Fermi's Golden Rule and Transition Rates

Modern absorption and fluorescence spectrometers report intensities and lifetimes that are Golden-Rule rates in disguise: a fluorophore's radiative lifetime (used to design fluorescent probes for cell imaging) is $1/A_{21}$, computed from the same transition dipole that appears in its absorption spectrum via the Strickler–Berg relation, itself a direct descendant of the Einstein $A$/$B$ connection derived above. Laser physics depends on population inversion built from exactly the competition between $B_{12}u(\omega)$ absorption, $B_{21}u(\omega)$ stimulated emission, and $A_{21}$ spontaneous decay — the three arrows in the figure above are literally the rate equations solved to design every laser gain medium, from He-Ne tubes to Ti:sapphire amplifiers used in ultrafast chemistry labs. In photochemistry and photovoltaics, non-radiative internal conversion and intersystem crossing rates — governing whether a photoexcited molecule fluoresces, undergoes charge separation in a solar cell, or triggers a photodynamic-therapy drug's toxic singlet-oxygen production — are computed with the same Golden Rule formula, replacing the radiation density with a Franck–Condon-weighted vibrational density of states (Marcus-type and Fermi-Golden-Rule electron-transfer theories share this exact mathematical skeleton). NMR relaxation times $T_1$ and $T_2$, central to every structural biology and clinical MRI measurement, are Golden-Rule rates driven by fluctuating local magnetic fields rather than an external laser, with the "transition dipole" replaced by nuclear spin coupling matrix elements and $\rho(E)$ replaced by the spectral density of molecular tumbling motion. Astrochemistry and radio astronomy depend on exactly the $A_{10}$ calculation of Example 3 to interpret interstellar CO emission maps used to trace molecular cloud mass and star-forming regions across the galaxy — a spontaneous-emission rate of a few nanohertz per molecule, made observable only by cosmological path lengths and huge column densities.

Ask the tutor: How would the Golden Rule rate change if the radiation were linearly polarized and the molecule's transition dipole were fixed in a crystal instead of freely tumbling in solution?

Picture a molecule in its ground electronic (or vibrational, or rotational) state $|i\rangle$, bathed in electromagnetic radiation. The oscillating electric field $\mathbf{E}(t)$ couples to the molecular dipole operator $\hat{\boldsymbol\mu}$, adding a small time-dependent term $H'(t) = -\hat{\boldsymbol\mu}\cdot\mathbf{E}(t)$ to the full Hamiltonian. Left alone, the molecule would sit forever in $|i\rangle$; under the perturbation, amplitude leaks into other stationary states $|f\rangle$, and if $|f\rangle$ differs in energy by close to $\hbar\omega$ — the photon energy of the driving field — that leakage is resonantly enhanced. If the field were a single, perfectly sharp frequency and lasted forever, the probability of finding the molecule in $|f\rangle$ would oscillate (Rabi-like) rather than grow steadily — this is the regime of coherent, monochromatic excitation such as a narrow-linewidth laser tuned exactly on resonance. But real light sources — thermal (blackbody) radiation, broadband lamps, or even "monochromatic" lasers viewed against a finite-linewidth absorption line — present the molecule with a continuum of frequencies or couple to a continuum of final states (vibrational density of states in a solid, or the continuum of directions for an emitted photon). Fermi's Golden Rule is the statement that once you integrate the transition probability over that continuum, the messy oscillatory quantum interference washes out, and what remains is a probability that grows linearly in time — i.e., a constant rate. That single fact — probability growing linearly instead of quadratically or obeying a exact `sin²` a Rabi formula — is why we are allowed to speak of a "rate constant" for a spectroscopic transition at all, exactly as chemists speak of first-order rate constants in kinetics. The rate constant is proportional to $|\langle f|H'|i\rangle|^2$, the squared transition matrix element, and to $\rho(E)$, the density of states or of field modes at the resonance energy.

The physical picture generalizes immediately to absorption and stimulated emission of light by extending $\rho(E)$ to the spectral energy density of the radiation field $u(\omega)$, and to spontaneous emission by invoking the density of electromagnetic field modes into which a photon can be emitted (vacuum fluctuations, properly treated with a quantized field, though the ratio of spontaneous to stimulated rates can be gotten from Planck's law without ever quantizing the field explicitly — this is Einstein's own 1917 argument, older than the Golden Rule itself).

The starting point, from time-dependent perturbation theory, is the first-order transition amplitude for a system prepared in $|i\rangle$ at $t=0$ and subjected to a perturbation $H'(t) = H'\,e^{-i\omega t} + H'^\dagger e^{i\omega t}$ (the standard decomposition of an oscillating field into rotating and counter-rotating parts):

where:

• $c_f^{(1)}(t)$ — first-order probability amplitude to be found in state $|f\rangle$ at time $t$
• $H'_{fi} = \langle f|H'|i\rangle$ — the matrix element of the perturbation between initial and final states
• $\omega_{fi} = (E_f - E_i)/\hbar$ — the Bohr transition frequency
• $\omega$ — the driving (photon) angular frequency

Reading: this is exactly the same integral that governs Rabi oscillations; squaring it gives a probability with the classic $\mathrm{sinc}^2$ resonance profile in $(\omega_{fi}-\omega)$, sharply peaked at exact resonance and narrowing (while growing taller) as $t$ increases. Carrying out the integral and taking the modulus squared gives the exact transition probability at time $t$ for a single sharp frequency $\omega$:

where:

• $P_{i\to f}(t)$ — transition probability after driving time $t$
• the $\mathrm{sinc}^2$-type function sharpens around $\omega=\omega_{fi}$ as $t$ grows, with peak height $\propto t^2$ and width $\propto 1/t$

Reading: the area under this resonance function in $\omega$ is what stays finite and grows linearly with $t$ — this is the seed of the Golden Rule, and it is why a continuum of frequencies (or of final states) is essential to get a genuine rate rather than an oscillation. Fermi's Golden Rule emerges by integrating $P_{i\to f}(t)$ against a continuous density of final states $\rho(E_f)$ (states per unit energy near resonance) or, equivalently, against a continuous spectral density of the driving field, and dividing by $t$ to extract the rate:

where:

• $k_{i\to f}$ — the transition rate (units s$^{-1}$), constant in time once the continuum limit is reached
• $\rho(E_f)$ — density of final states (or of field modes) evaluated at the resonance energy

Reading: this is Fermi's Golden Rule in its most general form; everything downstream in spectroscopy is a specific choice of what $H'_{fi}$ and $\rho(E_f)$ represent physically. For electric-dipole radiative transitions, the matrix element is built from the Transition Dipole Moment $\boldsymbol\mu_{fi} = \langle f|\hat{\boldsymbol\mu}|i\rangle$ dotted into the field, and $\rho(E_f)$ becomes the spectral energy density of the radiation $u(\omega)$ evaluated at $\omega=\omega_{fi}$. The resulting rate for absorption (or stimulated emission), averaged over random molecular orientation relative to an isotropic, unpolarized field, is

where:

• $\varepsilon_0$ — vacuum permittivity (SI units)
• $|\boldsymbol\mu_{fi}|^2 = |\langle f|\hat{\boldsymbol\mu}|i\rangle|^2$ — squared transition dipole matrix element
• $u(\omega_{fi})$ — spectral energy density of the radiation field per unit angular frequency at the resonance frequency
• the factor $1/3$ comes from orientational averaging over the three Cartesian components of an isotropic dipole in isotropic light

Reading: this is the master formula linking a computed or measured transition dipole directly to an absorption rate under any specified illumination — the exact bridge the Transition Dipole Moment article promises but does not by itself supply. Comparing this rate to the standard definition of the Einstein $B$ coefficient, $k_{i\to f} = B_{i\to f}\,u(\omega_{fi})$, identifies

where:

• $B_{i\to f}$ — Einstein coefficient for stimulated absorption/emission (SI convention, energy density per unit angular frequency)

Reading: the Golden Rule is precisely the microscopic derivation of the Einstein $B$ coefficient that Einstein himself, in 1917, had to introduce phenomenologically — quantum mechanics fills in what he could only postulate. ---

Example 1 — Foundational: rate for a model two-level system in a weak, broad field. Consider a model chromophore with transition dipole magnitude $|\mu_{fi}| = 2.4$ D (debye) between two non-degenerate electronic states separated by $\Delta E = 3.10$ eV (a mid-range, asymmetric, non-round choice typical of a visible-absorbing dye), illuminated by broadband, isotropic, incoherent light with spectral energy density $u(\omega_{fi}) = 3.5\times10^{-17}\ \text{J}\,\cdot\,\text{s/m}^3$ at the resonance frequency (a modest lamp intensity, deliberately not a "nice" number). First convert the dipole to SI: $1\ \text{D} = 3.336\times10^{-30}\ \text{C}\,\cdot\,\text{m}$, so

Then

Insert into the absorption rate formula:

Numerically, $\varepsilon_0 = 8.854\times10^{-12}\ \text{C}^2\text{N}^{-1}\text{m}^{-2}$ and $\hbar = 1.0546\times10^{-34}\ \text{J}\,\cdot\,\text{s}$, so $\hbar^2 = 1.1122\times10^{-68}\ \text{J}^2\text{s}^2$. Compute the prefactor:

Multiply by $|\mu_{fi}|^2$:

Finally multiply by $u(\omega_{fi})=3.5\times10^{-17}$:

So this moderately strong transition, under a fairly weak broadband lamp, absorbs at a rate of roughly $2.4\times10^3$ photons per second per molecule — corresponding to a mean time between absorption events of about 0.4 ms, entirely reasonable for dilute-solution UV-Vis excitation under lamp (not laser) illumination.

Limiting-case check: if the lamp is turned off, $u\to 0$, and $k_{i\to f}\to 0$ — no light, no absorption, exactly as required; the rate is strictly linear in $u(\omega_{fi})$, consistent with the definition of $B_{i\to f}$.

Example 2 — Applied: oscillator strength and molar absorptivity of a real vibrational overtone. The first overtone of a C–H stretch ($\Delta v = 2$) near $\tilde\nu \approx 5800\ \text{cm}^{-1}$ (about 1.72 μm, relevant to near-infrared spectroscopy of organic solvents and process analytics) typically has a weak integrated molar absorptivity because the transition is formally forbidden in the harmonic approximation and only appears through anharmonicity. Take a measured integrated band area of

(a realistic order-of-magnitude value for a weak NIR overtone, deliberately non-round). Using the Step 5 relation from the second derivation:

This tiny oscillator strength — four to five orders of magnitude below a fully allowed electronic transition ($f\sim0.1$–$1$) — is the quantitative signature of an overtone's forbidden character, entirely consistent with the fact that a strictly harmonic oscillator has zero transition dipole for $\Delta v = 2$ (a selection-rule result that anticipates Selection Rules from Symmetry). Now invert the oscillator-strength definition to extract the implied transition dipole. Rearranging

Convert $\tilde\nu = 5800\ \text{cm}^{-1}$ to angular frequency: $\omega = 2\pi c\tilde\nu = 2\pi\times(2.998\times10^{10}\ \text{cm/s})\times 5800\ \text{cm}^{-1} = 1.0925\times10^{15}\ \text{rad/s}$. Numerator: $3\hbar e^2 f_{if} = 3\times(1.0546\times10^{-34})\times(1.602\times10^{-19})^2\times(3.72\times10^{-8})$. Compute $e^2 = 2.566\times10^{-38}\ \text{C}^2$. Then

Denominator: $2m_e\omega_{fi} = 2\times(9.109\times10^{-31})\times(1.0925\times10^{15}) = 1.9903\times10^{-15}$.

Taking the square root, $|\mu_{fi}| = 1.232\times10^{-32}\ \text{C}\,\cdot\,\text{m}$, and converting to debye ($1\ \text{D}=3.336\times10^{-30}\ \text{C}\,\cdot\,\text{m}$):

A transition dipole of only 0.0037 D — two to three orders of magnitude smaller than a typical fundamental C–H stretch dipole (~0.1–0.3 D) — quantitatively captures why overtones are so much weaker than fundamentals in IR/NIR spectra, and it is exactly this number the Golden Rule turns into the measured band area.

Example 3 — Exam/competition level: Einstein A/B ratio and spontaneous vs. stimulated emission for a rotational microwave transition. Consider the $J=0\to1$ rotational transition of carbon monoxide (CO), a classic rigid-rotor system (foreshadowing Rotational Transitions and the Rigid Rotor), with transition frequency $\tilde\nu_{01} = 3.845\ \text{cm}^{-1}$ (i.e., $\nu_{01} = 115.27\ \text{GHz}$, the well-known CO $J=0\to1$ line used in radio astronomy) and permanent dipole moment $\mu = 0.112\ \text{D}$, giving a transition dipole (rigid rotor selection rule factor for $J=0\to1$) of $|\mu_{01}|^2 = \mu^2\cdot\frac{J+1}{2J+3}\Big|_{J=0} = \mu^2/3$.

Step A — transition dipole. $\mu = 0.112\ \text{D} = 3.736\times10^{-31}\ \text{C}\,\cdot\,\text{m}$, so $\mu^2 = 1.396\times10^{-61}\ \text{C}^2\text{m}^2$, and

Step B — Einstein $B$ coefficient. Using $B_{i\to f} = \dfrac{\pi}{3\varepsilon_0\hbar^2}|\mu_{fi}|^2$ from the same prefactor computed in Example 1 ($1.0553\times10^{78}$ in SI units):

Step C — angular frequency. $\omega_{01} = 2\pi\nu_{01} = 2\pi\times1.1527\times10^{11} = 7.243\times10^{11}\ \text{rad/s}$.

Step D — Einstein $A$ coefficient from detailed balance.

(Degeneracies enter here: $B_{01}=B_{10}$ per absorbing/emitting molecule only after accounting for the $2J+1$ degeneracies of each level; since we have already built the $J$-dependent dipole factor into $|\mu_{01}|^2$ using the standard rigid-rotor line-strength formula, $B_{01}$ and $B_{10}$ as computed are already the correctly normalized per-molecule coefficients for this specific line.) Compute $\omega_{01}^3 = (7.243\times10^{11})^3 = 3.800\times10^{35}\ \text{rad}^3/\text{s}^3$. Compute $c^3 = (2.998\times10^{8})^3 = 2.695\times10^{25}\ \text{m}^3/\text{s}^3$.

Numerator: $1.0546\times10^{-34}\times3.800\times10^{35} = 40.07$. Denominator: $9.870\times2.695\times10^{25} = 2.660\times10^{26}$.

The spontaneous emission rate is only $7.4\times10^{-9}\ \text{s}^{-1}$, i.e. a spontaneous radiative lifetime of order $1.35\times10^{8}\ \text{s} \approx 4.3$ years — utterly negligible compared to collisional relaxation in any real gas sample, which is exactly why pure rotational transitions are observed via absorption (stimulated processes, governed by $B$) rather than spontaneous glow, and why radio-astronomical CO emission from interstellar clouds, though extraordinarily weak per molecule, is detectable only because of the vast column densities and long path lengths in space.

Limiting-case check: the $A/B$ ratio scales as $\omega^3$; for a visible electronic transition ($\omega\sim 10^{15}\ \text{rad/s}$, roughly $10^4$ times larger than this microwave $\omega$), $A/B$ is larger by a factor of $10^{12}$, correctly predicting that electronic transitions relax spontaneously (fluoresce) on nanosecond timescales while rotational transitions essentially never do — a direct, checkable consequence of the $\omega^3$ scaling derived above.