jj Coupling and Intermediate Coupling

An atom's energy levels come with labels — the term symbols $^{2S+1}L_J$ — but which labels are good depends on a contest inside the atom. In light atoms the electrostatic repulsion between electrons wins, so their orbital momenta lock together into a total $L$ and their spins into a total $S$ before either notices the spin–orbit interaction: this is LS (Russell–Saunders) coupling. In heavy atoms the spin–orbit interaction — whose strength we take from aq-qm-spin-orbit-coupling-relativity, where the $\zeta_{nl}\,\mathbf{L}\!\cdot\!\mathbf{S}$ term is derived and shown to scale steeply with nuclear charge — wins first, so each electron's own orbit and spin lock into a single $j$ before the electrons talk to one another: this is jj coupling. Most real atoms live in between. What makes the whole picture hang together is a conservation law that neither scheme can break: the total angular momentum $\mathbf{J}$ and the parity are exact constants of motion regardless of which interaction is stronger. So the two schemes cannot disagree about how many levels a configuration has or which $J$ values appear — they only relabel and reorder the same states. This article makes that invariance quantitative, shows how the Landé interval rule signals the drift from LS toward jj as you march down a column of the periodic table, and equips you to read a real spectrum and say how far from pure LS coupling an atom has strayed.

Key takeaways

LS coupling couples $\mathbf{L}=\sum\mathbf{l}_i$ and $\mathbf{S}=\sum\mathbf{s}_i$ first, then $\mathbf{J}=\mathbf{L}+\mathbf{S}$ — valid when electrostatic repulsion ≫ spin–orbit (light atoms).
jj coupling couples each electron's $\mathbf{j}_i=\mathbf{l}_i+\mathbf{s}_i$ first, then $\mathbf{J}=\sum\mathbf{j}_i$ — valid when spin–orbit ≫ residual electrostatic (heavy atoms).
The number of levels and the set of $J$ values are identical in both schemes: for two non-equivalent $p$ electrons, 10 levels / 36 states either way.
The Landé interval rule ( $E_J-E_{J-1}\propto J$ , giving a 2:1 ratio for a $^3P$ term) breaks down as coupling drifts toward jj; real carbon already shows 1.646, not 2.
KEY IDEA: the coupling scheme is a choice of good basis, not a change of physics — $J$ and parity are conserved, so LS and jj describe the same levels; only the labels, orderings, and intervals differ.

The core insight

LS and jj coupling are the two limits of a single problem — which of the electrostatic and spin–orbit interactions is diagonalized first — and because $J$ and parity are conserved they always yield the same count of levels with the same $J$ values.

Where this lives today

Intermediate coupling is the daily bread of atomic-structure codes. Packages such as Cowan's suite and the multiconfiguration Dirac–Hartree–Fock programs used to feed the NIST Atomic Spectra Database build the energy matrix from both the electrostatic Slater integrals and the spin–orbit parameters, then diagonalize — automatically landing wherever the atom actually sits between LS and jj. The scheme matters far beyond bookkeeping: lighting engineers exploit mercury's spin-forbidden $6\,^3P_1\to6\,^1S_0$ intercombination line at 253.7 nm (the germicidal and fluorescent-lamp workhorse), which is only weakly allowed because intermediate coupling mixes a little singlet character into the triplet. The same physics sets the branching ratios in the heavy-ion clocks and highly charged ions now used for tests of fundamental constants, where jj-coupling labels are the honest description.

The physical picture

Picture two competing "handshakes" inside a multi-electron atom. The first is the residual electrostatic repulsion — the part of the electron–electron Coulomb interaction left over after the average central field is removed. It does not care about spin, but it strongly favors particular relative orientations of the electrons' orbital motions, so it tends to lock the individual $\mathbf{l}_i$ into a well-defined total $\mathbf{L}$ (and, through the Pauli principle and exchange, a well-defined total $\mathbf{S}$ ). The second handshake is the spin–orbit interaction $\zeta_{nl}\,\mathbf{l}_i\!\cdot\!\mathbf{s}_i$ acting on each electron separately; it wants to lock each electron's own $\mathbf{l}_i$ and $\mathbf{s}_i$ into a single $\mathbf{j}_i$ . Whichever handshake is stronger happens "first" and defines the good quantum numbers. When electrostatics wins we get LS coupling and the familiar term symbols $^{2S+1}L_J$ ; when spin–orbit wins we get jj coupling and levels labeled by the pair $(j_1,j_2)_J$ . The strength ratio is $\zeta_{nl}/F$ , where $F$ is a Slater electrostatic integral. Because $\zeta_{nl}\propto Z_{\text{eff}}^4$ (it rides on $\langle 1/r^3\rangle$ ) while $F$ grows slowly, the ratio climbs steeply with atomic number: LS coupling is excellent for the first row, marginal by the middle of the table, and gives way to jj for the heaviest elements and for deeply bound inner shells.

Figure

The two coupling trees: LS binds L and S first, jj binds each jᵢ first

The payoff is that we can count without solving. For a given electron configuration the total number of microstates is fixed by combinatorics, and the total $\mathbf{J}$ decomposition is fixed by angular-momentum algebra. The next figure carries the punchline: for two non-equivalent $p$ electrons both schemes deliver exactly 10 levels built from 36 microstates, joined by a correlation diagram, and a real atom slides along that diagram as $Z$ grows.

Figure

LS→jj correlation for (np,n'p): 10 levels / 36 states either end — carbon already intermediate

The core equations

In LS coupling we build the two collective momenta and only then combine them. The good quantum numbers are $L$ , $S$ , and $J$ , and the label is the term symbol.

\mathbf{L}=\sum_i \mathbf{l}_i,\qquad \mathbf{S}=\sum_i \mathbf{s}_i,\qquad \mathbf{J}=\mathbf{L}+\mathbf{S}\;\Rightarrow\;{}^{2S+1}L_J

where: $\mathbf{l}_i,\mathbf{s}_i$ are the orbital and spin momenta of electron $i$ ; $L$ and $S$ are the total orbital and spin quantum numbers; $J$ runs over $|L-S|,\dots,L+S$ . The superscript $2S+1$ is the spin multiplicity (the number of $J$ levels in the term when $L\ge S$ ). This basis diagonalizes the residual electrostatic interaction, so it is the good basis when that interaction dominates. In jj coupling we lock each electron's own momenta first, then add the resulting $j_i$ together.

\mathbf{j}_i=\mathbf{l}_i+\mathbf{s}_i\;\;(j_i=l_i\pm\tfrac12),\qquad \mathbf{J}=\sum_i \mathbf{j}_i\;\Rightarrow\;(j_1,j_2,\dots)_J

where: each $j_i$ takes the two values $l_i\pm\tfrac12$ ; the level is labeled by the set of $j_i$ and the total $J$ . This basis diagonalizes the one-electron spin–orbit interaction, so it is the good basis when spin–orbit dominates. Both bases span the same space of states — they are related by a unitary transformation that leaves $J$ and parity untouched. Within an LS term the fine-structure splitting follows a simple rule, useful precisely because its failure diagnoses the drift toward jj. The spin–orbit energy is $\langle A\,\mathbf{L}\!\cdot\!\mathbf{S}\rangle=\tfrac{A}{2}[J(J+1)-L(L+1)-S(S+1)]$ , whose successive differences give:

E_J-E_{J-1}=A\,J\qquad(\text{Landé interval rule})

where: $A=\zeta_{nl}/[\,\dots\,]$ is the term's fine-structure constant (positive for less-than-half-filled shells, negative above). The rule says the spacing between adjacent fine-structure levels is proportional to the larger $J$ . For a $^3P$ term ( $S=L=1$ , $J=0,1,2$ ) it predicts $E_2-E_1 : E_1-E_0 = 2:1$ ; measured departures from 2 are a direct readout of intermediate coupling.

Derivation ALevel and state count is the same in LS and jj coupling

We want to prove the invariance concretely rather than by hand-waving about "the same Hilbert space." The strategy: take two non-equivalent $p$ electrons (so the Pauli principle imposes no restriction), enumerate the levels and the microstates once in the LS scheme and once in the jj scheme, and check that both the number of levels and the multiset of $J$ values agree. Watch for one subtlety — non-equivalent electrons keep all combinations, whereas equivalent electrons (same $n$ and $l$ ) drop the Pauli-forbidden ones, changing both counts but preserving the invariance.

Step 1. Count microstates first. Each $p$ electron has $2\times(2\cdot1+1)=6$ single-particle states (three $m_l$ times two $m_s$ ). Two distinguishable $p$ electrons therefore span $6\times6=36$ microstates — a number fixed by combinatorics that neither coupling scheme can alter. Why this matters: the microstate count is the dimension of the space; any complete set of levels must account for exactly these 36 states via $\sum_{\text{levels}}(2J+1)=36$ .

Step 2. Enumerate the LS terms. With $l_1=l_2=1$ the total $L$ can be $0,1,2$ and total $S$ can be $0,1$ , and for non-equivalent electrons every combination survives: the terms are $^1S,\,^3S,\,^1P,\,^3P,\,^1D,\,^3D$ . Splitting each into its $J$ levels gives $^1S_0,\,^3S_1,\,^1P_1,\,^3P_{0,1,2},\,^1D_2,\,^3D_{1,2,3}$ — 10 levels. Their weights sum to $1+3+3+(1+3+5)+5+(3+5+7)=36$ .

Step 3. Enumerate the jj levels. Each electron has $j=\tfrac12$ or $\tfrac32$ , giving four pairs $(\tfrac32,\tfrac32),(\tfrac32,\tfrac12),(\tfrac12,\tfrac32),(\tfrac12,\tfrac12)$ . Adding the two $j$ 's: $(\tfrac32,\tfrac32)\to J=0,1,2,3$ ; $(\tfrac32,\tfrac12)\to J=1,2$ ; $(\tfrac12,\tfrac32)\to J=1,2$ ; $(\tfrac12,\tfrac12)\to J=0,1$ — again 10 levels, with weights $(1+3+5+7)+(3+5)+(3+5)+(1+3)=36$ . Why the match is forced: both enumerations are just different ways of decomposing the tensor product of the two single-electron spaces into eigenspaces of the exact operators $J^2$ and parity, so they must return the same list of $J$ values.

Step 4. Check the equivalent-electron case for contrast. For $p^2$ (same $n$ ) the Pauli principle keeps only $^1S_0,\,^3P_{0,1,2},\,^1D_2$ in LS — the multiset $\{0,0,1,2,2\}$ , 5 levels, 15 states — and jj keeps $(\tfrac12,\tfrac12)_0,(\tfrac12,\tfrac32)_{1,2},(\tfrac32,\tfrac32)_{0,2}$ , the same multiset $\{0,0,1,2,2\}$ . The invariance holds with or without the Pauli restriction. Verified in verify_aq-optics-jj-coupling-intermediate.py (36 microstates and 10 levels both schemes for np n'p; {0,0,1,2,2} both schemes for equivalent p²).

Go deeper · Derivation B · 1 minThe Landé interval rule and its breakdown as a coupling diagnostic

Derivation BThe Landé interval rule and its breakdown as a coupling diagnostic

Our goal is to show why adjacent fine-structure intervals stand in the ratio $2:1$ inside a pure-LS $^3P$ term, and why the measured ratio drifts below $2$ as spin–orbit coupling grows — turning a textbook rule into a quantitative gauge of intermediate coupling. The route is to evaluate $\langle\mathbf{L}\!\cdot\!\mathbf{S}\rangle$ from the identity $2\,\mathbf{L}\!\cdot\!\mathbf{S}=J^2-L^2-S^2$ , take differences, and then compare with real carbon.

Step 1. Write the spin–orbit energy of a term as $E_J=\tfrac{A}{2}\,[J(J+1)-L(L+1)-S(S+1)]$ , using $\langle 2\,\mathbf{L}\!\cdot\!\mathbf{S}\rangle=J(J+1)-L(L+1)-S(S+1)$ . The constants $L(L+1)$ and $S(S+1)$ are common to all $J$ in the term and cancel in any difference. Why we may do this: within a single $LS$ term $\mathbf{L}\!\cdot\!\mathbf{S}$ is the only $J$ -dependent operator, and $\mathbf{L}$ and $\mathbf{S}$ have fixed magnitudes, so the term splits purely according to $J(J+1)$ . The magnitude parameter $A$ carries the underlying $\zeta_{nl}\,\mathbf{L}\!\cdot\!\mathbf{S}$ scale supplied by aq-qm-spin-orbit-coupling-relativity.

Step 2. Difference adjacent levels: $E_J-E_{J-1}=\tfrac{A}{2}[J(J+1)-(J-1)J]=A\,J$ . For a $^3P$ term the two intervals are $E_2-E_1=2A$ and $E_1-E_0=A$ , so their ratio is exactly $2:1$ in pure LS coupling.

Step 3. Now compare with carbon. The $2p^2$ ground term $^3P$ sits at $0$ , $16.40$ , and $43.40\ \text{cm}^{-1}$ for $J=0,1,2$ . The intervals are $16.40$ ( $J{=}1{-}0$ ) and $27.00\ \text{cm}^{-1}$ ( $J{=}2{-}1$ ), a ratio $27.00/16.40=1.646$ — already noticeably below $2$ . Why it drops: the true eigenstates are superpositions of LS terms of the same $J$ mixed by the spin–orbit interaction; the admixture shifts levels away from the pure- $A\,J$ pattern. The deficit $2\to1.646$ is a direct, dimensionless measure of how far even a light atom like carbon has moved toward jj coupling. Verified in verify_aq-optics-jj-coupling-intermediate.py (pure-LS ratio 2; carbon 27.00/16.40 = 1.646 < 2).

Limiting cases

case	result	meaning
$\zeta_{nl}/F\to 0$ (light atom, inner shells shallow)	pure LS: good $L,S,J$ ; Landé ratio $2:1$	electrostatics diagonalized first; term symbols $^{2S+1}L_J$ exact
$\zeta_{nl}/F\to \infty$ (very heavy atom / deep shell)	pure jj: good $j_1,j_2,J$	spin–orbit diagonalized first; levels $(j_1,j_2)_J$
intermediate ( $\zeta_{nl}\sim F$ )	eigenstates mix terms of equal $J$ ; Landé ratio $<2$	neither $L,S$ nor $j_i$ strictly good — only $J$ , parity
any scheme	same # levels, same $J$ multiset	$J$ and parity conserved throughout
spin-forbidden line ( $\Delta S\ne0$ )	intensity $\propto\zeta^2$ , grows with $Z$	singlet–triplet admixture opens intercombination lines (e.g. Hg 253.7 nm)

Worked examples

Example 1 — Reading intermediate coupling out of carbon's ground term. Carbon's $2p^2$ ground term $^3P$ has its $J=0,1,2$ levels at $0$ , $16.40$ , and $43.40\ \text{cm}^{-1}$ . Quantify how far carbon is from pure LS coupling.

Step 1. Form the two fine-structure intervals: $E_1-E_0=16.40-0=16.40\ \text{cm}^{-1}$ and $E_2-E_1=43.40-16.40=27.00\ \text{cm}^{-1}$ .

Step 2. Take their ratio: $27.00/16.40=1.646$ .

Step 3. Compare with the Landé prediction. Pure LS gives $E_2-E_1:E_1-E_0 = 2A:A = 2.000$ . The measured $1.646$ falls short of $2$ .

Result: carbon's interval ratio is 1.646 versus the pure-LS value 2.000, a $\sim18\%$ deficit — even the second-row element carbon shows measurable intermediate coupling, and the deficit grows rapidly for Si, Ge, Sn, and Pb.

Example 2 — Counting the levels of a $(3p\,4p)$ configuration two ways. An excited configuration with one $3p$ and one $4p$ electron is non-equivalent. How many levels does it have, and do LS and jj agree?

Step 1. Microstates: each $p$ electron has 6 states, so $6\times6=\textbf{36}$ microstates.

Step 2. LS levels: terms $^1S,\,^3S,\,^1P,\,^3P,\,^1D,\,^3D$ split into $^1S_0,\,^3S_1,\,^1P_1,\,^3P_{0,1,2},\,^1D_2,\,^3D_{1,2,3}$ — 10 levels, weights summing to $1+3+3+9+5+15=36$ .

Step 3. jj levels: pairs $(\tfrac32,\tfrac32),(\tfrac32,\tfrac12),(\tfrac12,\tfrac32),(\tfrac12,\tfrac12)$ give $J=\{0,1,2,3\},\{1,2\},\{1,2\},\{0,1\}$ — 10 levels, weights $16+8+8+4=36$ .

Result: both schemes give 10 levels / 36 states; only the labels and orderings differ. (For the equivalent $3p^2$ configuration the Pauli principle trims both to the same 5 levels $\{{}^1S_0,{}^3P_{0,1,2},{}^1D_2\}$ , 15 states.)

Physics of the portrait

jj coupling is not named for a person but for the algebra it performs: it is the heavy-atom limit in which each electron's $\mathbf{j}=\mathbf{l}+\mathbf{s}$ is formed first and the atom is labeled by the set of $j$ 's. It was the natural language once spectroscopists pushed into the heavy elements, where the Russell–Saunders term symbols that had organized the light-atom spectra so beautifully began to fail: intervals stopped obeying the $2:1$ rule, intercombination lines brightened, and the "good" quantum numbers $L$ and $S$ lost their meaning. The scheme is the endpoint of a continuous story — most atoms are described by neither pure limit but by an intermediate-coupling calculation that diagonalizes the electrostatic and spin–orbit interactions together.

Origins · 1 minj-j coupling and how this was discovered

j-j coupling

the heavy-atom limit where each electron's j = l + s is formed before the electrons couple to one another · atomic structure and spectra

The complementary scheme to Russell–Saunders (1925) coupling, jj coupling became indispensable as spectroscopy reached the heavy elements in the 1920s–1930s. Its steep onset traces the $\zeta_{nl}\propto Z_{\text{eff}}^4$ growth of the spin–orbit interaction; lead's $6p^2$ ground configuration is the textbook near-jj case, while carbon's $2p^2$ already shows the first departures from pure LS.

History

1925

— Russell and Saunders formalize LS coupling for the alkaline-earth spectra (Astrophysical Journal)

1925

— Pauli states the exclusion principle, fixing the allowed terms of equivalent electrons

1926

— Heisenberg and others develop the exchange origin of the electrostatic term splitting

1927

— Hund's rules organize the ground terms within a configuration

1935

— Condon and Shortley's Theory of Atomic Spectra codifies intermediate-coupling matrices

Going further

References. Foot, Atomic Physics (2005), Ch. 5 (LS coupling) and §5.5–5.6 (jj and intermediate coupling) · Demtröder, Atoms, Molecules and Photons — angular-momentum coupling schemes · Cowan, The Theory of Atomic Structure and Spectra (1981) — intermediate-coupling matrices · Bransden & Joachain, Physics of Atoms and Molecules, Ch. 8 · Condon & Shortley, The Theory of Atomic Spectra (1935).

Provenance. Coupling recipes and the $\mathbf{l}\!\cdot\!\mathbf{s}$ expectation value are standard (Foot Ch. 5; Bransden & Joachain Ch. 8). This upgrade recomputes: the $(np\,n'p)$ level and microstate counts (10 levels / 36 states in both schemes) and the equivalent- $p^2$ multiset $\{0,0,1,2,2\}$ ; the Landé interval rule $E_J-E_{J-1}=AJ$ giving the pure-LS $2:1$ ratio; and carbon's measured $^3P$ interval ratio $27.00/16.40=1.646$ from the levels $0,16.40,43.40\ \text{cm}^{-1}$ . The $\zeta_{nl}\propto Z_{\text{eff}}^4$ scaling follows from hydrogenic $\langle 1/r^3\rangle$ (Foot; Condon & Shortley); the fine-structure origin $\zeta_{nl}\,\mathbf{L}\!\cdot\!\mathbf{S}$ is recalled by name from aq-qm-spin-orbit-coupling-relativity. The Hg 253.7 nm intercombination line is a standard value (Demtröder).

Builds on: Fine Structure and Spin–Orbit Coupling · LS Coupling and Term Symbols · Electron Spin and the Spin Magnetic Moment. Leads to: The Zeeman Effect · The Paschen–Back and Anomalous Zeeman Regimes · Hyperfine Structure.

If J and parity are always conserved, why bother with LS or jj labels at all?

Because $J$ and parity alone are not enough to organize a spectrum. A configuration like $np\,n'p$ has ten levels, several sharing the same $J$ ; you need extra labels to tell them apart, predict their orderings and intensities, and apply selection rules. LS labels ( $L,S$ ) work when electrostatics dominates; jj labels ( $j_1,j_2$ ) work when spin–orbit dominates. The labels are approximate "good enough" quantum numbers that make the level structure intelligible — even though the only exactly conserved quantities are $J$ and parity.

How do I know whether to use LS or jj coupling for a given atom?

Compare the spin–orbit parameter $\zeta_{nl}$ with the electrostatic (Slater) energy $F$ . Light atoms and outer shells have $\zeta_{nl}\ll F$ → use LS. Heavy atoms and deeply bound shells have $\zeta_{nl}\gg F$ → use jj. In practice you rarely commit to a pure limit: you diagonalize both interactions together (intermediate coupling). A quick empirical test is the Landé interval rule — if the fine-structure ratios are close to the LS predictions, LS labels are meaningful; if they are badly off, you are well into the intermediate regime.

Why do equivalent electrons (like 2p²) have fewer levels than non-equivalent ones?

The Pauli exclusion principle. Two equivalent electrons share the same $n$ and $l$ , so no two of them may occupy the same $(m_l,m_s)$ state. That forbids the symmetric combinations that non-equivalent electrons are free to use, cutting $2p^2$ from the 36 microstates of $np\,n'p$ down to 15, and the ten levels down to five: $^1S_0,\,^3P_{0,1,2},\,^1D_2$ . Remarkably, the same trimming happens identically in the jj scheme, so the LS↔jj level count still matches — here at 5 levels rather than 10.

What makes a "spin-forbidden" intercombination line show up in mercury lamps?

In pure LS coupling a transition with $\Delta S\ne0$ is forbidden because the electric-dipole operator does not act on spin. But intermediate coupling means mercury's nominal triplet state is really a superposition that contains a small singlet amplitude $\sin\theta\approx V/\Delta E$ . The electric-dipole transition proceeds through that admixed singlet, with intensity $\propto\sin^2\theta\propto\zeta^2$ . Because $\zeta$ grows steeply with $Z$ , the effect is large in a heavy atom like mercury — enough to make the 253.7 nm intercombination line the bright, useful workhorse of fluorescent and germicidal lamps.

Does the ordering of levels stay the same between the LS and jj ends?

Not necessarily. The correlation diagram connects each LS level to exactly one jj level of the same $J$ , but the lines can cross: a level that is low-lying in the LS limit may end up higher in the jj limit and vice versa, subject to the rule that levels of the same $J$ never cross. That is exactly why the diagram is worth drawing — it tells you which physical level (at intermediate coupling) evolves from which idealized label, and it explains anomalies in heavy-atom orderings that look baffling if you insist on pure LS labels.

Is there anything between LS and jj, like a "half-way" scheme?

Yes — several. The most common named intermediate is $jK$ (or "pair") coupling, used when one electron is highly excited: the ionic core's momenta couple to $\mathbf{J}_c$ , that couples with the outer electron's orbital $\mathbf{l}$ to form $\mathbf{K}=\mathbf{J}_c+\mathbf{l}$ , and finally $\mathbf{K}+\mathbf{s}=\mathbf{J}$ . It fits rare-gas excited and Rydberg states, where the core spin–orbit interaction beats the weak core–outer-electron electrostatic coupling. But the truly general answer is a numerical intermediate-coupling calculation that diagonalizes both interactions at once and needs no named limit.

Why does the Landé interval ratio drop below 2 rather than above it?

Because spin–orbit mixing between terms of the same $J$ generically pushes levels together in a way that shrinks the larger interval relative to the smaller, lowering the ratio from the pure-LS value of 2. Physically, the "no-crossing" repulsion between same- $J$ levels of different terms perturbs the neat $E_J-E_{J-1}=AJ$ ladder. The magnitude of the drop scales with the spin–orbit strength, so heavier atoms (bigger $\zeta$ ) show larger deficits — carbon's 1.646 becomes far more pronounced by tin and lead.