The Laplace–Runge–Lenz Vector in Quantum Mechanics

Kepler orbits have a suspicious perfection: they close. A planet in a $1/r$ potential retraces the same ellipse forever, perihelion pinned in space — almost any other force law precesses into a rosette. Perfection like that is never an accident: behind it stands a conservation law beyond energy and angular momentum — the Laplace–Runge–Lenz vector, an arrow of fixed length locked along the major axis. Mercury's perihelion, creeping 43 arcseconds per century under general relativity's non- $1/r$ correction, is this vector slowly failing to be conserved. The quantum Coulomb problem inherits the structure — and sharpens it. In 1926, months before Schrödinger's equation was in print, Pauli quantized the vector and extracted the complete hydrogen spectrum from algebra alone. But the first step demands care: the classical expression contains $\mathbf{p}\times\mathbf{L}$ , whose naive operator version is not even Hermitian. The repair — symmetrization — plants a purely quantum $\hbar^2$ into the vector's algebra, the fingerprint that makes the levels land on integer $n$ . This article builds the operator and proves its credentials: Hermiticity, conservation, orthogonality to $\hat{\mathbf{L}}$ , the $\hat{A}^2$ identity — and shows it working as the degeneracy shuffler that turns a $2s$ state into a $2p$ at the same energy. What the credentials buy — the $SO(4)$ algebra and the spectrum without a differential equation — is the story of Accidental Degeneracy and the SO(4) Symmetry.

Key takeaways

The Laplace–Runge–Lenz (LRL) vector is the Kepler problem's third conserved quantity: it points along the major axis with length $mke$ , and its constancy is the non-precession of closed orbits.
The naive $\hat{\mathbf{p}}\times\hat{\mathbf{L}}$ is not Hermitian; the honest operator is $\hat{\mathbf{A}} = \tfrac{1}{2m}(\hat{\mathbf{p}}\times\hat{\mathbf{L}} - \hat{\mathbf{L}}\times\hat{\mathbf{p}}) - k\,\hat{\mathbf{r}}/r$ , $k = Ze^2/4\pi\varepsilon_0$ .
For the Coulomb Hamiltonian — and only for $1/r$ — $[\hat{H}, \hat{\mathbf{A}}] = 0$ : genuinely conserved.
Credentials pack: $\hat{\mathbf{A}}\cdot\hat{\mathbf{L}} = \hat{\mathbf{L}}\cdot\hat{\mathbf{A}} = 0$ and $\hat{A}^2 = k^2 + \tfrac{2\hat{H}}{m}(\hat{L}^2 + \hbar^2)$ — the $+\hbar^2$ is symmetrization's quantum fingerprint, absent classically.
$\hat{\mathbf{A}}$ is a vector operator that commutes with $\hat{H}$ but not $\hat{L}^2$ : it shuffles $\ell$ within a shell — concretely, $\hat{A}_z|2s\rangle = \tfrac{k}{2}|2p_0\rangle$ .
Key idea: the Coulomb problem conserves a second vector; its quantum credentials, proven here, power the algebraic solution of hydrogen.

The core insight

The $1/r$ problem conserves, beyond $\hat{H}$ and $\hat{\mathbf{L}}$ , the Hermitian operator $\hat{\mathbf{A}} = \tfrac{1}{2m}(\hat{\mathbf{p}}\times\hat{\mathbf{L}} - \hat{\mathbf{L}}\times\hat{\mathbf{p}}) - k\hat{\mathbf{r}}/r$ — a vector operator commuting with the Hamiltonian, orthogonal to $\hat{\mathbf{L}}$ , satisfying $\hat{A}^2 = k^2 + \tfrac{2\hat{H}}{m}(\hat{L}^2 + \hbar^2)$ , and mixing $\ell$ within one energy shell.

Where this lives today

The LRL vector is the working tool wherever hydrogen's shell degeneracy is an engineering resource. Circular Rydberg states — the $\ell = n-1$ , $e_q \to 0$ corner of Derivation C's table — are sculpted through transition sequences most cleanly described in the LRL ladder basis; with tens-of-milliseconds lifetimes they are current hardware for quantum simulation and cavity QED. The same shell-mixing underlies the linear Stark effect: a field couples degenerate $\ell$ states through exactly the equation-(6) elements — why hydrogen responds linearly while non-degenerate atoms respond quadratically (theory in the approximation-methods chapter). As a diagnostic, non-conservation of $\mathbf{A}$ remains the cleanest way to say "not exactly $1/r$ ." Mercury's 43 arcseconds per century is the classical flagship; in atoms, alkali quantum-defect shifts measure how a finite core rotates the vector; in Rydberg molecules and exotic atoms the ellipse's precession rate probes the perturbing environment directly. The vector even organizes momentum space: Fock's 1935 construction — told at Accidental Degeneracy and the SO(4) Symmetry — makes hydrogen's bound states free rotations on a 3-sphere, with $\hat{\mathbf A}$ tipping the sphere's poles.

The physical picture

Classically the story fits in one paragraph. For $H = p^2/2m - k/r$ , the vector $\mathbf{A} = \mathbf{p}\times\mathbf{L} - mk\,\hat{\mathbf{r}}$ is constant along every trajectory: it points from focus to perihelion with magnitude $mk$ times the eccentricity. Constant direction means no precession; constant length, a shape that never breathes. Any deviation from pure $1/r$ — an oblate sun, a relativistic correction, a screening cloud — sets $\mathbf{A}$ rotating: that is how precession is diagnosed. The classical derivation belongs to the mechanics track; here we take the vector as found and ask what survives quantization.

Figure

The arrow that keeps orbits closed — and betrays every deviation

Quantum mechanically the vector earns its keep more subtly. $\hat{\mathbf{A}}$ commutes with $\hat{H}$ but not with $\hat{L}^2$ — conserved, yet it changes $\ell$ : on a hydrogen eigenstate it produces another of the same energy with $\ell$ shifted by one, rotating $2s$ into $2p$ without spending a joule. That is the mechanism behind the "accidental" degeneracy — one shell's states are connected by a conserved operator, as an $m$ -multiplet is connected by $\hat{L}_\pm$ (Degeneracy and the Role of the Magnetic Quantum Number, one rung down in ambition).

Figure

The degeneracy shuffler — one conserved operator walks the n = 2 shell

Why "hidden"? Rotations act on coordinates, and you can see them. This conservation law lives in the shape of the dynamics — it exists because $1/r$ is special, not because space is isotropic. The names record repeated rediscovery; the portrait band tells that story.

Building the operator

Fix conventions once, with the Coulomb Hamiltonian:

\hat{H} = \frac{\hat{p}^2}{2m} - \frac{k}{r}, \qquad k \equiv \frac{Ze^2}{4\pi\varepsilon_0} \qquad (1)

where: $m$ — the (reduced) electron mass; $k$ — the Coulomb coupling, $Z = 1$ for hydrogen. Classically $\mathbf{A} = \mathbf{p}\times\mathbf{L} - mk\,\hat{\mathbf{r}}$ ; the temptation is to hat every symbol and move on. But $\hat{p}_i$ and $\hat{L}_j$ do not commute, and Derivation A finds the adjoint of $\hat{\mathbf{p}}\times\hat{\mathbf{L}}$ is $-\hat{\mathbf{L}}\times\hat{\mathbf{p}}$ — a different operator. An observable must be Hermitian, so the honest vector splits the difference:

\hat{\mathbf{A}} = \frac{1}{2m}\left(\hat{\mathbf{p}}\times\hat{\mathbf{L}} - \hat{\mathbf{L}}\times\hat{\mathbf{p}}\right) - k\,\frac{\hat{\mathbf{r}}}{r} \qquad (2)

where: $\hat{\mathbf{L}} = \hat{\mathbf{r}}\times\hat{\mathbf{p}}$ ; $\hat{\mathbf{r}}/r$ — the unit radial operator (multiplication, unambiguous since $\hat{\mathbf r}$ commutes with itself). With this $1/m$ normalization $|\langle\hat{\mathbf A}\rangle| \to k\,e$ in the classical limit — the vector measures eccentricity in units of $k$ . Some texts (Pauli's original included) keep the classical $m\mathbf{A}$ normalization; identities convert by the obvious factor.

Derivation AWhy the naive operator fails and what the repair costs

The goal is the correct Hermitian quantization of $\mathbf{p}\times\mathbf{L}$ — and an exact accounting of the price. The strategy is bare-hands operator algebra: compute the naive product's adjoint, find it is a different operator, identify the Hermitian combination, and compress it to something computable. Watch $[\hat{p}_i, \hat{L}_j] = i\hbar\,\epsilon_{ijk}\hat{p}_k$ do all the work — the only quantum input, and the source of the leftover $\hbar$ .

Step 1. Take the adjoint of the naive product. With $(\hat{\mathbf{p}}\times\hat{\mathbf{L}})_i = \epsilon_{ijk}\hat{p}_j\hat{L}_k$ and both factors Hermitian,

\left[(\hat{\mathbf{p}}\times\hat{\mathbf{L}})_i\right]^\dagger = \epsilon_{ijk}\hat{L}_k\hat{p}_j = -(\hat{\mathbf{L}}\times\hat{\mathbf{p}})_i.

Why: daggering reverses the order; relabeling the summed indices costs one sign. The adjoint differs from the original — the naive quantization is not an observable.

Step 2. Measure the gap between the two orderings. Using $[\hat{p}_j, \hat{L}_k] = i\hbar\,\epsilon_{jkl}\hat{p}_l$ and the contraction $\epsilon_{ijk}\epsilon_{jkl} = 2\delta_{il}$ ,

(\hat{\mathbf{p}}\times\hat{\mathbf{L}})_i + (\hat{\mathbf{L}}\times\hat{\mathbf{p}})_i = \epsilon_{ijk}[\hat{p}_j, \hat{L}_k] = 2i\hbar\,\hat{p}_i.

Why: the two orderings differ by exactly $2i\hbar\hat{\mathbf p}$ — an anti-Hermitian offset. This identity lets us trade orderings at will.

Step 3. Keep the Hermitian half — self-adjoint by Step 1, compressed by Step 2:

\tfrac12\left(\hat{\mathbf{p}}\times\hat{\mathbf{L}} - \hat{\mathbf{L}}\times\hat{\mathbf{p}}\right) = \hat{\mathbf{p}}\times\hat{\mathbf{L}} - i\hbar\,\hat{\mathbf{p}}.

Why: the right side is the workhorse form — one cross product plus a counterterm that precisely cancels the anti-Hermitian part of $\hat{\mathbf p}\times\hat{\mathbf L}$ .

Step 4. Name the price. The counterterm propagates into every quadratic identity: squaring $\hat{\mathbf{A}}$ produces $\hat{L}^2 + \hbar^2$ where the classical calculation has $L^2$ (Derivation C). That $\hbar^2$ is no blemish — it shifts the algebraic spectrum onto the integers, as Accidental Degeneracy and the SO(4) Symmetry cashes out. Verified: Step 2's sum rule checked componentwise on a generic function (sympy); the adjoint relation and the naive operator's non-Hermiticity gap exhibited numerically on a spectral grid; the symmetrized $\hat A_z$ passes the same test at $10^{-9}$ .

Conservation — and why only the Coulomb potential

A conserved quantity must commute with the Hamiltonian; here the claim is exact, not shell-by-shell:

[\hat{H}, \hat{\mathbf{A}}] = 0 \qquad (3)

where: $\hat{H}$ is the Coulomb Hamiltonian (1) — the statement fails for any other central potential. Each term of $\hat{\mathbf{A}}$ is separately non-conserved; conservation is a conspiracy between the two failures.

Go deeper · Derivation B · 1 minConservation of the LRL operator — the Coulomb-only cancellation

Derivation BConservation of the LRL operator — the Coulomb-only cancellation

The goal is equation (3), the full operator statement $[\hat{H}, \hat{A}_z] = 0$ — and visibility into why the cancellation happens only for $1/r$ . The strategy: split the commutator into its two cross terms, compute each with the canonical rules, display the decisive cancellation, and hand the page of intermediate algebra to the machine verifier. One component suffices — the vector-operator property (equation 4 below) generates the rest by rotation.

Step 1. Split by source. Writing $\hat{H} = \hat{T} + \hat{V}$ with $\hat{T} = \hat{p}^2/2m$ , $\hat{V} = -k/r$ , and $\hat{A}_z = \tfrac{1}{m}[(\hat{\mathbf{p}}\times\hat{\mathbf{L}})_z - i\hbar\hat{p}_z] - k\hat{z}/r$ ,

[\hat{H}, \hat{A}_z] = \underbrace{\left[\hat{T},\, -k\,\frac{\hat z}{r}\right]}_{\text{kinetic hits the radial term}} + \underbrace{\left[\hat{V},\, \tfrac{1}{m}\big((\hat{\mathbf{p}}\times\hat{\mathbf{L}})_z - i\hbar\hat{p}_z\big)\right]}_{\text{potential hits the momentum term}},

since $[\hat{T}, \hat{\mathbf{p}}\times\hat{\mathbf{L}}] = 0$ (kinetic commutes with $\hat{\mathbf p}$ and $\hat{\mathbf L}$ ) and $[\hat{V}, \hat z/r] = 0$ (functions of position commute). Why: only the "mixed" pieces survive — each term of $\hat A_z$ is attacked by the part of $\hat H$ it does not resemble.

Step 2. Compute the first piece. Using $[\hat p^2, f(\hat{\mathbf r})] = -\hbar^2(\nabla^2 f) - 2i\hbar(\nabla f)\cdot\hat{\mathbf{p}}$ with $f = z/r$ :

\left[\hat{T}, -\frac{k\hat z}{r}\right] = \frac{k}{2m}\left(-\hbar^2\,\nabla^2\frac{z}{r} - 2i\hbar\left(\nabla\frac{z}{r}\right)\cdot\hat{\mathbf{p}}\right), \quad \nabla\frac{z}{r} = \frac{\hat{\mathbf{e}}_z}{r} - \frac{z\,\mathbf{r}}{r^3}.

Why: note the two structures — $\propto \hat p_z/r$ and $\propto z(\mathbf r\cdot\hat{\mathbf p})/r^3$ — that the second piece must match.

Step 3. Compute the second piece. With $[\hat p_j, g(\hat{\mathbf r})] = -i\hbar\,\partial_j g$ and $g = -k/r$ , each momentum factor in $(\hat{\mathbf{p}}\times\hat{\mathbf{L}})_z - i\hbar \hat p_z$ trades one $\hat{p}$ for $-i\hbar k\,\mathbf{r}/r^3$ . The cross product then yields exactly Step 2's two structures, plus matching $\hbar^2$ contact pieces — with opposite signs. Why: the match is not generic. For $V = -k/r^s$ the gradient scales as $r^{-(s+1)}$ while the kinetic piece keeps its $1/r^2$ weight from $\nabla(z/r)$ : alignment happens only at $s = 1$ ; any other power leaves a residue $\propto (s-1)$ — the operator echo of Bertrand's theorem.

Step 4. Add. Every term of Step 2 finds its negative in Step 3; the sum collapses to zero identically — not merely in expectation, not merely on a shell. Verified: $[\hat H, \hat A_z]\psi = 0$ ground out symbolically (sympy, atomic units) on three independent test functions — $e^{-r}$ , $z e^{-r/2}$ , $(2-r)e^{-r/2}$ — dozens of terms cancelling to exactly 0 each time; matches Pauli (1926) and Sakurai §3.9 conventions.

One structural fact matters constantly: $\hat{\mathbf{A}}$ is a vector operator — its components respond to rotations exactly as $\hat{\mathbf r}$ and $\hat{\mathbf p}$ do,

[\hat{L}_i, \hat{A}_j] = i\hbar\,\epsilon_{ijk}\hat{A}_k \qquad (4)

where: the bracket is the universal rotation grammar — an instance of the vector-operator definition owned by Angular Momentum Commutation Relations and systematized in Spherical Tensor Operators, not a new derivation. It is why one component's conservation proves all three, and it hands $\hat{\mathbf{A}}$ the $k = 1$ selection rules: $\Delta\ell = \pm 1$ (parity kills $\Delta\ell = 0$ ), $\Delta m = 0$ for $\hat{A}_z$ . With $[\hat H, \hat{\mathbf A}] = 0$ forbidding changes of $n$ , the shuffle is completely pinned: within a shell, along $\ell$ , nowhere else.

The credentials pack

Two operator identities carry the downstream weight. Accidental Degeneracy and the SO(4) Symmetry quotes both and spends them on the spectrum; they are proven here.

\hat{\mathbf{A}}\cdot\hat{\mathbf{L}} = \hat{\mathbf{L}}\cdot\hat{\mathbf{A}} = 0, \qquad \hat{A}^2 = k^2 + \frac{2\hat{H}}{m}\left(\hat{L}^2 + \hbar^2\right) \qquad (5)

where: both orderings vanish — not obvious, since $\hat A_i$ and $\hat L_i$ do not commute — and the $\hat{A}^2$ identity is the classical one with an extra $\hbar^2$ . On $|n\ell m\rangle$ it evaluates to the quantum eccentricity — worked in Derivation C, audited in Worked example 2.

Go deeper · Derivation C · 1 minOrthogonality to L and the A-squared identity

Derivation COrthogonality to L and the A-squared identity

The goal is both statements of equation (5): the double orthogonality and the closed form for $\hat{A}^2$ — the two credentials the algebraic solution of hydrogen consumes. The strategy is structure before grinding: orthogonality from triple-product geometry plus one commutator trade; the $\hat{A}^2$ identity as a mechanical expansion whose quantum surplus we isolate. Watch where each $\hbar$ enters — every one traces back to Derivation A's counterterm.

Step 1. Orthogonality, term by term. Dot the workhorse form into $\hat{\mathbf{L}}$ : $(\hat{\mathbf{p}}\times\hat{\mathbf{L}})\cdot\hat{\mathbf{L}} = \hat{\mathbf p}\cdot(\hat{\mathbf L}\times\hat{\mathbf L}) = i\hbar\,\hat{\mathbf{p}}\cdot\hat{\mathbf{L}}$ , using the operator curiosity $\hat{\mathbf{L}}\times\hat{\mathbf{L}} = i\hbar\hat{\mathbf{L}}$ . Meanwhile $\hat{\mathbf{p}}\cdot\hat{\mathbf{L}} = 0$ and $\hat{\mathbf{r}}\cdot\hat{\mathbf{L}} = 0$ exactly (triple products with a repeated vector, safe because the offending components commute). Every term of $\hat{\mathbf{A}}\cdot\hat{\mathbf{L}}$ vanishes. Why: classically $\mathbf A$ lies in the orbital plane, perpendicular to $\mathbf L$ ; the statement survives ordering because the dangerous terms carry their own $\epsilon$ -antisymmetry.

Step 2. The reversed order. $\hat{\mathbf{L}}\cdot\hat{\mathbf{A}} = \hat{\mathbf{A}}\cdot\hat{\mathbf{L}} + [\hat{L}_i, \hat{A}_i]$ , and equation (4) gives $[\hat{L}_i, \hat{A}_i] = i\hbar\,\epsilon_{iik}\hat{A}_k = 0$ . Why: the trace of an antisymmetric symbol — orthogonality is ordering-proof for any vector operator.

Step 3. Square the operator. Expanding $\hat{A}^2$ with the workhorse form gives three blocks — pure momentum, pure radial ( $k^2$ : the unit vector squares to one), and cross. Repeated use of $[\hat p_j, \hat L_k]$ , $[\hat p_j, f(r)]$ and the $\epsilon$ -contractions collapses them to

\hat{A}^2 = \frac{\hat{p}^2}{m^2}\left(\hat{L}^2 + \hbar^2\right) - \frac{2k}{m}\,\frac{1}{r}\left(\hat{L}^2 + \hbar^2\right) + k^2 = k^2 + \frac{2\hat{H}}{m}\left(\hat{L}^2 + \hbar^2\right).

Why: the second equality recognizes $\hat{p}^2/2m - k/r = \hat H$ in front of the common factor. Classically the same expansion gives $2HL^2/m$ — the $+\hbar^2$ is Derivation A's counterterm come home, and no Hermitian ordering removes it.

Step 4. Evaluate on a shell. On $|n\ell m\rangle$ , set $\hat H \to E_n = -mk^2/2\hbar^2n^2$ (the spectrum of The Hydrogen Atom: Coulomb Radial Equation, consumed by citation) and $\hat L^2 \to \hbar^2\ell(\ell+1)$ :

\langle\hat{A}^2\rangle_{n\ell} = k^2\left(1 - \frac{\ell(\ell+1) + 1}{n^2}\right) \equiv k^2\,e_q^2.

Circular states $\ell = n-1$ give $e_q^2 = (n-1)/n^2 \to 0$ — high circular Rydberg states really are round — while $2s$ carries $e_q^2 = 3/4$ : as eccentric as its shell allows. Verified: $\hat A^2\psi_{100} = 0$ and $\hat A^2\psi_{210} = \tfrac14 k^2\psi_{210}$ ground out symbolically (sympy) — exactly $e_q^2(1,0) = 0$ and $e_q^2(2,1) = \tfrac14$ ; both dot-product orderings annihilate three test states; the eccentricity table checked to $(3,2)$ .

The shuffle exhibit makes this concrete. On the real $2s$ wavefunction the $(\hat{\mathbf p}\times\hat{\mathbf L})_z$ term dies ( $\hat{\mathbf L}$ annihilates $s$ -states), leaving $-\tfrac{i\hbar}{m}\hat p_z - k\hat z/r$ — two terms that assemble, with no residue, into a multiple of the $2p_0$ wavefunction:

\hat{A}_z\,|2s\rangle = \frac{k}{2}\,|2p_0\rangle, \qquad \hat{A}_z\,|2p_0\rangle = \frac{k}{2}\,|2s\rangle \qquad (6)

where: the states are the standard real hydrogen wavefunctions ( $\psi_{200}$ positive at the origin, $\psi_{210}$ positive on the $+z$ lobe); the second relation is the first plus Hermiticity — at $n = 2$ there is no $\ell = 2$ level to reach. One exact matrix element $k/2$ completes the hero diagram, with the consistency check $\langle\hat{A}^2\rangle_{2s} = 3(k/2)^2 = \tfrac34 k^2$ landing on Derivation C's table.

Limiting cases

Case	Result	Meaning
$\hbar \to 0$	$\hat{A}^2 \to k^2 + 2HL^2/m$ , $\lvert\mathbf{A}\rvert = ke$	Goldstein's classical identity; the fingerprint term vanishes
circular states $\ell = n-1$	$e_q^2 = (n-1)/n^2 \to 0$	high circular Rydberg states approach round classical orbits
$s$ states $\ell = 0$	$e_q^2 = 1 - 1/n^2$ , maximal in shell	the degenerate "straight-line" orbits — most eccentric available
non-Coulomb $\delta V$	$[\hat{H}, \hat{\mathbf{A}}] \ne 0$	precession; $\ell$ -degeneracy splits — fine structure, quantum defects
$E > 0$	rescaling turns imaginary	the algebra changes character — scattering story at Accidental Degeneracy and the SO(4) Symmetry

Common trapHatting the classical formula p × L − mk r̂ produces a non-observable

The classical vector quantized symbol-for-symbol is not Hermitian: $(\hat{\mathbf{p}}\times\hat{\mathbf{L}})^\dagger = -\hat{\mathbf{L}}\times\hat{\mathbf{p}}$ , the gap being exactly $2i\hbar\hat{\mathbf{p}}$ — use the naive form and your "expectation values" grow spurious imaginary parts. Symmetrizing is mandatory and physical, not cosmetic: it plants the $\hbar^2$ in $\hat{A}^2 = k^2 + \tfrac{2\hat H}{m}(\hat{L}^2 + \hbar^2)$ , and that term is what makes the algebraic spectrum land at $-\mathrm{Ry}/n^2$ with integer $n$ . A second trap: do not conclude from $[\hat H, \hat{\mathbf A}] = 0$ that $\hat{\mathbf A}$ shares eigenstates with $\hat{L}^2$ — it does not commute with $\hat{L}^2$ ; its whole job is to move $\ell$ within a shell. Conserved and $\ell$ -mixing at once is the defining strangeness.

Worked examples

1 · Auditing the classical vector on a real orbit (confidence builder)

Integrate a Kepler orbit with $m = k = 1$ , semi-major axis $a = 1$ , eccentricity $e = 0.6$ , and verify the conservation and the eccentricity reading of $\mathbf{A} = \mathbf{p}\times\mathbf{L} - m k\,\hat{\mathbf{r}}$ .

Step 1. At perihelion $r_p = a(1-e) = 0.4$ ; vis-viva gives $v_p = \sqrt{(1+e)/[a(1-e)]} = 2.000$ , so $L_z = r_p v_p = 0.8$ .

Step 2. Evaluate at the start: $A_x = p_y L_z - x/r = 2.0\times0.8 - 1 = \mathbf{0.600}$ , $A_y = 0$ — the vector points at perihelion with magnitude $0.6 = mk\,e$ , the eccentricity read off before integrating anything.

Step 3. Integrate six orbits (high-order Runge–Kutta, tolerance $10^{-12}$ — the portrait's other namesake auditing his own vector). The components hold constant to better than $10^{-10}$ while position and momentum swing by order one: conservation to the integrator's floor.

Step 4. Deform the law: with $V = -k/r^{1.02}$ the same integration shows $\mathbf{A}$ rotating steadily — a rosette in the making. Conservation belongs to $1/r$ , not to central forces generally; Derivation B's Step 3 is that exclusivity in operator language.

2 · Quantum eccentricity and the shuffle element (exam level)

For the $n = 2$ shell: compute the quantum eccentricities, the matrix element connecting $2s$ and $2p_0$ , and check the two against each other.

Step 1. Eccentricities from Derivation C: $e_q^2(2,0) = 1 - \tfrac{0+1}{4} = \tfrac34$ gives $e_q(2s) = \mathbf{0.866}$ ; $e_q^2(2,1) = \tfrac14$ gives $e_q(2p) = \mathbf{0.500}$ — the $s$ state the shell's most eccentric object, the $p$ its roundest.

Step 2. The shuffle element. Acting on $\psi_{200} = (2-r)e^{-r/2}/\sqrt{32\pi a_0^3}$ (atomic units, $k$ restored at the end), the cross-product term of $\hat{A}_z$ vanishes on $\ell = 0$ ; the remaining terms give $\hat{A}_z\psi_{200} = \tfrac{k}{2}\,\psi_{210}$ exactly, so $\langle 2p_0|\hat{A}_z|2s\rangle = \mathbf{+\,k/2}$ in these real conventions.

Step 3. Cross-check with the identity. On $2s$ : $\langle\hat{A}^2\rangle = k^2 e_q^2 = \tfrac34 k^2$ . By spherical symmetry the three components contribute equally, and $\hat{A}_z|2s\rangle$ has nowhere to go but $|2p_0\rangle$ : $\langle\hat A_z^2\rangle = (k/2)^2$ , and $3\times\tfrac{k^2}{4} = \tfrac34 k^2$ . The shuffle element and the eccentricity are the same fact.

Step 4. Read it physically: sandwiching $[\hat H, \hat A_z] = 0$ between the states gives $(E_{2s} - E_{2p})\langle 2p_0|\hat A_z|2s\rangle = 0$ , and the element is $k/2 \ne 0$ — so $E_{2s} = E_{2p}$ is forced. The degeneracy has a generator. Counting how far the argument reaches — all $n^2$ states — needs the algebra of Accidental Degeneracy and the SO(4) Symmetry.

Origins · 1 minCarl Runge and how this was discovered

Carl Runge

1856–1927 · Hanover and Göttingen

Runge would be baffled to find his name on hydrogen's hidden symmetry. A spectroscopist before he was "the father of numerical analysis," he spent two decades in Hanover measuring spectral lines with Kayser and hunting series formulas — parallel to Rydberg. The vector he neither discovered nor claimed: it appears in Jakob Hermann's work of 1710, in Laplace's Mécanique céleste (1799), and in Hamilton's hodograph — one of history's most rediscovered results. Runge's contribution was pedagogical: his 1919 Vektoranalysis textbook worked the Kepler vector as a tidy exercise in the new notation. Wilhelm Lenz — whose person the glossary does not carry, so his side of the story lives here — cited "Runge's vector" in his 1924 old-quantum-theory paper; Pauli's 1926 tour de force cited Lenz, and "Runge–Lenz" fossilized, Laplace sometimes restored up front, Hermann almost never. Runge's deeper legacy runs through this article anyway: the Runge–Kutta integrator (1895) is the machine Worked example 1 uses to verify his textbook exercise to $10^{-10}$ — the numerical analyst auditing the geometer, a century on.

History

1710

— Jakob Hermann derives the conserved vector for inverse-square motion; Bernoulli generalizes it within the year.

1799

— Laplace presents the vector systematically in the Mécanique céleste; it becomes celestial-mechanics folklore.

1919

— Runge's Vektoranalysis textbook works the Kepler vector as an exercise in modern vector notation.

1924

— Lenz applies "Runge's vector" to the quantum theory of the hydrogen atom's Stark and Zeeman problems.

1926

— Pauli, citing Lenz, derives the hydrogen spectrum from the quantized vector by pure algebra — months before Schrödinger's solution appears.

Going further

Texts. Sakurai & Napolitano, Modern Quantum Mechanics, §3.9 · Shankar, Principles of Quantum Mechanics, Ch. 17 · Goldstein, Poole & Safko, §3.9 (the classical vector) · Pauli, Z. Phys. 36, 336 (1926).

Provenance. Operator definitions and commutators follow Sakurai & Napolitano Modern QM §3.9 and Shankar Ch. 17; a verifier should re-derive $[\hat H,\hat{\mathbf A}]=0$ , the closure relation, the identity $\hat A^2=k^2+\frac{2\hat H}{m}(\hat L^2+\hbar^2)$ , and the algebraic spectrum $E_n=-mk^2/2\hbar^2 n^2$ . Numerical checks: with $k=Ze^2/4\pi\varepsilon_0$ and reduced mass, $E_1=-13.606\,\mathrm{eV}$ for hydrogen ( $Z=1$ ); $\alpha\approx1/137.036$ ; degeneracy $\sum_{\ell=0}^{n-1}(2\ell+1)=n^2$ . Classical limit $|\mathbf A|=mk e$ per Goldstein §3.9; Fock $S^3$ mapping per Fock (1935) and Bargmann (1936). Edition note (2026-07-08). Per the chapter ownership map, the base's SO(4) derivations — the closure relation, the two-SU(2) decomposition, and the algebraic Bohr levels — are owned by Accidental Degeneracy and the SO(4) Symmetry (QUANT.HYDR.SO4.0943), which consumes the credentials proven here (Hermiticity, conservation, $\hat{\mathbf A}\cdot\hat{\mathbf L} = 0$ , the $\hat A^2$ identity) by citation; the division note appears at both ends. The eccentricity/classical-limit derivation stays here.

Prerequisites. The hydrogen Coulomb radial equation · the Bohr energy spectrum and Rydberg constant · degeneracy and the role of the magnetic quantum number · accidental degeneracy and the SO(4) symmetry (for where the credentials go). Leads to. Hydrogenic ions and scaling with nuclear charge · hydrogenic wavefunctions and quantum numbers · radial probability distribution and expectation values.

Why is the Hermitian combination the difference p × L − L × p rather than the sum?

A sign hides in the cross product. Daggering $\hat{\mathbf p}\times\hat{\mathbf L}$ reverses operator order and the index relabeling costs a minus: the adjoint is $-\hat{\mathbf L}\times\hat{\mathbf p}$ . So the difference is self-adjoint, while the innocent-looking sum is anti-Hermitian — Derivation A collapses it to $2i\hbar\hat{\mathbf p}$ . The "average $\hat X\hat Y + \hat Y\hat X$ " rule is for simple products; cross products carry their own antisymmetry and flip the recipe.

Is the LRL vector an observable I could actually measure?

In principle yes — Hermitian, commuting with $\hat H$ — and its matrix elements are measured all the time in disguise: hydrogen's linear Stark effect is the field coupling to the in-shell dipole (the equation-(6) structure), and quantum-defect precession in Rydberg atoms tracks $\langle\hat{\mathbf A}\rangle$ rotating under a non-Coulomb core. But $[\hat A_i, \hat A_j] \ne 0$ (closing on $\hat{\mathbf L}$ — the SO(4) article's story), so like an angular momentum: length and one component, never the full arrow.

Where exactly does the extra ħ² in A² come from, and does it matter?

Trace it: the workhorse form $\hat{\mathbf p}\times\hat{\mathbf L} - i\hbar\hat{\mathbf p}$ carries Hermiticity's counterterm, and squaring assembles its cross terms into exactly one $\hbar^2$ beside $\hat L^2$ . It matters enormously: run the algebraic spectrum with $\hat L^2$ alone and the Casimir bookkeeping returns half-integer-shifted levels, flatly contradicting experiment. It also survives in $\langle\hat A^2\rangle = k^2[1 - (\ell(\ell+1)+1)/n^2]$ : the $1s$ state has $e_q = 0$ only because the $+1$ upstairs matches $n^2 = 1$ exactly.

Conservation laws come from symmetries — what symmetry does the LRL vector generate?

A genuine one, but not geometric. Noether-style, $\hat{\mathbf A}$ generates transformations mixing position and momentum in a dynamics-dependent way — "dynamical symmetry" is the standard name, "hidden" the honest one, since nothing in the shape of space advertises it. They become ordinary rotations only in Fock's picture: in momentum space the bound states live on a 3-sphere, and $\hat{\mathbf A}$ moves its poles. That construction, the algebra's closure, and the payoff spectrum are the business of Accidental Degeneracy and the SO(4) Symmetry.

Does Â connect states of different n, the way ladder operators connect different m?

No — that is its defining feature. $[\hat H, \hat{\mathbf A}] = 0$ means it cannot change energy, so it never leaves an $n$ -shell. Within the shell, equation (4)'s rules permit $\Delta\ell = \pm 1$ , $\Delta m = 0, \pm 1$ by component. So $\hat{\mathbf A}$ is a horizontal ladder along $\ell$ , complementary to $\hat L_\pm$ along $m$ ; between them the two ladders reach all $n^2$ states from any seed — the operator-level meaning of the shell's degeneracy.

If only 1/r conserves this vector, is Bertrand's theorem the same statement?

Two faces of one fact, not identical. Bertrand's theorem (mechanics track) says only $1/r$ and $r^2$ close all bounded orbits; universal closure requires an extra conserved quantity locking the orientation. For $1/r$ it is the LRL vector; for the isotropic oscillator a symmetric tensor — why its hidden symmetry is SU(3), not SO(4), as the oscillator articles take up. Derivation B's Step 3 is the residue calculation: deform the exponent and conservation fails at first order in $(s-1)$ .