Tip: Increase T or decrease η/r — the particles wander more vigorously. The MSD slope = 4D in 2D (Einstein relation).
Section 02
The Idea, Step by Step
Catch a sunbeam slanting through a window and watch the dust motes inside it. They never hold still — they jitter, drift, and zig-zag as if something keeps nudging them. Drop a bead of ink into a still glass of water and, even without stirring, the colour slowly creeps outward in every direction. Nobody is pushing. So what is?
The answer is the fluid itself. Water and air are made of countless tiny molecules in ceaseless motion, and they are constantly slamming into any speck floating among them. A speck big enough to see gets hit from all sides at once — but never perfectly evenly. At each instant a few more molecules happen to strike one side than the other, and that tiny imbalance shoves the speck a little. The shoves point in random directions and arrive millions of times a second, so the speck takes a stumbling, drunken walk. This is Brownian motion, named for botanist Robert Brown, who in 1827 saw pollen grains doing exactly this and could not explain it.
Here is the surprising part. Because each nudge is random, the speck is just as likely to step left as right, so on average it goes nowhere — its average position stays put. What grows is how far it has typically strayed. The key quantity is the diffusion coefficient $D$, which measures how fast that spreading happens. The rule is not "distance grows with time" but "distance-squared grows with time": $\langle r^2\rangle = 4Dt$ in the two-dimensional view here. A micron pollen grain in water has $D \approx 0.22\ \mu\text{m}^2/\text{s}$, so after $1$ second it has typically wandered only about $1\ \mu\text{m}$ from where it started. To stray ten times farther it needs not ten times but a hundred times as long. That square-root creep — $\sqrt{\langle r^2\rangle}\propto\sqrt{t}$ — is the signature of a random walk.
Einstein's 1905 breakthrough was to pin down $D$ from the fluid's properties. Hotter, more energetic molecules kick harder; thicker, more viscous fluids and bigger grains resist more. He found $D = \dfrac{k_B T}{6\pi\eta r}$: diffusion rises with temperature $T$ and falls with viscosity $\eta$ and grain radius $r$. The sliders map straight onto this — T raises $D$, while η and r lower it — and the simulation moves each particle by a random Gaussian step of size $\sqrt{2D\,\Delta t}$ per axis, exactly as the theory prescribes.
Try this in the sim above: (1) Switch to MSD Mode and watch the green measured curve hug the dashed Einstein line $\langle r^2\rangle = 4Dt$ — a straight line, not a curve. (2) Crank T up and slide η down, and the cloud of particles explodes outward as $D$ jumps. (3) Drag r to its largest value and the big grains barely budge — proof that $D \propto 1/r$, the reason small molecules diffuse far faster than visible specks.
Section 03
Equation Derivation
In 1905 Einstein derived the diffusion coefficient $D$ of a Brownian particle from kinetic theory and statistical mechanics. His prediction, verified by Perrin (Nobel 1926), gave the first direct estimate of Avogadro's number and definitive proof that atoms exist.
A particle of mass $m$ in a fluid feels viscous drag plus rapid random kicks from molecular collisions:
$$m\frac{d\mathbf v}{dt} = -\gamma\, \mathbf v + \boldsymbol\xi(t)$$
Here $\boldsymbol\xi(t)$ is a stochastic force satisfying $\langle\boldsymbol\xi(t)\rangle = 0$ and $\langle\xi_i(t)\xi_j(t')\rangle = 2\gamma k_B T\,\delta_{ij}\,\delta(t-t')$. The amplitude is fixed by the fluctuation-dissipation theorem.
Step 2 — Equipartition fixes the noise strength
In equilibrium, each translational degree of freedom has energy $\frac12 k_B T$:
$$\frac12 m \langle v_x^2\rangle = \frac12 k_B T \;\Rightarrow\; \langle v_x^2\rangle = \frac{k_B T}{m}$$
For a Langevin process, integrating gives the same result if the noise correlator $\langle\boldsymbol\xi(t)\boldsymbol\xi(t')\rangle$ has strength $2\gamma k_B T$ — establishing the FDT.
Step 3 — Mean-square displacement
Multiply the Langevin equation by $\mathbf{r}$ and average. Using $\mathbf{r}\cdot\mathbf{v} = \tfrac12 d|\mathbf r|^2/dt$ and equipartition, after one trick of solving an ODE for $\langle r^2\rangle$:
Combined with $\langle r^2\rangle = 2dDt$ — this is the Einstein-Smoluchowski relation. Perrin used it (with $D$ measured under a microscope) to extract $k_B$ — and hence Avogadro's number $N_A = R/k_B$ — getting $N_A \approx 6 \times 10^{23}$, exactly today's value.
Mapping to the simulation
• Slider T sets temperature → controls noise amplitude via $\sqrt{2 D\,\Delta t}$ per step.
• Slider η sets fluid viscosity → controls drag $\gamma$ and hence $D = k_B T/\gamma$.
• The MSD plot is computed live by averaging $|\mathbf r(t)|^2$ over all particles. The dotted line shows the Einstein prediction $\langle r^2\rangle = 4 D t$ (2D).
Reference: Reif — Fundamentals of Statistical and Thermal Physics, §15.6: "Brownian Motion"; Pathria — Statistical Mechanics, 3rd Ed., Ch. 13: "Stochastic Processes"; Einstein A. — "Über die von der molekularkinetischen Theorie der Wärme geforderte Bewegung von in ruhenden Flüssigkeiten suspendierten Teilchen", Ann. d. Physik 17, 549 (1905).
Section 04
Frequently Asked Questions
A free particle moving ballistically (no collisions) covers distance proportional to $t$, so $\langle r^2\rangle \propto t^2$. But a Brownian particle suffers many random collisions per second, with the direction of velocity randomised on the timescale $\tau = m/\gamma$. After many such direction changes, the path is a random walk — the central-limit theorem then gives Gaussian-distributed displacement with variance growing linearly in time. Steps add in squares, not in lengths, so $\langle r^2\rangle \propto t$. At very short times ($t \ll \tau$), the velocity hasn't been randomised yet and you do see $\langle r^2\rangle \propto t^2$ (ballistic regime).
Key: Ballistic at short times → diffusive at long times. The crossover occurs at $\tau \sim m/\gamma$, typically $\sim$nanoseconds for water.
Everywhere matter is suspended in a fluid: pollen grains in water (Brown's original 1827 observation), milk droplets in coffee, smoke particles in air, ink in water. In modern science: single-molecule biophysics (DNA tracking, optical tweezers), atomic-force microscopy thermal noise, drug delivery (drugs diffuse to target cells via Brownian motion), and stock prices (Bachelier 1900 — financial mathematics is built on Brownian motion). It is also the foundational mathematical model of diffusion processes — heat, ions through membranes, atoms in alloys.
Key: If atoms exist, Brownian motion must exist. Perrin's measurements of Brownian motion settled the centuries-old debate about whether atoms were real.
Each coloured dot is a Brownian particle (think pollen grain) suspended in fluid. At each timestep, every particle is displaced by a random Gaussian vector with variance $2D\,\Delta t$ per axis. The MSD graph plots $\langle r^2(t)\rangle$ averaged over all particles — at long times this is exactly linear with slope $4D$ (in 2D). The histogram tab shows the spatial distribution $P(x,t)$, which is a Gaussian whose width grows as $\sqrt{2Dt}$. The single-trace mode highlights one particle so you can see the characteristic jagged, fractal random-walk path.
Key: The dots' "wiggling" is statistically identical to what Brown saw under his microscope — randomness from invisible molecular collisions.
Because "average" doesn't mean "exact". At any instant the number of molecules hitting one side of the particle slightly exceeds the number on the other side — by a small fluctuation $\sim \sqrt{N}$ out of total $N$ molecules. With $N \sim 10^{20}$ collisions per second on a micron particle, even a tiny relative imbalance produces a noticeable kick. These imbalances arrive in random directions, summing to a random walk. The thermal energy budget per fluid molecule is $\sim k_B T$, and equipartition guarantees the suspended particle ends up with the same average kinetic energy.
Key: Brownian motion is fluctuation made visible — the particle is large enough to see, but small enough to feel statistical noise.
The drag force on a sphere in a viscous fluid (Stokes' law) scales linearly with radius: $F = 6\pi\eta r v$. So the drag coefficient $\gamma = 6\pi\eta r$ grows with $r$. Larger particles are slowed more by viscous drag, but the random force driving them is fixed by temperature and FDT. The net result is $D = k_B T/\gamma \propto 1/r$. This is why pollen grains (~30 μm) show faint Brownian motion while small molecules (~0.1 nm) diffuse rapidly — a 300,000× difference in $D$.
Key: Smaller objects diffuse faster. This is why intracellular signalling is fast — proteins are small enough to diffuse across a cell in milliseconds.
Because zooming in reveals more wiggles — the path has structure at every scale. Mathematically, a Brownian path is continuous everywhere but differentiable nowhere: the velocity at a sharp instant is undefined. Its trajectory has Hausdorff fractal dimension 2 in 2D and beyond — meaning the path is so wiggly it almost fills space. This non-smooth behaviour is essential to stochastic calculus (Itô integral) and modern probability theory. Mandelbrot used Brownian-related processes to describe coastlines, mountain ranges, and price fluctuations.
Key: Brownian paths are "rough at every scale" — a path of total length infinity in finite time.
Perrin (1908) observed gamboge resin spheres of known size in water under a microscope, recording their positions every 30 seconds. Plotting $\langle r^2\rangle$ versus $t$ gave a straight line, slope $= 4D$. Substituting into Einstein's formula $D = k_B T/(6\pi\eta r)$, he extracted $k_B$ — and hence $N_A = R/k_B$. His result, $N_A \approx 6 \times 10^{23}$, agreed with values from totally different methods (kinetic theory, electrolysis, X-ray diffraction). This convergence finally convinced sceptics like Ostwald that atoms are real, not just calculational devices. Perrin received the 1926 Nobel Prize.
Key: Brownian motion was the smoking gun for atoms — visible motion with magnitude predicted exactly by atomic theory.
Best Resource: Khan Academy — "Brownian motion"; HyperPhysics — "Brownian Motion"; MIT OCW 8.044 Statistical Physics I, Lecture 23; Berg — Random Walks in Biology, Princeton 1993 (intuitive standard).
Section 05
Common Misconceptions
❌ Brownian motion is caused by living organisms or chemical reactions.
✅ It's purely thermal — caused by molecular collisions of any fluid.
Brown himself first thought pollen grains moved because they were alive. He then tested with grains from long-dead plants, with mineral dust, and even with pulverised meteorite — all wiggled identically. Brownian motion is universal: any micron-sized particle in any fluid at any temperature above absolute zero exhibits it. The cause is the random thermal motion of the surrounding molecules pushing the particle from all sides.
📖 Mazo — Brownian Motion: Fluctuations, Dynamics, and Applications, Oxford, 2002, Ch. 1.
❌ Brownian particles have a typical "Brownian velocity".
✅ The instantaneous velocity is undefined in idealised theory; the displacement has a distribution.
If you measure displacement over time interval $\Delta t$ and divide, you get an "apparent velocity" that scales as $\sqrt{2D/\Delta t}$ — it diverges as $\Delta t \to 0$ in the idealised model. In real systems, on the very short timescale $\tau = m/\gamma$ the particle's velocity does follow the Maxwell-Boltzmann distribution with $\langle v^2\rangle = k_BT/m$. But this regime is usually inaccessible to optical microscopes — only ultrafast experiments (Li & Raizen 2010) finally measured the true ballistic velocity directly.
📖 Li T., Raizen M. — "Brownian motion at short time scales", Annalen der Physik 525, 281 (2013).
❌ Brownian motion violates the second law of thermodynamics — energy is created from nothing.
✅ It's a continuous energy exchange with the heat bath at fixed $T$.
A Brownian particle constantly exchanges kinetic energy with the surrounding fluid: it gets random kicks (energy in) and loses energy to viscous drag (energy out). On average, the two balance — the particle's mean kinetic energy stays at $\tfrac12 dk_B T$ (equipartition). The fluid maintains the temperature by being a thermal reservoir. No net work is extracted; trying to do so would require a Maxwell demon, ruled out by Landauer's principle (the demon's memory must eventually be erased, which dissipates exactly the energy gained).
📖 Reif — Statistical Physics, §15.7; Landauer R., IBM J. Res. Dev. 5, 183 (1961).
❌ Brownian motion is a quantum-mechanical effect.
✅ It's a classical statistical phenomenon, not quantum.
Brownian motion is fully described by classical statistical mechanics — Langevin equation, fluctuation-dissipation theorem, Stokes drag. There's no $\hbar$ in Einstein's formula. Quantum effects do appear at very low temperatures, very small particles, and very short timescales (for example, zero-point motion remains as $T \to 0$), but typical Brownian motion of micron-sized objects in liquid water at room temperature is purely classical.
📖 Reif — Statistical Physics, §15.6; Kubo R. — "The fluctuation-dissipation theorem", Rep. Prog. Phys. 29, 255 (1966).
❌ The trajectory of a Brownian particle has a definite length over a given time.
✅ Mathematically, every Brownian path has infinite arc length.
In the standard mathematical model, a Brownian path is continuous but nowhere differentiable. At every point the path "wiggles" at all scales — zoom in, you see more wiggles. Summing the path lengths of all these wiggles diverges. In practice, a real particle's path becomes smooth at scales below $\tau v_{th}$, but in the idealised Wiener process the arc length is mathematically infinite over any nonzero time interval.
📖 Mörters & Peres — Brownian Motion, Cambridge UP, 2010, Ch. 1; Mandelbrot — The Fractal Geometry of Nature, 1982.
❌ Brownian motion eventually stops because the particle "uses up" its energy.
✅ It never stops as long as the fluid is at $T > 0$.
The fluid acts as a thermal reservoir — every collision returns kinetic energy to the suspended particle. Equipartition guarantees $\langle KE\rangle = \tfrac12 d k_B T$ at all times, regardless of how long you wait. The motion only stops at $T = 0$ K (and even there, quantum zero-point fluctuations remain). What does happen with time is that the position uncertainty grows as $\sqrt{2dDt}$ — the particle wanders further from its starting point, but its kinetic energy stays constant on average.
Misconception research: Brookes & Etkina — "Force, ontology, and language", Phys. Rev. ST Phys. Educ. Res. 5, 010110 (2009); Loverude — "Student understanding of basic probability concepts in an upper-division thermal physics course", AIP Conf. Proc. 1179 (2009); Driver et al. — Making Sense of Secondary Science.