Equipartition theorem — ½k_BT per quadratic degree of freedom: Visual walkthrough

Before we start, here is the vocabulary we will earn (never used before it is drawn):

Our one and only goal:

Step 1 — Draw the energy bowl

WHAT. We plot the energy $E_{q} = α q^{2}$ against the coordinate $q$ . This is a parabola: a bowl. The bead lives somewhere inside it, and the height of the wall at position $q$ tells you how much energy it costs to be there.

WHY. Everything downstream is about how often the bead visits each spot in this bowl. Before we can talk about probability, we must see the landscape the bead is exploring. The bowl is the stage; the bead is the actor.

PICTURE. The orange curve is $α q^{2}$ . At $q = 0$ (bottom) the cost is zero — cheap, easy to sit there. As $q$ grows either way, the wall rises steeply. Look at the two dashed heights: a stiffer $α$ (steeper bowl) makes the same $q$ far more expensive.

Figure — Equipartition theorem — ½k_BT per quadratic degree of freedom

Step 2 — The Boltzmann weight: how often does the bead visit $q$ ?

WHAT. Nature does not visit all spots equally. The rule (the Boltzmann distribution) is: the chance of finding the bead at coordinate $q$ is proportional to $P (q) \propto e^{- E_{q} / k_{B} T} = e^{- α q^{2} / k_{B} T} .$

WHY. Why this function and not, say, $1/ E$ or a straight line? Because a system in contact with a huge thermal bath spends time in each state in proportion to $e^{- E / k_{B} T}$ — high-energy states are exponentially rare, and the "exchange rate" between energy and rarity is set by $k_{B} T$ . We do not re-derive Boltzmann here; we use it as our one imported law. The minus sign matters: more energy $\Rightarrow$ smaller exponent $\Rightarrow$ less likely.

PICTURE. The blue curve is the visiting probability $e^{- α q^{2} / k_{B} T}$ — a bell (Gaussian) sitting on top of the same axis as the bowl. Where the bowl is cheap (near $q = 0$ ) the bell is tall: frequent visits. Where the bowl is expensive (large $∣ q ∣$ ) the bell hugs zero: rare visits. The bell is the upside-down shadow of the bowl.

Step 3 — Turn "average" into a fraction of two integrals

WHAT. The average energy is: (sum over all spots of energy-at-that-spot × how-often-we-visit) divided by (total visiting, so it normalizes to 100%). In the language of continuous $q$ , "sum" becomes an integral $\int$ — an infinite sum of infinitely thin slivers: $⟨ α q^{2} ⟩ = \frac{\int _{- \infty}^{\infty} energy here α q ^{2} weight here e ^{- α q^{2} / k_{B} T} d q}{\int _{- \infty}^{\infty} weight here e ^{- α q^{2} / k_{B} T} d q} .$

WHY. An average is always "weighted total ÷ total weight." The top adds up energy, giving each spot its Boltzmann vote. The bottom adds up the votes alone, so dividing turns votes into a proper probability. Crucially, every other coordinate in a big system contributes the same exponential factor to top and bottom, so it cancels — we are legitimately allowed to stare at just one variable $q$ .

PICTURE. Top panel: the numerator integrand $α q^{2} e^{- α q^{2} / k_{B} T}$ — a two-humped curve (zero at the middle because $q^{2} = 0$ there, zero far out because the bell kills it). Its shaded area is the weighted energy. Bottom panel: the denominator integrand, the plain bell, whose shaded area is the total weight. The answer is (top area) ÷ (bottom area).

Step 4 — The differentiation trick: manufacture the $q^{2}$

WHAT. The numerator is awkward because of the extra $q^{2}$ . We remove it with a trick. First rename the messy exponent: let $β = \frac{1}{k _{B} T}$ (Greek "beta"; small $β$ = hot, big $β$ = cold). Define the bottom integral its own name: $Z (β) = \int_{- \infty}^{\infty} e^{- β α q^{2}} d q .$ Now notice: if you differentiate $Z$ with respect to $β$ , the chain rule pulls a factor $- α q^{2}$ down from the exponent: $\frac{d Z}{d β} = \int_{- \infty}^{\infty} (- α q^{2}) e^{- β α q^{2}} d q .$ That is exactly the numerator, up to a minus sign! Therefore $⟨ α q^{2} ⟩ = \frac{- d Z / d β}{Z} = - \frac{d}{d β} ln Z .$

WHY. Why reach for a derivative? Because "multiply the integrand by $q^{2}$ " and "differentiate the whole integral by $β$ " do the same job here — and the second is far easier to carry out once we know $Z$ . The derivative is a machine that reads a factor of $q^{2}$ out of an exponential for free. That last equality uses $\frac{d}{d β} ln Z = \frac{1}{Z} \frac{d Z}{d β}$ , the log-derivative rule.

PICTURE. We show $Z (β)$ as a curve, and how a small nudge $Δ β$ (turning the temperature knob slightly colder) shrinks $Z$ . The slope of that shrink — the red tangent line — divided by the height $Z$ , is precisely our average energy. Cooling narrows the bell, so $Z$ (its area) drops; the rate of that drop encodes how much energy sat in the slot.

Step 5 — Do the one integral we need (the Gaussian)

WHAT. We only need to evaluate $Z (β) = \int_{- \infty}^{\infty} e^{- β α q^{2}} d q$ . This is the famous Gaussian integral. Its value is $\int_{- \infty}^{\infty} e^{- a x^{2}} d x = \frac{π}{a}, here a = β α,$ $Z (β) = \frac{π}{β α} = \frac{π}{α} β^{- 1/2} .$

WHY. Why does the area of a bell equal $π / a$ ? A taller, wider bell (small $a$ ) encloses more area; a narrow spike (large $a$ ) encloses less — and the square-root captures exactly how area trades against width. The exact constant $π$ comes from the classic "square it and switch to polar coordinates" proof, which you can take on faith here. The one fact we need: $Z$ scales as $β^{- 1/2}$ .

PICTURE. Three bells drawn for hot / medium / cold ( $a$ small / medium / large). The shaded area under each equals $Z$ . Hotter (smaller $a$ ) $\Rightarrow$ fatter bell $\Rightarrow$ larger area. The $π / a$ area is annotated on each.

Step 6 — Take the log-derivative and watch $α$ vanish

WHAT. Plug $Z = π / α β^{- 1/2}$ into $⟨ α q^{2} ⟩ = - d ln Z / d β$ . Take the log first — logs turn the product into a sum: $ln Z = constant in β ln \frac{π}{α} - \frac{1}{2} ln β .$ Now differentiate with respect to $β$ . The constant term (which is the only place $α$ lives) has zero slope and disappears: $- \frac{d l n Z}{d β} = - (0 - \frac{1}{2 β}) = \frac{1}{2 β} = \frac{1}{2} k_{B} T .$

WHY. This is the punchline of the whole page. Differentiation kills additive constants. Because $α$ entered only as an additive constant of $ln Z$ (thanks to the log splitting the product), the derivative erases it entirely. Whatever the stiffness, the answer is the same $\frac{1}{2} k_{B} T$ . Also note we used $1/ β = k_{B} T$ at the end to translate back to temperature.

PICTURE. We plot $ln Z$ against $β$ : a straight-ish descending curve of slope $- \frac{1}{2} / β$ . The horizontal offset (the $α$ part) is shown as a vertical shift of the whole line — and shifting a line up or down never changes its slope. Two lines for two different $α$ 's lie parallel; their common slope gives the same energy.

Step 7 — Edge case A: what if the slot is NOT quadratic?

WHAT. Suppose the energy were $E = α ∣ q ∣^{n}$ for some power $n$ instead of $n = 2$ . Re-running the same $- d ln Z / d β$ machine (the Gaussian becomes a general power integral, and $Z \propto β^{- 1/ n}$ ) gives $⟨ α ∣ q ∣^{n} ⟩ = \frac{1}{n} k_{B} T .$

WHY. We must show where the $\frac{1}{2}$ came from, or a reader will over-apply it. The $\frac{1}{2}$ is literally the $\frac{1}{n}$ with $n = 2$ . A stiffer-walled quartic bowl ( $n = 4$ ) confines the bead more, storing only $\frac{1}{4} k_{B} T$ . The theorem's magic number is a fingerprint of the shape of the energy well, not a universal constant.

PICTURE. Three bowls on one axis: $n = 2$ (parabola), $n = 4$ (flat-bottomed, steep walls), $n = 1$ (a V, sharp point). Beside each, the answer $\frac{1}{n} k_{B} T$ . The $n = 2$ curve is highlighted — that's the equipartition case.

Step 8 — Edge case B: the classical assumption ( $k_{B} T ≫$ level spacing)

WHAT. Our whole derivation replaced a sum over states with a smooth integral over $q$ . That is only legal when the allowed energy levels are packed far closer together than $k_{B} T$ — i.e. the ladder of quantum states looks like a smooth ramp. When the gap $Δ E$ between rungs satisfies $Δ E ≳ k_{B} T$ , the bead cannot even reach the first excited rung: the slot is frozen out and contributes zero, not $\frac{1}{2} k_{B} T$ .

WHY. We must cover the degenerate limit or the reader will wrongly predict $C_{V} = \frac{7}{2} R$ for room-temperature nitrogen. This is exactly why heat capacities rise in steps as you heat a gas — see Quantum freezing of degrees of freedom. The integral in Step 3 secretly assumed a continuum; when levels are coarse that assumption fails.

PICTURE. Left: closely spaced quantum rungs with $k_{B} T$ (a green bar) towering over the gap — the smooth-integral / equipartition regime, slot "on." Right: widely spaced rungs with $k_{B} T$ shorter than the first gap — the bead stuck on the ground rung, slot "off," contributing $0$ .

The one-picture summary

WHAT. One figure that chains every step: bowl → bell → weighted-energy humps → area $Z$ → slope of $ln Z$ → $\frac{1}{2} k_{B} T$ , with $α$ shown entering and then being crossed out.

PICTURE. Follow the arrows left to right. The parabola sets the stage; the Boltzmann bell says how often each spot is visited; the twin-humped product is where energy actually sits; its total normalises through $Z$ ; differentiating $ln Z$ by the cooling knob $β$ pulls out the slope $\frac{1}{2 β}$ ; and $α$ , having appeared only as a vertical shift, is struck through.

Recall Feynman retelling of the whole walkthrough

Picture a bead in a bowl. The bowl's steepness is the "stiffness" $α$ . The warmer world outside jiggles the bead, and a golden rule (Boltzmann) says the bead spends time near each spot in proportion to $e^{- cost / k_{B} T}$ — so it hangs around the cheap bottom and rarely climbs the expensive walls. To find the average energy the bead carries, we add up "energy here times time-spent-here" and divide by "total time" — that's the fraction of two integrals. The top one has an annoying extra $q^{2}$ , so we play a trick: differentiating the whole thing by the coldness-knob $β$ magically pulls a $q^{2}$ out of the exponential for us. That reduces everything to the area under a bell curve, a number that scales like $β^{- 1/2}$ . Take its log, take its slope, and out drops $\frac{1}{2 β} = \frac{1}{2} k_{B} T$ . The beautiful part: the bowl's steepness $α$ only ever showed up as a fixed shift of the log — and slopes ignore shifts — so a stiff bowl and a floppy bowl give the exact same energy. Two footnotes: if the bowl weren't a clean parabola (say $q^{4}$ ) the magic number changes to $\frac{1}{n}$ ; and if the world is too cold to lift the bead off the very bottom rung (quantum freezing), the slot simply switches off and stores nothing.