Translation · Brownian motion: from wandering to a measurable law

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

In this paper it is to be shown that, according to the molecular-kinetic theory of heat, bodies of microscopically visible size suspended in liquids must, as a consequence of the molecular motion of heat, carry out motions of such magnitude that these motions can easily be detected with the microscope. It is possible that the motions to be treated here are identical with the so-called “Brownian molecular motion”; the information about the latter that is available to me is, however, so imprecise that I could form no judgment on the matter.

If the motion to be treated here can actually be observed, together with the regularities to be expected for it, then classical thermodynamics can no longer be regarded as exactly valid even for spaces distinguishable under the microscope, and an exact determination of the true size of atoms is then possible. If, conversely, the prediction of this motion should prove incorrect, a weighty argument against the molecular-kinetic conception of heat would thereby be furnished.

§ 1. On the osmotic pressure to be ascribed to the suspended particles.

In the partial volume V∗V^* of a liquid of total volume VV, let zz gram-molecules of a non-electrolyte be dissolved. If the volume V∗V^* is separated from the pure solvent by a wall that is permeable to the solvent but not to the dissolved substance, then this wall is acted on by the so-called osmotic pressure, which, for sufficiently large values of V∗/zV^*/z, satisfies the equation: pV∗=RTz.p V^* = R T z.

If, on the other hand, instead of the dissolved substance, small suspended bodies are present in the partial volume V∗V^* of the liquid, bodies that likewise cannot pass through the wall permeable to the solvent, then according to the classical theory of thermodynamics one is not to expect — at least when gravity, which does not interest us here, is neglected — that a force acts on the wall; for on the usual view the “free energy” of the system appears to depend not on the position of the wall and of the suspended bodies, but only on the total masses and the qualities of the suspended substance, of the liquid and of the wall, and on pressure and temperature. It is true that the energy and entropy of the boundary surfaces (capillary forces) would also come into consideration in calculating the free energy; we can, however, disregard these, since in the changes of position of the wall and of the suspended bodies that are to be considered, changes in the size and nature of the surfaces of contact are supposed not to occur.

From the standpoint of the molecular-kinetic theory of heat, however, one arrives at a different view. According to this theory a dissolved molecule differs from a suspended body only in size, and one does not see why a number of suspended bodies should not correspond to the same osmotic pressure as the same number of dissolved molecules. One will have to assume that the suspended bodies, as a consequence of the molecular motion of the liquid, carry out an irregular motion in the liquid, even if a very slow one; if they are prevented by the wall from leaving the volume V∗V^*, they will exert forces on the wall, just as dissolved molecules do. If, therefore, nn suspended bodies are present in the volume V∗V^*, that is, n/V∗=νn/V^* = \nu in unit volume, and if neighboring ones among them are sufficiently far apart, then there will correspond to them an osmotic pressure pp of magnitude: p=RTV∗nN=RTN⋅ν,p = \frac{R T}{V^*} \frac{n}{N} = \frac{R T}{N} \cdot \nu, where NN denotes the number of actual molecules contained in one gram-molecule. In the next section it is to be shown that the molecular-kinetic theory of heat does in fact lead to this extended view of osmotic pressure.

§ 2. The osmotic pressure from the standpoint of the molecular-kinetic theory of heat.[1)]

If p1p2…plp_1 p_2 \ldots p_l are variables of state of a physical system that completely determine its instantaneous state (for example, the coordinates and velocity components of all the atoms of the system), and if the complete system of the equations of change of these variables of state is given in the form ∂pν∂t=φν(p1…pl)(ν=1,2…l)\frac{\partial p_\nu}{\partial t} = \varphi_\nu(p_1 \ldots p_l) \quad (\nu = 1, 2 \ldots l) where ∑∂φν∂pν=0\sum \frac{\partial \varphi_\nu}{\partial p_\nu} = 0, then the entropy of the system is given by the expression: S=EˉT+2ϰlg⁡∫e−E2ϰTdp1…dpl.S = \frac{\bar{E}}{T} + 2 \varkappa \lg \int e^{-\frac{E}{2 \varkappa T}} d p_1 \ldots d p_l. Here TT denotes the absolute temperature, Eˉ\bar{E} the energy of the system, and EE the energy as a function of the pνp_\nu. The integral is to be extended over all combinations of values of the pνp_\nu that are compatible with the conditions of the problem. ϰ\varkappa is connected with the constant NN mentioned above by the relation 2ϰN=R2 \varkappa N = R. For the free energy FF we therefore obtain: F=−RNTlg⁡∫e−ENRTdp1…dpl=−RTNlg⁡B.F = -\frac{R}{N} T \lg \int e^{-\frac{E N}{R T}} d p_1 \ldots d p_l = -\frac{R T}{N} \lg B.

We now imagine a liquid enclosed in the volume VV; in the partial volume V∗V^* of VV let there be nn dissolved molecules or suspended bodies, which are held in the volume V∗V^* by a semipermeable wall; the limits of integration of the integral BB that occurs in the expressions for SS and FF are thereby affected. Let the total volume of the dissolved molecules or suspended bodies be small compared with V∗V^*. Let this system be completely described, in the sense of the theory mentioned, by the variables of state p1…plp_1 \ldots p_l.

Even if the molecular picture were fixed down to every detail, the evaluation of the integral BB would still present such difficulties that an exact calculation of FF could hardly be contemplated. Here, however, we need only know how FF depends on the size of the volume V∗V^* in which all the dissolved molecules or suspended bodies (called “particles” for short in what follows) are contained.

We call x1,y1,z1x_1, y_1, z_1 the rectangular coordinates of the center of gravity of the first particle, x2,y2,z2x_2, y_2, z_2 those of the second, etc., xn,yn,znx_n, y_n, z_n those of the last particle, and we assign to the centers of gravity of the particles the infinitely small parallelepiped-shaped regions dx1dy1dz1,dx2dy2dz2…dxndyndznd x_1 d y_1 d z_1, d x_2 d y_2 d z_2 \ldots d x_n d y_n d z_n, all of which are to lie in V∗V^*. Let the value be sought of the integral occurring in the expression for FF, with the restriction that the centers of gravity of the particles lie in the regions just assigned to them. This integral can in any case be brought into the form dB=dx1dy1…dzn⋅Jd B = d x_1 d y_1 \ldots d z_n \cdot J where JJ is independent of dx1dy1d x_1 d y_1 etc., and of V∗V^*, that is, of the position of the semipermeable wall. But JJ is also independent of the particular choice of the positions of the regions for the centers of gravity, and of the value of V∗V^*, as is now to be shown. For let a second system of infinitely small regions for the centers of gravity of the particles be given, denoted by dx1′dy1′dz2′,dx2′dy2′dz2′…dxn′dyn′dzn′d x'_1 d y'_1 d z'_2, d x'_2 d y'_2 d z'_2 \ldots d x'_n d y'_n d z'_n, regions that are to differ from those originally given only in their position and not in their size, and that are likewise all to be contained in V∗V^*; then, analogously, dB′=dx1′dy1′…dzn′⋅J′,d B' = d x'_1 d y'_1 \ldots d z'_n \cdot J', where dx1dy1…dzn=dx1′dy1′…dzn′.d x_1 d y_1 \ldots d z_n = d x'_1 d y'_1 \ldots d z'_n. We therefore have: dBdB′=JJ′.\frac{d B}{d B'} = \frac{J}{J'}.

From the molecular theory of heat given in the papers cited, however, it can easily be inferred[1)] that dB/Bd B/B (respectively dB′/Bd B'/B) is equal to the probability that at an arbitrarily chosen instant the centers of gravity of the particles lie in the regions (dx1…dzn)(d x_1 \ldots d z_n) (respectively in the regions (dx1′…dzn′)(d x'_1 \ldots d z'_n)). Now if the motions of the individual particles are independent of one another (to a sufficient approximation), if the liquid is homogeneous, and if no forces act on the particles, then for regions of equal size the probabilities belonging to the two systems of regions must be equal to each other, so that: dBB=dB′B.\frac{d B}{B} = \frac{d B'}{B}. But from this equation and the one found last it follows that J=J′.J = J'.

It is thus proved that JJ depends neither on V∗V^* nor on x1,y1…znx_1, y_1 \ldots z_n. By integration one obtains B=∫Jdx1…dzn=JV∗nB = \int J d x_1 \ldots d z_n = J V^{*n} and from this F=−RTN{lg⁡J+nlg⁡V∗}F = -\frac{R T}{N} \{ \lg J + n \lg V^* \} and p=−∂F∂V∗=RTV∗nN=RTNν.p = -\frac{\partial F}{\partial V^*} = \frac{R T}{V^*} \frac{n}{N} = \frac{R T}{N} \nu.

This consideration shows that the existence of osmotic pressure is a consequence of the molecular-kinetic theory of heat, and that according to this theory dissolved molecules and suspended bodies present in equal numbers behave in exactly the same way with respect to osmotic pressure at great dilution.

§ 3. Theory of the diffusion of small suspended spheres.

Let suspended particles be distributed irregularly in a liquid. We shall investigate their state of dynamic equilibrium under the assumption that on the individual particles there acts a force KK that depends on position but not on time. For the sake of simplicity let it be assumed that the force everywhere has the direction of the XX-axis.

Let ν\nu be the number of suspended particles per unit volume; then in the case of thermodynamic equilibrium ν\nu is a function of xx such that, for an arbitrary virtual displacement δx\delta x of the suspended substance, the variation of the free energy vanishes. We therefore have: δF=δE−TδS=0.\delta F = \delta E - T \delta S = 0. Let it be assumed that the liquid has cross-section 1 perpendicular to the XX-axis and is bounded by the planes x=0x = 0 and x=lx = l. We then have: δE=−∫0lKνδxdx\delta E = -\int_0^l K \nu \delta x d x and δS=∫0lRνN∂δx∂xdx=−RN∫0l∂ν∂xδxdx.\delta S = \int_0^l R \frac{\nu}{N} \frac{\partial \delta x}{\partial x} d x = -\frac{R}{N} \int_0^l \frac{\partial \nu}{\partial x} \delta x d x. The equilibrium condition sought is therefore: −Kν+RTN∂ν∂x=0-K \nu + \frac{R T}{N} \frac{\partial \nu}{\partial x} = 0 or Kν−∂p∂x=0.K \nu - \frac{\partial p}{\partial x} = 0. The last equation states that the force KK is held in equilibrium by forces of osmotic pressure.

We use equation (1) to determine the diffusion coefficient of the suspended substance. We can regard the state of dynamic equilibrium just considered as the superposition of two processes running in opposite directions, namely
1. a motion of the suspended substance under the action of the force KK acting on each individual suspended particle,
2. a process of diffusion, which is to be regarded as a consequence of the irregular motions of the particles due to the molecular motion of heat.

If the suspended particles are spherical (sphere radius PP) and the liquid has the coefficient of friction kk, the force KK imparts to the individual particle the velocity[1)] K6πkP,\frac{K}{6 \pi k P}, and through the unit of cross-section there pass per unit time νK6πkP\frac{\nu K}{6 \pi k P} particles.

If, further, DD denotes the diffusion coefficient of the suspended substance and μ\mu the mass of a particle, then as a result of diffusion there pass through the unit of cross-section per unit time −D∂(μν)∂xGramm-D \frac{\partial (\mu \nu)}{\partial x} \quad \text{Gramm} or −D∂ν∂x-D \frac{\partial \nu}{\partial x} particles. Since dynamic equilibrium is to prevail, we must have: νK6πkP−D∂ν∂x=0.\frac{\nu K}{6 \pi k P} - D \frac{\partial \nu}{\partial x} = 0.

From the two conditions (1) and (2) found for dynamic equilibrium, the diffusion coefficient can be calculated. One obtains: D=RTN16πkP.D = \frac{R T}{N} \frac{1}{6 \pi k P}. The diffusion coefficient of the suspended substance therefore depends, apart from universal constants and the absolute temperature, only on the coefficient of friction of the liquid and on the size of the suspended particles.

§ 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 \Delta where ∫−∞+∞φ(Δ)dΔ=1\int_{-\infty}^{+\infty} \varphi(\Delta) d \Delta = 1 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)

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. 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}. 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.} 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} 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, and setting 1τ∫−∞+∞Δ22φ(Δ)dΔ=D\frac{1}{\tau} \int_{-\infty}^{+\infty} \frac{\Delta^2}{2} \varphi(\Delta) d \Delta = D 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}.

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 = n 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}}.

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}.

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}.

§ 5. Formula for the mean displacement of suspended particles. A new method for determining the true size of atoms.

In § 3 we found, for the diffusion coefficient DD of a substance suspended in a liquid in the form of small spheres of radius PP, the value: D=RTN16πkP.D = \frac{R T}{N} \frac{1}{6 \pi k P}. Further, in § 4 we found, for the mean value of the displacements of the particles in the direction of the XX-axis in the time tt: λx=2Dt.\lambda_x = \sqrt{2 D t}. By eliminating DD we obtain: λx=t⋅RTN13πkP.\lambda_x = \sqrt{t} \cdot \sqrt{\frac{R T}{N} \frac{1}{3 \pi k P}}. This equation shows how λx\lambda_x must depend on TT, kk and PP.

We shall calculate how large λx\lambda_x is for one second if NN is set equal to 6⋅10236 \cdot 10^{23} in accordance with the results of the kinetic theory of gases; let the liquid chosen be water at 17° C (k=1,35⋅10−2k = 1{,}35 \cdot 10^{-2}), and let the diameter of the particles be 0,001 mm. One obtains: λx=8⋅10−5 cm=0,8 Mikron.\lambda_x = 8 \cdot 10^{-5} \, \text{cm} = 0{,}8 \, \text{Mikron}. The mean displacement in 1 minute would therefore be about 6 microns.

Conversely, the relation found can be used to determine NN. One obtains: N=tλx2⋅RT3πkP.N = \frac{t}{\lambda_x^2} \cdot \frac{R T}{3 \pi k P}.

May a researcher soon succeed in deciding the question raised here, which is important for the theory of heat!

Bern, May 1905.

(Received 11 May 1905.)