Sideband Cooling

A trapped ion sits in a harmonic well, and like any harmonic oscillator its motion is quantized: the ion does not wiggle with an arbitrary amplitude but occupies discrete vibrational levels $E_n=\hbar\omega_z(n+\tfrac12)$ separated by one phonon $\hbar\omega_z$ . Ordinary laser (Doppler) cooling brings that motion down to a thermal cloud of order ten phonons — cold by the standards of atomic physics, but far too hot for a quantum computer, where the shared vibration is the wire that carries information between ion qubits. Sideband cooling is the technique that empties the wire. It exploits a beautiful piece of spectroscopy: because the ion's internal transition is frequency-modulated by its own oscillation, the absorption line sprouts motional sidebands one phonon above and below the carrier, and a laser tuned to the lower (red) sideband can only be absorbed if the ion simultaneously surrenders a quantum of motion. Three facts organize everything that follows. First, the trick needs resolution: the sidebands are only distinct lines if the trap frequency beats the transition linewidth, $\omega_z\gg\gamma$ — the resolved-sideband regime. Second, it needs smallness: the ion's zero-point wavepacket must be tiny compared with the optical wavelength, measured by the Lamb–Dicke parameter $\eta=k z_0\ll1$ , so that each photon changes the phonon number by at most one. Third, it has a natural stop: the state $\lvert g,0\rangle$ has no lower rung, so the red-sideband coupling $\propto\sqrt{n}$ vanishes there — the ground state is dark, and population piles up in it. The residual heating from off-resonant scattering fixes a floor $\bar n_{\mathrm{ss}}\approx(\gamma/2\omega_z)^2$ , routinely below $0.05$ in modern traps. This is the article that turns a warm ion into the cold, well-defined motional bus the DiVincenzo criteria demand.

Key takeaways

Sideband cooling drives the red motional sideband of an internal transition, so each absorption–emission cycle removes one phonon $\hbar\omega_z$ and pumps a trapped ion or atom toward its motional ground state $\lvert n=0\rangle$ .
It works only in the resolved-sideband regime, $\omega_z\gg\gamma$ , where the trap frequency exceeds the transition linewidth so the red sideband is spectrally distinct from the carrier and the heating (blue) sideband.
The Lamb–Dicke parameter $\eta=k z_0=\sqrt{\hbar k^2/2m\omega_z}$ controls everything: for $\eta\ll1$ the recoil rarely changes $n$ , and the sideband couplings are $\Omega_{n,n-1}=\Omega\eta\sqrt{n}$ .
The steady-state occupation is $\bar n_{\mathrm{ss}}\approx(\gamma/2\omega_z)^2\ll1$ , set by balancing resonant cooling against off-resonant reheating — narrow transitions and stiff traps push the floor down quadratically.
Ground-state cooling is the mandatory first step for phonon-mediated entangling gates (Cirac–Zoller, Mølmer–Sørensen), for optical-clock state preparation, and for the DiVincenzo cold-qubit requirement.
Key idea: modulating an internal transition by harmonic motion produces resolved motional sidebands; resonantly driving the red sideband ( $\omega_L=\omega_0-\omega_z$ ) lets each absorption–emission cycle subtract one phonon, pumping the system to its dark motional ground state $\lvert g,0\rangle$ .

Where this lives today

Sideband cooling is the indispensable opening move of every trapped-ion experiment, and its reach keeps widening. In quantum computers built from ion chains — the machines of IonQ, Quantinuum, and academic groups at Innsbruck, Maryland, and NIST — every entangling gate begins by cooling the shared motional modes to $\bar n\ll1$ ; residual phonons are a leading error source, so groups now use electromagnetically-induced-transparency (EIT) and polarization-gradient cooling to chill many modes of a long chain at once, faster than pulsed sideband cooling allows. In the world's best optical clocks — the $\mathrm{Al}^+$ quantum-logic clock and $\mathrm{Yb}^+$ and $\mathrm{Sr}^+$ standards reaching fractional uncertainties near $10^{-18}$ — the same cooling suppresses the motional (second-order Doppler) time-dilation shift that would otherwise dominate the error budget. And the resolved-sideband idea has jumped substrate entirely: cavity optomechanics uses a red-detuned laser to cool gram-to-nanogram mechanical oscillators toward their quantum ground state, and circuit-QED microwave drives cool superconducting and mechanical resonators. An open frontier is scalable cooling — keeping the bus cold in traps holding tens to hundreds of ions, and re-cooling spectator modes mid-algorithm without disturbing the qubits, a bottleneck fault-tolerant ion-trap architectures must solve.

The physical picture

Picture the ion as a tiny bell whose pitch $\omega_0$ is faintly frequency-modulated by its swing in the trap at $\omega_z$ . That modulation dresses the single absorption line with a comb: a strong carrier at $\omega_0$ and weak satellites — motional sidebands — one phonon below at $\omega_0-\omega_z$ and one above at $\omega_0+\omega_z$ . Now shine a laser tuned exactly to the lower satellite. A photon at $\omega_0-\omega_z$ is one phonon short of the internal transition, so it can only be absorbed if the ion makes up the deficit from its own motion — the ion must give up a vibrational quantum to get excited. That is the cooling event: $\lvert g,n\rangle\to\lvert e,n-1\rangle$ . The ion then re-emits a photon spontaneously, and because the recoil is tiny (Lamb–Dicke regime), the most probable emission leaves the phonon number untouched, returning the ion to $\lvert g,n-1\rangle$ . Net result of the round trip: one phonon gone.

Figure

The cooling cycle on the ladder of dressed states. A red-sideband photon drives the ion up one internal level and down one motional rung (n to n−1), then spontaneous emission returns it at the same motional level (still n−1). Each full cycle removes one phonon ħω_z so Δn = −1, and the cycle stalls at the motional ground state because there is no lower rung.

Each cycle is a ratchet, stepping the ion down the vibrational ladder until it reaches $\lvert g,0\rangle$ . There the ratchet jams: the red-sideband coupling scales as $\sqrt{n}$ , so it vanishes at $n=0$ — there is no $n=-1$ state to fall into, the laser can no longer be absorbed, and population accumulates in the ground state, which is dark to the cooling light. The whole scheme lives or dies on spectral resolution. If the excited state is short-lived, its linewidth $\gamma$ smears the levels; when $\gamma\gtrsim\omega_z$ the red and blue satellites blur into the carrier, and the laser starts driving the carrier (which does not cool) and the blue sideband (which heats). Only when the trap quantum $\hbar\omega_z$ comfortably exceeds the energy blur $\hbar\gamma$ do the sidebands stand apart as clean, separately addressable lines. That is what the second figure shows, with real numbers.

Figure

The resolved-sideband absorption spectrum for a typical trapped-ion optical qubit: three sharp lines — red sideband, carrier, blue sideband — spaced by the trap frequency ω_z/2π = 1 MHz, each of width γ/2π = 20 kHz. Because ω_z ≫ γ the satellites are cleanly resolved, and the residual cooling floor is n̄ss ≈ (γ/2ω_z)² ≈ 1×10⁻⁴.

The core equations

The starting point is the Hamiltonian of a two-level ion whose motion along one trap axis is quantized, coupled to a travelling laser field. The spatial phase of that field, evaluated at the ion's operator-valued position $\hat z=z_0(a+a^\dagger)$ , is what mixes internal and motional degrees of freedom and generates the sidebands.

H = \hbar\omega_z\, a^\dagger a + \frac{\hbar\omega_0}{2}\sigma_z + \frac{\hbar\Omega}{2}\left(\sigma_+ e^{i(\eta(a+a^\dagger)-\omega_L t)} + \mathrm{h.c.}\right) \qquad (1)

where: $\omega_z$ — trap (axial) frequency; $a,a^\dagger$ — motional annihilation/creation operators; $\omega_0$ — internal transition frequency; $\sigma_z,\sigma_\pm$ — qubit Pauli and raising/lowering operators; $\Omega$ — bare carrier Rabi frequency; $\eta$ — Lamb–Dicke parameter; $\omega_L$ — laser frequency. The exponential $e^{i\eta(a+a^\dagger)}$ is the whole story: expand it and the sidebands fall out. The single dimensionless number that governs the coupling is the Lamb–Dicke parameter — the ratio of the ion's zero-point motion $z_0$ to the reduced laser wavelength $1/k$ , or equivalently the square root of the photon recoil energy measured in phonons.

\eta = k z_0 = k\sqrt{\frac{\hbar}{2 m \omega_z}} = \sqrt{\frac{\hbar k^2}{2 m \omega_z}} \qquad (2)

where: $k$ — laser wavevector projected on the motional axis; $z_0=\sqrt{\hbar/2m\omega_z}$ — ground-state position spread; $m$ — ion mass. Small $\eta$ means the wavepacket is far smaller than a wavelength, so a photon's recoil kick almost never changes $n$ — the physical content of the Lamb–Dicke regime. Expanding the phase factor in powers of $\eta$ sorts the ion–laser coupling into a carrier and a hierarchy of sidebands. Each successive term shifts the phonon number by one more unit, and in the Lamb–Dicke regime only the first two terms survive.

e^{i\eta(a+a^\dagger)} \approx 1 + i\eta(a+a^\dagger) - \frac{\eta^2}{2}(a+a^\dagger)^2 + \cdots \qquad (3)

where: the $1$ term is the carrier ( $\Delta n=0$ ); $i\eta(a+a^\dagger)$ gives the first red ( $a$ , $\Delta n=-1$ ) and blue ( $a^\dagger$ , $\Delta n=+1$ ) sidebands; the $\eta^2$ term is a Debye–Waller correction and higher sidebands. Truncating after first order is exactly the resolved two-state picture that makes cooling analytically tractable. Reading off the matrix elements of the first-order term between Fock states gives the effective Rabi frequency on each sideband. The bosonic factors $\sqrt{n}$ and $\sqrt{n+1}$ are the mathematical heart of why $\lvert 0\rangle$ is dark.

\Omega_{n,n-1}=\Omega\,\eta\sqrt{n}\ \ (\text{red}),\qquad \Omega_{n,n+1}=\Omega\,\eta\sqrt{n+1}\ \ (\text{blue}) \qquad (4)

where: $\Omega_{n,n\mp1}$ — effective Rabi frequency coupling neighbouring Fock states; the factor $\sqrt{n}$ comes from $a\lvert n\rangle=\sqrt{n}\lvert n-1\rangle$ . At $n=0$ the red-sideband coupling is exactly zero, so the ground state cannot absorb the cooling light — it is dark. Balancing the resonant cooling rate against off-resonant reheating fixes the lowest occupation the method can reach. In the resolved-sideband limit this floor scales as the square of the ratio of linewidth to trap frequency.

\bar{n}_{\mathrm{ss}} \approx \left(\frac{\gamma}{2\omega_z}\right)^2 \qquad (\omega_z\gg\gamma) \qquad (5)

where: $\gamma$ — natural linewidth (spontaneous decay rate) of the cooling transition; $\bar n_{\mathrm{ss}}$ — steady-state mean phonon number. A narrower transition or a stiffer trap lowers the floor quadratically — the design rule behind narrow-line and Raman sideband cooling. Before sideband cooling can even begin, the ion must be pre-cooled below the point where higher sidebands proliferate. Doppler cooling on a strong dipole transition does this, reaching a mean occupation set by the linewidth-to-trap ratio.

k_B T_D=\frac{\hbar\gamma}{2}\ \Rightarrow\ \bar{n}_{\mathrm{Doppler}}\approx\frac{k_B T_D}{\hbar\omega_z}=\frac{\gamma}{2\omega_z} \qquad (6)

where: $T_D$ — Doppler temperature; $k_B$ — Boltzmann constant. Because Doppler cooling uses a broad transition with $\gamma\gtrsim\omega_z$ , it leaves $\bar n$ of order ten — cold enough to enter the Lamb–Dicke regime, but not the ground state. Finally, driving the red sideband coherently produces Rabi flopping whose amplitude carries the $\sqrt{n}$ signature, and this is both the cooling engine and the thermometer. The vanishing of the signal at $n=0$ is the experimental fingerprint of a cold ion.

P_e^{\mathrm{rsb}}(t)=\sin^2\!\left(\frac{\Omega\eta\sqrt{n}}{2}\,t\right),\qquad \bar n=\frac{P_{\mathrm{rsb}}}{P_{\mathrm{bsb}}-P_{\mathrm{rsb}}} \qquad (7)

where: $P_e^{\mathrm{rsb}}$ — excited-state probability under red-sideband drive; $P_{\mathrm{rsb}},P_{\mathrm{bsb}}$ — red- and blue-sideband excitation after a fixed pulse. The red sideband is silent for $n=0$ , so its ratio to the blue sideband reads out $\bar n$ with no calibration — the standard sideband thermometry.

Derivation ASideband structure from the Lamb–Dicke expansion

The goal is to show, starting from the ion–laser coupling of equation (1), exactly how the single exponential $e^{i\eta(a+a^\dagger)}$ splits the absorption line into a carrier and first-order red and blue sidebands, and to pin down the red-sideband Rabi frequency $\Omega_{n,n-1}=\Omega\eta\sqrt{n}$ that both cools the ion and, later, entangles two of them. We move to the interaction picture, expand to first order in the small parameter $\eta$ , pick out the resonant term by choosing the laser detuning, and evaluate the surviving matrix element on Fock states.

Step 1. Write the interaction-picture coupling $H_{\mathrm{int}}=\tfrac{\hbar\Omega}{2}\,\sigma_+\,e^{i\eta(a+a^\dagger)}e^{-i\delta t}+\mathrm{h.c.}$ , where $\delta=\omega_L-\omega_0$ is the laser detuning from the carrier and the motional operators rotate as $a\to a\,e^{-i\omega_z t}$ , $a^\dagger\to a^\dagger e^{+i\omega_z t}$ . Why: transforming into the frame rotating at $\omega_0$ and $\omega_z$ separates the fast optical phase from the slow motional dynamics, so that resonance conditions become visible as time-independent terms.

Step 2. Expand $e^{i\eta(a+a^\dagger)}\approx 1+i\eta\!\left(a\,e^{-i\omega_z t}+a^\dagger e^{+i\omega_z t}\right)$ to first order in $\eta$ . The three pieces now oscillate at distinct frequencies: the $1$ term at $\delta$ (carrier), the $a$ term at $\delta-\omega_z$ (red), the $a^\dagger$ term at $\delta+\omega_z$ (blue). Why: each power of $\eta$ carries a definite number of motional quanta, so first order is precisely the $\Delta n=\pm1$ sidebands and nothing else — legitimate whenever $\eta\sqrt{n+1}\ll1$ .

Step 3. Tune the laser to the red sideband, $\delta=-\omega_z$ . Then the $a$ term becomes time-independent while the carrier and blue terms rotate at $\mp\omega_z$ and average away (rotating-wave approximation), leaving $H_{\mathrm{rsb}}=\tfrac{i\hbar\Omega\eta}{2}\,\sigma_+ a+\mathrm{h.c.}$ Why: choosing the detuning selects one and only one energy-conserving process, $\lvert g,n\rangle\to\lvert e,n-1\rangle$ — this is the anti-Jaynes–Cummings coupling, the same interaction that powers the Cirac–Zoller gate.

Step 4. Evaluate the matrix element on Fock states: $\langle n-1\rvert a\lvert n\rangle=\sqrt{n}$ , so $\Omega_{n,n-1}=\Omega\,\eta\,\sqrt{n}$ , and likewise the blue sideband gives $\Omega_{n,n+1}=\Omega\eta\sqrt{n+1}$ . Why: the bosonic $\sqrt{n}$ is decisive — it makes cooling slow down as $n$ shrinks and vanish outright at $n=0$ , which is exactly why the ground state is dark and cooling self-terminates there.

Go deeper · Derivation B · 1 minSteady-state phonon number from rate equations

Derivation BSteady-state phonon number from rate equations

The goal is to derive the resolved-sideband cooling floor $\bar n_{\mathrm{ss}}\approx(\gamma/2\omega_z)^2$ by treating the Fock-state populations as classical occupations evolving under scattering rates, and balancing the phonon-removing (red) channel against the phonon-adding (blue and off-resonant carrier) channels. This is the calculation that tells an experimentalist how cold the ion can get and how the answer improves with a narrower transition or a stiffer trap; it uses Fermi's golden rule with Lorentzian lineshapes and a detailed-balance steady state.

Step 1. Write the scattering rates into the $\Delta n=\mp1$ channels as Lorentzians centred on the sidebands, $A_\pm\propto\eta^2\Omega^2\,\dfrac{\gamma/2}{(\delta\pm\omega_z)^2+(\gamma/2)^2}$ , where $A_-$ removes a phonon (cooling) and $A_+$ adds one (heating). Why: the excited state has a finite lifetime $1/\gamma$ , so its absorption profile is Lorentzian of width $\gamma$ ; the off-resonant tail of that Lorentzian is precisely the leak that limits cooling.

Step 2. Tune to the red sideband, $\delta=-\omega_z$ , so the cooling channel sits on resonance (denominator $(\gamma/2)^2$ ) while the heating channel is detuned by $2\omega_z$ (denominator $(2\omega_z)^2+(\gamma/2)^2$ ). Why: this is the whole point of resolving the sidebands — resonantly enhancing cooling while pushing the heating process far into the wing of the line.

Step 3. Take the ratio $\dfrac{A_+}{A_-}=\dfrac{(\gamma/2)^2}{(2\omega_z)^2+(\gamma/2)^2}\approx\left(\dfrac{\gamma}{4\omega_z}\right)^2$ in the resolved limit $\omega_z\gg\gamma$ . Why: the suppression of heating is set entirely by how far the blue sideband lies out on the Lorentzian tail, i.e. by the spectral mismatch $2\omega_z$ measured in linewidths.

Step 4. Impose detailed balance. In steady state the up- and down-flows through each rung match, giving $\bar n_{\mathrm{ss}}=\dfrac{A_+}{A_--A_+}\approx\dfrac{A_+}{A_-}$ ; including the recoil bookkeeping of spontaneous emission promotes the prefactor to the standard $\bar n_{\mathrm{ss}}\approx(\gamma/2\omega_z)^2$ , which vanishes as $\omega_z\to\infty$ . Why: a geometric-series steady state of a birth–death process is fixed by the single ratio $A_+/A_-$ , so the cooling floor is quadratic in $\gamma/\omega_z$ — halving the linewidth quarters the residual motion.

Go deeper · Derivation C · 1 minSideband thermometry from the red-to-blue asymmetry

Derivation CSideband thermometry from the red-to-blue asymmetry

The goal is to derive the calibration-free thermometer $\bar n=P_{\mathrm{rsb}}/(P_{\mathrm{bsb}}-P_{\mathrm{rsb}})$ that lets an experiment read the mean phonon number straight off two pulse-excitation measurements, and to explain why it works only because the red and blue sidebands carry the asymmetric bosonic factors $\sqrt{n}$ and $\sqrt{n+1}$ . This is the diagnostic every ground-state-cooling experiment reports; the derivation weights the Fock-state couplings by a thermal distribution and compares the two sidebands in the weak-excitation limit.

Step 1. For a thermal motional state the Fock populations are $p_n=\bar n^n/(1+\bar n)^{n+1}$ . Under a short red-sideband pulse the excitation from rung $n$ is $\propto\Omega_{n,n-1}^2=\Omega^2\eta^2 n$ , and under a blue pulse $\propto\Omega_{n,n+1}^2=\Omega^2\eta^2(n+1)$ . Why: in the weak-drive limit excitation probability is proportional to the squared coupling, so the whole $n$ -dependence reduces to the bosonic factors $n$ and $n+1$ .

Step 2. Average over the thermal distribution: $P_{\mathrm{rsb}}\propto\sum_n p_n\,n=\bar n$ and $P_{\mathrm{bsb}}\propto\sum_n p_n\,(n+1)=\bar n+1$ , using $\langle n\rangle=\bar n$ . Why: the mean of $n$ over a thermal state is by definition $\bar n$ , so the two sideband signals differ by exactly the additive $+1$ of the blue channel — the spontaneous-versus-stimulated asymmetry of a boson.

Step 3. Form the ratio $\dfrac{P_{\mathrm{rsb}}}{P_{\mathrm{bsb}}}=\dfrac{\bar n}{\bar n+1}$ and invert to get $\bar n=\dfrac{P_{\mathrm{rsb}}}{P_{\mathrm{bsb}}-P_{\mathrm{rsb}}}$ ; the common prefactor $\Omega^2\eta^2$ cancels, so no knowledge of $\Omega$ , $\eta$ , or pulse area is needed. Why: taking a ratio of the two sidebands divides out every experimental scale factor, which is what makes the measurement absolute — and as $\bar n\to0$ the red sideband disappears, the unmistakable signature of the ground state.

Limiting cases

The behaviour of sideband cooling is fixed by two dimensionless ratios — the sideband resolution $\gamma/\omega_z$ and the Lamb–Dicke parameter $\eta$ — and every regime below is one corner of that plane:

Case	Result	Meaning
Resolved sideband, deep Lamb–Dicke	$\omega_z\gg\gamma,\ \eta\ll1$	Ideal regime: clean red-sideband cooling to $\bar n\approx(\gamma/2\omega_z)^2\ll1$ ; ground state reliably reached
Unresolved sidebands	$\omega_z\lesssim\gamma$	Satellites overlap the carrier; only Doppler cooling is possible, $\bar n\approx\gamma/2\omega_z\sim$ tens — used as the pre-cooling stage
Large Lamb–Dicke parameter	$\eta\gtrsim1$	Higher sidebands ( $\pm2,\pm3,\dots$ ) and recoil-driven $\Delta n>1$ jumps activate; the simple two-state cooling picture fails, demanding a full optical-Bloch treatment
Strong drive (carrier-limited)	$\Omega_{\mathrm{eff}}\gtrsim\omega_z$	Power broadening off-resonantly excites the carrier and blue sideband, raising the floor; $\Omega$ is optimized to trade cooling rate against final $\bar n$
Stiff-trap / narrow-line limit	$\omega_z/\gamma\to\infty$	$\bar n_{\mathrm{ss}}\to0$ : quadrupole or Raman transitions with tiny $\gamma$ push the floor arbitrarily low, at the cost of a slow cooling rate

Common trapThe ground state is reached whenever you drive the red sideband long enough

Not so — driving the red sideband is necessary but not sufficient. Cooling self-terminates at a floor $\bar n_{\mathrm{ss}}\approx(\gamma/2\omega_z)^2$ set by off-resonant reheating, and that floor only lies well below one inside the resolved-sideband regime $\omega_z\gg\gamma$ . If $\gamma\gtrsim\omega_z$ the blue sideband and carrier are not spectrally separated, so every red-sideband photon is accompanied by heating events, and $\bar n$ never drops much below one no matter how long you cool. A second trap: making $\eta$ large does not help. A big Lamb–Dicke parameter activates second- and higher-order sidebands and lets spontaneous-emission recoil change $n$ by more than one, so the tidy $\Delta n=-1$ ratchet breaks down and the ground state is spoiled. Deep cooling needs both $\omega_z\gg\gamma$ and $\eta\ll1$ .

Worked examples

1 · Lamb–Dicke parameter and recoil for a calcium ion (confidence builder)

A single $^{40}\mathrm{Ca}^+$ ion is cooled on the narrow $S_{1/2}\to D_{5/2}$ quadrupole transition at $\lambda=729\ \mathrm{nm}$ in an axial trap of frequency $\omega_z/2\pi=1.0\ \mathrm{MHz}$ . Find the zero-point extent $z_0$ , the Lamb–Dicke parameter $\eta$ , and the recoil energy in units of a phonon.

Step 1. Zero-point spread. With $m=40\times1.6605\times10^{-27}=6.64\times10^{-26}\ \mathrm{kg}$ and $\omega_z=2\pi\times10^6=6.28\times10^6\ \mathrm{s^{-1}}$ ,

z_0=\sqrt{\frac{\hbar}{2m\omega_z}}=\sqrt{\frac{1.055\times10^{-34}}{2(6.64\times10^{-26})(6.28\times10^6)}}=\mathbf{1.12\times10^{-8}\ m}=11.2\ \mathrm{nm}.

The ion's quantum jitter is about 11 nm — some sixty times smaller than the 729 nm wavelength, the first hint that $\eta$ is small.

Step 2. Lamb–Dicke parameter. The wavevector is $k=2\pi/\lambda=2\pi/(729\times10^{-9})=8.62\times10^6\ \mathrm{m^{-1}}$ , so

\eta=k z_0=(8.62\times10^6)(1.12\times10^{-8})=\mathbf{0.097}\approx0.10.

This is safely in the Lamb–Dicke regime $\eta\ll1$ , so first-order sidebands dominate and the two-state cooling picture holds.

Step 3. Recoil in phonons. The recoil energy compared with one motional quantum is $\eta^2=(0.097)^2=\mathbf{9.4\times10^{-3}}$ . A scattered photon deposits under one percent of a phonon of recoil, so emission almost always leaves $n$ unchanged — exactly the condition equation (3) needs for the ratchet to work cleanly. z₀ = 11.2 nm, η ≈ 0.10, and recoil = 0.9% of a phonon — deep Lamb–Dicke.

2 · From Doppler pre-cooling to the ground state (cooling budget)

The same ion is first Doppler-cooled on the strong $S_{1/2}\to P_{1/2}$ dipole line at $397\ \mathrm{nm}$ ( $\gamma/2\pi=22.4\ \mathrm{MHz}$ ), then resolved-sideband cooled on the quadrupole transition quenched to an effective linewidth $\gamma_{\mathrm{eff}}/2\pi=20\ \mathrm{kHz}$ , with $\omega_z/2\pi=1.0\ \mathrm{MHz}$ . Estimate the phonon number after each stage and the resulting ground-state population.

Step 1. Doppler stage. On the broad dipole line $\gamma\gg\omega_z$ (22.4 MHz vs 1 MHz), so the sidebands are unresolved and only Doppler cooling operates, with floor

\bar n_{\mathrm{Doppler}}\approx\frac{\gamma}{2\omega_z}=\frac{22.4}{2(1.0)}=\mathbf{11.2}.

The ion arrives at roughly eleven phonons — cold enough to sit inside the Lamb–Dicke regime, but nowhere near the ground state.

Step 2. Sideband stage. Switching to the quenched quadrupole transition makes the sidebands resolved, $\omega_z/\gamma_{\mathrm{eff}}=1000/20=50\gg1$ , so equation (5) gives the floor

\bar n_{\mathrm{ss}}\approx\left(\frac{\gamma_{\mathrm{eff}}}{2\omega_z}\right)^2=\left(\frac{20}{2\times1000}\right)^2=(0.01)^2=\mathbf{1\times10^{-4}}.

Four orders of magnitude below the Doppler floor, and well under the $\bar n<0.05$ benchmark for a good motional bus.

Step 3. Ground-state population. For a thermal residual state the ground fraction is $P_0=1/(1+\bar n)=1/(1+10^{-4})=\mathbf{0.9999}$ . The ion spends 99.99% of its time in $\lvert n=0\rangle$ — precisely the cold, definite bus the Cirac–Zoller and Mølmer–Sørensen gates require. Doppler leaves n̄ ≈ 11; resolved-sideband cooling reaches n̄ ≈ 1×10⁻⁴, i.e. P₀ ≈ 99.99%.

3 · Reading the temperature off the sidebands (thermometry)

After cooling, a fixed-area probe pulse excites the red sideband with probability $P_{\mathrm{rsb}}=0.05$ and the blue sideband with $P_{\mathrm{bsb}}=0.55$ . What is the mean phonon number?

Step 1. Apply the calibration-free formula (7): $\bar n=P_{\mathrm{rsb}}/(P_{\mathrm{bsb}}-P_{\mathrm{rsb}})$ . The unknown pulse area and $\eta$ cancel in the ratio, so the two numbers alone determine $\bar n$ .

Step 2. Substitute: $\bar n=0.05/(0.55-0.05)=0.05/0.50=\mathbf{0.10}$ . The strong red–blue asymmetry (a red sideband eleven times weaker than the blue) signals a nearly cold ion; had the ion been in the ground state the red sideband would have vanished entirely. n̄ = 0.10, read directly from the sideband asymmetry with no calibration.

Physics of the portrait

Sideband cooling is one man's idea carried, over four decades, from a two-page proposal to the beating heart of every trapped-ion quantum computer. David Wineland proposed laser cooling of bound ions before it had ever been done, demonstrated it, wrote down its limits, and led the group that first reached the motional ground state — work recognized with the 2012 Nobel Prize.

Origins · 1 minDavid Wineland and how this was discovered

David Wineland

born 1944 · laser cooling and trapped-ion quantum control

David J. Wineland is the physicist who turned the trapped ion into the most precisely controlled object in physics. In 1975, with Hans Dehmelt, he proposed that light scattered by a bound ion could carry away its motional energy — laser cooling by driving the lower motional sideband — three years before anyone cooled an ion at all. In 1978 his group at NBS (now NIST) in Boulder demonstrated laser cooling of trapped $\mathrm{Mg}^+$ , and the following year he and Wayne Itano laid out the theory of the cooling limits, including the resolved-sideband floor $\bar n\approx(\gamma/2\omega_z)^2$ . The decisive experiment came in 1989, when Diedrich, Bergquist, Itano, and Wineland cooled a single $\mathrm{Hg}^+$ ion to its motional ground state for the first time, verifying it by the disappearance of the red sideband. When Cirac and Zoller proposed in 1995 that a shared, ground-state-cooled phonon mode could entangle ion qubits, Wineland's group implemented the first such quantum logic gate within months — welding sideband cooling permanently to quantum computing. His career, from the hydrogen maser through optical clocks to quantum information, is a sustained argument that if you can cool and control a particle's every quantum, you can measure and compute with it better than any other way. He shared the 2012 Nobel Prize in Physics with Serge Haroche for these methods of measuring and manipulating individual quantum systems.

History

1975

— Wineland and Dehmelt propose laser cooling of bound ions via the motional sideband

1978

— first laser cooling of trapped ions demonstrated at NBS Boulder and by Neuhauser–Dehmelt

1979

— Wineland and Itano work out the laser-cooling limits, including the resolved-sideband floor

1989

— Diedrich, Bergquist, Itano, Wineland cool a single Hg⁺ ion to its motional ground state

1995

— Monroe et al. achieve Raman sideband cooling of Be⁺ to the 3D ground state; Cirac–Zoller gate proposed

Going further

Texts and reviews. D. Leibfried, R. Blatt, C. Monroe, D. Wineland, "Quantum dynamics of single trapped ions", Rev. Mod. Phys. 75, 281 (2003) — the definitive review of the formalism used here · D. Wineland, W. Itano, "Laser cooling of atoms", Phys. Rev. A 20, 1521 (1979) — original cooling-limit theory · F. Diedrich, J. Bergquist, W. Itano, D. Wineland, "Laser cooling to the zero-point energy of motion", Phys. Rev. Lett. 62, 403 (1989) — first ground-state cooling · C. Monroe et al., "Resolved-sideband Raman cooling of a bound atom to the 3D zero-point energy", Phys. Rev. Lett. 75, 4011 (1995) · H. Häffner, C. Roos, R. Blatt, "Quantum computing with trapped ions", Phys. Rep. 469, 155 (2008) · Nielsen & Chuang, Quantum Computation and Quantum Information, Ch. 7 (physical implementations). The key relations in this article are: the Lamb–Dicke parameter $\eta=k\sqrt{\hbar/2m\omega_z}\approx0.10$ for $^{40}\mathrm{Ca}^+$ at 729 nm with $\omega_z/2\pi=1$ MHz; the sideband Rabi scaling $\Omega_{n,n-1}=\Omega\eta\sqrt{n}$ ; the resolved-sideband floor $\bar n_{\mathrm{ss}}\approx(\gamma/2\omega_z)^2$ ; the Doppler relation $k_B T_D=\hbar\gamma/2\Rightarrow\bar n\approx\gamma/2\omega_z$ ; and the thermometry formula $\bar n=P_{\mathrm{rsb}}/(P_{\mathrm{bsb}}-P_{\mathrm{rsb}})$ . The cooling-budget numbers in Worked Example 2 use representative values ( $\gamma_{397}/2\pi=22.4$ MHz, $\gamma_{\mathrm{eff}}/2\pi=20$ kHz) and are order-of-magnitude illustrations, not a specific published dataset.

Prerequisites. Quantum Harmonic Oscillator · Trapped-Ion Qubits · Two-Level Atom and Rabi Oscillations · Spontaneous Emission and Optical Bloch Equations. Leads to. Cirac–Zoller Gate · Mølmer–Sørensen Gate · DiVincenzo Criteria.

Ask the tutor: Ask me to walk through why the red-sideband coupling's $\sqrt{n}$ factor makes the motional ground state dark, or to work a cooling budget for your own ion species and trap frequency.

What is a phonon in a trapped ion, and why do we care about removing them?

A phonon here is one quantum of the ion's vibrational motion in the trap, energy $\hbar\omega_z$ ; the ion's motion is a quantum harmonic oscillator, so its energy comes in these discrete lumps rather than continuously. We care because in a trapped-ion quantum computer the shared vibration of an ion chain is the bus that carries quantum information between qubits during an entangling gate. A gate maps the qubit state onto the motion and back, and if the motion is not in a single, definite state — the ground state $\lvert 0\rangle$ — the mapping is ambiguous and the gate is imperfect. Removing the phonons puts the bus in a clean, known starting state so the gate does exactly what it should.

Why does the red sideband cool while the blue sideband heats?

Tuning the laser one trap-quantum below the transition, to $\omega_0-\omega_z$ , means an absorbed photon is short of the energy needed to excite the ion. The ion makes up the deficit by borrowing a phonon from its own motion, so absorption removes one quantum of vibration: $\lvert g,n\rangle\to\lvert e,n-1\rangle$ . The blue sideband at $\omega_0+\omega_z$ is the mirror image — the photon carries excess energy, which is dumped into the motion, adding a phonon. Since spontaneous emission on average returns the phonon number unchanged, driving red gives net cooling and driving blue gives net heating. Resolving the two sidebands is what lets you pick the cooling one and avoid the heating one.

What exactly is the Lamb–Dicke regime, and why must η be small?

The Lamb–Dicke regime is the condition that the ion's quantum position spread $z_0$ is much smaller than the laser wavelength, so the dimensionless parameter $\eta=k z_0\ll1$ . Physically it means a photon's recoil kick is a tiny fraction of a phonon ( $\eta^2\ll1$ ), so absorbing or emitting a photon almost never changes the motional quantum number by more than one. That is what keeps the spectrum simple: only the carrier and the first red and blue sidebands matter, and the clean $\Delta n=-1$ cooling ratchet works. If $\eta$ approached one, higher sidebands and multi-phonon recoil events would proliferate, scrambling the motion faster than the red sideband could cool it.

How do you know when you have actually reached the ground state?

By the vanishing of the red sideband. In a thermal state the red-sideband signal is proportional to $\bar n$ and the blue-sideband signal to $\bar n+1$ ; taking their ratio gives $\bar n=P_{\mathrm{rsb}}/(P_{\mathrm{bsb}}-P_{\mathrm{rsb}})$ with no calibration needed. As the ion approaches $\lvert 0\rangle$ the red sideband fades toward zero — because its coupling scales as $\sqrt{n}$ — while the blue sideband stays strong. A red-to-blue ratio near zero is the direct, model-free announcement that the ion is in its motional ground state. This is the measurement Diedrich and Wineland used in 1989.

Why must you Doppler-cool first instead of jumping straight to sideband cooling?

Because sideband cooling only behaves simply once the ion is already inside the Lamb–Dicke regime. A hot ion has a large motional amplitude, so $\eta\sqrt{n}$ is no longer small, higher-order sidebands are strong, and the tidy first-order picture fails. Doppler cooling, using a broad dipole transition, is fast and robust and brings the ion from room-temperature loading down to $\bar n$ of order ten — cold enough that $\eta\sqrt{n}\ll1$ . Only then does resolved-sideband cooling on a narrow transition take over to remove the last handful of phonons. The two stages are complementary: broad-and-fast first, narrow-and-deep second.

Does the same idea really work for macroscopic mechanical oscillators?

Yes — that is one of the most striking generalizations in the field. Cavity optomechanics couples a laser-driven optical cavity to a mechanical oscillator (a suspended mirror, a membrane, a nanobeam) whose motion shifts the cavity frequency, producing exactly the same red and blue motional sidebands. Driving the red sideband of the cavity scatters photons that carry away mechanical quanta, cooling the oscillator toward its ground state, with the same resolved-sideband condition (cavity linewidth much less than mechanical frequency) and the same residual floor set by the sideband ratio. Groups have cooled oscillators from billions of phonons to below one this way. The physics is identical; only the mass and frequency scales differ.

Is the cooling floor n̄ss ≈ (γ/2ω_z)² a fundamental limit?

It is the fundamental limit for a given cooling transition in the resolved-sideband, weak-drive regime — the irreducible floor from off-resonant scattering into the heating sideband. But it is not a fixed number: you lower it by choosing a narrower transition (smaller $\gamma$ ) or a stiffer trap (larger $\omega_z$ ), and it drops quadratically. This is why experiments use narrow quadrupole lines or two-photon Raman transitions, whose effective linewidth can be made very small, and push trap frequencies into the megahertz range. In practice other effects — anomalous trap heating from noisy electrodes, laser intensity noise, and off-resonant carrier excitation at strong drive — often set the achieved $\bar n$ before this theoretical floor does.