Visual walkthrough — Halo orbits — linearized motion near Lagrange points

Prerequisites we lean on (skim if rusty): Simple Harmonic Motion, Circular Restricted Three-Body Problem, Lagrange Points L1–L5.

---

Step 1 — The rotating frame flattens gravity into a hill-and-valley

WHAT. Instead of watching two big bodies whirl around, we ride along with them — we spin the whole camera at the same rate $\omega$ so the two primaries sit frozen. Take $\omega =1$ in normalized units (we pick our clock so one radian of rotation is one time-unit). In this frozen picture, all the forces on a small spacecraft can be packed into one landscape called the effective potential $\Omega$.

WHY. A ball rolling on a landscape is the simplest mental model we have: it rolls downhill. If we can turn "gravity plus rotation" into a single surface, then "where do forces balance?" becomes the easy question "where is the ground flat?" — that is, where $\nabla \Omega =0$.

PICTURE.

$$\Omega (x,y,z)=\underset{\text{centrifugal bowl (pushes out)}}{\underbrace{\frac{1}{2}(x^2+y^2)}}+\underset{\text{pull of big body}}{\underbrace{\frac{1-\mu }{r_1}}}+\underset{\text{pull of small body}}{\underbrace{\frac{\mu }{r_2}}}$$

• $x,y,z$ ::: the spacecraft's position in the frozen (rotating) frame.
• $\mu$ ::: the fraction of total mass in the smaller body (so $1-\mu$ sits in the big one).
• $r_1,r_2$ ::: distances from the spacecraft to the big and small primary.
• $\frac{1}{2}(x^2+y^2)$ ::: grows as you move outward — that is the centrifugal tendency to fling you away from the spin axis.

A Lagrange point is a flat spot on this landscape: $\nabla \Omega =0$. On the collinear line joining the two bodies there are three such flat spots, $L_1,L_2,L_3$.

---

Step 2 — Zoom in: a flat spot is a saddle, not a bowl

WHAT. We pick one collinear point (say $L_2$), place the origin there, and call any small displacement $(\xi ,\eta ,\zeta )$. We ask: what does the landscape look like right around that flat spot?

WHY. A flat spot can be a valley bottom (stable — a ball returns), a hilltop (unstable — a ball flees), or a saddle (a mountain pass: stable across the ridge, unstable along it). Which one it is decides everything about whether a spacecraft stays. To find out we only need the curvature at the point — the second derivatives.

PICTURE.

Because the point sits on the $x$-axis, the landscape is symmetric, and the mixed curvatures vanish (${\Omega }_{xy}={\Omega }_{xz}={\Omega }_{yz}=0$). The three surviving curvatures are:

$$\boxed{U_{xx}=1+2c(>0),U_{yy}=1-c(<0),U_{zz}=-c(<0)}c\equiv \frac{1-\mu }{r_1^3}+\frac{\mu }{r_2^3}$$

• $c$ ::: a single positive number measuring how sharply the combined gravity falls off here. For collinear points $c>1$.
• $U_{xx}=1+2c$ ::: positive curvature along the line to the bodies → the surface curves up → potential rises → force pushes you back... but watch, the Coriolis force in Step 4 flips this into instability along the ridge.
• $U_{yy}=1-c$ and $U_{zz}=-c$ ::: both negative (since $c>1$) → curves down → these look like downhill/repelling directions on the bare landscape.

---

Step 3 — The equations of motion, term by term

WHAT. Newton's law in the rotating frame, expanded to first order about the flat spot, gives three linear equations for $(\xi ,\eta ,\zeta )$.

WHY. Full CR3BP is nonlinear and unsolvable in closed form. But near the point the displacement is tiny, so the leading (linear) term of each force dominates and the rest we throw away. Linear equations with constant coefficients we can solve exactly — with exponentials and sines.

PICTURE.

$$\ddot{\xi }-\underset{\text{Coriolis}}{\underbrace{2\dot{\eta }}}=(1+2c)\xi$$ $$\ddot{\eta }+\underset{\text{Coriolis}}{\underbrace{2\dot{\xi }}}=(1-c)\eta$$ $$\ddot{\zeta }=-c\zeta$$

• $\ddot{\xi },\ddot{\eta },\ddot{\zeta }$ ::: accelerations along the three axes ($\ddot{}$ means "second time-derivative", i.e. acceleration).
• $-2\dot{\eta }$, $+2\dot{\xi }$ ::: the Coriolis force — a sideways kick proportional to your velocity, present only because the frame spins. It couples $\xi$ and $\eta$ together.
• Right-hand sides ::: the curvatures from Step 2 acting as spring/anti-spring constants.

---

Step 4 — The out-of-plane bob: pure simple harmonic motion

WHAT. Solve $\ddot{\zeta }=-c\zeta$ directly.

WHY. Compare it to a mass on a spring, $\ddot{x}=-kx$, whose solution oscillates at $\sqrt{k}$. Here $k$ plays the role of $c$, so we can read the answer off without any new machinery — this is why we invoke SHM rather than solving a general ODE.

PICTURE.

$$\zeta (t)=A_z\sin ({\omega }_{z}t+\psi ),\boxed{{\omega }_z=\sqrt{c}}$$

• $A_z$ ::: how high the spacecraft bobs above/below the orbital plane.
• ${\omega }_z$ ::: how fast it bobs. Since $c>0$, this is a real frequency → genuine back-and-forth, never runaway.
• $\psi$ ::: a phase (where in the bob it starts).

The out-of-plane direction is the tame one: it is a valley in disguise, and the spacecraft rocks through the plane like a pendulum.

---

Step 5 — The in-plane pair: guess $e^{\lambda t}$ and turn the ODE into algebra

WHAT. The $\xi ,\eta$ equations are tangled by Coriolis. We guess that both grow/oscillate exponentially, $\xi =A{e}^{\lambda t}$, $\eta =B{e}^{\lambda t}$, and substitute.

WHY. The exponential $e^{\lambda t}$ is the only function whose derivative is a copy of itself ($\frac{d}{dt}{e}^{\lambda t}=\lambda e^{\lambda t}$). That single property turns every derivative into a multiply-by-$\lambda$, converting the differential equations into ordinary algebra. This is the standard key for any linear constant-coefficient system.

PICTURE.

Substituting ($\dot{\xi }\to \lambda A$, $\ddot{\xi }\to {\lambda }^{2}A$, etc.):

$$\left(\begin{array}{cc}{\lambda }^2-(1+2c)& -2\lambda \\ 2\lambda & {\lambda }^2-(1-c)\end{array}\right)\left(\begin{array}{c}A\\ B\end{array}\right)=\left(\begin{array}{c}0\\ 0\end{array}\right)$$

• Each entry ::: what is left when you move everything to one side; ${\lambda }^2$ from the acceleration, the $-2\lambda$ / $+2\lambda$ from Coriolis, the constants from curvature.
• $(A,B)$ ::: the shape of the mode — the ratio $B/A$ tells you how elongated the in-plane wobble is.

A nonzero $(A,B)$ exists only if the matrix squashes some direction to zero, i.e. its determinant is zero. Setting $\det =0$:

$$({\lambda }^2-(1+2c))({\lambda }^2-(1-c))+4{\lambda }^2=0$$

Expand and collect (let $\Lambda ={\lambda }^2$):

$$\boxed{{\Lambda }^2+(2-c)\Lambda +(1+2c)(1-c)=0}$$

---

Step 6 — Two roots of opposite sign: the saddle × center structure

WHAT. Solve the quadratic in $\Lambda ={\lambda }^2$.

WHY. The sign of each $\Lambda$ decides the physics: $\Lambda >0$ gives real $\lambda =\pm \sqrt{\Lambda }$ (growth/decay), while $\Lambda <0$ gives imaginary $\lambda =\pm i\sqrt{-\Lambda }$ (oscillation). We must check both roots for every collinear point — none may be skipped.

PICTURE.

$$\Lambda =\frac{-(2-c)\pm \sqrt{(2-c{)}^2-4(1+2c)(1-c)}}{2}$$

For collinear points $c>1$, so $(1+2c)(1-c)<0$. That makes the product of the two roots negative — which forces one root positive and one negative:

$${\Lambda }_1>0\Rightarrow \lambda =\pm \gamma \text{ (real)}{\Lambda }_2<0\Rightarrow \lambda =\pm i{\omega }_{xy}\text{ (imaginary)}$$

• $\gamma =\sqrt{{\Lambda }_1}$ ::: instability rate. The mode $e^{+\gamma t}$ grows — displace along it and you flee the point. This is the saddle's runaway ridge. It is exactly why $L_1,L_2$ demand station-keeping, and it is the seed of the invariant manifolds.
• ${\omega }_{xy}=\sqrt{-{\Lambda }_2}$ ::: planar oscillation frequency. The mode $e^{\pm i{\omega }_{xy}t}$ is a sine/cosine — a stable in-plane loop.

---

Step 7 — The eigenvector fixes the ellipse: the ratio $\kappa$

WHAT. For the oscillatory root, plug $\lambda =i{\omega }_{xy}$ back into the matrix to get the ratio $B/A=\kappa$.

WHY. The determinant told us a mode exists; the eigenvector tells us its shape. We need $\kappa$ because the in-plane path is an ellipse, and $\kappa$ is how much taller it is in $\eta$ than in $\xi$.

PICTURE.

$$\xi (t)=-A_x\cos ({\omega }_{xy}t+\varphi ),\eta (t)=\kappa A_x\sin ({\omega }_{xy}t+\varphi ),\boxed{\kappa =\frac{{\omega }_{xy}^2+(1+2c)}{2{\omega }_{xy}}}$$

• $A_x$ ::: the in-plane amplitude along $x$.
• $\kappa$ ::: the stretch factor. Since typically $\kappa \approx 3$, the planar loop is an ellipse elongated along $y$, not a circle.
• The $\cos$/$\sin$ pairing ::: a quarter-period out of phase, which traces an ellipse (like the $x,y$ of a point going round).

---

Step 8 — Two frequencies that don't match: Lissajous, not yet halo

WHAT. We now have two independent frequencies: in-plane ${\omega }_{xy}$ and out-of-plane ${\omega }_z=\sqrt{c}$. In general ${\omega }_{xy}\ne {\omega }_z$.

WHY. A closed 3-D loop needs the wobble and the bob to sync up — to return to the same point at the same phase. If their periods differ, the path drifts forever and fills a tube: a Lissajous orbit, not a halo.

PICTURE.

---

Worked check — Earth–Moon $L_2$ ($c\approx 3.19$)

---

The one-picture summary

Recall Feynman retelling — say it back in plain words

Spin your camera with the two big bodies so they stop moving. Now every force is a hill-and-valley landscape. The Lagrange point is a flat spot — but not a valley: it's a mountain pass, a saddle. Roll off it one way (along the line to the bodies, mixed by the sideways Coriolis kick) and you can loop — that's your in-plane wobble, an ellipse stretched about three times taller than it is wide. Bob up and down through the plane and you get a clean spring oscillation, because that direction is secretly a valley. But there's also a nasty runaway direction where you slide off the pass and never come back — that's why these spots need constant nudging to stay put. The two nice oscillations have slightly different speeds, so their combined path drifts and never quite closes — a Lissajous scribble on a tube. Only when you add the fine nonlinear corrections do the two speeds lock into a perfect 1:1 match at just the right size — and that closed 3-D loop hovering around the point is the halo.

Recall Quick self-test

Why is a collinear Lagrange point unstable? ::: The characteristic quadratic has a positive root ${\Lambda }_1$, giving real $\lambda =+\gamma$; that mode grows like $e^{\gamma t}$. What makes the out-of-plane equation easy? ::: It decouples entirely ($\ddot{\zeta }=-c\zeta$, no Coriolis), so it is pure SHM with ${\omega }_z=\sqrt{c}$. Why does linear theory give only Lissajous? ::: Generically ${\omega }_{xy}\ne {\omega }_z$, so the loop never closes; a halo needs a nonlinear 1:1 resonance. What does $\kappa$ tell you? ::: The in-plane ellipse's $y$-to-$x$ stretch — for Earth–Moon $L_2$, about $2.9$.