Keeping Time with a Mercury Atom

Roland Piquepaille writes "The National Institute of Standards and Technology (NIST) has announced that a new experimental atomic clock based on a single mercury atom is now at least five times more precise than NIST-F1, the U.S. standard clock. This mercury atomic clock 'would neither gain nor lose a second in about 400 million years' while it would take 'only' 70 million years to NIST-F1, based on a 'fountain' of cesium atoms, to gain or lose a second. But even if this new kind of optical atomic clock is more accurate than cesium microwave clocks, it will take a while before such a design can be accepted as an international standard. A ZDNet summary contains pictures and more details about the world's most precise clock."

How much accuracy do you need?

[tylernt]

So... at what point do you say that a clock is accurate enough? I mean, yeah maybe this thing is more accurate than current technology, but if it turns out to be way more expensive, why bother? How often do you need the accuracy that current technology can't provide?

Why?

[mh101]

Can someone explain why we even need this sort of precision?

I Know I'm Missing Something Here...

[stuffman64]

I'm just curious about something here. If a second is defined to be 9,192,631,770 oscillations of a Caesium-133 atom, then why is it said atomic clocks are accurate to within a second over 70 million years? Isn't that lost/gained second itself defined by the Caesium atom's transitions? I hope this question makes sense...

Re:Universal clock?

[pontifier]

http://en.wikipedia.org/wiki/Clock_of_the_Long_Now [wikipedia.org]

Because there are limits to measuring cesium

[Anonymous Coward]

For any definition of a fundamental unit, there are (quantum-mechanical or practical) limitations on how accurately the specified measurement can be made. Thus, there is a small but finite spread in the effective values used by different laboratories. As long as the new standard is within that range, it is "exactly the same" in the sense of being indistinguishable, but nonetheless better, because its measurement uncertainty is smaller.

Whenver the definition is revised, the new proposed standard is compared to the old accepted standard as precisely as anyone has ever done. For example, Louis Essen measured the frequency of the Cesium hyperfine transition as 9,192,631,770 +/- 20 Hz relative to the old tropical year definition. Thus, 9,192,631,770 was picked as the definition.

However, there are quantum mechanical limitations on our ability to measure that. In particular, when we examine the atoms for a time t, there is an uncertainty proportional to 1/t in the frequency. With cesium atoms, which are electrically neutral, gravity poses a problem. There's no way to hold them up without disturbing them, so they fly through cesium beam clocks in a fraction of a second, giving a small uncertainty in the measurement frequency. Suppose this is +/-1 Hz; that then leads to an uncertainty of +/- 1/9192631770 in the duration of a second.

Cesium fountains slow the cesium atoms down as much as possible and thereby extend the measuring time and reduce the uncertainty. However, for any given measuring time, a higher frequency will always lead to a smaller relative uncertainty. The problem is that 9 GHz is accessible to fast electronics. The mercury clock generates a frequency of 1,064,721,609,899,143 Hz (+/-10 Hz as of current measurements) - that's 1.065 Petahertz! There's no electronics that can keep up, so the challenge of building such a clock is measuring its output frequency. Nonetheless, it should be obvious that with a base frequency some 100,000 times higher than the base frequency of a cesium clock, the potential measurement uncertainty is 100,000 times lower.

If the standard second is ever redefined, it will be to a value that is indistinguishable from the old one using any cesium clock ever built.

It's like drawing a line. Suppose you have a line in pencil, and need to know where it is as precisely as possible. After a while the width of the pencil gets annoying, so you sit down with a magnifying glass and measure it as precisely as possible, and draw a line with a super-sharp pen through the middle of the pencil line. But then that's too wide and fuzzy, so you use a microscope and score a line with a diamond-tip probe. But then that's too wide, so you use an atomic-force microscope and push individual atoms around. Then the atoms are too fuzzy, so you cryogenically cool it to reduce their motion. Etc. etc. Each standard is equal to the old one because it's inside the range of uncertainty of measurement.

Re:How accurate is accurate enough?

[rwwh]

Scientific American once had a nice paper about time. I remember these two facts:

• At an accuracy of 10^-17, the earths gravity makes that two identical clocks, one of which is 5cm higher up than the other one, will start deviating from each other (i.e. time really IS different 5 cm up, at this accuracy)
• At an accuracy for 10^-17, relativistic effects start playing a role at walking speeds (i.e. time really IS different at walking speed than at rest, at this accuracy).

I think 5cm and 5km/hour are reasonable usability limits, hence an accuracy of better than 1:10^17 would not make much sense to me.

Re:One small problem...

[oskay]

The clock is based on mercury-199. Yes, it's a stable isotope.

Re:How much accuracy do you need?

[Detritus]

We can always use more accuracy. Many communications systems rely on accurate clocks to keep the transmitters and receivers in synchronization. Frequency stability is also important for communications systems and test and measurement equipment. Any defects in the clock will degrade the performance of the equipment.