Landau Damping

Definition

Landau damping is the decay of a plasma wave caused by energy transfer to particles moving near the wave's phase velocity — with no collisions involved.

In 1946 Lev Landau showed that an electron plasma wave in a collisionless plasma still loses amplitude over time. This was startling: with no collisions, what could possibly turn the ordered, oscillating energy of the wave into something else? The answer is the resonant particles — the small population of electrons whose speed happens to match the wave's phase velocity, v ≈ ω/k.

Those particles "surf" the wave. The damping rate is governed by one quantity: the slope of the velocity distribution evaluated right at the phase velocity.

γ ∝ ( ∂f/∂v )  evaluated at  v = ω/k

If that slope is negative (the normal case for a thermal plasma), the wave damps. If it is positive, the wave grows — an instability.

How it works

The cleanest intuition is the surfer analogy. Imagine ocean waves rolling in and a crowd of swimmers at different speeds:

• A swimmer moving slightly slower than a wave (v < ω/k) gets caught on the back of the crest and pushed forward — the wave gives energy to the swimmer.
• A swimmer moving slightly faster than the wave (v > ω/k) climbs up the front and is held back — the swimmer gives energy to the wave.
• Swimmers far from the wave speed barely interact; they just bob up and down and average out.

So the only particles that exchange net energy with the wave are the resonant ones, clustered around v = ω/k. Whether the wave gains or loses energy depends entirely on whether there are more slightly-slower particles (gainers) or more slightly-faster particles (losers).

For a thermal, bell-shaped (Maxwellian) distribution, the curve is falling on the high-velocity side. At any phase velocity above the bulk, there are always a few more particles just below ω/k than just above it. More gainers than losers ⇒ net energy out of the wave ⇒ damping. That asymmetry — encoded as the negative slope ∂f/∂v < 0 — is the whole mechanism.

A worked example with numbers

Take a laboratory electron plasma with density n = 10¹⁶ m⁻³ and electron temperature T_e = 10 eV.

Step 1 — plasma frequency. The plasma frequency is

ω_p = √( n e² / (ε₀ m_e) ) ≈ 5.64 × 10⁴ · √(n_cm⁻³) rad/s
n = 10¹⁶ m⁻³ = 10¹⁰ cm⁻³  →  ω_p ≈ 5.6 × 10⁹ rad/s

Step 2 — Debye length and thermal speed.

v_th = √(k_B T_e / m_e) ≈ 1.33 × 10⁶ m/s   (at 10 eV)
λ_D  = v_th / ω_p ≈ 1.33e6 / 5.6e9 ≈ 2.4 × 10⁻⁴ m  (≈ 0.24 mm)

Step 3 — pick a wavelength. Say the wave has kλ_D = 0.3, comfortably in the weakly-damped regime. The Bohm–Gross dispersion relation gives the real frequency:

ω² ≈ ω_p² ( 1 + 3 k²λ_D² )  →  ω/ω_p ≈ √(1 + 3·0.09) ≈ 1.13

Step 4 — the damping rate. Landau's Maxwellian result is

γ / ω_p ≈ −√(π/8) · (1 / (kλ_D)³) · exp( −1/(2 k²λ_D²) − 3/2 )

Plug in kλ_D = 0.3:

1/(2·0.09) = 5.56 ,  +1.5  →  exp(−7.06) ≈ 8.6 × 10⁻⁴
(1/0.027) ≈ 37.0 ,  √(π/8) ≈ 0.627
γ/ω_p ≈ −0.627 · 37.0 · 8.6e-4 ≈ −0.020

So γ ≈ −0.02 ω_p ≈ −1.1 × 10⁸ s⁻¹. The wave amplitude falls as e^(γt), dropping to 1/e of its value in about 1/|γ| ≈ 9 ns — and zero collisions occurred. Push the wavelength shorter to kλ_D = 0.5 and the exponent jumps, sending |γ|/ω_p up to ~0.15: the wave now barely completes one oscillation before it is gone.

Regimes and variants

Variant	Resonant species	Sign of ∂f/∂v at ω/k	Result
Electron Landau damping	Electrons near ω/k	Negative (Maxwellian)	Langmuir wave damps
Ion Landau damping	Ions near ω/k of ion-acoustic wave	Negative	Strong when T_e ≈ T_i
Bump-on-tail instability	Beam electrons	Positive (rising slope)	Wave grows — inverse Landau damping
Nonlinear / trapping regime	Trapped resonant particles	Flattened locally	Damping saturates; BGK modes form
Magnetized (transit-time)	Particles resonant with parallel phase speed	Negative	Damps Alfvén / kinetic waves
Gravitational (galaxies)	Stars near pattern speed	Negative	Spiral density waves damp the same way

The same mathematics that damps an electron plasma wave appears, remarkably, in stellar dynamics: a spiral density wave in a galactic disk can be Landau-damped by stars whose orbital frequency resonates with the wave's pattern speed. The slope of the stellar distribution function plays the role of ∂f/∂v.

Weak vs strong damping

The single dimensionless control parameter is kλ_D — the wavelength measured in Debye lengths.

kλ_D	Phase velocity vs v_th	Resonant particles	|γ|/ω_p (approx)
0.1	≈ 10 v_th (far out on tail)	Vanishingly few	~10⁻⁸ (essentially undamped)
0.2	≈ 5 v_th	Very few	~10⁻⁴
0.3	≈ 3.3 v_th	Few	~0.02
0.4	≈ 2.5 v_th	Some	~0.07
0.5	≈ 2 v_th	Many	~0.15
≥ 1.0	≈ v_th (in the bulk)	Most particles	≳ 1 (overdamped — no real wave)

The pattern: long-wavelength waves have a phase velocity far out on the tail of the distribution where almost no particles live, so they hardly damp. Short-wavelength waves have a phase velocity down in the thermal bulk, where the slope ∂f/∂v is steep and there are plenty of resonant particles — so they damp away within an oscillation or two. This is why a plasma supports clean oscillations only at long wavelengths.

Derivation sketch — why Landau and not Vlasov

Vlasov (1938) wrote the collisionless kinetic equation and Fourier-analyzed it in both space and time. The resulting dispersion relation contains a singular integral:

1 + (ω_p²/k²) ∫ (∂f₀/∂v) / (v − ω/k) dv = 0

The denominator blows up exactly at the resonant velocity v = ω/k. Vlasov took a principal-value of this integral and found purely real frequencies — no damping. Landau's insight (1946) was that the problem is not a steady state but an initial-value problem: you specify a perturbation at t = 0 and ask how it evolves. Solving it by Laplace transform in time forces a specific contour around the pole. The contour picks up the imaginary part of the integral:

γ = (π/2) (ω_p²/k²) (ω/ |k| · something) · ∂f₀/∂v |_(v = ω/k)

The crucial fact is that γ is directly proportional to ∂f₀/∂v at the phase velocity. A falling distribution (negative slope) gives γ < 0 (damping); a rising one gives growth. The bulk of the distribution barely matters — only the local slope at ω/k does. That is the entire physical content, dressed in contour integration.

Where Landau damping shows up

• Magnetic-confinement fusion. Radio-frequency heating schemes (lower-hybrid, ion-cyclotron, electron-cyclotron) deliberately launch waves whose phase velocity sits where there are resonant particles, so the wave Landau-damps and dumps its energy as heat exactly where you want it.
• The solar wind. The wind is so dilute that collisions are negligible over astronomical distances; Landau and transit-time damping are leading candidates for how turbulent fluctuations heat the corona and solar wind to millions of kelvin.
• Laser–plasma interaction. In inertial-confinement fusion and laser wakefield accelerators, Landau damping limits how large an electron plasma wave can grow and how cleanly it accelerates electrons.
• Galactic dynamics. Spiral density waves and the response of stellar disks to perturbations are Landau-damped by stars resonating with the wave's pattern speed.
• Plasma echoes. Because the process is reversible, two pulses launched at the right times produce a spontaneous third wave — direct experimental proof that the "damped" energy was stored in particle phases, not thermalized.
• Diagnostics. Measuring the damping rate of a launched wave is a way to infer the slope of the electron distribution — a probe of non-thermal tails.

Common pitfalls and misconceptions

• "Damping must mean collisions / heat / friction." No. Landau damping is purely collisionless. Energy is conserved and goes into ordered particle motion; in principle the process is reversible (plasma echoes prove it).
• "The whole distribution does the damping." Only the resonant particles near v = ω/k matter. The damping rate depends on the local slope ∂f/∂v there, not on the bulk.
• "It always damps." If the distribution has a positive slope at the phase velocity (a beam or bump-on-tail), the wave grows — inverse Landau damping, the engine behind beam–plasma instabilities.
• "It's a linear effect that lasts forever." At finite amplitude the resonant particles get trapped in the wave's potential wells, flatten the local slope, and damping shuts off (saturation, BGK modes). The clean exponential is a small-amplitude result.
• "Vlasov already had it." Vlasov's steady-state Fourier treatment missed it entirely. Landau's initial-value (Laplace-transform) treatment is what selects the damped contour. The mathematics, not the physics, is the subtle part.
• "Phase velocity is the particle velocity." The wave's phase velocity ω/k is generally much larger than the thermal speed. The resonant particles are a rare tail population, which is exactly why long-wavelength waves are so weakly damped.