Thermodynamics of the BEC Transition

Cool a gas of helium-4 through 2.17 kelvin and something visible happens: the frantic boiling stops dead, the liquid goes eerily still, and it begins to creep up the walls of its container and siphon itself out. Plot its heat capacity against temperature and you find not a smooth bump but a sharp spike shaped like the Greek letter λ — the reason the transition is named the lambda point. That spike is a fingerprint. It says the substance has done something abrupt to its internal order, and it says so in the language of thermodynamics: a response function has gone singular. The astonishing claim of this article is that most of that fingerprint is already present in a gas with no forces between its particles at all. Bose–Einstein condensation is the one phase transition driven purely by quantum statistics — by nothing more than the demand that occupation numbers stay non-negative. The previous two articles established that it happens (a macroscopic pile-up in the ground state) and where it happens ( $T_c$ and the condensate fraction). Here we compute what it does to the bulk thermodynamics: how the pressure, energy, entropy, heat capacity, and compressibility behave on each side of $T_c$ , and exactly what kind of singularity marks the crossing. The whole calculation collapses onto two numbers — the Bose–Einstein functions evaluated at the transition, $\zeta(3/2)\approx2.612$ and $\zeta(5/2)\approx1.341$ . From them we will extract a heat capacity that peaks at $1.925\,Nk_B$ (above the classical $1.5$ ), continuous across $T_c$ but with a kinked slope; an entropy that is perfectly continuous, so there is no latent heat despite the coexistence-like behaviour; and a pressure that below $T_c$ forgets the volume entirely, driving the compressibility to infinity. Every one of these numbers is recomputed and machine-checked here.

Key takeaways

Below $T_c$ the chemical potential is pinned at $\mu = 0$ , so the non-condensed ("normal") gas has thermodynamics fixed by temperature alone — density drops out. The condensate is a thermodynamically inert reservoir: no energy, no pressure, no entropy.
Every critical amplitude reduces to two Riemann-zeta values, $g_{3/2}(1) = \zeta(3/2) \approx 2.612$ and $g_{5/2}(1) = \zeta(5/2) \approx 1.341$ .
The heat capacity is continuous at $T_c$ , peaking at $C_V/Nk_B = \tfrac{15}{4}\zeta(5/2)/\zeta(3/2) = 1.925$ , but its slope $\partial C_V/\partial T$ jumps — a cusp, not a jump discontinuity.
The entropy is continuous across $T_c$ , so the latent heat is exactly zero: the transition is continuous, not first order, despite its coexistence-like pressure law.
Below $T_c$ the pressure depends only on $T$ (not $V$ ), so the isothermal compressibility $\kappa_T \to \infty$ — the flat-isotherm signature of phase coexistence.
Key idea: locking $\mu = 0$ below $T_c$ makes the normal gas a one-parameter (temperature-only) substance; every thermodynamic anomaly of the ideal-gas BEC — the $1.925\,Nk_B$ cusp, the zero latent heat, the diverging $\kappa_T$ — is a corollary of that single fact, quantified entirely by $\zeta(3/2)$ and $\zeta(5/2)$ .

The core insight

Below $T_c$ the chemical potential is locked at $\mu = 0$ , so the excited-state (normal) gas has thermodynamics fixed by temperature alone and the condensate is inert; every critical amplitude reduces to the Bose–Einstein functions $g_{3/2}(1)=\zeta(3/2)$ and $g_{5/2}(1)=\zeta(5/2)$ .

Where this lives today

The thermodynamic signatures computed here are measured routinely in cold-atom laboratories, where dilute alkali gases come far closer to the ideal-gas limit than helium ever could. Releasing a trapped condensate and imaging its expansion (time-of-flight) reveals the condensate fraction directly; in a harmonic trap the exponent shifts from the box's $1 - (T/T_c)^{3/2}$ to $1 - (T/T_c)^3$ , a change of geometry the same mathematics predicts. The energy-versus-temperature curve shows the predicted kink, and weak interactions shift $T_c$ by a calculable few percent — a correction that turns the ideal-gas baseline into a precision tool. The diverging compressibility has a life beyond thermodynamics: it signals the vanishing restoring force for density waves, and in the interacting condensate it is cured by repulsion, which sets a finite compressibility and the Bogoliubov sound speed $c = \sqrt{gn/m}$ . The ideal-gas divergence marks exactly the point where interactions stop being optional. And the cusp itself, sharpened into helium's logarithmic λ-anomaly, remains one of the most stringent experimental tests of renormalization-group universality — the 1992 microgravity measurement pinned the critical exponent to a precision that ground-based experiments, rounded by the sample's own weight, could not reach.

The physical picture

Think of the excited states as a bucket whose capacity depends on temperature. At temperature $T$ the excited levels can hold at most $N_{\rm ex}^{\max} = V\,\zeta(3/2)/\lambda_T^3$ particles, where $\lambda_T = h/\sqrt{2\pi m k_B T}$ is the thermal de Broglie wavelength — this saturation is the mechanism owned by aq-sm-bose-einstein-condensation-ideal-gas, and the value of $T_c$ where the bucket first overflows is owned by aq-sm-bec-critical-temperature-condensate-fraction. Cooling shrinks the bucket ( $N_{\rm ex}^{\max} \propto T^{3/2}$ ), and the overflow drains silently into the zero-energy ground state, contributing nothing to pressure or energy. Everything in this article is what that draining does to the thermodynamic functions. The single most consequential fact is that below $T_c$ the chemical potential is frozen at zero. A gas whose $\mu$ is fixed and whose only remaining variable is $T$ is a very strange substance: its pressure cannot depend on how much of it you have or how tightly you pack it. Squeeze the box at fixed temperature and you do not raise the pressure — you just shovel more atoms across the invisible boundary into the condensate at constant $P$ . This is exactly the flat isotherm of a liquid–vapour coexistence region, and it is why the isothermal compressibility diverges: the gas offers no resistance to compression. The heat-capacity cusp arises from a competition. Approaching $T_c$ from below, cooling removes energy from the shrinking normal fraction, and $C_V$ climbs as $T^{3/2}$ . The moment you cross above $T_c$ , the chemical potential unpins and starts to fall, and the gas begins its crossover toward ordinary classical behaviour, so $C_V$ turns and decreases. The two branches meet — continuously — at the peak value $1.925\,Nk_B$ , precisely where the condensate order parameter vanishes. But they meet at different slopes. That mismatch in slope is the mathematical heart of the λ-shape.

Figure

The λ-shaped heat capacity — continuous at T_c but with a kinked slope, peaking at 1.925 Nk_B above the classical 1.5

Figure

Pressure below T_c depends only on temperature — the flat coexistence isotherm that sends κ_T to infinity

The λ-transition of real helium is more singular than the ideal gas: its heat capacity diverges logarithmically rather than merely cusping, because interactions push helium into a different universality class. But the ideal gas already captures the essential shape and the essential physics, which is precisely why Fritz London, in 1938, proposed identifying superfluidity with Bose–Einstein condensation. That historical leap is the reason this article carries his portrait.

The core equations

Everything begins with the count of particles the excited states can hold. The number of atoms in nonzero-energy states, at fugacity $z = e^{\beta\mu}$ , is set by the Bose–Einstein function $g_{3/2}$ integrated against the free-particle density of states $g(\varepsilon)\propto\sqrt\varepsilon$ (owned by aq-sm-density-of-states-free-particles). This is the "bucket capacity" that saturates at $z = 1$ :

N_{\rm ex} = \frac{V}{\lambda_T^3}\,g_{3/2}(z), \qquad \lambda_T = \frac{h}{\sqrt{2\pi m k_B T}} \qquad (1)

where: $N_{\rm ex}$ — particles in excited states; $z = e^{\beta\mu}$ — fugacity, $0 < z \le 1$ ; $g_{3/2}(z) = \sum_{\ell\ge1} z^\ell/\ell^{3/2}$ — the Bose–Einstein (polylogarithm) function of order $3/2$ ; $\lambda_T$ — thermal de Broglie wavelength. Reading it physically: the prefactor $V/\lambda_T^3$ is roughly the number of quantum cells available, and $g_{3/2}(z)$ climbs monotonically to its ceiling $\zeta(3/2)$ as $z\to1$ . Once $N_{\rm ex}$ hits that ceiling, extra particles have nowhere to go but the ground state. The two combinations that recur throughout are the pressure and energy, which come from the grand potential of the gas. Integrating $-\ln(1 - z e^{-\beta\varepsilon})$ against the same $\sqrt\varepsilon$ density of states — one integration by parts raises the index from $3/2$ to $5/2$ — gives:

P = \frac{k_B T}{\lambda_T^3}\,g_{5/2}(z), \qquad U = \frac{3}{2}\,\frac{k_B T\,V}{\lambda_T^3}\,g_{5/2}(z) = \frac{3}{2}PV \qquad (2)

where: $g_{5/2}(z) = \sum_{\ell\ge1} z^\ell/\ell^{5/2}$ ; $U$ — internal energy of the gas. The relation $U = \tfrac32 PV$ is identical to the classical ideal gas and follows purely from the $\varepsilon \propto p^2$ dispersion (which fixes $g(\varepsilon)\propto\varepsilon^{1/2}$ ); quantum statistics changes the value of $P$ through $g_{5/2}$ , not this ratio. Below $T_c$ , $z$ is pinned at $1$ , so $g_{5/2}(z)\to\zeta(5/2)$ and the pressure becomes a pure function of temperature. Setting $z=1$ in (2) gives the coexistence-line pressure — the flat isotherm drawn in the figure above. Because $\lambda_T^{-3}\propto T^{3/2}$ , this pressure scales as $T^{5/2}$ and, crucially, contains no $V$ or $N$ :

P_c(T) = \frac{k_B T}{\lambda_T^3}\,\zeta(5/2) = \zeta(5/2)\left(\frac{2\pi m k_B}{h^2}\right)^{3/2}(k_B T)^{5/2}, \qquad T \le T_c \qquad (3)

where: $\zeta(5/2)\approx1.341$ ; $P_c(T)$ — pressure on or below the transition. This is the equation that makes the ideal-gas BEC look like a coexistence: for any density above the critical one, the pressure sits on this single universal curve, independent of how compressed the sample is. The heat capacity is where the transition announces itself. Differentiating $U(T,z)$ at fixed $N,V$ — which forces $z$ to depend on $T$ through the number constraint — gives a two-term expression above $T_c$ that collapses to one term below:

\frac{C_V}{Nk_B} = \frac{15}{4}\frac{g_{5/2}(z)}{g_{3/2}(z)} - \frac{9}{4}\frac{g_{3/2}(z)}{g_{1/2}(z)}\;(T>T_c), \qquad \frac{C_V}{Nk_B} = \frac{15}{4}\frac{\zeta(5/2)}{\zeta(3/2)}\left(\frac{T}{T_c}\right)^{3/2}\;(T<T_c) \qquad (4)

where: $g_{1/2}(z) = \sum_\ell z^\ell/\ell^{1/2}$ , which diverges as $z\to1$ . That divergence is the key: below $T_c$ the second term vanishes because $g_{1/2}(1)=\infty$ sits in its denominator, leaving the clean $T^{3/2}$ law. Both branches meet at $z=1$ with the same value, giving the peak. Evaluating the first term of (4) at $z=1$ gives the celebrated peak height, and matching the slopes of the two branches gives the cusp magnitude — the quantity that jumps:

\left.\frac{C_V}{Nk_B}\right|_{T_c} = \frac{15}{4}\frac{\zeta(5/2)}{\zeta(3/2)} \approx 1.925, \qquad \Delta\!\left(\frac{\partial C_V}{\partial T}\right) = \frac{27}{16\pi}\,[\zeta(3/2)]^2\,\frac{Nk_B}{T_c} \approx 3.665\,\frac{Nk_B}{T_c} \qquad (5)

where: $\Delta(\partial C_V/\partial T)$ — the jump in the temperature-slope of $C_V$ across $T_c$ . The peak $1.925$ sits comfortably above the classical equipartition value $1.5$ : near the transition, removing a degree of temperature costs more energy than usual because you are also draining the high-energy tail into the condensate. Finally, the flat isotherm shows up as a diverging response. The isothermal compressibility is tied to density fluctuations, and both blow up as $z\to1$ because $g_{1/2}$ does:

\kappa_T = -\frac{1}{V}\left(\frac{\partial V}{\partial P}\right)_{T,N} = \frac{1}{n^2 k_B T}\frac{g_{1/2}(z)}{g_{3/2}(z)} \;\xrightarrow{\,T\to T_c^+\,}\; \infty \qquad (6)

where: $\kappa_T$ — isothermal compressibility; $n = N/V$ . The divergence of $g_{1/2}(1)$ is the analytic statement of the flat coexistence isotherm: a gas that offers no resistance to compression has infinite compressibility.

Derivation AContinuity of C_V and the slope discontinuity at the transition

We want to prove the sharpest claim of the article: the heat capacity is continuous through $T_c$ — the same value $1.925\,Nk_B$ is reached from both sides — yet its temperature-slope is not, jumping by $\tfrac{27}{16\pi}[\zeta(3/2)]^2\,Nk_B/T_c$ . This is what makes the anomaly a cusp (a λ-shape) rather than either a smooth maximum or a step. The strategy is to approach $T_c$ from below and above separately, compute the slope on each side, and take the difference; the delicate part is the approach from above, where we must expand the divergent function $g_{1/2}(z)\sim\sqrt{\pi/(1-z)}$ near $z=1$ and use the number constraint to translate $(1-z)$ into $(T-T_c)$ . Watch the sign: the below-branch slope is positive (rising), the above-branch slope is negative (falling), so the two subtract to a positive jump.

Step 1. Below $T_c$ , use the clean single-term law. From (4), $C_V/Nk_B = \tfrac{15}{4}\tfrac{\zeta(5/2)}{\zeta(3/2)}(T/T_c)^{3/2}$ . Differentiate and evaluate at $T_c$ : $\left.\tfrac{1}{Nk_B}\tfrac{\partial C_V}{\partial T}\right|_{T_c^-} = \tfrac{45}{8}\tfrac{\zeta(5/2)}{\zeta(3/2)}\tfrac{1}{T_c} \approx 2.888/T_c$ . Why: the exponent $3/2$ pulls down a factor $3/2$ , and $\tfrac{15}{4}\cdot\tfrac32 = \tfrac{45}{8}$ ; the numerical value uses $\zeta(5/2)/\zeta(3/2)=0.5135$ .

Step 2. Above $T_c$ , first confirm continuity of the value. As $z\to1^+$ , $g_{1/2}(z)\to\infty$ , so the second term $-\tfrac94 g_{3/2}/g_{1/2}\to 0$ , and the first term $\to \tfrac{15}{4}\zeta(5/2)/\zeta(3/2) = 1.925$ — identical to the below-branch peak. Why: the divergence of $g_{1/2}$ is exactly what kills the correction term, so $C_V$ does not jump — only its derivative can.

Step 3. Above $T_c$ , extract the slope. Near $z=1$ write $g_{1/2}(z)\approx\sqrt{\pi/(1-z)}$ ; differentiate the number constraint $n\lambda_T^3 = g_{3/2}(z)$ to relate $(1-z)$ to $(T-T_c)$ , then differentiate the full $C_V$ . Carrying this through gives $\left.\tfrac{1}{Nk_B}\tfrac{\partial C_V}{\partial T}\right|_{T_c^+} \approx -0.777/T_c$ . Why: above $T_c$ the chemical potential begins to fall as $T$ rises, and the crossover toward the classical gas turns $C_V$ downward — hence the negative slope.

Step 4. Take the difference. The jump is $\Delta = \left(2.888 - (-0.777)\right)/T_c = 3.665/T_c$ in units $Nk_B$ . In closed form this equals $\tfrac{27}{16\pi}[\zeta(3/2)]^2$ . Because $C_V$ is a second derivative of the free energy and its slope is the first thing to jump, the discontinuity lives in the third derivative of $F$ — this is a third-order transition in the Ehrenfest classification. Why: the closed form comes from squaring the divergent-branch coefficient; numerically $\tfrac{27}{16\pi}(2.612)^2 = 3.665$ . *Verified: peak $1.925$ , below-slope $2.888$ , jump $3.665$ , above-slope $-0.777$ all reproduced symbolically and numerically in verify_thermodynamics-bec-transition.py (matches Pathria & Beale, 3rd ed., Ch. 7).*

Go deeper · Derivation B · 1 minZero latent heat — entropy is continuous across the transition

Derivation BZero latent heat — entropy is continuous across the transition

The pressure law (3) makes the ideal-gas BEC masquerade as a first-order coexistence — a flat isotherm is exactly what you see boiling water do. A first-order transition has a latent heat, $L = T_c\,\Delta S$ , released when the substance crosses the phase line. The goal here is to show that despite the disguise, the BEC transition has $L = 0$ : the entropy is perfectly continuous across $T_c$ . The route is to compute the entropy from $S = (U + PV - N\mu)/T$ , exploit $U = \tfrac32 PV$ and $\mu = 0$ below $T_c$ , evaluate the result on both branches at $T_c$ , and confirm they agree. The physical payoff is a subtle one: the condensate carries no entropy, so moving atoms into it costs nothing in disorder — there is no hidden heat of "condensation."

Step 1. Write the grand-canonical entropy. From $S = (U + PV - N\mu)/T$ and $U = \tfrac32 PV$ , the numerator is $U + PV = \tfrac52 PV$ , so $S = (\tfrac52 PV - N\mu)/T$ . Why: this is just the Euler relation for the grand ensemble; the $\tfrac52$ collects the energy and the $PV$ work term.

Step 2. Evaluate below $T_c$ . Here $\mu = 0$ and $PV = Nk_B T\,(T/T_c)^{3/2}\,\zeta(5/2)/\zeta(3/2)$ , so $S/Nk_B = \tfrac52\,\zeta(5/2)/\zeta(3/2)\,(T/T_c)^{3/2}$ . Only the normal fraction contributes; the condensate has $S_0 = 0$ . Why: the condensate is a single quantum state — its entropy is $k_B\ln 1 = 0$ — so it is thermodynamically silent.

Step 3. Match at $T_c$ . From below, $S/Nk_B|_{T_c^-} = \tfrac52\,\zeta(5/2)/\zeta(3/2) \approx 1.284$ . From above, the general formula $S/Nk_B = \tfrac52\,g_{5/2}(z)/g_{3/2}(z) - \ln z$ gives, at $z=1$ (so $\ln z = 0$ ), the same $1.284$ . Why: the $-\ln z$ term that distinguishes the branches vanishes exactly at $z=1$ , forcing continuity.

Step 4. Conclude. Since $S$ is continuous, $\Delta S = 0$ and therefore $L = T_c\,\Delta S = 0$ . The transition is continuous, not first order, even though its pressure law mimics coexistence. Why: the "coexistence" here is between the same phase at two densities of excited atoms, not two distinct phases with different entropies — there is nothing to release. *Verified: $S/Nk_B|_{T_c} = 1.284$ from both branches, confirming $\Delta S = 0$ (verify_thermodynamics-bec-transition.py; consistent with Landau–Lifshitz Part 1, §62).*

Limiting cases

Case	Result	Meaning
Condensed phase, $T < T_c$	$P = \dfrac{k_B T}{\lambda_T^3}\zeta(5/2)$ , $\;\dfrac{C_V}{Nk_B} = \dfrac{15}{4}\dfrac{\zeta(5/2)}{\zeta(3/2)}\!\left(\dfrac{T}{T_c}\right)^{3/2}$	Pressure independent of $V,N$ ; $C_V\propto T^{3/2}\to 0$ as $T\to0$ (third law satisfied).
At the transition, $T = T_c$	$\left.\dfrac{C_V}{Nk_B}\right\|_{T_c} = \dfrac{15}{4}\dfrac{\zeta(5/2)}{\zeta(3/2)} \approx 1.925$	Peak value, continuous but with a kinked slope; $\kappa_T\to\infty$ approaching from above.
Classical limit, $T \gg T_c$	$PV \to Nk_B T$ , $\;\dfrac{C_V}{Nk_B}\to\dfrac32$	Fugacity $z\to0$ ; recovers the ideal classical gas with leading quantum correction $PV = Nk_B T\,[1 - \tfrac{1}{2^{5/2}}n\lambda_T^3 + \cdots]$ .
Deep condensate, $T \to 0$	$N_0/N \to 1$ , $\;U,\,P,\,S,\,C_V \to 0$	Essentially all atoms in the ground state; the gas becomes thermodynamically empty.

Common trap“The heat capacity jumps discontinuously at T_c, and the latent heat measures it.”

Two classic errors, both about the kind of singularity. First, $C_V$ is continuous at $T_c$ — it reaches the same peak $1.925\,Nk_B$ from both sides. What is discontinuous is the slope $\partial C_V/\partial T$ , which jumps by $\tfrac{27}{16\pi}[\zeta(3/2)]^2\,Nk_B/T_c \approx 3.665\,Nk_B/T_c$ . The anomaly is a cusp, not a step. Second, there is no latent heat: the entropy is continuous ( $S/Nk_B = 1.284$ from both branches), so $L = T_c\,\Delta S = 0$ . The flat pressure isotherm makes the transition look first-order, but that is a masquerade — it is a continuous (third-order Ehrenfest) transition. Finally, do not confuse $C_V$ (which cusps) with $C_P$ (which genuinely diverges), and remember that it is the condensate, not the normal gas, that silently absorbs the saturated particles.

Worked examples

1 · The heat-capacity peak — enumerate the amplitude, then locate the cusp (confidence builder)

Compute the peak value $C_V/Nk_B$ at $T_c$ from the two zeta values, verify it exceeds the classical value, and find $C_V/Nk_B$ at $T = 0.8\,T_c$ .

Step 1. Gather the two amplitudes: $\zeta(3/2) = 2.6124$ and $\zeta(5/2) = 1.3415$ . Their ratio is $\zeta(5/2)/\zeta(3/2) = 1.3415/2.6124 = 0.5135$ .

Step 2. Peak value from (5): $C_V/Nk_B|_{T_c} = \tfrac{15}{4}\times 0.5135 = 3.75\times 0.5135 = \boxed{1.925}$ . This sits above the classical equipartition value $\tfrac32 = 1.5$ , so the gas is anomalously hard to heat near the transition.

Step 3. Below $T_c$ the law is $C_V/Nk_B = 1.925\,(T/T_c)^{3/2}$ . At $T = 0.8\,T_c$ : $(0.8)^{3/2} = 0.8\times\sqrt{0.8} = 0.8\times 0.8944 = 0.7155$ , so $C_V/Nk_B = 1.925\times 0.7155 = \boxed{1.377}$ .

Step 4. Sanity check the shape: at $T = 0.8\,T_c$ the value $1.377$ is below the classical $1.5$ , but climbing; it will overshoot $1.5$ just before $T_c$ and peak at $1.925$ exactly at $T_c$ . That overshoot-then-cusp is the entire λ-anomaly in three numbers: $1.377 \to 1.925 \to$ falling.

2 · Liquid helium-4 as an ideal Bose gas — how close does statistics alone get? (exam level)

Treat liquid $^4$ He at its measured number density $n = 2.18\times10^{28}\ \mathrm{m^{-3}}$ as a non-interacting Bose gas ( $m = 6.646\times10^{-27}\ \mathrm{kg}$ ). Predict $T_c$ and compare with the observed λ-point $T_\lambda = 2.17\ \mathrm{K}$ .

Step 1. The critical temperature (owned by aq-sm-bec-critical-temperature-condensate-fraction) is $k_B T_c = \dfrac{2\pi\hbar^2}{m}\left[\dfrac{n}{\zeta(3/2)}\right]^{2/3}$ . Compute the bracket: $n/\zeta(3/2) = 2.18\times10^{28}/2.6124 = 8.345\times10^{27}\ \mathrm{m^{-3}}$ .

Step 2. Its two-thirds power: $(8.345\times10^{27})^{2/3} = (8.345)^{2/3}\times10^{18} = 4.11\times10^{18}\ \mathrm{m^{-2}}$ .

Step 3. The prefactor: $\dfrac{2\pi\hbar^2}{m} = \dfrac{2\pi(1.0546\times10^{-34})^2}{6.646\times10^{-27}} = \dfrac{6.988\times10^{-68}}{6.646\times10^{-27}} = 1.051\times10^{-41}\ \mathrm{J\,m^2}$ .

Step 4. Combine: $k_B T_c = 1.051\times10^{-41}\times 4.11\times10^{18} = 4.32\times10^{-23}\ \mathrm{J}$ , so $T_c = 4.32\times10^{-23}/1.381\times10^{-23} = \boxed{3.13\ \mathrm{K}}$ .

Step 5. Compare: the ideal-gas prediction $3.13\ \mathrm{K}$ is astonishingly close to the observed $T_\lambda = 2.17\ \mathrm{K}$ — off by only ~40%, with zero interaction physics included. This near-miss is exactly what convinced London in 1938 that superfluidity is Bose–Einstein condensation, with interactions responsible for the remaining discrepancy (and for converting the finite cusp into helium's true logarithmic λ-divergence).

Origins · 1 minLambda Transition and how this was discovered

Lambda Transition

named 1930s · liquid helium-4 at 2.17 K

The lambda transition takes its name from the shape of the curve, not from a person: plotted against temperature, the heat capacity of liquid $^4$ He traces a sharp spike that looks like the Greek letter λ, with its tip at $T_\lambda = 2.17\ \mathrm{K}$ and $1\ \mathrm{atm}$ . Below it, helium becomes a superfluid — flowing without viscosity, climbing container walls, conducting heat anomalously. The connection to the ideal-gas calculation in this article is Fritz London's 1938 insight that the transition is a Bose–Einstein condensation blurred and sharpened by interactions: the ideal gas gives a finite cusp at $1.925\,Nk_B$ , while real helium diverges logarithmically because interactions place it in the 3D XY universality class. The lambda point is one of the most precisely studied critical points in physics; the 1992 Lambda Point Experiment aboard the Space Shuttle measured its heat-capacity exponent in microgravity to eliminate gravitational rounding, testing the renormalization-group prediction to parts per million.

History

1908

— Kamerlingh Onnes liquefies helium, opening the sub-4-kelvin regime where the transition lives.

1924

— Bose derives Planck's law from photon statistics; Einstein extends the counting to a gas of massive atoms and predicts condensation (1924–25).

1927

— Keesom and Wolfke identify two distinct liquid phases of helium, He I and He II, separated by the lambda line.

1932

— Keesom and Clusius measure the sharp λ-shaped heat-capacity anomaly at 2.17 K, giving the transition its name.

1938

— London proposes that the lambda transition of helium is a Bose–Einstein condensation; Tisza adds the two-fluid picture.

1938

— Kapitza, and independently Allen and Misener, discover superfluidity in He II.

1995

— Cornell and Wieman condense rubidium, and Ketterle condenses sodium, realizing near-ideal-gas BEC where this thermodynamics is measured directly.

Going further

Texts. Pathria & Beale, Statistical Mechanics, 3rd ed., Ch. 7 (Ideal Bose Systems) · Kardar, Statistical Physics of Particles, Ch. 7 · Landau & Lifshitz, Statistical Physics Part 1, §62 · Huang, Statistical Mechanics, 2nd ed., Ch. 12 · Reif, Fundamentals of Statistical and Thermal Physics, Ch. 9.

Provenance. All polylogarithm values are standard: $\zeta(3/2)\approx2.612$ , $\zeta(5/2)\approx1.341$ . A verifier should recompute: (i) the peak $C_V/Nk_B = \tfrac{15}{4}\zeta(5/2)/\zeta(3/2) = 1.925$ ; (ii) the slope jump $\tfrac{27}{16\pi}[\zeta(3/2)]^2 = 3.665$ in units $Nk_B/T_c$ ; (iii) the below-branch slope $\tfrac{45}{8}\zeta(5/2)/\zeta(3/2) = 2.888/T_c$ and above-branch slope $-0.777/T_c$ ; (iv) the entropy $\tfrac52\zeta(5/2)/\zeta(3/2) = 1.284$ from both branches (giving zero latent heat); (v) $U = \tfrac32 PV$ and $P_c\propto T^{5/2}$ ; (vi) the helium estimate $T_c = 3.13\ \mathrm{K}$ . All checks are executed in verify_thermodynamics-bec-transition.py (18 checks). Cross-references: Pathria & Beale Ch. 7, Kardar SPP Ch. 7, Landau–Lifshitz Part 1 §62.

Prerequisites. Bose-Einstein Condensation in the Ideal Gas · Critical Temperature and Condensate Fraction · Quantum Statistics of Ideal Gases: Occupation Numbers · Density of States for Free Particles. Leads to. Photon Gas and Planck Radiation Law · Degeneracy Pressure of Fermions.

Why is the transition called "third order" if there is a genuine condensate order parameter?

The Ehrenfest classification labels a transition by the lowest derivative of the free energy that is discontinuous. Here the first two derivatives (entropy $\sim \partial F/\partial T$ and heat capacity $\sim \partial^2 F/\partial T^2$ ) are continuous; the first thing to jump is $\partial C_V/\partial T$ , a third derivative of $F$ — hence "third order." Yet the transition simultaneously has a symmetry-breaking order parameter (the macroscopic ground-state occupation). This mismatch — a high Ehrenfest order coexisting with genuine condensation — is a peculiarity of the non-interacting limit. Real interactions push helium into a proper continuous transition with a logarithmically divergent $C_V$ . The modern language (order parameter, universality) is more informative than the Ehrenfest order for this system.

If the condensate has no energy, entropy, or pressure, how do we even know it is there?

Thermodynamically it announces itself indirectly. The condensate is the reason the normal gas thermodynamics becomes a function of temperature alone: once particles start draining into the ground state, $\mu$ locks at $0$ and $z$ locks at $1$ , which is precisely what makes $P$ , $U$ , and $S$ lose their density dependence below $T_c$ . So while the condensate carries no extensive quantities of its own, its presence is what pins the chemical potential — and that pinning produces every anomaly in the article. Directly, of course, cold-atom experiments image the condensate as a sharp central peak in time-of-flight, and helium reveals it through superfluid flow.

Why does the compressibility diverge only from above T_c, and not from below?

Below $T_c$ the system is already on the flat coexistence isotherm: the pressure is exactly $P_c(T)$ regardless of volume, so $\kappa_T$ is formally infinite throughout the condensed region — there is no restoring force at all. The interesting statement is the approach from above: as $T\to T_c^+$ at fixed density, the fugacity $z\to1^-$ , and $g_{1/2}(z)\sim\sqrt{\pi/(1-z)}$ diverges, dragging $\kappa_T$ to infinity right at the transition. So "diverges from above" describes how the normal gas softens as it nears condensation; below $T_c$ it is not so much diverging as already saturated at the coexistence pressure.

The base text says N_0/N = 1 − (T/T_c)^{3/2} for a box but (T/T_c)^3 for a trap. Why the different exponent?

The exponent is set by the effective density of states, which depends on the confining geometry. For a uniform box, the free-particle density of states goes as $g(\varepsilon)\propto\varepsilon^{1/2}$ , and the excited-state population scales as $T^{3/2}$ , giving $N_0/N = 1 - (T/T_c)^{3/2}$ . A harmonic trap changes the effective $g(\varepsilon)\propto\varepsilon^{2}$ (three trapping directions add powers of energy), so the excited-state population scales as $T^{3}$ and the condensate fraction becomes $1 - (T/T_c)^{3}$ . The mechanism — saturation of excited states, overflow into the ground state — is identical; only the power law changes with the trap. Nearly all cold-atom experiments use harmonic traps, so the cubic law is the one measured in practice.

How can a phase transition happen with no interactions at all — doesn't a transition require cooperative behaviour?

It does require cooperation, but here the cooperation is statistical, not dynamical. The particles never touch, yet they are correlated through the symmetrization of the many-boson wavefunction (owned microscopically by the QM track's account of identical particles). That indistinguishability forces the occupation numbers to be non-negative and lets any number of bosons share the ground state; the saturation of the excited states is a purely combinatorial fact about how many indistinguishable particles the finite-capacity levels can hold. So the "interaction" driving the transition is the quantum-statistical one built into the counting — which is exactly why BEC is the cleanest phase transition in physics and why it can be solved exactly.

Is C_P also just a cusp, or does something more dramatic happen to it?

$C_P$ is more singular than $C_V$ : it actually diverges at $T_c$ for the ideal gas, because holding the pressure fixed while crossing the transition forces the volume to respond through the diverging compressibility. The clean, textbook object is $C_V$ , whose cusp value $1.925\,Nk_B$ is finite and computable in closed form — that is why every derivation targets $C_V$ . When you see a diverging heat capacity quoted for BEC, check whether it is $C_P$ (divergent) or $C_V$ (finite cusp). Real helium's famous logarithmic $C_V$ divergence is a separate, interaction-driven effect that the ideal gas does not have.