The Dipole Operator & Transition Moments

Visible light has a wavelength around 5000 Å. A molecule is about 1 Å across. From the molecule's point of view the electric field of the light wave doesn't vary in space at all — it is a uniform field $\mathbf{E}(t)$ that just oscillates in time, the same at every electron. A uniform field can't grab a molecule by its shape; the only handle it has is the total charge displacement,

That is the electric-dipole approximation: keep the first term in the expansion of the field across the molecule and drop the rest (quadrupole, magnetic dipole, …) as smaller by powers of the size ratio — three or four orders of magnitude each. So when light and matter talk, they talk through $\hat{\mu }$. The entire theory of spectra hangs off one operator.

Take the energy eigenstates $|0⟩,|1⟩,|2⟩,\dots$ and write $\hat{\mu }$ as a matrix in that basis. The diagonal elements $⟨n|\hat{\mu }|n⟩$ are the permanent dipole moments — the static quantity a polar molecule like water carries around. The off-diagonal elements $⟨f|\hat{\mu }|i⟩$ are the transition moments, and they answer a different question: how strongly does the field mix state $i$ with state $f$?

The absorption rate for the $i\to f$ line is proportional to $|⟨f|\hat{\mu }|i⟩{|}^2$. A zero there means the line does not exist, no matter how hard you pump — a dark transition. So a spectrum is really a picture of one matrix: line positions from the diagonal of $\hat{H}$, line intensities from the off-diagonals of $\hat{\mu }$. The tables below print that matrix, entry by entry, for two systems where every entry is known in closed form.

Why should $⟨f|\hat{\mu }|i⟩$ of all things set the brightness? Put the system in a superposition $\psi =c_i{\psi }_i+c_f{\psi }_f$ and watch the charge density: the cross term oscillates as ${\psi }_f^{\ast }{\psi }_i{e}^{-i{\omega }_{fi}t}$ at exactly the transition frequency. Charge sloshing back and forth at frequency ${\omega }_{fi}$ is an antenna, and the amplitude of that antenna — the dipole moment of the oscillating overlap density ${\psi }_f^{\ast }{\psi }_i$ — is precisely the transition moment.

This picture earns its keep twice. Absorption: the incoming field drives the antenna resonantly. Emission: an excited state decays by radiating through the same antenna, which is why the spontaneous lifetime of a state is set by the same matrix element (squared) that sets its absorption strength. Bright in, bright out. A state with no dipole coupling to anything below it can't radiate — it is metastable, and has to wait for collisions or weaker multipoles to get home.

Most transition moments are zero, and you can know it without integrating anything. In 1D with a symmetric potential, every eigenstate is even or odd. The operator $x$ is odd. So the integrand of $⟨n|x|m⟩$ has parity (even)·(odd)·(even) — odd overall — whenever $n$ and $m$ have the same parity, and an odd function integrates to zero over a symmetric interval. Done: same-parity transitions are dark, exactly, before any arithmetic.

Every selection rule you have met is this argument with a bigger symmetry group. Hydrogen's $\Delta l=\pm 1$ is the same integral-vanishes-by-symmetry statement with spherical harmonics in place of even/odd (that one also carries angular momentum bookkeeping: the photon takes one unit with it). In the particle in a box, states 1 and 3 are both even about the center, so $⟨1|x|3⟩=0$ — the grid below confirms it to 14 digits. Symmetry sorts the matrix into structural zeros and everything else; dynamics only fills in the "everything else."

The formal slot this all plugs into is first-order time-dependent perturbation theory: a weak perturbation $\hat{V}$ oscillating at $\omega$ drives transitions at the rate

Fermi's golden rule. The delta function is the eigenvalue side (energy must match); the squared matrix element is the dipole side (the states must be connected). With $\hat{V}=-\hat{\mu }\cdot \mathbf{E}$, the rate carries $|⟨f|\hat{\mu }|i⟩{|}^2$ — which is why the transition moment always shows up squared, and why its sign never matters for a single line (it starts mattering when two pathways interfere).

Packaged into a dimensionless number per transition, this squared moment becomes the oscillator strength, and computing it from many-electron eigenvectors is the final step of Casida's equation. This page is the one-particle root of that machinery.