Halo orbits — linearized motion near Lagrange points: Question bank

Quick symbol reminder so nothing here is unearned:

$(ξ, η, ζ)$ = small displacement from the Lagrange point along the $x$ (Sun–planet line), $y$ (in-plane sideways), and $z$ (out-of-plane) directions.
$c = \frac{1 - μ}{r _{1}^{3}} + \frac{μ}{r _{2}^{3}} > 1$ at a collinear point — a positive number measuring how strong the leftover gravity gradient is.
$λ$ = growth/oscillation rate you get from trying $e^{λ t}$ ; $Λ = λ^{2}$ .
$ω_{x y} = - Λ_{2}$ = planar oscillation frequency; $ω_{z} = c$ = out-of-plane oscillation frequency.

True or false — justify

A linear system near

L_{1}

has only oscillatory solutions

False — the in-plane quartic gives one positive

Λ

, so a real

\pm γ

pair exists; that saddle direction grows/decays exponentially alongside the oscillations.

Setting the unstable-mode amplitude to zero is a physical choice, not a mathematical trick

True in spirit — it means you launch exactly onto the center manifold with no component along the

+ γ

eigenvector, so nothing grows; physically it demands infinite precision, which is why real missions still need station-keeping.

The out-of-plane motion

\ddot{ζ} = - c ζ

is genuine simple harmonic motion

True — it has the exact form

\ddot{ζ} = - ω^{2} ζ

with

ω^{2} = c > 0

, the defining equation of Simple Harmonic Motion, so

ζ

bobs sinusoidally at

ω_{z} = c

In pure linear theory a halo orbit is a closed 3-D loop

False — generally

ω_{x y} \neq = ω_{z}

, so the in-plane wobble and out-of-plane bob never share a period and the path is a non-closing Lissajous; true halos need nonlinear frequency corrections to force

ω_{x y} = ω_{z}

The Coriolis terms

- 2 \overset{η}{˙}

+ 2 \dot{ξ}

come from the effective potential

Ω

False —

Ω

only produces the

\nablaΩ

forces (gravity + centrifugal); the Coriolis terms are velocity-dependent and appear separately because we sit in a rotating frame.

Because

L_{1}

and

L_{2}

host oscillating spacecraft, they are stable equilibria

False — oscillation lives on the center manifold, but the co-existing real eigenvalue

+ γ

makes them saddle-unstable; any tiny error along that direction blows up like

e^{γ t}

The cross-derivatives

Ω_{x y}, Ω_{x z}, Ω_{y z}

vanish at every Lagrange point

False — they vanish at the collinear points

L_{1}, L_{2}, L_{3}

by the on-axis symmetry; at the triangular

L_{4}, L_{5}

points

Ω_{x y} \neq = 0

and the analysis is different.

Larger

c

means a more strongly unstable collinear point

True —

Λ_{1}

(the positive root, hence

γ

) grows with

c

, so a bigger gravity-gradient number means faster exponential escape and harder station-keeping.

Spot the error

"Since

Ω

has a maximum-like curvature, the point is a potential well and objects fall into it."

The curvatures have mixed signs:

U_{xx} = 1 + 2 c > 0

but

U_{z z} = - c < 0

, so it is a saddle of the effective potential, not a well — you fall in along some axes and out along others.

ω_{x y} = Λ_{2}

since

Λ_{2}

is the oscillation root."

Wrong sign —

Λ_{2} < 0

, so

Λ_{2}

is imaginary; the real frequency is

ω_{x y} = - Λ_{2}

, taking the magnitude of the negative root.

"We linearize by keeping the constant term of the Taylor expansion of

\nablaΩ

The constant term is

\nablaΩ

evaluated at the point, which is zero by definition of an equilibrium; linearization keeps the first-derivative (linear-in-displacement) term.

"The characteristic equation is quadratic in

λ

, giving two roots."

It is quartic in

λ

(degree 4); it becomes quadratic only after substituting

Λ = λ^{2}

, and each

Λ

still yields two

λ

values, so four roots total.

"Because both primaries pull inward, the leftover force near

L_{1}

is purely restoring."

Gravity plus centrifugal nearly cancel, and the leftover is restoring along some directions but repelling along others — that anti-spring behaviour is exactly the

+ γ

instability.

"The amplitude ratio

κ

makes the in-plane path a circle."

κ \approx 2.9

means the

η

amplitude is ~2.9× the

ξ

amplitude, so the planar path is an ellipse elongated along $y$ , not a circle.

"Halo orbits are found by solving

\ddot{ζ} = - c ζ

alone."

That equation only fixes the out-of-plane bob; a halo also needs the in-plane

(ξ, η)

motion and the nonlinear closing condition

ω_{x y} = ω_{z}

— the

ζ

equation is one ingredient, not the whole recipe.

Why questions

Why does the

ζ

equation decouple from the

ξ, η

equations?

At a collinear point the cross-derivatives

Ω_{x z} = Ω_{y z} = 0

, so

z

never appears in the in-plane equations and vice versa; this clean split is what makes the out-of-plane frequency independently readable.

Why do we bother linearizing instead of solving the full CR3BP?

The full Circular Restricted Three-Body Problem is nonlinear with no closed-form solution; near an equilibrium the displacement is small, so the linear term dominates and gives us exact exponentials and sinusoids we can read off directly.

Why must the two in-plane roots

Λ_{1}, Λ_{2}

have opposite signs at a collinear point?

Their product equals

(1 + 2 c) (1 - c)

, which is negative when

c > 1

; a negative product forces one positive and one negative root, guaranteeing the saddle-plus-oscillation structure.

Why does linear theory give Lissajous but not halos?

Linear theory fixes

ω_{x y}

and

ω_{z}

as independent numbers that are generally unequal, so the two motions never re-sync; only nonlinear amplitude-dependent frequency shifts can drag them into the 1:1 resonance that closes the loop.

Why do collinear-point missions need active Station-keeping and Orbit Maintenance?

The real eigenvalue

+ γ

amplifies any residual error along the unstable direction exponentially, so without periodic corrections the spacecraft drifts off the center manifold and escapes.

Why is

ω_{z} = c

rather than something involving

μ

and

r

explicitly?

Because

c

already packages the gravitational geometry (

c = \frac{1 - μ}{r _{1}^{3}} + \frac{μ}{r _{2}^{3}}

); the out-of-plane restoring coefficient

U_{z z} = - c

makes

ω_{z}^{2} = c

automatically.

Why can we relate this whole topic to a spring–mass system at all?

Each decoupled or diagonalized direction reduces to

\overset{q}{¨} = k q

; where

k < 0

it is a restoring spring (oscillation) and where

k > 0

it is an anti-spring (exponential runaway), the two building blocks of the linearized dynamics.

Edge cases

What happens to the analysis if

c = 1

exactly?

Then

(1 - c) = 0

, the product

(1 + 2 c) (1 - c) = 0

, so one root

Λ

becomes zero — a degenerate borderline case where the oscillation frequency vanishes and the clean saddle-center split breaks down.

What if you start with a component along the unstable eigenvector, however tiny?

It grows like

e^{γ t}

and eventually dominates, carrying the spacecraft off along the unstable manifold (the launch pad for Invariant Manifolds and Low-Energy Transfers); nothing bounded survives without correction.

What does zero amplitude in every mode (

A_{x} = A_{z} = 0

) describe?

The spacecraft sitting exactly at the Lagrange point with zero velocity — the trivial equilibrium solution; any nonzero mode amplitude is what turns it into an orbit.

What is the limiting shape when

A_{z} \to 0

but

A_{x} \neq = 0

A purely planar oscillation — the elongated in-plane ellipse with no out-of-plane bob, a planar Lyapunov orbit rather than a 3-D halo.

What happens at the triangular points

L_{4}, L_{5}

where cross-terms don't vanish?

The neat diagonal

U_{xx}, U_{y y}, U_{z z}

form fails, the eigenvalue structure changes, and (for small enough

μ

) those points can actually be linearly stable — the opposite of the collinear saddles.

What if the discriminant

(2 - c)^{2} - 4 (1 + 2 c) (1 - c)

were negative?

You would get complex

Λ

and thus complex-conjugate quartets of

λ

(spiralling growth); but at collinear points

c > 1

keeps this discriminant positive, so this case does not arise there — worth knowing it is excluded by the geometry.

Recall Self-test before you leave

Name the two eigenvalue types near $L_{1}$ and what each means physically. ::: A real pair $\pm γ$ (unstable saddle, exponential escape/approach) and an imaginary pair $\pm i ω_{x y}$ (bounded planar oscillation) — plus the decoupled out-of-plane $\pm i ω_{z}$ . State the single condition that turns a Lissajous into a halo. ::: The frequency match $ω_{x y} = ω_{z}$ , achievable only through nonlinear amplitude-dependent corrections.