Linear Stark Effect in Hydrogen

Most atoms respond to an electric field only at second order: their non-degenerate levels have definite parity, the expectation of the odd operator $z$ vanishes, and the leading shift is the quadratic $-\tfrac12\alpha\mathcal{E}^2$ of Quadratic Stark Effect and Polarizability. Hydrogen breaks the rule. Its Coulomb $1/r$ potential produces the accidental $\ell$ -degeneracy — every state of a given $n$ shares one energy regardless of $\ell$ — so a single level holds states of opposite parity at the same energy. Into that degenerate manifold the perturbation $H'=e\mathcal{E}z$ can reach with nonzero off-diagonal elements, and the response becomes linear in the field. The machinery is degenerate perturbation theory: diagonalize $H'$ inside the fixed- $n$ shell rather than evaluate a diagonal expectation value that is guaranteed to be zero. Two selection rules do almost all the work. Because $[\hat L_z,z]=0$ , the field cannot change $m$ , so the matrix block-diagonalizes by magnetic quantum number. Because $z\propto rY_1^0$ is parity-odd, it connects only $\ell\to\ell\pm1$ . For the $n=2$ shell these rules collapse a $4\times4$ problem to one $2\times2$ block spanned by $2s$ and $2p_0$ . Diagonalizing that block gives eigenvalues $\pm3ea_0\mathcal{E}$ and eigenstates $(\,|2s\rangle\mp|2p_0\rangle)/\sqrt2$ — parity-mixed hybrids that carry a permanent electric dipole of magnitude $3ea_0$ . A permanent dipole in a field has energy $-\mathbf p\cdot\boldsymbol{\mathcal E}$ , linear in $\mathcal{E}$ : this is the physical origin of linearity. The two $m=\pm1$ states stay unshifted at first order, having no opposite-parity partner at their $m$ . The whole pattern for any $n$ is cleanest in parabolic coordinates, in which the Coulomb-plus-field problem separates exactly and the first-order shift is $\Delta E^{(1)}=\tfrac32 n(n_1-n_2)ea_0\mathcal{E}$ , an evenly spaced ladder of $2n-1$ rungs. Every number below is machine-verified: the $-3a_0$ matrix element and its sign, the $\pm3ea_0\mathcal{E}$ splitting, and the parabolic formula's reduction to $n=2$ .

Key takeaways

Hydrogen shows a Stark shift linear in the field $\mathcal{E}$ — unlike generic atoms, whose parity-definite non-degenerate states shift only quadratically.
The cause is the Coulomb accidental degeneracy: states of the same $n$ but different $\ell$ (hence opposite parity) share an energy, so $H'=e\mathcal{E}z$ has nonzero matrix elements inside one level.
Solve by degenerate perturbation theory — diagonalize $H'$ in the $n$ -shell — using $[\hat L_z,z]=0$ ( $\Delta m=0$ ) and the parity rule ( $\Delta\ell=\pm1$ ).
The only nonzero $n=2$ element is $\langle 2,0,0|z|2,1,0\rangle=-3a_0$ (radial $-3\sqrt3\,a_0$ times angular $1/\sqrt3$ ; sign machine-checked).
The $n=2$ level splits into three: $-3ea_0\mathcal{E},\,0,\,0,\,+3ea_0\mathcal{E}$ , the $m=\pm1$ pair unshifted.
The shifted eigenstates $(|2s\rangle\mp|2p_0\rangle)/\sqrt2$ carry a permanent dipole $\pm3ea_0$ — the reason the effect is linear, not quadratic.
For general $n$ the good labels are parabolic $(n_1,n_2,m)$ and $\Delta E^{(1)}=\tfrac32 n(n_1-n_2)ea_0\mathcal{E}$ : $2n-1$ equally spaced sublevels.
Key idea: hydrogen's $\ell$ -degeneracy lets the parity-odd $e\mathcal{E}z$ couple opposite-parity states within one $n$ -shell, so degenerate perturbation theory yields a splitting linear in $\mathcal{E}$ whose eigenstates carry permanent dipoles.

The core insight

Hydrogen's $\ell$ -degeneracy lets the parity-odd perturbation $e\mathcal{E}z$ couple opposite-parity states within a single $n$ -shell, so degenerate perturbation theory produces a splitting linear in $\mathcal{E}$ — with eigenstates that carry permanent electric dipole moments.

Where this lives today

The linear Stark effect is the operating principle of Rydberg-atom technology. Because the parabolic shift grows as $\tfrac32 n(n_1-n_2)ea_0\mathcal{E}$ — linear in $\mathcal{E}$ but scaling as $n^2$ with the shell — highly excited hydrogenic states are extraordinarily field-sensitive, and Rydberg-atom electrometers exploit exactly this ladder to measure DC and microwave fields with single-atom sensitivity, now a workhorse of quantum sensing and of SI-traceable field metrology. The same $n^2$ -amplified shift underlies field ionization: raise $\mathcal{E}$ until $e\mathcal{E}a_0\sim|E_n|$ and the tilted Coulomb saddle lets the electron tunnel out, the state-selective detection method behind most Rydberg experiments and behind the Rydberg blockade used in neutral-atom quantum computers. In precision spectroscopy the effect runs the other way: the $2s$ metastable state of hydrogen is quenched by even a modest field, which mixes in the short-lived $2p$ (the mechanism and lifetime owned by Lamb Shift), a systematic that hydrogen-spectroscopy experiments and the proton-radius measurements must control to sub-kHz accuracy. And the parabolic-state picture itself lives on in strong-field and attosecond physics, where the tilted-barrier tunnelling of this article's ionization limit is the first step of high-harmonic generation.

The physical picture

Start from the fact that ordinary atoms are boring in a field. A non-degenerate state has definite parity, its charge cloud is symmetric under $\mathbf r\to-\mathbf r$ , and $\langle z\rangle=0$ : no permanent dipole to grab. The field must first build a dipole by distorting the cloud, and only then pay energy for it — two powers of $\mathcal{E}$ , the quadratic Stark effect owned by Quadratic Stark Effect and Polarizability. Hydrogen escapes this because at $n\ge2$ two clouds of opposite parity sit at the same energy, and a field can rotate them into a lopsided hybrid for free.

Figure

The two mechanisms side by side: a symmetric cloud must have its dipole induced (quadratic), a degenerate parity-mixed cloud already carries one (linear)

Picture the $2s$ and $2p_0$ states as a balanced pair at one energy. The field tilts the board; because the two have opposite parity, it can mix them into hybrids that push their charge toward or against the field. One hybrid drops in energy (cloud on the low-potential side), the other rises — a symmetric splitting, linear in $\mathcal{E}$ . The energetics is dipole-in-field: each hybrid carries a fixed $\mathbf p=\mp3ea_0\hat z$ , and $-\mathbf p\cdot\boldsymbol{\mathcal E}$ is linear. The selection rules narrow the field's reach to almost nothing. Because $z$ commutes with $\hat L_z$ , it preserves $m$ ; because $z$ is parity-odd it changes $\ell$ by exactly one. Within $n=2$ only the single pair $(2s,2p_0)$ satisfies both, so the $4\times4$ matrix reduces to one $2\times2$ block plus two spectator $m=\pm1$ states that have nothing to couple to.

Figure

The n=2 Stark splitting: one fourfold-degenerate level fans into three, spacing 3 e a₀ ℰ, with 2p_±1 riding through unshifted

One habit for the whole Stark family: the field defines the axis. Choosing $\boldsymbol{\mathcal E}=\mathcal{E}\hat z$ promotes one $\hat L_z$ out of the democratic $m$ -family of Degeneracy and the Role of the Magnetic Quantum Number — until the field arrives there is nothing for $m$ to measure, and afterward $m$ is a conserved label of the split spectrum.

The Stark Hamiltonian and its selection rules

A uniform field $\boldsymbol{\mathcal E}=\mathcal{E}\hat z$ adds to the atomic Hamiltonian the potential energy of the electron (charge $-e$ ) in that field. Writing $z=r\cos\theta$ exposes its angular content immediately:

\hat H' \;=\; e\,\mathcal{E}\,z \;=\; e\,\mathcal{E}\,r\cos\theta \;\propto\; r\,Y_1^0 \qquad (1)

where: $e>0$ is the elementary charge (the electron's is $-e$ ); $\mathcal{E}$ the field magnitude along $\hat z$ ; $z=r\cos\theta$ the coordinate along the field. Read it physically: $H'$ is an odd, $\ell=1$ (dipole) operator — it shifts the cloud along the field, and its transformation content dictates exactly which states it can connect. The unperturbed spectrum is what makes hydrogen special. Its Coulomb levels

E_n^{(0)} \;=\; -\frac{13.6\,\text{eV}}{n^2},\qquad \text{degeneracy} \;=\; \sum_{\ell=0}^{n-1}(2\ell+1) \;=\; n^2 \qquad (2)

where: $n$ is the principal quantum number and the $n^2$ degeneracy lumps all $\ell$ from $0$ to $n-1$ at one energy. This $\ell$ -degeneracy — special to the $1/r$ potential, and traced to a hidden symmetry in Accidental Degeneracy and the SO(4) Symmetry — is the entire reason a first-order effect exists. Two rules then prune the $n^2\times n^2$ matrix of $H'$ :

[\hat L_z,\,z]=0 \;\Rightarrow\; \langle n\ell'm'|z|n\ell m\rangle \propto \delta_{m'm};\qquad \langle\ell'm|\cos\theta|\ell m\rangle\neq 0 \iff \ell'=\ell\pm1 \qquad (3)

where: the first (an operator identity, since $z$ has no $\phi$ -dependence) forbids $\Delta m\ne0$ ; the second is the electric-dipole parity rule — $\cos\theta\propto Y_1^0$ raises or lowers $\ell$ by one. Together they mean $H'$ connects only opposite-parity, same- $m$ partners. Consume the parity doctrine of Parity Symmetry of One-Dimensional Eigenstates and the $m$ -conservation of Degeneracy and the Role of the Magnetic Quantum Number here rather than re-proving them; what is new is that hydrogen supplies such partners inside one energy. The single nonzero coupling in the $n=2$ shell is the workhorse number of the whole subject:

\langle 2,0,0|z|2,1,0\rangle \;=\; \Big(\!\int_0^\infty\! R_{20}R_{21}\,r^3\,\mathrm dr\Big)\Big(\!\int Y_0^{0*}\cos\theta\,Y_1^0\,\mathrm d\Omega\Big) \;=\; (-3\sqrt3\,a_0)\!\left(\tfrac{1}{\sqrt3}\right) \;=\; -3a_0 \qquad (4)

where: $R_{20},R_{21}$ are the $n=2$ radial functions and $a_0\approx0.529\,\text{\AA}$ the Bohr radius. The factorization into a radial integral $-3\sqrt3\,a_0$ and an angular integral $1/\sqrt3$ is worth internalizing: the sign lives in the radial part and is easy to drop. Both pieces — and the product — are machine-verified below, along with a direct three-dimensional cross-check.

Derivation AFirst-order Stark splitting of the n=2 level

The goal is to obtain the symmetric splitting $\pm3ea_0\mathcal{E}$ by degenerate perturbation theory, treating the four $n=2$ states on an equal footing and letting the selection rules collapse the matrix. The subtlety degenerate perturbation theory exists to handle is precisely this: the naive diagonal shift $\langle n\ell m|H'|n\ell m\rangle$ vanishes by parity for every state, yet the level manifestly splits — because the physics lives entirely in the off-diagonal elements that couple the degenerate partners. We build the $4\times4$ matrix, watch three of four rows die, and diagonalize what remains.

Result.

\Delta E \;=\; \{\,+3ea_0\mathcal{E},\;\;0,\;\;0,\;\;-3ea_0\mathcal{E}\,\}

Step 1. Order the degenerate basis $(|2,0,0\rangle,\,|2,1,0\rangle,\,|2,1,1\rangle,\,|2,1,-1\rangle)$ — the $2s$ and the three $2p$ states. Every diagonal element $\langle 2\ell m|z|2\ell m\rangle=0$ because each basis state has definite parity and $z$ is parity-odd. So the matrix is purely off-diagonal.

Step 2. Apply the $m$ -rule. Since $[\hat L_z,z]=0$ , $z$ cannot change $m$ : the only states sharing an $m$ are $2s$ and $2p_0$ (both $m=0$ ). The $m=\pm1$ states have no same- $m$ partner of opposite parity anywhere in $n=2$ , so their entire rows and columns vanish — two immediate zero eigenvalues.

Step 3. Insert the one surviving coupling, $\langle 2,0,0|z|2,1,0\rangle=-3a_0$ from equation (4). The $4\times4$ problem reduces to a $2\times2$ block in the $\{2s,2p_0\}$ subspace,

\hat H'_{n=2}\big|_{m=0} \;=\; -3ea_0\mathcal{E}\begin{pmatrix}0 & 1\\ 1 & 0\end{pmatrix}.

Step 4. Diagonalize the block: $\det(-3ea_0\mathcal{E}\,\sigma_x-\lambda\mathbb 1)=\lambda^2-(3ea_0\mathcal{E})^2=0$ , so $\lambda=\pm3ea_0\mathcal{E}$ with eigenvectors $(|2,0,0\rangle\mp|2,1,0\rangle)/\sqrt2$ (equal-weight, verified $|c_{2s}|=|c_{2p_0}|=1/\sqrt2$ ).

Step 5. Assemble the spectrum. The fourfold-degenerate level splits into three: a shifted pair at $\pm3ea_0\mathcal{E}$ and a doubly degenerate unshifted level at $0$ (the $m=\pm1$ spectators). Total splitting $6ea_0\mathcal{E}$ — the fan of the figure above. Verified: the radial integral $-3\sqrt3\,a_0$ , angular factor $1/\sqrt3$ , product $-3a_0$ (and a direct 3D integral cross-check), the $\pm3ea_0\mathcal{E}$ eigenvalues, and the equal-weight eigenvectors all reproduced symbolically in this article's verify script against Griffiths §7.3 and Sakurai §5.2.

Go deeper · Derivation B · 1 minThe permanent dipole that makes the shift linear

Derivation BThe permanent dipole that makes the shift linear

The goal is to show that the hybridized eigenstates carry a genuine permanent electric dipole moment, so that the linear shift is not a formal accident but the ordinary energy $-\mathbf p\cdot\boldsymbol{\mathcal E}$ of a dipole in a field. We compute $\langle z\rangle$ in a mixed state and watch the cross term — which parity forbids for pure states — survive because the state is not a parity eigenstate. This is the conceptual payload of the whole article: degeneracy manufactures a dipole out of two states that individually have none.

Result.

\langle\psi_\pm|(-e z)|\psi_\pm\rangle \;=\; \pm3ea_0,\qquad \Delta E_\pm=-\langle p_z\rangle\,\mathcal{E}=\pm3ea_0\mathcal{E}

Step 1. Take $|\psi_\pm\rangle=(|2s\rangle\mp|2p_0\rangle)/\sqrt2$ and expand

\langle\psi_\pm|z|\psi_\pm\rangle=\tfrac12\big(\langle 2s|z|2s\rangle+\langle 2p_0|z|2p_0\rangle\mp2\langle 2s|z|2p_0\rangle\big).

Step 2. The two diagonal terms vanish by parity, leaving only the cross term — exactly the element that is nonzero because $2s$ and $2p_0$ have opposite parity.

Step 3. Insert $\langle 2s|z|2p_0\rangle=-3a_0$ : $\langle\psi_\pm|z|\psi_\pm\rangle=\mp(-3a_0)=\pm3a_0$ , so the electron dipole $\langle p_z\rangle=-e\langle z\rangle=\mp3ea_0$ — a permanent moment of magnitude $3ea_0\approx2.5\times10^{-29}\,\text{C}\,\cdot\,\text{m}$ (about $7.6$ debye), fixed independent of field strength.

Step 4. The energy of that dipole in the field, $\Delta E_\pm=\langle\psi_\pm|e\mathcal{E}z|\psi_\pm\rangle=\pm3ea_0\mathcal{E}$ , is linear in $\mathcal{E}$ and matches the eigenvalue derivation term for term. The dipole-in-field cartoon $-\mathbf p\cdot\boldsymbol{\mathcal E}$ is not an analogy here — it is the exact result. Verified: $\langle\psi_+|z|\psi_+\rangle=+3a_0$ , $\langle\psi_-|z|\psi_-\rangle=-3a_0$ , and $\Delta E_+=+3ea_0\mathcal{E}$ recomputed symbolically; dipole magnitude cross-checked against Cohen-Tannoudji Complement $E_{XII}$ .

The general pattern: parabolic coordinates

The $n=2$ block is the whole story in miniature, but the elegant general result comes from abandoning spherical coordinates. Because a uniform field plus a Coulomb centre share a common axis but no common centre, the natural coordinates are the confocal parabolic ones $(\xi,\eta,\phi)$ , in which the Stark Hamiltonian separates exactly. The good quantum numbers become the parabolic $(n_1,n_2,m)$ with $n_1+n_2+|m|+1=n$ , and the first-order shift is a strikingly simple linear formula:

\Delta E^{(1)}_{n_1n_2m} \;=\; \tfrac32\,n\,(n_1-n_2)\,e a_0\,\mathcal{E} \qquad (5)

where: $n_1,n_2\ge0$ count nodes along the "up-field" and "down-field" parabolic axes and $n_1-n_2$ ranges over the $2n-1$ integers from $-(n-1)$ to $+(n-1)$ . Each shell therefore fans into $2n-1$ equally spaced sublevels with spacing $\tfrac32 n\,ea_0\mathcal{E}$ . Set $n=2$ : the allowed $(n_1,n_2,m)$ are $(1,0,0),(0,1,0),(0,0,\pm1)$ , giving $n_1-n_2\in\{+1,-1,0,0\}$ and shifts $\{+3ea_0\mathcal{E},-3ea_0\mathcal{E},0,0\}$ — exactly the spherical-basis answer, now read off a formula. The parabolic states are the correct zeroth-order Stark states; the $s$ – $p$ hybrids of the $2\times2$ block are simply the $n=2$ parabolic states written in the spherical basis.

Limiting cases

Which Stark regime governs hydrogen depends on how the field energy $ea_0\mathcal{E}$ compares to the internal fine-structure splitting:

case	result	meaning
strong field, $ea_0\mathcal{E}\gg\Delta E_{\text{fs}}$	pure linear splitting $\pm3ea_0\mathcal{E}$	the field beats fine structure; diagonalize $e\mathcal{E}z$ in the $\ell$ -degenerate shell first — the treatment of this article
weak field, $ea_0\mathcal{E}\ll\Delta E_{\text{fs}}$	leading shift $\propto\mathcal{E}^2$	fine structure lifts the $\ell$ -degeneracy first; states regain definite parity within each $j$ -multiplet, so the response becomes quadratic — owned by Quadratic Stark Effect and Polarizability
ground state $n=1$	$\Delta E^{(1)}=0$ , $\Delta E^{(2)}=-\tfrac12\alpha_{1s}\mathcal{E}^2$	non-degenerate, definite parity: no linear term; polarizability $\alpha_{1s}=\tfrac92 a_0^3$ , purely quadratic
field-ionization limit, $e\mathcal{E}a_0\sim\|E_n\|$	Stark resonances (finite widths)	the tilted Coulomb potential develops a saddle; bound levels become quasi-bound and tunnel out — perturbation theory fails

Common trapDo not compute a diagonal expectation value and declare the effect zero

The tempting mistake is to evaluate $\langle n\ell m|z|n\ell m\rangle$ , find it zero by parity, and conclude hydrogen has no linear Stark effect. Those diagonal elements are zero — but the level is degenerate, so first-order theory requires diagonalizing the whole degenerate block, and the physics lives in the off-diagonal $\ell\to\ell\pm1$ elements. Equally, never forget that real hydrogen has fine structure: the pure linear result holds only when $ea_0\mathcal{E}$ exceeds the fine-structure scale ( $\sim10^{-4}$ eV at $n=2$ , i.e. fields above $\sim10^5$ V/cm). Below that threshold, fine structure restores definite parity within each $j$ -multiplet and the leading response is quadratic. Decide the regime by the ratio, never by whether the field "sounds strong."

Worked examples

Example 1 — the size of the $n=2$ splitting in a strong lab field. Place hydrogen in $\mathcal{E}=10^{7}$ V/m ( $=10^{5}$ V/cm, a strong but attainable field). Find the splitting and confirm which regime applies.

Step 1. Half-split: $3ea_0\mathcal{E}=3\times(1.602\times10^{-19}\,\text{C})(5.29\times10^{-11}\,\text{m})(10^{7}\,\text{V/m})=2.543\times10^{-22}$ J $=1.5875\times10^{-3}$ eV (computed).

Step 2. As a frequency, $3ea_0\mathcal{E}/h=383.9$ GHz (computed) — a large, easily resolved shift. The full pattern is three lines at $-1.59,\,0,\,+1.59$ meV, total span $6ea_0\mathcal{E}=3.18\times10^{-3}$ eV.

Step 3. Regime check: compare to the $n=2$ fine-structure gap $\Delta E_{\text{fs}}=13.6\,\text{eV}\cdot\alpha^2/16=4.53\times10^{-5}$ eV (computed). Here $ea_0\mathcal{E}/\Delta E_{\text{fs}}\approx35\gg1$ : deep in the strong-field, purely-linear regime, so the $\pm3ea_0\mathcal{E}$ answer is legitimate. Answer: a symmetric triplet at $0,\pm1.59\times10^{-3}$ eV ( $\pm383.9$ GHz), with the $m=\pm1$ states pinned at the centre — and the field is $\sim35\times$ past the linear-regime threshold.

Example 2 — the parabolic ladder for $n=3$ . Predict the first-order Stark pattern of the $n=3$ shell.

Step 1. Enumerate parabolic states with $n_1+n_2+|m|+1=3$ : $(n_1,n_2,m)$ running over $(2,0,0),(1,1,0),(0,2,0),(1,0,\pm1),(0,1,\pm1),(0,0,\pm2)$ , giving $n_1-n_2\in\{+2,+1,0,-1,-2\}$ (each value with its own $m$ -degeneracy).

Step 2. Apply equation (5) with $n=3$ : $\Delta E^{(1)}=\tfrac32(3)(n_1-n_2)ea_0\mathcal{E}=\tfrac92(n_1-n_2)ea_0\mathcal{E}$ . The five distinct shifts are $\{+9,+\tfrac92,0,-\tfrac92,-9\}\,ea_0\mathcal{E}$ — a five-rung ladder, equally spaced by $\tfrac92 ea_0\mathcal{E}$ (computed).

Step 3. Count sublevels: $2n-1=5$ energies, confirming the general rule, with the central $n_1-n_2=0$ rung carrying the extra $m$ -degeneracy. Answer: $n=3$ splits into five equally spaced lines at $0,\pm\tfrac92 ea_0\mathcal{E},\pm9ea_0\mathcal{E}$ , spacing $\tfrac92 ea_0\mathcal{E}$ — the Stark ladder Stark himself resolved in the Balmer series.

Origins · 1 minParabolic Coordinates and how this was discovered

Parabolic Coordinates

1913–1926 · Munich, Göttingen

The hydrogen Stark effect was old quantum theory's last and cleanest triumph. Within months of Stark's 1913 measurement, Paul Epstein in Munich and Karl Schwarzschild in Göttingen independently solved it — by recognising that a Coulomb centre and a uniform field share an axis but not a centre, so the problem separates not in spherical but in parabolic coordinates $(\xi,\eta,\phi)$ , the confocal paraboloids that thread the field lines. In those coordinates the Hamilton–Jacobi equation splits cleanly, the action integrals quantize into two parabolic quantum numbers, and out drops the first-order shift $\tfrac32 n(n_1-n_2)ea_0\mathcal{E}$ — the $2n-1$ evenly spaced components Stark had photographed. Schwarzschild finished the work while serving at the front and died of illness months later, in 1916, having also written down the first exact solution of Einstein's field equations that same year; his Stark paper is a coda to a career compressed into eighteen months of war. Epstein carried the parabolic method forward for decades. When Schrödinger rebuilt quantum mechanics in 1926, the very same separation reappeared in wave mechanics — the parabolic states are exact solutions of the Stark Schrödinger equation to first order — so the coordinate trick outlived the theory that spawned it. The lesson the coordinate teaches is permanent: pick the coordinates the symmetry prefers, and a hard perturbation becomes a separable one.

History

1913

— Stark resolves hydrogen Balmer lines splitting into evenly spaced components in a strong field

1916

— Epstein and Schwarzschild independently solve the hydrogen Stark effect in parabolic coordinates

1916

— Schwarzschild dies; his Stark and general-relativity papers both date to this year

1926

— Schrödinger re-derives the parabolic separation within wave mechanics

1927

— Unsöld and others compute the second-order (quadratic) shift, completing the field-response picture

1974

— Zimmerman and collaborators map Rydberg Stark manifolds, exhibiting the

2n-1

ladders at high

n

Going further

Griffiths & Schroeter, Introduction to Quantum Mechanics, 3rd ed., §7.3 — the Stark effect as the degenerate-perturbation-theory showcase.
Sakurai & Napolitano, Modern Quantum Mechanics, 2nd ed., §5.2 — the $n=2$ diagonalization and the parity selection rule.
Bethe & Salpeter, Quantum Mechanics of One- and Two-Electron Atoms, §51 — the full parabolic-coordinate treatment for arbitrary $n$ .
Landau & Lifshitz, Quantum Mechanics, §77 — the parabolic separation and the general first-order formula.
Gallagher, Rydberg Atoms — the experimental Stark maps and field-ionization physics.

Provenance note: A verifier should reconfirm: (1) the radial integral $\int_0^\infty R_{20}R_{21}r^3\,\mathrm dr=-3\sqrt3\,a_0$ and angular factor $1/\sqrt3$ giving $\langle2,0,0|z|2,1,0\rangle=-3a_0$ (Griffiths §7.3); (2) the eigenvalues $\pm3ea_0\mathcal{E}$ of the $2\times2$ block; (3) the parabolic first-order formula $\Delta E=\tfrac32 n(n_1-n_2)ea_0\mathcal{E}$ (Bethe–Salpeter §51, Landau–Lifshitz §77); (4) the $1s$ polarizability $\alpha_{1s}=\tfrac92 a_0^3$ in Gaussian units (standard). Numbers: $E_n=-13.6\,\mathrm{eV}/n^2$ , $a_0\approx0.529\,\text{\AA}$ . Selection rules from $[L_z,z]=0$ and the parity of $z=r\cos\theta\propto Y_1^0$ . Cross-checked against Sakurai §5.2 and Cohen-Tannoudji Complement $E_{XII}$ . — All of items (1)–(3) and the $n=2$ / $n=3$ field conversions run in this article's verify script, including a direct three-dimensional evaluation of the matrix element that independently pins its sign; item (4) is derived and verified where it is owned, in Quadratic Stark Effect and Polarizability. Ownership note: the degenerate-perturbation-theory machinery (the $W$ -matrix and its eigenvalue formula) is derived in Degenerate Perturbation Theory and consumed here as its named $n=2$ application; the quadratic/non-degenerate branch and the polarizability $\tfrac92 a_0^3$ belong to Quadratic Stark Effect and Polarizability; the $2s$ -quenching lifetime detail belongs to Lamb Shift.

Why does degenerate perturbation theory give a linear shift when ordinary perturbation theory gives a quadratic one?

Ordinary (non-degenerate) first-order theory returns $\langle n|H'|n\rangle$ , which vanishes by parity for $H'=e\mathcal{E}z$ — so the first nonzero term is second order, hence quadratic. Degenerate theory instead diagonalizes $H'$ within the degenerate subspace, and its eigenvalues are already first order in $\mathcal{E}$ because $H'$ has nonzero off-diagonal elements there. The difference is not a different physics but a different bookkeeping forced by the degeneracy: when several states share an energy, first-order theory is a matrix eigenvalue problem, not a number.

Where does the sign of the −3a₀ matrix element come from, and does it matter?

The sign lives entirely in the radial integral $\int R_{20}R_{21}r^3\,\mathrm dr=-3\sqrt3\,a_0$ ; the angular factor $1/\sqrt3$ is positive. It arises because $R_{20}$ changes sign (the $2s$ radial function has a node at $r=2a_0$ ) while $R_{21}$ does not, so the weighted overlap comes out negative. For the splitting $\pm3ea_0\mathcal{E}$ the sign is immaterial — it only swaps which hybrid $(|2s\rangle\pm|2p_0\rangle)/\sqrt2$ goes up and which goes down. But it does fix the sign of each eigenstate's permanent dipole, which matters the moment you ask which way a given Stark state points. This article's verify script pins the sign with a direct 3D integral to remove any convention ambiguity.

Are the parabolic states and the s–p hybrids the same thing?

Yes — they are the same states in two languages. The $2\times2$ diagonalization produces $(|2s\rangle\mp|2p_0\rangle)/\sqrt2$ ; these are the $n=2$ parabolic states $(n_1,n_2,m)=(1,0,0)$ and $(0,1,0)$ written in the spherical basis. Parabolic coordinates are simply the basis in which the Stark Hamiltonian is already diagonal, so no diagonalization is needed — the good quantum numbers $(n_1,n_2,m)$ label the exact first-order eigenstates directly. For $n=2$ the two pictures are trivially equivalent; for larger $n$ the parabolic route is the only tractable one.

Does the linear Stark effect really give hydrogen a permanent dipole moment? Doesn't that violate something?

The Stark eigenstates carry a permanent dipole $\pm3ea_0$ , but the isolated atom does not: in zero field the true stationary states are the parity eigenstates $2s$ and $2p_0$ , each with $\langle z\rangle=0$ . The permanent dipole appears only once the field selects the hybrid basis — the field breaks the parity symmetry that otherwise forbids a ground-state dipole. Nothing is violated: a permanent dipole is forbidden only for a non-degenerate parity eigenstate, and the $n=2$ states are degenerate, so the theorem simply does not apply.

Why does fine structure turn the linear effect into a quadratic one at weak fields?

Fine structure (spin–orbit plus relativistic corrections) already lifts the $\ell$ -degeneracy by $\sim10^{-4}$ eV, splitting $n=2$ into $2s_{1/2}$ , $2p_{1/2}$ , $2p_{3/2}$ of definite $j$ . When $ea_0\mathcal{E}\ll\Delta E_{\text{fs}}$ , these $j$ -levels are effectively non-degenerate and parity-definite, so $\langle z\rangle=0$ again and the leading Stark shift is second order — quadratic. The linear result is recovered only when the field is strong enough to override fine structure and restore the effective $\ell$ -degeneracy. So "hydrogen has a linear Stark effect" is a strong-field statement; the honest weak-field answer is quadratic.

How does this connect to field ionization of Rydberg atoms?

The same tilted potential that produces the linear shift also produces a saddle: $-e^2/4\pi\varepsilon_0 r - e\mathcal{E}z$ has a maximum on the down-field side, and once $e\mathcal{E}a_0$ approaches $|E_n|$ that barrier drops below the level and the electron escapes. Because $|E_n|\propto n^{-2}$ and the relevant length scales as $n^2$ , the critical field for ionization scales as $n^{-4}$ : a state with $n=30$ ionizes in a field $\sim10^6$ times weaker than the ground state. This $n^{-4}$ sensitivity is what makes state-selective field ionization the standard Rydberg detection method — the linear Stark shift and field ionization are two faces of the same tilted Coulomb potential.