A Week in the Lab: Just-So Stories
The second graph is the most complicated of the three, but it's probably the easiest place to start. This is a plot of the initial amplitude of my curve fits as a function of pressure. That amplitude is basically a measure of how much stuff hits the detector when I don't put any voltage on the screening wires to steer charged particles away. There are two important features of this graph: at high pressure, if you increase the pressure, you increase the signal, and at low pressure, if you increase the pressure you decrease the signal. This suggests that there are two different effects at work here.
The high pressure result is easy to explain: as you increase the pressure, you increase the number of atoms passing through the discharge region. More atoms passing through means more atoms excited to various states, and more atoms ionized, thus, more free electrons to hit the detector.
The lower pressure result is a little trickier, but the key to figuring that part out is that the pressure measured is the pressure in the whole chamber, not just the discharge region. Higher pressure means more stuff between the discharge region and the detector, which means there are more chances for an electron to hit something, get deflected, and not make it to the detector. As you increase the pressure, you decrease the chances that a free electron created in the discharge region will get detected, and thus you decrease the signal.
When you put these two competing (and very non-linear) effects together, you get the "U" shape seen in the graph of amplitude vs. pressure. At high pressure, the sheer number of particles created is large enough that substantial numbers get through, even though there's a lot of gas in the way. At lower pressure, the rapid increase in the distance an electron can move without being deflected makes up for the fact that fewer electrons are being created. The total number is lower, but a larger fraction of them survive to be detected, so the net signal is higher.
The survival effect can also be used to explain the third graph, of the characteristic stopping voltage vs. pressure. Here we see a smooth decrease in stopping voltage as we increase the pressure. That sort of makes sense in terms of electrons suffering collisions as they come across the chamber. At higher pressure, the electrons hit more things as they pass through the chamber, and they end up losing a bunch of energy. The lower-energy electrons are more easily steered away from the detector, so the stopping voltage is lower. At lower pressure, the electrons cruise right across without hitting anything, so it takes a more substantial screening potential to ward them off.
That leaves only the graph of final current vs. pressure to explain. This is a little trickier, but my current theory is this: In addition to producing a lot of electrons, the discharge operation also produces metastable atoms and positive ions. metastables or positive ions striking the detector plate will also cause a current to flow, but you can argue that the current ought to have the opposite sign from that caused by the electrons. Metastables and positive ions are also a little more fragile than the electrons are-- they're bigger, move more slowly, and can be de-excited or neutralized before making it to the detector.
So, the idea is this: The discharge spits out enough stuff that there's always something hitting the detector. At high pressure, this signal is mostly electrons, because they're the only thing that survives across the chamber. As you decrease the pressure, you get more positive ions and metastable atoms surviving the trip, which pulls the final current down, and eventually causes it to change sign. At low pressure and high (negative) screening potential, the signal on the detector should be dominated by positive ions and metastable atoms.
That's the theory, here. The problem I have now is that I don't know how to sort those two out from one another. All I really care about is the metastable atoms, and if I could get rid of the positive ions, that would make my life easier. At the moment, though, I don't have a way of getting rid of both positive ions and electrons, so I'm sort of stuck.
Or, I could be telling the wrong just-so stories about these graphs, and what's going on here could be something completely different. There are a couple of things that hint that this "I'm an idiot" theory might be the right one, but I'll have to do some more tests before I know for sure. Of course, then the question becomes "How much time do I want to spend trying to understand the plasma discharge source, when that's not really the point of this experiment?"
A Week in the Lab: It's Not Science Without Graphs
Today's project was to test out the operation of the detector I installed yesterday. I always approach this sort of thing with a little trepidation, as I'm never really sure that I'm not going to short something out and destroy the entire lab. As a result, I end up doing a bunch of silly preliminary measurements that basically amount to stalling, such as checking the calibration on the panel meters of the two voltage supplies I dragged out of the stockroom to supply the screening potential...
(To be fair, an initial check made it look like there was something funny with the gauges. It turned out to be a dying battery in the digital multimeter I was using to check the voltage...)
The detector, as I said in yesterday's post, is a square bit of sheet steel, placed in the path of the atom beam. When metastable atoms hit the surface, they can give up their internal energy, and knock electrons loose. The lost electrons go off somewhere, and are replaced with other electrons flowing up from the ground, so if you attach a very sensitive meter (or, in my case, a very sensitive current amplifier) between the steel plate and ground, you should be able to measure the current, and thus determine how many metastable atoms are hitting the plate per second.
There are also a couple of longer copper wires extending out past the plate into the vacuum system. These wires get connected to voltage supplies, and act to deflect any charged particles (electrons or positive ions) that are headed toward the plate. Electrons or ions striking the plate will also cause a small current to flow, and we don't want that. Metastable atoms are neutral, and will be unaffected by the wires, so with the right choice of screening voltage on the wires, we should see current only from metastable atoms.
The very first test of any detector is to make sure that when there's nothing to detect, it detects nothing. I shut down everything that might produce charged particles in the vacuum system, hooked up the current amplifier on the 1 volt per picoampere setting, and looked at the output on an oscilloscope. The scope showed a current of -0.1 pA, which is pretty darn close to zero. I could generate similar currents by just walking around near the wires, so I was pretty happy to call that zero. Then I started turning other things on, to see whether any of them made a difference. Running the screening wires up to +/- 100V made no difference, turning the vacuum gauge back on made no difference (that was a little surprising, as it should produce some ions, but it's almost a meter away, and around a couple of corners, so it's not that surprising that it didn't produce a signal), and turning the chopper wheel in the laser beam on didn't make any difference (hey, you never know...). Turning on the RF supply used to run a plasma discharge created some pick-up, but that was mostly broadcast noise, and didn't change the zero point.
The next step was to plug in the 123 nm lamp, and see what that did. That made a big difference-- something like 80 pA of current. It was a nice steady signal, so it didn't look like RF pickup from the discharge in the lamp, and I'm ashamed to admit that it took me a minute to figure out what was going on. I must not have been sufficiently caffeinated.
I figured it out pretty quickly, and ran a couple of tests to confirm my theory, by putting different voltages on the screening wires. The result was this graph, duly pasted into the lab notebook. When I put a positive voltage on one of the wires, the signal shot up dramatically. When I put a negative voltage on the other wire, the signal dropped off. Putting a negative voltage on both wires killed the signal really quickly.
The signal being detected here is from the photoelectric effect. The 123 nm photons in the lamp have enough energy to knock electrons out of the steel chamber walls, and some of those electrons fly down the length of the apparatus to strike the detector. A positive voltage on the screening wire draws them in more effectively, while a negative voltage steers them away.
This is a useful result, as it shows that the detector works, and the screening system can get rid of electrons in the chamber. It's a little annoying to have those electrons around in the first place, but there's nothing that can be done about that-- the laws of physics always win.
The next step was to start looking for metastable atoms. It wouldn't make sense to go directly to trying to do optical excitation, though, because I need to confirm that the detector works before trying that, so I started out with a plasma discharge (which is made with an RF coil wound around the outside of the glass tube I feed the gas in through). That should produce a little of everything-- electrons, ions, and metastable atoms. So I turned off the lamp, let a little gas flow into the chamber, and started up the plasma discharge
Sure enough, the signal shot way up-- about 200 pA with a moderate screening potential, and it overloaded the amplifier at lower voltages. So far, so good.
In this situation, there are two things that can be adjusted: the screening potential, and the gas pressure. So I varied the pressure, and recorded the detector current vs. screening voltage for several different pressures. The resulting graph is kind of a mess-- about the only thing you can learn from it is that the lowest pressure I tested (the data shown as open triangles) was radically different than the others. Of course, I hadn't done a very good job of changing the pressure, so I went back and took a second set of data to fill in the middle part. This graph is a little clearer (the pressures are given in slightly different units, but the two lowest values are the same (one micro-Torr is 10-6 Torr)), and shows a bit more of a trend. It's still not completely obvious, though.
To try to get more information out of this, I did a bit of curve fitting. If you look at the different sets of data, you can see that they all have the same basic shape-- they start off high, and drop down smoothly as the voltage increases. They start at different values at low screening potential, they drop off at different rates, and they reach different final values, but they all follow the same basic trend.
The default fit function for this sort of curve is a decaying exponential with an offset, and that provides three fit parameters: the final value at high screening potential, the initial value at low screening potential, and the characteristic voltage at which the signal drops by a factor of e (a bit less than three). Those three graphs are at least a little simpler.
So, what's going on here? I'm not entirely sure. I do have some guesses, and I can at least construct just-so stories to explain each of the graphs, but I need to do some more testing. I'll explain my guesses tomorrow, but feel free to speculate in the comments...
What about the optical excitation, you ask? Well, I did make a quick run at it, turning the 819 nm laser on and pulling it to the right wavelength, but I didn't see anything. That was pretty disappointing, but after I had shut everything down for the night, I noticed something that I'd forgotten: when I was aligning the laser yesterday, I had put a big filter in the beam to knock the intensity down to a level that wouldn't saturate the camera. I forgot to take the filter out before today's test, and that factor of 100 might make a tiny bit of difference in the effectiveness of the source...
A Week in the Lab: Cleanliness Is Next to Somethingorother
If you had been in my lab this afternoon, you might very well have asked "Why on earth do you have two nearly identical Welch Duo-Seal pumps sitting right next to one another?" The answer, ultimately, is "Cleanliness," but that didn't really come up until the end of the day.
The project for today was to install a new detector, which you can see in this picture. The dark grey square in the center is a half-inch square of sheet steel, connected to an electrical feed-through running out of the vacuum system. The idea is that metastable atoms hitting the plate can release their internal energy, and that 10 eV is enough to knock an electron off one of the atoms in the surface. When that happens, a small amount of current will flow into the plate to replace the lost electrons. The current involved is tiny-- on the order of picoamperes-- but I've got an amplifier that will measure it, and measuring that current provides a measurement of the metastable atom flux.
The shiny copper wires extending upward from the flange are there to steer charged particles away. You can also get a current by bombarding the plate with ions or free electrons, as are produced in a plasma discharge source, and I don't really want to detect those. Hence, the two copper wires, one of which will be put at +100 V, the other at -100 V, which should hopefully be enough to steer charged particles away from the plate. The metastable atoms are neutral, and thus shouldn't feel any force from the potential on the wires, so that signal should remain.
As long as I had to open the vacuum chamber up (I turned the turbopump off before I left last night), I took a little time to re-align the optical system. It's very important that the 819 nm laser intersects the atomic beam at a point just past the end of the glass tube that we feed gas in through, and in the same area illuminated by the 123 nm lamp (about which more later). The atoms have to absorb a 123 nm photon followed immediately by an 819 nm photon in order to end up in the metastable state, so both those light sources need to hit the same point in space.
Of course, that point is a point hanging in space right in the middle of the vacuum system, which means it's kind of tough to verify that everything's aligned. The way I checked the alignment was to run a piece of copper wire down from the open end of the vacuum system to right where I wanted. The wire is the bright orange line coming in from the right of this picture, while the glass tube is a little hard to make out to the left. For reference, the window in the blurry flange in the foreground is about an inch and a half across.
With the wire positioned where I wanted it, it was easy to verify that the 819 nm laser was passing through the right region of space-- I could see the shadow cast by the wire in the beam as it exited the chamber on the other side. I also aligned an 811 nm laser to hit the same spot, so I'll be able to see fluorescence as well. I put a chopper wheel in the 811 nm laser beam, and set up a lens to image the end of the wire onto a photodiode, and moved the lock-in from yesterday's installment over to the rack by the chamber, and confirmed that the lock-in could see the laser light scattered from the wire. It worked very nicely, actually, which was a relief.
All of that was completed by lunchtime, so I left things there, and went off to play basketball. When I got back from the field house, it was time to install the detector (a simple matter of bolting a flange onto the chamber), and deal with the lamp.
The lamp that I use for this is a ridiculously expensive piece of apparatus-- the final version was $7,000. This seems preposterous, but it's the going rate, because 123 nm is not an easy wavelength to deal with. Light at that wavelength won't propagate very far through air, for one thing, and it definitely doesn't pass through ordinary glass. This means that the window has to be made of a special material-- lithium fluoride, in this case-- and the lamp can only be operated in vacuum.
Another charming feature of 123 nm light is that the photons are energetic enough to break down a lot of organic molecules, and turn them into a nasty film bonded to the front window of the lamp. This film, in turn, absorbs ultraviolet light, meaning that you don't get enough photons out to make the source work.
In order to avoid this, the lamp must be kept very clean, so every time I open the system up, I need to remove the lamp and clean it by polishing the window with a spongy little swab dipped into a special "polishing powder" (a granular white powder that came in a little vial, and led my summer students to make a bunch of jokes about drugs...). Here's a picture of the front face of the lamp just prior to cleaning. The aluminum bit extends about 30 cm into the vacuum chamber, allowing me to get the lamp up close to the beam to maximize the number of photons hitting the atoms. The brass housing in the background contains the electronics for running the lamp (which is basically a fancy RF discharge). The lamp is being held by an Alliance assassin who was sent after me because I can kill people with my brain.
The other tricky thing here is that the roughing pumps used to do the initial pump-out of the system are oil-based pumps, and oil is famously an organic molecule. That means that steps have to be taken to ensure that the lamp chamber stays oil-free. This is the reason for the two pumps in the picture above. The two aren't quite identical-- the one in the foreground has a liquid nitrogen cold trap attached. This is basically a bucket with hollow walls that we fill with liquid nitrogen. There are two openings into the space in the walls, one of which is connected to the chamber, the other to the pump. That way, any oil molecules trying to make their way from the pump into the chamber have to pass by a large expanse of really cold metal, and the odds are very good that they'll stick.
The pump with the cold trap is borrowed from our particle accelerator lab, and used to do the initial pump-down only. Once the turbopump is spinning, there's not much danger of oil from the backing pump getting through to the chamber, but the turbopump can't be started when the chamber is at atmospheric pressure. So, I use the cold-trap pump to get the pressure down to a level where the turbo can be started (around 50 mTorr), then close a valve between the pump and the system, replace the cold-trap pump with the dirtier pump that's a permanent part of my lab, pump down the area behind the valve, and re-open the valve. The swap takes maybe five minutes, and then everything just runs.
So, to recap: 1) I opened up the vacuum chamber, 2) realigned the optical system, 3) confirmed that the lock-in still works, 4) played basketball for an hour, 5) installed the new detector, 6) cleaned and re-installed the lamp, and 7) closed everything up and re-started the vacuum pumps. That's what I call a busy day. When I left, the system was down to 10-6 Torr, and I'll let it pump overnight to see how low it goes (I got into the mid-10-7's with the previous arrangement, so I don't expect it to go much lower).
Tomorrow, I'll see if I can see some metastables.
A Week in the Lab: Lock and Load
So, let's say you have a small signal-- the tiny amount of light scattered by a collection of metastable atoms passing through a laser, for example. And let's say that you need to detect this signal, in spite of a large amount of noise-- the much larger signals caused by the lights in your room, for example. What do you do?
In principle, you could just measure the total amount of light with and without your signal of interest, and take the difference between the two. For really small signals on top of really large backgrounds, though, that tends not to work too well. In some cases, the rounding error you get when you digitize the large signal is at least as big as the small one, and you'll never see anything other than noise.
The thing to do in this situation is to modulate the signal of interest at a known frequency-- say, by chopping your laser on and off-- and only look at signals that oscillate at that frequency. This may seem like a counter-intuitive thing to do ("I'm going to make my signal stand out by only having it on half the time?"), but it works very well. There's a dedicated electronic device for this sort of thing, called a lock-in amplifier. You give it an input consisting of the small signal of interest plus the large amount of noise, and you give it a reference signal at the modulation frequency, and it give you back whatever part of the input signal oscillates at the same frequency as the reference signal, amplified by a substantial amount. Unless you're incredibly unlucky (or just plain dumb), the output should be just the signal you're looking for, without any of the noise you don't want.
Lock-in amplifiers are standard equipment in physics departments, but they also tend to be kind of expensive-- I couldn't find a new one in a self-contained package for under $1000 (you can get lock-in boards that are meant to plug into other systems for less, but I don't have any of the systems they're meant to plug into...). Scrounging around through the department, I found four, none of which could be certified to work entirely as designed... One of the four works all right some of the time, but freezes up at odd intervals. Another does an admirable job of removing noise from the input signal, but weirdly introduces noise on the input signal, so you end up with something that's not all that much better than where you started (I suspect this is an impedence issue, but I can't be bothered to track it down). The other two appear to work correctly, but every time I want to use one, I end up spending a whole day convincing myself that it really is working.
Today was that day. I hooked up a lock-in to look for metastable atoms in the source chamber yesterday, and it gave me all sorts of screwy output signals. I've learned the hard way that when this happens, it's better to immediately go back to first principles, and look at the performance in a controlled experiment, rather than continuing to putter around for a day and a half looking for a signal that may or may not be there with a lock-in that may or may not be working. I'll be looking at it on a lab bench sooner or later, so it might as well be sooner.
The test set-up is really very simple: I have a laser that passes through the reference krypton cell I mentioned yesterday, and falls onto a photodiode. When I scan the frequency of the laser back and forth across the resonance frequency of the atoms, the intensity of the light increases and decreases as the atoms absorb light when the laser is on resonance, and don't absorb when it's off resonance. This scanning is fairly slow-- something like 10 Hz.
Then I put a chopper wheel (a rapidly spinning wheel with holes cut in it) in front of the cell, so the laser beam is blocked and unblocked rapidly (about 400 Hz with the wheel I'm using). This means that the signal from the laser on the photodiode now blinks on and off at 400 Hz, with a slower modulation due to the laser scan. When the laser is far from the atomic resonance, there's a big modulation (as the laser beam passes through the cell unchanged), and when it's on resonance, there's a small modulation (even when the laser is unblocked, the light is mostly absorbed by the gas in the cell).
Then I put a big stack of filters in front of the laser, to reduce the intensity by a factor of 10,000. That way, even when the laser is unblocked, the photodiode signal from the laser is at least a factor of 100 smaller than the signal from the room lights. Then I use the lock-in amplifier to recover the original 10 Hz signal due to laser absorption in the gas cell.
As is often the case, I wound up spending a bunch of time chasing a phantom problem. In this case, it was a beautifully sinusoidal noise signal on the output at about 66 Hz. The noise was there even when I blocked the laser beam.
"Aha!", you say if you've ever done any electronics, "That's line noise!" (The voltage on electrical lines in the US alternates at 60 Hz, and if you do anything with electronics, you find 60 Hz noise creeping in all over the damn place.)
Line noise is the obvious guess, but it's also wrong. I didn't figure out the problem until I started playing with the chopping frequency-- when I reduced the speed of the wheel to 200 Hz, the noise signal dropped to 33 Hz. It turned out to be some sort of strange pick-up effect from having the chopper wheel too close to the discharge cell. Moving the chopper wheel back six inches made the 66 Hz noise disappear completely.
Experimental physics is forever throwing up this sort of thing. There's probably some moderately interesting physics behind this noise-- I suspect that the wheel was modifying the plasma discharge characteristics in a way that made the amount of light from the discharge itself fluctuate-- but it's only interesting to people who are stupid enough to butt their chopper wheel right up against a plasma discharge cell. There isn't a Journal of Strange Noise Sources that I could write this up for-- it's just a nuisance, and cost me most of a day.
Once I figured out the source of the problem, and verified that the lock-in was working correctly, I took a little time to make some power measurements, and figure out how sensitive the diode I was using will be in the actual experiment. Running some back-of-the-envelope numbers shows that I get a nice signal out of the lock-in from about a nanowatt of laser power, or about a billion laser photons per second hitting the diode. For the optical system I have set up to put that many photons on the diode will require something like 1011 photons per second to be emitted in the source, which would correspond to something like 1010 metastables per second. You probably need to bump that up by another factor of ten or so to account for all the factors I've left out. (These are all back-of-the-envelope numbers, mind.)
That's actually kind of marginal as a detector system-- a good plasma discharge source puts out something like 1013 metastables per second, and I probably won't get that many on the first pass. I wouldn't want it to be my only means of detecting metastable atoms in the source, because I'd never be sure if there was no signal because the source wasn't working, or if there was no signal because it was just too small to detect. Happily, I have another means of detecting metastable flux, and I got the detector assembly back from the shop today. Tomorrow's project will be to install the new detector in the vacuum system, and then I can make a real run at testing the source.
At least, that's the plan until something else stupid comes up.
A Week in the Lab: Lasers, Eight O'Clock, Day One
Well, OK, maybe nine o'clock... Nine-thirty, at the latest. I did start with lasers, though.
There are two lasers involved in the optical excitation source development project. One is the 819 nm laser needed for the second step of the excitation process (the right-hand side of this handy diagram), the other is an 811 nm laser that will eventually be used for detecting the metastable atoms that are (hopefully) produced.
The 811 nm laser system is a little more complicated, as it's a grating-locked laser. The light comes out of the laser, and bounces off a diffraction grating. Most of the beam just reflects straight off the grating, and goes out to the optical system, but some of the light is reflected back into the diode laser itself. That may seem like an odd thing to do, but it allows us to do fine adjustments of the laser wavelength by changing the orientation of the grating-- sending light back into the cavity causes the laser to run at the specific wavelength sent back, rather than emitting light over a relatively a broad range, as diode lasers are wont to do. Changing the grating alignment changes the wavelength sent into the cavity, and allows us to tune the laser to where we want it.
These are nice systems in general, but they need to be tweaked up every so often, and this one hadn't been touched since October, so it needed a fair amount of adjustment. This is accomplished by adjusting two small screws on the grating mount, while keeping track of the wavelength. We use a fiber-coupled wavemeter for this job, so the first task was to get the laser back into the optical fiber, and then start tweaking the grating. Of course, every time you move the grating, you change the alignment of the laser, so this is an iterative process...
After about an hour of that, I had the laser wavelength close to where it needed to be. Here's a picture of the laser diagnostics: the wavemeter is the middle box with the LED readout saying "811.51 nm," while the top and bottom oscilloscopes are other signals that I use to make sure the laser is functioning correctly.
The wavemeter doesn't have good enough resolution to tune the laser perfectly into resonance, so at this point, I moved on to the ultimate test: comparing the frequency to the krypton resonance line. For this, I've got a small krypton cell (pictured here) that I run an RF plasma discharge in to produce metastable atoms as a reference. I shine a bit of the laser light through, and look for absorption of the light with a photodiode. Finding these absorption lines can be a hugely time-consuming process, if you don't know exactly what you're looking for (ask my summer students), but I've adjusted this laser to this particular line a half-dozen times before, so it wasn't that difficult. Here's the scope trace showing absorption-- the triangle wave is the signal I use to sweep the laser frequency back and forth, while the big dips in the lower trace show the laser light being absorbed.
I had the laser tuned up to the right line by lunchtime, so I went off to the gym to play basketball (MWF at noon). When I got back, I spent the afternoon doing a bunch of the little annoying calibration measurements that you always need to do at the start of a project.
One of these was tuning up the 819 nm laser, which is just a free-running diode, with no grating. There's still a wide tuning range available, though, as the peak wavelength emitted by a diode laser depends on the temperature of the diode, and the amount of current you run through it. If you're willing to make drastic changes in the temperature, you can pull the wavelength by several nanometers. This is a Good Thing, especially because the lasers I'm trying to use for this want to run at about 822 nm at room temperature. To get them down to 819 nm, I need to run the temperature down below 10 C. Back in the summer when we started this, this led to frost forming on the laser mounts-- now that winter is here, the humidity in the lab had dropped from 75% to 20%, so frost isn't really an issue any more (though static shocks are a big problem). (That seasonal humidity swing has been the subject of several discussions between me and the facilities people, but there doesn't seem to be much that can be done about it...)
I've made a couple of changes in the laser system since the summer, so I needed to re-do the wavelengths vs. temperature calibration. This is a tedious process, as it takes a few minutes for the temperature to settle down after you make a change, so I alternated between making temperature measurements and pressure measurements.
Another important element in this project is the pressure of krypton gas that we feed into the system. The effectiveness of the optical excitation mechanism is somewhat pressure-dependent, and we need to be able to control that pressure, and know what we're doing. I've got two gauges in the system, neither of which is exactly where I'd like it to be (there's nothing I can do about that), so there's a little guesswork involved in this process, and I needed to do a consistency check on the gauge readings. On top of that, the one that's closer to the region of interest (a thermocouple gauge) has a tendency to go all wacky when I switch on the RF for the discharge source, so it would be nice to know what reading on the ion gauge corresponds to what reading on the thermocouple gauge.
A good chunk of the afternoon was spent making alternating between noting the laser wavelength and changing the temperature, and noting the two pressure readings and changing the gas pressure. Lather, rinse, repeat. Happily, the laser can still be tuned to the right wavelength (more or less-- it has a 2 nm bandwidth, so it's not that sensitive), and a graph of thermocouple gauge pressure vs. ion gauge pressure was plausibly linear. Graphs were made, and dutifully pasted into the lab notebook, and by then it was five o'clock, so I headed home.
Tuesday, the new detector system should be finished (it's in the shop now), and I need to spend some time figuring out why my lock-in amplifier isn't working right (Again. Damnit.). I hope to start making a run at detecting metastable atoms on Wednesday.
A Week in the Lab: The Big Picture
Before diving into details, I should probably provide a little context for the experiment. So, what is it that I'm working on down in the basement of the science and engineering building?
I posted a brief description of the project back when Dan and I were writing up the original proposal paper. The basic idea is to measure the amount of krypton in some other gas (neon and xenon are the main ones that people are interested in) by trapping single krypton atoms out of a stream of gas passing into our vacuum system. The exceptional frequency selectivity of laser cooling lets you do this, and there's a group at Argonne National Laboratory that pioneered the single-atom trapping of krypton, and have used it for radioactive dating.
There's one little quirk of the krypton system that makes it different than the other rare-isotope trapping experiments out there: krypton atoms in their lowest energy state (the "ground state" in physics jargon) can't be trapped with currently available lasers. The laser wavelength you would need to trap ground-state krypton is around 120 nanometers, and you can't get a laser at that wavelength (and even if you could, it wouldn't propagate through air).
The way around this problem is to put the atoms in a metastable state, a state above the ground state with a very long lifetime. "Very long" in this case means "about thirty seconds," which is effectively forever in atomic physics terms (a more typical atomic lifetime would be thirty nanoseconds (0.00000003 s)). The whole trapping and detection process takes only a few seconds, so atoms will stay in the metastable state through the entire experiment, and once they're there, there's a convenient transition to another excited state with a wavelength in the infrared region of the spectrum (811 nm), where diode lasers are (relatively) cheap and readily available. You can't cool ground-state krypton, but once you get it into the metastable state, it works just as well as any other atom.
Somewhat surprisingly, this is a feature, not a bug, at least when it comes to measuring krypton as a background contaminant. The energy needed to excite the atoms to the metastable state is about 10 electron volts (eV), which sounds like a trivial amount of energy to particle physicists (who regard energies of a million eV as beneath notice), but is pretty significant in atomic physics-- another way to think about it is that 10 eV is roughly equal to the kinetic energy of an atom in a gas at a temperature of about 12,000 Celsius. In other words, there aren't a lot of ways for a krypton atom wandering around my lab to end up in the metastable state. And since we can only trap and detect atoms that are in the metastable state, this eliminates a lot of the worries you might otherwise have about contamination in the system. Any krypton atoms in the air in the lab (air is about one part per million Kr) would need to both get into the system somehow (not a trivial matter) and also acquire 10 eV of internal energy before they could possibly be detected and compromise the measurement.
Of course, for this to work, we need to be able to put the krypton atoms in our samples of neon or xenon into this metastable state. There's a well-established technique for doing this, which involves exciting the atoms in a plasma discharge. The trick is to ionize a few atoms in the sample, and then do something to drive the free electrons through the gas at high speed (either by putting a high voltage across the sample, or by applying a radio-frequency oscillating electric field). As the electrons pass through the gas, they will slam into some of the gas atoms, and excite them to, well, pretty much every possible state in the atom. Some of them will inevitably end up in the right metastable state, and then we're in business.
There are two major problems with this method. First, it's inefficient-- if you're lucky, one atom in a thousand ends up in the right state. More importantly, it's dirty. You excite pretty much everything in the source, and you even ionize some of the atoms. Those ions can be driven into the walls of the source chamber at high speed, and become buried deeply enough that only another impact will knock them loose. This turns out to be a major source of contamination-- if you run a sample of gas containing a large amount of krypton through the system, you'll be seeing atoms from that sample coming out of the walls until the end of time, and they will completely obscure your attempts to measure lower abundances of krypton in later samples.
So, we'd like a better way of preparing our metastable atoms, preferably one that excites only metastables, and doesn't produce any ions. Some sort of optical technique would seem ideal for this, say, using a laser to put the atoms into the metastable state. It's a nice idea, but you quickly run into a problem: at the microscopic level, physics is reversible, and the same rules that make the metastable state have such a long lifetime make it almost impossible to directly excite atoms into that state.
The way around this is to use an indirect method to excite atoms to the metastable state. If you use two different light sources (one ridiculously expensive lamp at 123 nm, and a laser at 819 nm), you can get some of the krypton atoms to absorb one photon from each beam, and end up in a state above the metastable state. About three-quarters of those atoms will spontaneously decay into the metastable state, where they get stuck for the next thirty seconds or so, and we're in business. This may seem like a ridiculously roundabout way of doing things, but it has the potential to be more efficient than the plasma discharge method (maybe one percent of the atoms ending up in the metastable state), and more importantly, it will only excite krypton atoms, and only to the one metastable state. This optical excitation method should completely eliminate the problem of contamination from ions embedded in the walls.
So, this is the project I had a student working on this past summer, and that I'm going to try to tackle this week. We're attempting to build a better metastable krypton source. The basic optical excitation scheme has been demonstrated in a closed cell containing krypton, but nobody has yet managed to make an atomic beam source by this method, and you need a beam of metastable atoms for the single-atoms trapping scheme to work.
In subsequent posts, I'll explain a little about the apparatus we're using to try to make this work, and how we plan to detect the metastable atoms when we get them, and all that fun stuff.
A Week in the Lab: Slow-Motion Experimental Physics Live-Blogging
It's been a while since I've posted much about physics, and I sometimes feel guilty about that. Particularly since Mixed States is showing me lots of posts by physics types talking about their research into the deep structure of the universe, and all that.
There are a couple of reasons for the relative lack of physics content here. One is philosophical-- I've committed myself to the idea of making my science posts accessible to a non-technical audience (as much as possible), which means those posts are a great deal more difficult to write than if I could just throw in all the technical jargon and equations and whatnot. I'm not sure I actually have a non-technical audience, at this point, but that's the goal I've set for myself, and I try to hold to it.
The bigger problem, though, is that most of the people posting regularly about research stuff are theorists, while I'm an experimental atomic physicist. The business of experimental atomic physics, at least in my experience, involves a whole lot of thinking "Where's the goddamn socket wrench, anyway?", and not so much the Deep Thoughts about Big Questions in physics. Not on a daily basis, anyway.
Theoretical work, at least from my outsider's perspective, provides many more opportunities for thinking about big picture issues. Particularly in the less computational areas of theory (people doing complicated molecular modelling spend a lot of time thinking "Why won't this goddamn thing compile, anyway?" Some of them give research talks that are like that...)-- A field like string theory or quantum information theory appears to be nothing but Deep Thoughts (though that might be an illusion caused by the fact that I don't understand about 80% of what they say).
As I tell my students when they prepare to give talks, the big picture stuff is a lot more interesting than the nuts-and-bolts daily grind of an experiment. But, as I said, I do feel guilty sometimes about not posting more about my own research.
As it happens, we're between classes at the moment, and I find myself with a sizeable block of dedicated lab time. My rough plan is to spend this week working on a relatively self-contained project, that was started by one of my summer research students. I have a fairly concrete list of tasks to work on, and there's a reasonable chance of getting actual data in the next week or two.
So here's the idea: Slow-Motion Experimental Physics Live-Blogging! (Yeah, that'll draw a crowd...) I'll attempt to post a series of articles explaining what I'm doing on this specific project, and how it's going. If I remember to bring the digital camera in to work, I may even post pictures of the apparatus. It'll be an all-access look into the glamorous world of lost tools, bloodied knuckles, minor electrocutions, and vacuum plumbing that is the development of an optically excited metastable krypton atomic beam source.
I'll try to keep this up for one whole week, on the usual one-post-a-day schedule, though it's possible I won't last that long-- if things go very badly, posting about work on the Web may be the last thing on Earth that I want to do, while if they go very well, I might end up being too tired to talk about it. But I'll at least give it a go for a few days before returning to the endlessly fascinating topic of what's on my iPod...
Also, I'd Like a Pony. A QUANTUM Pony!
We recently had a joint colloquium between Computer Science and Physics (OBInsideJoke: You can just hear the technologies converging...) in which Ivan Deutsch gave a talk about the general principles of quantum computing. I thought it was a very nice talk on the big-picture side, and most people agreed.
An unscientific survey of the attendees afterwards reveals that many of the people hearing the talk would've liked a more detailed explanation of the workings of quantum algorithms, and exactly why it is that the ability to have bits in superposition states lets you factor numbers quickly. I couldn't really help much with this, because I don't understand it myself.
Thinking about it, I don't know that I've ever heard a talk that really explained this. (I'm told that my old group at NIST once got someone to factor 15 on a blackboard, but I was away at a meeting when that happened.) I think I understand why, in terms of the physics market: it's a really complicated process, and difficult to explain, and it was worked out by Peter Shor a decade ago, so doing an hour-long seminar on how the factoring algorithm works isn't going to get anybody a job.
Still, it'd be really nice if something along these lines could be arranged. So, since my posts are being syndicated to smart people all over the world, I'll throw this out there: can anyone suggest the name of a person who:
- can walk through one of the big quantum algorithms (Shor's factoring algorithm, Grover's search algorithm, something else that's not trivial) explaining how quantum effects speed things up, and
- can do this at a level comprehensible to undergraduate physics and computer science majors, and
- might reasonably be enticed to come to Schenectady and do this for us?
We can't offer vast sums of money for expenses or honoraria (so Europeans are right out, unless they're going to be in the eastern US anyway), but we could probably come up with something. I'd also be haappy with a more general research talk, plus a guest lecture for my senior-level "Quantum Optics" class, if such a thing could be arranged in late May.
The odds of this aren't good, I realize, but what the hell...
There have been vague reports of booklog activity in Chateau Steelypips, dealing with important issues like the similarity between George R. R. Martin and Robert Jordan, and what books to read as an introduction to series fantasy.
If you'd like some physics-type content, well, I hope to provide some later this week. In the interim, there are posts by Dave Bacon and Rod Van Meter talking about exciting new developments in ion trap quantum information processing. I probably ought to do a post about that, too, one of these days. In my copious free time.