Atom interferometer gravimeter

Three laser pulses split, mirror and recombine a falling rubidium cloud; the fringe phase is k g T^2.

Switch to 2D

Presets

Playback

Interferometer

Run

Camera

Layers

Quality

Model check

The science

What is modelled

A cold-atom gravimeter turns free fall into a matter-wave interferometer. A cloud of rubidium-87 atoms is cooled in a magneto-optical trap, released, and hit by three laser pulses separated by a free-fall time T. The first pi/2 pulse puts each atom into a superposition of two momentum states that differ by the two-photon recoil, the pi pulse swaps them so the arms converge, and the last pi/2 pulse recombines them. The fraction of atoms found in the excited state is read out by fluorescence. The number of fringes is huge, which is why the instrument can resolve parts in 109 of g.

Equations

  • Counter-propagating Raman beams: keff = 4π/λ (twice the single-photon wave number). The slider is the laser wavelength.
  • Two-photon recoil velocity: vr = ħkeff/m, about 11.8 mm/s for Rb-87 at 780 nm. The arms separate as s(t) = vrt for t ≤ T, then close again, so the largest separation is vrT (about 1.2 mm at T = 100 ms).
  • Both arms fall classically, z(t) = z0 - ½gt2. Each laser pulse writes keffz(ti) onto the atoms, and the interferometer phase is Δφ = keff(z(0) - 2z(T) + z(2T)) = keff g T2.
  • Readout probability: P(e) = ½[1 - cos Δφ]. One fringe corresponds to a change of g of Δg = 2π/(keffT2) = λ/(2T2).
  • With N atoms the shot-noise-limited phase uncertainty is 1/√N, so δg = 1/(√N keff T2). 1 µGal = 10-8 m/s2.

Checks run in the page

  • An RK4 integration of the falling cloud gives the laser phases, and the combination k(z0 - 2z(T) + z(2T)) is compared with k g T2.
  • 40,000 simulated single-shot readouts against P(e), judged in binomial standard deviations.
  • The fringe spacing found by bisecting P(g) against lambda/(2T2).
  • The recoil velocity computed from hbar k_eff/m and independently from 2h/(m lambda), compared with the textbook 11.8 mm/s.

Honest simplifications

  • The arms are drawn side by side and exaggerated about 150 times so they can be seen; in a real instrument they differ by about a millimetre along the beam. Untick "Exaggerate arm separation" for the true scale.
  • The model is the ideal three-pulse sequence: no laser phase noise, vibrations, Coriolis or gravity-gradient terms, finite pulse duration, or cloud temperature. These dominate the error budget of real devices.
  • At 105 to 106 radians the fringe is read out only modulo 2 pi; real gravimeters resolve the fringe order by scanning T or chirping the laser frequency.
  • The atom cloud shown is a few hundred markers; the atom number slider only changes the shot-noise estimate.

What to try

  • Press "Scan g across fringes" and watch single-shot dots fill in the cosine fringe.
  • Halve T: the fringe spacing grows four times and the resolution gets four times worse.
  • Switch the laser to 1500 nm: keff halves, the recoil and fringe spacing change accordingly.
Loading model...
Loading Blender model...
Fringe scan: single shots vs P(e)
Shot-noise resolution vs T
Arm offset in time (worldlines)
Made with 🟠 Blender · Three.js