Fine Structure and Spin-Orbit Coupling

A sodium streetlight is yellow twice. Put its D line through even a modest spectrometer and it resolves into two lines $0.6$ nm apart — a doublet no amount of Coulomb physics can make, because the Bohr and Schrödinger hydrogen levels depend on $n$ alone. The crack is fine structure: splittings smaller than the gross intervals by roughly $\alpha^2\approx5\times10^{-5}$ , the fingerprint of an electron that is relativistic and carries spin. The three corrections that produce it — the relativistic kinetic term, the spin-orbit coupling $\xi(r)\,\hat{\mathbf L}\cdot\hat{\mathbf S}$ , and the Darwin contact term — and the way they collapse into a single formula in $n$ and $j$ are built in aq-qm-fine-structure-hydrogen; the $\hat{\mathbf L}\cdot\hat{\mathbf S}$ machinery, the Thomas half, and the $\langle1/r^3\rangle$ moment are owned by aq-qm-spin-orbit-coupling. Quantum mechanics derives those operators; this article evaluates them into a spectrum — the hydrogen $2P$ gap in gigahertz, the sodium doublet in wavenumbers, the alkali ladder that climbs like $Z^4$ . Once spin-orbit coupling is present, $\hat{\mathbf L}$ and $\hat{\mathbf S}$ precess about their sum $\hat{\mathbf J}=\hat{\mathbf L}+\hat{\mathbf S}$ and are no longer separately conserved; states are organised by the total $j$ , written $n\,^{2S+1}L_J$ . In hydrogen the three corrections conspire so the total shift depends only on $n$ and $j$ — an exact coincidence of the Dirac equation that leaves $2S_{1/2}$ and $2P_{1/2}$ degenerate until the Lamb shift, one power of $\alpha$ deeper, splits them. Fine structure is the template for every finer splitting that follows: hyperfine structure, smaller by the electron-to-nucleon mass ratio, and the field regimes where the Zeeman and Paschen-Back patterns are set by whether the applied field beats the spin-orbit interval.

Key takeaways

Fine structure is the $\alpha^2$ -scale splitting of atomic levels from three relativistic corrections; the dominant one for $\ell\ge1$ is spin-orbit coupling, $\hat H_{SO}=\xi(r)\,\hat{\mathbf L}\cdot\hat{\mathbf S}$ , an internal Zeeman effect from the field the electron sees moving through the nucleus's Coulomb field.
Because $\hat{\mathbf L},\hat{\mathbf S}$ precess about $\hat{\mathbf J}$ , the good labels become $n,\ell,j,m_j$ ; the coupling eigenvalue is $\langle\hat{\mathbf L}\cdot\hat{\mathbf S}\rangle=\tfrac{\hbar^2}{2}[j(j+1)-\ell(\ell+1)-\tfrac34]$ .
Hydrogen $2P$ : the $2P_{3/2}$ – $2P_{1/2}$ gap is $\mathrm{Ry}\,\alpha^2/(2n^3)=4.53\times10^{-5}$ eV $=10.95$ GHz $=0.365$ cm⁻¹; at $n=3$ it is $0.108$ cm⁻¹.
The sodium D doublet — $D_2$ 588.995 nm, $D_1$ 589.592 nm, $\Delta\lambda=0.597$ nm — measures the $3p$ spin-orbit gap: $17.2$ cm⁻¹ $=515.5$ GHz $=2.13$ meV, giving $\zeta_{3p}=11.46$ cm⁻¹.
Spin-orbit grows like $Z^4$ : the alkali doublet climbs Li $0.34\to$ Na $17.2\to$ K $57.8\to$ Rb $237.6\to$ Cs $554.0$ cm⁻¹, a factor $32$ from Na to Cs.
Key idea: spin-orbit coupling and relativistic corrections split levels by $\sim\alpha^2 E_n$ ; the good quantum number becomes $j$ because $\hat{\mathbf J}=\hat{\mathbf L}+\hat{\mathbf S}$ is conserved while $\hat{\mathbf L}$ and $\hat{\mathbf S}$ separately are not, and the doublet size runs as $Z^4/n^3$ .

The core insight

Spin-orbit coupling and relativistic corrections split levels by $\sim\alpha^2 E_n$ ; the good quantum number becomes $j$ because $\hat{\mathbf J}=\hat{\mathbf L}+\hat{\mathbf S}$ is conserved while $\hat{\mathbf L}$ and $\hat{\mathbf S}$ are not — and the doublet spacing runs as $Z^4/n^3$ , from hydrogen's microwave $2P$ gap to caesium's $554$ cm⁻¹.

Where this lives today

Fine structure is now a working tool more than a puzzle. Astronomers read stellar magnetic fields and cosmic abundances through fine-structure multiplets; the far-infrared [C II] 158 μm line, the same physics in many-electron dress, is among the brightest single lines galaxies emit and is tracked across cosmological distances. In the laboratory the $Z^4$ scaling turns hydrogenic ions into stress tests — in hydrogen-like uranium the "fine" structure reaches hundreds of eV and probes quantum electrodynamics where $Z\alpha$ is no longer small. The sodium doublet underwrites laser cooling and the sodium-guide-star lasers that sharpen large telescopes, both of which must address $D_2$ and $D_1$ separately. And $\alpha$ itself — named for this splitting — is now measured to better than $10^{-10}$ : every precision run at these intervals is Michelson's 1887 afternoon with ten more digits.

The physical picture

Ride along with the electron. In its instantaneous rest frame the nucleus orbits it, and a circulating charge is a current loop with a magnetic field $\mathbf B$ at the electron's position, pointing along $\hat{\mathbf L}$ . The spin moment then has an orientation energy $-\boldsymbol\mu_s\cdot\mathbf B$ that reads the angle between spin and orbit — spin parallel or antiparallel to $\hat{\mathbf L}$ costs different energy. That difference is the spin-orbit splitting, and it splits every $\ell\ge1$ term into a $j=\ell\pm\tfrac12$ doublet whose centre of gravity does not move.

Figure

The hydrogen n = 2 fine structure: relativistic corrections leave 2S₁/₂ and 2P₁/₂ exactly degenerate, with 2P₃/₂ above them by 10.95 GHz = 0.3652 cm⁻¹. The residual 2S₁/₂–2P₁/₂ gap (1058 MHz, ghosted) is not fine structure at all — it is the Lamb shift

Two facts organise everything downstream. First, the internal field strengthens sharply as the electron probes closer to the nucleus — the coupling runs as $\langle1/r^3\rangle$ , so penetrating orbits and heavy nuclei feel it most, and the doublet climbs like $Z^4$ until it stops being "fine" at all. Second, because the electron's charge is negative the moment is antiparallel to the spin, so in the attractive Coulomb field the parallel orientation $j=\ell+\tfrac12$ is the expensive seat and $j=\ell-\tfrac12$ the cheaper one. The alkali doublets are this $Z^4$ law drawn on a photographic plate.

Figure

The punchline ladder: the alkali D-line doublet splitting climbs steeply with the valence element (log scale, cm⁻¹). Na is 17.2 cm⁻¹, Cs is 554.0 cm⁻¹ — a factor of 32 — tracking the Z⁴ growth of spin-orbit coupling

The core equations

Fine structure is the sum of three corrections, all of order $\alpha^4 mc^2$ ; each is derived and evaluated in aq-qm-fine-structure-hydrogen, and recalled here by name to be turned into a spectrum.

\hat H_{\text{fs}}=\hat H_{\text{rel}}+\hat H_{SO}+\hat H_{\text{Darwin}} \qquad (1)

where: $\hat H_{\text{rel}}=-\hat p^4/8m^3c^2$ — relativistic kinetic correction; $\hat H_{SO}=\xi(r)\hat{\mathbf L}\cdot\hat{\mathbf S}$ — spin-orbit coupling ( $\ell\ge1$ ); $\hat H_{\text{Darwin}}\propto\delta^3(\mathbf r)$ — contact term, nonzero only for $\ell=0$ . The three carry explicit $\ell$ -dependence that cancels in the sum. The spin-orbit term is the doublet-maker; its prefactor carries the Thomas factor of $\tfrac12$ , a relativistic kinematic effect that halves the naive rest-frame estimate. This operator is owned by aq-qm-spin-orbit-coupling.

\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} \qquad (2)

where: $V(r)$ — central potential; $\xi(r)$ — radial coupling function; the $\tfrac12$ is Thomas precession, without which every splitting comes out twice too large. Forget it and you re-live the crisis of 1925–26. Because $\hat{\mathbf L}\cdot\hat{\mathbf S}$ is not diagonal in $m_\ell,m_s$ but is diagonal in the coupled basis $|n\ell j m_j\rangle$ , its eigenvalue is pure quantum-number bookkeeping from $\hat{\mathbf L}\cdot\hat{\mathbf S}=\tfrac12(\hat J^2-\hat L^2-\hat S^2)$ .

\langle\hat{\mathbf L}\cdot\hat{\mathbf S}\rangle=\frac{\hbar^2}{2}\big[j(j+1)-\ell(\ell+1)-\tfrac34\big] \qquad (3)

where: $j=\ell\pm\tfrac12$ for $s=\tfrac12$ ; the two branches give $+\tfrac{\hbar^2}{2}\ell$ and $-\tfrac{\hbar^2}{2}(\ell+1)$ . Weighted by $2j+1$ these sum to zero — the centre-of-gravity rule any doublet must obey. In hydrogen the three corrections merge into the Dirac fine-structure formula, depending only on $n$ and $j$ — the coincidence that keeps $2S_{1/2}$ and $2P_{1/2}$ degenerate.

E_{nj}=-\frac{\mathrm{Ry}}{n^2}\left[1+\frac{\alpha^2}{n^2}\left(\frac{n}{j+\tfrac12}-\frac34\right)\right] \qquad (4)

where: $\mathrm{Ry}=13.6057$ eV; $\alpha\approx1/137.036$ ; $E_{nj}$ — energy including gross plus fine structure. Every state with the same $n,j$ is degenerate, an exact consequence of the Dirac equation broken only by QED. Differencing (4) at fixed $n=2$ , $\ell=1$ gives the quantity this article owns as a spectrum — the hydrogen $2P$ doublet, evaluated.

\Delta E(nP)=E(2P_{3/2})-E(2P_{1/2})=\frac{\mathrm{Ry}\,\alpha^2}{2n^3} \qquad (5)

where: at $n=2$ this is $4.53\times10^{-5}$ eV $=10.95$ GHz $=0.3652$ cm⁻¹; at $n=3$ , $0.108$ cm⁻¹. The $1/n^3$ closes the doublet for high shells; the leading $\alpha^2$ is why it is five orders below the gross structure. For a many-electron atom the same physics is packaged as a spin-orbit constant $\zeta_{n\ell}$ , and the $p$ -electron doublet spans exactly $\tfrac32\zeta$ — the form spectroscopists tabulate.

\Delta E_{\text{doublet}}=\tfrac32\,\zeta_{n\ell}\quad(p\text{-electron}),\qquad \zeta_{3p}(\mathrm{Na})=\tfrac23\times17.20=11.46\ \mathrm{cm^{-1}} \qquad (6)

where: $\zeta_{n\ell}=\hbar^2\langle\xi(r)\rangle$ — the spectroscopic spin-orbit parameter; the pattern is $j=\ell+\tfrac12$ up by $\zeta/2$ , $j=\ell-\tfrac12$ down by $\zeta$ , gap $\tfrac32\zeta$ . Reading $\zeta$ off a measured doublet is how one gets the coupling for atoms the hydrogen formula cannot touch.

Derivation AEvaluating the hydrogen 2P splitting into gigahertz

The goal is to take the Dirac fine-structure formula recalled from aq-qm-fine-structure-hydrogen and turn it into the actual hydrogen $2P_{3/2}$ – $2P_{1/2}$ interval a spectroscopist measures — in eV, GHz, and cm⁻¹ — then read off the $1/n^3$ scaling that makes higher shells' doublets shrink. The formula's derivation (the $\ell$ -cancellation of the three corrections) is owned upstream; what is owned here is the evaluation, the conversions, and the instrument-scale reading. We difference the two $j$ branches symbolically first, so the algebra is exact, then insert numbers.

Step 1. Write $E_{nj}=-\tfrac{\mathrm{Ry}}{n^2}[1+\tfrac{\alpha^2}{n^2}(\tfrac{n}{j+1/2}-\tfrac34)]$ and difference the $j=\tfrac32$ and $j=\tfrac12$ levels at fixed $n,\ell=1$ . Only the $\tfrac{n}{j+1/2}$ term survives, giving

\Delta E(nP)=\frac{\mathrm{Ry}\,\alpha^2}{n^3}\left(\frac{1}{1}-\frac{1}{2}\right)=\frac{\mathrm{Ry}\,\alpha^2}{2n^3}.

Why the $\tfrac34$ and the leading $1$ drop out: they are common to both branches, so the doublet gap keeps only the $j$ -dependent piece — a clean $\mathrm{Ry}\,\alpha^2/2n^3$ , verified symbolically.

Step 2. Insert $n=2$ : $\Delta E=\mathrm{Ry}\,\alpha^2/16=(13.6057)(5.325\times10^{-5})/16=4.53\times10^{-5}$ eV. Why this is "fine": it is the gross scale $\mathrm{Ry}$ times $\alpha^2/16$ , five orders of magnitude down — a $3.4$ eV level cracked by $45$ μeV.

Step 3. Convert. As a frequency, $\Delta E/h=10.95$ GHz; as a wavenumber, $\Delta E\times8065.54=0.3652$ cm⁻¹. These are microwave and far-sub-optical scales. Why the units matter: $10.95$ GHz is why the $2P$ fine structure is a microwave measurement, not an optical one — the same reasoning that pushed Lamb to microwaves for the residual gap.

Step 4. Scale to $n=3$ : $\Delta E\propto1/n^3$ gives $0.3652\times(8/27)=0.108$ cm⁻¹. Higher shells barely visit the core, so their doublets close up. Verified in verify_aq-optics-fine-structure-spin-orbit.py.

Go deeper · Derivation B · 1 minThe doublet from ζ L·S, and reading ζ off the sodium D lines

Derivation BThe doublet from ζ L·S, and reading ζ off the sodium D lines

The goal is to show that a $p$ -electron term splits into a doublet of total width $\tfrac32\zeta$ , prove the centre-of-gravity rule that disciplines it, and invert a measured splitting — sodium's D doublet — to extract the spin-orbit constant $\zeta_{3p}$ that no hydrogenic formula can supply for an alkali. The $\hat{\mathbf L}\cdot\hat{\mathbf S}$ eigenvalues and the coupling operator are recalled from aq-qm-spin-orbit-coupling; the evaluation into wavenumbers and the extraction are owned here as spectroscopy.

Step 1. From (3), a $p$ -electron ( $\ell=1$ ) has $\langle\hat{\mathbf L}\cdot\hat{\mathbf S}\rangle=+\tfrac{\hbar^2}{2}$ for $j=\tfrac32$ and $-\hbar^2$ for $j=\tfrac12$ . With $\hat H_{SO}=(\zeta/\hbar^2)\hat{\mathbf L}\cdot\hat{\mathbf S}$ the levels sit at $+\tfrac{\zeta}{2}$ and $-\zeta$ . Why the antiparallel level drops twice as far: it holds only $2\ell=2$ states against the parallel level's $2\ell+2=4$ , so it must sit proportionally farther from the centroid.

Step 2. The gap is therefore $\Delta E=\tfrac{\zeta}{2}-(-\zeta)=\tfrac32\zeta$ . Check the centre of gravity: $4(+\tfrac{\zeta}{2})+2(-\zeta)=0$ , so spin-orbit only redistributes states about an unmoved centroid. Why the sum rule is a shield: any proposed doublet pattern that fails $\sum(2j+1)\delta E=0$ is wrong before a single constant is checked.

Step 3. Sodium's yellow doublet is the $3p\to3s$ transition; the lower $3s$ is an $\ell=0$ state with no spin-orbit splitting, so the whole doubling lives in the $3p$ term. From the NIST air wavelengths $D_2=588.9950$ nm and $D_1=589.5924$ nm,

\Delta\tilde\nu=\frac{1}{\lambda_{D_2}}-\frac{1}{\lambda_{D_1}}=17.20\ \mathrm{cm^{-1}}=515.5\ \mathrm{GHz}=2.13\ \mathrm{meV}.

Why this is a measurement of the electron, not a prediction: sodium's $3p$ electron penetrates a ten-electron core, so no $Z=1$ formula applies — the honest move is to read the gap and extract the coupling.

Step 4. Invert with $\Delta E=\tfrac32\zeta$ : $\zeta_{3p}(\mathrm{Na})=\tfrac23\times17.20=11.46$ cm⁻¹, matching the tabulated value. Verified in verify_aq-optics-fine-structure-spin-orbit.py.

Go deeper · Derivation C · 1 minThe Z⁴ ladder: why the doublet explodes from lithium to caesium

Derivation CThe Z⁴ ladder: why the doublet explodes from lithium to caesium

The goal is the scaling that turns a fine correction into a coarse one — that spin-orbit coupling grows as $Z^4$ — and its display as the measured alkali ladder from lithium to caesium, ending where the perturbative "fine" language breaks down and LS coupling gives way to jj. The $\langle1/r^3\rangle$ moment and the $Z$ -dependence are owned by aq-qm-spin-orbit-coupling; the empirical ladder, read from D-line wavelengths, is owned here.

Step 1. In a hydrogenic field the coupling collects one power of $Z$ from the potential gradient $dV/dr\propto Z$ and three from the radial moment $\langle1/r^3\rangle\propto Z^3/a_0^3$ , so $E_{SO}\propto Z^4/n^3$ . Scaling hydrogen's $2p$ gap, He⁺ ( $Z=2$ ) is $16\times$ larger, $7.25\times10^{-4}$ eV. Why $Z^4$ and not $Z^2$ : raising $Z$ both shrinks the orbit (three powers, via $\langle1/r^3\rangle$ ) and steepens the field the electron rides through (one power) — the two multiply.

Step 2. Real alkalis are hydrogenic only outside their closed core; inside, the valence electron penetrates to feel a much larger effective charge, so the true growth is steeper than a single-electron $Z=1$ estimate and must be read from data. Why the penetration matters: spin-orbit is a core tax weighted by $1/r^3$ , levied exactly where the shielding fails — so it magnifies the innermost part of the orbit.

Step 3. Read the ladder from the measured D-doublets: Li $0.34$ , Na $17.2$ , K $57.8$ , Rb $237.6$ , Cs $554.0$ cm⁻¹. From sodium to caesium the splitting grows by a factor $554.0/17.2=32$ . Why this is the headline: a "fine" structure of hundreds of cm⁻¹ in caesium is no longer a small correction — it is a gross feature of the spectrum, and the coupling scheme itself must change.

Step 4. By the heaviest atoms $\hat{\mathbf L}\cdot\hat{\mathbf S}$ rivals the electrostatic term, so $LS$ (Russell-Saunders) coupling breaks down and each electron's $\mathbf l_i,\mathbf s_i$ couple first — the $jj$ regime developed in the coupling-scheme serials that follow. Verified in verify_aq-optics-fine-structure-spin-orbit.py.

Limiting cases

Case	Result	Meaning
$\alpha\to0$	$E_{\text{fs}}\to0$	fine structure is a relativistic effect; the Schrödinger spectrum is the $c\to\infty$ world
$\ell=0$ (S states)	$\hat{\mathbf L}\cdot\hat{\mathbf S}=0$	no orbit, no internal field; the S-state shift is the kinetic + Darwin terms, not spin-orbit
same $j$ , different $\ell$	$2S_{1/2}=2P_{1/2}$ exactly	Dirac-protected degeneracy, broken only by QED — the residual gap is the Lamb shift
high $n$ , fixed $\ell$	$\Delta E\propto1/n^3$	Rydberg doublets close up; the electron barely visits the $1/r^3$ core
large $Z$	$\Delta E\propto Z^4$	fine structure stops being fine; heavy atoms abandon $LS$ for $jj$ coupling
$\mu_B B\gtrsim\Delta E_{\text{fs}}$	$\hat{\mathbf L},\hat{\mathbf S}$ decouple	an external field beating the spin-orbit interval forces the Paschen-Back regime

Common trapFine structure is the same thing as hyperfine structure

It is not — they differ by the nuclear-magneton suppression and by which spins are involved. Fine structure comes from the electron's spin coupling to its own orbital motion, at the scale $\alpha^2 E_n$ (tens of μeV to hundreds of cm⁻¹). Hyperfine structure comes from the nucleus's moment coupling to the electron, smaller by roughly the electron-to-nucleon mass ratio $m_e/m_p\approx1/1836$ — kilohertz-to-gigahertz, three orders finer, the subject of a separate serial. A second trap: never quote the spin-orbit formula for an $\ell=0$ state. There $\hat{\mathbf L}\cdot\hat{\mathbf S}=0$ identically, so the $\langle1/r^3\rangle$ divergence never fires; the S-state fine structure comes entirely from the kinetic and Darwin terms.

Worked examples

1 · The sodium D doublet, read as a spin-orbit measurement (a real instrument number)

A sodium lamp emits the D doublet at $D_2=588.9950$ nm and $D_1=589.5924$ nm (NIST air wavelengths). Find the splitting in nm, cm⁻¹, GHz, and meV, and extract the $3p$ spin-orbit constant $\zeta_{3p}$ .

Step 1. Wavelength gap: $\Delta\lambda=589.5924-588.9950=\mathbf{0.5974\ nm}$ — the yellow line's famous doubling, resolvable on a bench grating.

Step 2. Wavenumber: $\Delta\tilde\nu=1/\lambda_{D_2}-1/\lambda_{D_1}=\mathbf{17.20\ cm^{-1}}$ . As a frequency $c\,\Delta\tilde\nu=\mathbf{515.5\ GHz}$ ; as an energy $hc\,\Delta\tilde\nu=\mathbf{2.13\ meV}$ .

Step 3. Both components share the lower $3s$ level ( $\ell=0$ , no splitting), so the $17.20$ cm⁻¹ is entirely the $3p$ fine structure. With $\Delta E=\tfrac32\zeta_{3p}$ , $\zeta_{3p}=\tfrac23\times17.20=\mathbf{11.46\ cm^{-1}}$ .

Step 4. Sanity: this is $\sim47\times$ the hydrogen $2p$ gap ( $0.365$ cm⁻¹) for "one valence electron" — because sodium's $3p$ penetrates the core and samples a far larger effective charge. The D doublet is the electron's spin-orbit coupling, photographed: 17.20 cm⁻¹, ζ₃ₚ = 11.46 cm⁻¹.

2 · What Michelson needed to resolve: the Hα doublet

Hydrogen's red Balmer line Hα ( $n=3\to2$ ) carries fine structure dominated by the $n=2$ splitting. Find the wavelength, the fine-structure $\Delta\lambda$ , and the resolving power required.

Step 1. Photon energy $E_\gamma=\mathrm{Ry}(1/4-1/9)=1.889$ eV, so $\lambda=hc/E_\gamma=\mathbf{656.3\ nm}$ .

Step 2. A shift $\delta E$ of a terminal level moves the line by $\Delta\lambda=\lambda\,\delta E/E_\gamma$ . With the dominant $2P$ interval $\delta E=4.53\times10^{-5}$ eV: $\Delta\lambda=656.3\times(4.53\times10^{-5}/1.889)=\mathbf{0.0157\ nm}$ .

Step 3. Resolving power: $R=\lambda/\Delta\lambda\approx\mathbf{4.2\times10^{4}}$ . A prism manages hundreds; 1880s gratings a few thousand in practice.

Step 4. Michelson's interferometer read fringe visibility as the mirror separation grew — no such ceiling — which is how 1887 could see a splitting theory would not explain until 1928. His reported width $\sim0.014$ nm sits honestly beside the leading-order $0.0157$ nm. Fine structure was measured 41 years before it was explained; it takes R ≈ 42 000 to see it.

Physics of the portrait

Fine structure is a concept with a four-act human story: a number measured before anyone could explain it, an old-quantum near-miss, the spin that supplied the mechanism, and the equation that made it exact.

Origins · 1 minfine structure and how this was discovered

fine structure

the relativistic anatomy of spectral lines · Michelson 1887 → Sommerfeld 1916 → Uhlenbeck–Goudsmit 1925 → Dirac 1928

The name and the number came first. In 1887 — the year his interferometer failed to find the ether — Albert Michelson turned the same instrument on a hydrogen lamp and, reading how the interference fringes washed out as the path difference grew, found the red Hα line was a close doublet about $0.014$ nm wide. He had no theory and wanted none. The first computation came from Arnold Sommerfeld, whose 1916 relativistic old-quantum orbits produced a splitting formula and, with it, a new dimensionless constant $\alpha$ — the fine-structure constant, named for exactly this splitting — though his integer index only mimicked the right answer by a fortunate cancellation. The mechanism arrived in 1925 with George Uhlenbeck and Samuel Goudsmit's electron spin: the doublets were spin reading its orientation against the orbit. That picture at first gave splittings twice too wide, until Llewellyn Thomas's 1926 relativistic-precession factor of one-half repaired them. Finally, in 1928, Dirac's relativistic equation delivered the exact spectrum depending only on $n$ and $j$ , with the three perturbative corrections revealed as its expansion — and left standing the one degeneracy, $2S_{1/2}=2P_{1/2}$ , that the Lamb shift would break in 1947. Fine structure is thus the rare concept whose measurement, naming, and explanation were separated by four decades and four physicists.

History

1887

— Michelson and Morley resolve the red Hα line into a close doublet by interferometric fringe visibility

1916

— Sommerfeld's relativistic old-quantum orbits give a splitting formula and the constant α

1925

— Uhlenbeck and Goudsmit's electron spin supplies the doublet mechanism

1926

— Thomas's relativistic precession halves the naive spin-orbit coupling, fixing the doublet widths

1928

— Dirac's equation delivers the exact n, j spectrum; the three corrections become its expansion

1947

— the residual 2S₁/₂–2P₁/₂ degeneracy falls: the Lamb shift opens the QED era

Going further

Texts. Foot, Atomic Physics, ch. 2 (fine structure and the alkalis) · Demtröder, Atoms, Molecules and Photons, ch. 5 · Bransden & Joachain, Physics of Atoms and Molecules, ch. 5 · Griffiths & Schroeter, Introduction to Quantum Mechanics, 3rd ed., §7.3 · Bethe & Salpeter, Quantum Mechanics of One- and Two-Electron Atoms.

Provenance. The three fine-structure corrections, the $n,j$ master formula and its Dirac origin are recalled by name from aq-qm-fine-structure-hydrogen; the spin-orbit operator $\hat H_{SO}=\xi(r)\hat{\mathbf L}\cdot\hat{\mathbf S}$ with the Thomas $\tfrac12$ , the eigenvalue $\tfrac{\hbar^2}{2}[j(j+1)-\ell(\ell+1)-\tfrac34]$ , the centre-of-gravity rule, and the $\langle1/r^3\rangle$ moment from aq-qm-spin-orbit-coupling (not re-derived here). A verifier should confirm: (1) the $2P_{3/2}$ – $2P_{1/2}$ gap $=\mathrm{Ry}\,\alpha^2/2n^3=0.365$ cm⁻¹ $=10.95$ GHz at $n=2$ , $0.108$ cm⁻¹ at $n=3$ ; (2) the Na D doublet $0.5974$ nm $=17.20$ cm⁻¹ $=515.5$ GHz $=2.13$ meV and $\zeta_{3p}=11.46$ cm⁻¹; (3) the alkali ladder K $57.8$ , Rb $237.6$ , Cs $554.0$ cm⁻¹ from air wavelengths, Cs/Na $=32$ ; (4) the Hα resolving power $R\approx4.2\times10^{4}$ . Values are NIST/CODATA/textbook; none fabricated.

Prerequisites. Solution of the Schrödinger Equation for Hydrogen · Electron Spin and the Spin Magnetic Moment · Quantum Numbers and Orbital Angular Momentum · Hydrogenic Wavefunctions and Radial Distributions. Leads to. The Lamb Shift · Hyperfine Structure · The Zeeman Effect · The Paschen-Back and Anomalous Zeeman Regimes.

Why does each of the three corrections depend on ℓ but their sum depends only on j?

Because their $\ell$ -dependences are not independent — all three are built from the same Coulomb expectation values, and those specific forms conspire so the $\ell$ 's cancel across both $j=\ell\pm\tfrac12$ branches, with the Darwin term covering $\ell=0$ . This is a special property of the $1/r$ potential and, more deeply, of the Dirac equation, whose exact Coulomb spectrum is $\ell$ -blind. In any non-Coulomb potential the conspiracy fails and levels genuinely depend on both $\ell$ and $j$ — which is exactly why sodium's doublets do not follow the hydrogen pattern and its $\zeta$ must be read from data.

Is the 2S₁/₂–2P₁/₂ degeneracy exact or just a first-order accident?

Exact — within the Dirac theory of a point Coulomb charge, at every order of the $\alpha$ expansion. Same $n$ , same $j$ , different $\ell$ : the master formula cannot tell the two apart, and the full Dirac spectrum protects the coincidence to all orders. What finally splits them, by about $1058$ MHz, is physics outside the Dirac equation — the quantized radiation field's radiative corrections. That is the Lamb shift, and it is why the degeneracy is famous rather than merely tidy.

Sodium's doublet is 17.2 cm⁻¹ but hydrogen's 2p gap is only 0.365 cm⁻¹ — why 47 times bigger for "one valence electron"?

Sodium's $3p$ electron is hydrogenic only outside the core. Inside, it penetrates ten shielding electrons and sees close to the full $Z=11$ nuclear field, and spin-orbit is a core tax weighted by $1/r^3$ — levied precisely where the shielding fails. The $Z=1$ hydrogen formula therefore badly underestimates alkali splittings, and the honest procedure is the worked example's: read the doublet and extract $\zeta_{3p}$ . The same penetration argument, scaled up, is what makes caesium's doublet $554$ cm⁻¹.

Why is spin parallel to the orbit the higher-energy state, not the lower?

Bar-magnet intuition misleads here because the electron's moment is antiparallel to its spin (negative charge): $\boldsymbol\mu_s=-(e/m)\hat{\mathbf S}$ . The internal field points along $\hat{\mathbf L}$ , and $-\boldsymbol\mu_s\cdot\mathbf B$ rewards $\boldsymbol\mu_s$ parallel to $\mathbf B$ , i.e. spin antiparallel to $\hat{\mathbf L}$ . So $j=\ell-\tfrac12$ is the cheap seat and $j=\ell+\tfrac12$ the expensive one in an attractive potential. (The ordering flips in nuclei, where the strong-force spin-orbit term has the opposite sign — the origin of the nuclear magic numbers.)

Do I ever need Clebsch–Gordan coefficients to get these 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 substitute $j,\ell,s$ and read off the eigenvalue. The actual coefficients are needed only for state questions: transition amplitudes, Zeeman mixing at intermediate fields, or expectation values of operators not built from $\hat J^2,\hat L^2,\hat S^2,\hat J_z$ . That machinery lives in its own serials and is consumed by name here.

Why did Michelson measure the doublet in 1887 but nobody explained it until 1928?

The explanation needed three revolutions that did not yet exist: relativity (1905), quantum theory of the atom (1913–16), and electron spin (1925). Sommerfeld's 1916 relativistic orbits got the formula's shape right by a fortunate cancellation — his integer index playing the role Dirac's $j+\tfrac12$ would later justify — but the correct mechanism required spin, then Thomas's factor of one-half, and finally the Dirac equation. Forty-one years from measurement to explanation: precision data patiently outwaiting theory, which is the recurring shape of atomic spectroscopy.