Section 43–6 Thermal conductivity

Idealizations / Approximations / Limitations

 

In this section, Feynman discusses the idealized conditions needed to derive the thermal conductivity of a gas, presents approximate formulas, and examines their limitations. The section could therefore be appropriately titled “Thermal conductivity of a gas.” Although we can use Fourier’s law of heat conduction for solids, liquids, and gases, the underlying mechanisms and assumptions differ among these states of matter. The distinction is particularly important for gases, where the behavior depends on the molecular mean free path. In essence, modeling the thermal conductivity of a dilute gas requires distinct frameworks depending on how the mean free path compares to the physical dimensions of the system.

 

1. Idealizations

“The transfer of heat from the hotter gas to the colder gas is by the diffusion of the “hot” molecules—those with more energy—downward and the diffusion of the “cold” molecules upward. To compute the flow of thermal energy we can ask about the energy carried downward across an element of area by the downward-moving molecules, and about the energy carried upward across the surface by the upward-moving molecules. The difference will give us the net downward flow of energy (Feynman et al, 1963).”

Feynman’s explanation of heat transfer in terms of the diffusion of “hot molecules” and “cold molecules” can be misleading if taken literally. Strictly speaking, heat conduction does not involve a net transport of mass: molecules move in all directions, and in a stationary medium the upward and downward molecular flux are equal. The terms “hot molecules” and “cold molecules” are also imprecise: temperature is a macroscopic statistical property of a local population of molecules, not a permanent property of individual molecules. Molecules possess molecules in both regions have a broad energy distribution, and thus, molecules crossing the plane in opposite directions do not necessarily belong to two sharply defined regions of “hot” and “cold” molecules. More importantly, heat conduction occurs through the transfer of energy during the collisions of molecules moving in opposite directions instead of simply a net diffusion of molecules or mass. From the perspective of conventional elementary theory of heat conduction, the macroscopic heat flux is associated with a temperature gradient, whereas the microscopic mechanism is the diffusion of thermal energy through molecular collisions.

 

Fourier’s law of heat conduction states that the heat flux Q, which is the flow of thermal energy per unit area and per unit time, is directly proportional to the negative temperature gradient: Q = -kA(dT/dx), where k is the thermal conductivity. The minus sign means that heat flows in the direction of decreasing temperature. In the context of a gas, this law is strictly valid only under a set of idealized conditions: (1) Homogeneous and Isotropic: The thermal conductivity is independent of spatial position (homogeneous) and independent of the direction of heat flow (isotropic). (2) Linear response: The temperature gradient must be sufficiently small so that the transport coefficient can be regarded as approximately constant. (3) Continuum or local-equilibrium: There should be no convectional current and the molecular mean free path must be shorter than the characteristic length of the system. When these conditions are satisfied, Fourier’s law provides a reliable description of conductive heat transport in gases.

 

2. Approximations

“Since the details of the calculations are quite similar to those we have done above in considering molecular diffusion, we shall leave it as an exercise for the reader to show that κ=knlv/(γ−1), where (γ−1)kT is the average energy of a molecule at the temperature T (Feynman et al., 1963).”

 

Deriving the Thermal Conductivity of a Dilute Gas

Feynman’s problem on the thermal conductivity of a dilute gas is closely analogous to his earlier derivation of the diffusion coefficient. Mathematically, the transport equations for mass transport and thermal energy share the same underlying structure. The primary difference lies in the physical quantity being transported: diffusion transport number density of molecules (or mass), whereas heat conduction transports energy.

 

The derivation can be summarized in four-steps:

Step 1: Relate Energy Flux to Molecular Transport

Consider an imaginary horizontal plane at position z. The net thermal energy flux q_z (energy per unit area per unit time) transported across this plane can be estimated as:

q_z = nv[E(z - l) - E(z + l)]

where n is the number density, v is the mean molecular speed, E(z) is the average energy per molecule at z, and l is the mean free path.

 

Step 2: Use Taylor’s Expansion

Assuming the average molecular energy varies slowly over a mean free path l, we can approximate E(z ± l) using a first-order Taylor expansion:

E(z ± l) » E(z) ± (l)(dE/dz)

Substitution gives the one-dimensional estimate:

q_z = (nv){[E(z) - (l)(dE/dz)] - [E(z) + (l)(dE/dZ)]} = -2(nvl)(dE/dZ)

A more complete three-dimensional treatment introduces a geometric factor would yield:

q_z = -(1/3)(nvl)(dE/dZ)

 

Step 3: Relate Energy Gradient to Temperature Gradient and Heat Capacity

The average molecular energy depends on the local temperature. Using the chain rule:

dE/dZ = (dE/dT)(dT/dZ) = c_v (dT/dZ),

where c_v is the thermal capacity per molecule at constant volume.

Substituting c_v into the heat flux equation gives:

q_z = -(1/3)(nvl)c_v(dT/dZ)

Comparing this with Fourier’s Law q_z = -k(dT/dZ), we identify

k = (1/3)nvlc_v

 

Step 4: Express the Molecular Heat Capacity in terms of Adiabatic Index

For an ideal gas, Mayer’s relation for a single molecule is:

c_p - c_v = k,  where k is Boltzmann’s constant.

Since g = c_p/c_v

Þ g - 1 = (c_p - c_v)/c_v = k/c_v

Þ c_v = k/(g - 1)

Substituting c_v into k = (1/3)(nvl)c_v, completes the derivation: k = (1/3)(nvl)k/(g - 1)

 

Feynman’s expression omits the geometric factor, so it is simply k = (nvl)k/(g - 1). This derivation is an order-of-magnitude estimate rather than a rigorous result. It captures the essential dependencies on number density, mean free path, and molecular speed, but it relies on Talor’s expansion and simplified assumptions (such as an ideal gas, a single mean molecular speed, and local equilibrium).

 

3. Limitations

“The formula (43.43) was derived, as were all the others in this chapter, under the assumption that the mean free path between collisions is much smaller than any of the dimensions of the container. Whenever the gas density is so low that a molecule has a fair chance of crossing from one wall of its container to the other without having a collision, none of the calculations of this chapter apply (Feynman et al., 1963).”

 

Limitations of Fourier’s Law: The Knudsen Regime

Fourier’s law applies to heat conduction in solids, liquids, and gases, provided that a well-defined local temperature field exists and the heat flux responds approximately locally and linearly to the temperature gradient. In gases, however, this continuum description has its limitations when the molecular mean free path (l) is comparable, or larger than the characteristic length scale (L) of the system. This breakdown is most pronounced in a Knudsen gas—a gas at such low density or confined to such a small space where l >> L. In this regime, intermolecular collisions are negligible, and the transport is governed almost entirely by collisions between molecules and the container walls. In short, Fourier’s law fails because heat transfer can no longer be represented by a local temperature gradient and thermal conductivity; instead, the shape and size of the container must be considered in detail. Thus, while Fourier’s law remains a good continuum approximation when l << L, it breaks down in the Knudsen regime, where heat transfer becomes a free-molecular, geometry-dependent process.

 

Thermal conductivity is commonly defined as a material’s ability to conduct thermal energy via lattice vibrations, molecular collisions or free-electron motion. However, the phrase “thermal conductivity of a dilute gas” can be misleading when a confined gas is at sufficiently low pressure to approach vacuum conditions. At ordinary dilute-gas pressures, a bulk thermal conductivity still provides a useful macroscopic description. But when the molecular mean free path becomes comparable to or exceeds the container size, the assumptions underlying a local, continuum description break down. In this regime, molecules tend to travel from one boundary to another with few or no molecular collisions, so heat transfer is governed primarily by molecule-wall collisions rather than molecular collisions. Thus, the process becomes geometry-dependent, requiring a different formula and geometrical factor. In the limiting near-vacuum case, heat transfer must be described by physical principles different from the conventional heat conduction used for solids and liquids.

 

Key Takeaways (This Section):

Universal Forms, Distinct Physical Realities: Thermal Conductivity

The mathematical form of a transport law can be broadly universal, while its physical realization and range of validity are not. Fourier’s law of heat conduction can describe heat transport in solids, liquids, and gases when its idealized assumptions are satisfied. However, the microscopic mechanisms that determine the thermal conductivity (k), as well as the conditions required for the equation to hold, differ among materials and physical regimes. In an ordinary gas, thermal conductivity arises from molecular motion and intermolecular collisions, with the mean free path determining the transport of thermal energy. In the Knudsen regime, where the mean free path becomes comparable to or larger than the system’s characteristic dimension, Fourier’s law breaks down, and the geometry of the system must be taken into account. Thus, the term “thermal conductivity” should not be taken to imply a single microscopic mechanism with a wide range of applicability; the same term does not imply the same underlying physics.

 

Key Takeaways (This Chapter):

Shared Mathematical Structure across Transport Phenomena

Ohm’s law of electrical conduction, Fick’s law of diffusion, and Fourier's law of heat conduction are examples of a broader class of linear transport laws that share the same mathematical structure:

[Flux] = [Transport Coefficient] ´ [Driving Force / Gradient].

Although the mathematical equations may appear different, each law describes the transport of a distinct physical quantity:

• Fick's law (mass transport): J = -D(dn/dx) or J = -DÑn

where n is particle number density and D is the diffusion coefficient.

• Fourier's law (thermal energy transport): q = -k(dT/dx) or q = -kÑT,

where T is temperature and k is thermal conductivity.

• Ohm's law (electrical charge transport): J = -s(dV/dx) or J = -sÑV

where V is electric potential and s is electrical conductivity.

The deeper lesson is that a common mathematical structure can describe radically different physical phenomena. What distinguishes these physical laws is not their mathematical equation alone, but the definitions of their variables—the quantity being transported, the driving force or gradient, and the transport coefficient. In short, we can use the same mathematical idea to model different physical phenomena, provided one interprets the symbols correctly.

 

The Moral of the Lesson: Maxwell’s Error and Scientific Humility

At the end of the audio recording of the lecture, Feynman recounts an anecdote about Maxwell that appears to serve a pedagogical purpose, but it was omitted from the edited text. After deriving electrical conductivity, diffusion, and thermal conductivity using simplified kinetic-theory arguments, Feynman was not simply making fun of Maxwell. Rather, the anecdote underscores an important lesson: having the right physical insight does not guarantee that every numerical result will be correct. Maxwell’s initial calculation of the ratio of the thermal conductivity of copper to that of air was wrong because of unit-conversion errors—specifically, failure to convert kilograms to pounds and hours to seconds. This is a striking reminder that even a great physicist can obtain a wildly incorrect numerical result through seemingly mundane mistakes. The story also illustrates scientific humility: Maxwell acknowledged his errors and recognized Clausius’s contribution to correcting his work.

 

There is, therefore, a broader moral to the chapter: physics is built not only from profound ideas, but also from rigorous attention to seemingly small details. Approximate reasoning may reveal the essential physics, but careful attention to the details is what turns physical insight into reliable quantitative knowledge. In science, there are many “little” things, but being meticulous about the little things is a very big thing.

 

Historical Note:

In his paper On the Dynamical Theory of Gases, Maxwell (1867) writes: “It is to Professor CLAUSIUS, of Zurich, that we owe the most complete dynamical theory of gases. His other researches on the general dynamical theory of heat are well known, and his memoirs “On the kind of Motion which we call Heat,” are a complete exposition of the molecular theory adopted in this paper. After reading his investigation of the distance described by each molecule between successive collisions, I published some propositions on the motions and collisions of perfectly elastic spheres, and deduced several properties of gases, especially the law of equivalent volumes, and the nature of gaseous friction. I also gave a theory of diffusion of gases, which I now know to be erroneous, and there were several errors in my theory of the conduction of heat in gases which M. CLAUSIUS has pointed out in an elaborate memoir on that subject.”

 

Clausius’s Comment on Maxwell’s Numerical Error

Feynman’s remark was based on a footnote (See below) in Rudolf Clausius’s (1862) paper on the thermal conductivity of gases.

Thus, the discrepancy in Maxwell’s numerical result arose from unit conversion mistakes instead of incorrect physical idea underlying his model of thermal conduction.

 

Review Questions

1. What idealized assumptions underlie the derivation of thermal conductivity for a dilute gas? In particular, explain why a well-defined local temperature, a sufficiently small temperature gradient, and a mean free path much smaller than the characteristic system length are each necessary for the derivation.

2. Feynman presents two approximate formulas for the thermal conductivity of a dilute gas—k = (nvl)k_B/(g - 1) and k = k_Bv/s_c(g - 1). What does each formula reveal about the microscopic physics of heat conduction? In particular, why does the second expression imply that thermal conductivity is independent of gas density, and under what conditions does that conclusion hold?

3. Under what conditions do these approximate formulas, and the Fourier-law description more generally, become unreliable for a dilute gas? Identify the relevant physical regimes—such as the Knudsen regime, high-density non-ideal gases, or convection-dominated transport—and explain what physical descriptions replace them.

 

References:

Clausius, R. J. E. (1862). Ueber die wärmeleitung gasförmiger körper. (On the Conduction of Heat in Gases). Annalen der Physik, 191(1), pp.1-56.

Feynman, R. P., Leighton, R. B., & Sands, M. (1963). The Feynman Lectures on Physics, Vol I: Mainly mechanics, radiation, and heat. Reading, MA: Addison-Wesley.

Fick, A. (1855). On Liquid Diffusion. Philosophical Magazine, Vol. 10, pp. 30–39.

Grattan-Guinness, I. (2005). Joseph Fourier, Théorie analytique de la chaleur (1822). In Landmark Writings in Western Mathematics 1640-1940 (pp. 354-365). Elsevier Science.

Maxwell, J. C. (1867). On the dynamical theory of gases. Philosophical transactions of the Royal Society of London, 157, 49-88.

Reif, F. (1965). Fundamentals of Statistical and Thermal Physics. McGraw-Hill.

Section 43–5 Molecular diffusion

Fick’s diffusion law / Einstein’s diffusion relation / Boltzmann distribution

 

In this section, Feynman derives Fick’s first law of diffusion, then proceeds with Einstein's diffusion relation, and shows why the Boltzmann distribution is essential for completing the derivation. However, the title “Molecular Diffusion” is overly narrow, describing only a physical mechanism. A more revealing title could be “Einstein’s Diffusion Relation: Balancing Drift and Diffusion,” which better captures the central concept and logical structure. The section’s deeper theme is not merely how molecules spread out, but the profound connection between random diffusive transport and force-induced drift—two seemingly distinct processes linked by the thermal energy. At thermal equilibrium, diffusion and drift are opposing transport tendencies whose balance reveals a fundamental relation between thermal fluctuations and dissipative response, sometimes known as the fluctuation-dissipation theorem.

 

1. Fick’s diffusion law

“In terms of n_a we can express the difference (n₊−n₋) as (n₊−n₋) = (dn_a/dx)Δx = (dn_a/dx)×2l (43.23). Substituting this result in Eq. (43.22) and neglecting the factor of 2, we get J_x=−lvdn_a/dx (43.24). We have found that the flow of special molecules is proportional to the derivative of the density, or to what is sometimes called the ‘gradient’ of the density (Feynman et al., 1963).”

 

Perhaps Feynman could have stated Fick’s first law of diffusion, which describes the net transport of particles through a medium. Specifically, diffusion is the spontaneous spreading of particles—typically in a liquid or gas—moving from regions of higher concentration to regions of lower concentration. In one dimension, Fick’s first law is expressed as J_x = -D(dn/dx), which means that the diffusive flux (J_x) is proportional to the magnitude of the concentration gradient (dn/dx) and the diffusion coefficient (D). The negative sign does not mean the gradient itself is negative; rather, it indicates that the flux moves down from regions of higher concentration to lower concentration. Strictly speaking, Fick’s first law is applied under steady-state conditions, where the concentration remains constant and the flux into and out of each volume element is equal.

 

Note: For simplicity, Feynman’s notation n_a is replaced by n. Some authors prefer the symbol c to represent concentration.

 

A 3-step Derivation of Fick’s First Law of Diffusion

Step 1: Relate Particle Flux to Particle Densities

Consider an imaginary plane perpendicular to the x-axis. The net diffusive flux J_x across the plane is the difference between the particles moving from the left and the right:

J_x = (n_- - n₊)v

where n_- and n₊ are the particle number densities on the left and right sides of the plane, respectively, and v is the average molecular speed. In general, particle flux is defined as the product of the number density and the velocity component normal to the surface.

 

Step 2: Evaluate the Densities One Mean Free Path Away

A molecule crossing the plane typically made its last collision, on average, one mean free path l away. Thus, n₋ and n₊ correspond to the number densities at x - l and x + l, respectively.

Assuming a smooth density gradient and using Taylor’s expansion, the density difference is:

n_- - n₊ » -2l (dn/dx)

Substituting this approximation into the particle flux expression gives:

J_x » - lv(dn/dx)

 

Step 3: Refine the Geometry and Identify the Diffusion Coefficient

The above estimate assumes that all molecules move perpendicular to the plane. In reality, molecules travel in random directions in three-dimensional space. A more rigorous treatment requires a geometric factor of 1/3 for the particle flux moving along the x-axis:

J_x » -(lv/3)(dn/dx)

Comparing this expression with Fick’s first law J_x » -D(dn/dx) allows us to identify the diffusion coefficient: D = lv/3.

 

“All of these refinements can be made; the result of a more careful analysis shows that the right-hand side of Eq. (43.24) should be multiplied by 1/3. So a better answer is J_x = −(lv/3)dn_a/dx (Feynman et al, 1963).”

Feynman briefly notes that a more careful analysis introduces a geometric factor of 1/3, giving the diffusion coefficient D = lv/3. This factor arises from the isotropic three-dimensional molecular motion: only the component of a molecule’s motion along the concentration gradient contributes to the net diffusive flux. However, this is based on a simplified model because it assumes a single molecular speed and isotropic spreading.

In physical systems, more rigorous frameworks, such as the Chapman–Enskog theory, yield transport coefficients that depends on the temperature and complex intermolecular potentials. For example, the Enskog-type corrections refine the diffusion coefficients as:

where D₀ is a constant, n is the number density, g(s) is the radial distribution function evaluated at the molecular contact diameter (s), and W(T) is a collision integral that characterizes the temperature-dependent interaction potential.

 

Note: In the standard elementary derivation, the one-way particle flux crossing a reference plane in the +x direction is written as nv/6 instead of nv/3 (Reif, 1965), because only one-sixth of the molecules—those with velocity components in the +x direction—contribute equally in the three orthogonal dimensions (±x, ±y, ±z). The six faces of a die serve as a useful analogy for these six equally probable directions.

 

2. Einstein’s diffusion relation

“We find that D, the diffusion coefficient, is just kT times μ, the mobility coefficient: D=μkT (43.31) (Feynman et al, 1963).”

 

A Condensed 3-Step Derivation of Einstein’s Diffusion Relation

This section could be confusing because Feynman uses nearly 20 equations to develop Fick’s first law of diffusion and then derive Einstein’s diffusion relation. To make this development more accessible, the derivation is reduced to three crucial steps:

 

Step 1. Two Opposing Fluxes

Particle motion in this context arises from two competing mechanisms:

Diffusion flux (J_diffusion): Random thermal motion causes particles to spread from regions of higher concentration to lower concentration. According to Fick’s First Law of Diffusion, the particle flux is proportional to the density gradient:

J_diffusion = -D(dn/dx)

Drift flux (J_drift): An external force (e.g., gravity) causes a systematic drift velocity v_d = mF, where m is generalized mobility. The resulting particle flux is:

J_drift = nv_d = nmF

 

Step 2. Zero Net Flux

At thermodynamic equilibrium, the diffusive and drift fluxes exactly balance:

J_diffusion + J_drift = 0

Substituting the expressions for both fluxes gives:

-D(dn/dx) + nmF = 0 Þ D(dn/dx) = nmF

The balance equation links the diffusion coefficient D to the mobility m and density gradient.

 

Step 3. Apply the Boltzmann distribution

At thermal equilibrium, the particle density follows the Boltzmann distribution:

n = n₀e^-U/kT

Differentiating this expression with respect to x and using F = -dU/dx gives

dn/dx = nF/kT (see equation 43.38)

By comparing the two expressions for dn/dx, we have nmF/D = nF/kT

By canceling out the density and force, we get Einstein’s diffusion relation:

D = mkT

 

Note: By balancing drift and diffusion, Feynman demonstrates that mobility (m) and diffusion (D) are not independent properties—they are two sides of the same coin, linked directly by thermal energy (kT).

 

“We now adjust the force F so that the drift current due to F just balances the diffusion, so that there is no net flow of our special molecules. We have J_x+J_drift = 0, or Ddn_a/dx = n_aμF (43.35) (Feynman et al., 1963).”

 

Reconciling Feynman’s Remark: Does Diffusion Stop at Equilibrium?

In the audio recording (48 min: 45 sec) of his lecture, Feynman remarks that it is possible to adjust a force so that there is “no diffusion” to achieve equilibrium—an intriguing comment that was omitted from the published text. This seems paradoxical because Feynman also states: “The current J_x (=J_diffusion) and the density gradient dn_a/dx can be measured by macroscopic observations.” However, the apparent paradox could be resolved by distinguishing between two different definitions of diffusion.

 

1. The Operational view: “No Diffusion”

The “no diffusion” view is defensible if diffusion is defined operationally as the observable net spreading or redistribution of particles. At thermodynamic equilibrium, the particle-density distribution is stationary and the net particle flux vanishes (J_net = 0). There is therefore no observable net transport or sustained redistribution of particles. In principle, this condition can be verified experimentally by monitoring the number of particles crossing a specified surface—e.g., counting particles in successive microscope images. In this operational sense, one may say that there is no diffusion at equilibrium because there is no observable net transport from one region to another.

 

2. The Transport-Theory view: “Diffusion Persists”

The “diffusion persists” view focuses on the individual transport mechanisms that contribute to the observed flux. Here, diffusion is defined as the spreading of particles driven by random thermal motion, where a density gradient supports a diffusive flux J_diffusion = -D(dn/dx). At thermodynamic equilibrium, the diffusion flux is exactly opposed by a force-induced drift flux J_drift = nmF, such that J_diffusion + J_drift = 0. Thus, the absence of net transport does not imply that each transport contribution has vanished. This is analogous to a book resting on a table: a downward gravitational force and an upward normal force continue to act even though the net force is zero. This view analyzes the underlying transport mechanisms and leads directly to Einstein’s diffusion relation.

 

Bridging the Two perspectives: Measurement vs. Mechanism

The apparent disagreement can be framed as a philosophical distinction between an operational view and a mechanistic view of diffusion—a distinction rooted in the difference between what is measured and how the measurement is interpreted. The operational view defines diffusion in terms of its observable macroscopic consequences, such as net particle flux or redistribution. The mechanistic view focuses on the underlying physical process: continuous random thermal motion. Although Feynman explains that diffusion can be measured macroscopically via the density gradient, an experiment can only measure the net flux—it cannot directly isolate the “diffusive” and “drift” fluxes. Thus, Feynman’s explanation reveals a subtle ambiguity in the word diffusion: the experimentally measured quantity is the net flux, while the transport theory decomposes that flux into diffusive and drift components.

 

Expanding the Definition: Normal versus Anomalous Diffusion

There is an additional complication when we are considering the precise definition of diffusion in the context of Brownian motion. The classical theory describes a transport regime where the mean-square displacement scales linearly with time, which underlies the familiar Einstein relation. More generally, if the mean-square displacement scales as <r²> µ t^a, then a = 1 corresponds to normal diffusion, whereas a ¹ 1 indicates anomalous diffusion (e.g., subdiffusion for a < 1, superdiffusion for a > 1). In  the anomalous case, a constant diffusion coefficient defined via the Einstein relation is no longer adequate; the effective diffusivity may depend on time or length scale, reflecting the complex underlying stochastic process. One might argue that Feynman could have problematized the definition of diffusion explicitly, or critiqued contemporary definitions of diffusion.

Source: Classification of stochastic processes by convolutional neural networks - IOPscience

 

Conclusion: A Semantic Distinction with Physical Significance

The apparent conflict is partly semantic because it depends on how one defines diffusion. However, the physical distinction is important: macroscopically, there is no observable diffusion as net spreading because the diffusive and drift fluxes exactly balance; microscopically, random thermal motion persists, and a diffusive flux contribution can still be mathematically identified. Thus, we may conclude that diffusion persists as an underlying microscopic tendency, while there is no macroscopic net redistribution of particles at thermodynamic equilibrium. Feynman’s phrase “no diffusion” is therefore best understood as referring to the absence of observable net transport, rather than the disappearance of microscopic random molecular motion.

 

Note: For clarity, Feynman’s equation J_x + J_drift = 0 is rewritten here as J_diffusion + J_drift = 0. Because this equilibrium balance is central to the derivation, it would be reasonable to assign it a numbered equation.

 

3. Boltzmann distribution

“[This is just exactly Eq. (40.2), from which we deduced e^−U/kT in the first place, so we have come in a circle] (Feynman et al., 1963).”

Addressing the “Circular Logic” Critique

The editors’ inserted comment—“…we have come in a circle”—might lead some readers (e.g., philosophers) to suspect that Feynman’s derivation is logically circular, as if he assumed what he set out to prove. A more appropriate interpretation, however, is that the argument functions as a consistency check: it shows that the Boltzmann distribution is compatible with the dynamical balance between diffusive and drift fluxes at thermodynamical equilibrium. Furthermore, the Boltzmann distribution is not a fixed, universal formula; its precise form depends on the physical constraints and boundary conditions of a given system. Historically, Einstein did not anchor his argument explicitly in terms of Boltzmann distribution. Instead, he referred to distribution of random errors, linking Brownian motion to the statistics of random walk. In Feynman’s derivation, the Boltzmann distribution is valuable not merely as an exponential function that simplifies the algebra, but because it connects the microscopic nature of thermal motion with the macroscopic equilibrium states that emerge from that randomness.

 

“We have shown that Eq. (43.31), which gives the diffusion current in terms of the mobility, has the correct coefficient and is very generally true. Mobility and diffusion are intimately connected. This relation was first deduced by Einstein (Feynman et al., 1963).”

 

In his 1905 paper on Brownian motion, Einstein did not express the diffusion relation in the modern form D = mkT. He did not formulate his theory in terms of a generalized “mobility” coefficient (m)—he used the symbol m for the mass of the suspended particle. Instead, Einstein derived the relation by balancing the osmotic force against the Stokes drag force on spherical particles suspended in a liquid, obtaining D = (RT/N)(1/6pkP). In this equation, R is the gas constant, N is the Avogadro’s constant, P is the radius of the suspended particle, and k is the coefficient of viscosity of the liquid—not the Boltzmann constant. According to Stokes’ law, the drag force is given by F = 6pkPv_d. From this, the mechanical mobility (m) can be identified as the inverse of the drag coefficient: m = v_d/F = 1/6pkP. Modern physics textbooks (including Feynman’s lectures) substitute the mobility coefficient m and Boltzmann’s constant (k = R/N) to rewrite Einstein’s relation in its generalized form D = mkT. Thus, the modern form is best considered not as the equation Einstein originally wrote, but as a reformulation of the relation he deduced.

 

The Complex History of the Diffusion Relation

William Sutherland had already presented his results on the diffusion coefficient in June 1904 at the meeting of the Australian Association for the Advancement of Science held in Dunedin, New Zealand. Independently, Marian Smoluchowski arrived at a similar diffusion relation around the same time, employing a random walk framework. Einstein subsequently published his derivation in 1905 using a different but closely related theoretical approach. Thus, the diffusion relation has a more complex history than the familiar label “Einstein relation” suggests. To properly acknowledge these independent contributions, it is sometimes referred to as the Sutherland-Einstein relation or Smoluchowski-Einstein relation.

 

Key Takeaways:

1. Fick’s First Law of Diffusion: Diffusion originates from random thermal motion and molecular collisions, rather than being caused by the density gradient (which determines the direction and magnitude of net flow). According to Fick’s diffusion law, the diffusion flux is directly proportional to the magnitude of density gradient and the diffusion coefficient. Furthermore, the geometrical factor of 1/3 arises from three-dimensional angular averaging, rather than serving as an arbitrary numerical correction.

2. Einstein’s diffusion relation: Thermodynamic equilibrium is a dynamic steady-state, not a static cessation of microscopic motion. Einstein’s diffusion relation can be derived by establishing an equilibrium condition, where the diffusive flux and the drift flux cancel exactly. This directly links the diffusion coefficient to the generalized mobility and thermal energy.

3. We need not simply criticize Feynman for saying that it is possible to adjust a force so that there is “no diffusion” and thereby achieve equilibrium. The more useful approach is to distinguish between two levels of descriptions. From an operational perspective, diffusion may be measured by observable net transport or spreading. From a mechanistic perspective, random thermal morion continues indefinitely. Thus, the apparent ambiguity is a matter of how diffusion is defined and at what level it is being measured, rather than a contradiction in the underlying physics.

 

The Moral of the Lesson:

Einstein’s diffusion relation provides a useful framework for understanding how molecular size, mobility, thermal motion, and transport properties are interconnected. Although this relation is not a routine quality-control equation used in food manufacturing, its underlying principles help explain how molecular transport changes during processes such as food fermentation and how such changes can be investigated using appropriate analytical methods.

 

1. Yogurt Production and Protein Mobility

During yogurt fermentation, starter cultures (e.g., Lactobacillus bulgaricus) acidify the milk and contribute to the proteolysis of milk proteins, producing smaller peptides and free amino acids. Based on Einstein’s relation, smaller solutes generally diffuse faster because their hydrodynamic drag is lower. This provides a useful conceptual link between fermentation-induced proteolysis and molecular transport.

 

2. Greek Yogurt vs. Greek-Style Yogurt

Traditional Greek yogurt is produced by fermenting milk and straining away much of the whey—a process that concentrates the milk solids and proteins to create a naturally thick texture. However, the term “Greek-style” does not inherently imply an inferior yogurt. Greek-style yogurts can achieve a comparable texture and nutritional profile using milk-protein concentrates, milk solids, starches, gelatin, or stabilizers, and many have an excellent nutritional profile. A more meaningful comparison relies on the nutrition and ingredient label: protein, carbohydrates, sugars, fat, sodium, and calcium content. For a simple choice, plain, unsweetened Greek yogurt with a relatively high protein content and minimal additives remains a reasonable option.

 

3. Kefir versus Yogurt

Kefir is produced using kefir grains, which contains a more diverse strains of bacteria and yeasts than typical yogurt. However, claiming that kefir is better for gut health is an oversimplification. The health effects of fermented foods depend on the specific product, microbial strains, individual physiology, and overall diet (including prebiotics). Greek yogurt is advantageous when a concentrated, high-protein food with a thick texture is preferred, whereas kefir provides a drinkable fermented food with a more diverse  microbial community. They are best viewed as complementary rather than competing probiotics.

 

4. Smoluchowski and Transport Theory

Einstein’s and Smoluchowski’s treatments of Brownian motion were closely related but differed in their mathematical formulations and assumptions. Smoluchowski, a professor of physics at the University of Lemberg (now Lviv, Ukraine), died in 1917 during a dysentery epidemic. Dysentery is an intestinal infection causing severe diarrhea (often containing blood or mucus), abdominal cramps, and fever. The primary measures for preventing such infection are safe food preparation, clean drinking water, proper sanitation, and good hygiene. Modern probiotics such as yogurt and kefir contribute to a healthy dietary pattern and support a balanced gut microbiome, but they should not be regarded as substitutes for basic hygiene, sanitation, or medical care.

 

5. A Notable Quotation

While Einstein developed his theory of translational Brownian motion in 1905, he expanded his theory in 1906 to include rotational Brownian motion. Although Einstein initially doubted the feasibility of measuring rotational Brownian motion, his analysis showed that thermal fluctuations should produce not only translational random motion, but also the rotational random motion of suspended particles. Jean Baptiste Perrin later experimentally verified Einstein’s theory of Brownian motion, helping to prove the physical reality of atoms.

In his Nobel Lecture, Perrin reflected on a broader philosophical lesson between the observable and the invisible: “Lastly, and doubtless always, but particularly at the end of the last century, certain scholars considered that since the appearances on our scale were finally the only important ones for us, there was no point in seeking what might exist in an inaccessible domain. I find it very difficult to understand this point of view since what is inaccessible today may become accessible tomorrow (as has happened by the invention of the microscope), and also because coherent assumptions on what is still invisible may increase our understanding of the visible.” Perrin’s quotation provides an especially fitting conclusion to the discussion of diffusion, Brownian motion, and transport theory.

 

Review Questions:

1. Why does the kinetic-theory expression for the diffusion coefficient contain the factor 1/3? Explain why it should not be interpreted as meaning that only one-third of the molecules participate in diffusion.

2. What does diffusion mean in the context of Einstein’s relation? Explain whether a diffusive flux can be present while the net flux is zero at equilibrium when it is exactly balanced by an opposing drift flux.

3. Why does Feynman invoke the Boltzmann distribution in deriving Einstein’s diffusion relation? Explain why should this use of the Boltzmann distribution be regarded as an equilibrium constraint or consistency condition?

 

References:

 

Einstein, A. (1905). On the Motion of Small Particles Suspended in Liquids at Rest Required by the Molecular-Kinetic Theory of Heat. Annalen der Physik, 17, 549–560.

Feynman, R. P., Leighton, R. B., & Sands, M. (1963). The Feynman Lectures on Physics, Vol I: Mainly mechanics, radiation, and heat. Reading, MA: Addison-Wesley.

Fick, A. (1855). On Liquid Diffusion. Philosophical Magazine, Vol. 10, pp. 30–39.

Perrin, J. B. (1926). Discontinuous structure of matter. In Nobel Lectures in Physics (Vol. 2). World Scientific.

Reif, F. (1965). Fundamentals of Statistical and Thermal Physics. McGraw-Hill.

Spiechowicz, J., Marchenko, I. G., Hänggi, P., & Łuczka, J. (2022). Diffusion coefficient of a Brownian particle in equilibrium and nonequilibrium: Einstein model and beyond. Entropy, 25(1), 42.

Sutherland, W. (1905). LXXV. A dynamical theory of diffusion for non-electrolytes and the molecular mass of albumin. The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science, 9(54), 781-785.

Sutherland, W. (1904). The measurement of large molecular masses. In Report of the 10th Meeting of the Australasian Association for the Advancement of Science, Dunedin (pp. 117-121).