Translation

On the Motion of Particles Suspended in Liquids at Rest, as Required by the Molecular-Kinetic Theory of Heat

Über die von der molekularkinetischen Theorie der Wärme geforderte Bewegung von in ruhenden Flüssigkeiten suspendierten Teilchen

by A. Einstein

§ 4. On the irregular motion of particles suspended in a liquid and its relation to diffusion.

We now turn to a closer investigation of the irregular motions which, produced by the molecular motion of heat, give rise to the diffusion investigated in the last section.

It must evidently be assumed that each individual particle carries out a motion that is independent of the motion of all the other particles; the motions of one and the same particle in different time intervals are also to be regarded as mutually independent processes, as long as we think of these time intervals as not chosen too small.

We introduce into the discussion a time interval τ\tau, which is to be very small compared with the observable time intervals, but still so large that the motions carried out by a particle in two successive time intervals τ\tau are to be regarded as mutually independent events.

Now let a total of nn suspended particles be present in a liquid. In a time interval τ\tau the XX-coordinates of the individual particles will increase by Δ\Delta, where Δ\Delta has a different (positive or negative) value for each particle. A certain frequency law will hold for Δ\Delta; the number dnd n of particles that in the time interval τ\tau undergo a displacement lying between Δ\Delta and Δ+dΔ\Delta + d \Delta will be expressible by an equation of the form dn=nφ(Δ)dΔd n = n \varphi(\Delta) d \DeltaParticle countJump densityDisplacement in one intervalExplain this equation where ∫−∞+∞φ(Δ)dΔ=1\int_{-\infty}^{+\infty} \varphi(\Delta) d \Delta = 1Jump densityDisplacement in one intervalExplain this equation and φ\varphi differs from zero only for very small values of Δ\Delta and satisfies the condition φ(Δ)=φ(−Δ)\varphi(\Delta) = \varphi(-\Delta)Jump densityDisplacement in one intervalExplain this equation

We now investigate how the diffusion coefficient depends on φ\varphi, again restricting ourselves to the case in which the number ν\nu of particles per unit volume depends only on xx and tt.

Let ν=f(x,t)\nu = f(x, t) be the number of particles per unit volume; we calculate the distribution of the particles at the time t+τt + \tau from their distribution at the time tt. From the definition of the function φ(Δ)\varphi(\Delta) one easily obtains the number of particles that at the time t+τt + \tau are located between two planes perpendicular to the XX-axis with the abscissas xx and x+dxx + d x. One obtains: f(x,t+τ)dx=dx⋅∫Δ=−∞Δ=+∞f(x+Δ)φ(Δ)dΔ.f(x, t + \tau) d x = d x \cdot \int_{\Delta = -\infty}^{\Delta = +\infty} f(x + \Delta) \varphi(\Delta) d \Delta.Number densityTimeStep intervalDisplacement in one intervalJump densityExplain this equation But since τ\tau is very small, we can now set: f(x,t+τ)=f(x,t)+τ∂f∂t.f(x, t + \tau) = f(x, t) + \tau \frac{\partial f}{\partial t}.Number densityTimeStep intervalExplain this equation Further, we expand f(x+Δ,t)f(x + \Delta, t) in powers of Δ\Delta: f(x+Δ,t)=f(x,t)+Δ∂f(x,t)∂x+Δ22!∂2f(x,t)∂x2… in inf.f(x + \Delta, t) = f(x, t) + \Delta \frac{\partial f(x, t)}{\partial x} + \frac{\Delta^2}{2!} \frac{\partial^2 f(x, t)}{\partial x^2} \ldots \text{ in inf.}Number densityDisplacement in one intervalTimeExplain this equation We can carry out this expansion under the integral, since only very small values of Δ\Delta contribute anything to the latter. We obtain: f+∂f∂t⋅τ=f⋅∫−∞+∞φ(Δ)dΔ+∂f∂x∫−∞+∞Δφ(Δ)dΔ+∂2f∂x2∫−∞+∞Δ22φ(Δ)dΔ…\begin{aligned} f + \frac{\partial f}{\partial t} \cdot \tau = f \cdot \int_{-\infty}^{+\infty} \varphi(\Delta) d \Delta &+ \frac{\partial f}{\partial x} \int_{-\infty}^{+\infty} \Delta \varphi(\Delta) d \Delta \\ &+ \frac{\partial^2 f}{\partial x^2} \int_{-\infty}^{+\infty} \frac{\Delta^2}{2} \varphi(\Delta) d \Delta \ldots \end{aligned}Number densityTimeStep intervalJump densityDisplacement in one intervalExplain this equation On the right-hand side, because φ(x)=φ(−x)\varphi(x) = \varphi(-x), the second, fourth, etc. terms vanish, while of the first, third, fifth, etc. terms each successive one is very small compared with the one before. From this equation we obtain, taking into account that ∫−∞+∞φ(Δ)dΔ=1,\int_{-\infty}^{+\infty} \varphi(\Delta) d \Delta = 1,Jump densityDisplacement in one intervalExplain this equation and setting 1τ∫−∞+∞Δ22φ(Δ)dΔ=D\frac{1}{\tau} \int_{-\infty}^{+\infty} \frac{\Delta^2}{2} \varphi(\Delta) d \Delta = DStep intervalDisplacement in one intervalJump densityDiffusion coefficientExplain this equation and taking into account only the first and third terms on the right-hand side: ∂f∂t=D∂2f∂x2.\frac{\partial f}{\partial t} = D \frac{\partial^2 f}{\partial x^2}.Number densityTimeDiffusion coefficientExplain this equation

This is the well-known differential equation of diffusion, and one sees that DD is the diffusion coefficient.

Another important consideration can be attached to this development. We have assumed that the individual particles were all referred to the same coordinate system. This is not necessary, however, since the motions of the individual particles are independent of one another. We shall now refer the motion of each particle to a coordinate system whose origin coincides with the position of the center of gravity of the particle in question at the time t=0t = 0, with the difference that f(x,t)dxf(x, t) d x now denotes the number of particles whose XX-coordinate has increased from the time t=0t = 0 to the time t=tt = t by an amount lying between xx and x+dxx + d x. In this case too, then, the function ff changes according to equation (1). Further, for x≷0x \gtrless 0 and t=0t = 0 we must evidently have f(x,t)=0und∫−∞+∞f(x,t)dx=nf(x, t) = 0 \quad \text{und} \quad \int_{-\infty}^{+\infty} f(x, t) d x = nNumber densityTimeParticle countExplain this equation The problem, which coincides with the problem of diffusion from a single point (neglecting the interaction of the diffusing particles), is now mathematically completely determined; its solution is: f(x,t)=n4πDe−x24Dtt.f(x, t) = \frac{n}{\sqrt{4 \pi D}} \frac{e^{-\frac{x^2}{4 D t}}}{\sqrt{t}}.Number densityTimeParticle countDiffusion coefficientExplain this equation

The frequency distribution of the changes of position that have taken place in an arbitrary time tt is therefore the same as that of random errors, as was to be expected. Of importance, however, is how the constant in the exponent is connected with the diffusion coefficient. With the help of this equation we now calculate the displacement λx\lambda_x in the direction of the XX-axis that a particle undergoes on average, or — more precisely expressed — the square root of the arithmetic mean of the squares of the displacements in the direction of the XX-axis; it is: λx=x2‾=2Dt.\lambda_x = \sqrt{\overline{x^2}} = \sqrt{2 D t}.Coordinate RMS displacementMean-square displacement (one dimension)Diffusion coefficientObservation intervalExplain this equation

The mean displacement is therefore proportional to the square root of the time. It can easily be shown that the square root of the mean of the squares of the total displacements of the particles has the value λx3\lambda_x \sqrt{3}.

Bern, May 1905.

(Received 11 May 1905.)