The Laplace-Runge-Lenz Vector

Energy and angular momentum are the celebrities of conservation, but the Kepler problem hides a third conserved quantity that most courses never name — and it is the deepest of the three. For a generic attractive central force, conserving $E$ and $\mathbf{L}$ is enough to trap the motion in a plane and bound the radius, but not enough to close the orbit: the generic result is a rosette, an ellipse-like loop whose long axis slowly pinwheels around the centre. Yet planets do not pinwheel. Kepler orbits close, perihelion after perihelion landing on the same spot, and that stubborn fixity is the signature of an extra frozen quantity. The frozen quantity is a vector: $\mathbf{A}=\mathbf{p}\times\mathbf{L}-mk\,\hat{\mathbf{r}}$ , the Laplace–Runge–Lenz vector. It is a strange construction — assembled from a momentum and a unit vector that individually swing wildly around the orbit — yet for the inverse-square force, and only the inverse-square force, the combination sits nailed in space, pointing from the focus to perihelion with magnitude $mke$ . Its constancy is a five-line consequence of Newton's law, and it repays the effort instantly: dot $\mathbf{A}$ into $\mathbf{r}$ and the conic-section orbit equation falls out in two lines, no differential equation solved, Binet's machinery bypassed entirely. The deeper story is symmetry. Noether ties $\mathbf{L}$ to ordinary rotations; $\mathbf{A}$ signals a hidden four-dimensional rotational symmetry of the Kepler problem, the $SO(4)$ structure that Pauli exploited in 1926 to crack the hydrogen spectrum before Schrödinger's equation was even solved for it. This article builds the vector, proves it frozen, spends it on the orbit, and opens the door that the hidden-symmetry article walks through.

Key takeaways

The Laplace–Runge–Lenz vector $\mathbf{A}=\mathbf{p}\times\mathbf{L}-mk\,\hat{\mathbf{r}}$ is a conserved vector of the Kepler problem, constant only for the exact inverse-square force $V=-k/r$ .
It lies in the orbital plane ( $\mathbf{A}\cdot\mathbf{L}=0$ ), points from the force centre toward perihelion, and its magnitude is $A=mke$ — the eccentricity in dynamical units.
Its conservation is why Kepler orbits close: any deviation from $1/r^2$ sets $\mathbf{A}$ rotating, and the rotation rate of $\mathbf{A}$ is literally the perihelion precession rate.
Dotting $\mathbf{A}$ with $\mathbf{r}$ yields the conic orbit equation $r=p/(1+e\cos\theta)$ algebraically — the shortest route to Kepler's first law in all of mechanics.
Of the seven scalars in $(E,\mathbf{L},\mathbf{A})$ only five are independent: $\mathbf{A}\cdot\mathbf{L}=0$ and $A^2=m^2k^2+2mEL^2$ eat two, leaving exactly the phase along the orbit undetermined.
Key idea: $\mathbf{A}=\mathbf{p}\times\mathbf{L}-mk\,\hat{\mathbf{r}}$ is conserved iff the force is exactly inverse-square; it points to perihelion with magnitude $mke$ , so dotting it with $\mathbf{r}$ gives the orbit equation in one line.

Where this lives today

The vector's quantum afterlife is its most celebrated: promoted to a (symmetrized) operator, $\mathbf{A}$ and $\mathbf{L}$ close into the same $SO(4)$ algebra, and Pauli's 1926 purely algebraic derivation of $E_n=-13.6\ \mathrm{eV}/n^2$ — including the "accidental" $n^2$ -fold degeneracy that plain rotational symmetry cannot explain — remains the standard first exhibit whenever hidden dynamical symmetry is taught. The same structure resurfaces wherever a $1/r$ problem does: in Rydberg-atom physics, where weak external fields lift the Kepler degeneracy and the eccentricity vector becomes the natural coordinate for the slow evolution of near-classical electron ellipses under engineered microwave control. Classically, the eccentricity vector $\mathbf{e}=\mathbf{A}/(mk)$ is everyday infrastructure. Planetary-dynamics codes track the secular evolution of exoplanet systems in $(\mathbf{e},\mathbf{i})$ -vector components, in which Laplace's own perturbation equations become nearly linear; and satellite operators flying today's large low-Earth-orbit constellations screen and design close formations using relative eccentricity/inclination vector separation — a collision-avoidance geometry whose state variables are, name unchanged, the frozen arrow of this article, allowed to creep slowly under $J_2$ and drag and steered back by station-keeping burns.

The physical picture

Watch a planet round its ellipse and track the two ingredients of $\mathbf{A}$ separately. The vector $\mathbf{p}\times\mathbf{L}$ is the momentum rotated $90°$ in the orbital plane and stretched by $L$ ; as the planet swings around, this vector swings too, growing at perihelion where the planet is fast, shrinking at aphelion. The second ingredient, $-mk\,\hat{\mathbf{r}}$ , is a constant-length arrow pointing from the planet back through the Sun. Each ingredient rotates through a full turn every orbit. And yet their sum never moves:

Figure

Two wildly swinging vectors, one frozen sum — the Runge–Lenz vector around an e = 0.6 orbit

At each of the three marked points the blue arrow ( $\mathbf{p}\times\mathbf{L}$ ) and the red arrow ( $-mk\hat{\mathbf{r}}$ ) are different — different lengths, different directions — but the green sum is the identical arrow, aimed at perihelion, always. That is what "conserved vector" means made visible, and it is worth pausing on how strange it is: nothing about either ingredient suggests permanence. The permanence is a conspiracy of the force law. As the derivation below shows, the rate of change of $\mathbf{p}\times\mathbf{L}$ involves the force's $r$ -dependence, while the rate of change of $mk\hat{\mathbf{r}}$ involves pure kinematics ( $\hat{\mathbf{r}}$ turns as fast as the angle sweeps). For $F\propto1/r^2$ — and for no other power law — the two rates match identically, term by term, everywhere on every orbit. The physical payoff of a frozen in-plane arrow is orientation. Energy fixes the ellipse's size, angular momentum its width, but only $\mathbf{A}$ fixes which way the long axis points. A conserved pointer means the axis can never drift: the orbit closes. Break the conspiracy — add the general-relativistic correction, the Sun's oblateness, a third body — and $\mathbf{A}$ starts to creep around the plane. Its creep is the perihelion precession:

Figure

When the force is not exactly inverse-square, A rotates — and its rotation is the perihelion advance

This picture converts a hard question — "how fast does a perturbed ellipse precess?" — into an easy one: "what torque-like drive does the perturbation exert on $\mathbf{A}$ ?" Mercury's famous 43 arcseconds per century, computed in the worked examples, is exactly this arrow creeping.

The definition and the conservation law

Set the stage precisely. A body of mass $m$ (the reduced mass of the two-body pair) moves under the attractive inverse-square force, and out of its instantaneous state we assemble the candidate vector:

\mathbf{A} = \mathbf{p}\times\mathbf{L} - mk\,\hat{\mathbf{r}} \qquad (1)

where: $\mathbf{p}=m\dot{\mathbf{r}}$ — momentum of the (reduced-mass) motion; $\mathbf{L}=\mathbf{r}\times\mathbf{p}$ — the angular momentum, itself conserved; $k$ — the force constant of $V=-k/r$ , with $k=Gm_1m_2$ for gravity and $\mathbf{F}=-k\hat{\mathbf{r}}/r^2$ ; $\hat{\mathbf{r}}$ — the unit vector from force centre to body. Dimensionally $\mathbf{A}$ carries momentum-times-angular-momentum; keep that in mind, because the literature also circulates the rescaled versions $\mathbf{A}/m$ and the dimensionless $\mathbf{A}/(mk)$ . The claim that earns the vector its fame is stark — a full extra conserved vector, beyond $E$ and $\mathbf{L}$ , but strictly conditional on the force law:

\dot{\mathbf{A}} = \mathbf{0}\qquad\text{for }\;\mathbf{F}=-\frac{k}{r^2}\hat{\mathbf{r}}\;\text{ exactly} \qquad (2)

where: the dot is the time derivative along the actual motion. The conditionality is the physics: this is not a geometric identity but a fingerprint test. Conservation holds for the inverse-square law and fails for every neighbouring force law, which is why measuring a perihelion drift is measuring a deviation from $1/r^2$ .

Derivation AProof That the Vector Is Conserved

The goal is equation (2): the time derivative of $\mathbf{A}=\mathbf{p}\times\mathbf{L}-mk\hat{\mathbf{r}}$ vanishes along any trajectory of the inverse-square force. The strategy is direct assault — differentiate the definition, feed in Newton's second law for $\dot{\mathbf{p}}$ , and grind the cross products down with the BAC-CAB identity until both terms of $\dot{\mathbf{A}}$ are visibly the same object with opposite signs. Two supporting facts get used without ceremony: $\mathbf{L}$ is constant for any central force (so it passes through time derivatives), and the kinematic identity for $\dot{\hat{\mathbf{r}}}$ , which is worth deriving en route because the whole cancellation hinges on it. Watch where the inverse-square law enters — exactly once, at Step 2 — because that single entry point is what makes the conservation so fragile to perturbations.

Step 1. Differentiate the definition. Since $\dot{\mathbf{L}}=\mathbf{0}$ for any central force, $\dot{\mathbf{A}}=\dot{\mathbf{p}}\times\mathbf{L}-mk\,\dot{\hat{\mathbf{r}}}$ . Why: the product rule would add $\mathbf{p}\times\dot{\mathbf{L}}$ , but a central force exerts no torque about the centre, so that term is identically zero.

Step 2. Insert the force law: $\dot{\mathbf{p}}=-\dfrac{k}{r^2}\hat{\mathbf{r}}$ , so $\dot{\mathbf{p}}\times\mathbf{L}=-\dfrac{k}{r^2}\,\hat{\mathbf{r}}\times(m\,\mathbf{r}\times\dot{\mathbf{r}})$ , using $\mathbf{L}=m\,\mathbf{r}\times\dot{\mathbf{r}}$ . Why: this is the only step where the specific power $1/r^2$ enters — every later cancellation is bookkeeping that either works or fails depending on this exponent.

Step 3. Expand with the BAC-CAB identity: $\hat{\mathbf{r}}\times(\mathbf{r}\times\dot{\mathbf{r}})=\mathbf{r}\,(\hat{\mathbf{r}}\cdot\dot{\mathbf{r}})-\dot{\mathbf{r}}\,(\hat{\mathbf{r}}\cdot\mathbf{r}) = r\dot r\,\hat{\mathbf{r}}-r\,\dot{\mathbf{r}}$ . Why: $\hat{\mathbf{r}}\cdot\mathbf{r}=r$ trivially, and $\hat{\mathbf{r}}\cdot\dot{\mathbf{r}}=\dot r$ because only the radial component of velocity changes the length of $\mathbf{r}$ .

Step 4. Assemble the first term: $\dot{\mathbf{p}}\times\mathbf{L}=-\dfrac{mk}{r^2}\left(r\dot r\,\hat{\mathbf{r}}-r\dot{\mathbf{r}}\right)=mk\,\dfrac{r\dot{\mathbf{r}}-\dot r\,\mathbf{r}}{r^2}$ . Why: just distributing the prefactor; the surviving structure $(r\dot{\mathbf{r}}-\dot r\mathbf{r})/r^2$ should look familiar — it is begging to be recognized as a derivative.

Step 5. Recognize the kinematic identity. Differentiating $\hat{\mathbf{r}}=\mathbf{r}/r$ by the quotient rule gives $\dot{\hat{\mathbf{r}}}=\dfrac{r\dot{\mathbf{r}}-\dot r\,\mathbf{r}}{r^2}$ — precisely the object in Step 4. Hence $\dot{\mathbf{p}}\times\mathbf{L}=mk\,\dot{\hat{\mathbf{r}}}$ . Why: this is the conspiracy in full view — the dynamical rate of change of $\mathbf{p}\times\mathbf{L}$ equals the kinematic rate of change of $mk\hat{\mathbf{r}}$ , and only the $1/r^2$ prefactor makes the powers of $r$ match.

Step 6. Conclude: $\dot{\mathbf{A}}=mk\,\dot{\hat{\mathbf{r}}}-mk\,\dot{\hat{\mathbf{r}}}=\mathbf{0}$ . Why: the two terms of the definition were engineered to cancel — the constant $mk$ in front of $\hat{\mathbf{r}}$ in (1) is exactly the coefficient Step 5 produces, and nothing else would work. *Verified: repeating Step 2 with $F=-k/r^{2+\epsilon}$ leaves a residue $\dot{\mathbf{A}}\propto\epsilon$ , confirming conservation is exclusive to inverse-square; matches Goldstein §3.9 and Taylor Ch. 8.*

Geometry of the frozen arrow

A conserved vector is only useful once you know where it points and how long it is. Both questions have exact answers, and both are one-line consequences of the definition. First, the vector never leaves the plane of the motion:

\mathbf{A}\cdot\mathbf{L} = 0 \qquad (3)

where: the notation is as above. Both ingredients are separately perpendicular to $\mathbf{L}$ — the cross product $\mathbf{p}\times\mathbf{L}$ by construction, and $\hat{\mathbf{r}}$ because the position always lies in the orbital plane to which $\mathbf{L}$ is normal. So $\mathbf{A}$ is an in-plane arrow, as a perihelion-pointer had better be. Next, its length. Squaring the definition and reducing with the scalar triple product compresses beautifully into the two conserved scalars:

A^2 = m^2k^2 + 2mEL^2 \qquad (4)

where: $E=\dfrac{p^2}{2m}-\dfrac{k}{r}$ — the conserved energy; $L=|\mathbf{L}|$ . The derivation is three moves: $(\mathbf{p}\times\mathbf{L})^2=p^2L^2$ since $\mathbf{p}\perp\mathbf{L}$ ; the cross term $-2mk\,\hat{\mathbf{r}}\cdot(\mathbf{p}\times\mathbf{L})=-2mkL^2/r$ by cycling the triple product $\hat{\mathbf{r}}\cdot(\mathbf{p}\times\mathbf{L})=\mathbf{L}\cdot(\hat{\mathbf{r}}\times\mathbf{p})=L^2/r$ ; then the combination $p^2L^2-2mkL^2/r=2mL^2E$ recognizes the energy. Two consequences follow immediately. Comparing with the Kepler-problem result $e^2=1+2EL^2/(mk^2)$ gives the headline identification

A = mke,\qquad \mathbf{e}=\frac{\mathbf{A}}{mk} \qquad (5)

where: $e$ — the orbital eccentricity, and $\mathbf{e}$ — the dimensionless eccentricity vector, the form of $\mathbf{A}$ that orbital mechanics uses daily. The frozen arrow's length is the shape of the orbit. And counting constants: $E$ , $\mathbf{L}$ , $\mathbf{A}$ are seven scalars, but (3) and (4) are two constraints, leaving five independent conserved quantities — the maximum a bounded system with a periodic orbit can have, pinning everything about the trajectory except the body's phase along it.

Go deeper · Derivation B · 1 minThe Arrow Points at Perihelion with Length mke

Derivation BThe Arrow Points at Perihelion with Length mke

The goal is the geometric content of equation (5): $\mathbf{A}$ aims from the focus toward the perihelion point, and its magnitude is $mke$ . The strategy is to evaluate the definition (1) at the one point of the orbit where the kinematics is maximally simple — perihelion, where the radial velocity vanishes and the momentum is purely tangential — and read off both direction and magnitude by explicit computation. Since $\mathbf{A}$ is conserved, whatever we find there holds everywhere. The inputs: at perihelion $r_p=\dfrac{L^2/(mk)}{1+e}$ from the conic equation, and $L=mr_pv_p$ because velocity and radius are perpendicular there.

Step 1. Set up perihelion kinematics. At the closest approach $\dot r=0$ , so $\mathbf{p}=mv_p\,\hat{\boldsymbol{\theta}}$ with $\hat{\boldsymbol{\theta}}$ the in-plane unit vector perpendicular to $\hat{\mathbf{r}}$ in the direction of motion, and $L=mr_pv_p$ . Why: a turning point of $r$ has purely tangential velocity by definition — the simplification that makes this evaluation clean.

Step 2. Compute the cross product there: $\mathbf{p}\times\mathbf{L}=(mv_p\,\hat{\boldsymbol{\theta}})\times(L\hat{\mathbf{L}})=mv_pL\,\hat{\mathbf{r}}$ . Why: $\hat{\boldsymbol{\theta}}\times\hat{\mathbf{L}}=\hat{\mathbf{r}}$ in a right-handed orbital triad — the rotated momentum points radially outward at perihelion.

Step 3. Assemble $\mathbf{A}$ at perihelion: $\mathbf{A}=\left(mv_pL-mk\right)\hat{\mathbf{r}}$ — the vector is purely radial, aimed along the focus-to-perihelion direction. Why: both surviving terms are proportional to $\hat{\mathbf{r}}$ ; direction established, sign pending the bracket.

Step 4. Evaluate the bracket with $v_p=L/(mr_p)$ : $mv_pL-mk=\dfrac{L^2}{r_p}-mk$ , and inserting $r_p=\dfrac{L^2/(mk)}{1+e}$ gives $\dfrac{L^2(1+e)mk}{L^2}-mk=mk(1+e)-mk=mke$ . Why: the conic supplies the perihelion radius; positivity of the result ( $e>0$ for any non-circular orbit) confirms $\mathbf{A}$ points toward perihelion, not away.

Step 5. Conclude $\mathbf{A}=mke\,\hat{\mathbf{r}}_{\text{perihelion}}$ , everywhere and forever, by conservation. Why: a constant vector evaluated at one convenient instant is evaluated for all time — the entire strategy of the derivation. *Verified: magnitude agrees with the square-root of (4) using $e^2=1+2EL^2/(mk^2)$ ; for a circular orbit $e=0$ the bracket in Step 4 vanishes and $\mathbf{A}=\mathbf{0}$ , consistent with a circle having no perihelion to point at.*

The orbit equation in two lines

Here is the vector's party trick, and it genuinely is two lines once the tools are on the table. In the Kepler problem, extracting the orbit meant transforming to $u=1/r$ , deriving Binet's differential equation and solving it. With $\mathbf{A}$ conserved, all of that becomes a single scalar product. Dot the definition with $\mathbf{r}$ , cycle one triple product, and the conic appears:

\mathbf{r}\cdot\mathbf{A}=rA\cos\theta = L^2-mkr \;\;\Longrightarrow\;\; r=\frac{L^2/(mk)}{1+e\cos\theta} \qquad (6)

where: $\theta$ — the angle between $\mathbf{r}$ and the fixed direction $\mathbf{A}$ , i.e. the true anomaly measured from perihelion; $L^2/(mk)$ — the semi-latus rectum $p$ ; $e=A/(mk)$ — the eccentricity, closing the loop with (5). This is Kepler's first law: a conic section with the force centre at a focus, derived without integrating anything. The angle $\theta$ acquired its physical meaning for free — it is measured from $\mathbf{A}$ , and $\mathbf{A}$ points at perihelion, which is exactly where the denominator maximizes and $r$ minimizes, as it should.

Go deeper · Derivation C · 1 minThe Conic from One Scalar Product

Derivation CThe Conic from One Scalar Product

The goal is equation (6), the full orbit equation, using nothing but the conservation of $\mathbf{A}$ and three lines of vector algebra — the most economical derivation of Kepler's first law known. The strategy: compute the scalar $\mathbf{r}\cdot\mathbf{A}$ twice. Once geometrically, as $rA\cos\theta$ with $\theta$ the angle from the fixed arrow $\mathbf{A}$ ; once algebraically, by dotting $\mathbf{r}$ into the definition of $\mathbf{A}$ term by term. Equating the two readings and solving for $r$ delivers the conic. The one manipulation worth savouring is the cyclic shuffle of the scalar triple product, which converts $\mathbf{r}\cdot(\mathbf{p}\times\mathbf{L})$ into $L^2$ — the moment the dynamics hands the geometry its answer.

Step 1. Geometric reading: $\mathbf{r}\cdot\mathbf{A}=rA\cos\theta$ , defining $\theta$ as the angle between the instantaneous position and the frozen perihelion direction. Why: this is just the definition of the dot product — legitimate as a global angle variable only because $\mathbf{A}$ is constant; for a precessing force law there would be no fixed reference arm.

Step 2. Algebraic reading, first term: $\mathbf{r}\cdot(\mathbf{p}\times\mathbf{L})=\mathbf{L}\cdot(\mathbf{r}\times\mathbf{p})=\mathbf{L}\cdot\mathbf{L}=L^2$ . Why: the scalar triple product is invariant under cyclic permutation, and $\mathbf{r}\times\mathbf{p}$ is $\mathbf{L}$ itself — the dynamics collapses into one conserved number.

Step 3. Algebraic reading, second term: $mk\,\mathbf{r}\cdot\hat{\mathbf{r}}=mkr$ , so altogether $\mathbf{r}\cdot\mathbf{A}=L^2-mkr$ . Why: $\mathbf{r}\cdot\hat{\mathbf{r}}$ is just the length $r$ — no content, only tidiness.

Step 4. Equate and solve: $rA\cos\theta=L^2-mkr\;\Rightarrow\;r\,(mk+A\cos\theta)=L^2\;\Rightarrow\;r=\dfrac{L^2/(mk)}{1+(A/mk)\cos\theta}$ . Why: pure algebra; comparing with the standard focal conic identifies the semi-latus rectum $p=L^2/(mk)$ and the eccentricity $e=A/(mk)$ — which is the independent, orbit-shape confirmation of the magnitude result $A=mke$ . *Verified: reproduces exactly the Binet-route result of the Kepler problem article, including $p=L^2/(mk)$ and $e=\sqrt{1+2EL^2/(mk^2)}$ via (4); Goldstein §3.9.*

Why inverse-square is special — and what breaks

Step 2 of the conservation proof is the pressure point: the force law entered exactly once. Perturb it and track the damage. Write the force as inverse-square plus a small extra piece; the inverse-square part still cancels against $mk\dot{\hat{\mathbf{r}}}$ as before, and what remains is an exact drive equation for the arrow:

\dot{\mathbf{A}} = \mathbf{F}_{\text{pert}}\times\mathbf{L} \qquad (7)

where: $\mathbf{F}_{\text{pert}}$ — the non-Keplerian part of the force (a GR correction, an oblateness term, a third body's tug). Since $\mathbf{F}_{\text{pert}}\times\mathbf{L}$ is perpendicular to $\mathbf{L}$ , the drive keeps $\mathbf{A}$ in the orbital plane: the arrow does not shrink or tip out of plane at leading order — it rotates. Averaged over one revolution, its rotation angle is the apsidal advance per orbit, which is how the second figure's picture becomes a calculational tool: Mercury's GR precession, the textbook $1/r^3$ apsidal drift, and satellite $J_2$ perigee rotation are all this one equation with different $\mathbf{F}_{\text{pert}}$ plugged in. The same specialness has a group-theoretic voice, and it is worth hearing even in passing. The Poisson brackets of the components of $\mathbf{A}$ with each other do not close on themselves — they return angular momentum, weighted by the energy:

\{A_i,A_j\}=-2mE\,\epsilon_{ijk}L_k \qquad (8)

where: $\{\ ,\ \}$ — the Poisson bracket; $\epsilon_{ijk}$ — the Levi-Civita symbol. On a bound-energy surface ( $E<0$ ) the rescaled vector $\mathbf{D}=\mathbf{A}/\sqrt{-2mE}$ satisfies $\{D_i,D_j\}=\epsilon_{ijk}L_k$ , and the six quantities $(\mathbf{L},\mathbf{D})$ close into the Lie algebra of $SO(4)$ — rotations of a four-dimensional space, one dimension more than the rotations $\mathbf{L}$ alone generates. This hidden four-dimensional symmetry is the structural reason the Kepler problem is "too solvable": why its orbits close (Bertrand's theorem singles out $1/r$ and $r^2$ as the only central potentials with universally closed bound orbits), and why hydrogen's energy levels are degenerate beyond what rotational symmetry explains. The full story — stereographic projections, $SU(2)\times SU(2)$ , and Pauli's 1926 tour de force — belongs to the hidden-symmetry article; here, simply register that the frozen arrow you can draw on an ellipse is the classical shadow of that larger rotational structure.

Which tool answers which question?

you are asked for	reach for
is the orbit's axis fixed, and why	conservation of $\mathbf{A}$ , equation (2)
the orbit equation, fastest route	dot $\mathbf{A}$ with $\mathbf{r}$ , equation (6)
eccentricity from dynamical data	$e=A/(mk)$ with $A^2=m^2k^2+2mEL^2$
which way perihelion points	the direction of $\mathbf{A}$ (eccentricity vector)
precession under a perturbation	the drive equation $\dot{\mathbf{A}}=\mathbf{F}_{\text{pert}}\times\mathbf{L}$
hydrogen's extra degeneracy	the $SO(4)$ algebra of $(\mathbf{L},\mathbf{D})$ , equation (8)

Limiting cases

Case	Result	Meaning
Circular orbit	$e=0\Rightarrow\mathbf{A}=\mathbf{0}$	At the energy minimum $E=-mk^2/(2L^2)$ , (4) gives $A=0$ : no perihelion exists, so no direction is singled out — full rotational symmetry in the plane.
Bound ellipse	$0<e<1$ , $A=mke$ , $E<0$	The generic case: a frozen in-plane arrow of fixed length pointing at perihelion; the orbit closes exactly.
Parabolic escape	$e=1$ , $A=mk$ , $E=0$	$\mathbf{A}$ survives unbound motion, pointing along the parabola's axis of symmetry.
Hyperbolic flyby	$e>1$ , $A=mke$ , $E>0$	Still conserved; $\mathbf{A}$ points at the closest-approach direction and fixes the scattering geometry — the hidden backbone of the Rutherford cross-section.
Non-Kepler force	$\dot{\mathbf{A}}=\mathbf{F}_{\text{pert}}\times\mathbf{L}\neq\mathbf{0}$	The arrow rotates in-plane; its per-orbit rotation is the apsidal (perihelion) precession — zero exactly and only for pure $1/r^2$ .

Common trapEvery textbook's A is the same vector

It is the same direction but rarely the same object, and mixed conventions quietly corrupt homework. Goldstein's $\mathbf{A}=\mathbf{p}\times\mathbf{L}-mk\hat{\mathbf{r}}$ has magnitude $mke$ ; Landau–Lifshitz uses the per-mass version $\mathbf{v}\times\mathbf{L}/m$ -flavoured constructions; orbital mechanics texts use the dimensionless eccentricity vector $\mathbf{e}=\mathbf{A}/(mk)$ with magnitude plain $e$ ; and quantum texts symmetrize the operator ordering. Before comparing any two formulas, check whether the object in hand has dimensions of momentum-times-angular-momentum, velocity-times-angular-momentum, or nothing. A second convention trap: $m$ here is the reduced mass of the two-body problem, not the orbiting body's mass — for planets the difference is small, for binary stars it is not. And a physics trap to finish: $\mathbf{A}$ is conserved only for exactly $V=-k/r$ ; the moment your problem adds any other central term, reach for $\dot{\mathbf{A}}=\mathbf{F}_{\text{pert}}\times\mathbf{L}$ , not for conservation.

Worked examples

1 · Earth's eccentricity from one perihelion snapshot (confidence builder)

At perihelion (early January), Earth is $r_p=1.4710\times10^{11}\ \mathrm{m}$ from the Sun moving at $v_p=3.029\times10^{4}\ \mathrm{m/s}$ . Using $\kappa=GM_\odot=1.3271\times10^{20}\ \mathrm{m^3/s^2}$ , find the eccentricity vector of Earth's orbit from the Runge–Lenz construction (work per unit mass, so $\mathbf{e}=(\mathbf{v}\times\mathbf{h})/\kappa-\hat{\mathbf{r}}$ with $\mathbf{h}=\mathbf{r}\times\mathbf{v}$ ).

Step 1. At perihelion the velocity is purely tangential, so the specific angular momentum is simply $h=r_pv_p=1.4710\times10^{11}\times3.029\times10^{4}=4.456\times10^{15}\ \mathrm{m^2/s}$ .

Step 2. The vector $\mathbf{v}\times\mathbf{h}$ at perihelion points radially outward (velocity tangential, $\mathbf{h}$ out of plane) with magnitude $v_ph=3.029\times10^{4}\times4.456\times10^{15}=1.350\times10^{20}\ \mathrm{m^3/s^2}$ .

Step 3. Assemble the eccentricity vector: both terms are radial, so $\mathbf{e}=\left(\dfrac{v_ph}{\kappa}-1\right)\hat{\mathbf{r}}=\left(\dfrac{1.350\times10^{20}}{1.3271\times10^{20}}-1\right)\hat{\mathbf{r}}=\boxed{0.0170\,\hat{\mathbf{r}}}$ — pointing along the perihelion direction, magnitude matching Earth's tabulated $e=0.0167$ to the precision of the inputs.

Step 4. Cross-check through the energy route: $\varepsilon=\tfrac12v_p^2-\kappa/r_p=\tfrac12(3.029\times10^4)^2-\dfrac{1.3271\times10^{20}}{1.4710\times10^{11}}=-4.43\times10^{8}\ \mathrm{J/kg}$ , and $e=\sqrt{1+2\varepsilon h^2/\kappa^2}=\boxed{0.0170}$ ✓ — equation (4) per unit mass, agreeing with the direct construction. The moral: one instantaneous state, no orbit integration, and the frozen arrow hands you both the shape of the orbit and the direction of its axis.

2 · Mercury's 43 arcseconds (exam level)

General relativity adds an effective perturbation to the Sun's Newtonian pull, breaking the conservation of $\mathbf{A}$ and rotating it by $\Delta\phi=\dfrac{6\pi GM_\odot}{c^2a(1-e^2)}$ per orbit. For Mercury ( $a=5.791\times10^{10}\ \mathrm{m}$ , $e=0.2056$ , $T=87.97$ days), find the perihelion advance per orbit and per century, with $c=2.998\times10^{8}\ \mathrm{m/s}$ .

Step 1. Assemble the denominator's pieces: $1-e^2=1-0.2056^2=0.9577$ , and $c^2a(1-e^2)=(2.998\times10^{8})^2\times5.791\times10^{10}\times0.9577=4.985\times10^{27}\ \mathrm{m^3/s^2}$ .

Step 2. The numerator: $6\pi GM_\odot=6\pi\times1.3271\times10^{20}=2.502\times10^{21}\ \mathrm{m^3/s^2}$ .

Step 3. Divide: $\Delta\phi=\dfrac{2.502\times10^{21}}{4.985\times10^{27}}=5.02\times10^{-7}\ \mathrm{rad}$ per orbit. Tiny — the Runge–Lenz arrow rotates by half a microradian per revolution.

Step 4. Convert one orbit's creep to arcseconds: $5.02\times10^{-7}\times\dfrac{180}{\pi}\times3600=\boxed{0.104''\ \text{per orbit}}$ .

Step 5. Count Mercury's orbits in a century: $N=36{,}525/87.97=415.2$ orbits, so the accumulated rotation of $\mathbf{A}$ is $0.104''\times415.2=\boxed{43.0''\ \text{per century}}$ — precisely the anomalous residue that Le Verrier could not explain with planetary perturbations and that Einstein's 1915 calculation recovered. In this article's language: measure how fast the frozen arrow actually creeps, subtract every known Newtonian $\mathbf{F}_{\text{pert}}$ , and the leftover creep is new physics. The two examples bracket the vector's career: when it is conserved it hands you orbits for free; when it is not, its drift rate is one of the most sensitive probes of the force law in physics.

Origins · 1 minPierre-Simon Laplace and how this was discovered

Pierre-Simon Laplace

1749–1827 · Paris

Laplace put the vector to systematic work in his Traité de mécanique céleste (1799), deriving all three of its components as integrals of the two-body motion and wielding them in the perturbation theory with which he set out to prove the Solar System stable. Characteristically, he was neither the first nor the last: Jakob Hermann had exhibited the conserved quantity in 1710, Johann Bernoulli generalized it within the year, Hamilton rediscovered it as his "eccentricity vector" in 1845, and the modern name attached through Carl Runge's popular vector-analysis textbook and Wilhelm Lenz's 1924 use of it in the old quantum theory of hydrogen — a chain of rediscovery so tangled that the vector is physics' favourite example of Stigler's law. Laplace's broader programme survives him: the eccentricity-vector methods of his Mécanique céleste are, almost unchanged, the secular perturbation theory running inside modern planetary-dynamics codes.

History

1710

— Jakob Hermann exhibits the conserved vector for inverse-square motion; Johann Bernoulli generalizes it the same year.

1799

— Laplace derives it systematically in Mécanique céleste as three scalar integrals of the two-body problem.

1845

— Hamilton independently rediscovers it as the "eccentricity vector" via his hodograph — the velocity-space circle of Kepler motion.

1924

— Lenz applies the vector (from Runge's textbook treatment) to hydrogen in the old quantum theory, welding the modern three-part name.

1926

— Pauli uses the operator algebra of A and L to derive the hydrogen spectrum before Schrödinger's equation is solved for it.

Going further

References. Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., §3.9 · Morin, Introduction to Classical Mechanics, Ch. 7 (problems) · Landau & Lifshitz, Mechanics, Vol. 1, §15 (problem on conserved vector) · Taylor, Classical Mechanics, Ch. 8 · Goldstein §9 (Poisson brackets) and Appendix on the symmetry group

Provenance. Definition and conservation derivation cross-checked against Goldstein §3.9 (3rd ed.) and Taylor Ch. 8. The verifier should re-derive $\dot{\mathbf{A}}=0$ via the BAC-CAB step, confirm $A^2=m^2k^2+2mEL^2$ by squaring the definition, and verify $e=\sqrt{1+2EL^2/(mk^2)}$ . The $SO(4)$ Poisson-bracket algebra matches Goldstein Ch. 9 and Pauli's 1926 hydrogen treatment. Mercury's 43 arcsec/century is the standard textbook GR value; numbers used are textbook-standard. Worked-example numbers recomputed: Earth perihelion snapshot ( $r_p=1.4710\times10^{11}$ m, $v_p=3.029\times10^4$ m/s) gives $e=0.0170$ by both the eccentricity-vector and energy routes; Mercury GR advance $5.02\times10^{-7}$ rad/orbit $=0.104''$ /orbit $\times415.2$ orbits/century $=43.0''$ /century.

Builds on: Central Force Fundamentals · Angular Momentum Conservation in Central Fields · The Kepler Problem · The Orbit Equation and Binet's Equation Leads to: Conic-Section Orbits · Kepler's First Law: Elliptical Orbits · Classification of Orbits by Energy

How can a vector built from rapidly changing quantities be constant?

Because conservation is a property of combinations, not ingredients — energy already works this way ( $\tfrac12mv^2$ and $-k/r$ both vary wildly; their sum does not), and $\mathbf{A}$ is the vector-valued sibling of that idea. The mechanism is visible in the proof: the dynamical rate of change of $\mathbf{p}\times\mathbf{L}$ , driven by the force, happens to equal $mk$ times the purely kinematic rate at which $\hat{\mathbf{r}}$ swings around. The definition (1) subtracts exactly the right multiple of $\hat{\mathbf{r}}$ to cancel that motion. What is genuinely special is not that some frozen combination exists but that a frozen combination exists pointing in a fixed spatial direction — that is one conserved direction more than a generic central force can afford, and it is paid for by the $1/r^2$ conspiracy.

Why is it named after Laplace, Runge and Lenz when none of them discovered it?

Because naming followed usage, not priority — the vector is a textbook case of Stigler's law. Jakob Hermann found the conserved quantity in 1710 and Johann Bernoulli generalized it immediately; Laplace made it systematic in 1799; Hamilton rediscovered it in 1845 through his hodograph. The modern name congealed in the 1920s: Carl Runge's widely used vector-analysis textbook presented the construction, Wilhelm Lenz cited Runge when applying it to hydrogen in the old quantum theory (1924), and Pauli's famous 1926 paper cited Lenz. Physicists learned it from that chain, so the last three links got the credit. It is a good calibration lesson: the history of a name and the history of an idea are different documents.

Is A a new piece of information, or is it secretly just E and L again?

Partly new — and the "partly" is precise. Its magnitude is old news: $A=mke$ is fixed by $E$ and $L$ through (4). Its direction within the orbital plane is genuinely new information: no function of $E$ and $\mathbf{L}$ can tell you which way the ellipse's long axis points, because $E$ and $\mathbf{L}$ are unchanged if you rotate the whole orbit within its plane. That one new angle is exactly what upgrades the conserved-quantity count from four ( $E$ , $\mathbf{L}$ ) to five, and five is the magic number: it pins the orbit completely, leaving only the body's phase along the curve — which no constant could ever pin — undetermined.

Does the vector still mean anything for a hyperbolic flyby?

Yes — conservation never required a bound orbit, and $\mathbf{A}$ is arguably more useful for scattering. For $E>0$ equation (4) gives $A=mke$ with $e>1$ , and the arrow points along the hyperbola's symmetry axis toward the point of closest approach. Fix the incoming asymptote and the impact parameter, and $\mathbf{A}$ tells you the orientation of the outgoing asymptote — i.e. the scattering angle — algebraically. That is the classical skeleton inside the Rutherford cross-section, and it is also why the (repulsive-force) Coulomb scattering problem is exactly solvable: the same hidden symmetry that closes bound orbits organizes unbound ones, with the $SO(4)$ algebra of the bound problem trading into $SO(3,1)$ for $E>0$ .

Why does a circular orbit have A equal to zero rather than pointing somewhere?

Because there is nowhere for it to point. $\mathbf{A}$ 's job is to mark the perihelion direction, and a circle has no perihelion — every point is equidistant. The algebra enforces this democratic verdict: at the circular-orbit energy $E=-mk^2/(2L^2)$ , equation (4) gives $A^2=m^2k^2+2mEL^2=0$ identically. There is a symmetry reading too, and it runs deep: a nonzero $\mathbf{A}$ breaks rotational symmetry within the orbital plane by selecting an axis, so the circle — the one orbit that keeps full in-plane rotational symmetry — is exactly the one where $\mathbf{A}$ must vanish. Small $e$ then measures small symmetry breaking, which is why the eccentricity vector is the natural "order parameter" of orbit shape in perturbation theory.

What is actually four-dimensional about the SO(4) symmetry?

Not physical space — the fourth dimension lives in phase space. The statement is algebraic: the six conserved quantities $\mathbf{L}$ and $\mathbf{D}=\mathbf{A}/\sqrt{-2mE}$ have exactly the Poisson-bracket relations that the six rotation generators of a four-dimensional Euclidean space obey. There is a beautiful geometric realization behind the algebra: for bound orbits, the momentum vector traces a perfect circle (Hamilton's hodograph), and stereographically projecting momentum space onto a 3-sphere turns Kepler motion into free rotation on that sphere — genuine $SO(4)$ rotations of a genuine (abstract) 4-D object. The upshot is that the Kepler problem has more symmetry than it looks like it has, and symmetries that do not come from spacetime transformations are called dynamical or hidden — the theme the hidden-symmetry article develops properly.

How exactly does this classical vector explain hydrogen's degeneracy?

Hydrogen is a quantum Kepler problem ( $V=-e^2/4\pi\epsilon_0 r$ is a $-k/r$ potential), so the quantum versions of $\mathbf{L}$ and $\mathbf{A}$ (suitably symmetrized) are still conserved — they commute with the Hamiltonian. Rotational symmetry alone forces states of the same $n$ and $\ell$ but different $m$ to share an energy: a $(2\ell+1)$ -fold degeneracy. But hydrogen's levels are degenerate across different $\ell$ at fixed $n$ — the 2s and 2p states coincide — which rotations cannot explain. The operator $\mathbf{A}$ is precisely the missing symmetry generator: it moves between different- $\ell$ states of the same energy, gluing them into single multiplets of the larger $SO(4)$ group, of dimension $n^2$ . Pauli inverted the logic in 1926: from the algebra of $\mathbf{L}$ and $\mathbf{A}$ alone he derived $E_n\propto-1/n^2$ , months before wave mechanics could.

If A is conserved, why do real planetary perihelia precess at all?

Because no real orbit feels an exactly $-k/r$ potential, and $\mathbf{A}$ 's conservation has zero tolerance. Real perturbations — the pulls of the other planets (by far the largest term for Mercury, some $530''$ /century), the Sun's slightly oblate mass distribution, and general relativity's effective correction — all enter $\dot{\mathbf{A}}=\mathbf{F}_{\text{pert}}\times\mathbf{L}$ and set the arrow creeping. The historical drama is in the accounting: Le Verrier summed every known Newtonian creep for Mercury and found the observed total exceeded it by $\approx43''$ per century — so stubborn a residue that a new planet ("Vulcan") was hunted for decades. Einstein's 1915 computation of the GR term landed on the missing $43''$ exactly, the first great test of general relativity, and the worked example above reproduces it in five steps.