The Einstein Relation

The Einstein relation is the simplest and most fundamental statement of the fluctuation–dissipation principle. A particle immersed in a thermal medium experiences two manifestations of the same microscopic collisions: a randomly fluctuating force that drives diffusion, and an average drag force proportional to velocity that dissipates energy. The relation $D=\mu k_B T$ asserts that these two effects are not independent — they are rigidly tied together by the temperature, because both arise from a single equilibrium ensemble of molecular impacts. It is the archetypal fluctuation–dissipation identity, the special zero-frequency case of the full theorem developed in The Fluctuation-Dissipation Theorem. Historically the result appeared in Einstein's 1905 paper on Brownian motion, where it provided the crucial bridge between the macroscopic, measurable diffusion coefficient and the microscopic Boltzmann constant $k_B=R/N_A$ . William Sutherland reached the identical relation independently in 1905, and Smoluchowski by a different route in 1906 — so the Stokes form is properly the Sutherland–Einstein relation. Because $D$ and the viscosity $\eta$ are both directly measurable, the relation gave Jean Perrin an experimental route to Avogadro's number, furnishing decisive evidence for the atomic hypothesis and earning Perrin the 1926 Nobel Prize. Mathematically the relation can be reached from several equivalent starting points: balancing fluxes in sedimentation equilibrium (Einstein's argument), comparing the drift and noise terms of the Langevin equation, requiring the Boltzmann distribution to be the stationary solution of the Fokker–Planck equation, or integrating the velocity autocorrelation via the Green–Kubo formula. Each route exposes a different facet of why diffusion and mobility must be locked together at thermal equilibrium. As a near-equilibrium identity the Einstein relation underlies an enormous range of transport phenomena. In its generalized forms it governs ionic mobility in electrolytes (the Nernst–Einstein relation), carrier transport in semiconductors, rotational Brownian motion of molecules, and the response of colloids and polymers. It is the gateway to linear response theory, of which it is the canonical first example.

Key takeaways

The Einstein relation $D = \mu k_B T$ links the diffusion coefficient $D$ (spontaneous spreading) to the mobility $\mu$ (drift per unit force), with temperature $T$ as the proportionality constant.
It is the first and prototypical fluctuation–dissipation relation: the same molecular collisions that randomly kick a Brownian particle also produce the systematic friction that resists steady motion.
Einstein derived it in 1905 by balancing diffusive flux against drift flux in sedimentation equilibrium, requiring the barometric (Boltzmann) distribution to be stationary.
For a sphere of radius $a$ in a fluid of viscosity $\eta$ , combining with Stokes' law $\mu=1/(6\pi\eta a)$ gives the Stokes–Einstein relation $D=k_BT/(6\pi\eta a)$ , which let Perrin measure Avogadro's number.
The relation holds near equilibrium for any linear-response transport: ionic conductivity, electron mobility, rotational diffusion all obey an analogous $D=\mu k_BT$ once $\mu$ is defined consistently.
Key idea: diffusion and drag are two faces of the same thermal noise — a force $F$ produces drift $\mu F$ , and the very same molecular kicks that exert that friction also drive diffusion at rate $D$ , with the equilibrium constraint forcing $D=\mu k_BT$ .

The core insight

Diffusion and drag are two faces of the same thermal noise: a force $F$ produces drift velocity $\mu F$ , and the very same molecular kicks that exert that friction also drive diffusion at rate $D$ , with the equilibrium constraint forcing $D = \mu k_B T$ .

Where this lives today

The Einstein relation is a daily workhorse of transport physics. In semiconductor device modeling, electron and hole currents combine a drift term $nq\mu_E E$ and a diffusion term $qD\nabla n$ ; the relation $D=\mu_E k_BT/q$ ensures these are consistent, so that at equilibrium the Fermi level is flat and no net current flows. The thermal voltage $k_BT/q\approx25.7\,\mathrm{mV}$ appears in every diode and transistor equation. In electrochemistry and battery science, the Nernst–Einstein relation $\sigma=nq^2D/(k_BT)$ links measured ionic conductivity to self-diffusion, and its deviations (the Haven ratio) diagnose correlated ion motion in solid electrolytes. In soft-matter and biophysics, microrheology inverts the Stokes–Einstein relation: by measuring the diffusion coefficient of tracer beads or labeled proteins one extracts the local viscosity or hydrodynamic radius, while rotational diffusion measured by fluorescence anisotropy or NMR reports molecular size and binding state. As the zero-frequency limit of the full fluctuation–dissipation theorem, the Einstein relation is also the conceptual entry point to Linear Response Theory and, through it, to Kubo's formulas for all transport coefficients.

The physical picture

Imagine a colloidal particle in a beaker. With no applied force it wanders by diffusion, exploring a region of size $\sqrt{2Dt}$ as described in Brownian Motion. Now let gravity pull the particles down; they drift downward at speed $\mu F$ but pile up until the upward diffusive flux exactly cancels the downward drift. The height of the resulting atmosphere is set by $k_BT$ . For this equilibrium to be the Boltzmann distribution, $D$ and $\mu$ cannot be chosen independently — their ratio is forced to be $k_BT$ . The deeper message is economy: a single bath of jiggling molecules supplies both the random force (which spreads the particle, i.e. $D$ ) and the systematic drag (which resists pushing it, i.e. $1/\mu$ ). Because both come from the same molecular velocity distribution at temperature $T$ , knowing one determines the other. Strong friction means frequent, vigorous collisions — which also means strong random kicks — so a sluggish (low- $\mu$ ) particle is, per unit mobility, kicked just as hard, keeping $D/\mu$ fixed. The punchline is what the relation predicts for the temperature dependence. A purely mechanical picture of drift — push with force $F$ , drift at $\mu F$ , spread by some intrinsic $D$ — would let $D$ be temperature-independent. The Einstein relation says instead that $D$ rises linearly with $T$ through the origin, with slope $\mu k_B$ . Diffusion vanishes as $T\to0$ because the thermal kicks that drive it die away. That linear law, not a constant, is the experimental signature of a genuinely thermal origin.

Figure

D versus T: the Einstein relation predicts a straight line through the origin with slope μk_B, against the temperature-independent constant a purely mechanical picture would give

One can read the relation as a thermometer in disguise: $T = D/(\mu k_B)$ . Measuring how fast something spreads ( $D$ ) and how readily it responds to a force ( $\mu$ ) tells you the temperature of the bath, because equilibrium ties these macroscopic observables to the microscopic energy scale $k_BT$ . Dimensionally this is airtight: $D$ has units $\mathrm{m^2/s}$ and $\mu$ has units $\mathrm{(m/s)/N}=\mathrm{s/kg}$ , so $D/\mu$ has units of energy, matching $k_BT$ exactly.

Figure

Flux balance in sedimentation: downward drift μF cancels the upward diffusive flux −D dn∕dx, and matching to the barometric Boltzmann profile forces D = μk_BT

The core equations

At the heart of the topic is a single algebraic identity connecting a fluctuation quantity to a dissipation quantity through the thermal energy scale. It states that the diffusion coefficient and the mobility are proportional, with $k_BT$ the constant of proportionality.

D = \mu\, k_B T \qquad (1)

where: $D$ — diffusion coefficient, defined by $\langle[\Delta x(t)]^2\rangle=2Dt$ in one dimension; $\mu$ — mobility, the drift velocity per unit applied force, $v_{\mathrm{drift}}=\mu F$ ; $k_B$ — Boltzmann constant, $1.380649\times10^{-23}\,\mathrm{J/K}$ ; $T$ — absolute temperature of the surrounding bath. Physically: the left side measures spontaneous spreading, the right side measures forced response — equilibrium ties them. Writing the mobility as the inverse of a friction coefficient recasts the relation in the form that connects directly to the Langevin equation.

D = \frac{k_B T}{\gamma},\qquad \gamma = \frac{1}{\mu} \qquad (2)

where: $\gamma$ — friction (drag) coefficient, drag force $=\gamma v$ , units $\mathrm{kg/s}$ . This is exactly the diffusion coefficient that The Langevin Equation derives from the Green–Kubo integral of the velocity autocorrelation. For the specific case of a sphere in a viscous fluid, inserting Stokes' law for the friction yields the form Perrin used to weigh the atom.

D = \frac{k_B T}{6\pi\eta a} \qquad (3)

where: $\eta$ — dynamic viscosity of the fluid; $a$ — hydrodynamic radius of the spherical particle. This connects four independently measurable quantities $D$ , $\eta$ , $a$ , $T$ , allowing $k_B$ — and hence $N_A=R/k_B$ — to be extracted from a microscope. The relation is a statement about the long-time diffusive regime; before that, at times shorter than the velocity relaxation time, motion is ballistic.

\langle [x(t)-x(0)]^2\rangle = 2 d\, D\, t \quad (t\gg \tau) \qquad (4)

where: $d$ — spatial dimensionality (so $2Dt$ in 1D, $6Dt$ in 3D); $\tau=m/\gamma=m\mu$ — velocity relaxation time. Below $\tau$ the mean-square displacement grows as $t^2$ ; above it, as $t$ . The relation generalizes cleanly to charged carriers, where the mobility is usually defined per unit field rather than per unit force, giving the Nernst–Einstein relation for the conductivity.

\sigma = \frac{n q^2 D}{k_B T} = n q^2 \mu \qquad (5)

where: $\sigma$ — electrical conductivity; $n$ — number density of charge carriers; $q$ — charge per carrier. The thermal voltage $k_BT/q\approx25.7\,\mathrm{mV}$ at room temperature that this implies is ubiquitous in semiconductor device analysis.

Derivation AEinstein's sedimentation-equilibrium argument, 1905

This derivation balances the diffusive flux against the drift flux of particles settling under a constant force, and demands that the equilibrium density be the barometric (Boltzmann) distribution. It is Einstein's original 1905 route and the cleanest way to see why the applied force cancels out, leaving a property of the medium and temperature alone. Consider particles in a column subject to a constant downward force $F$ (gravity minus buoyancy); two opposing fluxes compete, and in steady state they must cancel everywhere.

Step 1. Write the drift flux: each particle drifts at velocity $\mu F$ , so $J_{\mathrm{drift}}=-n\mu F$ with $n(x)$ the number density. Why: mobility is defined as drift velocity per unit force.

Step 2. Write the diffusive flux (Fick's law): $J_{\mathrm{diff}}=-D\,\mathrm{d}n/\mathrm{d}x$ . Why: diffusion drives particles down the concentration gradient, the content of The Diffusion Equation.

Step 3. Impose steady state $J_{\mathrm{drift}}+J_{\mathrm{diff}}=0$ , giving $\mu F\,n=-D\,\mathrm{d}n/\mathrm{d}x$ . Why: in equilibrium the net particle flux vanishes at every height.

Step 4. Integrate to get the density profile $n(x)=n_0\,e^{-\mu F x/D}$ , an exponential atmosphere with decay length $D/(\mu F)$ .

Step 5. Require the equilibrium density to follow the Boltzmann factor for the potential $U=Fx$ : $n(x)\propto e^{-U/k_BT}=e^{-Fx/k_BT}$ , the barometric law from Boltzmann Distribution in External Potentials.

Step 6. Compare exponents: $\mu F/D=F/k_BT$ , so $F$ cancels and $D=\mu k_BT$ . Why: the relation is a property of the medium and temperature, independent of how hard you push. Verified in verify_aq-sm-einstein-relation.py: the profile $n_0 e^{-\mu Fx/D}$ makes the drift+diffusion flux vanish, and matching to $e^{-Fx/k_BT}$ returns $D=\mu k_BT$ .

Go deeper · Derivation B · 1 minThe Langevin and Green–Kubo routes to the same relation

Derivation BThe Langevin and Green–Kubo routes to the same relation

This derivation reaches $D=\mu k_BT$ from the dynamics rather than from equilibrium fluxes, first through the overdamped Langevin equation and then through the Green–Kubo integral of the velocity autocorrelation. The point is that three logically distinct arguments — flux balance, trajectory integration, and correlation integration — all land on the identical relation, demonstrating its robustness. We use only the thermally calibrated noise strength $2\gamma k_BT$ established in The Langevin Equation.

Step 1. In the overdamped limit inertia is negligible, so $\gamma\dot x=\xi(t)$ with $\langle\xi(t)\xi(t')\rangle=2\gamma k_BT\,\delta(t-t')$ . Why: the noise strength is fixed by the requirement that velocity satisfy equipartition.

Step 2. Integrate the displacement: $x(t)-x(0)=\gamma^{-1}\int_0^t\xi(s)\,\mathrm{d}s$ , then square and average using the delta correlation:

\langle[x(t)-x(0)]^2\rangle=\frac{1}{\gamma^2}(2\gamma k_BT)\,t=\frac{2k_BT}{\gamma}\,t.

Why: the delta collapses the double time integral to a single one of length $t$ .

Step 3. Compare with $\langle\Delta x^2\rangle=2Dt$ to read off $D=k_BT/\gamma=\mu k_BT$ .

Step 4. The Green–Kubo route: $D=\int_0^\infty\langle v(0)v(t)\rangle\,\mathrm{d}t$ with the Ornstein–Uhlenbeck correlation $\langle v(0)v(t)\rangle=(k_BT/m)e^{-\gamma t/m}$ . Integrating gives $D=(k_BT/m)(m/\gamma)=k_BT/\gamma=\mu k_BT$ . Why: diffusion is the time-integral of velocity correlations, and equipartition fixes the $t=0$ value. Verified in verify_aq-sm-einstein-relation.py: the overdamped MSD gives $2(k_BT/\gamma)t$ , and the Green–Kubo integral of the OU correlation returns $k_BT/\gamma=\mu k_BT$ .

Limiting cases

The relation's form depends on how mobility is defined and on the geometry of the problem:

Case	Result	Meaning
Stokes sphere (translational)	$D=\dfrac{k_BT}{6\pi\eta a}$	Rigid sphere radius $a$ , viscosity $\eta$ , no-slip; the Stokes–Einstein relation used by Perrin.
Charged carrier in field	$D=\dfrac{\mu_E k_BT}{q}$	With mobility $\mu_E$ defined by $v=\mu_E E$ for charge $q$ , the force is $qE$ ; the Nernst–Einstein relation for ions and semiconductor carriers.
Rotational diffusion (sphere)	$D_{\mathrm{rot}}=\dfrac{k_BT}{8\pi\eta a^3}$	Angular analogue with rotational friction $\zeta_{\mathrm{rot}}=8\pi\eta a^3$ , governing dielectric and NMR relaxation.
High- $T$ / overdamped core	$D=\mu k_BT=\dfrac{k_BT}{\gamma}$	With $\gamma=1/\mu$ ; valid whenever linear response and a well-defined bath temperature hold — the universal form.

Worked examples

Example 1 — Stokes–Einstein diffusion coefficient of a micron bead (the three-article closure).

Step 1. Take a sphere of radius $a=1\,\mu\mathrm{m}$ in water at $T=300\,\mathrm{K}$ , $\eta=10^{-3}\,\mathrm{Pa\,s}$ . Stokes' law gives mobility $\mu=1/(6\pi\eta a)=1/(1.885\times10^{-8})=5.31\times10^{7}\,\mathrm{s/kg}$ .

Step 2. The Einstein relation gives $D=\mu k_BT=(5.31\times10^{7})(1.380649\times10^{-23})(300)=\mathbf{2.197\times10^{-13}\ m^2/s}$ . This is the identical number obtained in Brownian Motion and The Langevin Equation — the sedimentation-balance derivation, the Green–Kubo integral, and the direct random-walk observation all agree, which is exactly what a genuine fluctuation–dissipation identity requires.

Step 3. Sanity check via the Stokes form directly: $D=k_BT/(6\pi\eta a)=(1.380649\times10^{-23})(300)/(1.885\times10^{-8})=2.197\times10^{-13}\,\mathrm{m^2/s}$ . Identical, confirming $\mu=1/\gamma$ with $\gamma=6\pi\eta a$ .

Example 2 — The thermal voltage for a semiconductor carrier.

Step 1. For an electron of charge $q=e=1.602\times10^{-19}\,\mathrm{C}$ the Einstein relation reads $D/\mu_E=k_BT/e$ .

Step 2. At $T=298\,\mathrm{K}$ : $k_BT/e=(1.380649\times10^{-23})(298)/(1.602\times10^{-19})=\mathbf{0.0257\ V}=25.7\,\mathrm{mV}$ . This "thermal voltage" sets the scale of diffusion potentials in pn-junctions and is the diode ideality voltage $V_T$ ubiquitous in transistor analysis — a direct, everyday consequence of $D=\mu k_BT$ .

Origins · 1 mineinstein relation and how this was discovered

einstein relation

1905 · Independently established by William Sutherland (1905), Albert Einstein (1905) and Marian Smoluchowski (1906)

The relation $D=\mu k_BT$ carries Einstein's name, but its history is a striking case of simultaneous discovery. Albert Einstein derived it in his March 1905 paper on the molecular-kinetic theory of Brownian motion (Annalen der Physik 17, 549), obtaining $D=RT/(6\pi\eta a N_A)$ and proposing it as a way to determine Avogadro's number. In the same year the Australian physicist William Sutherland published an essentially identical result in the Philosophical Magazine by the same flux-balance argument, so the Stokes form is often called the Sutherland–Einstein relation. In 1906 Marian Smoluchowski reached the same relation through a distinct kinetic calculation. The experimental confirmation came from Jean Perrin (1908–1913), whose measurements of colloidal diffusion yielded a consistent value of $N_A$ , clinching the reality of atoms and earning him the 1926 Nobel Prize in Physics.

History

1827

— Robert Brown observes the ceaseless motion of suspended particles

1905

— Einstein derives D = RT ∕ (6πηa N_A), the diffusion–mobility relation

1905

— Sutherland independently publishes the same Stokes–Einstein relation

1906

— Smoluchowski reaches the relation by an independent kinetic argument

1908

— Perrin begins measuring N_A from colloidal Brownian motion

1926

— Perrin awarded the Nobel Prize for the discontinuous structure of matter

Going further

References. Pathria & Beale, Statistical Mechanics, 3rd ed., Ch. 13 (Brownian motion) · Kardar, Statistical Physics of Particles, Ch. 9 · Reif, Fundamentals of Statistical and Thermal Physics, Ch. 15 · Sethna, Statistical Mechanics: Entropy, Order Parameters, and Complexity, Ch. 2 & 10 · Landau & Lifshitz, Statistical Physics Part 1, §124 · Einstein, Annalen der Physik 17, 549 (1905) · Sutherland, Philosophical Magazine (1905).

Provenance. The Einstein relation $D=\mu k_BT$ and Stokes–Einstein $D=k_BT/(6\pi\eta a)$ from Pathria & Beale Ch. 13, Kardar Ch. 9, Reif Ch. 15, Sethna Ch. 2 & 10, Landau–Lifshitz §124. A verifier should: (1) reproduce the sedimentation flux-balance argument and confirm force cancellation; (2) check the overdamped-Langevin MSD gives $2(k_BT/\gamma)t$ with noise strength $2\gamma k_BT$ ; (3) evaluate the Green–Kubo integral of the OU velocity correlation; (4) confirm $D=2.197\times10^{-13}\,\mathrm{m^2/s}$ for the worked bead and $k_BT/e\approx25.7\,\mathrm{mV}$ at 298 K with $k_B=1.380649\times10^{-23}\,\mathrm{J/K}$ . Stokes drag $6\pi\eta a$ and rotational drag $8\pi\eta a^3$ are standard hydrodynamic results.

Builds on: Brownian Motion · The Langevin Equation · The Diffusion Equation · Boltzmann Distribution in External Potentials. Leads to: The Fluctuation-Dissipation Theorem · The Fokker-Planck Equation · Linear Response Theory · The Ornstein-Uhlenbeck Process.

Is the Einstein relation an exact law or an approximation?

It is exact within linear-response, near-equilibrium theory — which is where "diffusion coefficient" and "mobility" are themselves well-defined. It assumes a single well-defined bath temperature and a linear relation between drift velocity and force (constant mobility). It breaks down far from equilibrium, at very high fields where the mobility becomes field-dependent (hot carriers in semiconductors), or when the bath is not thermal. Within its domain, though, it is a rigorous identity, not a fitting formula, and the same $k_BT$ appears no matter which of the four derivations you use.

Why is it called a fluctuation–dissipation relation?

Because $D$ measures a fluctuation — the spontaneous random spreading of a particle with no force applied — while the friction $\gamma=1/\mu$ measures a dissipation — the energy lost per unit velocity when you drive the particle. The relation $D\gamma=k_BT$ says these two are not independent; the strength of the spontaneous fluctuations is fixed by the dissipative response and the temperature. This is the zero-frequency, static special case of the general fluctuation–dissipation theorem, which relates the full frequency-dependent response function to the spectrum of equilibrium fluctuations.

How did Perrin actually get Avogadro's number from this?

He measured everything in $D=RT/(6\pi\eta a N_A)$ except $N_A$ . Using a microscope he tracked colloidal particles of carefully calibrated radius $a$ in fluids of known viscosity $\eta$ and temperature $T$ , and extracted $D$ from their mean-square displacement $\langle\Delta x^2\rangle=2Dt$ . With $R$ the (macroscopic) gas constant known, the only unknown was $N_A$ , which he could then solve for. His value agreed with independent determinations from radioactivity and blackbody radiation — the concordance across unrelated methods is what convinced the last skeptics that atoms are real.

What is the difference between mobility μ and the electric mobility μ_E?

Convention. The "mechanical" mobility $\mu$ in $D=\mu k_BT$ is drift velocity per unit force, units $\mathrm{s/kg}$ . The "electrical" mobility $\mu_E$ used in semiconductor physics is drift velocity per unit electric field, units $\mathrm{m^2/(V\,s)}$ . They are related by $\mu_E=q\mu$ for a carrier of charge $q$ , because the force is $F=qE$ . So the Einstein relation in electrical variables is $D=\mu_E k_BT/q$ . Keeping the two straight is the single most common source of factor-of- $q$ errors in transport calculations.

Does the relation hold for rotational diffusion too?

Yes, with the appropriate rotational friction. A sphere reorienting in a viscous fluid has rotational friction $\zeta_{\mathrm{rot}}=8\pi\eta a^3$ , and the rotational diffusion coefficient obeys the same Einstein form $D_{\mathrm{rot}}=k_BT/\zeta_{\mathrm{rot}}=k_BT/(8\pi\eta a^3)$ . This governs how fast a molecule loses memory of its orientation and is measured by dielectric relaxation, fluorescence anisotropy decay, and NMR. The strong $a^3$ dependence makes rotational diffusion a sensitive probe of molecular size and of binding events that change the effective radius.

Why does diffusion vanish at zero temperature?

Because the Einstein relation $D=\mu k_BT$ is linear in $T$ through the origin. Diffusion is driven by thermal fluctuations — the random molecular kicks — and as $T\to0$ those kicks die away, so there is no thermal spreading. The mobility $\mu$ (the response to a force) need not vanish, but the diffusive spreading does. This linear-through-origin behavior, rather than a temperature-independent constant, is the experimental fingerprint that the transport is thermally driven, and it is the punchline figure of this article.