🎲Statistical Mechanics Unit 8 Review

Diffusion describes the net movement of particles from regions of high concentration to regions of low concentration, driven by random thermal motion. It governs transport phenomena in gases, liquids, and solids at the molecular level, and it connects directly to the statistical foundations of kinetic theory.

Fick's laws

Fick's first law relates the diffusive flux to the concentration gradient. In steady state, the flux of particles is proportional to how steeply concentration changes in space:

$J = -D \nabla c$

Here $D$ is the diffusion coefficient (with units $m^2/s$ ), $J$ is the particle flux, and $\nabla c$ is the concentration gradient. The negative sign captures the physical fact that particles flow down the gradient, from high to low concentration.

Fick's second law (the diffusion equation) describes how concentration evolves in time and is covered in the next section.

Brownian motion

Brownian motion is the erratic, random motion of particles suspended in a fluid, caused by incessant collisions with the surrounding molecules. Robert Brown first observed it in 1827 watching pollen grains in water, but it was Einstein's 1905 theory that provided the quantitative framework.

The key result: the mean square displacement of a Brownian particle grows linearly with time, connecting the microscopic randomness of molecular collisions to the macroscopic diffusion coefficient. This linear growth is the hallmark of normal (Fickian) diffusion.

Random walk model

The random walk is the simplest microscopic model of diffusion. A particle takes a series of steps in random directions, with each step independent of the previous ones (a Markovian process).

The walk can be 1D, 2D, or 3D
After many steps, the probability distribution of the particle's position converges to a Gaussian (by the central limit theorem)
The mean displacement is zero (no net drift), but the mean square displacement grows linearly with the number of steps

This model provides the microscopic justification for Fick's laws and the diffusion equation.

Diffusion equation

The diffusion equation describes how a concentration profile spreads out over space and time. It's the same mathematical form as the heat equation, so techniques you learn here apply directly to heat conduction problems as well.

Derivation from Fick's laws

The derivation combines two ingredients:

Fick's first law: $J = -D \nabla c$ (flux is proportional to concentration gradient)
The continuity equation: $\frac{\partial c}{\partial t} = -\nabla \cdot J$ (particles are conserved locally)

Substituting the first into the second gives the diffusion equation:

$\frac{\partial c}{\partial t} = D \nabla^2 c$

This assumes $D$ is constant. If $D$ varies in space or time, you'd need to keep it inside the divergence operator: $\frac{\partial c}{\partial t} = \nabla \cdot (D \nabla c)$ .

Solutions for various geometries

Analytical solutions exist for simple geometries:

Infinite domain: The fundamental solution (Green's function) is a spreading Gaussian: $c(x,t) = \frac{N}{\sqrt{4\pi D t}} \exp\left(-\frac{x^2}{4Dt}\right)$ in 1D
Finite domains (slab, sphere, cylinder): Separation of variables yields solutions as series of eigenfunctions (sines, Bessel functions, etc.)
Periodic boundary conditions: Fourier series solutions are natural

For complex or irregular geometries, numerical methods (finite difference, finite element) are typically required.

Boundary conditions

To solve the diffusion equation, you need both boundary conditions and an initial condition:

Dirichlet: concentration specified at the boundary (e.g., surface held at fixed concentration)
Neumann: flux specified at the boundary (e.g., impermeable wall means zero flux)
Robin (mixed): a linear combination of concentration and flux at the boundary (models surface reactions or partial permeability)
Periodic: used for systems with repeating structure

The choice of boundary conditions reflects the physical setup and strongly affects the solution.

Microscopic theory of diffusion

This is where kinetic theory meets diffusion. The goal is to derive macroscopic quantities like $D$ from microscopic particle properties.

Einstein-Smoluchowski relation

This relation connects the diffusion coefficient to the mobility of a particle and the thermal energy:

$D = \mu k_B T$

Here $\mu$ is the particle mobility (drift velocity per unit force), $k_B$ is Boltzmann's constant, and $T$ is temperature. The physical content is profound: diffusion and drift are two manifestations of the same underlying random thermal motion. This is an example of a fluctuation-dissipation relation, and it holds for systems in thermal equilibrium.

For a spherical particle of radius $a$ in a fluid of viscosity $\eta$ , the Stokes drag gives $\mu = 1/(6\pi \eta a)$ , yielding the Stokes-Einstein relation: $D = \frac{k_B T}{6\pi \eta a}$ .

Diffusion coefficient

The diffusion coefficient $D$ quantifies how fast particles spread through a medium. Several factors control its value:

Temperature: higher $T$ means faster diffusion. In many solids, the temperature dependence follows an Arrhenius form: $D = D_0 \exp(-E_a / k_B T)$ , where $E_a$ is the activation energy
Particle size: larger particles diffuse more slowly (see Stokes-Einstein above)
Medium properties: viscosity, density, and intermolecular interactions all matter

Typical orders of magnitude: $\sim 10^{-5}$ $m^2/s$ in gases, $\sim 10^{-9}$ $m^2/s$ in liquids, and $\sim 10^{-10}$ to $10^{-15}$ $m^2/s$ in solids.

Mean square displacement

The mean square displacement (MSD) is the central observable connecting microscopic motion to the diffusion coefficient:

$\langle r^2 \rangle = 2dDt$

where $d$ is the spatial dimensionality (1, 2, or 3). For a 3D system, $\langle r^2 \rangle = 6Dt$ .

Linear growth of MSD with time signals normal diffusion
Any deviation from linearity indicates anomalous diffusion (see below)
Experimentally, you can extract $D$ by plotting MSD vs. time and measuring the slope

Diffusion in different media

Diffusion rates and mechanisms vary enormously depending on the state of matter. The differences trace back to how tightly packed the molecules are and how strongly they interact.

Gases vs liquids

In gases, molecules travel relatively long distances between collisions (large mean free path), so diffusion is fast. Typical gas-phase diffusion coefficients are on the order of $10^{-5}$ $m^2/s$ at standard conditions.

In liquids, molecules are much closer together and interact more strongly. Diffusion coefficients drop to roughly $10^{-9}$ $m^2/s$ . Temperature and pressure effects are more dramatic in gases because the mean free path is sensitive to both.

Fick's laws, Fick's laws of diffusion - Wikipedia

Solids and crystal lattices

Diffusion in solids is much slower still and proceeds through specific mechanisms:

Vacancy diffusion: an atom jumps into a neighboring empty lattice site
Interstitial diffusion: small atoms (like carbon in iron) hop between interstitial sites
Grain boundary diffusion: faster transport along defects and interfaces

Because each jump requires overcoming an energy barrier, solid-state diffusion is strongly temperature-dependent and well-described by the Arrhenius equation.

Porous materials

In porous media, the geometry of the pore network modifies diffusion:

When pore diameters are much larger than the mean free path, ordinary (Fickian) diffusion dominates
When pore diameters are comparable to the mean free path, Knudsen diffusion takes over: molecules collide with pore walls more often than with each other
The effective diffusion coefficient accounts for the porosity (void fraction) and tortuosity (how winding the paths are)

Models like the dusty gas model and Maxwell-Stefan equations handle multicomponent diffusion in porous media. Applications include catalysis, gas separation membranes, and contaminant transport in soils.

Anomalous diffusion

Normal diffusion predicts $\langle r^2 \rangle \propto t$ . In many real systems, this breaks down. Anomalous diffusion is characterized by:

$\langle r^2 \rangle \propto t^\alpha$

where $\alpha \neq 1$ .

Subdiffusion vs superdiffusion

Subdiffusion ( $\alpha < 1$ ): particles spread more slowly than expected. This happens in crowded or disordered environments where particles get temporarily trapped. A classic example is protein diffusion in the crowded cytoplasm of a cell.
Superdiffusion ( $\alpha > 1$ ): particles spread faster than expected. This occurs in systems with long-range correlations, active transport, or Lévy flight-like dynamics. Turbulent transport is a common physical example.

Fractional diffusion equation

To model anomalous diffusion mathematically, the standard diffusion equation is generalized using fractional calculus:

$\frac{\partial^\beta c}{\partial t^\beta} = D_\alpha \nabla^\alpha c$

Here $\beta$ and $\alpha$ are fractional orders of the time and space derivatives, respectively. When $\beta = 1$ and $\alpha = 2$ , you recover the standard diffusion equation. Fractional time derivatives introduce memory effects (the system's future depends on its entire history), while fractional space derivatives capture long-range spatial correlations.

Examples in complex systems

Protein diffusion in cell membranes often shows subdiffusion due to membrane heterogeneity and crowding
Transport in fractal structures (percolation clusters, porous rocks) displays anomalous scaling
Contaminant spreading in heterogeneous aquifers can deviate from Fickian predictions
Turbulent flows produce superdiffusive particle transport
Financial market price fluctuations have been modeled with superdiffusive frameworks

Collective diffusion

So far, most of the discussion has focused on single-particle (tracer) diffusion. In real many-body systems, particle-particle interactions change the picture.

Self-diffusion vs collective diffusion

These are two distinct quantities:

Self-diffusion tracks the motion of a single tagged particle in a sea of identical particles. It's measured by following individual trajectories (e.g., with fluorescent labels).
Collective diffusion describes how density fluctuations relax. It's what Fick's law describes macroscopically, and it governs how concentration gradients decay.

In an ideal (non-interacting) system, the two are identical. In interacting systems, they can differ substantially. For example, in a concentrated colloidal suspension, repulsive interactions can make collective diffusion faster than self-diffusion.

Onsager reciprocal relations

In systems with multiple coupled transport processes (e.g., simultaneous diffusion of heat and mass), the Onsager relations constrain the transport coefficients:

$L_{ij} = L_{ji}$

These symmetry relations hold near equilibrium and follow from microscopic reversibility (detailed balance). They're essential for multicomponent diffusion and thermodiffusion (the Soret effect), where a temperature gradient drives mass transport.

Kubo formula

The Kubo formula connects transport coefficients to equilibrium fluctuations via time correlation functions:

$D = \frac{1}{d} \int_0^\infty \langle \mathbf{v}(0) \cdot \mathbf{v}(t) \rangle \, dt$

This is a Green-Kubo relation: the diffusion coefficient equals the time integral of the velocity autocorrelation function, divided by the dimensionality $d$ . It's derived from linear response theory and is the standard way to compute $D$ from molecular dynamics simulations. Analogous formulas exist for viscosity (stress autocorrelation) and thermal conductivity (heat flux autocorrelation).

Applications of diffusion

Materials science

Semiconductor doping: controlled diffusion of dopant atoms (e.g., phosphorus into silicon) creates the p-n junctions that make transistors work
Surface hardening: carburization and nitriding rely on diffusing carbon or nitrogen into metal surfaces
Diffusion bonding: joining materials at high temperature by allowing atoms to interdiffuse across an interface
Diffusion barriers: thin layers in microelectronics that prevent unwanted atomic migration between device layers
Solid-state batteries: ion diffusion through solid electrolytes determines charge/discharge rates

Biological systems

Gas exchange: oxygen diffuses from alveoli into blood in the lungs; $CO_2$ diffuses the other way
Synaptic transmission: neurotransmitters released into the synaptic cleft diffuse to receptors on the postsynaptic cell
Morphogen gradients: during embryonic development, signaling molecules spread by diffusion to create concentration patterns that guide cell differentiation
Passive membrane transport: small nonpolar molecules cross cell membranes by simple diffusion down their concentration gradient

Fick's laws, pde - Deriving Fick's Principle from the Equation of Conservation of Matter - Mathematics Stack ...

Chemical engineering

Catalysis: in heterogeneous catalysis, reactants must diffuse to the catalyst surface and products must diffuse away; pore diffusion can be rate-limiting
Separation processes: dialysis, osmosis, and membrane separations exploit differences in diffusion rates
Controlled drug release: polymer matrices release drugs at rates governed by diffusion
Reactor design: mass transfer (diffusion) limitations often determine reactor performance and scaling

Experimental techniques

Neutron scattering

Neutron scattering probes atomic and molecular motions on length scales of angstroms and time scales of picoseconds to nanoseconds. Quasielastic neutron scattering (QENS) measures the broadening of the elastic peak, which is directly related to the self-diffusion coefficient. It's particularly powerful for studying hydrogen diffusion in metals and other materials, since hydrogen has a large neutron scattering cross-section.

Fluorescence correlation spectroscopy

FCS measures diffusion by monitoring fluorescence intensity fluctuations in a tiny observation volume (typically a femtoliter). As fluorescently labeled molecules diffuse in and out of this volume, the intensity fluctuates. The autocorrelation function of these fluctuations yields the diffusion coefficient and concentration. FCS is widely used in biophysics to study molecular diffusion in solutions and cell membranes, and it can detect anomalous diffusion when the autocorrelation deviates from the standard model.

Pulsed field gradient NMR

PFG-NMR is a non-invasive technique that measures self-diffusion coefficients by applying magnetic field gradient pulses. The basic idea:

A gradient pulse labels the spatial position of nuclear spins
Spins diffuse during a waiting period
A second gradient pulse detects how far the spins have moved
The signal attenuation depends on the diffusion coefficient

This technique works well for liquids, polymers, and molecules confined in porous materials. It can also measure anisotropic diffusion (different rates in different directions) in oriented systems like liquid crystals.

Computational methods

Molecular dynamics simulations

MD simulations integrate Newton's equations of motion for a system of interacting particles. To extract diffusion coefficients, you can either:

Compute the MSD over time and use $D = \lim_{t \to \infty} \frac{\langle r^2 \rangle}{2dt}$
Compute the velocity autocorrelation function and integrate it (Kubo formula)

Modern MD can handle millions of atoms, but the results are only as good as the interatomic potentials used. These simulations reveal diffusion mechanisms at atomic resolution that experiments alone cannot access.

Monte Carlo methods

Monte Carlo approaches use random sampling to simulate diffusion:

Kinetic Monte Carlo (KMC): particularly efficient for rare-event dynamics like vacancy diffusion in solids, where the time between jumps is long compared to vibrational periods
Metropolis algorithm: samples equilibrium configurations and can be used to study thermodynamic properties of diffusing systems
Lattice gas models: simulate particles hopping on discrete lattice sites, useful for studying diffusion in porous media and on surfaces

Lattice Boltzmann models

The lattice Boltzmann method discretizes the Boltzmann transport equation onto a lattice. Rather than tracking individual particles, it evolves distribution functions at each lattice node through streaming and collision steps. It handles complex boundary conditions naturally and is well-suited for modeling diffusion coupled with fluid flow, multiphase systems, and multicomponent transport.

Advanced topics

Diffusion-limited reactions

When reactants diffuse slowly compared to the intrinsic reaction rate, the overall kinetics become diffusion-limited. Smoluchowski theory predicts the rate constant for such reactions by solving the diffusion equation with an absorbing boundary at the reaction radius.

Reaction-diffusion equations couple local reaction kinetics with spatial diffusion:

$\frac{\partial c}{\partial t} = D \nabla^2 c + R(c)$

where $R(c)$ describes the reaction terms. These equations can produce spontaneous pattern formation (Turing patterns), relevant to morphogenesis, chemical oscillations, and ecological spatial dynamics.

Diffusion in non-equilibrium systems

Far from equilibrium, diffusion can behave in counterintuitive ways. External driving forces, chemical potential gradients, or energy input can produce phenomena like:

Uphill diffusion: particles moving against their concentration gradient (driven by coupling to other gradients)
Pattern formation: sustained non-equilibrium conditions can stabilize spatial structures
Active matter: self-propelled particles (bacteria, molecular motors) exhibit enhanced or anomalous diffusion

Non-equilibrium statistical mechanics provides the theoretical framework, but many problems remain open.

Stochastic differential equations

At the microscopic level, diffusion is inherently stochastic. Two key equations capture this:

Langevin equation: $m\ddot{x} = -\gamma \dot{x} + \xi(t)$ , where $\gamma$ is the friction coefficient and $\xi(t)$ is a random force (white noise). This describes the trajectory of a single Brownian particle.
Fokker-Planck equation: describes the time evolution of the probability density $P(x,t)$ for the particle's position. It's the deterministic counterpart to the stochastic Langevin equation.

The Itô and Stratonovich interpretations of stochastic calculus give different prescriptions for handling the noise term, which matters when the noise amplitude depends on position (multiplicative noise). Applications extend beyond physics to financial modeling, population dynamics, and chemical kinetics.