Doppler-Free and High-Resolution Laser Spectroscopy

The clearest descendant of these ideas is the optical atomic clock. A clock laser is pre-stabilised to an ultra-narrow, Doppler-free reference and locked to a forbidden transition in a single trapped ion or a lattice of neutral strontium or ytterbium atoms cooled to microkelvin — cooling being the modern way to erase both first- and second-order Doppler shifts at once. The best reach fractional uncertainties near $10^{-18}$, a thousand-fold beyond the caesium standard, with a frequency comb tying their ticking to the SI second. The same comb, seeded by Hänsch's programme, lets hydrogen 1S–2S be measured absolutely to about $4\times10^{-15}$, feeding the Rydberg constant and, with muonic-hydrogen data, the proton-radius determination. Sub-Doppler methods also remain the everyday workhorse of the laboratory. Saturated-absorption spectroscopy of a rubidium or caesium vapour cell — a diode laser, a beamsplitter, a retroreflected probe — is the standard way labs lock a laser to an atomic line, resolving hyperfine dips and crossovers on an oscilloscope. Molecular iodine and methane Lamb dips serve as portable frequency anchors, and King-plot analyses of isotope shifts have become a frontier search for physics beyond the Standard Model.

Picture the absorbers as a crowd of runners on a straight track and the laser as a fixed-pitch whistle blown from one end. A runner sprinting toward the whistle hears it sharpened; one fleeing hears it flattened. Because the crowd carries every speed at once, no single pitch is "in tune" with all of them — as you sweep the pitch you excite one speed group after another, and the collective response smears into a broad band. That band is the Doppler profile, and its width has nothing to do with how sharply any one runner hears: it is a spread over runners, which is what inhomogeneous means and why the trick that follows works. Saturation spectroscopy blows the same whistle from both ends of the track. A runner standing still hears both whistles at the true pitch; a moving runner hears one sharpened and the other flattened, never both in tune. So the only group that responds to both beams is the stationary one, and if the first (pump) beam has already bleached those still runners, the second (probe) beam sails through unattenuated — a sharp notch of extra transmission at the true pitch on the broad Doppler background: the Lamb dip, whose width is the homogeneous width the still runners had all along.

Two-photon spectroscopy defeats the Doppler shift by a subtler route: a runner absorbs one photon from each direction. Moving toward one beam they see it blue-shifted by $+\omega v_z/c$, but the counter-propagating beam is equally red-shifted by $-\omega v_z/c$; the two shifts appear with opposite sign in the sum of the photon energies and cancel exactly to first order. Every runner, whatever their speed, is resonant at the same tuning $\omega=\omega_{ab}/2$. Nothing is bleached and no velocity class is thrown away — the entire crowd contributes to one narrow line, which is why two-photon signals can be strong from a dilute vapour. Whatever the method, the residual width is now homogeneous — the same for every atom because it no longer comes from motion: the finite excited-state lifetime (natural width), the transit time across the beam, collisions, power broadening, or the laser's own jitter. These are the true floors that cold atoms, large beams and stable cavities are engineered to lower. The second figure shows why two-photon absorption sees no Doppler shift for any atom.

The whole subject starts from one distribution: the beam-component of an atom's velocity is Gaussian-distributed, and it is this spread — not any single-atom property — that the laser scans through. The one-dimensional Maxwell–Boltzmann law fixes it.

where: $v_z$ — velocity component along the laser axis; $m$ — mass; $T$ — absolute temperature; $k_B$ — Boltzmann constant. The $1/e$ velocity spread $\sqrt{2k_BT/m}$ is hundreds of metres per second for light atoms at room temperature. An atom moving with that $v_z$ does not absorb at the rest frequency $\omega_0$: in its frame the lab laser is shifted, so in the lab it is resonant at a shifted frequency. This linear-in-$v_z/c$ shift is the entire engine of Doppler broadening.

where: $\omega_0$ — rest-frame transition frequency; $\omega_{\mathrm{lab}}$ — laser frequency at which an atom of velocity $v_z$ is resonant; $c$ — speed of light. An atom moving toward the source is resonant with a higher laser frequency, and vice versa. Folding the sharp single-atom response over the Gaussian velocity distribution turns the delta-function absorber into a Gaussian line, whose width scales as the optical frequency times the thermal velocity spread. This is the central formula of Doppler-limited spectroscopy.

where: $g_D$ — normalised Gaussian profile; $\delta_D$ — $1/e$ half-width. Setting the exponent to $-\ln2$ gives the FWHM $\Delta\omega_D=2\sqrt{\ln2}\,\delta_D=(\omega_0/c)\sqrt{8k_BT\ln2/m}$, growing as $\sqrt{T/m}$ and linearly with optical frequency. Against that Gaussian sits the homogeneous line every atom shares. Spontaneous emission from a state of lifetime $\tau$ gives a Lorentzian of full width $\Gamma=1/\tau$ — the width the Doppler-free techniques are trying to recover.

where: $\Gamma$ — natural linewidth (FWHM) $=1/\tau$; $\tau$ — excited-state radiative lifetime. A few nanoseconds of lifetime gives a few megahertz of width, roughly a hundredth of the Doppler width. Saturation spectroscopy relies on bleaching — a strong beam cycling population fast enough to empty the ground state of the class it addresses. How strong that must be is set by the saturation intensity, and how much a beam saturates by the saturation parameter.

where: $I$ — laser intensity; $I_{\mathrm{sat}}$ — saturation intensity; $\lambda$ — transition wavelength; $h$ — Planck constant. At $I=I_{\mathrm{sat}}$ the transition is half-bleached on resonance, and the Lamb dip then broadens as $\Gamma\sqrt{1+s_0}$. The Lamb dip's location is pure geometry: pump and probe share one lab frequency $\omega$ but travel oppositely, so their resonance conditions pick mirror-image velocity classes that coincide only at line centre.

where: $\omega_{\mathrm{pump}},\omega_{\mathrm{probe}}$ — resonance conditions for the two counter-propagating beams of a single laser of frequency $\omega$. Both hold for one class only when $\omega=\omega_0$, i.e. $v_z=0$; there the bleached atoms make the probe more transmitted — the dip. Two-photon spectroscopy skips velocity selection entirely. Absorbing one photon from each direction, the atom's resonance condition contains the sum of the two Doppler-shifted frequencies, and the linear terms cancel for every $v_z$.

where: $\omega_{ab}$ — two-photon ($a\!\to\!b$) transition frequency; $\omega_1,\omega_2$ — the counter-propagating photon frequencies. The cancellation is exact for all $v_z$, so the full ensemble contributes and the line is Doppler-free — at the cost of a weak, still-broadened pedestal from same-beam absorption. The cancellation is only first order. Relativistic time dilation survives at second order and cannot be geometry-cancelled — the ultimate thermal limit, and the reason the best experiments cool their atoms.

where: $\langle v^2\rangle=3k_BT/m$ — mean square speed. This red shift (a moving clock runs slow) is of order $10^{-11}$–$10^{-13}$ at room temperature and shrinks linearly with $T$ under cooling.

1 · How badly does Doppler broadening beat the natural width? (confidence builder)

The rubidium-87 D₂ transition is at $\lambda=780\ \mathrm{nm}$ with excited-state lifetime $\tau=26.2\ \mathrm{ns}$. Take a room-temperature vapour, $T=300\ \mathrm{K}$, $m=87\,u=1.44\times10^{-25}\ \mathrm{kg}$. Find the Doppler FWHM, the natural width, their ratio, and the mean atomic speed.

Step 1. Work in ordinary frequency using $\Delta\nu_D=(1/\lambda)\sqrt{8k_BT\ln2/m}$ (the $2\pi$ cancels between $\Delta\omega$ and $\Delta\nu$). The bracket $8k_BT\ln2/m=1.59\times10^{5}\ \mathrm{m^2/s^2}$ has square root $399\ \mathrm{m/s}$; dividing by $\lambda$ gives $\Delta\nu_D=399/(780\times10^{-9})=\mathbf{511\ MHz}$.

Step 2. The natural width is $\Gamma/2\pi=1/(2\pi\tau)=1/(2\pi\cdot26.2\times10^{-9})=\mathbf{6.07\ MHz}$. This is the homogeneous Lorentzian floor a Lamb dip or two-photon line can recover.

Step 3. The ratio is $511/6.07=\mathbf{84}$ — Doppler broadening hides the natural line by nearly two orders of magnitude, so nothing below ∼500 MHz reads without a sub-Doppler method. The mean speed behind it is $\bar v=\sqrt{8k_BT/\pi m}=\mathbf{270\ m/s}$, whose projection on the beam is the entire source of the 511 MHz. A single laser scan sees 511 MHz of Gaussian; saturation spectroscopy carves it to the 6 MHz natural width.

2 · Hydrogen 1S–2S: the two-photon flagship

The 1S–2S transition frequency is $\nu_{ab}=2.466\times10^{15}\ \mathrm{Hz}$, and the 2S state is metastable with lifetime $\tau_{2S}\approx0.12\ \mathrm{s}$. Find the two-photon laser wavelength, the natural width, the resulting quality factor, and the room-temperature second-order Doppler shift.

Step 1. Each photon carries half the energy, $\nu=\nu_{ab}/2=1.233\times10^{15}\ \mathrm{Hz}$, so $\lambda=c/\nu=\mathbf{243\ nm}$ — the deep-UV drive for 1S–2S.

Step 2. The natural width is $\Delta\nu_{\mathrm{nat}}=1/(2\pi\tau_{2S})=1/(2\pi\cdot0.12)=\mathbf{1.3\ Hz}$ — the metastable 2S state makes this among the sharpest transitions in reach.

Step 3. The intrinsic quality factor is $Q=\nu_{ab}/\Delta\nu_{\mathrm{nat}}=2.466\times10^{15}/1.3\approx\mathbf{1.9\times10^{15}}$ — a resonance sharp enough to underpin measurements of the Rydberg constant and tests of QED.

Step 4. The residual relativistic shift at $T=300\ \mathrm{K}$ is $-3k_BT/2mc^2=-3(1.381\times10^{-23})(300)/[2(1.67\times10^{-27})(2.998\times10^8)^2]=\mathbf{-4.1\times10^{-11}}$. First-order cancellation gets the line Doppler-free, but this term survives; cooling the beam to a few kelvin shrinks it proportionally — exactly what the precision experiments do. A 243 nm two-photon drive turns a 1.3 Hz natural line into a $Q\approx2\times10^{15}$ standard, limited at last only by second-order Doppler and technical noise.

Doppler-free spectroscopy is a field where one physicist's career traces almost the whole arc — from the first sub-Doppler hydrogen spectra to the frequency comb that finally read them absolutely.

Texts. Demtröder, Laser Spectroscopy (Vol. 2, Ch. 2 and 7 — saturation and two-photon methods) · Foot, Atomic Physics (Ch. 8, Doppler-free spectroscopy) · Loudon, The Quantum Theory of Light (Ch. 2, line broadening) · Saleh and Teich, Fundamentals of Photonics (laser spectroscopy) · Hecht, Optics (the Doppler effect and coherence). The key relations are $\Delta\omega_D=(\omega_0/c)\sqrt{8k_BT\ln2/m}$, the Lamb dip at $\omega_0$ with power-broadened width $\Gamma\sqrt{1+s_0}$, the two-photon cancellation $\omega(1-v_z/c)+\omega(1+v_z/c)=2\omega$, and the second-order Doppler shift $-3k_BT/2mc^2$ (from $\gamma^{-1}\approx1-\tfrac12\beta^2$ with $\langle v^2\rangle=3k_BT/m$). The anchor numbers: Rb D₂ at 780 nm, 300 K gives $\Delta\nu_D\approx511$ MHz vs $\Gamma/2\pi\approx6.07$ MHz (ratio ∼84), $I_{\mathrm{sat}}\approx1.67\ \mathrm{mW/cm^2}$, $\bar v\approx270$ m/s; hydrogen 1S–2S is a 243 nm line of ∼1.3 Hz natural width, $Q\approx2\times10^{15}$.

Prerequisites. Selection Rules for Molecular Transitions · Rotational Spectra and the Rotational Constant · Vibration–Rotation Spectra and P, Q, R Branches. Leads to. Electronic Transitions in Molecules · The Franck–Condon Principle.

Ask the tutor: Ask me to walk through why a crossover resonance lands exactly halfway between two hyperfine lines, or to estimate the Lamb-dip width and saturation intensity for any transition you name.