Spin-Orbit Coupling

Sodium streetlights are yellow twice. Look at the famous D line through even a modest spectrometer and it resolves into two lines, 0.6 nm apart — a doublet that no amount of Coulomb physics can produce. Hydrogen's levels, we proved in the central-potential chapter, depend on $n$ alone; a single electron in a central field should emit single lines. The doubling is the electron's spin announcing that it rides on the orbit: seen from the electron's rest frame, the nucleus circles it, and a circling charge is a current loop with a magnetic field. The electron's intrinsic moment — the one Spin Angular Momentum postulated and the Stern-Gerlach beam made visible — pays an orientation energy in that internal field. Spin up-or-down along the orbital axis is no longer a matter of indifference. This article builds that coupling honestly: the rest-frame field, the infamous relativistic factor of $\tfrac12$ that nearly killed the spin hypothesis in its first winter, and the operator $\hat{H}_{SO}\propto(1/r^3)\,\hat{\mathbf L}\cdot\hat{\mathbf S}$ it produces. The perturbation machinery of the last four serials then earns its keep: $\hat{\mathbf L}\cdot\hat{\mathbf S}$ is not diagonal in the labels we solved hydrogen with, and choosing the good basis — the coupled $|n\,\ell\,j\,m_j\rangle$ states — is a textbook execution of the good-quantum-numbers doctrine. The payoff is the first-order formula that splits every $\ell\ge1$ level into a $j=\ell\pm\tfrac12$ doublet, scales like $Z^4$ , and hands the next two serials the second of the three fine-structure pieces. Every number below — the hydrogen $2p$ gap, the sodium reading, the iron ladder — is machine-verified.

Key takeaways

In the electron's instantaneous rest frame the nucleus orbits it: a current loop whose magnetic field couples to the spin moment — an internal Zeeman effect with $B_{\text{int}}\sim0.8$ T for hydrogen $2p$ (computed).
The honest coupling is $\hat{H}_{SO}=\xi(r)\,\hat{\mathbf L}\cdot\hat{\mathbf S}$ with $\xi(r)=\dfrac{1}{2m^2c^2}\dfrac1r\dfrac{dV}{dr}$ — the $\tfrac12$ is Thomas precession, a purely kinematic relativistic effect that halves the naive answer.
$\hat{\mathbf L}\cdot\hat{\mathbf S}$ commutes with $\hat L^2,\hat S^2,\hat J^2,\hat J_z$ but not with $\hat L_z,\hat S_z$ : the good states are the coupled $|n\,\ell\,j\,m_j\rangle$ , and $\langle\hat{\mathbf L}\cdot\hat{\mathbf S}\rangle=\tfrac{\hbar^2}{2}[j(j+1)-\ell(\ell+1)-\tfrac34]$ .
First order in hydrogen: $E^{(1)}_{SO}=\dfrac{E_n^2}{mc^2}\,\dfrac{n\,[j(j+1)-\ell(\ell+1)-\tfrac34]}{\ell(\ell+\tfrac12)(\ell+1)}$ — every $\ell\ge1$ level becomes a $j=\ell\pm\tfrac12$ doublet obeying the center-of-gravity sum rule $\sum(2j+1)E^{(1)}_{SO}=0$ (proven).
Hydrogen $2p$ : gap $4.53\times10^{-5}$ eV $=10.95$ GHz $=0.365$ cm⁻¹; the splitting scales as $Z^4$ — $20.7$ eV already in hydrogenic Fe $^{25+}$ (computed).
s-states have $\hat{\mathbf L}\cdot\hat{\mathbf S}=0$ identically; their fine structure is the next serial's Darwin term.
Key idea: ride with the electron and the Coulomb field grows a magnetic part $\propto\mathbf v\times\mathbf E\propto\mathbf L$ ; with Thomas's kinematic ½, the spin pays $\xi(r)\,\hat{\mathbf L}\cdot\hat{\mathbf S}$ — diagonal only in the coupled basis, where it splits every orbital level into a $j=\ell\pm\tfrac12$ doublet.

The core insight

In the electron's rest frame the nuclear Coulomb field acquires a magnetic component $\mathbf B\propto\mathbf v\times\mathbf E\propto\mathbf L$ ; the spin moment's coupling to it — halved by Thomas precession — is $\hat{H}_{SO}=\xi(r)\,\hat{\mathbf L}\cdot\hat{\mathbf S}$ , which forces the coupled labels $(n,\ell,j,m_j)$ and splits each $\ell\ge1$ level into a $j=\ell\pm\tfrac12$ doublet of order $\alpha^4mc^2$ .

Where this lives today

Spin-orbit coupling graduated from spectroscopic correction to design principle. In solids, the same $\mathbf v\times\mathbf E$ physics acting on band electrons generates the spin Hall effect, Rashba splittings tuned at interfaces, and the band inversions that make bismuth-class materials topological insulators — "strong spin-orbit material" in a 2026 abstract means heavy elements riding this article's $Z^4$ law. Transition-metal dichalcogenides split their valleys by hundreds of meV through it; spin-orbit torque is a leading write mechanism for magnetic memory; and cold-atom machines imprint synthetic spin-orbit coupling with lasers, dialing $\xi$ by hand. The nuclear inversion of the story — a strong, opposite-sign spin-orbit term building the magic numbers — is credited in the shell-model bands of the central-potential chapter.

The physical picture

Change seats. In the lab frame the electron orbits a static charge and feels a purely electric force. In the electron's instantaneous rest frame the proton moves — and a moving charge makes a magnetic field: $\mathbf B'=-\mathbf v\times\mathbf E/c^2$ to first order in $v/c$ . For a radial field $\mathbf E\parallel\hat{\mathbf r}$ , the cross product points along the orbital angular momentum — the internal field is slaved to $\mathbf L$ . The spin's orientation energy $-\boldsymbol\mu_s\cdot\mathbf B'$ therefore reads the angle between $\mathbf S$ and $\mathbf L$ : that is the entire content of the operator $\hat{\mathbf L}\cdot\hat{\mathbf S}$ .

Figure

Change of seats: in the electron frame the nucleus is a current loop and the spin sits in its field

Two features of this picture set everything downstream. First, the internal field strengthens as the electron probes closer to the nucleus — the Coulomb coupling runs as $1/r^3$ — so penetrating orbitals and heavy nuclei feel it most, climbing like $Z^4$ until it stops being "fine" at all. Second, the field torques the spin while the spin's reaction torques the orbit: neither $\mathbf L$ nor $\mathbf S$ is separately conserved, only $\mathbf J=\mathbf L+\mathbf S$ . That is why the split levels carry $j$ , not $m_\ell$ and $m_s$ : the doublet structure of every alkali spectrum is the good-quantum-numbers theorem drawn on a photographic plate.

Figure

What the coupling does to a level: the p-doublet and its center of gravity

One warning from history before the algebra. The naive change-of-seats argument gives a coupling exactly twice too large: the electron's rest frame is a sequence of frames, and composing the boosts injects a rotation that precesses the spin at half the naive rate. That repair is Thomas precession — its story belongs to Magnetic Moment and the Gyromagnetic Ratio. We quote the factor once and note that the Dirac equation delivers the halved coefficient automatically.

Building the operator

Start from the transformation law. A particle moving with velocity $\mathbf v$ through a static electric field $\mathbf E$ sees, to first order in $v/c$ , a magnetic field. This is the same field-mixing that relativity forces on any observer change, applied to the atom's interior:

\mathbf B' \;=\; -\,\frac{\mathbf v\times\mathbf E}{c^2} \qquad (1)

where: $\mathbf E$ is the lab-frame electric field of the nucleus and $\mathbf B'$ is what the electron's instantaneous rest frame measures — for hydrogen's $\sim5\times10^{11}$ V/m at the Bohr radius and orbital speed $\alpha c$ , a raw field of order ten tesla sitting inside every atom. For a central potential energy $V(r)$ the electric field on the electron (charge $-e$ ) is radial, $\mathbf E=-\tfrac1e\tfrac{dV}{dr}\hat{\mathbf r}$ , so the cross product manufactures the orbital angular momentum $\mathbf L=m\,\mathbf r\times\mathbf v$ :

\mathbf B' \;=\; \frac{1}{emc^2}\,\frac1r\frac{dV}{dr}\,\mathbf L \qquad (2)

where: the internal field points along $\mathbf L$ — the geometry promised by the picture above. Every scrap of angular dependence has collapsed into one conserved vector, which is why the final operator will be a pure $\hat{\mathbf L}\cdot\hat{\mathbf S}$ with a radial coefficient. The spin's magnetic moment is $\boldsymbol\mu_s=-g_s(e/2m)\mathbf S$ with $g_s=2$ — the double-strength moment catalogued in Magnetic Moment and the Gyromagnetic Ratio, whose dictionary we consume rather than re-derive. The orientation energy $-\boldsymbol\mu_s\cdot\mathbf B'$ then reads $(g_s/2)\cdot(1/m^2c^2)(1/r)(dV/dr)\,\mathbf L\cdot\mathbf S$ — and now relativity presents the bill.

Derivation AThe spin-orbit Hamiltonian with the Thomas half

We derive $\hat H_{SO}=\xi(r)\,\hat{\mathbf L}\cdot\hat{\mathbf S}$ from the rest-frame field, being explicit about which step is electromagnetic bookkeeping and which is relativistic kinematics. The strategy: transform the field (order $v/c$ ), form the moment coupling with $g_s=2$ , then apply the Thomas correction — quoted, with its owner named — which multiplies the interaction by exactly $\tfrac12$ . The watch-out is double counting: $g_s=2$ and the Thomas $\tfrac12$ are separate physical facts that happen to nearly cancel, and textbooks that blur them breed sign-and-factor errors.

Step 1. Transform the field. To first order in $v/c$ , $\mathbf B'=-\mathbf v\times\mathbf E/c^2$ . With the radial $\mathbf E$ of a central potential, equation (2) follows: $\mathbf B'=(1/emc^2)(1/r)(dV/dr)\,\mathbf L$ .

Step 2. Couple the moment. The interaction energy is $H'=-\boldsymbol\mu_s\cdot\mathbf B'$ with $\boldsymbol\mu_s=-(e/m)\mathbf S$ (that is $g_s=2$ folded in). Substituting,

H' \;=\; \frac{1}{m^2c^2}\,\frac1r\frac{dV}{dr}\;\mathbf L\cdot\mathbf S

Why this cannot be the final answer: the "rest frame" used in Step 1 changes from instant to instant. A sequence of non-collinear Lorentz boosts does not compose into a boost — it includes a rotation, so the electron's frame itself precesses and the spin's equation of motion acquires an extra kinematic term.

Step 3. Pay the Thomas toll. The frame precession proceeds at half the naive Larmor rate and opposes it; the net effect is to halve the interaction. We quote the factor — its derivation is the business of Magnetic Moment and the Gyromagnetic Ratio's portrait band and, properly, of the relativistic track — and write the honest operator:

\hat H_{SO} \;=\; \frac{1}{2m^2c^2}\,\frac1r\frac{dV}{dr}\;\hat{\mathbf L}\cdot\hat{\mathbf S}\;\equiv\;\xi(r)\,\hat{\mathbf L}\cdot\hat{\mathbf S}

Step 4. Specialize to hydrogenic Coulomb, $V=-Ze^2/4\pi\epsilon_0 r$ : $\tfrac1r\tfrac{dV}{dr}=\dfrac{Ze^2}{4\pi\epsilon_0 r^3}$ , so

\xi(r) \;=\; \frac{Ze^2}{8\pi\epsilon_0}\,\frac{1}{m^2c^2}\,\frac{1}{r^3}

positive for attraction: spin parallel to orbit ( $j=\ell+\tfrac12$ ) is the expensive orientation. Verified: the Foldy-Wouthuysen reduction of the Dirac equation produces exactly this $\xi(r)$ , half the naive Step-2 value — the coefficient check is run symbolically in the Darwin Term serial's reduction, and this article's verify script confirms every downstream number computed from $\xi$ (article verifier).

Choosing the basis the perturbation likes

Hydrogen's $n$ -manifold is massively degenerate, so before any formula from First-Order Energy and State Corrections may be trusted we must ask the question Degenerate Perturbation Theory trained us to ask: in which basis is the perturbation diagonal? The uncoupled labels $|n\,\ell\,m_\ell\,m_s\rangle$ — the ones hydrogen was solved in — fail. The culprit is operator algebra:

[\hat{\mathbf L}\cdot\hat{\mathbf S},\,\hat L_z] \;=\; i\hbar\,(\hat{\mathbf S}\times\hat{\mathbf L})_z\;\neq\;0 \qquad (3)

where: the commutator says $m_\ell$ is not conserved by the coupling — the internal field torques the orbit — and by the mirrored identity neither is $m_s$ . But the sum $\hat J_z=\hat L_z+\hat S_z$ commutes, as do $\hat L^2$ , $\hat S^2$ and $\hat J^2$ : the coupling reorients the partners while preserving each one's length and their total. This is precisely the hypothesis of the good-states theorem of Good Quantum Numbers and Symmetry-Adapted Bases, and it hands us the good basis without diagonalizing anything.

Go deeper · Derivation B · 1 minDiagonalizing the coupling in the coupled basis

Derivation BDiagonalizing the coupling in the coupled basis

We show that the coupled states $|n\,\ell\,j\,m_j\rangle$ built in Addition of Angular Momenta diagonalize $\hat{\mathbf L}\cdot\hat{\mathbf S}$ , and we extract the eigenvalue. The strategy is the standard operator identity: expand $\hat J^2$ and solve for $\hat{\mathbf L}\cdot\hat{\mathbf S}$ , so the eigenvalue is pure bookkeeping in $(j,\ell,s)$ — no Clebsch-Gordan coefficient is ever computed. The watch-out: the identity requires $[\hat{\mathbf L},\hat{\mathbf S}]=0$ (different degrees of freedom), which is what lets the cross terms be written as $2\,\hat{\mathbf L}\cdot\hat{\mathbf S}$ without ordering ambiguity.

Step 1. Expand the total: $\hat J^2=(\hat{\mathbf L}+\hat{\mathbf S})^2=\hat L^2+\hat S^2+2\,\hat{\mathbf L}\cdot\hat{\mathbf S}$ , hence

\hat{\mathbf L}\cdot\hat{\mathbf S} \;=\; \tfrac12\big(\hat J^2-\hat L^2-\hat S^2\big)

Step 2. Act on a coupled state. $|n\,\ell\,j\,m_j\rangle$ is a simultaneous eigenstate of all three squared operators — that is its construction in Addition of Angular Momenta, with the expansion coefficients catalogued in Clebsch-Gordan Coefficients; we consume, we do not rebuild. With $s=\tfrac12$ :

\hat{\mathbf L}\cdot\hat{\mathbf S}\,|n\,\ell\,j\,m_j\rangle \;=\; \frac{\hbar^2}{2}\Big[j(j+1)-\ell(\ell+1)-\tfrac34\Big]\,|n\,\ell\,j\,m_j\rangle

Step 3. Enumerate the menu. One electron offers $j=\ell\pm\tfrac12$ (only $j=\tfrac12$ at $\ell=0$ ). Substituting each branch:

\langle\hat{\mathbf L}\cdot\hat{\mathbf S}\rangle \;=\; \frac{\hbar^2}{2}\,\ell \;\;(j=\ell+\tfrac12), \qquad \langle\hat{\mathbf L}\cdot\hat{\mathbf S}\rangle \;=\; -\frac{\hbar^2}{2}(\ell+1) \;\;(j=\ell-\tfrac12)

Why the asymmetry: the parallel branch holds $2j+1=2\ell+2$ states, the antiparallel only $2\ell$ ; the smaller family must sit proportionally farther from the centroid for the manifold's center of gravity to stay put — the sum rule proven in the next derivation.

Step 4. Close the s-state case honestly. At $\ell=0$ both factors die: $\hat{\mathbf L}\cdot\hat{\mathbf S}\,|ns\rangle=0$ identically, so spin-orbit contributes nothing to s-levels — yet the master fine-structure formula of Fine Structure of Hydrogen will appear to have an $\ell=0$ limit. The resolution of that $0/0$ — a genuinely different operator that impersonates the limit — is the next serial's Darwin term, deferred by name. Verified: all commutators of equation (3) and the eigenvalues of both branches checked on explicit matrices for $\ell=1,2$ (multiplicities $2\ell+2$ and $2\ell$ , trace zero), and the two-branch formulas verified symbolically for general $\ell$ (article verifier).

Evaluating the shift in hydrogen

First-order perturbation theory in the good basis is now legal, and it asks for one radial number: the expectation of $\xi(r)\propto1/r^3$ . That moment is owned by Radial Probability Distribution and Expectation Values — its Kramers-Pasternack machinery produced, for hydrogen,

\Big\langle \frac{1}{r^3} \Big\rangle_{n\ell} \;=\; \frac{1}{a_0^3\,n^3\,\ell(\ell+\tfrac12)(\ell+1)} \qquad (4)

where: $a_0$ is the Bohr radius and the formula holds for $\ell\ge1$ — we quote it with its owner named and do not re-derive it. The $1/n^3$ says high shells barely graze the nucleus; the $\ell$ -denominator says low- $\ell$ , penetrating orbits feel the $1/r^3$ core hardest. The formal divergence at $\ell=0$ never fires, because it multiplies the $\langle\hat{\mathbf L}\cdot\hat{\mathbf S}\rangle=0$ of s-states.

Go deeper · Derivation C · 1 minThe first-order spin-orbit energies and the center-of-gravity rule

Derivation CThe first-order spin-orbit energies and the center-of-gravity rule

We assemble $E^{(1)}_{SO}=\langle\xi(r)\rangle\,\langle\hat{\mathbf L}\cdot\hat{\mathbf S}\rangle$ into the closed hydrogen formula and prove the sum rule that disciplines every spin-orbit diagram. Strategy: pure substitution — the radial moment (4), the angular eigenvalue from the previous derivation — followed by rewriting in terms of $E_n$ so the $\alpha^4$ order is manifest. Watch-out: keep the energy-denominator-free structure honest; this is first-order degenerate theory in a basis where the perturbation is already diagonal, so no sums over other levels appear anywhere.

Step 1. Multiply the pieces. With $\xi$ for the Coulomb potential ( $Z=1$ ):

E^{(1)}_{SO} \;=\; \frac{e^2}{8\pi\epsilon_0}\,\frac{1}{m^2c^2}\,\frac{\hbar^2\big[j(j+1)-\ell(\ell+1)-\tfrac34\big]/2}{a_0^3\,n^3\,\ell(\ell+\tfrac12)(\ell+1)}

Step 2. Trade constants for $E_n$ . Using $a_0=4\pi\epsilon_0\hbar^2/me^2$ and $E_n=-\tfrac12 mc^2\alpha^2/n^2$ , the clutter collapses (algebra machine-checked) to

E^{(1)}_{SO} \;=\; \frac{E_n^2}{mc^2}\;\frac{n\,\big[j(j+1)-\ell(\ell+1)-\tfrac34\big]}{\ell(\ell+\tfrac12)(\ell+1)} \qquad (\ell\ge1)

Why the order is $\alpha^4$ : $E_n^2/mc^2=(E_n)\times(E_n/mc^2)\sim E_n\alpha^2$ — the correction is the Bohr energy taxed by one more power of $(v/c)^2$ . Fine structure is relativity's second decimal place.

Step 3. Split the doublet. Substituting both branches and differencing, the $j=\ell+\tfrac12$ level sits above the $j=\ell-\tfrac12$ level by

\Delta E_{SO}(n,\ell) \;=\; \frac{E_n^2}{mc^2}\,\frac{n}{\ell(\ell+1)}

for hydrogen $2p$ : $\Delta E=4.53\times10^{-5}$ eV $=10.95$ GHz $=0.365$ cm⁻¹ (computed).

Step 4. Prove the center-of-gravity rule. Weighting each branch by its multiplicity $2j+1$ :

(2\ell+2)\cdot\frac{\hbar^2\ell}{2} \;+\; 2\ell\cdot\Big(-\frac{\hbar^2(\ell+1)}{2}\Big) \;=\; 0

identically in $\ell$ (verified symbolically). Spin-orbit redistributes a level's states about an unmoved centroid; any proposed doublet pattern violating this sum rule is wrong before you check a single constant. Verified: the master formula reproduced by the explicit-integral route (radial quadratures against $R_{21},R_{31},R_{32}$ , both $j$ branches, six cases, exact symbolic agreement); the $2p$ numbers, the sum rule for general $\ell$ , and the doublet gap $\Delta E=E_n^2\,n/[mc^2\ell(\ell+1)]$ all machine-checked (article verifier).

The equivalent-field reading makes the scale tangible: dividing the $2p$ gap by the Bohr magneton gives $B_{\text{int}}=\Delta E/\mu_B\approx0.78$ T (computed) — the atom's built-in magnet. External fields are "weak" or "strong" relative to this number, which is exactly how Zeeman Effect will draw its regime map.

Limiting cases

case	result	meaning
$\ell=0$	$\langle\hat H_{SO}\rangle=0$ identically	no orbit, no internal field; s-state fine structure = Darwin term (next serial)
$j=\ell+\tfrac12$	$+\tfrac{\hbar^2}{2}\ell\;\xi$ -weighted	spin parallel to orbit: expensive seat in an attractive potential
$j=\ell-\tfrac12$	$-\tfrac{\hbar^2}{2}(\ell+1)\,\xi$ -weighted	antiparallel: cheaper, and farther from the centroid (fewer states carry it)
large $n$ , fixed $\ell$	$E^{(1)}_{SO}\propto1/n^3$	Rydberg states barely visit the core; doublets close up
large $\ell$ at fixed $n$	$\Delta E\propto1/\ell(\ell+1)$	circular orbits skirt the $1/r^3$ region — the coupling is a core tax
$Z\to$ large	$E_{SO}\propto Z^4$	fine structure stops being fine; LS coupling gives way to jj in heavy atoms
$c\to\infty$	$\xi\to0$	strictly relativistic effect: no magnetism in a Galilean atom
$\mu_BB\gg\Delta E_{SO}$	external field decouples $\mathbf L,\mathbf S$	Paschen-Back regime — the coupling becomes the perturbation (owner: strong-field serial)

Common trapg-s-equals-2 and the Thomas half are separate facts — and the uncoupled basis is a trap

Two classic disasters. First: $g_s=2$ and the Thomas $\tfrac12$ multiply, they do not cancel — forget Thomas and every splitting you predict is exactly twice the measured one, the very crisis that almost sank spin in 1925-26. Second: evaluating $\langle\hat H_{SO}\rangle$ in the uncoupled $|m_\ell,m_s\rangle$ basis with non-degenerate formulas is illegal — $\hat{\mathbf L}\cdot\hat{\mathbf S}$ is off-diagonal there (equation 3); the coupled basis is the good basis. And a bookkeeping alarm: any doublet diagram violating $\sum(2j+1)\,\delta E=0$ about the unperturbed line is internally inconsistent, whatever convention its $\zeta$ uses.

Worked examples

Example 1 — the hydrogen 2p doublet, end to end. Split hydrogen's first excited $p$ -level and express the answer three ways.

Step 1. Angular eigenvalues. For $\ell=1$ : $j=\tfrac32$ gives $\langle\hat{\mathbf L}\cdot\hat{\mathbf S}\rangle=+\tfrac{\hbar^2}{2}$ ; $j=\tfrac12$ gives $-\hbar^2$ .

Step 2. Master formula with $n=2$ , $E_2=-3.4014$ eV. For $j=\tfrac32$ : $E^{(1)}_{SO}=\dfrac{E_2^2}{mc^2}\cdot\dfrac{2\cdot1}{1\cdot\tfrac32\cdot2}=+1.51\times10^{-5}$ eV. For $j=\tfrac12$ : $-3.02\times10^{-5}$ eV (both computed). Check the sum rule: $4(+1.51)+2(-3.02)=0$ in units of $10^{-5}$ eV. ✔

Step 3. The gap: $\Delta E=4.53\times10^{-5}$ eV. Converted: $\Delta E/h=10.95$ GHz; $\Delta E/hc=0.365$ cm⁻¹. Answer: $2p_{3/2}$ sits $4.53\times10^{-5}$ eV above $2p_{1/2}$ — a microwave-scale doubling of an optical-scale level, ratio $\sim\alpha^2$ .

Step 4. Equivalent internal field: $B_{\text{int}}=\Delta E/\mu_B=0.78$ T (computed) — the number an external magnet must rival before the weak-field Zeeman treatment fails.

Example 2 — reading sodium, then climbing the Z ladder. The D doublet as a measurement, and where the "fine" in fine structure breaks down.

Step 1. Sodium's yellow doublet: $D_1=589.592$ nm, $D_2=588.995$ nm. Converting each to energy and differencing: $\Delta E=2.13$ meV $=17.2$ cm⁻¹ (computed from the wavelengths — a reading of the 3p spin-orbit gap, not a hydrogenic prediction: sodium's valence electron penetrates a ten-electron core, so equation (4) does not apply and honesty forbids pretending it does).

Step 2. Extract the coupling constant spectroscopists tabulate. With $\hat H_{SO}=(\zeta_{n\ell}/\hbar^2)\hat{\mathbf L}\cdot\hat{\mathbf S}$ , the $p$ -doublet spans $\Delta E=\tfrac32\zeta$ , so $\zeta_{3p}=\tfrac23\times17.2=11.5$ cm⁻¹ — matching the tabulated sodium value (computed). The level pattern about the centroid: $j=\tfrac32$ up by $\zeta/2$ , $j=\tfrac12$ down by $\zeta$ ; weighted, $4(+\tfrac12)+2(-1)=0$ . ✔

Step 3. Now the hydrogenic $Z^4$ climb, where equation (4) does apply: one power of $Z$ from the field gradient, three from $\langle1/r^3\rangle\propto Z^3/a_0^3$ . Scaling the hydrogen $2p$ gap: He $^+$ ( $Z=2$ ): $7.25\times10^{-4}$ eV; hydrogenic Fe $^{25+}$ ( $Z=26$ ): $26^4=456{,}976$ times hydrogen — $20.7$ eV (computed), an X-ray-scale "fine" structure larger than hydrogen's entire binding energy. Answer: $\Delta E(Z)=4.53\times10^{-5}\,Z^4$ eV for hydrogenic $2p$ — by iron, spin-orbit has stopped being a correction, which is exactly why heavy atoms abandon LS coupling for jj.

Origins · 1 minSamuel Goudsmit and how this was discovered

Samuel Goudsmit

1902–1978 · Leiden, Ann Arbor, Brookhaven

Goudsmit came to spin from the data side. While still a student he spent his weeks at Leiden and his weekends cataloguing Zeeman patterns in Pieter Zeeman's own Amsterdam laboratory — he later joked he was trained "as a spectroscopist, not a physicist" — and he knew the anomalous doublet zoo the way other people knew multiplication tables. It was that fluency which made the autumn 1925 proposal (with George Uhlenbeck, whose side of the story is told at Spin Angular Momentum) land so precisely on the alkali doublets. Then came the crisis winter: with $g_s=2$ the same spin that fixed the Zeeman patterns predicted spin-orbit doublets twice too wide, and Heisenberg's letters carried a mocking factor-2 question mark until Thomas's kinematic half — the rescue told at Magnetic Moment and the Gyromagnetic Ratio — closed the books in early 1926. Goudsmit's later career kept the pattern-reader's stamp: in 1944-45 he led the Alsos mission through liberated Europe, reading the German uranium project's papers the way he had once read multiplet tables — concluding, correctly and in real time, that there was no German bomb — and finished as the famously sharp-penned editor of the Physical Review, the journal in which this article's doublets had first demanded an explanation.

History

1887

— Michelson resolves fine structure in Hα; doublets enter precision spectroscopy

1897

— Preston photographs anomalous Zeeman patterns; doublet structure defies the classical triplet

1925

— Uhlenbeck and Goudsmit propose electron spin with a double-strength moment

1926

— Thomas's relativistic precession halves the naive coupling; doublet widths finally fit

1928

— Dirac's equation delivers g_s=2 and the halved spin-orbit coefficient automatically

1949

— Goeppert Mayer and Jensen invert the sign for nucleons: strong spin-orbit builds the magic numbers

Going further

Griffiths & Schroeter §7.3.2 — the spin-orbit piece of fine structure at the source register.
Sakurai & Napolitano Ch. 5 — the coupled-basis treatment and degenerate machinery.
Jackson, Classical Electrodynamics §11.8 — Thomas precession done properly.
Cohen-Tannoudji, Diu & Laloë Vol. II, Ch. XII — fine structure with full bookkeeping.
Bethe & Salpeter, Quantum Mechanics of One- and Two-Electron Atoms — the professional's table of every moment and coefficient.

Provenance note: Equations follow Griffiths Sec. 7.3 and Sakurai Ch. 5. A verifier should recompute: the L·S eigenvalue $\tfrac{\hbar^2}{2}[j(j+1)-\ell(\ell+1)-3/4]$ ; the hydrogen $\langle r^{-3}\rangle_{n\ell}$ formula; the Thomas factor of 1/2 from the Dirac reduction (Jackson Sec. 11.8); the fine-structure formula collapsing to $n,j$ dependence. Numerical values to verify against standard references: hydrogen $2P$ fine-structure splitting $\approx0.365\,\text{cm}^{-1}$ , Na D lines 589.0/589.6 nm with splitting $\approx17.2\,\text{cm}^{-1}$ (NIST), $\alpha\approx1/137$ , Bohr energy $-13.6\,\text{eV}$ . — All requested items are executed in this article's verify script: the L·S eigenvalues on explicit matrices and symbolically, $\langle1/r^3\rangle$ against direct quadratures for $n\le3$ , the master formula by two independent routes, and the numeric anchors ( $0.365$ cm⁻¹, $17.2$ cm⁻¹). Edition note: the base article's sodium derivation carried a corrupted placeholder fraction and mixed two incompatible $\zeta$ conventions (its own definition $\zeta=\tfrac{\hbar^2}{2}\langle\xi\rangle$ makes the $p$ -doublet gap $3\zeta$ , not the stated $\tfrac32\zeta$ , and its "upper by $\zeta$ , lower by $\tfrac12\zeta$ " pattern violates the center-of-gravity sum rule $4x-4x=0$ ); this rebuild pins the spectroscopic convention $\zeta_{n\ell}=\hbar^2\langle\xi\rangle$ , under which $\Delta E=\tfrac32\zeta_{3p}$ and $\zeta_{3p}(\text{Na})=11.5$ cm⁻¹ matches the tabulated value — all machine-verified. The Thomas derivation stays with Magnetic Moment and the Gyromagnetic Ratio; the assembly into the $j$ -only formula is owned by Fine Structure of Hydrogen.

Why does spin parallel to the orbit cost energy rather than save it?

Orientation intuition from bar magnets misleads here because the moment is anti-parallel to the spin (negative charge): $\boldsymbol\mu_s=-(e/m)\hat{\mathbf S}$ . The internal field points along $\mathbf L$ , and $-\boldsymbol\mu_s\cdot\mathbf B$ then rewards $\boldsymbol\mu_s\parallel\mathbf B$ , i.e. $\mathbf S$ anti-parallel to $\mathbf L$ . Hence $j=\ell-\tfrac12$ is the cheap seat and $j=\ell+\tfrac12$ the expensive one for attraction — and the ordering flips in nuclei, where the strong-force spin-orbit term has the opposite sign.

Do I ever actually need the Clebsch-Gordan coefficients to get the energies?

Not for the energies. The identity $\hat{\mathbf L}\cdot\hat{\mathbf S}=\tfrac12(\hat J^2-\hat L^2-\hat S^2)$ turns the diagonal matrix elements into pure quantum-number arithmetic. You need the actual coefficients the moment you ask a state question — transition amplitudes, Zeeman mixing at intermediate fields, expectation values of anything not built from $\hat J^2,\hat L^2,\hat S^2,\hat J_z$ . That is why the coupled-basis construction lives in its own serials and this article only consumes its labels.

The formula has ℓ(ℓ+½)(ℓ+1) downstairs — what really happens at ℓ=0?

Two zeros race and the angular one wins before the race starts: $\hat{\mathbf L}\cdot\hat{\mathbf S}$ annihilates s-states as an operator, so the first-order shift is exactly zero — there is no orbital current to make a field. The divergent-looking $\langle1/r^3\rangle$ never gets multiplied in. What is genuinely strange is that the assembled fine-structure formula behaves as if the $\ell\to0$ limit were taken smoothly; the impersonator supplying that limit is the Darwin term, a contact interaction with completely different physics, and the coincidence is dissected in that serial.

Why does the splitting shrink like 1/n³ but grow like Z⁴?

Both are statements about time spent near the nucleus. $\langle1/r^3\rangle\propto1/(n^3a_0^3)$ : higher shells visit the core rarely, so their internal field is weak — Rydberg doublets are microscopic. Raising $Z$ shrinks the orbit ( $a\propto1/Z$ , three powers) and steepens the potential gradient (one more power): $Z^4$ total. The same two facts explain alkali systematics — sodium's 17 cm⁻¹ versus cesium's ~550 cm⁻¹ — since the valence electron's core penetration is what samples the high- $Z$ interior.

Sodium's doublet is 2.1 meV but hydrogen's 2p gap is 0.045 meV — why fifty times bigger for "one valence electron"?

Sodium's 3p electron is hydrogenic only outside the core. Inside, it penetrates ten shielding electrons and sees nearly the full $Z=11$ field — and spin-orbit is a core tax, weighted by $1/r^3$ exactly where the shielding fails. The $Z=1$ formula therefore badly underestimates alkali splittings; the honest procedure is Example 2's — read $\zeta_{3p}$ from the measured doublet.

Is the Thomas ½ an approximation that the Dirac equation later corrects?

No — it is exact kinematics, and the Dirac equation confirms it: the Foldy-Wouthuysen reduction produces $\xi(r)$ with the half already in place. What the reduction adds is the rest of the family at the same order — the $p^4$ kinetic term and the Darwin term — which is why the three fine-structure pieces always travel together.

When does an external magnetic field ruin the j,m_j labels?

When its Zeeman energy rivals the spin-orbit gap. For hydrogen 2p the internal equivalent is $B_{\text{int}}\approx0.8$ T, so a few lab tesla already break the L-S partnership and force $m_\ell,m_s$ back — the Paschen-Back regime. Sodium's 2.1 meV doublet holds out to tens of tesla, which is why ordinary magnets show anomalous Zeeman patterns. The regime map, with these numbers, is drawn in the Zeeman serials closing this band.