The classical synthesis and its limits

Three theories, each complete in its own domain and joined at the edges. Newton's laws with universal gravitation take a state — every position and momentum now — and return every later state by integrating F = dp/dt. Maxwell's four equations do the same for fields, and hand back a wave speed c = 1/√(ε₀μ₀) = 3.00 × 10⁸ m s⁻¹ assembled from two bench constants. Statistical mechanics closes the gap: when tracking 10²³ trajectories is hopeless, S = k ln W and the equipartition theorem's ½kT per quadratic degree of freedom deliver the thermodynamics anyway. The Lorentz force F = q(E + v×B) stitches the first two together. Call this the classical synthesis. It is not a slogan about determinism; it is a working prediction machine, and for planets, bridges, engines, radio and dilute gases it still returns the right number to more figures than most experiments can check.

The synthesis assumes velocities add the way Galileo said, which is what would let a light wave have a different speed in a moving frame. Maxwell's equations disagree: they contain c and no frame for it to be measured against. The dimensionless β = v/c decides which of the two matters here. Expand the relativistic kinetic energy: (γ − 1)mc² = ½mv²(1 + ¾β² + …), so the classical value is low by 0.75% at β = 0.1 and by 19% at β = 0.5. Concretely, an electron pushed through 100 V reaches β = 0.0198, and there the classical formula overstates β by only 0.015%; pushed through 100 kV, the classical √(2eV/m) predicts β = 0.626 where the true value is 0.548 — 14% too fast, and above 256 kV it returns speeds greater than c. Keep β under about 0.1 and Newtonian dynamics survives; past that, the corrections are the physics.

Equipartition assumes energy is continuous, so a mode can accept any amount however small and therefore averages kT in equilibrium whatever its frequency. Compare hν, the smallest lump the mode can actually take, with kT, the energy on offer: x = hν/kT. At 300 K, kT/h = 6.25 THz, so any mode above roughly 6 THz — any wavelength shorter than 48 μm — is starved. Nitrogen makes the point inside one molecule. Its rotational levels sit at x ≈ 0.02 and behave classically; its 2359 cm⁻¹ vibration sits at x = 11.3, and its mean energy hν/(e^x − 1) is 1.4 × 10⁻⁴ of kT, one part in seven thousand of what equipartition promises. So the measured molar heat capacity of N₂ at room temperature is 5R/2 = 20.8 J K⁻¹ mol⁻¹, not the 7R/2 = 29.1 J K⁻¹ mol⁻¹ that counting degrees of freedom predicts. Heat it to 3000 K, x drops to 1.13, and the vibration wakes up.

The third assumption is that a particle has a definite path, so position and momentum can both be sharp at once. De Broglie's λ = h/p sets the scale on which that fails, and what matters is λ compared with d, the length the system actually cares about — a slit, a well, a mean spacing. A nitrogen molecule in room air carries λ = 28 pm against a mean spacing of 3.4 nm, so Λ = 0.008 and the ideal-gas law survives. Conduction electrons in copper are the counter-example. With n = 8.5 × 10²⁸ m⁻³ the spacing is 0.23 nm, while the thermal wavelength h/√(3mkT) for an electron at 300 K is 6.2 nm, giving Λ = 27. That 6.2 nm is itself a classical estimate, and the fact that it dwarfs the spacing is exactly the signal that the estimate had no right to be made: the electrons overlap, classical state counting is simply wrong, and the electronic heat capacity measured at 300 K is about 1% of the 3k/2 per electron that Drude's classical gas assumed.

The three ratios are three faces of one comparison: the system's characteristic action against Planck's constant, h = 6.63 × 10⁻³⁴ J s. For a bound orbit that action is S = ∮ p dq. The old quantum condition of Bohr and Sommerfeld set it to an integer multiple of h, and the semiclassical estimate S ≈ (n + ½)h refines the same idea; neither is a law of quantum mechanics, but both say the classical regime is S/h ≫ 1 — equivalently, large quantum numbers. That is what makes the boundary one-way. Relativity must return Newton as β → 0, and Planck's mean mode energy hν/(e^x − 1) must return kT as x → 0; expand the exponential, e^x − 1 ≈ x, and it does. Any successor theory is obliged to reproduce the classical answers wherever these ratios are small, because that is exactly where the classical answers are already known to be right to several figures. The correspondence principle is not politeness; it is a hard constraint on what a new theory is permitted to say.

The procedure is short. Write down the system's speed v, the frequency ν (or level spacing ΔE/h) of the mode you care about, the temperature T, and the length d that confines it. Form β = v/c, x = hν/kT and Λ = λ/d. If all three sit well below 1 — say under 0.1 — use classical physics and expect errors of a few percent or less. If one exceeds 1, that ratio names which part of the account has failed and therefore which repair to reach for: Lorentz kinematics, quantized modes, or a wave description of the particle. Two can fail together and often do; inside a hot plasma β and x are both large. And a small ratio is a licence, not a guarantee — the classical model can still fail for reasons no ratio here tracks, such as the indistinguishability of identical particles or exponential sensitivity to initial conditions.