Random Walks Statistical Mechanics

Random Walks in Statistical Mechanics describe the cumulative motion of a system that takes successive random steps. The central result is that the mean-squared displacement grows linearly with the number of steps,

⟨ R^{2} ⟩ = N a^{2}

, and that the displacement distribution converges to a Gaussian. They apply to diffusion, Brownian motion, polymer conformations, and lattice models. Foundational work by Pearson, Rayleigh, Einstein, and Smoluchowski (1900–1906).

A random walk is a stochastic process formed by summing independent random displacement vectors, modeling how microscopic randomness produces macroscopic transport. It matters because it is the simplest non-trivial model bridging probability theory and physical phenomena: diffusion of particles, heat conduction, and the shape of polymer chains all emerge from it. The key quantity is the displacement after $N$ steps, whose variance grows like $N$ rather than $N^{2}$ — randomness suppresses net progress. A surprising fact: in one and two dimensions a random walker returns to its origin with probability 1 (recurrence), but in three or more dimensions it may wander off forever (transience). This single model unifies the diffusion equation, the central limit theorem, and the statistical thermodynamics of fluctuating systems.

The Core Insight

Adding $N$ independent, identically distributed random steps produces a displacement whose typical magnitude scales as $N$ , not $N$ . The direction averages to zero, but the spread accumulates. This $N$ law is the signature of diffusive behavior and is a direct consequence of the additivity of variances for independent random variables.

Formal Statement

Discrete Random Walk

Let ${s_{i}}_{i = 1}^{N}$ be independent, identically distributed (i.i.d.) random step vectors in $d$ dimensions with $⟨ s_{i} ⟩ = 0$ and step length $∣ s_{i} ∣ = a$ . The position after $N$ steps is

R_{N} = i = 1 \sum N s_{i}

⟳The walk is defined as the running sum of independent displacements; this is the definitional starting point.

Symbols: $R_{N}$ = net displacement vector; $s_{i}$ = the $i$ -th step; $a$ = fixed step length; $N$ = number of steps; $d$ = spatial dimension.

Mean-Squared Displacement Law

For a random walk with zero-mean, uncorrelated steps of length $a$ , the mean displacement vanishes and the mean-squared displacement grows linearly in $N$ :

⟨ R_{N} ⟩ = 0, ⟨ R_{N}^{2} ⟩ = N a^{2}

⟳Expand

⟨ R_{N} \cdot R_{N} ⟩ = \sum_{i, j} ⟨ s_{i} \cdot s_{j} ⟩

; cross terms vanish by independence, diagonal terms each give

a^{2}

Gaussian Limit / Central Limit Theorem

For large $N$ , the probability density of the endpoint $R$ approaches a Gaussian:

P (R, N) = (\frac{d}{2 π N a ^{2}})^{d /2} exp (- \frac{d R ^{2}}{2 N a ^{2}})

⟳Apply the multivariate central limit theorem; the variance per dimension is

⟨ R^{2} ⟩ / d = N a^{2} / d

, fixing the normalization and exponent.

Why It Holds / Derivation

Mean-Squared Displacement

Step 1. Start from the displacement as a sum:

R_{N} = \sum_{i = 1}^{N} s_{i}

. Step 2. Take the expectation of the displacement. By linearity and

⟨ s_{i} ⟩ = 0

⟨ R_{N} ⟩ = i = 1 \sum N ⟨ s_{i} ⟩ = 0

⟳Expectation is linear; each isotropic step has zero mean by symmetry.

Step 3. Compute the squared magnitude:

R_{N}^{2} = R_{N} \cdot R_{N} = i = 1 \sum N j = 1 \sum N s_{i} \cdot s_{j}

⟳Dot the sum with itself and expand into a double sum.

Step 4. Average term by term. Split diagonal (

i = j

) and off-diagonal (

i \neq = j

⟨ R_{N}^{2} ⟩ = i = 1 \sum N ⟨ s_{i} \cdot s_{i} ⟩ + i \neq = j \sum ⟨ s_{i} \cdot s_{j} ⟩

⟳Separate the double sum; the second group contains all distinct-index pairs.

Step 5. Each diagonal term equals

a^{2}

(there are

N

of them). Each off-diagonal term factorizes by independence:

⟨ s_{i} \cdot s_{j} ⟩ = ⟨ s_{i} ⟩ \cdot ⟨ s_{j} ⟩ = 0

. Step 6. Therefore

⟨ R_{N}^{2} ⟩ = N a^{2}

. The root-mean-square displacement is

R_{rms} = a N

■

Continuum Diffusion Limit

Step 1. Let

τ

be the time per step and

a

the step length. After time

t = N τ

, the variance is

⟨ R^{2} ⟩ = (t / τ) a^{2}

. Step 2. Define the diffusion coefficient so the variance is finite as

a, τ \to 0

D = \frac{a ^{2}}{2 d τ}

⟳Demand

⟨ R^{2} ⟩ = 2 d D t

to hold; solve for

D

from

⟨ R^{2} ⟩ = (t / τ) a^{2}

Step 3. In this limit the discrete master equation for

P (R, t)

Taylor-expands to the diffusion equation:

\frac{\partial P}{\partial t} = D \nabla^{2} P

⟳Expand

P (R, t + τ)

and

P (R \pm a \overset{e}{^}, t)

to second order; the leading surviving terms give

\partial_{t} P = D \nabla^{2} P

Step 4. The Gaussian

[3]

(with

N a^{2} = 2 d D t

) is exactly the fundamental solution of

[8]

■

Albert Einstein | 1879–1955

In his 1905 paper Über die von der molekularkinetischen Theorie der Wärme geforderte Bewegung..., Einstein derived that the mean-squared displacement of a Brownian particle grows linearly in time, $⟨ x^{2} ⟩ = 2 D t$ , and connected $D$ to measurable quantities via $D = k_{B} T / (6 π η r)$ (the Stokes–Einstein relation). This gave Perrin a route to measuring Avogadro's number and provided decisive evidence for the atomic hypothesis. Einstein's treatment fused random-walk statistics with thermodynamics, making the random walk a cornerstone of physical kinetics.

Building Intuition

Imagine a drunkard taking steps of length $a$ in random directions. After $N$ steps where will they be? On average, nowhere — they're equally likely to drift any way, so $⟨ R ⟩ = 0$ . But "nowhere on average" does not mean "near the start." The spread of possible positions grows: after 100 steps the typical distance is $10 a$ , after 10,000 steps it is $100 a$ . The walker covers distance proportional to $N$ , far slower than the $N$ they would travel walking straight. The reason is cancellation: forward and backward steps partially undo each other, but only partially, because fluctuations don't fully cancel. The leftover imbalance scales like $N$ — exactly the standard deviation of a sum of independent coin flips.

Three random walk trajectories from a common origin, illustrating spread growing as √N

Worked Examples

Diffusion of a Pollen Grain

Given: A pollen grain in water at

T = 300 K

, radius

r = 1 μ m

, water viscosity

η = 1.0 \times 1 0^{- 3} Pa\cdots

. Find the rms displacement after

t = 10 s

(3D). Step 1. Stokes–Einstein diffusion coefficient:

D = \frac{k _{B} T}{6 π η r} = \frac{( 1.38 \times 1 0 ^{- 23} ) ( 300 )}{6 π ( 1.0 \times 1 0 ^{- 3} ) ( 1.0 \times 1 0 ^{- 6} )}

Step 2. Numerator

= 4.14 \times 1 0^{- 21}

; denominator

= 1.885 \times 1 0^{- 8}

D = 2.20 \times 1 0^{- 13} m^{2} / s

Step 3. In 3D,

⟨ R^{2} ⟩ = 6 D t = 6 (2.20 \times 1 0^{- 13}) (10) = 1.32 \times 1 0^{- 11} m^{2}

. Step 4.

R_{rms} = 1.32 \times 1 0^{- 11} = 3.63 \times 1 0^{- 6} m

R_{rms} \approx 3.6 μ m

End-to-End Distance of a Polymer

Given: An idealized freely-jointed polymer chain of

N = 10, 000

segments, each of length

a = 0.25 nm

. Find the rms end-to-end distance. Step 1. Model the chain as a 3D random walk; each segment is a step of length

a

. Step 2. Apply

⟨ R^{2} ⟩ = N a^{2}

⟨ R^{2} ⟩ = (10, 000) (0.25 \times 1 0^{- 9})^{2} = (10, 000) (6.25 \times 1 0^{- 20}) = 6.25 \times 1 0^{- 16} m^{2}

Step 3.

R_{rms} = 6.25 \times 1 0^{- 16} = 2.5 \times 1 0^{- 8} m

. Step 4. Compare to the contour length

L = N a = 2.5 μ m

. The coil is

100 \times

more compact.

R_{rms} = 25 nm

Common Misconceptions

"The walker drifts away at constant speed." No — net displacement scales as

N

, not

N

. The walker's average velocity of net progress actually decreases over time as

N / N = 1/ N

. Confusing total path length (

N a

) with net displacement (

a N

) is the single most common error. " $⟨ R ⟩ = 0$ means the walker stays put." The vector mean is zero by symmetry, but the magnitude

⟨ ∣ R ∣ ⟩

is nonzero and grows with

N

. Always distinguish the mean of the vector from the mean of the distance.

Self-Avoiding Walks Break the Simple Law

The clean result $⟨ R^{2} ⟩ = N a^{2}$ assumes steps are independent. A self-avoiding walk (SAW) — modeling a real polymer that cannot overlap itself — violates this. Excluded-volume interactions correlate steps, and the scaling changes to $⟨ R^{2} ⟩ \sim N^{2 ν}$ with the Flory exponent $ν \approx 0.588$ in 3D (not $ν = 1/2$ ). The chain swells beyond the ideal prediction. This shows the $N$ law is a consequence of independence, not a universal truth of all walks.

Connections

Random walks sit at the crossroads of several fields. The continuum limit yields the Diffusion Equation, identical in form to the heat equation. The Gaussian endpoint distribution is a direct manifestation of the Central Limit Theorem. The model underlies Brownian Motion as its continuous-time, continuous-space counterpart (the Wiener process). In polymer physics, the ideal chain connects to the Freely-Jointed Chain Model, while corrections lead to the Self-Avoiding Walk. The full probabilistic framework generalizes to Markov Chains, since each step depends only on the current state.

Real-World Applications

Brownian motion of colloids — Perrin's experiments (1908) used random-walk statistics to measure Avogadro's number, confirming atomic theory.
Polymer conformation — The radius of gyration of DNA, proteins, and synthetic polymers is predicted by random-walk scaling, central to gel electrophoresis and chromatography.
Diffusion in materials — Dopant diffusion in semiconductor manufacturing follows random-walk/diffusion-equation kinetics governing transistor fabrication.
Financial mathematics — The Black–Scholes model treats log-prices as a continuous random walk (geometric Brownian motion).
Population genetics — Genetic drift of allele frequencies is modeled as a random walk (the Wright–Fisher model).
Neutron transport — Neutrons scattering through a reactor moderator perform random walks; mean free path and diffusion length determine criticality.
Search algorithms — PageRank and Monte Carlo methods (e.g., Markov chain Monte Carlo) exploit random walks on graphs to sample distributions.

Historical Development

1827 — Robert Brown observes the erratic motion of pollen grains in water, later named Brownian motion.
1900 — Louis Bachelier, in Théorie de la spéculation, models stock prices as a random walk, anticipating diffusion theory.
1905 — Albert Einstein derives $⟨ x^{2} ⟩ = 2 D t$ and the Stokes–Einstein relation, linking Brownian motion to molecular kinetics.
1905 — Karl Pearson coins the term "random walk" in a letter to Nature, posing the drunkard's-walk problem.
1905 — Lord Rayleigh notes the connection to his 1880 work on the sum of random sinusoidal vibrations.
1906 — Marian Smoluchowski independently develops a statistical theory of Brownian motion.
1908 — Jean Perrin experimentally verifies Einstein's predictions, measuring Avogadro's number.
1921 — George Pólya proves recurrence in 1D/2D and transience in $d \geq 3$ ("a drunk man finds his way home, a drunk bird may not").
1923 — Norbert Wiener gives the rigorous mathematical construction of Brownian motion (the Wiener process).
1949 — Paul Flory introduces excluded-volume theory, giving the swollen scaling exponent for self-avoiding walks.

Subtleties & Limits

The connective constant $μ$ governing the number of $N$ -step self-avoiding walks ( $c_{N} \sim μ^{N} N^{γ - 1}$ ) is not known in closed form for the square lattice; only rigorous bounds and high-precision numerical estimates exist. The exact value of the Flory exponent $ν$ in three dimensions remains analytically open — its $d = 2$ value ( $ν = 3/4$ ) was conjectured by Nienhuis and connected to SLE (Schramm–Loewner Evolution) theory, but a fully rigorous derivation of SAW scaling in $d = 3$ is still missing. More broadly, anomalous diffusion (where $⟨ R^{2} ⟩ \sim t^{α}$ with $α \neq = 1$ ) in disordered and fractal media resists complete classification.