Chapter 18: Degenerate Perturbation Theory and Fine Structure of Hydrogen

"The spectrum of hydrogen has proved to be the Rosetta Stone of modern physics: once this was clearly understood, the door was opened to other advances." — Arthur Schawlow, Nobel Lecture (1981)

"God made the bulk; surfaces were invented by the devil." — Wolfgang Pauli (adapted)

In Chapter 17, we developed a powerful machine: perturbation theory. Given a Hamiltonian $\hat{H} = \hat{H}_0 + \lambda \hat{H}'$ where $\hat{H}_0$ is solvable and $\hat{H}'$ is small, we systematically calculated energy corrections and state corrections order by order. The first-order energy correction was beautifully simple: $E_n^{(1)} = \langle n^{(0)} | \hat{H}' | n^{(0)} \rangle$. The second-order correction involved a sum over all other states. Everything worked.

Until it does not.

The formula for the second-order energy correction contains terms of the form:

$$E_n^{(2)} = \sum_{m \neq n} \frac{|\langle m^{(0)} | \hat{H}' | n^{(0)} \rangle|^2}{E_n^{(0)} - E_m^{(0)}}$$

What happens when $E_n^{(0)} = E_m^{(0)}$ for some $m \neq n$? The denominator vanishes. The correction diverges. The perturbation series explodes.

This is not a rare pathology. It is the generic situation in quantum mechanics. The hydrogen atom has $n^2$-fold degeneracy at each energy level (ignoring spin, $2n^2$ with spin). The three-dimensional harmonic oscillator has degeneracies that grow quadratically. Any system with continuous symmetry — rotational symmetry, for instance — has degenerate energy levels. Standard perturbation theory, as developed in Chapter 17, fails for the majority of physically interesting problems.

This chapter solves the problem. Section 18.1 diagnoses why degeneracy breaks perturbation theory and reveals that the cure is not a new formalism but a careful choice of basis. Section 18.2 develops the degenerate perturbation theory algorithm. Sections 18.3 through 18.6 apply it to the crown jewel of atomic physics: the fine structure of hydrogen. Section 18.7 descends to an even finer scale — hyperfine structure and the famous 21 cm line. Section 18.8 adds an external magnetic field, unfolding the full richness of the Zeeman effect.

By the end, you will understand why the "simple" hydrogen atom actually has an extraordinarily complex energy spectrum — and why that complexity tells us something profound about the structure of nature.

🔗 Connection: This chapter builds directly on the non-degenerate perturbation theory of Chapter 17. The hydrogen atom eigenstates from Chapter 5, the spin formalism from Chapter 13, and the addition of angular momenta from Chapter 14 are all essential. If any of these feel rusty, review the key results before proceeding.

18.1 When Perturbation Theory Meets Degeneracy

The Problem: Which State Do You Perturb?

Let us return to the core assumption of non-degenerate perturbation theory. In Chapter 17, we wrote the unperturbed state as $|n^{(0)}\rangle$ and assumed this state was unique — that there was exactly one eigenstate of $\hat{H}_0$ with eigenvalue $E_n^{(0)}$. The first-order correction was then:

$$E_n^{(1)} = \langle n^{(0)} | \hat{H}' | n^{(0)} \rangle$$

But suppose $E_n^{(0)}$ is $g$-fold degenerate. Then there are $g$ linearly independent eigenstates $|n, 1\rangle, |n, 2\rangle, \ldots, |n, g\rangle$ all sharing the same energy $E_n^{(0)}$. Any linear combination:

$$|\psi\rangle = \sum_{i=1}^{g} c_i |n, i\rangle$$

is also an eigenstate of $\hat{H}_0$ with the same eigenvalue. So which state should we use as our unperturbed state? The naive formula $E_n^{(1)} = \langle \psi | \hat{H}' | \psi \rangle$ gives a different answer depending on which linear combination we choose!

This is the crisis. Non-degenerate perturbation theory assumed a unique starting state. Degeneracy means there is an entire subspace of valid starting states, and the perturbation itself must tell us which starting state to use.

💡 Key Insight: The problem is not that perturbation theory fails. The problem is that we are perturbing from the wrong states. There exists a special basis within the degenerate subspace — the one that diagonalizes $\hat{H}'$ — and if we start from those states, everything works. Finding that basis is the entire content of degenerate perturbation theory.

A Simple Analogy

Before diving into the mathematics, consider an analogy. Imagine you are standing on a perfectly flat hilltop — a summit with exactly zero slope in every direction. If someone asks "which way is downhill?", the question has no unique answer. Every direction looks the same. But now suppose a gentle breeze begins to blow from the north. Suddenly, the hilltop is no longer perfectly symmetric — and if you roll a ball, it will tend to roll southward. The breeze has broken the symmetry and selected a preferred direction.

In quantum mechanics, the degenerate subspace is the flat hilltop, and the perturbation is the breeze. Without the perturbation, all linear combinations of degenerate states are equally valid eigenstates. The perturbation breaks the symmetry and selects specific linear combinations — the ones that diagonalize $\hat{H}'$. These "preferred" states are the ones that Nature actually uses.

A Concrete Example: Why the Wrong Basis Fails

Consider a two-fold degeneracy. The unperturbed states $|a\rangle$ and $|b\rangle$ share energy $E^{(0)}$. In this basis, the perturbation has matrix elements:

$$W_{aa} = \langle a | \hat{H}' | a \rangle, \quad W_{ab} = \langle a | \hat{H}' | b \rangle, \quad W_{ba} = \langle b | \hat{H}' | a \rangle, \quad W_{bb} = \langle b | \hat{H}' | b \rangle$$

If we naively apply first-order non-degenerate perturbation theory, we get $E_a^{(1)} = W_{aa}$ and $E_b^{(1)} = W_{bb}$. But the first-order state correction has terms proportional to:

$$\frac{\langle b | \hat{H}' | a \rangle}{E^{(0)} - E^{(0)}} = \frac{W_{ba}}{0}$$

This diverges whenever $W_{ba} \neq 0$ — that is, whenever the perturbation has off-diagonal matrix elements connecting degenerate states. In a randomly chosen basis, this is essentially guaranteed.

⚠️ Common Misconception: Students sometimes think that degenerate perturbation theory is a fundamentally different formalism from non-degenerate perturbation theory. It is not. It is the same perturbation theory, but applied after first finding the "right" basis within the degenerate subspace. The "right" basis is the one that makes $W_{ba} = 0$ — the one that diagonalizes $\hat{H}'$ within the degenerate subspace.

Physical Intuition: The Perturbation Breaks the Symmetry

Why does the perturbation prefer certain states over others? Because the perturbation breaks the symmetry that caused the degeneracy in the first place.

Consider the hydrogen atom's $n = 2$ level. Without any perturbation, the four states $|2,0,0\rangle$, $|2,1,0\rangle$, $|2,1,1\rangle$, $|2,1,-1\rangle$ (using $|n, l, m_l\rangle$) are all degenerate. This degeneracy exists because the Coulomb potential has perfect $\mathrm{SO}(3)$ rotational symmetry and a hidden $\mathrm{SO}(4)$ dynamical symmetry (the Runge-Lenz vector). Any perturbation — relativistic corrections, spin-orbit coupling, an external field — breaks some or all of this symmetry, lifting the degeneracy and selecting preferred states.

The preferred states are those that commute with both $\hat{H}_0$ and $\hat{H}'$. These are the good quantum numbers — the quantum numbers that remain well-defined even in the presence of the perturbation.

✅ Checkpoint: Before proceeding, make sure you can answer: (1) Why does the second-order energy formula diverge when degeneracy is present? (2) What does it mean to "diagonalize $\hat{H}'$ within the degenerate subspace"? (3) Why does the perturbation prefer certain linear combinations of degenerate states?

18.2 The Degenerate Case: Diagonalize $\hat{H}'$ in the Degenerate Subspace

The Algorithm

Here is the complete recipe for degenerate perturbation theory:

Step 1: Identify the degenerate subspace $\mathcal{D}$ of $\hat{H}_0$ with energy $E^{(0)}$. Let the degeneracy be $g$, and let $\{|1\rangle, |2\rangle, \ldots, |g\rangle\}$ be any convenient basis for $\mathcal{D}$.

Step 2: Construct the $g \times g$ matrix $\mathbf{W}$ of the perturbation restricted to $\mathcal{D}$:

$$W_{ij} = \langle i | \hat{H}' | j \rangle, \qquad i, j = 1, 2, \ldots, g$$

Step 3: Diagonalize $\mathbf{W}$. Find its eigenvalues $E^{(1)}_1, E^{(1)}_2, \ldots, E^{(1)}_g$ and eigenvectors $|\psi_1^{(0)}\rangle, |\psi_2^{(0)}\rangle, \ldots, |\psi_g^{(0)}\rangle$.

Step 4: The first-order corrected energies are:

$$E_k = E^{(0)} + E_k^{(1)}$$

where $E_k^{(1)}$ are the eigenvalues of $\mathbf{W}$.

Step 5: The "correct" zeroth-order states — the ones from which standard perturbation theory may proceed — are the eigenvectors $|\psi_k^{(0)}\rangle$.

That is the entire algorithm. The hard work is computing the matrix elements $W_{ij}$ and diagonalizing $\mathbf{W}$.

Worked Example: Two-Level System

Let us work through the simplest case in full detail. Suppose two states $|a\rangle$ and $|b\rangle$ share energy $E^{(0)}$, and the perturbation matrix is:

$$\mathbf{W} = \begin{pmatrix} W_{aa} & W_{ab} \\ W_{ab}^* & W_{bb} \end{pmatrix}$$

where we have used the Hermiticity of $\hat{H}'$ (which guarantees $W_{ba} = W_{ab}^*$).

The eigenvalues are found from the characteristic equation:

$$\det \begin{pmatrix} W_{aa} - E^{(1)} & W_{ab} \\ W_{ab}^* & W_{bb} - E^{(1)} \end{pmatrix} = 0$$

$$(W_{aa} - E^{(1)})(W_{bb} - E^{(1)}) - |W_{ab}|^2 = 0$$

Solving the quadratic:

$$E_{\pm}^{(1)} = \frac{W_{aa} + W_{bb}}{2} \pm \frac{1}{2}\sqrt{(W_{aa} - W_{bb})^2 + 4|W_{ab}|^2}$$

Several important features emerge:

1. The degeneracy is always lifted (at first order) unless $W_{aa} = W_{bb}$ and $W_{ab} = 0$ — that is, unless $\hat{H}'$ happens to be proportional to the identity within the degenerate subspace.

2. The splitting depends on $|W_{ab}|$, not on the phase of $W_{ab}$. This means the energy levels are gauge-invariant.

3. The "good" states are the eigenvectors of $\mathbf{W}$. If $W_{aa} = W_{bb}$ (the symmetric case), the good states are $(|a\rangle \pm e^{-i\phi}|b\rangle)/\sqrt{2}$ where $\phi$ is the phase of $W_{ab}$.

4. When $W_{ab} = 0$, we recover the non-degenerate result: $E_a^{(1)} = W_{aa}$ and $E_b^{(1)} = W_{bb}$, and the original states $|a\rangle$, $|b\rangle$ are already the good states. This is why non-degenerate perturbation theory works when the perturbation has no off-diagonal elements between degenerate states.

🔴 Warning: The formula above assumes exact degeneracy ($E_a^{(0)} = E_b^{(0)}$). If the states are nearly but not exactly degenerate, and if $|W_{ab}|$ is comparable to or larger than $|E_a^{(0)} - E_b^{(0)}|$, then non-degenerate perturbation theory still fails (the denominator, while not zero, is too small). In this case, one should use the "near-degenerate" variant — essentially diagonalizing the full $2 \times 2$ Hamiltonian including the unperturbed energy difference. See Exercise B.4 for the details.

Formal Derivation

Let us derive this more carefully. The corrected state $|\psi_k\rangle$ solves:

$$(\hat{H}_0 + \lambda \hat{H}') |\psi_k\rangle = E_k |\psi_k\rangle$$

Expand in powers of $\lambda$:

$$|\psi_k\rangle = |\psi_k^{(0)}\rangle + \lambda |\psi_k^{(1)}\rangle + \cdots$$ $$E_k = E^{(0)} + \lambda E_k^{(1)} + \lambda^2 E_k^{(2)} + \cdots$$

At zeroth order, we know $|\psi_k^{(0)}\rangle$ lives in $\mathcal{D}$:

$$|\psi_k^{(0)}\rangle = \sum_{i=1}^{g} c_i^{(k)} |i\rangle$$

At first order, collecting terms proportional to $\lambda$:

$$\hat{H}_0 |\psi_k^{(1)}\rangle + \hat{H}' |\psi_k^{(0)}\rangle = E^{(0)} |\psi_k^{(1)}\rangle + E_k^{(1)} |\psi_k^{(0)}\rangle$$

Project onto any degenerate state $\langle j |$ (where $\langle j | \hat{H}_0 = E^{(0)} \langle j |$):

$$E^{(0)} \langle j | \psi_k^{(1)} \rangle + \langle j | \hat{H}' | \psi_k^{(0)} \rangle = E^{(0)} \langle j | \psi_k^{(1)} \rangle + E_k^{(1)} \langle j | \psi_k^{(0)} \rangle$$

The $\langle j | \psi_k^{(1)} \rangle$ terms cancel:

$$\langle j | \hat{H}' | \psi_k^{(0)} \rangle = E_k^{(1)} \langle j | \psi_k^{(0)} \rangle$$

Substituting the expansion of $|\psi_k^{(0)}\rangle$:

$$\sum_{i=1}^{g} W_{ji} \, c_i^{(k)} = E_k^{(1)} \, c_j^{(k)}$$

This is precisely the eigenvalue equation for the matrix $\mathbf{W}$. The first-order energies $E_k^{(1)}$ are the eigenvalues, and the coefficients $c_i^{(k)}$ define the correct zeroth-order states.

When Degeneracy Is Not Fully Lifted

If the eigenvalues of $\mathbf{W}$ are all distinct, the degeneracy is completely lifted at first order, and we may proceed with standard (non-degenerate) perturbation theory for the higher-order corrections. But what if some eigenvalues of $\mathbf{W}$ coincide? Then we have residual degeneracy — states that remain degenerate even after first-order corrections. In this case, we must go to second order within the remaining degenerate subspace, repeating the diagonalization procedure with the second-order perturbation matrix.

For hydrogen fine structure, it turns out that the degeneracy is not fully lifted at first order: states with the same total angular momentum quantum number $j$ but different $l$ values remain degenerate. This is a deep result related to the specific form of the Coulomb potential and is only fully explained by the Dirac equation.

Good Quantum Numbers: A Shortcut

In practice, there is often a shortcut that avoids explicit matrix diagonalization. If we can find a complete set of operators that commute with both $\hat{H}_0$ and $\hat{H}'$, then the simultaneous eigenstates of these operators are automatically the "good" states — the basis that diagonalizes $\hat{H}'$ within the degenerate subspace.

For the hydrogen fine structure:

• $\hat{H}_0$ (Coulomb) commutes with $\hat{L}^2$, $\hat{L}_z$, $\hat{S}^2$, $\hat{S}_z$. The states $|n, l, m_l, s, m_s\rangle$ (the uncoupled basis) are "good" for $\hat{H}_0$.

• The fine structure perturbation $\hat{H}'_{\text{FS}}$ (dominated by spin-orbit coupling) does not commute with $\hat{L}_z$ or $\hat{S}_z$ individually, but it does commute with $\hat{J}^2$ and $\hat{J}_z$ where $\hat{\mathbf{J}} = \hat{\mathbf{L}} + \hat{\mathbf{S}}$.

Therefore the good quantum numbers for fine structure are $n$, $l$, $j$, and $m_j$. We should work in the coupled basis $|n, l, j, m_j\rangle$ — exactly the basis we constructed in Chapter 14 using Clebsch-Gordan coefficients.

💡 Key Insight: "Good quantum numbers" are not just a mathematical convenience — they encode the physical symmetries that survive the perturbation. The perturbation breaks some symmetries of $\hat{H}_0$ but preserves others. The quantum numbers labeling the surviving symmetries are the good ones.

🔗 Connection: The Clebsch-Gordan coefficients from Chapter 14 connect the uncoupled basis $|l, m_l; s, m_s\rangle$ to the coupled basis $|j, m_j\rangle$. For hydrogen with $s = 1/2$, the relevant coefficients for $j = l \pm 1/2$ are given explicitly in Section 14.4. You will need them throughout this chapter.

18.3 Fine Structure of Hydrogen: Three Corrections

The hydrogen atom, solved exactly in Chapter 5, has energy levels:

$$E_n^{(0)} = -\frac{13.6 \text{ eV}}{n^2}$$

These depend only on the principal quantum number $n$, not on $l$ or $m_l$. The $n^2$-fold orbital degeneracy (and $2n^2$ including spin) is a consequence of the hidden $\mathrm{SO}(4)$ symmetry of the pure Coulomb potential.

In reality, the hydrogen spectrum has much finer structure. Three corrections — all of the same order of magnitude — collectively produce what is called the fine structure:

1. Relativistic correction to the kinetic energy
2. Spin-orbit coupling — the interaction between the electron's spin magnetic moment and the magnetic field seen in its rest frame due to orbital motion
3. The Darwin term — a relativistic correction arising from the "Zitterbewegung" (trembling motion) of the electron

All three corrections are of order $\alpha^2 E_n^{(0)}$, where:

$$\alpha = \frac{e^2}{4\pi\epsilon_0 \hbar c} \approx \frac{1}{137.036}$$

is the fine structure constant — one of the most important dimensionless numbers in all of physics.

📊 By the Numbers: The fine structure constant $\alpha \approx 1/137$ controls the strength of the electromagnetic interaction. For hydrogen: $E_n^{(0)} \sim \alpha^2 m_e c^2 / (2n^2)$. The fine structure corrections are of order $\alpha^4 m_e c^2 \sim \alpha^2 E_n$. For $n = 2$: $E_2^{(0)} = -3.4$ eV, so the fine structure splitting is $\sim \alpha^2 \times 3.4 \text{ eV} \approx 1.8 \times 10^{-4}$ eV $\approx 1.5$ cm$^{-1}$. This is tiny but spectroscopically resolvable — and it is precisely what Michelson observed in the 1890s.

🔵 Historical Note: Arnold Sommerfeld introduced the fine structure constant $\alpha$ in 1916 while extending Bohr's model to include relativistic corrections and elliptical orbits. His formula for the fine structure agreed with experiment to remarkable precision — a lucky success, since his method (the "old quantum theory") was fundamentally flawed. Dirac later rederived the same formula from his relativistic wave equation in 1928. Pauli famously obsessed over the number 137, declaring that a complete physical theory should explain its value. It remains unexplained.

The total fine structure Hamiltonian is:

$$\hat{H}'_{\text{FS}} = \hat{H}'_{\text{rel}} + \hat{H}'_{\text{SO}} + \hat{H}'_{\text{Darwin}}$$

Let us derive each term.

18.4 The Relativistic Correction

Motivation

In special relativity, the kinetic energy of a particle with rest mass $m$ and momentum $p$ is:

$$T = \sqrt{p^2 c^2 + m^2 c^4} - mc^2 = mc^2\left(\sqrt{1 + \frac{p^2}{m^2c^2}} - 1\right)$$

Expanding in powers of $p/mc$ (valid when $v \ll c$):

$$T = \frac{p^2}{2m} - \frac{p^4}{8m^3 c^2} + \cdots$$

The first term is the familiar non-relativistic kinetic energy. The second term is the leading relativistic correction. In the quantum-mechanical version, we replace $p \to \hat{p}$:

$$\hat{H}'_{\text{rel}} = -\frac{\hat{p}^4}{8m_e^3 c^2}$$

Evaluating the Expectation Value

We need $\langle \hat{H}'_{\text{rel}} \rangle = -\frac{1}{8m_e^3 c^2} \langle \hat{p}^4 \rangle$. A direct calculation using momentum-space wave functions is possible but tedious. The elegant approach uses the unperturbed Schrodinger equation:

$$\hat{H}_0 |n, l, m\rangle = E_n^{(0)} |n, l, m\rangle$$

where $\hat{H}_0 = \hat{p}^2/(2m_e) + V(r)$ with $V(r) = -e^2/(4\pi\epsilon_0 r)$. Therefore:

$$\hat{p}^2 = 2m_e(\hat{H}_0 - V)$$

and:

$$\hat{p}^4 = 4m_e^2(\hat{H}_0 - V)^2$$

Taking the expectation value in the state $|n, l, m\rangle$:

$$\langle \hat{p}^4 \rangle = 4m_e^2 \langle (\hat{H}_0 - V)^2 \rangle = 4m_e^2 \left[ (E_n^{(0)})^2 - 2 E_n^{(0)} \langle V \rangle + \langle V^2 \rangle \right]$$

We need two expectation values in hydrogen eigenstates (derived in Chapter 5 and collected here):

$$\left\langle \frac{1}{r} \right\rangle_{nl} = \frac{1}{n^2 a_0}$$

$$\left\langle \frac{1}{r^2} \right\rangle_{nl} = \frac{1}{n^3(l + 1/2) a_0^2}$$

Since $V = -e^2/(4\pi\epsilon_0 r)$, we have $\langle V \rangle = -e^2/(4\pi\epsilon_0) \cdot \langle 1/r \rangle$ and $\langle V^2 \rangle = [e^2/(4\pi\epsilon_0)]^2 \langle 1/r^2 \rangle$.

After careful algebra (using $E_n^{(0)} = -e^2/(8\pi\epsilon_0 n^2 a_0)$ and $a_0 = 4\pi\epsilon_0\hbar^2/(m_e e^2)$), the result is:

$$\boxed{E_{\text{rel}}^{(1)} = -\frac{(E_n^{(0)})^2}{2m_e c^2}\left(\frac{4n}{l + 1/2} - 3\right)}$$

Several features deserve comment:

1. The correction is always negative (since $(E_n^{(0)})^2 > 0$ and the factor in parentheses is positive for all valid $l$). The relativistic correction lowers all energy levels.

2. It depends on $l$ through the factor $1/(l + 1/2)$. States with lower $l$ (more eccentric orbits in the Bohr-Sommerfeld picture) spend more time near the nucleus where the electron moves fastest — and relativistic corrections are largest for the fastest electrons.

3. The correction scales as $\alpha^2 E_n^{(0)}$ — we can write it as:

$$E_{\text{rel}}^{(1)} = -\frac{\alpha^2}{4n^2} E_n^{(0)} \left(\frac{4n}{l + 1/2} - 3\right)$$

confirming that fine structure corrections are order $\alpha^2$ relative to the gross structure.

⚠️ Common Misconception: The operator $\hat{p}^4$ is not the same as $(\hat{p}^2)^2$ acting naively. In coordinate representation, $\hat{p}^2 = -\hbar^2\nabla^2$, so $\hat{p}^4 = \hbar^4\nabla^4$. The expectation value must be evaluated carefully — the trick of using $\hat{p}^2 = 2m(\hat{H}_0 - V)$ avoids all the subtleties.

✅ Checkpoint: Verify that for $n = 1$, $l = 0$: $E_{\text{rel}}^{(1)} = -(E_1^{(0)})^2 \cdot 5/(2m_e c^2)$. Using $E_1^{(0)} = -13.6$ eV and $m_e c^2 = 511{,}000$ eV, this gives $E_{\text{rel}}^{(1)} \approx -9.06 \times 10^{-4}$ eV. This is the right order of magnitude for fine structure.

Worked Example: Relativistic Correction for $n = 2$

For the $n = 2$ level ($E_2^{(0)} = -3.4$ eV):

Case $l = 0$ ($2s$ state): $$E_{\text{rel}}^{(1)} = -\frac{(-3.4)^2}{2 \times 511{,}000}\left(\frac{4 \times 2}{0.5} - 3\right) = -\frac{11.56}{1.022 \times 10^6}(16 - 3) = -1.47 \times 10^{-4} \text{ eV}$$

Case $l = 1$ ($2p$ state): $$E_{\text{rel}}^{(1)} = -\frac{(-3.4)^2}{2 \times 511{,}000}\left(\frac{4 \times 2}{1.5} - 3\right) = -\frac{11.56}{1.022 \times 10^6}(5.33 - 3) = -2.64 \times 10^{-5} \text{ eV}$$

Notice that the $2s$ correction is about 5.6 times larger than the $2p$ correction. This makes physical sense: $s$-states ($l = 0$) have significant probability density near the nucleus where the electron moves fastest and relativistic effects are most important. $p$-states ($l = 1$) are kept away from the nucleus by the centrifugal barrier, so they sample lower-velocity regions of the orbit.

In the Bohr-Sommerfeld picture (not rigorous but useful for intuition), $s$-orbits correspond to highly eccentric ellipses that plunge close to the nucleus, while higher-$l$ orbits are more circular. The plunging orbits achieve the highest speeds and therefore receive the largest relativistic corrections. This correlation between $l$ and orbit shape persists in the full quantum treatment through the expectation values $\langle 1/r \rangle$ and $\langle 1/r^2 \rangle$.

18.5 Spin-Orbit Coupling

The Physical Picture

An electron orbiting the nucleus "sees," in its own rest frame, a nucleus circling around it. This orbiting positive charge creates a magnetic field at the electron's position. The electron's spin magnetic moment interacts with this field, creating an energy shift that depends on the relative orientation of the orbital angular momentum $\hat{\mathbf{L}}$ and the spin $\hat{\mathbf{S}}$.

In the electron's rest frame, the magnetic field due to the nucleus is:

$$\mathbf{B} = -\frac{\mathbf{v} \times \mathbf{E}}{c^2}$$

where $\mathbf{v}$ is the electron's velocity and $\mathbf{E}$ is the electric field from the nucleus. For a central potential $V(r)$:

$$\mathbf{E} = -\frac{1}{e}\frac{dV}{dr}\hat{\mathbf{r}}$$

The interaction Hamiltonian between the electron's magnetic moment $\hat{\boldsymbol{\mu}}_s = -g_s \mu_B \hat{\mathbf{S}}/\hbar$ (with $g_s \approx 2$) and this field is:

$$\hat{H}'_{\text{SO}} = -\hat{\boldsymbol{\mu}}_s \cdot \mathbf{B}$$

However, there is a crucial subtlety. The electron's rest frame is not inertial — it is accelerating. The transformation from the lab frame to the electron's frame introduces an additional precession called Thomas precession, which reduces the spin-orbit coupling by exactly a factor of 2. Including this factor:

$$\hat{H}'_{\text{SO}} = \frac{1}{2m_e^2 c^2} \frac{1}{r}\frac{dV}{dr} \hat{\mathbf{L}} \cdot \hat{\mathbf{S}}$$

For the Coulomb potential $V(r) = -e^2/(4\pi\epsilon_0 r)$:

$$\frac{1}{r}\frac{dV}{dr} = \frac{e^2}{4\pi\epsilon_0 r^3}$$

Therefore:

$$\hat{H}'_{\text{SO}} = \frac{e^2}{8\pi\epsilon_0 m_e^2 c^2} \frac{\hat{\mathbf{L}} \cdot \hat{\mathbf{S}}}{r^3}$$

🔵 Historical Note: Llewellyn Thomas resolved the factor-of-two discrepancy in 1926. Before Thomas, the spin-orbit calculation gave twice the experimentally observed splitting. Thomas showed that the non-inertial frame transformation contributes an additional precession (Thomas precession) that halves the result. Reputedly, when Thomas presented his result, Bohr's response was: "I never thought that a simple kinematic argument could resolve this problem."

Evaluating $\langle \hat{\mathbf{L}} \cdot \hat{\mathbf{S}} \rangle$

This is where the coupled basis $|n, l, j, m_j\rangle$ becomes essential. Since $\hat{\mathbf{J}} = \hat{\mathbf{L}} + \hat{\mathbf{S}}$:

$$\hat{\mathbf{J}}^2 = \hat{\mathbf{L}}^2 + \hat{\mathbf{S}}^2 + 2\hat{\mathbf{L}} \cdot \hat{\mathbf{S}}$$

Solving for $\hat{\mathbf{L}} \cdot \hat{\mathbf{S}}$:

$$\hat{\mathbf{L}} \cdot \hat{\mathbf{S}} = \frac{1}{2}\left(\hat{\mathbf{J}}^2 - \hat{\mathbf{L}}^2 - \hat{\mathbf{S}}^2\right)$$

In the state $|n, l, j, m_j\rangle$, the expectation value is:

$$\langle \hat{\mathbf{L}} \cdot \hat{\mathbf{S}} \rangle = \frac{\hbar^2}{2}\left[j(j+1) - l(l+1) - s(s+1)\right]$$

with $s = 1/2$ for an electron. For $j = l + 1/2$:

$$\langle \hat{\mathbf{L}} \cdot \hat{\mathbf{S}} \rangle = \frac{\hbar^2}{2} l$$

For $j = l - 1/2$ (with $l \geq 1$):

$$\langle \hat{\mathbf{L}} \cdot \hat{\mathbf{S}} \rangle = -\frac{\hbar^2}{2}(l+1)$$

We also need:

$$\left\langle \frac{1}{r^3} \right\rangle_{nl} = \frac{1}{n^3 l(l+1/2)(l+1) a_0^3}$$

which is valid only for $l \geq 1$. For $l = 0$, $\langle \hat{\mathbf{L}} \cdot \hat{\mathbf{S}} \rangle = 0$, so spin-orbit coupling vanishes for $s$-states (as it should — an electron with no orbital angular momentum has no orbital magnetic field to interact with).

Combining everything, the spin-orbit energy correction is:

$$\boxed{E_{\text{SO}}^{(1)} = \frac{(E_n^{(0)})^2}{m_e c^2} \cdot \frac{n[j(j+1) - l(l+1) - 3/4]}{l(l + 1/2)(l+1)}} \quad (l \geq 1)$$

For $l = 0$: $E_{\text{SO}}^{(1)} = 0$.

Sign and Magnitude

For $j = l + 1/2$ (spin and orbital angular momenta aligned): the energy is raised (less negative).

For $j = l - 1/2$ (spin and orbital angular momenta anti-aligned): the energy is lowered (more negative).

The energy difference between the two $j$ levels for a given $n$ and $l$ is:

$$\Delta E_{\text{SO}} = \frac{(E_n^{(0)})^2}{m_e c^2} \cdot \frac{n(2l+1)}{l(l + 1/2)(l+1)} \cdot \frac{1}{2}$$

This splitting is directly observable spectroscopically — it is what gives the sodium D lines their famous doublet structure.

🧪 Experiment: The sodium D lines at 589.0 nm and 589.6 nm arise from the $3p \to 3s$ transition. The $3p$ level splits into $3p_{3/2}$ ($j = 3/2$) and $3p_{1/2}$ ($j = 1/2$) due to spin-orbit coupling, while the $3s_{1/2}$ level has no splitting. The measured separation of $\sim 17$ cm$^{-1}$ matches the spin-orbit prediction beautifully. This doublet, visible even with a modest spectrometer, was one of the earliest confirmations of electron spin.

Worked Example: Spin-Orbit Coupling for $n = 2$

For $n = 2$, $l = 1$ (the $2p$ states), with $E_2^{(0)} = -3.4$ eV:

Case $j = 3/2$ ($2p_{3/2}$): $$\langle \hat{\mathbf{L}} \cdot \hat{\mathbf{S}} \rangle = \frac{\hbar^2}{2}[3/2 \cdot 5/2 - 1 \cdot 2 - 1/2 \cdot 3/2] = \frac{\hbar^2}{2}\left[\frac{15}{4} - 2 - \frac{3}{4}\right] = \frac{\hbar^2}{2} \cdot 1 = \frac{\hbar^2}{2}$$

This equals $+\frac{\hbar^2}{2} l = +\frac{\hbar^2}{2}$ (as expected for $j = l + 1/2$).

Case $j = 1/2$ ($2p_{1/2}$): $$\langle \hat{\mathbf{L}} \cdot \hat{\mathbf{S}} \rangle = \frac{\hbar^2}{2}[1/2 \cdot 3/2 - 1 \cdot 2 - 1/2 \cdot 3/2] = \frac{\hbar^2}{2}\left[\frac{3}{4} - 2 - \frac{3}{4}\right] = -\hbar^2$$

This equals $-\frac{\hbar^2}{2}(l+1) = -\hbar^2$ (as expected for $j = l - 1/2$).

Computing the full spin-orbit correction:

$$E_{\text{SO}}^{(1)}(2p_{3/2}) = \frac{(3.4)^2}{511{,}000} \times \frac{2 \times 1}{1 \times 1.5 \times 2} = \frac{11.56}{511{,}000} \times \frac{2}{3} = 1.51 \times 10^{-5} \text{ eV}$$

$$E_{\text{SO}}^{(1)}(2p_{1/2}) = \frac{(3.4)^2}{511{,}000} \times \frac{2 \times (-2)}{1 \times 1.5 \times 2} = \frac{11.56}{511{,}000} \times \left(-\frac{4}{3}\right) = -3.01 \times 10^{-5} \text{ eV}$$

The $j = 3/2$ state (spin and orbit aligned) is raised by $1.51 \times 10^{-5}$ eV, while the $j = 1/2$ state (anti-aligned) is lowered by $3.01 \times 10^{-5}$ eV. The splitting between them is $4.52 \times 10^{-5}$ eV. Note the ratio of the shifts: $|E_{\text{SO}}(j=1/2)| / |E_{\text{SO}}(j=3/2)| = 2$, which is the ratio $(l+1)/l = 2/1$. This ratio is a general property of spin-orbit coupling and is known as the interval rule (or Lande interval rule): the splitting between adjacent $j$ levels is proportional to the larger $j$ value.

💡 Key Insight: The interval rule states that the energy interval between levels $j$ and $j - 1$ is proportional to $j$. Deviations from the interval rule in multi-electron atoms indicate that the spin-orbit coupling is no longer well-described by a single-electron $\hat{\mathbf{L}} \cdot \hat{\mathbf{S}}$ interaction — signaling the presence of configuration mixing or strong electron-electron correlations.

18.6 The Darwin Term

Origin: Zitterbewegung

The Darwin term is the strangest of the three fine structure corrections. It arises from a subtle relativistic effect: the Dirac equation predicts that the electron undergoes rapid oscillatory motion called Zitterbewegung ("trembling motion") with an amplitude of order the Compton wavelength $\lambda_C = \hbar/(m_e c) \approx 3.86 \times 10^{-13}$ m.

This trembling effectively "smears out" the electron over a region of size $\sim \lambda_C$ around its mean position. Instead of feeling the potential $V(r)$ at a single point, the electron samples $V$ over a small volume. The correction is proportional to $\nabla^2 V$:

$$\hat{H}'_{\text{Darwin}} = \frac{\hbar^2}{8m_e^2 c^2} \nabla^2 V = \frac{\hbar^2 e^2}{8m_e^2 c^2 \cdot 4\pi\epsilon_0} \nabla^2\left(\frac{1}{r}\right)$$

For the Coulomb potential:

$$\nabla^2 \left(\frac{1}{r}\right) = -4\pi \delta^3(\mathbf{r})$$

(This is a result from electrostatics — the Laplacian of $1/r$ is a delta function.) Therefore:

$$\hat{H}'_{\text{Darwin}} = \frac{\pi\hbar^2 e^2}{2m_e^2 c^2 \cdot 4\pi\epsilon_0} \delta^3(\mathbf{r})$$

Evaluating the Expectation Value

The delta function makes the expectation value trivial — it selects the probability density at the origin:

$$E_{\text{Darwin}}^{(1)} = \frac{\pi\hbar^2 e^2}{2m_e^2 c^2 \cdot 4\pi\epsilon_0} |\psi_{nlm}(0)|^2$$

Now, $|\psi_{nlm}(0)|^2 \neq 0$ only for $l = 0$ states (all other states have $\psi_{nlm}(0) = 0$ due to the centrifugal barrier). For $l = 0$:

$$|\psi_{n00}(0)|^2 = \frac{1}{\pi n^3 a_0^3}$$

The result:

$$\boxed{E_{\text{Darwin}}^{(1)} = \frac{2(E_n^{(0)})^2}{m_e c^2 n} \quad (l = 0 \text{ only})}$$

For $l \neq 0$: $E_{\text{Darwin}}^{(1)} = 0$.

💡 Key Insight: The Darwin term affects only $s$-states ($l = 0$), while spin-orbit coupling affects only states with $l \geq 1$. They are complementary corrections. This is not a coincidence — both arise from the Dirac equation, and together they ensure that the total fine structure energy depends only on $j$ (not on $l$ separately), as the Dirac equation predicts.

Worked Example: Darwin Correction for $n = 2$

For $n = 2$, $l = 0$ ($2s$ state):

$$E_{\text{Darwin}}^{(1)} = \frac{2(E_2^{(0)})^2}{m_e c^2 \cdot 2} = \frac{(-3.4)^2}{511{,}000} = 2.26 \times 10^{-5} \text{ eV}$$

This is positive — the Darwin correction raises the energy of $s$-states. Physically, the Zitterbewegung smearing causes the electron to sample a slightly less negative average potential (the potential is less negative when averaged over a small sphere around the nucleus than at the nucleus itself). This explains why the correction is positive.

For $n = 2$, $l = 1$ ($2p$ states): $E_{\text{Darwin}}^{(1)} = 0$ (the $2p$ wave function vanishes at the origin).

A Satisfying Conspiracy

It is remarkable that the relativistic correction, spin-orbit coupling, and the Darwin term — three physically distinct effects — combine to produce a result that depends only on $n$ and $j$. Let us verify this explicitly for the $n = 2$ states.

For $2s_{1/2}$ ($l = 0$, $j = 1/2$): - Relativistic: $-1.47 \times 10^{-4}$ eV - Spin-orbit: $0$ eV (because $l = 0$) - Darwin: $+2.26 \times 10^{-5}$ eV - Total: $-1.24 \times 10^{-4}$ eV

For $2p_{1/2}$ ($l = 1$, $j = 1/2$): - Relativistic: $-2.64 \times 10^{-5}$ eV - Spin-orbit: $-3.01 \times 10^{-5}$ eV - Darwin: $0$ eV (because $l = 1$) - Total: $-5.65 \times 10^{-5}$ eV

Wait — these do not appear to be equal! But they should be, because both states have $j = 1/2$. The issue is rounding in our intermediate calculations. Using the exact formulas:

$$E_{\text{FS}}(n=2, j=1/2) = -\frac{(E_2^{(0)})^2}{2m_ec^2}\left(\frac{8}{1} - 3\right) = -\frac{(E_2^{(0)})^2}{2m_ec^2} \times 5$$

This is exactly the same for both states, independent of $l$. The individual corrections differ — $2s_{1/2}$ gets a large relativistic correction offset by the Darwin term, while $2p_{1/2}$ gets a moderate relativistic correction enhanced by spin-orbit — but they sum to the same total. This is the "miraculous" cancellation.

⚠️ Common Misconception: Students sometimes think that the $l$-independence of the total fine structure is approximate. It is not — it is exact to order $\alpha^4$. The cancellation occurs because all three corrections arise from expanding a single relativistic effect (the Dirac equation), and the expansion artificially separates terms that the full theory naturally combines. When we encounter the Dirac equation in Chapter 29, the $j$-only dependence will emerge automatically without any need for separate calculations.

18.7 The Combined Fine Structure: A Miraculous Simplification

Adding all three contributions (carefully accounting for the $l = 0$ and $l \geq 1$ cases separately), the total fine structure correction is:

$$\boxed{E_{\text{FS}}^{(1)} = -\frac{(E_n^{(0)})^2}{2m_e c^2}\left(\frac{4n}{j + 1/2} - 3\right)}$$

This is the fine structure formula. Notice:

1. It depends on $n$ and $j$ only. The orbital quantum number $l$ has completely dropped out! States with the same $n$ and $j$ but different $l$ remain degenerate. For example, the $2s_{1/2}$ ($l = 0$, $j = 1/2$) and $2p_{1/2}$ ($l = 1$, $j = 1/2$) have exactly the same fine structure correction.

2. It can be rewritten as:

$$E_{\text{FS}}^{(1)} = \frac{\alpha^2}{n^2} |E_n^{(0)}| \left(\frac{3}{4n} - \frac{1}{j + 1/2}\right)$$

3. For a given $n$, states with smaller $j$ are lower in energy (more negative correction). The state $j = 1/2$ lies lowest, $j = 3/2$ next, and so on up to $j = n - 1/2$.

The $n = 2$ Fine Structure

Let us work out the $n = 2$ level in detail. The unperturbed energy is $E_2^{(0)} = -3.4$ eV. The eight degenerate states (including spin) are:

Fine structure corrections:

• $j = 1/2$: $E_{\text{FS}}^{(1)} = -\frac{(3.4)^2}{2 \times 511{,}000}(8 - 3) = -\frac{11.56}{1{,}022{,}000}(5) \approx -5.66 \times 10^{-5}$ eV

• $j = 3/2$: $E_{\text{FS}}^{(1)} = -\frac{(3.4)^2}{2 \times 511{,}000}(4 - 3) = -\frac{11.56}{1{,}022{,}000}(1) \approx -1.13 \times 10^{-5}$ eV

The splitting between $j = 1/2$ and $j = 3/2$ is:

$$\Delta E_{\text{FS}}(n=2) = 4.53 \times 10^{-5} \text{ eV} \approx 0.365 \text{ cm}^{-1}$$

This corresponds to a frequency of $\sim 10.9$ GHz — easily measurable by microwave spectroscopy.

📊 By the Numbers: The measured fine structure splitting of the $n = 2$ level in hydrogen is $0.365$ cm$^{-1}$, in excellent agreement with our formula. The Lamb shift (discovered in 1947, explained by QED) further splits the $2s_{1/2}$ and $2p_{1/2}$ levels by about $0.035$ cm$^{-1}$ — an effect our perturbative treatment cannot capture but the full quantum electrodynamic calculation does. This was one of the triumphs that established QED as the correct theory of the electromagnetic interaction.

Comparison with the Dirac Equation

The result we derived perturbatively — that the fine structure depends only on $n$ and $j$ — is not a coincidence. The Dirac equation (Chapter 29) provides the exact energy levels of hydrogen:

$$E_{nj} = m_e c^2 \left[ 1 + \left(\frac{\alpha}{n - (j + 1/2) + \sqrt{(j+1/2)^2 - \alpha^2}}\right)^2 \right]^{-1/2}$$

Expanding this to order $\alpha^4$:

$$E_{nj} \approx m_e c^2 - \frac{\alpha^2 m_e c^2}{2n^2} - \frac{\alpha^4 m_e c^2}{2n^4}\left(\frac{n}{j + 1/2} - \frac{3}{4}\right) + \cdots$$

The first term is the rest energy, the second is the Bohr energy $E_n^{(0)}$, and the third is exactly our fine structure correction $E_{\text{FS}}^{(1)}$. Our perturbative result is exact to order $\alpha^4$.

💡 Key Insight: The fact that the fine structure depends on $j$ alone (not on $l$) is a deep consequence of the full Lorentz symmetry of the Dirac equation. The three corrections we calculated separately (relativistic, spin-orbit, Darwin) are artificially separated pieces of a single relativistic effect. The Dirac equation treats them all at once and reveals the underlying simplicity.

⚖️ Interpretation: The $l$-degeneracy of the fine structure (i.e., $2s_{1/2}$ and $2p_{1/2}$ having the same energy) is sometimes called an "accidental degeneracy." But in the context of the Dirac equation, it is not accidental at all — it is protected by the symmetry of the exact relativistic Coulomb problem. The Lamb shift, which lifts this degeneracy, arises from quantum electrodynamic effects (vacuum fluctuations, self-energy) that are not present in the Dirac equation for a single electron in an external potential.

Energy Level Diagram

The fine structure splits each $n$ level into $n$ sub-levels labeled by $j$:

For $n = 1$: One level only, $1s_{1/2}$ ($j = 1/2$).

For $n = 2$: Two levels — $2s_{1/2}, 2p_{1/2}$ ($j = 1/2$, still degenerate) and $2p_{3/2}$ ($j = 3/2$).

For $n = 3$: Three levels — $3s_{1/2}, 3p_{1/2}$ ($j = 1/2$); $3p_{3/2}, 3d_{3/2}$ ($j = 3/2$); $3d_{5/2}$ ($j = 5/2$).

The pattern is that states with the same $j$ but different $l$ remain degenerate. Each $j$-level has degeneracy $2j + 1$ (from the $m_j$ values).

18.8 Hyperfine Structure and the 21 cm Line

Beyond Fine Structure: The Nuclear Magnetic Moment

The fine structure corrections account for the electron's relativistic motion and its spin-orbit interaction. But we have been treating the nucleus as a mere point charge. In reality, the proton has its own magnetic moment:

$$\hat{\boldsymbol{\mu}}_p = g_p \frac{e}{2m_p} \hat{\mathbf{I}}$$

where $\hat{\mathbf{I}}$ is the proton spin ($I = 1/2$), $m_p$ is the proton mass, and $g_p \approx 5.586$ is the proton's anomalous $g$-factor. Because $m_p \approx 1836 m_e$, the proton's magnetic moment is roughly 1000 times smaller than the electron's. The resulting energy splittings are correspondingly tiny — hence the name hyperfine structure.

The Contact Interaction

For $s$-states ($l = 0$), the dominant hyperfine interaction is the Fermi contact interaction — the direct magnetic coupling between the electron and proton when the electron probability density is nonzero at the nucleus:

$$\hat{H}'_{\text{HF}} = \frac{2\mu_0}{3} g_p \mu_N \frac{\mu_B}{\hbar^2} |\psi_{n00}(0)|^2 \hat{\mathbf{I}} \cdot \hat{\mathbf{S}}$$

where $\mu_N = e\hbar/(2m_p)$ is the nuclear magneton and $\mu_B = e\hbar/(2m_e)$ is the Bohr magneton.

The operator $\hat{\mathbf{I}} \cdot \hat{\mathbf{S}}$ has eigenvalues determined by the total spin $\hat{\mathbf{F}} = \hat{\mathbf{I}} + \hat{\mathbf{S}}$:

$$\hat{\mathbf{I}} \cdot \hat{\mathbf{S}} = \frac{1}{2}(\hat{\mathbf{F}}^2 - \hat{\mathbf{I}}^2 - \hat{\mathbf{S}}^2)$$

Since $I = 1/2$ and $S = 1/2$, the total spin quantum number is $F = 0$ (singlet) or $F = 1$ (triplet):

$$\langle \hat{\mathbf{I}} \cdot \hat{\mathbf{S}} \rangle = \frac{\hbar^2}{2}[F(F+1) - I(I+1) - S(S+1)] = \begin{cases} +\hbar^2/4 & F = 1 \text{ (triplet)} \\ -3\hbar^2/4 & F = 0 \text{ (singlet)} \end{cases}$$

The 21 cm Line

For the ground state ($n = 1$, $l = 0$), the hyperfine splitting between $F = 1$ and $F = 0$ is:

$$\Delta E_{\text{HF}} = \frac{4}{3} g_p \alpha^2 \frac{m_e}{m_p} |E_1^{(0)}| = 5.88 \times 10^{-6} \text{ eV}$$

This corresponds to a frequency:

$$\nu = \frac{\Delta E_{\text{HF}}}{h} = 1420.405 \text{ MHz}$$

and a wavelength:

$$\lambda = \frac{c}{\nu} = 21.106 \text{ cm}$$

This is the famous 21 cm line (or hydrogen line) — one of the most important spectral lines in all of astrophysics.

📊 By the Numbers: The 21 cm transition frequency has been measured to extraordinary precision: $\nu = 1{,}420{,}405{,}751.768 \pm 0.001$ Hz. This is one of the most precisely known quantities in physics. The transition is magnetic dipole (M1), making it extremely slow: the spontaneous emission lifetime is $\tau \approx 1.1 \times 10^7$ years. This means an isolated hydrogen atom in the $F = 1$ state will wait, on average, eleven million years before spontaneously flipping to $F = 0$ and emitting a 21 cm photon. Despite this tiny transition rate, the sheer abundance of hydrogen in the universe makes the 21 cm line readily observable.

🧪 Experiment: The 21 cm line was predicted by Hendrik van de Hulst in 1944 and first detected by Harold Ewen and Edward Purcell at Harvard in 1951 using a horn antenna pointed out of a window in the physics building. Today, 21 cm observations are used to map the distribution of neutral hydrogen throughout the Milky Way and other galaxies, to measure galactic rotation curves (providing evidence for dark matter), and to probe the epoch of reionization in the early universe. The Square Kilometre Array (SKA) and its precursors are designed to detect 21 cm emission from the cosmic dawn.

Hierarchy of Energy Scales

Let us pause and appreciate the hierarchy of corrections to hydrogen's energy levels:

Each level of structure is roughly two orders of magnitude smaller than the previous one. The fact that hydrogen's spectrum has this hierarchical structure — with each level revealing new physics — is what makes it the "Rosetta Stone" of modern physics.

18.9 The Zeeman Effect: Weak, Strong, and Intermediate Fields

The Perturbation

When a hydrogen atom is placed in an external magnetic field $\mathbf{B} = B\hat{\mathbf{z}}$, the Hamiltonian acquires an additional term:

$$\hat{H}'_Z = -(\hat{\boldsymbol{\mu}}_L + \hat{\boldsymbol{\mu}}_S) \cdot \mathbf{B}$$

where $\hat{\boldsymbol{\mu}}_L = -\mu_B \hat{\mathbf{L}}/\hbar$ and $\hat{\boldsymbol{\mu}}_S = -g_s \mu_B \hat{\mathbf{S}}/\hbar$ (with $g_s \approx 2$) are the orbital and spin magnetic moments. The Bohr magneton is $\mu_B = e\hbar/(2m_e) = 5.788 \times 10^{-5}$ eV/T. Therefore:

$$\hat{H}'_Z = \frac{\mu_B}{\hbar} B(\hat{L}_z + 2\hat{S}_z) = \frac{\mu_B B}{\hbar}(\hat{J}_z + \hat{S}_z)$$

The behavior depends critically on whether $\hat{H}'_Z$ is large or small compared to the fine structure perturbation $\hat{H}'_{\text{FS}}$.

The Weak-Field (Anomalous) Zeeman Effect

When $\mu_B B \ll \Delta E_{\text{FS}}$ (typically $B \lesssim 0.3$ T for low-$n$ states), the fine structure splitting dominates. The atom is well-described by the coupled basis $|n, l, j, m_j\rangle$, and $\hat{H}'_Z$ is treated as a small perturbation on top of the fine structure.

In this regime, $j$ and $m_j$ are good quantum numbers. We need:

$$E_Z^{(1)} = \langle n, l, j, m_j | \hat{H}'_Z | n, l, j, m_j \rangle = \frac{\mu_B B}{\hbar} \langle \hat{J}_z + \hat{S}_z \rangle$$

Now $\langle \hat{J}_z \rangle = m_j \hbar$ is trivial. For $\langle \hat{S}_z \rangle$, we use the projection theorem (a consequence of the Wigner-Eckart theorem): for any vector operator $\hat{\mathbf{V}}$ within a subspace of fixed $j$:

$$\langle j, m_j | \hat{V}_z | j, m_j \rangle = \frac{\langle j, m_j | \hat{\mathbf{V}} \cdot \hat{\mathbf{J}} | j, m_j \rangle}{j(j+1)\hbar^2} m_j \hbar$$

Applying this to $\hat{\mathbf{S}}$:

$$\langle \hat{S}_z \rangle = \frac{\langle \hat{\mathbf{S}} \cdot \hat{\mathbf{J}} \rangle}{j(j+1)\hbar^2} m_j \hbar$$

Using $\hat{\mathbf{S}} \cdot \hat{\mathbf{J}} = \hat{\mathbf{S}} \cdot (\hat{\mathbf{L}} + \hat{\mathbf{S}}) = \hat{\mathbf{L}} \cdot \hat{\mathbf{S}} + \hat{S}^2$ and the relation $\hat{\mathbf{L}} \cdot \hat{\mathbf{S}} = \frac{1}{2}(\hat{J}^2 - \hat{L}^2 - \hat{S}^2)$:

$$\langle \hat{\mathbf{S}} \cdot \hat{\mathbf{J}} \rangle = \frac{\hbar^2}{2}[j(j+1) - l(l+1) + s(s+1)]$$

Therefore:

$$\langle \hat{S}_z \rangle = \frac{j(j+1) - l(l+1) + s(s+1)}{2j(j+1)} m_j \hbar$$

The total Zeeman energy is:

$$\boxed{E_Z^{(1)} = g_j m_j \mu_B B}$$

where the Lande $g$-factor is:

$$\boxed{g_j = 1 + \frac{j(j+1) - l(l+1) + s(s+1)}{2j(j+1)}}$$

For $s = 1/2$:

$$g_j = 1 + \frac{j(j+1) - l(l+1) + 3/4}{2j(j+1)}$$

Some specific values:

The Zeeman energy is proportional to $m_j$, so each $j$-level splits into $2j + 1$ equally spaced sublevels with spacing $g_j \mu_B B$.

⚠️ Common Misconception: Students sometimes confuse the "anomalous" Zeeman effect with an effect that violates theory. It is called "anomalous" for purely historical reasons: before spin was discovered, the "normal" Zeeman effect (splitting into three lines) was the expected pattern. The more complex splitting observed for most spectral lines — now understood as due to spin and the $g_j$-factor — was called "anomalous" simply because it was unexplained at the time. There is nothing anomalous about it.

Worked Example: Zeeman Effect on the Sodium D Lines

The sodium D lines are among the most-studied spectral lines in physics. They arise from the $3p \to 3s$ transition in sodium. Let us work out their Zeeman splitting in the weak-field regime.

The $D_1$ line ($3p_{1/2} \to 3s_{1/2}$):

The upper state $3p_{1/2}$ has $g_j = 2/3$ and $m_j = \pm 1/2$. The lower state $3s_{1/2}$ has $g_j = 2$ and $m_j = \pm 1/2$.

The transition energies (relative to the zero-field line center) are:

$$\Delta E = (g_{j,\text{upper}} m_{j,\text{upper}} - g_{j,\text{lower}} m_{j,\text{lower}}) \mu_B B$$

Applying the selection rule $\Delta m_j = 0, \pm 1$:

So the $D_1$ line splits into four components at $\Delta E / \mu_B B = -4/3, -2/3, +2/3, +4/3$. In a field of $B = 0.5$ T, the splitting is $g_j m_j \mu_B B \sim 10^{-5}$ eV, easily resolved by a Fabry-Perot interferometer.

The $D_2$ line ($3p_{3/2} \to 3s_{1/2}$) splits into six components (you should verify this using $g_j = 4/3$ for the $3p_{3/2}$ state). The pattern is more complex, and the unequal spacing of the components is what historically made the "anomalous" Zeeman effect so puzzling.

🔵 Historical Note: Pieter Zeeman discovered the magnetic splitting of spectral lines in 1896, earning the 1902 Nobel Prize. Hendrik Lorentz provided the classical explanation (the "normal" Zeeman effect), predicting three equally spaced lines. But most spectral lines showed more complex patterns. The full explanation came only after Goudsmit and Uhlenbeck proposed electron spin in 1925, and Lande introduced the $g$-factor. The anomalous Zeeman effect was the strongest early evidence that something was fundamentally missing from classical and early quantum physics.

The Strong-Field (Paschen-Back) Effect

When $\mu_B B \gg \Delta E_{\text{FS}}$ (typically $B \gtrsim 3$ T), the magnetic field dominates over the fine structure. The atom "forgets" about spin-orbit coupling: $\hat{\mathbf{L}}$ and $\hat{\mathbf{S}}$ decouple and precess independently around $\mathbf{B}$. The good quantum numbers are $m_l$ and $m_s$ (not $j$ and $m_j$).

In this regime, we first apply $\hat{H}'_Z$ (the dominant perturbation), then treat $\hat{H}'_{\text{FS}}$ as a correction:

$$E^{(1)}_Z = \mu_B B(m_l + 2m_s)$$

The fine structure correction in the uncoupled basis is:

$$E^{(1)}_{\text{FS}} = -\frac{(E_n^{(0)})^2}{2m_e c^2}\left(\frac{4n}{l + 1/2} - 3 + \frac{n \cdot m_l m_s}{l(l+1/2)(l+1)} \cdot 2\right)$$

The last term arises from the spin-orbit contribution evaluated in the uncoupled basis: $\langle \hat{\mathbf{L}} \cdot \hat{\mathbf{S}} \rangle_{m_l, m_s} = m_l m_s \hbar^2$ (since $\hat{L}_z \hat{S}_z$ is the only surviving term when $m_l$ and $m_s$ are good quantum numbers, and the perpendicular components average to zero).

In the Paschen-Back regime, the primary splitting pattern is $(m_l + 2m_s)$ — a much simpler structure than the anomalous Zeeman effect. For a transition where $\Delta m_l = 0, \pm 1$ and $\Delta m_s = 0$, the observed pattern collapses to exactly three lines: the "normal" Zeeman triplet. This is ironic: the strong-field limit gives the "normal" pattern, while the weak-field limit gives the "anomalous" one.

The Intermediate-Field Regime

When $\mu_B B \sim \Delta E_{\text{FS}}$, neither approximation works. We must diagonalize the combined Hamiltonian:

$$\hat{H} = \hat{H}_0 + \hat{H}'_{\text{FS}} + \hat{H}'_Z$$

within each $n$-manifold. This requires constructing and diagonalizing a $(2n^2) \times (2n^2)$ matrix.

For the $n = 2$ manifold (8 states), the problem simplifies because $m_j$ is always a good quantum number (the total Hamiltonian commutes with $\hat{J}_z$), so states with different $m_j$ do not mix. The largest submatrix to diagonalize is $2 \times 2$ (for $m_j = \pm 1/2$, where $s_{1/2}$ and $p_{1/2}$ states can mix, or $p_{1/2}$ and $p_{3/2}$ states).

For the ground state ($n = 1$), the intermediate-field behavior of the hyperfine structure is described by the Breit-Rabi formula:

$$E(F, m_F) = -\frac{\Delta E_{\text{HF}}}{4} + g_I m_F \mu_N B \pm \frac{\Delta E_{\text{HF}}}{2}\sqrt{1 + \frac{4m_F}{2I+1}x + x^2}$$

where $x = (g_j - g_I) \mu_B B / \Delta E_{\text{HF}}$ and the $\pm$ corresponds to $F = I + 1/2$ and $F = I - 1/2$ (for $m_F$ values shared by both $F$ manifolds). For extreme values of $m_F$ (belonging to only one $F$), the formula reduces to the weak-field or strong-field result as appropriate.

The Breit-Rabi diagram — a plot of energy vs. magnetic field strength — reveals the full richness of the problem: levels that are degenerate at zero field split, cross, and rearrange as $B$ increases. It is one of the most beautiful and instructive diagrams in all of atomic physics.

📊 By the Numbers: The critical magnetic field where the Zeeman energy equals the $n = 2$ fine structure splitting is $B_{\text{crit}} \sim \Delta E_{\text{FS}}/\mu_B \approx 4.5 \times 10^{-5} \text{ eV} / 5.8 \times 10^{-5} \text{ eV/T} \approx 0.8$ T. This is easily achievable in the laboratory, making the intermediate-field regime experimentally accessible. For the ground-state hyperfine splitting, the critical field is much smaller: $B_{\text{crit}} \sim 0.05$ T.

Worked Example: $n = 2$ in the Intermediate Regime

To illustrate the intermediate field, consider the $n = 2$ hydrogen states in a magnetic field of $B = 1$ T. At this field strength, $\mu_B B \approx 5.79 \times 10^{-5}$ eV, which is comparable to the $n = 2$ fine structure splitting of $\Delta E_{\text{FS}} \approx 4.5 \times 10^{-5}$ eV. We are squarely in the intermediate regime.

The key simplification is that $m_j$ (the projection of total angular momentum on the field axis) remains a good quantum number in all regimes, because the total Hamiltonian $\hat{H}_0 + \hat{H}'_{\text{FS}} + \hat{H}'_Z$ commutes with $\hat{J}_z$. States with different $m_j$ never mix.

For $m_j = +3/2$: Only the $|2p_{3/2}, m_j = +3/2\rangle$ state has this $m_j$ value. Its energy is simply:

$$E = E_2^{(0)} + E_{\text{FS}}(j = 3/2) + g_{3/2} \cdot (3/2) \cdot \mu_B B$$

No diagonalization needed — it is a $1 \times 1$ "matrix."

For $m_j = +1/2$: Three states have $m_j = +1/2$: $|2s_{1/2}, +1/2\rangle$, $|2p_{1/2}, +1/2\rangle$, and $|2p_{3/2}, +1/2\rangle$. However, because $l$ is also conserved by the total Hamiltonian (only states with the same $l$ and $j$ are mixed by the field), the practical subspace requiring diagonalization is at most $2 \times 2$ for states that differ in $j$ but share $l$ and $m_j$. Working through the detailed matrix elements (which requires the Clebsch-Gordan coefficients from Chapter 14) produces energy levels that smoothly interpolate between the weak-field Zeeman pattern and the strong-field Paschen-Back pattern.

The result is a set of energy curves — the Zeeman diagram — that show level crossings and avoided crossings as $B$ varies. Plotting these curves (as the code in example-01-fine-structure.py does) is one of the most instructive exercises in this chapter.

Summary of Zeeman Regimes

🔗 Connection: The weak-to-strong field transition is a beautiful example of a theme we have encountered repeatedly: symmetry dictates which quantum numbers are "good." When $\hat{H}'_{\text{FS}}$ dominates, the rotational symmetry of the spin-orbit coupling makes $j$ and $m_j$ good. When $\hat{H}'_Z$ dominates, the axial symmetry of the magnetic field makes $m_l$ and $m_s$ good. In the intermediate regime, neither set is perfectly good, and we must work harder.

Selection Rules and Spectral Lines

The Zeeman effect does not merely split energy levels — it splits spectral lines. The selection rules for electric dipole (E1) transitions in a magnetic field are:

• $\Delta l = \pm 1$ (parity must change)
• $\Delta j = 0, \pm 1$ (but $j = 0 \to j = 0$ is forbidden)
• $\Delta m_j = 0$ ($\pi$ polarization, linear along $\mathbf{B}$)
• $\Delta m_j = \pm 1$ ($\sigma^{\pm}$ polarization, circular perpendicular to $\mathbf{B}$)

In the strong-field limit, the selection rules on $m_l$ and $m_s$ are: - $\Delta m_l = 0, \pm 1$ - $\Delta m_s = 0$ (the spin does not change in an E1 transition)

These selection rules mean that in the Paschen-Back regime, the primary splitting produces exactly three spectral components ($\Delta m_l = 0, \pm 1$), all equally spaced by $\mu_B B / h$ in frequency. This is the "normal" Zeeman triplet — and it is deeply satisfying that the strong-field limit recovers the simple classical prediction, while the weak-field limit gives the complex "anomalous" pattern that so puzzled early spectroscopists.

🧪 Experiment: Modern magneto-optical spectroscopy uses laser-based techniques to resolve Zeeman structure with extraordinary precision. In a magneto-optical trap (MOT), atoms are cooled to microkelvin temperatures, eliminating Doppler broadening and revealing Zeeman sublevels as sharp, well-separated spectral features. This technology underlies atomic clocks, magnetometers, and quantum computing with trapped ions.

18.10 Summary and Progressive Project

What We Learned

This chapter addressed two interconnected problems: the technical challenge of degeneracy in perturbation theory and the physical application to hydrogen's fine and hyperfine structure.

Degenerate perturbation theory resolves the divergences that arise when non-degenerate perturbation theory is applied to systems with degenerate energy levels. The cure is to find the "good" basis — the one that diagonalizes the perturbation within each degenerate subspace. This basis corresponds to the eigenstates of the operators that commute with both the unperturbed and perturbing Hamiltonians. Finding good quantum numbers is often the key to simplifying degenerate perturbation problems.

The fine structure of hydrogen arises from three corrections of order $\alpha^2 E_n$: the relativistic kinetic energy correction (which depends on $l$ and is always negative), spin-orbit coupling (which splits levels by $j$ value and vanishes for $s$-states), and the Darwin term (which affects only $s$-states). Remarkably, these three physically distinct effects combine to give a correction depending only on $n$ and $j$:

$$E_{\text{FS}}^{(1)} = -\frac{(E_n^{(0)})^2}{2m_e c^2}\left(\frac{4n}{j + 1/2} - 3\right)$$

This agrees exactly with the order-$\alpha^4$ expansion of the Dirac equation.

Hyperfine structure arises from the interaction between the electron and nuclear magnetic moments. For the hydrogen ground state, the singlet-triplet splitting produces the 21 cm line ($\nu = 1420$ MHz), one of the most important spectral lines in astrophysics.

The Zeeman effect in hydrogen involves a rich interplay between the fine structure and the external magnetic field. In weak fields, $j$ and $m_j$ are good quantum numbers and the splitting involves the Lande $g$-factor. In strong fields (Paschen-Back), $m_l$ and $m_s$ are good and the pattern simplifies. In intermediate fields, one must diagonalize the full Hamiltonian.

Looking Ahead

• Chapter 19 introduces the variational principle — a completely different approach to approximation that does not require a solvable $\hat{H}_0$.
• Chapter 22 (Scattering Theory) will use the hydrogen potential again, this time for scattering calculations.
• Chapter 29 (Relativistic Quantum Mechanics) will derive the Dirac equation and show how all three fine structure corrections emerge naturally from a single relativistic framework.
• Chapter 38 (Capstone: Hydrogen) will integrate everything from this chapter into a complete hydrogen atom simulation.

Progressive Project Checkpoint: Fine Structure Calculator

In this chapter's project component, you will add the degenerate_perturb() and fine_structure() functions to your quantum toolkit.

Specifications:

1. degenerate_perturb(H0_eigenstates, H_prime, degenerate_indices) — General-purpose degenerate perturbation theory solver. Constructs the $W$-matrix within the degenerate subspace, diagonalizes it, and returns the corrected energies and good states.

2. fine_structure(n, j) — Returns the fine structure energy correction for hydrogen, given $n$ and $j$.

3. spin_orbit(n, l, j) — Returns the spin-orbit contribution alone.

4. zeeman_splitting(n, l, j, B) — Returns Zeeman-shifted energies for all $m_j$ values in the weak-field regime.

5. plot_breit_rabi(n, B_range) — Generates the Breit-Rabi diagram for the specified $n$-manifold.

See code/project-checkpoint.py for the implementation.

🔗 Connection: This project checkpoint builds on the perturbation engine from Chapter 17 (perturbation_1st(), perturbation_2nd(), energy_correction()), the hydrogen radial functions from Chapter 5 (hydrogen_Rnl()), and the angular momentum algebra from Chapter 14 (clebsch_gordan()). The Zeeman module will be extended in Chapter 21 when we study time-dependent perturbation theory and radiative transitions.