Electric Dipole Approximation

When light drives an atomic transition, the electron feels the wave's oscillating electric field — but not all of it at once. The field varies over a wavelength of hundreds of nanometres, while the atom it is pushing on is a fraction of a nanometre across. From the electron's point of view the wave is essentially uniform: the same field everywhere across the orbit, merely turning on and off in time. Replacing the true spatially-varying field by this uniform one is the electric dipole approximation, the single most important simplification in the theory of light–matter interaction, and the reason atomic spectroscopy is dominated by a short list of "allowed" lines. We take the derivation of the approximation from quantum mechanics and keep its optical consequences here. The expansion $e^{i\mathbf k\cdot\mathbf r}=1+i\mathbf k\cdot\mathbf r+\dots$ , the minimal-coupling Hamiltonian $(\mathbf p-q\mathbf A)^2/2m$ , and the equivalence of the length and velocity gauges are owned by aq-qm-dipole-approximation and aq-qm-semiclassical-atom-radiation; we recall them by name. What this article owns is the physical smallness parameter that justifies keeping only the leading " $1$ " — the ratio $ka_0 = 2\pi a_0/\lambda \approx 6.65\times10^{-4}$ for an atom in optical light — and what happens when that leading term vanishes. The payoff ties this whole chapter together. Because the next term in the expansion is smaller by $ka_0\sim10^{-3}$ , the electric-quadrupole and magnetic-dipole transitions it generates are weaker in rate by $(ka_0)^2\sim4\times10^{-7}$ . That one number explains why "forbidden" lines are $10^6$ – $10^7$ times slower than allowed ones, why metastable states live for seconds instead of nanoseconds, and why the 21 cm hydrogen line takes ten million years to fire. The electric dipole approximation is not just a convenience — its breakdown is the origin of the entire forbidden spectrum.

Key takeaways

The electric dipole (E1) approximation replaces the spatially-varying field $\mathbf E(\mathbf r,t)$ across an atom by a uniform field $\mathbf E(t)$ , giving the interaction $H' = -\mathbf d\cdot\mathbf E$ with $\mathbf d = -e\mathbf r$ .
It is justified by expanding $e^{i\mathbf k\cdot\mathbf r} = 1 + i\mathbf k\cdot\mathbf r + \dots$ and keeping only the " $1$ ", valid because the smallness parameter $ka_0 = 2\pi a_0/\lambda \approx 6.65\times10^{-4}$ for an atom ( $a_0\approx0.05$ nm) in optical light ( $\lambda\approx500$ nm).
The neglected next term generates electric-quadrupole (E2) and magnetic-dipole (M1) transitions, weaker in rate by $(ka_0)^2 \approx 4.4\times10^{-7}$ .
So when E1 is forbidden by a selection rule, the surviving transitions are $\sim 10^6$ – $10^7$ times slower — the origin of metastable states and forbidden lines (H 2s/2p lifetime ratio $\approx 7.5\times10^7$ ; the 21 cm M1 line lives $\sim10^7$ yr).
The length gauge ( $-\mathbf d\cdot\mathbf E$ ) and velocity gauge ( $-\tfrac{e}{m}\mathbf A\cdot\mathbf p$ ) give identical E1 rates — a consequence of the commutator $[\mathbf r, H_0] = i\hbar\mathbf p/m$ .
Key idea: because an atom is $\sim10^{-3}$ of an optical wavelength, the field is uniform across it and only the leading dipole term survives; the tiny next term, suppressed by $(ka_0)^2\sim10^{-7}$ in rate, is exactly what makes forbidden transitions and metastable states.

The core insight

Because an atom ( $a_0\approx0.05$ nm) is only $\sim10^{-3}$ of an optical wavelength, the field is essentially uniform across it, so the interaction reduces to $-\mathbf d\cdot\mathbf E$ (the electric dipole approximation); the neglected next multipole is suppressed in rate by $(ka_0)^2\sim4\times10^{-7}$ , which is the origin of forbidden lines and metastable states.

Where this lives today

The electric dipole approximation is the silent workhorse of every calculation in atomic, molecular, and optical physics — from designing laser gain media to computing the light shifts in an atomic clock, one writes $-\mathbf d\cdot\mathbf E$ and proceeds. But the frontier is increasingly in its corrections. Optical atomic clocks exploit transitions that are E1-forbidden precisely because the resulting metastable states have millihertz linewidths; the Sr and Yb clock transitions, and the octupole (E3) transition in Yb⁺ with a lifetime of years, are engineering triumphs built on the multipole hierarchy this article quantifies. In astrophysics, the forbidden [O III], [N II], and [S II] lines remain the primary density–temperature diagnostics of ionized nebulae, and the 21 cm hydrogen line — a magnetic-dipole transition ten million years in the waiting — is the tool with which radio astronomers map neutral hydrogen across the galaxy and back to the cosmic dawn. Even in strong-field and attosecond physics, where intensities are enormous, the question of when the dipole approximation breaks down (as $ka_0$ or the ponderomotive excursion grows) is an active research topic. Knowing the leading term, and knowing exactly how small its corrections are, is what lets physicists trust the simple picture — and exploit its failures.

The physical picture

Imagine standing in an ocean swell whose wavelength is a kilometre. Over the span of your body the water surface is essentially flat — it rises and falls in time, but it does not vary in space across you. That is exactly an atom's view of a light wave. The optical wavelength is $\sim500$ nm; the atom is $\sim0.1$ nm. The ratio is about a thousand to one, so the electric field of the wave has the same value and direction across the entire electron orbit at any instant, changing only in time. The atom feels a spatially-uniform, temporally-oscillating field — a "dipole" drive. Mathematically, the spatial variation of the field is carried by the phase factor $e^{i\mathbf k\cdot\mathbf r}$ , where $\mathbf k$ is the wavevector and $\mathbf r$ ranges over the atom. Expanding it, $e^{i\mathbf k\cdot\mathbf r} = 1 + i\mathbf k\cdot\mathbf r - \tfrac12(\mathbf k\cdot\mathbf r)^2 + \dots$ , the successive terms are the electric-dipole, then the magnetic-dipole and electric-quadrupole, then higher multipoles. The size of each term relative to the last is $\mathbf k\cdot\mathbf r \sim k a_0 = 2\pi a_0/\lambda$ , which for optical light is about $6.65\times10^{-4}$ — a thousandth. Keeping only the " $1$ " is the electric dipole approximation, and the error is genuinely of order $10^{-3}$ in amplitude. Now the crucial consequence. Rates go as the square of the amplitude, so the next multipole (E2 or M1) produces transitions weaker by $(ka_0)^2 \sim 4\times10^{-7}$ than an allowed E1 transition. As long as E1 is allowed, this correction is utterly negligible — nobody measures a $10^{-7}$ admixture on a bright line. But when a selection rule makes the E1 dipole matrix element vanish (recall aq-optics-selection-rules-atomic-transitions), the E2 and M1 terms are all that remain, and the transition proceeds at $10^{-6}$ – $10^{-7}$ of the allowed rate. A state whose only decay is so suppressed becomes metastable, living not nanoseconds but seconds or longer. The breakdown of the very approximation that simplifies everything is the origin of the entire forbidden spectrum.

Figure

Atom in a wave: over the atom (a₀ ≈ 0.05 nm) the 500 nm field is uniform — ka₀ ≈ 6.65×10⁻⁴

Figure

Multipole suppression ladder: E1 rate 1, E2 ~ (ka₀)² = 4.4×10⁻⁷ ⟹ forbidden lifetimes ~10⁶–10⁷× longer

The core equations

An atom in a light wave couples to the field through the phase factor that carries the field's spatial variation. Expanding that factor in the small quantity $\mathbf k\cdot\mathbf r$ is the multipole expansion; the leading term is the electric dipole.

e^{i\mathbf k\cdot\mathbf r} = 1 + i\,\mathbf k\cdot\mathbf r - \tfrac12(\mathbf k\cdot\mathbf r)^2 + \dots \tag{1}

where: $\mathbf k$ — wavevector of the light, $|\mathbf k|=2\pi/\lambda$ ; $\mathbf r$ — electron coordinate, of order $a_0$ . The " $1$ " gives the electric-dipole (E1) interaction, the $i\mathbf k\cdot\mathbf r$ term gives magnetic-dipole (M1) and electric-quadrupole (E2), and so on. Each term is smaller than the last by the dimensionless parameter that governs the whole expansion.

ka_0 = \frac{2\pi a_0}{\lambda} \approx 6.65\times10^{-4}\quad(\lambda=500\ \mathrm{nm}),\qquad \frac{a_0}{\lambda}\approx1.06\times10^{-4} \tag{2}

where: $a_0=0.0529$ nm — the Bohr radius (atomic scale); $\lambda$ — optical wavelength. Because $ka_0\sim10^{-3}\ll1$ , dropping everything but the " $1$ " is an excellent approximation. Keeping only that term, the interaction Hamiltonian becomes the dipole coupling.

H' = -\,\mathbf d\cdot\mathbf E(t),\qquad \mathbf d = -e\,\mathbf r \tag{3}

where: $\mathbf d$ — electric-dipole operator; $\mathbf E(t)$ — the (now spatially-uniform) field at the atom. This is the length-gauge form; the transition rate is $\propto|\langle f|\mathbf d|i\rangle|^2$ , and its selection rules ( $\Delta l=\pm1$ , etc.) are exactly those of aq-optics-selection-rules-atomic-transitions. The size of the next term sets how weak a forbidden line is.

\frac{W_{\text{E2}}}{W_{\text{E1}}} \sim (ka_0)^2 \approx 4.4\times10^{-7},\qquad \frac{\tau_{\text{forbidden}}}{\tau_{\text{allowed}}} \sim \frac{1}{(ka_0)^2} \approx 2.3\times10^{6} \tag{4}

where: $W$ — transition rate; $\tau$ — lifetime. A forbidden (E2/M1) transition is slower than an allowed E1 one by roughly $(ka_0)^2$ , so its upper state lives $\sim10^6$ times longer — the quantitative origin of metastability. Finally, the choice of gauge does not matter for the E1 rate, because the length and velocity forms are linked by a commutator identity.

-\mathbf d\cdot\mathbf E \;\longleftrightarrow\; -\frac{e}{m}\mathbf A\cdot\mathbf p,\qquad [\mathbf r, H_0] = \frac{i\hbar}{m}\mathbf p \tag{5}

where: $\mathbf A$ — vector potential; $\mathbf p$ — momentum; $H_0$ — atomic Hamiltonian. The identity $\langle f|\mathbf p|i\rangle = im\omega_{fi}\langle f|\mathbf r|i\rangle$ (from the commutator) converts the velocity-gauge matrix element into the length-gauge one, so both gauges give identical E1 rates — the equivalence proved in aq-qm-dipole-approximation.

Derivation AThe smallness parameter that justifies the dipole approximation

The goal is to establish, quantitatively, why keeping only the leading term of $e^{i\mathbf k\cdot\mathbf r}$ is an excellent approximation for atoms in optical (and even X-ray) light — by computing the dimensionless expansion parameter $ka_0$ and showing it is of order $10^{-3}$ . The strategy is to compare the two length scales in the problem — the wavelength of the light and the size of the atom — since their ratio is exactly what controls the size of the correction terms. Watch that the rate correction is the square of the amplitude correction, which is what makes forbidden lines so extraordinarily weak.

Step 1. Identify the expansion parameter. The spatial variation of the field across the atom is $\mathbf k\cdot\mathbf r$ , with $|\mathbf k|=2\pi/\lambda$ and $|\mathbf r|\sim a_0$ . So the relevant small number is $ka_0 = 2\pi a_0/\lambda$ .

Step 2. Evaluate for optical light. With $a_0 = 0.0529$ nm and $\lambda = 500$ nm: $ka_0 = 2\pi(0.0529)/500 = 6.65\times10^{-4}$ . The field varies by less than a part in a thousand across the atom. Why this is small even for X-rays: the atom is bound by Coulomb physics at the scale $a_0$ , while the wavelength that resonantly drives a transition is set by the transition energy; for a valence transition $\lambda\gg a_0$ always, and even for a 1 keV X-ray ( $\lambda\sim1$ nm) $ka_0\sim0.3$ — only then does the approximation start to fail.

Step 3. Amplitude vs rate. The first correction to the E1 amplitude is of order $ka_0\sim10^{-3}$ . Because the transition rate is the amplitude squared, the correction to the rate — the strength of the E2/M1 channel relative to E1 — is $(ka_0)^2\approx4.4\times10^{-7}$ .

Step 4. Consequence for lifetimes. When E1 is allowed, this $10^{-7}$ admixture is invisible. When E1 is forbidden and only E2/M1 survives, the transition rate is $\sim(ka_0)^2$ of a typical allowed rate, so the lifetime is longer by $\sim1/(ka_0)^2\approx2.3\times10^6$ — nanoseconds become milliseconds to seconds. Verified: $ka_0=6.65\times10^{-4}$ , $a_0/\lambda=1.06\times10^{-4}$ , $(ka_0)^2=4.4\times10^{-7}$ , $1/(ka_0)^2=2.3\times10^6$ (verify_aq-optics-electric-dipole-approximation.py).

Go deeper · Derivation B · 2 minFrom the multipole hierarchy to real forbidden-line lifetimes

Derivation BFrom the multipole hierarchy to real forbidden-line lifetimes

The goal is to connect the abstract suppression factor $(ka_0)^2$ to the measured lifetimes of real metastable states, showing that the electric-dipole approximation's breakdown quantitatively predicts why forbidden lines are $10^6$ – $10^8$ times slower than allowed ones. The strategy is to take the E1 lifetime scale (nanoseconds), apply the multipole suppression, and compare with the actual lifetimes of hydrogen's 2s state and the astrophysical forbidden lines. Watch that a two-photon or higher-order process can suppress even further than a single E2/M1 step.

Step 1. Set the allowed scale. A typical dipole-allowed optical transition has $\tau_{\text{E1}}\sim1$ –10 ns (Na D: 16 ns; H 2p: 1.6 ns). This is the benchmark against which "forbidden" is measured.

Step 2. Apply the E2/M1 suppression. A single-step forbidden transition should be slower by $\sim1/(ka_0)^2\sim10^6$ , giving lifetimes of order milliseconds to seconds. The [O I] 630.0 nm auroral line (a forbidden E2/M1 decay of the metastable $^1D_2$ level) has $\tau\approx110$ s — consistent with a strongly suppressed rate. Why some are even longer: the 21 cm hydrogen line is a magnetic-dipole transition between hyperfine levels with a tiny energy (and hence tiny $\omega^3$ in the rate as well as the multipole suppression), giving a lifetime of $\sim10^7$ years — the multipole suppression compounds with the $A\propto\omega^3$ frequency scaling (recall aq-optics-spontaneous-emission).

Step 3. Hydrogen 2s as the cleanest case. The $2s\to1s$ transition is E1-forbidden ( $\Delta l=0$ ) and single-photon-forbidden ( $J$ and parity), so it proceeds by two-photon emission — a second-order process suppressed beyond a single multipole step. Its lifetime is 0.12 s versus the 2p state's 1.6 ns, a ratio of $7.5\times10^7$ .

Step 4. The unifying statement. Every long-lived atomic state exists because the electric dipole approximation forbids its fast decay, forcing it onto a channel suppressed by powers of $ka_0$ (and sometimes by extra factors from multi-photon order). The approximation's success on bright lines and its breakdown on faint ones are two faces of the same small number. Verified: H 2s/2p lifetime ratio $7.5\times10^7$ ; E2/E1 rate ratio $\sim(ka_0)^2\sim4\times10^{-7}$ ; 21 cm M1 lifetime $\sim10^7$ yr order (verify_aq-optics-electric-dipole-approximation.py).

Limiting cases

case	result	meaning
$ka_0\ll1$ (optical, atomic)	E1 dominates; $H'=-\mathbf d\cdot\mathbf E$	the standard regime; corrections $\sim10^{-7}$ in rate, negligible
E1 allowed	rate $\propto\|\langle f\|\mathbf d\|i\rangle\|^2$	bright line, $\tau\sim$ ns; multipole corrections invisible
E1 forbidden	E2/M1 survive at $\sim(ka_0)^2$	metastable state, $\tau\sim$ ms–s (forbidden/nebular lines)
$ka_0\gtrsim1$ (hard X-ray, $\lambda\sim a_0$ )	approximation fails	multipole expansion no longer converges quickly; E2 comparable to E1
nuclear $\gamma$ transitions	$kR\sim1$ common	nuclei are large relative to $\gamma$ wavelengths; higher multipoles routine

Common trapForbidden is set by the dipole operator, not by energy conservation

A common confusion is to think "forbidden" transitions violate some conservation law or are energetically blocked. They do not and are not — the energy is conserved and a lower state exists. What is "forbidden" is only the leading electric-dipole term of the interaction: the dipole matrix element $\langle f|\mathbf r|i\rangle$ vanishes by symmetry. The transition still proceeds through the higher multipole terms that the dipole approximation discarded, at a rate suppressed by $(ka_0)^2\sim10^{-7}$ . So a "forbidden" line is one for which the biggest term in the multipole expansion happens to be zero, throwing the transition onto the next, much weaker term. It is a statement about which approximation applies, not about whether the process can occur.

Worked examples

Example 1 — how good is the dipole approximation for visible light? Compute the smallness parameter $ka_0$ for a hydrogen atom driven by 500 nm light and state the fractional error of the dipole approximation.

Step 1. The wavevector is $k = 2\pi/\lambda = 2\pi/(500\times10^{-9}\ \mathrm m) = 1.26\times10^{7}\ \mathrm{m^{-1}}$ .

Step 2. Multiply by the atomic size: $ka_0 = (1.26\times10^{7})(5.29\times10^{-11}) = 2\pi(0.0529)/500 = \mathbf{6.65\times10^{-4}}$ .

Step 3. The field varies by only $\sim0.07\%$ across the atom, so the amplitude error from keeping only the dipole term is $\sim7\times10^{-4}$ , and the rate error (what an intensity measurement sees) is $(ka_0)^2 \approx \mathbf{4.4\times10^{-7}}$ . The dipole approximation is good to about one part in a million in rate for optical light — which is why it is used almost universally in atomic and molecular spectroscopy, and why the corrections only matter when the leading term vanishes.

Example 2 — the lifetime price of a forbidden transition. Estimate how much longer a forbidden (E2/M1) transition lives than an allowed one, and check against hydrogen's 2s state.

Step 1. The forbidden rate is suppressed relative to allowed by $\sim(ka_0)^2 = 4.4\times10^{-7}$ , so the lifetime is longer by $\sim1/(ka_0)^2 = \mathbf{2.3\times10^6}$ .

Step 2. Apply to an allowed benchmark $\tau_{\text{E1}}\sim1.6$ ns (the H 2p lifetime): a single-step forbidden transition would then live $\sim1.6\ \mathrm{ns}\times2.3\times10^6 \approx \mathbf{4\ ms}$ — already millions of times longer.

Step 3. Hydrogen's actual 2s state lives 0.12 s, a factor $7.5\times10^7$ beyond the 2p's 1.6 ns — even longer than a single E2/M1 step, because the $2s\to1s$ decay is doubly forbidden and proceeds by two-photon emission, a second-order process with extra suppression. The lifetime enhancement of a forbidden line is $\sim10^6$ – $10^8$ , exactly the range of measured metastable lifetimes — from the 110 s [O I] auroral line to the $10^7$ -year 21 cm line — all traceable to the small parameter $ka_0$ that the dipole approximation exploits.

Physics of the portrait

The electric dipole approximation is worth dwelling on as an example of how physics uses the hierarchy of small parameters to make hard problems tractable. The full interaction of an atom with the quantized field is formidable; the observation that $ka_0\sim10^{-3}$ collapses it to a single term, $-\mathbf d\cdot\mathbf E$ , that a student can evaluate by hand. Almost all of atomic spectroscopy, laser physics, and quantum optics is built on this one simplification. But the mark of a deep approximation is that its breakdown is also physics: the terms it discards are not merely errors to be bounded, they are the electric-quadrupole and magnetic-dipole transitions that produce the forbidden spectrum, the metastable states that power optical clocks and lasers, and the 21 cm line that maps the neutral hydrogen of the cosmos. Understanding when to keep only the leading term — and what it means when that term vanishes — is a template for approximation throughout physics.

Origins · 1 minelectric dipole approximation and how this was discovered

electric dipole approximation

the leading long-wavelength term −d·E of the light–matter interaction, valid because ka₀ ≈ 6.65×10⁻⁴ · atomic physics and quantum optics

The approximation that replaces the spatially-varying field of a light wave by a uniform field across the atom, reducing the interaction to $-\mathbf d\cdot\mathbf E$ . Justified by the smallness of $ka_0 = 2\pi a_0/\lambda \sim 10^{-3}$ for atoms in optical light, it underlies essentially all of atomic and laser spectroscopy. Its systematic corrections — electric-quadrupole and magnetic-dipole terms suppressed by $(ka_0)^2 \sim 10^{-7}$ — are the origin of the entire forbidden spectrum and of metastable states. Formalised through Dirac's 1927 radiation theory.

History

1900

— Planck quantizes cavity radiation, setting the stage for light–matter interaction theory

1917

— Einstein's A/B coefficients treat emission and absorption phenomenologically

1927

— Dirac quantizes the electromagnetic field and derives the electric-dipole radiation rate

1930

— Weisskopf and Wigner give the natural linewidth from spontaneous emission

1945

— forbidden (E2/M1) lines established as astrophysical diagnostics of nebulae and aurorae

1951

— Ewen and Purcell detect the 21 cm magnetic-dipole line of neutral hydrogen

Going further

Texts. Foot, Atomic Physics (2005), Ch. 7 (interaction of atoms with radiation) · Loudon, The Quantum Theory of Light, 3rd ed., Ch. 2 & 8 · Sakurai & Napolitano, Modern Quantum Mechanics (radiation/multipole section) · Demtröder, Laser Spectroscopy, Vol. 1, Ch. 2 · Saleh & Teich, Fundamentals of Photonics, Ch. 13.

Provenance (verbatim facts preserved from the base). Equations and numbers are cross-checked against Foot (Ch. 7), Loudon (Ch. 2 & 8), Saleh & Teich (Ch. 13), and Sakurai. The verifier recomputes: $ka_0 = 6.65\times10^{-4}$ at 500 nm; $a_0/\lambda = 1.06\times10^{-4}$ ; the E2/E1 rate suppression $(ka_0)^2 = 4.4\times10^{-7}$ and lifetime enhancement $1/(ka_0)^2 = 2.3\times10^6$ ; the H 2s/2p lifetime ratio $7.5\times10^7$ ; the expansion $e^{i\mathbf k\cdot\mathbf r}=1+i\mathbf k\cdot\mathbf r-\tfrac12(\mathbf k\cdot\mathbf r)^2+\dots$ ; and the length↔velocity gauge identity via $[\mathbf r,H_0]=i\hbar\mathbf p/m$ . The $e^{i\mathbf k\cdot\mathbf r}\approx1$ expansion, minimal coupling, and gauge equivalence are owned by aq-qm-dipole-approximation and aq-qm-semiclassical-atom-radiation, recalled here by name. No non-standard numbers are asserted.

Builds on: Selection Rules for Atomic Transitions · Einstein A and B Coefficients · Fermi's Golden Rule. Leads to: Transition Dipole Moment · Oscillator Strength · Radiative Lifetime and Natural Linewidth.

Does the dipole approximation mean the magnetic field of the light does nothing?

For the leading interaction, essentially yes. The magnetic force on the electron is smaller than the electric force by a factor $v/c$ , where $v\sim\alpha c$ is the electron's orbital speed — so it is of the same order as the neglected multipole terms, $\sim ka_0$ . Keeping only the electric-dipole term therefore consistently drops both the field's spatial variation and its magnetic force. The magnetic field reappears exactly where the discarded terms do: it drives magnetic-dipole (M1) transitions, which are the same order, $(ka_0)^2$ in rate, as electric-quadrupole ones. So "the magnetic field does nothing" is true only at leading order; at the next order it produces the M1 forbidden lines, like the 21 cm line.

When does the electric dipole approximation actually break down?

When $ka_0$ stops being small — that is, when the wavelength becomes comparable to the atomic size. For valence-electron transitions in the optical and even ultraviolet, $ka_0\sim10^{-3}$ – $10^{-2}$ and the approximation is superb. It only begins to fail for hard X-rays, where $\lambda\sim$ ångströms approaches $a_0$ and $ka_0\sim0.1$ –1, so quadrupole and higher transitions become genuinely competitive — which is why X-ray spectroscopy of inner shells routinely sees "forbidden" lines with appreciable strength. In nuclei the effect is even more dramatic: nuclear sizes are not negligible compared to $\gamma$ -ray wavelengths, so $kR\sim1$ and high multipole transitions are the norm, not the exception.

Why does keeping only the "1" in the expansion give a dipole and not something else?

Because the " $1$ " makes the interaction $-\mathbf d\cdot\mathbf E$ , which is literally the energy of an electric dipole $\mathbf d=-e\mathbf r$ in a uniform field $\mathbf E$ . Physically, replacing $e^{i\mathbf k\cdot\mathbf r}$ by 1 says the field has the same value everywhere across the atom, i.e. it is uniform; a charge distribution in a uniform field feels only its dipole moment (a uniform field exerts no net force on a neutral atom, only a torque/energy through the dipole). The next term, $i\mathbf k\cdot\mathbf r$ , encodes the first spatial gradient of the field, which is what a quadrupole (charge distribution's second moment) and the magnetic field couple to. So the multipole names are not arbitrary — each term of the expansion picks out the corresponding moment of the charge distribution.

If forbidden transitions are 10⁷ times weaker, why do we ever see them?

Because there is a lot of gas and a lot of time, and because the competition matters more than the absolute rate. In a dilute nebula or the upper atmosphere, a metastable atom has seconds to decay radiatively before a collision disturbs it, so its slow forbidden decay actually happens rather than being quenched — and with $10^{50}$ atoms in a nebula, even a $10^{-7}$ rate produces a bright line. In an atomic clock, the weakness is the point: a transition that takes seconds has a linewidth of millihertz, which is exactly the ultra-narrow reference a clock needs. And the 21 cm line, hopeless for a single atom, is detectable because the Galaxy holds an enormous column of neutral hydrogen. Weak per atom does not mean invisible in aggregate — and slow is often precisely what makes a transition useful.

How can two different Hamiltonians (length and velocity gauge) give the same physics?

Because they are related by a gauge transformation, which cannot change any observable. The length gauge writes the interaction as $-\mathbf d\cdot\mathbf E$ ; the velocity gauge writes it as $-(e/m)\mathbf A\cdot\mathbf p$ . These look different, but the transition matrix elements are connected by the operator identity $\langle f|\mathbf p|i\rangle = im\omega_{fi}\langle f|\mathbf r|i\rangle$ , which follows from $[\mathbf r, H_0] = i\hbar\mathbf p/m$ . Substituting converts one form into the other and yields identical rates for the E1 transition. (Subtleties arise if you truncate the two expansions inconsistently or work off-resonance, which is a known source of apparent gauge discrepancies — but the exact, consistently-computed rates always agree.) It is a healthy check on any calculation to verify both gauges give the same answer.

Is the electric dipole approximation the same as "long-wavelength" or "point-dipole"?

They are closely related names for the same idea seen from different angles. "Long-wavelength" refers to the condition — the wavelength is long compared to the atom, $\lambda\gg a_0$ , i.e. $ka_0\ll1$ — which is what makes the field uniform across the atom. "Electric dipole" refers to the result — with a uniform field, only the dipole moment $\mathbf d=-e\mathbf r$ of the charge distribution couples, giving $H'=-\mathbf d\cdot\mathbf E$ . "Point-dipole" is the mental picture: because the atom is so small compared to the wavelength, it acts like a single oscillating dipole located at a point, radiating and absorbing the dipole radiation pattern. All three describe the same approximation; the first names its validity condition, the second the surviving interaction, and the third the physical caricature. They coincide precisely because $ka_0\sim10^{-3}$ is small.