Transition Dipole Moment

Ask why an atom radiates at all and the answer is not in the atom's energy levels but in the bridge between two of them. This article consumes one result from quantum mechanics — first-order time-dependent perturbation theory, whose Fermi golden rule says a transition rate is set by the squared matrix element of the perturbation — and one from the electromagnetics of light–matter coupling: in the electric dipole approximation the perturbation is $H'=-\boldsymbol{\mu}\cdot\mathbf{E}$ , with $\boldsymbol{\mu}=-e\mathbf{r}$ the dipole operator. Put the two together and a single object falls out. The rate of $|i\rangle\to|f\rangle$ is proportional not to the operator $\boldsymbol{\mu}$ itself but to its off-diagonal matrix element between the two stationary states, $\boldsymbol{\mu}_{fi}=\langle f|\,\boldsymbol{\mu}\,|i\rangle$ . That complex vector is the transition dipole moment. It is a strange hybrid: not a property of one state but of a pair — the quantum stand-in for the amplitude of a classical radiating dipole. Unlike a permanent moment $\langle i|\boldsymbol{\mu}|i\rangle$ , which many atoms lack by symmetry, it connects two levels and oscillates at the Bohr frequency $\omega_{fi}=(E_f-E_i)/\hbar$ . Its magnitude squared, $|\boldsymbol{\mu}_{fi}|^2$ , is the one number carrying every dynamical fact about an electric-dipole transition: it fixes the Einstein $A$ coefficient (spontaneous emission), the $B$ coefficients (absorption and stimulated emission), the oscillator strength, the radiative lifetime, the natural linewidth and the Rabi frequency. Selection rules are just the statement of which of these matrix elements vanish by parity and angular-momentum symmetry — so a "forbidden" line is simply a $\boldsymbol{\mu}_{fi}$ that is exactly zero.

Key takeaways

The transition dipole moment $\boldsymbol{\mu}_{fi}=\langle f|\,-e\,\mathbf{r}\,|i\rangle$ is the matrix element of the electric-dipole operator between initial and final states — an off-diagonal object belonging to a pair of levels, not one.
Its magnitude squared $|\boldsymbol{\mu}_{fi}|^2$ sets every E1 radiative rate: the Einstein $A$ and $B$ coefficients, the oscillator strength, the radiative lifetime and the natural linewidth.
When $\boldsymbol{\mu}_{fi}=0$ the transition is dipole-forbidden — this vanishing, forced by parity and angular momentum, is the electric-dipole selection rule.
Allowed atomic transitions give $|\boldsymbol{\mu}_{fi}|\sim ea_0\approx 8.478\times10^{-30}\ \mathrm{C\,m}\approx 2.54\ \mathrm{D}$ ; Rydberg transitions scale as $n^2a_0$ and become enormous.
Because $\mathbf{r}$ is a rank-1 operator, the Wigner–Eckart theorem splits $\boldsymbol{\mu}_{fi}$ into a geometric Clebsch–Gordan part (polarisation and $\Delta m$ ) and a reduced radial matrix element carrying the dynamical scale.
Key idea: every electric-dipole radiative rate — spontaneous, stimulated, absorptive — is proportional to $|\langle f|-e\mathbf{r}|i\rangle|^2$ ; this off-diagonal matrix element is the transition dipole moment, and its vanishing is exactly a selection rule.

The core insight

Every electric-dipole radiative rate — spontaneous, stimulated and absorptive — is proportional to $|\langle f|-e\mathbf{r}|i\rangle|^2$ ; this off-diagonal matrix element is the transition dipole moment, and its vanishing by symmetry is exactly a selection rule.

Where this lives today

Every time an atomic clock ticks, a quantum computer runs a gate, or an astronomer reads a stellar spectrum, a transition dipole moment is doing the work. In neutral-atom quantum processors, arrays of rubidium or caesium atoms in optical tweezers are entangled by exciting them to Rydberg states whose transition dipoles — thousands of debye, per Worked Example 3 — let one excited atom shift its neighbours' levels enough to block a second excitation, the "Rydberg blockade" behind platforms from QuEra and Pasqal. In metrology the same matrix element sets the natural linewidth $\Gamma=A_{21}$ that limits a clock's stability, which is why optical-clock designers chase weakly allowed lines with tiny $|\boldsymbol{\mu}_{fi}|$ — the strontium and ytterbium clock transitions are nominally forbidden, giving millihertz linewidths. And in astrophysics the relative intensities of a multiplet, read straight from the Wigner–Eckart factorisation into Clebsch–Gordan coefficients, let spectroscopists infer the temperature, density, field and abundances of a plasma they will never touch.

The physical picture

A stationary state does not radiate: its charge density $-e|\psi_i|^2$ is time-independent — a frozen cloud with no oscillating current to feed a wave. The radiation comes instead from a superposition. Prepare an atom in a coherent mixture of two eigenstates,

\psi(\mathbf{r},t)=c_i\,\psi_i(\mathbf{r})\,e^{-iE_it/\hbar}+c_f\,\psi_f(\mathbf{r})\,e^{-iE_ft/\hbar},

and its charge density $-e|\psi|^2$ acquires a cross term $-2e\,\mathrm{Re}\big[c_i^*c_f\,\psi_i^*\psi_f\,e^{i\omega_{fi}t}\big]$ that sloshes at the Bohr frequency $\omega_{fi}$ . The cloud's centroid oscillates — a genuine dipole whose amplitude is precisely $\boldsymbol{\mu}_{fi}$ , which is why an atom mid-transition radiates like a miniature Hertzian antenna. The first figure makes this concrete: two symmetric clouds that each sit still, and the beating superposition whose charge centre swings.

Figure

An s-state and a p-state are each spatially symmetric and radiate nothing on their own, but their superposition has a charge centroid that oscillates at the Bohr frequency ω_fi = (E_f − E_i)/ℏ — a microscopic dipole whose amplitude is exactly the transition dipole moment μ_fi

Why does the integral $\int\psi_f^*(-e\mathbf{r})\psi_i\,\mathrm{d}^3r$ so often vanish? Because it is an overlap weighted by position, and position is odd under inversion. If $\psi_f^*\psi_i$ is even, the odd weighting by $\mathbf{r}$ makes the integrand odd and it vanishes identically; a non-zero result demands two states of opposite parity. That is the parity selection rule. The vector nature matters too: the Cartesian components of $\boldsymbol{\mu}_{fi}$ couple to different polarisations, so a transition's geometry is written into the direction of its dipole, not only its length. The second figure computes the piece carrying the magnitude — the hydrogen radial overlap that fixes the scale at a Bohr radius.

Figure

The hydrogen 2p → 1s radial overlap integrand r³ R₁₀ R₂₁ plotted against r in Bohr radii: the shaded area is the radial matrix element ⟨r⟩ = 1.29 a₀, and multiplied by e this fixes the transition-dipole scale near ea₀ = 8.478×10⁻³⁰ C·m = 2.54 D

The core equations

The definition is the whole idea in one line: sandwich the electric-dipole operator between the final and initial states and integrate. Everything downstream is a consequence or a computation of it. Because the two states differ, it is an off-diagonal element — a permanent moment would have $f=i$ .

\boldsymbol{\mu}_{fi}=\langle f|\,\boldsymbol{\mu}\,|i\rangle=\int \psi_f^*(\mathbf{r})\,(-e\mathbf{r})\,\psi_i(\mathbf{r})\,\mathrm{d}^3r \qquad (1)

where: $\boldsymbol{\mu}_{fi}$ — transition dipole moment (complex vector); $\boldsymbol{\mu}=-e\mathbf{r}$ — electric-dipole operator for one electron; $e=1.602\times10^{-19}\ \mathrm{C}$ ; $\psi_i,\psi_f$ — initial and final stationary states. Read it as a position-weighted overlap of two clouds: large when mutually displaced, zero when symmetry forbids it. Why this matrix element governs radiation is the coupling itself. In the E1 approximation the atom feels the light through the energy of its dipole in the field, so the perturbation driving transitions is $\boldsymbol{\mu}\cdot\mathbf{E}$ .

H'(t)=-\boldsymbol{\mu}\cdot\mathbf{E}(t) \qquad (2)

where: $H'$ — perturbing Hamiltonian in the E1 approximation; $\mathbf{E}(t)$ — external field at the atom (uniform over its size). Because $H'$ is linear in $\boldsymbol{\mu}$ , first-order perturbation theory picks out exactly $\boldsymbol{\mu}_{fi}$ . The headline application is spontaneous emission. Summing the golden-rule rate over every vacuum photon mode (Derivation 2) gives the Einstein $A$ coefficient — an isolated excited atom's decay probability per unit time — scaling as the cube of frequency and the square of the dipole.

A_{21}=\frac{\omega_0^3\,|\boldsymbol{\mu}_{21}|^2}{3\pi\varepsilon_0\hbar c^3} \qquad (3)

where: $A_{21}$ — spontaneous decay rate $2\to1$ ( $\mathrm{s^{-1}}$ ); $\omega_0=(E_2-E_1)/\hbar$ ; $\varepsilon_0$ — vacuum permittivity; $c$ — speed of light. The radiative lifetime is $\tau=1/\sum_f A_{2f}$ , so a larger $|\boldsymbol{\mu}_{21}|$ means a shorter-lived state and a broader natural line. The stimulated processes — absorption and stimulated emission — are governed by the same squared dipole through the Einstein $B$ coefficient. Because $A$ and $B$ carry the identical $|\boldsymbol{\mu}|^2$ , their ratio is independent of the transition, fixed by constants and frequency alone — the signature of detailed balance with the Planck spectrum.

B_{12}=\frac{\pi\,|\boldsymbol{\mu}_{12}|^2}{3\varepsilon_0\hbar^2},\qquad \frac{A_{21}}{B_{12}}=\frac{\hbar\omega_0^3}{\pi^2 c^3} \qquad (4)

where: $B_{12}$ — stimulated absorption/emission coefficient (spectral-energy-density convention); the ratio $A/B$ contains no dipole at all. That cancellation is how Einstein deduced the existence of spontaneous emission in 1916 without knowing a single matrix element. Spectroscopists often quote the dimensionless oscillator strength $f_{fi}$ instead of $|\boldsymbol{\mu}_{fi}|^2$ — the same information rescaled so a classical single-electron oscillator has $f=1$ , obeying the Thomas–Reiche–Kuhn sum rule $\sum_f f_{fi}=Z$ .

f_{fi}=\frac{2m_e\omega_{fi}}{3\hbar e^2}\,|\boldsymbol{\mu}_{fi}|^2 \qquad (5)

where: $f_{fi}$ — oscillator strength (dimensionless); $m_e$ — electron mass; $\omega_{fi}$ — transition angular frequency. An allowed optical line has $f\sim0.1$ – $1$ ; a weak line $f\ll1$ . Finally, the vector $\boldsymbol{\mu}_{fi}$ factorises. Because $\mathbf{r}$ is a rank-1 operator, the Wigner–Eckart theorem separates its matrix element into a Clebsch–Gordan coefficient — pure geometry, fixing $\Delta m$ and polarisation — times a reduced matrix element holding all the dynamical strength.

\langle j'm'|\,T^{(1)}_q\,|jm\rangle=\langle jm;1q|j'm'\rangle\;\frac{\langle j'\|T^{(1)}\|j\rangle}{\sqrt{2j'+1}},\qquad \mu_{\pm}=\mp\tfrac{1}{\sqrt2}(\mu_x\pm i\mu_y),\ \ \mu_0=\mu_z \qquad (6)

where: $T^{(1)}_q$ — spherical component $q=0,\pm1$ of the dipole; $\langle jm;1q|j'm'\rangle$ — Clebsch–Gordan coefficient (geometry); $\langle j'\|T^{(1)}\|j\rangle$ — reduced matrix element (dynamical scale). Here $\mu_0$ drives $\pi$ ( $\Delta m=0$ ) transitions with $z$ -polarised light, $\mu_\pm$ drive $\sigma^\pm$ ( $\Delta m=\pm1$ ) with circular polarisation. For scale, the natural atomic dipole unit is $ea_0$ and the spectroscopic unit is the debye:

1\ \mathrm{D}=3.336\times10^{-30}\ \mathrm{C\,m},\qquad ea_0=8.478\times10^{-30}\ \mathrm{C\,m}=2.542\ \mathrm{D} \qquad (7)

where: $a_0=5.292\times10^{-11}\ \mathrm{m}$ — Bohr radius. Allowed atomic transitions cluster near a few debye; Rydberg and molecular transitions can be far larger.

Derivation AWhy the rate is quadratic in the transition dipole

The goal is to show from first principles that the rate of a driven transition is proportional to $|\boldsymbol{\mu}_{fi}|^2$ and not to $\boldsymbol{\mu}_{fi}$ itself — the single fact that makes the squared magnitude, not the vector, the physically meaningful quantity. We take a two-level atom driven by $\mathbf{E}(t)=\mathbf{E}_0\cos\omega t$ under $H'=-\boldsymbol{\mu}\cdot\mathbf{E}$ , run first-order perturbation theory, square the amplitude, and take the long-time limit to Fermi's golden rule. The squared modulus is the crux — it is why probability, not amplitude, is what nature reports.

Step 1. Write the first-order amplitude for finding the atom in $|f\rangle$ : $c_f(t)=-\tfrac{i}{\hbar}\int_0^t\langle f|H'(t')|i\rangle\,e^{i\omega_{fi}t'}\,\mathrm{d}t'$ , with $\langle f|H'|i\rangle=-\boldsymbol{\mu}_{fi}\cdot\mathbf{E}_0\cos\omega t'$ . Why: the amplitude is the time integral of the coupling weighted by the free phase, and the coupling is linear in $\boldsymbol{\mu}_{fi}$ , so the amplitude inherits that single power.

Step 2. Pull the constant vector out: $c_f(t)=\tfrac{i}{\hbar}\,\boldsymbol{\mu}_{fi}\cdot\mathbf{E}_0\int_0^t\cos\omega t'\,e^{i\omega_{fi}t'}\,\mathrm{d}t'$ , keeping only the near-resonant term $e^{i(\omega_{fi}-\omega)t'}$ (rotating-wave approximation). Why: $\boldsymbol{\mu}_{fi}$ is time-independent, so it factors out — atomic structure and field time-dependence separate cleanly.

Step 3. Square the modulus: $|c_f(t)|^2=\dfrac{|\boldsymbol{\mu}_{fi}\cdot\mathbf{E}_0|^2}{4\hbar^2}\,\dfrac{\sin^2[(\omega_{fi}-\omega)t/2]}{[(\omega_{fi}-\omega)/2]^2}$ . Why: the probability is $|c_f|^2$ , and squaring a quantity linear in $\boldsymbol{\mu}_{fi}$ produces the quadratic $|\boldsymbol{\mu}_{fi}|^2$ — the Born rule is where the dipole enters as a square.

Step 4. Take $t\to\infty$ : the $\mathrm{sinc}^2$ sharpens into $2\pi t\,\delta(\omega_{fi}-\omega)$ , so the probability grows linearly in $t$ and the rate is constant: $W_{i\to f}=\dfrac{\pi}{2\hbar^2}\,|\boldsymbol{\mu}_{fi}\cdot\mathbf{E}_0|^2\,\delta(\omega-\omega_{fi})$ . Why: a constant rate — Fermi's golden rule — needs the probability linear in time; broadening later replaces the delta by a line shape.

Go deeper · Derivation B · 1 minSpontaneous emission from the vacuum modes

Derivation BSpontaneous emission from the vacuum modes

The goal is to build the Einstein $A$ coefficient of equation (3) by summing the emission rate over the continuum of empty vacuum photon modes — the QED-flavoured route showing spontaneous emission is really emission stimulated by vacuum fluctuations. We quantise the field, apply the golden rule to one mode, sum over polarisations and directions, then fold in the mode density. The payoff: the $\omega_0^3$ scaling emerges from the geometry of the mode sum, not from the atom.

Step 1. For a single vacuum mode $\mathbf{k}$ with polarisation $\hat{\boldsymbol{\epsilon}}$ , the golden-rule emission rate carries $|\langle f,1_\mathbf{k}|H'|i,0\rangle|^2\propto|\boldsymbol{\mu}_{21}\cdot\hat{\boldsymbol{\epsilon}}|^2$ , with the zero-point amplitude $\sqrt{\hbar\omega/2\varepsilon_0V}$ setting the scale. Why: even with no photons present the field has a non-zero matrix element for creating one, so the atom emits into the vacuum — again quadratic in the dipole, projected on the mode polarisation.

Step 2. Sum over polarisations and integrate over solid angle: $\sum_{\mathrm{pol}}\int|\boldsymbol{\mu}_{21}\cdot\hat{\boldsymbol{\epsilon}}|^2\,\mathrm{d}\Omega=\tfrac{8\pi}{3}\,|\boldsymbol{\mu}_{21}|^2$ . Why: averaging over all directions and both transverse polarisations replaces the direction-dependent coupling by the rotationally invariant $|\boldsymbol{\mu}_{21}|^2$ with a geometric factor $8\pi/3$ .

Step 3. Multiply by the mode density $\rho(\omega_0)=\omega_0^2/(\pi^2c^3)$ and collect constants to obtain $A_{21}=\dfrac{\omega_0^3\,|\boldsymbol{\mu}_{21}|^2}{3\pi\varepsilon_0\hbar c^3}$ . Why: mode number rises as $\omega^2$ and per-mode energy adds one more power, so the spontaneous rate scales as $\omega_0^3$ — why blue/UV/X-ray lines decay so much faster than low-frequency ones.

Go deeper · Derivation C · 1 minAngular and radial factorisation for a one-electron atom

Derivation CAngular and radial factorisation for a one-electron atom

The goal is to split $\boldsymbol{\mu}_{fi}$ for a hydrogenic atom into an angular integral that decides whether the transition is allowed at all and a radial integral that sets how strong it is — the concrete face of the Wigner–Eckart theorem. We write $\psi_{nlm}=R_{nl}(r)Y_{lm}(\theta,\phi)$ , expand $\mathbf{r}=r\hat{\mathbf{r}}$ in rank-1 spherical harmonics, and separate the $r$ -integral from the angular one. The separation is exact, and it is where the selection rules physically live.

Step 1. For the $z$ -component, factor the matrix element: $\langle n'l'm'|z|nlm\rangle=\Big(\int_0^\infty R_{n'l'}\,r\,R_{nl}\,r^2\mathrm{d}r\Big)\Big(\int Y_{l'm'}^*\cos\theta\,Y_{lm}\,\mathrm{d}\Omega\Big)$ . Why: because $z=r\cos\theta$ splits into a radial factor and an angular factor, the six-dimensional integral collapses into a product of a radial and an angular integral.

Step 2. Evaluate the angular integral. Since $\cos\theta\propto Y_{10}$ , the triangle rule forces $\int Y_{l'm'}^*\cos\theta\,Y_{lm}\,\mathrm{d}\Omega\neq0\ \Rightarrow\ l'=l\pm1,\ m=m'$ . Why: this is the selection rule in the flesh — geometry decides which lines exist, so a $\Delta l=0$ or $\Delta l=2$ transition is E1-forbidden however large the radial overlap.

Step 3. Evaluate the radial integral $\langle r\rangle_{n'l',nl}=\int_0^\infty R_{n'l'}R_{nl}\,r^3\,\mathrm{d}r$ , which for hydrogen 2p → 1s equals $1.29\,a_0$ (the shaded area in the second figure). Why: this dynamical factor carries the magnitude — order $a_0$ for low-lying states but $n^2a_0$ between Rydberg levels, explaining both the debye scale of ordinary lines and the giant Rydberg dipoles.

Limiting cases

Case	Result	Meaning
Forbidden transition	$\boldsymbol{\mu}_{fi}=0$	Parity, $\Delta l\neq\pm1$ , or a spin change forces the matrix element to vanish; the line is E1-forbidden and proceeds, if at all, through far weaker magnetic-dipole (M1) or electric-quadrupole (E2) channels.
Hydrogenic / allowed	$\vert\boldsymbol{\mu}_{fi}\vert\sim ea_0$	The length scale of $\mathbf{r}$ is the Bohr radius, so $\vert\boldsymbol{\mu}_{fi}\vert\approx8.5\times10^{-30}\ \mathrm{C\,m}\approx2.5\ \mathrm{D}$ — the natural order of magnitude of an optical atomic transition.
Rydberg limit	$\vert\boldsymbol{\mu}_{fi}\vert\propto n^2$	Between adjacent high- $n$ levels the orbit scales as $n^2a_0$ , giving huge transition dipoles and the strong microwave/THz coupling exploited in Rydberg quantum technology.
Classical correspondence	$f_{fi}\to1$	When the oscillator strength approaches unity the quantum transition mimics one classical Lorentz oscillator; the Thomas–Reiche–Kuhn sum rule guarantees $\sum_f f_{fi}=Z$ .

Common trapA transition dipole moment is not a permanent dipole moment

The common confusion is to picture $\boldsymbol{\mu}_{fi}$ as a little static arrow the atom carries. It is not. A permanent dipole is the diagonal value $\langle i|\boldsymbol{\mu}|i\rangle$ , which vanishes for any parity eigenstate — most atoms have none. The transition dipole is the off-diagonal element $\langle f|\boldsymbol{\mu}|i\rangle$ between two levels, and it can be large precisely when both permanent moments are zero, because the states have opposite parity. Two further traps: $\boldsymbol{\mu}_{fi}$ is a complex vector, so a rate depends on the full $|\boldsymbol{\mu}_{fi}|^2=\sum_q|\mu_q|^2$ , not one Cartesian component; and "forbidden" means E1-forbidden, not impossible — residual M1/E2 pathways are simply orders of magnitude weaker.

Worked examples

1 · How long does an excited atom live? (confidence builder)

An allowed optical transition sits at $\lambda=589\ \mathrm{nm}$ (the sodium-D scale) with a transition dipole of order $|\boldsymbol{\mu}_{21}|=ea_0=8.478\times10^{-30}\ \mathrm{C\,m}$ . Find the spontaneous rate $A_{21}$ , the radiative lifetime, and the oscillator strength.

Step 1. Get the angular frequency: $\nu=c/\lambda=5.09\times10^{14}\ \mathrm{Hz}$ , so $\omega_0=2\pi\nu=3.198\times10^{15}\ \mathrm{rad\,s^{-1}}$ , and $\omega_0^3=3.27\times10^{46}\ \mathrm{rad^3\,s^{-3}}$ .

Step 2. Assemble equation (3). Numerator $\omega_0^3|\boldsymbol{\mu}_{21}|^2=(3.27\times10^{46})(8.478\times10^{-30})^2=2.35\times10^{-12}$ ; denominator $3\pi\varepsilon_0\hbar c^3=2.37\times10^{-19}$ . Dividing gives $A_{21}=\mathbf{9.9\times10^{6}\ s^{-1}}$ .

Step 3. Invert for the lifetime: $\tau=1/A_{21}=\mathbf{1.0\times10^{-7}\ s\approx100\ ns}$ . Real alkali lines decay in a few to a few tens of nanoseconds because their radial matrix elements are a few $a_0$ , not one — a factor of a few in $|\boldsymbol{\mu}|$ becomes an order of magnitude in $A\propto|\boldsymbol{\mu}|^2$ .

Step 4. Cross-check with the oscillator strength, equation (5): $f=\dfrac{2m_e\omega_0}{3\hbar e^2}|\boldsymbol{\mu}_{21}|^2=\mathbf{0.052}$ . A sub-unity value for a single $ea_0$ dipole is right — the full sodium doublet sums to $f\approx1$ once the larger real dipole and both D-lines enter, consistent with the sum rule.

2 · The Rabi frequency in a real laser field

The same transition ( $|\boldsymbol{\mu}|=ea_0$ ) is now driven by a laser of intensity $I=100\ \mathrm{W\,cm^{-2}}=1.0\times10^6\ \mathrm{W\,m^{-2}}$ . How fast does the atom cycle between the two levels?

Step 1. Convert intensity to field amplitude via $I=\tfrac12 c\varepsilon_0 E_0^2$ , so $E_0=\sqrt{2I/(c\varepsilon_0)}=\mathbf{2.74\times10^{4}\ V\,m^{-1}}$ .

Step 2. The Rabi frequency is $\Omega=\boldsymbol{\mu}_{fi}\cdot\mathbf{E}_0/\hbar=(8.478\times10^{-30})(2.74\times10^4)/(1.055\times10^{-34})=\mathbf{2.2\times10^{9}\ rad\,s^{-1}}$ , i.e. $\Omega/2\pi\approx\mathbf{350\ MHz}$ .

Step 3. Interpret it. A full absorption–emission cycle takes $2\pi/\Omega\approx2.9\ \mathrm{ns}$ — far shorter than the $\sim100\ \mathrm{ns}$ spontaneous lifetime, so coherent Rabi flopping outpaces spontaneous decay and clean $\pi$ -pulses are possible. Since $\Omega\propto\boldsymbol{\mu}$ , a larger transition dipole gives faster gates — which is why they are prized in quantum control.

3 · Why Rydberg dipoles are gigantic

Estimate $|\boldsymbol{\mu}_{fi}|$ between two neighbouring Rydberg levels near $n=50$ , and compare it to a ground-state optical line.

Step 1. The orbit radius scales as $n^2a_0$ , so the transition dipole between adjacent Rydberg states scales as $|\boldsymbol{\mu}_{fi}|\sim n^2\,ea_0$ . For $n=50$ that is $(50)^2=2500$ times the atomic scale, i.e. $|\boldsymbol{\mu}_{fi}|\sim2500\times2.54\ \mathrm{D}\approx\mathbf{6\times10^{3}\ D}$ — four thousand times larger than a ground-state line.

Step 2. Consequence: since coupling goes as $\boldsymbol{\mu}\cdot\mathbf{E}$ , a Rydberg atom reaches the same Rabi frequency at a field thousands of times weaker, and Rydberg–Rydberg interactions ( $\propto\mu^2\propto n^4$ ) blockade a whole micron-scale ensemble. This $n^2$ growth of the transition dipole underlies Rydberg quantum computing and single-photon nonlinear optics.

Physics of the portrait

The transition dipole moment is not a footnote to quantum mechanics — it is one of the objects out of which quantum mechanics was built.

Origins · 1 minWerner Heisenberg and how this was discovered

Werner Heisenberg

1901–1976 · matrix mechanics and the origin of the transition amplitude

In the summer of 1925, recovering from hay fever on the island of Helgoland, Werner Heisenberg resolved to build a mechanics from observable quantities alone. The old Bohr model spoke of electron orbits no one could see; what one could measure were the frequencies and intensities of spectral lines — and each line belonged to a pair of states, not to one. Heisenberg therefore replaced the classical position $x(t)$ by an array of transition amplitudes $x_{nm}$ , one per pair of levels, oscillating at $\omega_{nm}$ . Max Born recognised these arrays as matrices, and within months Born, Heisenberg and Jordan had a complete matrix mechanics. The crucial claim was that a line's intensity is proportional to $|x_{nm}|^2$ — which, multiplied by charge, is exactly $|\boldsymbol{\mu}_{fi}|^2$ . The transition dipole moment is thus the direct descendant of Heisenberg's original observable: the off-diagonal matrix element whose square is a measurable line strength. Dirac completed the picture in 1927 by quantising the field and showing these same elements drive spontaneous emission, and the Wigner–Eckart theorem soon exposed their geometry. The whole edifice of selection rules, oscillator strengths and lifetimes rests on the object Heisenberg put at the centre of physics that summer.

History

1913

— Bohr quantises atomic energy levels and their transition frequencies

1916

— Einstein introduces the A and B coefficients for spontaneous and stimulated radiation

1925

— Heisenberg recasts position as transition amplitudes whose squares give line intensities

1927

— Dirac quantises the radiation field and derives spontaneous emission from the dipole matrix element

1930

— the Wigner–Eckart theorem factorises matrix elements into geometry and a reduced element

Going further

Texts. Foot, Atomic Physics, ch. 2 and 7 (E1 interaction, selection rules, lifetimes) · Loudon, The Quantum Theory of Light, 3rd ed., ch. 1–2 (golden rule and Einstein coefficients) · Sakurai and Napolitano, Modern Quantum Mechanics, ch. 5 (Wigner–Eckart theorem) · Demtröder, Laser Spectroscopy, ch. 2 · Saleh and Teich, Fundamentals of Photonics, ch. 13 · Hecht, Optics, ch. 3 (radiation from oscillating dipoles). Key numerical anchors used throughout this article: the golden-rule transition rate is $\propto|\boldsymbol{\mu}_{fi}\cdot\mathbf{E}_0|^2$ ; the spontaneous-emission coefficient is $A_{21}=\omega_0^3|\boldsymbol{\mu}_{21}|^2/(3\pi\varepsilon_0\hbar c^3)$ ; the atomic dipole scale is $ea_0=8.478\times10^{-30}\ \mathrm{C\,m}=2.542\ \mathrm{D}$ , obtained from $a_0=5.292\times10^{-11}\ \mathrm{m}$ , $e=1.602\times10^{-19}\ \mathrm{C}$ , and $1\ \mathrm{D}=3.336\times10^{-30}\ \mathrm{C\,m}$ ; the worked results are $A_{21}\approx9.9\times10^6\ \mathrm{s^{-1}}$ , $\tau\approx100\ \mathrm{ns}$ , $\Omega/2\pi\approx350\ \mathrm{MHz}$ ; the hydrogen 2p → 1s radial matrix element is $\langle r\rangle_{2p,1s}=1.29\,a_0$ ; and the $\Delta l=\pm1$ selection rule follows from the $Y_{lm}$ triangle rule. All constants are CODATA values.

Prerequisites. Electric Dipole Approximation · Resonant Absorption. Leads to. Einstein A and B Coefficients · Oscillator Strength · Selection Rules for Dipole Transitions · Radiative Lifetime and Natural Linewidth.

Ask the tutor: Want to work through the Wigner–Eckart factorisation of a real multiplet — say the sodium D doublet or a Zeeman pattern — and predict the relative line intensities from Clebsch–Gordan coefficients alone?

Is the transition dipole moment a real physical dipole I could measure with a voltmeter?

Not directly — you cannot put leads on an atom and read off $\boldsymbol{\mu}_{fi}$ , because it is not a static charge separation. It is the amplitude of a time-dependent charge oscillation that exists only in a superposition of the two states. What you measure instead are the consequences of $|\boldsymbol{\mu}_{fi}|^2$ : the decay rate (hence lifetime), the absorption cross-section, the oscillator strength, the Rabi frequency in a known field. Any one, inverted through its formula, hands you back $|\boldsymbol{\mu}_{fi}|$ . So it is thoroughly physical and precisely measurable — through radiation, not electrostatics.

How can two states with no permanent dipole still have a large transition dipole?

Because they are completely different objects. A permanent dipole is $\langle i|\boldsymbol{\mu}|i\rangle$ , the average charge position within a single state; for any state of definite parity this is exactly zero, because the cloud is symmetric. The transition dipole is $\langle f|\boldsymbol{\mu}|i\rangle$ , involving the product $\psi_f^*\psi_i$ of two different clouds. If those clouds have opposite parity — an s-state and a p-state — their product is odd, matches the oddness of $\mathbf{r}$ , and the integral is large. So the very symmetry that kills both permanent moments permits a strong transition between them.

Why does spontaneous emission scale as ω³ while absorption does not?

The $\omega^3$ in the Einstein $A$ coefficient comes from the vacuum, not the atom. Spontaneous emission is emission into all the empty photon modes: two powers of frequency come from the mode density ( $\propto\omega^2$ ) and one more from the zero-point energy per mode ( $\propto\omega$ ). Absorption and stimulated emission respond to an externally supplied field whose mode structure you set, so the $B$ coefficient has no such vacuum factor — which is why $A/B\propto\omega^3$ , and why UV and X-ray lines decay far faster than microwave ones for the same dipole.

What is the difference between the transition dipole moment and the oscillator strength?

They carry the same physics in different clothing. The transition dipole $|\boldsymbol{\mu}_{fi}|^2$ is the fundamental object with units of (charge × length)². The oscillator strength $f_{fi}$ is a dimensionless rescaling, equation (5), defined so a single classical electron oscillator has $f=1$ . Its advantage is the Thomas–Reiche–Kuhn sum rule: strengths from a given level add to the electron number, $\sum_f f_{fi}=Z$ — an instant sanity check. Spectroscopists often quote $f$ ; atomic physicists quote $|\boldsymbol{\mu}_{fi}|$ or its reduced matrix element. Converting needs only the transition frequency.

If a line is "forbidden", does the atom simply never make that transition?

No — "forbidden" is shorthand for "electric-dipole-forbidden", meaning $\boldsymbol{\mu}_{fi}=0$ so the leading E1 channel is off. The atom can still decay through weaker interactions: magnetic-dipole (M1) or electric-quadrupole (E2) radiation from the next multipole terms. These are typically $10^{-5}$ to $10^{-8}$ times slower, giving metastable states that live from milliseconds to minutes rather than nanoseconds. Far from useless, these long-lived transitions are the backbone of optical atomic clocks and of the green and red auroral lines of atomic oxygen.

Why does the rate depend on the whole vector magnitude and not one component?

Because a real atom, or ensemble, is not aligned to your axes. The transition dipole is a vector with three spherical components $\mu_0,\mu_\pm$ , each coupling to a different polarisation. A rate summed over final magnetic sublevels, or averaged over an unpolarised sample, involves $|\boldsymbol{\mu}_{fi}|^2=\sum_q|\mu_q|^2$ — the rotationally invariant total. Only when you deliberately select one polarisation and one $\Delta m$ (say $z$ -polarised light driving $\pi$ transitions) does a single component appear alone. The Wigner–Eckart theorem is the bookkeeping for how that total shares among the geometric channels — which predicts the relative intensities of Zeeman and fine-structure multiplets.