The trick is that the second is defined to be the frequency of an unperturbed cesium atom, which is about as real as that "frictionless plane" that you might have had in high-school physics.
An example of the problem is this: for technical reasons, a small magnetic field is needed inside a cesium clock. Magnetic fields change the spacing between all atomic energy levels to some degree. For cesium, the relevant change is very small, but it is still there. What you need to do is measure the magnetic field, calculate how much it affects the frequency of the atomic transition, and correct your output frequency by the required amount. What ultimately sets the accuracy level of a given clock is how well the magnetic field shift (and dozens of others) can be corrected for.
The same is true for the mercury clock. The difference is that the systematic frequency shifts that can affect accuracy of the clock are now understood, and controllable, at a higher level of precision.
The first picture in the ZDNet article is not actually of the mercury clock, it's of the strontium atomic clock under development at JILA. JILA is associated with NIST, but they are not even on the same campus. The strontium clock uses a competing technology and is at a much earlier stage of development-- and performance-- when compared to the mercury clock!
Calibration is a process that evaluates the possible sources of frequency shifts, measuring how strong each type of stimulus and response is. For example, how strongly is the transition frequency affected by magnetic fields, and how much magnetic field is there?
It's actually the same procedure that's already used for the best cesium clocks-- there's isn't (or wasn't anyway) anything better to compare those to, and yet they've been making great strides forward for fifty years now.
As for the second question, we're not running a clock for 400 million years-- what counts is that it's that much more accurate in the lab today.
Of course not!
It's an atomic clock, not an atomic watch!
(Actually, the entire atom trap is only about 1 cm across. The problem is the three rooms full of support equipment.)
The 400-million year figure is still limited by technical issues, not fundamental physics. It is expected that once a few more calibration methods are tried out, that it will be able to reach its theoretical limit, which actually does turn out to be pretty close to one second in five billion years.
In any case, these millions-of-years figures are not really practical-- they're just the way that clock people phrase things so that they sound good in the popular press. What really matters is that the precision that can be obtained in a much shorter period of time is much higher. Right now the mercury clock has errors at the level of about a second in 400 million years-- but a second is a lot of timing error! Perhaps a more useful (but equivalent) figure would be 2.3 ns per year, or perhaps you would rather use 44 picoseconds per week.
Slashdot Mirror
I'd do it for only 25 million dollars.
Will someone please tell me which Dept. this article is from? Shouldn't it be from the don't-tailgate-Starbuck Dept. or something?
An example of the problem is this: for technical reasons, a small magnetic field is needed inside a cesium clock. Magnetic fields change the spacing between all atomic energy levels to some degree. For cesium, the relevant change is very small, but it is still there. What you need to do is measure the magnetic field, calculate how much it affects the frequency of the atomic transition, and correct your output frequency by the required amount. What ultimately sets the accuracy level of a given clock is how well the magnetic field shift (and dozens of others) can be corrected for.
The same is true for the mercury clock. The difference is that the systematic frequency shifts that can affect accuracy of the clock are now understood, and controllable, at a higher level of precision.
Looks like Roland Piquepaille failed to RTFA?
It's actually the same procedure that's already used for the best cesium clocks-- there's isn't (or wasn't anyway) anything better to compare those to, and yet they've been making great strides forward for fifty years now.
As for the second question, we're not running a clock for 400 million years-- what counts is that it's that much more accurate in the lab today.
Of course not! It's an atomic clock, not an atomic watch! (Actually, the entire atom trap is only about 1 cm across. The problem is the three rooms full of support equipment.)
The 400-million year figure is still limited by technical issues, not fundamental physics. It is expected that once a few more calibration methods are tried out, that it will be able to reach its theoretical limit, which actually does turn out to be pretty close to one second in five billion years. In any case, these millions-of-years figures are not really practical-- they're just the way that clock people phrase things so that they sound good in the popular press. What really matters is that the precision that can be obtained in a much shorter period of time is much higher. Right now the mercury clock has errors at the level of about a second in 400 million years-- but a second is a lot of timing error! Perhaps a more useful (but equivalent) figure would be 2.3 ns per year, or perhaps you would rather use 44 picoseconds per week.
The clock is based on mercury-199. Yes, it's a stable isotope.
That's Edgar Allan Poe, not "Allen."
Since Neptune and Uranus are about the same size, it looks like the units were chosen precisely to avoid that particular lame joke. =)
So the only question remaining is: Am I supposed to bring my own vegetable oil?
Actually, it's an offshoot of the Beanie Babies line. Here is an example. Collect them all!