Bose-Einstein Condensate

When the de Broglie wavelength outgrows the spacing

Atoms in a dilute classical gas are like billiard balls — well-separated point objects with momenta thermally distributed around √(k_B T m). Each atom has an associated thermal de Broglie wavelength λ_dB = h/√(2π m k_B T). At room temperature this length is a tiny fraction of the inter-atomic spacing — the wavefunctions of adjacent atoms barely overlap, and quantum statistics is irrelevant.

Cool the gas down. λ_dB grows as 1/√T. Eventually the de Broglie wavelength becomes comparable to the inter-atomic spacing n^(−1/3). At that point the wavefunctions overlap, the particles are no longer distinguishable, and quantum statistics matters. For bosons (integer-spin particles) the statistics is symmetric under exchange; for fermions (half-integer) it is antisymmetric. The boson case allows arbitrarily many atoms in the same single-particle state, and Bose-Einstein statistics says they prefer to crowd in.

The condition for condensation is essentially "phase-space density of order one":

n λ_dB³ ≈ ζ(3/2) ≈ 2.612

Solving for T gives the critical temperature:

T_c = (h² / 2π m k_B) · (n / ζ(3/2))^(2/3)

Below T_c the Bose-Einstein distribution can no longer accommodate all the atoms in excited states. Any additional atom must go into the single-particle ground state, and the ground-state population N₀ jumps to a macroscopic value:

N₀ / N = 1 − (T / T_c)³   (for a uniform 3D ideal gas)

At T = 0 every atom is in the ground state and the condensate fraction is 100 %. The whole many-body system is described by a single complex-valued wavefunction Ψ(r,t), the "macroscopic wavefunction" or order parameter, with |Ψ|² giving the local condensate density.

Worked example: ⁸⁷Rb at JILA densities

Take the canonical Cornell-Wieman conditions: ⁸⁷Rb (m = 87 × 1.66 × 10⁻²⁷ kg = 1.44 × 10⁻²⁵ kg), peak density n ≈ 2.5 × 10¹⁸ m⁻³ in a magnetic trap. Plug into the T_c formula:

h² / (2π m k_B) = (6.626e−34)² / (2π · 1.44e−25 · 1.38e−23)
                ≈ 3.51 × 10⁻¹⁸ K · m²
n / ζ(3/2)      = 2.5e18 / 2.612 ≈ 9.57 × 10¹⁷ m⁻³
(n/ζ(3/2))^(2/3) ≈ 9.71 × 10¹¹ m⁻²
T_c ≈ 3.51e−18 · 9.71e11 ≈ 3.4 × 10⁻⁶ K = 3.4 µK ?

That's roughly the BEC threshold for that density. The original 1995 cloud actually condensed at about 170 nK because the working density was a few times lower. To get T_c into the 100 nK ballpark you tune the density and the trap geometry; tighter traps with higher central density reach higher T_c, looser ones lower.

The thermal de Broglie wavelength at T_c is roughly:

λ_dB(T_c) = h / √(2π m k_B T_c)
          ≈ 6.626e−34 / √(2π · 1.44e−25 · 1.38e−23 · 3.4e−6)
          ≈ 0.5 µm

Half a micron — about a wavelength of red light. The atoms are spaced ~0.7 µm apart at this density, so their wavefunctions overlap by exactly the right amount.

The two-stage cooling pipeline

You cannot reach T_c with a refrigerator; the lowest achievable lab temperatures with conventional methods are millikelvin in dilution refrigerators, far above 100 nK. The BEC pipeline therefore uses a two-stage cooling sequence specific to neutral atoms.

Stage 1: laser cooling. A magneto-optical trap (MOT) intersects six counter-propagating laser beams (one pair per axis) tuned just below an atomic resonance. By the Doppler-cooling mechanism, atoms moving toward a beam preferentially scatter that beam's photons, decelerating in 1–10 ms. Coupled with a magnetic field gradient that adds a position-dependent restoring force, the MOT holds 10⁹ atoms at 50–200 µK. The recoil limit (the energy of one absorbed-and-emitted photon) caps Doppler cooling around 100 µK for alkalis; sub-Doppler techniques like polarization-gradient cooling push it down to a few µK.

Stage 2: evaporative cooling. Transfer the laser-cooled atoms to a purely conservative trap (magnetic or far-detuned optical), where there is no spontaneous emission. Apply an RF or microwave knife that flips the spins of atoms in the high-energy tail, ejecting them. The remaining atoms rethermalize at lower temperature; lower the cut and repeat. After 10–30 seconds of evaporation the cloud has lost ~99 % of its atoms but reached 100 nK. Forcing the cloud through T_c by lowering the cut just below the condensation point is what produces the BEC.

The whole sequence runs at a few-second cycle in a typical lab. A modern compact BEC machine fits on an optical table; the original JILA experiment occupied a room, but used the same components.

How you know you have a BEC

The diagnostic signature is bimodal time-of-flight imaging. Switch off the trap. The cloud expands ballistically; atoms in thermal momentum states fly out isotropically while atoms in the condensate keep their narrow ground-state momentum distribution. After 10–30 ms of free fall, image the cloud by absorption or fluorescence. A pure thermal cloud gives a single Gaussian; below T_c you see a sharp, narrow peak (the condensate) sitting on top of a wider Gaussian thermal pedestal. As you lower temperature further, the thermal pedestal shrinks and the central peak grows. At T = 0 ideally only the narrow peak remains.

A condensate is also coherent. Two BECs released simultaneously interfere, producing high-contrast matter-wave fringes — the famous Andrews et al. 1997 image is the matter-wave equivalent of a double-slit experiment. The fringe spacing depends on the relative-momentum-spread between the two condensates and is independent of how the two were prepared.

A periodic table of cold-atom condensates

Species	Mass (u)	Year first BEC	Typical T_c	Typical N atoms	Notable feature
⁸⁷Rb	87	1995 (Cornell, Wieman)	~170 nK	10⁵–10⁶	Workhorse alkali, simple optics at 780 nm
²³Na	23	1995 (Ketterle)	~600 nK	10⁶–10⁷	Strong cooling lasers (yellow), large condensates
⁷Li	7	1995 (Hulet)	~300 nK	10³–10⁵	Negative scattering length → unstable BEC ("bosenova")
¹H (atomic hydrogen)	1	1998 (Kleppner)	~50 µK	10⁹	First BEC proposed; cooled by surface evaporation
⁵²Cr	52	2005 (Pfau)	~700 nK	10⁵	Strong magnetic dipoles → dipolar quantum gases
¹⁶⁴Dy / ¹⁶⁸Er	164/168	2011/2012	~100 nK	10⁵	Largest magnetic dipoles, supersolid phases
⁸⁷Rb (microgravity)	87	2018 (CAL on ISS)	~100 nK	10⁴	Free fall extends interrogation time 10×

The earliest BECs were the alkalis with simple electronic structure and convenient laser wavelengths. Later experiments expanded to magnetic atoms (Cr, Dy, Er) for dipolar interactions, to fermionic species (⁶Li, ⁴⁰K) for BEC-BCS studies, and to molecules (formed from atom pairs) for chemistry-relevant condensates.

Where BECs show up

• The 2001 Nobel Prize. Eric Cornell, Carl Wieman and Wolfgang Ketterle shared the prize "for the achievement of Bose-Einstein condensation in dilute gases of alkali atoms, and for early fundamental studies of the properties of the condensates." The 1995 experiments are now the canonical reference for new states of matter.
• Atom interferometry and precision measurement. A BEC has a single coherent phase you can manipulate with mirror and beam-splitter laser pulses. BEC interferometers in microgravity (NASA's Cold Atom Lab on ISS, 2018–) and on rocket flights are testing the Equivalence Principle to ~10⁻¹³.
• Quantum simulation. Loaded into optical lattices, BECs realize textbook condensed-matter Hamiltonians (Bose-Hubbard, Hubbard-fermion) without disorder. Greiner et al. 2002 first imaged the superfluid-to-Mott-insulator transition; modern simulators reach 100+ sites in 2D.
• Vortex lattices and topological quantum matter. Stir a condensate and quantized vortices nucleate, arranging into Abrikosov-style triangular lattices. The same lattices appear in spin-orbit-coupled BECs and reveal topological phases otherwise hard to study.
• Slow light and Bose stimulation. A BEC can be coupled to electromagnetic-induced transparency to slow light to a few m/s, even temporarily storing a photon's quantum state in atomic excitations and releasing it. Hau et al. 1999 reported 17 m/s; later work has stopped light entirely on millisecond timescales.

Interactions and the Gross-Pitaevskii equation

The ideal-gas formulas above ignore atom-atom interactions. Real atoms scatter off each other with a low-energy s-wave scattering length a, typically 5 nm for ⁸⁷Rb. At dilute-gas densities, the mean-field interaction term is small but nonzero and qualitatively important. The condensate wavefunction Ψ(r,t) satisfies the Gross-Pitaevskii equation:

iℏ ∂Ψ/∂t = [ −(ℏ²/2m) ∇² + V_trap(r) + g |Ψ|² ] Ψ

where g = 4π ℏ² a / m

This is a nonlinear Schrödinger equation. The g|Ψ|² term gives the condensate finite size in a trap (Thomas–Fermi profile), supports solitons (in 1D), and produces the speed of sound c_s = √(g n_0 / m). At unitarity (a → ∞ near a Feshbach resonance) the gas becomes strongly interacting and the GP equation breaks down; that regime is where the BEC-BCS crossover lives and where exact ground-state energies are known only numerically.

Variants and extensions

• Fermi degenerate gas. Cooling a dilute gas of fermionic atoms below their Fermi temperature T_F produces a degenerate Fermi gas — every momentum state up to k_F is filled, and there is no condensation but the gas becomes "quantum cold" by the Pauli exclusion principle. Realized in ⁴⁰K (DeMarco-Jin 1999) and ⁶Li.
• BEC-BCS crossover. Tune the inter-atom scattering length of a fermion gas across a Feshbach resonance. On the molecular side, atoms pair into bosonic dimers that BEC. On the BCS side, weakly attractive pairs form a Cooper-pair superfluid. The crossover regime is the cleanest analogue of high-Tc superconductivity available in the lab.
• Spinor and dipolar condensates. Optically-trapped atoms with internal spin freedom (e.g. ⁸⁷Rb F=2) form spinor BECs with multiple order parameters. Magnetically dipolar species (Cr, Dy, Er) add long-range anisotropic interactions that produce supersolid stripes and droplet phases.
• Polariton condensates. Strongly coupled cavity-photon / quantum-well exciton hybrids in semiconductor microcavities can condense at temperatures up to ~100 K because the polaritons are very light. Not a strict BEC of structureless bosons but the same macroscopic phase coherence.
• Microgravity and matter-wave space experiments. NASA's Cold Atom Lab (ISS, 2018) and the German Bundeswehr-led MAIUS rocket flights produce BECs in free-fall, removing trap-induced phase shifts and enabling longer interrogation times for fundamental-physics tests.

Common pitfalls

• Confusing BEC with low temperature alone. A cold thermal gas is not a BEC; you need the phase-space density n λ_dB³ to cross ζ(3/2) ≈ 2.6. Ultracold mercury-vapour at high temperature can have higher n than a BEC and not condense, while a sparse but cold rubidium cloud does.
• Treating helium-4 as a textbook BEC. Helium-4 is a strongly-interacting liquid, not a dilute gas. Even at T = 0 only ~10 % of its atoms are in the ground state — the rest are pushed out by interactions. The other 90 % still participate in superfluidity through their phase coherence, but the simple condensate fraction formulas don't apply.
• Assuming all bosons can BEC. Photons in a vacuum cannot Bose-condense because their number is not conserved (they freely radiate away). Photons trapped in a dye-filled microcavity at room temperature have been condensed (Klaers et al. 2010), but only because the dye keeps a thermal photon-number reservoir.
• Misreading the bimodal time-of-flight image. A narrow peak by itself doesn't prove a BEC; you also need to verify it is anisotropic in the trap-aspect ratio (condensates expand inversely to the trap, while thermal clouds expand isotropically) or use phase-coherent measurements like absorption interferometry.
• Forgetting interactions in the GP equation. The ideal-gas T_c is shifted by interactions. The Thomas-Fermi profile is broader than the harmonic-oscillator ground state because the interaction term g|Ψ|² pushes density outward. Quantitatively comparing experiments to the ideal-gas formulas without this correction gives systematic errors of ~10–30 %.