Chapter 4: Random walks, diffusion

Key themes: Brownian motion Random walks Diffusion laws Biological applications of diffusion Biological question: If everything is so random in the nanoworld of cells, how can we say anything predictive about whats going on there? Physical idea: The collective activity of many randomly moving actors can be effectively predictable, even if the individual motions are not.

Friction
Friction ? It is characterized by dissipation that reduces order
Electrical resistance makes heat as it runs down flashlight batteries Diffusion of ink in wet paper erases order Friction erases order in the directed motion of a particle

That is, organized kinetic or potential energy gets degraded into disorganized motion. That in turn is characterized by the random walk The random walk is an example of diffusion, which in cells turns out to be the dominant form of material transport on submicrosecond scales we will soon see why the scale is important

Brownian motion

Tiny solvent molecules interacting with the colloid through collisions Colloid mass much larger than the mass of solvent molecules While the effect of each single collision is almost negligible, the large number of them gives rise to macroscopic observable motion Brownian motion is thus driven by thermal energy

Brownian motion & random walk

Un-biased Random Walk on a 2D P = 1/4 square lattice 1. 2. 3. 4. Start from the origin
P = 1/4 P = 1/4

Self-Avoiding Random Walk on a 2D lattice 1. 2. 3. Start from the origin Choose one of 4 nearest neighbors randomly Make a jump to the chosen nearest neighbor site provided that it has never been visited before Go to 2.

Choose one of 4 nearest neighbors randomly Make a jump to the chosen nearest neighbor site Go to 2.

P = 1/4

4.

Coupling to polymer models

Random walks leading to diffusion

Diffusion. Way to find how far one is likely to go in a random walk.

a) b)

1D random walk through three steps j = 1, 2, 3. Step j makes a displacement of kj = 1. Examples of eight distinct 3-step walks with step length L = 1 cm.

Random walk leading to diffusion

Random walk in d = 1 dimension:
For simplicity, consider a random walk in 1D. Position after n steps xn, x0 = 0, is constant, and jumps take place with an interval of . x1 = x2 = x1 xN = xN-1 Clearly < xn > = 0 for all n. The second moments, however, yield < x12 > = < xN2 > = N
2 2 2

Step n

< x22 > = < x12 > < x1 > +

=2

Here the discrete step number N relates to time t, thus we have < x2(t) > = ( t / ) Hence
2

. Diffusion coefficient by Einsteins definition

Random walk and diffusion

Note dimensions: [D] = Length2 / time

Slope of the mean-squared displacement at long times is proportional to the diffusion coefficient.
Diffusion coefficient by Einsteins definition

Single walk

Average over many walks

Diffusion and diffusion length

Random walk in D dimensions:
Generally in d dimensions, < r2(t) > = 2d D t , and one can define the diffusion coefficient via the Einsteins relation D = (1 / 2d) lim < r2(t) > / t .
t0

Diffusion length:
To characterize the length scale crossed by a diffusing particle during a time scale t, define the diffusion length lD = ( 2d t D )1/2 . For typical solvents D 105 cm2 / s. Thus in 3D they cover 2.4 nm in 1 ns 7.7 nm in 10 ns 24.5 nm 100 ns 77.5 nm in 1000 ns . The cell diameter would be crossed in about 0.1 seconds.

Importantly, there is the limit of long times there are reasons for the prefactor of 2d the above coefficient describes the motion of single particles only those who are interested, can recall how the random walk gives rise to a Gaussian distribution in time and space

Friction coupled to diffusion

Consider a time period t Suppose only one collision to occur during t Between collisions, there is free flow subject to force f (e.g. gravitation). Then dvx / dt = f / m , thus vx(t) = v0x + f t /m The resulting uniformly accelerating motion of the particle is then x = v0x t + (1/2)(f/m) (t)2 The average < x > = (f / 2m) (t)2 This implies that there is a net drift velocity equal to (x/t), which is vdrift = f / , = 2m / t. The friction coefficient hence depends on e.g. the inter-particle collision rate (the period of free flow). The well-known relation is the Stokes formula = 6 R, for size R and viscosity .

Friction coupled to diffusion

To couple diffusion and friction, Einstein further used the equipartition theorem < (v0x)2 > = kBT / m Then combining the equipartition theorem with the above equations D = L2 / t = 2m / t , gives us D = (L / t )2 t = < v0x2 > t = (kBT / m) (2m / ) . Einstein therefore found D kBT. This is an example of the famous fluctuation-dissipation theorem, which couples microscopic fluctuations with macroscopic transport coefficients. t = 2m /

Friction & Diffusion: Langevin equation

Brownian diffusion for a colloidal particle, using the Langevin equation
Diffusion coefficient by Einsteins definition

Langevins form

White noise

Equipartition theorem Useful tips

Brownian motion & random walk

Un-biased Random Walk on a 2D P = 1/4 square lattice 1. 2. 3. 4. Start from the origin
P = 1/4 P = 1/4

Self-Avoiding Random Walk on a 2D lattice 1. 2. 3. Start from the origin Choose one of 4 nearest neighbors randomly Make a jump to the chosen nearest neighbor site provided that it has never been visited before Go to 2.

Choose one of 4 nearest neighbors randomly Make a jump to the chosen nearest neighbor site Go to 2.

P = 1/4

4.

Coupling to polymer models

Self-avoiding random walk

Polymers as random walks

Polymerization The word (poly)-(mer) means (many)-(parts) and refers to molecules consisting of many repeating units, called monomers. Monomers are structural repeating units of a polymer connected by covalent bonds (during polymerization). The number of monomers in a polymer is called its degree of polymerization N. The molar mass M of a polymer is equal to N times the molar mass of its monomer: M = N Mmon .

Two sequence isomers of polypropylene

Ideal chain
A flexible polymer of n+1 monomers (n bonds). The bond vector ri goes from monomer Ai-1 to monomer Ai. Ideal state corresponds to a situation where there are no interactions between Ai and Aj for which |i j| >> 1. li = | ri |.

For a freely jointed chain model l = constant and

Polymers and random walks

Random walk in 2 dimensions

Constructing polymers can be interpreted as an unbiased / biased random walk depending on the model considered. For an ideal polymer without any correlations, one has a pure random walk. Then what is the distribution of the end-to-end distance? For a one-dimensional case, it turns out to be

Please find the Gaussian nature!

Radius of gyration
Polymer of N monomers whose positions are given by We assume all monomers to have the same mass m. As a measure of size, one can use the end-to-end distance for linear chains, or the radius of gyration for any polymer.
Center of mass position

For an ideal chain one can derive Rg2 = Ree2 / 6 (prove it)

Real chains: example

Scaling of polymer size:

Rg ~ N x
The exponent x 0.79 +/ error for single DNA molecules confined to 2D.
Maier & Rdler, Phys. Rev. Lett. 82, 1911 (1999).

Diffusion equation
How to determine D? 1. 2. Follow a single moving particle Follow the distribution of many (many!) diffusing particles

Assume non-uniform distribution in x, but uniform in y and z.

N(x) Total number of particles in a slot centered at x

Assume that at every time step t every particle moves a distance L to the left or the right, at random. In the slot represented by N(x L), there is an excess of particles. This leads to a flow to the right. The net number of particles crossing the boundary a from left to right is then [ N(x L) N(x) ] / 2

Diffusion equation
Considering L to be very small, we get N(x L) N(x) = L (dN/dx). Dividing N(x) by the volume of the slice (V = XYZ), we are given the number density of particles, c(x). Then the flux (number of particles crossing a (from left to right) per unit area per unit time) is

j=

1 YZ t

1 2

d dx

LYZ c(x)

1 t

L2 2

dc dx

This corresponds to Ficks first law j= D dc dx The flux is hence driven by the density gradient

Diffusion equation
This corresponds to Ficks first law j= D dc dx

For practical purposes, this is easier to measure in a different form using the continuity equation Net change of particles across the surface a is d N(x) / dt = [ YZ j(x (L/2)) YZ j(x + (L/2)) ] In the limit of small L, this leads to known as the continuity equation. dc/dt = d/dx ( D dc / dx) ) dc/dt = d j /dx

Combining the above two yields the diffusion equation dc/dt = D d2 c / dx2
Provided that D is constant

Diffusion equation & Ficks laws

Macroscopically the diffusion of particles (mass) is traditionally described in terms of the diffusion equation and Ficks laws. Let the flow of matter be Assume linear response, when c is the concentration, A some constant, and a volume element.

Ficks 1. law

c x
comes out goes in

Continuity equation

Ficks 2nd law

Usually this is not the tracer (single-particle) diffusion coefficient

Biological applics. of tracer diffusion

T. Schmidt et al. Biophys. J. 77, 2638 (1999).

Lateral diffusion of lipids in the plane of the membrane Single-particle tracking

Fluorescence image of an individual labeled lipid.

Fluorescent probe

Do fluorescent probes perturb the structure and dynamics of lipid membranes? See: J. Repakova et al. Biophys. J. (2005).

Trajectory of the lipid in the plane of the membrane (from above, every 5 ms).

Example: motion of individual proteins

A 30-second movie of fluorescencelabeled proteins moving on a cell membrane. Courtesy of Thomas Schmidt.

Kusumi group / Nagoya

Example: fluctuations of ordered domains

Movie based on atomistic simulations for density fluctuations in a lipid membrane Courtesy of Teemu Murtola

Teemu Murtola / TKK

Lateral Lipid Diffusion

Lateral lipid trajectories in a DPPC bilayer of 1152 lipids over a period of 30 ns
4 systems with number of lipids ranging between 128 4096 Time scale 10 100 ns Less than 10 events observed where a lipid moves ~0.7 nm in a short period of ~100 ps. That is, the simulations indicate that there are no single-particle jumps
Falck, Rog, Karttunen, Vattulainen, J Am Chem Soc 130, 44 (2008)

Lateral Lipid Diffusion

Lateral lipid displacement s over t = 1 ns

A more detailed consideration reveals that all diffusive motions are collective ones, as nearby lipids move in unison as loosely defined clusters.
Falck, Rog, Karttunen, Vattulainen, J Am Chem Soc 130, 44 (2008)

Lateral Lipid Diffusion

t = 0.05 ns t = 0.5 ns Lateral displacements of individual lipids during a period of t
On a molecular scale, lipids move in unison as loosely defined clusters. On larger scales, the intimately correlated motions of neighboring lipids manifest themselves as 2D flow patterns
Falck, Rog, Karttunen, Vattulainen, J Am Chem Soc 130, 44 (2008)

t = 5 ns

t = 30 ns