Вы находитесь на странице: 1из 69

Biological Physics

Ray Goldstein
(Zachary Ulissi)
Michaelmas Term, 2009-2010
Contents
1 Introduction 3
1.1 Recommended books . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Overview of Course . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Microscopic Physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.2 Fluctuations and Fluctuation Induced Forces . . . . . . . . . . . . . . . . 3
1.2.3 Elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.4 Chemical Kinetics and Pattern Formation . . . . . . . . . . . . . . . . . . 3
1.2.5 Bioconvection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Microscopic Physics 5
2.1 Review of Molecular Physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Quantum-mechanical argument for attraction among neutral molecules (Dipole-
dipole interactions) [8] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.1 Attraction of other neutral objects . . . . . . . . . . . . . . . . . . . . . . 8
2.2.2 Attraction of Finite Slabs and Spheres . . . . . . . . . . . . . . . . . . . . 8
2.2.3 Competition between electrostatic attraction and van der Waal interactions 9
2.3 Screening Eect of Water on Ion-Ion Interactions (Debye-Huckel) . . . . . . . . . 10
2.3.1 Potential Between Two Surfaces . . . . . . . . . . . . . . . . . . . . . . . 11
2.3.2 Fixed charge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Surface tension and wetting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5 Long, linear, charged objects (e.g. DNA) [14] . . . . . . . . . . . . . . . . . . . . 13
2.6 Geometrical aspects of screened electrostatic interaction . . . . . . . . . . . . . . 14
2.6.1 Geometric comparison method [20] . . . . . . . . . . . . . . . . . . . . . . 14
2.6.2 Surface free energy in Fourier space . . . . . . . . . . . . . . . . . . . . . 15
2.6.3 Perturbation method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3 Fluctuations and Fluctuation-Induced Forces 17
3.1 Brownian Motion[2, 5] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 Review of Statistical Physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3 Review of Polymer Physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.3.1 Simplest model of polymers: Freely-Jointed Chain . . . . . . . . . . . . . 19
3.3.2 More realistic freely jointed chain . . . . . . . . . . . . . . . . . . . . . . . 20
3.3.3 Entropic surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3.4 Energy Calculations in Fourier Space . . . . . . . . . . . . . . . . . . . . . 21
3.3.5 Check against physical systems . . . . . . . . . . . . . . . . . . . . . . . . 22
3.3.6 Fluctuations of an interface in a gravitational eld . . . . . . . . . . . . . 22
3.4 Brownian Motion and Diusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.4.1 Brownian particle in harmonic force eld . . . . . . . . . . . . . . . . . . 23
3.4.2 Brownian Diusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.4.3 Brownian Motion of Polymers . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.4.4 Example of Nearest Neighbor Interaction . . . . . . . . . . . . . . . . . . 26
3.4.5 Impenetrability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.4.6 Flory Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1
4 Elasticity 29
4.1 Curve Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.1.1 Lagrangian Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.1.2 Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.1.3 General 2D curve dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.1.4 Example: The curve shortening equation . . . . . . . . . . . . . . . . . . 31
4.1.5 Global constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2 Space Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2.1 Example: Vortex Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2.2 Example with a Variational Principle . . . . . . . . . . . . . . . . . . . . 33
4.2.3 Hasimotos trick . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2.4 Generalized Frenet-Servet equations . . . . . . . . . . . . . . . . . . . . . 34
4.3 Viscous Drag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.3.1 Typical elastic coecients for biological systems . . . . . . . . . . . . . . 34
4.4 Elastohydrodynamics [19, 9] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.5 Comparison to Stokes Problems (Stokes, 1851) . . . . . . . . . . . . . . . . . . . 35
4.5.1 Post-Transient Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.5.2 Biophysics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.6 Elastohydrodynamics, continued . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.6.1 Reversibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.7 Euler Buckling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.7.1 Propulsion Mechanisms of Flagella . . . . . . . . . . . . . . . . . . . . . . 40
4.8 Further Elasticity [9] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.8.1 The Strain Tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.8.2 Stress Tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.8.3 Generalization of Hookes Law . . . . . . . . . . . . . . . . . . . . . . . . 42
4.8.4 Homogeneous Deformations . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.9 Torsion of Rods [9] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.9.1 Viscously induced twist . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.10 More General Elasticity Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.11 Twisting and Bending . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.11.1 More general equilibrium conditions . . . . . . . . . . . . . . . . . . . . . 48
4.11.2 Moments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.11.3 Circular rods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.11.4 Torsional Instability [9, 7] . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5 Chemical Kinetics and Pattern Formation 50
5.1 Michaelis-Mentin kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.1.1 Biologists method of solution . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.1.2 Example: Cooperativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.1.3 Example: Slaving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.1.4 Diusive eects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.2 Phenomenology of Reaction-Diusion Systems . . . . . . . . . . . . . . . . . . . . 57
5.3 Fitz-Hugh Nagumo Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.3.1 Variants of the FHN model . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.4 Separation of Timescales / Method of Multiple Scales . . . . . . . . . . . . . . . 59
5.5 Front Velocity Via a Solvability Condition . . . . . . . . . . . . . . . . . . . . . . 60
5.6 Pendulum, continued . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.7 Application to PDEs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.8 Front Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6 Bioconvection 64
6.1 Gyrotaxis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.2 Ordinary Diusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2
Chapter 1
Introduction
1.1 Recommended books
1. What is life? (Schrodinger)
2. Stochastic Processes
3. The Theory of Polymer Dynamics
4. Mathematical Biology (Murray)
5. Molecular Biology of the Cell
6. Lectures on the Physiology of Plants
1.2 Overview of Course
1.2.1 Microscopic Physics
Inter-molecular attraction (Van der Waals, Lennard Jones potentials, uctuating dipoles,
etc.)
Charged Particles in Solution (Debye-Huckel theory)
Membranes of a cell (deformed, charged sheets of positive ions)
Bending energy of membranes
1.2.2 Fluctuations and Fluctuation Induced Forces
Dynamics of polymer chains and DNA (entropic springs)
Brownian Motion (stochastic dierential equations)
1.2.3 Elasticity
Curve dynamics, elasticity, bending energies, curve shortening, elasticity in higher dimen-
sions (membranes, etc.)
1.2.4 Chemical Kinetics and Pattern Formation
Simple kinetic models (Michaelis Mentin, etc.)
Reaction/diusion equations, slaving, multiple time scales
Interface dynamics
Pattern formation in the Fitz-Hugh Nagumo model
3
1.2.5 Bioconvection
Gyrotaxis
bio-convection/diusion equations
Instabilities
4
Chapter 2
Microscopic Physics
2.1 Review of Molecular Physics
The ideal gas law, as normally used, is:
PV = nRT = Nk
B
T
P
k
B
T
=
Where P is the pressure, V is the volume, T is the absolute temperature, is the number density,
R is the ideal gas constant, N is the number of molecules, n is the number of moles, and k
B
is
the Boltzmann constant. However, is often not linearly related to the pressure. Expanding out
the ideal gas law as a function of (a Virial expansion)
P
k
B
T
= +B
2
(T)
2
+B
3
(T)
3
+. . .
Intuitively, the rst added term refers to pairwise coupling eects, the second to triplet eects
etc. Truncating after the second term:
P
k
B
T
= +B
2
(T)
2
Experimental determination of the coecient B
2
for relatively ideal gases yields the following
qualitative curve (left):
Initial attempts to understand molecular interactions resulted in a general understanding of
interaction potential (above,right). Van der Waal made a mean eld averaging argument in the
derivation of his eponymous equation of state. He assumed that below a certain critical radius
the interaction potential was essentially innite, and above the attraction beyond that radius
could be characterized by an integral term. Essentially:
a
1
2
_
d
3
rU
attr
(r)
5
Where a is then a characteristic constant for a given species, temperature, etc. The total energy
of interaction for a system of N particles is:
U
Attraction
= aN = N
1
2
_
d
3
rU
attr
(r)
The attraction pressure will be:
P
attr
=
U
attr
V
= a
2
The rst guess at a solution is then:
P = k
B
T a
2
+. . .
Further, Van der Waal realized that a central region of the attraction potential should not
be considered in the volume integrals from before. He thus added another term to his expansion
to arrive at the familiar form for weakly interacting molecules:
(P +a
2
)(V Nb) = Nk
B
T
where Nb is eectively the total volume of the particles (the appropriate excluded volume
due to strong repulsion). Expanding for small densities:
P
k
B
T
+ (b
a
k
B
T
)
2
+. . .
This suggests the Boyle point will be at
T
B
=
a
k
B
b
Another way of viewing this is in terms of the radial distribution function g(r). The RDF is
shown below for a semi-realistic (Lennard Jones) ideal gas and for the assumed gas of Van der
Wall.
2.2 Quantum-mechanical argument for attraction among
neutral molecules (Dipole-dipole interactions) [8]
One way to obtain a semi-quantitative understanding of the attraction between two polarizable
(but neutral) molecules is viewing the problem as two linear springs:
6
The Hamiltonian for the system is:
1 = 1
0
+1
1
= Spring Energy + Coulombic Energy
1
0
=
p
2
1
2m
+
1
2
m
2
0
x
2
1
+
p
2
2
2m
+
1
2
m
2
0
x
2
2
1
1
= e
2
_
1
R
+
1
R +x
1
+x
2

1
R x
1

1
R +x
2
_

2e
2
x
1
x
2
R
3
, [x
1
[, [x
2
[ R
To simplify the situation, a coordinate change is made:
x

=
x
1
x
2

2
x
1
=
x
+
+x

2
x
2
=
x
+
x

2
1 =
p
2
1
2m
+
1
2
_
m
2
0

2e
2
R
3
_
x
2
+
+
p
2
2
2m
+
1
2
_
m
2
0
+
2e
2
R
3
_
x
2

The two terms in parenthesis are essentially adjusted frequencies:

2
+
=
2
0

2e
2
mR
3

2

=
2
0
+
2e
2
mR
3
1 =
p
2
1
2m
+
1
2
_
m
2
+
_
x
2
+
+
p
2
2
2m
+
1
2
_
m
2

_
x
2

The energy is then:


U(r) =
1
2

+
+
1
2

2
1
2

0

e
4
2m
2

2
0
R
6
=
1
2

0
(e
2
/m
2
0
)
2
R
6
=
1
2

0
(e
2
/m
2
0
)
2
R
6
noting that e
2
/m
2
0
is a characteristic volume. Thus, the energy is related to the diculty
of polarizing the atom, as well as an eective volume. This can be checked using standard
electromagnetics. The normal equation for magnetic displacement is:
d = E
The units on d are Q L, the units on E are Q/L
2
, making the units of L
3
(an eective volume).
The Hamiltonian is:
1 = 1
0
+eE
0
x
1
+eE
0
x
2
=
1
2
m
2
0
_
x
2
1
+
2eE
0
x
1
m
2
0

_
eE
0
m
2
0
_
2
_
+ (1 2)
=
1
2
m
2
0
z
2
1
+
z
1,2
= x
1,2
+
eE
0
m
2
0
The dipole moment (i.e. the displacement due to the electric eld) is then:
=
e
2
m
2
0
which is the same as seen previously. The energy of attraction in either case is then:
U(r) =
1
2

2
r
6
7
2.2.1 Attraction of other neutral objects
The attraction between two neutral sheets of atoms can be inferred from the calculated scaling
through a progression of geometries:
First, the attraction between two point molecules is referred to as:
V
11
(r) =
C
r
6
Then, the attraction between a neutral atom and a slab is calculated:
V
1S
(h) =
_

h
dz
_
2
0
d
_

0
rdrV
11
(
_
z
2
+r
2
)
=
_

h
dz
_
2
0
d
_

0
rdr
C
(z
2
+r
2
)
3
=
C
6h
3
(atom-slab)
Finally, extending to two neutral sheets (Adz atoms each per thickness dz):
V
ss
= A
_

h
V
1s
(z)dz
V
ss
A
=
A
H
12
1
h
2
where A
H
is the Hamaker constant
A
H
=
2

2
C
12

2
()
2
A
H
scales with
2
()
2
, allowing an estimate of 5 10
20
J, which is an order of magnitude
larger than thermal energy at 300K. Just as importantly, this interaction scales with the square
of the distance, permitting relatively large range forces. This work is related to DLVO theory
(Derjagum, London, Verway, Overbeck) [18].
2.2.2 Attraction of Finite Slabs and Spheres
The point-point interaction is given as
V (r) =
C
r
6
The point-plane interaction is just an integral of this
V (d) =
_
/2
0
d(d tan )2
_
C
(d cos )
6
_
=
2C
d
5
_
/2
0
d cos
5
sin
=
2C
d
5
_
cos
6

6
_
/2
0
=
2C
6d
5
=
C
3d
5
8
A slab-point interaction is just an integral over plane-point interactions with multiple distances
_

1
0
dz
C
3(d +z)
5
=
C
12
_
1
(d +
1
)
4

1
d
4
_
The interaction of a slab with a plane will just be the eect of all the individual points and the
slab
V =
C
12
A
Finally, the slab-slab interaction will be
V =
C
2
12
_

2
0
dz
_
1
(d +z +
1
)
4

1
(d +z)
4
_
=
C
2
36
_
1
(d +z +
1
)
3

1
(d +z)
3
_

2
0
=
C
2
36
_
1
(d +
2
+
1
)
3

1
(d +
2
)
3

1
(d +
1
)
3
+
1
d
3
_
As a check, if either
1
or
2
go to zero, the solution does as well.
Add sphere-
Sphere inter-
action from
the hw solu-
tions
2.2.3 Competition between electrostatic attraction and van der Waal
interactions
The total interaction between surfaces is a combination of screened electrostatics and van der
Waal interactions. As the electrostatics fall o exponentially (to be shown later) and are nite at
r = 0, the van der Waal forces are more important at very short and very long ranges. However,
for intermediate r, electrostatics can be important:
A naive estimate of the energy of electrostatic interactions is obtained using Coulombic in-
teractions:
Energy c = e(r) = e
2
/r
Where r is usually on the order of 0.1 nm. Thus, the electrostatic energy, compared to
thermal energy, is:
c
k
B
T
=
(4.8 10
10
)
2
4 10
14
10
8
cm
=
580
r[

A]
However, the dielectric constant of water is about 80 (before counting the eect of screening
charges). Thus, the ratio is more on the order of 7/r

A. This suggests a characteristic length, the


9
Bjerum length, which represents the separation at which the electrostatic interaction between
two elemental charges is comparable to the thermal energy scale k
B
T
Bjerum Length =
B
, l
B
=
e
2
k
B
T
7

A
At 300K, l
B
is about 7

A.
2.3 Screening Eect of Water on Ion-Ion Interactions (Debye-
Huckel)
Even pure water has small amounts of ionic species (H
+
/OH

) at equilibrium. These ions will


have a screening eect on any interactions, eectively forming a neutralizing layer around any
charged species. This interaction can be modeled through the Poisson-Boltzman equation, named
after the two equations on which it is based:
Poisson Equation
2
=
4e

, E =
Boltzman Distribution C
s
= C
0
e
E/k
B
T
= C
0
e
z
s
e/k
B
T
Combining these two equations into a self consistent equation (and letting = 1/k
B
T):

2
=
4

spec.
z
s
eC
0
e
z
s
e
Consider the specic case of a z : z electrolyte (1:1, NaCl, 2:2, CuSO
4
etc.):

2
=
8zeC
0

sinh(ze)
In the weak eld or small limit:
sinh(x) =

n=1
x
2n+1
(2n + 1)!
x
2
=
8z
2
e
2
C
0
k
B
T
+. . .
This suggests a characteristic length scale, the Debye-Huckel length:

2
DH
=
k
B
T
8z
2
e
2
C
0

10nm
C/1[mM]
This then suggests an equation:
(
2

2
DH
) = 0
In one dimension away from a surface, this becomes:
_

2
x
2

1

2
DH
_
= 0
The general solution for this is:
= Ae
x/
DH
+Be
x/
DH
Since the eld must decay to zero at innity, the rst coecient A = 0:
=
0
e
x/
DH
The induced charge on the surface is related to the associated electric eld:
n [
surf
=
4
e

0
=

4
DH

0
10
In this case, the relationship is linear, but this is not always the case.
It is interesting to note that
2

2
DH
= 0 is the Euler-Lagrange equation for the free
energy T functional:
T =

4
_
d
3
r
_
1
2
()
2
+
1
2
2
DH

2
_
=

4
_
d
3
r [Electric. Energy + Entropy Cont.]
with the general Euler-Lagrange formula being

x
/

x
+
/

= 0
Thus, the free energy is just the sum of the contributions of the electrical energy and the entropy
contribution.
() = () +
2

_
bulk
() =
_
bulk
() +
_
bulk

_
surf
n =
_
bulk
()
2
+
_
bulk

_
()
2
=
_
surf
n
_
bulk

So that the free energy is just


F =
_
surf
n
1
2
_

_

2

1

2
DH

_
=
1
2
_
dS (Poisson term = 0)
This makes intuitive sense, and is the appropriate solution for a surface of xed potential.
surface for
xed charge;
which needs
Legendre
transforms
of xed
energy be-
tween vari-
ables
For situations with xed charge rather than xed potential, the surface energy will istead be:
F
C
= F +
_
=
1
2
_
dS or
_
ds
_

0
(
1
)d
1
2.3.1 Potential Between Two Surfaces
For two surfaces of equal potential
0
:
Assuming the potential decays exponentially from both surfaces, the (symmetric) potential dis-
tribution will be:
=
0
cosh(x/)
cosh(d/2)
11
The charge at each surface can then be calculated from the divergence of the potential:
(d/2) =
c
4

tanh(d/2)
And the free energy by integrating up over the charge on the surface:
F
2A
=
1
2
_
dS =
c
2
8
tanh(d/2)
A more interesting quantity is the dierence between the free energy and the free energy of
the two surfaces at innite distance:
F(d) F()
2A
=
c
2
0
8
[1 tanh(d/2)]
At very large separations (d ), the potential will be:
F
2A
=
c
2
0
8
e
d/
+. . .
However, for very small separations, the potential instead goes to zero, since there is no break
in the material.
2.3.2 Fixed charge
As shown before, there will be an exponentially decaying potential
=
0
e
x/
The potential at the surface can then be calculated
4
0
c
=

0


0
=
4
0
c
=
4
0
c
e
x/
The energy of interaction, as shown before, will then be:
F
A
=
1
2

0
=
2
2
0
c
Where the factor of two comes from the build-up of charges (similar to the integrals resulting in
capacitor voltages).
The charge between two surfaces of xed charge is then:
F
2A
=
2
2
0

c
[coth(d/2) 1]
0
As expected, this energy blows up to innity, since the charges are forced together. For more
reasonable distances, how strong are these interactions? A typical lipid has a cross sectional area
of about 50-100 cm
2
. If each head holds a single charge, and the DH length is about 1nm,
then the typical energy is:
F
A
50
erg
cm
2
Surface Tension of Water (80[erg/cm
2
])
So, these eects are comparable to surface tension etc., and thus quite relevant.
2.4 Surface tension and wetting
12
The energy per cross sectional unit in the third, perpendicular dimension (E/L) is dened as:
E
L
=
_
dx
_
_

1 +
_
h
x
_
2
+
1
2
gh
2
_
_
where is the surface tension. For small displacements [h(x)[ 1, the square root can be
expanded:
_
1 +h
2
x
= 1 +
1
2
h
2
x
+. . .
E[h] E
0
L

_
dx
_
1
2
h
2
x
+
1
2
gh
2
_
Generalizing this to a continuous planar surface, h(x, y) (dropping the assumption of translational
invariance)
E =
_
dx
_
dy
_

2
(h)
2
+
1
2
gh
2
_
This is the same as the equation above for forces near a charged/conducting wall. There is then
a characteristic length, called the capillary length,
1
l
2
c

l
c

_

g
which, for water/air, is about 3mm (
_
100/1/1000).
2.5 Long, linear, charged objects (e.g. DNA) [14]
As a rst approximation, consider a line of even charge:
Charge Density = =
z
p
e
b
where z
p
is the valence and e is the charge of an electron. In 2D cylindrical coordinates, the
energy of a test charge is:
U
ip
= z
i
e
2

ln(r)
If we assume a probability density concentration based on this energy:
e
U
ip
(r)
= f
2z
i
z
p
(e
2
/k
B
T)/b
= r
2z
i
z
p
l
B
/b
where l
B
is the Bjerum length. The integral of this probability density is then:
_
2dre
U
ip
(r)

_
drr
1+2z
i
z
p
lB/b
If the test charge is a counterion (z
i
z
p
< 0) and l
B
/b [z
i
z
p
[
1
the normalization will
fail at the origin. This is accounted for with Manning condensation (counter-ion collapse or
condensation) [14] which cancels the bare charges, reducing to the edge of convergence.
Read this
paper
13
2.6 Geometrical aspects of screened electrostatic interac-
tion
The idea is to use a perturbation method assumes weak curvature /R (in 1D, , in 2D H =
/R
1
+ /R
2
, = 1/R
1
R
2
) where H = 1/R
1
+ 1/R
2
, = 1/R
1
R
2
. There are three possible
levels of computing the surface free energy:
c =
_
dr
_
1
2
k
c
(H H
0
)
2
+
1
2
k
c
k
_
1. Compute the energy for regular shapes ([20, 11, 15]) and compare to nd the relative terms.
2. Construct a perturbation theory around a at surface [6, 4]
3. Multiple scattering method (very hard, not covered here)
2.6.1 Geometric comparison method [20]
Debye Huckel theory relies on a simplied version of the Poisson-Boltzmann equation. In the
weak eld limit, the dierential equation simplies to
(
2

2
DH
) = 0
Solution for a Plane The solution for the potential outside of a plane was developed in
class, and led to
=
_

0
e
(xa)/
DH
x > a

0
e
(ax)/
DH
x < a
For a surface of xed charge, the potential at the surface will be

0
=
DH

The associated energy, as discussed in class, is determined through the energy necessary to build
up the relevant charge
E =
_

0
d =
_

0

DH

e
(xa)/
DH
d =

2
=

2

DH

Solution for a Cylinder


Expanding the Laplacian (for only the radial term, due to symmetry)

2
=
1
r

r
_
r

r
_
=

r
=

DH
14
Solutions to dierential equations of this form are Bessel functions. The solutions will be
=
0
I
0
(r/
DH
)
I
1
(r/
DH
)
The coecient
0
(from Hunter) for xed charge is

0
=
DH

As before, the electrostatic energy is just the integral from the buildup of charge
E =
_

0
d =
_

0

DH

I
0
(r/
DH
)
I
1
(r/
DH
)
d =

2

Solution for a Sphere


Expanding the Laplacian (for only the radial term, due to symmetry)

2
=
1
r
2

r
_
r
2

r
_
=

+ 2

r
=

DH
The general solution is
= A
e
r/
DH
r
Since the solution that decays at innity is desired (outside the sphere), one solution can be
eliminated so that
= A
e
r/
DH
r
For a sphere of radius b with xed potential
0
, this becomes:
=
0
b
e
(rb)/
DH
r
Inside of the sphere, the potential must be bounded at zero so that the two terms cancel.
=
0
a
e
a/
DH
e
a/
DH
_
e
r/
DH
+e
r/
DH
r
_
Simplifying to hyperbolic functions and incorporating solutions for both inside and outside the
sphere
=
_
=
0
a
sinh(r/
DH
)
r cosh(a/
DH
)
r < a
=
0
b
exp[(rb)/
DH
]
r
r > b
For xed charge, the potential (taken from Hunter) will be

0
=

DH
(1 +
DH
/a)
How are the
terms de-
rived from
these poten-
tials?
2.6.2 Surface free energy in Fourier space
In fourier space:

(n)
(k, y) =
_
dxe
ikx

(n)
(x, y)
The transform of the height function is:

h(q) =
_
dxe
iqx
h(x)
15
Looking at the 2D Debye-Huckel equation:
(
xx
+
yy

2
) = 0
In Fourier space this becomes:
(
yy

2
q
)

(n)
(q, y) = 0
where
2
q
=
2
+q
2
. This is esentially just an ODE in Fourier space. The solution is:

(n)
(q, y) =

(n)
(q, 0)e

q
y
The inverse Fourier transform of this may get tricky, but it should be possible for long
wavelength modes of . The O(
1
) term immediately becomes

h(q) and the O(
2
) term is just a
convolution. Some comments:
(x) =

4
n (x, h(x))
where the normal is:
n =
hx e
x
+ e
y
_
1 +h
2
x
E(h(x)) will then yield the energy.
2.6.3 Perturbation method
Perturbation
method - in
the notes,
from [6], or
from the
HW solu-
tions
16
Chapter 3
Fluctuations and
Fluctuation-Induced Forces
3.1 Brownian Motion[2, 5]
The common diusion equation, for a species concentration, is:
C
t
= D
2
C = (DC) = J
where J is the ux from Ficks law (J = DC). Why is the diusion coecient of a molecule
in water 10
5
cm
2
/s? From the book by Doi and Edwards [3], in 1D, the total ux is:
J = J
Diusive
+J
Force
= D
dC
dx
+Cv = D
dC
dx
+C
F

where v is the speed. That is, the total ux is equal to a contribution from the concentration
gradient and a contribution from a limiting force. Since this is at small length scales (and thus
low Reynolds number), inertia should be negligible and the forces should balance. can be
dened as a drag coecient:
v = F =
d
dx
= F
If the force of the particle is just Stokesian drag on a sphere:
F = 6R
J = D
dC
dx

c

d
dx
At equilibrium, the total ux is 0 so the two terms can be integrated up:
D
dC
dx
=
c

d
dx
C = C
0
e
/D
e
E/k
B
T
That is, the concentration is related to some potential, which can be thought of as an entropic
contribution. This leads to the famous Stokes-Einstein relation:
D = k
B
T D =
k
B
T

=
4 10
14
6R
= 10
5
cm
2
/s
Thus, once the diusion coecient is measured (not hard), as well as the gas constant and
the Boltzman constant, it is possible to infer Avagadros number and show that there must be
molecules within solution (which was still controversial at the time) [5].
read Ein-
steins paper
17
3.2 Review of Statistical Physics
The probability p(E) of nding a system in a state of energy E is given by:
p(E) e
E
To normalize this probability, the partition function : is dened:
: =

i
e
E
i
where i is the index of all possible states. The real probability is then:
p(E) =
e
E
:
And the average of a measure A (related to the energy) is:
A =

i
A(E
i
)
e
E
i
:
: can be used as a generating function. For example, in the case of a quantum mechanical
system (single particle), the energy is:
E =
p
2
2m
+U(r)
The partition function is then:
: = A
_
dp
x
_
dp
x
_
dp
z
_
d
3
re
[p
2
/2m+U(r)]
:
mom
:
conf
Where the integrals run over all possible momentums and positions and additivity allows the
splitting of partition function into :
mom
and :
conf
. There are several coecients hidden in A,
but as the probability is usually normalized by the total sum, these are not necessary.
The equipartition theorem states that all energy will be distributed evenly among available
modes. For example, in the case of intermolecular interactions with a quadratic energy relation-
ship:
E =
1
2
kx
2
:
conf
=
_
dxe
kx
2
/2
The average kinetic energy is then:
_
1
2
kx
2
_
=
_
dx
1
2
kx
2
e
kx
2
/2
_
dxe
kx
2
/2
=
ln :
conf

The partition function : can be simplied by changing variables from x


2
to q
2
, removing the
constants in the exponential (the trick is simply multiplying by 1):
:
conf
=
_
dxe
kx
2
/2
_
k
2
_
2
k
=
_
2
k
_
dqe
q
2
ln : can then be calculated, yielding one term with and other terms independent of , and
dierentiated to yield the average energy
ln :
conf
=
ln
2
+ Stu
ln :
conf

=
1
2
=
1
2
k
B
T
Thus, the energy per mode or degree of freedom is
_
1
2
kx
2
_
=
1
2
k
B
T
which is precisely the equipartition theorem. For three dimensions of freedom (real space), this
would be 3k
B
T/2.
18
3.3 Review of Polymer Physics
Polymers are long chain molecules with repeating units. A few examples are:
1. DNA: Covalently bonded polymer with nucleotide units
2. Microtubules: Aggregates of protein monomers, held together with electrostatics and van
der Waal forces. Due to the weak bonding these polymers uctuate often in length.
3. Linear/branched: Most proteins are single chains, but molecules with multiple attachments
are possible.
4. Cross-linked networks: interlocked chains of polymers are necessary for many cellular func-
tions.
An example of a polymer is styrene, with monomers of ethyl benzene. Since rotation around
single bonds is possible, there are degrees of freedom for internal rotation.
Both short and long range interactions are possible. For example, one region might interact
with a polymer region far down chain, but locally chemical interactions are more important. The
connections between monomers determine things like stiness and self avoidance.
Missing
background
information
on chem-
istry, rota-
tional de-
grees of free-
dom, etc.
3.3.1 Simplest model of polymers: Freely-Jointed Chain
The simplest model of a polymer is a jointed chain of N links, each of which can be up or down:
The energy of a conguration is related to the length of the chain z
E = mgz z =
N

n=1
bs
n
where s
n
is 1 The total number of possibilities is then:
:
n
=

s
n
=
e
bmgs
n
= [2 cosh(mgb)]
N
The probability of nding a particular extension state is then:
z = b

N
s
n
=
ln :
n
(mg)
= Nb tanh(mgb) = Ltanh(mgb)
resulting in the following curve (approximate):
For a weak gravitational force, the length z can be Taylor expanded
z =
Nb
2
mg
k
B
T
+
_
Nb
2
k
B
T
_
mg
19
This is analagous to a Hookian spring force (F = kx, with linear constant k):
k =
k
B
T
Nb
2
This result is especially interesting because the spring force comes entirely from entropic forces.
Work is done to reduce the entropy when the chain is extended. For example, at full extension
there is only one conguration, so a minimum in entropy.
3.3.2 More realistic freely jointed chain
A more realistic model is to have every link freely jointed, with an arbitrary force F:
The new length and energy are:
x =
N

n=1
b cos
N
E[
N
] = Fb
N

n=1
cos
N
The partition function (analagous to above) is:
:
N
=

N
_
d
N
sin
n
2e
F cos
N
f =
Fb
k
B
T
:
N
=
_
4 sinh(f)
f
_
N
This can then be used to nd the average extention
x
b
=
(ln :
N
)
f
x = bN

f
ln
sinh(f)
f
= NbL(f) L(f) = coth(f)
1
f
The spring constant for this new chain is then:
k =
3k
B
T
Nb
2
which makes sense, since there are two more degrees of freedom than in the simple case above.
This is exactly analagous to molecular velocity distributions, with each degree of freedom having
an energy of k
B
T/2.
3.3.3 Entropic surfaces
This model can be generalized to continuous surfaces. Starting with a simple 1D string:
The energy of this surface, relative to the system energy of a at surface, is:
E =
_
_
L
0
dx
_
1 +h
2
x
L
_
20
Where h
x
is basically the arc length of the segment and is the equivalent of a surface tension
coecient. Expanding the square root
E

2
_
L
0
dxh
2
x
+. . .
If we then take the Fourier transform of the height function,
h(x) =

n
A
n
sin
_
nx
L
_
the energy of the surface is then just a sum over the energy of each constituent mode.
E =
1
2

m=1

n=1
A
n
A
m
_
n
L
__
m
L
_
_
L
0
dxcos
_
nx
L
_
cos
_
mx
L
_
=
1
2

m=1

n=1
A
n
A
m
_
n
L
__
m
L
_
_

mn
L
2
_
=
L
4

n=1
_
n
L
_
2
A
2
n
Thus, there is a system-sized dependence on the energy change, which is characteristic of diusive
systems. If the string is in thermal equilibrium (e.g. connected to a heat bath), then the average
of each coecient will be:
A
n
= 0
A
2
n
=
2k
B
T

2
n
2
L
where the A
2
n
comes from the energy term above
E
n
=
k
B
T
2
=

2
n
2
4L
A
2
n

The average deviation of the string can then be calculated:


h
2
(x) =

m
A
m
A
n
sin
_
mx
L
_
sin
_
nx
L
_
=

mn
A
2
m
sin
_
mx
L
_
sin
_
nx
L
_
=
2k
B
TL

n=1
sin
2
_
mx
L
_
n
2
As a check, has units of Energy/Length, so that the sum has units of Length
2
.
3.3.4 Energy Calculations in Fourier Space
Rather than doing summations over discrete modes, the same problem can be treated in Fourier
space. Starting with the initial energy again and introducing the Fourier representation of h
x
:
E =

2
_
dxh
2
x
h
x
=

q
e
iqx

h(q)
=

2
_
dx

(iq)(iq

)e
iqx
e
iq

h(q)

h(q

)
This can be simplied with:
not sure
about this
simplica-
tion, 0 might
be q in L
p,q
_
L
0
dxe
ipx
= L
p,0

h(q)

h(q

) =
L
2

q
q
2
[

h(q)[
2
21
Using the equipartition theorem:
_
[

h(q)[
2
_
=
k
B
T
L
1
q
2
(there might be another constant near the q in the denominator). The term h
2
is denoted as
the thermal average, while h
2
is the system average. The thermal average of the system average
is
_
h
2
_
=

q
[

h(q)[
2

In the continuum limit, 1/L

q

_
dq/2,
_
h
2
_
=
k
B
T
2
_
dq
q
2
Since this is poorly behaved as q 0, so a cuto is necessary. These usually occur in either of
two forms:
1. Small scale: Molecular lengths a limit small scale behavior
2. Large scale: The limited system size L prevents some behavior
Accounting for one such cuto:
_
h
2
_

k
B
T
2
_
1
q
min

1
q
max
_
q
min
= /L q
max
= /a
Considering the energy then:
E h
2
x

_
h
2
x
_

k
B
T

_
dq
q
2
q
2

k
B
T

(q
max
q
min
)
As we let L , q
min
0, and a 0, q
max
, so that
E
k
B
T

a
3.3.5 Check against physical systems
Since k
B
T 4 10
14
erg, 1 100erg/cm, a 10
8
cm, the average displacement energy
h
2
x
is on the order of 10
3
10
2
which is quite small.
3.3.6 Fluctuations of an interface in a gravitational eld
The new surface function in Fourier space and the associated energy is
h(r) =

q
e
iqr

h(q) =
_
d
2
re
i(q+q

)r
= A
q+q

,0
E =
_ _
dxdy
_
(h)
2
+
1
2
gh
2
_
=
A
2

q
(q
2
+l
2
c
)[

h(q)[
2
22
where l
c
is once again the capillary length. The average displacement is then:
_
h
2
_
=
k
B
T
2
=
k
B
T
4
ln
_
1 + 2
_
l
c
l
_
2
1 + 2
_
l
c
L
_
2
_
In the thermodynamic limit (L ), which is now possible at nite g,
h
2

k
B
T
4
ln
_
1 + 2
_
l
c
l
_
2
_
3.4 Brownian Motion and Diusion
The diusion equation and Stokes-Einstein relation, discussed earlier, was:
C
t
= D
2
C D =
k
B
T

where was a drag coecient. The diusion equations suggests natural time and length scales
of l

Dt. Diusion of a small molecule in water (D = 10


5
cm
2
/s, l = 30m

t shows that in
one second, each molecule will naturally move around 30m, comparable to the length of a cell.
However, a cell, which is more on the order of m, would only diuse very slowly. Thus, active
propulsion is necessary for cells to move any reasonable distance.
3.4.1 Brownian particle in harmonic force eld
Brownian motion can be investigated in modern laser trapping apparatuses, rst invented in the
1970s at Bell Labs [1]. The focus beam naturally converges on a small diraction limited region:
The force acting on the particle is then a combination of two phenomenes:
Stokes Drag = Optical Force + Fluctuating Forces
x

= kx +(t)
which is the Langevin equation. This essentially leads to a classical spring well with small
perturbations from Brownian motion
For m sized spheres and moderate lasers, the eective spring coecient k is 10 fN/nm. Laser
tweezers can also be used to measure the forces of moving particles. For example, attaching
spheres onto motor proteins allows the strength of interaction to be determined.
23
The stall force for motor proteins just happens to be on the order of a few pN. The time scale
of Brownian movement in the wells comes from the spring constant, = k
1
4ms. At the
same time, thermal energy is equivalent to k
B
T 4pN nm, which allows the length scale of
movement to be obtained. Rescaling the Langevin equation and using the integrating factor of
e
t/
allows a solution to be found
x +
1

x = (t) ( rescaled)
e
t/
_
x +
1

x
_
= e
t/
(t)
x(t) x
0
e
t/
=
_
t
0
dt

e
(tt

)/
(t

)
The noise, which is the deviation from the normal relaxation path, is obtained by taking the
thermal average
x(t) x
0
e
t/
=
_
t
0
dt

e
(tt

)/
(t

) = 0 x(t) = x
0
e
t/
The integral term is zero since we expect the thermal average of the noise to be zero. However,
the square of the deviation will be nonzero
(x(t) x
0
e
t/
)
2
=
_
t
0
dt

_
t
0
dt

e
(tt

)/
e
(tt

)/
(t

)(t

)
Assuming that the last term is a sharply peaked function of [t t

[, it can be represented as a
function (t

). With the change of variables


s = t

+t

q = t

The right hand side of the previous equation will then be


_
t
0
dt

_
t
0
dt

e
(tt

)/
e
(tt

)/
(t

)(t

) =
1
2
e
2t/
_
2t
0
dse
s/
_

dq(q)
where the nal integral is just a number (). The average deviation squared is then
(x(t) x
0
e
t/
)
2
=

2
(1 e
2t/
)
In the long time limit (t/ )
(x(t) x
0
e
t/
)
2
= x(t)
2
=

2
However, using the equipartition argument
1
2
kx
2
=
1
2
k
B
T =
2k
B
T

Logically, we assume that there is a correlation


(t)(t

) =
2k
B
T

(t t

)
24
The resulting time scale is on the order of kHz for normal biological systems. In order to conrm
this result, the implication for small time movement is derived. For no initial displacement and
short t,
x
2
(t)

2
_
2t

+. . .
_

2k
B
T

which is just a random walk in 1D (x


2
= 2Dt). Thus
D =
k
B
T

and the Stokes-Einstein relation is recovered.


As an aside, we can justify the assumption of no inertia for small particles. If a 1m bacteria
moving under its own power at 10m/sec stops propulsion, the bacteria will stop within a
subatomic distance. Thus, there is negligible inertia and no drifting at these length scales.
3.4.2 Brownian Diusion
The diusion coecient is just the average of the movement rate per time at long times
D = lim
t
1
6t
[(r(t) r(0)[
2

where r(t) is
r(t) = r(0) +
_
u(t

)dt

The diusion coecient term above holds provided that the correlation of velocities (u(t

)u(t

)
falls o fast enough. This yields
D =
1
3
_

0
dtu(t) u(0)
For E. Coli the average velocity is about 20m/s, and the bacteria executes 1s of movement
before randomly changing direction. This yields a diusion coecient of 410
6
cm
2
/s, which is
approximately the diusion coecient of a small molecule in water. Thus, E. Coli use propulsion
to simulate the diusion that their large size does not allow, allowing them to eectively explore
a region.
3.4.3 Brownian Motion of Polymers
Consider an arbitrary free polymer
with each segment labeled as r
n
. Each segment is followed by another random segment of equal
length
r
n+1
= r
n
+
n
The total and average length of the polymer is then
r
N
r
0
=
N

n=1

n
r
N
r
0
=
N

n=1

n
= 0
25
by symmetry. The average of the displacement squared is
(r
N
r
0
)
2
=
N

n=1
N

m=1

n
=
N

n=1
N

m=1

mn
b
2
= Nb
2
The probability that a polymer will have segment positions at r
k
is
p = F(r
k
=
1
Z
G(r
k
) G = e
U({r
k
})
Consider when
U(r
k
) =
N

j=1
U
j
(r
j1
, r
j
) +W(r
k
)
where W represents an external potential (electric, gravitational, etc.. For W = 0, this is just
a random ight model. Either way, this is a local model for the total energy, as it only relies
on nearest neighbor interactions. We then dene a term to represent the energy exponential
between two elements

j
(R
j
) = exp[U
j
(R
i
)] R
j
= r
j
r
j1
The vector dierence R
j
is assumed to be normalized, and can also be used to calculate the end
position, assuming the starting position is the origin 0
_
dR
i

i
(R
i
) = 1 r
N
=
N

j=1
R
j
The xed end to end vector partition function, the integral over all degrees of freedom for which
the end position is R, is then
G(R; N) =
_
dR
k
G(R
k
)(r
N
R) =
_
dR
k

j=1
(R
j
)
_
_
N

j
R
j
R
_
_
Thus, molecular level structures are eectively abstracted away:
For example, the position distribution for a spring is
p(x) = exp(kx
2
/2k
B
T)
3.4.4 Example of Nearest Neighbor Interaction
An example of for a xed length polymber
(R
j
) =
1
4l
2
([R
j
[ l)
26
which is the simplest model for a self-avoiding chain. The trick is to write

_
_
N

j+1
R
j
R
_
_
=
_
d
3
k
(2)
3
exp
_
_
ik
_
_
N

j1
R
j
R
_
_
_
_
The distribution function is then
G =
_
d
3
(2)
3
e
ikR
__
dR
j
(R
j
) exp(ik R
j
)
_
N
The bracketed term is a characteristic function K(k; N), and in this case equivalent to
K(k; N) =
_
sin(kl)
kl
_
N
We expect N to be on the order of R
2
if dominated by diusive behavior, and thus quite large.
In the limit of large N (small k)
K(k; N)
_
1
k
2
l
2
6
+. . .
_
N
exp(Nk
2
l
2
/6)
When inverse transformed, this yields a Gaussian relation. We would have anticipated a radius-
length relation R
2
N assuming random walk behavior. Instead
_
d
3
k
(2)
3
e
ikR
exp
_
N ln
_
dR
j
(R
j
)[1 +ik R
j
(k R
j
)
2
/2 +. . .
_
but the expansion in the integral goes to zero by symmetry. We then dene

,
l
2
3
=
_
dR
j
(R
j
)R

j
R

j
Which simplies G to
G(R
j
; N) =
_
d
3
k
(2)
3
exp(ik R) exp(Nl
2
k
2
/6) =
_
3
2l
2
N
_
3/2
exp
_
3R
2
2Nl
2
_
3.4.5 Impenetrability
Excluded volume interactions (impenetrability) can be included through an extremely crude
interaction energy
k
B
T(R
m
R
n
) k
B
T(R
m
R
n
)
The external potential term is then
w =
1
2
k
B
T
_
N
0
dn
_
N
0
dm(R
n
R
m
)
This energy should be related to the concentration of local segments
c(r) =

n
(r R
n
) =
_
N
0
dn(r R
n
)
The external potential is then
w =
_
dr
_
1
2
k
B
T
_
c(r)
2
However, the concentration is not known beforehand. From a random walk model, wed expect
R
2
N
2
, = 1/2
However, in actuality due to self avoidance this is slightly higher (3/5). The concentration of
segments is then
C
seg

N
R
3
N
13
_
N
1/2
ideal
N
4/5
reality
1
Thus, the probability of segment interactions is quite small.
27
3.4.6 Flory Theory
We saw previously
P(R) e
3R
2
/2Ne
2
e
Energy

/k
B
T
In this case the energy term can be interpreted as the free energy, equal to
k
B
T ln P(R) = k
B
T
3R
2
2Ne
2
where the energy is similar to the spring force constant kx
2
/2.
Estimate of the Repulsive Contribution
The segment density scales as
N
R
3
vk
B
T N
N
R
3
k
B
T
vN
2
R
3
and the total force is then
F
tot
= k
B
T
_
3R
2
2Nl
2
+
vN
2
R
3
_
Thus, there is a trade-o between the two competing eects
For a xed R(N) R
2

1/2
(in the mean-squared sense)
d
dR
R
N

N
2
R
4
R N
3/5
which matches the previous results and shows that self avoidance leads to expansion. Generalizing
this for three dimensions
R
2
N

N
2
R
d
R
d+2
N
3
R N
3/(d+2)
This relation has varying results for dierent dimensions
d = 1 R N correct
d = 2 R N
3/4
exact (solved)
d = 3 R N
3/5
Numerical Solution (0.589)
d = 4 R N
1/2
correct
d > 4 R N
3/(d+2)
wrong (should be N
1/2
)
A radius of gyration R
G
could be a mean-square measure of polymer coil dimensions.
28
Chapter 4
Elasticity
4.1 Curve Dynamics
4.1.1 Lagrangian Dynamics
The Lagrangian for a classical system is given as
/(q, q) = T U
The equations of motion are then
s = 0, S =
_
dt/
d
dt
/
q

/
q
= 0
The kinetic energy and potential energy are just
T =
m q
2
2
V (q) m q =
U
q
However, these equations do not hold for viscous ow as there is an implied conservation of
energy. If the viscous force is q, the rate of energy dissipation will be proportional to q
2
. A
generalization of this for cases without conservation of energy is
d
dt
/
q

/
q
=

q
! =

q
_
1
2
q
2
_
where ! is the Rayleigh dissipation function (the rate at which viscous forces do work). This
leads to
q =
U
q think more
about this
4.1.2 Geometry
Consider a curve in a plane that can be mapped to another coordinate
The derivative with respect to alpha and the unit vector

t are dened as
r

t =
r

[r

[
29
A section of the arc is then
ds = [r

[d = [dr[
and is the tangent angle. The Frenet-Servet equations are

s
_

t
n
_
=
_
0
0
__

t
n
_
Where is the curvature. As an exercise, it can be shown that
=

s
The goal is to express the energy functional in a reparameterization-invariant form . For both
look this up
open and closed curves the length is just
L =
_
L
0
ds =
_
1
0
[r

[ d =
_
1
0
d

g = L[r]
where g is a new metric
g r

Now, we can formulate the Lagrangian in terms of r and its derivatives


L =
_
1
0
d

L[r, r

, r

, . . . ]
In classical mechanics, the equivalent is /(q, q, q, . . . ). By analogy then, q r and The
generalized force associated with this functional is L/r (a functional derivative). Recalling
that
s =
_
dt/(q, q) s =
_
dt
_
/
q
q +
/
q
q
_
Integrating by parts
s =
_
dt
_
d
dt
/
q
+
/
q
_
q
Taking the derivative is simple
S
q
=
d
dt
/
q
+
/
q
The equivalent between the classical Lagrangian and this one is
L
r
=

. . .
r

+
. . .
r
L =
_
d (

g)
Since the length depends only on r

L
r
=

g =

1
2
1

g
2r

g
=

_
r

g
_
=

t
Since

g =
s
with

t = r
s
. The interesting quantity (L/r)/

g is
1

g
L
r
= s

t = n
by the Frenet-Servet equations. The force is

g
L
r
= n
30
The energy comes from L where is the line tension. The real force is n. The Rayleigh
dissipation function is then
! =

2
_
d

gr
2
t

!
r
t
=

gr
t
=
(L)
r
with
r
t
=
1

g
()
r
= n
Interestingly, r
t
ts into the equation of motion of , . However, we need a constraint to prevent
the curve from collapsing through minimization of energy, such as a constraint on the total surface
area of a surface (think of a cellular membrane).
4.1.3 General 2D curve dynamics
Consider a generalized frame of motion U = U(r, , ,
s
, . . . ). The rate of change of the curve
can be parameterized as
r
t
= U n +W

t
For closed curves
L =
_
1
0
d

g
The rate of change of the length will then be
dL
dt
=
_
1
0
d
_
1
2
__
2

g
_
r

r
t
=
_
ds
d

g
_
1

g
r

_
1

r
t
_
=
_
ds

t
r
t
s
The Frenet-Servet equations for 2D are

s
_

t
n
_
=
_
0
0
__

t
n
_
Combining this with the rate of change of the length, r
t
= U n +W

t,

s
r
t
= U
s
n +U

t +W
s

t W n
and thus the time derivative of the length is
dL
dt
=
_
ds(W
s
+U)
For a closed curve, W, U are functions of shape alone and
_
W
s
= 0. Thus
L
t
=
_
dsU
4.1.4 Example: The curve shortening equation
For example, if an energy functional
U =
The time derivative of the length is
L
t
=
_
ds
2
0
Which leads to continual decrease (consider Lyapunov function). If
r
t
= u n +W

t
31
then

t
= U
s
+W
t
= (
ss
+
2
)u +
s
W
Interestingly, derivatives do not commute, except for W
s
+U; that is

t
,=
t

s
=
t
1

Choosing W = 0 in the curve shortening equation


U =

t
= U
s
=
s
=
ss

t
= (
ss
+
2
) = (
ss
+
3
)
Rescaling
t
=
ss
+
2
, there will be a singularity at nite time as the length goes to zero (and
innite curvature). The area of the curve A can be calculated with
A =
_
d
1
2
; r r

A
t
=
_
dsU =
_
ds =
_
ds
d
ds
= 2
Since the rate of change of area is constant, the singularity occurs in nite time.
4.1.5 Global constraints
We need to determine global constraints on the evolution equations to prevent these singularities.
Recalling from before (the case when U = 0
s
f())
L
t
=
_
dsU A
t
=
_
dsU g
t
= w
s
+U
If instead U =
s
,
A
t
=
_
ds
s
= 0
L
t
=
_
ds
s
=
_
ds
_
1
2

2
_
= 0
L
t
= 0 and A
t
= 0. It is then necessary to choose a guage w that conserves the local arclength
w
s
= U =
1
2

2
w =
1
2

2
The time derivative of is then

t
= (
ss
+
2
)
s
+
s
(
2
/2) =
sss

3
2

s
Which is a modied Korteweg-deVries equation (mKdV). This equation has an innite num-
ber of conserved quantities. The global constraints are then achieved by introducing Lagrange
multipliers. Introducing the bare energy c

c = c pA r
t
= U n +w

t
U = U
0
+ (p) w = w
0
+ (p)
Thus
= c L = c
_
dA()

g
A() corresponds to imposing a local tension constraint on the bend to bend distance in a
polymer model (beads on springs).
32
4.2 Space Curves
The fundamental work in this eld came from:
1. Hasimoto, JFM 51,477 1972
2. Petrick PRL 67,3203 1991
3. Segur, Wadaki PRL 69, 2603 (1992)
Beginning with the denition of the curve
r
t
= U n +V

b +w

t
and its time derivative:

s
_
_

t
n

b
_
_
=
_
_
0 0
0
0 0
_
_
_
_

t
n

b
_
_
where once again the curvature is and the tension . Note that is dened here to be positive,
but the normals are dened for each tangent circle dening the curvature. However, there is a
problem at any inection points: the curvature needs to be positive, but switches directions. In
this case

t
= OU PV +
s
w
t
=
s
_
1

(OU +PV )
_
+zU
s
V +
s
w
where
O =
ss
+
2

2
P =

s
+
s
When = 0,
t
has a singularity, there is an inection point, and the normal vector ips.
4.2.1 Example: Vortex Rings
Consider a circular vortex ring. The velocity is proportional to

b/R. In this case, U = W = 0
and V is a constant (since there is constant curvature).
4.2.2 Example with a Variational Principle
For an elastic lament, the energy will be
c =
A
2
_
ds
2
We then can nd c/r (a functional derivative) leading to
U A(
ss
+
1
2

2
) V A(2
s
+
s
) w = 0
33
4.2.3 Hasimotos trick
Hasimotos trick was to realize that these equations could be formulated in terms of a wave
equation through a transformation. Instead of considering , ,
(s, t) = (s, t) exp
_
i
_
s
ds

(s

, t)
_
Where
_
ds

(s

, t) is the accumulated phase angle, and is the angular rate of rotation. In


this form, there is no singularity across changes in the curvature around inection points. The
general rules for transformations are:
= (U +iV )e
i
w = ( n +ib)e
i

t
= (
ss
+[[
2
) +/
m
_
s
ds

+
s
w
where for U = 0, V = , obeys the nonlinear Shroedinger equation. In addition:
i
t
=
ss
Const.[[
2

this can be
shown
4.2.4 Generalized Frenet-Servet equations
The general form of the F-S equations is then
w
s
=

t w

s
=

t

t
s
=
1
2
(

w +w

)
Thus, given a function , the general curve can be obtained.
4.3 Viscous Drag
The viscous drag on a sphere at low Reynolds numbers is given by stokes drag:
= 6a r
t
= F
That is, when representing the 3D motion of a particle as a space curve, the force is related to
the drag times the time derivative of the motion curve.
For a rod lament, which doesnt have the many symmetries of a sphere, this is slightly more
complicated. Three drags will be necessary, for perpendicular motion

, for parallel motion


||
and for rotational motion (torque based)
t
. These are then fundamentally non-local forces, and
trying to nd an expression like that for the sphere simply wont work. However, slender body
hydrodynamics shows that a local model is justied (via an expansion in the aspect ratio).
4.3.1 Typical elastic coecients for biological systems
When the elastic energy is expressed in the form
c =
1
2
A
_
ds
2
A bending modulus (units energy times length) is necessary. This can be dened in terms of a
persistence length L
p
A = k
B
TL
p
For DNA, L
p
is about 50 nm, or 150 base pairs. For actin, L
p
is about 10 m. The energy
required to bend a lament into a circular shape is 1/R where the radius of curvature is
proportional to the length of the lament. The energy will thus be proportional to
c A
_
ds
2
34
The behavior of a lament thus depends on the ratio of the length to the persistence length. For
small ratios, the rods are essentially sti, while for large ratios the rods are oppy. Finally, the
bending modulus can be related to Youngs modulus through
A E r
4
needs to be
expanded
4.4 Elastohydrodynamics [19, 9]
For an elastic rod in a viscous uid, the elastic energy will be given by
c =
A
2
_
ds
2
and the uid viscosity will contribute drag forces as described above. Only two dimensional
weak deformations are considered. If the uctuations from linearity are small ([h
x
[ 1), the
curvature is just h
xx
and the dierential length will be ds dx
_
1 +h
2
xx
. The energy is
then
c =
A
2
_
dsh
2
xx
The obvious question is how does the elastic energy depend on the height function h (c/h). If
c[h] =
_
dxe(h, h
x
, h
xx
, . . . ) =
c
h
=
e
h


x
e
h
x
+

2
x
2
e
h
xx
+. . .
and thus
c
h
= Ah
xxxx
=

h
t
which is also known as a hyper-diusive equation. The dominant motion is perpendicular to the
curve. If the height function looks like a traveling wave e
ikx+t
,

t
Ak
4
=
A

t
k
2
so that modes will grow/die with e
Ak
4
t/
t
. If the rod oscillates with frequency ,
t
A/l
4
and
l()
_
A

_
1/4
4.5 Comparison to Stokes Problems (Stokes, 1851)
The typical form of the Stokes equation is
(u
t
+u u) = p +
2
u
For wall driven ow, the pressure gradient can be ignored. The problem is solved in two stages:
1. Starting at t = 0, the wall begins to move with velocity U, parallel to itself.
2. The walls velocity oscillates, as
U = U
0
e
it
From symmetry,
u u = 0 u = f(x, y) e
x
35
Rather than dealing with the velocity vector eld, we assume unidirectional ow and only care
about the function f(x, y) = u(y). By inspection, the only length scale will be diusion, so that
u
t
= u
yy
= / u = u(y) e
x
Thus
u
t
= u
yy
1
t
= /l
2
l

t
Searching for a similarity solution of the form
u = Uf
_
y

t
_
u
yy
= Uf

()
_
1
t
_
u
t
= Uf

()
_
1
2t

_
u
t
= u
yy
then becomes
u
t
= u
yy
= U
_
1
2t

_
f

=
1
t
Uf

This simplies to an ODE in

1
2
f

= f

Letting g = f

1
2
g = g

=g = Ae

2
/4
as 0, g 1, so that A = 1. f() is then
f() =
_

d

2
/4
and thus the similarity solution exists.
4.5.1 Post-Transient Behavior
At long times after the cyclic motion has begun, the velocity prole will look like
u Ue
it
F
_
y
_
/
_
The scaling term
_
/ is known as the viscous penetration depth. For water, 0.01 cm
2
/s,
and = 2, so that
_
/ 0.1cm. Substituting u into u
t
= u
yy
Uie
it
f = Ue
it
_
1
_
/
_
2
F

iF = F

Since we expect that F e

i =
2
= = e
i/4
=
(1 +i)

2
For convergence at , the solution with a negative real part is chosen. After transients have
died out, the similarity solution will take the form
f = e
/

2
e
i/

2
=u Ue
/

2
e
i(/

2t)
the real part of which is
Re[u] = Ue
/

2
cos
_

2
t
_
yielding a solution that looks like
36
4.5.2 Biophysics
A frequency of interest might be the beating of agella/cilia. A typical length scale of a bacteria
is 50-1000 m, each cilia is about 10-20m long, and this leads to the important quantity
_
2/.
4.6 Elastohydrodynamics, continued
Starting from before, the important dierential equation for an elastic rod in a viscous medium
h
t
= Ah
xxxx
h(0, t) = h
0
cos(t)
If we assume hinged boundary conditions at the left, h
xx
(0, t) = 0. From before, the eective
persistence length is
l() =
_
A

_
1/4
A = k
B
TL
p
= 4
We then guess a similarity solution
h(x, t) = h
0
Re
_
e
it
f
_
x
l()
__
For a hinged left end and a free right end, the boundary conditions are
f(0) = 1, f

(0) = 0, f() = 0, f

() = 0
If we let = x/l()
if = f

For the case of an innite lament, the free end should eventually settle to a straight end so that
a resulting boundary condition is f

() = 0. The case for a nite lament is considered in the


rst example sheet. For an innite lament, f will have solutions of the form
include this
solution here
f e

4
= i
with solutions

2
= e
/8
= (C
8
iS
8
)
3
= e
3i/8
= S
8
iC
8
where C
8
= cos(/8) and S
8
= sin(/8). The solution is then
h(x, t)
h
0
=
1
2
_
e
C
8

cos(t +S
8
) +e
S
8

cos(t C
8
)
_
. Thus, there are two traveling waves, one in each direction, with C
8
> S
8
. However, the right
wave dominates. This was rst proposed by K. Machin [13]. He argued that since a rod driven
by one end has a quickly decaying oscillation, and the amplitude of oscillation of agella was
approximately xed (a sinusoid down the length), there must be force capabilities distributed
along the length of the agella. This was later veried.
The important ratio in the problem is L/l(). For small ratios, the solution will be like that
of a sti rod. For large ratios, the rods will deform and bend with oscillations.
37
4.6.1 Reversibility
A common consequence of low Reynolds number ow is that reversible movement provide no
net movement. For example, a device with only one degree of freedom will not be able to move
continuously, as the two possible motions accomplish opposite eects. For small L/l(), there
will only be one degree of freedom as the rod swings back and forth. However, for large ratios the
traveling wave breaks symmetry and allows for a net movement. Exactly solving this problem
shows a force distribution of
4.7 Euler Buckling
Consider the case of a buckling elastic lament
several ap-
plications
are cited (l-
aments, etc.)
Where the curvature can be directly related to the angle =
s
. The energy functional associated
with this conguration is just
c =
_
L
0
ds
_
A
2

2
s
+F cos
_
The amount of compression can be calculated from
X(L) X(0) =
_
L
0
dx
dy
ds =
_
L
0
cos((s))
The components of the curve in the x, y directions are just
_
dx/ds
dy/ds
_
=
_
cos((s))
sin((s))
_
So we can see that the horizontal force is just F cos . The Euler-Lagrange equation is
c

=
d
ds
[]

s
+
[]

A
ss
= F sin
Since we know that the force is small (below any critical value), a straight lament will be the
solution. Near the critical force, we expect bending as in the above gure, albeit with a small
amplitude. We can guess a solution
(s) = a sin
_
2s
L
_
In order for this to work, we need
ss
F/A for 1, so that
F
c
=
4
2
A
L
2
Near the bifurcation,
38
By ansatz, we use a single mode approximation and variational approach. We require a fourth
order energy correction initially moving away from a = 0 has an energy penalty. We then expand
near F
c
F = F
c
(1 +f) cos = 1
1
2

2
+

4
24
+. . .
The energy functional is then
c[] =
_
L
0
_
A
2

2
s
+F
c
(1 +f)
_
1
1
2
+

4
24
+. . .
__
a sin
_
2s
L
_
c = c
c
+F
c
(1 +f)L +
F
c
L
2
_
fa
2
2
+
a
4
32
+. . .
_
For the two cases f > 0 and f < 0, c will look like
Thus, weve achieved an expansion in the order parameter a. Minimizing with respect to a
2
, we
nd
a =
_
8f
which is the bifurcation that we expected. After buckling, the deformed rod behaves like a
Hookian spring, which we can nd by comparing the compression length to the applied force.
The dierence in length (from before) is just
X(L) X(0) =
_
L
0
ds cos =
_
L
0
ds
_
1
1
2

2
+. . .
_
L
La
2
4
+. . .
So, the displacement due to a force beyond F
c
is La
2
/4. With the knowledge that a =

8f,
x
L
4
8f = 2L
_
F F
c
F
c
_
= F F
c
=
F
c
2L
x
Thus, beyond the critical length the rod will behave like a Hookian spring with eective spring
constant F
c
/2L. Full solutions farther from the bifurcation are possible through a numerical
solver. For L = 1, the solutions for various values of a with clamped ends (left) and hinged ends
(right) are
39
4.7.1 Propulsion Mechanisms of Flagella
Eukaryotic agella
Eukaryotic agella are arranged as a bundle of microtubules. Each microtubule is composed of
a bundle of tubulin. Each agella has 9 microtubule doublets arranged in a circle, with a nal
doublet inside. Molecular motos are used to cross-link the microtubules and provide a local force.
Prokaryotic Flagella [21] To be com-
pleted
4.8 Further Elasticity [9]
Much of this section comes from Landau/Lifshitz book on elasticity. The basic theory is com-
posed of denitions of stresses and strains, boundary conditions, and geometrical considerations.
4.8.1 The Strain Tensor
The strain tensor tracks deformations. For a given position vector x, after deformation the same
innitesimal element will be at position x

. The displacement u is then


u
i
= x

i
x
i
The vector dierence between two nearby points will then be deformed as:
dx
i
dx

i
= dx
i
+du
i
In three dimensions, the distance is then
dl =
_
dx
2
1
+dx
2
2
+dx
2
3
dl

=
_
dx

2
1
+dx

2
2
+dx

2
3
Using summation notation the distance can be expanded
dl
2
= dx
2
i
dl

2
= dx

2
i
= (dx
i
du
i
)
2
, du
i
=
u
i
x
k
dx
k
= dl
2
+ 2
u
i
x
k
dx
i
dx
k
+
u
i
x
k
u
i
x
l
dx
l
dx
k
However, since there is a summation over i and k in the second term,
2
u
i
x
k
=
u
i
x
k
+
u
k
x
i
So that
dl
2
= dl
2
+ 2u
ik
dx
i
dx
k
u
ik
=
1
2
_
u
i
x
k
+
u
k
x
i
+
u
l
x
i
u
l
x
k
_
The last term in u
ij
is nonlinear, and thus assumed to be small. Furthermore, u
ik
is the strain
tensor. It is possible to diagonalize this matrix at every point, so that
u
ik

_
_
u
11
0 0
0 u
22
0
0 0 u
33
_
_
Labeling these as u
(1)
, u
(2)
, u
(3)
, the deformation of distance can be represented with
dl

2
= (
ik
+ 2u
ik
)dx
i
dx
k
= (1 + 2u
(1)
)dx
2
1
+ (1 + 2u
(2)
)dx
2
2
+ (1 + 2u
(3)
)dx
2
3
40
Thus
_
1 + 2u
(i)
1 u
(i)
=
dx

i
dx
i
dx
i
The deformation in each component is then
check this
dx

i
= (1 +u
(i)
)dx
i
The change in volume of the element due to deformations is then just
dV

= (1 +u
(1)
+u
(2)
+u
(3)
)dV
where the (1 +u
(1)
+. . . ) term is the dilation.
4.8.2 Stress Tensor
The stress tensor in equilibrium situations is a mechanical problem where the resultant of all
forces is zero. The stress tensor under deformation comes from molecular properties and internal
stresses. The total force on a body is
Total Force =
_
dV F
where F is a force density. The forces of various parts of a continuous body with respect to each
other must sum to zero by Newtons 3rd law. Thus, the total force is due to surrounding regions,
which interact through surfaces. We recall that if
_
dV (scalar) =
_
dS(something)
the scalar and the other term are related through
scalar = (something)
since F
i
is a vector, something must be a rank two tensor. Thus,
_
F
i
dV =
_

ik
x
k
dV =
_

ik
df
k
where f
k
is a surface element. We then expect to be able to write
F
i
=

ik
x
k
and the condition for equilibrium is
F
i
=

ik
x
k
= 0
Suppose there are pressures P
i
exerted on the surface (with n
k
being a unit normal). The
boundary conditions will be

ik
n
k
= P
i
The dierential work (pressure times movement in surface) will be
w = F
i
u
i
=

ik
x
k
u
i
Integrating
_
wdV =
_

ik
u
i
df
k

_

ik
u
i
x
k
dV
Since the body will be undeformed at some innite distance, the rst integral must be zero
(u
i
0). Thus
_
wdV =
1
2
_

ik

_
u
i
x
k
+
u
k
x
i
_
w =
ik
u
ik
which is just the Helmholtz free energy. The stress tensor is then

ik
=
F
u
ik
41
4.8.3 Generalization of Hookes Law
In Hookes law, the energy is proportional to the square of the displacement (kx
2
/2). The
question is, which invariants and scalars are important? The force can be expanded in terms of
the displacements
F = F
0
+
1
2
u
2
ii
+u
2
ik
where , are called the Lame coecients. Using the identity
u
ik
=
_
u
ik

1
3

ik
u
ll
_
+
1
3

ik
u
ll
the force is just
F =
_
u
ik

1
3

ik
u
ll
_
2
+
1
2
u
2
ll
= +
2
3

where is the shear modulus and is the bulk modulus. The strain tensor is then calculated
from

ik
=
f
u
ik
to give:
dF = u
ll
du
ll
+ 2
_
u
ik

1
3

ik
u
ll
_
d
_
u
ik

1
3

ik
u
ll
_
= u
ll
du
ll
+ 2
_
u
ik

1
3

ik
u
ll
_
du
ik
= u
ll

ik
du
ik
+ 2
_
u
ik

1
3

ik
u
ll
_
du
ik
=
_
u
ll

ik
+ 2
_
u
ik

1
3

ik
u
ll
__
du
ik
So that the stress is given in terms of the strains u. If
ik
= u
ll

ik
+2(u
ik
u
ll

ik
/3), then

ii
= 3ku
ll
= 3ku
ii
u
ii
=
ii
/3

ik
=

ii
3

ik
+ 2(u
ik

ll

ik
/9)
Solving for the strain
u
ik
=

ik

ll
9
+
1
2
(
ik

ik

ll
)
Finally, since F/u
ik
=
ik
F =
1
2

ik
u
ik
4.8.4 Homogeneous Deformations
Simple Compression
To demonstrate these principles, the case of homogeneous deformations of a rod will be consid-
ered. For simple compression, u
ik
=constant, and
ik
is determined by boundary conditions. In
this case, the stress in the z direction (along the rod) will be
zz
= p. The strain will be
u
ik
=

ik

ll
9
+
1
2
_

ik

1
3

ik

ll
_
42
If i ,= k, u
ik
= 0 and thus
u
xx
= u
yy
=
p
9
+
p
6
=
1
3
_
1
2
+
1
3
_
p
We can dene a material property called the Poisson ratio (usually positive, but sometimes
negative)so that
u
xx
= u
zz
It is also possible to dene Youngs modulus E in this case from:
u
zz
=
p
E
=E =
9
3 +
The Poisson ratio then follows
=
1
2
_
3 2
3 +
_
with bounds of 1 1/2.
Rod Bending
For the alternate case of the bending of rods, a bending modulus proportional to Ea
4
is expected.
As an aside, this makes hollow tubes an ecient use of materials. The coordinates for this system
are:
where the dotted line represents the neutral surface. The incremental change in curve length as
a function of the radius goes as
dz

=
R +x
R
dz
and the corresponding relative extension is
dz

dz
dz

x
R
The following theory is valid for a/R 1 (a thin rod compared to the radius of curvature). The
strain tensor is then
u
zz
= x/R
zz
= Ex/R
Without compression in the rod, the total internal stress will be zero and
_

zz
df
_
xdf = 0
gives the neutral surface (where df is the cross section. The free energy per volume is then
Free Energy per Volume =
1
2

ik
u
ik
=
1
2

zz
u
zz
=
1
2
Ex
2
R
2
43
Alternatively, the free energy per unit length is
Free Energy per Length =
1
2
E
R
2
_
x
2
df
where the integral is similar to a moment of inertia of the cross-section I
y
. Also, R could be
a function of position (R(z)). The moment of inertia is given as
I
_
x
2
df
so that the force per unit length is just
F
Length
=
1
2
I
y
R
2
If the cross section is circular, I = a
4
/4 where a is twice the radius. The force per length can
also be expressed as
F
Length
=
1
2
A
2
where is the curvature (1/R), leading to the relation
F
Length
=
1
2
I
y
R
2
=
1
2
A
1
R
2
= A = EI
This is useful, as it is easy to measure A, but generally the Youngs modulus is more desirable.
So, we can measure A, calculate I from the geometry, and thus infer E.
4.9 Torsion of Rods [9]
After considering the bending and compression of rods, the next step is to consider the torsion
of rods. The torsion angle is dened as the ratio of the angle of rotation to the length of the
rod.
For two cross sections dz apart then rotate through an angle of d = dz so that
=
d
dz
For most problems, we assume the limit of small deformations a 1. We can nd the displace-
ment u of an element
r = r
If the rotation is about the z-axis, the angle is just z. deformation in the x,y,z directions are
then
u
x
=
zy
u
y
=
zx
u
z
= (x, y)
Where u
z
comes from linearity, and is the torsion factor. For small deformations ([u[ 1)
using the denition of u
ik
:
u
ik
=
1
2
_
u
i
x
k
+
u
k
x
i
_
many components of strain are just zero:
u
xx
= u
yy
= u
xy
= u
yx
= u
zz
= 0
44
This leaves just u
xz
, u
yz
, which are just
u
xz
=
1
2

x
y
_
u
yz
=
1
2

y
+x
_
Since this is pure shear, there is no volume change and u
ll
= 0 (no compression). The components
of the stress tensor are then:

xx
=
yy
=
zz
=
xy
= 0

xz
= 2u
xz
= (
x
y)

yz
= 2u
yz
= (
y
+x)
In equilibrium the torsion angle will be constant along the length of the rod, so that

ik
x
k
= 0 =

zx
x
+

zy
y
= 0 =
2
= 0
By denition, is then a harmonic function. For convenience, we let

xz
= 2
d
dy

yz
= 2
d
dx
=
2
= 1
The boundary conditions are

ik
n
k
= 0
zx
n
x
+
zy
n
y
= 0 n
x
=
dy
ds
, n
y
=
dx
ds
This leads to

x
dx +

y
dy = 0
and thus = 0 on the boundary. The free energy F can then be calculated
F =
1
2

ik
u
ik
=
xz
u
xz
+
yz
u
yz
=
1
2
(
2
xz
+
2
yz
) =
Alternatively, the free energy of a cross section can be calculated from
F = 2
2
[[
2
_
F(over cross-section) =
1
2
C
2
where C is the twist modulus and analogous to A from before (A
2
/2). The coecient C is then
calculated as
C = 4
_
()
2
df = 4
_

2

n
ds + 4
_
df
where the last term is analogous to the ux in pipe ow. Finally,
C = 4
_
dxdy
The energy can then be calculated as an integral over the free energy of each cross section
_
ds
1
2
C
2
=
d
ds
If we hold one end of the rod xed and use a couple M to twist the other end of the rod, the
free energy will be
Free Energy, F = F
elastic
+F
external
45
The change in free energy can easily be calculated from previous expressions
F =
_
1
2
C
_
dz
_
d
dt
_
2
_
+U
=
_
dzC
d
dz
d
_

dz
_
+U
=
_
C
d
dz
dz + [C] +U = 0
where the bracketed term represents surface terms (and the last line was achieved through inte-
gration by parts). By setting = 0 at the left end (where the rod is held), the surface terms
will involve only the twisted end and we then know
U = M
This allows the equation for free energy to be simplied
_
C
d
d
dz + [(C M)] = 0
with the boundary condition that M = C at the end. In the bulk (at equilibrium), the twist
density will be uniform as discussed previously
C
d
dz
= 0 =
M
C
= const
With more complex total energy functionals, its possible to have ground state energies.
4.9.1 Viscously induced twist
A rod-like object will have a rotational drag coecient of

r
= 4a
2
such that the torque/length on a rod rotating at angular velocity will be
r
. From geometry,
the twist will increase from the uid torque as

t
=

z
and if
r
Cd/dz

t
=

z
_
C

r
d
dz
_

z
2
which looks like a diusion term, with a twist diusion coecient of
D
twist
=
C

r
=
k
B
TL
twist
p
4a
2
where the twist persistence length is usually comparable to the normal persistence length. For
DNA, L
p
150 base pairs, or 500 angstroms and a 20nm, so that D 10
6
cm
2
/s. The
twisting behavior of DNA was conrmed experimentally by attaching strands to a substrate and
small magnetic spheres that could then be used to control twisting. This was rst considered for
DNA in the contest of how DNA assembly occurs (two straight strands inter-twining) [12]
4.10 More General Elasticity Theory
46
An alternative formulation is to describe the rotation of a material coordinate system. We let
d describe rotation, which is a vector parallel to axis of rotation. The components are angles
about each of the three coordinate axes. The twist density is dened as
=
d
ds
For pure rotation/torsion, will be parallel to the rod axis. However, more generally
[[ =
z
= Torsion angle
If

t is the tangential unit vector,
d

t
ds
= n = d

t = d

t =
d

t
ds
=

t
Even more generally
d e
i
ds
= e
i
for all i. Crossing with

t yields
=

t
d

t
ds
+ (

t )

t =

t n +
z

t =

b +
z

t
The goal is now to express the free energy in terms of .
4.11 Twisting and Bending
We expect the energy to be approximately quadratic in the components of , assuming a straight
untwisted ground state. (as a side note, not all bers will have an untwisted ground state; for
example DNA has a natural twist, intrinsic bends (non-random sequences), and chirality). The
free energy of the rod is then
F
rod
=
_
ds
_
1
2
EI
1

+
1
2
EI
2

+
1
2
C
2

_
In the weak-bending (no torsion) limit,
=

t n =

t
d

t
ds
d

t
ds
=
d
2
r
ds
2
then,
d
2
r
ds
2

d
2
r
dz
2

t e
z
and
e
z

d
2
r
dz
2
= X
zz
e
y
Y
zz
e
x

= Y
zz

= X
zz
where the direction z is the primary direction
So, the free energy of the rod is just
F
rod
=
1
2
E
__
(I
1
Y
2
zz
+I
2
X
2
zz
)ds
_
This suggests the existence of a complex function = X +iY .
47
4.11.1 More general equilibrium conditions
A few examples. A hanging chain:
where
dT +gdz = 0
dT
dz
= g
or an elastic rod
where F is the force per length, and thus
dF
ds
= K
4.11.2 Moments
In equilibrium, the sum of all moments will be zero. if M is the moment of the internal stresses
on a cross-section about a point O(origin) in the cross section, then the moment at one point is
M, at another M +dM, and the total moment is
dM
ds
= F

t
4.11.3 Circular rods
For a circular rod, the moment of inertia in both axis will be the same (I
1
= I
2
= I). Then,

= M
s
/C
Now, consider
d
ds
_
M

t
_
= C
d

ds
=
dM
ds

t +M
d

t
ds
= M
d

t
ds
so, we can deduce that
C
d
ds
= M
d

t
ds
48
and also that
=

t
d

t
ds
+ (

t )

t M

= EI

= EI

so in fact
M = EI

t
ds
+C

t
4.11.4 Torsional Instability [9, 7]
When applying a given twist angle (torsion), the moment will be
M = EI

t
d

t
ds
+C

t
Assuming no external forces, then the assumption of weak (small) displacements reduces the
equation to
EI

t
d
3

t
ds
3
+C
d
2

t
ds
2
= 0
where

t
0
is the reference direction. Then,
Y
zzzz

C
EI
X
zzz
= 0 X
zzzz
+
C
EI
Y
zzz
= 0
Dening a complex representation
= X +iY
we can deduce that

zzzz
i
zzz
= 0 =
C
EI
and the appropriate boundary conditions that
z
= 0 at z = 0, L. The solutions are then of the
form
= +z +z
2
+e
iz
Thus, there is a bifurcation at which there is a critical angle for non-trivial solutions. We then
see that CL/A is just a number that depends on the material.
This is especially relevant for buckling in DNA, called the Zayac instability.
49
Chapter 5
Chemical Kinetics and Pattern
Formation
We need to rst understand the basic laws of mass action for chemical reactions. The key article
that started the eld was by Alan Turing [17]. Turing showed that the combination of nonlinear
chemical kinetics and diusion can produce patterns, e.g. stationary patterns in time. This was
experimentally observed one year after the prediction [16]
5.1 Michaelis-Mentin kinetics
The essential idea behind this simple model of enzyme-substrate interactions is the reaction
scheme
E +S
k
1
GGGGGGB
F GGGGGG
k
1
ES
k
2
GGGA E +P
where E is the enzyme, S is the substrate, P is the product, and ES is the enzyme-substrate
complex. k
1
, k
1
, k
2
are the rate constants (empirically measured quantities). For notation, let
s = [S], e = [E], c = [ES], p = [P]. The concentration is dened as the average concentration
(for a CSTR). The law of mass action is then that the reaction rate is proportional to the
concentration, so that
de
dt
= k
1
es +k
1
c +k
2
c
ds
dt
= k
1
es +k
1
c
dc
dt
= k
1
es k
2
c k
1
c
dp
dt
= k
2
c
The initial conditions are:
s(0) = s
0
c(0) = 0 e(0) = 0 p(0) = 0
Note that p is completely decoupled from the other reactions, so that
p(t) = k
2
_
t
0
c(t

)dt

It also follows that the quantity e +c must also be conserved (conservation of the enzyme/cata-
lyst), so that
de
dt
+
dc
dt
= 0
50
which is indeed satised. Thus, e + c = e
0
at all times. The system thus reduces from four
coupled dierential equations to just two
ds
dt
= k
1
e
0
s + (k
1
s +k
1
)c s(0) = s
0
dc
dt
= k
1
e
0
s (k
1
s +k
1
+k
2
)c c(0) = 0
It is useful to identify appropriate scalings. The concentration of the enzyme/catalyst is generally
much smaller concentration of the substrate, so that the small parameter is = e
0
/s
0
. Thus, we
let the rescaled time be = k
1
e
0
t, and
u() = s(t)/s
0
=
k
2
k
1
s
0
v() = c(t)/e
0
k =
k
1
+k
2
k
1
s
0
The equation system thus reduced further to an equation for the substrate and the complex:
du
d
= u + (u +k )v u(0) = 1

dv
d
= u (u +k)v v(0) = 0
Although is small, neglecting it would remove the leading order term.
5.1.1 Biologists method of solution
Suppose that 1, we can assume it will reach a quasi-equilibrium, then assume that c is
temporarily independent of time and
k
1
es (k
1
+k
2
)c e = e
0
c c = e
0
s
s +k
m
where k
m
= (k
2
+k
1
)/k
1
. Or, equivalently, v = u/(u +k), yielding the simple curve
However, this only works if there is a large separation in time scales between the two reactions.
The fast variable
is then coupled to the slow one, which obeys a nonlinear relation. The rate of reaction would
then be
dp
dt
= k
2
c = k
2
e
0
s
s +k
m
= V
max
s
s +k
m
The two empirical terms would be tted by plotting 1/V = (1 +k
m
/s)/V
max
vs 1/S
51
5.1.2 Example: Cooperativity
Unfortunately, reaction schemes are usually more complex and often involve multiple complexes.
For example, the binding of oxygen to hemoglobin involves a tetramer of four proteins that each
bind oxygen. However, the binding of each site is linked to the number of bound oxygens, so
that the second oxygen is easier to bind than the rst, and so on. This allows for cooperativity
and very large change in absorbed oxygen over a relatively small shift in oxygen concentration
(perfectly tuned for lungs).
Consider another reaction network
E +S
k
1
GGGGGGB
F GGGGGG
k
1
C
1
k
2
GGGA E +P
C
1
+S
k
3
GGGGGGB
F GGGGGG
k
3
C
2
k
4
GGGA C
1
+P
The reduced equations are
du
d
= f(u, v
1
, v
2
)

dv
1
d
= g
1
(u, v
1
, v
2
)

dv
2
d
= g
2
(u, v
1
, v
2
)
where v
1
= c
1
/e
0
, v
2
= c
2
/e
0
. Then,
ds
dt

t=0
= e
0
s
0
( +s
0
)
+s
0
+s
2
0
with a general result of a rate
rate
s
n
k
m
+s
n
which is known as the hill equation, with n being the Hill coecient. This gives a sigmoidal rate
like the example from above.
5.1.3 Example: Slaving
Suppose that there are two degrees of freedom p, q, with a slow reaction and a fast one:
dq
dt
= q pq slow

dp
dt
= p q
2
fast
52
where p is an autocatalytic term. In the steady state
p q
2
p

q
2
so that
dq
dt
=
V (q)
q
V (q) =
1
2
q
2
+

4
q
4
which looks like a potential. This corresponds to the physical scenario of overdamped dynamic,
and in fact produces a bifurcation
The enslaving of slow to fast species can cause bifurcations, bistabilities, or multistabilities.
5.1.4 Diusive eects
Now we include diusive eects with these nonlinearities. When considering pattern forming
systems we xate on boundaries between dierent states
The ingredients for a 1D 1-component model is
u
t
= u
xx
+f(u) = Di. Part + Nonlinearity
Suppose that f(u) =
V
u
. Then, if
Thus, frontal propagation is possible
53
The questions remain: how fast does the front move? are there other instabilities? Some
examples of f(u) are also based on transitions from locally stable states to other locally stable
states. For:
f(u) = u(u r)(u r) for 0 < r < 1
f(u) = F

(u)
For
F(u) =
1
4
u
2
(1 u)
2
+
_
r
1
2
__
1
2
u
2

1
3
u
3
1
12
_
The stability of the system is regulated by the parameter r. If r = 1/2, u = 0, 1 are equally
stable and the dierence in energy between the two states is
f[
0,1
=
1
6
_
r
1
12
_
With a linear stability analysis
Then,
u
t
= mu
xx
u(u r)(u 1)
Near u = 0,
u
t
= mu
xx
ru +. . .
and let u = e
ikx
e
t
. If < 0, u is stable and for > 0, u is unstable. Substituting for u, we can
deduce that = mk
2
r. Graphically,
Thus, is always negative, regardless of r (thus, stable).
Near u = 1,
u
t
mu
xx
(1 r) u
= mk
2
(1 r)
54
which is also always stable. Thus, both global minima are stable for all k.
Near u = r,
u
t
= m u
xx
+r(1 r) u
= mk
2
+r(1 r)
and thus there is a band of unstable modes below a critical k. The obvious question is what
happens between u = 0 and u = 1 when r 1/2.
Stationary front (r=1/2)
mu
xx
u
_
u
1
2
_
(u 1) = 0
multiplying through by u
x
we nd
1
2
mu
2
x
F(u) +C = 0
where C can be set to 0 through boundary conditions. This yields
u =
1
2
_
1 tanh
_
x
2

2m
__
yielding a transition with a width controlled by m.
To determine the behavior of the case r ,= 1/2,a systematic perturbation theory is necessary to
obtain the rst correction to the shape of the front. In doing so, a sickness of the perturbation
theory is found whose removal yields a front speed and shape. The goal is then heuristically
derive the front motion of a 1D PDE with a generic nonlinearity. Consider
u
t
= mu
xx
F

(u)
Imagine, after some transient period, a steady uniformly moving solution (see previous plot, for
the case where F(u) has 2 minima). We then seek a traveling solution of the form
u(x, t) = U(x vt)
for some unknown v. The simplest case is for an F(u) that looks like
55
From the ansatz (traveling wave solution), we use U(x vt), so that
mU
zz
+vU
z
= (F

(U))
which is similar to Newtons second law (m q+b q =force) with m being the mass of the particle,
U the position, and z the time. We can identify F

(u) with potential/position. We can


thus interpret F as a potential. Consider a front with competing states at x = 0, 1.
Now, looking at F instead of F, the situation can be viewed as a ball moving down a hill.
There exists a unique v, that is, a unique damping coecient to achieve u 0 as t . This
is only for the case of a transition from one stable state to another. If instead the front consists
of a stable to unstable situation, the analogy will be
In this case, any damping coecient v greater than a critical value v
c
will ensure u 0 as
t .
Unfortunately, not every class of problems in pattern formation can be reduced to a system
as simple as this one.
56
We now seek a rst integral to the dierential equation
mU
zz
+vU
z
= (F

(U))
mU
z
U
zz
+vU
2
z
= F

(u)u
z
1
2
mU
2
z

+v
_

dzU
2
z
=
_

dF
dU
dU
dz
dz = F(0) F(1)
which is precisely the energy dierence between the two locally stable minima. It can also be
viewed as the energy gradient which induces the front velocity
v =
F(0) F(1)
_

dzU
2
z
=
F(0) F(1)
Friction Coecient
Note that the integral term is dominated by the front region
If we approximate the solutions using U
stationary
, which holds when F = 0, we can then
approximate the velocity. Note that for the stationary solution F = 0, v = 0
mU
zz
= F

(u)
mU
zz
+F

(u) = 0
which is just the Euler-Lagrange equation for
_
dz
_
1
2
mU
2
z
+F(u)
_
There exists two penalties: U is not a minima of F and U
z
,= 0. There also exists a balance.
The two extremes for m = 0 or F 0 are
So, we can see that the fronts are controlled by the diusion constant m, reecting a balance
between diusion and nonlinearities.
5.2 Phenomenology of Reaction-Diusion Systems
We consider equations of the form
u
t
= Du
xx
+. . . (some nonlinearity)
Consider

U(k, t) =
_
dxe
ikx
u(x, t)

t
u(k, t) = Dk
2
u +. . .
If we consider a solution of the form u e
ikx+t
, then (k) = Dk
2
. In k-space, the structure
is simple
57
Experimentally, this corresponds to excitations of initial Fourier modes: unstable modes corre-
spond to rapid growth of a particular Fourier mode, for example in the case of the approximately
uniform state. A more interesting possibility is when both long and short wavelength are damped
(see the second plot). In this case, there is a well dened k

corresponding to the fastest grow-


ing mode. Under those conditions, there exists a k

that leads to a periodic pattern with a


wavelength = 2/k.
This leads to a fundamental question: How can diusion (a second derivative) produce a
k-dependence other than k
2
?. So, in fact = (k
2
), so we want
(k) +k
2
k
4
+. . .
u
t
= u u
xx
u
4x
+. . .
but such nonlinearities are quite rare. Instead, two coupled reaction-diusion equations can
produce this behavior.
5.3 Fitz-Hugh Nagumo Model
The Fitz-Hugh Nagumo model was rst developed as a simplication of neuronal dynamics. Two
chemical species are involved: u, an actuator, and v, an inhibitor (assuming suitable scalings
already).
u
t
= D
2
u +f(u) v
v
t
=
2
v +u v
Notice the second diusion constant has been rescaled to 1, with 0 < < . The various terms
on the RHS of the equations are:
f(u) Autocatalysis
Coupling constant
u Stimulation
v Self-limitation
This system exhibits feedback dynamics (and there might also exist nonlinearities that pro-
duce singularities). The presence of the relative diusion constant D can produce lateral inhibi-
tion for D 1 (dierent length scales), for example:
Depending on the terms f, D, , the FHN model can produce homogeneous states, strips, or
other periodic patterns, spiral waves, etc.
5.3.1 Variants of the FHN model
It is sometimes useful to instead write
u
t
= D
2
u +f(u) (v u)
v
t
=
2
v +u v
The goal is then to see if there exists a variational structure or if we can dene an energy
functional.
u
t
=
E
u
u
+. . . (v)
v
t
=
E
v
v
+. . . (u)
58
In the stationary case, this leads to
u
t
= v
v
t
= u
which is a Hamiltonian structure, thus leading to oscillations. As it turns out, is key to these
oscillations. Consider the simplest limit (like in Michaelis-Mentin kinetics). The fast inhibitor
limit basically sets = 0, so that
(
2
1)v = u (local in time, but not in space)
Since there exists a Greens function of the operator (
2
1), we can deduce
v(x) =
_
dx

G(x x

)u(x

)
For example, in one dimension,
G(x x

) =
1
2
e
|xx

|
and in two dimensions,
G(x x

) =
1
2
K
0
([x x

) Note: Complete elliptic integral of the rst kind


So, in the fast inhibitor limit we have
u
t
= D
2
+f(u) +u
_
dx

F(x x

)u(x

)
which is a closed, nonlocal equation of motion. We can solve this in Fourier space, where the
convolutions become products (much simpler). In fact, u is variational,
u
t
=
E

u
, E

=
_
dx
_
1
2
D[u[
2
+F(u)
1
2
u
2
_
+
1
2

_
dx
_
dx

u(x)G(x x

)u(x

)
which is similar to electrostatics, where the last term appears similar to a non-locality terms in
electrostatics and QFT (nonlocality elds) [16]. In a later paper [10], a non-locality was observed
where interfaces dont cross and the action of inhibitor fronts are also sharp.
read papers,
gure this
out
5.4 Separation of Timescales / Method of Multiple Scales
The classic example for separation of timescales is the pendulum.
The usual formulation is

=
2
0
sin (0) =
0

(0) = 0
We explore the behavior for the case
0
is not innitesimal (so that we cannot assume sin ).
We let = (t)/
0
then

=

2
0

0
sin(
0
) (0) = 1

= 0
59
We try a perturbation theory in
0
:


2
0
_

1
6

3
0

0
+. . .
_

2
0
+

2
0

2
0
6

3
+. . .
We rescale time (
0
t) = , then dene
2
0
(the small parameter in the perturbation theory). We
then nd

+

6

3
+. . .
Since not multiplied by the highest derivative, we can try
=
(0)
+
(1)
+. . .
and substituting

(0)

+
(1)

+ =
(0)

(1)
+

6
_

(0)
3
+. . .
_
At O(
0
)

(0)

=
(0)
=
(0)
= cos
which satises (0) = 1 and

(0) = 0 even at the zeroth order. At O(
1
):

(1)

=
(1)
+
1
6

(0)
3
Recalling that cos
3
=
1
4
cos(3) +
3
4
cos ,

(1)

+
(1)
=
1
6
_
1
4
cos 3 +
3
4
cos
_
with a particular solution
p
that solves

(1)
p,
+
(1)
p
=
1
24
cos 3
(1)
= Acos 3 =
1
36.2
cos 3
However, the general solution to

(1)

+
(1)
=
3
24
cos
is secular. So, the resulting solution is

(1)
=
(1)
homog.
+
1
16
sin
1
192
cos 3
which grows, unbounded and is thus unacceptable (it should also be a function of . The aw
of the problem lies in the fact that for non-innitesimal amplitudes, the frequency operator is
dependent on the amplitude but the perturbation remains about the natural frequency (which
is wrong!).
5.5 Front Velocity Via a Solvability Condition
In the pendulum problem, the secular term arose because we did not account for change in the
frequency of oscillations with amplitude. A useful example is the dierential equation
y + 2 y + (1 +
2
)y = 0
the exact solution of which is
y(t, ) = e
t
sin t
60
This implies that there are two timescales in the problem: t and t. If we did a naive perturbation
theory,
y(t, ) = y
0
(t) +y
1
(t) +
2
y
2
(t) + = sin t t sin t +
1
2

2
t
2
sin t +. . .
just like in the pendulum problem (and secular terms appear). So, the secular term is not a
problem if t 1, but doesnt work for long times. But, if instead we recognize that there are
two timescales so that
y = f(t, t, )
we can instead expand as
f(t, , ) = f
(0)
(t, ) +f
(1)
(t, ) +. . .
so that t becomes
t
+.
5.6 Pendulum, continued
If we use the previous , and

+
T

T

T
(, T) can be expanded as before. At zero order

(0)

+
(0)
= 0
as before, so that
= A(T) sin +B(T) sin A(0) = 1 B(0) = 0
At rst order,

(1)

+
(1)
=
1
6
(Acos +Bsin )
3
2(A
T
sin +B
T
cos )
=
_
2B
T
+
1
8
(A
3
+AB
2
)
_
cos + [2A
T
+
1
8
(B
3
+A
2
B)] sin
+
1
24
(A
3
3AB
2
) cos 3
1
24
(B
3
3A
2
B) sin 3
We seek a solution
(1)
. The last two terms in the expansion are non-resonant. The secularity
condition (to make things right in the perturbation theory) is
[2B
T
+. . . ] = 0 = [2A
T
+. . . ]
For a general operator on the LHS of the equation, we orthagonalize with respect to the null
space (of the adjoint operator; a Sturm-Liouville problem). Near = 0:
A
2
+B
2
=
16B
T
A
2A
T
+
2BB
T
A
(A
2
+B
2
)
T
= 1 A
2
+B
2
= 1
Thus, A(T) = cos I/16, B(T) = sin I/16. Then,
(, T) = cos
_

2
0
16

_
cos + sin
_

2
0
16

_
sin
= cos
__
1

2
0
16
_

_
And thus the frequency decreases with
0
.
61
5.7 Application to PDEs
Consider the PDE
u
t
= /(
x
)u +(u)
which is a general PDE (e.g.
xx
+
4x
+ ). As before, we can adjust the parameter k such
that a band of unstable modes just begins to appear
When 1, the growth rate of unstable modes is slow ( 1/). There will be a range of length
scales with closely spaced modes:
cos
__
k

+
k
2
_
x
_
which leads to long-wavelength modulations
cos(k

x) + cos(kx)
so that there are two timescales and two length scales, requiring multiple scalings.
5.8 Front Dynamics
Consider the Fitz-Hugh Nagumo model
u(u r)(u 1) = Du
xx
(v u) u
t
v
xx
v +u = v
t
with control parameter 0 < r < 1 where r = 1/2 represents a stationary front. For the case
when D 1, 1, if [r 1/2[ 1, then there is a small velocity to the moving front (u
t
1)
which represents a steady state solution.
Sharp interface moves, and the u-eld is seen only at scales larger than

D (the outer scale).


Thus, u is approximately constant. u(x) can represent two states (black or white)
u(x) = u
0
(x) =
_
1 black
0 white
then
v
0
xx
v
0
= u
0
=
_
1
0
62
A solution is achieved by demanding that there is continuity in v
0
and v
0
x
. At the inner scale
(O(

D), we can let


=
x Q
j

D
where Q
j
represents the location of the jth interface/front. We could say that
u(x = Q
j
) =
1
2
At this scale, U(, t) (a shifting coordinate with Q
j
= Q
j
(t). We then dene
s[U] = U

u(u 1/2)(u 1)
for r = 1/2, noting that r
1
2
= 6F (the small parameter, perturbed about the stationary
point). The inner equation is then
s[U] =
Q
j
t
U

D
6FU(U 1) +(V U) +U
t
v

= D(V U) +DV
t
where the second term is the small perturbation and the third is the small coupling. The
homogeneous solution of the these equations are
U
(0)
() =
1
2
_
1 tanh
_

2
__
V
(0)
= v
(0)
We expect that
Q
j
t

D
F 1 but we still have to assume that there is an expansion in
so we want a stability condition for
U() = U
(0)
() +U
(1)
() +. . .
Substituting into s[U] will invoke the operator
S
1
[U
1
] = U
1

(U
0
, 1/2)U

and using

2U
0

= u
0
(u
0
1), this will yield:
S

[U

] = [Q
j
t
/

2D + 6F]U
0
(U
0
1) +[V
0
(x = Q
j
) u
0
]
At rst order on the left hand side is a linear operator on U
(1)
, which is much more complicated
than the pendulum example. So, the solvability condition is to integrate the RHS against the null
space of the LHS operator, then set the result equal to zero. This gives a condition on Q
j
t
which
is approximately the equation of motion for the moving interface. So, what is the null-space of
S ?
In fact, U
0

is the null space by translational symmetry (think of an innitessimal generator


of translations). So, the solution is
Q
j
t
= 6

2D
_
F
_
v
0
(x = Q
j
) 1/2
_
63
Chapter 6
Bioconvection
6.1 Gyrotaxis
not sure how
this relates,
still in the
notes
Consider a suspension of aerobic bacteria
Interesting behavior can arise from two eects:
1. Chemotaxis - the motion of organisms along concentration gradients
2. Bacteria are denser than pure water
The bacteria are aected by the gradient they themselves have created in their search for oxy-
gen, creating a classic Rayleigh-Taylor instability. As a simple model, consider three species
C Oxygen Concentration
n Bacterial Concentration
u Velocity Field
The convection/diusion equations are then
C
t
+u C = D
c

2
C nf(c)
n
t
+u n = D
n

2
n (r(c)nC)
The nonlinearity in the nal term (nC) is important. The overall behavior is based on the
quasi run and tumble of bacteria, leading to eective diusion on large scales. The eective
diusion constant is D u
2
( is a persistence time, a measure of how long the bacteria stay
in a straight line). So, the conservation law suggests that
D
n

2
n = J J = J
D
+J
chem
= D
n
n +r(c)nC
Together with the Navier-Stokes equations:

w
(u
t
+u u) = p +
2
u + gn e
z
v
b
where the last term is a gravity driven forcing on the ow (see the paper by Childress, Levandowsky,
and Spiegel). Using the Boussinesq approximation of thermal convection (i.e. the density on the
LHS of the uid will be amused to be that of pure water, and the contribution due to the bugs
will be accounted for by the last term on the RHS of the equation),
64
Here we suppose that L is the depth of the uid, and we have the appropriate rescalings
x = r/L s =
D
n
L
2
t = C/C
s
v =
L
D
n
u
= n/n
0
=
L
2

D
n
=
D
c
D
n
1 =
aV
0
D
n
10
where C
s
is the saturation concentration. The rate of tumbling is varied based on receptor
measurements. Using = av
1
/C
s
,
=
n
0
L
2
C
s
D
n
Sc =

D
c
1000 =
V
b
n
0
gL
3
D
n
Rescaling the bioconvection equations then

s
+v =
2
f()

s
+v =
2
[r()]
1
Sc
(v
s
+v v) = +
2
v e
z
what is V
s
?
right sym-
bol?
This leads to several possible questions:
1. Linear Stability Analysis: Pedley+Kessler
2. Nonlinear Analysis: Pedley+Hill etc.
3. Self-concentration in drop: ZBN (Zooming Bionematic Turbulance)
4. Nematic Crystals: align in 3D structure (there exists an orientational order without a
translational one, with a characteristic coherence length).
A small example calculation is the steady state distribution of oxygen and bugs prior to the
formation of Rayleigh-Taylor instabilities. Setting the time derivatives to zero:
0 = D
c
d
2
C
dz
2
kn = O
2
Diusivity Consumption
The diusivity of the bugs produces
0 = D
n
d
2
n
dz
2

d
dz
_
n
dC
dz
_
=
d
dz
_
D
n
dn
dz
n
dC
dz
_
In the most symmetric situation,
The solution is quasi-Boltzmann (of the Poisson-Boltzmann formulation):
n(z) = n
0
e
C/D
n
65
so that
d
2
C
dz
2

kn
0
C
e
C/D
n
= 0
qualitatively, we can let C/D
n
= and Z =
_
kn
0
/D
n
D
c
so that
d
2

dz
2
e

= 0
This is testable by theory. We let = e

and solve. Recent measurements in thin lms have


observed this prole of instabilities in atmospheric dynamics.
6.2 Ordinary Diusion
In ordinary diusion (Fickian),
J = DC C
t
= J = D
2
C
A common characteristic of diusion is that it spreads arbitrarily far away instantaneously (C > 0
for all t > 0, without an edge). But, this is no longer true if the diusion is nonlinear. An example
is the spreading of viscous droplets, which obey
D(n) = D
0
n
p
n
t
= D
0
(n
p
n
x
)
x
for p > 1, which is known as a degenerate diusion equation. As n 0, D(n) 0. The key is
that this does admit solutions with compact support for all time. Check that there are solutions
of the form
n(x, t) =
M
2/(2+p)
F()
D
0
t
1/(2+p)
which is a similarity solution with
M =
_

dxn(x, t) =
x
(M
p
D
0
t)
1/(2+p)
and
F() =
_
_
_
_
A
P
2
2(2+p)
_
1/p
0 < <
_
2(2+p)A
p
_
1/2
0 Otherwise
which is like a spreading drop (a spreading parabolic prole).
66
Bibliography
[1] A. Ashkin. Acceleration and trapping of particles by radiation pressure. Phys. Rev. Lett.
24(4):156159, Jan 1970, doi:10.1103/PhysRevLett.24.156.
[2] R. Brown. A brief account of microscopical observations: made in the months of June,
July, and August, 1827, on the particles contained in the pollen of plants and on the general
existence of active molecules in organic and inorganic bodies. Not published, 1828.
[3] M. Doi and S. Edwards. The theory of polymer dynamics. Oxford University Press, USA,
1988.
[4] B. Duplantier, R. Goldstein, V. Romero-Rochn, and A. Pesci. Geometrical and topological
aspects of electric double layers near curved surfaces. Physical review letters 65(4):508511,
1990, doi:10.1103/PhysRevLett.65.508.
[5] A. Einstein and R. F
urth. Investigations on the Theory of the Brownian Movement. Dover Pubns, 1956.
[6] R. Goldstein, A. Pesci, and V. Romero-Rochn. Electric double layers near modulated sur-
faces. Physical Review A 41(10):55045515, 1990, doi:10.1103/PhysRevA.41.5504.
[7] R. Goldstein, T. Powers, and C. Wiggins. Viscous nonlinear dynamics of twist and writhe.
Physical Review Letters 80(23):52325235, 1998, doi:10.1103/PhysRevLett.80.5232.
[8] B. Holstein. The van der Waals interaction. American Journal of Physics 69:441, 2001,
doi:10.1119/1.1341251.
[9] L. Landau and E. Lifshitz. Theory of Elasticity: Volume 7. Butterworth-Heinemann, 1986.
[10] K. Lee, W. McCormick, Q. Ouyang, and H. Swinney. Pattern formation by interacting
chemical fronts. Science 261(5118):192194, 1993, doi:10.1126/science.261.5118.192.
[11] H. Lekkerkerker. Contribution of the electric double layer to the curvature elasticity of
charged amphiphilic monolayers. Physica A: Statistical and Theoretical Physics 159(3):319
328, 1989, doi:10.1016/0378-4371(89)90400-7.
[12] C. Levinthal and H. Crane. On the unwinding of DNA. Proceedings of the National Academy
of Sciences of the United States of America 42(7):436, 1956.
[13] K. Machin. Wave propagation along agella. J. exp. Biol 35(4):796806, 1958.
[14] G. Manning. Limiting laws and counterion condensation in polyelectrolyte solutions I. Col-
ligative properties. The Journal of Chemical Physics 51:924, 1969, doi:10.1063/1.1672157.
[15] D. Mitchell and B. Ninham. Curvature elasticity of charged membranes. Langmuir
5(4):11211123, 1989, doi:10.1021/la00088a044.
[16] Q. Ouyang and H. Swinney. Transition from a uniform state to hexagonal and striped
Turing patterns. Nature 352(6336):610612, 1991, doi:10.1038/352610a0.
[17] A. Turing. The chemical basis of morphogenesis. Bulletin of Mathematical Biology 52(1):153
197, 1990, doi:10.1007/BF02459572.
[18] E. Verwey. Theory of the Stability of Lyophobic Colloids. The Journal of Physical Chemistry
51(3):631636, 1947, doi:10.1021/j150453a001.
67
[19] C. Wiggins, D. Riveline, A. Ott, and R. Goldstein. Trapping and wiggling: Elas-
tohydrodynamics of driven microlaments. Biophysical Journal 74(2):10431060, 1998,
doi:10.1016/S0006-3495(98)74029-9.
[20] M. Winterhalter and W. Helfrich. Eect of surface charge on the curvature elas-
ticity of membranes. The Journal of Physical Chemistry 92(24):68656867, 1988,
doi:10.1021/j100335a004.
[21] C. Wolgemuth, T. Powers, and R. Goldstein. Twirling and whirling: Viscous dy-
namics of rotating elastic laments. Physical Review Letters 84(7):16231626, 2000,
doi:10.1103/PhysRevLett.84.1623.
68

Вам также может понравиться