the Qz direction is obtain by the following argument: if we write
1
V n + km = kj, where j is again an integer, then n = k(j m)/V , where j m has

to be chosen such, that it makes n an integer, too. The allowable values of j m must
therefore be independent of j. It follows that the periodicity in the Qz direction must be
k.
3.5 Scattering intensity of a discrete double helix

Let us now analyze the scattering from a double helix. Consider first a contiuous helical
line. Since the scattering density of the two lines is additive, so is - by definition -
the scattering amplitude. Let us assume that the two helical lines are shifed one with
respect to the other by an angle in the plane perpendicular to the long axis. The total
Formuli CCV, ki je napisana v prejšnjem razdelku dodamo še faktor:
G(f ) = 1 + e2i⇡f l , (4)
kjer lahko namesto l vstavimo tudi l = P2⇡

kz
in za f vstavimo vrednost 3/8.
Zopet spreminjamo kz = 2⇡j/P za j [ 10, 9, .., 9, 10], izračunamo A(k)
vzdolž osi k? in narišemo grafe A A = S(k).
41
⇤
15
10
F(Q) Q||Q||
kz
0
10
15
10 5 0 5 10
Q⊥
Slika 3: Sipanje na dvojni verigi DNA - 3D prikaz. Z rdečo sta označeni črti, ki
Q⊥
Slika 4: Prikaz z izohipsami. Tudi tu vidimo, da 4. črta manjka.
zaradi zamika f = 3/8 izgineta.
Figure 17. The structure factor Eq. 89 as a function of the wave vector for a double
helix with P = 3.4nm, h = 0.34nm and R = 1nm. The two helices are displaced by
4 around the long axis of the molecule. (Left) Structure factor for different Qz have
3⇡
beenAprildisplaced
Thursday, 28, 2011 along the
5
y axis. (Right) Contour plot of the scattering intensity as a
function of longitudinal and transverze components of the scattering wave vector. One
can discern the emergence of St. Andrea’s cross for asymptotically large values of n in
the scattering intensity of continuous helical lines. The fourtn line is missing because of
the destructive interference in the structure function. Courtesy of L. Curovic.
6
scattering amplitude of this double helix is thus obtained as

1
X
+⇡ 2⇡`(n, m)
F(Q) = Fc (Q) Jn (QR) ein( 2) (Qz )(1 + ein ). (97)
n= 1
p
1
By the famous Euler formula ei(m+ 2 )2⇡ + 1 = 0 we obtain a total extinction of the
n th line if and only if
n = (m + 12 )2⇡. (98)
Obviously when is half a turn, = ⇡, all the odd numbered lines dissappear
since in that case n = 2m + 1. If we would like the k th line to dissappear, then the
corresponding relative angle between the two helices should be
(m + 12 )2⇡
. = (99)
k
For DNA fibers one observes empirically that the 4 th line is extinguished. This would
correspond to mutual orientation angle of 4 , 4 , . . . The smallest angle 4
= ⇡4 , 3⇡ 5⇡ ⇡
is unacceptable in the case of DNA because of the sterical resons since the two helices
would have to interpenetrate. The first acceptable value is thus 3⇡ 4 . This means that the
two helices are displaced angularly by 3/8 of the helical pitch.
If we would have a mutiple, e.g. L-tiple helix, we would need to insert a factor
containing a sum over all the strands. For many strands this would lead to a complete
cancellation of scattered intensity except for the n = 0 equatorial line.
42 Rudolf Podgornik
15
10
2π/P
5
δ 2π/h
kz
0
10
15
10 5 0 5 10
k
Slika 4: Prikaz z izohipsami. Tudi tu vidimo, da 4. črta manjka.
Figure 18. The Franklin-Gosling experimental diffractogram (Photo 51) of B-DNA and
theoretical scattering intensity comparison for a double helical, discrete helix. The pri-
mary structural parameters deducible form diffractogram are: P, h, = arctan 2⇡R P
,
and .
A double helix with the same angular displacement as DNA would thus have the
scattering amplitude of
Thursday, April 28, 2011 1
X
+⇡ 2⇡`(n, m) 3⇡
F(Q) = Fc (Q) Jn (QR) ein( 2) (Qz )(1 + ein 4 ). (100)
n,m= 1
p
For DNA with the mutual orientational angle of the two 6 strands in the double helix
equal to 3⇡
4 and addditionally taking into account the discrete nature of the DNA helix
as schematically represented on Fig. 15 , the corresponding orientationally averaged
scattering intensity assumes the final form
1
X n m
hI(Q)i = |Fc (Q)|2 Jn2 (QR) (Qz )(1 + cos n 3⇡
4 ). (101)
n,m= 1
P h
This is the prediction that we have to compare with Franklin - Gosling scattering inten-
sity. For DNA we have P = 34 Å and h = 3.4 Å. The correposndence between the main
features of Eq. 101 and the measured X-ray scattering intensity of DNA are preseneted
on Fig. 69. We can discern the St. Andrea’s cross shape due to the helical symmetry
of the molecule and the missing fourth layer line because of the relative orientation of
the two helices in the double helical DNA which are displaced by an angle of 3⇡ 4 . While
Crick and Watson were not aware of this peculiar feature of the B-DNA diffraction pat-
tern, Franklin and Gosling noted the extinction of the 4th order line and duly drew from
it the correct conclusion on the shift between the two strands 28 .
This creates a major and a minor grove in the DNA. The layer lines of the 34 Å DNA
pitch, the layer lines of the 3.4 Å nucleotide repeat along the axis of the double helix
28 A. A. Lucas, A-DNA and B-DNA: Comparing Their Historical X-ray Fiber Diffraction Images, J. Chem.
Educ. 85 737 (2008).

43
and the cross angle that via = arctan 2⇡R P

provides the radius of the DNA R = 10
Å. It also follows immediately that the periodicity of DNA is ten residues per pitch. All
this is valid for the B-form of DNA which is the common variety in ordinary solution
conditions.
3.6 Details of B-DNA structure

Different atoms scatter remarkably differently. While hydrogen (1) scatters only barely,
carbon (6), nitrogen (7) and oxygen (8) that make up the bulk of biological matter in-
cluding DNA, scatter X-ray with comparable strengths, whereas phosphorus (15) which
is present in the phosphate (P O4 ) groups of the DNA backbone, scatters approximately
four times more intensely than all the other relevant atoms.This feature of the phosphate
group scattering allows for a rather straightforward interpretation of the X-ray scattering
fetures of DNA.
X-ray scattering off of DNA, described in details above, gives us thus the positions of
the main scatterers along DNA which are the phosphate groups making up the phosphate
backbone of the molecule. The backbone is formed by a sequence of phosphate groups
and deoxy-ribose sugars, and is oriented because on one side of the sugar the phosphate
group is linked to a carbon atom that does not belong to the sugar ring while, on the
other side, it is linked to a carbon atom which is part of the sugar ring. A sequence of
nucleotides, i.e. monomer units of DNA, is then attached to the phosphate backbone. A
nucleotide is composed of three fundamental parts: a P O4 phosphate group, a five-atom
cyclic sugar ring and a nitrogen-rich base, that can consist of either one or two nitrogen
containing rings. The two types are referred to as purine and pyrimidine bases.
The DNA backbone, formed by a sequence of phosphates and sugars, is oriented
because on the "incoming" side the phosphate is linked to the 5’ carbon of the sugar ring
while, on the "outgoing" side it is linked to the 3’ carbon of the sugar ring. By convention,
since this is the way the synthesis of DNA in vivo is conducted, the directionality of DNA
is assumed to be 50 ! 30 . Schematically DNA could thus be represented as two directed
helices running in opposite directions, see Fig. 19.
The four nitrogen-rich bases are: adenine, cytosyne, guanine and thymine (A,C,G,T)
forming a monophosphate-deoxyribonucleotide usually abbreviated as dAMP, dCMP,
dGMP, dTMP. The four bases belong to two different types: purines (A,G) and pyrim-
idines (C,T). The realization that in double-stranded DNA the nitrogen bases assemble
in pairs through hydrogen bonds was one of the important steps in the elucidation of the
DNA structure. The hydrogen bonded pairs are formed always by a purine and a pyrim-
idine and are almost matched in size thus providing a very regular periodic backbone
structure to the two phosphate chains constituting a double-stranded DNA molecules.
The molecular weights (MW) of these four monophosphate-deoxyribonucleotides are
not identical. In fact
MW(dAMP) = 331.2 MW(dCMP) = 307.2

MW(dGMP) = 347.2 MW(dTMP) = 322.2
but we usually average the molecular weights of all four to 330 daltons per nucleotide
44 Rudolf Podgornik
29
in DNA (=330 grams/mole of individual nucleotides). Since DNA is double stranded,
i.e., one nucleotide pairs with a nucleotide on the opposing strand, then we can say the
average molecular weight of a nucleotide pair in DNA is 660 daltons.
Figure 19. A schematic representation of the Watson-Crick DNA double-helix. The two
antiparallel phospahate backbone are joined by the hydrogen bonds between purine (one
nitrogen containing ring) and pyrimidine (two nitrogen containing rings) bases across
the symmetry axis of the molecule. Graphics courtesy of A. Siber.
Purine-pyrimidine pairs of bases, viz. guanine and cytosyne, adenine and thymine,
make base pairs, where attractive hydrogen bond interactions hold the bases together
giving the molecule its transverse stability. The base pairs are in planar configurations
with ⇡-bond electron and hydrophobic base stacking interactions between them stabi-
lizing the double helix longitudinally.
These stacking interactions are mostly responsible for the helical nature of the DNA
molecule according to an argument of Calladine and Drew. This argument goes as fol-
lows 30 : The phosphate-phosphate bond along the phosphate backbone has a length of
DP P = 6 Å, basically determined by the structural properties of phosphate and ribose
sugars. This distance is basically set and can not vary. The nitrogen bases are pla-
nar and interact with strong hydrophobic attractions, that keep them at a separation of
DBB = 3.4 Å. They were recently measured directly between bases in a poly(dA) single
stranded DNA 31 where they are the strongest. The estimate of the base-stacking energy
for adenines is ⇠
= 3.2 4.0 kcal per mole of base pairs. If DNA would be aladder of
29 The exact number being 327.0 daltons per nucleotide.
30 C.R. Calladine adn H. R. Drew, Understanding DNA: The Molecule and How It Works, Academic Press
(1992).
31 Changhong Ke, M. Humeniuk, Hanna S-Gracz, and P. E. Marszalek, Direct Measurements of Base Stack-
ing Interactions in DNA by Single-Molecule Atomic-Force Spectroscopy, Phys. Rev. Lett. 99 018302 (2007).
45
Figure 20. Left: the Carradine-Drew argument for the emergence of helix. Directional
ladder leads to a skewed stacking of base pairs,whereas stacking with a twist leads to a
helix. The latter is observed in DNA. Right: The propeller-twist of the bases in a base-
pair alleviates the unfavorable energy due to hydrophobic interaction between bases in
the helical local geometry.
stacked bases that would not satisfy the strong hydrophobic stacking interactions. If the
ladder is tilted , Fig. 20, at an angle that could satisfy the phosphate-phosphate bond
length as well as the stacking separation between bases. However, this type of ladder
conformation is not observed in DNA. The other possibility is that instead of tilting the
ladder of bases in one direction, the bases twist at the same time, making a helical con-
figuration that is indeed observed in DNA. This twisting should be about
Tuesday, November 2, 2010 p
DP2 P DBB 2
2 ⇥ arcsin = 32o
(2 ⇥ R)
that leads to about 11 bases per turn of DNA, which is very close to 10 observed in X-
ray scattering of B-DNA. Because of the twist of the base-pairs around the axis of the
molecule parts of the bases get exposed to water because of uncomposensated hydropho-
bic interactions. The remedy for this is the propeller-twist of the bases that allows the
bases in the base-pairs to reorient in such a way that they minimize the extent of exposure
to water, Fig. 20. Twist and propeller-twist thus act in accord to promote the formation
of the helix with minimum uncompensated hydrophobic interactions between the bases.
There are many other subtle modifications in the geometry of the bases that allow DNA
to adjust its structure and minimize the total free energy 32 .
The canonical structure of B-DNA is presented in Fig. 21 in the space filling, ball
and stick and wireframe models. We see the double helix with assymetrically apposed
phosphate backbones. This assymetry gives rise to the major an minor grove along the
DNA surface. The base-pairs are almost perpendicular to the axis of the molecule and
do not tilt along this axis. Though on the average DNA does resemble a cylinder of
radius 10 Å, it nevertheless contains a lot of accessible space within both groves. The
axial separation between the base pairs is 3.4 Å, and there are 10 base pair per pitch of
periodicity 34 Å.
DNA is the genetic material and carries the blueprint of an organism. In the human
genome DNA is partitioned into 46 distinct fragments in each cell. These fragments are
called chromosomes. Each chromosome is made of a single continuous piece of DNA
whose length is on the order of 4-10 cm. Thus, the genome stored in every single human
32 V. A. Bloomfield, D.M. Crothers, I. Tinoco, Nucleic Acids: Structures, Properties, and Functions, Univer-
sity Science Books; (2000).

46 Rudolf Podgornik
cell has is about 2 m long. A species of salamanders has a genome that is almost 1000
times longer then the human genome.
Figure 21. Space filling, ball and stick and wireframe models of the structure of B-DNA.
The base pairs in the B-DNA have different degrees of freedom that can be quan-
tified by translational and rotational modes of motion of the bases within the DNA du-
plex. These degrees of freedom that show a pronounced variation in fiber diffraction
results have been defined at an international workshop 33 and are given by the following
schematics. Translational motion of the bases within a base pair is quantified by two
displacements perpendicular to the axis of the molecule, consisting of stagger, stretch,
shear, rise, slide and shift. The rotational motion is quantified by the tip, inclination,
opening, propeller twist, buckle, twist, roll and tilt.
3.7 Comparison of B-DNA and A-DNA structure

Apart from the B conformation which is of utmost relevance for DNA in vivo, DNA can
assume everal other types of configurations at appropriate solution conditions that play
a role in vivo. Among them most notably the A-DNA conformation that differs from
B-DNA due to the degree of hydration of the molecule and/or the ionic strength of the
solution. The Z conformation is a left-handed helix, contrary to the two other DNA con-
33 R.E. Dickerson, Definitions and nomenclature of Nucleic Acid Structure Parameters, EMBO J. 8 (1998)
1-4.
47
formations. As such it might be involved in twist relaxation during DNA transcription.

Though all three forms are biologically relevent, In vivo the typical form of DNA is the
B-DNA form.
Figure 22. Osmotic pressure of the DNA subphase as a function of DNA concentration.
The break in the curve signals a transition between two ordered tyeps of DNA arrays as
well as possibly a transition from the B to A conformational forms. The two lone points
at the extremely large osmotic pressure correspond to an incipient breakdown of the A-
DNA helical form. Adapted from S. M. Lindsay et al., The origin of the A to B transition
in DNA fibers and films, Biopolymers 27 1015 (1988).
Fig. 22 shows the equation of state of DNA in an ionic solution with 1 M NaCl. This
dependence, to be elaborated upon in a later chapter, can be obtained by changing the rel-
ative humidity of the sample 34 or by changing the concentration of a crowding polymer
such as PEG, and concurrently measuring the concentration of the in the sample. The
connection between the two types of experiments is provided by the chemical potential
of water µw . On one side the chemical potential of water is given by the standard relation
µw = µ0 + kB T log aw ,
34 S. M. Lindsay, S. A. Lee, J. W. Powell, T. Weidlich, C. Demarco, G. D. Lewen, N. J. Tao, A. Rupprecht,
The origin of the A to B transition in DNA fibers and films, Biopolymers 27 1015 (1988).
48 Rudolf Podgornik
where aw is the absolute activity of water. Furthermore the water activity of the sample
is equal to the relative humidity of air surrounding the aqueous solution in a sealed mea-
surement chamber, i.e. aw = Rh. Since the connection between osmotic pressure ⇧ of
an aqueous solution and the chemical potential of water is given by
V ⇧( s , b)
= (µw µ0 ) = log Rh, (102)
kB T
where V is the volume of a single water molecule in solution. The relative humidity
of a DNA sample can thus be directly associated with its osmotic pressure and thus its
equation of state.
The experiments show that at a certain value of DNA concentration the molecules
probably undergo an B ! A transition as well as a transition from a relatively dis-
ordered liquid (crystalline) phase to an ordered (orthorhombic) crystal. This transition,
observed first by R. Franklin in her seminal X-ray diffraction studies of DNA fibers,
appears to be first order and is associated with a substantial volume change. As we will
delve into the structure and symmetries of DNA condensed phases at a later point, we will
only address the differences between the B and A conformations. The transition between
the two is most clearly indicated by changes in the vibration spectrum in light scattering
studies 22, which remains the strongest evidence for this transition. One notes, Fig. 22,
that after the B ! A the DNA assembly is almost incompressible as the changes in
DNA concentration are almost negligible as the osmotic pressure continues to increase,
see Fig. 22. This attests to the fact that at the DNA densities conducive to the A form
the DNA array must be in a form with pronounced crystalline order. While the transi-
tion from the disordered liquid-crystalline phase to an ordered orthorhombic crystal for
large concentrations of DNA is well established directly through detailed SAXS 35 , the
B ! A is less so. In fact SAXS experimental diffraction patterns at densities higher
then the putative A ! B transition, obtained by Livolant and coworkers, showing a
strong diffuse arc at 0.336 nm could be fitted also with a B-DNA structure and do not
contain a sufficient number of Bragg reflections that would unequivocally point to an A
conformation of DNA.
Ideally, however, the X-ray diffraction intensities of the two conformations of DNA
are strikingly different 36 . First of all the A-DNA diffraction intensity contains peak-like
diffraction maxima in the vicinity of the center of the diffractogram. This is due to the
pronounced amount of positional order in the plane perpendicular to the long axes of the
molecules. In this plane A-DNA at large osmotic pressures obviously retains a highly
crystalline character that is lost at lower osmotic pressures in the case of B-DNA form
where only blurred maxima due to the helical nature of the molecule are retained.
The presence of Polanyi layer lines in the A form of DNA indicates that this con-
formation of DNA too has periodicity along the long axis. The corresponding value of
the pitch is 2.46 nm for A-DNA as opposed to 3.32 nm for B-DNA, being thus about 25
% more compact in the axial direction. In addition there are close to 11 base pairs in
the pitch for A-DNA, as opposed to 10 for B-DNA. The former thus needs to be more
35 D. Durand, J. Doucet and F. Livolant, A study of the structure of highly concentrated phases of DMA by
X-ray diffraction, J. Phys. II France 2, 1769 (1992).

36 For a more detailed description see A. A. Lucas, A-DNA and B-DNA: Comparing Their Historical X-ray
Fiber Diffraction Images, J. Chem. Educ. 85 737 (2008).

49
10
8
4 4
00 0
Figure 23. Comparison of diffractions patters of A-DNA and B-DNA conformations at

75 % and 90 % relative hiumidity, respectively. After R. Franklin (1953).
compact in the axial direction and one of the consequences of the large osmotic pressures
Monday, September 19, 2011
to which A-DNA is subjected is its compactification in the axial direction.

On the other hand the big meridional arcs at 10th layer lines of B-DNA are completely
absent in the case of A-DNA. This can be ascribed to the fact that the base-pair planes
are not parallel between themselves nor perpendicular to the axis of the molecule. The
appearance of the arcs at the 6th, 7th and 8th layer lines furthermore indicates that the
base-pair planes tilt along the axis of the molecule with a maximum at 20 . Of the 11
base pairs that make up a complete turn only 4, i.e. two pairs, are parallel, leading to a
vestigial St. Andrew’s cross for high order (6,7 and 8) diffraction lines. This vestigial
cross allows to determine the radius of the molecule which is slightly larger then in the
B-DNA case, being This indicates that there are fundamental differences between the two
forms that can not be reduces to mere quantitative differences between their respective
geometrical parameters. This differences actually led Franklin away from the (double)
helical form in her attempts to understand the patterns seen in X-ray diffraction without
any a-priori model building.
Another important feature of the A-DNA diffraction pattern is the absence of St.
Andrew’s cross and the emergence of multiple spots that on closer inspection show
consistent splittings for the lowest order lines. The distribution of spots is consistent
with the monoclinic crystalline symmetry of the unit cell that contains a DNA molecule
with phosphate chains running in opposite directions. They also indicate the pronounced
crystallinity of the A-DNA assembly in the plane perpendicular to the long axes of the
molecules. The orthorhombic packing symmetry is obtained when the hexagonal local
symmetry is perturbed and the 2D lattice is deformed in the direction perpendicular to
the sides of the hexagon. This deformation is due to the complicated nature of the in-
teractions between the DNA molecules that do not depend only on the axial separation
50 Rudolf Podgornik
Figure 24. A comparison between A-DNA (left) and B-DNA (right) conformations. The
former one is thicker, more tightly packed in the axial direction, with tilted base pairs
and a hole running down the middle.
between the molecules but also upon the polar angle between them 37 .
The scattering pattern of A-DNA thus attests to a rather different geometry of this
form of DNA, Fig. 24. While B-DNA is ubiquitous in vivo, the A-DNA conformation
presumably does not have a well defined biological role, though a form similar to A-DNA
is adopted by the double helical portions of t-RNA molecules and by the double-stranded
RNA genome of reoviruses. The A-DNA is first of all right handed double helix, but a
bit thicker then the B-DNA form, with a radius of 1.15 nm, compared to 1 nm for the
B-DNA form. This thickening of the molecule leads to a formation of a pronounced hole
running down th emiddle of the helix. It is also compressed along the axis of the molecule
leading to 11 bp per turn as opposed to 10.5 bp per turn for B-DNA, whereas the rise per
base pair is 0.24 nm for A-DNA and 0.34 nm fir B-DNA. This gives the pitch of A-DNA
of 2.46 nm as opposed to 3.32 nm for B-DNA. While in B-DNA the plane of the nitrogen
bases in base pairs is almost perpendicular to the axis of the molecule with an inclination
of only 1.2 , A-DNa has pronouncly inclined base pair planes with an inclination of 19 .
This deformation of the whole molecule as it goes through the B ! A transition thus
leads to a more pronounced surface hydration of the molecule for it to oppose - in the Le
Chatelier sense - the enormous osmotci pressure that pushes the molecules together in an
ordered array.
37 V. Lorman, R. Podgornik and B. Zeks, Positional, Reorientational, and Bond Orientational Order in DNA
Mesophases PRL 87 218101 (2001).

51
4 BASE-PAIR INTERACTIONS AND DNA MELTING

The canonical B structure of DNA is stabilized by various interactions that one could
decompose somewhat arbitrarily into the primary interactions and auxiliary interactions
stabilizing the double helix. The primary interactions act perpendicular to the long
axis of the molecule as well as between the base pairs along the axial direction of the
molecule.. They are the
• covalent bond interactions between the phosphate groups in the phosphate back-
bone running along the long axis of the molecule
• base stacking interactions between the bases in the stack along the axis of the
molecule
• hydrogen bond interactions between complementary bases perpendicular to the
axis of the molecule
Though these are certainly the primary interactions that set the overall feature of the
B-DNA geometry, there are other interactions that further stabilize it and adjust all the
important details of the structure. These auxiliary interactions are composed of
• water-DNA surface hydrogen bonding interactions between the polar moieties

of the DNA surface and the water molecules as well as the
• counterion-mediated electrostatic interactions along the charged phosphate groups
along the DNA
The auxiliary interactions are connected with the biological environment of DNA, i.e.
an aqueous solution with cationic counterions that ensure the overall electroneutrality of
the system. The importance of auxiliary interactions is attested by the fact that severe
dehydration of the DNA molecule such as severe osmotic stress on the order of a 1000
atm, leads invariably to a degradation of the double-helical structure. Less drastic de-
hydration leads to a conformational change from a B-DNA to an A-DNA form. On the
other hand the dilution of the salt in the aqueous medium, leading to weaker screening
of electrostatic interactions, leads to local opening of the DNA structure that eventually
engenders a complete melting of the molecule.
Interactions stabilizing the B-form of DNA are not infinitely strong and, since hydro-
gen bonding makes an important part of these interactions, can be broken, dissociating
the two strands of DNA at sufficiently high yet practically attainable temperatures. In
that case DNA melts and dissociates into two separate single strands in solution. In or-
der to describe this phenomenon one needs to assume some quantitative features of the
stabilizing interactions.
4.1 A model for primary stabilizing interactions

The hydrogen bonding acts only between complementary bases in a base-pair. The
canonical purine-pyrimidine Watson-Crick pairing is between A and T which form two
hydrogen bonds and between G and C which form three hydrogen bonds. Because of a
52 Rudolf Podgornik
Figure 25. Separation dependence of the hydrogen bond energy in a GC (left) and AT
(right) Watson-Crick base-pair, vide K. Brameld et al. op.cit., calculated by various
quantum mechanical methods. The distance between heteroatoms of the central hydrogen
bond is defined to be the one between N1 atom of the pyrimidines, and N3 atom of the
purines.
Monday, November 1, 2010
different number of hydrogen bonds the G-C base-pair is the more robust one 38 . The
covalent chemical bonding between the phosphates as well as atomic bonds in the nitro-
gen bases are of course the strongest interactions stabilizing the double-helical structure
of DNA. The covalent bond has a typical length of 0.154 nm and an energy of about 145
kB T at room temperature. The covalent bond is quite rigid at room temperature and one
has to input an energy of about 6 kB T to stretch the bond by about 0.01 nm. The hydro-
gen bond on the other hand has a length of about 0.275 nm and a typical strength of about
10-20 kB T at room temperature, depending also strongly on the nature of the hydrogen
bond. Contrary to the covalent bond the hydrogen bond is quite soft and at room temper-
ature one has to input an energy of about 0.16 kB T to stretch the bond by about 0.01 nm
39
. At room temperature the hydrogen bonds can deform, brake and reform any number
of times. They are thus ideal for holding together biological macromolecules that need
to be relatively robust but also pliable at the same time.
The detailed dependence of the hydrogen bond interaction energy on the separation
between the two bases in a Watson-Crick base-pair is very complicated, see Fig. 25. A
simple yet useful parametrization of this interaction potential is provided by the Morse
potential. Assuming that the separation between the bases in athe n th Watson-Crick
base-pair along the DNA chain is yn , the form of the Morse potential is given by
ayn 2
V (yn ) = D e 1 . (103)
The Morse potential depends on only two phenomenological parameters: the dissociation
energy of the bond, D, and an inverse length a which sets the length scale of the bond.
In principle both parameters depend on the nature of the base-pair and thus on base-pair
position n. This form can be fitted to more complicated y dependence such as presented
on Fig. 25. The Morse form of the interaction potential in spite of all the simplifications
38 K. Brameld, S. Dasgupta, and W. A. Goddard III, Distance Dependent Hydrogen Bond Potentials for
Nucleic Acid Base Pairs from ab Initio Quantum Mechanical Calculations (LMP2/cc-pVTZ), J. Phys. Chem.
B 101 4851-4859 (1997).
39 M. Peyrard, Nonlinear dynamics and statistical physics of DNA, Nonlinearity 17 (2004) R1-R40.
53
has nevertheless an appropriate qualitative shape, viz. it has a strong steric repulsive
branch for negative y, corresponding to the sterical hindrance mentioned earlier, it has
a well developed minimum at an appropriately defined equilibrium position y = 0, and
leads to a vanishing force between the bases when they are far apart, allowing for a
complete dissociation of the base pair.
The base-pair stacking interactions between the disks formed by the base-pairs in a
stack along the long axis of the DNA molecule have a complex origin and can not be
subsummed under a single heading. Partly they stem from the overlap of the ⇡ electrons
of the nucleotide rings of the nitrogen rich bases, partly from the non-local van der Waals
interactions between the rings and partly from the hydrophobic interactions mediated by
the water molecules of the bathing aqueous solution. What one can calculate these days
relatively accurately are the first two types of the interactions 40 while the hydrophobic
contribution to the total interaction can be estimated only very roughly. While the hy-
drogen bonding within a base-pair depends almost solely on the nature of the purine and
pyrimidine in that base-pair, the stacking interaction depends on the local base sequence
41
.
In experiments of Ke et al. 42 it was possible to measure stacking interactions di-
rectly for a ssDNA attached to a gold surface and pulled with an AFM tip. The relative
extension per base-pair of single-stranded poly(dA) obtained in this way show that it
displays two force plateaus, at 23 pN and 113 pN, that appear to be due to breaking the
stacking interactions between consecutive adenines. The total energy, which is necessary
to overstretch poly(dA) and thus break the stacking interactions, is inferred to be -6 kB T
per base. It is quite extraordinary that this value that was deduced from direct ssDNA
stretching experiments is consistent with the reported value obtained by looping kinetics
of ssDNA which gives an estimate of a stacking energy for poly(dA) of -6.5 kB T per
base 43 .
The detailed dependence of the stacking interaction energy on the separation between
two neighboring Watson-Crick base-pairs is also very complicated. Apart from the sep-
aration between the two base-pairs it also depends on the separation between the bases
within a base-pair. Nevertheless, on a simplest level we will assume that the stacking
potential close to its minimum can be approximated in its simplest form by a nearest
neighbor 1D interaction of the form W (yn , yn+1 ). The functional dependence is chose
to be of a simple harmonic form
1 2
W (yn , yn+1 ) = K (yn+1 yn ) . (104)
2
The harmonic form of the stacking potential describes several salient features of this in-
teraction evident from the detailed quantum calculations, see Fig. 26. In order to take into
account the sequence specificity of the stacking energy as evident from Fig. 26 the stef-
ness of the stacking potential K should also depend on the base-pair position n. It takes
40 M. Elstner et al., Hydrogen bonding and stacking interactions of nucleic acid base pairs: A density-
functional-theory based treatment, J. Chem. Phys.114 5149-5155 (2001).

41 V.R. Cooper et al., Stacking Interactions and the Twist of DNA, JACS 130 1304-1308 (2008).
42 Changhong Ke, M. Humeniuk, Hanna S-Gracz, and P. E. Marszalek, Direct Measurements of Base Stack-
ing Interactions in DNA by Single-Molecule Atomic-Force Spectroscopy, Phys. Rev. Lett. 99 018302 (2007).
43 A. Sain, Bae-Yeun Ha, Heng-Kwong Tsao, and J. Z. Y. Chen, Chain persistency in single-stranded DNA,
Phys. Rev. E 69, 061913 (2004).

54 Rudolf Podgornik
Figure 26. Left: stacking energy between two guanine disks as a function of separation.
The calculation is performed with two different methods, adapted from M. Elstner et
al. op.cit. Right: stacking energy between base-pairs at a twist angle corresponding to
the B-DNA geometry, ı.e. 36 , adapted from V.R. Cooper et al. op.cit. Only the base
sequence on one of the strands is shown.
into
Monday,account the overlap of the ⇡ electron orbitals of the nitrogen rings of the bases as
November 1, 2010
well as the presence of the rather rigid sugar-phosphate covalent link between the neigh-
boring bases that couples the neighbors. There are several possible deformations, most
notably base rotations leading to DNA twisting, along the chain that can not be properly
represented in the harmonic parametrization involving only nearest neighbors. The po-
tential W (yn , yn+1 ) can thus be viewed upon as an effective potential, pre-averaged over
various different displacements of the bases.
The parameters of the phenomenological forms for the two primary interaction poten-
tials are not known to any accuracy but can be calibrated by comparison with experiments
such as the thermal denaturatio 44 . The strength of the hydrogen bonding potential can
thus be estimated as D = 1.2 kB T , and a = 45nm 1 . It follows from these two numer-
ical values that stretching of a base pair by 0.01nm, changes the energy by 0.24kB T , a
value consistent with theproperties of hydrogen bonds. The harmonic spring stiffness of
the stacking interaction can be estimated as K = 240kB T nm 2 , which corresponds to
a rather weak coupling between the bases. T
4.2 Peyrard-Bishop-Dauxois model of DNA melting

The two approximate forms for the primary stabilizing interactions of the B-DNA double
helix can be used in order to parameterize a model of DNA melting 45 . This implies a
very simplified model of stabilizing interactions where the DNA double helix can be
represented in the straightened form with hydrogen bonds acting between the strands
and stacking interactions acting along the strands, see Fig. 27. This simplified view of
the DNA DNA we refer to as the Peyrard-Bishop-Dauxois model of DNA.
44 M.Peyrard, op.cit.
45 M. Peyrard and A.R. Bishop, Statistical Mechanis of a non-linear Model for DNA Denaturation, Phys.
Rev. Letts. 62 2755-2758 (1989), T. Dauxois, M. Peyrard and A.R. Bishop, Entropy driven DNA denaturation,
47 R44-R47 (1993).
55
Figure 27. A schematic representation of the Peyrard-Bishop-Dauxois model of DNA.

DNA double helix is represented in the straightened form with hydrogen bonds (white
lines) acting between the strands and stacking interactions (not shown explicitly) acting
along the strands. Graphics courtesy of A. Siber.
Assuming that there are N base-pairs along the DNA chain, the configurational
Hamiltonian is assumed to be of the form
NX1 X1 ✓ 1
N
2 2
◆
H[yn ] = (W (yn , yn+1 ) + V (yn )) = K (yn+1 yn ) + D e ayn 1 .
n=1 n=1
2
(105)
The configurational integral corresponding to the above Hamiltonian can then be written
in the form
Z Y Z Y "N 1 #
Y
H[yn ] f (yn+1 ,yn )
Z( ) = dyn e (yN y0 ) = dyn e (yN y0 ),
n n n=1
(106)
where
1 2 ayn 2
f (yn+1 , yn ) = W (yn , yn+1 ) + V (yn ) = K (yn+1 yn ) + D e 1 . (107)
2
The delta function takes care of the periodic boundary condition. The configurational
integral can thus be written explicitly in the form that couples only the first neighbors
along the chain. Since the problem is one-dimensional the transfer integral method can
be used to evaluate the configurational integral exactly at least in principle. Introducing
first the transfer operator T (y, x) as
Z
+1 Z
+1
[ 12 K(y x)2 +V (y)] ✏i
dx T (y, x) i (x) = dx e i (x) =e i (y). (108)
1 1
The solution of this equation provides the eigenenergy as well as the eigenfunction of the
problem. If not by other means, it can be solved numerically. Since the basis functions of
56 Rudolf Podgornik
P
the transfer operator are orthonormal one can expand (yN y0 ) = i ⇤i (yN ) i (y0 ).
Thus one obtains the partition function in the form
XZ Z Z
+1 +1 +1
f (y1 ,y0 ) f (y2 ,y1 ) f (yN ,yN 1)
Z= dy0 e i (y0 ) dy1 e ... dyN 1e ⇤i (yN ).
i 1 1 1
(109)
Evaluating the integrals starting from the l.h.s. of the above equation and taking into
account Eq. 108 at each step one can move through the whole sequence of integrations
finally remainig with the following form of the partition function
X Z X
Z( ) = e N ✏i dyN | i (yN )|2 = e N ✏i ! e N ✏0 . (110)
N !1
i i
where for an infinitely long chain one assumes the ground state dominance and thus ✏0
is the smallest eigenvalue of the transfer operator. The corresponding free energy per
monomer of the chain is then
kB T
f( ) = log Z( ) = ✏0 . (111)
N
Assuming that the harmonic spring stiffness K is large one can Taylor expand the integral
in Eq. 108 around the minimum of the harmonic term, i.e. x = y + z, and furthermore
assuming that z is small this leads to
Z
+1 ✓ ◆
1
Kz 2 1 0
dz e 2 (y) + z (y) + z 2 ”(y) + . . . =e ✏0
(y). (112)
2
1
One can now perform the Gaussian integral in the variable z and expand the exponen-
tial of the Morse potential since it is bounded for positive values of y, corresponding
to complete dissociation of the chains. introducing the displaced eigenenergy ✏0 =
✏0 + 21 log 2⇡/( K) one then derives for the eigenfunctions the following differential
equation
1
”(y) + 2 V (y) (y) = 2 ✏0 (y), (113)
2K
which is just the standard Schrödinger equation corresponding to the Morse potential.
The spectrum of this Schrödinger equation contains one or more localized states if 2 ✏0 <
D and a continuous spectrum in the opposite limit. However, since the potential is not
bounded on both sides of y, the localized ground state exists only when the potential is
deep enough 46 . A detailed analysis leads to the following scenario involving a melting
transition 47 of the whole chain
46 M. Peyrard, op.cit.
47 The existence of a melting transition in this 1D model is not contrary to the general Landau-Peierls argu-
ment that there should be no phase transitions in 1D systems. This is due to the fact that the interactions contain
a term viz. V (yn ) that stems from the interaction between the two strands of DNA. This term plays the same
role as an external field in a magnetic system. Furthermore the energy of a domain wall between a melted and
a compact region of DNA is infinite for an infinitely long chain. This is again contrary to the Landau-Peierls
argument where the domain wall energy is assumed to be constant. More on this, see below.
57
• for low temperatures, the depth of the potential proportional to 2 D, is large and
there exists at least one localized state. Consequently the average base-pair sepa-
ration Z +1
hyi = dy| 0 (y)|2
1
is finite and the molecule is not denatured. This expression can be derived by the
same leapfrog integrations as in Eq. 109.
• for sufficiently high temperatures, the depth of the potential decreases and there
exists a critical temperature Tc where the localized state merges into the contin-
uum. The ground state now has a non-localized eigenfunction, consequently the
average base-pair separation diverges and the molecule is denatured.
Luckily this scenario does not depend crucially on the details of the interaction potential
V (y), as long as it remains qualitatively similar to the Morse potential. We thus need not
worry about the exact form of the interaction potential but only on its salient features that
can be parameterized by various analytical forms. What we failed to do is to evaluate
the transition tenperature exactly as a function of the interaction potential parameters.
Though this is possible from the analysis of the solutions of the Morse equation, we will
proceed via a more intuitive argument involving the stability of a domain wall.
4.3 The DNA melting temperature

Based on the general analysis of the Peyrard-Bishop-Dauxois model in the previous sec-
tion we have demonstrated that there must exist a temperature-driven transition in the
behavior of DNA that corresponds to a denaturation or melting of the chain. Let us now
calculate the temperature at which this happens 48 .
We first of all write down the complete Hamiltonian of the chain, containing the
kinetic part with the momentum per base-pair deonted by pn
X1 ✓
N
1 2 1 2 2
◆
ayn
H[pn , yn ] = p + K (yn+1 yn ) + D e 1 . (114)
n=1
2m n 2
where we assumed that all the base-pair masses are the same, m. The Hamiltonian
equations of motion have an interesting static solution for pn = 0. Assuming that the
index n can be treated as a continuous variable, the variation w.r.t. yn one obtains
@H[pn = 0, yn ] d 2 yn d ayn 2
=0 wherefrom K D e 1 . (115)
@yn dn2 dyn
Evaluating the first integral of this equation and assuming that for n ! 1 all base-
pairs are closed, leads to an explicit domain wall (yDW ) solution of the form
✓ p 2 ◆ (
2Da
(n n ) 0 n ⌧ n0
ayDW (n) = ayn = log 1 + e K 0
, or yDW (n) ⇠
1 n n0
(116)
48 M. Peyrard, Nonlinear dynamics and statistical physics of DNA, Nonlinearity 17 (2004) R1-R40.
58 Rudolf Podgornik
where the integration constant n0 centers the solution. The r.h.s. of the above equation
represents a simple limiting behavior that clearly shows the main features of the domain
wall solution. For sites on the chain with n ⌧ n0 we have yn ⇠ 0. The Morse potential
V (yn ) and the coupling energy W (yn+1 , yn ) between adjacent sites vanish. For sites
with n n0 , yn ! 1 and the Morse potential takes the value V (yn ) = D. Calculating
the coupling energy in the same limit with a continuous index n, we derive that indeed
in this limit we also have W (yn+1 , yn ) = D. The solution thus obviously represents
a domain wall located at n0 separating open base-pair configurations from closed ones.
The total energy of an N base-pairs long chain corresponding to the domain wall solution
centered at n0 is thus 2D(N n0 ).
We now decompose the total free energy of the chain F with a single domain wall
into a mean-field contribution FM F , which we just calculated and that by construction
does not take into account thermal fluctuations, and the fluctuation part FF L , so that
F = 2D(N n0 ) + FF L . (117)
For the fluctuational part we first need to analyze the solutions of the equation of motion
that can be represented as
yn (t) = yDW (n) + f (n, t), pn (t) = mf˙(n, t) (118)
where time independent part yDW (n) is given by Eq. 116 and the fluctuations around the
domain wall solution, f (n, t), are assumed to be small. Expanding the Hamiltonian Eq.
114 with respect to f (n, t) up to the second order so that one has H[f˙(n, t), f (n, t)] =
H[pn = 0, yDW (n)] + H2 [f˙(n, t), f (n, t)] with the harmonic deviation from the domain
wall value of the total energy in the form
N ✓ ◆2 !
X1 m ˙ 2 K df (n, t) 1
H2 [f˙(n, t), f (n, t)] = f (n, t) + 2
+ V ”(yDW (n))f (n, t) + . . . ,(119)
n=1
2 2 dn 2
where the first order is zero in view of Eq. 115. The equation of motion is then
✓ 2 ◆
d f (n, t)
mf¨(n, t) K + V ”(yDW (n))f (n, t) = 0. (120)
dn2
Since the domain wall solution has the behavior described by Eq. 116 the corresponding
stiffness of the hydrogen bond potential is obtained as
(
2Da2 n ⌧ n0
V ”(yDW (n)) ⇠ . (121)
0 n n0
Let us assume that there is a sharp transition between the two regimes right at the domain
wall boundary. The calculation will be simpler with this proviso though no essential
physics will be lost. The solution of the equation of motion can be written in terms of
propagating modes f (n, t) ⇠ ei(qn !t) . The dispersion relation that follows from Eq.
120 can be obtained in the two limiting cases of n as
(
2Da2
2 !< 2 (q) = K 2
mq + m ; n ⌧ n0
! (q) = 2 K 2
(122)
!> (q) = m q ;n n0
59
and gives us the possible phonon modes left and right of the domain wall. The fluctuation
part of the partition function, ZF L , is then obtainedQfrom the partition function of har-
monic oscillators each at wave-vector q as ZF L = q Z(q) 49 . From here one obtains
for the fluctuational contribution to the free energy
X X
FF L = kB T log ZF L = kB T log Z(q) = kB T log ( h !(q)). (123)
q q
The sum over the q space can now be transformed into an integral for a closely spaced
lattice
P as is standardly
R done in the solid state physics. This transformation can be written
as q ! 2⇡ V
dq, where V is the "volume" corresponding to the harmonic modes.
The way we have set up the calculation this "volume" is n0 for the !< (q) modes and
N n0 for the !> (q) modes. The fluctuational free energy this assumes the form
Z1 Z1
k B T n0 kB T (N n0 )
FF L = dq log ( h !< (q)) + dq log ( h !> (q)). (124)
⇡ ⇡
0 0
The part of the fluctuational free energy that does not depend on n0 is irrelevant for the
consideration of the domain wall effects and thus we remain with
Z1 r
k B T n0 !< 2 (q) kB T 2Da2
FF L = dq log = n 0 . (125)
2⇡ !> 2 (q) 2 K
0
The total free energy of the domain wall, Eq. 117, including the mean field and the
fluctuations can thus be obtained in the form
r !
kB T 2Da2
F = 2D(N n0 ) + FF L . = n0 2D + FN . (126)
2 K
Obviously for small temperatures, when the term in the brackets is negative, the system
can lower its free energy by increasing n0 so that no base-pairs are dissociated. Contrary,
for large temperatures the term in the brackets is positive, and the system is in a dissoci-
ated state reached via a decrease in n0 . The melting temperature Tm is thus defined as
the temperature at which the properties of the system do not depend on the position of
the domain wall
r p
kB Tm 2Da2 2 2KD
2D = 0 or else Tm = . (127)
2 K akB
The melting temperature derived through the small fluctuations around the domain wall
solution turn out to be the exact value that can be obtained also via the transfer operator
method by a much more involved procedure. The melting temperature depends on the
square root of the product of the depth of the hydrogen bond potential minimum and the
stiffness of the stacking potential and the length scale of the hydrogen bond potential.
The stronger the hydrogen bond potential or the steeper the stacking potential, the higher
is the melting temperature.
49 The quantum partition function for each q mode of a harmonic oscillator is given by the standard expression
Z(q) = (1 exp h !(q)) 1 , obtained by summing the Boltzmann factor over all the phonon occupancies.
Expanding this in the classical limit of high temperatures yields lim !0 Z(q) ⇡ ( h !(q)) 1 .
60 Rudolf Podgornik
Figure 28. The melting profile of various sequences of DNA. Poly(AT) has the lowest
melting temperature and Poly(GC) the highest. The melting profile is defined as =
1 ✓, where ✓ is the average fraction of bonded base pairs. Adapted from R. Horton,
L. A. Moran, G. Scrimgeour, M. Perry, D. Rawn, Principles of Biochemistry, Pearson
Prentice and Hall (2006), Fig. 19-17.
4.4 Observing DNA melting

In an experiment one would typically monitor the changes in the UV absorbance as a
function of temperature. The DNA bases absorb UV radiation because the nitrogen rich
rings contain alternating single and double bonds which are known to absorb in the UV
region of the spectrum if the system is small enough. In the opposite case, the absorption
peak is displaced towards the visual part of the spectrum. Each of the four nucleotide
bases has of course a somewhat different absorption spectrum 50 , and the absorption
spectrum of DNA is obtained as an appropriate compositional average. In an aqueous
DNA solution a measurable absorption kicks in at about 320 nm leading to a peak at
approximately 260 nm and then a dip between 220 and 230, raising again to a very
pronounced peak just below 200 nm and then fading off into the far UV region 51 . Of the
four nucleotides adenine has a most pronounced peak at 260 nm, followed by guanine,
thymine and cytosine. The energy absorbed by photochemical reactions in the UV region
by the DNA bases causes changes in the carbon double bonds on the bases that lead to
formation of pyrimidine dimers. Unrepaired dimers essentially destroy DNA by causing
mutations that interfere with regulatory mechanisms involved in cell division 52 .
The melting temperature can be defined as the temperature at half maximal DNA UV
absorption. DNA melting depends strongly on the base-pair composition of DNA. DNA
that consists mostly of A-T base-pairs melts at about 70 and DNA that is predominantly
composed of G-C base-pairs melts at above 100 , see Fig. 28. Thus by knowing the
50 A. Pinchuk, Optical constants and dielectric function of DNA’s nucleotides in UV range, Journal of Quan-
titative Spectroscopy & Radiative Transfer 85 211-215 (2004).

51 By definition the absorbance of native DNA at 260 nm in a 1-cm quartz cuvette of a 50Âţg/ml solution of
double stranded DNA is equal to 1.

52 The UV light can thus be used for sterilization but can also cause skin cancer.
61
exact base composition of DNA one can calculate the melting temperature. The major
cause for variation of DNA melting temperature with composition is the stacking energy
for different base-pairs. On Fig. 26 one can discern how the stacking energy varies
as a function of base-pair step 53 , with the C-G sequence having the lowest - strongest
stacking - energy. The base-pair composition of DNA could be included in a generalized
theoretical model by invoking the n dependence of K = K(n) and D = D(n) but in
that case the analytical feature of the Peyrard-Bishop model would not be available and
the model would have to be solved numerically from the start 54 .
Comparison with experiments also shows that an anharmonic form of the stacking
potential is needed in order to describe properly the sharpness of the melting transition.
The standard anharmonic generalization of Eq. 105 is given by
X1 ✓
N
1 ⇣ ⌘
2 2
◆
b(yn yn 1) ayn
H[yn ] = K 1 + ⇢e (yn+1 yn ) + D e 1 . (128)
n=1
2
In order to take into account the sequence specificity of the chain Hamiltonian, all the
elastic constants can exhibit some dependence on the base-pair sequence. The most
important heterogeneity of the DNA base-pair sequence are the different values of the
Morse potential for adenine-thymine and guanine-cytosine base pairs. In order to take
into account the heterogeneous sequence of DNA one needs to introduce two different
values of D for two different Watson-Crick base pairs: the AT base pair has two hydrogen
bonds and the GC base pair has three.
K ⇢ b D a
0.025 eV/Å 2
2 b = 0.35Å 1
DAT = 0.05 eV aAT = 4.2Å 1
1
DGC = 0.075 eV aGC = 6.9Å
Table 3. A table of the Campa-Giansanti values for the Peyrard-Bishop-Dauxois model

parameters. They give very good agreement between the observed and calculated melting
profiles.
The most accurate parametrization of the PBD model 55 , see Table 4.4, adjusts the
parameter values in order to reproduce the experimentally observed melting temperatures
of long homogeneous DNAs at the standard solvent conditions and chooses the depth for
the GC Morse potential as 1.5 times that of the AT Morse potential. The depth of the
Morse potential is thus the major difference between the AT and GC base pairs. One can
write for the sequence dependent pairing strength of the Morse potential
D(n) = D̃(1 + p⌘(n)), (129)

53 Only the base-pair sequence on one of the strands is shown.
54 M. Peyrard, S. Cuesta-Lopez and D. Angelov, Experimental and theoretical studies of sequence effects on
the fluctuation and melting of short DNA molecules, J. Phys.: Condens. Matter 21 034103 (2009).
55 The values of the parameters that most closely fit the experimental data were obtained in A. Campa and A.
Giansanti, Experimental tests of the Peyrard-Bishop model applied to the melting of very short DNA chains,
PHYSICAL REVIEW E 58 3585 (1998).
62 Rudolf Podgornik
where D̃ = 0.0625 eV and p = 0.2 while the field ⌘(n) describes the DNA sequences
being +1 for GC base pair and -1 for AT base pair 56 .
DNA melting is used standardly in the polymerase chain reaction (PCR) used in
molecular biology to amplify thousand or million-fold a single or several copies of
a piece of DNA. Usually thermal cycling is used as a part of PCR by sequentially
heating and cooling the DNA sample to a defined series of temperature steps and then
enzymatically replicating the DNA. Standardly the application of PCR is based on a
heat-stable DNA polymerase, such as T aq polymerase, isolated from the bacterium
T hermusaquaticus. This DNA polymerase enzymatically assembles a new DNA strand
from dissociated single strands of DNA, by using single-stranded DNA as a template and
DNA oligonucleotides a.k.a DNA primers in order to initiate the DNA synthesis. As
PCR steps are repeated the DNA generated is itself used as a template for replication in
the following step, setting up a chain reaction in which the DNA template is amplified
in a geometric progression. Thermal cycling is necessary in order to physically sepa-
rate the two strands of a DNA double helix at above melting temperature. After a lower
temperature quench each strand is then used as a DNA polymerase template in order to
synthesize the target DNA. The selectivity of PCR results from the use of primers that
are complementary to the DNA region targeted for amplification under specific thermal
cycling conditions.
4.5 DNA denaturation bubbles

Inspite of the fact that DNA denaturation is a temperature driven collective transition
in the conformation of the chain, partly melted sequences are observed at any non-zero
temperature even well below the denaturation transition. These transient local openings
of the DNA chain, see Fig. 29 are usually referred to as denaturation bubbles. Bubbles
are observed as dynamic fluctuations of the chain e.g. by dielectric spectroscopy 57 as
they change the effective extension of the chain or are hypothesized to exist in the case
of tight bending of DNA, locally relaxing the curvature stresses 58 . The opening and
subsequent closing of DNA bubbles is often referred to as DNA breathing.
Let us calculate the distribution of the size of these bubbles in DNA below the melting
transition. In order to do that we need the free energy of the bubble formation F(n) that
has n open base pairs under the assumption that the DNA chain has a homogeneous
sequence, thus ignoring any sequence dependence of the interaction potential constants.
We will calculate this free energy by decomposing into an energy part due to base pair
opening and the entropy of a single DNA strand loop n segments long. Energy first.
Let us first define the model that will serve to obtain an estimate for this energy. Sung
and Jeon 59 introduced a model very similar to Dauxois-Peyrard-Bishop model differing
from it only in the interpretation of various constants. The Sung-Jeon Hamiltonian then
56 H.-H. Jeon, P.J. Park and W. Sung, The effect of sequence correlation on bubble statistics in double-
stranded DNA, J. Chem. Phys. 125 164901 (2006).

57 S. Tomic, S. Dolanski Babic, T. Ivek, T. Vuletic, S. Krca, F. Livolant and R. Podgornik, Short-fragment
Na-DNA dilute aqueous solutions: Fundamental length scales and screening, EPL, 81 68003 (2008).
58 R. Podgornik, DNA off the Hooke, Nature Nanotechnology 1 100 (2006).
59 W. Sung and J.-H. Jeon, Conformation and local denaturation in double-stranded DNA, PRE 69 031902
(2004).
63
Figure 29. A schematic depiction of a denaturation bubble in a model DNA chain. Partly
opened, breathing DNA helix, with a stretch of dissociated base pairs. Graphics courtesy
of A. Siber.
reads
X1 ✓
N
1 2 2
◆
ayn
H[yn ] = K (yn+1 yn ) + D e 1 , (130)
n=1
2
where now y(n) is the separation between the two strands and the stiffness K can be
interpreted as arising not so much from stacking interactions but from the flexibility of
the DNA strands. In that case it would be equal to K = 32 kB T
`2 , where ` is the segment
length between two base pairs along the axis of DNA, i.e. 0.34 nm from the B-DNA
configuration. The minimum of the interaction potential is obviously D. We want to
calculate the energy penalty for a local denaturation of the chain.
The free energy corresponding to the above Hamiltonian can be evaluated exactly by
analyzing the solutions of Eq. 113. The solution can correspond either to a bound state
or a continuum of unbound states, depending on the sign of the eigen-energy ✏ obtained
by solving
1
”(y) + D(e ay 1)2 (y) = ✏ (y). (131)
2 2K
This is in fact a Morse equation of quantum mechanics whose eigen-function solutions
are known explicitly 60 and have the form
z(y)/2
p
n (y) =e z(y)s L2s
n (z(y)) with z(y) = 2t e ay
and2DK, t=
a
(132)
where L2sn (z) is an associated Laguerre polynomial. The above solution remains finite
for positive values of the argument only if s is positive. From this solution the eigen-
60 T. Dauxois, N. Theodorakopoulos and M. Peyrard, Thermodynamic Instabilities in One Dimension: Cor-
relations, Scaling and Solitons, Journal of Statistical Physics, 107 869 (2002).
64 Rudolf Podgornik
energies are then obtained in the form

✓ ◆2
n + 12
✏n = D D 1 with n = 0, 1, . . . Int(t 1
2 ), (133)
t
where Int(b) is the integer part of the argument. Thus as long as t < 12 the solution indi-
cated a bound state of negative eigen-energy, the limiting value being tc = 12 wherefrom,
by taking into account the definition of t it follows
p
2 2DK
Tc = with ✏0 = D, (134)
kB a
where ✏0 is the lowest lying eigen-energy. Tc is of course nothing but the melting temper-
ature, Eq. 127, calculated by a different path before. The gap between the ground state
and the positive energy unbound states measured from the minimum of the interaction
potential of magnitude D, is then
✓ ◆
T
(T ) = (✏0 (T ) D) = D 1 , (135)
Tc
and zero gap of course defines the melting temperature. (T ) is thus the energty penalty
of dissociating a base pair. The larger it is the more DNA with a homogeneous sequence
is stable against the loop formation. Its inverse 1
is then the correlation size for the
base pair dissociation, i.e. the number of base pairs over which a disturbance is felt,
and amounts to about 5 base pairs. We can furthermore write the free energy of an n
segments long loop as
F(n) = (T ) n kB T log ⌦(n) + const., (136)
where ⌦(n) is the number of conformation of the chain in a closed polymer ring con-
stituting a loop, while the constant at the end describes the free energy barrier for initial
bubble formation. For a sufficiently long loop one can approximate ⌦(n) by its ideal ran-
dom walk form ⌦(n) ⇠ n 3/2 , which equals the statistical weight of all configurations
of an ideal chain that start and end at the same position. A more detailed analysis in fact
yields a general dependence ⌦(n) ⇠ n ↵ , where ↵ can be different form the ideal ran-
dom walk value. Introducing now the bubble length distribution function as the average
number of bubbles of length n, N (n), on a DNA sequence of total length N ! 1, we
end up with
( (T )n
e
hN (n)i F (n) T  Tm
(137)
n 3/2
P(n) = lim ⇠e = 1
.
N !1 N 3/2 T ! Tm
n
One can show that the non-harmonic nature of the base-pair interaction is directly re-
sponsible for this distribution function. Because the correlation size 1
depends on D
there is obviously an effect of the sequence specificity. The DNA sequences of higher
GC content, with stronger interaction potential as codified by D, thus exhibit a faster de-
cay of bubble sizes. From the bubble length distribution one can deduce also the average
bubble length LB , which is given by the total number of base pairs in bubbles divided by
the total number of bubbles P
nP (n)
LB = P n , (138)
n 1 P (n)
65
which shows a strong exponential dependence on n 61 . For any finite temperature be-
low the melting temperature the bubble length distribution function decays exponentially
with a correlation number of base pair on the order of 5-10. When the temperature ap-
proaches the denaturation temperature the bubble size starts growing algebraically and
the chain eventually melts completely. This result can be also obtained in the framework
of the Poland-Sheraga model 62 of the DNA melting which in some respects represents
a simplified version of the Dauxois-Peyrard-Bishop model. Numerical simulations cor-
roborate this distribution function quite well 63 .
61 S. Ares and G. Kalosakas, Distribution of bubble lengths in DNA, Nano Letts bf 7 307 (2007).
62 D. Poland D, H.A. Scheraga, Phase transitions in one dimension and the helix-coil transition in polyamino
acids., J Chem Phys. 45 1456 (1966).
63 S. Ares and G. Kalosakas, ibidem.
66 Rudolf Podgornik
5 ELASTICITY OF DNA
In aqueous solutions at biologically relevant conditions DNA often behaves as an elas-
tic rod whose deformation energy can be sufficiently well approximated by the classical
continuum theory of deformable elastic filaments 64 . In fact, packing of DNA, such as
packing in viral capsids or packing within the chromatin fiber, usually involves large
elastic deformation energies. On the other hand various forms of DNA, such as plec-
tonemes, circular plasmids or suporcoils, are connected with a redistribution of different
types of deformation energy leading to a conversion of the twist deformation into the
writhe deformation.
Trough any deformation of the DNA double helix involves local molecular displace-
ment and should thus be per force described at an atomic level, quite often the change
of shape or form of DNA can be described surprisingly well with a macroscopic theory
that averages over all the atomic degrees of freedom. The continuum description and
the corresponding bending and torsion deformation (free) energies 65 have their origin
in the same nanoscale inter-atomic forces that give rise to the double-helical form of the
molecule. This signifies that the deformation energy of the DNA molecule is governed
by the
• base stacking energy that is sensitive to the amount of overlap between neighbor-
ing basepairs whic in turn depends on local deformation of DNA
• water-DNA surface interactions between the polar moieties of the DNA surface,
especially the major and minor grooves
• electrostatic interactions along the charged phosphate groups along the DNA that
depend on the shape of the molecule
The overlap between the types of interactions involved in DNA elastic deformation and
DNA melting is obvious and therefore it is no wander that the two are closely related.
Through modeling of DNA deformation and shape variation on the atomic or base-
pair level is feasible, and is in fact vigorously pursued 66 , we will delve only in the
continuum theory of DNA elasticity that relies on average nano-scale DNA properties,
being independent of the very complicated description at the atomic level. The macro or
meso-scopic continuum elasticity of DNA depends on several phenomenological param-
eters such as the various types of persistence lengths and spontaneous curvatures, which
are usually determined from suitable experiments. From the value of these mesoscopic
parameters DNA would be a relatively stiff molecules, positioned somewhere between
the elasticity of the very flexible hylauronic acid 67 and that of the extremely stiff micro-
tubules 68 .
64 A.E.Love, Treatise on the Mathematical Theory of Elasticity, Dover Books (1927).
65 Usually one describes the DNA molecule at thermal equilibrium where the thermodynamic potential is its
free energy.
66 A.A. Travers, J.M. Thompson, An introduction to the mechanics of DNA, Philos Transact A Math Phys
Eng Sci. 362 1265 (2004).

67 Persistence length on the order of ten nm.
68 Persistence length on the order of several mm.
67
The continuum description of the deformable elastic filament assumes that its arbi-
trary deformation can be decomposed into bending and torsional modes. Their energy
depends on the shape of the molecule that has to be described within an appropriate frame
of reference. Local changes in this frame of reference describe the deformation.
5.1 Elastic deformation energy

In a deformed elastic filament the Darboux vector is a continuous function of the contour
length along the filament
⌦(`) = ⌧ d3 (`) + 2 d2 (`) + 1 d1 (`).
In an undeformed filament it is by definition a constant. The elastic energy of this defor-

mation can only depend on the derivative of the Euler angles, thus on the Darboux vector,
but not on the values themselves. The deformation energy per unit length or the elastic
energy density of the filament f , will thus depend on ⌦ and possibly higher derivatives
of the Euler angles, thus f = f (⌦).
The direction of the tangent vector to the filament at any point along its length can be
chose either as d3 or as d3 . Therefore f (⌦) should be invariant under a transformation
d3 ! d3 , which means that ⌦3 should enter at most quadratically into the deformation
energy, thus
1
f = f0 + C⌦23 + g(⌦2 , ⌦1 ). (139)
2
Here we assumed that the stress-free state has no twist density, ⌦03 = 0. If this is not the
case, than the above formula should read
1
f = f0 + C(⌦3 ⌦03 )2 + g(⌦2 , ⌦1 ).
2
What about the dependence of the elastic energy on ⌦2 in ⌦1 ? It shoudl not show any
linear dependence since this would lead to a deformed state of the filament even in the
absence of any external stresses. In general the dependence of the elastic deformation
energy on ⌦2 in ⌦1 can be therefore written to the lowest order as
X 2
1 1
f = f0 + C⌦23 + E Iij ⌦i ⌦j , (140)
2 2 i,j=1
where Iij shoudl be taken as the components of the moment tensor and E is the Young
modulus of the material of which the filament is composed. Let us now write the elastic
energy density in a local coordinate system that coincides with the principal directions
of the tensor Iij , with values I1 and I2 . In this case we remain with f = f0 + 12 C⌦23 +
2 EI2 ⌦2 + 2 EI1 ⌦1 . We need to remember the connection between ⌦i,2 and the two
1 2 1 2
principal curvatures 1,2 , Eq. 24. For a filament with a symmetric corss section (I1 =
I2 = I) the total deformation elastic energy can be therefore written as
Z L Z L
1 1
F = F0 + C ⌧ 2 d` + Kc 22 + 21 d`, (141)
2 0 2 0
68 Rudolf Podgornik
where we used Kc = EI. A standard value for the bend elastic modulus would be
⇠ 46 50 kB T nm, while the torsional elastic modelus would be ⇠ 86 kB T nm, with
a larger scatter in value 69 . This is the final form of the elastic deformation energy of
a deformable filament with symmetric cross-section. It can be written in various forms
depending on the context as we shall see in what follows.
5.2 Elastic equation of state

What we need to know now is how the elastic energy gives rise to forces and torques
within the elastic material, in other words we need to know and understand the elastic
equation of state of the medium. Let us assume first that the rod is acted upon by ex-
ternal torques. The work done by the change in external torques dM on a rod in a state
described via the local angles is given by
dW = dM · (`). (142)
We assume furthermore that the torque is distributed linearly with the linear density
dM
.µ=
d`
Instead of En. 140 we thus obtain for the total energy of the rod the expression
Z L✓ ◆ Z L
1 1
F = C⌦23 + Kc (⌦22 + ⌦21 ) d` µ · d`. (143)
0 2 2 0
The elastic equilibrium ios obtain from the minimum o fthe above free energy functional
with respect to small variations of the shape of the filament. Now since the "velocity" of
the deformation is defined as ⌦ = ˙ , the free energy is a functional of the form in ˙ :
F [ , ˙ ], or equivalently
Z L
F = f ( , ˙ )d`. (144)
0
This form of the energy functional reminds us of the Lagrange formalism of analytical
mechanics where the extremum of the action (the Hamilton principle) is given by the
corresponding Euler-Lagrange equations, except that here the parameter is not time but
arclength 70 The role of tiem in Hamiltonian formalism is thus played by the arclength `
in the theory of elastic filaments. Thus by analogy with classical mechanics the minimum
of the elastic energy functional Eq. 143 is given by
✓ ◆ ✓ ◆
d @f @f d @f @f
= = 0. (145)
d` @ ˙ @ d` @⌦ @
Let us now recall the form of the elastic energy Eq. 143 and evaluate its derivative with
respect to the components of vector ⌦, yielding
@f
= (C⌦3 , Kc ⌦2 , Kc ⌦1 ) = C⌧ t + Kc 2 d2 + Kc 1 d1 , (146)
@⌦
69 T.
70 In
R proteins, Rep. Prog. Phys. 66 (2003) 1
R. Strick et al. Stretching of macromolecules and
classical mechanics the action is given as S = L(r, ṙ, t)dt, where t is time. This is very similar to
d @L @L
the form Eq. 144. The extremum of this action satisfies the Euler-Lagrange equation dt @ ṙ @r
= 0. In
the framework ofthe theory of elastic filaments this becomes Eq. 145.
69
Figure 30. Torque around one of the Cosserat directions leads to bending and vice versa.
Each direction along the rod is prone to bending.
where we have taken into account the definition of the "velocity of deformation" or the
Darboux vector. On the other hand, the elastic Euler-Lagrange equation gives explicitly
for the derivative with respect to the Euler angle
✓ ◆
d @f @f
= = µ, (147)
d` @⌦ @
@f
which means that @⌦ is nothing but the internal equilibrium elastic torque within the rod;
the above equation thus states the condition of the elastic equilibrium within the rod: the
sum of the internal and external linera torque density within the rod have to be zero! The
linear elastic torque can be thus obtained as
@f
M(`) = = C⌧ d3 + Kc 2 d2 + Kc 1 d1 . (148)
@⌦
This is the sought for elastic equation of stet of the rod, valid for the elastic equilibrium
of the rod. It states that the torque around any direction of the Cosserat triad along the
elastic rod leads to a non-zero curvature in that direction, as depicted schematically in
Fig. 5.2
To this we can add also the external torque due to the force couple acting on the ends
of the chain that contributes the term
MF = (r(`) r(0)) ⇥ F = F ⇥ (r(`) r(0)). (149)
to the total moment of forces. Since
Z L
(r(`) r(0)) = ṙ(`)d` (150)
0
70 Rudolf Podgornik
we then remain with

Z L Z L
dMF
MF = F ⇥ ṙ(`)d` = (F ⇥ ṙ) d` or else = F ⇥ ṙ. (151)
0 0 d`
Adding now all the linear densities of the moment of forces together, the final equation
for the elastic equilibrium can be written in the form
dM
+ ṙ ⇥ F + µ = 0. (152)
d`
This equation describes the equilibrium of the moment of force density along the rod.
The first term is the intrinsic moment density and the last two are the external force
moment densities.
5.3 Kirchhoff’s equation

Writing equation Eq. 152 in extenso we then remain with
✓ ◆
dM d @f
= = C ⌧˙ d3 +C⌧ ḋ3 +Kc ̇2 d2 +Kc 2 ḋ2 +Kc ̇1 d1 +Kc 1 ḋ1 . (153)
d` d` @⌦
If we now take into account En. 156 and then multiply the elastic equilibrium condition
consecutively with d3 , d2 and d1 , we finally remain with
dM
· d3
C ⌧˙ =
dl
dM
Kc ̇2 + (Kc C)⌧ 1 = · d2
dl
dM
Kc ̇1 + (C Kc )⌧ 2 = · d1 . (154)
dl
These three equations are referred to as the Euler-Kirchhoff equations of the elastic
equilibrium of a thin rod.
Let us assume now that the only source of external torque is the torque density due
to the tractions on the ends of the chain, Eq. 151. In that case remembering that d3 = ṙ,
the first of Eqs. 154 becomes C ⌧˙ = 0 and the second one is transformed into
⇣ ⌘
Kc ̇1 + (C Kc )⌧ 2 = (F ⇥ ṙ) · d1 = d3 ⇥ ḋ1 · F = F · d2 . (155)
The following analysis could be just as well followed by taking the third of equations
Eqs. 154. In nay case, the above equation can re rewritten as follows 71
⇣ ⌘
Kc (̇1 ⌧ 2 ) + C⌧ 2 = Kc d3 ⇥ d̈3 · d1 + C⌧ ḋ3 · d1 = (F ⇥ ṙ) · d1 .
(158)
71 We take into account that
ḋ3 = 2 d 1 1 d 2
ḋ2 = ⌧ d 1 + 1 d 3
ḋ1 = ⌧ d2 2 d 3 . (156)
knowing that the three vectors d3 , d2 in d1 are orthonormal, we remain with
ḋ3 · d2 = 1 ,
71
Remembering also that d3 is simply the tangent vector t = ṙ, it then follows that
...
C⌧r̈ + Kc (ṙ⇥ r ) = F ⇥ ṙ. (159)
This is the Kirchhoff equations that can be solved explicitly in terms of elliptic functions
for various boundary conditions 72 . In the case of external tractions operating at the
ends of the filament this is the most succinct way to write down the equations of elastic
equilibrium.
5.4 Kirchhoff kinematic analogy

Elastic equilibrium equations lead to an interesting interpretation that we tackle next,
in fact hey bear an uncanny similarity to the Euler equations of motion of a symmetric
top. If we assume that the symmetric top has the eigenvalues of the moment of inertia
I1 = I2 = I and I3 in the direction of the long axis, then the equations of motion for the
components of the angular velocity can be derived in the form
!˙ 3 = 0
I3 I
!˙ 1 + !˙ 3 !˙ 2 = M · d1
I
I3 I
!˙ 2 !˙ 3 !˙ 1 = M · d2 , (160)
I
where M · d1 in M · d2 are the components of the torque around the transverse axes.
Euler’s equations of a spinning symmetrical top are thus completely equivalent to the
Kirchhoff’s equations describing the equilibrium deformation of a spherical filament Eq.
154. This equivalence is usually referred to as the Kirchhoff kinematic analogy and is
reiterated in the Table 5.4. This analogy allows us to associate each type of rotation of a
symmetric top with a certain type of deformation of an elastic filament, e.g.
• a sleeping top is equivalent t torsionally deformed straight filament
• precession of a symmetric top is analogous to helical deformation of an elastic

filament
• pendular motion is analogous to sinusoidal deformation of a filament (see Fig. 31)
Euler himself was able to construct a zoo of planar equilibrium shapes of an elastic
filament following this kinematic analogy. Non-planar shapes are a lot more difficult to
construct even with the help of this analogy. Some fo them are presented on Fig.
ḋ3 · d1 = 2 ,
d3 ⇥ d̈3 · d2 = ̇2 + 1 ⌧,
d3 ⇥ d̈3 · d1 = ̇1 2 ⌧. (157)
The last equations follow from the derivative of : d3 ⇥ ḋ3 = 2 d2 + 1 d1 w.r.t `

72 D. Swigon, The Mathematics of DNA Structure, Mechanics, and Dynamics, IMA Volumes in Mathematics
and Its Applications, 150, 293 (2009).

72 Rudolf Podgornik
Symbol Elastic rod Symbol Spinning top
d3 Unit tangent vector d3 Unit director

(d1 , d2 , d3 Orthonormal basis (d1 , d2 , d3 Orthonormal basis
` Contour length t Time
(1 , 2 , ⌧ ) Darboux vector (!1 , !2 , !3 ) Angular velocity vector
EI Principal bending stiffnesses I Principal moments of inertia
dM(`)
d` Linear torque density M(t) Torque
Table 4. A vocabulary of the Euler - Kirchhoff analogy.
5.5 Elastic equilibrium with external stretching forces

Let us start with elastic energy of a filament with symmetric cross section. In this case it
can be written in a symmetric form as derived in Eq. 141.
Furthermore, we note that the third Cosserat vector d3 corresponds to the local tan-
gent to the curve d3 = t and we can thus obtain the first of Eqs. 156 in the form
ṫ = 2 d1 1 d2 , which of course immediately leads to
22 + 21 = ṫ2
since both d1 and d2 are orthogonal unit vectors. Also, by following the definitions of
the three Cosserat unit vectors d3 = t, d2 and d1 we end up with the following identity
73
:
⌧ = t · (d1 ⇥ ḋ1 ) = t · (d2 ⇥ ḋ2 ).
We know, Eq. 61, that this is nothing but the twist density
According to the definition of the twist density, Eq. 50, we can thus associate either
d2 or d1 with the normal n and thus obtain for the square of the twist density
⌧ 2 = (t · (n ⇥ ṅ))2 .
Also, let us add to the free energy a stretching term and assume that the stretching force
is constant throughout the filament and acts only in the direction of the local tangent to
the filament, so that the total elastic energy Eq. 141, with all the inessentail constants
omitted, assumes the final form
Z L Z L Z L
1 2
F = C (t · (n ⇥ ṅ)) d` + 12 Kc ṫ2 d` F · t d`. (161)
2 0 0 0
Here we have written Kc for the curvature elastic modulus of a thin rod, instead of
EI. An open filament under external tension can be shown to be equivalent to a closed
filament loop, with F playing a role of a Lagrange multiplier. How does one see this?
73 This follows from d ⇥ ḋ = ⌧ d (d · t) + ⌧ t(d · d ) +  d and d ⇥ ḋ = ⌧ d2 (d2 · t) +
1 1 1 1 1 1 2 2 2 2
⌧ t(d2 · d2 ) + 1 d1 , as well as the fact the fundamental Cosserat triad is orthonormal.
73
Figure 31. Euler’s classification of the solutions to elastic equilibrium of a filament

constrained to a plane (2D). Euler, L., Methodus inveniendi lineas curvas maximi min-
imive proprietate gaudentes, sive solutio problematis isoperimitrici latissimo sensu ac-
cepti, Lausanne: Bousquet, 1744.
Let us chose a coordinate frame for this closed loop such that the filament is closed in
the (x, y) plane. Therefore in order to enforce the closure of the loop one needs to have
Z L
(z(L) z(0)) = t · nz d` = 0.
0
This constraint can be inforced via a Lagrange multiplier F in the free energy functional
that thus obtains the form
Z L Z L Z L
1 2
F = C (t · (n ⇥ ṅ)) d` + 12 Kc ṫ2 d` F t · nz d`. (162)
2 0 0 0
Obviously the free energy functional of an open filament under external tension Eq. 161
is completely equivalent to a closed filament Eq. 348, except that in the first case F =
(0, 0, F ) is a given external tension and in the second case F is a Lagrange multiplier
ensuring the constraint (z(L) z(0)) = 0.
In order to minimize this free energy functional we have to parameterize r(`), the
shape of the DNA molecule in space. We choose the Euler parametrization in terms of
the three Euler angles that also give explicit forms for the three curvatures ⌧ and 1,2 , see
Eq. 30. Assuming furthermore that the stretching force is in the z directions, we remain
with the elastic free energy functional Eq. 161 parameterized with Euler angles of the
74 Rudolf Podgornik
form
Z ✓ ◆
˙ ˙, ˙] =
L
1 ⇣˙ ⌘2 ⇣ ⌘
F [✓, ✓, d` C cos ✓ + ˙ + 12 Kc ✓˙2 + ˙ 2 sin2 ✓ F cos ✓ .
0 2
(163)
In addition to the elastic free energy we also have the constraint that the total linking
number of the molecule should remain the same, irrespective of its deformations, i.e.
we assume that the deformatins can not change the topology of the molecule. This con-
straint can be expressed via the Călugăreanu-White-Fuller theorem in the Euler angle
parametrization as
Z L
1
Lk[ ˙ , ˙ ] = T w + W r = ( ˙ + ˙ )d`, (164)
2⇡ 0
the form which was derived in Eq. 62. There are several limitations to the use of this
expression that we discussed in the first chapter. In the case where it is appropriate,
this constraint has to be included into the total free energy in the form of a Lagrange
multiplier term, so that we need to minimize the functional
Z L Z L
˙ ˙ ˙ ˙ ˙ ˙
F[✓, ✓, , , ] = F [✓, ✓, , ] p ˙ ˙
( + )d` = ˙ ˙ , ˙ )d`,
f (✓, ✓, (165)
0 0
where p is the Lagrange multiplier chosen in such a way that the linking number com-
puted from Eq. 164 is set and f is given by
⇣ ⌘2 ⇣ ⌘
˙ ˙ , ˙ ) = 1 C ˙ cos ✓ + ˙ + 1 Kc ✓˙2 + ˙ 2 sin2 ✓
f (✓, ✓, F cos ✓ p( ˙ + ˙ ).
2 2
(166)
Obviously the Lagrange multiplier p is nothing but the torque needed to fix the linking
number of the filament. The free energy functional Eq. 165 defines a fundamental elastic
model of a flexible filament under external tension with a set value of the linking number
and is particularly suitable for felxible plymers like DNA with caveats listed at the end
of this chapter.
The minimization problem is in general complicated and can be formulated in terms
of three Euler - Lagrange equations for each of the variables ✓, , in the form
✓ ◆
d @f @f
= 0, and analogously for ✓ ! , ✓ ! , (167)
d` @ ✓ ˙ @✓
that have to be solved simultaneously together with the constraint Eq. 164. We will not
go into technical details that are given by Fain et al. and Jülicher. Let us juts add at the
end that we have not included any effects of thermal fluctuations which for soft filament
like DNA are never negligible and have not taken into account any (self)interactions
between the differrent segments of the elastic filament which for a charged molecule like
DNA is also not particularly realistic.
5.5.1 Torsional instability of a filament
Let us investigate the deformation of a filament firmly embedded into a support on one
side and under an external traction of magnitude F . Assume first that the shape is con-
strained to coincide with a twisted straight line, = 0 and ✓ = 0, with a total twist given
75
by the linking number Lk and its writhe obviously vanishing W r = 0. In this case the
free energy Eq. 165 is given by the expression
Z L  Z L
˙ 1 ˙2 ˙ 1 ˙ (`)d`.
F[ , ✓ = 0, = 0] = d` C (`) F p (`) and Lk =
0 2 2⇡ 0
(168)
The Euler-Lagrange equation for this free energy reduces to
˙ (`) = p and Lk =
1 p
L. (169)
C 2⇡ C
From this one deduces that the equilibrium elastic free energy of the twisted straight rod
2
is F[ ˙ , ✓ = 0, = 0] = 2LC ⇡ Lk L F L. Clearly the value of the elastic free energy
depends on external traction and is as small as we want for large enough traction. It can
be shown quite generally that the the twisted straight line with specified linking number
always corresponds to the minimum of the elastic free energy 74 . For small force it is not
clear that the above solution remains stable, though.
The stability of the twisted horizontal line can be assessed by perturbing its shape
into the third dimension . The general shape equations are obtained from minimizing the
complete free energy Eq. 166. This leads to the following Euler-Lagrange equations for
the Euler angles that can be cast into the form
˙ p p(1 cos ✓) cos ✓

=
C Kc sin2 ✓
˙ p (1 cos ✓)
=
Kc sin2 ✓
1 p2
Kc ✓˙2 + F cos ✓ + = E0 , (170)
2 Kc (1 + cos ✓)
where E0 is the total elastic energy of the filament. Note the change in sign of the
"potential energy", F cos ✓. We also note that the total elastic free energy density Eq. 163
is a functional that does not contain ` explicitly, only , and ✓ with their derivatives.
This energy functional is thus analogous to the action integral in classical mechanics that
gives rise to the Euler-Lagrange equations. It then follows by the well known theorem
of classical mechanics that there exists a first integral of the Euler-Lagrange equations
which is exactly the third line in Eq. 170. These equations are general, valid for any
shape that minimizes the elastic energy Eq. 166.
We assume now that the simplest perturbation from a twisted straight line is given
by ✓(`) with ✓(`) ⌧ 1. The corresponding form of the Euler-Lagrange equations then
becomes to the lowest order in the perturbation
˙+ ˙ p
=
C
˙ p
=
2Kc
✓ ◆
p2
Kc ✓¨ F ✓ = 0, (171)
4Kc
76 Rudolf Podgornik
(a) q0 = 0.0 (b) q0 = 0.32
(c) q0 = 0.35 (d) q0 = 0.4
Figure 32. Formation of a plectoneme via torsonal instability. The Coyne solution cor-
responds to the sequence a, b, c, d. Courtesy L. Curovic.
where the connection between p and the linking number was derived in Eq. 169. Thus for
2
1 > 4Kp c F the twisted straight rod solution is stable, giving rise to a harmonic dependence
⇣ ⌘
p2
of the deviation on the arclength, ✓(`) ⇠ e±i!` , where ! 2 = F 4K c
/Kc . On the
2 2 2
contrary, for 1 < 4Kp c F = KCc F ⇡ Lk
L the solution is unstable, ✓(`) ⇠ e±|!|` , and the
rod escapes into the third dimension. The bifurction point between the two regimes is
then given by
p2
1= . (172)
4Kc F
Though this result was obtained from a linear stability analysis we will show later that it
is consistent also with complete non-linear analysis.
This scenario is accomplished by the filament first looping over, shedding and con-
verting some torsion into writhe. A fully unstable filament is finally converted into a
single and later a multitude of plectoneme(s) 75 along its length, see Fig. 32. This type
75 meaning ”braided string”, from Greek ⌫⌘µ↵ - thread, ⇡ ⌘⌧ o⇣ - twisted.
77
of transition can be often observed also in telephone wires that are unconsciously twisted
each time the phone is used, eventually creating a profusion of plectonemes.
The shape phase diagram describing the sequence of equilibrium shapes of the fila-
ment in this case therefore consists of
• a twisted straight line with W r = 0

• a plectoneme with W r 6= 0
where the straight twisted line is stable only if the parameters describing the solution
C2 2
obey the relation F < K c
⇡ Lk
L .
Similar conclusions can be reached also in the case of a closed elastic filament, e.g. as
in the case of a circular DNA. In that case the solution of the Euler- Lagrange equations
give rise to various closed shapes, depending on the values of the linking number dif-
ference Lk and the ratio C/Kc . The shape phase diagram here contains four different
regions
• planar circles with W r = 0

• non-planar rings with 0 < W r < 1
• self-interacting rings with 1 < W r < 2
• interwound configurations with W r > 2.
For large values of the ratio C/Kc transitions between these various shapes, like the
buckling transition of the circle into a non-planar ring, are continuous but they become
discontinuous for
p small enough C/Kc . In general the loop remains a circle, W r = 0<
until Lk = 3 KCc , at which point it undergoes a transition into a non-planar form.
For KCc  12 the linking number can be increased until it reaches Lk = 1 of the figure
eight and then further proceeds into a plectonemic form. For 12  KCc  1 the curve
continuously distorts to a shape with writhe between 0 and 1. And finally for KCc 1
p K
as soon as Lk 3 C the planar circle snaps into a plectoneme. The length of the
c
filament does not feature in any of these considerations 76 .
5.5.2 Coyne solution
We now address finite deformation of a straight twisted line close to the instability point
calculated in the previous section. The equations giving the profile of the deformed
filament were derived in the previous section and have the form
˙ p (1 cos ✓)
=
Kc sin2 ✓
1 p2
Kc ✓˙2 + F cos ✓ + = E0 . (173)
2 Kc (1 + cos ✓)
76 B. Fain, J. Rudnick and S. Östlund, Conformations of linear DNA, PRE 55 (1997) 7364. F. Jülicher,
Supercoiling transitions of closed DNA, PRE 49 (1994) 2429. B. Fain, J. Rudnick, Conformatins of closed
DNA, PRE 60 (1999) 7239.
78 Rudolf Podgornik
2
p
The total elastic energy per unit arclength has to be equal to E0 = F + 2K c
, since at
˙
the boundaries the filament is straight, implying ✓, ✓ = 0. Combining the two we then
remain with an equation that contains only ✓ and its first derivative
r r
F ✓ ✓
✓˙ = 2 tan cos2 1+ 2
Kc 2 2
r p
F 1 2
˙ = , (174)
Kc cos2 2 ✓
with s◆2 ✓
p
= 1 . p (175)
2 Kc F
q
Introducing the dimensionless arclength as u = ` KFc , the two equations can be inte-
grated straightforwardly yielding the Coyne solution 77 of the form
1
✓(u) = 2 sin
cosh u
p tanh u
(u) = 1 2 u + arctan p (176)
1 2
and is presented on Fig. 32. The origin of ` has been chosen so as to coincide with the
midpoint along the filament so that L/1 < ` < L/2. We observe that the total curvature
of this shape is maximal at the midpoint and then decays exponentially towards the two
ends. Because of this curvature the projection of the curve onto the z axis is shortened.
The above solution does not take into account any boundary conditions at the two
ends of the filament. In this respect it is approximate, but in every other respect, ex-
act. The effect of end-point boundary conditions will be in general negligible for long
filaments, which is the limit we consider here.
This shortening of the projection can be calculated as follows. The difference be-
tween the distance along the filament and the projection on the z axis, b(`), satisfies
d db(`) ✓(`)
(` z(`)) = 1
2 =1 cos ✓(`) = 2 sin2 , (177)
d` d` 2
where ✓(`) is given in Eq. 176. Since ✓(`) is symmetrical around the midpoint of the
filament it is obvious that z(`) is antisymmetrical and is obtainedqby a straightforward
integration noting furthermore that z(0) = 0 yielding b(`) = 4 KFc tanh `. One
thus obtains for the displacement of the two end-points of the filament b(±L/2)/2 the
following result
r
b(L/2) b( L/2) L Kc L
= = z(L/2) = 2 tanh . (178)
2 2 2 F 2
77 J. Coyne, Analysis of the Formation and Elimination of Loops in Twisted Cable, IEEE J. Oceanic Eng. 15
72 (1990).
79
This means that

q for a very long filament the end displacement reaches a finite limit
b(L!1)
2 = 2 KFc . The total elastic energy of the deformed filament can be obtained
from Eq. 166 when one takes into account the Euler-Lagrange equations Eq. 173
Z 
1 ⇣ ⌘2 ⇣ ⌘
F(p, F ) = d` C ˙ cos ✓ + ˙ + 12 Kc ✓˙2 + ˙ 2 sin2 ✓ F cos ✓ =
2
Z  2
p p2 L
= d` F + 2F (1 cos ✓) = F L + 2F b(L/2), (179)
2C C
according to the definition of b(L/2). The equilibrium value of the torque p is then given
by
dF(p, F ) pL db
=0 or + 2F = 0. (180)
dp C dp
q
For a very long chain we obtain from Eq. 178 that b(L/2) = 4 KFc (p), where (p) is
p q
F b2
given by Eq. 175 and p = 2Kc F 1 16K c
. Solving the above differential equation
for p(b) we are left with
r
4C b F
p(b) = p0 arcsin , (181)
L 4 Kc
where the value of torque at zero deformation of the filament, p0 = p(b = 0) = 2⇡LC Lk,
can be taken from the linear
q stability analysis and depends on the imposed linking number
Lk. Apparently when 4b KFc = 1 we have p(b) = p0 2⇡LC = 2⇡LC (Lk 1). It thus
follows that one complete twist is removed from the filament when b reaches this value.
From here we can easily derive also the expression for the force F as a function of
the extension b. The force needed to maintain a straight configuration of the filament,
b = 0, is then seen to satisfy
p20
= 1.
4Kc F
which is exactly what we derived in Eq. 172 within a linear stability analysis. As soon
s the external traction F is smaller than this limiting value, the filament develops a loop
as described by the Coyne solution. One needs to remember here that this solution is not
an exact solution of the plectoneme formation problem but is asymptotically valid in the
limit of long filaments.
Apart from that, there are other inherent limitations for the Coyne solution. First
of all by construction this solution does not take into account the impenetrability of the
chain and allows phantom self-crossings. These crossings have to be dealt with if one
wants to explore not just incipient pkectoneme formation, but also the actual growth of
the plectoneme. The other important and fundamental limitation of the Coyne solution is
the implementation of the Fuller theorem, Eq. 164. It turns out 78 that the implementation
that we use is inappropriate for a fully developed plectoneme, it is however acceptable
for an incipient plectoneme that does not show any full plectonemic turns yet.
78 S. Neukirch and E. L. Starostin, Writhe formulas and antipodal points in plectonemic DNA configurations
, Phys Rev E 78 041912 (2008).

80 Rudolf Podgornik
Figure 33. Formation of a plectoneme in an optical tweezer experiment. Adapted from

Fast dynamics of supercoiled DNA revealed by single-molecule experiments, A. Crut et
al., and Discrete Elastic Rods M. Bergou et al. For larger values of the twist a more
complicated configuration of many plectonemes represents an elastic equilibrium.
5.5.3 Plectoneme elasticity
One can deduce from the approximate Coyne solution of the previous section that for
a constant force clamp the relation between the end-point displacement and the applied
torque is given by the implicit equation
✓ ◆ r !2
b2 F 4C b F
4F Kc 1 = p0 arcsin . (182)
16Kc L 4 Kc
Just above the plectoneme bifurcation and in the limit of small external traction F , the
shortening of the chain b and the externally imposed twist, as quantified by the linking
number, are thus approximately linearly related as
✓ ◆
2LKc p0
b' p 1 . (183)
C 4Kc F
p0
Therefore for subcritical values of twisting p4K < 1 there is no shortening of the
cF
filament, while for supercritical values b depends linearly in the amount of supertwisting.
As the linking number of the filament is increased at a constant force, the overall length of
the filament diminishes since some of the torsional deformation is converted into bending
and proliferation of one and then many plectonemes. One should add here that this is a
picture obtained by ignoring bending and torsional fluctuations of the molecule as well
as the fact that the linking number is expressed by a local formula, with a limited range
of validity.
Let us therefore formulate the elastic energy of a plectoneme based on a general
expression for the linking number that is not dependent on the validity of the local form
81
of the writhe 79 . For an open filament the Călugăreanu-White-Fuller theorem can be in

general written as
Z L
1 sin 2↵ LP
Lk = T w + W r = !(s)ds =
2⇡ 0 4⇡a
✓ ◆
1 sin 2↵ LP
= ⌧L , (184)
2⇡ 2ia
where we assumed that the twist is uniform, !(s) = ⌧ , so that the torque exerted on
the end of the filament is the same over its whole length. We also took into account the
general definition of writhe for a helical curve, Eq. 55, and assumed that of the total
length of the filament, L, only the portion LP , pertains to the plectoneme. The rest are
undeformed tails or so we assume. Here a is the radius of the plectoneme helix and
↵ is the angle between the local direction of the helix and the average direction of the
plectoneme.
The elastic energy of the plectoneme under an external traction F and an external
torque M with undeformed tails is then obtained as
Z L Z L Z L Z L
1 2 1 2
E = 2 Kc ⌧ (s) ds + 2 C !(s) ds F · dr(s) M · d (s) =
0 0 0 0
sin ↵4
= 1
2 K c LP + 12 C⌧ 2 L F (L LP ) M (2⇡ Lk). (185)
a2
The curvature in the above expression was derived from Eq. 36 and we assumed that the
tails of the plectoneme are straight and undeformable so that the effective length of the
plectoneme + tails is (L LP ). This of course means that we completely ignored the
effect of thermal fluctuations, which is of course what this chapter is all about.
The above plectoneme energy still lacks an interaction term between the helices in
the plectoneme proper. This energy is doe to interactions between DNA molecules and
is in general quite complicated, see the chapter on DNA-DNA interactions. At this point
we will simplify the calculation and assume that the radius of the plectoneme equals the
radius of the DNA molecule, a0 , i.e. that the two strands of the plectoneme touch. Putting
everything together we then remain with the following total energy of the plectoneme
✓ ◆
sin ↵4 sin 2↵ LP
E(⌧, ↵, LP , a) = 12 Kc 2 LP + 12 C⌧ 2 L F (L LP ) M ⌧ L (a a0 ),
a 2⇡a
(186)
where is the Lagrange parameter setting the value of the radius of the plectoneme.
In equilibrium this free energy has to be minimized w.r.t. all of its arguments. What
we are after is then the dependence of the extension of the plectoneme, (L LP ), as a
function of the turns imposed on the chain, Lk. Using Eq. 184 and setting the derivative
of E(⌧, ↵, LP , a) w.r.t. ↵ to zero, we remain with
✓ ◆
2Kc sin ↵2 4⇡a
b(Lk) = (L LP ) = 1 + L Lk. (187)
C cos 2↵ sin 2↵
79 N. Clauvelin, B. Audoly, and S. Neukirch, Mechanical Response of Plectonemic DNA: An Analytical
Solution, Macromolecules (2008), 41, 4479. N. Clauvelin, B. Audoly, and S. Neukirch, Elasticity and Electro-
statics of Plectonemic DNA, Biophysical Journal Volume 96, 2009 3716.
82 Rudolf Podgornik
We should not forget here that the above relation was derived for zero temperature and
does not contain any effect of the fluctuations, so the direct comparison with experiment
is a bit tricky, but feasible 80 . The especially interesting conclusion of the above analysis
is that the slope of (L LP ) as a function of Lk is linear, in fact
d b(Lk) 4⇡a
= . (188)
d Lk sin 2↵
In fact experiments on DNA stretching and twisting show that there is a regime for suffi-
ciently large twisting angles where the above linear relationship holds true, see Fig. 34.
Later, when we will be analyzing the effect of thermal fluctuations we will see that for
small values of externally imposed twist the corresponding relation is quadratic.
5.5.4 Twisting DNA
A simple experiment where the extension of DNA fixed to a magnetic bead and a solid
support is monitored while it is twisted reveals the emergence of a plectoneme. The
schematics of the experiment is shown on Fig. 35. Ignoring the fact for a moment, that
DNA in its relaxed configuration is already twisted because of its double helical nature,
additional twisting of a constrained DNA decreases the separation between the magnetic
bead and the solid support to which the DNA is attached.
Experimentally observed changes in the length of the DNA under positive or negative
twisting are not symmetric. Only overtwisted DNA shows a linear relationship between
Lk and b, whereas undertwisted DNA shows very little dependence between the two.
This is due to the fact that with undertwisting it is energetically more favorable for the
chain to locally melt then escape into the third dimension via plectoneme proliferation.
This approximate relation between b and torque is born out also buy experiments at small
values of the external traction.
Wat one can note is also that there is a small regime of quadratic dependence of
b(Lk) which can not be adequately explained by the present theory. In order to do that
one would need to take into account also the effect os thermal fluctuations on the elastic
equations of state, an exercise we shall set ourselves to in a later chapter.
5.6 Buckling instability
Elastic filament subject to an external compressional load directed along a straight fil-
ament is subject to an instability first investigated by Euler. It arises from the fact that
it is easier for the filament to escape into the third dimension then remain subject to
compressional forces. Another variant on this theme comes about when the external
compressional load is substituted by the attractive self-interactions along the filament. In
both cases the final state of the filament is presented by a buckled shape.
80 N. Clauvelin, B. Audoly, and S. Neukirch, Mechanical Response of Plectonemic DNA: An Analytical
Solution, Macromolecules (2008), 41, 4479.

83
Figure 34. Experimental dependence of the extension of DNA, L = 48kbp, as a function

of the number of turns imposed on the magnetic bead, Lk, for positive values only. The
imposed external traction F for each curve is: 0.25, 0.33, 0.44, 0.57, 0.74, 1.10, 1.31,
2.20, 2.95 pN. Triangles represent the fit for the slope in the linear region. For details see
N. Clauvelin, B. Audoly, and S. Neukirch, Mechanical Response of Plectonemic DNA:
An Analytical Solution, Macromolecules (2008), 41, 4479. Schematic representation of
the plectoneme adopted from B. Daniels, Cornell University.
5.6.1 Euler buckling
For a filament confined to a plane, = 0, = 0, the corresponding free energy is

simplified to
Z L 
˙ 1
F[✓, ✓] = d` Kc ✓˙2 F cos ✓ . (189)
0 2
Because of this confinement of the filament it is torsionally relaxed and the linking con-
strained does not exist anymore. The corresponding Euler - Lagrange equation has the
form
¨
Kc ✓(`) F sin ✓(`) = 0 or ✓00 (s) q 2 sin ✓(s) = 0 (190)
if we introduce s = `/L and q 2 = F L2 /Kc with prime standing for the derivative with
respect to s. The Euler-Lagrange equation is thus just the Euler equation for a planar
torsionally relaxed elastic filament. The Euler’s equation, or equations closely related to
it, feature prominently in many areas of physics such as statistical mechanics (Poisson-
Boltzmann equation), field theory (sine-Gordon equation) etc.
Apart from the trivial solution that always exists, ✓ = 0, the Euler equation can be
84 Rudolf Podgornik
Figure 35. Equilibrium shape of an elastic filament with different external loads F at the
free end corresponding to  = 0.1[1], 0.2[2], 0.3[3], ..., 0.9[9]. The lower end is clamped.
The filament is bend homogeneously at small loads, while the bending at lower end is
increased when the magnitude of the load increases.
solved also by a nontrivial solution that can be expressed in terms of the Jacobian elliptic
function sinus amplitudinis as 81
✓(s) = 2 arcsin ( sn(, qs + A)) , (191)
where  and A are the two integration constants. These solutions have already been
investigated in detail by Euler himself 82 , and have the form as shown on Fig. 35. As-
suming that the filament is tightly clamped at the begininng (✓(` = 0) = 0) and free at
˙ = L) = 0), we obtain two equations for the two constants, i.e.
its end (✓(`
0 = sn(, A) and 0 = cn(, qL + A), (192)
thus A has to be a zero of the sinus amplitudinis and qL + A has to be a zero of the
cosinus amplitudinis functions. From here, by taking note of the connections between
the different elliptic functions, we deduce that
q = K[]. (193)
Analyzing the functional dependence of K[x] on its argument it is found that K[x] ⇡2
for any value of the argument. Invoking then the definition of q, it is thus clear that the
81 We define standardly the sinus amplitudinis function implicitly as F [sn(x, ), ] = x, with F [k, ] =
R p
dx/ 1 k2 sin2 x and K[k] = F [k, ⇡/2].
0
82 Max Born in his Ph.D. thesis among other things investigated the forms of Euler solutions via mechanical
models, M. Born, Stabilitaet der elastischen Linie in Ebene und Raum, Preisschrift und Dissertation, Goettin-
gen, Dieterichsche Universitaets-Buchdruckerei Goettingen, 1906. Reprinted in: Ausgewaehlte Abhandlungen,
Goettingen, Vanderhoeck & Ruppert, 1963, Vol. 1, 5-101.
85
nontrivial solution of the Euler’s equations exists only if
⇡ 2 Kc
F = Fc . (194)
4L2
When the force reaches this critical value, Fc , there is a bifurcation and the filament
buckles from a straight shape - the trivial solution - into a curved one. This is the well
known phenomenon of Euler buckling.
Since this kind of instability is seldom seen in DNA world, we will analyze a related,
Manning instability, more in detail in what follows. The Manning instability is not driven
by external force but by intrinsic DNA self-interactions but witha similar outcome.
5.6.2 Manning buckling
If the force pressing on the ends of a filament exceeds a certain limit, the filament buckles
in what is usually referred to as the Euler’s buckling instability. For DNA this type of ex-
ternal traction driven instability is not particularly relevant, but an associated instability,
that I refer to as the Manning instability 83 , is. In this case the filament self-interacts with
an attractive potential V (r(`) r(`0 )) that plays a role similar to an external compressing
force thus leading to an self.interaction driven elastic instability.
The idea of an interaction-induced buckling instability is really quite intuitive. At the
point where the elastic energy of bending can no longer compensate for the change of the
interaction energy due to diminished separations between the interacting segments of the
polymer chain, the polymer will buckle. Clearly, this buckling depends on the persistence
length of the polymer as well as on the “strength” of the attractive interactions between
its segments.
Assume that the elastic filament is not torsionally stressed and that the elastic free
energy is given by
Z L ✓ ◆2 Z L Z L
Kc @2r 1
F= d` + d`d`0 V (|r(`) r(`0 )|) , (195)
2 0 @`2 2 0 0
where r(`) denotes a general parametrization of the the filament, ` the contour length,
and |r(`) r(`0 )| is the distance in embedding space betweeen two monomers at ` and
`0 . It is not convenient in this case to go to the Euler angle parametrization because
the interaction potential depends on the values of the coordinates and not their arclength
derivatives.
In the mean-field approximation we determine a typical configuration of the polymer
by functionally minimizing the free energy w.r.t. r(`) 84 . The result of such a variation
is:
Z
@ 4 r(`)
Kc d`0 F(|r(`) r(`0 )|) = 0 , (196)
@s4
83 It was first described by G. Manning in Packaged DNA: an Elastic Model, Cell Biophysics 7 57 (1985).
At that time it was not clear that the self-interactions along the DNA filament can be attractive, but the idea of
the interaction driven instability is clearly there.
84 For a Lagrangian of the form L(r̈, ṙ, r) the EL equation has the form d2 @L d @L
+ @L = 0.
d`2 @r̈ d` @ ṙ @r
86 Rudolf Podgornik
where F(|r(`) r(`0 )|) = @V /@r(`) is the local linear force density. This equation
determines the typical polymer configuration subject to appropriate boundary conditions.
For small deviations from a straight configuration, the filament can be parameterized
as r(`) = (z, ⇢(z)), where z is the direction of the axis of the molecule and |⇢| is the
radial distance from that axis. In this parameterization a linearized form of the Euler-
Lagrange equation can be derived in the form 85 deformations, where
Z L ✓ ◆ Z L
d4 ⇢(z) 0 d d⇢(z) 0 0 @V (|r(z) r(z 0 )|)
Kc + dz V (|r(z) r(z )|) = dz .
dz 4 0 dz dz 0 @⇢(z)
(197)
The first term on the l.h.s. of Eq. (197) stems from the curvature energy. The second is
the longitudinal stress acting along the deformed rod. The term on the r.h.s. corresponds
to the transverse bending force. Both of the last two terms depend on the characteristics
and the details of the interaction potential.
In the case of short range interactions, the r.h.s. term of the Euler-Lagrange equation,
Eq. (197), can be simplified further. Significant contributions to the integral on the r.h.s.
of Eq. (197) come only from the points along the polymer for which z and z 0 are not very
far apart. Therefore one can develop V (r(z) r(z 0 )) in powers of z z 0 , truncating the
Taylor expansion at the first order term. To the lowest order in ⇢0 (z), we then obtain the
Euler-Lagrange equation in the form
Z L
d4 ⇢(z) d2 ⇢(z)
Kc + dz 0 V (|z z 0 |) '0. (198)
dz 4 0 dz 2
Eq. (198) is the fundamental mean-field equation describing the shape of the elastic rod
in the limit of small deformations and short range self-interactions. For the Eulerian
solution that is contained within a plane, the above equation can be transformed into an
Ermakov-Pinney equation and solved exactly as we show below.
d2 ⇢(z)
In the Eulerian limit where the filament remains coplanar, dz2 = u(z) is also
coplanar and can be treated as a scalar. The mean-field equation is not easily solvable
except for a limited variety of potentials, V (t). Nevertheless, a formal solution can be
obtained by introducing the following ansatz for u(z)
u(z) = C⌘(z) sin ( (z, 0)) . (199)
C is a normalization constant introduced in order to make ⌘(z) dimensionless. The func-

tions ⌘(z) and (z, 0) are easily shown to satisfy the following Ermakov-Pinney equa-
tions:
0 2
⌘ 00 (z) + ⌦2 (z)⌘(z) ⌘(z)( ) (z, z 0 ) = 0
2 0 0 0
⌘ (z) (z, z ) = 0. (200)
where Z L
⌦2 (z) = Kc 1
dz 0 V (|z z 0 |).
0
85 P.L. Hansen,D. Svensek, V.A. Parsegian, and R. Podgornik, Buckling, fluctuations, and collapse in semi-
flexible polyelectrolytes, PRE 60 1956 (1999). Note a typo in the sign of the basic equation in this paper.
87
A general stability analysis shows that the filament buckles when the following identity
is fulfilled: Z L
dt
2
= ⇡ L, (201)
0 ⌘ (t)
which follows from the fact that the non-trivial solution should satisfy the boundary
condition u(z = 0, L) = 0, which is translated into sin (L, 0) = 0. Eq. (201) selects
the lowest mode compatible with this condition. Because the function ⌘(z) still has to
be evaluated, this is just a formal statement. It then follows from the above equation that
the necessary condition for the existence of the instability is
Z L
dz 0 V (|z z 0 |) < 0, (202)
0
i.e. the interaction potential needs to have an attractive branch. Assuming a short range
attractive interaction potential, e.g. V (|z z 0 |) = 2 3 (z z 0 ), one remains with
⌦2 (z) = 2 . In this case a deformation grows exponential as a function of z. The
properties of this instability (bifurcation) point are the same as in the case of the simpler
Euler instability where the elastic rod is simply compressed at both ends by a transverse
force. The Manning buckling in an attractive self-interaction potential represents the first
steps to an ordered collaps of the DNA molecule into a toroidal shape, to be discussed
later.
5.7 Euler angle of the pinched teardrop configuration

Assume now a deformed elastic filament with coincident ends that makes a teardrop-
like configuration showing an interesting universality. The filament length is L, and we
deform it in a plane in such a way that both ends coincide. At the end-points there is by
definition no torque and the corresponding curvature has to be zero. where ✓0 = ✓(` = 0)
Figure 36. Euler’s angle and the teardrop configuration of a bent lastic filament.
is introduced as the Euler’s angle. Solving this equation gives a teardrop shape, as shown
88 Rudolf Podgornik
on Fig. 5.7 with a closing angle at the junction of the two ends as ⇡ 2✓0 . By symmetry
the x-coordinate at the middle of the filament has to coincide with the coordinate of the
end (beginning) of the filament, therefore
Z L/2
cos ✓(`)d` = 0. (203)
0
It then follows from Eq. 193 that

Z L/2
1 ˙ 2 d` = F L cos ✓0 ,
✓(`) (204)
2 0 Kc 2
wherefrom
Z L/2 Z ⇡ Z L/2
˙✓(`)2 d` = ˙ F L F ˙ 1
✓(`)d✓(`) =2 cos ✓0 = cos ✓0 ✓(`) d✓(`).
0 ✓0 Kc 2 Kc 0
(205)
˙
Inserting now ✓(`) as a function of ✓(`) from Eq. 193, one obtains an implicit equation
for the Euler’s angle ✓0 that does not contain F or Kc any more 86 . In this sense its solu-
tion is a constant of Nature. Numerically the Euler’s angle is obtained as approximately
25o . The closing angle between both ends of the teardrop shaped filament is then 130o .
From the above analysis one can also calculate the corresponding force F as a func-
tion of the length of the filament
✓ Z ⇡ p ◆2
2Kc 1 Kc
F (L) = cos ✓0 cos ✓d✓ ' 15.508 (206)
L2 cos ✓0 ✓0 L2
one needs to apply to the touching ends of the filament in order for it to be in the teardrop
configuration. The corresponding elastic free energy of the filament then follows as
Z L ✓Z ⇡ p ◆2
˙ Kc 2 Kc
F(L) = f (✓, ✓)d` = cos ✓0 cos ✓d✓ ' 14.055 . (207)
0 L cos ✓0 ✓0 L
If we introduce the persistence length of DNALP via the thermal energy as Kc =

kB T LP , then it is obvious that in units of thermal energy the elastic energy of a teardrop
configuration becomes comparable with the thermal energy for length on teh order of 10
times the persistence length.
The elastic free energy Eq. 207 is obtained by dividing the loop into two halves at the
(symmetry) mid-point and taking into account Eqs. 203, 204, as well as calculating the
integrals numerically. Obviously the elastic energy of a teardrop loop diverges for small
L. The teardrop bending energy is about 71% of the bending energy of the circle which
is given straightforwardly by 2⇡ 2 KLc 87 .
86 This
R⇡ p R⇡ d✓
equation has the form: cos ✓0 cos ✓d✓ = cos ✓0 p .
✓0 ✓0 cos ✓0 cos ✓
87 J. F. Marko, INTRODUCTION TO SINGLE-DNA MICROMECHANICS, Department of Physics, Uni-
versity of Illinois at Chicago preprint (AUGUST 9, 2004).

89
6 STATISTICAL MECHANICS OF DNA

DNA usually lives at or around the room temperature and is a quintessential soft object.
If we look at the elastic free energy of an e.g. teardrop configuration, Eq. 207, we
see that for sufficiently long DNAs it becomes comparable and eventually smaller then
the thermal energy. That means that long enough DNA becomes susceptible to thermal
fluctuations and its equilibrium configurational properties derive in part from its thermal
motion.
Apart from simulations one can formulate several coarse grained models of DNA
elasticity that allow for a complete statistical mechanical treatment. These models are
• Inextensible semiflexible chain model or the Kratky-Porod model that takes

into account DNA bending but not torsional elasticity, one thus assumes that DNA
is torsionally relaxed and the torsion term can be omitted from the elastic energy,
leaving only with the bending term.
• Extensible semiflexible chain model allows also for the effects of local exten-
sibility of the DNA chain which are important for large external tractions on the
chain and lead to Hookian elasticity
• Inextensible rod-like chain model takes into account DNA bending as well as
torsional elasticity. This model applies to the most complete experimental probing
of DNA elasticity.
Within each of these model one can derive an appropriate elastic equation of state for
the polymer chain valid in a certain regime of externally imposed forces and torques.
DNA elastic equation of state, apart from its osmotic equation of state, is one of the most
studied DNA properties and allows us to understand its behavior in terms of a few elastic
parameters such as the bending and the torsional rigidities.
While the value of the DNA bending rigidity as quantified by its persistence length
was first established in the ’50s, it is only recently that its importance for DNA packing in
chromatin, bacterial nucleoid and viruses has become the focus of experimental as well
as theoretical attention.
In what follows we will try to analyze some effects of thermal fluctuations assuming
that in every configuration the total energy of the DNA filament is given by its elastic
deformation energy. Also we will ignore some microscopic details of the DNA structure
such as its double helical nature and assume that it is a featureless circular filament. The
conseqiences of the helical nature of DNA on its thermal properties have been investi-
gated extensively by Yamakawa 88 .
6.1 Kratky-Porod model

For a torsionally relaxed filament, viz. a filament whose elastic energy does not depend
on the torsional deformation, with symmetric cross section we just found out that the
88 H. Yamakawa, Helical Wormlike Chains in Polymer Solutions, Springer; 1 edition (1997).
90 Rudolf Podgornik
elastic energy can be written as

Z L Z L ✓ ◆2
dt
F [ṫ(`)] = 12 Kc ṫ2 d` = 12 Kc d`. (208)
0 0 d`
This expression look like a kinetic energy of a particle with "position" t and "time" `. In
addition such a particle would have to move on a unit sphere since t2 = 1. For a filament
that is at a constant non-zero temperature one should study its free energy as oppose to
its energy. It can be obtained from the partition function that is defined as
Z
Z(T, L) = D[t(`)] e F [ṫ(`)] with t2 (`) = 1. (209)
where is the invesre thermal energy, 1/ = kB T and the integral has to be performed
over all the unit vector t. The partition function is thus analogous to the probability am-
plitude of a quantum mechanical particle with kinetic energy Eq. 208, living on a unit
sphere, if one identifies the arclength of the filament with "imaginary time" 89 . Calcu-
lations of the partition function of this model of DNA can be obtained expoliting this
fundamental analogy.
Formally we can write down the relationship between the partition function Eq. 243
and the probability density, that starting with a tangent vector t1 (0) the filament ends up
with a tangent vector t2 (L) which we denote as Z(t2 (L), t1 (0); L), as
Z Z
Z(T, L) = dt1 dt2 Z(t2 (L), t1 (0)). (210)
From this formal relationship can derive an equation for the probability density Z(t2 (L), t1 (0); L)
analogous to the Schrödinger equation in the form
@Z(t2 (L), t1 (0); L) 1 2
= L Z(t2 (L), t1 (0); L), (211)
@L 2Lp
where Lp = Kc is the persistence length and L2 is the angular part of the Laplace
operator which is given by
✓ ◆
2 1 @ @ 1 @2
L = sin ✓ + . (212)
sin ✓ @✓ @✓ sin ✓2 @ 2
This equation is a straightforward consequence of the analogy between an elastic filament
and a quantum mechanical particle on a unit sphere. Since the filament looks statistically
the same along all of its contour one can write
ht(`2 ) · t(`1 )i = ht(`2 `1 ) · t(0)i = hcos ✓(`2 `1 )i . (213)
90
If we now take note of the statistical definition of hcos ✓(`2 `1 )i we can derive
straightforwardly from Eq. 211 that
@ hcos ✓(`2 `1 )i 1 ⌦ 2 ↵ 1
= L cos ✓(`2 `1 ) = hcos ✓(`2 `1 )i . (214)
@L 2Lp Lp
89 Randall D. Kamien: The geometry of soft materials: a primer,R Rev. Mod. Phys. 74, 953âĂŞ971 (2002).
+1
d(cos ✓) cos ✓ Z(t2 (`2 ),t1 (`1 ))
90 This is given in the standard fashion as hcos ✓(`2 `1 )i = R1 +1 .
d(cos ✓)Z((t2 (`2 ),t1 (`1 ))
1
91
The last equality follows from the fact that L2 is the angular part of the Laplace operator
and thus that L2 cos ✓(`2 `1 ) = 2 cos ✓(`2 `1 ). Therefore one obtains that
(`2 `1 )/Lp
hcos ✓(`2 `1 )i = e . (215)
The directional correlations along an elastic filament thus decay exponentially, where the
decay length is equal to the persistence length. The persistence length is thus the correla-
tion length for elastic correlations along the filament. This can be seen straightforwardly
from the length of the chain directed along the direction of the beginning of the chain i.e.
* Z !+ Z
L L
ht(0) · (R(L) R(0))i = t(0) · t(`0 )d`0 = e `/Lp
d`0 =
0 0
⇣ ⌘
L/Lp
= Lp 1 e . (216)
RL RL
Here we have taken into account that R(L) R(0) = 0 dr(`) d` d` = 0 t(`)d`. For
long filaments the length of the chain directed along the direction of the beginning of the
chain thus saturates at the value of the persistence length.
We are now in a position to calculate what is the statistical shape of an elastic filament
in a brownian thermal bath. Let us just remind ourselves that the average square of the
end to end separation of the filament is given by
D E Z LZ L ⌧ Z LZ L
2 dr(`) dr(`0 )
(R(L) R(0)) = d`d`0 = d`d`0 ht(`)t(`0 )i =
0 0 d` d`0 0 0
Z LZ L
= d`d`0 hcos ✓(` `0 )i . (217)
0 0
Using now Eq. 215 for the angular average we remain with
D E ⇣ ⌘
2 L/Lp
(R(L) R(0)) = 2Lp L Lp + Lp e . (218)
This general result was first derived by Kratky and Porod91 and thus the model of an
elastic filament based on their calculation is usually reffferred to as the Kratky-Porod
model or the wormlike chain model or the semiflexible chain model, the nomenclature
really varies with the author. It represents the most fundamental insight into the thermal
fluctuations of an elastic filament. Except for very tight curvatures the Kratky-Porod
model seems to be consistent with the bahavior of DNA in a thermal bath.
6.1.1 The limit of a flexible filament
The Kratky-Porod model has two important limits dependnig on the ratio L/Lp that we
shall analyse in what follows. They are imbedded in the Kratky-Porod model but can be
derived even without invoking it. First of all we have the free-flight chain limit, that can
be derived as D E
2
lim (R(L) R(0)) = 2Lp L, (219)
L/Lp !1
91 O. Kratky, G. Porod (1949), "Röntgenuntersuchung gelöster Fadenmoleküle." Rec. Trav. Chim. Pays-Bas.
68: 1106-1123.
92 Rudolf Podgornik
Figure 37. Statistical shape (exaggerated) of an elastic filament in the Kratky-Porod

model. Short filamnts really look like stiff rods. The longer they are, the more convoluted
they become, eventually becoming completely disordered free-flight chains.
The length 2Lp is usually referred to sa the Kuhn length and describes the independent
unit of the polymer chain. Obviously in this limit the thermal bath completely destroys
the correlations between far away segments along the filament.
This result can be obtainde also from a simplified consideration along the following
lines. Assume the chain is composed of segments of length b directed along the local
tangent t(`), so that the end to end vector is defined as
Z L
R(L) R(0) = t(`) d`.
0
The segments are assumed to be orientationally completely uncorrelated so that
ht(`) · t(`0 )i = b (` `0 ). (220)
This definition of the orientational correlation function gives for the length of the chain
directed along the direction of the beginning of the chain i.e.
* Z !+
L
ht(0) · (R(L) R(0))i = t(0) · t(`0 )d`0 =b
0
exactly the length of the segment b. It follows straightforwardly in this case that the
average size of the chain squared is
D E Z LZ L Z LZ L
2 0 0
(R(L) R(0)) = d`d` ht(`) · t(` )i = b (` `0 ) d`d`0 = b L.
0 0 0 0
(221)
93
Clearly by comparing Eqs. 219 and 221 we can identify the Kuhn length with the length
of the statistically independent unit of the polymer chain, thus b = 2Lp . It equals exactly
twice the persistence length.
6.1.2 The limit of a stiff filament
The other limit of the general Kratky-Porod result can be derived for a very stiff or a very
short chain in the form
D E
2
lim (R(L) R(0)) = L2 , (222)
L/Lp !0
Obviously the thermal bath does not play any role at all in this case. Let us derive this
result again by a different route. For a very stiff chain the only configuration surviving
the thermal average over all configurations Eq. 243 is the one that minimizes the elastic
energy
Z L ✓ ◆2
dt
F = 12 Kc d`. (223)
0 d`
leading to the Euler - Lagrange equation of the form
d2 t
= 0, with solution t(`) = a, (224)
d`2
where a is a unit vector, a · a = 1, specifying the direction of the rod in space and thus
t(`) = a obviously describes a straight line in space. This means that for a very stiff
filament the orientational correlation function becomes
ht(`) · t(`0 )i = 1. (225)
Thus we obtain for the orientational persistence of the chain that

Z L
ht(0) · (R(L) R(0))i = ht(0) · t(`0 )i d`0 = L.
0
The corresponding end to end separation si given by

D E Z LZ L Z LZ L
2
(R(L) R(0)) = d`d`0 ht(`) · t(`0 )i = d`d`0 = L2 . (226)
0 0 0 0
Here we have ignored the thermal average since as we noted for a very stiff chain the only
possible configuration is the one corresponding to the solution of the Euler-Lagrange
equation Eq. 224. The statistical avergae thus reduces to a single value. In this limit the
elastic properties of the chain obviously do not figure any more in its statistical descrip-
tion.
6.2 Light scattering from a Kratky-Porod filament in solution

Let us start from a definition of the total scattering factor of a filament in solution which
is given by ⌦ ↵
hI(Q)i = |F(Q)|2 (227)
94 Rudolf Podgornik
since we have to avaluate the statistical average over all the conformations of the filament
in solution. In the scattering geometry the magnitude of the scattering wave vector Q in
the above formula is given by
✓ ◆2
2⇡ ✓
Q2 = 2 sin ,
2
with the wavelength of light and ✓ the scattering angle. Assuming that the filament is
linear, and thus does not posses any transverze dimension, this leads to
*Z Z +
0
hI(Q)i = ⇢(r) eiQ·r d3 r ⇢(r0 ) e iQ·r d3 r0
(V ) (V )
Z LZ L D E
1 r(`0 ))
= eiQ·(r(`) d`d`0 . (228)
L2 0 0
Here we have assumed that the filament is threadlike and we have conventionally nor-
malized the result by dividing it with L2 . It is very difficult to calculate this quantity
exactly for the Kratky-Porod chain, though there is no shortage of approximate results
that interpolate between two obvious limiting cases given by Eqs. 221 and 226. Let us
evaluate these limits explicitly.
For a free-flight chain the probability distribution of segments with length r(`) r(`0 )
is Gaussian due to the central limiti theorem that follows from statistical independence
of the segments. For a Gaussian distribution one can furthermore derive 92
D E
D 0
E 1
(Q · (r(`) r(` 0
)))
2
eiQ·(r(`) r(` )) = e 2 . (229)
⌦ ↵
We already know that for a free-flight chain (xi (`) xi (`0 ))2 = 13 b|` ⌦ `0 |. This fol-
↵
lows from the fact the the chain is isotropic and that we already derived (r(`) r(`0 ))2 =
b|` `0 |. Therefore we finally remain with
Z LZ L 2
1 b 2 0 2(e (QRg ) 1 + (QRg )2 )
hI(Q)i= 2 e 6 Q |` ` | d`d`0 = = f ((QRg )2 ),
L 0 0 (QRg )4
(230)
where f (x) is the Debye scattering function and the radius of gyration Rg is given by
Rg2 = 16 bL. The Debye scattering function can be often conveniently approximated by
1
hI(Q)i = 1 2 2
1+ 2 Q Rg
with about 15 % accuracy for the whole range of Q values. The other limit is again
obtained by treating the filament as a rigid rod. In this case r(`) r(`0 ) = t(` `0 ). where
t is the constant unit direction tengential vector of the rod. For the rod the statistical
average is translated directly into the integral over all the orientations of the rod with
respect to Q. Thus we remain with
Z LZ L D E
1 iQ·(r(`) r(`0 ))
hI(Q)i = e d`d`0
L2 0 0
92 M. Doi and S.F. Edwards, The Theory of Polymer Dynamics, Oxford (1986), p. 23
95
Figure 38. Exact scattering intensity for a Kratki-Porod chain calculated in A.J. Spakow-
icz and Z-G. Wang, J. Chem.Phys. 37 5814 (2004). The graphs essentially show
Q hI(Q)i for different values of the length N of the cain. For short chains (left) the exact
result is very close to the stiff rod limit (also shown as dotted curve). Fro long chains
(right) the exact result is very close to the free-flight chain limit, exept for asymptotically
long chains, where it again approaces the rigid rod limit.
Z LZ L Z LZ L
1 `0 | cos ✓ 2 sin Q|` `0 |
= eiQ|` d(cos ✓) d`d`0 = d`d`0 .
L2 0 0 L2 0 0 Q|` `0 |
(231)
One can now introduce an auxiliary variable u = ` `0 . The domain o fintegration now
decomposes into a stripe from u to L for the variable ` and from 0 to L for u. We can
this finally derive for the total scattering factor of an orientationally averaged thin rod
Z Z !
L QL
2 sin Qu 2 sin z 1 cos QL
hI(Q)i = 2 (L u) du = dz . (232)
L 0 Qu (QL) 0 z (QL)
The two forms of the scattering function Eqs. 230 and 232 also present two asymptotic
limits for the scattering from an elastic Kratki-Porod filament. For small values of L/Lp
we have the rigid rod result and for large values of L/Lp we have the Debye free-flight
chain result. The limiting values of the two scattering intensities for large Q are ⇡/2 for
the rigid rod and 2 for the free-flight chain.
Exact calculations of the scattering function for the Kratki-Porod model, i.e. its form
between the limits of Eqs. 230 and 232, are extremely difficult to calculate and there
have been various attempts at it. In the Kratki-Porod model the scattering function can
be obtained exactly from
Z LZ L D E Z LZ L
1 r(`0 )) 1
hI(Q)i = eiQ·(r(`) d`d`0 = G (Q; ` `0 ) d`d`0 .
L2 0 0 L2 0 0
(233)
Here G (Q; ` `0 ) is the Fourier transform of the Green function of the Kratki-Porod
96 Rudolf Podgornik
Figure 39. Two limiting shapes - schematically - used to calculate the light scattering off
a flexible polymer chain in solution. A Gaussian, completely disordered, coil and a stiff
rod. They represent two extremes of the Kratky-Porod statistics of a filament in solution.
chain defined as
Z Z ` RL 2
G (r 0
r ;` 0
`)= D(t) 3
(r r 0
t(s)ds) e
1
2 Kc
0
( dt
d` ) d`
. (234)
`0
where the summation is over all the configurations of the chain with t2 (`) = 1 as in Eq.
243. This average is difficult to evaluet explicitly. This feat was nevertheless accom-
plished by a mathematical tour de force 93 that provided the exact result. Still the final
dependence of hI(Q)i on the magnitude of the wave-vector Q can only be evaluated
numerically.
A good analytical approximation that interpolates between the random walk and the
stiff rods limit can be obtained as follows. Starting from the scattering intensity in the
form
Z LZ L
1 0
I(Q) = <eiQ·(r(`) r(` )) > d`d`0 =
L2 0 0
Z LZ L 2
1 1 0
= 2
e 2 <(Q · (r(`) r(` ))) > d`d`0 =
L 0 0
Z LZ L
1 Q2 0 2
= 2 e 6 <(r(`) r(` )) > d`d`0 . (235)
L 0 0
by implementing the Gaussian ansatz Eq. 229 and the fact that the statistical distribu-
tion of the polymer chain is isotropic. The difficult part in the above integration is to
93 A.J. Spakowicz and Z-G. Wang, J. Chem.Phys. 37 (2004) 5814-5823.
97
Figure 40. Experimental points show light scattering intensity as a function of (QL)2
from different DNA samples. Lines are calculated scattering intensities from Eq. 237.
1
The plot of hI(Q)i as a function of (QL)2 .
get the appropriate form of <(r(`) r(`0 ))2 > in the Kratky-Porod model. Note that
here the arguments ` and `0 do not belong to the beginning and the end of the chain, but
to intermediate positions. and so the Kratky-Porod result Eq. 218 can not be used di-
rectly. One can either evaluate it explicitly and remain with a complicated integration, or
one can come up with some suitable approximation and hopefully evaluate the integral
analytically, which is the path we will follow below 94 .
The main approximation that we introduce at this point is
<(r(`) r(`0 ))2 > = 2Lp 2 u 1+e u
, (236)
with x the ratio of the separation between the segments r(`), r(`0 ) along the chain and
its persistence length, i.e. u = |` `0 |/Lp . The essence of this approximation is to
use the Kratky-Porod form of the end-to-end separation of the chain also for the local
segment-to-segment separation among any two segments along the chain. Though this is
not valid exactly, it is certainly plausible. We assume that the statistical deviation from
this assumption, expected for a Kratky - Porod chain, should have little consequences
on his final conclusions. The local approximate relation Eq. 236 should be the more
accurate the larger the ratio L/Lp .
Taking this closed form expression for <(r(`) r(`0 ))2 > the two integrals in Eq.
235 can now be evaluated analytically, yielding the following closed form expansion for
the scattering intensity

u p p2
I(Q) = P (p, x) = e F (p, x) F (p + 1, x) + F (p + 2, x) + . . . (237)
1! 2!
94 A. Peterlin, Nature 171, 259 - 260 (1953). Peterlin, A. , J. chim. Phys., 47, 669 (1950); 48, 13 (1951); J.
Polymer Sci., 8, 173 (1952).

98 Rudolf Podgornik
Here we used the following abbreviations

✓ ◆2
4⇡ ✓ 2
p= 1
3 Lp sin2 and F (p, x) = px 1+e px
,
2 (px)2
where x = L/Lp . The above scattering intensity reduces directly to the Debye result
valid for a Gaussian chain. In fact for the limit x ! 1 we obtain
✓ ◆2
2 4⇡ ✓
lim P (p, x) = 2 w 1 + e w with w = 16 R sin2 ,
x !1 w 2
which is indeed the Debye result with R2 the average square end-to-end distance for a
Gaussian chain.
Since experimentalists usually plot the scattering data in terms of the s.c. Zimm
plots, where one plots not the scattering intensity but rather its inverse as a function
of sin2 ✓2 , we rewrite these results in an alternative form that would be in accord with
this convention. We evaluate the expansion of the inverse scattering intensity, noticing
that the Zimm plot should show a convex curvature close to the origin for a Gaussian
chain and should show a concave curvature for a stiff rod. Our calculation should fall
right somewhere in between these two limits. One can derive the following form for the
inverse scattering intensity
1 p
=1+ [x + 3(F (1, x) 1)] + . . . . (238)
P (p, x) 3 1!
2
This form is obviously linear in p = 13 4⇡ Lp sin2 ✓2 and its L dependent coefficient,
i.e. [x + 3(F (1, x) 1)], should be easily extractable from the Zimm plot. Let us check
the limiting forms of this expression. First of all one can write down the Debye limit of
a Gaussian chain which now has the form of
✓ ◆2
1 w 4⇡ ✓
lim = 1 + + ... again with w = 16 R sin2 . (239)
x !1 P (p, x) 3 2
Here again R2 is simply the mean square end-to-end separation of a Gaussian chain.
Then one can derive also the form of the Zimm plot for a stiff rod. This limit corresponds
to the general case of a length camparable to its persistence length. Formally this limit
is obtained by taking x ! 0 in the general formula Eq. 237 yielding the following
expression
Z ! 1
QL
1 (QL) sin z 1 cos QL y
lim = dz =1 + + . . . , (240)
x !0 P (p, x) 2 0 z (QL) 9
with ✓ ◆2
4⇡ ✓
y= L sin2 ,
2
which is completely in accord with Eq. 232. The Zimm plot for all x of the computed
scattering intensity remaina concave.
99
Historically light scattering of DNA solutions has been used for a first estimate of the
DNA persistence length Lp . Peterlin 95 evaluated the scattering intensity approximately
and compared his calculations with experiments of Dotty et al.. This allowed him to
extract a value for the persistence length of the chain. What is plotted on Fig. 6.2 is the
inverse of the scattering intensity as a function of (QL)2 . The DNA data fall right in
between the free-flight chain and the rod models. Fitting the calculated scattering curve
to the light-scattering experiments yields numerical values for the presistence length.
Various preparations of nucleic acids give the numbers presented in the table Fig. 6.2.
The average persistence length of DNA obtained by this fitting is given by
DNA preparation M ⇥ 10 6
a[nm] L[nm]
Signer 6.7 28.5 4300
Bunce-Geiduschek 4 26 2600
Gülland 4 40 1000
Bunce-Geiduschek I 2.64 40 1000
Bunce-Geiduschek II 2.1 37 1100
Bunce-Geiduschek III 2.7 54 1080
Varin I 7.7 60.6 2100
Table 5. Measured persistence length and length of the DNA in nm for various molecular
weight M DNA preparations (leftmost column) in light scattering experiments.
Lp = 40.6 (1 ± 0.28)[nm] or approximately 120 base pairs, (241)
where the contour length of the base pair is taken standardly as 0.34 nm. This makes
DNA a moderately stiff molecule. The fitted value is indeed very close, but somewhat
smaller, then the modern accepted value 96 of 46 50 nm or 140 150 base pairs, thus
very close to the length of the nucleosomal DNA fragment. However, as it has been
realized for a while, it depends crucially on the ionic solution conditions and can vary
significantly with these conditions. The most accurate values for DNA persistence length
are obtained from atomic force spectroscopy (AFM) 97 which is the modern method of
choice for measuring elastic properties of single macromolecules that we consider next.
6.3 Elastic response of a Kratky-Porod filament

We have thus far analysed the bahavior of an unconstrained semiflexible polymer chain.
Let us now analyze also what happens to this chain when it is under the influence of an
external tension applied to both ends of the chain. In this case one can easily generalize
95 A. Peterlin, Nature 171, 259 - 260 (1953). Peterlin, A. , J. chim. Phys., 47, 669 (1950); 48, 13 (1951); J.
Polymer Sci., 8, 173 (1952).

96 P. J. Hagerman, Flexibility of DNA, Annu. Rev. Biophys. Biophys. Chem. 17 265-286 (1988).
97 M. C. Williams and Ioulia Rouzina, Force spectroscopy of single DNA and RNA molecules, Current
Opinion in Structural Biology 12 330-336 (2002).

100 Rudolf Podgornik
the Hamiltonian of the chain Eq. 208 to iclude the action of the external force f as
Z L ✓ ◆2 Z L ✓ ◆2 Z L
1 dt 1 dt
F = 2 Kc d` + f · (R(L) R(0)) = 2 Kc d` + f · t d`.
0 d` 0 d` 0
(242)
The external force thus acts as a linear source term in the hamiltonian F . From the
definition of the partition function
Z 1 R L dt 2 R L
Z(f ) = D[t(`)] e 2 c 0 ( d` )
K d`+ f ·t d`
0 with t2 (`) = 1, (243)
we obtain immediately that

*Z +
L
@F(f ) @ log Z(f )
= kB T = t(`) d` = h(R(L) R(0))i . (244)
@f @f 0
The free energy and the partition function are connceted via F(f ) = kB T log Z(f ). In
general the partition function entering the above elastic equation of state is very difficult
to evaluate in closed form. This is due principally to the non-linear constraint that the
tangent vector has to be a unit vector. There are nevertheless two important limiting cases
that allow for a simple analytical form of the equation of state.
6.3.1 The limit of small stretching force
We already know that the Kratky-Porod model has two important limits dependnig on
the ratio L/Lp that we shall analyze also in the context of elastic stretching of the chain.
Assume that the ratio L/Lp 1. If the external force is small, the local segments of the
chain will be completely decorrelated, see Eq. 220. In this case the approximate form of
the partition function Z Eq. 243 will be
Z RL
f ·t d`
Z⇠ = D[t(`)] e 0 with t2 (`) = 1. (245)
The explicit form of this approximate partition function is obtained by discretizing the
contour of the chain in N segments of length a, so that the total length is L = N a and
realizing that the functional measurefor each of the segments i is sin ✓i d✓i . In this case
one can derive 98
R⇡
@ log Z ⇠ N a 0 cos ✓i e f a cos ✓i sin ✓i d✓i
= R⇡ = h(z(L) z(0))i , (246)
@f 0
e f a cos ✓i sin ✓i d✓i
according to Eq. 244, where in addition we have assumed that the axis z is now orineted
in the directin of the external force, i.e. f = (0, 0, f ). Invoking now the definition of the
Langevin function L(x) = coth(x) 1/x and remembering that the derivation is valid
in the limit of small external forces, we thus finally remain with
f 2LP
h(R(L) R(0))i · = h(z(L) z(0))i = L L( f a) ⇠
=L f. (247)
|f | 3 kB T
98 P.J. Flory, Principles of polymer chemistry, Cornell University Press, 1953, p.428.
101
where we invoked Eq. 219 in order to define the segment a with the Kuhn’s length that
equals twice the persistence length. The above result i.e. the linear relationship between
the force and extension where the elastiuc coefficient is inversely proportional to the
temperature, is usually referred to as the polymer entropic elasticity.
6.3.2 The limit of large stretching force
In the opposite limit of large external forces, the elastic equation of state is more difficult
to derive. In this case the chain is almost fully extended in the direction of the external
force so that to the lowest order the unit tangent vector t = f /|f |. The elastic Hamiltonian
of the chain can be written as
Z L ✓ ◆2 Z L Z L ✓ ◆2 Z L
1 dt 1 dt
F = 2 kB T L P d` f · t d` = 2 kB T LP d` f tk d`,
0 d` 0 0 d` 0
(248)
where tk is the magnitude of the tangent vector in the direction of the external force. The
corresponding partition function is of the form
Z 1 R L dt(`) 2 RL
L d`+ f tk (`) d`
Z = D[t(`)] e 2 P 0 d` 0 with t2 (`) = 1. (249)
If we simply put the ansatz tk = f /|f | into the above expression we end up with
h(z(L) z(0))i = L, i.e. the chain is indeed fully extended in itself not a particularly
interesting result. It would be interesting if we could evaluate the subdominant order in
the expansion of the elastic equation of state for large external forces, which has to be
inverse in the external force. Let us derive it following Marko and Siggia 99 .
Assume first that the tangent vector has a component tk in the direction of f and a
component perpendicular to it t? . The constraint on the magnitude of the tangent vector
now reads
q
t2 (`) = t2k (`) + t2? (`) = 1 and thus tk (`) = 1 t2? (`) ⇠ = 1 12 t2? (`),
where we assumed that the deviations from the completely extended chain configuration
are small, allowing us to stop at the lowest order expansion of the square root in the above
equation. The elastic Hamiltonian thus assumes the form
Z L ✓ ◆2 Z L
dt
1
F = 2 kB T L P d` + f tk d` ⇠
=
0 d` 0
Z L✓ ◆2 Z L !
⇠ dt ? f 2
= 1
2 kB T L P d` + t d` f L, (250)
0 d` LP 0 ?
where obvioulsy to the lowest order dt

2
⇠ dt? 2 . The last term in the elastic Hamil-
=
d` d`
tonian obviously corresponds to the case tk = f /|f |, and the rest are deviations from this
case. Let us now apply the Fourier transform to t? (`) obtaining
Z L
1
t? (Q) = eiQ` t? (`)d`,
2⇡ 0
99 J.F. Marko and E.D. Siggia, Macromolecules, 28, 8759 (1995)
which in the limit of an infinitely long chain becomes simply a regular Fourier integral.
The part of the elastic Hamiltonian that depends on t? can be rewritten in the space of
Fourier components t? (Q) in the form
Z ✓ ◆
dQ f
F = 12 kB T LP Q2 + |t? (Q)|2 . (251)
2⇡ LP
Noticing that the energy is now a quadratic function of the Fourier components of the
transverse part of the tangent vector, we can straightforwardly apply the equipartition
theorem and obtain
Z Z +1
⌦ 2↵ dQ ⌦ ↵ dQ 1 1
t? = |t? (Q)|2 = 2 ⇣ ⌘=p . (252)
2⇡ 1 2⇡ L Q2 + f LP f
P LP
In the above integral the factor of 2 counts the two components of the vector t? . The
elastic equation of state for the strongly extended chain can then be obtained from Eq.
246 and Eq. 250 as
Z L
@kB T log Z @F 1
⌦ 2↵ L
hz(L) z(0)i = = =L 2 t? d` = L p .
@f @f 0 4 LP f
(253)
We note here that the first order correction to the completely stretched chain result goes
as an inverse power of the stretching force and reduces to it in the limit of an infinitely
large force.
6.4 Extensible semiflexible elastic filament

Up to now we have only considered a chain that is flexible but is not really extensible.
This condition was coded in the constraint that t(`)·t(`) = 1 at any point along the chain.
For small external tractions this constraint is of course always valid. For large tractions
it need not be so, however, as amply demonstrated by experiments on DNA 100 .
In the case of a stretched chain the arclength is no longer a good parameter to describe
the state of the stretched chain. Let us assume that in the unstretched configuration the
arclength is given by ` = n`0 , where `0 is the unstretched bond length, n is the bond
index and the total number of the bonds along the chain is N0 101 .
RN
The total length of the stretched chain is given by L = 0 0 (1 + ✏(n))dn where for
stretched bonds we now introduce the strain along the bond vector t(`) as
b(`) b0
✏(`) =
b0
so that the unstretched and the stretched bonds, r(`) and r0 (`) respectively, are at posi-
tions given by
Z ` Z `
r(`) = r(0) + t(`)d` and r0 (`) = r0 (0) + (1 + ✏(`))t(`)d`. (254)
0 0
100 J. F. Marko and E. D. Siggia, Stretching DNA, Macromolecules 28 8759 (1995).
101 J. Kierfeld, O. Niamploy, V. Sa-yakanit, and R. Lipowsky, Stretching of semiflexible polymers with elastic
bonds, Eur. Phys. J. E 14 17 (2004).

103
Just as in the unstretched chain here too the tangent vector is a unit vector, |t(`)| = 1.
Furthermore we assume that the bonds are harmonic elastic springs that can be charac-
terized by the appropriate bond axial stretching (Lamé) elastic modulus of the molecule
, which in principle can depend on the bond index n or equivalently on the unstretched
arclengt ` to model longitudinal heterogeneity of the chain. The elastic energy of the
bonds is then
Z L
Fs [✏(`)] = d` ✏(`)2 . (255)
2 0
The other component of the total elastic energy of a semiflexible stretchable chain is the
bending energy. It is proportional to the square of the local curvature. For the n th
bond the curvature is proportional to t(n + 1) t(n) and thus the bending energy os
proportional to t(n + 1) · t(n) for each of the bonds. In the continuum representation the
total bending energy then amounts to
Z L ✓ ◆2
Kc dt(`)
Fb [t(`)] = d` . (256)
2 0 d`
. Since the bending energy depends only on the relative angle between the two neigh-
boring tangent vectors, it does not depend on the deformation of the bond ✏(`). The last
conponent of the total deformation free energy is the energy due to external traction f on
the chain. Taking into account Eq. 254 it can be written as
Z `
Ft [✏(`), t(`), f ] = (1 + ✏(`))f · t(`)d`. (257)
0
. The total deformation energy is then given by
F [✏(`), t(`), f ] = Fb [t(`)] + Fs [✏(`)] + Ft [✏(`), t(`), f (`)] =

Z ✓ ◆2 Z L Z L
Kc L dt(`) 2
= d` + d` ✏(`) (1 + ✏(`))f · t(`)d`.
2 0 d` 2 0 0
(258)
The relative extension of the chain due to the external traction, Lf , can be obtained as
the average of the end-end separation projected onto the direction of the traction
⌧
f @F (f ) @ log Z(f )
(r(L) r(0)) · = h(z(L) z(0))i = = kB T , (259)
f @f @f
where F (f ) is the free energy of the chain at constant external traction, assumed to
act in the direction of the z axis. The average in this equation is defined as a double
average: over the spatial configurations of the chain as well as over all the local stretching
deformations of the chain. Formally this means that measure in the partition function is
D[t(`)] (|t(`)| 1)⇥D[✏(`)], with explicitly taking into account the fact that the tangent
vector is a unit vector. The partition function is thus
Z
Z(f ) = D[t(`)]D[✏(`)] e F [✏(`),t(`),f ] with t2 (`) = 1. (260)
The partition function for the elastic Hamiltonian Eq. 258 can not be evaluated ana-
lytically 102 but several approximations can be employed to advantage 103 . From the
definition Eq. 261 we derive
Z L0 Z L0
hz(L) z(0)i = L0 hf · t(`)i d` h✏(`) f · t(`)i d`, (261)
0 0
We now again express the tangent vector in terms of its fluctuating perpendicular
component, t? in exactly the same way as in Eq. 252. The lowest order term in ✏(`)
is uncoupled from t? and alows for an explicit evaluation 104 . Thus one finally remains
with s
hz(L) z(0)i 1 kB T f
=1 + , (262)
L0 2 LP f
where we took into account that the banding modulus Kc and the persistence length LP
are connected as Kc = kB T LP . The axial stretching thus simply adds a term linear
in external traction to the elastic equation of state. This linear Hookian term is due to
the enthalpic stretching elasticity and depends inversely on the stretching (Lamé) elastic
modulus of the molecule. The validity of this expression is unfortunatley quite restricted
105
, however it does not fare poorly when compared with DNA stretching experiments.
6.5 Approximate elastic filament equation of state

The equation of state for the semiflexible chain under the influence of an external force
can thus be written explicitly and analytically in the limit of small and large force as
follows 8
> 2 LP f ; LP f
kB T ⌧ 1
hz(L) z(0)i < 3 kB T
= 1 q 1 LP f
; kB T 1. (263)
L >
: L f
4 P kB T
The general form of the elastic equation of state for any value of force and especially
right in between the two limits is in general difficult to calculate. Explicit approximate
but nevertheless quite accurate elastic equations of state were nevertheless obtained by
Marko and Siggia 106 and Blundell and Terentjev 107 . In the first case a simple interpola-
tion formula that covers the whole regime of external stretching forces was proposed in
the form
LP f z 1
= + 1
4. (264)
kB T L 4 1 ( Lz )2
102 However, one notes that the Hamiltonian Eq. 258 is a harmonic function of the local strain function ✏(`)
and can be thus integrated out explicitly. This leads to an effective elastic free energy functional of the form
RL dt(`) 2
RL RL (f ·t(`))2
Fef f [t(`), f ] = K2c d` d`
f · t(`)d` 12 d` .
0 0 0
103 Kierfeld et al. ibidem.
104 T. Odijk, Stiff Chains and Filaments under Tension, Macromolecules 28 7016 (1995).
105 As shown by Kierfeld et al. ibidem. In fact it is valid only for f Kc /LP , Kc /L20 .
106 J.F. Marko and E.D. Siggia, Op.cit.
107 J. R. Blundell and E.M. Terentjev, Buckling of semiflexible filaments under compression, Soft Matter
(2009).
105
Figure 41. Traction as a function of relative deformation per base pair in 1000 mM
(fullbullet), 100 mM (circle), and 10 mM (X) [Na] buffer at pH 7.5. Optical tweezers
y, January 27, 12
experiment. The solid lines represent the extensible WLC (left) and FJC (right) from
Wenner et al. (2002). Right: a schematic representation of the experiment, adapted
from: G. Bao and S. Suresh, Nature Materials 2, 715 - 725 (2003).
where z = hz(L) z(0)i. This is the s.c. Marko-Siggia elastic equation of state that
is applicable for any semiflexible polymer under external traction. It has been tested
repeatedly and found to describe the single-molecule stretching quite accurately. In the
second approach a very accurate analytic approximation was derived directly for the
partition function of the semiflexible chain under external traction. The ensuing Blundell-
Terentjev elastic equation of state can then be obtained in the form
✓ ◆2
LP f 2 LP z 4 z
=⇡ . (265)
kB T L L ⇡ 1 ( z )2 2 L
L
Fitting these two equations to the DNA stretching data gives equally accurate versions of
the persistence length of DNA. The Marko-Siggia interpolation formula provides a better
model fro DNA stretching at smaller extensions (in the flexible regime). The Blundell-
Terentjev elastic equation of state is based on the mean-field approximation, which works
worse at lower extensions.
The stretched chain elastic equation of state can be dealt with in the same way. Ob-
viously, from Eq. 262, by introducing Lz f as a new variable in the denominator of the
second term, the only term that gets modified for large external tractions, we can derive
a slightly modified form of the Marko-Siggia interpolation formula as
LP f z 1
= + ⇣ ⌘ 1
4. (266)
kB T L 4 1 ( Lz f 2
)
Experimentally the transition from the small stretching force to large stretching force and
into the regime of enthalpic stretching has been clearly demonstrated in experiments on
DNA 108
6.6 Stretching DNA

In a stretching experiment a single double stranded -phage DNA molecule of contour
length of 16.3 µm is first biotinylated at both 5’ ends. Biotin is a chemical that sticks to
specific receptor molecules such as streptavidin. This labelled DNA is then brought in
contact with two streptavidin-covered polystyrene beads to which it sticks with biotiny-
lated ends. One bead is then held fixed by the glass micropipette and the second one is
captured with optical tweezers. The stretching force on the DNA molecule can be then
measured by the deviation of the bead from the center position in the laser beam. In this
way it is possible to study the force-versus-extension behavior of a single DNA molecule
wherefrom elasticity and structure parameters can be deduced accurately 109 .
Stretching experiments on DNA confirmed that the B-DNA form is a classical eule-
rian elastic rod whose force-extension behavior can be well described using the worm-
like chain model and enthalpic stretching as in Eq. 265. In a typical stretching experiment
DNA shows an elastic response at forces below ' 5 pN dominated by entropic effects,
first term on the r.h.s. of Eq. 265, whereas at forces above ' 5 pN enthalpic contributions
start to play an important role, denominator of the second term on the r.h.s. of the same
equation.
The experimentally determined values of the persistence length and the stretching
modulus are then LP = 46 nm and = 1256 pN 110 . These values describe the elastic
properties of DNA in the case of high salt, i.e. when the long range interactions along
the DNA backbone are completely screened out. If this is not the case, it is necessary to
consider the effect of long range interactions on the mesoscopic parameters LP and as
will be done later. The persistence length also allows us to calculate the bending rigidity
Kc = kB T LP , wherefrom by applying classical elasticity theory the Young’s modulus
E ' 1 nN follows. For a classical elastic rod one has E = 4Kc /(⇡a4 ), where a is
the DNA radius. This gives an estimate of E ' 300M P a, comparable to the Young’s
modulus of plexiglass.
6.7 Rod-like chain model of DNA

In what follows we will now try to introduce thermal fluctuations into the elastic confor-
mations of an elastic rod with bending and twisting fluctuations. This will then lead us to
the rod-like chain model of DNA as formulated by Bouchiat and Mézard 111 and Moroz
and Nelson 112 .
The worm-like chain model starts with an elastic Hamiltonian that includes bending
108 Jay R. Wenner, Mark C. Williams, Ioulia Rouzina, and Victor A. Bloomfield, Biophysical Journal Volume
82 June 2002.
109 Smith SB, Cui Y, Bustamante C., Overstretching B-DNA: the elastic response of individual double-
stranded and single-stranded DNA molecules. Science (1996) 271 795.

110 Wenner et al., op. cit.
111 C. Bouchiat and M. Mézard, Elastic rod model of a supercoiled DNA molecule, EPJE 2 377 (2000).
112 D. Moroz and P. Nelson, Entropic Elasticity of Twist-Storing Polymers, Macromolecules, 3 6333 (1998).
107
rigidity but does not take into account the torsion. Therefore it does not follow directly
from the elastic theory of an Euler-Kirchhoffian filament. We will now try to go past this
approximation and start with a complete Hamiltonian of a bendable twistable rod under
external traction.
6.7.1 Partition function and its quantum analogy
The point of departure is thus the mesoscopic elastic (free) energy Eq. 163 parameterized
with Euler angles
Z ✓ ⌘◆
˙ ˙, ˙] =
L
1 ⇣˙ ⌘2 ⇣
F [✓, ✓, d` C cos ✓ + ˙ + 12 Kc ✓˙2 + ˙ 2 sin2 ✓ , (267)
0 2
to this we add the contribution of external force F and external torque ⌧ as

Z L
F (z(L) z(0)) + 2⇡ Lk ⌧ = F d` cos ✓ + 2⇡ Lk ⌧, (268)
0
assuming that the force acts in the ẑ direction. The total linking number of the molecule is
then given by the Călugăreanu-White-Fuller theorem in Euler parameterization as written
down in Eq. 62 113 , i.e.
Z L
1
Lk[ ˙ , ˙ ] = T w + W r = ( ˙ + ˙ )d`, (269)
2⇡ 0
where the local twist and writhe have been defined in the form
Z L
1
Wr = (1 cos ✓(s)) ˙ (s)ds. (270)
2⇡ 0
and Z
1
Tw = (cos ✓(s) ˙ (s) + ˙ (s))ds. (271)
2⇡ C
Since the Euler parameterization is singular at the two poles, ✓ = 0, ⇡, where the z axis,
taken to point in the direction of external traction, pierces the unit sphere. The regions
around the two poles have thus to be pierced from the domain of definition which ensures
that the local and non-local writhe formulas give the same result. This means that the
Euler parametrization can only be valid for a highly stretched molecule, while in general
one should use the more complicated non-local formulation of the Călugăreanu-White-
Fuller theorem.
Inserting the unity in the form of an integral fo a delta function, the partition function
of the rod-like chain model can be transformed to
Z Z Z L !
˙ ˙ ˙
Z(⌧, L) = D[✓, , ] dLk 2⇡Lk + ( ˙ + ˙ )d` e F [✓,✓, , ] . (272)
0
Writing the Dirac delta function in the integral representation 114 we remain with
Z Z Z
˙ ˙ ˙
Z(⌧, L) = dLk dk D[✓, , ] e G[k,✓,✓, , ]+2⇡ Lk (⌧ ık) , (273)
where
Z L
˙ ˙ , ˙ ] = F [✓, ✓,
G[k, ✓, ✓, ˙ ˙, ˙] ık ( ˙ + ˙ )d` (274)
0
Obviously this expression is related to the constrained elastic free energy Eq. 165 that
we were analysing in the case of no thermal fluctuations. In this case k was constant
and equal to the Lagrange multiplier setting the value of the linking number of the chain,
which in its turn was nothing but the torque needed to fix the linking number of the
filament.
The above form of the partition function for the rod-like chain model can be further
decomposed by taking into account the Călugăreanu-White-Fuller decomposition of the
linking number, Eq. 269, as well as the form of the elastic energy of the rod. In fact one
can write Z Z
Z(⌧, L) = dLk dk Z̃(k, L) e2⇡ Lk (⌧ ık)
, (275)
where Z̃(k, L) can be decomposed into
Z R L ⇣1 ⌘
˙2 ˙ 2 2
0
d` 2 Kc (✓ + sin ✓ ) F cos ✓ +ık 2⇡W r
Z̃(k, L) = D[✓, ]e ⇥
✓ ⇣ ⌘2 ◆
RL ˙ cos ✓ + ˙
Z 1
0
d` 2C +ık 2⇡T w
⇥ D[ ] e (276)
Since twist is given by the integral Eq. 271, the second functional integral being of a
general Gaussian form can be readily evaluated yielding the following identity
Z R L ⇣1 ⌘
kB T 2
d` 2 Kc (✓˙ 2 + ˙ 2 sin2 ✓ ) F cos ✓ ık (1 cos ✓) ˙ 2C k L
0
Z̃(k, L) = D[✓, ]e ,
(277)
where we substituted the definition for local writhe, Eq. 270, and we also omitted a trivial
multiplicative constant.
This expression has a very illuminating interpretation if one analytically continues
the ` integral along the imaginary axis as first ascertained by Bouchiat and Mezard 115 .
After this Wick rotation one recognizes the action integral of a particle with unit charge
moving on the unit sphere under the joint action of the electric field proportional to
external traction F and the magnetic potential of a monopole k (1 cos ✓) with charge
k. Indeed the Wick rotated partition function Eq. 277 correponds to a particle on a unit
sphere interacting with an electric field in the z direction and a point magnetic monopole
at the origin.
114
R
That is: (x) = dk e ıkx .
115 C.
Bouchiat and M. Mézard, Elasticity Model of a Supercoiled DNA Molecule, PRL 80 1556 (1998).
109
6.7.2 Evaluation of the partition function
We note first that the functional integral over is again Gaussian and can be evaluated,
so that up to an irrelevant constant and a redefinition of the integration measure, the final
partition function for the Euler-Kirchhoff filament is then obtained from Eq. 275 as a
function of the external torque and external force
Z Z Z RL
˙
d`L(✓,✓,k)+2⇡ Lk (⌧ ık)
Z(⌧, F ) = dLk dk ⇥ D[✓]e 0 . (278)
We notice at this point that both integral in Lk and k can be trivially evaluated leading
finally to Z RL
˙
d`L(✓,✓, ı⌧ )
Z(⌧, F ) = D[✓]e 0 , (279)
where
˙ ✓, ı⌧ ) = ⌧ 2 (1 cos ✓)2 ⌧2
L(✓, 1
Kc ✓˙2 F cos ✓ 1 1
. (280)
2 2 Kc sin2 ✓ 2 C
Once the above integral is evaluated the elastic equation of state for the chain can be
obtained by calculating
@ log Z(⌧, F )
<z(L)> = kB T (281)
@F
and
@ log Z(⌧, F )
2⇡<Lk> = kB T . (282)
@⌧
This is the complete formulation of the model. The validity of this model is tied to the
assumptions of the elastic filament theory as well as the limits of applicability of the
Euler parametrization of the Călugăreanu-White-Fuller theorem.
In order to insure the applicability of Eq. 269 we delimit the discussion to large forces
that prevent the molecules to have any antipodal points. In that case we can look at the
perturbation expansion around the straight rod case defined by ✓ = 0. In that case to the
second order in ✓ we get
2
✓ ◆
˙ 1 ⌧ ˙ 2 ⌧2
L(✓, ✓, ı⌧ ) ' F 2 1
+ 2 Kc ✓ + 2 1
F ✓2 + O(✓4 ). (283)
C 4 Kc
This furthermore implies that at the same level of approximation Eq. 279 becomes
Z R L ⇣1 1 2
⌘
1 ⌧2L d` 2 Kc ✓˙ 2 + 2 F 4 ⌧Kc ✓ 2
Z(⌧, F ) ' e F L+ 2 C D[✓]e 0 . (284)
The functional integral remaining corresponds to the partition function of a harmonic

oscillator which can be evaluated analytically so that up to an irrelevant constant and for
a sufficiently long chain
2
✓ 2
◆
1⌧ L Kc ! 1 ⌧
log Z(⌧, F ) = F L + 2 1
+ 2 log ' F+2 1
2 ! L, (285)
C sinh !L C
where we defined
F ⌧2
! 2 (⌧, F ) = .
Kc 4( Kc )2
From the equations of state for the semi flexible chain with torsion, Eqs. 281, 282, we
then deduce that
<z(L)> kB T
= 1
L 4Kc !(⌧, F )
✓ ◆
<Lk> ⌧ 1 kB T
= + . (286)
L 2⇡ 2 C 8Kc2 !(⌧, F )
This is the lowest order result and more accurate expressions in terms of a systematic
expansion in ! 1 can be found in the literature 116 . From the above equations of state
we can derive the dependence of the extension of the chain as a function of the imposed
twist at a fixed external traction. In fact it follows from Eq. 286 that up to and including
second order
✓ ◆
<z(L)> kB T kB T
= 1 p p ⌧ 2. (287)
L 4 Kc F 32 Kc 2 F Kc F
For zero twist this equation reduces to the worm-like chain model result, as it should.
Twist dependence can furthermore be converted to the linking number dependence by
noting that
✓ ◆ 1
<Lk> C
⌧ = 2⇡ 2 C̃ with C̃ = C 1 + p . (288)
L 8 Kc K c F
Here C̃ is the effective twist rigidity. This result tells us that the effective twist rigidity
is softened for small applied tensions due to the thermal configurational fluctuations and
their geometric coupling implicit in White’s formula for the rod-like chain model. The
rang eof validity of this softening effect of thermal fluctuations is of course the same as
the range of validity of the local writhe formula used in its derivation.
6.8 Stretching and twisting DNA

The results of the rod-like chain model can now be compared directly to experiments
with DNA where one can now exert force and torque at the same time on a torsionally
constrained DNA molecule for which both strands are now immobilized at one end of the
molecule. Magnetic tweezers allow to rotate a magnetized bead attached to the end of a
DNA molecule with both strands and to a glass coverslip through its other end. Under the
application of constant force the extension of the DNA molecule changes as the bead is
rotated clockwise or counterclockwise, and the twist increases or decreases. Under twist
various structural transitions are observed that I will not describe here. At low forces the
DNA forms plectonemic supercoils if overtwisted and denatured bubbles if undertwisted.
All these behaviors have been extensively described 117 .
116 J.D. Moroz and P. Nelson, Entropic Elasticity of Twist-Storing Polymers, Macromolecules 1998, 31, 6333-
6347.
117 F. Ritort, Single-molecule experiments in biological physics: methods and applications, Journal of Physics:
Condensed Matter (2006) 18 R531.

111
Figure 42. Relative extension of a DNA molecule vs its degree of supercoiling for
various stretching forces F . For small degree of supercoiling the quadratic regime is
clearly seen. Right: a schematic representation of the experiment, adapted from: G. Bao
Friday, January 27, 12
and S. Suresh, Nature Materials 2, 715 - 725 (2003).
As the torque is proportional to the changes in the linking number the above expres-
sion means that the extension of the chain at a fixed external force diminishes quadrati-
cally with the changes in the linking number. This is in fact what is observed in experi-
ments on DNA stretching and twisting 118 . For small values of over twist the dependence
of the extension is quadratic and for larger values reverts to a linear form.
The experimental data are presented on Fig. 42 that shows the relative extension
of DNA as a function of the degree of supercoiling, Eq. 56, and supercoiling itself is
proportional to the twist of the molecule. To measure and control twist and stretch of
a DNA molecule one can use a very effective magnetic trapping technique. It consists
ï£ijof a single DNA molecule bound at one end to a surface and at the other to a magnetic
micro-bead a few microns in diameter. Small magnets are then used to pull on and
rotate the micro-bead and thus stretch and twist the molecule. This system allows forces
ranging from a few femtonewtons to almost 100 pN with a relative accuracy of about
10%.
118 T. R. Strick, M.-N. Dessinges, G. Charvin, N. H. Dekker, J.-F. Allemand, D. Bensimon and V. Croquette,
Stretching of macromolecules and proteins, Rep. Prog. Phys. 66 (2003) 1-45.

7 ELECTROSTATICS OF DNA
At ordinary pH all the P O4 groups along the phosphate backbone are dissociated giving
DNA a large net negative charge. In the B conformation there is an elementary charge
every 1.47 Å along the contour of DNA, or two per each base pair every 3.4 Å. This
makes DNA one of the most charged molecules in Nature and also does not allow us to
simply forget or gloss over these large and fundamental interactions in DNA.
Electrostatic interactions between charged macromolecules are one of the two pil-
lars of the DLVO theory of colloid stability 119 . The other one being the Lifshitz - van
der Waals electromagnetic fluctuation forces. In DLVO theory both contributions are
assumed to be additive, though in general the van der Waals forces and the electrostatic
interactions are coupled via the s.c. zero-order Lifshitz term.
7.1 Weak coupling (Poisson-Boltzmann) theory

In our investigation of the electrostatics of DNA we first note that the large fixed charged
along the DNA generates a large electrostatic potential. But electrostatic potential is
also generated by the mobile counterions and salt ions in the aqueous solution sorround-
ing DNA. Instead of dealing with DNA directly let us first consider a simpler system,
composed of two planar charged surfaces with mobile charges confined to the region
in between. We will first deal separately with the electrostatic energy and then ewith
the entropy of the system, finally combining the two into the s.c. Poisson - Boltzmann
theory.
7.1.1 Electrostatic energy
Let us assume that ni is the density of the charged species of type i, i denoting either
a DNA counterion or salt ions, of charge zi ei . In the space between the two charged
surfaces there is a mean electrostatic potential that we denote by (r) and by E(r) =
r (r) the corresponding electrostatic field. In the bulk the mean potential is zero by
symmetry. To this potential there corresponds an electrostatic field energy which is given
as a volume integral of the electrostatic energy density density w(E(r), ni (r)) in the
form
Z Z !
X
3 2
W = w(E(r), ⇢i (r)) d r = 1
2 ✏✏0 r (r) + zi ei ni (r) (r) d3 r,
(V ) (V ) i
(289)
where ✏ is the static dielectric constant of the aqueous solution. In fact, since there is no
mean electrostatic potential in the bulk, this could also be viewed upon as the difference
between the electrostatic energy
P in the spaceP between the surfaces and the bulk. The
total charge density is thus i ⇢i (r) = i zi ei ni (r). By minimizing this electrostatic
free energy first with respect to the electrostatic potential one first obtains the Euler -
119 DLVO stands for Derajguin-Landau-Verwey-Overbeek. The classical exposition o fthe stability of colloids
is given in: E.G. Verwey and J.T.G. Overbeek, The Theory of the Stability of Lyophilic Colloids (Elsevier,
Amsterdam, 1948)
113
Figure 43. Distribution of counterions between two charged plates governed solely by
electrostatic energy. The counterions are adsorbed onto oppositely charged surfaces.
Lagrange equation in the form

✓ ◆ X
@w @w
r + =0 wherefrom ✏✏0 rE(r) = zi ei ni (r). (290)
@r @ i
This is obviously nothing but the Poisson equation. The energy density (w), Eq. 289,
pertains only to electrostatic interactions. If we now insert the Poisson equation Eq. 290
back into the energy Eq. 289, we obtain the equilibrium electrostatic energy, proportional
to the square of the electrostatic field, if we disregard the boundary effects. For mobile
charges between two charged interfaces the minimum of electrostatic energy is when all
the counterions are adsorbed onto the surfaces, and salt ions are paired.
7.1.2 Configurational entropy
At a finite temperature we must also consider the entropy of a gas of mobile charged
particles i.e. counterions and salt ions, assumed to be describable by the ideal entropy
density of the mobile charge carriers (s). Let us assume furthermore that there are N
mobile ionic species with charges zi ei , whose densities are ni , while n0i is the particle
density of the same mobile charged species in the bulk with which the system is in chem-
ical equilibrium. Then the entropy difference S between the volume under consideration
and the bulk is given by the integral of the entropy volume density s(ni (r)) as
Z XZ ✓ ✓
ni (r)
◆ ◆
3 0
S= s(ni (r)) d r = kB ni (r) log 0 ni (r) ni d3 r,
(V ) i (V ) n i
(291)
where kB is the Boltzmann constant. The Euler - Lagrange equation for the above en-
tropy is given simply as
✓ ◆
@s ni (r)
=0 wherefrom kB T log = 0. (292)
@ni n0i
The minimum of configuraional entropy alone is thus given by a uniform distribution

of charged species, ni (r) = n0i , in the space that is available to them. Thus while the
electrostatic energy wants the counterions to be localized at the charged surfaces, the
entropy wants them uniformly distributed.
Figure 44. Uniform distribution of counterions between two charged plates governed
solely by entropy.
7.1.3 Poisson - Boltzmann equation
At any finite temperature the equilibrium configuration of this system is given by mini-
mizing the free energy F of the system which is given by the standard thermodynamic
definition F = W T S. The volume density of the free energy (f ) is then defined as a
volume integral
Z Z
F = d3 r f (E(r), ni (r)) = d3 r (w(E(r), ni (r)) T s(ni (r))). (293)
Now that we have the complete free energy correposnding to all the charged species in
the system, we can determine the equilibrium configuration of the mobile charges. This
is obtained by minimizing the free energy Eq. 293 with respect to the mean electrostatci
115
potential as well as the ionic densities ni . There are thus two Euler - Lagrange equations.
The first one is of the form
✓ ◆ X
@f @f
r + =0 wherefrom ✏✏0 rE(r) = zi ei ni (r). (294)
@r @ i
which is again nothing but the Poisson equation, while the second one is given by
✓ ◆
@f ni (r)
=0 wherefrom zi ei (r) + kB T log = 0, (295)
@ni n0i
which is nothing but the Boltzmann distribution for the i th species, set by the value
of the corresponding potential energy, i.e. the electrostatic interaction energy with the
mean field. The above equation
⇣ can
⌘ also be interpreted as stating that the electrochemical
kB T ni (r)
potential (r) + zi ei log n0 for every mobile ion species is a constant (zero) across
i
system.
Taking into account the Poisson equation Eq. 294 and the Boltzmann distribution
Eq. 295, one can put the two together obtaining the Poisson - Boltzmann equation, that
governs the equilibrium spatial distribution of the mean electrostatic potential
1 X e0 X
r2 (r) = rE(r) = zi ei ni (r) = zi n0i e zi e0 (r)
, (296)
✏✏0 i ✏✏0 i
If one now integrates the Poisson equation over the whole volume available to the
mobile ionic species and takes into account the theorem of Gauss- Ostrogradsky, one
gets the condition fo r the overall electroneutrality of the system
I Z X I
✏✏0 (E · n) d2 r = zi ei ni (r)d3 r = d2 r. (297)
@V V i @V
Thus the total volume charge of the mobile charged species should be matched by the
neutralizing surface charge of surface charge density on the surface of the DNA, pro-
vided by the dissociated phosphate groups. The solutions of the Poisson - Boltzmann
equation present a compromise between the demands of the electrostatic energy, com-
plete localization of counterions at the charged surfaces, and the demands of the entropy,
a uniform profile of charges between the surfaces. They always yield a non-uniform
profile of the mobila charge density in the space between charged bodies.
7.1.4 Linearized Poisson - Boltzmann equation
If the magnitude of the surface charges is not too large, one has e0 (r) ⌧ 1 in the
whole accessible volume. Assume further the we have only two mobile charged species:
e1 = e0 and e2 = e0 , together with n1 = n+ and n2 = n , in equilibrium with a bulk
reservoir with n0i = n0 . In this case the Poisson-Boltzmann equation can be linearized
and reduced to the Debye - Hückel equation
e 0 n0 X e 0 n0 X
r2 (r) = zi e zi e0 (r) ⇡ (zi (zi )2 e0 (r)+. . .) = 2D (r).
✏✏0 i ✏✏0 i
(298)
P P
By assumption i zi = 0 and i (zi )2 = 2. Above we have introduced 2D = 8⇡`B n0
where `B = e20 /4⇡✏✏0 kB T is the Bjerrum length, defined to be the separation between
two elementary charges at which their electrostatic interaction energy equals the thermal
energy. At room temeperature and water as solvent it’s value is 7.1.
The linearized PB equation is reduced to the Poisson equation, in the case of no
mobile charges. If we compare the solution of the Poisson equation (left column below)
and the linearized Poisson - Boltzmann equation (right column below) in one, two and
three dimensions we get for the electrostatic potential of a fixed point charge e0 located
at r0 the following expressions 120
e0 e0 D |z z0 |
(z) = |z z0 | vs. (z) = e
✏✏0 ✏✏0
e0 e0
(x, y) = log |⇢ ⇢0 | vs. (x, y) = K0 (D |⇢ ⇢0 |)
2⇡✏✏0 2⇡✏✏0
e0 1 e0 e D |r r0 |
(x, y, z) = vs. (x, y, z) = (299)
4⇡✏✏0 |r r0 | 4⇡✏✏0 |r r0 |
In all three cases we see that the long rangeelectrostatic potential is modified by the pres-
ence of mobile ions into a short range potential of a typical length D . Therefore the
constant D = D1 is interpreted as the inverse Debye screening length. For mono-
valent salts at
p room temperature with concentration expressed in moles [M] per liter,
D = 3.05/ [M ] in Å. The salient feature of electrostatic interaction on the linearized
PB equation level is thus the screening quantified by the Debye length D .
This picture of screened interactions has to be drastically modified in the case of
polyvalent counterions. The linearization of the PB equation is usually justified at asymp-
totic conditions, meaning usually small fixed charges and/or large separations between
charged macroions as well as by the fact that it is easily amenable to analytic solutions.
The full non-linear PB equation represents a much tougher mathematical problem ana-
lytically solvable only for special geometries.
7.2 Cell model and Manning condensation

In many colloidal systems, most notably in the case of ordered DNA phases, one seldom
deals with isolated molecules in ionic solutions. Quite often one has a phase of densely
packed macroions which complicates the problem of evaluating the electrostatic interac-
tions even further. A simple way around this problem is the polyelectrolyte cell model
121
, being a variant of the Wigner - Seitz model of electrons in the crystalline lattice,
which substitutes the complicated colloidal geometry with a cell, containing a single col-
loid as shown in FIg. 7.2.1. The effect of the rest is assumed to be mimicked by the cell
wall where the electrostatic potential should have a zero derivative by symmetry. Let us
furthermore assume that positively charged uni-valent counterions are the only mobile
component of the system comprised of a negatively charged cylinder of radius a and
counterions assumed to roam the space a < r < R, where R is the radius of the cell.
120 K (x) is the standard
0 p 2⇡cylindrical Bessel’s function of the second kind of order zero, that behaves asymp-
totically as K0 (x) ⇠ x
e x.
121 The cell model for DNA was introduced in R. Fuoss, A. Katchalsky, and S. Lifson, Proc. Natl. Acad. Sci
USA 37 (1951) 579.

117
Figure 45. Schematic presentation of a cell model of DNA. DNA is blue, counterions
are white and the outer wall of the cell is yellow. The density of counterions is a con-
tinuous function of the spatial position. On this wall the radial component (which is
perpendicular to the wall) of the mean electric field is zero.
This radius mimicks the effects of the other DNAs present in the system at a finite con-
centration with volume ratio v = a2 /R2 . The Poisson - Boltzmann equation for this case
can be formulated in terms of a dimensionless electrostatic potential u(r) = e0 (r) as
✓ ◆
1 @ @u(r) 1 X e20 n0
r2 u(r) = r = ⇢i ( (r)) = e u(r) . (300)
r @r @r ✏✏0 i ✏✏0
Here we assumed that due to the cylindrical symmetry of the cell the electrostatic poten-
tial depends only on the radial coordinate r. Differential equation of this form has been
first studied by Liouville already in 1853 122 .
The solution of this equation in the region a < r < R was first given by Fuoss,
Katchalski and Lifson as
!
2
(r) r
u(r) = log cos2 (z log( )) , (301)
(2z)2 RM
e20 n0
where we set 2 /2 = ✏✏0 . The integration constants z, RM , , are obtained from three
boundary conditions:
du du
|r=a = 2Q/a |r=R = 0 and u(r = R) = 0, (302)
dr dr
122 J. Liouville, J. Math. Pures Appl. 18 (1853) 71.
where we introduced the dimensionless linear charge density Q = lB /lDN A which for
the case of DNA takes on the value Q ' 4.353, determined by the charge separation
lDN A ' 1.7 and the Bjerrum length lB ' 7.14 for water at room temperature. Evaluating
now the derivatives if the dimensionless electrostatic potential at the external and the
internal boundary of the cell, we are led to
✓ ◆ ✓ ◆
a R
Q = 1 z tan z log( ) and 0 = 1 z tan z log( ) (303)
RM RM
while the third boundary condition gives 2 R2 = 4(1+z 2 ). The constant RM is reffered
to as the Manning radius. From these equations we determine z as a solution of
✓ ✓ ◆ ✓ ◆◆
a 1 1 Q 1
log( ) = arctan arctan . (304)
R z z z
This equation is solvable for any value of Q. For Q > 1 the ensuing z is real, while for
Q < 1 it becomes imaginary.
The solution of this equation gives z = z(a/R). It can only be obtained numerically.
Explicit analytical expressions exist only for limiting values of this ratio. Before ana-
lyzing them let us introduce the osmotic coefficient defined as the ratio of the actual
osmotic pressure to the osmotic pressure of a hypothetical gas of uniformly distributed
counterions or, equivalently, the number density of ions n(R) divided by the total density
nDN A = 1/(lDN A ⇡R2 ) of charges
n(R) 2 1
= = ⇡R2 lDN A = (1 + z 2 ) (305)
nDN A 8⇡lB 2Q
where, as noted before, z 2 (Q > 1) > 0 while z 2 (Q < 1) < 0. The sign of z 2 in
the above expression for the osmotic coefficient thus depends on whether the Manning
parameter Q = lB /lDN A is larger or smaller then 1. The nature of the solution thus
depends crucially on the value of this ratio.
In the asymptotic limit of large r and at infinite dilution, the approximate solution of
the PB equation up to an irrelevant constant has a form of a 2D electrostatic potential,
but with a renormalized charge
(
2Q log ar , Q<1
u(r) = r r
(306)
2 log a + 2 log 1 + (Q 1) log a , Q > 1.
This can be interpreted in the following way: first note that the potential 2 log ar is
the potential of an infinite cylinder of unit line charge density. Therefore for Q < 1 the
charged cylinder behaves as an infinitely long rod of linear charge density Q. On the other
hand for Q > 1 the electrostatic potential looks as if the rod would have an effective unit
linear charge density. If Q > 1 then the counterions condense on the rod until it has the
charge density Q = 1. This is also referred to as the Manning condensation. The physical
content of the above equations was noticed from a different perspective by Onsager in
1967 who observed an anomaly in the partition function of a Coulomb gas in 2D, where
the Coulomb interaction depends logarithmically on separation. G. Manning realised
in 1969 that Onsager’s insight is relevant for polyelectrolytes and in the ’70 Oosawa,
119
Figure 46. Numerical solution for the counterions density in the case of a single cylin-
ders below and above the Manning condensation treshhold, at Q = 1.5 and 0.5. Con-
densed layer is clearly seen in the forst case. The magnitude is indicated by the color
scaling from blue (small) to red (large).
Manning and Dolar realised that the Fuoss - Katchalski - Lifson solution of the Poisson
- Boltzmann equation gives the same behavior.
The corresponding values of z or better still the physically observable osmotic coef-
fcicient in the two limits are
!
2
1 ⇡
(Q > 1) ⇠ = 1+
2Q log2 ( R
a)
0 !2 1
✓ ◆ ⇣ a ⌘2(1 Q)
Q 1 Q
(Q < 1) ⇠ = 1 @1 + A. (307)
2 1 Q 2
R
In the case of the Manning condensation due to the slowly varying nature of ⇡ 2 /log2 ( R
a)
the infinite dilution limit might be very difficult to approach and is in fact reached
only asymptotically for infinitely large cell size. For DNA the Manning parameter is
Q = 4.2 > 1 and thus we have the case of Manning condensed counterions. For the
polysaccharide hyaluronic acid Q = 0.72 < 1, so not all polyelectrolytes have Manning
condensed counterions. For DNA the osmotic coefficient is thus ⇠ = 1/(2Q) ⇠ 0.12
and the fraction of condensed charge is (1 1/Q) ⇠ 0.76. On the average thus 3 out of
4 phosphates are "neutralized" by counterions. The rest are acting as "free", i.e. osmoti-
cally active counterions.
7.2.1 Effect of salt
Let us shortly asses also the solution of the PB theory in the presence of additional 1-
1 salt. The cell is now assumed to have the cylindrical macroion in the middle and to
contain coions as well as counterions. They both come from the added salt. Retaining
Figure 47. Schematic presentation of a cell model of DNA. DNA is blue, counterions are
white and coinons are black. The density of counterions and coions is assumed to be a
continuous function of the spatial position. On this wall the radial component (which is
perpendicular to the wall) of the mean electric field is zero.
the cell model formulation, the basic equation for the mean potential now reads
✓ ◆
1 @ @u(r) 1 X
= ⇢i ( (r)) = 2D sinh u(r). (308)
r @r @r ✏✏0 i
This equation belongs to the class of Painlevé III equations whose solutions are signifi-
cantly more complicated then in the case of no added salt. Mathematical theory of these
equations has been recently applied also to the case of counterion condensation 123 and
we give a brief account of the main results.
First of all the boundary conditions are exactly the same as before, Eq. 302 and thus
imply an asymptotic form of the solution as
u(r) = 4 K0 (D r). (309)
The parameter reduces to Q/(2D a K1 (D a)) for vanishingly small linear charge
density, but has to be computed from the solution of Eq. 308 in the general case. In the
asymtotic limit of D a ! 0, the critical value of Q is no more 1, but is given rather by
1
Qc = 1 + , (310)
log D a + log 8
with the Euler constant 124 and reduces to the previous result of Qc = 1 asymptotically
123 E. Trizac and G. Téllez, Phys. Rev. Lett 96 (2006) 038302.
124 ⇡ 0.5772 . . . .
121
in the limit of no salt. However, the correction to this limit is logarithmic in D a and
persists for all realistic values of salt concentration: for D a = 10 3 we get Qc = 0.881
and for D a = 10 1 we get Qc = 0.737. in the asymptotic dependence on D r, Eq.
309, which is valid with very good accuracy for D r > 1 is obtained from a coupled set
of equations
(2 2 )(D a)2 2
= Q+ ⇣ ⌘2
(12 )
(D a)2 2 16(1 2 )2 6
( 1+
2 )
✓ ◆
(D a)
2µ = (Q 1) tan 2µ log + 2µ (311)
8
for auxiliary variables and µ, that finally yield for the effective linear charge density in
the asymptotic limit of large D r the expression
1 ⇣⇡ ⌘
= sin Q < Qc
⇡ 2
1
= cosh (⇡µ) Q > Qc , (312)
⇡
that then features in the asymptotic electrostatic potential Eq. 309. It should be noted
here that non-linearities implied by the PB equation with finite salt lead to that does
not reduce to Q/(2D a K1 (D a)) for vanishing salt concentration but approaches a
value of Q = ⇡2 independent of bare Q > Qc . This means that in the Manning-Debye-
Hückel picture the interaction energy between two charged rods in the asynptotic regime
is usually overestimated.
The numerical solution of the above equations gives very close but always larger
then Q. Close to Qc both and µ behave as
⇠
= 1 + |Q Qc |1/2 and µ⇠
= |Q Qc |1/2 . (313)
For highly charged cylinder, Q >> 1, the solution of Eqs. 311 approaches
⇡
µ⇠
= 1)
. (314)
2 (log D a + log 8 (Q 1)
This well all results that can be obtained via an asymptotic analysis of the PB equation
in the case of added salt. They are obviously more complicated then in the case of no
salt, but of course that is to be expected. For computational purposes one is best advised
to solve the PB equation numerically from the start. In cylindrical geometry and the cell
model this is well-nigh a manageable task.
7.3 Forces and interactions

Interactions and thus forces between charge distributions in a medium containing mo-
bile charged species can be obtained by proiperly generalizing the Coulomb force. The
Coulomb force on a charge distribution contained in volume V with charge density ⇢(r)
is given by Z
F= ⇢(r)r (r) d3 r. (315)
(V )
⇣ ⌘
kB T ni (r)
In the case of mobile ions, the electrochemical potential (r) + z i ei log n0i
, Eq.
295, takes
P on the role of the electrostatic potential (r) and the total charge density equals
⇢(r) = j zj ej nj (r). Tthus the above equation should actually read
Z ✓ ✓ ◆◆
kB T ni (r)
F = ⇢(r)r (r) + log d3 r =
(V ) zi e i n0i
0 1
Z X
= ⇢(r)E(r) d3 r kB T r @ nj (r)A . (316)
(V ) j
This form of the force, where the volume integration is over theat part of the total volume
of the system that contains the charges on which the force we are about to calculate, can
be further cast into a form that contains the stress tensor pik in the form
Z I
@pik 3
Fi = d r= pik nk dS, (317)
(V ) @xk S
where the force is now obtained by integrating the stress tensor along the area S that
contains the whole body on which we want to calculate the force. The pressure tensor pik
is composed of two terms, since the free energy of the system is also composed of the sum
of the electrostatic energy and the ideal gas entropy Eq. 293. Comparing Eqs. 316 and
317, the pressure thus contains the Maxwell stresses part pik = ✏✏0 Ei Ej 12 E 2 ij
and the osmotic pressure of the idealPgas of all the N mobile ionic species given by the
van’t Hoff expression pik = kB T j nj ( (r)). The total stress tensor is thus
X
pik = ✏✏0 Ei Ej 1 2
2 E ij kB T ik nj ( (r)), (318)
j
where we explicitly indicated that the local density of the mobile charged species is a
function of the local electrostatic potential. In order to get the force acting on a charged
body in an ionic solution one first has to calculate the electrostatic potential and the
desnity profiles, insert them into the above equation, and integrate this pressure tensor
along a surface that contains only the body on which we want to calculate the force.
Depending on the symmetry of arrangement of the interacting charged bodies one
can conveniently chose various forms of the bounding surface S. Chosing the surface of
integration in Eq. 317 to coincide with the symmetry surface bisecting the line joining
the two cylinders. On this plane Ez , the component of the electric field parallel to the
normal to this surface is zero by symmetry. Thus we remain with
0 1
I X
F·n= @kB T nj (S) 12 ✏✏0 E 2 A dS. (319)
S j
This form of the force corresponds to repulsions between two eqally charged cylinders.
It turns out that in fact for any symmetric distribution of charges the interaction is always
repulsive though sporadically there are contraray claims in the literature. This fact s now
sufficiently well established.
123
7.4 Strong coupling theory

The traditional approach to charged (bio)colloidal systems has been the mean-field Poisson-
Boltzmann (PB) formalism applicable at weak surface charges, low counter-ion valency
and high temperature. The limitations of this approach become practically important in
highly-charged systems where counterion-mediated interactions between charged bod-
ies start to deviate substantially from the mean-field accepted wisdom. This is certainly
the case for DNA, one of the most highly charged molecules in Nature, in solutions
of polyvalent salts. One of the fundamental recent advances in this field has been the
systematization of these non-PB effects based on the notions of weak (WC) and strong
coupling (SC) approximations. The latter approach has been pioneered by Rouzina and
Bloomfield, elaborated later by Shklovskii et al., Levin et al., and brought into final form
by Netz et al. 125 . These two approximations allow for an explicit and exact treatment of
charged systems at two disjoint limiting conditions whereas the parameter space in be-
tween can be analysed only approximately and is mostly accessible solely via computer
simulations.
7.4.1 Electrostatic coupling parameter
Both the weak and the strong coupling approximations are based on a functional inte-
gral or field-theoretic representation of the grand canonical partition function of a sys-
tem composed of fixed surface charges with intervening mobile counterions, and depend
on the value of a single dimensionless coupling parameter ⌅. The distance at which
two unit charges interact with thermal energy kB T is known as the Bjerrum length
`B = e20 /(4⇡""0 kB T ) (in water at room temperature, one has `B ' 0.7nm). If the
charge valency of the counterions is q then the aforementioned distance scales as q 2 `B .
Similarly, the distance at which a counterion interacts with a macromolecular surface
(of surface charge density ) with an energy equal to kB T is called the Gouy-Chapman
length, defined as µ = e0 /(2⇡q`B ). A competition between ion-ion and ion-surface
interactions can be quantitatively measured with a ratio of these characteristic lengths,
that is
⌅ = q 2 `B /µ = 2⇡q 3 `2B /e0 ,
which is known as the (Netz-Moreira) electrostatic coupling parameter. The weak cou-
pling (WC) regime ⌅ ⌧ 1 (appropriate for low valency counterions and/or weakly
charged surfaces), is characterized by the fact that the width of the counterion layer µ
is much larger than the separation between two neighbouring counterions in solution and
thus the counterion layer behaves basically as a three-dimensional gas. Each counte-
rion in this case interacts with many others and the collective mean-field approach of the
Poisson-Boltzmann (PB) type is completely justified. On the other hand in the strong
coupling (SC) regime ⌅ 1 (appropriate for high valency counterions and/or highly
charged surfaces), the mean distance p between counterions, a? = /qe0 , is much larger
than the layer width (i.e., a? /µ ⇠ ⌅ 1), indicating that the counterions are highly
localized laterally and form a strongly correlated quasi-two-dimensional layer next to a
charged surface. In this case, the weak-coupling approach breaks down due to strong
125 For a review of these developments see: H. Boroudjerdi, Y.W. Kim, A. Naji, R.R. Netz, X. Schlagberger
and A. Serr, Phys. Rep. 416, 129 (2005).

counterion-surface and counterion-counterion correlations. Since counterions can move

almost independently from the others along the direction perpendicular to the surface,
the collective many-body effects that enable a mean-field description are absent, neces-
sitating a complementary SC description.The range of validity of both limiting theories
at intermediate values of the coupling parameter has been explored thoroughly in the
literature 126 .
Formally the weak coupling limit can be straightforwardly identified with the mean-
field PB theory in the lowest order for ⌅ ! 0. The quadratic fluctuations around the
mean field provide a second-order correction to the mean-field solution for small finite
⌅ < 1. The strong coupling approximation has no PB-like correlates since it is formally
equivalent to a single particle description obtained from a systematic 1/⌅ expansion in
the limit ⌅ ! 1, and corresponds to two lowest order terms in the virial expansion of
the grand canonical partition function. The consequences and the formalism of these two
limits of the Coulomb fluid description have been explored widely and in detail 127 .
7.4.2 Partition function in the strong coupling theory
In many respects the SC theory of electrostatic interactions between charged macro-

molecules is simpler then the PB theory. In effect the partition function can be reduced
to the lowest order in ⌅ to a simple quadrature.
In the strong coupling limit the partition function is approximated by the lowest order
term in the virial expansion
1
X
0N 0 02
ZG = ZN = Z0 + Z1 + O( ), (320)
N =0
where 0 is proportional to fugacity and is in fact given by = 0 e µ , where µ is the

chemical potential. Z0 is the partition function of the system with no counterions and Z1
is te partition function with a single counterion. Higher order terms contain the partition
function with two, three etc. counterions. In the SC limit, one cuts the fugacity expansion
at the first order term. That is why the SC theory is effectively a one particle theory. By
definition Z0 and Z1 are given by
1 RR
drdr0 ⇢0 (r)u(r,r0 )⇢0 (r0 )
Z0 = e 2 = e U0 (321)
where U0 is the electrostatic interaction energy due to ⇢0 (r) and certainly contains a part
that depends on the separation between the two macromolecules, charge distribution and
Z R 0 0 0
Z1 = Z0 dR e q dr u(R,r )⇢0 (r ) = Z0 U1 . (322)
U1 is just the partition function of a single counterion in the field of the two cylindrical
macroions. Here ⇢0 (r) is the charge density of external charges. Assuming we have two
charged cylinders of radius R, separated by an interaxial spacing D in the direction of
x axis with the same surface charge density s , the external charge density. u(r, r0 ) =
126 Boroudjerdi et al. op.cit.
127 Boroudjerdi et al. op.cit.
125
(4⇡""0 |r r0 |) 1 is the Coulomb potential and q is the charge of the counterions. The
partition function in the strong coupling theory can then be obtained in the form
0 02
F= kB T log Z0 kB T U1 + µN + O( ), (323)
where N is the number of counterions in the system that neutralize exactly the charge
on the cylindrical surfaces. Taking into account the expression for the chemical potential
that can be obtained from
@F 0
=0 and thus U1 = N
@ 0
and one remains with the following form of the free energy in the strong coupling ap-
proximation
F(D) = U0 (D) kB T N log U1 (D), (324)
where the dependence on the separation between the cylinder was written explicitly. Ob-
viously the second term contains the single-particle interaction energy as well as its en-
tropy due to the confinement of the counterion between the bounding surfaces of the two
cylindrical macromolecules.
7.4.3 Forces and interactions
Assume one has two charged macroions immersed in a bath of polyvalent neutralizing
counterions. When the separation between the macroions, D, is changed in the direction
n, the component of the force between them in this direction can be calculated from
@F( , D)
F( , D) · n = , (325)
@D
where describes the mutual orientation of the two macroions if they do not posses
spherical symmetry. Obviously the mutual forces contain two parts: one due to the
direct electrostatic interaction and the other one due to the interactions mediated by single
counterion
@U0 ( , D) @ log U1 ( , D)
F( , D) · n = + kB T N . (326)
@D @D
The first term, for two macroions with identical charge distribution, is always repulsive.
The sign of the second one depends on the interplay of several mechanisms: the con-
finement entropy of the counterion in the space between the macroions, depletion of the
finite size counterion in the space between the macroions and the counterion mediated
electrostatic interactions. Of these three contributions the last two are attractive and the
first on is repulsive 128 .
7.5 DNA-DNA electrostatic interactions

We now analyze interactions between two parallel DNA molecules described either as
uniformly charged cylinders or as cylinders with a helical structural charge motif. The
128 M. Kanduc, J. Dobnikar and R. Podgornik, Counterion-mediated electrostatic interactions between helical
molecules, Soft Matter DOI: 10.1039/b811795k (2008).

structural parameters of ds-DNA in the B conformation, with two unit charges per base
pair, set the effective surface charge density of the uniformly charged cylinder. For the
ds-DNA the structural parameters are: cylinder diameter of a = 1 nm and helical pitch
of H = 3.4 nm, so that the Manning parameter is Q = 4.1q where q is the counterion
valency. The Manning parameter with mean surface charge density , is defined as
e0 qa
Q= . (327)
2""0 kT
Note that at fixed surface charge the Manning parameter Q is proportional to counterion
valency q. The electrostatic coupling parameter ⌅ for the ds-DNA case is ⌅ = 2.8 q 3 in
the presence of counterions with valency q. The results for WC as well as SC approxi-
mations do not depend directly on ⌅ but this parameter determines the range of validity
of these approximate theories. For monovalent counterions the coupling parameter has a
value of ⌅ = 2.8 which is small enough to justify the WC approach. But this approach
certainly breaks down for tri- and tetra-valent case where ⌅ = 76 and 180, respectively.
7.5.1 Charged cylinders, weak coupling, no salt
Let us now examine the case of the interaction of two charged parallel cylinders as a
model system to study interactions between two DNA molecules that we will later treat
in more detail. The structural charge of the cylinders is assume dto be the same as that of
DNA. In case the DNAs are in the presence of their own monovalent counterions only,
the Poisson - Boltzmann equation has the form of Eq. 296
✓ ◆
@2 @2 e0 0 ⇢(x, y)
r2 (r) = 2
+ 2 (x, y) = n e e0 (x,y)
= . (328)
@x @y ✏✏0 ✏✏0
Due to assumed cylindrical symmetry the mean electrostatic potential (x, y) as well
as the charge density ⇢(x, y) depend only on the x, y coordinates. The value of n0 is
obtained from the electroneutrality condition Eq. 297. The form of the electrostatic
potential and the charge density is given in Fig. 7.5.2.
The force in the direction of the line joining the two molecules can now be obtained
from Eq. 319. Since the system is homogeneous along the axis of the two parallel
molecules the force between them must scale linearly with the length `, thus F · n = F`.
The force along the direction joining the centers of the two cylinders is obtained as
I I
2 0
1
F · n = 2 ✏✏0 E(S) dS + kB T n e e0 (S) dS. (329)
(S) S
For two charged cylinders the force can be cast into a dimensionless form by defining
2
F̃ = F ⇥ 2✏✏0 µ2 , where is the surface charge desnity of the cylinders, and µ is the
Gouy-Chapman length. The dimensionless separation is defined as R̃ = R/µ. All
numerics was performed for DNA structural charge ' 0.94 e0 /nm2 and monovalent
counterions in aqueous solution with µ ' 0.24 nm. The size of the computational box
was taken large enough so that it does not influence the results. Since we computed the
force per unit length between the two charged cylinders, it is now trivial to construct the
127
Figure 48. Numerical solution of the Poisson - Boltzmann equation for two cylinders at
a separation R. Left is the density of counterions ⇢(x, y) and right is the electrostatic
potential (x, y). The magnitude is indicate by the color scaling from blue (small) to red
(large).
Figure 49. Dimensionless force between two charged cylinders. It scales approximately
as R̃ 1 . The approach to this exponent is very slow and is in fact reached only in the
asymptotic limit of infinitely large cell.
corresponding DNA-DNA interaction potential per unit length V (R). It is defined in a

standard way, so that
@V (R) @ Ṽ (R̃)
F= = i.e. Ṽ (R̃) = V (R̃)/µ, (330)
@R @ R̃
where Ṽ (R̃) is now the dimensionless interaction potential per unit length of the inter-
acting molecules.
We can obtain an analytic form for the separation dependence of the interaction po-
tential only in the limit of vary large separations R. In this case we already know that the
asymptotic forms of the electrostatic potentials are given by expressions Eq. 306 and do
not contain any explicit presence of the counterions. Concentrating again on the force in
the direction of the line joining the two cylinders, F · n, which equals the surfaec integral
of the zz component of the pressure tensor, Eq. 318,
I I
2 0 e0 (S)
F·n= 1
2 ✏✏0 E(S) dS + kB T n e dS, (331)
(S) S
which contains only the square of the electric field along the symmetry surface at the
midpoint between the cylinders. The surfaec integral can be performed explicitly and
from this we can obtain the dimensionless interaction potential per unit length of the
interacting molecules as
(
kB T (Q/2)2 log R̃, Q<1
Ṽ (R̃) = (332)
µ`B 2
(1/2) log R̃, Q > 1.
Take note that Q = `B /`DN A = 12 ã and defined `˜B = `B /µ. Clearly the asymptotic
dimensionless interaction potential depends crucially on the magnitude of Q, reflecting
the effects of Manning condensation of counterions. Better analytical approximations
can be devised easily 129 .
Figure 50. Potential, left, and density, right, evaluated at R̃ = 10 with 0.1M salt. Obvi-
ously the potential as well as the density decay faster away from the charged cylinders.
129 F. Oosawa, Polyelectrolytes, M. Dekker, New York, 1971.

129
7.5.2 Charged cylinders, weak coupling, effects of salt
In the case of added salt the above discussion has to be modified in accordance with the
results of the salt effects on the electrostatic potential profile. Numerical solutions of
the PB equation show mostly the effects of screening on the decay of the potential and
density away from the charged cylinders, see Fig. 7.5.2. A simple analytical approxi-
mation for the interaction potential can again be obtained only in the asymptotic case of
R >> a.
Writing down the normal component of the force at the symmetry plane between the
two parallel molecules within a 1:1 salt bath at a saeparation R one has
I I
2 0
1
F · n = 2 ✏✏0 E(S) dS + 2kB T n cosh ( e0 (S)) dS. (333)
(S) S
We now approximate cosh ( e0 (S)) at the symmetry plane with its lowest order ex-
pansion, thus cosh ( e0 (S)) = 1 + 12 ( e0 (S))2 + . . ., and discard all the inessential
constants that do not depend explicitly in the separation between the molecules. In the
asymptotic regime of large separations between the two cylinders we also assume the
superposition approximation, where the potential at the midplane is simply equal to
the sum of the potentials of two isolated cylinders Eq. 309. Thus we obtain
I p
F · n = kB T n0 (4 )2 (K0 (D x2 + (R/2)2 ))2 dS + . . . . ⇠ =
S
⇠ D
= L kB T (2 )2 K0 (D R), (334)
`B
with L the length of the cylinders. x is the coordinate on the plane of symmetry perpen-
dicular to the line joining the two cylinders. The dimensionless interaction potential per
unit length of the interacting molecules, Ṽ (R̃) now assumes the form
Z ( 2
kB T D R 2 ⇡
⇡ sin 2 , Q < Qc
Ṽ (R̃) = K0 (t)dt ⇥ 2 (335)
˜
`B µ 0 2
⇡ cosh (⇡µ) , Q > Qc .
where and µ are obtained from Eq. 311 and Qc is given in the limit of small salt by
Eq. 310. As already discussed, close to Qc we have and µ given by Eq. 313 and in the
complementary limit of Q Qc µ approaches the expression Eq. 314.
In the general case one should best rely on the numerical solutions of the PB equation
vwhich can be obtained very straightforwardly for cylindrical geometry. The main effect
of salt is thus to screen the interactions between charged cylindrical molecules, and in
particularly between DNA moelcules. The screening effect of added salt has been well
studied also experimentally and is in broad outlines consistent with the theoretical picture
emerging from the PB theory. This is however only for monovalent salts. The case of
polyvalent counterions is vastly different and can not in general be treated via the PB
formalism.
7.5.3 Charged cylinders, strong coupling, no salt
In the case of two charged cylindrical macroions the external charge density can be writ-
ten as
⇢0 (r) = s (r1 R) + s (r2 R). (336)
Figure 51. Dimensionless force between two charged cylinders at different values of the
salt concentrations. Inverse Debye length is expressed in terms of the Gouy-Chpaman
length. For 0.1M salt, e.g.  = 1.1/nm and µ = 0.24nm, so that the dimensionless ̃
equals 0.25. The potential and density are evaluated at R̃ = 10.
where s is the constant surface charge density and the two radii are defined as
p p
(x D/2)2 + y 2 = r1 and (x D/2)2 + y 2 = r2 .
Assuming furthermore that the linear charge density of the cylinders is ⌧ , the surface
charge density can be written as s = 2⇡R ⌧
and the corresponding Gouy-Chapman
parameter for cylindrical macroions assumes the form µ = `bRq⌧ . If one furthermore
rescales the radius of the cylinder by the Gouy-Chapman parameter one obtain exactly
the Manning condensation parameter
R
⇠= = `b q⌧. (337)
µ
The free energy per unit length of the two cylindrical macroions as derived in Eq. 324 up
to an irrelevant constant can now be written explicitly for the two cylindrical macroions
as 130 Z ✓ ◆ 2⇠ ✓ ◆ 2⇠
2 D r1 r2
f = 2⇠ log 2⇠ log dS , (338)
µ (S) µ µ
where S is the cross-sectiona area available to the counterions between the two hard
cylinders. The counterions are assumed to be of vanishing hard core radius.
The first term in the above expression is of course nothing else but the direct electro-
static repulsion between two homogeneously charged cylinders and scales as the log of
the separation between the two. The second term is the leading entropic as well as en-
ergetic term describing the counterion mediated strong coupling attraction between the
130 Naji and Netz, Europhys Lett ().
131
two macroions. Obviously it corresponds to an attraction mediated by a single counte-

rion, which is strictly what the SC theory is.
The second term also reflects directly the Manning condensation transition transition.
Obviously if ⇠  1/2 the integral in Eq. 338 diverges in the limit of an infinite volume
available to the two macroions (the thermodynamic limit) and the strong coupling coun-
terion mediated interaction between them vanishes also. On the other hand for ⇠ > 1/2
counterion distribution becomes strongly localized in a narrow region between the rods
and gives rise to strong coupling attraction. The total free energy in this case shows an
attractive branch at sufficiently large D and a minimum at small values of the interaxial
separation D⇤ ⇠ 2R.
Figure 52. The rescaled strong-coupling interaction free energy of the two-cylinder sys-
tem as a function of the interaxial spacing for Manning parameter ⇠ = 1.0 above and
⇠ = 0.3. Adapted from A. Naji and R.R. Netz, Attraction of like-charged macroions in
the strong-coupling limit , Eur. Phys. J. E 13, 43-59 (2004).
For ⇠ < 1, the integral in Eq. 338 is mostly dominated by the the contribution of the
bounding region around the charged cylinders which gives a contribution independent of
the macroion separation. The corresponding free energy can be derived in the form
D
f= 2⇠(2 3⇠) log . (339)
µ
On the other hand for ⇠ > 1 the integral in Eq. 338 is dominated by the region proximal
to both cylinders and the corresponding free energy turns out to be

D
f = 2⇠ 2 log . (340)
µ
Obviously the two charged cylinders experience repulsion at large separations, when the
Manning parameter, ⇠, becomes smaller than the threshold
⇠c = 2/3
and they attract each other for larger Manning parameters ⇠ > ⇠c . This result is obtained
also by exact numerical calculation of the free energy 131 . The attractive force predicted
in the strong-coupling limit is long-ranged and scales inversely with the interaxial sepa-
ration between the rods. It originates mainly from the energetic interaction mediated by
counterions sandwiched between the two rods.
Thus for ⇠c > 2/3 there is always asymptotic attraction between the two cylinders.
However at sufficiently small separations an additional repulsion sets in due primarily
to the direct interaction between the surface charges on the cylinders. The two together
then conspire and yield a minimum of the interaction free energy at a finite value of the
interaxial separation as is clearly seen from numerical results, see FIg. 7.5.3. In the
asymptotic limit of ⇠ ! 1 the corresposnding free energy can be evaluated exactly,
leading to
D D 2R
f = 6⇠ 2 log 2⇠ log + const.. (341)
µ µ
Obviously the ïňArst
˛ term in the above equation contributes a dominant energetic attrac-
tive force and the second term contribute a repulsive force between the cylinders at small
separations. The equilibrium separation D⇤ between the cylinders, correspnding to the
minimum in the interaction free energy, can be easily obtained as
2
D⇤ ⇠ 2R + µ + O(µ2 /R). (342)
3
As stated this result is true only in the limit of very large Manning parameters. For
⇠ < ⇠c the attractive branch of the interaction vanishes and the two cylinders experience
monotonic electrostatic repulsion due to their surface charge, see FIg. 7.5.3.
7.5.4 Cylinders with a helical charge motif, weak coupling, no salt
We now expand our investigation of the interactions by considering an explicit helical

charge motif wrapped around the cylinders, see Figure 7.5.4. The only difference with
respect to the previous analysis for uniformly charged cylinders is that now the charge
density as well as the mean potential are functions of all three coordinates, (x, y, z) and
⇢(x, y, z), and that the interactions depend also on the mutual azimuthal orientation of
the two helices. It is still given by the general expression Eq. 329.
The force of interaction between two helically wrapped cylinders can be normalez
by length and is given in Fig. 7.5.4. On the mean-field level the force between charged
131 A. Naji and R.R. Netz, Attraction of like-charged macroions in the strong-coupling limit , Eur. Phys. J. E
13, 43-59 (2004).

133
Lz
!0
Ly
R
Lx
Figure 53. Schematic presentation of two cylinders with a (single) helical charge motif.
helices is always repulsive and decays with increasing interaxial distance. In a symmetric
system on the mean-field level it is well known that interactions are always repulsive, a
result that can be formulated as a general theorem 132 .
On Fig. 7.5.4 we present the results of numerical calculations of the interaction force
between two cylinders with a helical charge motif as a function of the interaxial sep-
aration as well as the azimuthal angle of mutual orientation. Increasing the Manning
parameter (corresponding to higher counterion valencies) decreases the repulsive inter-
action though it always remains repulsive. The maximal angular variation at a separation
of R = 2.3 nm amounts to only 16% of the total force per unit length, and is a lot smaller
if the cylinders are even further apart.
The azimuthal angle dependence of the force for two helical charge distributions thus
shows only a slight variation, being larger at smaller interaxial separations. This is a
completely general conclusion of the mean-field analysis. The azimuthal modulation of
the force has a maximum at a configuration corresponding to '0 = ⇡ and a minimum at
'0 = 0. It also shows mirror symmetry along '0 = ⇡ line, so that configurations '0 and
2⇡ '0 have exactly the same interaction.
Because the potential depends on all the coordinates, so does the mean density of the
counterions, see Fig. 7.5.5. If we compare it to the uniformly charged cylinders case,
Fig. 7.5.2, we see that most of the counterions are concentrated in the vicinity of the
helical stripes. The interaction between the helices is thus largest if they are oriented
mirror symmetrically, '0 = ⇡, since in this case the concentration of the counterions
at the mid-plane is largest, giving concurrently the largest osmotic contribution to the
interaction.
The angular variation of the interaction force for helical charge distributions obvi-
ously makes up for only a slight angular modulation, between 10% and 20%, if compared
with the uniformly charged case. The largest repulsive force appears at '0 = ⇡ mutual
cylindrical orientation and the smallest at '0 = 0. Increasing the counterion valencies q
tends to decrease the amount of counterions in the system and the corresponding osmotic
132 J. E. Sader and D. Y. C. Chan, J. Colloid Interf. Sci., 1999, 213(1), 268.
Figure 54. Left, force per unit length - separation dependence for two helices with
ds-DNA parameters at various mutual azimuthal angles '0 and valencies. The corre-
sponding Manning parameter for q = 1 is Q = 4.1 and for q = 2 it is Q = 8.2. Right,
azimuthal angle dependence of the force per unit length at different interaxial distances
for ds-DNA parameters and monovalent counterions (Q = 4.1).
contribution as well as the net interaction is therefore smaller but again, it never changes
the sign. If the two helices can freely rotate around their long axes they will settle in the
orientation that minimizes their free energy. At large distances most unfavorable con-
figuration on the mean-field level corresponds to '0 = ⇡ and the most favorable one to
'0 = 0, 2⇡. Counterions that mediate electrostatic attraction between both helices tend
to orient helices so that the local helical segments are as far away as possible.
7.5.5 Cylinders with a helical charge motif, strong coupling, no salt
In the SC limit the free energy Eq. 324 depends on the bare electrostatic interaction
energy between the two charged helices U0 (D) as well as on the partition function of
a single counterion between them, U1 (D). The bare electrostatic interaction energy per
counterion can be evaluated straightforwardly by the integration of charge contributions
on both helices with the unscreened Coulomb kernel
Z 1Z H
U0 Q dz dz 0
= . (343)
N 2H 1 0 s(z, z 0 )
Here, s(z, z 0 ) represents the distance between two points on different helices with coor-
dinates z and z 0 , respectively. Helices are represented as (infinitely) thin charged lines.
The single counterion partition function is then obtained simply by integration of the
Coulomb kernel along both helices as
Z
U1 (D) = e u(r) dV
V
where V is the volume accessible to counterions. Here the single particle interaction
potential is Z 1✓ ◆
1 1
u= Q + dz 0 , (344)
1 s1 (z 0 ) s2 (z 0 )
135
Figure 55. Left: SC force per unit length-separation curves for various relative orienta-
tions, '1 '2 , in the case of two cylinders with ds-DNA-like parameters: radius a = 1
nm and pitch H = 3.4 nm in the presence of strongly coupled polyvalent counterions.
The corresponding Manning parameter is Q = 4.1 and the coupling parameter ⌅ = 3.
The inset shows force per unit length-separation curves for orientation '1 '2 = ⇡
for different cylinder radii. Qualitatively the interactions are the same. Right: Force per
unit lengthâĂŞangle curves for various interaxial distances. ds-DNA-like parameters are
taken and q = 1.
where s1 and s2 are distances from the counterion-coordinate to the point on helix 1 and
helix 2 defined as
q
s1,2 (⇢, ', z; z 0 ) = a2 + ⇢2 2a⇢ cos(kz 0 + '1,2 ') + (z z 0 )2 (345)
The angles '1 and '2 represent orientations of both cylinders around their symmetry
axes. Here, again as in the WC case, the convention '1 = 0 and '2 = '0 is used without
loss of generality.
The integral fro U1 (D) should be in principle evaluated over the entire space available
to the counterions which converges rapidly since the counterion density decays very fast
with the radial distance from the cylinders. In the case of a single uniformly charged
cylinder the density decays as r 2Q , so the asymptotic behaviour at large distances for
two cylinders with a helical charge motif behaves as r 4Q . Typically, Q > 4 for DNA
so the volume integral converges very rapidly.
The interaction force between both cylinders is obtained by taking the derivative of
the free energy with respect to the interaxial separation
@F('1 , '2 , D)
F ('1 , '2 , D) = . (346)
@D
Obviously, the force scales with the length of the cylinders, therefore we express the
force per length of cylinders, F/L. It depends on the separation between the helices and
their respective orientations.
The interaction force in the case of the strong coupling limit shows a much richer
behaviour than in the weak coupling (PB) case. As seen from Fig. 7.5.5 the interaction
force for various parameters, exhibits similar qualitative behaviour. Again we will use
ds-DNA parameters for illustration. We present the SC results for monovalent as well as
polyvalent counterions even though small valencies q that correspond to small coupling
parameter ⌅ are not relevant for the SC theory. The aim here is to illustrate the differ-
ent behaviour of the interaction for counterions of different valencies, i.e. of different
Manning parameters.
We find three different regimes in the behaviour of the interaction as a function of in-
teraxial separation, Fig. 7.5.5. At large separations the SC force is always attractive. This
is due to the fact that each counterion is localized at one or the other cylinder, effectively
creating a correlation hole, causing net correlation attraction between the helices. At
larger interaxial spacings the SC theory breaks down and the WC repulsion takes over.
Similarly to the case of interacting planar surfaces, for very small interaxial spacings
the counterions accumulate in the space between the two cylinders, creating a repulsive
component to the total interaction due to their osmotic pressure.
The osmotic contribution can be larger than the electrostatic correlation attraction and
a net repulsion ensues for the total interaction between helices at very small separations.
The highest value of osmotic pressure is reached when the separation between cylin-
ders becomes equal to the counterion diameter 2Rc , see Fig. 7.5.5. This consequently
leads to the highest repulsive force between the cylinders. At yet smaller separations,
the depletion effect sets in, in the sense that the counterions cannot penetrate any more
the inter-cylinder space, and are thus depleted from the spatial region between the two
apposed cylinders, diminishing their osmotic pressure in that region. Thus, the counte-
rion contribution to the osmotic pressure is reduced, i.e. the osmotic repulsion as well as
counterion electrostatic interactions are both reduced.
Figure 56. Potential, left, and density, right, at a cross-section along the cylinders with
a (single) helical charge motif evaluated at R̃ = 10 with counterions only. The points
where the helical charges cross the plane perpendicular to the axis of the cylinder are
clearly visible. The density and the electrostatic potential are substantial only close
to the helical stripe. The z axis of the helices is directed out of the paper. Dielectric
inhomogeneities were ignored.
In all these situations the bare electrostatic repulsion between the two helices remains
unaffected. Because of this a bound state between the cylinders occurs at a separation
137
R* where the net force is zero F = 0. The relative orientation of the two cylinders plays
a crucial role in determining the nature of the inter-helical interaction. The force at
orientation ⇡ now appears to be an order of magnitude larger than at orientations 0 or
⇡/2. This is quite different to the WC case, where the analogous variation was much
smaller, and indeed almost negligible. The case of two uniformly charged cylinders is
found somewhere in between both extremal behaviours, see Fig. 6.
From Fig. 8 we also deduce that the extremum of the force is reached at ⇡. This could
be explained by the fact that at this orientation the charges on helices are apposed and can
approach at a smallest local separation. The electrostatic potential between the cylinders
is therefore higher than for other values of '1 , '2 , which causes stronger localization of
counterions in between. Stronger localization in its turn leads to larger attraction at large
distances as well as greater osmotic repulsion at smaller distances.
7.5.6 Kornyshev - Leikin theory

8 DNA COLLAPSE AND DNA MESOPHASES

Though ordered phases of DNA in aqueous solutions are not difficult to prepare or ob-
serve it took some time until Conmar Robinson reported the first sighting of a DNA
cholesteric phase in 1961 133 in a solution containing about 6 % of DNA in 0.1 NaCl.
The cholesteric order observed was very similar to the cholesteric order in polypeptide
solutions which were studied in detail. The observed pitch was ⇠ 1µm and the concen-
tration range of the cholesteric phase was small. Later Lerman 134 noticed that the samo
type of cholesteri corder can be seen also in polymer and salt (acronym ) induced con-
densation of DNA. The cholesteric pitch seen in these samples in the form of cholesteric
spherulites was the same as in Robinson’s case.
Detailed investigation of the properties of DNA mesophases started with the obser-
vation of Robinson in 1961 that DNA in aqueous solutions can make similar liquid crys-
talline phases as some filamentous proteins. Later this observation was systemitized in
the work of Rill, Livolant and Parsegian. They deciphered the most important features
of the DNA phase diagram and characterised the order and thermodynamic properties
of DNA within the various regions of the phase diagram. Our current understanding of
the sequence of phases in DNA solutions, going from the dilute isotropic solution to the
cristalline phase are summarized in Fig.
L. Onsager presented detailed arguments and calculations on how long thin rods ori-
entationally order at sufficiently high densities. His arguments, at least to some extent,
remain valid also for DNA but also with important caveats. Recent work on short frag-
ment DNA (s-DNA) 135 with as few as 6 base pairs in a row show that even in this case
the DNA can nematically order, which is in complete contradiction to the Onsager theory
that demands that the length of the nematogens should be at least four times their diam-
eter. There is subtle interplay between the electrostatic nad sterical interaction of the
double helix periphery and the stacking and hydrophobic interactions of the DNA inte-
rior that shows in the nature and characteristics of the DNA liquid crystalline transitions
and phases.
In the phase diagram of DNA in aqueous solutions there is a range extending from the
DNA crystal at densities corresponding to interaxial separations below ⇡ 24 all the way
to the nematic - isotropic transition at densities corresponding to interaxial separations
⇡ 120 (for a bathing solution of 0.5 M NaCl) where DNA is orientationally ordered or
even shows long range hexatic order perpendicular to the long axes of the molecules but
is positionally still a liquid with only short range positional order 136 . This is in particu-
lar valid for long ⇠ µm fragments of DNA. The local structure of these DNA arrays is
presented on Fig. ??. In this part of the phase diagram, the equation of state, i.e. the de-
pendence of the osmotic pressure on the macromolecular (DNA) concentration, has been
studied very carefully 137 . In this lecture I will give a comprehensive introduction to the
equation of state for DNA under various solution conditions with monovalent salts and
133 C. Robinson, Liquid crystalline structures in polypeptide solutions, Tetrahedron 13 (1961) 219-234.
134 L.S. Lerman, Chromosomal Analogs - Long-Range Order in Psi-Condensed DNA, Cold Spring Harbor
symposia on quantitative biology, 1974;38:59-73.
135 M. Nakata, G. Zanchetta, et al. B.D. Chapman, C.D. Jones, J.O. Cross, R. Pindak, T. Bellini, and N.A.
Clark, Science 318, 1276 (2007).

136 R. Podgornik, H.H. Strey and V.A. Parsegian, Curr. Op. in Colloid and Interf. Sci. 3 (1998) 534.
137 R. Podgornik, et al., op. cit.
139
counterions and make an attempt to explain in some detail the theoretical underpinnings
of its equation of state.
8.1 Collapse of a single DNA molecule

Let us start with an elastic filament subject to self non-local interactions along the fila-
ment. This model has been treated on various levels of approximation and we will choose
the one that appears the most straightforward. Non-local interactions in this context
means that two segments that are quite apart along the chain, can be in close proximity
in the embedding space of the chain. Neglecting torsional rigidity we can write the total
mesoscopic Hamiltonian Eq. 161 in this case as
Z L Z LZ L
H[t, ṫ] = 12 Kc ṫ2 d` + 12 V (r(`), r(`0 ))d`d`0 . (347)
0 0 0
We assumed in addition that the interaction energy between two segments along the
filament depends only on their position and not e.g. their orientation. This is of course
an approximation and we could refine our analysis later to including this part of the
interaction. Furthermore we assume that the non-local interaction energy depends only
on the absolute value of the relative separation between segments r(`) and r(`0 ) along
the filament, i.e.
Z LZ L Z LZ L
0 0
V (r(`), r(` ))d`d` = V (|r(`) r(`0 )|)d`d`0 =
0 0 0 0
Z LZ L Z `0
= V (| t(s)ds|) d`d`0 , (348)
0 0 `
where we took into account the definition of the tangent vector and its integral. Param-
eterizing the Hamiltonian Eq. 347 with Euler angles Eq. 27, (`), ✓(`), (`), where the
unit tangent vector has the form
t(`) = (sin ✓(`) cos (`), sin ✓(`) sin (`), cos ✓(`)), (349)
leads to the energy functional obtained in the form 138

Z L⇣ ⌘
˙ ˙ ˙
H[✓, ✓, , ] = 2 Kc 1
✓˙2 + ˙ 2 sin2 ✓ d` +
0
Z LZ Z !
L `0
+ 12 V (sin ✓(s) cos (s), sin ✓(s) sin (s), cos ✓(`))ds d`d`0 .
0 0 `
(350)
This is obviously a non-local functional since it contains a term witha double integration
along the flament. The partition function would then be given by the standard form
Z
⌅(L) = D[t(`)] exp ( H[t, ṫ]).
138 Ishimoto, Y; Kikuchi, N: J. .Chem. Phys. 125 074905 (2006).

Ignoring for the moment the thermal effect, thus assuming zero temperature, the equilib-
rium configuration is given as an extremum of the action in the above functional integral,
i.e. by the minimum of the free energy functional Eq. 350. This is not a difficult under-
taking but demands some caution and we will proceed in steps.
Let us first forget about the interaction term. Minimization of the elastic part of the
Hamiltonian Eq. 350 gives two Euler-Lagrange equations of the form
✓¨ ˙ 2 sin ✓ cos ✓ = 0
( ˙ sin2 ✓) = const. (351)
The solutions of these two equations are of only two types 139 :
⇡
✓ = 0, ⇡ and ✓˙ = 0 or ✓= and ˙ = const. = a. (352)
2
and correspond either to a constant tangent vector, but since the tangent vector is a unit
vector, to a constant direction, or to a unit vector that circles around the big circle of the
R`
sphere. Solving now also for r(`) via r(`) = 0 t(s)ds, we obtain circular solutions in
the r(`) space 0 1
sin (a` + b) sin b
B C
r(`) = 1 B cos (a` + b) + sin b C (353)
a @ A,
const.
where a is obviously the inverse radius of the circle. We will take a circle in this context
as a torus with very small thickness. We now analyze the effects of the interaction term.
Assume first of all the the interactions are attractive and of very short range on the scale
of the persistence length of the chain. In this case a reasonable approximate form would
be given by a Dirac function potential
Z 0 !
`
V (|r(`) r(`0 )|) = w (|r(`) r(`0 )|) = w t(s)ds =
`
Z !
`0
= w (sin ✓(s) cos (s), sin ✓(s) sin (s), cos ✓(`))ds .
`
(354)
If the argument of the delta function is to be zero, all the components of the vector
r(`) r(`0 ) must be identically equal to zero. It can be easily seen that the first solution
of the purely elastic problem Eq. 352 can not satisfy this constraint. The second one can,
however. Writing the second solution in the form ✓ = ⇡2 and (`) = a` + b, the delta
function has the argument
✓ ◆
0 2 a(` `0 )
V (|r(`) r(` )|) = w sin . (355)
a 2
thus it follows that for the short range interaction potential one should have `0 ` = 2⇡n
a ,
where n is obviously an integer. Now we need to calculate the total contribution of
139 Ishimoto, Y; Kikuchi, N: op.cit.
141
this solution to the Hamiltonian Eq. 350. The first step in this direction is the explicit
calculation of the double integral
Z L Z s Z L
0 0
ds V (|r(`) r(` )|) d` = ds Ṽ (s).
0 0 0
with the interaction potential given by Eq. 354. A little caution is in order to compute
Ṽ (s). Using the well known properties of the Dirac delta function
X (s si )
(g(s)) = ,
i
|g 0 (si )|
0
where si are the zeros of g(s), and taking g(s) = 2
a sin a(s 2 s ) , we obtain that Ṽ (s) is
simply w times the winding number N (s) of solution Eq. 353, i.e. how many times
does this solution pass past the point r(s). Thus we get for the final Hamiltonian on this
level of approximation
Z L Z L
H[a, L, w] = 2 Kc 1 ˙ 2 2
ds (s) sin ✓(s) w ds N (s) =
0 0
⇣ ⇡ ⌘
= 12 Kc a2 L wN (L) L (N (L) + 1) . (356)
a
This integral can be evaluated easily if one makes a graph of the integrand and takes
into account the summation formula for the first N (L) 1 integers and the fact that
after the last wound the remainder in the length of the filament is L (2⇡/a)N (L). The
corresponding Hamiltonian for the other solution, given by a constant tengent vector t(`),
would simply be zero. By introducing the dimensionless parameters c = w/2Kc (L/2⇡)2
Figure 57. The structure of the model DNA toroidal aggregate and the corresponding
free energy of formaton. The model is idealized and does not contain any finite size ef-
fects. We delimit ourselves to the analysis of the N = 1 case, which is best approximated
by the model. After Ishimoto and Kikuchi, arXiv:cond-mat/0701737v1 [cond-mat.soft]
2007.
and x = aL/(2⇡), so that the winding number in the toroidal state is given by [x], the
Hamiltonian can be written as
 2
x [x]([x] + 1)
H[a, L, w] = wL [x] + = wL H̃[c, x], (357)
4c 2x
and can lead to multiple minima as seen from the figure, each minimum representing a
toroidal state with a different winding number. Let us concentrate on the case of a single
loop formation, i.e. on N (L) = 1. In this case the dimensionless Hamiltonian reads
 2
x 1
H̃[c, x] = 1+ , (358)
4c x
and has a minimum at xc = (2c)1/3 which leads to the scaling of the critical inverse
curvature radius of the toroid ac ⇠
= L1/3 . The value of the Hamiltonian at xc equals
2
Hc = 1.
(2c)1/3
Thus a transition from the coil state, that has no contribution from attractive short range
interactions, to a toroidal state comes about when the above Hamiltonian becomes nega-
tive, thus for c 4. Tori with higher winding numbers would be problematic to analyse
within the confines of the present simplified model since it does not take properly into ac-
count the finite size effects: the finite size of the DNA radius, as well as the finite range of
the interaction potential. Nevertheless the above simplified model has a heuristic value in
showing how the bending energy and the attraction conferred by short range interactions
conspire to give rise to a toroidal DNA condensate.
In order to get a more realistic model for the single DNA collapse especially in the
case of large winding numbers, a more detailed picture of the DNA-DNA interactions
should be taken into account. These interactions have been measured in the case of
various amounts of added CoHex3+ condensed DNA 140 . Complete force curves have
been measured for concentrations in 0.25 M NaCl salt, at 20 C. The added 0 and 20 mM
CoHex3+ force curves represent two extremal behaviors: the first shows a net repulsion
at all DNA densities and the second one shows a condensation to an equilibrium density
corresponding to an interaxial DNA-DNA separation of about 2.7 nm.
8.2 DNA toroidal globule

For realistic sizes of the toroidal globules, that have typically an inner radius of ⇠ 30 nm
and an outer radius of ⇠ 60 nm, it is more appropriate to get their shape via a mesoscopic
model 141 . In a mesoscopic model of a nematic DNA aggregate one assumes that DNA
is condensed, i.e. exhibits a condensed phase with a finite DNA density and a finite size.
8.2.1 Free energy and its first integral
In the mesoscopic model the free energy of the aggregate is by assumption composed
of three terms: the bulk free energy that depends on the local density f0 (⇢) and is due
to direct DNA-DNA interactions as measured in the bulk experiment, the surface part
that depends on the surface area of the aggregate fS and the nematic deformation free
140 Rau DC, Parsegian VA., Direct measurement of the intermolecular forces between counterion-condensed
DNA double helices. Evidence for long range attractive hydration forces. Biophys J. 61 246 (1992).
141 Antonio Siber, Miran Dragar, V.Adrian Parsegian and Rudolf Podgornik, Packing nanomechanics of viral
genomes, EJP E(2008).

143
energy fD (⇢, ni , @l nk ) that depends on the director profile as well as the density. Here
one assumes that the aggregate is completely ordered and n is thus a unit director field.
Let us formalize the model by introducing different analytic expressions for various
contributions to the free energy of the aggregate and solve it for the density and director
field in the aggregate. The deformation free energy density is given by the Frank-Oseen
ansatz as
fD (⇢, ni , @l nk ) = 12 K1 (⇢)(r · n)2 + 12 K2 (⇢)(n · (r ⇥ n))2 + 12 K3 (⇢)(n ⇥ (r ⇥ n))2 ,

(359)
just as in the case of bulk DNA nematic liquid crystals. The bending elastic modulus of
the DNA nematic has the form
V (D)
K3 (⇢) = Kc ⇢(2) + = kB T LP ⇢(2) + K0 (⇢), (360)
D
where Kc = kB T LP is the intrinsic elastic modulus of the polymer with LP its
persistence length. ⇢(2) is the 2D density of the polymers perpendicular to the di-
rector, D is the average axial separation between DNAs and V (r) is the interaction
potential. The connection between the 3D monomer density ⇢ and ⇢(2) is given by
⇢ = (⇢(2) /Lbp ) ⇥ (L/Lbp ), where Lbp is the phosphate-phosphate separation along the
DNA and L is its total length. The other two terms in the elastic free energy are assumed
to be negligible. In the range of densities inside the toroidal aggregate the K0 term is
much smaller then the first one and remains only a small perturbation.The splay term,
proportional to K1 , is small because the splay modulus in nematic polymers is renor-
malized due to the coupling between local polymer density and local nematic director,
in such a way to penalize any appreciable splay deformation. Also the cylindrical spool
ansatz that is used below automatically satisfied the condition r · n = 0. The twist term
in Eq. 459, proportional to K2 , is also small on the length scale of a DNA aggregate. Of
the three terms in the Frank - Oseen elastic energy only the bending term is significant.
The total free energy of the aggregate can be thus written to the lowest order as a sum
of volume and surface parts
Z Z Z Z
F[⇢, ni , @l nk ] = f0 (⇢)dV + fD (⇢, ni , @l nk )dV µ ⇢ dV + fS dS, (361)
V V V @V
Here both, the density as well as the director field are by assumption position dependent,
⇢ = ⇢(r) and n = n(r). The first term in the above free energy represents the part
that depends only on the density and thus describes bulk DNA at density ⇢. The surface
part of the free energy fS can be assumed as proportional to the surface tension of the
aggregate, , since the aggregate presents a separate phase in the solution and the surface
tension correposnds to the surface free energy of the phase separation. The index 0 will
be used to stand for the bulk properties of DNA.
The thermodynamic equilibrium conditions are obtained from the Euler - Lagrange
(EL) equations that are obtained from the variation of the free energy by taking into
account the fact that n is a unit vector, i.e. n · n = 1. The EL equations are the explicit
equilibrium conditions that determine the director and the density equilibrium profiles of
the DNA nematic aggregate: ⇢(r) and n(r). The EL equations have a first integral that
can be derived directly by rewriting the EL equations in the form of a total derivative. It
is of the form
p0 (⇢) + fD (⇢, ni , @l nk ) = p ⌘ 0. (362)
valid anywhere within the aggregate. Here p0 (⇢) is the osmotic pressure of DNA at
bulk conditions, where the deformation free energy is zero, evaluated at density ⇢ and
p is the total osmotic pressure. The total osmotic pressure of DNA within the sponta-
neously formed aggregate, brought about by adding condensing agents like Co(Hex) or
polyamines to a dilute solution of DNA, has to be zero. The DNA density profile can
then be calculated by assuming an appropriate ansatz for the director field, based on the
assumed geometry of the aggregate, and then solving Eq. 362 for the density field. Eq.
362. Obviously, since the elastic energy can only be positive, Eq. 362, has a solution
only if the bulk osmotic pressure p0 (⇢) has a negative branch.
Figure 58. Right: a schematic representation of the toroidal geometry and microscopic
structure of the toroidal DNA aggregate(right). Adapted from P. Linse.
In fact the lowest order solution in the elastic deformation follows simply from
p0 (⇢c ) = 0, (363)
assuming that the bulk DNA collapses at a critical density ⇢c . Making use of the general
thermodynamic relation 142 ⇢µ f0 = p(= 0) which, combined with the first integral
p0 + fD = p , yields
f0 (⇢) + fD (⇢, ni , @l nk ) ⇢µ = p0 , (364)
one can write the complete free energy of the aggregate Eq. 361 in two equivalent forms
Z I Z I
F= p0 (⇢c )d3 r + fS dS = fD (⇢c , ni , @l nk )d3 r + fS dS (365)
V @V V @V
to the lowest order in the density. Additionally a constraint on the length of the DNA in
the aggregate L (and consequently its volume) has to be imposed, of the form that the
length of the DNA within the aggregate is fixed
Z Z
L 1
N= ⇢(r) d3 r = = ⇢(2) d3 r (366)
V Lbp Lbp V
where N is the number of base pairs in the DNA molecule and we made the obvious
connection between the 2D and 3D DNA densities, ⇢ = ⇢(2) /Lbp , where Lbp is the
142 A. Losdorfer, unpublished (2010).
145
length of a single DNA base pair. When minimizing the functional Eq. 365, we have
to take into account the above volume constraint by adding an additional term equal to
Lagrange multiplicator ⇤ , times Eq. 365. This is then the final free energy functional
that needs to be minimized.
8.2.2 Nematic field configuration
A reasonable assumption for the nematic director profile corresponding to the toroidal
shape would have to be consistent with cylindrical symmetry appropriate for a toroidal
aggregate and should only have a non-zero equatorial component with respect to the z
axis, which is also the axis of symmetry of the aggregate, n̂ = ê , where we have
introduced cylindrical coordinates (r, , z) . In these coordinates,
1
r ⇥ n̂ =
êz , (367)
r
1 1
n̂ ⇥ (r ⇥ n̂) = ê ⇥ êz = êr . (368)
r r
The elastic energy Eq. 459 in the limit of no splay and no twist deformation, which is an
appropriate limit for polymer nematics in small enclosures, can be written as
1 ⇢c (2)
fD (⇢, ni , @l nk ) = K3 (⇢)(n ⇥ (r ⇥ n))2 ' 12 kB T Lp 2 (369)
2 r
Furthermore we assume that the surface free energy equals a constant surface tension
of the aggregate fS = . Thus we can write the free energy functional in the form 143
Z Z I
1 ⇢c (2) 3 (2) 3
F = kB T Lp 2
d r ⇤ ⇢c d r + dS, (370)
2 V r V @V
with the constraint Eq. 366. The only effect of the surface term in the above free energy
is to chose that toroidal form of the aggregate that corresponds to the minimal surface
area. We note again that the above analysis is first order in the deformatin energy and
does not take into account the density changes as a result of elastic stresses.
Parameterizing now the shape of the toroid by h(r) as coordinate of the upper half
of the toroid’s shape in z direction, bounded by the inner and the outer radiu ri and ro
respectively, one remaina with the following form of the free energy
Z ro ✓ p ◆
1 (2) h(r) (2)
F = 4⇡ k B T ⇠ p ⇢c ⇤⇢c h(r)r + r 1 + h 02 dr, (371)
ri 2 r
with the length constraint
Z ro
L = 4⇡ ⇢(2)
c h(r)rdr . (372)
ri
Introducing now the dimensionless parameter
↵ = 4 V 1/3 /(⇢c (2) Lp )
the minimization of the above free energy leads to solutions h(r) that can be parameter-
ized by ↵. Some DNA toroids for various values of ↵ are presented in Fig. 59. Obviously
the smaller the ↵ the more the toroid looks like a perfect torus.
143 J. Ubbink and T. Odijk, Europhys. Lett. 33, 353 (1996).
Figure 59. The solution of the minimization problem for the free energy ansatz Eq. 472
for different values of the parameter ↵ = 0.15, 0.5, 20 (outer, middle and inner toroid).
The outer radii are approximately 160 nm, 120 nm and 50 nm, respectively. Graphics
courtesy of A. Siber.
8.2.3 Scaling of torus size
Obviously for smaller ↵ corresponding roughly to a smaller toroid volume we get tori
with larger radii but thinner then in the case for large ↵. The DNA toroid diameters
observed in cryoelectron micrographs of phage DNA 144 are about 90 nm outer and 30
nm inner radius which corresponds to the case between ↵ = 0.5 and ↵ = 20 on Fig. 59.
One derive two limiting forms for the scaling of the inner and outer radii of the torus with
its volume, or equivalently with the amount of DNA packed inside them 145 . We refer to
them as thick and thin toroids.
For thick toroids that correspond to large ↵ values, the surface tension energy domi-
nates and the toroids must resemble more and more a sphere for which the outer radius
1/3
should scale as V 1/3 , in fact r0 = ((3V )/(4⇡)) , while the inner radius should go
to zero. From Fig. 60 it can be seen that in fact the inner radius does increase, but only
logarithmically with V , while the outer radius increases as V 1/3 for sufficiently large vol-
umes (V > 3.97108nm3 for this choice of .), as discussed above. For smaller volumes,
the increase is slower.
In the opposite limit of large ↵ values we are in the thon toroid scaling, corresponding
to the dominant volume elastic energy. In this regime the toroid becomes very thin since
the surface energy cost is smaller. In that case the inner and outer radius are comparable
and scale as V 1/5 . The dependence on volume the slows down for smaller ↵. The
approach to the V 1/5 limiting scaling for both the inner as well as the outer torus radii is
clearly seen form exact numerical solution of the problem
144 Nicholas V. Hud and Igor D. Vilfan, Toroidal DNA condensates: Unraveling the Fine Structure and the
Role of Nucleation in Determining Size, Annu. Rev. Biophys. Biomol. Struct. 34 295 (2005).
145 M. R. Stukan, V. A. Ivanov, A. Yu. Grosberg, W. Paul and K. Binder, Chain length dependence of the
state diagram of a single stiff-chain macromolecule: Theory and Monte Carlo simulation, JOURNAL OF
147
Figure 60. Scaling of the torus size with the enclosed volume for slim and thick torus.
For thick torus the outer radius scales as V 1/3 , while the inner radius varies only loga-
rithmically. Inner and outer radii of the toroid dependence on volume for = 0.1nm 2 .
The thick and this torus scaling is clearly visible. Graphics courtesy of A. Siber.
8.3 Nematic LC transition in a DNA solution

Let us consider a semiflexible chain with persistence length LP whose segments in-
teract via a non-local nematic potential that depends not only on the separation be-
tween segments at r(`) and r(`0 ) as well as on the angle between the two segments
as V (ṙ(`), ṙ(`0 ); r(`), r(`)0 ), with t(`) = ṙ(`) the tangent vector 146 . The mesoscopic
Hamiltonian of the chain is thus given by
Z L ✓ ◆2 Z LZ L
d2 r(`)
H[r(`)] = 1
2 kB T LP d` + 1
2 V (ṙ(`), ṙ(`0 ); r(`), r(`)0 )d`d`0 .
0 d`2 0 0
(373)
Again the tangent vector is a unit vector while the partition function is given by
Z ✓ ◆2
H[r(`)] dr(`)
⌅= D[r(`)] e with = ṙi (`)ṙi (`) = 1. (374)
d`
CHEMICAL PHYSICS VOLUME 118 3392 (2003).

146 A.M. Gupta and S.F. Edwards, J. Chem. Phys. 98 1588 (1993).
The explicit dependence of the non-local nematic interaction potential is assumed to be

of the form
V (ṙ(`), ṙ(`0 )) = |ṙ(`) ⇥ ṙ(`0 )|2 w(|r(`) r(`0 )|). (375)
We have disregarded the possible non-orientationally dependent part of the potential that
however would not play a significant role in what we are to do below. The form of the
nematic potential is approximate and takes into account the orientational interaction to
the lowest level in the local orientation of the two segments at ` and `0 . The argument of
nematic part of the Hamiltonian can be furthermore decomposed into
|ṙ(`) ⇥ ṙ(`0 )|2 = ṙi (`)ṙa (`)ṙj (`0 )ṙb (`0 ) ( ia jb ib ja ) . (376)
The exact evaluation of the partition function for this particular interaction Hamiltonian
is hopeless. We have to introduce first of all an alternative representation of the partition
function via s.c. auxiliary fields and then take recousre to additional approximations to
obtain a solvable model that would allow for an approximate solution.
Following closely Gupta and Edwards we now introduce two local variables: the
local monomer density ⇢(r) and the local orientation tensor ij (r) as
Z L Z L
3
⇢(r) = (r(`) r)d` ij (r) = ṙi (`)ṙj (`) 3 (r(`) r)d`. (377)
0 0
By construction ii (r) = ⇢(r). The orientational tensor is the measure of the ordering
of the polymer chain and in the ordered state the thermodynamic average h ij (r)i 6=
ij h⇢(r)i. In the ordered i.e. nematic phase the three eigenvalues of the orientation
tensor are not identical.
With the constraints Eq. 377 and Eq. .374 the Hamiltonian Eq. 373 can now be
transformed via non-local tensorial Lagrange multiplier that would take into account the
definition of the orientation tensor. Thus we get an effective Hamiltonian of the form
Z
3
H[r(`), ij (r), ij (r)] = H[r(`)] ij (r) ij (r)d r +
ZZ ⇣ ⌘
+ 2 1
d3 rd3 r0 ii (r) w(|r r0 |) jj (r0 ) ij (r)w(|r r0 |) ij (r0 ) ,
(378)
with
Z L ✓ ◆2
1 d2 r(`)
H[r(`)] = 2 kB T LP d` +
0 d`2
Z L
+ (r(`)) (ṙi (`)ṙi (`) 1) d` +
0
Z L
+ ij (r(`))ṙi (`)ṙj (`)d`. (379)
0
Inspection of the above Hamiltonian shows that it can be evaluated explicitly if one as-
sumes that the two auxiliary Lagrange fields are constant in space, i.e. (r(`)) =
and ij (r(`)) = ij . This approximation is very similar to the mean-field approxima-
tion standardly used in statmech. The idea would now be to evaluate the trace over the
149
polymer degrees of freedom in Eq. 374 explicitly, while the rest is dealt with on the
mean-field level.
The part of the partition function that can be evaluated explicitly is obtained by intro-
ducing the Rouse modes , i.e. the Fourier modes of the chain as
Z 1
dQ
r(`) = r(Q) ei`Q . (380)
0 2⇡
The Fourier modes of the chain can be introduced only in the limit of an infinitely long
chain. If this would not be the case one would have to deal with a discrete Fourier series.
Since H[r(`)] is a quadratic functional of the Fourier modes r(Q) the trace over these
modes can be done explicitly and leads too
Z Z
L X L
kB T log D[r(`)] e H[r(`)] = dQ log LP Q4 + Q2 ( + ↵ ) =
2⇡ ↵ 0
s Z
L X 2( + ↵ ) L L
= + dQ log LP Q4 .
2 ↵ LP 2⇡ 0
(381)
where the index ↵ runs over all the eigenvalues of the matrix ij . The last term in the
above equation can be dropped since it amounts to only an additive constant. Assuming
furthermore that the range of the nematic interaction is small, we can approximate w(|r
r0 |) = w 3 (r r0 ). This gives us finally the free energy of the form
s 0 !2 1
L X 2( + ↵ ) X X X
1 @ 2A
F= L ↵ ↵V + 2 w ↵ ↵ V,
2 ↵ LP ↵ ↵ ↵
(382)
where V is the total volume of the system. The equilibrium values of the fields ij , ij
and are now obtained via minimization of this free energy. For the orientation tensor
we now take the ansatz
0 1
a+b 0 0
c B C
B
ij = @ 0 a b 0 C. (383)
ij
3 A
0 0 2a
Here c is obviously the isotropic concentration since ii = c and a, b are a measure of

projected monomer concentrations in a given direction over its isotropic value. Minimiz-
ing Eq. 382 w.r.t. ↵ , ↵ and we obtain the following set of equations
!
X
↵ = w ↵ w ↵
↵
r
L 2 1
↵ = p
4V LP ( + ↵ )
r
1 2 X 1
1 = p . (384)
4 LP ↵ ( + ↵)
Figure 61. The dependence of the order parameter S on reduced density c/c? . Apart
from the nematic phase the theory also gives a discotic phase, characterised by the neg-
ative values of the order parameter.
From the last two lines it follows that

X L
↵ = ii =c= . (385)
↵
V
These equations have different solutions. We chose the one representing the uniaxial
nematic phase described by a 6= 0 and b = 0. For this particular phase the orientational
tensor is of the form
0 1
1 S 0 0
cBB 0
C
C,
ij = @ 1 S 0 A (386)
3
0 0 1 + 2S
where the eigenvalue perpendicular to the nematic axis, z in this case, is obviously doubly
degenerate, see Eq. 386 and we introduced S = 3 ac as the standard orientational order
parameter which can be defined also as 147
k ?
S=L = 1
2 3 < cos2 ✓ > 1 . (387)
⇢
Here ✓ is the angle between the nematic axis z and the polymer director. This order
parameter can have values in the range from -1/2 to 1. In a totally disordered system
where the polymer director points randomly in all possible directions, < x2 >= 13 ,
and the orientational order parameter vanishes, S = 0. In the completely ordered case,
S = 1 and < x2 >= 1, so that all polymer directors are parallel with the z axis. The
other extremum, S = 12 , corresponds to < x2 >= 0 so that all polymer directors are
within a plane perpendicular to the nematic axis, see Fig. 62.
147 A. Tkachenko and Y. Rabin, Coupling between Thermodynamics and Conformation in Worm-like Polymer
Nematics, Macomol 28 8646 (1995).

151
Figure 62. Schematic presentation of nematic ordering in a polymer solution with vari-
ous values of the nematic order parameter, from left to right: S = 1/2, 1/2 < S < 0,
S = 0, 0 < S < 1, S = 1. S = 1 corresponds to perfect order in the direction of the
nematic axis and S = 12 corresponds to ordering within the plane perpendicular to the
nematic axis. Graphics courtesy M. Kanduc.
The minimization equations Eq. 384 now lead to the following condition for the
orientational order parameter
2 2
(1 + 2S) (1 S) 27 c⇤
= = . (388)
(2 + S) 8 LP w c c
Only the solutions of this equations for which 12  S  1 so that all the eigenvalues of
ij are positive. One can now define a critical concentration for which the p
above equation
is first satisfied. This can be readily calculated to yield S ? = 14 ( 5 + 33) = 0.186
which then via Eq. 388 leads to the critical concentration
p
? 189 + 33 33 5.915
c = ' . (389)
64 LP w LP w
For concentrations below c? Eq. 388 has no solutions, so at the critical concentrations the
order parameter S jumps from zero to S ? and thus describes a first order nematic ordering
transition. The dependence of the order parameter on reduced density is given in Fig. 8.3.
For concentrations above the critical value of the polymer density the equilibrium state
of the system can be described with an average value of the orientational order parameter
and corresponds to a nematic ordering of the polymer solution. This only happens for
sufficiently stiff polymers such as DNA but not for very flexible one such as hyaluronic
acid.
8.4 Free energy of a DNA nematic phase

For a polymer nematic the Frank elastic free energy can be written in the form 148
Z h i
2 2 2
FN = 12 d3 r K1 (r · t) + K2 (t · (r ⇥ t)) + K3 ((t · r)t) +
Z ✓ ◆2
3 ⇢
+ 12 d rB , (390)
⇢0
148 Gregory M. Grason, Braided bundles and compact coils: The structure and thermodynamics of hexagonally
packed chiral filament assemblies, Phys. Rev. E 79 041919 (2009).

where t is the local director field of the polymers and K1 , K2 and K3 are the respective
splay, twist and bend elastic constants. We omitted the s.c. saddle-splay term since it
only contributes surface terms and we deal with an infinite system. To the Frank elastic
energy one needs to add also the free energy of isotropic compression quadratic in the
local density change since the nematic phase is a positional liquid and isotropic stresses
are the only deformations that it can sustain. B is the bulk modulus for compressions.
Delimiting ourselves to small deviations of the director field t(r) around its average
orientation assumed to be along the z axis we can write 149
t(r) ⇠
= ( tx (r), ty (r), 1) and ⇢(r) = ⇢0 + ⇢(r), (391)
and decoupling the volume integral into an integral along the average director of the
chains and the surface perpendicular to it, the free energy assumes the form
Z " ✓ ◆2 #
1 2 2 2 2 ⇢
FN = 2 d r? dz K1 (r? · t) + K2 (r? ⇥ t) + K3 (@z t) + B ,
⇢0
(392)
where r = (r? , @z ). For polymer nematics we now have to consider that the director
field and the density of polymers in the (x,y)-plane are coupled. Imagine a bundle of long
flexible rods all oriented in the same direction. If you squeeze the bundle and thus locally
change their average nematic director, the local density of the rods perpendicular to their
directior has to change too. This is only true for long molecules. Short nematogens can
change their director without any concomitant change in local density. This is the fun-
damental difference between long and short nematogens. If the polymers were infinitely
long and stiff the coupling is given by the continuity equation:
@ z ⇢ + ⇢0 r ? · t = 0 (393)
This constraint, however, is softened if the polymer has a finite length L or a small
persistence length Lp but in the following we assume that this constrained is satisfied
exactly in a DNA nematic. In general one can derive that @z ⇢ + ⇢0 r? · t equals the
difference between the number of polymer heads and tails 150 .
The above constraint can be satisfied identically if one assumes that
⇢= ⇢0 r ? · u and t = @z u. (394)
This decomposition allows us to write the free energy Eq. 392 in the form
Z h i
2 2 2 2
FN = 12 d2 r? dz K1 (r? · @z u) + K2 (r? ⇥ @z u) + K3 @z2 u + B (r? · u) ,
(395)
This form of the deformation free energy leads to the structure factor proposed by
Selinger and Bruinsma 151
⌦ ↵ ⌦ ↵
S(q? , qz ) = | ⇢(q? , qz )|2 = ⇢20 |q? · u(q? , qz )|2 =
149 H.H. Strey, V.A. Parsegian and R. Podgornik, Equation of state of polymeric liquid crystals, Phys. Rev. E
59 999-1008 (1999).
150 Meyer
151 J. V. Selinger and R. F. Bruinsma, Hexagonal and nematic phases of chains. I. Correlation functions, Phys.
Rev. A 43 2910(1991).
153
⇢20 q?
2
= kB T 2 + K q2 q2 + K q4 . (396)
Bq? 1 ? z 3 z
where we defined
2
K1 q ? + K3 qz2
K(q) = . (397)
kB T
In order to calculate the contribution to the free energy due to fluctuations in nematic
order we have to sum over all the density modes. To the lowest order in interactions the
nematic director modes do not give any contribution to the density dependence of the
free energy and thus to osmotic pressure.
The density dependence of the free energy stems primarily from the bulk compress-
ibility term in the free energy that modifies the density fluctuations. Integrating out the
Gaussian fluctuations in the density deviation from a homogeneous equilibrium, i,.e.
⇢(r? , z) = ⇢(r? , z) ⇢0 , one remains with the following from of the fluctuation free
energy ZZ 2
d q? dqz 2 2
1
F = 2 kB T log K1 q? qz + K3 qz4 + Bq? 2
. (398)
(2⇡)3
The problem here is that the above integral requires a cutoff and that the higher order
terms in q? are more important than the ones we have kept here. For a moment let us
assume that this free energy is valid and we may calculate
ZZ 2 Z
@F 1 kB T V q? dq? dqz q? ⇡ kB T V q 3 dq
=2 2 2 2 4 2 = 4 2 p q ? ? p .
@B (2⇡) K1 q? qz + K3 qz + Bq? (2⇡) 2 2 + 2 BK q 2
Bq? K1 q? 3 ?
(399)
The last integral depends essentially on the upper cutoff for q? = q?max and we obtain
0 1
@F V BK3 q ?max A
= kB T p F@ q , (400)
@B 4⇡ K12 BK1 2 BK23K1
where the function F (x) has been defined as

Z x 3/2 (
u du 1 p
p p 2 5/2
5x ;x ⌧ 1
F (x) = p =4 x 1 + x(2x 3) + 3 areasinh x = .
0 1+u 1 2
2x ;x 1
(401)
From here we obtain the two limiting forms of the free energy as
r s
kB T V 4 B 5/2 BK3
F' q + ... ; q?max ⌧ 2
5 ⇥ 23/2 ⇡ K3 ?max K12
r s
kB T V B 2 BK3
F' q + ... ; q?max 2 . (402)
16⇡ K1 ?max K12
Obviously the long-wavelength physics is very complicated and depends crucially on
the values of typical polymer length and the ratios of elastic constants. However it is
also dependent on the q? cutoff. We have to either eliminate the cutoff by including
higher order terms in the original Hamiltonian or choose a meaningful cutoff. Higher
order terms will capture the short-wavelength physics and remove the divergence 152 .
152 Strey
One can show that a consistent value of the cutoff has to be proportional to the Brillouin
zone radius q?max ' D ⇡
, where D is the effective separation between the polymers in
the nematic phase. This is a physically meaningful and appropriate cutoff because the
underlying macroscopic elastic model has, by definition, to break down at wavelengths
comparable to the distance between molecules.
For DNA arrays one one can show 153 that in the regime of the nematic (cholesteric)
phase we are always in the Eq. 411 limit. We would now have to derive the mesoscopic
elastic moduli from the microscopic interactions described via a pair potential F(R) (the
interaction free energy) between the segments of the macromolecules. At present this
program is too ambitious and we simply exploit the standard ansatz for the different
elastic moduli 154 expressed via the cell model free energy
K1 ' K2 ' F(R)/R

K3 ' ⇢0 KC + F(R)/R
✓ ◆
@ 2 F(V ) 1 @ 2 (F/L) 1 @(F/L)
B ' V = . , (403)
@V 2 4⇡ @2R R @R
where ⇢0 is the 2D density of the macromolecules perpendicular to their long axes, KC
is the elastic rigidity modulus of a single polymer molecule and we assumed that the
polymers have an average separation R between first neighbors 155 .
The macroscopic free energy Eq. 411 together with the values of elastic constants
Eq. 460 already points to the salient features of the fluctuation modified equation of
state. Obviously the thermal fluctuations make the free energy much longer ranged than
the underlying microscopic interaction potential. If the interaction potential decays ex-
ponentially with characteristic length , then the free energy Eq. 411 decays with four
times the characteristic length! The factor of four is a simple consequence of mesoscopic
elasticity.
8.4.1 Positional and Orientational Order in a Nematic Array
The orientational order fluctuations can be obtained by taking the decomposition of t

into a longitudinal and transverse forms. This leads us to the following expression
⌦ ↵ kB T kB T qz2
| t|2 = 2 2
+ 2 2 + K q2 (404)
K2 q ?+ K3 q z (K1 qz + B)q? 3 z
The longitudinal part of the director field fluctuations thus shows and effective wave
vector dependent splay elastic constant K1 ! K1 + Bqz 2 . The splay deformation
elastic constant thus diverges in the limit qz ! 0 which means that one can not impose
a uniform splay deformation on a polymer nematic composed of long chains. This is
obviously a fundamental difference between the polymer nematic and ordinary short
chain nematics. It comes about because a large splay deformation due to the coupling
Eq. 393 induces also a large density change which is energetically prohibitively costly. In
this respect the splay modes in a polymer nematic are simply frozen out and thus do not
153 Helmut
154 De Gennes C
155 Strey
155
contribute to the light-scattering intensity diminishing its value if compared to ordinary

nematics.
Just as in ordinary nematics the orientational fluctuations diverge approximately as
the inverse second power of the wave-vector. This has a pronounced effect on the ⌦ light-
↵
scattering from nematics since the scattered light intensity is proportional to | t|2 .
Ordinary short chain nematics thus appear turbid. In polymer nematics the orientational
fluctuations still diverge with the second power of the wave vector because of twist and
bend fluctuations even if the splay fluctuations are frozen out and are thus also turbid.
8.5 Free energy of a DNA hexagonal columnar phase

The hexagonal columnar phase of DNA, being also orientationally ordered, differs from
the nematic or the cholesteric phase in the fact that it can support shear stresses within
the plane perpendicular to the average nematic director and one should thus consider the
in-plane elastic energy as opposed to the isotropic 3D compression energy. The total
elastic free energy of the hexagonal columnar phase must thus be of the form
Z h i
2 2 2
FC = 12 d3 r K1 (r · t) + K2 (t · (r ⇥ t)) + K3 ((t · r)t) +
Z
⇥ ⇤
+ 12 d3 r ? u2kk + 2µ? uij uij . (405)
Here ? and µ? are the Lamé elastic constants that penalize isotropic and shear defor-
mations of the array in the plane perpendicular to the average nematic director t of the
hexagonal columnar phase and uij (⇢, z) are the components of the 2D elastic strain ten-
sor. Its components are given by the displacements of the DNA filament positions in the
(x, y) plane, u(⇢, z), and the projection of the local filament orientation, t(⇢, z), onto
the (x, y) plane, as
uij (⇢, z) = 1
2 @xi uj (⇢, z) + @xj ui (⇢, z) ti (⇢, z)tj (⇢, z) . (406)
The non-linear contribution to strain from filament tilt is a direct consequence of the in-
variance of the elastic energy under rotation about in axis in (x, y) plane and measures
the distances of closest approach between neighbor filaments perpendicular to the fila-
ment tangent 156 . The last term in the above equation, allowed by symmetry, contributes
higher orders to the free energy and will be ignired in what follows.
For small deformation of the assembly we again assume that
t(⇢, z) ⇠
= ( t? (⇢, z), 1) and t? (⇢, z) ⇠
= @z u(⇢, z), (407)
so that to the lowest order in the deviations from a homogeneous state the free energy
corresponding to small elastic fluctuations can be written as
ZZ h
2 2 2
FC =2 1
d2 r? dz K1 (r? · t) + K2 (r? ⇥ t) + K3 (@z t) +
⇤
+ ( ? + 2µ? )(r? · u)2 + µ? (r? ⇥ u)2 =
156 J. V. Selinger and R. F. Bruinsma, Hexagonal and nematic phases of chains. I. Correlation functions, Phys.
Rev. A 43 2910(1991).
ZZ h
2 2 2
= 1
2 d2 r? dz K1 (r? · @z u(⇢, z)) + K2 (r? ⇥ @z u(⇢, z)) + K3 @z2 u(⇢, z) +
⇤
+ ( ? + 2µ? )(r? · u)2 + µ? (r? ⇥ u)2 . (408)
From here one can derive the corresponding structure factor which is obtained by taking
into account Eq. 407 in the form
⌦ ↵ ⌦ ↵
S(q? , qz ) = | ⇢(q? , qz )|2 = ⇢20 (q? · u(q? , qz ))2 =
⇢20 q?
2
= kB T 2 + k T K(q)q 2 , (409)
( ? + 2µ? )q? B z
where we defined
2
K1 q ? + K3 qz2
K(q) = . (410)
kB T
Obviously the structure factor for a hexagonal columnar phase reduces to the one for
the nematic phase, Eq. 396, in the case of very long chains and if we identify B !
? + 2µ? .
The free energy corresponding to elastic fluctuations in the hexagonal columnar phase
can now be evaluated in analogy with the calculation in the nematic case performed in the
previous section. The only difference is that we have to take into account the longitudinal
and the transverse fluctuations in u since both contain the contribution of the interactions
embedded in the 2D elastic constants ? and µ? . This gives rise to two additive terms
in the free energy which now assumes the form
0s 1 s
r
kB T V @ 4 ( ? + 2µ? ) µ ? A 5/2 µ ? K3
F' + 4
q?max ; q?max ⌧ 2
5⇥2 ⇡ 3/2 K3 K3 K12
0s 1 s
r
kB T V @ ( ? + 2µ? ) µ? A 2 ( ? + 2µ? )K3
F' + q?max ; q?max 2 .
16⇡ K1 K1 K12
(411)
Again the long-wavelength physics of the hexagonal columnar phase is very complicated
and depends crucially on the values of the elastic constants as well as the q? cutoff. As for
the nematic case one can show that a consistent value of the cutoff has to be proportional
to the Brillouin zone radius q?max ' D ⇡
, where D is the effective separation between
the polymers in the hexagonal columnar phase. The above free energy as regards the
scaling with the effective separation D is thus very similar to the nematic case. The only
difference is that the interaxial spacings in the hexagonal columnar phase are smaller
then in the nematic phase.
The elastic moduli entering the above free energy can be guesstimated in the same
general way as in the case of the nematic phase, Eq. 460. The only difference is that now
the elastic constants for the 2D elasticity as opposed to the isotropic volume compress-
ibility contain directly the interaction potential between the molecules. This leads to the
same scaling form of the fluctuation free energy with respect to DNA density as in the
case of the nematic phase. This is no wander since the mechanism is the same.
157
The fluctuation free energy for a hexagonal phase clearly reduces to the free energy
of the nematic phase in the limit when the shear modulus of the hexagonal phase µ? goes
to zero.
8.5.1 Positional and Orientational Order in a Columnar Hexagonal Arrays
From the free energy of the columnar phase Eq. 408 one can now obtain the fluctuations
in the positional and orientational order in the columnar hexagonal phase. Going to the
Fourier space and taking into account the equipartition theorem one obtains for the mean
square fluctuation in the transverse displacement
ZZ 2 ✓
⌦ 2↵ d q? dqz kB T
|u| = 2 + K q2 q2 + K q4 +
(2⇡)3 ( ? + 2µ? )q? 1 ? z 3 z
◆
kB T
+ 2 + K q2 q2 + K q4 . (412)
µ? q? 1 ? z 3 z
⌦R ↵ ⌦ ↵
Here one has standardly V 1 u2 (r)d3 r = |u|2 . Again the long-wavelength
physics of the hexagonal columnar phase depends crucially on the values of the elastic
constants as well as the q? cutoff. A consistent value of the cutoff has to be proportional
to the Brillouin zone radius q?max ' D ⇡
, where D is the effective separation between
the polymers in the hexagonal columnar phase. One thus obtains 157
✓ q ◆
⌦ 2↵ kB T 2
|u| = p log 1 + L + 2 L + L +
4⇡ K1 ( ? + 2µ? )
✓ q ◆
kB T
+ p log 1 + T + 2 T + T2 . (413)
4⇡ K2 µ?
p
where we introducedp the cutoff related parameters: L = (K 1 )/ K3 ( ? + 2µ? ) q?max
and T = (K1 )/ K3 µ? q?max . Since for realistic systems the bulk elastic modulus
? + µ? is much bigger then the shear modulus, the size of the displacement fluctu-
⌦ations
↵ should be governed by the second term and obviously the fluctuations are finite
|u|2 < 1.
For
⌦ the ↵ nematic phase, being a liquid in terms of the positional order one of course
has |u|2 ! 1 and its behavior is fundamentally different from the behavior of the
hexagonal columnar phase, which si a solid in the directions perpendicular to the average
axes of the molecules.
The corresponding orientational fluctuations can be obtained in a similar way as in the
case of a nematic phase. One again decouples the director fluctuations into a longitudinal
and tangential components and eventually derives the form
⌦ ↵ kB T qz2 kB T qz2
| t|2 = 2 + K q 2 + (K q 2 + ( 2 . (414)
(K2 qz2 + µ? )q? 3 z 1 z
2
? + 2µ? ))q? + K3 qz
Comparing this result with the nematic case, Eq. 404, it is obvious that in the hexagonal
columnar phase not only K1 is renormalized but also K2 . In fact we can obviously write
K2 ! K2 + µ ? q z 2 and K 2 ! K2 + ( ? + 2µ? )qz 2 .
157 Jonathan V. Selinger and Robijn F. Bruinsma, Hexagonal and nematic phases of chains. I. Correlation
functions, Phys. Rev. A 43 2910âĂŞ2921 (1991).

What are the consequences of orientational fluctuations in hexagonal columnar DNA

phases on light scattering? It is clear from the above expression that in the case of q? =
⇥qz , for ⇥ 6= 0, the fluctuations remain finite and the scattering intensity would thus not
diverge just as in the case of solid crystals and the hexagonal columnar phase should thus
not be turbid. This is true for any finite ⇥. However in the case of ⇥ ! 0 implying
q? ! 0, the fluctuations and thus the light scattering intensity diverges. This situation
corresponds to a strong specular reflection in the plane perpendicular to the long axes of
the molecules.
8.6 Line Hexatic DNA Phase

159
9 DNA ARRAYS
In aqueous solutions DNA often makes dense phases exhibiting various types of order.
For theses dense DNA phases in solution we will use the term DNA arrays. Dense
DNA phases usually have local hexagonal packing symmetry, see Fig.9, unless they are
crystalline. In that case the local packing symmetry is orthorhombic 158 . DNA arrays can
be induced by either of the two methods:
i. through attractive interactions mediated by polyvalent counter ions or
ii. by external osmotic pressure of osmoticants provided through depletion interac-

tions
In the first case, extensively studied by Livolant and coworkers 159 , the condensed phase
itself phase separates from the rest of the solution and is held together by counter ion-
mediated attractions at zero solution osmotic pressure. It is thus by definition in a spon-
taneous thermodynamic equilibrium state. We could say that the effective free energy
in this case is composed of a repulsive and attractive part so that the thermodynamic
equilibrium ensues as
@FR (VDN A ) @FA (VDN A )

F (VDN A ) = FR (VDN A ) FA (VDN A ) with =
@VDN A @VDN A
(415)
giving us the equilibium molecular volume.
Figure 63. The schematic local geometry of an array of DNA molecules with long
range orientational order and short range positional order at two values of DNA density.
Drawn approximately to scale.
158 Bernard Lotz, Frustration and Frustrated Crystal Structures of Polymers and Biopolymers, Macro-
molecules (2012).
159 Livolant F., Leforestier A. (1996) Condensed phases of DNA: Structure and phase transitions. Progress in
Polymer Sciences 21, 1115-1164.
Wednesday, November 24, 2010

DNA array formation by externally imposed osmotic stress, studied by Parsegian and
Rau 160 , is brought about by a demixing of DNA and a neutral polymer such as poly-
ethylene-glycol (PEG), that then acts like an osmotic agent. In the latter case the DNA
array is under osmotic stress provided by the excluded PEG. The free energy in this case
is given by
@FR (VDN A )
F (VDN A ) = FR (VDN A ) ⇧o VDN A with = ⇧o , (416)
@VDN A
where ⇧o is the osmotic pressure of the polymer rich sub phase.
There is also a third method that we will not introduce separately, that actually com-
bines the two previous ones and was used by Rau 161 to study the equation of state of
DNA in the case of polyvalent counter ions.
Though the equilibrium conditions in both cases look very different, they nevertheless
correspond to the equilibrium between repulsive and attractive parts of the free energy,
except that in the latter case the interactions are mediated by excluded polymers and
not by included polyvalent counterions. What is more, the second type of DNA arrays
obviously leads to a straight forward osmotic stress method of measuring the repulsive
part of the free energy by simply varying the externally imposed osmotic pressure ⇧o
and thus obtaining its equation of state.
In what follows we will describe some of the salient features of the equation of state
of a DNA array and connect them to theory where such a theory can be set up.
9.1 Osmotic stress method

Equation of state of DNA, i.e. the dependence of its osmotic pressure on the DNA den-
sity, provides essential thermodynamic data on DNA in solution. From the equation of
state one can reconstruct the interactions operating between DNA molecules as well as
its phase diagram. The osmotic stress method is essentially an osmotic balance method
where one balances out the (unknown) osmotic pressure of a macromolecular solution,
such as an aqueous DNA solution, with a solution of known osmotic pressure, such as
an aqueous solution of polyethylene glycol (PEG). Be varying the osmotic pressure of
PEG and concurrently measuring the density of the macromolecular subphase one can
get the equation of state directly. The osmotic stress method is usually applied by settin
the two phases apart with a semipermeable membrane that is permeable to water and
smaller ions, but is impermeable either to PEG or to macromolecules. Nevertheless, for
many macromolecules from which PEG readily phase separates, among them DNA, one
can use the osmotic stress method without the semipermeable membrane. A schematic
presentatio fo the osmotic stress method is given on Fig. 9.1.
The physical basis for using the osmotic stress method in order to asses the interac-
tions between macromolecules is straightforward. For the sake of the argument, let us
concentrate on two macromolecular surfaces in an aqueous solution. One can measure
the interactions between the two by simply changing their separation and mechanically
measuring the change of the force or pressure between them. An alternative to this is
160 Stanley, C., and Rau, D. C. (2011) Curr. Opin. Colloid Interface Sci., (2011).
161 D.C. Rau, V.A. Parsegian, Biophys. J. 61, 246 (1992).
161
Figure 64. A schematic presentation of the osmotic stress method, where the osmotic
pressure of the PEG subphase is balanced against the unknown osmotic pressure of a
macromolecular solution.
varying the amount of water between the two macromolecular surfaces by exchanging it
with the with the bulk compartment and again measuring the respective interaction pres-
sure. The two approaches are of course completely equivalent since the corresponding
change in Gibbs free energy dG in thermodynamic equilibrium has to be zero 162 . Thus
one obtains a thermodynamic identity between the change of mechanical work and the
change of chemical work
dG = pdV + µw dN = 0 wherefrom pdV = µw dN, (417)
where µw is the chemical potential of water, p is the osmotic pressure, dV is the change
of volume and dN the change in the number of water molecules between macromolecular
surfaces. The osmotic pressure of the macromolecular solution is of course just the rate
of change of energy with respect to the volume of all exchangeable species between the
two compartments, i.e. water. It also follows from the above relation that the osmotic
pressure of the macromolecular solution in chemical equilibrium with the PEG subphase
is given by
pVw = µw , (418)
where Vw is the volume of a single water molecule. One should note here that osmotic
stress is not equivalent to hydrostatic pressure where the mass of the solvent is kept fixed.
162 V. A. Parsegian, R. P. Rand, N. L. Fuller, and D. C. Rau, Osmotic Stress for the Direct Measurement of
Intermolecular Forces, in Methods in Enzymology, edited by L. Packer (Academic, New York, 1986), Vol.
127, p. 400.
In somotic pressure measurements, on the contrary, the solvent subphase is allowed to

exchange between the macromolecular solution and the PEG subphase.
Figure 65. Osmotic pressure of PEG aqueous solutions for various molecular weights of
PEG from 300 to 20000 Da. Data vs. theoretical equation of state, Eq. 419, composed
of a weighted sum of the van’t Hoff ideal osmotic pressure and des Cloizeaux semidilute
solution osmotic pressure. The fit is quite accurate and produces the value ↵ = 0.48 for
the weight coefficient in the osmotic equation of state. The two limiting slopes belong to
the first and the second term of Eq. 419. Adapted from J.A. Cohen, R. Podgornik, P.L.
Hansen, V.A. Parsegian, unpublished (2008).
In order to apply the osmotic stress method one needs to measure the osmotic pressure
of the stressing solutions quite accurately. The most commonly used osmoticant is PEG
and tables of its osmotic pressure are readily available. The osmotic pressure of PEG as a
function of its concentration can be readily represented in a rather simple form as a sum
of the van’t Hoff law valid for dilute polymer solutions and the des Cloizeaux law valid
for semidilute polymer solutions 163 . If N is the number of monomers in PEG, Mm is
the monomer molecular weight and Cp the PEG mass concentration
"✓ ◆ ✓ ◆9/4 #
9/5 RT Cp Cp
pN = ⇤
+↵ . (419)
Mm V Cp Cp⇤
Here of course R is the universal gas constant and T is the temperature. Cp⇤ is a character-
istic polymer concentration associated with the crossover between the dilute and semidi-
lute regimes, and ↵ is an undetermined numerical prefactor. Cp⇤ is traditionally taken as
163 J.A. Cohen, R. Podgornik, P.L. Hansen, V.A. Parsegian, unpublished (2008).
163
the polymer overlap concentration, i.e., Cp⇤ ⇠ N 4/5 /V , where V is the polymer partial
specific volume, at which the statistical polymer volumes per each chain start to overlap.
The fit of this analytic equation of state to the measured osmotic pressure of PEG
for various concentrations and various molecular weights is presented on Fig. 9.1. The
analytical representation of Eq. 419 is thus quite accurate. Balancing a solution of DNA
with a solution of PEG in chemical equilibrium with a known osmotic pressure, given by
Eq. 419, thus allows us to measure the equation of state for DNA. This method has been
used exhaustively to asses the validity of various theories of DNA-DNA interactions 164 .
9.2 Measured DNA Equation of State

9.2.1 Monovalent salts
Figure 66. Osmotic pressure of DNA aqueous solutions as a function of DNA concentra-
tion for various uni-valent (NaCl) salt concentrations. This is a composite graph showing
measurements done by different grioups as indicated by arrows. . The solid line repre-
sents the osmotic pressure obtained in the PB theory with only uni-valent counterions
present. Above log ⇧ of about 8, there is a transition into a crystalline state with very
small osmotic compressibility. The last two points at extremely high osmotic pressure
already correspond to the breaking of the DNA helix.
164 H.H. Strey, R. Podgornik, D.C. Rau and V.A. Parsegian: DNA-DNA Interactions, Curr. Opin. Struc. Biol.
8 309 (1998).
The outcome of these osmotic stress experiments on solutions of DNA can be summa-
rized in the following equation of state for DNA, Fig. 9.2.1, giving the osmotic pressure
of the DNA solution as a function of its concentration, CDN A , and the concentration of
salt in the bathing solutions, Cs , i.e. ⇧ = ⇧(CDN A , Cs ). In various regimes of the DNA
density the equation of state is governed by different physics: electrostatic interactions,
conformational fluctuations and short range water structural forces.
The main feature of the DNA equation of state is that the counterions and aqueous
solvent give a substantial contribution to the osmotic pressure. The first one is discerned
at very small amounts of added salt and the latter at very large amounts of added salt.
There is also a very pronounced specific effect of various counterions. Even in the
case of monovalent salts the magnitude of the osmotic pressure is starkly different for
e.g. Li+ counterion as opposed to N a+ counterion.
One of the salient features of the osmotic pressure dependence on the DNA density, or
equivalently on the average separation between the DNAs assumed to be locally packed
hexagonally is the strong repulsive components for very large densities. This strong re-
pulsion is attributed to the energy associated with a tight bonding of water molecules to
the highly hydrophillic surface of DNA and is usually referred to as the hydration force.
I. Langmuir 165 was the first to surmise that the hydration of (bio)molecular surfaces
is the most reasonable explanation why macromolecules remain separated by consider-
able distances and do not come into contact even at extremely large values of osmotic
pressure.
The hydration interaction is one of the reasons that the osmotic pressure values for
various ionic types and concentrations in the bathing solution all converge to comparable
magnitudes at high DNA concentrations. Though the osmotic pressure of counterions
alone would lead to the same conclusion, it does not follow the same scaling form as
observed experimentally. In addition, the fact that these large values of osmotic pressure
persist even for uncharged molecular surfaces at zero salt gives a strong evidence of
solvent mediated interactions. Though easily measure the hydration interaction eludes
simple understanding on the level usually achieved for electrostatic interactins even on
th emean-field level.
9.2.2 Polyvalent salts
For DNA arrays with a mixture of polyvalent and monovalent salts it is appropriate to
use a combination of methods i) and ii). If enough polyvalent salt is added, DNA simply
precipitates into a condensed array at an equilibrium spacing given by Eq. 415. At
smaller concentrations DNA does not precipitate and additional PEG osmotic stress can
be applied. Typically the iso-ionic strength lines in the equation of state look like van der
Waals - loops with a Maxwell construction.
Protamines are small, arginine rich peptides used to package DNA in vertebrate
sperm. Salmon protamine is 32 amino acids long with 21 arginines. The attractive force
between protamine-DNA helices closely resembles the forces mediated by much smaller
multivalent ions, as CoHex3+ .
165 I.J. Langmuir, The Role of Attractive and Repulsive Forces in the Formation of Tactoids, Thixotropic Gels,
Protein Crystals and Coacervates, J. Chem. Phys. 6 873 (1938).

165
8.0
7.5
(ergs/cm3)
7.0
6.5
6.0
log
0 mM NaCl
5.5 50 mM NaCl
150 mM NaCl
300 mM NaCl
5.0
500 mM NaCl
25 26 27 28 29 30 31 32
Dint (Å)
Figure 67. Equation of state for polyvalent counter ions. Left: CoHex3+ at various con-
centrations with monovalent NaCl salt. Right: 100 µM salmon protamine with different
values of monovalent NaCl salt concentration: 0 mM, 50 mM, 150 mM, 300 mM, and
500 mM.
Figure 3
9.3 Cell model electrostatics

Thursday, April 5, 12
In many colloidal systems, most notably in the case of ordered DNA phases, one sel-
dom deals with isolated molecules in ionic solutions. Quite often one has a phase of
densely packed macroions which complicates the problem of evaluating the electrostatic
interactions even further. A simple way around this problem is the polyelectrolyte cell
model [?], being a variant of the Wigner - Seitz model of electron in the crystalline
lattice, which substitutes the complicated colloidal geometry with a cell, containing a
single colloid. The effect of the rest is assumed to be mimicked3 by the cell wall where
the electrostatic potential should have a zero derivative by symmetry.
For a long cylindrical molecule such as DNA in a dense phase where molecules are
on the average oriented in one direction, a cylindrical cell model should capture the main
features of the molecular environment. There are nevertheless important caveats that one
has to be aware of in the context of the cell model [?]. The linearized PB equation in
cylindrical coordinates (r, ✓, z) reads
✓ ◆
1 d d
r = 2D , (420)
r dr dr
with the boundary condition at the inner wall, (i.e. at the surface of the central molecule
with a radius a) being
d e0
(r = a) = = , (421)
dr ✏✏0 2⇡✏✏0 ab
wher b is the length of the molecule per one elementary charge on the surface. For DNA
its structural charge would correspond to one charge per 1.7 Å. The boundary condition
at the outer surface of the cell, assumed to be located at r = R, is by symmetry

d
(r = R) = 0. (422)
dr
The radius of the cell R is obtained from the macromolecular density of the system. If the
density of the macromolecules is nM , then nM 1 = ⇡(R2 a2 )b. The macromolecular
density thus determines the radius of the cell.
Obtaining a solution of the linearized Poisson - Boltzmann equation in cylindrical
cell model is quite straightforward and leads to the electrostatic potential that can be
expressed via the cylindrical Bessel functions K0 (x) and I0 (x) as
e0 K0 (D r)I1 (D R) + I0 (D r)K1 (D R)
(r) = . (423)
2⇡✏✏0 b (D a) K1 (D a)I1 (D R) I1 (D a)K1 (D R)
From here we get for the radial component of the electrostatic field the expression
d e0 K1 (D r)I1 (D R) I1 (D r)K1 (D R)
Er (r) = = . (424)
dr 2⇡✏✏0 ba K1 (D a)I1 (D R) I1 (D a)K1 (D R)
One should note here that for a single cylinder, thus in the limit of R ! 1, the elec-
trostatic potential and the electrostatic field reduce to
e0 K0 (D r) e0 K1 (D r)
lim (r) = and lim Er (r) = .
R !1 2⇡✏✏0 b (D a) K1 (D a) R !1 2⇡✏✏0 ba K1 (D a)
(425)
Additional terms in the cell model solution Eq. 423 and 424 are thus due to the finite
concentration of the DNAs and obviously depend on its density via the radius of the
outer cell wall R. One should note here that the linearized solutions of the PB equation
are quite accurate in the case that the salt concentration is not too small [?]. The non-
linear PB equation has an analytical solution [?] only for a single cylinder in infinite ionic
solution.
9.4 Osmotic pressure of a DNA array

The forces between macromolecules mediated by the equilibrium distribution of counte-
rions and salt ions between them can be obtained via the stress tensor at the outer surface
of the cell, which gives the force acting on this surface. This force per unit area of the cell
is obviously nothing but the osmotic pressure. The stress tensor contains the Maxwell
electrostatic part and the osmotic part [?],
X
ij = ✏✏0 Ei Ej
1 2
2 E ij kB T ni ( (r)) ij . (426)
i
The last term is simply the van’s Hoff ideal gas pressure, corresponding to the ideal gas
entropy in the free energy ansatz. The negative sign comes from the general continuum
mechanics argument that positive pressure should lead to a decrease in volume.
We can evaluate the stress tensor on any plane of cylindrical symmetry witin the cell.
In many cases it is simplest to take the stress tensor at the surface of the central macro-
molecule, in which case the forces are obtained via a contact theorem [?]. Equivalently
167
we can take the stress tensor at the outer wall of the cell where by symmetry the electric
field is zero and thus the stress tensor contains only the van’t Hoff part. Since the latter
case is simpler we thus find for the radial components of the stress tensor (all the other
ones vanish at the outer wall)
X
rr = kB T ni (r = R). (427)
i
The osmotic pressure ⇧(R) in the cell is via the mechanical equilibrium, equal to minus
the force per surface area on the outer wall, thus
X
⇧(R) = rr = kB T ni (r = R). (428)
i
We can use this expression to derive the form of the osmotic pressure in the case of the
linearized PB equation in a cylindrical cell. Remember that we have assumed that we
have only two mobile charged species: e1 = e0 and e2 = e0 , together with n1 = n+
and n2 = n , in equilibrium with a bulk reservoir where n01 = n02 = n0 . The osmotic
pressure difference between the cell and the bulk reservoir, which alone is measurable,
can be evaluated via the expression Eq. 428
⇧ = kB T (n+ (r = R) + n (r = R)) 2kB T n0 =
= 2kB T n0 (cosh e0 (r = R) 1) . (429)
This is the complete expression for osmotic pressure difference between the inside of the
cell and the bulk. Obviously here the only molecular species contributing to the osmotic
pressure are the mobile ions. The macromolecule itself does not contribute to the osmotic
pressure. We will see later that this point of view is not entirely correct.
Another approach to the evaluation of the osmotic pressure, independent of the cell
model, would be to calculate the complete equilibrium free energy F from Eq. 293 via a
volume integral over all the space available to the mobile charges
Z
F[nM ] = d3 r f (E(r), ni (r)) (430)
where the electrostatic field and the densities of the mobile charges are obtained from the
solution of the full or linearized Poisson - Boltzmann equation Eq. ??. This free energy
is of course a function of the concentration of the macromolecules nM . The osmotic
pressure in the system would then be obtained from the standard thermodynamic relation
✓ ◆
@F(nM )
⇧= . (431)
@VM T,µ
Constant temperature and chemical potential of the mobile charged species are of course
assumed in the above expression. In the case that the macromolecular solution is ordered
and exhibits certain symmetries, this expression can be simplified even further. If we
again assume that the effects of the packing symmetry of the molecules can be captured
by a cylindrical cell model of radius R and that the length of the molecules is L, we can
write for the osmotic pressure
✓ ◆ ✓ ◆
@F(nM ) @(F(R)/L)
⇧= = . (432)
@VM T,µ 2⇡ R@R T,µ
Eq. 432 connects the expression for the osmotic pressure in the cell model with the
one obtained from the complete partition function. Discrepancies between the values for
osmotic pressure obtained from the two expressions are due to the approximate nature of
the cell model.
9.5 Electrostatic part of the osmotic pressure

Let us now investigate the osmotic pressure in the cylindrical cell model on the level of
the linearized PB equation. In this case Eq. 429 can be written as
⇧ = 2kB T n0 (cosh e0 (r = R) 1) =
2 2
= 2kB T n0 ( e0 ) (r = R) + . . . = 2D ✏✏0 2
(n = R) + . . . , (433)
and represents the osmotic pressure difference between the wall of the cell and the bulk
reservoir. Using now the solution of the linearized PB equation Eq. 423, we end up with
the following expression for ⇧
2
⇧= K02 (D R) p2 (D R, D a), (434)
✏✏0 K12 (D a)
where we introduced the correction factor p(x, y) as
K1 (D R)I0 (D R)
1+ I1 (D R)K0 (D R)
p(D R, D a) = K1 (D R)I1 (D a)
. (435)
1 I1 (D R)K1 (D a)
The factor p2 (D R, D a) obviously represents the effect of the finite concentration of
the macromolecules, i.e. DNA, or in other words the effect of the walls of the cell. For
small DNA densities this correction factor goes to unity, since in that limit only the first
neighbors of the central molecule are important. This would be equivalent to taking only
the first neighbors into account in the evaluation of the free energy.
Eq. 434 for osmotic pressure is relatively complicated. In order to avoid the details
that will later prove to be irrelevant, we investigate only its asymptotic form, valid for
large values of D R or small values of the macromolecular concentration. In this limit
we have p(D R, D a) ! 1 and the approximate form of the osmotic pressure can be
derived as
2 2
⇡ e 2D R
⇧⇠ K02 (D R) = , (436)
✏✏0 K12 (D a) ✏✏0 K12 (D a) 2 D R
where
p ⇡ we have taken into account the asymptotic form of the Bessel function K0 (x) !
2 exp( x)x
1/2
. From the expression for osmotic pressure Eq. 432 we can now
derive the interaction free energy per unit length between the central molecule and its
neighbors. We obtain
⇡2 2
2D R `B 2D R
F(R)/L ⇠ e = kB T e . (437)
2 ✏✏0 2D K12 (D a) b2
Here wepintroduced an effective separation b between charges along the cylinder 1/b =
⇡ D / 2e0 K1 (a/ D ) [?] ( for DNA b ⇠ lP O4 , where lP O4 ⇠ 1.7 Å, is the separa-
tion between the phosphate charges along DNA [?]). This expression derived via the
169
cell model with cylindrical symmetry is very close to the expression derived via a pair-
interaction energy evaluated on the linearized PB level [?]. This is not surprising since
apart from geometric factors the cell model and the pairwise free energy should give the
same result at vanishing macroion concentrations.
In the case of very large surface charges the nonlinearities of the Poisson-Boltzmann
equation effectively change the surface charge density entering the above equation Eq.
437 via b ! `B while leaving the separation dependence largely unaltered [?]. Some-
times these non-linearity effects, that become pronounced depending on whether the s.c.
Manning parameter ⇣ = B /b is larger or smaller than one, would be referred to as Man-
ning condensation [?].
9.6 Equation of state of a DNA array

An equation of state in general means a connection between the osmotic pressure ⇧ and
the macromolecular density nM of an assembly
⇧ = ⇧(nM ) (438)
and can in principle be obtained exactly via a complete statistical mechanical treatment
of a solution composed of cylindrical macromolecules and the bathing medium. Since
this is usually not feasible, a helpful shortcut is to evaluate the osmotic pressure in the
cell model, that mimmicks the finite meacromolecular density as explained above.
The system that we are studying is composed of macomolecules that have intermolec-
ular as well as intramolecular degrees of freedom since they are usually not infinitely
rigid. A most naive approach to the equation of state would be to simply forget the in-
tramolecular degrees of freedom, assume that the macromolecules are ideally rigid and
that they assemble into a crystal of hexagonal symmetry with perfect positional order. In
this case the osmotic pressure of such a system can be obtained via Ewald force sum-
mation (in the case of short range interactions even this is dispensable) involving the
intermolecular potentials leading directly to the equation of state or again via the cell
model; the latter approach being certanly simpler then the Ewald summation since it is
effectively a single particle model. The osmotic pressure, i.e. the equation of state, in the
cell model would be given by
✓ ◆
@(F(R)/L) `B
⇧(nM ) = F(R) = kB T 2 e 2D R . (439)
2⇡ R@R T b
L is the length of the molecules in the array. As already stated in writing this free en-
ergy we assumed that the van der Waals interactions as well as the short range, non-
electrostatic structural nteractions do not make an essential contribution to the equation
of state. For the latter this is true only at not very high ionic concentrations and not too
high concentration of the macromoleculs (for details see [?]).
We can now take this equation of state and compare it to experiments performed
on DNA [?] at various solution conditions. After performing this type of exercise, one
is immediately convinced that something crucial is missing as there is practically no
correspondence (except at very high densities) between the experiment and this type of
simpleminded theory, see Fig. 68. In the case of DNA the magnitude of the surface
charge is taken at the Manning value [?], thus b = `B . Obviously this equation of
state underestimates the energetics of the system. In the following section we will try to
ammend it by taking into consideration also the intramolecular degrees of fredom of the
macromolecular ions.
Figure 68. A comparison of the equation of state Eqs. 439 with DNA experimental data
[?]. This form of the equation of state assumes that the molecules are infinitely rigid
and crystalline. The magnitude of the electrostatic part of the interaction is obtained by
assuming that the charge density of the DNA is correctly given by the Manning value [?].
Obviously there are large discrepancies on this level between the theoretical predictions
and actual data.
While the intermolecular degrees of freedom are taken into account through interac-
tion potentials described in the previous section, the intramolecular degrees of freedom
are usually treated within a mesoscopic elastic model that substitutes macroscopic elas-
ticity for the complicated short-range intramolecular potentials acting between different
segments of the macromolecules [?]. Elastic models for cylindrical macromolecules have
been well worked out [?]. The general idea is that the trace over all the microscopic de-
grees of freedom is assumed to result exactly in the mesoscopic Hamiltonian itself, that
one then uses in the partition function to evaluate the trace over mesoscopic degrees of
freedom.
Because in the regime of relevant densities the hexagonal array is either in the line
hexatic or the cholesteric phases [?] I can use the mesoscopic elastic Hamiltonians per-
taining to them. The goal here will be to combine the effects of thermally driven elastic
fluctuations of the macromolecules with the interactions between them and calculate their
combined effect on the equation of state. This approach is based on ideas first introduced
by Helfrich in the ’70’s and later worked out in detail by Lipowsky, Leibler and others
in the ’80’s [?], who showed in the context of membranes that thermal conformational
fluctuations can have a profound effect on interactions between flexible macromolecules.
We now use the form of the bare interaction free energy F(R) appropriate for a DNA
array Eq. 439. One can see that at all relevant densities K3 ' ⇢0 KC . We are now able
171
to fit the calculated equation of state obtained from Eq. 411 to the experimental equation
of state 166 , see Fig. 69. The values for the DNA bending rigidity and the Debye length
obtained from such a fit are comfortably within the expected range [?]. Let me just
mention here that the effective charge evaluated from the fit is about half the amount
expected on the basis of the Manning condensation theory.
Figure 69. A fit of the lowest order fluctuation equations of state, Eqs. 411, to DNA [?]
data. Dashed line - theory without fluctuations as on FIg. 68. Full line - theory with
conformational fluctuations of the molecules taken into accoumt on a harmonic level,
411. The value of the effective charge on the DNA surface is obtained from the fit to
experiment and is found to be about half the Manning condensation value.
A fundamental drawback of this formulation for the equation of state in an assembly

of flexible molecules is most clearly seen in the ansatz Eq. 460. The elastic moduli
are not really calculated on the same level as the free energy but are assumed to have a
form, that at least for the compressibility modulus, would be strictly valid only for rigid
molecules. The above formulation is thus not completely self-consistent and we will
make an attempt to improve it in the next section. The failure of this attempt will make
us aware of some fundamental properties of the nature of the positional order in DNA
arrays.
166 Strey
10 DNA ORGANIZATION IN CHROMATIN

DNA can be compacted with polyvalent counterions quite efficiently decreasing its vol-
ume by a factor of 105 . In the process of compaction the interplay of polycounterion-
mediated attraction and DNA elasticity conspires in order to bring about nicely ordered
structures of which the toroidal structure was already analyzed in detail. Apart from
toroidal condensates one can observe also spherical globules, rods, ratchets and mul-
titoroidal aggregates 167 . This type of compaction is observed in vivo specifically in
viruses, bacteria, prokaryotes in general and in eukaryotic sperm cells. The packing
agents in these cases are usually polyamines (e.g. spermine 4+, spermidine 3+) and
protamines.
Figure 70. Different morphologies of DNA - nanoparticle interactions. From mere dec-
oration of the DNA chain, to DNA wrapping around the nanoparticles, depending on
their size, charge and ionic concentration. . Adapted from A. Zinchenko, D. Baigl, K.
Yoshikawa, Nanostructures and Organization of Compacted Single Chains of Polyelec-
trolytes, in Polymeric Nanostructures and Their Applications, Edited by: H.S. Nalwa
(2006).
A radically different mechanism of DNA compaction is used in eukaryotic cells.

In this case the polycounterions are more akin to charged nanoparticles with their size
playing a crucial role in the compaction mechanism.In vivo DNA is condensed via the
positively charged histone proteins that self-assemble into nanostructures such as nucle-
osomal core particles. In vitro other types of condensing nanoparticles such as cationic
micelles, globular proteins, dendrimers and matallic nanoparticles 168 have been proven
to condense DNA just as well. The size of the condensing nanoparticles is however cru-
cial in determining the morphology of the condensate. Small nanoparticles of size 3 -
4.5 nm merely decorate DNA in aqueous solution. Larger nanoparticles of size 7 - 8
nm induce DNA to wrap around them but the details depend strongly on the amount of
charge on the nanoparticle, their size and the salt concentration in the bathing solution.
The schamtics of the different types of DNA - nanoparticle interaction are presented in
Fig.
167 A. Zinchenko, D. Baigl, K. Yoshikawa, Nanostructures and Organization of Compacted Single Chains of
Polyelectrolytes, in Polymeric Nanostructures and Their Applications, Edited by: H.S. Nalwa (2006).
168 Ibidem.
173
Figure 71. A cartoon view of packing of nucleosomes along the chromatin fiber . Adapted
from Karolin Luger, The tail does not always wag the dog, Nature Genetics, 32 221
(2002).
Wrapping of DNA around charged nanoparticles is particularly germane for under-

standing of the DNA compaction in chromatin. In a typical mmetaphase eukaryotic
chromatin the compaction ratio of DNA is somewhat smaller then in prokaryotes and
viral capsids, reaching about 103 . Furthermore the compaction process in chromatin is a
multilevel process where each of the levels corresponds to a different type of compaction
mechanism.
10.1 Nucleosomes
Nucleosomal particles contain unfolded parts of the histone protein chains in the vicinity
of the N-terminal that can not be detected by X-ray scattering and dangle out of the
compact NCP structure in a disordered way. The N-tails exit the protein core of the
NCP most frequently through the minor groove of the DNA wrapped around the histone
octamer. The four H3 and H2B N-tails, being also the longest ones, exit through the
channels formed by the minor grooves of the wrapped DNA and three of the H4 and
H2A tails exit through the minor grooves on either top or bottom of the NCP. Apart from
providing stabilization for the nucleosome structure, these tails, being rich in arginines
and lysines and thus highly basic, accunting for about 50 % of the positive charges of the
whole histone octamer that contain in all about 300 charges, are at least in part available
for longer ranged electrostatic and polyelectrolyte bridging interactions that have been
shown to contribute significantly to the interactions between NCPs in solution. The exit
positions of the NCP tails from the histone core are not isotropically distributed along
the circumference of the nucleosome particle wrapped with DNA but show a pronounced
polar symmetry congruent with the location of the dyad axis which is itself connected
with the entry-exit point of the DNA.
While N-tails associated with H2A and H2AC histones have exit points on the dyadic
symmetry axis, the N-tails associated with the H3, H4 and H2B histones are distributed
symmetrically on both sides of this axis with the center of mass of the N-tail charge
displaced slightly towards the dyadic entry-exit point of DNA. The exit points of the
N-tails are located partly on top and partly on the bottom of the NCP. While it is well-
neigh impossible to calculate the detailed spatial distribution of charge on the disordered
and flexible N-tails it is clear that it must have a pronounced polar structure that could
be quantified by the dipolar and the uniaxial component of the quadrupolar moment.
Though the details of the interaction between two closely apposed NCPs are difficult to
fathom several general conclusions follow based on the symmetry of the mobile charge
attached to the N-tails. Since the longest N-tail have exit points close to the dyadic
position of the NCP it seems reasonable to assume, being also consistent with what we
know about the polyelectrolyte bridging interaction in general, that at small separations
between two coplanar NCPs these longest N-tails would tend to repel most and thus exert
forces that would tend to orient the two NCPs in antiparallel directions.
10.2 Positioning of nucleosomes along DNA

Genome wide experimental mappings 169 of nucleosome occupancy along the genome
point to a very patchy landscape, with regions that are completely nucleosome-free
flanked by either positioned nucleosomes with a well defined periodicty of ⇠ 165 bp
or disordered regions with no apparent nucleosome periodicity. The nucleosome de-
pleted regions in fact correlate with regions active in functional and structural protein
regulation.
It seems in fact 170 that the nucleosome depleted regions induce periodic ordering
of nucleosomes in the intervening regions through a non local effect induced by the
boundaries of the depleted regions. This kind of ordering can be observed clearly in a
one-dimensional system of hard rods - the Tonks gas - confined between hard boundaries.
Similar mechanism might also drive the ordering of the nucleosomes along a 1D stretch
of genome termined by a region, either with stably bound proteins and/or stably bound
nucleosomes on specific sequences, that prevent any reordering of the nucleosomes along
the genome. In yeast these nucleosome depleted regions coincide with transcription start
and termination sites. If we define the nucleosome occupancy as
Y (`) = log ⇢(`),
where ⇢(`) is the nucleosmoe density along the genome at the position `, then its form
169 In the case of yeast genome G.C. Yuan et al., Genome-Scale Identification of Nucleosome Positions in S.
cerevisiae, Science (2005) 309 626-630.
170 G. Chevereau et al., Thermodynamics of Intragenic Nucleosome Ordering, PRL 103, 188103 (2009).
175
Figure 72. Average in vivo nucleosome occupancy Y (`), aligned on the first flanking
nucleosome (+1) downstream the transcription start and (-1) upstream the transcription
termination sites. Adapted from G. Chevereau et al., Thermodynamics of Intragenic
Nucleosome Ordering, PRL 103, 188103 (2009). Red line.
around transcription start and/or termination sites shows a typical decaying order profile
as shown on Fig. 72.
10.2.1 Tonks gas with soft boundaries
The form of the nucleosome occupancy profile suggest a simple possible model that
would describe the ordering of nucleosomes close to depleted regions. The model ef-
fectively coincides with the grand canonical 1D Tonks gas 171 , except that the boundary
potential can be soft as opposed to hard wall type. We thus allow for all the nucleosomes
to excganhe with a "bulk" at a specified value of the chemical potential.
The model system is composed of N + 2 nucleosome particles on a line interact-
ing through hard sphere potentials with the core diameter equal to a. Additionally the
boundary particles ”0” and ”N + 1” interact with the boundaries at R = 0 and R = L
through soft potentials W (R0 ) and W (RN +1 L). We shall treat these two particles as
separate entities, belonging to the walls rather than to the intervening nucleosome fluid
so that only N particles are allowed to exchange with the bulk reservoir.
Particles on a 1D line can be ordered over the subregion R : R0 > R1 > R2 >...>
RN > RN +1 . The partition function of this system of N hard spheres can therefore be
obtained in the form
Z
+1 Z
+1 Z
+1 Z
+1
W (R0 ) W (L RN +1 )
⌅N (L) = e dR0 dR1 ... dRN e dRN +1 ,
1 R0 +a RN 1 +a RN +a
(440)
171 Mathematical Physics in One Dimension, Eds. E.H. Lieb and D.C. Mattis (Academic Press, New York
1966).
P1
so that the grand canonical partition function is then ⌅(L) = N =0
N
⌅N (L). By
inspecting the above definition we can derive the grand canonical partition function in
the form
Z
+1 Z
+1
W (R0 ) W (L RN +1 )
⌅N (L) = e dR0 ⌅T (RN +1 R0 ) e dRN +1 , (441)
1 RN +a
where ⌅T (u) is the grand canonical partition function of a 1D Tonks gas on a segment
of total length u with hard boundaries that can be evaluated in the closed form 172
1
X N
⌅T (L) = (L (N + 1)a)N H(L (N + 1)a) (442)
N!
N =0
where is the absolute activity of the confined 1D nucleosome fluid. The above sum
is in fact finite with the upper bound of Nm , which is the largest integer satisfying the
inequality L (Nm + 1)a 0. The grand canonical partition function in this case is
therefore a polynomial in , with the degree of the polynomial depending on the value of
L. This property is a consequence of a simple physical fact: as we increase L more and
more particles can be packed inside the volume L a.
The corresponding distribution of density ⇢(R) of the nucleosome particles between
the two soft walls can be obtained from the standard definition as
N
X Z
+1 Z
+1
W (R0 ) W (L RN +1 )
⇢(R) = < (Ri R)>= e dR0 ⇢T (R) e dRN +1 .
i=1 1 RN +a
(443)
where again ⇢T (R) is the Tonks gas one-particle distribution function, given by 173
⌅T (R R0 ) ⌅T (RN +1 R)
⇢T (R) = . (444)
⌅N (L)
For a hard boundary potential mimicking excluding barriers at gene extremities, e W (R) =
(R), the standard confined Tonks gas that has been discussed in detail within the context
of nucleosome positioning 174 . The hard confining potential is however not particularly
realistic and it is maybe advisable to assume it is of a more soft, Gaussian form in order
to describe the interaction of nucleosoms with the exploded regions. In this case W (R) is
approximated in the form of a Gaussian W (R) = 12 KR2 where we can interpret K as an
effective one - dimensional "elastic modulus". Obviously this form approaches the hard
core result of Eq. 442 when K ! 1, which corresponds to the hard core interaction
175
.
172 A. Robledo & J.S. Rowlinson, The distribution of hard rods on a line of finite length, Molecular Physics58
711 (1986).
173 A. Robledo & J.S. Rowlinson, ibidem.
174 G. Chevereau et al., Thermodynamics of Intragenic Nucleosome Ordering, PRL 103, 188103 (2009).
175 For the Gaussian choice of the boundary potential there exists a special scaling relationship connecting the
magnitude of the "effective" elastic constant and the extension of the system in the form ↵N +1 ⌅N (L, K, a) =
⌅N (↵L, ↵K2 , ↵a) or equivalently ↵⌅(L, K, a, ↵ ) = ⌅(↵L, ↵K2 , ↵a, ).
177
Figure 73. Left: A stretch of DNA bordered by two hard walls and µ = 1.5kB T : (a) box
large enough to shelter n = 5 nucleosomes; (c) larger stretch allowing for n = 6 nucleo-
somes. Right: Experimental data (red) and model predictions (blue) for the one-particle
distribution function for different size genes L; horizontal grey-shaded bands correspond
to some bi-stable L domains. A horizontal line on the r.h.s. figure would correspond to
a density distribution function on the left with a different n. Fuzzy occupancy profile
corresponds to cases of coexistence between several optimal nucleosome configurations.
Adapted
Tuesday, May 29,from
12 G. Chevereau et al., Thermodynamics of Intragenic Nucleosome Ordering,
PRL 103, 188103 (2009). Red line.
By assuming that the nucleosome formation energy landscape is flat and their po-
sitioning along the gene has no sequence dependence, and by choosing a value for the
chemical potential or equivalently the absolute activity one can calculate from the
above model the nucleosome density distribution as a function of the size of the DNA
region accessible to them 176 . The problem of stacking a variable number of nucleosomes
in a 1D box with infinite wall boundaries is then analytically tractable from Eq. 443.
For a particular choice of chemical potential and the size of the nucleosome accessible
region L, one can observe density distributions with varying number of particles. The
cases where there is only one dominating value for the number of nucleosomes in the
gene stretch, e.g. 5 or 6 on Fig. 73, are referred to as "crystal" genes. For some values
of the size of the DNA stretch there can be a coexistence domain where two (or maybe
more) crystalline "bistable" configurations contribute statistically to a seemingly fuzzy
occupancy profile as shown on the same figure. IN any case there need not be any specific
information along an accessible DNA stretch for nucleosome positioning. According to
the above thermodynamic model, the existence of a chemical potential of nucleosomes
and a confining potential along DNA is enough to engender a periodic positioning of the
nucleosomes along DNA accounting quite well for the observed nucleosome statistics in
vivo. This statistical distribution probably plays a role also in gene expression regulation
in particular the bistable genes apparently show more transcriptional plasticity then the
crystal ones 177 .
176 G.Chevereau et al.,Thermodynamics of Intragenic Nucleosome Ordering, PRL 103, 188103 (2009).
177 A.Arneodo et al., Multi-scale coding of genomic information: From DNA sequence to genome structure
and function, Physics Reports 498 45 (2011).
11 DNA ORGANIZATION IN VIRUSES

Viruses are complexes of nucleic acids, proteins and in some cases lipids. They are
obligate intracellular parasites, lacking any mechanism for their own reproduction since
they contain no ATP generating system and no means of proteins synthesis. The nucleic
acid component of viruses can be either ds-DNA, ss-DNA or RNA, either in one or many
pieces. The protein component is usually assembled in a capsid that provides complex
protection of enclosed nucleic acids and is used for attachement of some viruses to their
cellular hosts. The lipid component is organized in a viral membrane, enclosing the
capsid, and is formed from host membrane lipids as well as viral proteins. Though viruses
can be quite complex they are in many respect just self-assembled macromolecules.
Viruses are a prime example of precise spontaneous macromolecular self-assembly.
In fact the very term self-assembly was invented by Donald Caspar and Aaron Klug,
who discovered the principles of organization of identical protein subunits in viral cap-
sids, of which more later. More than fifty years ago Fraenkel-Conrat and Williams 178
demonstrated that fully infectious tobacco mosaic viruses (TMV) could be created sim-
ply by mixing the viral RNA component with viral proteins. Under the right conditions,
such as pH and salinity, viruses formed spontaneously, i.e. without any special external
constraints or external energy input. This would mean that viruses could be viewed as
complicated chemicals with a life cycle coupled to the life cycle of a cell. In this view
e.g. a poliovirus is a complicated macromolecule with atomic formula
C332,652 H492,388 N98,245 O131,196 P7501 S2340 ,
that can easily reproduce in a test tube, containing a cell-free extract devoid of nuclei,
mitochondria, and other cellular organelles, seeded only with viral RNA 179 . The question
of whether viruses belong to living or non-living matter is thus ill-concieved.
Viruses are differentiated as either prokaryotic bacteriophages, infecting mostly bac-
teria, or eukaryotic, infecting plant and animal cells. Bacteriophages, estimated to be in
total on the order of ⇠ 1031 , amount to the most abundant organism in the biosphere
180
. Many phages have a tubular tail structure through which the viral genome passes
into the cell. The bacteriophage capsid is not internalized, in contrast to most animal
viruses that are internalized into the host cell essentially intact through endocytosis or
membrane fusion mechanisms. In the latter case, the capsid is covered with a lipid en-
velope that fuses with the host cell plasma membrane. Plant viruses utilize insects as
vectors that aid in mechanical penetration of the plant cell wall. Once the viral genome
has entered the cell, the virus hijacks the host cell’s machinery and synthesizes multiple
copies of viral genome. Transcription and translation of viral genes yields proteins that
assemble into new viral particles that exit the host cell on rupturing its cell wall. In what
follows I will discuss only certain aspects of ds-DNA icosahedral viruses which include
virus families Asfarviridae, Poxviridae, Iridoviridae, Herpesviridae, Adenoviridae, Pa-
povaviridae, Polyomaviridae, Papillomaviridae and many tailed bacteriophages, which
have been studied extensively by physical methods such as osmotic encapsidation and
ejection studies.
178 H.Fraenkel-Conrat and R.C. Williams, Proc. Natl. Acad. Sci. 41 690 (1955).
179 MollaA, Paul AV, Wimmer E., Cell-free, de novo synthesis of poliovirus, Science 254 1647-1651 (1991).
180 Wommack and Colwell, (2000).
179
Figure 74. A zoo of simple viruses obtained from Protein Data Bank with the UCSF
Chimera software. PDB identifiers are given below the structures.. Adapted from T. D.
Goddard, C. C. Huang, and T. E. Ferrin, Structure, 13, 473-482, (2005).
Capsids of simple viruses show pronounced icosahedral symmetry as is clear from

the viral zoo presented on Fig. 74. There are also many exceptions such as the core of the
HIV virus which displays a markedly conical shape with quantized conical angle values.
The structural principles of non-icosahedral viruses are not well understood. As the size
of the virus is increased, again compare Fig. 74, one observes also pronounced faceting
of the shape. Here one should clearly differentiate between the symmetry of the capsid,
which is icosahedral, and the characteristics of its shape, which can be spherical or to a
more or less pronounced degree faceted.
11.1 Caspar-Klug theory end elaborations

In 1956 Crick and Watson 181 argued that capsids of small viruses, with relatively short
genomes, have to be composed of numerous identical protein subunits arranged either
in helical motifs or in the shapes of regular polyehedra. Multiple redundancy of protein
subunits in viral shells, i.e. the fact that they are present multiple copies, is evolutionally
favored since damage to one subunit may render that subunit non-functional, but does not
destroy the infectivity of the whole particle. In 1962 Caspar and Klug 182 elaborated these
ideas further and argued that the regular polyhedra should have icosahedral symmetry
since icosahedron has the largest volume-to-surface ratio of all the regular polyhedra.
They also proposed an aufbau prinzip 183 for viral assembly, i.e. a model of mapping a 2D
hexagonal sheet composed of identical capsid proteins onto an icosahedron surface. This
can be accomplshed by cutting exactly 12 60 sectors from a 2D hexagonal structure and
stitching the remaining 2D mesh together by preserving the bonding pattern of the protein
subunits in the lattice. This process generates a 12 fivefold disclinations on a hexagonal
lattice that relieve the ensuing elastic stresses by an escape into a higher dimension via
buckling. This buckling in turn generates an icosahedral closed surface with exactly 12
fivefold disclinations.
This geometric construction would lead to an icosahedral shell composed of exactly
60 subunits, 5 per each five-fold vertex, distributed in completely equivalent positions by
symmetry and preserving the bnding pattern of the 2D hexagnal lattice. As many simple
viruses are composed of a lot bigger number of subunits, Caspar and Klug devised an
extension of their model by introducing the notion of quasi-equivalence. This would be
done by cutting out triangular sections that would not preserve the sixfold symmetry of
the lattice. Stitching the remaining 2D mesh together in this case would not preserve
the bonding pattern of the protein subunits in the 2D lattice exactly. The sites of protein
subunits in the icosahedral structure would be in this case only quasi equivalent, since
they would not preserve the symmetry of the original 2D lattice but they would in this
way allow for the shell to be compose dof more then 60 subunits. In fact the total number
of protein subunits in the quasi-equivalent positions would be
60 T where T = h2 + k 2 + hk,
with h, k non-negativer integers. T is referred to as the triangulation number of the virus.

h, k measure the number of steps in the direction of the two unit vectors of the hexagonal
lattice that connect one five-fold symmetric site to another one. In order to get from
one five-fold symmetric site to another one , one has to move over h six-fold symmetric
site along a row of nearest neighbor bonds on the hexagonal lattice, then turn 120 and
move another k six-fold symmetric steps. If the icosahedron has a higher triangulation
number, even if all proteins are chemically identical, some will be in an environment of 5
neighbors (5-fold vertex area: pentagons) and others will be in a 6 neighbors environment
181 F.H.C. Crick and J.D. Watson, Nature 177 473 (1956).
182 D.L.D. Caspar and A. Klug, Cold Spring Harbor Symp. Quant. Biol. 27 1 (1962).
183 In analogy to the famous aufbau prinzip for the construction of multielectron atoms that states a set of rules
for the buildup of complex atoms.

181
Figure 75. Two Caspar-Klug structures with T = 3 and T = 4 triangulation nubers.

The total number of protein subunits is 60 T . The values of h and k thatt set the trian-
gulation number are also shown for both cases. They correspond to the number of steps
on a 2D lattice that connect two neighboring 5-fold symmetric vertices. Not all protein
subunits are in symmetrically equivalent positions since some are in an environmenta of
5 neighbors and others are in a 6 neighbors.
(hexagons). Hence the postioning of each protein is not equivalent between each protein
and any other but only quasi-equivalent. The triangulation number thus measures the
number of quasi-equivalent positions in the viral capsid.
Though many virus types can be described via the Caspar-Klug construction not all
of them conform to the principles of quasi-equivalence. In order to put the classification
and structural principles of viral design onto a broader basis, Lorman and Rochal 184
devised a more general approach based on the notion introduced by Caspar and Klug in
their seminal work, viz. that "the viral self-assembly is a process akin to crystallization
and is governed by the laws of statistical mechanics. The protein subunits and the nucleic
acid chain spontaneously come together to form a simple virus particle because this is
their lowest energy state" 185 .
The theory of crystallization was put fourth by Landau and gives accurate and clear
predictions close to the crystallization point. Since the shape of small viruses with icosa-
hedral symmetry is close to spherical, one could treat the viral self-assembly as crystal-
lization on a spherical surface with protein positions given by the maxima in the corre-
spondnig probability density, consistent with the symmetry requirements of the system.
In thsi way the limitations of the Caspar-Klug geometrical construction can be thror-
oughly avoided.
184 V.L.Lorman and S.B. Rochal, Phys. Rev. E 77 224109 (2008).

185 Citedfrom G.J. Morgan Hostorical review: Viruses, crystals and geodesic domes, TRENDS in Biochemi-
cal Sciences, 28 86 (2003).
Figure 76. Viruses with different triangulation numbers (left to right): T = 1 (satel-
lite tobacco mosaic virus), T = 3 (cowpea chlorotic mottle virus), T = 3 (Norwalk
virus) and T = 4 (Nudaurelia capensis ! virus). From top to bottom are shown a 5 Å-
resolution rendering of the subunit coordinates and a radial rendering view of the virus
particle. The images were obtained with the viper particle explorer software (VIPER) at
http://viperdb.scripps.edu/. Adapted from P. Natarajan et al., Nature Reviews Microbi-
ology 3 809-817 (2005).
In the Landau theorythe probability density of protein distribution is given by ⇢ =

⇢0 + ⇢, where ⇢0 is an isotropic density of proteins in the solution and ⇢ corresponds
to density deviations from the isotropic solution case on viral self-assembly or crystal-
lization. The order parameter describing this crystalization is given by the critical system
of density waves and the corresponding Landau free energy is given as a function of the
invariant amplitudes of the order parameter. The critical part of ⇢, the one that gives
rise to crystallization, can be obtained by symmetry considerations as a series of spherical
harmonics Yl,m (✓, ) that has the form
m=l
X
⇢l (✓, ) = Al,m Yl,m (✓, ), (445)
m= l
where Alm are the amplitudes of the spherical harmonics Yl,m (✓, ). They are solutions
of the angular part of the Laplace equation and can be written in a canonical form as
s
2l + 1 (l m)! m
Yl,m (✓, ) = P (cos ✓) eim ,
4⇡ (l + m)! l
where Plm (x) are the associated Legendre polynomials. The critical part of ⇢ de-
scribing the distribution of proteins in the viral capsid has the icosahedral symmetry that
should not contain spatial inversion and mirror planes, i.e. it can only contain rotational
symmetry. This is a consequence of the fact the in general capsid proteins are asymmet-
ric. This effectrively constraints l the expasion over spherical functions, Eq. 445, to odd
values only - even more, only to particular odd vales: l = 15, 21, 25, 27, 31, 33 . . ..
183
Figure 77. Function fl (✓, ) across a sphere for l = 31 which corresponds to N = 4.

The maxima in this distribution correspond to positions of the protein subunits on the
capsid. Because of the symmetry of this function, only the positive part fl (✓, ) > 0 is
shown. Adapted from V.L. Lorman and S.B. Rochal, Phys. Rev. E 77 224109 (2008).
The absence of spatial inversion and mirror planes is due to the asymmetry of the
capsid proteins as these two symmetry operations would impose a higher symmetry on
capsid proteins. The corresponding Landau free energy close to the crystallization point
contains all the rotational invariants with icosahedral symmetry that one can form from
the powers of Al,m . To the lowest order it has the form given by
m=l
X
F = F0 + A(T, C) Al,m1 Al,m2 (m1 +m2 ) +
m= l
X X
+ Ck (T, c) akm1 ,m2 ,m3 ,m4 Al,m1 Al,m2 Al,m3 Al,m4 (m1 +m2 +m3 +m4 ) + . . .
k m1 ,m2 ,m3 ,m4
(446)
where ai are the weight coefficients required by the rotational symmetry and A(T, C) and
Ck (T, c) are temperature and composition dependent phenomenological coefficiencts of
the Landau theory. It thus contains only invariants of second and fourth order in the co-
efficients Al,m . For odd l the third order term in the Landau free energy is absent, which
means that the virus self-assembly would be similar to a second order phase transition
that proceeds without any nucleation intermediates. This would mean that no partially
formed capsids shoudl be observed on self-assembly. Even if thius is not true exactly, the
viral self-assembly process can correspond to at most a very weakly first order transition.
For small viruses Lorman and Rochal proved that the order parameter of the viral
self-assembly can be written in a much simplified form as
⇢l (✓, ) = const. fl (✓, ), (447)
where the function fl (✓, ) for a given l is obtained by averaging the spherical harmonics
over the icosahedral symmetry group that does not contain spatial inversion and mirror
Figure 78. Comparison of the positions of protein centers (colored circles) predicted
by the Lorman-Rochal theory with experimental viral structures for capsids satisfying
selection rules of the Caspar-Klug geometrical model. (a) Satellite Tobacco Mosaic
virus, (b) Cowpea Chlorotic Mottle Virus, (c) Sindbis virus. Protein centers (colored cir-
cles) predicted by the Lorman-Rochal model compared to experimental viral structures
for capsids which cannot be explained by the Caspar-Klug geometrical model. (a) L-A
virus, (b) Dengue virus, (c) Murine Polyoma virus. Adapted from V.L. Lorman and S.B.
Rochal, Phys. Rev. E 77 224109 (2008).
planes, as discussed above. The positions of the capsid paroteins are obtained as maxima
of the density function fl (✓, ) for different l. There are alltogether 60N of these maxima
but contrary to the Caspar-Klug theory, N can be any integer and not just those given by
the triangulation number T . Fig. 77 shows the variation of fl (✓, ) across a sphere for
l = 31 which corresponds to N = 4 and thus describes a Caspar - Klug structure of 240
protein subunits. fl (✓, ) is an antisymmetric function and changes sign under inversion
of coordinates or under mirror reflection across reflection planes of a regular icosahedron.
The figure represents only the positive part of fl (✓, ).
Functions fl (✓, ) generate capsid protein distributions that can be obtained via the
Caspar-Klug geometric construction as well as those that can not. Fig. 78 presents
examples of both cases. Protein positions correspond to maxima in the order parameter
distribution fl (✓, ).
185
11.2 Continuum elasticity of viral capsids

One can discern from Fig. 74 that deviations from an approximately spherical shape
become more pronounced as the radius of the capsid grows. If the surface density of
the capsomeric proteins is assumed constant, then the radius should scale as the square
root of the number of protein subunits. Also there is a tendency towards faceting as the
size of the casid is increased, with pronounced ridges between five-fold symmetric sites.
Lidmar et al. 186 applied continuum leasticity in order to understand the shape of viral
capsids as a function of their mean size.
Assuming that the viral caspid can be modelled as a thin elastic shell, an approxima-
tion that becomes more realstic as the mean size of the capsid grows, one can write down
its elastic energy as composed of two part: the curvature elasticity as well as the in-plane
stretching elasticity. The total free energy of deformation thus reads
Z ✓ ◆2 ! Z
1 1 1
1
FD = 2 dS K + + KG + 2 dS 2µ u2ij + u2kk ,
1
R1 R2 R 1 R2
(448)
where R1 and R2 are the two principal radii of curvature of the surface and uik is the
in-plane deformation tensor. K is the bending rigidity, KG the Gaussian rigidity and
, µ are the 2D Lamé coeffcients. The first term in the equation above is of course the
curvature energy and the second one is the in-plane stretching energy. The term linear
in the Gaussian curvature 1/R1 R2 integrates out to a constant dependent only on the
topological gender of the surface, due to the Gauss-Bonet theorem. Instead of the Lamé
coeffcients one can also use the 2D Young’s modulus Y and Poisson ratio ⌫ defined as
4µ(µ + )
Y = ⌫= .
(2µ + ) (2⌫ + )
In the Monge parametrization uz = ⇣(x, y) for a plane-like shell perpendicular to the axis
z, the in-plane deformation tensor can be written in terms of the 2D in-plane displacement
vector as ✓ ◆
@ui @uk @⇣ @⇣
uik = 12 + + 12 .
@xk @xi @xi @xk
The in-plane and curvature deformations are thus strongly and non-linearly coupled. The
minimum of the elastic free energy Eq. 448 is obtained via the Euler-Lagrange equations
that result in a system of non-linear Föppl-von Kàrmàn differential equations that are
quite difficult to solve. Instead of solving the FvK equations directly one can minimize
the appropriately discretized ftotal free energy of deformation itself.
Instead thus one usually takes recourse in the discretized version of the elastic energy
Eq. 448 that is standardly written in the form
X 2
X 2
F = 12  (nI nJ ) + 12 ✏ (|ri rj | a) . (449)
I,J i,j
Here i, j denote a pair of nearest-neighbor vertices with coordinates ri , rj and I, J a pair

of nearest-neighbor plaquettes with external normals nI , nJ in the discretized version
186 J. Lidmar, L. Murny and D.R. Nelson, Phys. Rev. E 68 051910 (2003).
of the shell. a is the mesh-size of a closed triangular surface of icosahedral symme-

try obtained by a Caspar-Klug construction. The discretized elastic energy Eq. 449 is
equivalent to the continuous version if one makes the following identification
p
2 1 3 4
Y = p ✏, ⌫ = K= , KG = K.
3 3 2 3
Results of numerical minimization of the above deformation free energy of the viral shell
can be understood in terms of a single parameter, referred to as the Föppl-von Kàrmàn
number ( ) defined as
= Y hRi2 /. (450)
Here hRi is the mean radius of the viral capsid. Different equilibrium shapes are thus
describable by different values of and were classified by Lidmar et al.. The energy
E corresponding to minimization of Eq. 449 is shown on Fig. 79. If is smaller
than ⇠ 150, the equilibrium shape of the capsid is a perfect sphere. For 250  
5000, a continuous ”buckling transition” in equilibrium shape takes place and the capsids
assume a more aspherical shape. The buckling referrs to regions surrounding the five-
Figure 79. The energies of elastic shells described by elastic energy Eq. 449 as a
function of . The regimes in which the shell can be represented as a sphere, assembly
of cones or ridges are indicated. Analytical expressions for the shell deformation energy
are indicated by full (sphere), dashed (cones) and dotted (ridges) lines, see main text. The
mixed regime can not be described with any simple dominant geometric shape. Adapted
from A. Siber and R. Podgornik (2008).
fld symmetric sites that ”buckle out” from the sphere so that the surrounding surface is
187
nearly conical. Furthermore for 250   104 , the capsids can be concieved as a union
of twelve conical frusta, with apices at the icosahedron vertices, that are fastened together
at their bases 187 . For 104   106 , the conical description of the shell becomes less
satisfactory with regard to the shell energetics since another creeping transition takes
place that flattens the icosahedron faces so that the regions around their edges begin
to sharpen. This effect has been explained by Witten and Li 188 and Lobkovsky 189 as
originating from the stretching energy along the edges that becomes prohibitively large
as the shell size. i.e. , increases. In the region 104   106 neither cones nor ridges
can provide an adequate representation of the shell shape and energetics. One should
note here that the borders of different regions are quite smeared, especially towards
the ridge sharpening regime, since the transitions in shell shapes are continuous. Since
scales as hRi2 , larger viruses should look more polyhedral, and this is indeed a general
trend observed in the experimental data, see Fig. 74. One should note finally that the
shapes with very large are not realistic for viral capsids but are possible for spherical
lipid vesicles with crystalline order.
Figure 80. Numerically obtained shapes for = 45 and 694. Even higher values of ,
and therefore more pronounced faceting, are possible in viruses e.g. the HK96 virus has
= 1480. The cases presented in this figure are thus quite realistic for viral shapes.
Adapted from J. Lidmar, L. Mirny and D.R. Nelson, Phys. Rev. E 68 051910 (2003)..
The energy of the various equilibrium shapes can be estimated by considering elastic
energies of the most important geometrical components of the shapes. Thus for almost
sphere-like shapes at < 150 = B the elastic energy can be estimated as
E
⇠A + B, (451)
K B
where A and B are numerical constants. In the regime of shapes dominated by cones the
elastic energy can be estimated as
✓ ◆
E
⇠ A 1 + log + C. (452)
K B
187 A. Siber, Nanotechnology 17, 3598 (2006).
188 T.A. Witten and H. Li, Europhys. Lett. 23, 51 (1993).
189 A.F. Lobkovsky, Phys. Rev. E 53, 3750 (1996).
And finally for the faceted regime dominated by the elastic energy of the ridges the elastic
energy should scale as
E
⇠ D 1/6 + E. (453)
K
The scaling of the minimal energy with the Föppl-von Kàrmàn number isindicated on
Fig. 79 and there is obviously very good correspondance with numerical results.
As is increased the faceting of the shape characterised by the elastic energy of
the ridges between five-fold symmetric Caspar-Klug sites becomes progressively more
pronounced. The final ideally sharp faceting is nevertheless observed in the numerical
solutions of the Föppl-von Kàrmàn model only for extremely high values of > 104
which are unrealistic for viral capsids. In spite of this one can certainly discern a trend
in the size progression of various viruses, see Fig. 74, that small viruses are round and
larger viruses are more faceted, just as one can conclude from the numerical studies of
the Föppl-von Kàrmàn model, see Fig. 80.
11.3 Viral capsids under mechanical stress

Everything said until now was true for unstressed viral shells. They can be stressed
either by the internal osmotic pressure of the encapsidated DNA, contributing a positive
mechanical stress on the capsid, or by the external osmotic pressure of osmoticants like
PEG, which can not penetrate the capsid, contributing a negative mechanical stress on the
capsid. Both cases have been analyzed ad the correspodinig free energies can be written
as
F = FD pV, (454)
where FD is given by Eq. 448 and V is the volume enclosed by the shell. The pressure
p above can be either positive or negative. For large positive internal pressures, the
shell energies increase signiïňAcantly
˛ above the values that obtained when there is no
mechanical stress acting on the shell. When the constant internal pressure is applied to
the shell changes together with the volume enclosed by the shell. Internal pressure
necessarily induces the increase of the stretching energy of the shell and leads to a larger
enclosed volume. For large internal pressures, the shell energies increase signiïňAcantly
˛
above the values that are obtained when there is no pressure acting on the shell and grow
linearly with even in the regime where unstressed shells do not show a linear variation
with , i.e. are not a regime that can be approximated by sphere. The shape of the shell
changes in the sense that the regions between five-fold symmetric sites buckle outwards
to allow for the volume increase of the shell driven by the internal pressure. In a limiting
case where the pressure difference becomes large enough the capsid can rupture and the
encapsidated DNA spills out. This effect also known as osmotic shock has been known
for a long time though it has not been investigated systematically 190 .
Fig. 81 shows the distribution of the deformation energy in a shell under mechanical
stress provided by internal pressure. For unstressed shells, the largest energy is contained
in the vicinity of the five-fold symmetric sites, while the opposite is observed for the
mechanically stressed shapes. In that case the largest energy is contained in the regions
between the five-fold symmetric sites that ”bulge out”. Practically all of this energy is
190 E.C. Pollard, The Physics of Viruses, Academic Press (1953).
189
of the stretching type. This may have some implications to bursting of viral capsids, in
particular, the points at which the cracks in the capsid initiate.
Figure 81. Equilibrium shape of an unstressed (a) elastic shell and an elastic shell
under mechanical stress (c) generated by internal pressure. The triangular faces in the
shapes are colored according to the total energy that they contain. This was calculated
as one half of the energy contained in the three edges of a particular triangle. The largest
energy, predominately of the stretching type, is contained in the regions between the five-
fold symmetric sites that ”bulge out” due to the mechanical stress. Adapted from A.
Siber, Phys. Rev E 73 061915 (2006).
These results have implications for viral assembly. Experimentally one sees that
mature viruses with encapsidated DNA enclose larger volumes, which is consistent with
the effects of a positive pressure of encapsidated DNA on the walls of the capsid. In the
process of packaging of the lambda genome the capsid shell expands from ⇠ 50 nm to
⇠ 60 nm. In order to strengthen the capsid after this dilation a cementing protein (gpD)
attaches to the icosahedral sites of the capsid 191
Also, mature viral shapes are observed to be more facetted from their precursor
shapes. This could be due to the concomitant increase of following the volume increase
because of the mechanical stress of encapsidated DNA. In most cases, DNA packaging
triggers an expansion process where the roughly spherical procapsid takes on the faceted
icosahedral shape of the mature virion and roughly doubles its volume. Capsid expansion
is typically accompanied by the addition of âĂŸ decoration âĂŹ proteins to the capsid
surface that serve to stabilize the expanded capsid shell against the internal pressures
generated by encapsidated DNA 192 .
The situation with negative pressures is drastically different. In this case for some
critical value of , depending on the magnitude of the applied pressure, the shell abruptly
collapses and crumples. In general deformations of elastic shells under external stress be-
long to two distinct categories, depending on their elastic parameters. Either the energy
of deformation is dominantly of the bending or of the stretching type, where the deforma-
tion is mostly concentrated in narrow stretching ridges. In the latter case the energetics
191 G.C.Lander, A. Evilevitch, M. Jeembaeva, C.P. Potter, B. Carragher, J.E. Johnson, J. Mol. Biol. (2008).
192 E.
Nurmemmedov, M. Castelnovo, C.E. Catalano and A. Evilevitch, Quarterly Reviews of Biophysics 40
327-356 (2007).)
of deformation is still dominated by the bending contribution, but the stretching energy
becomes of comparable magnitude. In non-stressed shells, both types of deformations
can be seen but in diïňĂerent regimes of . One thus expects that the same types of
deformations should also exist in the energetics of the shell collapse. We already know
that the deformation energy of equilibrium shapes can be written as E = K f ( ), where
f ( ) takes on diïňĂerent forms depending on the value of , see Eqs. 451, 452 and
453. One can thus assume that there is universality also in the collapsing pressure of
the shell that should manifest itself in a critical value of the collapse pressure pc of the
form pc = pc hRi3 /K, where again hRi is the mean radius of the non-pressurized shell.
This assumption is born out by numerical calculation presented on Fig. 82. The critical
pressure corresponding to the collapse of the elastic shells can be obtained by following
the eigenmodes of shells for a given pressure. These are obtained by diagonalization of
the Hessian, i.e. the matrix of second derivatives, of the elastic Hamiltonian. Near the
collapsing pressure, one of the eigenmodes becomes soft and its frequency approaches
zero indicating an instability of the shell with respect to the motion pattern speciïňAed
˛
by the eigenvector of the soft eigenmode.
This figure presents the scaled critical pressures pc as a function of in the non-
pressurized state. Note that the thus rescaled collapsing pressures for shells characterised
by different elastic moduli and average radius all fall onto the same universal curve given
by
K
pc = U ( ). (455)
hRi3
For smaller than ⇠ 250 the equilibrium shape of the unpressurized shell is a sphere.
Therefore, the critical pressures should be proportional to the collapse pressure of an thin
elastic sphere which is given by
K 1/2 1/2
pc ⇠ or equivalently U( ) = . (456)
hRi3
This is indeed what can be discerned from numerical results in the region 90  
250. There is some discrepancy at smaller values of but in that case the thin shell
approximation breaks down anyhow, and this scaling should not be expected.
In the range of where the dominant shapes of the elastic shell are conical, the critical
pressure should be given by its value for a cylinder, which can be conceived as deformed
conus. In that case the critical pressure should scale as
K
pc ⇠ or equivalently U ( ) = const. (457)
hRi3
where hRi is now proportional the maximal radius of the cylindrical regions in the shell
(this suggest that the critical points of collapse are located in the middle of icosahedron
edges). In this regime obviously the critical pressures should be independent of , which
is indeed born out by numerical investigation in the interval 300   3000.
For even larger values of , the equilibrium shell shape just prior to collapsing shows
ïňĆattening of the faces and concentration of curvatures along the ridges, so that the
radius of cylindrical regions immediately before collapse notably decreases from its non-
pressurized value as a result of applied pressure. This can also be seen as an increase
191
Figure 82. Panel (a): Scaled critical pressures pc hRi3 /K (symbols) of the icosadelta-
hedral shells with ✏ = 1 and T = 441 (+), ✏ = 10 and T = 441 (⇥), ✏ = 5 and T = 100
(squares), ✏ = 5 and T = 49 (circles), and ✏ = 4 and T = 73 (triangles) as a function of
. In these calculations, u = 0.005a, ✏ was kept fixed and  was varied so to produce a
variation in . Thick dashed lines show scalings with as discussed in the text. The in-
sets show the buckled shell shapes for different FvK numbers as denoted. Panel (b): For
comparison, the energetics of non-pressurized shells as a function of , Fig. 79. Adapted
from A. Siber and R. Podgornik, PHYSICAL REVIEW E 79, 011919 (2009).
in the collapse pressure, i.e. its deviation from the constant value predicted by Eq. 457.
In the intermediate region 104   106 neither ridges nor cones provide an adequate
description of the shell shape. However, when > 106 , the ridge-sharpening transition
gradually kicks in and the critical pressure assumes the scaling form
K 1/2
pc ⇠ 1/6 )3
or equivalently U( ) = . (458)
(hRi
Here the characteristic radii of curvature for non-pressurized ridges scale as hRi 1/6 .
The functional form of U ( ) is thus exactly the same as in the case of a sphere. Again
numerical results corroborate the above scaling analysis.
In order to apply our results to viruses one needs to know first of all the elastic moduli
associated with capsid shells. They are not known very precisely, but can be estimated
to be of the order of K ⇠ 10 40kB T from the studies on bacteriophage capisds under
internal pressure generated by encapsidated DNA. Assuming a radius of about hRi ⇠ 30
nm, typical for e.g. -bacteriophage gives rise to collapsing pressures of about pc ⇠ 5
atm. The pressures of this order of magnitude can be easily achieved in osmotic stress ex-
periments on empty capsids in an aqueous solution of PEG of the type already performed
by Evilevitch and coworkers on native bacteriophages.
11.4 Osmotic encapsidation of DNA

Recent studies of packing of viral DNA in ds-DNA viruses such as , T4, T7, epsilon15
and 29, are starting to provide a very detailed picture of the genome organization inside
bacteriophage capsids. At elevated densities DNA appears to be wrapped into a coaxial
inverse spool, with pronounced ordering and high density close to the capsid wall that
both appear to decay towards the center of the capsid. This packing allows DNA to act
like a coiled osmotic spring piled up against the inner surface of the capsid ready to
release its chemical and mechanical energy through the portal complex on docking onto
a bacterial wall. Cryo-EM of 29 bacteriophage, Fig. 83, shows ordered spooling of
DNA inside the capsid spanning the whole available space from the capsid wall to the
core. It is packed at elevated densities corresponding to the line-haxatic liquid crystalline
phase observed in the bulk. Obviously microns long DNA has to be confined under very
large values of osmotic pressure reaching tens of atmospheres and the capsid has to be
sufficiently robust to withstand these elevated pressures. Young’s modulus on the order
of 100 MPa would be sufficient to preserve structural integrity of the capsid at these
mechanical stresses. The high pressure is necessary in order to initiate the ejection of
the viral genome into the bacterial cell. On the other hand, active packaging mechanisms
involving portal motor proteins are obviously needed for the maturation of a virus. In the
case of 27 the work done by the portal motor complex in the process of packaging of
DNA is ⇠ 20000kB T .
The energetics of packaging has been investigated thoroughly by direct single-molecule
measurements of the in vitro packaging rates in phage 29 by the optical tweezer method
193
. These experiments made it possible to start thinking about the encapsidation process
in purely physical terms. An internal force of up to ⇠ 50 pN was detected, resisting
packaging that provided quantitative estimates of the energy of packaging which in turn
spawned several theoretical models of DNA packaging in bacteriophages. A different
approach was pursued by Gelbart and coworkers who applied the principles of osmotic
stress experiments, amply used to study the packaging of DNA in the bulk, to the problem
of packaging and ejection of DNA from bacteriophages such as and T5 194 .
They used PEG of sufficient molecular weight so that it can not penetrate the viral
193 Smith,D. E. et al. Nature 413 748âĂŞ752 (2001).
194 E.
Nurmemmedov, M. Castelnovo, C.E. Catalano and A. Evilevitch, Quarterly Reviews of Biophysics 40
327-356 (2007).
193
Figure 83. Three-dimensional reconstruction of the full-length 29 packaged genome.

Surface rendering of the reconstructed density maps, for the symmetrical refinement, at
two slightly different isosurface values, A and B on the figure. Adapted from Comolli,
LR; Spakowitz, AJ; Siegerist, CE, et al., Virology 371 267-277 (2008).
capsid. By adding DNAase and LamB receptor to the bathing solution, the tail complex
of the virus opens up and DNA can be ejected into the solution by uncoiling its osmotic
spring i.e. the energy of confinement to the capsid interior. The amount of ejected DNA
can be regulated by varying the concentration of PEG in the bathing solution and thus
the osmotic pressure of the solution that counteracts the osmotic pressure of DNA in
the capsid. In experiments with bacteriophage the fraction of DNA, of total length
41.5 kb, ejected from the capsid decreased monotonically with increasing PEG osmotic
pressure. Similar results were obtained with bacteriophage T5. It thus appears that the
same physical principles control DNA ejection from two quite diïňĂerent bacteriophages
and suggests that there is a general mechanism of DNA ejection that applies to many
dsDNA viruses that use molecular motors to package their genomes.
Osmotic encapsidation experiments nicely complement the osmotic stress experi-
ments on DNA in the bulk. In the latter case they provide the dependence of DNA
osmotic pressure on its density: p0 (⇢). The osmotic pressure in the bulk case is usually
measured in nematic hexagonal or line hexatic arrays via the osmotic stress method as a
function of the interaxial spacing D, and is usually referred to as the equation of state of
DNA. This equation of state has been investigated very thoroughly in different salt envi-
ronments and a wealth of osmotic pressure data in the bulk has been assembled. Just as
one measures the bulk DNA equation of state one can refer to the osmotic encapsidation
experiments as measuring the encapsidation equation of state for viral DNA. It would
connect its osmotic pressure inside the capsid with the total length of DNA encapsidated.
Figure 84. Schematic presentation of the osmotic stress experiment on DNA ejection
from viral capsids. The bathing solution contains PEG at varying concentrations, pro-
viding the osmotic reaction force that confines DNA to an open capsid. The result is the
dependence of the ejected fraction of DNA on external osmotic pressure. Adapted from
E. Nurmemmedov, M. Castelnovo, C.E. Catalano and A. Evilevitch, Quarterly Reviews
of Biophysics 40 327-356 (2007).
11.5 The inverse spool model

The spooling of DNA inside the viral capsid gives rise to a molecular model of packing
that would take into account its salient features. Inside the capsid, the viral DNA seems to
be coiled against the internal surface of the capsid, showing pronounced order throughout
its interior, see Fig. 85. The DNA is not in close conatct with the caspid surface since it
is negatively charged.
Proposed first in physical terms by Grosberg and later elaborated by Odijk and Gel-
bart inverse spool model is based on the idea of the decomposition of DNA packing en-
ergy within the viral capsid into an interaction term and a curvature term. The structure
of HK97 and viral capsids indicates that their interior is negatively charged, preventing
a direct contact between DNA and the capsid surface. This model is known to describe
the viral genome ejection reasonably well. Apart from the simulations of the genome
packing within the capsid, all theoretical approaches to DNA viral packing are based
on certain assumptions regarding the form of the curvature energy of the DNA forced
to reside within the confines of the capsid, as well as the interactions among the highly
charged and hydrated DNA segments packed at high densities within the capsid.
Elastic curvature energy appears to be the lesser of the two unknowns alluded to
above. It is proportional to the square of local DNA curvature and in fact follows from
the Euler-Kirchhoff model of an elastic filament. This model appears to be a consistent
description of DNA on mesoscopic scales. The parameters of the Euler-Kirchhoffian
model of DNA, such as its persistence length, are well established and have been mea-
sured by a variety of methods with satisfactory consensus among the results as discussed
above. The interaction energy enters the inverse spool model rather differently. It is
measured in osmotic stress experiments directly and can be deconvoluted into a longer
ranged electrostatic contribution and a shorter ranged hydration component, that have
195
Figure 85. Three-dimensional density of mature bacteriophage lambda reconstructed

from CryoEM micrographs. Left: the subnanometer-resolution map of bacteriophage
lambda colored radially from the phage center (red to blue). Right: the DNA spool and
the dimensions of the capsid. Adapted from G.C. Lander, A. Evilevitch, M. Jeembaeva,
C.P. Potter, B. Carragher, J.E. Johnson, J. Mol. Biol. (2008).
been quantified in terms of magnitudes and decay lengths. The experimental variable is
thus the osmotic pressure of DNA, either in bulk arrays or in the osmotic encapsidation
experiments. In these experiments one assumes that the viral genome is in osmotic equi-
librium with an external PEG solution of known osmotic pressure and known activity of
the salt. The viral capsid is fuerthermore assumed to be permeable to small salt ions but
is impermeable to PEG.
According to cryo-EM studies the packing of DNA within bacteriophages seems to
be governed by the tendency of DNA to form local alignments (nematic liquid crystals)
whose orientational order is constrained by the bacteriophage capsid. This constrained
nematic crystallization within the bacteriophage capsid suggests to treat the viral genome
packing in the framework of the liquid crystalline nanodroplet model that can be formu-
lated as a local thermodynamic mesoscopic theory based on the director field and the
density field of the polymer.
11.5.1 The free energy in the inverse spool model
Since the polymer shows local nematic order, the corresponding deformation energy den-
sity is given by the Frank-Oseen ansatz as
fD (⇢, ni , @l nk ) = 12 K1 (⇢)(r · n)2 + 12 K2 (⇢)(n · (r ⇥ n))2 + 12 K3 (⇢)(n ⇥ (r ⇥ n))2 ,
(459)
just as in the case of bulk DNA nematic liquid crystals. The bending elastic modulus of
the DNA nematic has the form
V (D)
K3 (⇢) = Kc ⇢(2) + = Kc ⇢(2) + K0 (⇢), (460)
D
where Kc = kB T LP is the intrinsic elastic modulus of the polymer with LP its per-
sistence length. ⇢(2) is the 2D density of the polymers perpendicular to the director,
D is the average axial separation between the polymers and V (r) is the interaction
potential. The connection between the 3D monomer density ⇢ and ⇢(2) is given by
⇢ = (⇢(2) /Lbp ) ⇥ (L/Lbp ), where Lbp is the phosphate-phosphate separation along
the DNA and L is its total length. The other two terms in the elastic free energy are as-
sumed to be negligible. In the range of densities inside viral capsids the K0 term is much
smaller then the first one and remains only a small perturbation.The splay term, propor-
tional to K1 , is small because the splay modulus in nematic polymers is renormalized
due to the coupling between local polymer density and local nematic director, in such a
way to penalize any appreciable splay deformation. Also the cylindrical spool ansatz that
is used below automatically satisfied the condition r · n = 0. The twist term in Eq. 459,
proportional to K2 , is also small on the length scale of a viral capsid. Of the three terms
in the Frank - Oseen elastic energy only the bending term is significant. The deformation
free energy can be thus written to the lowest order as
⇢ L2bp
fD (⇢, ni , @l nk ) = 1
2 kB T LP (n ⇥ (r ⇥ n))2 ), (461)
L
with the total free energy thus of the form
f (⇢, ni , @l nk ) = f0 (⇢) + fD (⇢, ni , @l nk ) ⇢µ, (462)
where both, the density as well as the director field are by assumption position dependent,
⇢ = ⇢(r) and n = n(r). The first term in the above free energy represents the part that
depends only on the density and thus describes bulk DNA at density ⇢, while µ is the
(constant) chemical potential associated with the equilibrium between the capsid and the
external PEG solution, depending on the osmotic pressure in the external compartment.
The index 0 will be used to stand for the bulk properties of DNA that is not enclosed and
thus elastically deformed inside a capsid. The thermodynamic equilibrium conditions are
obtained from the Euler - Lagrange (EL) equations that can be derived as
@ @
f0 (⇢) + fD (⇢, ni , @l nk ) = µ, (463)
@⇢ @⇢
and
✓ ◆
@ @
@k fD (⇢, ni , @l nk ) fD (⇢, ni , @l nk ) = ni , (464)
@(@k ni ) @ni
where we have taken into account the fact that n is a unit vector. is the Lagrange
multiplier associated with the constraint n · n = 1. The EL equations are the explicit
equilibrium conditions that determine the director and the density equilibrium profiles of
the DNA nematic droplet inside the viral capsid: ⇢(r) and n(r).
11.5.2 Osmotic pressure in the inverse spool model
The EL equations have a first integral of the form
p0 (⇢) + fD (⇢, ni , @l nk ) = p. (465)

197
where p is the external osmotic pressure. Explicitly, by taking into account the form of
the deformation free energy Eq. 461 we end up with
⇢ L2bp 2
p0 (⇢) + 12 kB T LP (n ⇥ (r ⇥ n)) = p. (466)
L
This equation states that for a deformed DNA nematic nanodrop in the capsid the total
osmotic pressure p is composed of an interaction contribution to the osmotic pressure,
p0 (⇢), that can be measured separately in an undeformed nematic DNA phase as e.g.
in the osmotic stress experiments, and an elastic stress contribution to the total osmotic
pressure, which is simply equal to the elastic free energy density.
Figure 86. A three-dimensional cut-through view of the DNA density distribution in the
capsid of R = 30 nm under osmotic pressure of p=0.5 atm in a solution that contains
0.15 M of NaCl. Left: A schematic presentation of the inverse spool of DNA inside
the capsid (left). High density yellow and low density blue. A tiny depleted region of
cylindrical symmetry is shown on the cylindrical axis. Right: density profiles along the
radial directions as shown on the left. The length of encapsidated DNA is 68.5 nm.
Adapted from Siber et al., Packing nanomechanics of viral genomes, Eur. Phys. J. E 26
317 (2008).
11.5.3 DNA director field and density profile in the inverse spool model
Instead of solving the EL equations directly for the spatial director and density profiles,
one can assume a certain nematic director profile and deduce the corresponding density
profile. In an inverse spool DNA is packed with cylindrical symmetry as shown by
experiments. The director configuration with cylindrical symmetry has only an equatorial
component with respect to the z axis, which is also the axis of symmetry of the inverse
spool, n = ˆ , where ˆ is the unit vector in the equatorial direction. With this ansatz one
has
cos ✓ ˆ 1
n ⇥ (r ⇥ n) = ✓ + r̂, (467)
r sin ✓ r
where ✓ˆ and r̂ are the appropriate unit vectors. The first integral of the EL equations
within this ansatz now assumes the form
L2bp ⇢(r, ✓)
p0 (⇢(r, ✓)) + 12 kB T LP = p. (468)
L r2 sin2 ✓
To this equilibrium condition we have to add the constraint on the length of DNA inside
the capsid, L(p). This is defined as
Z ⇡ Z R
L(p)
= 2⇡ ⇢(r, ✓)r2 sin ✓d✓ dr. (469)
Lbp 0 0
Above we have acknowledged that the DNA density within the capsid has the form
⇢(r) = ⇢(r, ✓) due to cylindrical symmetry. The mechanical equilibrium condition Eq.
468 has a zero-radius limit viz.
2pL
r !0: ⇢(r, ✓) = r2 sin2 ✓ + . . . (470)
kB T LP L2bp
The curvature stress obviously depletes the core (r ! 0) of the DNA liquid crystalline
nano-drop within the capsid, though the region of this depletion is relatively small since
the rather small curvature pressure has to compete with huge interaction pressure. Fig.
86 schematically shows the density profile of DNA within a viral capsid obtained directly
from solving Eq. 468.
The density profile of the encapsidated DNA as a function of external osmotic pres-
sure p then gives the length of encapsidated DNA as a function of external osmotic pres-
sure, which is exactly the quantity measured in experiments of Evilevitch et al. The
results for a capsid of radius R = 10 nm in two different concentrations of NaCl salt are
shown in Fig. ??. The full lines were obtained by neglecting completely the DNA bend-
ing energy contribution, i.e. by setting ⇢(r, ✓) = ⇢0 from p0 (⇢0 ) = p, and evaluating
the encapsidated length as L(p) = 43 R3 ⇡⇢0 (p)Lbp . Note that the encapsidated lengths
obtained by this simple method are practically indistinguishable from the full numerical
results which indicates that the DNA bending energy contribution to the total energy bal-
ance for this choice of parameters is obviously small. Similar results are obtained also for
capsids of larger radius, e.g. R = 30 nm. Of course, the total encapsidated lengths are
much larger in this case, but again, approximate values obtained disregarding completely
the bending energy are in excellent agreement with numerical results. We thus conclude
that DNA bending does not contribute significantly to the energetics of viral packing. For
bacteriophage EMBL3, that has a mean capsid radius close to 30 nm, the theoretical
results predict that the full length of its genome (14.4 µm) at 0.15 M NaCl would be
packed at osmotic pressure of 24.4 atm, in excellent agreement with experimental data
of Evilevich et al. 195 . One should also note the convergence of the encapsidation curves
in Fig. 87 at high osmotic pressures. This is a straightforward consequence of the behav-
ior of DNA in the bulk, where the equation of state at high values of osmotic pressure
gradually ceases to be dependent on the salt concentration.
One should note nevertheless, that although the energetics of DNA bending is of
minor importance concerning the encapsidated DNA length, it has nevertheless a visible
195 A. Evilevitch, J.W. Gober, M. Phillips, C.M. Knobler, and W.M. Gelbart, Biophys. J. 88, 751 (2005).
199
Figure 87. The length of the encapsidated DNA as a function of the osmotic pressure in
the solutions containing 0.15 M (squares) and 2 M of NaCl (circles). The symbols denote
numerical results obtained by solving Eq. 468. Full lines are predictions of Eq. ??. The
capsid radius in these calculations is R=10 nm. The region of bulk osmotic pressures
left from the vertical dashed line is not determined in the osmotic stress experiments and
was obtained by extrapolating the existing results in the way indicated in Ref. [?]. The
relative error in L changes with pressure and depends on the relative errors of osmotic
pressure and DNA density measurements in the bulk. It is smallest at high pressures
(typically below 7 %). Adapted from A SIber et al. (2008).
and non-negligible influence on the density distribution of encapsidated DNA, i.e. on

the DNA density profile within the capsid. This is clear from Fig. 86. One can clearly
discern a cylindrical region of depleted DNA density. The depletion effect is always
more prominent for smaller radial distances, but is observed only for sufficiently low
pressures (small genome packing fractions). For high enough pressures, the depletion
region becomes ”compressed” and its effective radius small on the scale set by the capsid
radius. Note that the radius of a cylinder of depleted DNA density is about 0.05 of
the capsid radius, and its volume is thus 0.016 R3 which is only 0.4 % of the capsid
volume. All these conclusions stemming from the numerical analysis of our model are
completely consistent with recent investigations of the three-dimensional architecture of
the bacteriophage 29 packaged genome 196 .
11.6 DNA toroids inside viral capsids

The formation of toroidal globules of single DNA chains trapped inside a confining vol-
ume has been reported recently, where the confining space is provided by a virus capsid
itself, either upon addition of spermine (Spm4+ ) or in a monovalent salt buffer contain-
196 Comolli, LR; Spakowitz, AJ; Siegerist, CE, et al., Virology 371 267-277 (2008).
ing PEG. These experiments raise additional theoretical and biological questions. Indeed
DNA is always confined in vivo, whatever the system under consideration: each DNA
chain with its associated proteins forming chromatin is confined in a chromosomal terri-
tory inside the nucleus of the eukaryotic cell, itself limited by the nuclear envelope; the
bacterial nucleoid is confined in the central volume of the bacteria by the macromolecu-
lar crowding effect arising from other macromolecules and the viral genome is confined
within the protein capsid. This confinement is particularly important in this latter case
because the capsid is rigid and entraps the entire genome under a very dense and still
unknown conformation. It has become clear recently that after a partial ejection of the
viral DNA, the fragment of the chain that is kept inside can be collapsed into a toroid.
This transition between a confined coil and a confined toroid, triggered by changes in the
environment of the virus, raises the interest to understand how the capsid can influence
the organization of the hedged in DNA chain.
In order to analyze the toroidal globule formation of DNA inside a rigid capsid, we
set up a theory of a confined condensed inverse spool. We assume that the surface free
energy equals a constant surface tension of the aggregate fS = . Thus we can write the
free energy functional in the form 197
Z Z I I
1 ⇢c (2) 3 (2) 3
F = kB T Lp 2
d r ⇤ ⇢ c d r + dS a dSa , (471)
2 V r V @V @V
with the constraint Eq. 366. The first two terms, the elastic energy of the wound-up DNA
and the surface free energy of the toroid, are the same as in the model proposed by Ubbink
and Odijk 198 . The third term complicates the model and makes it less tractable due to the
introduction of surface energy parameter, a , that depends on the strength of the effective
short-range DNA-capsid interaction. The area ofHthe contact region between the DNA
condensate and the inside capsid wall is given by @V dSa . This is an unknown quantity
whose value is obtained only after the minimization of the functional. We consider the
free energy functional on a hyperplane of constant DNA volume, V , the same as in the
UO case. In our case, we additionally restrict the problem by the introduction of the
capsid which acts as a constraint, allowing only solutions contained within the sphere of
radius R.
Following closely the analysis of a toroid formation in free solution we parametere
again the shape of the toroid by h(r) as coordinate of the upper half of the toroid’s shape
in z direction, we remain with the following form of the free energy and the total length
of condensed DNA constraint
Z ro ✓ p ◆
1 h(r)
F = 4⇡ kB T ⇠p ⇢c (2) ⇤⇢(2)
c h(r)r + r 1 + h 02 dr, (472)
ri 2 r
and Z ro
L = 4⇡ ⇢(2)
c h(r)rdr . (473)
ri
Introducing now the dimensionless parameter ↵t = 4 V 1/3 /(⇢c (2) ⇠p ) the minimization
of the above free energy leads to solutions h(r) that can be parameterized by ↵t .
197 J.
Ubbink and T. Odijk, Europhys. Lett. 33, 353 (1996).
198 Ubbink,J., T. Odijk. Deformation of toroidal DNA condensates under surface stress, Europhys. Lett., 33,
353-358 (1996).
201
a) 0 1 10 100 1000 a
10
40
20
y [nm]
0.1 0
0 20 40
t
x [nm]
b) c)
Figure 88. The shapes of DNA toroids in [↵t , a / ] parameter space. Blue dashed lines
in a) indicate the inner surface of the spherical container with nm. The volume of the
DNA in these calculations was fixed to 1.4 105 nm3 . b) Three-dimensional rendering of
the DNA toroid with ↵t = 1, a / = 0. c) Three-dimensional rendering of the DNA
toroid with ↵t = 10, a / = 100.
A shape phase-diagram of the problem representing the optimal toroid cross-sections

for particular combination of ↵t and a / parameters is shown in Fig. 88. Columns
of cross-sections corresponding to a / = 0 case, i.e. without attraction between the
DNA and the capsid, are presented in Fig. 88a). Note that the top-left shape ( a / = 0
and ↵t = 10) is the same as would be obtained in the UO model since the toroid does
not touch the capsid in this case, i.e. the capsid impenetrability constraint is not active.
For sufficiently large adhesion energy parameter (last column of images in Fig. 88a),
the cross-section of the DNA toroid only weakly depends on ↵t . Obviously the presence
of attractive interactions between the capsid wall and the toroidal condensate leads to
a deformation of its cross-section that becomes markedly concave, crescent-like, Fig.
88c). By analyzing numerical solutions for very small values of the parameter, without
any attraction between the capsid wall and the DNA toroid ( a / = 0), one finds that
the limiting shape for small values of is a plan-convex lens.
This leads to the conclusion that the concavity of the inner surface of the DNA toroid
section can be observed only when the attraction between the DNA and the capsid is
present. Otherwise the cross- sections are always flat or convex i.e. the shape constraint
in combination with elasticity and the surface energy minimization cannot produce a
concave shape. The crescent-like form of the condensate cross-section thus appears to be
the distinguishing feature of the attractive interaction between the DNA condensate and
the capsid wall. The experiment shows two markedly different types of toroid shapes. In
numerical calculations the characteristic crescent-like shapes of the cross-sections with
concave inner surface (see Fig. 88), similar to those found in experiments with PEG, are
obtained only in the case of a sufficiently strong attractive interaction between the DNA
condensate and the inner capsid wall as codified by the parameter . These distinguishing
features are certainly observed in the case of PEG condensation but are absent in the case
of Spm4+ condensation. In that case one can identify more circle-like cross-sections
consistent with model calculations for ↵t ⇠ 1 and small enough a / < 1.
11.7 Nematic DNA Ordering in viro

Since DNA is a long polyelectroyte molecule its connectivity introduces additional fea-
tures into a consistent continuum description of its nematic ordering that are not ad-
dressed by the standard approach of the liquid crystal physics [?]. Based on previous
work by Kamien et al. 199 we thus first introduce and formulate the concept of the
polymer current and then construct an appropriate nematic ordering free energy that we
solve in confined geometry of a spherical capsid.
11.7.1 Polymer "current density"
In a polymer nematic liquid crystal, splay deformation becomes progressively more ex-
pensive with increasing chain length, as it imposes costly local changes in polymer den-
sity. In the continuous limit of long chains this coupling between the polymer density and
orientational fields is described by an analogue of the continuity equation for the nematic
director field n, 200 r · (⇢s n) = 0, where ⇢s is the surface number density of chains
crossing the plane perpendicular to the director field. The above equation can be inter-
preted as the continuity condition for a “polymer current density” j = ⇢s n ,defined by
a length derivative rather than the time derivative. The only difference between Eq. (??)
and the usual continuity equation is that in this case j does not describe a rate (there is
no time derivative involved in its definition), i.e., we are observing the number of chains
199 R.D. Kamien, P. LeDoussal, D.R. Nelson, Phys. Rev. A 45 8727 (1992).
200 P.G. de Gennes, Mol. Cryst. Liq. Cryst. Lett. 34 177 (1977); R. Meyer in Polymer Liquid Crystals, A.
Ciferri et. al. (eds.), Academic NY (1982).
203
perforating the perpendicular plane rather than the number of particles crossing it per
unit time.
The analogy comes fully into life if one relaxes the condition |n| = 1 and takes into
account the degree of nematic order. Let us stress that at this stage we will consider
polar ordering rather than quadrupolar nematic ordering. It is not yet understood how to
construct a replacement for the polymer current in case of the nematic order tensor. We
will introduce a description of nematic ordering through a non-unit nematic order vector
a, a = hcos ✓i n, with ✓ the angle between the local tangent on a chain and the average
vicinal director field represented by n. Let us furthermore define the following quantity
* +
XZ
3
t(r) = dì ṙ(ì ) (r r(ì )) . (474)
i i
The sum over i goes over all the polymer molecules in the system, and the integrals over
ì go over the length of each of the polymers. ṙ(ì ) = dr(` i)
dì is the local unit tangent
vector of the chain. Evaluating the divergence of this vector, we get
* +
XZ
3
r · t(r) = dì ṙ(ì ) · r (r r(ì )) =
i ì
* +
XZ d 3
= dì (r r(ì )) = ⇢+ (r) ⇢ (r) .
i ì dì
(475)
where
* + * +
X X
+ 3 3
⇢ (r) = ⇢(r; L) = (r ri (L)) and ⇢ (r) = ⇢(r; 0) = (r ri (0))
i i
are the total densities of the ends and beginnings of the chains in the sample. In deriving
the above result we took into account that r ! ri , since the Dirac delta function
has an argument r r(ì ). Also, ṙ(ì ) · ri = d`d
.
This constraint, Eqs. 475, on the director field in polymer nematics has been discov-
ered by de Gennes and Meyer 201 and bears some similarity with the differential form
of the Gauss theorem in electrostatics. The divergence of the field is proportional to the
density of the sources. Here the orientational field t plays the role of the electrostatic
field and the density of beginnings and ends of the chain play the role of positive and
negative charges. Only if there is a mismatch of ⇢+ (r) and ⇢ (r) in a certain volume
can there be a splay deformation of the polymer nematic.
11.7.2 Free energy
To determine the equilibrium configuration of the director and density fields we set up a
free energy density following the Landau approach 202 . The appropriate variables in the
201 P. G. de Gennes, Mol. Cryst. Liq. Cryst. Lett. 34 177 (1977); R. Meyer in Polymer Liquid Crystals, A.
Ciferri et. al. (eds.), Academic NY (1982).

202 P. M. Chaikin and T. C. Lubensky, Principles of Condensed Matter Physics, (Cambridge University Press;
1st edition (2000)).

polymer case are the complete director field a, describing the orientation and the degree
of order, and the polymer density field ⇢. Both fields are coupled by the continuity
requirement (475). The conservation of polymer mass will be satisfied globally. The
phase transition will be controlled by the density (concentration) of DNA as is the case
for lyotropic nematic liquid crystals.
Let us stress that for computational reasons all equations must remain regular also for
vanishing order, i.e., they must be expressed by the full vector a. Decomposition of the
form a = an, where a is the degree of order, would result in a singularity of the form 0
times 1 taking place in centers of defects, where rn diverges while the degree of order
vanishes. In contrast, a and its derivatives remains regular everywhere.
Taking into account the definition (??), Eq. (475) is already of the correct form. In
the elastic free energy, instead of using the usual Frank terms for splay, twist, and bend
of the director,
1 1 1
f F rank = K1 (r · n)2 + K2 [n · (r ⇥ n)]2 + K3 [n ⇥ (r ⇥ n)]2 , (476)
2 2 2
a new set of elastic terms must be used 203 :
1 0 1 1 1
f el = L (@i aj )2 + L02 (@i ai )2 + L03 ai aj (@i ak )(@j ak ) + L04 (✏ijk ak @i aj )2 .
2 1 2 2 2
(477)
Unlike the Frank elastic parameters Ki , the elastic constants L0i do not depend on the de-
gree of order. To keep the number of elastic parameters at minimum, among all possible
terms quadratic in the derivative we have retained only those non-vanishing in the limit
of a fixed degree of order. Comparison of Eqs. (476) and (477) in this limit leads to a
rewriting of the Frank constants Ki in terms of the constants Li 204 .
One observes that the dependence on the degree of order is different than in the case
of the nematic tensor order parameter, where the corrections that break the degeneracy
of splay and bend are cubic in the degree of order. Note that all three deformation modes
(splay, twist, and bend) are degenerate in the limit of small degree of order, as indicated
by the leading order dependence Ki / a2 . This would not be the case if one just used
a in the Frank expression (476). Yet it must be so, because the direction n necessary to
distinguish between splay, twist, and bend, is not defined as a ! 0. Generally, the Frank
expression (476) can be used only perturbatively, i.e., for |a| ⇡ 1, whereas the expression
(477) or similar can be used to describe the full range 0  a  1.
To establish a one-to-one correspondence between the two sets of elastic parameters,
we make a further simplification of our model free energy expression (477), while retain-
ing full elastic anisotropy known to be significant in lyotropic liquid crystals. One of the
possibilities is omitting the L04 term. Hence, the minimal free energy model in this case
reads
1 ⇢⇤ ⇢ 2 1
f = ⇢C ⇤ a + ⇢Ca4
2 ⇢ +⇢ 4
203 A. dé Lozar, W. Schöpf, I. Rehberg, D. Svensek, L. Kramer, Phys. Rev. E 72, 051713 (2005). C. Blanc,
D. Svensek, S. Zumer, M Nobili, Phys. Rev. Lett. 95, 097802 (2005).
204 This relationship is of the form: K = a2 L0 + a2 L0 , K = a2 L0 + a4 L0 and K = a2 L0 + a4 L0 .
1 1 2 2 1 4 3 1 3
205
1 1 1
+ ⇢L1 (@i aj )2 + ⇢L2 (@i ai )2 + ⇢L3 ai aj (@i ak )(@j ak )
2 2 2
1 ⇥ ⇤ 2
+ G r · (⇢a) ⇢±
2
1 1
+ (⇢ ⇢0 )2 + L⇢ (r⇢)2 , (478)
2 2
where C is a positive Landau constant describing the isotropic-nematic phase transition
and ⇢⇤ is the transition density. Elastic constants Li for i = 1, 2, 3 can be expressed via
the Frank elastic moduli Ki , the bulk equilibrium density ⇢0 and a 205 , while and L⇢
are the density compressibility and the density variation correlation length. The nonlinear
density factor in the first term of (478) guarantees that the bulk nematic ordering stays
limited to |a| < 1. In the first five terms of the total free energy, corresponding to the
ordering and elastic parts of the free energy, we have taken into account the fact that they
need to be proportional to the number of molecules, i.e. to the local density.
The continuity requirement (475) is taken into account by means of the penalty poten-
tial proportional to a coupling constant G. For most of the time, the density of beginnings
and ends of chains, ⇢± , will not be considered as a variable but as a fixed external param-
eter. What physical meaning can be attributed to the parameter G that defines the rigidity
of the continuity constraint (475), as this parameter does not exist if the continuity con-
straint (475) is satisfied exactly? pOne notes that for the case ⇢± = 0, there emerges a
splay relaxation length scale l⇢ = G/ that controls the rigidity of the continuity con-
straint. Splay deformation of the director field at length scales much larger than l⇢ does
not bring about any substantial density variation, i.e. at this scale splay and density are
decoupled. On the other hand, on length scales much shorter than l⇢ , the coupling be-
tween splay and density becomes strong. This makes sense physically, as on large length
scales the (weak) divergence of the polymer current is compensated by a spontaneous
rearrangement of chain ends - which we are not taking into account - while this is not
effective at shorter length scales. It seems natural that l⇢ should be in direct relation with
the chain length.
11.8 Orientational configurations of DNA in a nematic nanodrop

Con- figurations like those in Figs. 89 and 90 are typical for regular nematics in spherical
confinement, e.g., nematic droplets. The variable density and its coupling to the splay
deformation in case of nematic polymers opens the door to a variety of new features
and configurations. The spontaneous appearance of depleted regions (phase separation)
is particularly interesting in this context. On the other hand we must not forget that by
using the polar ordering instead of the usual nematic quadrupolar ordering, we give up a
class of configurations containing defects of half-integer strengths.
T5 cryoelectron microscopy experiments show that in case of a loose packing inside a
bacteriophage capsid, the DNA toroids condensed with spermine sp4+ often occupy only
a part of the available volume rather than being spread over the entire capsid. Contrary
to the toroids observed in the bulk, the toroids confined to a viral capsid often show
no polar symmetry. The nematic and density field configurations obtained in our work,
205 In the form: L1 = K2 /(⇢0 a2 ), L2 = (K1 K2 )/(⇢0 a2 ) and L3 = (K3 K2 )/(⇢0 a4 ).
Figure 89. Depleted regions without polar symmetry in the cases with larger confin-
ing spheres. The same conclusion regarding the absence of polar symmetry in the en-
capsidated toroidal aggregate holds also in the case of anisotropic elasticity, where the
polymer is expelled from one polar region but not from the other. We note here that our
simplified macroscopic free energy, can not capture all the subtleties of the inter- seg-
ment interactions of a real nematic polymer such as DNA. Density variation cheaper, no
sources. Density variation cheaper, single source as in Fig. 3. Density variation cheaper,
sources placed on the z and x axes. The radius of the confining sphere is 32 (large).
containing pronounced depletion regions, show that the partial packing scenario could be
in principle attributed already to a macroscopic free energy of the type considered above,
as an alternative to the strictly cohesive energy between DNA segments.
We obtain depleted regions without polar symmetry in the cases with larger con-
fining spheres. The same conclusion regarding the absence of polar symmetry in the
encapsidated toroidal aggregate holds also in the case of anisotropic elasticity, where the
polymer is expelled from one polar region but not from the other. We note here that our
simplified macroscopic free energy cannot capture all the subtleties of the intersegment
interactions of a real nematic polymer such as DNA that are still not quantitatively known
with great precision. The mechanism of polyvalent- counterion mediated attraction keeps
defying proper understanding since apart from the charge of the counterions it appears to
be related also to ion-specific nonelectrostatic interactions that are notoriously difficult
to parameterize and model.
Though these results definitely point to the existence of states with partial depletion
regions in polymer packing with no polar symmetry, they also indicate that the inverse-
spool family of states is nevertheless quite robust within our model assumptions. This
finding is completely consistent also with recent molecular simulations. Strong confine-
ment for smaller confining spheres appears to induce toroidal packing irrespective of all
the other terms in the free energy and points to a universality of the inverse-spool con-
figuration. The observed noninverse-spool states might thus originate in other types of
interaction energy, e.g., between the DNA and various molecular moieties or charged
groups distributed along the inner surface of the capsid, effects which are not considered
in this work.
207
Figure 90. Strong confinement for smaller confining spheres appears to induce toroidal
packing irrespective of all the other terms in the free energy and points to a universal-
ity of the inverse spool configuration. The observed non-inverse spool states might thus
originate in other types of interaction energy e.g. between the DNA and various molecu-
lar moieties or charged groups distributed along the inner surface of the capsid, effects
which are not considered in this work. The director and density fields in the case of no
sources, ⇢± = 0. Dipolar sources: ⇢± = 1 in the green region, ⇢± = 1 in the red
region. A ⇢± = 1 (green) and ⇢± = 1 (red) source placed on the z and x axes. The
radius of the confining sphere is 10 (small).

Physics of DNA 1.0 PDF

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Physics of DNA 1.0 PDF

Diunggah oleh

Hak Cipta:

Format Tersedia

1

5.7 Euler angle of the pinched teardrop configuration . . . . . . . . 87

11.3 Viral capsids under mechanical stress . . . . . . . . . . . . . . 188

1 A.Vologodskii, Topology and Physics of Circular DNA, CRC Press (1992).

2.1 Mathematical description of spatial curves

ṫ(`) = (`)n(`) where ṫ(`) · t(`) = 0. (3)

(ṫ, ṅ, ḃ) = K (t, n, b), (9)

where the curvature matrix is given by

|t(`) ⇥ ṫ(`)| = |r̈(`)| = (`). (11)

Using the Frenet-Seret equations this leads to

2.2 Material frame of reference

2.3 Darboux vector

⌦(`) = ⌧ d3 (`) + 2 d2 + 1 d1 . (25)

2.4 Euler angle parametrization

(d1 (`), d2 (`), d3 (`)) = R(`)(d1 (0), d2 (0), d3 (0)) (26)

Figure 3. Definition of the precession, nutation and self-rotation Euler angles , ✓,

where the rotational matrix R is given by

⌦1 = 1 = ✓˙ sin ˙ sin ✓ cos

2.5 Single and double helix

r( ) = (R cos , R sin , c ) (32)

This is then the general definition of a helical curve in space.

2.5.2 Continuous double helix

r1 (`) = (a cos `1 , a sin `1 , c`1 ) and r2 (`) = (a cos `2 , a sin `2 , c(`2 +

DNA handedness n ! h[nm] a[nm] ✓

2.5.3 Left and right-handed helix

r( ) = (R cos , ±R sin , c ) (39)

2.5.4 Rotational symmetry and handedness

2.6 Călugăreanu-White-Fuller theorem

2.6.1 The linking number

r0 (s) r(s) = ✏n(s), (42)

2.6.2 Derivation within a magnetic analogy

Ba = (0, 0, Bz (r)) and Bm = (0, B✓ (r), 0). (44)

where we introduced the writhe of the curve

where we introduced the twist of the helix defined by

and Its Applications, 150, 293 (2009).

2.6.3 Twist and writhe of a continuous helix

2.6.4 DNA linking number

198107-1 (2004). E. L. Starostin, arXiv:physics/0212095v1.

2.6.5 Twist and writhe of open curves

DNA, Ann. Rev, Biphys. Biomol. Struct. 23 609 (1994).

2.6.6 A local expression for writhe

Since furthermore in this parametrization

we obtain for the writhe of the curve

Phys Rev E 78 041912 (2008).

⌧ = t(s) · (n(s) ⇥ ṅ(s)) (61)

expression by a Lagrange multiplier and inserting it in the complete Hamiltonian of the

2.6.7 Writhe and magnetic monopoles

3 MOLECULAR STRUCTURE OF DNA

• DNA is composed of a periodic array of molecular residues that scatter X-rays

Reviews of Modern Physics 79 943 (2007).

3.1 X-ray scattering

I(Q) = F(Q)F ⇤ (Q) = |F(Q)|2 . (69)

Assume now that we have N pointlike scatterers in crisytalline arrangement located at

and thus that Q is a reciprocal lattice vector. It can be written as

Q = Qnmj = na⇤ + mb⇤ + jc⇤ , (76)

n = 2dnmj sin ✓. (80)

Here we introduced also the scattering intensity of the crystalline lattice as

3.2 Structure factor of a linear polymer chain

where we used the identity

3.3 Structure factor of a single helix

on a helix, Acta Crystallogr 5 581-586 (1952).

3.4 Scattering intensity of an orientationally averaged single helix

Še ContourPlot narisan z Mathematico.