The Brownian motion paper is usually presented as the one that proved atoms exist. That is roughly what it accomplished, but it badly misdescribes what Einstein was doing. He was not trying to settle the atomic question. He was working out what would have to be true, and measurable, if the kinetic theory he already believed were correct — and the fact that this delivered a verdict on atoms was almost incidental to the argument.

The Problem

In 1905 the atomic hypothesis was still contested. Boltzmann had built statistical mechanics on it and had spent decades defending it against a serious opposition — Mach, Ostwald, and the energeticists, who regarded atoms as a computational convenience rather than physical objects. Their objection was not unreasonable. Nobody had seen one, and thermodynamics worked fine without them.

The deeper issue was that kinetic theory's central claim seemed untestable in principle. If heat is molecular motion, molecules move at hundreds of meters per second. But they are far too small to see, and their collective effect on anything large enough to observe should average out to nothing. The theory made a claim about a scale that appeared permanently out of reach.

Background Science

Robert Brown had reported in 1827 that pollen grains suspended in water jitter continuously and erratically. The motion never stops, does not depend on the grain being organic, and had no accepted explanation seventy-five years later. It was a known curiosity, not a research frontier.

The obvious guess — that molecular collisions push the grain around — fails on inspection. A particle is struck from all sides by an enormous number of molecules every second, and those impacts very nearly cancel. The intuition that individual kicks visibly shove the grain around is wrong by many orders of magnitude.

What survives is the imbalance. With something like 10²⁰ collisions per second, the number arriving from the left and right differ slightly and randomly. Fluctuations scale as the square root of the number of events, so relative to the total they are vanishingly small — but they are not zero, and they never average away completely. Einstein's insight was that the residue is the entire phenomenon.

What He Did

Einstein reasoned that a suspended particle is thermodynamically no different from a solute molecule, just larger. It therefore contributes to osmotic pressure and diffuses under the same statistical rules.

He then balanced two descriptions of the same system. From the thermodynamic side, an osmotic pressure gradient drives particles from high to low concentration. From the mechanical side, that drift is opposed by viscous drag, given by Stokes' law for a sphere. Setting these against each other yields a diffusion coefficient in terms of measurable quantities: temperature, viscosity, particle radius, and Avogadro's number.

The critical step is what he does with it. Rather than predicting a velocity — which turns out to be meaningless here, since the trajectory is so jagged that measured speed depends on how finely you sample it — he predicts the mean squared displacement:

⟨x²⟩ = 2Dt

Displacement grows with the square root of time, not linearly. The particle wanders rather than travels. And rearranging gives Avogadro's number in terms of things a microscope and a stopwatch can measure.

Why It Mattered

This converted a metaphysical dispute into a measurement. Perrin carried it out between 1908 and 1911, tracking particles under a microscope, extracting Avogadro's number, and getting agreement with values from entirely unrelated methods. Ostwald conceded. The atomic debate ended not by argument but by arithmetic.

The square-root scaling also became foundational well beyond physics. The same relation governs diffusion in cells, polymer conformation, heat flow, and — through Bachelier, independently and five years earlier — the mathematics of financial markets.

Some Thoughts

Two things about this paper stay with me.

The first is Einstein's hedging. He opens by saying that the motion he predicts might be Brownian motion, but that his information is too imprecise to judge. He had derived a phenomenon and was not certain whether it had already been observed for eighty years. That is an odd position for a paper to take, and it tells you the work was theory-first: the prediction came from kinetic theory, not from staring at pollen.

The second is the choice of observable. Velocity is the natural quantity to reach for, and it is exactly the wrong one — anyone trying to measure the speed of a Brownian particle gets a different answer at every magnification. Einstein saw that the question was ill-posed and replaced it with one that was not. Recognizing that your instinctive measurement is meaningless, and knowing what to substitute, seems to me the harder and more transferable skill.

There is also a lesson in what made this work decisive when a century of direct argument had not. The atomic hypothesis became credible when someone found a case where its statistical consequences were visible at human scale. The evidence was never going to come from seeing an atom. It came from finding the one place where an enormous number of invisible things fails, just barely, to cancel out.