Orbital Contributions to Magnetic Moment

A cobalt(II) complex in an octahedral field should behave almost identically to a chromium(III) complex. This assumes the simplest counting rules apply. Both would have the same number of unpaired electrons. It does not work out that way. Measured room-temperature moments for high-spin octahedral Co(II) routinely fall in the range 4.7–5.2 μB. This is far above the spin-only prediction of 3.87 μB for three unpaired electrons. Cr(III), by contrast, sits right on top of its spin-only value of 3.87 μB. Something beyond simple electron counting is at work. That something is orbital angular momentum surviving the crystal field. It couples back onto the spin through relativistic spin-orbit interaction. This matters because magnetic susceptibility is one of the cheapest, fastest structural probes a coordination chemist has. A single Gouy or Evans-method measurement can be compared against the Spin-Only Magnetic Moment Formula. This comparison is often used to assign oxidation state, coordination geometry, and even high-spin versus low-spin configuration in seconds. But the spin-only formula is only an approximation. It is valid only when orbital angular momentum is quenched by the ligand field. This holds true for most first-row 3d ions in weak or moderate fields with A or E ground terms. It is false for ions with T ground terms. It becomes increasingly false as you descend to 4d and 5d metals, where spin-orbit coupling is enormous. Knowing when and why the spin-only formula fails lets you read a magnetic moment correctly. Otherwise, you risk being misled by it. The big picture is a three-way competition. The free ion has full orbital angular momentum. This is described by Russell-Saunders terms, built from the Free-Ion Terms and Russell-Saunders Coupling framework. The ligand field partially or fully quenches that angular momentum. It does this by mixing real, phase-locked d-orbital combinations. These combinations no longer support circulating charge. Spin-orbit coupling then reintroduces a small amount of orbital character. It does this by mixing the ground term with excited terms of the same spin multiplicity. These excited terms lie nearby in energy. The net result depends sensitively on the symmetry label of the ground state. A and E terms are ideally quenched, though residual coupling still perturbs the g-value slightly. T terms retain full first-order orbital angular momentum and show the largest deviations. This article builds the physical machinery for that competition. It covers orbital quenching by real orbitals. It covers the group-theoretic criterion for which terms retain orbital angular momentum. It covers the Kotani/Figgis correction formula for T-term ions. It covers the sign rules that make τ2g and eg configurations behave oppositely. It ends with fully worked numerical predictions for Co(II), Ni(II), and a low-spin d⁵ Fe(III) case. It also covers common traps. One trap is confusing "high spin-orbit coupling constant" with "large deviation." Another is forgetting that covalency reduces λ below its free-ion value.

Key takeaways

The spin-only formula μ_S = 2√[S(S+1)] μB assumes zero orbital contribution; real ions often violate this.
Orbital angular momentum survives the ligand field only in orbitally degenerate ground terms (E or T symmetry, or more precisely T terms — E terms are quenched to first order but show second-order effects).
Spin-orbit coupling (constant λ, or ζ for a single electron) mixes the ground term with excited terms of the same spin, adding a correction ±(some multiple of λ)/Δ to μ.
The sign of the correction depends on whether the ground term arises from a less-than-half-filled or more-than-half-filled shell (i.e., the sign of λ).
Octahedral Co(II) (⁴T₁g) and Ni(II) in tetrahedral fields show the largest positive deviations; octahedral Ni(II) (³A₂g) is nearly spin-only.
Key idea: Magnetic moment deviates from the spin-only value exactly when the ground-state orbital symmetry allows angular momentum, and the size of the deviation is set by λ/Δoct via second-order spin-orbit mixing.

Where this lives today

Magnetic moment measurements remain a front-line diagnostic in synthetic inorganic and bioinorganic chemistry: a chemist characterizing a new cobalt or iron complex routinely runs an Evans-method NMR measurement or a SQUID magnetometry scan and compares the result against exactly the tables built from this theory, to assign spin state, geometry (tetrahedral vs octahedral vs square planar), and sometimes oxidation state, before ever growing a crystal for X-ray diffraction. In bioinorganic chemistry, the anomalously high moments of high-spin Fe(II)/Fe(III) heme centers and non-heme iron enzyme active sites are interpreted using precisely this spin-orbit correction framework, distinguishing genuine spin-state changes (spin crossover) from purely orbital effects. The same physics, pushed to its extreme, is the working principle behind modern single-molecule magnets (SMMs) and molecular qubits: lanthanide and some heavy transition-metal (4d/5d) complexes are deliberately designed to retain large, anisotropic orbital angular momentum (unquenched, unlike the mild first-row 3d T-term corrections discussed here) so that spin-orbit coupling produces a large zero-field splitting and a substantial energy barrier to spin reversal — the entire field of molecular magnetism for quantum information storage grows directly out of the competition between ligand-field quenching and spin-orbit coupling described in this article. Understanding when orbital angular momentum survives a ligand field is also essential background for interpreting EPR g-values (which deviate from the free-electron value $g_e=2.0023$ by amounts directly related to the same λ/Δ mixing), and for anticipating the reliability of a simple magnetic moment measurement before moving on to full variable-temperature susceptibility fitting, the subject taken up under Measuring Magnetic Moments.

Ask the tutor: Why does covalent metal-ligand bonding (the orbital reduction factor k) always shrink the magnitude of the orbital contribution rather than sometimes enhancing it?

The physical picture

Start from the free ion. A d^n configuration in Russell-Saunders coupling has a ground term ²S⁺¹L with both spin angular momentum S and orbital angular momentum L well defined, and the free-ion magnetic moment includes both, given by the Landé formula with gyromagnetic factor g_J. Place that ion in an octahedral or tetrahedral ligand field, and the field splits the (2L+1)-fold orbital degeneracy according to how the real d-orbitals transform under the octahedral group O_h. The crucial question for magnetism is: after this splitting, does the ground orbital level still support a nonzero expectation value of the orbital angular momentum operator L_z? The answer is almost always no if the ground orbital wavefunction is non-degenerate (an A₁g, A₂g, B₁g, or B₂g type function in the relevant point group) or doubly degenerate in a way built from real, standing-wave orbital combinations (Eg). A single real orbital such as d_{z²} or d_{x²−y²} is a standing wave — a superposition of +m and −m circulating components with equal weight and fixed phase — and has ⟨L_z⟩ = 0 identically. Superpositions of Eg orbitals behave the same way: no combination of real dz² and dx²−y² functions can generate net orbital current. This is orbital quenching, and it is why almost all high-spin 3d ions with A or E ground terms (Cr³⁺, Mn²⁺ high-spin, Ni²⁺ octahedral, Cu²⁺) sit close to their spin-only moments. The situation is entirely different for T-symmetry ground terms (T₁g or T₂g in O_h). These triply degenerate levels are built from combinations of real orbitals that can be recombined into complex, angular-momentum-bearing combinations — because the T₂g set (dxy, dxz, dyz) transforms exactly like the p-orbital set (px, py, pz) under rotation, and p-orbitals are well known to carry orbital angular momentum ℓ=1. Formally, the T₂g representation in O_h is isomorphic to the P representation (L=1) of the full rotation group; this is often summarized by saying "T₂g behaves like an atomic L=1 state," so an electron or hole in a t2g^n configuration with a T ground term retains an effective orbital angular momentum with quantum number ℓ_eff = 1. Spin-orbit coupling then acts as a second-order perturbation, mixing the ground T term with excited terms of the same spin multiplicity lying at energy Δ above it (typically another component of the same free-ion term, split by the field). This mixing adds a correction term to the magnetic moment, and — this is the subtle and testable part — the sign of the correction depends on the sign of the one-electron spin-orbit coupling constant ζ, which is positive for shells less than half full and negative for shells more than half full. That sign rule is why Co²⁺ (d⁷, more than half full, negative λ, ground term ⁴T₁g) shows a large positive deviation from spin-only, while Ti³⁺ (d¹, less than half full, positive λ) shows a negative deviation.

Figure

Quenching: a T2g "p-like" orbital set vs an Eg standing-wave set

The magnitude of the residual orbital contribution is controlled by the ratio of the spin-orbit coupling constant to the crystal-field (or term) splitting, λ/Δ. Because Δ (typically thousands of cm⁻¹, e.g. Δoct ≈ 10Dq) is usually much larger than λ (tens to a few hundred cm⁻¹ for 3d ions), the correction is a modest but experimentally decisive shift — enough to move Co²⁺ from 3.87 μB to ~4.7–5.2 μB, but not enough to double it. For 4d/5d ions λ is one to two orders of magnitude larger and can be comparable to Δ, breaking the perturbative treatment altogether and requiring full intermediate-coupling or relativistic (j-j) treatments.

Figure

Energy-level cartoon: free ion → octahedral field → spin-orbit mixing

The core ideas

The spin-only starting point, valid when orbital angular momentum is fully quenched:

\mu_{S} = g_e\sqrt{S(S+1)}\ \mu_B \approx 2\sqrt{S(S+1)}\ \mu_B

where: $g_e \approx 2.0023$ is the free-electron g-factor, $S$ is total spin quantum number, $\mu_B$ the Bohr magneton. Reading: this is the zeroth-order answer — correct only if $\langle L_z\rangle = 0$ in the ground state, i.e. the ground orbital term is orbitally non-degenerate (A or E) and any residual coupling is negligible. The full free-ion (Landé) moment, needed when orbital angular momentum is not quenched at all (e.g., lanthanides, where the ligand field is a weak perturbation on strong intrinsic spin-orbit coupling):

\mu_{J} = g_J\sqrt{J(J+1)}\ \mu_B, \qquad g_J = 1 + \frac{J(J+1)+S(S+1)-L(L+1)}{2J(J+1)}

where: $J$ is the total angular momentum quantum number from Russell-Saunders coupling, $L$ the total orbital angular momentum, $g_J$ the Landé g-factor. Reading: this is the opposite limit from spin-only — full orbital and spin angular momentum both survive, appropriate to 4f ions where the ligand field is weak compared to spin-orbit coupling (the crystal field acts on a shielded 4f shell). The transition-metal working formula (Figgis/Kotani correction), valid for octahedral or tetrahedral 3d ions with orbitally degenerate ground terms, treating spin-orbit coupling as a perturbation of order λ/Δ:

\mu_{\text{eff}} = \mu_S\left(1 - \frac{\alpha\lambda}{\Delta}\right)

where: $\mu_S$ is the spin-only value, $\lambda$ is the spin-orbit coupling constant for the free ion (sign included: positive for half-filled shells), $\Delta$ is the energy gap to the nearest excited term of the same spin multiplicity connected by the orbital angular momentum operator, and $\alpha$ is a numerical constant (typically 2, 4, or a fraction thereof) fixed by the specific T term and the orbital reduction factor. Reading: since $\lambda$ is negative for more-than-half-filled shells, $-\alpha\lambda/\Delta$ becomes positive, raising μ above spin-only (Co²⁺, Fe²⁺); since λ is positive for less-than-half-filled shells, the correction lowers μ below spin-only (Ti³⁺, V³⁺). This single sign rule organizes essentially the whole magnetochemistry of first-row T-term ions.

Derivation AWhy T2g orbitals carry angular momentum but Eg orbitals do not

Goal. Show explicitly, from the definitions of the real d-orbitals, that $\langle L_z\rangle=0$ for any real combination of $d_{z^2}, d_{x^2-y^2}$ (Eg) but that the t2g set can be recombined into eigenstates of $L_z$ with eigenvalue ±1 (times ℏ), justifying the "T looks like P (L=1)" isomorphism.

Step 1 — complex m-basis. The five hydrogenic d-orbitals are most naturally complex functions $Y_2^m$ with $m=-2,-1,0,1,2$ , each an eigenfunction of $L_z$ with eigenvalue $m\hbar$ . Real orbitals are formed by taking symmetric and antisymmetric combinations:

d_{xz} = \frac{1}{\sqrt2}\left(Y_2^{-1}-Y_2^{1}\right), \quad d_{yz} = \frac{i}{\sqrt2}\left(Y_2^{-1}+Y_2^{1}\right)

d_{xy} = \frac{i}{\sqrt2}\left(Y_2^{-2}-Y_2^{2}\right), \quad d_{x^2-y^2} = \frac{1}{\sqrt2}\left(Y_2^{-2}+Y_2^{2}\right), \quad d_{z^2}=Y_2^{0}

Step 2 — act with L_z on the Eg pair. Since $Y_2^0$ is already an $L_z$ eigenstate with eigenvalue 0, $d_{z^2}$ trivially has $\langle L_z\rangle=0$ . For $d_{x^2-y^2}=\tfrac{1}{\sqrt2}(Y_2^{-2}+Y_2^{2})$ ,

L_z\, d_{x^2-y^2} = \frac{1}{\sqrt2}\left(-2\hbar\, Y_2^{-2} + 2\hbar\, Y_2^{2}\right) = 2\hbar\, d_{xy}

This is not proportional to $d_{x^2-y^2}$ itself — $L_z$ rotates the orbital into a different real orbital rather than returning an eigenvalue, so $\langle d_{x^2-y^2}|L_z|d_{x^2-y^2}\rangle = 0$ by orthogonality of $d_{xy}$ and $d_{x^2-y^2}$ . The same holds for $d_{z^2}$ and any linear combination confined to the Eg pair: no nonzero diagonal matrix element of $L_z$ survives within the Eg manifold at first order, because $L_z$ always maps an Eg function outside the Eg subspace. This is the operational meaning of "quenching."

Step 3 — act with L_z on the T2g set. Apply $L_z$ to $d_{xy} = \tfrac{i}{\sqrt2}(Y_2^{-2}-Y_2^{2})$ :

L_z\, d_{xy} = \frac{i}{\sqrt2}\left(-2\hbar\, Y_2^{-2} - 2\hbar\, Y_2^{2}\right) = -2\hbar\, d_{x^2-y^2}

Again this maps out of the t2g set — but crucially, within the t2g set itself we can form new complex combinations that are closed under $L_z$ . Define

d_{+1} = \frac{1}{\sqrt2}\left(d_{xz}+i\,d_{yz}\right), \qquad d_{-1} = \frac{1}{\sqrt2}\left(d_{xz}-i\,d_{yz}\right)

Direct substitution shows $d_{+1}\propto Y_2^{-1}$ and $d_{-1}\propto Y_2^{1}$ (up to phase), so

L_z\, d_{\pm 1} = \mp\hbar\, d_{\pm1}

These are true $L_z$ eigenstates built entirely from t2g orbitals, with eigenvalues $\pm\hbar$ — exactly the pattern of an $\ell=1$ (p-like) system missing its $m=0$ partner in the relevant subspace (the $d_{xy}$ -like $m=0$ combination inside t2g plays an analogous role in the full treatment through the T2g×T2g coupling). This demonstrates concretely that the t2g manifold, unlike eg, contains orbital combinations that support real orbital angular momentum, and is the microscopic root of the isomorphism T2g ≅ P (L_eff=1) used throughout ligand-field magnetism.

Step 4 — consequence for terms. Any ground term whose orbital wavefunction is built from odd/hole occupation of the t2g set (T1g or T2g symmetry overall) inherits nonzero $\langle L_z\rangle$ in the appropriate real (unquenched) combination, and this is what spin-orbit coupling can subsequently exploit. Eg-based terms (and non-degenerate A-based terms) cannot, at first order, because the argument in Step 2 generalizes: $L_z$ always connects Eg to T2g, never Eg to Eg.

Go deeper · Derivation B · 2 minSecond-order correction to μ from spin-orbit mixing of T terms

Derivation BSecond-order correction to μ from spin-orbit mixing of T terms

Goal. Derive the Figgis-type correction $\mu_{\text{eff}} = \mu_S(1 - \alpha\lambda/\Delta)$ for a T-term ground state perturbed by an excited term of the same spin at energy $\Delta$ , and fix the sign/magnitude of $\alpha$ for the simplest illustrative case.

Step 1 — set up the perturbation. The spin-orbit Hamiltonian for a term is written in the coupled orbital-spin operator form

H_{SO} = \lambda\, \mathbf{L}\cdot\mathbf{S}

acting within the manifold of states built from the ground T term (orbital angular momentum operator $L$ , with effective $L=1$ inside T2g/T1g manifolds) and any excited term of the same S connected to it by nonzero matrix elements of $\mathbf{L}$ . Because $\mathbf{L}$ is a vector operator, only states of the same spin multiplicity are mixed at this order (Wigner-Eckart / spin selection rule), and the mixing coefficient between ground state $|0\rangle$ at energy 0 and excited state $|n\rangle$ at energy $\Delta$ is, from ordinary perturbation theory,

|\psi_0\rangle = |0\rangle + \sum_n \frac{\langle n|H_{SO}|0\rangle}{-\Delta}\,|n\rangle

Step 2 — magnetic moment operator. The magnetic moment operator along z is

\hat{\mu}_z = -\mu_B\left(g_e \hat S_z + \hat L_z\right)

so its expectation value in the perturbed ground state picks up a cross term between the zeroth-order $|0\rangle$ (which has $\langle 0|\hat L_z|0\rangle=0$ if $|0\rangle$ itself is a real, quenched combination, or a definite nonzero value if complex — for the thermally averaged, field-split case one instead evaluates the trace over the perturbed multiplet) and the admixed $|n\rangle$ . Carrying this through for an isotropic powder average (the standard treatment, e.g. as done originally by Kotani and Figgis) gives an effective moment of the form

\mu_{\text{eff}}^2 = \mu_S^2\left(1 - \frac{\alpha\lambda}{\Delta}\right)^2 \approx \mu_S^2\left(1-\frac{2\alpha\lambda}{\Delta}\right)

to first order in λ/Δ, which on taking the square root reproduces the working formula

\mu_{\text{eff}} \approx \mu_S\left(1 - \frac{\alpha\lambda}{\Delta}\right)

Step 3 — fix α for the T1g case relevant to d⁷ (Co²⁺) and d² (V³⁺). Detailed group-theoretical evaluation of the L-matrix elements connecting the ⁴T1g ground term to the ⁴T2g excited term in an octahedral d⁷ ion gives $\alpha = 4$ (this is the textbook value quoted for A2g/T1g-type d⁷ and d² systems; some texts write the equivalent form $\mu_{\text{eff}}=\mu_S(1-4\lambda/10Dq)$ identifying $\Delta\approx 10Dq$ as the relevant splitting for weak-field d⁷). For d³/d⁸ systems whose ground term is orbitally non-degenerate (⁴A2g, ³A2g) the leading correction instead appears only at order $(\lambda/\Delta)^2$ and is far smaller — consistent with the observation that octahedral Cr³⁺ (d³, ⁴A2g) and octahedral Ni²⁺ (d⁸, ³A2g) sit almost exactly on their spin-only values.

Step 4 — sign check. For d⁷ (more than half-filled t2g^5eg^2 configuration), $\lambda<0$ , so $-\alpha\lambda/\Delta>0$ and $\mu_{\text{eff}}>\mu_S$ — matching the experimental fact that Co²⁺ moments (4.7–5.2 μB) exceed the spin-only value (3.87 μB). For d² (V³⁺, less than half filled), $\lambda>0$ , giving $\mu_{\text{eff}}<\mu_S$ , consistent with observed V³⁺ moments (2.6–2.8 μB) sitting slightly below the spin-only value of 2.83 μB.

Limiting check. As $\lambda/\Delta \to 0$ (weak spin-orbit coupling or very large field splitting, as is nearly the case for 4d/5d ions with huge Δ, or formally in the limit of infinitely strong ligand field), $\mu_{\text{eff}}\to\mu_S$ : the spin-only formula is recovered exactly, as it must be, since it is the zeroth-order (unperturbed) result.

Worked examples

Example 1 — Foundational: predicting the direction of deviation for high-spin Fe²⁺ (d⁶) in an octahedral field. High-spin octahedral Fe²⁺ has configuration $t_{2g}^4 e_g^2$ , giving $S=2$ and a ⁵T2g ground term (T symmetry because the hole in the otherwise-filled t2g^5-analogue set — here four electrons among three degenerate orbitals — leaves orbital degeneracy). The spin-only moment is

\mu_S = 2\sqrt{S(S+1)} = 2\sqrt{2(3)} = 2\sqrt{6} = 4.90\ \mu_B

Since the ground term is ⁵T2g (orbitally degenerate, T symmetry) and d⁶ is more than half-filled ( $\lambda<0$ ), the correction $-\alpha\lambda/\Delta$ is positive, so we predict $\mu_{\text{eff}} > 4.90\ \mu_B$ . Experimentally, high-spin octahedral Fe²⁺ complexes show moments in the range 5.1–5.7 μB, confirming both the direction and rough scale of the prediction. As a limiting check: if Fe²⁺ were instead low-spin ( $t_{2g}^6$ , $S=0$ , ¹A1g ground term, e.g. in $[\text{Fe(CN)}_6]^{4-}$ ), the spin-only moment would trivially be $\mu_S = 2\sqrt{0(1)} = 0$ , and since the ground term is now non-degenerate (A1g), no first-order orbital contribution is possible — matching the observed diamagnetism of ferrocyanide.

Example 2 — Applied: comparing octahedral Ni²⁺ (should be near spin-only) against tetrahedral Co²⁺ (should show large deviation). Octahedral Ni²⁺ is d⁸, configuration $t_{2g}^6e_g^2$ , giving $S=1$ and ground term ³A2g (orbitally non-degenerate — the eg² configuration with both electrons unpaired in different eg orbitals has no orbital degeneracy left to exploit). Spin-only:

\mu_S = 2\sqrt{1(2)} = 2\sqrt{2} = 2.83\ \mu_B

Because the ground term is A-symmetry, only a small, second-order correction survives, and observed values for octahedral Ni²⁺ salts cluster tightly at 2.9–3.3 μB — close to, though systematically a little above, spin-only (the small excess here comes from second-order mixing with the excited ³T2g term via λ²/Δ² type terms, and from a small orbital contribution from the excited ³T1g term admixture, plus the fact that $\Delta$ is not infinite). Tetrahedral Co²⁺ is d⁷ in a tetrahedral field. Because tetrahedral splitting inverts the ordering relative to octahedral (e below t2), a d⁷ tetrahedral ion has configuration $e^4t_2^3$ and ground term ⁴A2 — formally non-degenerate, so one might naively expect near-spin-only behavior. In fact, because the tetrahedral crystal-field splitting $\Delta_{\text{tet}}$ is small (roughly $\Delta_{\text{tet}} = \tfrac{4}{9}\Delta_{\text{oct}}$ for the same metal-ligand pair), and because higher T2-symmetry terms lie close in energy and mix strongly via spin-orbit coupling even though the ground term itself is A2, tetrahedral Co²⁺ still shows substantially enhanced moments, typically 4.4–4.8 μB, well above the spin-only value of

\mu_S = 2\sqrt{\tfrac32\left(\tfrac32+1\right)} = 2\sqrt{3.75} = 3.87\ \mu_B

This illustrates an important subtlety: the nominal ground-term symmetry label is not the whole story when Δ is small enough that low-lying excited terms of the same spin are thermally and spin-orbitally accessible — the "T-term rule" is really shorthand for "orbital angular momentum survives when Δ is not enormous compared to λ, and especially when the ground term itself is T," but small-Δ A-term cases can still show meaningful deviations through nearby T-term admixture.

Example 3 — Exam/competition level: estimating μ_eff for octahedral Co²⁺ numerically from the Figgis formula, and checking self-consistency against a measured value. Take high-spin octahedral Co²⁺ (d⁷, $t_{2g}^5e_g^2$ , ⁴T1g ground term), with typical literature values $\lambda_{\text{free-ion}}(\text{Co}^{2+}) \approx -172\ \text{cm}^{-1}$ (negative, since d⁷ is more than half-filled) and a crystal-field splitting $10Dq \approx 9300\ \text{cm}^{-1}$ (representative for a weak/moderate-field octahedral Co²⁺ complex such as $[\text{Co(H}_2\text{O)}_6]^{2+}$ ). Using $\alpha=4$ and identifying $\Delta \approx 10Dq$ : Spin-only starting point:

\mu_S = 2\sqrt{\tfrac32\left(\tfrac52\right)} = 2\sqrt{3.75} = 3.873\ \mu_B

Correction factor:

\frac{\alpha\lambda}{\Delta} = \frac{4\times(-172)}{9300} = \frac{-688}{9300} = -0.0740

Corrected moment:

\mu_{\text{eff}} = \mu_S\left(1-\frac{\alpha\lambda}{\Delta}\right) = 3.873\times(1-(-0.0740)) = 3.873\times 1.0740 = 4.16\ \mu_B

This first-order estimate (4.16 μB) already shows a clear enhancement over spin-only, and undershoots the experimental range (4.7–5.2 μB) precisely because the true treatment for ⁴T1g ions requires the full multi-term intermediate-coupling calculation (Figgis's more complete tables, which include mixing with more than one excited quartet term, second-order terms in λ/Δ, and orbital reduction factors $k<1$ from covalency) rather than the single-term, first-order perturbative estimate used here for transparency. As a limiting-case sanity check: if we imagine artificially switching off spin-orbit coupling ( $\lambda\to 0$ ), the formula correctly collapses to $\mu_{\text{eff}}\to\mu_S = 3.87\ \mu_B$ , reproducing pure spin-only behavior — confirming the formula's internal consistency even though the magnitude of the real correction for T1g ions needs the fuller treatment to match experiment quantitatively. This is exactly why Co²⁺ is the textbook example chosen to teach students that simple first-order formulas correctly predict direction and rough scale but should not be trusted for precise numbers without consulting full Figgis tables or fitting susceptibility data directly.

Limiting cases, scope & common traps

Situation	Ground term type	Expected behavior	Representative μ_eff (μB)
Cr³⁺, oct., d³	⁴A2g	Spin-only, tiny 2nd-order correction	3.7–3.9 (spin-only 3.87)
Mn²⁺/Fe³⁺ high-spin, oct., d⁵	⁶A1g	Essentially exact spin-only (L=0 term)	5.7–6.0 (spin-only 5.92)
Ni²⁺, oct., d⁸	³A2g	Near spin-only, small positive excess	2.9–3.3 (spin-only 2.83)
V³⁺, oct., d²	³T1g	Below spin-only (λ>0, less than half-filled)	2.6–2.8 (spin-only 2.83)
Fe²⁺ high-spin, oct., d⁶	⁵T2g	Above spin-only (λ<0)	5.1–5.7 (spin-only 4.90)
Co²⁺ high-spin, oct., d⁷	⁴T1g	Large excess above spin-only	4.7–5.2 (spin-only 3.87)
Co²⁺, tetrahedral, d⁷	⁴A2 (nominal)	Still enhanced via nearby T-term mixing	4.4–4.8 (spin-only 3.87)
Lanthanides (most), 4f^n	Full J-term (weak field)	Landé/free-ion formula, not spin-only at all	e.g. Dy³⁺: 10.5–10.6 (calc. 10.63)
4d/5d ions	Any	λ huge; perturbative T-term formula breaks down	Requires full intermediate coupling

The chemist behind it

Origins · 1 minMasao Kotani and how this was discovered

Masao Kotani

1912–1986 · Japan, theoretical/molecular physics, mid-20th century

Kotani developed the group-theoretical and perturbative machinery for computing magnetic susceptibilities of transition-metal complexes with orbitally degenerate ground terms, publishing the key theoretical papers in the early 1950s that quantified how spin-orbit coupling modifies the spin-only formula for T-term ions. His treatment, extended and tabulated in practical form by Brian Figgis and collaborators, remains the standard reference calculation cited whenever a chemist needs to rationalize an "anomalous" magnetic moment for a first-row transition-metal ion — precisely the Co²⁺/Ni²⁺/Fe²⁺ comparisons worked through above.

History

- 1929 — Bethe formulates crystal-field theory, giving the group-theoretic language (A, E, T labels) needed to classify ground terms.

- 1932–1937 — Van Vleck develops the general quantum theory of magnetic susceptibilities, including the framework for temperature-independent paramagnetism and orbital contributions.

- 1951–1954 — Kotani derives perturbative expressions for μ_eff of T-term 3d ions in terms of λ/Δ.

- 1960s — Figgis and Lewis compile systematic experimental and theoretical tables of μ_eff across the first transition series, cementing the standard diagnostic tables still used today.

- 1970s–present — Extension to 4d/5d and lanthanide/actinide systems, and to single-molecule magnets, where large orbital contributions and magnetic anisotropy (not just isotropic μ_eff) become the central design parameter.

Why does the spin-only formula work so well for so many 3d ions if orbital angular momentum is "generally" present in the free ion?

Because in a strong-to-moderate octahedral or tetrahedral ligand field, most ground terms end up being orbitally non-degenerate (A-type) or doubly degenerate in the quenched Eg sense, not triply degenerate T-type. The ligand field, not the free ion, sets the relevant symmetry for magnetism in 3d complexes, since Δoct/Δtet (thousands of cm⁻¹) vastly exceeds λ (tens to a couple hundred cm⁻¹) for first-row ions. Only when the ground term specifically retains T symmetry (d¹, d², d⁶ high-spin, d⁷ high-spin in Oh, and various Td cases) does first-order orbital angular momentum survive to produce a sizeable correction.

How is this different for lanthanides — why can't I use the same octahedral T-term reasoning there?

For lanthanides the energy ordering is reversed: the 4f orbitals are core-like and shielded by filled 5s/5p shells, so the ligand field is a weak perturbation (hundreds of cm⁻¹) compared to spin-orbit coupling (thousands of cm⁻¹). Russell-Saunders J is therefore a good quantum number essentially undisturbed by the ligand environment, and the free-ion Landé formula $\mu_J = g_J\sqrt{J(J+1)}$ applies almost unmodified. This is the opposite hierarchy from 3d ions, where the ligand field dominates and spin-orbit coupling is the small perturbation — the same physics, evaluated in the opposite perturbative order.

Does the orbital contribution show temperature dependence?

Yes, and this is a diagnostic tool in its own right. For T-term ions the second-order spin-orbit mixing populates spin-orbit sublevels that are thermally accessible at comparable energies to kT, so μ_eff for ions like Co²⁺ genuinely varies with temperature (often decreasing somewhat on cooling as only the lowest spin-orbit component remains populated), unlike A/E-term ions (Cr³⁺, Mn²⁺, Ni²⁺) whose moments are essentially temperature-independent over normal ranges. Variable-temperature SQUID measurements can therefore distinguish orbital contributions from other sources of temperature dependence such as spin crossover or antiferromagnetic coupling.

What is an "orbital reduction factor" k, and why does it matter?

The free-ion value of λ used in these formulas assumes a purely ionic d-orbital with no covalent delocalization onto ligands. In real complexes, metal-ligand covalency spreads the metal d-electron density partly onto ligand orbitals, reducing the effective orbital angular momentum and the effective spin-orbit coupling actually felt by the "metal-centered" electron. This is captured by multiplying λ (or the orbital angular momentum operator) by an empirical factor $k<1$ (often 0.7–0.9 for 3d complexes with moderately covalent ligands like Cl⁻ or CN⁻), which brings the first-order perturbative predictions into closer agreement with experiment and is itself a useful covalency probe (the nephelauxetic-adjacent use of magnetic data).

Can a low-spin complex still show large orbital contributions?

Yes — what matters is the symmetry label of the ground term, not simply high-spin versus low-spin status. Low-spin octahedral Fe(III) (d⁵, $t_{2g}^5$ , ²T2g ground term) is a classic example: despite being low-spin with only one unpaired electron (spin-only prediction 1.73 μB), the ²T2g ground term is orbitally degenerate and shows substantial orbital contribution, with measured moments often in the 1.8–2.5 μB range and pronounced temperature dependence — one of the most dramatic deviations from spin-only in all of first-row coordination chemistry.

How would I quickly check, given only a dⁿ count and geometry, whether to expect a large deviation?

Work out the ground term symmetry from the standard octahedral/tetrahedral splitting diagrams (t2g/eg or e/t2 filling with Hund's rule), then check whether the resulting many-electron ground term carries a T (triply degenerate) label. If yes, expect a sizeable, sign-predictable deviation set by whether the shell is more or less than half-filled. If the ground term is A or E, expect near-spin-only behavior except for the ⁶S-type d⁵ case, which is exactly on spin-only, and be alert to nearby low-lying T terms in weak-field or tetrahedral geometries that can still mix in some orbital character even when the formal ground term is A.