What Do You Need to Make Cold Atoms? Part 2: Lasers and Optics

Following on yesterday’s discussion of the vacuum hardware needed for cooling atoms, let’s talk about the other main component of the apparatus: the optical system. The primary technique used for making cold atoms is laser cooling, and I’m sure it will come as no surprise that this requires lasers, and where there are lasers, there must also be optics.

There are lots of different types of lasers used for laser cooling experiments, but they all need to have certain properties: tunability, stability, and adequate power. Tunability is important because laser cooling requires light at exactly the right frequency to be absorbed by the atoms you’re trying to cool; stability is important because you need the laser to stay at the right frequency once you get it there; and power is important because you need a decent amount of light, and would prefer not to have to use multiple independent laser systems.

The particular lasers I use are diode lasers, like the ones found in CD and DVD players, or laser pointers:

(The cylindrical black thing on the foil is a collimation tube, which holds a lens in front of the laser to bend the widely diverging output of the diode laser into a more or less constant width beam. the battery is for scale.)

Diode lasers are mass produced for optical data storage purposes, so they’re relatively cheap– the ones shown here are something like $100-200– and they produce a decent amount of power, around 150 mW per laser. They’re tunable over a wide range, as such things go– the lasers I buy ordinarily put out light with a wavelength of about 808 nm, but I run them at 811 nm by controlling the temperature and current.

The disadvantage of diode lasers is that they can be a little unstable, and also fragile. Static electricity can kill them completely (the ones in the picture are all dead), so you need to be careful to ground yourself to something else before touching the laser system, and they’re prone to “mode-hopping,” moving to a different wavelength a small fraction of a nanometer away, out of the range where laser cooling will work.

You can improve the stability and provide some fine tuning by putting the laser in an “external cavity,” typically involving a diffraction grating that selects light of a particular wavelength and sends it back into the laser cavity. The external cavity laser I use for my main light source is shown at right. The laser itself is inside a collimation tube that is mounted inside the aluminum block– you can see the power leads going into the block at the top of the picture. The black anodized alumninum mount bolted to the near side of the block holds a diffraction grating mounted on a piezoelectric stack (the little green thing that’s just visible in the center), a material that expands or contracts when you apply a voltage across it. The mount lets us tune the wavelength that is sent back into the laser in a very coarse way (you can see an allen wrench coming out the back– the smallest turn you can reproducibly make with the wrench shifts the wavelength by around 0.1 nm), and we can make very fine adjustments by controlling the voltage sent to the piezo. The light from the laser exits the aluminum block heading up and to the left in the picture, and passes out through a hole cut in the outer aluminum block (which is covered by a microscope slide).

This whole assembly sits in an aluminum box, as you can see, that it usually kept closed when things are working well. The wavelength of the laser is highly sensitive to temperature (there are thermoelectric coolers sandwiched between the aluminum block and the copper baseplate to control the temperature of the block to with a small fraction of a degree), and closing it in a box helps avoid air currents that might upset the temperature. It also reduces sensitivity to sound that might cause the laser wavelength to jump slightly due to vibrations of the grating.

The biggest hassle with these, as you might imagine, is getting the grating aligned just right to send light back into the laser. The laser itself is something like 100 microns in size– about the thickness of a human hair– so this is a really fiddly process. It’s sensitive enough to the alignment that after making a big change in the tilt of the grating, the tiny change in angle caused by relaxation of the metal “spring” holding the mount in place will make the laser drift off the desired wavelength over a period of days or weeks. The second-biggest hassle is getting the temperature tuning just right, because anything involving thermal processes is sloooow, and something that looks right a minute or two after making a change can settle to a new equilibrium an hour later that doesn’t work at all. Once you get everything working, you tend to leave these on 24-7, because they’re more stable that way.

There are two other major technologies used for laser-cooling lasers, those being solid-state lasers like the Ti:Sapph laser I used in grad school, which uses atoms in a crystal as the laser medium, and dye lasers which use organic molecules in a jet of liquid as the laser medium. Both of these offer wide tunability– tens to hundreds of nm– using the fact that molecules and solids have broad emission and absorption bands rather than narrow lines. They require external pump lasers, though– typically either an argon ion laser at 514 nm or a diode-pumped doubled YAG laser at 532 nm–
and are extremely expensive. Dye lasers are also messy, and often use toxic chemicals, which are great reasons to avoid the damn things.

Once you have a laser at the right wavelength, you need optics. Lots and lots of optics. This is a slightly outdated picture showing a part of the optical system for my lab, with the beam lines traced out:

A more complete explanation of what the different colors mean can be found in this old post explaining the importance of cosmetics to laser cooling. The master laser shown above is in the grey box at the top of the picture. The cardboard box toward the bottom contains a “slave” laser whose frequency is locked to that of the master laser by feeding a small amount of light into it (the orange beam line). This is only about half of the optics layout I use (the rest is on a different table), and this is a relatively simple system by current standards.

The main optical elements involved are:

• Mirrors: most of the elements you see in these pictures are mirrors used to direct the beam to where it needs to be. These are mounted in sturdy aluminum mounts with two tilt screws to control the horizontal and vertical tilt of the mirror. A good and stable mount will let you adjust the position of a spot a few meters away by a millimeter or so. I have dozens and dozens of these, and am always scrounging more.
• Lenses: sometimes you want a beam to be a little bigger than it is (to provide a large trapping volume, say), and sometimes a little smaller (to get it through some narrow opening). We use a whole bunch of different lenses for this.
• Optical Fibers: You can direct light all over the place with mirrors, but the problem with this is that a small shift in the tilt of one mirror early in the chain can produce an enormous change a few meters downstream. This problem can be reduced by coupling the light into optical fibers, whose ends will always be fixed. Of course, getting light through the fiber is a hassle, and you generally lose half of it, so this is to be used sparingly. For some things, though, like the alignment of the injection lock of the slave laser, it’s absolutely critical.
• Waveplates: The polarization of the light is critical for laser cooling, and we use devices called waveplates to control that polarization. These come in two types: “quarter-wave,” which turn linear polarization into circular polarization and vice versa, and “half-wave,” which rotate linear polarization from horizontal to vertical, or any angle in between.
• Beamsplitters: These are partially reflecting elements that split a beam into two parts, typically mounted in glass cubes. These also come in two types, polarizing and non-polarizing. A non-polarizing beamsplitter gives you a fixed ratio of intensities (typically close to 50/50) between the output beams, while a polarizing beamsplitter sends horizontally polarized light in one direction, and vertically polarized light in another. A half-wave plate followed by a polarizing beamsplitter gives you a variable intensity ratio between the output beams, which is extremely useful. If you only want a small amout of light in one beam, you can use a glass microscope slide as a cheap beamsplitter, and this is done in several places for beam diagnostics.
• Optical Isolators: This is a crystal of special material mounted inside a strong magnet, sandwiched between two polarizers. This combination lets light pass in only one direction, and is important to protect the diode lasers from reflected light that might screw up the wavelength. These are pricey, but essential– they’re the two clunky gold things visible in the picture with the beam lines, one for each laser.
• Acousto-Optic Modulators: An AOM is a block of glass with a small speaker mounted to it. The speaker vibrates at very high frequency– typically 80 MHz– setting up a sound wave in the glass. This sound wave acts like a diffraction grating, and shifts both the direction and the frequency of the diffracted light. These are used for fine control of the laser frequency, and fast switching of the frequency for various experiments. We’ve got two in my lab; other labs with bigger budgets use half a dozen or more.

All of this stuff is expensive– I’ve probably dropped over $50,000 on miscellaneous optical elements over the years, and I could always use more.

All of these elements are arranged in very particular ways to get the beams where they need to go. Of course, this is incredibly sensitive to small changes in the position of the optical elements, which is why everything is bolted to the surface of an optical table. These tables are large, heavy tables designed to minimize vibrations, and with steel tops having 1/4″-20 holes drilled in a 1″ grid over the whole surface. The mounts holding the mirrors and lenses and so forth are bolted to the table, so they’re fixed in position unless you do something really stupid to knock them out of place.

(One of the standard horror stories in this business involves cleaning or maintenance workers who come into research labs to do something, and climb on top of the optical tables, kicking mounts in the process. This is why our labs at NIST had big signs saying “DO NOT CLEAN” on the door.)

So that’s the quick guide to the toys used in getting light to where it’s needed for using lasers to cool atoms. Questions, comments, things I left out?