Skip to content

The Closure Problem

Source: PHY653B Ch. 1

Intuition

Everything in plasma physics descends from one equation for the distribution function \(f(\mathbf{x},\mathbf{v},t)\). Solving it in six dimensions plus time is out of reach for most problems, so the standard move is to take moments — integrate over velocity to get density, then momentum, then energy. The trouble is structural and it never goes away: every moment equation contains the next moment.

Why it never terminates

Multiply the Boltzmann equation by \(1, \mathbf{v}, \tfrac12 v^2, \dots\) and integrate:

\[\frac{\partial n}{\partial t} + \nabla\cdot(n\mathbf{u}) = 0 \qquad\text{— density needs } \mathbf{u}\]
\[mn\frac{d\mathbf{u}}{dt} = qn(\mathbf{E} + \mathbf{u}\times\mathbf{B}) - \nabla\cdot\overleftrightarrow{P} \qquad\text{— momentum needs the pressure tensor}\]
\[\frac{\partial}{\partial t}\left(\tfrac32 p\right) + \dots = -\nabla\cdot\mathbf{q} \qquad\text{— energy needs the heat flux}\]

The \(n\)-th moment equation always contains the \((n+1)\)-th. The hierarchy is infinite, and no amount of algebra closes it, because the information genuinely lives in the shape of \(f\), and moments are a lossy summary of that shape.

To get a solvable system you must assert something about a high moment that the equations do not give you. That assertion is a closure, and it is a physical assumption wearing the costume of an algebraic step.

Three closures you will meet

Closure Assertion Valid when Cost of being wrong
Cold \(\overleftrightarrow{P} = 0\) thermal speed ≪ phase speed misses all pressure-driven waves, all Landau damping
Isothermal / adiabatic \(p = nT_0\), or \(p n^{-\gamma}\) = const strong collisions, or fast compared with heat transport wrong \(\gamma\) gives wrong sound speed
Chapman–Enskog expand \(f\) about a Maxwellian in the Knudsen number mean free path ≪ gradient scale breaks down in shocks and collisionless regions

Chapman–Enskog is the honest one: rather than guessing, it expands \(f = f^{(0)} + \varepsilon f^{(1)} + \dots\) about local equilibrium, and the first-order term derives the pressure tensor and heat flux — giving Navier–Stokes with transport coefficients that are predictions rather than parameters. This is exactly the machinery that makes lattice Boltzmann work, where the emergent viscosity \(\nu = c_s^2(\tau - \tfrac12)\) is a Chapman–Enskog result you can measure to three digits.

The two escapes

If closure is a lie, you have two options, and this course takes both:

  • Do not close. Keep all of \(f\): the Vlasov solver on a phase-space grid, or PIC with markers. You pay in dimensionality or in noise, and you get kinetic physics — Landau damping, beam instabilities, trapping — that no closure reproduces.
  • Close, but know where. Use MHD where it applies, and be explicit about the regime. Fluid models are not inferior; they are correct in their domain and hugely cheaper. The sin is not closing, it is closing silently.

Common mistakes

  • Treating a closure as a mathematical step. It is a physical claim about \(f\), and it should be stated and justified in your methods section, not buried.
  • Assuming more moments is always better. A 10-moment model still needs a closure at the 11th, and can be worse behaved than a well-chosen 5-moment one.
  • Expecting a fluid code to show kinetic effects. If your closure discarded the resonant particles, no amount of resolution will produce Landau damping. Resolution does not fix a model error — a distinction the verification/validation framework exists to enforce.

Knowledge graph position

Prerequisites: kinetic theory, Vlasov equation (PC368). Leads to: every method in this course — the choice of method is the choice of how to handle closure.

Quiz

Q1 (conceptual). Why can the moment hierarchy never be closed by algebra alone?

Answer

Because the \(n\)-th moment equation contains the \((n+1)\)-th as a structural consequence of the \(v\cdot\nabla f\) term. Truncating requires supplying information about the shape of \(f\) that the moments have already discarded — a physical assumption, not an algebraic manipulation.

Q2 (conceptual). A cold-plasma model is used to study a wave whose phase velocity is close to the electron thermal speed. What goes wrong?

Answer

The cold closure sets \(\overleftrightarrow{P} = 0\), which removes the resonant particles near \(v \approx \omega/k\) entirely. The model cannot produce Landau damping, so it predicts an undamped wave where the real one may damp strongly. This is a model error: refining the grid makes the wrong answer more precise.

Q3 (MCQ). Chapman–Enskog differs from the other closures because it:

  • (a) requires no assumptions
  • (b) derives the transport coefficients from a controlled expansion rather than asserting them
  • (c) works in collisionless plasma
  • (d) closes at infinite order
Answer

(b). It still assumes small Knudsen number, but within that assumption the viscosity and heat conductivity come out as predictions — which is exactly why LBM's \(\nu = c_s^2(\tau-\frac12)\) can be checked against a measurement.