Intuition What this page is for
The parent note built the machinery: the pseudopotential U , the Lagrange points, and the characteristic equation
λ 4 + ( 4 − U xx − U y y ) λ 2 + ( U xx U y y − U x y 2 ) = 0.
Here we use it until nothing can surprise you. We march through every kind of input the theory can throw: both families of Lagrange point, the whole allowed range of the mass parameter μ , the degenerate boundary case, the two limiting behaviours (μ → 0 and μ → 2 1 ), and two real-world word problems. Guess before you compute — that is how the reflex gets built.
Before anything, let us re-anchor four symbols so nothing appears unearned.
Definition The rotation rate
ω and "units where ω = 1 ", in plain words
The two primaries orbit their common centre once every orbital period T orb . Their angular rate — how many radians of angle they sweep per second — is
ω = T orb 2 π ( radians per second ) .
We are free to choose our clock so that ω = 1 : one "tick" of dimensionless time equals the time for the primaries to sweep one radian of their orbit (about T orb /2 π seconds — e.g. ≈ 58 days for the Sun–Earth system). This is the standard CR3BP time-normalisation; from here on "one time unit" always means "one radian of orbital angle," and every λ , ω , growth rate is measured on that clock.
Definition The perturbation
( ξ , η ) and e λ t , in plain words
A Lagrange point sits at ( x 0 , y 0 ) . Nudge the little body a tiny bit off it; write its displacement as
ξ = x − x 0 , η = y − y 0 .
So ξ , η are just the small offsets from the balance point. We test each offset with the trial motion ξ ∝ e λ t : an exponential in time whose exponent λ tells the fate — if λ is a real positive number the offset grows (runaway), if λ is purely imaginary the offset oscillates (stays bounded). Solving for the allowed λ is exactly the characteristic equation.
Definition The mass parameter
μ , in plain words
The two big bodies (the primaries ) have masses m 1 ≥ m 2 . The mass parameter is simply the fraction of the total mass carried by the smaller primary:
μ = m 1 + m 2 m 2 , 0 < μ ≤ 2 1 .
When μ is tiny the second body is a mere speck (Sun–Jupiter: μ ≈ 0.00095 ); when μ = 2 1 the two bodies are equal twins. Every example on this page is really just "pick a μ , read off the fate," so this single number drives everything.
Definition The three second-derivatives, in plain words
Imagine the pseudopotential U ( x , y ) as a landscape of height above the rotating plane. At a Lagrange point the ground is flat (slope zero). To know its shape we need the curvatures:
U xx = how the ground curves as you walk in the x (Sun–Earth line) direction. Positive = valley-shaped that way; negative = ridge-shaped.
U y y = the same curvature in the y direction.
U x y = the "twist" — how the x -slope changes as you step sideways in y . Zero when the landscape is symmetric about the x -axis right at the point.
These three numbers are the entire input to the characteristic equation. Everything below is: find those three numbers, feed them in, read off the fate.
Every worked example below is tagged with the cell it fills.
Cell
Case class
Distinguishing feature
Example
A
Collinear points L 1 , L 2 , L 3 , generic μ
U x y = 0 , c < 0 → saddle (ALL three)
Ex 1 (+ note on L 2 , L 3 ), Ex 8 (L 2 )
B
Triangular point, small μ
c > 0 , b 2 − 4 c > 0 → stable
Ex 2
C
Triangular point, large μ
b 2 − 4 c < 0 → unstable
Ex 3
D
Degenerate boundary μ = μ crit
b 2 − 4 c = 0 (repeated root)
Ex 4
E
Limit μ → 0
one primary vanishes
Ex 5
F
Limit μ → 2 1
equal masses, most unstable triangular
Ex 6
G
Real-world word problem
pick a real system, decide fate
Ex 7 (Sun–Jupiter)
H
Exam twist / trap
recover the growth rate , not just sign
Ex 8 (L 2 )
Intuition Why all three collinear points share cell A
L 1 , L 2 , L 3 all lie on the x -axis (the primary line), so by up–down symmetry the twist U x y = 0 at each of them. And at every one of them the potential curves up along the line (U xx > 0 ) but curves down across it (U y y < 0 ) — a saddle. Hence c = U xx U y y < 0 for all three , giving one real positive λ every time: L 1 , L 2 , L 3 are always unstable , regardless of μ . We work L 1 in Ex 1 and L 2 in Ex 8; L 3 behaves identically (only the numbers change), so we state it here rather than repeat the arithmetic.
The triangular curvatures are fixed by geometry (parent note): at L 4 , L 5
U xx = 4 3 , U y y = 4 9 , U x y = ± 4 3 3 ( 1 − 2 μ ) .
So for ALL triangular examples: b = 4 − 4 3 − 4 9 = 1 , and
c = 4 3 ⋅ 4 9 − 16 27 ( 1 − 2 μ ) 2 = 16 27 [ 1 − ( 1 − 2 μ ) 2 ] = 16 27 ⋅ 4 μ ( 1 − μ ) = 4 27 μ ( 1 − μ ) .
And the stability discriminant is b 2 − 4 c = 1 − 27 μ ( 1 − μ ) , which is the parent's condition. Keep those two facts in view; the triangular examples are all one formula with different μ .
The figure below is our map for the whole page: its horizontal axis is the mass parameter μ (running from 0 to 2 1 , dimensionless), and its vertical axis is the stability discriminant 1 − 27 μ ( 1 − μ ) (also dimensionless — a pure number). Where the yellow curve sits above zero (blue shading) triangular points are stable; below zero (pink shading) they are unstable. The pink dotted line marks the critical mass μ crit where the curve crosses zero, and each coloured dot marks the μ of one worked example so you can see where every case lives before we compute it.
Figure s01 — Triangular-point stability across all μ : the discriminant curve, the stable/unstable shading, μ crit , and the marked example cases.
Worked example Ex 1 — Cell A · Collinear point
L 1 (Earth–Moon), a saddle
Statement. For Earth–Moon L 1 the parent note quotes U xx ≈ + 11.3 , U y y ≈ − 4.2 , and U x y = 0 (collinear points sit on the symmetry axis). Classify the equilibrium and find one growth rate λ .
Forecast: U y y is negative — the landscape is a ridge in the y direction. Guess: unstable. But how unstable — real λ of what size?
Compute c = U xx U y y − U x y 2 = ( 11.3 ) ( − 4.2 ) − 0 = − 47.46 .
Why this step? c is the product of the two Λ -roots; its sign alone decides saddle vs centre.
Since c < 0 , the roots Λ ± have opposite signs. Compute b = 4 − 11.3 − ( − 4.2 ) = − 3.1 .
Why this step? We need b to solve the quadratic and pull out the actual positive Λ .
Solve Λ 2 − 3.1 Λ − 47.46 = 0 : Λ = 2 3.1 ± 3. 1 2 + 4 ( 47.46 ) = 2 3.1 ± 199.45 .
199.45 = 14.123 , so Λ + = 8.611 , Λ − = − 5.511 .
Why this step? The positive root gives the real exponential; the negative gives an oscillation.
Growth rate λ = Λ + = 8.611 = 2.935 (on the ω = 1 clock defined above — so λ is measured per radian of orbital angle).
Why this step? The small offset obeys ξ ∝ e λ t (recall ξ is the tiny displacement off L 1 ). With λ = 2.935 , after one time unit — one radian of the Earth–Moon orbit — the nudge grows by e 2.935 ≈ 18.8 × . That relentless amplification is why L 1 needs active station-keeping .
Verify: Product of roots Λ + Λ − = 8.611 × ( − 5.511 ) = − 47.46 = c . ✓ Sum = 8.611 − 5.511 = 3.1 = − b . ✓ Opposite signs confirm the saddle.
Worked example Ex 2 — Cell B · Triangular
L 4 (Earth–Moon), stable
Statement. Earth–Moon μ = 0.0123 . Classify L 4 and give the two oscillation frequencies.
Forecast: L 4 is a maximum of U (a hilltop). Naively "balls roll off hilltops" ⇒ unstable. But Coriolis rescues small-μ triangular points. With μ = 0.0123 well under 0.0385 , guess: stable , two distinct real oscillation frequencies.
b = 1 (always, for triangular). Compute c = 4 27 μ ( 1 − μ ) = 4 27 ( 0.0123 ) ( 0.9877 ) = 0.08202 .
Why this step? c > 0 is the first stability gate; here it passes.
Discriminant b 2 − 4 c = 1 − 4 ( 0.08202 ) = 1 − 0.32808 = 0.67192 > 0 .
Why this step? Equivalent to 1 − 27 μ ( 1 − μ ) ; positive ⇒ both Λ real. Combined with c > 0 , b > 0 ⇒ both negative ⇒ stable.
Roots: Λ ± = 2 − 1 ± 0.67192 = 2 − 1 ± 0.81971 , giving Λ + = − 0.09015 , Λ − = − 0.90985 .
Why this step? Both negative confirms pure imaginary λ .
Frequencies ω 1 , 2 = − Λ ± : ω 1 = 0.09015 = 0.3003 (long-period libration), ω 2 = 0.90985 = 0.9539 (short-period). These are measured on the same ω = 1 clock, so ω 2 ≈ 0.95 means the fast wobble is almost as quick as one orbital radian.
Why this step? These are the two libration frequencies real Trojan-like bodies wobble at.
Verify: Λ + Λ − = ( − 0.09015 ) ( − 0.90985 ) = 0.08202 = c . ✓ Sum = − 1 = − b . ✓ Both negative ⇒ stable, matching Ex 2 of the parent (27 μ ( 1 − μ ) = 0.328 < 1 ).
Worked example Ex 3 — Cell C · Triangular point, mass just over critical → unstable
Statement. A hypothetical binary has μ = 0.05 (above μ crit ≈ 0.03852 ). Classify L 4 .
Forecast: μ exceeds the critical value, so guess: the discriminant flips negative → complex Λ → unstable (spiralling growth).
c = 4 27 ( 0.05 ) ( 0.95 ) = 0.320625 ; b = 1 .
Why this step? Set up the quadratic in Λ .
Discriminant b 2 − 4 c = 1 − 1.2825 = − 0.2825 < 0 .
Why this step? Negative discriminant ⇒ Λ are complex conjugates , not real. Then λ = Λ has a nonzero real part ⇒ growth.
Complex roots Λ = 2 − 1 ± i 0.2825 = − 0.5 ± 0.26575 i .
Why is there always a growing direction? Write a complex Λ in polar form as Λ = R e i θ with R > 0 and angle − π < θ < π . Its two square roots are λ = ± R e i θ /2 . Their real parts are ± R cos ( θ /2 ) , and since − 2 π < θ /2 < 2 π we have cos ( θ /2 ) > 0 — so one root has a strictly positive real part whenever θ = 0 (i.e. whenever Λ is not real). That positive real part is the exponential growth. This is oscillatory instability : the body spirals outward, unlike the pure exponential saddle of the collinear points.
Verify: discriminant sign check: 27 μ ( 1 − μ ) = 27 ( 0.05 ) ( 0.95 ) = 1.2825 > 1 , so 1 − 27 μ ( 1 − μ ) = − 0.2825 < 0 . ✓ Unstable confirmed.
Worked example Ex 4 — Cell D · The degenerate boundary
μ = μ crit
Statement. At exactly μ crit = 2 1 ( 1 − 23/27 ) the discriminant is zero. Find μ crit numerically and the (repeated) root Λ .
Forecast: Boundary between stable and unstable — expect a repeated real root, marginal case. Guess μ crit ≈ 0.0385 .
Solve 1 − 27 μ ( 1 − μ ) = 0 , i.e. 27 μ 2 − 27 μ + 1 = 0 . Then μ = 2 ⋅ 27 27 − 2 7 2 − 4 ⋅ 27 = 54 27 − 621 .
Why this step? The stability discriminant, set to zero, is a plain quadratic in μ .
621 = 24.9199 , so μ crit = 54 27 − 24.9199 = 54 2.0801 = 0.038520 .
Why this step? Confirms the parent's 0.03852 and shows the algebraic form 2 1 ( 1 − 23/27 ) is identical (since 621 = 2 7 2 − 108 and 23/27 = 621/729 ).
At the boundary b 2 − 4 c = 0 , so Λ = − b /2 = − 1/2 (double root). Then λ = ± i / 2 (repeated).
Why this step? A repeated imaginary pair means the two libration frequencies merge ; beyond this μ they split into a complex pair and destabilise. This coalescence is the mathematical signature of the stability boundary.
Verify: 2 1 ( 1 − 23/27 ) = 2 1 ( 1 − 0.922958 ) = 0.038521 . ✓ matches step 2. And 27 μ crit ( 1 − μ crit ) = 1 . ✓
Worked example Ex 5 — Cell E · Limit
μ → 0 (second primary vanishes)
Statement. Let μ → 0 + . What happens to c and to the triangular-point stability? Physically interpret.
Forecast: With m 2 → 0 we are back to (nearly) a two-body / one-body situation. Guess: L 4 becomes marginally, then trivially stable; the libration frequencies approach 0 and 1 .
c = 4 27 μ ( 1 − μ ) → 0 as μ → 0 .
Why this step? c is the product of roots; if it → 0 , one root Λ → 0 .
With b = 1 and c → 0 : roots Λ + → 0 and Λ − → − 1 . Frequencies ω 1 = − Λ + → 0 , ω 2 = − Λ − → 1 .
Why this step? ω 2 → 1 is the orbital frequency itself (on the ω = 1 clock) — the tiny body just orbits the lone primary at the mean motion. ω 1 → 0 means the long-period libration becomes infinitely slow (no second mass to hold it).
Discriminant 1 − 27 μ ( 1 − μ ) → 1 > 0 : deeply inside the stable region.
Why this step? Confirms every real planetary system with a tiny secondary keeps stable Trojans.
Verify: at μ = 0 , c = 0 , roots 0 and − 1 ; ω 2 = 1 = ω (orbital rate). ✓
Worked example Ex 6 — Cell F · Limit
μ → 2 1 (equal masses)
Statement. Equal-mass binary, μ = 0.5 . Classify L 4 and find the complex Λ .
Forecast: Two equal stars — the most symmetric, and (since 0.5 > μ crit ) the most unstable triangular case. Guess: strongly complex Λ .
c = 4 27 ( 0.5 ) ( 0.5 ) = 16 27 = 1.6875 ; b = 1 .
Why this step? c is largest possible here, since μ ( 1 − μ ) peaks at μ = 2 1 .
Discriminant 1 − 4 ( 1.6875 ) = 1 − 6.75 = − 5.75 < 0 ⇒ complex conjugate Λ .
Why this step? Confirms strong oscillatory instability; note U x y = 4 3 3 ( 1 − 2 μ ) = 0 here, so the twist vanishes yet instability remains — instability is not about the twist term.
Λ = 2 − 1 ± i 5.75 = − 0.5 ± 1.19896 i .
Why this step? Nonzero real part in λ = Λ ⇒ runaway (by the polar-form argument of Ex 3). Equal-mass binaries have no stable Trojan points.
Verify: ∣ Λ ∣ = 0. 5 2 + 1.1989 6 2 = 0.25 + 1.4375 = 1.6875 = c . ✓ (product of a conjugate pair = ∣Λ ∣ 2 = c = 1.6875 .)
Worked example Ex 7 — Cell G · Real-world word problem: Sun–Jupiter Trojans
Statement. Jupiter is 9.55 × 1 0 − 4 of the Sun's mass. Real asteroids (the Trojans ) sit at L 4 , L 5 . Show the theory permits them and find the two libration periods (in units of Jupiter's orbital period).
Forecast: μ ≈ 0.00095 ≪ 0.0385 , so stable — matches observation. Expect one very long-period and one nearly-orbital-period libration.
μ = 1 + 9.55 × 1 0 − 4 9.55 × 1 0 − 4 = 9.541 × 1 0 − 4 .
Why this step? Convert mass ratio to the mass parameter μ = m 2 / ( m 1 + m 2 ) .
c = 4 27 μ ( 1 − μ ) = 4 27 ( 9.541 × 1 0 − 4 ) ( 0.99905 ) = 6.4327 × 1 0 − 3 ; b = 1 .
Why this step? Tiny c ⇒ one root near 0 , one near − 1 .
Roots Λ ± = 2 − 1 ± 1 − 4 c = 2 − 1 ± 0.974269 . 0.974269 = 0.987051 . So Λ + = − 0.0064746 , Λ − = − 0.993525 .
Why this step? Both negative ⇒ stable; feed to frequencies.
Frequencies ω 1 = − Λ + = 0.0064746 = 0.080465 , ω 2 = − Λ − = 0.993525 = 0.996757 (on the ω = 1 clock).
Why this step? These are the two libration rates; now convert to periods.
On the ω = 1 clock the primaries' orbital period is T orb = 2 π / ω = 2 π . A libration period is T = 2 π / ω i , so its ratio to the orbital period is T / T orb = ω / ω i = 1/ ω i . Hence the long-period libration lasts 1/ ω 1 = 1/0.080465 = 12.43 Jupiter orbits, and the short one 1/ ω 2 = 1/0.996757 = 1.003 Jupiter orbits.
Why this step? The ~12.4-orbit long libration matches the observed slow "tadpole" wobble of the real Trojans.
Verify: Λ + Λ − = ( − 0.0064746 ) ( − 0.993525 ) = 6.4327 × 1 0 − 3 = c . ✓ Both roots negative ⇒ stable, μ ≪ μ crit . ✓ Long-period ratio 1/ ω 1 ≈ 12.43 . ✓
Worked example Ex 8 — Cell H (and cell A, point
L 2 ) · Exam twist: get the number of e-foldings at L 2
Statement. Sun–Earth L 2 (home of space telescopes) has effective curvatures U xx ≈ + 7.38 , U y y ≈ − 2.19 , U x y = 0 . A spacecraft is nudged and left uncontrolled. By what factor has the displacement grown after one dimensionless time unit (t = 1 , i.e. one orbital radian)?
Forecast: L 2 is collinear ⇒ (by cell A) unstable ⇒ real positive λ . The trap: many stop at "unstable." The exam wants the growth factor e λ .
c = U xx U y y − U x y 2 = ( 7.38 ) ( − 2.19 ) − 0 = − 16.1622 < 0 ⇒ saddle, exactly like the other collinear points.
Why this step? Confirms a real positive λ exists (cell A behaviour for L 2 ).
b = 4 − 7.38 − ( − 2.19 ) = − 1.19 . Solve Λ 2 − 1.19Λ − 16.1622 = 0 : Λ + = 2 1.19 + 1.4161 + 64.6488 = 2 1.19 + 66.0649 = 2 1.19 + 8.1281 = 4.6591 .
Why this step? Positive root gives the unstable direction's rate.
λ = 4.6591 = 2.1585 . Growth factor of the offset ξ after t = 1 : e λ ⋅ 1 = e 2.1585 = 8.658 .
Why this step? In one time unit (one radian of the Sun–Earth orbit, ~58 s-scaled days) an unmanaged deviation grows ~8.7×. That is the quantitative reason station-keeping burns happen roughly every few weeks.
Verify: Λ + = 4.6591 ; check Λ + 2 − 1.19 Λ + + ( − 16.1622 ) ≈ 0 : 21.707 − 5.544 − 16.162 = 0.001 ≈ 0 . ✓ Growth factor e 2.1585 = 8.66 . ✓
Recall Self-test (reveal after guessing)
Collinear points L 1 , L 2 , L 3 are always ::: unstable saddles, because U x y = 0 with U xx > 0 , U y y < 0 gives c < 0 at each of them.
Triangular points are stable iff ::: μ ≤ μ crit ≈ 0.03852 , equivalently 27 μ ( 1 − μ ) ≤ 1 .
What single quantity decides saddle vs centre for a collinear point? ::: the sign of c = U xx U y y (product of the Λ -roots).
Why don't equal-mass binaries have stable Trojans? ::: μ = 0.5 > μ crit , so the discriminant 1 − 27 μ ( 1 − μ ) = − 5.75 < 0 , giving complex Λ and oscillatory runaway.
The "4" in the characteristic equation comes from ::: the Coriolis coupling terms 2 η ˙ , 2 ξ ˙ squared into the determinant.
"One time unit" means ::: one radian of the primaries' orbital angle (the clock where ω = 1 ).
Mnemonic Two gates, one grid
C then D. First check c (the corner value): c < 0 ⇒ dead (saddle, unstable). If c > 0 , check the D iscriminant 1 − 27 μ ( 1 − μ ) : positive ⇒ delightful (stable), negative ⇒ doomed (spiral).
See also: Lagrange points , Jacobi constant & zero-velocity curves , Rotating reference frames — Coriolis & centrifugal forces .