Imagine, as I sometime ask, that you are doing an experiment which is very sensitive to external magnetic fields. (Like, oh, I don’t know, an atomic clock). And you find evidence of some stray field gremlin that has taken up residence. Since every previous time this has happened it has been the result of a thermoelectric current, you might be lulled into thinking that it’s the same thing, and fixing it will be a piece of cake. A colleague might even announce something to that effect: “It’s a current. It’s always a current. If there’s one thing you can depend on, it’s that stray fields are always currents.”

Welcome to the phenomenon of the sportscaster’s curse.

The sportscaster’s curse, in case you’re not familiar with it, is a phenomenon seen during sportscasts, in which the announcer will basically guarantee an outcome, which then dooms the effort to failure. The athlete is tagged as “Mister Automatic” in some way, with a mention of how he hasn’t missed a free throw/short putt/chip-shot field goal in X attempts, at which point the attempt clangs off the upright or rim, or lips out of the cup. (I’m sure a fair bit of confirmation bias is present here, since the curse doesn’t strike every time, but I cringe nonetheless if it’s a player on my team being lauded for his reliability)

So this is what happened. After we convinced ourselves that it was a simple problem and a quick fix, as happened earlier, we took all the steps to fix it. These steps include taking off the nested layers of magnetic shielding which make the fountain look like a Russian-doll hot-water heater. (or really just a water heater, because you don’t need to heat the water if it’s already hot)

(Such a device might look something like this)

Nada. No obvious connections. We let it cool down to room temperature to minimize gradients and reassembled, but there was no change in the signal. OK, disassemble again and start checking for some magnetic component. But we’re looking for a milligauss-ish field, which isn’t going to be seen amidst the half a gauss of the earth’s field, so the only real way to do this systematically is to change one thing and reassemble it so we can look at the signal in a shielded environment.


We did that a lot over the past few days.


We finally decided that looking with a magnet might be a good idea — a strong one might stick to the offending component. There shouldn’t be any downside — the nonmagnetic materials aren’t going to become magnetized, and if there is a magnetic part, it will only change the scale of the already-existing problem. The latter is exactly what happened. The magnet didn’t stick to anything, but all of the sudden (after yet another reassembling and degaussing of the shields) the problem was much bigger — we had induced more magnetization, and that made it easier to find the offending component. Kinda like finding a needle in a haystack by being able to make the needle a lot bigger.

It was the salmon mousse a washer on a bolt in the vacuum system. Somehow a shiny stainless steel washer had successfully been hiding among the copper ones, and nobody noticed; it either had acquired a similar-looking tarnish, or because of the shininess it looked coppery when it was in the bin. In any event, transplant surgery was indicated and carried out successfully without a vacuum breach (which is good because losing vacuum would have sucked in all the wrong ways)

While I'm In Fist-Shaking Mode …

as I was yesterday, it’s time to stand on the porch and curse the Seebeck effect, which is the phenomenon behind thermoelectric currents. If you take two different conductors and make them touch each other, you get cooties a current if there is a temperature gradient present in the system. You can run this in the other direction and create a temperature difference, which is why thermoelectric coolers/heaters work; this is the Peltier effect and is the technology in most portable active coolers (if it plugs in and doesn’t have a compressor, then it’s probably a Peltier cooler).

Now, imagine you have a vacuum system running an experiment which is very sensitive to magnetic fields. Because you are trapping atoms, you have MOT coils dissipating a few Watts of power (while the trap is on) and you are heating your alkali metal reservoir to facilitate their introduction into the vacuum system. You will have a temperature gradient along your nicely conductive vacuum system. So if you have some dissimilar metals touching the vacuum system, somewhere, you’ll get a tiny amount of current flowing through it, creating a tiny magnetic field. But since such magnetic fields are phenomena non grata, having a thermoelectic connection — which you did not expect to have — is a bit of a pain.

Found it though. Something was inappropriately touching the vacuum system, so we had it arrested and now we make sure there’s a few mils of kapton between it and the vacuum chamber.

Vintage Lab Pics: The Laser System

I already showed the vacuum system from my grad school days. This is the laser system that drove it. Slowing and trapping a thermal beam of atoms and then creating a new cold beam requires several lasers at different frequencies.

This first picture is a diode laser system, obviously home-built; this pre-dates any sort of commercially available system by several years, and perhaps a decade. On the far left are the electrical connections and the on/off switch. Power is needed for the laser, a piezoelectric transducer stack and the thermoelectic cooler, and a signal from a thermistor is fed back for coarse temperature tuning/stabilization of the wavelength. The diode is mounted in the thin rectangular block and has a collimating lens mounted in the thicker one; as I recall the lens position was adjusted with an external jig and then glued down.


Light leaves the diode and hits the grating, reflecting off to the bottom, but the grating is blazed — the lines are angled, and in this position one of the diffraction orders is reflected back into the laser, which forces the laser to operate at that frequency. Thus, by changing the angle slightly, the wavelength can be tuned over some small range, perhaps a few nanometers. The grating is mounted on a small kinematic mount, and this obscures a gap between two parts of the mount. At the lower right you can see the gap where the piezo stack is, and at the upper left, near the screw, is the pivot point.

The entire block is mounted on a thermoelectric cooler to stabilize the temperature. Laser diodes are coarsely temperature-tunable, so the temperature is chosen to get you close to the desired wavelength (780.24 nm for Rb-85). When operating, this would be covered with a plexiglass housing to act as a thermal barrier and a baffle for air currents, and on later designs contained a tray for a desiccant to help prevent condensation on the cold laser.

Here is the table, with a couple of lasers in the foreground. Light goes out of the side of the boxes and hits the turning mirrors; some of the light is picked off and sent into the spectroscopy cells visible just past the lasers (the one on the left is closer) This ensured the lasers were on resonance.


This is the whole laser table, showing the vacuum system on the right. Much of the equipment is on shelves above the optical table, and this design ensured the attraction of shorter personnel to the lab, as might be predicted by Murphy’s law.


The blue boxes on the center-line of the table are fast photodiodes; we used a beat between lasers to generate a locking signal for those not locked to the spectroscopy signals, and could tune the other lasers several MHz away using this technique. This allowed us to tune the trapping laser beams such that the trap was not stationary in the lab frame. The atoms would feel a force to eject them from the atomic funnel, and the lasers would become equal due to the Doppler shift, once they atoms were moving at the right speed.

Vintage Lab Pictures

I was decreasing the local entropy in a small part of my abode and found a shoebox full of photos which happened to contain a few shots of my grad school lab, in all its glory. We were building an interferometer which would use cold atoms, which means relatively large deBroglie wavelengths and a correspondingly small system. But one has to trap the atoms and cool them down first, and then generate the cold beam of atoms to feed into the diffraction gratings that comprise the interferometer, so the system is still quite complex. Since I was the first PhD candidate in the group, it meant I was involved in the construction of most of the components of the lab apparatus pictured here. It also meant a lot of fumbling in the dark, both figuratively and literally, since the only one with any experience with doing this kind of work was the PI, who had other duties (like teaching, writing grant proposals, etc.) It was a big day when the group finally reached the point where we didn’t want him playing with the experiment, because we knew more about the details than he did. That took a couple of years.

Here’s the vacuum system.


At the bottom of the picture is the oven, which was basically a pipe bomb with a hole in it (under vacuum, of course), and when heated sufficiently, would spray Rubidium atoms out. These were collimated with a second hole a few cm away (the extra Rb was collected on some cold metal and sent back into the oven during a refill cycle we ran each night after running the oven). Underneath is a turbo pump and a roughing pump; the small green hose was for adding dry nitrogen if we needed to open the system up. The nitrogen was supposed to keep the Rb from reacting with water in the air. It never worked.

The long pipe leads down to the next part of the vacuum system, and it has a “wobble stick” that blocks the atomic beam, if desired, and a valve to isolate the two parts of the system. You can also see an ion gauge on the left which was normally off so the light from the filament wouldn’t register on any photodiodes or the photomultiplier tube.

The tube is about a meter or so long, and needed to be so we could slow the atoms down. A laser was sent down the tube from the other end, and “chirped” in frequency to account for the changing Doppler shift — you wanted the laser to track the resonance of the atoms as they slowed down, so they would keep scattering photons. Once the atoms were slow enough, the laser turned off, and the atoms drifted into the bigger chamber, where they were trapped in a two-dimensional trap called an “atomic funnel.” They were able to move along the axis of the funnel, and were forced out to the left into the region where the interferometer would be. I made the gratings for the interferometer, but never got to the point where the apparatus was finished.

Here is an early version of the funnel:


It’s a “hairpin” made of copper tubing so we could water-cool it while sending several amps through it, and it creates a quadrupole field in two dimensions, so that atoms would be trapped into a pencil shape along the central axis, left-to-right. It’s supported from below by two insulated standoffs — you can see the ceramic, i.e. insulating, part — to keep it straight. The dark hole straight through it is the tube leading to the oven and the lighter window on the left would be where one of the six trapping lasers would come in. Up top there is a tube with some lenses in it for imaging the trap, and on the right is a target for aligning the lasers. The target is on a linear feedthrough, and the target could be inserted into the center and all of the lasers sent through it to make sure they all overlapped where they should.

The funnel tubes were originally soldered together. What you see here is version 2, because at one point we lost cooling water while the system was energized, and the solder melted. Which meant that lots of water was introduced to the system and it became a giant fish tank, sans fish. Fortunately, we had interlocks in the system so that if the pressure rose too high the pneumatic valves would shut (that’s what the rest of the green tubing is for in the first picture), and the damage was limited. The white blobs at the ends of the little elbows are torr-seal, which we used to repair the trap. This eventually failed, too, and we replaced it with a much better trap fashioned from a single piece of copper tubing.


I graduated after having built much of this and characterizing the atomic beam. We were able to extract atoms going up to 10 m/s — this was adjustable, depending on the laser frequencies of the different beams, but atoms going too slow would miss the interferometer because of gravity, so we didn’t bother trying to generate a beam going below a couple of m/s. The atoms were also somewhat cold — less than a milliKelvin — so that the beam didn’t spread out too much. At the time, one other funnel had been built, but it had on of its lasers along the axis of the funnel, which precluded putting any sort of target there. This was novel enough for a publication (T. B. Swanson, N. J. Silva, S. K. Mayer, J. J. Maki, and D. H. McIntyre, “Rubidium atomic funnel,” J. Opt. Soc. Am. B 13, 1833-1836 (1996)) and more importantly, a degree.

Two For One

One of the things we’re investigating is pulsed laser systems, because they’re fun, but (especially for funding purposes) also because they are the basis of optical frequency combs (as I’ve mentioned). And things are pulsing along. One of the things that was noticed was that light from the pulsed system, running at 1560 nm, was showing up on a Silicon CCD camera. The Silicon response peaks at 900 nm and drops pretty sharply, petering out at 1100-1200 nm. There’s no way it should respond to a 1560 nm photon.

And it isn’t. It’s responding to pairs of 1560 nm photons. This is a pulsed system, so you have high peak power making it a lot easier to see nonlinear responses like two-photon transitions, because they scale as the square of the intensity. (more photons incident per unit time means a better chance to have two interacting at once, Having n photons means that if you look at any photon, the chance of another photon being around is n(n-1)) Two photons have enough energy for the interaction, since that’s the same as having a 780 nm photon, which is well above the “to be detected you must be this tall” energy cutoff

Here are two images. The square is a beamsplitter cube, and the white blob is the light. The top image is the pulsed laser, and the bottom one is a CW beam, both with around 10 mW average power.



The pulsed laser is saturating the heck out of the CCD, so the spot is really a lot brighter than from the CW beam, though we can’t say for sure based on this quick look. Even though the average power is about the same, though, the pulsed laser is repeating at about 10 MHz, and the pulses are less than a picosecond, so all of the light is being delivered in less than 10-5 of the time, so the peaks have powers measure in kW.


The fastest clock in the world, my ass.

You need to a flashplayer enabled browser to view this YouTube video

Oooooh. It displays six whole digits past the decimal. Down to the microsecond. (can you sense the sarcasm?) It’s a display. Just because it reads that many digits doesn’t mean the measurement actually has that precision.

I’ve wanted to get a display that went to the picosecond for the lab, but have it flash 12:00:00.000000000000 the whole time. Add it to the list of my unadopted suggestions.

“I see no progress in this industry. These clocks are no faster than the ones they made a hundred years ago.” — Henry Ford

Screw Archimedes

Who needs a lever, man? I’m Marty McFly.

Let’s say, for the sake of argument, that you needed to move a heavy piece of equipment you’re helping to build. And you have the constraints of not wanting to tip it, and the course you need to navigate has some spots with less than 10 cm of overhead space. (This ignores one doorway that’s actually 10 cm too low — yes, we’ve done the equivalent of having built a boat in our basement. The solution there will either be slapping Daffy Duck with a frying pan in the shape of the equipment and having him run through the door, or a sawzall.) You need a smooth ride, because it’s an expensive, somewhat delicate trinket representing a several dollars and a few person-years of labor.

I tried designing a cart, but couldn’t meet all of the constraints — anything low enough would tend to bottom out on the incline (oh, that’s right, the path isn’t level the whole way. Is that a problem? I need this soon.) One day I was stressing and kvetching about it in front of the right person, who suggested air bearings/air casters. The heretofore unconnected link between the physics and the application clicked, and I knew that was the answer. Float the sucker on air. A small industrial blower and lots of small holes.

Here’s a demo of the system with a dummy load. There are also some lead blocks there, too. (cue rimshot). 160 kg for the optical table, lead and support structure, and another 110 kg or so for me.

You need to a flashplayer enabled browser to view this YouTube video

(if you want a soundtrack, turn on a vacuum cleaner. Any “Ishmael” wisecracks about my pasty-white legs will be subject to retaliatory editing)
Continue reading

And Now For Something Slightly Different

The previous MOT video showed the atoms squirting out of the side when the trapping field was turned off. In this video things are more balanced, and you can see the atoms remaining in the beam overlap region, and fluorescing quite brightly. The trap is cycled on and off and you can see the trap “grab” the atoms and pull them back to the center; when the trapping field is left off it takes several seconds for the cloud to dim as atoms diffuse out, and that’s a qualitative sign that the atoms in the molasses are pretty cold. Probably tens of microKelvin.

You need to a flashplayer enabled browser to view this YouTube video

“Mr. Hands” is pointing out the trap axis at the beginning, as a cue to the person adjusting the trimming magnetic field. This kind of adjustment can be very laborious, as there are several parameters which need to be optimized, and they aren’t independent of each other. Beam alignment, magnetic field and beam intensity all need to be optimized, but all exert forces which can be offset by one of the the parameters, e.g. a slight imbalance in intensity can be offset by a small magnetic field, and the small amount of swirling of the atoms when the trap is turned off is likely an indication that this is the case.

However, at this level of adjustment, the atoms are the best indicators. An optical power meter or a magnetic field probe aren’t going to yield the precision necessary — they can only get you close. At this point you just have to wander around phase space, checking that you aren’t merely at a locally optimum signal. The true test comes when you can actually measure the temperature of the atoms, by imaging them in time-of-flight and seeing how much the cloud has expanded.

Little Infrared Riding Hood

My, what bright, glowing optical fibers you have.

One of my online compatriots recently explained a quick and easy way to do some IR photography. I felt compelled to try, and it was pretty easy. Cheap webcams are the most direct way to do this for a few reasons:

— they’re cheap. If you mess it up, you’re only out a few simoleans.

— they have manual focus. Modifying an autofocus camera requires you replace the IR filter with a glass plate, because removing it changes the optical path length. It’s a much trickier operation.

— it’s usually a fast modification

Just remove the lens — some of them simply unscrew — and check to see if the filter is mounted on the back. (If not, you’ll have to take the assembly apart. No biggie, though, it’s likely just one or two screws. You’ll need a jeweler’s screwdriver, probably phillips-head). Pop the filter off with a small screwdriver or equivalent; the filter may not survive in one piece, so don’t go into this expecting it to survive. Reassemble. You’re done. If the filter isn’t there, it’ll be covering the CCD/CMOS chip, but my extensive data (three points) says that it’s mounted on the back of the lens.

Plug it in to your computer and start taking pictures.

Expectations: This isn’t thermal imaging, so don’t expect bodies to show up glowing. Silicon, the element of choice, has a pretty sharp cutoff starting at about 950 nm, so what you’ll see in the near-IR. Something would have to be about 3000 K to be peaked at that wavelength and thermal images of body temperature targets peak between 9 and 10 microns. Also, the images will be small, since cheap webcams generally run only about a megapixel.

I just happen to have access to several infrared lasers (852 nm and 780 nm, the images use the latter), to give extreme examples of what you can see. This first picture is a laser table with the room lights off. You can see scattered light from several optical components, as well as light emanating from two optical fibers — not all of the light gets coupled into the fibers, and you’re seeing some of what leaks out (some probably in the wrong mode, since these are single-mode fibers, and the bending probably contributes)

IR laser table photo

In this second photo, there are two images of the same scene, taken with the room lights on. On the left, some shutters are shut, and on the right they are open, and you can see two fibers lit up. Also note the cylinder to the left — that’s a vapor cell with rubidium gas in it, set up for spectroscopy for servo-locking the laser. The laser is on resonance, so you can see the fluorescence as the beam passes through it.

As you can see, there’s quite a lot of scattered light, so normally this is encased in opaque plexiglass. None of the bright features shown are visible with the naked eye.