Dipole Force and the Optical Dipole Potential

Point a laser slightly off an atom's resonance and something remarkable happens: the light stops merely pushing the atom and starts holding it. This article takes the AC Stark shift — the light-induced shift of atomic energy levels from aq-optics-ac-stark-dressed-states — and shows that when that shift depends on where the atom sits, it becomes a potential landscape the atom can be trapped in. The beam's oscillating field induces a dipole $\mathbf{p}=\alpha\mathbf{E}$ ; the cycle-averaged energy of that dipole, $U_{\mathrm{dip}}\propto-\mathrm{Re}(\alpha)\,I(\mathbf{r})$ , depends on position only through the intensity, and its gradient is a conservative force — the dipole, or gradient, force. This is a different animal from the scattering force of aq-optics-radiation-pressure-scattering-force, which comes from the momentum of absorbed and re-emitted photons: dissipative, saturating near resonance, and the basis of laser cooling. The dipole force does no net work around a closed loop — it can build a bowl to hold atoms but cannot damp their motion — and its sign is set entirely by the detuning $\Delta=\omega_L-\omega_0$ . Red-detuned light ( $\Delta<0$ ) lowers the ground-state energy where the beam is bright, drawing atoms to intensity maxima — a laser focus becomes a trap; blue-detuned light ( $\Delta>0$ ) raises it, pushing atoms into the dark. The whole reason this is useful is a mismatch in scalings: trap depth falls as $I/\Delta$ but heating from residual photon scattering falls as $I/\Delta^2$ . Detune far, raise the power, and you get a nearly lossless conservative trap. Optical dipole traps, lattices, and tweezers built on this one idea carry atoms to Bose–Einstein condensation, realise optical lattices for quantum simulation, and assemble single-atom qubit arrays.

Key takeaways

An atom in a light field acquires an induced dipole $\mathbf{p}=\alpha\mathbf{E}$ ; that dipole in the intensity gradient is a conservative force — the dipole force, distinct from the dissipative scattering force.
The force is the gradient of a potential $U_{\mathrm{dip}}(\mathbf{r})\propto-\mathrm{Re}(\alpha)\,I(\mathbf{r})$ equal to the AC Stark (light) shift of the ground state.
Red detuning ( $\Delta<0$ ) makes it attractive toward intensity maxima; blue detuning ( $\Delta>0$ ) makes it repulsive toward minima.
Depth scales as $I/\Delta$ while the heating scattering rate scales as $I/\Delta^2$ , so far detuning plus high intensity gives a nearly conservative far-off-resonance trap (FORT).
Dipole traps give spin-independent trapping — the platform for all-optical evaporation to BEC, optical lattices, tweezer arrays, and Feshbach studies.
Key idea: the dipole force is the gradient of the light-shifted ground-state energy, $\mathbf{F}=-\nabla U_{\mathrm{dip}}$ with $U_{\mathrm{dip}}=-\tfrac{1}{2\varepsilon_0 c}\,\mathrm{Re}(\alpha)\,I$ ; its sign follows the sign of the detuning, and because depth ( $\propto I/\Delta$ ) and heating ( $\propto I/\Delta^2$ ) scale differently, one can build a nearly conservative trap far from resonance.

The core insight

The dipole force is the gradient of the light-shifted ground-state energy: $\mathbf{F}=-\nabla U_{\mathrm{dip}}$ , with $U_{\mathrm{dip}}(\mathbf{r})=-\tfrac{1}{2\varepsilon_0 c}\,\mathrm{Re}(\alpha)\,I(\mathbf{r})$ . Its sign follows the sign of the detuning, and depth ( $\propto I/\Delta$ ) versus heating ( $\propto I/\Delta^2$ ) scale differently, enabling nearly conservative traps far from resonance.

Where this lives today

Optical dipole traps are the quiet backbone of modern atomic physics. In an all-optical BEC machine, atoms pre-cooled in a magneto-optical trap are loaded into a crossed pair of far-detuned beams; ramping down the power lets the hottest atoms escape over the lowered lip — forced evaporation that reaches quantum degeneracy in seconds, holding every magnetic sublevel equally, which is what makes spinor condensates and two-component Fermi gases possible. Because the trap ignores the internal state, a bias field can be swept across a Feshbach resonance to tune interactions from weak to unitary — the standard route to the BEC–BCS crossover. The same force, sculpted differently, powers quantum simulation and computing. Retro-reflected beams build optical lattices whose depth in recoil units $E_R=\hbar^2k^2/2m$ (for ⁸⁷Rb at 1064 nm, $E_R/h\approx2.0$ kHz) is set by laser power, realising Hubbard models, Mott insulators, and quantum-gas microscopes. At the sharpest focus, arrays of diffraction-limited tweezers — steered by acousto-optic deflectors and rearranged in real time — hold individual atoms as qubits, the platform behind today's neutral-atom Rydberg quantum processors.

The physical picture

Treat the atom as a tiny driven oscillator. The laser forces its bound electron to swing at $\omega_L$ , and the induced dipole $\mathbf{p}=\alpha\mathbf{E}$ carries an energy $-\tfrac12\langle\mathbf{p}\cdot\mathbf{E}\rangle$ in the field. Below resonance ( $\Delta<0$ ) the oscillator swings in phase with the drive, so the dipole lowers its energy where the field is strong — the atom is pulled toward the bright spot. Above resonance ( $\Delta>0$ ) it responds out of phase, its energy rises in bright regions, and it is expelled. That in-phase, dispersive part is $\mathrm{Re}(\alpha)$ ; the quadrature part $\mathrm{Im}(\alpha)$ is what absorbs photons. Put another way, the light shift lowers (red) or raises (blue) the ground-state energy wherever the beam is bright, and a ball rolls downhill. Because it is energy, not momentum kicks, that varies in space, the force is conservative — you can shape a well and hold atoms, but the same static field cannot remove their kinetic energy.

Figure

A focused Gaussian beam makes a red-detuned potential well of depth U₀ ≈ 130 μK that pulls atoms to the bright focus, while the same beam blue-detuned makes a repulsive hump that pushes them to the dark — the sign of the light shift follows the sign of the detuning Δ

The two figures of merit trade off through the detuning. Push $|\Delta|$ larger and $U\propto I/\Delta$ shrinks linearly while the destructive scattering $R_{\mathrm{sc}}\propto I/\Delta^2$ shrinks quadratically; raise $I$ to compensate and you keep the trap while quenching the heating — which is why FORTs run tens or hundreds of terahertz from resonance at watt-level powers. Geometry does the rest: a single beam gives a cigar; two crossed beams a tight dimple; a standing wave the periodic wells of an optical lattice; blue-detuned hollow beams box atoms in the dark.

Figure

Holding intensity fixed and detuning farther from resonance, the trap depth U falls only as one over Δ while the scattering rate R_sc falls as one over Δ² — the widening gap is the far-off-resonance win, and at the 1064 nm rubidium point the atom scatters ≈ 1 photon per second while sitting in a ≈ 130 μK well

The core equations

Because the potential depends on position only through the intensity, the force is its gradient — conservative, curl-free, trapping but not dissipating.

\mathbf{F}_{\mathrm{dip}}(\mathbf{r})=-\nabla U_{\mathrm{dip}}(\mathbf{r}) \qquad (1)

where: $\mathbf{F}_{\mathrm{dip}}$ — conservative dipole (gradient) force; $U_{\mathrm{dip}}$ — dipole potential, equal to the ground-state light shift; $\mathbf{r}$ — atomic centre-of-mass position. Physical reading: the atom rolls downhill in $U_{\mathrm{dip}}$ , finding equilibrium at an energy extremum of the intensity. The potential is the cycle-averaged energy of the induced dipole. Only the real (in-phase) part of the polarizability carries conservative energy; the imaginary part appears separately as absorption.

U_{\mathrm{dip}}(\mathbf{r})=-\frac{1}{2\varepsilon_0 c}\,\mathrm{Re}(\alpha)\,I(\mathbf{r}) \qquad (2)

where: $\alpha$ — complex polarizability, $\mathbf{p}=\alpha\mathbf{E}$ ; $\mathrm{Re}(\alpha)$ — dispersive (in-phase) response setting the potential; $I(\mathbf{r})$ — local intensity, $I=\tfrac12\varepsilon_0 c\,|E|^2$ ; $\varepsilon_0,\,c$ — permittivity and light speed. Physical reading: below resonance $\mathrm{Re}(\alpha)>0$ , so $U_{\mathrm{dip}}<0$ where $I$ is large — an attractive well at the focus. The same polarizability's imaginary part absorbs. Dividing the absorbed power by $\hbar\omega$ gives the residual scattering rate — the sole source of heating in a dipole trap.

R_{\mathrm{sc}}(\mathbf{r})=\frac{1}{\hbar\varepsilon_0 c}\,\mathrm{Im}(\alpha)\,I(\mathbf{r}) \qquad (3)

where: $R_{\mathrm{sc}}$ — scattering rate (absorption then spontaneous emission); $\mathrm{Im}(\alpha)$ — absorptive (out-of-phase) part; $\hbar$ — reduced Planck constant. Physical reading: each scattered photon is a random recoil kick, so $R_{\mathrm{sc}}$ sets the heating rate — trap design makes it negligible while keeping $U_{\mathrm{dip}}$ deep. For a two-level atom far from resonance and below saturation, the Lorentz-oscillator polarizability collapses to a compact potential — the everyday FORT formula.

U_{\mathrm{dip}}(\mathbf{r})=\frac{\hbar\Gamma^{2}}{8\Delta}\,\frac{I(\mathbf{r})}{I_{\mathrm{sat}}}\qquad (|\Delta|\gg\Gamma,\ s\ll1) \qquad (4)

where: $\Delta=\omega_L-\omega_0$ — detuning (its sign sets the trap type); $\Gamma$ — natural linewidth; $I_{\mathrm{sat}}=\pi h c\,\Gamma/3\lambda^3$ — saturation intensity; $s=I/I_{\mathrm{sat}}$ — saturation parameter (small). Physical reading: $\Delta<0$ gives $U_{\mathrm{dip}}<0$ (red, attractive), $\Delta>0$ gives $U_{\mathrm{dip}}>0$ (blue, repulsive). The scattering rate in the same limit reveals the crucial ratio — at fixed depth it falls one extra power of $\Delta$ faster.

R_{\mathrm{sc}}(\mathbf{r})=\frac{\Gamma^{3}}{8\Delta^{2}}\,\frac{I(\mathbf{r})}{I_{\mathrm{sat}}}=\frac{\Gamma}{\hbar\Delta}\,U_{\mathrm{dip}}(\mathbf{r}) \qquad (5)

where: $R_{\mathrm{sc}}$ — far-detuned scattering rate; $\Gamma/\Delta$ — small parameter suppressing heating. Physical reading: $\hbar R_{\mathrm{sc}}/U_{\mathrm{dip}}=\Gamma/\Delta$ , so at fixed depth heating drops as $1/\Delta$ — detune farther, raise the power, and the trap becomes ever more conservative. Real traps use focused Gaussian beams, so the potential has a spatial profile whose curvature at the bottom sets the oscillation frequencies.

U(r,z)=-U_0\,\frac{1}{1+(z/z_R)^2}\,\exp\!\left[-\frac{2r^2}{w_0^2\,(1+(z/z_R)^2)}\right] \qquad (6)

where: $U_0=|U_{\mathrm{dip}}(\text{peak})|$ — depth at the focus; $w_0$ — beam waist (1/e² intensity radius); $z_R=\pi w_0^2/\lambda$ — Rayleigh range; $r,z$ — radial and axial coordinates. Physical reading: the well is deepest at the focus and elongated along the beam by $z_R/w_0\gg1$ , which is why single-beam traps are cigar-shaped.

Derivation AThe dipole potential from the induced-dipole interaction energy

The goal is to turn "an atom in a light field has some energy" into the exact formula $U_{\mathrm{dip}}=-\tfrac{1}{2\varepsilon_0 c}\,\mathrm{Re}(\alpha)\,I$ , and to make transparent why only the real part of the polarizability contributes to the conservative energy while the imaginary part is sidelined into absorption. We start from the interaction energy of an induced dipole, insert the linear response $\mathbf{p}=\alpha\mathbf{E}$ , average over the optical cycle, and convert the field amplitude into intensity.

Step 1. Write the interaction energy as $U=-\tfrac12\langle\mathbf{p}\cdot\mathbf{E}\rangle$ , the brackets denoting a time average over one optical cycle. Why the factor one-half: the dipole is induced, not permanent — as the field ramps up the moment grows with it, and integrating $-\mathbf{p}\cdot d\mathbf{E}$ from zero costs half of what a rigid dipole would, as a spring stores $\tfrac12 kx^2$ .

Step 2. Substitute the linear response $\mathbf{p}=\alpha\mathbf{E}$ for a field $\mathbf{E}=\mathrm{Re}[\mathbf{E}_0 e^{-i\omega t}]$ , and use $\langle E^2\rangle=\tfrac12|E_0|^2$ for the cycle average. Why: the polarizability is complex because the driven electron lags the drive; only the part of $\mathbf{p}$ in phase with $\mathbf{E}$ survives the average, and that is $\mathrm{Re}(\alpha)$ .

Step 3. Collect the result: $U=-\tfrac14\,\mathrm{Re}(\alpha)\,|E_0|^2=-\tfrac{1}{2\varepsilon_0 c}\,\mathrm{Re}(\alpha)\,I$ , using $I=\tfrac12\varepsilon_0 c\,|E_0|^2$ . Why this is the payoff: the energy depends on the field only through $I(\mathbf{r})$ , so its gradient is a genuine mechanical force — the dipole force — and $\mathrm{Im}(\alpha)$ has dropped out entirely, reappearing only in the absorbed power that drives $R_{\mathrm{sc}}$ . *Verified: $U_{\mathrm{dip}}=-\tfrac{1}{2\varepsilon_0 c}\,\mathrm{Re}(\alpha)\,I$ matches Grimm et al. (2000) and Foot Ch. 9; only $\mathrm{Re}(\alpha)$ enters the conservative energy.*

Go deeper · Derivation B · 1 minDressed states and the light shift

Derivation BDressed states and the light shift

The goal is to derive the same potential from the quantum two-level atom, showing that $U_{\mathrm{dip}}$ is literally the shift of the ground-state energy when the atom is dressed by the laser — the AC Stark shift — and to recover equation (4). Diagonalizing the atom-plus-field Hamiltonian in the rotating frame gives the dressed-state energies; the branch connecting adiabatically to the bare ground state carries the trapping potential, its sign fixed by the detuning.

Step 1. In the rotating-wave approximation, in the frame rotating at the laser frequency, the two-level Hamiltonian in the basis $\{|g\rangle,|e\rangle\}$ is $H=\tfrac{\hbar}{2}\begin{pmatrix}0 & \Omega\\ \Omega & -2\Delta\end{pmatrix}$ , where $\Omega=dE/\hbar$ is the Rabi frequency and $d$ the dipole matrix element. Why: the rotating frame removes the fast optical oscillation, leaving a static two-by-two problem whose eigenvalues are the light-shifted energies; $\Omega^2\propto I$ carries the intensity dependence.

Step 2. Diagonalize to get the dressed energies $E_\pm=-\tfrac{\hbar\Delta}{2}\pm\tfrac{\hbar}{2}\sqrt{\Delta^2+\Omega^2}$ . Why: the square root is the hallmark of an avoided crossing — the levels repel most strongly near resonance, which is why the light shift is large close in and small far away.

Step 3. The branch connected to $|g\rangle$ is shifted by $\delta E_g=\tfrac{\hbar\Delta}{2}\big(\sqrt{1+\Omega^2/\Delta^2}-1\big)$ , which is $U_{\mathrm{dip}}$ . Why the sign matters: for $\Delta<0$ the shift is negative where $\Omega$ (hence $I$ ) is large, pulling the ground state down at the bright focus — attractive; for $\Delta>0$ it is pushed up and repelled.

Step 4. Expand for $|\Delta|\gg\Omega$ : $\delta E_g\approx\hbar\Omega^2/4\Delta$ , and with $\Omega^2/\Gamma^2=I/2I_{\mathrm{sat}}$ this becomes $U_{\mathrm{dip}}=\hbar\Gamma^2 I/(8\Delta I_{\mathrm{sat}})$ . Why: this reproduces equation (4) exactly, tying the dressed-state picture to the practical FORT formula. *Verified: reduces to the two-level far-detuned formula; matches Foot Ch. 9 and Grimm et al. (2000); $\Delta<0$ gives an attractive potential.*

Go deeper · Derivation C · 1 minFrom Gaussian beam to trap frequencies

Derivation CFrom Gaussian beam to trap frequencies

The goal is to predict how tightly a trapped atom oscillates — the numbers an experiment measures — by expanding the Gaussian potential of equation (6) about its minimum. Near the bottom every smooth well looks harmonic, so the curvature of $U(r,z)$ at the focus sets a radial frequency $\omega_r$ and an axial frequency $\omega_z$ , whose large ratio quantifies the cigar shape of a single-beam trap.

Step 1. Taylor-expand $U(r,z)$ about $r=z=0$ to second order, using $e^{-2r^2/w_0^2}\approx 1-2r^2/w_0^2$ and $1/(1+(z/z_R)^2)\approx 1-(z/z_R)^2$ , giving $U\approx -U_0\big(1-2r^2/w_0^2-z^2/z_R^2\big)$ . Why: only the quadratic terms survive at the bottom, so near equilibrium the atom feels a linear restoring force — harmonic confinement.

Step 2. Match the radial term to a harmonic oscillator, $\tfrac12 m\omega_r^2 r^2=U_0\,(2r^2/w_0^2)$ , giving $\omega_r=\sqrt{4U_0/(m w_0^2)}$ ; the axial term gives $\omega_z=\sqrt{2U_0/(m z_R^2)}$ . Why: reading the spring constant off the curvature turns a potential minimum into a measurable oscillation frequency, set here by $U_0$ and the length scales $w_0$ and $z_R$ .

Step 3. Form the aspect ratio $\omega_r/\omega_z=\sqrt{2}\,z_R/w_0=\sqrt{2}\,\pi w_0/\lambda$ . Why: since $z_R=\pi w_0^2/\lambda\gg w_0$ , the radial frequency vastly exceeds the axial one — a single beam is a weakly confined cigar along its axis, which is why experiments cross a second beam for a tight three-dimensional dimple. *Verified: $\omega_r=\sqrt{4U_0/mw_0^2}$ , $\omega_z=\sqrt{2U_0/mz_R^2}$ by direct Taylor expansion of equation (6); aspect ratio $\sqrt2\,z_R/w_0$ follows from $z_R=\pi w_0^2/\lambda$ .*

Limiting cases

The far-detuned two-level potential $U_{\mathrm{dip}}=\tfrac{\hbar\Gamma^2}{8\Delta}\tfrac{I}{I_{\mathrm{sat}}}$ shows its behaviour cleanly in the regimes below.

Case	Result	Meaning
Red detuning $\Delta<0$	$U_{\mathrm{dip}}<0$	attractive at high intensity: atoms trapped at intensity maxima, e.g. a Gaussian focus
Blue detuning $\Delta>0$	$U_{\mathrm{dip}}>0$	repulsive at high intensity: atoms pushed to intensity minima (dark traps), minimising heating
Far-detuned $\vert\Delta\vert\gg\Gamma$	$U_{\mathrm{dip}}/\hbar R_{\mathrm{sc}}=\Delta/\Gamma$	depth-to-heating ratio grows linearly with detuning — the FORT regime, nearly conservative
Near resonance $\vert\Delta\vert\lesssim\Gamma$	$R_{\mathrm{sc}}\sim\Gamma$ (saturated)	scattering dominates, the force turns dissipative, and the $I/\Delta$ scaling fails (full Lorentzian and saturation required)
Static limit $\omega_L\to0$	$U\to-\tfrac12\alpha_0 E^2$	the DC Stark / dielectric-force limit; $\alpha_0$ the static polarizability

Common trapConfusing the conservative dipole force with the dissipative scattering force

These are two physically distinct forces and mixing them is the classic error. The dipole force is conservative, $\mathbf{F}=-\nabla U_{\mathrm{dip}}$ ; it does no net work around a closed loop, scales as $\nabla I$ , points toward bright or dark by the sign of $\Delta$ , and is strongest off resonance — it traps but never cools. The scattering force comes from photon momentum, points along $\mathbf{k}$ , saturates near resonance, and is the one made velocity-dependent for laser cooling. Two further traps: $U\propto I/\Delta$ is only the far-detuned, unsaturated limit — near resonance keep the full Lorentzian and watch for saturation; and the depth-to-heating ratio for a real 1064 nm rubidium trap is $\Gamma/\Delta\approx6\times10^{-8}$ , not the $10^{-4}$ sometimes misquoted from confusing "detuning in gigahertz" with "detuning in units of $\Gamma$ ".

Worked examples

1 · A real far-off-resonance trap for rubidium (confidence builder)

A 1064 nm laser of power $P=1$ W is focused to a waist $w_0=25\ \mu\mathrm{m}$ onto ⁸⁷Rb (D2 line $\lambda_0=780$ nm, $\Gamma/2\pi=6.07$ MHz, $I_{\mathrm{sat}}=1.67\ \mathrm{mW/cm^2}$ ). Find the peak intensity, the trap depth, the radial and axial trap frequencies, and the photon-scattering rate.

Step 1. Peak Gaussian intensity is $I_0=2P/(\pi w_0^2)=2(1)/[\pi(25\times10^{-6})^2]=\mathbf{1.02\times10^{9}\ W/m^2}$ $(=1.02\times10^{5}\ \mathrm{W/cm^2})$ ; the peak is twice the area-averaged value.

Step 2. The detuning is enormous: $\Delta/2\pi=(c/\lambda_L)-(c/\lambda_0)=281.8-384.3=-102.6$ THz, so equation (4) applies. With $I_0/I_{\mathrm{sat}}=6.1\times10^7$ , the depth is $U_0=\dfrac{\hbar\Gamma^2}{8|\Delta|}\dfrac{I_0}{I_{\mathrm{sat}}}=1.81\times10^{-27}\ \mathrm{J}$ , i.e. $U_0/k_B=\mathbf{131\ \mu K}$ . (The full D1+D2 doublet and counter-rotating terms deepen this by 15–20% in practice.)

Step 3. Trap frequencies, with $m=87\,u=1.44\times10^{-25}$ kg and $z_R=\pi w_0^2/\lambda_L=1.85$ mm: $\omega_r=\sqrt{4U_0/mw_0^2}$ gives $\nu_r=\mathbf{1.43\ kHz}$ ; $\omega_z=\sqrt{2U_0/mz_R^2}$ gives $\nu_z=\mathbf{13.7\ Hz}$ . The aspect ratio $\nu_r/\nu_z\approx100$ is the cigar shape.

Step 4. Scattering from equation (5): $R_{\mathrm{sc}}=(\Gamma/|\Delta|)\,(U_0/\hbar)$ with $\Gamma/|\Delta|=5.9\times10^{-8}$ gives $R_{\mathrm{sc}}=\mathbf{1.0\ photon/s}$ . A watt of light in a 25 μm waist makes a 131 μK trap oscillating at 1.4 kHz radially, in which an atom scatters only about one photon per second — conservative confinement for seconds.

2 · How far to detune, at fixed trap depth

You need a red-detuned trap $U_0/k_B=200\ \mu\mathrm{K}$ deep on the rubidium D2 line. Compare running it at $\Delta/2\pi=-10$ THz versus $\Delta/2\pi=-100$ THz.

Step 1. At fixed depth, equation (4) forces $I\propto\Delta$ : detuning ten times farther (10 → 100 THz) demands ten times more intensity to hold the same 200 μK — the far trap is power-hungry.

Step 2. Scattering at fixed depth follows $\hbar R_{\mathrm{sc}}=(\Gamma/\Delta)\,U_0$ , so $R_{\mathrm{sc}}\propto1/\Delta$ . With $U_0/\hbar=3.44\times10^7$ rad/s: at $\Delta/2\pi=-10$ THz, $R_{\mathrm{sc}}=\mathbf{15.9\ photon/s}$ ; at $-100$ THz, $R_{\mathrm{sc}}=\mathbf{1.6\ photon/s}$ .

Step 3. The far trap costs ten times the power but heats ten times more slowly for the identical depth. Detuning farther deliberately trades power for coherence — you spend watts to buy a nearly lossless trap, which is why FORTs use high-power infrared lasers far from resonance.

Physics of the portrait

The dipole force sits between Maxwell's radiation pressure and the quantum light shift, and one experimenter turned it into a tool that now holds single atoms, cells, and viruses.

Origins · 1 minArthur Ashkin and how this was discovered

Arthur Ashkin

1922–2020 · optical trapping and the gradient force

Arthur Ashkin spent his career at Bell Labs asking whether the momentum and energy of light could push matter around in a controlled way. In 1970 he showed that a focused laser beam exerts two distinct forces on a small dielectric particle — a scattering force along the beam and a gradient force toward highest intensity — and used them to levitate micron-sized spheres. That gradient force is the macroscopic cousin of the atomic dipole force: both pull an induced polarization up an intensity gradient. Ashkin realised the same physics should trap neutral atoms, and in a landmark 1986 experiment with Steven Chu, John Bjorkholm and Alex Cable he trapped sodium in a single focused red-detuned beam — the first optical dipole trap. That same year he built the single-beam gradient trap now called optical tweezers, which went on to grip bacteria, measure the piconewton forces of molecular motors, and stretch single DNA strands. For decades the atomic dipole potential and the optical tweezer were separate fields; today they have merged, as single-atom tweezer arrays assemble defect-free qubit registers atom by atom. Ashkin received the 2018 Nobel Prize at the age of 96, the oldest laureate at the time, for a force he first measured almost half a century earlier.

History

1961

— Cohen-Tannoudji measures the light shift, the AC Stark shift of atomic energy levels

1968

— Letokhov proposes confining atoms in a standing light wave using the dipole force

1970

— Ashkin demonstrates optical acceleration and trapping of particles by the gradient force

1978

— Ashkin and Gordon analyse dipole traps and radiation forces on neutral atoms

1986

— Chu, Bjorkholm, Ashkin and Cable trap sodium in a single focused red-detuned beam

2001

— all-optical Bose-Einstein condensation is achieved in a CO₂-laser dipole trap

Going further

Texts. Grimm, Weidemüller and Ovchinnikov, Optical Dipole Traps for Neutral Atoms, Adv. At. Mol. Opt. Phys. 42, 95 (2000) — the definitive review · Foot, Atomic Physics (2005), Ch. 9 · Metcalf and van der Straten, Laser Cooling and Trapping (1999) · Loudon, The Quantum Theory of Light, 3rd ed. · Cohen-Tannoudji, Dupont-Roc and Grynberg, Atom–Photon Interactions · Saleh and Teich, Fundamentals of Photonics, 2nd ed.

Provenance. The core formulas — $U_{\mathrm{dip}}=-\tfrac{1}{2\varepsilon_0 c}\mathrm{Re}(\alpha)I$ , $R_{\mathrm{sc}}=\tfrac{1}{\hbar\varepsilon_0 c}\mathrm{Im}(\alpha)I$ , the Lorentz-oscillator polarizability, and the far-detuned two-level $U$ and $R_{\mathrm{sc}}$ — follow Grimm et al. (2000) and Foot Ch. 9; the dressed-state derivation is checked against Cohen-Tannoudji et al. and Metcalf and van der Straten. Numerical values use standard ⁸⁷Rb D-line data ( $\lambda_0=780$ nm, $\Gamma/2\pi=6.07$ MHz, $I_{\mathrm{sat}}=1.67\ \mathrm{mW/cm^2}$ ; Steck). A verifier should confirm: (1) $U_{\mathrm{dip}}=\hbar\Gamma^2 I/(8\Delta I_{\mathrm{sat}})$ from the dressed-state light shift; (2) $\hbar R_{\mathrm{sc}}/U_{\mathrm{dip}}=\Gamma/\Delta$ ; (3) the worked-example chain $I_0=1.02\times10^9\ \mathrm{W/m^2}$ , $U_0/k_B=131\ \mu K$ , $\nu_r=1.43$ kHz, $\nu_z=13.7$ Hz, $R_{\mathrm{sc}}=1.0/\mathrm{s}$ for $P=1$ W, $w_0=25\ \mu$ m; and (4) the corrected ratio $\Gamma/\Delta\approx6\times10^{-8}$ for the 1064 nm rubidium trap (the earlier $10^{-4}$ figure was an error). No fabricated constants are used.

Prerequisites. Radiation Pressure and the Scattering Force · AC Stark Shift and Dressed States · Atomic Polarizability. Leads to. Evaporative Cooling · Bose-Einstein Condensation · Feshbach Resonances · Optical Molasses · BEC-BCS Crossover.

Ask the tutor: Ask me to size a crossed-beam dipole trap for your atom and laser, or to work through why a magic wavelength makes the light shift vanish for an optical-lattice clock.

How is the dipole force different from the scattering force if both come from light?

They come from the two parts of the same complex polarizability. The scattering force comes from $\mathrm{Im}(\alpha)$ — light absorbed and re-emitted, each cycle transferring momentum $\hbar k$ along the beam; it is dissipative, saturates near resonance, and drives laser cooling. The dipole force comes from $\mathrm{Re}(\alpha)$ — the induced polarization stores field energy set by the local intensity, and the gradient of that energy is a conservative force pointing toward bright (red) or dark (blue). One pushes and cools; the other traps but cannot. In any real beam both act at once, which is why designers detune far enough that the conservative part dominates.

Why does red detuning attract atoms and blue detuning repel them?

It follows the sign of the light shift. Below resonance ( $\Delta<0$ ) the ground state is pushed down wherever the field is strong, so the atom lowers its energy by moving into the bright region — attraction to intensity maxima. Above resonance ( $\Delta>0$ ) the ground state is pushed up in bright regions, so the atom flees to the dark — repulsion. In the oscillator picture, below resonance the induced dipole swings in phase with the field and is drawn into it; above resonance it swings out of phase and is expelled. The dressed-state derivation makes this exact: the ground-state branch shifts by $\tfrac{\hbar\Delta}{2}(\sqrt{1+\Omega^2/\Delta^2}-1)$ , whose sign is the sign of $\Delta$ .

Why can't a dipole trap cool atoms, only hold them?

Because the force is conservative: $\mathbf{F}=-\nabla U_{\mathrm{dip}}$ is the gradient of a potential, so the work around any closed path is exactly zero. An atom that falls into the well and climbs back out keeps the kinetic energy it started with — the trap stores and reshapes energy but cannot remove it. Cooling needs a dissipative, velocity-dependent force that carries entropy away, which in laser cooling is the spontaneous-emission part of the scattering force. So a dipole trap is always paired with a cooling mechanism: atoms are first cooled in a magneto-optical trap, then loaded into the dipole trap and cooled further by evaporation, which throws energy out with the escaping atoms rather than with light.

What exactly is a FORT and why go so far from resonance?

A FORT is a far-off-resonance trap — a dipole trap run tens or hundreds of terahertz from resonance, at high intensity. The motivation is the scaling mismatch: depth falls as $I/\Delta$ while heating falls as $I/\Delta^2$ , so the heating-to-depth ratio is $\Gamma/\Delta$ . Detune ten times farther and heating per unit depth drops tenfold; you pay with ten times the intensity to keep the well deep. For a 1064 nm rubidium trap, $\Delta/2\pi\approx-100$ THz gives $\Gamma/\Delta\approx6\times10^{-8}$ , so a 130 μK trap scatters only about one photon per second. That near-lossless confinement is what makes all-optical evaporation to BEC and long-lived qubit storage possible.

How deep and how tight is a typical dipole trap, in real numbers?

For 1 W of 1064 nm light focused to a 25 μm waist on rubidium, the depth is about 130 μK, so atoms colder than that stay bound. The radial frequency is about 1.4 kHz but the axial only about 14 Hz, because a single beam is a long cigar: the Rayleigh range (1.85 mm) hugely exceeds the waist. Experiments fix this by crossing a second beam through the focus, making a tight three-dimensional dimple with kilohertz frequencies in all directions. Depth scales linearly with power and inversely with waist area.

Does the dipole force need the atom to be in its ground state?

The everyday FORT picture uses the ground-state light shift, $U_{\mathrm{dip}}\propto\mathrm{Re}(\alpha_g)I$ , and for ground-state atoms far below saturation that is what matters. But every level has its own polarizability, so the excited state is shifted too, generally the opposite way for red detuning. This state dependence is a feature: it underlies state-dependent lattices that trap two internal states in different places, and it forces the choice of "magic" wavelengths where two clock states shift identically, so an optical-lattice clock runs without the trap perturbing the transition.