“Quantum Mechanics Is Magic”: The Making of “Spin polarization and quantum statistical effects in ultracold ionizing collisions”

This was the last of the experiments that I did for my thesis (it’s not the last xenon paper I’m an author on, but the work for that one was done while I was writing up), so my memories of it are bound up with the thesis-writing process.

My favorite story about this stuff was when I gave a talk about this work at NIST– I don’t recall if it was before or after my defense– and somebody asked the obvious question about how the quantum statistical rules are enforced. That is, how is it that you never get two identical fermions colliding in an s-wave state? Since an s-wave collision is just a head-on collision, shouldn’t there be some fermions somewhere in the sample that are headed straight at one another?

I shrugged, and said, “Quantum mechancis is magic.” Then I went on to give the more serious explanation, namely that it’s a mistake to think of the atoms as having definite positions and momenta within the sample, which is what you’re doing when you imagine head-on collisions. Instead, you’ve got fuzzy probability clouds moving around, and you never find two fuzzy fermionic probability clouds in the s-wave configuration.

Then I turned to Bill Phillips, who had won a Nobel Prize by that point, and said “Anyway, that’s how I think about it. How would you explain it.”

“Quantum mechanics is magic,” he said.

This also led to one of the best moments in my thesis defense. It started out seeming like the worst moment, though– I was going along through my prepared talk, if anything too quickly (I was starting to worry that I was going to leave too much time for questions), when I got to the expansion cooling trick for obtaining lower effective temperatures. I described the flight of the atoms as a “ballistic expansion,” and the outside-the-program representative on the committee, a rather scary condensed matter physicist, absolutely blew up.

“Who came up with this term, ‘ballistic expansion,’ anyway?!?” he bellowed angrily. I had no idea what he was talking about (“I don’t know. Galileo?”), and for an instant, I saw my academic career going down in flames. He was angry enough that other members of the committee were starting to intervene, but as he started ranting, he said something that made it clear what his real objection was, and I was able to wave off my bosses and say “No, I’ve got this one…”

The issue was that people in his field frequently misuse the term “ballistic” to refer to things like the motion of large clouds of electrons. That’s inappropriate, though, because “ballistic” flight implies that there are no interactions between particles, and electron clouds are always interacting (which is why condensed matter theory is so freakin’ difficult). Since I was talking about collisions in an expanding cloud, he assumed I was making the same mistake, and blew up.

Of course, the difference is that electrons interact via Coulomb forces, over very large distances. I was dealing with neutral atoms at low densities, so the collision rate was actually very small– at our highest density, there would be roughly 0.4 collisions per atom in the full duration of the experiment, and that’s without accounting for the fact that the density drops by better than three orders of magnitude during the expansion. The vast majority of atoms make it through the entire expansion without colliding with anything, so there’s no problem with using “ballistic” to describe the process.

(I knew this right away because one of the referees had asked the same basic question in a much less combative manner, and I’d worked out the answer before resubmitting the paper. Thank God for peer review.)

After I made it through that moment, I wasn’t too worried about the rest of the questions. He didn’t ask anything tough in the free-form question period, either, so I guess I had impressed him as not an idiot after all.

The other big thing I remember about this experiment is the data collection process, which took forever. In order to get enough counts to get a useful signal from the long expansion times (the lowest temperatures), each data run added together a thousand trapping cycles, which took 40-45 minutes. They needed to be taken in pairs, too, one polarized and one unpolarized, so a single useful data set took about an hour and a half. That still wasn’t all that great, in terms of statistics, so the final data sets presented in the paper are a weighted average of several individual data sets.

It took months to acquire all the data we needed (some of the sets would turn out to be useless, as well, which was never obvious until the scan was done), once the system was working right. The experiment was mostly conducted late at night, too– it would take most of a working day to look at the previous night’s results and do the preliminary measurements we needed, so the good data runs were generally taken between 7pm and 7am. And let me tell you, NIST was a pretty creepy place at three in the morning, when I was generally the only person in the building.

I also associate this experiment with McDonald’s cheeseburgers. There was a McDonald’s across the road from NIST, and I figured out at some point that I could start a data run going, run out to McDonald’s and get a Value Meal, come back to the office, and finish eating it just in time to catch the end of the run and start the next run. Provided I ordered something they already had on hand, which generally meant the two-cheeseburger value meal. Every now and then, they had McNuggets ready to go, but for a period of a couple of months when the experiment was running, I pretty much lived off of McDonald’s cheeseburgers.

(I swore off their burgers after that. It’s been almost ten years since the last time I ate one of their burgers– I’ll get chicken products sometimes when we’re on the Mass Pike, but no more cheeseburgers for me.)

The data collection and writing of this were pretty much all me. The experiment was originally started by Matt Walhout, and Uwe Sterr was a post-doc who did some of the initial collision measurements, but they had moved on by the time I really started taking data, and I did pretty much everything myself (I even wrote some C++ code to do the binning and averaging for the data analysis, the first and last time I’ve ever done that…). Paul Julienne is a theorist who came up with the really simple model we used to generate the theory curves in the paper (the paper describes how it works), and Steve Rolston was the PI for all the xenon stuff, and provided general guidance.

Shimizu’s group in Japan had done a similar experiment in krypton a while before we got our experiment working, and while we covered a much wider range of temperatures than they did, and measured a bigger effect, we didn’t really have anything all that new, and so decided not to try to squeak it into PRL. It was written for Phys. Rev. A from the start, and as such was a much more relaxed process, as I didn’t need to worry about the length. Paper torture was relatively un-torturous, as a result.

The down side of publishing it in Phys. Rev. A is that it’s not all that prominent a paper (ADS lists 12 citations), which is kind of a shame, as I think it’s some of the coolest science I’ve been involved with. Those are the breaks, though. It was a fun experiment in a lot of ways, and I enjoyed doing it.

That’s the last of the experiments that went into my thesis. I am an author on one more metastable xenon paper, though, and I’ll write that up next.

4 thoughts on ““Quantum Mechanics Is Magic”: The Making of “Spin polarization and quantum statistical effects in ultracold ionizing collisions”

  1. (I swore off their burgers after that. It’s been almost ten years since the last time I ate one of their burgers– I’ll get chicken products sometimes when we’re on the Mass Pike, but no more cheeseburgers for me.)

    There’s your next silly/informal poll: What did you eat so goddam much of during grad school that you just can’t stand the taste of it any mroe?

    (Not to be confused with, “What did you drink so much of as an undergrad that the smell of it makes you puke?” inspired by a recent encounter with peppered vodka.)

  2. Reality does not cower before Euclid, Newton, or Descartes. Reality is jazz, theory is Mozart. Build with approximation, discover with exactitude.

  3. The experiment was mostly conducted late at night, too– it would take most of a working day to look at the previous night’s results and do the preliminary measurements we needed, so the good data runs were generally taken between 7pm and 7am.

    Yup, this sounds familiar! How much published atomic physics data do you think was taken late at night? And for that matter, how much of published data in other fields was taken after hours (astronomy excepted)?

    That might be an interesting thing to know, since one of the biggest differences I’ve found between working in policy and working in science is that you can’t do policy work at night — if it doesn’t get done by the end of the day, it’s got to wait until tomorrow.

    I wonder if this is one of the most important differences between scientists (or STEM workers in general) and more mainstream office workers that makes it hard for both groups to understand the other?

  4. How much published atomic physics data do you think was taken late at night?

    A few years ago I heard Wolfgang Ketterle give a talk about his work on Bose-Einstein condensates for a general audience. He showed a photo of the lab book from when his lab produced its first BEC. IIRC the time stamp was between 5 and 6 in the morning, after they had been running all night.

    you can’t do policy work at night

    Depends what sort of policy work you are talking about. Certain things that require consulting with others have to wait until business hours, but internet research and writing can happen at any time of day.

Comments are closed.