Trapping of Neutral Mercury Atoms and Prospects for Optical Lattice Clocks

I'm not hugely enthusiastic about the ResearchBlogging.org project, but it's a little ridiculous that they've been active for weeks now, and there still isn't a single post in the "Physics" category. If they're going to offer the category link, something ought to come up when you click it, so let's give them some blogging on peer reviewed physics research.

The recent paper that most seems to lend itself to a quick explanation is this Phys. Rev. Letter from the Katori group at the University of Tokyo on the trapping of neutral mercury atoms. Full disclosure: Dr. Katori was a student of Prof. Shimizu, who I worked with when I was there in 1998, and I toured Katori's lab that year, where he was working on laser cooling of strontium.

As the title suggests, this is a two-part paper: first, they describe the laser cooling and magneto-optical trapping of neutral mercury, then they discuss its possible use in an atomic clock. It turns out that with some fairly reasonable-sounding assumptions, they expect to be able to reach a really impressive level of accuracy in the clock-- around one part in 10¹⁸. This would open up all sorts of interesting possibilities for tabletop measurements of fundamental physics-- for example, by comparing a mercury clock with a clock based on some other atom, you can search for time variation of the fine-structure constant, which turns up in some theories of new physics. There are also a number of exotic effects that become more likely in heavier nuclei, and mercury is the heaviest non-radioactive element trapped to date.

If you follow atomic physics at all, it might seem surprising that magneto-optical trapping of a new element rates a PRL-- after all, we're up to something like two dozen elements that have been cooled and trapped. Why is mercury worth a spot in the best journal in physics?

This is noteworthy, because trapping neutral mercury is a pain in the ass. Laser cooling depends on tuning a laser very precisely near an atomic transition, and the available transition in mercury is at the extremely inconvenient wavelength of 253.7 nm. Just getting enough laser light that far into the UV is an achievement-- they do it by quadrupling the frequency of a diode laser. They start with 1W of laser power at 1014.9 nm (which they need an amplifier in a build-up cavity to get), and end up with about 10mW of light at the cooling transition, which is just barely enough to do the cooling and trapping.

The cooling is done on a so-called "intercombination line," which means that an electron needs to both change energy levels within the atom, and flip its spin. This is normally forbidden, making the lifetime of the upper state longer than normal, allowing them to reach a lower minimum temperature than usual with Doppler cooling (the sub-Doppler cooling techniques people use in the alkali metals don't work for mercury), ending up with atoms at about 31 microkelvin. They trap their atoms out of a room-temperature vapor, catching the small fraction of atoms whose velocities happen to be low enough to be cooled effectively by their lasers, which means that the apparatus is relatively simple. Using fairly standard techniques, they catch around a million atoms in their trap (estimating the number from the amount of light scattered by the trapped atoms as they absorb and re-emit light from the trapping laser). That's not all that many as such things go, but given the wavelength they're working with, and the relatively simple apparatus, it's pretty darn good.

There are a lot of things to like about mercury, from an atomic physics perspective. It's got six stable isotopes, so if you want to look at effects that depend on the mass of the nucleus, you have a little room to play with. Four of the isotopes are bosons, and two are fermions, so you can look at different quantum statistical effects, if you like. And all of the isotopes are heavy, ranging from 198-204 atomic units, which means that all sorts of exotic effects are magnified.

Another nice feature of the system is that it has a "clock transition," a set of two states whose energy separation is very well defined, and can serve as a frequency reference. In this case, the states are connected by light with a wavelength of 265.6 nm, corresponding to a frequency of something like 1.1 x 10¹⁵ Hz. This is five orers of magnitude bigger than the frequency used in current cesium-based atomic clocks, which is good, because the accuracy of the clock increases as the frequency increases.

The rest of the paper deals with looking at the possible performance of such a clock, based on holding the atoms in an optical lattice, an arrangement of laser beams that creates a periodic crystal-like arrangement of sites at which atoms can be trapped. The tight confinement of the lattice allows the atoms to be cooled to even lower temperatures, improving the poential performance of the clock.

Loading atoms into a lattice can be tricky, because the lattice is created by something called a "light shift," which changes the energy of atomic states when light is applied to those atoms. This shift is different for different levels, and depends on the difference between the frequency of the light and the frequency needed to drive a transition to some other level so most of the time, putting a lattice on on top of a magneto-optical trap changes the separation of the states in the cooling transition (the upper state experiences a different light shift than the lower state), and screws up the normal laser cooling process. This means that you lose some of the atoms you started with. In this case, though, they use a neat trick that was (as far as I know) first developed by Katori for loading strontium into a lattice. He realized that there are lots of different levels around, and if you take all of them into account, you can find a "magic wavelength" for your optical lattice that causes the same shift for both states in the laser cooling transition. That works out to be around 360 nm for mercury, and if they operate their lattice there, they can substantially improve the performance.

They spend the last page or so evaluating various sources of uncertainty in the clock frequency. These include the "natural linewidth" due to the finite lifetime of the upper state in the clock transition (the shortest lifetime is around 100s, so this isn't a big limitation), residual light shifts due to the lattice light, shifts due to small magnetic fields, and the contribution of black-body radiation from the room-temperature environment (basically a light shift due to the presence of thermal radiation). These are all small, and the black-body contribution impressively so, working out to an error of about 10^-19, due to the short wavelength (high frequency) of the clock transition. It's an order of magnitude better than expected for clocks based on ytterbium or strontium, which were already pretty impressive on this account.

In the end, they estimate that the accuracy of their clock would be around 10^-18, which means that the frequency should be stable to an astonishing eighteen decimal places. That's something like one tenth of a second error over the age of the universe, which is pretty impressive. The current best I could find in a little quick Goggling is a claim of 5 x 10^-16 for NIST-F1 (though there may be a more recent number than that one).

That sort of level of accuracy offers all sorts of possibilities for exotic physics tests. Put it together with the fact that mercury is an attractive atom for other reasons, and, well, you've got a paper in Physical Review Letters.

(It should be noted that there are a number of previous experiments involving trapped and laser-cooled mercury ions. These also have a good deal of promise as possible frequency standards, and a lot of work has been done on that by the Wineland group at NIST in Boulder. The current paper is the first trapping of neutral mecury atoms, which have a different set of advantages and disadvantages from the mercury ions, the most obvious difference being that as neutral systems, they're not nearly as sensitive to other nearby charges as the ions are. This means they can be accumulated in greater numbers without messing things up. Ions can be trapped more strongly, and for longer periods of time, though, so there are trade-offs. My background is in neutral atoms, so I'm impressed by the current work; ion people may feel differently.)

Hachisu, H., Miyagishi, K., Porsev, S.G., Derevianko, A., Ovsiannikov, V.D., Palââ¬â¢chikov, V.G., Takamoto, M., Katori, H. (2008). Trapping of Neutral Mercury Atoms and Prospects for Optical Lattice Clocks. Physical Review Letters, 100(5) DOI: 10.1103/PhysRevLett.100.053001