Optical Conveyor Belt Mechanisms

• Optical conveyor belt is a transport mechanism that uses moving standing-wave dipole potentials to shift trapped particles in space.
• It enables precise transfer of atoms over distances from micrometers to centimeters, bridging free-space and integrated photonic setups.
• Various implementations achieve efficient single-atom loading and controlled adiabatic transport, critical for advancing cold-atom and nanophotonic experiments.

An optical conveyor belt is a transport scheme in which a moving optical potential carries trapped particles or quasiparticles through space. In its standard cold-atom realization, two counter-propagating optical fields form a standing-wave dipole lattice, and a controlled frequency or phase offset translates the lattice sites so that atoms remain confined while being displaced. This architecture has been used to move ultracold atoms over distances from hundreds of micrometers to 37.2 cm, to transfer single atoms from a magneto-optical trap (MOT) to a tightly confined optical tweezer, and to deliver atoms toward nanofibers, hollow-core fibers, and photonic chips (Xu et al., 2023, Matthies et al., 2023, Xu et al., 29 Jul 2025). The same phrase also appears in distinct photonic and molecular contexts, including blue-detuned conveyor-belt MOTs, chiral single-beam transport, plasmonic hot-spot transport, and traveling optical potentials for exciton-polaritons (Yu et al., 2024, Fernandes et al., 2016, Xu et al., 8 Oct 2025).

1. Core operating principle

The canonical optical conveyor belt is a moving standing-wave dipole trap. In one widely used formulation, two linearly polarized, counter-propagating Gaussian beams with equal intensity and a small frequency difference $\Delta\nu$ create a potential

$$U(z,t)=U_0 \cos^2\!\left(\frac{2\pi \Delta\nu}{2}t-\frac{2\pi}{\lambda}z\right),$$

so that atoms trapped at the antinodes are transported as the interference pattern moves. The corresponding lattice velocity is

$v=\frac{1}{2}\lambda \Delta\nu .$

This is the essential conveyor-belt mechanism: transport is produced by translating the standing-wave lattice itself, rather than by mechanically moving an optical element or by pushing atoms with resonant radiation pressure (Xu et al., 2023).

A closely related description appears in single-atom transfer experiments, where the frequency difference $\delta$ between two counterpropagating Gaussian beams is controlled with phase-locked AOMs, giving

$v=\frac{1}{2}\lambda \delta .$

In this form, the conveyor belt functions as a transport stage between spatially separated trapping zones, such as a MOT region and a tightly focused optical tweezer (Xu et al., 29 Jul 2025).

In guided-wave implementations, the same principle survives but the propagation medium modifies the kinematics. For a nanofiber-based optical conveyor belt, the mutual detuning of the two counter-propagating red-detuned guided fields gives

$$v = \frac{1}{2}\,\frac{\lambda_0^{\rm red}}{n_{\rm eff}}\,\Delta \nu_{\rm opt},$$

with the lattice period set by the guided wavelength. In free-space-to-chip proposals, a $2\pi$ phase change shifts the standing-wave lattice by one lattice period and moves the atom from one antinode to the next (Schneeweiss et al., 2012, Liu et al., 2023).

These formulations share a common structure: an optical interference pattern defines a periodic confining potential, and control of relative phase or frequency turns that periodic structure into a translating transport channel.

2. Experimental architectures and transport range

The experimental literature spans free-space, guided-wave, fiber-integrated, and hybrid photonic geometries. The common element is the use of a moving optical potential to bridge a mismatch between where particles can be prepared and where they are needed.

Platform	Transport geometry	Representative result
Long-distance transport of ultracold $^{133}$Cs and $^{87}$Rb atoms	two counter-propagating 1064 nm Gaussian beams with a controllable frequency difference	37.2 cm in under 25 ms; efficiency up to 75% (Matthies et al., 2023)
Transporting cold atoms towards a GaN-on-sapphire chip	two counter-propagating, red-detuned dipole-trap beams	maximum transport efficiency about 50% with a transport distance of 500 $\mu\mathrm m$ (Xu et al., 2023)
Efficient single-atom transfer to a tightly confined optical tweezer	transport from MOT region to an overlap region 0.6 mm away from the tweezer	single-atom loading probability 77.6% with feedback control (Xu et al., 29 Jul 2025)
Nanofiber-based transport of cold cesium atoms	two-color evanescent-field trap surrounding a nanofiber	transport over distances of up to about 0.3 mm; millimeter-scale distances along the nanofiber (Schneeweiss et al., 2012)
Transport into a hollow-core photonic crystal fiber	moving one-dimensional optical lattice formed by two counter-propagating 805 nm laser beams	more than 40% transport efficiency into the fiber (Langbecker et al., 2018)
Levitated nanoparticles inside an HCPCF	standing-wave trap from two counterpropagating 1064 nm laser beams	transport length 15 cm (Grass et al., 2016)

In free space, a notable development is the use of two static Gaussian beams with displaced foci to maintain sufficient trap depth over a 37.2 cm path. With about 18 W per beam at 1064 nm, optimized beam parameters and a minimum jerk trajectory enabled transport of up to $$U(z,t)=U_0 \cos^2\!\left(\frac{2\pi \Delta\nu}{2}t-\frac{2\pi}{\lambda}z\right),$$0 cesium or rubidium atoms with efficiency up to 75%, while preserving conditions compatible with Bose-Einstein condensation after transport (Matthies et al., 2023).

A shorter-distance but technologically important example is transport toward a GaN-on-sapphire chip. There, atoms cooled to about $$U(z,t)=U_0 \cos^2\!\left(\frac{2\pi \Delta\nu}{2}t-\frac{2\pi}{\lambda}z\right),$$1 are loaded into a standing-wave trap formed by two counter-propagating 781 nm beams and carried over a maximum distance of $$U(z,t)=U_0 \cos^2\!\left(\frac{2\pi \Delta\nu}{2}t-\frac{2\pi}{\lambda}z\right),$$2 toward the chip surface, with maximum transport efficiency about 50% (Xu et al., 2023).

3. Fiber- and chip-integrated implementations

Optical conveyor belts are especially valuable when the target region is embedded in a waveguide or photonic structure and cannot be directly overlapped with a MOT. In these systems the conveyor belt becomes an interface technology, coupling free-space preparation to guided-wave or near-surface trapping.

In a nanofiber-based implementation, cold cesium atoms are trapped in a two-color evanescent-field potential surrounding a nanofiber with a 500 nm diameter waist and a 5 mm long uniform section. The red-detuned field at $$U(z,t)=U_0 \cos^2\!\left(\frac{2\pi \Delta\nu}{2}t-\frac{2\pi}{\lambda}z\right),$$3 nm forms the axial standing wave, the blue-detuned field at $$U(z,t)=U_0 \cos^2\!\left(\frac{2\pi \Delta\nu}{2}t-\frac{2\pi}{\lambda}z\right),$$4 nm provides radial repulsion, and the local minima lie about 230 nm from the nanofiber surface with trap depth $$U(z,t)=U_0 \cos^2\!\left(\frac{2\pi \Delta\nu}{2}t-\frac{2\pi}{\lambda}z\right),$$5. The experiment reported transport over distances of up to about 0.3 mm, with the general capability to move atoms over millimeter scales along the nanofiber (Schneeweiss et al., 2012).

A hollow-core photonic crystal fiber interface uses two counter-propagating 805 nm beams to form a red-detuned standing-wave transport lattice for $$U(z,t)=U_0 \cos^2\!\left(\frac{2\pi \Delta\nu}{2}t-\frac{2\pi}{\lambda}z\right),$$6 atoms. The MOT is placed at a distance $$U(z,t)=U_0 \cos^2\!\left(\frac{2\pi \Delta\nu}{2}t-\frac{2\pi}{\lambda}z\right),$$7 from the fiber tip, and the transport potential can be shaped by varying both the frequency ramp and the trap depth. The reported transport efficiency into the fiber exceeds 40%, and adaptation of the transport potential lowers the temperature of the transported atoms by a factor of 6 while reducing the transport efficiency only by a factor of 2 (Langbecker et al., 2018).

Transport toward photonic chips introduces an additional near-surface problem: reflection from the chip can distort the standing-wave lattice. A GaN-on-sapphire experiment used a standard optical conveyor belt to bring a cloud from free space toward the chip, while a separate proposal addressed the final free-space-to-chip handoff with a two-layer photonic chip architecture. In that proposal, a Gaussian beam from free space interferes with a diffraction beam from an integrated apodized grating, and TM-polarized incidence at the Brewster angle $$U(z,t)=U_0 \cos^2\!\left(\frac{2\pi \Delta\nu}{2}t-\frac{2\pi}{\lambda}z\right),$$8 suppresses reflection from the chip surface. Numerical simulation showed that the antinode trajectories become continuous and that atoms can be brought to the chip surface with a distance of only 100 nm, allowing smooth delivery to the evanescent-field trap of waveguides (Xu et al., 2023, Liu et al., 2023).

Levitated-optomechanics experiments adapt the same moving-standing-wave idea to nanoparticles in a hollow-core photonic crystal fiber. Two equally polarized counterpropagating 1064 nm beams create a standing-wave trap, a Rayleigh nanoparticle is carried with velocity

$$U(z,t)=U_0 \cos^2\!\left(\frac{2\pi \Delta\nu}{2}t-\frac{2\pi}{\lambda}z\right),$$9

and the reported transport length is 15 cm, with step sizes about 12 $v=\frac{1}{2}\lambda \Delta\nu .$0m and sub-micron positioning about 0.5 $v=\frac{1}{2}\lambda \Delta\nu .$1m (Grass et al., 2016).

4. Deterministic single-atom loading and transfer

A particularly stringent use case is deterministic single-atom preparation in a static trap that is spatially separated from the atom source. In one experiment, an optical conveyor belt built from two counterpropagating Gaussian beams at 852 nm, focused through a shared objective lens of $v=\frac{1}{2}\lambda \Delta\nu .$2, formed a one-dimensional standing-wave trap lattice with a beam waist of about 10 $v=\frac{1}{2}\lambda \Delta\nu .$3m. A separate 852 nm optical tweezer with a waist of about 2 $v=\frac{1}{2}\lambda \Delta\nu .$4m was located 0.6 mm from the MOT. The conveyor belt transported atoms from the MOT region into the overlap region with the tweezer, and transfer into the static trap was achieved by adiabatically ramping the conveyor depth down to zero over 2 ms (Xu et al., 29 Jul 2025).

The decisive improvement came from real-time feedback control of atom number in the overlap region. Fluorescence on the 780 nm D2 cycling transition was measured in two 50 ms probe intervals: one before ramp-down, yielding $v=\frac{1}{2}\lambda \Delta\nu .$5, and one after ramp-down, yielding $v=\frac{1}{2}\lambda \Delta\nu .$6. An FPGA compared $v=\frac{1}{2}\lambda \Delta\nu .$7 with a threshold $v=\frac{1}{2}\lambda \Delta\nu .$8; if $v=\frac{1}{2}\lambda \Delta\nu .$9, a 10 ms frequency-sweeping pulse shifted the conveyor lattice by one step, and the procedure was repeated until $\delta$0 or a maximum feedback duration of 1 s was reached. The optimized operating point used $\delta$1 mK, $\delta$2 mK, $\delta$3, and a lattice step size of 3.75 $\delta$4m. Under these conditions the single-atom loading probability increased from 57.5% without feedback to

$\delta$5

that is, 77.6% with a 1.2% uncertainty (Xu et al., 29 Jul 2025).

This protocol also makes explicit the role of parity projection in a tightly confined static trap. The second fluorescence measurement leaves either one or zero atoms after light-assisted collisions eliminate multi-atom occupancy. The paper identified three main failure channels: about 8.6% of the time, no atom was initially loaded into the overlap region; about 4% single-atom loss occurred during ramp-down, mainly from non-adiabaticity and beam misalignment; and fluorescence fluctuations could prolong the feedback sequence until the atom was pushed out of the overlap region (Xu et al., 29 Jul 2025).

For cavity-based systems and hybrid atom-photon interfaces, this is significant because it addresses a practical loading bottleneck: a static trap can remain fixed in the optically constrained region of a Fabry–Perot cavity trap, nanophotonic chip trap, or related hybrid interface, while the conveyor belt supplies the missing transport degree of freedom.

5. Control theory, adiabaticity, and loss mechanisms

The principal control problem is to move the lattice rapidly without exceeding the restoring capability of a given well or driving parametric and anharmonic excitation. A recurring acceleration bound is

$\delta$6

which marks the point at which the accelerating-frame potential ceases to support a trap minimum. In an optical conveyor transporting atoms near a fiber Fabry–Perot cavity, a fitted effective depth $\delta$7K gave $\delta$8 and a minimum one-way sine-ramp transport time $\delta$9 (Hickman et al., 2020).

Waveform smoothness is a central experimental variable. In back-and-forth transport of $v=\frac{1}{2}\lambda \delta .$0Rb over $v=\frac{1}{2}\lambda \delta .$1, a smooth sine velocity profile allowed transport about 25% faster than a triangle profile for the same retention, and the triangle waveform could heat ground-state atoms by up to $v=\frac{1}{2}\lambda \delta .$2K. The same work showed resonances associated with parametric heating from digitally stepped DDS updates, with a practically safe choice

$v=\frac{1}{2}\lambda \delta .$3

and fitted trap frequency $v=\frac{1}{2}\lambda \delta .$4 (Hickman et al., 2020).

At much larger scale, the 37.2 cm transport apparatus compared linear, minimum acceleration, minimum jerk, and minimum snap trajectories. All trajectories degraded at high acceleration because the accelerating lattice tilts the axial potential, and all degraded at very low speeds because the RF update rate approached the axial trap frequency and caused parametric heating. The best performance occurred around $v=\frac{1}{2}\lambda \delta .$5–$v=\frac{1}{2}\lambda \delta .$6 maximum acceleration, and the minimum jerk trajectory showed the broadest and clearest optimal plateau, roughly $v=\frac{1}{2}\lambda \delta .$7–$v=\frac{1}{2}\lambda \delta .$8. For a fixed speed of $v=\frac{1}{2}\lambda \delta .$9, reducing the RF update rate from 333 kHz to 167 kHz dropped efficiency from 66% to 38% (Matthies et al., 2023).

Theoretical control work has made the same problem more precise for realistic three-dimensional lattices. A study of single-atom transport in optical conveyor belts treated the full anharmonic potential with shortcuts to adiabaticity (STA) and enhanced STA (eSTA), validated by Fourier split-operator propagation of the time-dependent Schrödinger equation. The central result was that eSTA is better than STA for roughly $$v = \frac{1}{2}\,\frac{\lambda_0^{\rm red}}{n_{\rm eff}}\,\Delta \nu_{\rm opt},$$0, while in the narrower window around $$v = \frac{1}{2}\,\frac{\lambda_0^{\rm red}}{n_{\rm eff}}\,\Delta \nu_{\rm opt},$$1 eSTA can underperform STA. This formalizes a point already visible in experiments: short-time transport is limited not only by nominal acceleration bounds but also by three-dimensional anharmonicity (Hauck et al., 2021).

Noise and lifetime impose an additional limit. In the nanofiber-based conveyor belt, the trap lifetime dropped from the previous stationary case $$v = \frac{1}{2}\,\frac{\lambda_0^{\rm red}}{n_{\rm eff}}\,\Delta \nu_{\rm opt},$$2 ms to $$v = \frac{1}{2}\,\frac{\lambda_0^{\rm red}}{n_{\rm eff}}\,\Delta \nu_{\rm opt},$$3 ms. The reported explanation was relative phase noise between the counter-propagating red-detuned fields, which jitters the standing wave and drives resonant heating. The estimated heating rate was about $$v = \frac{1}{2}\,\frac{\lambda_0^{\rm red}}{n_{\rm eff}}\,\Delta \nu_{\rm opt},$$4, corresponding to a phase-noise-limited lifetime $$v = \frac{1}{2}\,\frac{\lambda_0^{\rm red}}{n_{\rm eff}}\,\Delta \nu_{\rm opt},$$5 (Schneeweiss et al., 2012).

Trap-depth shaping can partially convert transport control into temperature control. In the hollow-core-fiber transport experiment, reducing the final depth lowered the transported-atom temperature by a factor of 6 while reducing the transport efficiency only by a factor of 2. This makes explicit a general tradeoff: deeper traps relax confinement and loading constraints, whereas shallower final traps suppress final temperature at the price of atom-number loss (Langbecker et al., 2018).

6. Broader uses and conceptual boundaries

Although the standard optical conveyor belt is a moving standing-wave dipole trap, the term has broadened to cover several distinct mechanisms. In a conveyor-belt magneto-optical trap of CaF, moving standing-wave light fields and a magnetic-field gradient pile molecules into a much smaller region than a conventional red-detuned MOT. The optimized cloud had mean radius $$v = \frac{1}{2}\,\frac{\lambda_0^{\rm red}}{n_{\rm eff}}\,\Delta \nu_{\rm opt},$$6m, peak number density $$v = \frac{1}{2}\,\frac{\lambda_0^{\rm red}}{n_{\rm eff}}\,\Delta \nu_{\rm opt},$$7 cm$$v = \frac{1}{2}\,\frac{\lambda_0^{\rm red}}{n_{\rm eff}}\,\Delta \nu_{\rm opt},$$8, and subsequent loading into a 1064 nm optical dipole trap yielded up to $$v = \frac{1}{2}\,\frac{\lambda_0^{\rm red}}{n_{\rm eff}}\,\Delta \nu_{\rm opt},$$9 trapped molecules at $2\pi$0K with phase-space density $2\pi$1 (Yu et al., 2024).

A later theoretical study of heavy molecules emphasized the same blue-detuned, dipole-force-dominated picture. In this CB-MOT language, two moving standing-wave polarization gradients carry molecules toward the field zero while blue-detuned Sisyphus cooling removes kinetic energy. The calculated capture velocity for $2\pi$2 at the experimental CB-MOT condition $2\pi$3 was $2\pi$4, compared with a typical red-MOT value of about $2\pi$5, while the same framework predicted broad nonzero capture-velocity regions for $2\pi$6 and $2\pi$7 (Yang et al., 20 Jun 2026).

Other uses are more strongly nonstandard. A single-beam optical conveyor belt for small chiral particles uses a circularly polarized plane wave, an opaque mirror, and a chiral metamaterial particle engineered so that one helicity produces positive radiation pressure and the opposite helicity produces negative radiation pressure, with no optical traps required. In the balanced case $2\pi$8, the force becomes position independent and reverses sign with helicity (Fernandes et al., 2016).

At the nanoscale, a nano-optical conveyor belt has been proposed using U-shaped Au–VO$2\pi$9–Au nanoholes. There the transport mechanism is a sequence of movable plasmonic hot spots, rather than a free-space or guided standing-wave lattice, and the direction of transport is reversed by the VO$^{133}$0 phase transition (Ouyang et al., 2023). In exciton-polariton systems, the conveyor belt becomes a traveling optical potential

$^{133}$1

which drags a polariton wavepacket with drift speed tied to the drive through $^{133}$2 and produces approximately linear dispersion for the transported packet (Xu et al., 8 Oct 2025). In structured-light work on bipolar coordinates, conveyor-belt modes are optical modes whose intensity follows an extended trajectory connecting two charge centers (Strohaber, 18 Jun 2026).

This suggests that “optical conveyor belt” now functions as a family-resemblance term rather than a single device class. In one branch it denotes a moving standing-wave dipole lattice for atoms or nanoparticles; in another it describes moving standing-wave polarization gradients in a blue-detuned MOT; in still others it refers to helicity-controlled push–pull transport, plasmonic hot-spot transport, traveling potentials for quasiparticles, or structured-light modes whose intensity traces a transport-like path. The unifying idea is directed conveyance by an optical field whose spatial structure moves, switches, or biases the trapped object’s motion.