Blackbody Radiation and the Planck Quantum

A steelworker can read temperature by eye: dull red near $900\ \mathrm{K}$ , orange at $1200$ , white heat beyond $1500$ . The remarkable fact is that this color–temperature dictionary is universal — the glow of molten steel, a kiln's interior, a star's surface, and the afterglow of the Big Bang all follow one curve, fixed by temperature alone. Kirchhoff proved in 1859 that the spectrum of a perfect absorber (a blackbody, realized in the lab as a pinhole in a heated cavity) cannot depend on material, shape, or size, and challenged theory to produce the function. It became the most consequential curve-fitting problem in the history of science. The previous article, Breakdown of Classical Physics, showed what the classical toolkit does with the challenge: the mode counting is flawless, the equipartition filling rule is fatal, and the predicted spectrum diverges — off by a factor of ten million for green light in an ordinary kiln. This article builds the repair. Planck's move looks almost too small to matter: let each cavity oscillator of frequency $\nu$ hold energy only in integer multiples of $h\nu$ , then redo the same Boltzmann average. One geometric series later, the divergence is gone, the curve fits every dataset ever taken, and a new constant of nature is on the books. From that one factor you will extract the whole phenomenology of thermal light — Wien's displacement law, the Stefan–Boltzmann $T^4$ law, the color of the Sun and of the cosmic microwave background — each as a two-line corollary, every number checked.

⚡ Key Takeaways

A blackbody's spectrum is universal: only temperature matters. The cavity mode count $8\pi\nu^2/c^3$ (derived in Breakdown of Classical Physics) is classical and survives; what changes is the energy per mode.
Planck's hypothesis: a mode of frequency $\nu$ holds energy only in lumps, $E_n = nh\nu$ . Boltzmann-averaging the lumps gives $\langle E\rangle = h\nu/(e^{h\nu/k_BT}-1)$ instead of the classical $k_BT$ .
The Planck law $u(\nu,T) = \frac{8\pi h\nu^3}{c^3}\frac{1}{e^{h\nu/k_BT}-1}$ reduces to Rayleigh–Jeans at low frequency and cuts off exponentially at high frequency — the catastrophe is cured.
Maximizing the wavelength spectrum gives Wien's law $\lambda_{\max}T = 2.898\times10^{-3}\ \mathrm{m\,K}$ ; integrating gives Stefan–Boltzmann $P = \sigma T^4$ with $\sigma$ built entirely from $h$ , $k_B$ , $c$ .
The Sun ( $5772\ \mathrm{K}$ ) peaks at $502\ \mathrm{nm}$ ; the cosmic microwave background ( $2.725\ \mathrm{K}$ ) peaks near $1.06\ \mathrm{mm}$ and is the most perfect blackbody ever measured.
Key idea: keep the classical mode count, quantize the energy exchange — expensive quanta freeze out, and one new constant $h$ fixes the entire spectrum.

💡 The Core Insight

Counting cavity modes is classical and correct; the quantum step is that each mode of frequency $\nu$ can gain or lose energy only in lumps of $h\nu$ , so high-frequency modes are thermally frozen out instead of each carrying $k_BT$ .

The physical picture

Picture the cavity field as an orchestra of independent oscillators, one per standing-wave mode — a few long-wavelength double basses, more cellos, and an unbounded crowd of piccolos at ever higher frequency. Classical equipartition is a flat salary: every player receives $k_BT$ , and since the piccolo section is infinite, the payroll diverges. Planck's rule changes the currency, not the orchestra: each mode can only be paid in coins of its own denomination $h\nu$ . The basses' coins are cheap — thermal fluctuations hand them out freely, and those modes behave classically. A piccolo's coin costs many times the thermal budget $k_BT$ , and the Boltzmann factor $e^{-h\nu/k_BT}$ makes even a single coin exponentially rare. The high-frequency crowd sits silent. The payroll converges. That freeze-out picture localizes all the new physics in one dimensionless ratio, $x = h\nu/k_BT$ — the coin price measured in thermal budgets. Modes with $x \ll 1$ are classical; modes with $x \gg 1$ are frozen; and the crossover region $x \sim$ a few is where the spectrum peaks. Everything quantitative in this article — the shape of the curve, the position of the peak, the $T^4$ energy budget — is bookkeeping downstream of that one idea. Even the average number of coins in a mode, $\langle n\rangle = 1/(e^x - 1)$ , will matter far beyond this problem: it is the Bose–Einstein occupation, your first sight of quantum statistics.

Figure

Planck curves at three temperatures — the peak slides with T along Wien's locus

From modes to a spectrum

Two ingredients build any thermal spectrum: how many modes there are, and how much energy each holds. The first ingredient is settled classical geometry — the standing-wave count in a cavity, derived step by step in Breakdown of Classical Physics and recalled here as a result:

\rho_{\text{modes}}(\nu) = \frac{8\pi\nu^2}{c^3} \qquad (1)

where: $\rho_{\text{modes}}$ — modes per unit volume per unit frequency, two polarizations included. Quantum mechanics never touches this factor; every dispute is about the second ingredient. Classically, equipartition fills each mode with $\langle E\rangle_{\text{cl}} = k_BT$ and the divergent Rayleigh–Jeans law follows. Planck's postulate replaces the allowed energies of a mode:

E_n = n\,h\nu, \qquad n = 0, 1, 2, \dots \qquad (2)

where: $n$ — the number of quanta in the mode; $h$ — a new constant, fixed by fitting the measured spectrum ( $6.626\times10^{-34}\ \mathrm{J\,s}$ ). Nothing else changes — the same Boltzmann statistics, the same cavity, the same temperature. The entire quantum revolution, at this stage, is the replacement of an integral over energies by a sum.

Derivation AMean energy of a mode that trades in whole quanta

We want the thermal average energy of one cavity mode when its energy ladder is $E_n = nh\nu$ , because this single quantity, multiplied by the classical mode count, is the whole spectrum. The route is the standard machinery of statistical mechanics: build the partition function as a Boltzmann-weighted sum over the ladder, recognize a geometric series, and extract the mean energy by differentiating its logarithm. Keep two things in view: the classical limit must return $k_BT$ (equipartition was right for cheap quanta), and the high-frequency limit must die exponentially (that is what cures the catastrophe).

Step 1. Write the partition function. With $\beta = 1/k_BT$ and $x = \beta h\nu$ , $Z = \sum_{n=0}^{\infty} e^{-n\beta h\nu} = \sum_{n=0}^{\infty}(e^{-x})^n = \dfrac{1}{1 - e^{-x}}$ . Why: the Boltzmann weight of ladder rung $n$ is $e^{-nx}$ ; the rungs form a geometric series with ratio $e^{-x} < 1$ , which sums in closed form.

Step 2. Differentiate for the mean. The standard identity $\langle E\rangle = -\partial_\beta \ln Z$ with $\ln Z = -\ln(1 - e^{-\beta h\nu})$ gives $\langle E\rangle = \dfrac{h\nu\,e^{-\beta h\nu}}{1 - e^{-\beta h\nu}}$ . Why: differentiating the log of a sum of Boltzmann factors automatically produces the weighted average — the bookkeeping is packaged in one derivative.

Step 3. Tidy the ratio. Multiply numerator and denominator by $e^{\beta h\nu}$ :

\langle E\rangle = \frac{h\nu}{e^{h\nu/k_BT} - 1} = h\nu\,\langle n\rangle, \qquad \langle n\rangle = \frac{1}{e^{h\nu/k_BT} - 1} \qquad (3)

Why: the factored form names the physics — $\langle n\rangle$ is the mean number of quanta in the mode, the Bose–Einstein occupation you will meet again for photons, phonons, and every boson field.

Step 4. Check the cheap-coin limit. For $x = h\nu/k_BT \ll 1$ , expand $e^x \approx 1 + x$ : $\langle E\rangle \to h\nu/x = k_BT$ . Why: equipartition is recovered exactly where it was experimentally correct — Planck's rule contains the classical answer rather than contradicting it.

Step 5. Check the expensive-coin limit. For $x \gg 1$ , $\langle E\rangle \to h\nu\,e^{-h\nu/k_BT}$ : exponentially small. Why: a mode that cannot afford one coin is silent — this Boltzmann suppression is the cutoff classical physics could not supply. *Verified: the geometric sum, the $-\partial_\beta \ln Z$ derivative, and both limits recomputed symbolically with sympy; matches Reif Ch. 9 and the Shankar historical chapter.*

Multiplying the surviving classical count (1) by the new mean energy (3) assembles the law that started quantum theory:

u(\nu, T) = \frac{8\pi h\nu^3}{c^3}\,\frac{1}{e^{h\nu/k_BT} - 1} \qquad (4)

where: $u$ — spectral energy density per unit frequency. Read its anatomy: the $\nu^3$ out front is mode counting times one quantum, pushing the spectrum up; the Planck factor drags it down exponentially once $h\nu \gtrsim k_BT$ ; the contest between them carves the peak. At low frequency it is Rayleigh–Jeans; at high frequency it is Wien's empirical exponential — the two pre-1900 fragments were each half of one function. For laboratory work one usually quotes the same law per unit wavelength. Substituting $\nu = c/\lambda$ and — the step everyone forgets — transforming the interval via $u(\lambda)\,d\lambda = u(\nu)\,d\nu$ , which brings a Jacobian $|d\nu/d\lambda| = c/\lambda^2$ :

u(\lambda, T) = \frac{8\pi hc}{\lambda^5}\,\frac{1}{e^{hc/\lambda k_BT} - 1} \qquad (5)

where: the $\lambda^{-5}$ (not $\lambda^{-3}$ ) records that a fixed wavelength interval holds more frequency room at short $\lambda$ . That Jacobian is physical, and it moves the peak — as the warning below insists.

Figure

One mode's thermal energy — the classical salary versus the quantum freeze-out

Two laws for the price of one

The Planck law is a machine for corollaries. Ask where the wavelength spectrum peaks and you get Wien's displacement law; ask how much total energy comes out and you get the Stefan–Boltzmann law. Both were known empirically before 1900; both now acquire coefficients built from $h$ , $k_B$ , and $c$ — which is how Planck, in reverse, extracted the values of $h$ and $k_B$ from radiation data alone.

Derivation BWien's displacement law from the peak of the Planck curve

We want the wavelength at which $u(\lambda, T)$ of equation (5) is maximal, and we want to see why $\lambda_{\max}T$ is a pure constant of nature. The strategy is ordinary calculus made dimensionless: write the law in terms of $x = hc/\lambda k_BT$ , so that temperature scales out entirely and the maximization becomes a single transcendental equation solved once and for all. The same procedure applied to the frequency form (4) will give a different root — keep both numbers, because their mismatch is a classic trap.

Step 1. Go dimensionless. With $x = hc/\lambda k_BT$ , equation (5) is proportional to $T^5\,x^5/(e^x - 1)$ . Why: all $\lambda$ - and $T$ -dependence collapses into $x$ ; the peak of $x^5/(e^x-1)$ is a number, so $\lambda_{\max}T$ must be constant — the displacement law's form is forced before any arithmetic.

Step 2. Differentiate and set to zero. $\dfrac{d}{dx}\dfrac{x^5}{e^x - 1} = 0$ gives $5(e^x - 1) = x e^x$ , i.e. $x = 5(1 - e^{-x})$ . Why: the product rule pits the $x^5$ growth against the exponential decay; the balance point is the peak.

Step 3. Solve the transcendental equation. Iterating or root-finding gives $x^\ast = 4.965114$ . Then

\lambda_{\max} T = \frac{hc}{x^\ast k_B} = 2.898\times10^{-3}\ \mathrm{m\,K} \qquad (6)

Why: the constant is built from three fundamental constants — measuring a glow color is measuring temperature.

Step 4. Repeat for the frequency spectrum. Maximizing (4) means maximizing $x^3/(e^x - 1)$ , whose root solves $x = 3(1 - e^{-x})$ : $x^\ast_\nu = 2.821$ . Why: the two peaks are genuinely different features — $\nu_{\max} \ne c/\lambda_{\max}$ , because equal frequency intervals and equal wavelength intervals slice the same spectrum differently. For the Sun, $\lambda_{\max} = 502\ \mathrm{nm}$ but $c/\nu_{\max} = 883\ \mathrm{nm}$ . *Verified: both roots recomputed by numerical root-finding ( $4.965114$ , $2.821439$ ) and $b = hc/x^\ast k_B = 2.8978\times10^{-3}\ \mathrm{m\,K}$ from CODATA constants.*

Derivation CThe Stefan–Boltzmann law and the value of sigma

We want the total energy density at temperature $T$ — the integral of the Planck law over all frequencies — both to confirm it is finite (the catastrophe's official death certificate) and to predict the radiated power. The route: the same dimensionless substitution as before turns the integral into a pure number times $T^4$ ; the number is a classic definite integral worth knowing on sight. The bridge from energy density in the cavity to power emitted by a surface is a standard kinetic-theory factor of $c/4$ picked up by counting outward-moving photons.

Step 1. Scale out the temperature. With $x = h\nu/k_BT$ , $\displaystyle\int_0^\infty u(\nu,T)\,d\nu = \frac{8\pi h}{c^3}\left(\frac{k_BT}{h}\right)^4 \int_0^\infty \frac{x^3}{e^x - 1}\,dx$ . Why: the quartic prefactor is pure dimensional analysis once the integrand is dimensionless — the $T^4$ law is fixed before the integral is evaluated.

Step 2. Evaluate the Bose integral. $\displaystyle\int_0^\infty \frac{x^3}{e^x - 1}\,dx = \Gamma(4)\,\zeta(4) = 6 \cdot \frac{\pi^4}{90} = \frac{\pi^4}{15}$ . Why: expanding $1/(e^x - 1) = \sum_{n\ge1} e^{-nx}$ turns the integral into $\sum 6/n^4$ — each mode of the geometric expansion contributes a term of the zeta function.

Step 3. Convert density to emitted flux. Radiation streams out of a small hole at $c$ in all outward directions; averaging the angular projection gives flux $= \tfrac{c}{4}\,u_{\text{total}}$ , so

P = \sigma T^4, \qquad \sigma = \frac{2\pi^5 k_B^4}{15\,h^3 c^2} = 5.670\times10^{-8}\ \mathrm{W\,m^{-2}K^{-4}} \qquad (7)

Why: $\sigma$ is not an independent constant of nature — it is $h$ , $k_B$ , and $c$ in a trench coat. Its measured value was one of Planck's two equations for extracting $h$ and $k_B$ simultaneously; the modern CODATA value of $\sigma$ is exact by definition. *Verified: the Bose integral checked symbolically ( $\Gamma(4)\zeta(4) = \pi^4/15$ ) and numerically; $\sigma = 5.6704\times10^{-8}$ recomputed from CODATA $h$ , $k_B$ , $c$ .*

Limiting cases

Case	Result	Meaning
$h\nu \ll k_BT$	$u \to \frac{8\pi\nu^2}{c^3}k_BT$	Rayleigh–Jeans recovered — classical physics is the low-frequency face of the Planck law
$h\nu \gg k_BT$	$u \to \frac{8\pi h\nu^3}{c^3}e^{-h\nu/k_BT}$	Wien's empirical exponential — the frozen-out tail
$h \to 0$	$\langle E\rangle \to k_BT$ at every $\nu$	the catastrophe returns; $h$ is literally the knob measuring quantum-ness
$\langle n\rangle \gg 1$	mode $\approx$ classical wave	radio engineering never notices photons: at 100 MHz and 300 K, $\langle n\rangle \sim 10^5$
$T \to 0$	$u \to 0$ at every $\nu$	no thermal photons at absolute zero (the field's zero-point story belongs to the oscillator articles)

⚠️ “The spectrum peaks at the same place whether you plot against frequency or wavelength”

It does not, and the mismatch is large. The peak of $u(\lambda)$ solves $x = 5(1-e^{-x})$ , giving $\lambda_{\max}T = 2.898\ \mathrm{mm\,K}$ ; the peak of $u(\nu)$ solves $x = 3(1-e^{-x})$ , and $c/\nu_{\max}$ works out to $\lambda' T \approx 5.10\ \mathrm{mm\,K}$ — for the Sun, $502\ \mathrm{nm}$ versus $883\ \mathrm{nm}$ , nearly a factor of two. Neither peak is "the real one": they answer different questions, because $u(\lambda)\,d\lambda = u(\nu)\,d\nu$ forces a Jacobian $c/\lambda^2$ that reshapes the curve when you change variables. Quote a peak only together with the variable it maximizes, and never compute $\nu_{\max}$ as $c/\lambda_{\max}$ . The physics — where the energy actually sits — is the integral over any band, which is variable-independent.

Worked examples

1 · Weighing the Sun's furnace by its color (confidence builder)

The Sun's surface temperature is $T = 5772\ \mathrm{K}$ and its radius $R = 6.96\times10^{8}\ \mathrm{m}$ . Find the peak wavelength of its spectrum, the power radiated per square metre of surface, and the total luminosity.

Step 1. Wien's law (6): $\lambda_{\max} = \dfrac{2.898\times10^{-3}}{5772} = \boxed{502\ \mathrm{nm}}$ — blue-green, near the peak sensitivity of the human eye, which evolved under exactly this lamp.

Step 2. Stefan–Boltzmann (7): $P = \sigma T^4 = (5.670\times10^{-8})(5772)^4 = \boxed{6.29\times10^{7}\ \mathrm{W\,m^{-2}}}$ — sixty-three megawatts from every square metre, a nuclear power station per bedsheet of surface.

Step 3. Luminosity: $L = 4\pi R^2\,\sigma T^4 = 4\pi(6.96\times10^8)^2(6.29\times10^7) = \boxed{3.8\times10^{26}\ \mathrm{W}}$ , matching the measured solar luminosity.

Step 4. Sensitivity check: because $L \propto T^4$ , a 1% error in temperature becomes a 4% error in power — which is why astronomers obsess over blackbody fits. This chain (color → temperature → power) is precisely how effective temperatures of stars are assigned.

2 · The oldest light in the universe (exam level)

The cosmic microwave background is a blackbody at $T = 2.725\ \mathrm{K}$ . Find its peak wavelength, its total energy density, and the number density of its photons ( $n_\gamma = 16\pi\zeta(3)(k_BT/hc)^3$ , $\zeta(3) \approx 1.202$ ).

Step 1. Wien's law: $\lambda_{\max} = \dfrac{2.898\times10^{-3}}{2.725} = \boxed{1.06\ \mathrm{mm}}$ — millimetre waves, which is why it took microwave receivers (and two puzzled radio astronomers) to find it.

Step 2. Energy density: $u = \dfrac{4\sigma}{c}T^4 = \dfrac{4(5.670\times10^{-8})}{3.00\times10^8}(2.725)^4 = 4.17\times10^{-14}\ \mathrm{J\,m^{-3}} = \boxed{0.26\ \mathrm{eV\,cm^{-3}}}$ — a quarter of an electron-volt per cubic centimetre, everywhere.

Step 3. Photon count: $n_\gamma = 16\pi(1.202)\left(\dfrac{(1.381\times10^{-23})(2.725)}{(6.626\times10^{-34})(3.00\times10^8)}\right)^3 = \boxed{411\ \mathrm{photons\,cm^{-3}}}$ . Each cubic centimetre of the universe — including the one at the tip of your nose — holds about four hundred Big-Bang photons.

Step 4. Occupation check: at the peak, $x^\ast = 4.965$ gives $\langle n\rangle = 1/(e^{4.965}-1) \approx 0.007$ quanta per mode — deeply quantum. At radio frequencies ( $x = 0.01$ ), $\langle n\rangle \approx 100$ — that tail is honestly classical, which is why the Rayleigh–Jeans formula still earns its keep in radio astronomy as the "brightness temperature."

Max Planck

1858–1947 · Berlin

Planck came to the blackbody problem as a conservative — a thermodynamics purist who had spent years trying to derive the spectrum without Boltzmann's statistical methods, which he distrusted. On 7 October 1900, the experimenter Heinrich Rubens visited Planck's home with fresh far-infrared data that broke Wien's formula; by that evening Planck had interpolated a new curve between the Wien and Rayleigh–Jeans behaviors, and it fit — perfectly, at every frequency, as he announced to the German Physical Society on 19 October. Only then came the desperate part: eight weeks hunting for what his lucky formula could mean, which forced him — the anti-Boltzmannian — to adopt Boltzmann's own counting of discrete energy elements, presented on 14 December 1900. The curve came first; the quantum was the price of explaining it. Planck spent years afterwards trying to unhappen his own revolution, calling the quantum "a purely formal assumption," and received the 1918 Nobel Prize for it anyway.

⏳ History

1859

— Kirchhoff proves the blackbody spectrum is a universal function of temperature and challenges theory to find it.

1879

— Stefan infers the

T^4

law from experiment; Boltzmann derives it thermodynamically in 1884.

1893

— Wien derives the displacement law

\lambda_{\max}T = \text{const}

from thermodynamic scaling.

1896

— Wien proposes his exponential distribution; it fits the visible but fails in the far infrared.

1900

— Rubens' data, Planck's October interpolation, and the December quantum hypothesis

E_n = nh\nu

1924

— Bose re-derives the Planck law from photon statistics alone; quantum statistics is born.

1990

— COBE/FIRAS measures the CMB as a

2.725\ \mathrm{K}

Planck curve — the most perfect blackbody ever observed.

Where this lives today

The Planck law is working infrastructure. Every non-contact thermometer — the clinical forehead scanner, the steel-mill pyrometer, the satellite instruments mapping sea-surface temperature — inverts equation (5) to turn spectral radiance into temperature. Thermal cameras image the $\sim300\ \mathrm{K}$ Planck curve of the human body, which peaks near $10\ \mathrm{\mu m}$ ; incandescent light bulbs died commercially because a $2900\ \mathrm{K}$ filament peaks at $1\ \mathrm{\mu m}$ and wastes most of its Planck curve in the infrared. In cosmology, the FIRAS instrument's measurement of the CMB fits a Planck spectrum with deviations below fifty parts per million — the most precise confirmation of any thermal-radiation prediction — and modern experiments hunt for the tiny spectral distortions that would betray energy injected into the early universe. The constant born here now anchors the units themselves. Since 2019 the SI fixes $h$ exactly, and the kilogram is realized by a Kibble balance that weighs mass against electrical power calibrated in quantum units — Planck's "purely formal assumption" is literally the definition of a kilogram of flour. And the mean occupation $\langle n\rangle$ from equation (3) went on to run quantum optics, laser physics (stimulated emission bookkeeping), and the thermal-noise floor of every radio telescope and gravitational-wave detector.

Check yourself

Going further

Texts. Griffiths, Introduction to Quantum Mechanics, 3rd ed., historical afterword · Shankar, Principles of Quantum Mechanics, 2nd ed., Ch. 7 historical introduction · Cohen-Tannoudji, Diu & Laloë, Vol. 1 · Townsend, A Modern Approach to Quantum Mechanics · Planck, Ann. Phys. 4, 553 (1901) · Reif, Fundamentals of Statistical and Thermal Physics, Ch. 9.

Provenance. Equations follow Planck (1901) and standard treatments in Griffiths, Shankar, Cohen-Tannoudji, and Reif. A verifier should: (1) recompute the geometric-series partition function and the $-\partial_\beta\ln Z$ derivation of the mean energy with both limits; (2) recall the cavity mode density $8\pi\nu^2/c^3$ from the octant argument (owned by Breakdown of Classical Physics); (3) verify the Bose integral $\int_0^\infty x^3/(e^x-1)\,dx = \pi^4/15$ and the resulting $\sigma = 2\pi^5k_B^4/15h^3c^2$ ; (4) solve the peak equations $x = 5(1-e^{-x})$ and $x = 3(1-e^{-x})$ ( $x^\ast = 4.965114$ , $2.821439$ ) and confirm Wien's $b = 2.898\times10^{-3}\ \mathrm{m\,K}$ ; (5) check the worked numbers — Sun: $502\ \mathrm{nm}$ , $6.29\times10^7\ \mathrm{W\,m^{-2}}$ , $3.8\times10^{26}\ \mathrm{W}$ ; CMB: $1.06\ \mathrm{mm}$ , $0.26\ \mathrm{eV\,cm^{-3}}$ , $411\ \mathrm{cm^{-3}}$ — against CODATA constants: $h = 6.626\times10^{-34}\ \mathrm{J\,s}$ , $k_B = 1.381\times10^{-23}\ \mathrm{J/K}$ , $\sigma = 5.670\times10^{-8}\ \mathrm{W\,m^{-2}K^{-4}}$ , CMB $T = 2.725\ \mathrm{K}$ .

Prerequisites. Breakdown of Classical Physics · thermal equilibrium and the Boltzmann distribution · electromagnetic standing waves. Leads to. Photoelectric Effect and the Photon (the quantum becomes a particle of light) · Wave–Particle Duality · de Broglie Matter Waves.

Where exactly does Planck's derivation stop being classical? Everything in it looks like ordinary statistical mechanics.

Sharp eye — the Boltzmann weighting, the partition function, the cavity, the mode count are all impeccably classical. The single non-classical move is the domain of the energy: classical statistical mechanics integrates $E\,e^{-E/k_BT}$ over a continuum of energies (yielding $k_BT$ for an oscillator), while Planck sums over the ladder $E_n = nh\nu$ . Replacing $\int dE$ by $\sum_n$ is the entire revolution. Everything downstream — freeze-out, the finite spectrum, Wien's constant — is classical machinery digesting that one substitution.

Did Planck believe light itself comes in photons?

No — and he resisted the idea for years. Planck quantized the energy exchange between the cavity-wall oscillators and the field, treating $E = nh\nu$ as a property of the material oscillators' bookkeeping, not of light. The claim that radiation itself is grainy — real flying quanta — is Einstein's 1905 photon hypothesis, and Planck explicitly demurred from it as late as 1913. The distinction matters conceptually: the Planck law alone can be derived either way; it is the photoelectric effect and Compton scattering (next articles) that force the photon to be real.

Is the mode count still valid in quantum mechanics, or does it get corrected later?

It survives untouched — and that is worth appreciating. The count $8\pi\nu^2/c^3$ is pure wave geometry: how many standing-wave patterns fit in a box per frequency interval. Quantum field theory re-derives exactly the same density of modes; what it changes is that each mode becomes a quantum oscillator with occupation $\langle n\rangle = 1/(e^{h\nu/k_BT}-1)$ . Even the modern refinement — zero-point energy $\tfrac12 h\nu$ per mode — cancels out of thermal spectra, because only changes in mode energy are radiated.

Why does the total energy scale as exactly the fourth power of temperature?

Dimensional analysis does it before the integral is touched. Once the spectrum depends on frequency only through $x = h\nu/k_BT$ , the integral over $\nu$ must pull out a factor $(k_BT/h)^4$ : three powers from converting $\nu^3$ and one from $d\nu$ . The remaining number, $\pi^4/15$ , just normalizes. Boltzmann had in fact derived $T^4$ thermodynamically in 1884 from radiation pressure alone — the Planck law's contribution is the coefficient $\sigma$ , tying the law to $h$ , $k_B$ , and $c$ .

If the CMB peak occupation is only 0.007 photons per mode, how is it the "most perfect blackbody"?

Perfection here means the shape of the spectrum, not the brightness. FIRAS compared the sky against an on-board calibrated cavity and found the Planck form followed to better than 50 parts per million across the peak — no other thermal source, laboratory ones included, matches its own Planck curve that well (real ovens have emissive walls, gradients, and holes). The low occupation is simply what $T = 2.725\ \mathrm{K}$ demands; sparse photons still trace an exquisitely thermal distribution because the early universe was a genuinely equilibrated plasma.

The kiln example says classical physics overpredicts green light by ten million — but a kiln obviously does glow. What am I seeing?

You are seeing the honest Planck tail, which is small but not zero. At $1500\ \mathrm{K}$ and $500\ \mathrm{nm}$ , $x \approx 19$ , so each green mode holds $\langle n\rangle \sim e^{-19} \approx 5\times10^{-9}$ photons on average — yet there are astronomically many modes and your eye is a superb photon counter, so the kiln looks orange-white. The factor of $10^7$ is the ratio between that faint reality and the classical prediction, which would have every kiln (and every warm object, including you) blazing in the visible and beyond. The glow you see is the cure; the blaze you don't see is the catastrophe.

Planck "fit the curve first and explained it later" — is that bad physics?

It is superb physics, and a useful lesson in how theory actually advances. The October interpolation was constrained guessing: the formula had to reduce to Wien at high frequency, to Rayleigh–Jeans at low, and to fit Rubens' data in between — a tight corridor. What elevates it beyond curve-fitting is the December step: demanding a mechanism, and accepting the mechanism's alien price (discrete energy) when it was the only one that worked. Fit-then-explain has a distinguished record since — Balmer's formula preceded quantum mechanics by forty years. The sin is stopping at the fit.