One of the questions one asks when trapping atoms in a magneto-optic trap (MOT) is “What shall we do with the atoms?” You often have an idea before you do the trapping — it’s not like we’re trophy hunters, trapping just to have something on the wall. Trapping in and of itself hasn’t been the goal for quite some time now, at least in experimental labs; one wants to do some kind of experiment with the atoms. Some of the time that can be done in the trap, but quite often it involves moving the atoms somewhere else. Sometimes you actually wanted an atomic beam of some sort, instead of a collection of atoms just sitting there, suspended in space — the trapping environment involves bright, near-resonant laser light and magnetic fields and those could be undesirable. The atom beam gets you away from this, and if you look at the beam from a perpendicular direction, the Doppler shift is very small. Perhaps you want low-speed collisions, and tuning the speed of the beam allows you to do your experiment. There are also a number of atom-optics experiments that can be done, e.g. sending the atoms through transmission gratings comprising an interferometer. The problem could also be the relatively high vapor pressure of the gas in your vapor cell giving you excessive background signals, or collisions with that background vapor could be the problem, limiting the trap lifetime. So you need to move the atoms, transporting them to a region that is better-suited for the experiment you are doing.

When I was at TRIUMF, the problem was the background and trap lifetime. We were trapping radioactive atoms, and the idea was that when an atom decayed, the beta would go one way and the atom would recoil, and each could be detected. But a vapor-cell MOT captures only the small percentage of atoms stupid enough moving slowly enough to get trapped, leaving the majority of the zipping around in the cell or sticking to the walls (or worse, attaching themselves to detectors). Not only did this mean they would be swamping the signal from the trapped atoms, the signals would be coming from different directions and originating from different points.

About the time we started fretting about this problem (you have to trap them first before you worry about the next step, and nobody had trapped these isotopes before) we got a visit from Zheng-Tian Lu, then at JILA/NIST, and he had come up with an ingenious method of generating a low-velocity atomic beam and shared the details with us (the paper was in the pipeline but had not yet been published at the time)

A typical vapor-cell MOT uses three beams along the cartesian axes, and it’s possible to do this by retroreflecting each of these beams — the vapor is dilute, so with decent mirrors there isn’t a large drop in intensity (any imbalances will push the trap slightly off-center as the effect of the magnetic field compensates). You get the proper polarization of the beams by placing a quarter-wave plate in front of the retroreflection mirror (this changes the circular light to linear and then back to circular of opposite helicity; if you started with linear it would circularize it and change it back to linear, perpendicular to the original. Ah the fun you can have with waveplates)

To extract a beam of atoms, one needs an unbalanced force to eject them from the trapping potential. Lu did this by drilling a hole in one of the wavepalte/mirror combinations (which was a waveplate with a layer of gold deposited on it), leaving a thin virtual “pipe” where the force from the trapping laser was unbalanced since there was no retroreflection. Atoms would be pushed along in that direction, but if they left the pipe they would again be subject to trapping forces. This meant that atoms with any significant transverse velocity would not escape the trap, and the beam that managed to escape would not tend to expand very quickly.

A few effects would tend to limit how fast the atoms would be pushed. As they sped up, they would move further from resonance due to the Doppler effect, which would lower the scattering rate and thus the force they felt. The magnetic field gradient present in the trap would also change the relative detuning of the transition owing to the Zeeman shift. The combination of these two effects would greatly diminish the scattering rate of the photons, dramatically weakening the push they felt, once they had achieved a certain velocity. And once the atoms had left the trap, the pushing laser could optically pump them into a dark state (inside the trap there would be a repump laser to put them back in a state where they could interact with the laser). Once in this dark state, they would not interact with the laser at all. You can see that the pipe can effectively be plugged with an extra beam that pushes the atoms out before they are able to escape.

Properly choosing the original detuning of the trap laser would mean that the atoms’ low velocities would be similar to the capture velocity of a MOT, so this concept was ideal for transferring atoms into a second trap. It was also fairly efficient, so the beam had a lot of atoms in it. It was dubbed the Low-Velocity Intense Source, or LVIS (pronounced “Elvis”)

This was great news for us, except for one teensy, tiny problem. Our trap geometry, which was partly dictated by being hooked up to a cyclotron, target and isotope separator, meant that we needed a large hole where one of the mirrors was, in order to allow atoms to enter the cell, and the extraction of atoms would have to be perpendicular to the entrance hole (otherwise we would run into huge pieces of vacuum hardware) In other words, everything was rotated 45º to the LVIS geometry, and we couldn’t just copy and implement it. We needed to recreate the functionality and adapt it to our system We needed an LVIS impersonator. (did you see that one coming?)

What we did was add a push beam that was tightly focused; the goal was to have the narrowest, and therefore most intense part of the beam near the MOT, and have it slowly expand as it moved away, so while the pushing was not unopposed, it was much stronger than the trapping force. This was aided by turning down the trapping beam intensity while pushing, which would also have the effect of minimizing lateral forces present from mismatched beam intensities at the edge of the trap (which is a much smaller issue for just trapping the atoms). The atoms we were transferring were Potassium, and for several of these isotopes, the hyperfine splitting is very small; in most alkali atoms you can almost isolate one transition such that other possible transitions are far from resonance so that optical pumping is weak, but this wasn’t possible for us. But the strong optical pumping meant that essentially as soon as the atom left the trapping volume, the atoms were pumped to a dark state and stopped interacting with the laser. The velocity spread was still several m/s, though, because the transition has a ~6 MHz linewidth, and the absorption probability depends on speed through the Doppler shift. 6 MHz translates into almost 5 m/s for Potassium’s wavelength of 766 nm, i.e. an atom could be moving faster or slower by several m/s before it starts “noticing” the difference in the photon’s Doppler-shifted frequency.

Still, the system worked, and atoms could be trapped in a second MOT. We added some transverse confinement along the way ( a 2-D MOT) to recompress the beam, and were able to get our atoms away from the background vapor, and into a trap whose lifetime was significantly longer than the ~1-second half-life of the atoms, so that when they decayed, they were almost always in the trap, which minimized noise in the data.

Low-velocity intense source of atoms from a magneto-optical trap. Z.-T. Lu et al. Phys. Rev. Lett., 77, 3331 (1996)
Efficient transfer in a double magneto-optical trap system. T. B. Swanson et al. JOSA B, 15, 2641 (1998)