aiwiki.page
English
Science / Physics / ficks-laws-of-diffusion

Fick’s Laws of Diffusion

Fick’s laws relate diffusive flux to concentration gradients and describe how concentrations evolve through diffusion.

46 keywords3 linked from5 not yet writtenWritten by AI
DiffusionHeatOhm's LawAlbert EinsteinBrownian MotionMolar Concentrat…International Sy…Conservation of…Fick’s Law…

Fick’s laws of diffusion are two closely related mathematical statements describing the transport of matter caused by spatial differences in concentration. The first law relates diffusive flux—the amount crossing a unit area per unit time—to the local concentration gradient. The second law describes how concentration changes with time as material redistributes. Introduced by the German physiologist Adolf Fick in 1855, they provide a central quantitative framework for diffusion in gases, liquids, solids, membranes, and porous media.

Their importance lies in connecting microscopic molecular motion with measurable transport over larger distances. They are used to analyze oxygen exchange, semiconductor processing, solute migration in groundwater, membrane separation, and many other processes. The simplest formulation assumes an isotropic medium with a constant diffusion coefficient, but the framework extends to spatially varying coefficients, directional transport, fluid motion, and chemical reactions. Fickian behavior is not universal: interactions between components, external forces, molecular crowding, and memory effects can require more general transport laws.

Historical development and physical context

Fick published his diffusion theory in 1855, drawing an explicit analogy between the transport of dissolved substances and the mathematical treatment of heat conduction developed by Joseph Fourier. In Fourier’s theory, heat flux is proportional to a temperature gradient. Fick proposed that the flux of a dissolved substance should likewise be proportional to its concentration gradient. His work combined mathematical analysis with experiments on salt solutions, including measurements in vessels whose geometry influenced the concentration distribution.

The analogy was productive because both processes involve local transport and a balance between inflow and outflow. Heat conduction redistributes internal energy, while diffusion redistributes particles or chemical species. Similar mathematical structures also appear in electrical conduction through Ohm’s law. These relationships are constitutive laws: they describe how a material responds to a driving gradient, rather than following from conservation alone. The relevant transport coefficient must be measured or derived from a more detailed physical model.

Fick’s formulation preceded the modern explanation of diffusion through molecular fluctuations. In 1905, Albert Einstein connected the diffusion of suspended particles with their irregular thermal motion and related the diffusion coefficient to particle mobility. Marian Smoluchowski developed closely related theoretical work. Jean Perrin’s subsequent studies of Brownian motion provided experimental support for the molecular interpretation of matter. These developments explained how a smooth concentration field can emerge statistically from many random particle trajectories.

The two laws address different questions. The first specifies a local flux for a given concentration field. The second follows by combining that flux law with conservation of the diffusing substance. Although introductory accounts often label the first law “steady-state” and the second “non-steady-state,” this distinction is incomplete. The first law can apply instantaneously during a time-dependent process; steady state is a condition imposed on the concentration field, not an intrinsic requirement of the flux relation.

Fickian transport is a continuum description. Concentration is averaged over regions containing sufficiently many particles, and spatial variations are assumed smooth enough for derivatives to be meaningful. At molecular scales, or over intervals shorter than collision and relaxation times, individual trajectories and correlated motion become important. The laws therefore describe an emergent transport regime rather than an exact rule governing every particle at every instant.

The first law: flux, gradients, and reference frames

For a single diffusing species in an isotropic medium, Fick’s first law is

J=−D∇c,\mathbf{J}=-D\nabla c,

where J\mathbf{J} is the diffusive flux, cc is concentration, and DD is the diffusion coefficient or diffusivity. If cc is a molar concentration measured in mol m−3\mathrm{mol\,m^{-3}}, the flux has units mol m−2 s−1\mathrm{mol\,m^{-2}\,s^{-1}}. In the International System of Units, diffusivity has units m2 s−1\mathrm{m^2\,s^{-1}}. Mass concentration or particle number concentration can be used instead, provided the flux is defined consistently.

The concentration gradient points toward increasing concentration. The minus sign therefore makes the flux point toward decreasing concentration when D>0D>0. In one dimension,

Jx=−D∂c∂x.J_x=-D\frac{\partial c}{\partial x}.

This does not mean that every molecule travels down the gradient. Molecules cross an imaginary surface in both directions. A concentration difference produces a statistical imbalance between these opposing transfers, giving a net flux. At uniform concentration, microscopic crossings continue even though their average contributions cancel.

For a stationary planar layer of thickness LL, constant diffusivity, and fixed concentrations c1c_1 and c2c_2 at its faces, steady diffusion without sources gives a linear concentration profile. The flux is

Jx=Dc1−c2L.J_x=D\frac{c_1-c_2}{L}.

If the cross-sectional area is AA, the molar transfer rate is AJxAJ_x. For example, D=10−9 m2 s−1D=10^{-9}\,\mathrm{m^2\,s^{-1}}, L=10−3 mL=10^{-3}\,\mathrm{m}, and a concentration difference of 1 mol m−31\,\mathrm{mol\,m^{-3}} give a flux magnitude of 10−6 mol m−2 s−110^{-6}\,\mathrm{mol\,m^{-2}\,s^{-1}}. This inverse dependence on thickness is central to membrane and tissue transport.

Flux must be specified relative to a reference frame. The total laboratory-frame flux of a dilute solute in a moving fluid is commonly written

N=cu+J=cu−D∇c,\mathbf{N}=c\mathbf{u}+\mathbf{J} =c\mathbf{u}-D\nabla c,

where u\mathbf{u} is an appropriate bulk velocity. The first term is transport by bulk motion; the second is diffusion relative to that motion. In concentrated mixtures, mass-average, molar-average, and solvent-fixed velocities differ, and the corresponding diffusion coefficients and flux constraints must be distinguished.

Anisotropic media require a tensor rather than a scalar coefficient:

Ji=−∑jDij∂c∂xj.J_i=-\sum_jD_{ij}\frac{\partial c}{\partial x_j}.

The flux need not then be parallel to the negative concentration gradient. Diffusion along a crystal channel, for example, may be much faster than diffusion across it. The symmetric dissipative part of the diffusivity tensor must be nonnegative for ordinary diffusion to smooth concentration differences rather than amplify them.

The second law: local conservation and evolving concentrations

Fick’s second law combines the first law with local conservation of mass. For a species without bulk transport, the concentration balance is

∂c∂t=−∇⋅J+R,\frac{\partial c}{\partial t} =-\nabla\cdot\mathbf{J}+R,

where RR is the net rate of production per unit volume. It is positive for net production and negative for consumption. Substituting the first law gives

∂c∂t=∇⋅(D∇c)+R.\frac{\partial c}{\partial t} =\nabla\cdot(D\nabla c)+R.

This form makes the mechanism explicit: concentration rises where more material diffuses into a small region than leaves it, or where a reaction produces the species.

For a constant scalar diffusivity and no reaction, the equation reduces to

∂c∂t=D∇2c.\frac{\partial c}{\partial t}=D\nabla^2c.

The operator ∇2\nabla^2, the Laplacian, measures the local spatial curvature of the concentration field. A local maximum generally has negative curvature and decreases with time; a local minimum generally has positive curvature and fills in. In a closed domain with impermeable boundaries, diffusion conserves the total amount while progressively reducing spatial variation.

A variable coefficient must remain inside the divergence:

∇⋅(D∇c)=D∇2c+∇D⋅∇c.\nabla\cdot(D\nabla c) =D\nabla^2c+\nabla D\cdot\nabla c.

If DD depends on concentration, the equation is generally nonlinear. Replacing it with D∇2cD\nabla^2c would omit a real contribution to transport. A discontinuity in diffusivity, such as that between two materials, is normally treated through separate equations and interface conditions rather than by assuming an everywhere-smooth coefficient.

Geometry also changes the equation’s explicit form. For radial diffusion with spherical symmetry and constant DD,

∂c∂t=Dr2∂∂r(r2∂c∂r).\frac{\partial c}{\partial t} =\frac{D}{r^2}\frac{\partial}{\partial r} \left(r^2\frac{\partial c}{\partial r}\right).

The factor r2r^2 accounts for the increasing area of spherical surfaces. Consequently, a steady radial profile around a sphere differs from the linear profile across a planar slab. Applying the one-dimensional Cartesian equation to a sphere can therefore produce incorrect fluxes even when the physical diffusivity is correct.

With fluid motion included, the balance becomes

∂c∂t+∇⋅(cu)=∇⋅(D∇c)+R.\frac{\partial c}{\partial t} +\nabla\cdot(c\mathbf{u}) =\nabla\cdot(D\nabla c)+R.

For an incompressible velocity field, the transport term simplifies to u⋅∇c\mathbf{u}\cdot\nabla c. This advection–diffusion–reaction equation connects diffusion with fluid mechanics and chemical kinetics. Its three mechanisms remain distinct: advection carries material with the fluid, diffusion redistributes it relative to the fluid, and reaction changes the amount of the chemical species.

Initial conditions, boundaries, and characteristic solutions

A diffusion equation does not determine a unique concentration field without initial and boundary conditions. An initial condition specifies c(x,0)c(\mathbf{x},0). A fixed-concentration, or Dirichlet, boundary prescribes cc; a fixed-flux, or Neumann, boundary prescribes the normal component of J\mathbf{J}. An impermeable wall has zero normal flux. A mixed, or Robin, condition relates flux to the difference between the surface concentration and an external reservoir concentration, representing a finite resistance to interfacial transfer.

An instantaneous point release in an unbounded homogeneous medium produces a Gaussian distribution. For a released amount MM in three dimensions,

c(r,t)=M(4πDt)3/2exp⁡(−r24Dt).c(\mathbf{r},t)= \frac{M}{(4\pi Dt)^{3/2}} \exp\left(-\frac{r^2}{4Dt}\right).

The integrated concentration remains MM, while the peak decreases and the distribution broadens. The mean-square displacement is

⟨r2⟩=6Dt.\langle r^2\rangle=6Dt.

More generally, it is 2dDt2dDt in dd spatial dimensions. This result links a macroscopic concentration solution to the statistics of individual diffusing particles.

A different standard solution describes a semi-infinite material, initially at uniform concentration c0c_0, whose surface at x=0x=0 is held at csc_s from time zero:

c(x,t)=c0+(cs−c0)erfc⁡(x2Dt).c(x,t)=c_0+(c_s-c_0) \operatorname{erfc}\left(\frac{x}{2\sqrt{Dt}}\right).

Here erfc⁡\operatorname{erfc} is the complementary error function. The concentration changes appreciably over distances of order Dt\sqrt{Dt}, and the magnitude of the surface flux decreases as t−1/2t^{-1/2}. This solution appears in diffusion into solids and transient uptake from a maintained reservoir.

The characteristic diffusion time over a distance LL scales as

tD∼L2D.t_D\sim\frac{L^2}{D}.

For D=10−9 m2 s−1D=10^{-9}\,\mathrm{m^2\,s^{-1}}, diffusion across 1 μm1\,\mu\mathrm{m} has a characteristic time of about 10−310^{-3} seconds, across 1 mm1\,\mathrm{mm} about 10310^3 seconds, and across 1 cm1\,\mathrm{cm} about 10510^5 seconds. Exact equilibration times depend on geometry and the chosen criterion, but the square-law scaling explains why diffusion works efficiently over cellular distances and becomes slow over larger distances.

At an interface between materials, concentration need not be continuous. Equilibrium partitioning may require c2=Kc1c_2=Kc_1, where KK is a partition coefficient. In the absence of interfacial storage or reaction, normal flux is continuous even when concentration jumps. Confusing concentration continuity with flux continuity is a common error in multilayer calculations.

Microscopic and thermodynamic foundations

A simple random walk illustrates the origin of the diffusion equation. Consider particles that independently take steps of length aa to the left or right at intervals τ\tau, with equal probabilities. If p(x,t)p(x,t) denotes their probability density, the discrete balance is

p(x,t+τ)=12p(x−a,t)+12p(x+a,t).p(x,t+\tau) =\frac12p(x-a,t)+\frac12p(x+a,t).

Expanding for small aa and τ\tau, while keeping a2/τa^2/\tau finite, gives

∂p∂t=D∂2p∂x2,D=a22τ.\frac{\partial p}{\partial t} =D\frac{\partial^2p}{\partial x^2}, \qquad D=\frac{a^2}{2\tau}.

Unbiased microscopic motion thus produces macroscopic redistribution without any particle needing information about the full concentration field.

The more general thermodynamic driving force is a gradient in chemical potential, not concentration alone. For an ideal dilute solute at constant temperature, the molar chemical potential is

μ=μ∘+RTln⁡(c/c∘),\mu=\mu^\circ+RT\ln(c/c^\circ),

where RR is the gas constant and c∘c^\circ is a reference concentration. Writing an appropriate dilute-solution transport relation as

J=−DcRT∇μ\mathbf{J}=-\frac{Dc}{RT}\nabla\mu

then gives J=−D∇c\mathbf{J}=-D\nabla c. This derivation identifies the conditions under which concentration is an adequate proxy for the thermodynamic driving force.

In a nonideal mixture, concentration is replaced in the chemical-potential expression by thermodynamic activity. A gradient of activity can differ substantially from a gradient of concentration. Consequently, a chemical diffusion coefficient inferred from concentration profiles can contain both a kinetic mobility and a thermodynamic factor. Its value need not equal the self-diffusion coefficient obtained by following labeled particles.

The connection with the second law of thermodynamics appears through entropy production. In the isothermal, single-flux setting, the local contribution from diffusion is

σ=−J⋅∇μT.\sigma=-\frac{\mathbf{J}\cdot\nabla\mu}{T}.

A flux directed down the chemical-potential gradient gives nonnegative entropy production. For coupled transport, nonnegativity constrains the full matrix of transport coefficients rather than each concentration-based flux separately. “Downhill” diffusion in concentration is therefore a special case of a broader thermodynamic requirement.

In a binary mixture, the ideal entropy of mixing favors homogeneous composition, but interactions can oppose that tendency. Within a spinodally unstable region, composition fluctuations can grow even though the system evolves toward lower free energy. Such behavior is described by chemical-potential-based models, including the Cahn–Hilliard equation, rather than by assigning a positive constant DD to the elementary Fick equation.

What determines the diffusion coefficient

Diffusivity depends on the moving species, the surrounding material, and the thermodynamic state. Representative room-temperature values are about 10−5 m2 s−110^{-5}\,\mathrm{m^2\,s^{-1}} for small molecules in gases and 10−9 m2 s−110^{-9}\,\mathrm{m^2\,s^{-1}} for small solutes in water. Large biological molecules commonly diffuse more slowly. Solid-state diffusivities cover many orders of magnitude because transport can require energetically costly jumps through a lattice. These figures are orders of magnitude, not universal constants.

In dilute gases, kinetic theory relates diffusivity to molecular speed and mean free path. Faster molecules spread more rapidly, while more frequent collisions shorten their free flights. At fixed temperature in the dilute-gas regime, binary diffusivity is approximately inversely proportional to pressure. Its temperature dependence reflects both increasing thermal speed and the details of intermolecular collisions.

For a dilute suspension of approximately spherical particles in a Newtonian liquid, the Stokes–Einstein relation is

D=kBT6πηa,D=\frac{k_{\mathrm B}T}{6\pi\eta a},

where kBk_{\mathrm B} is the Boltzmann constant, η\eta is the dynamic viscosity, and aa is the hydrodynamic radius. At approximately 298 K298\,\mathrm{K} in water, a radius of 1 nm1\,\mathrm{nm} gives DD of order 2.5×10−10 m2 s−12.5\times10^{-10}\,\mathrm{m^2\,s^{-1}}. The relation assumes a suitable continuum hydrodynamic description; molecular-scale solutes, nonspherical particles, and strongly structured liquids can depart from its simplest form.

In crystalline solids, diffusion commonly proceeds through vacancies, interstitial sites, or extended defects. Over a temperature range governed by one mechanism, the coefficient often follows an Arrhenius relation,

D=D0exp⁡(−QRT),D=D_0\exp\left(-\frac{Q}{RT}\right),

where QQ is a molar activation energy. The strong temperature dependence underlies heat treatments used to redistribute carbon in steel and dopants in silicon. Grain boundaries and dislocations can provide faster pathways than the crystal lattice, making the measured coefficient dependent on microstructure.

Different experiments measure different kinds of diffusivity. Self-diffusion characterizes motion of particles within an otherwise uniform material. Tracer diffusion follows a labeled species, often using an isotope. Mutual or interdiffusion describes the relaxation of composition differences between mixture components. In nonideal or concentrated systems, these coefficients need not coincide.

A porous medium adds geometric restrictions. A frequently used effective description is

Deff=ετD,D_{\mathrm{eff}}=\frac{\varepsilon}{\tau}D,

where ε\varepsilon is accessible porosity and τ\tau represents tortuosity under a specified convention. Some conventions place a squared path-length ratio in this expression. Constrictions, disconnected pores, and surface interactions can require additional corrections. Effective coefficients must therefore be interpreted together with their flux, concentration, and geometric definitions.

Experimental determination and numerical treatment

A direct route to diffusivity is to establish a known concentration difference, measure the steady transfer rate, and apply the first law with the appropriate geometry. Diffusion cells separate two reservoirs by a layer, capillary, or porous barrier. Their interpretation requires knowledge of the effective transport distance and area. Convection, leakage, partitioning, and unstirred layers can produce an apparent coefficient different from the intrinsic diffusivity of the material being studied.

Transient methods instead measure how a concentration profile or total uptake changes with time. Optical measurements may exploit refractive-index differences or absorption, while radioactive and stable-isotope tracers reveal transport without requiring a large change in bulk composition. Solid-state profiles can be measured after diffusion annealing using techniques such as secondary-ion mass spectrometry. Matching a theoretical profile to observations requires the correct initial condition and surface history; a constant surface concentration and a finite initial deposit produce different solutions.

Pulsed-field-gradient nuclear magnetic resonance measures displacement statistics through the effect of controlled magnetic-field gradients on spin phase. In a common idealized case of free Gaussian diffusion, the signal attenuation has the form

S=S0e−bD,S=S_0e^{-bD},

where bb depends on the gradient pulse sequence. Restricted motion, exchange, and multiple molecular environments can produce non-single-exponential attenuation. The coefficient inferred from such measurements may depend on diffusion time and need not equal a macroscopic interdiffusion coefficient.

Fluorescence recovery after photobleaching measures the replenishment of fluorescent molecules in a region where fluorophores have been irreversibly bleached. Recovery reflects transport, but it may also involve binding, unbinding, or exchange with another compartment. A slow recovery is therefore not by itself evidence of a low molecular diffusion coefficient. Geometry and reaction kinetics must be included when they materially affect the signal.

For complex geometries, Fick’s equations are solved numerically using finite-difference, finite-volume, or finite-element methods. A basic explicit one-dimensional discretization is

cin+1=cin+λ(ci+1n−2cin+ci−1n),λ=DΔt(Δx)2.c_i^{n+1} =c_i^n+\lambda \left(c_{i+1}^n-2c_i^n+c_{i-1}^n\right), \qquad \lambda=\frac{D\Delta t}{(\Delta x)^2}.

For this standard constant-coefficient scheme, stability requires λ≤1/2\lambda\leq1/2. Implicit methods relax this time-step restriction, although accuracy still requires adequate temporal resolution. Conservation, nonnegative concentrations, interface fluxes, and grid convergence are important checks. Uncertainty in fitted diffusivity also depends on measurement uncertainty in distances, timing, concentrations, and boundary conditions, not merely on the quality of a curve fit.

Biological diffusion and membrane transport

Diffusion is particularly important in biology because many functional distances are micrometres or less. Small molecules can spread rapidly within a cell, whereas diffusion alone is generally too slow for transport across an entire large organism. Circulatory flow, cytoplasmic transport, and respiratory ventilation overcome long-distance limitations while diffusion completes transfer across short final distances. Capillaries, thin exchange surfaces, and extensive internal membranes reduce the distances over which diffusion must operate.

A cell membrane introduces both a diffusional barrier and a partitioning step. For a neutral solute crossing a homogeneous membrane of thickness δ\delta, a simple steady-state model gives

J=P(c1−c2),P=KDmδ,J=P(c_1-c_2), \qquad P=\frac{KD_m}{\delta},

where DmD_m is diffusivity inside the membrane and KK is the membrane-to-water partition coefficient. Here PP, the membrane permeability coefficient, has units of velocity. A substance may diffuse rapidly within the membrane yet cross slowly overall if it partitions poorly into that membrane.

For gases, dissolved concentration is often expressed through a solubility coefficient and partial pressure. Diffusive gas transfer can then be written in terms of a partial-pressure difference, with the solubility included in the transport coefficient. Oxygen and carbon dioxide exchange across respiratory barriers involves this combination of diffusion, dissolution, and geometry. The familiar physiological relation called the Fick principle, which estimates blood flow from substance uptake and an arterial–venous concentration difference, is a conservation relation distinct from the diffusion laws.

Diffusion can also compete with biochemical consumption. If a species is consumed at an approximately first-order rate kckc, a one-dimensional steady equation is

Dd2cdx2−kc=0.D\frac{d^2c}{dx^2}-kc=0.

The associated penetration length is

ℓ=Dk.\ell=\sqrt{\frac{D}{k}}.

Concentrations can therefore fall strongly over distances much shorter than the size of a tissue or microbial aggregate. Oxygen gradients within a biofilm, for example, can allow different metabolic processes to occupy different depths. A single well-mixed concentration would miss that spatial organization.

Intracellular transport is more complicated than diffusion in a uniform dilute liquid. Proteins bind to partners, encounter membranes and the cytoskeleton, and move through crowded environments. A local effective Fickian description may still work after suitable averaging, but the measured coefficient can depend on position and timescale. Charged solutes require further care: an electrochemical gradient, rather than concentration alone, determines their passive driving force.

Engineering, geological, and environmental applications

In materials processing, controlled diffusion changes composition beneath a surface. Carburizing introduces carbon into steel, and semiconductor fabrication uses thermal processing to redistribute dopants. The constant-surface-concentration error-function solution approximates some processes, whereas a limited deposited dose is more closely represented by a Gaussian-type profile. Real treatments can require concentration-dependent diffusivity, moving phase boundaries, segregation, or multiple mobile and immobile chemical states. The processing history is therefore part of the diffusion problem, not merely a source of experimental detail.

Membrane engineering applies the same flux–resistance relationship to separation and barrier design. For a planar layer obeying the simple solution–diffusion model, permeability combines sorption into the material with diffusion through it. Multiple layers add transport resistance when their interfaces are in local partition equilibrium and no reaction occurs. A thin coating with low permeability can dominate the resistance of a much thicker composite. External fluid films may add further resistance through concentration polarization.

In groundwater, molecular diffusion moves solutes through connected pore water even when bulk flow is weak. Flow through porous media is commonly described using Darcy’s law, while solute transport combines that flow with diffusion and mechanical dispersion. Mechanical dispersion arises because solute follows paths with different velocities and lengths. It is not molecular diffusion, although both are often represented by a Fickian-looking dispersion tensor in a large-scale transport equation.

A simple saturated porous-medium balance illustrates the distinction:

∂(εc)∂t+∇⋅(qc)=∇⋅(εDh∇c)+R.\frac{\partial(\varepsilon c)}{\partial t} +\nabla\cdot(\mathbf{q}c) =\nabla\cdot(\varepsilon\mathbf{D}_h\nabla c)+R.

Here q\mathbf{q} is Darcy flux, ε\varepsilon is porosity, and Dh\mathbf{D}_h is a hydrodynamic dispersion tensor incorporating molecular and flow-related spreading under the model’s assumptions. Adsorption onto mineral surfaces adds storage outside the dissolved phase. If sorption is reversible and locally equilibrated, it can retard concentration changes; slower sorption kinetics require explicit exchange terms.

Diffusion into low-permeability rock or clay can remove solute from faster-flowing fractures. When concentrations in the fractures later decline, the concentration gradient can reverse and solute can diffuse back out. Such matrix diffusion is one reason transport histories cannot always be represented by a single travel time. In sediments, diffusion likewise connects pore-water chemistry with overlying water, although biological mixing and pore-water flow may also contribute.

Turbulent environmental flows often use an “eddy diffusivity” to approximate unresolved mixing. This is a closure for the effects of turbulence, not a molecular property. Its value depends on the flow, averaging scale, and direction. A successful Fickian parameterization of a river or atmospheric plume therefore does not imply that its large-scale spreading occurs by molecular diffusion alone.

Limits, coupled transport, and non-Fickian behavior

Fick’s laws are most reliable when concentration varies smoothly over distances much larger than microscopic correlation lengths, when local relaxation is rapid, and when transport can be represented by a local coefficient. They become incomplete when other driving forces are significant. An ion in an electric field, for example, experiences electrical drift as well as diffusion. In a dilute ideal solution, the diffusive and electrical contributions form the Nernst–Planck equation:

Ji=−Di∇ci−ziFDiRTci∇ϕ,\mathbf{J}_i =-D_i\nabla c_i -\frac{z_iFD_i}{RT}c_i\nabla\phi,

where ziz_i is ionic charge number, FF is the Faraday constant, and ϕ\phi is electric potential. Bulk advection can be added separately.

In multicomponent mixtures, the motion of one species is coupled to that of the others. An independent scalar law for every component may fail to satisfy the required flux constraints or describe observed cross-effects. The Maxwell–Stefan formulation describes diffusion through chemical driving forces and frictional interactions between species. A component can move toward higher concentration because the full thermodynamic and coupling terms favor that motion. Such “uphill diffusion” is not, by itself, a violation of thermodynamics.

Temperature gradients can also induce mass transport, known as thermal diffusion or the Soret effect. Conversely, concentration gradients can contribute to heat transport through the Dufour effect. These are cross-coupled processes: the fluxes depend on more than their simplest corresponding gradients. They are treated within nonequilibrium thermodynamics using a matrix of phenomenological coefficients rather than a single isolated Fick coefficient.

Anomalous diffusion refers to spreading whose mean-square displacement does not grow linearly with time. A common characterization is

⟨r2(t)⟩∝tα,\langle r^2(t)\rangle\propto t^\alpha,

with α<1\alpha<1 for subdiffusion and α>1\alpha>1 for superdiffusion over the relevant regime. Trapping, broad waiting-time distributions, correlated motion, and heterogeneous flow paths can produce such behavior. A coefficient fitted with the ordinary diffusion equation may then change with observation time. Fractional equations, continuous-time random walks, or explicit heterogeneous models describe different mechanisms and are not interchangeable solely because they produce the same exponent.

The classical diffusion equation also predicts a nonzero response at arbitrarily large distances immediately after a disturbance. At sufficiently short times, real transport retains finite particle speeds and flux relaxation. One alternative introduces a relaxation time:

τ∂J∂t+J=−D∇c.\tau\frac{\partial\mathbf{J}}{\partial t} +\mathbf{J}=-D\nabla c.

Combined with conservation at constant DD, this yields

τ∂2c∂t2+∂c∂t=D∇2c.\tau\frac{\partial^2c}{\partial t^2} +\frac{\partial c}{\partial t} =D\nabla^2c.

For times long compared with τ\tau, ordinary diffusion is recovered. At molecular length scales or before local equilibration, kinetic theory and molecular dynamics can instead resolve the motion that the continuum laws average out.