[time-nuts] Re: leap seconds finally being retired?

[Magnus Danielson]

Hi Chris,

On 2022-11-22 19:21, Chris Caudle via time-nuts wrote:
> Did I miss some messages?  I did not see any discussion of the vote to
> retire leap seconds which took place last Friday.  I just saw this in my
> news headlines today:
> https://gizmodo.com/leap-second-time-1849807606
>
> "A global group of scientists and government officials voted (almost
> unanimously) Friday to axe the small time adjustment method, in a change
> scheduled to take place by 2035."
>
> Official resolution (resolution 4):
> https://www.bipm.org/documents/20126/64811223/Resolutions-2022.pdf/281f3160-fc56-3e63-dbf7-77b76500990f
>
> Also resolution 5 sets out a timeline for redefining the second with
> something other than cesium (would like to have the choice for the
> preferred atom species decided in 2026, and vote on the new standard
> definition in 2030).
Do notice that it acknowledge that this could either be a single spieces 
or an ensemble definition.
>
> I thought Magnus wrote previously that he was going to be at that meeting
> last week, but perhaps I misunderstood and he was only providing input,
> not physically attending.
>
No, at the best I could be there as a visitor, and that would be great 
fun for sure. In those meetings I will have no official say in anything.

I have however provided input to the work of CCTF in proposing a palette 
of different proposals, including an ensamble definition for a future SI 
second. The ensemble definition has an interesting elegance to it, but 
breaks the comfort zone.

When the can of worm was opppend in 2015 of a redefinition of the SI 
second in terms of an optical transition, it was already clear that the 
clocks where heading towards 100 times better than the cesium fountains. 
The line of the stability improved very quickly at that time, and even 
if the slope changed there is still improvements occurring.

What was also already known was that a handful of optical transitions 
was occurring, such as Aluminium, Strontium etc. At the same time, what 
was apparent was that multiple physical package approaches was being 
used. Both neutral atoms and ions was used. Laser cooling with various 
methods was applied. Single atom, lattice, etc. State-squeezing also 
applied. We see super-Ramsey interrigation being used to compensate 
first degree doppler. There is a richness of methods and tricks that is 
applied to further improve the performance. They all end up in a similar 
range of performance.

Now, this is in contrast with the times when Cesium was selected. 
According to the criterions at the time, Thallium 205 was actually 
considered, but in the only physical realisation of atomic beam device, 
Thallium 205 had one performance edge over all other, in the magnetic 
shift sensitivity is 20 times better than the next competitor of Cesium 
133. However, Thallium had two techical drawbacks at the time making 
physical implementation difficult, the transition frequency of 
21.3108339461 GHz was at the time considerably harder to achieve than 
the Cesium 9.192631770 GHz. Secondly the ionization was harder to do in 
the masspectrometer needed for detection, resulting in weaker signal, 
worsening the signal to noise ratio and predicted lower stability 
properties.

Notice how the performance of a single type of physical package shaped 
the field. Today we have removed ionization as measure for state-flip in 
the Ramsey interigation, we can do that using lasers. Optical readout 
was done at NIST-7, and the continuation of that is in the form of 
todays fountain clock realizations that also address some of the other 
issues of the original beam clocks. Observation time has been 
considerably improved in the folded beam of the fountains. Some effects 
is first-degree compensated for but subtle micro-lensing effects occurs 
and is compensated for. Also, the sensitivity to magnetic field has been 
removed as effect. Through observation of the splitting of fields, the 
magnetic field can be measured and even stabilized or compensated, to 
create a stable and well known shift. This have even made it into the 
commercial clocks such as FTS-4040, FTS-4065, HP5071A and others.

For some time, it was even thought that Rubidium was better suited for 
fountain operation than Cesium, but learning to control the issues the 
shift can be estimated and compensated for and they are more or less on 
par after that.

So, what we can see is that the realization form now varies greatly and 
that then in combination with the spieces (atom and its particular 
isotope) it becomes a race to combine all the tricks and perfecting it. 
The exact frequency in the optical range varies a little, and higher 
should be better, but other factors also combine to a rather complex 
mixture. It's not a very clear race of speices, it's a much more 
multi-dimensional field.

So, history teaches us that our selection criterions may be completely 
wrong down the line, and that development can take interesting turns 
that we do not fully can forsee.

Further, there is additional concerns here. One such concern is one of 
both political and practical aspect. If we come up with a definition of 
a single spieces that for it's realization turns out to be hard to 
realize or duplicate, with the needed repeatability (our ability to 
build multiple giving the same measure), we can end up in a new kg 
reference unit situation, in which only one lab (or possibly very few) 
have access to the actual reference. This creats both practical issues, 
you can say it is a problem of democratic access even, and there can 
become political conotations to that which we really want to avoid.

Another aspect is that you can end up locking yourself to a definition 
that actually turns out to be a particular bad choice, that you can have 
trouble to develop further, and then you end up having to redefine later.

In metrology, there is a concept of primary reference and secondary 
reference which is often misunderstood or even missused. The primary 
reference is a definition you can realize with a repeatability that is 
well understood and require no corrections. Secondary references may 
perform well, but need calibration towards the primary to correct for 
it. Rubidum have had a secondary reference relationship to cesium, 
despite that for compareable realizations they perform similar enough, 
not withstanding the normal incorrecty distinction of cesium vs. 
rubidium when they really differen in beam device vs. gas cell device. 
So, we now have multiple ways of realizing an optical transition 
standard for multiple spieces, which all is potential primary references.

One way of approaching this is to take a pick of a spieces and define 
that, but then keep maintaing the comparison measurments that is done 
between spieces and define and refine relationships to these other 
spieces and let them be very well known secondary standards. That could 
work well. However, we now have the technological bet here. If we now 
choose the wrong spieces, it could turn out that we have trouble to 
develop it's performance at the same speed as the other. Then it wasn't 
as good candiate for primary standard as we thought, and we bet on the 
wrong horse.

However, look at the EAL/TAI/UTC realization of SI second and 
coordinated time-scales. We've learend to ensamble dissimilar spieces 
and clock types to provide a really good realisation and overall 
performance. We already do a range of relative clock comparisions 
between the optical clocks that currently beat any measurement back 
towards cesium with two digits since 1E-16 and 1E-18 has that factor of 
100 in it. Actually, the limitation in realization of cesium transition 
prohibit us of knowing two digits for any of our potential definitions, 
we have essentially random numbers there. As we do the redefinition we 
will end up guessing those and it is likely that we will not be able to 
ever measure and estabilsh the actual offset we create. So, now comes 
the mental leap of consider the actual definition to be an ensemble of a 
number of species, in which the relative comparison measurments is 
continued and the definition then keeps changing with known 
relationships. As our technology develop, we can harvest the benefit of 
that and the ensemble update will slowly shift in smaller and smaller 
fractions at the edge of our continued development of technology. If a 
spieces turns out to be a limit, its contribution will become lower, if 
another spieces turns out particularly useful it's contribution will 
become stronger, all as consequence of the ensemble balance.

It should also be stated that the constant of nature approach that has 
been used succsessfully in the redefinition of the other SI base units 
does not work for frequency, because we currently lack the sufficient 
clarity in theoretical models that come even close to match the 
precision we can realize in our labs. Therefore that option was dropped.

So, when Fritz Riele of PTB presented the concept of redefinitions and 
asked openly for any input on how to approach this, then I presented 
this view in November 2015. He was friendly enough to also point out 
that his article do provide a reference to my email to him.

Having an ensamble definition is however not in most physics people 
comfort zone. 7 years later, it has however not been written of, because 
it does have interesting merits going forward.

While my own contribution is naturally something I enjoy pointing out, I 
think it is really a very interesting field and that we do need many 
solutions that we can rule out in order to learn deeper what the many 
aspects of concerns is. As I have been tracking the issue since, much of 
the analysis still holds. I do not see any of the spieces having a 
significant benefit in performance or realizations to others. As 
development continues, the research groups learn of each other and keep 
reiterate their realizations to squeeze more performance, easier to 
maintain, better availability (PTB reported an optical clock that was 
over a two-week time had 99.5% availability of observing the optical 
transition, and they did not see any major obsticle maintaining that for 
periods of years).

So, while this post turns out in the long form, I think it is a very 
interesting little corner of problems, and it covers more than just pure 
physics or skill of making physical devices, is a whole range of issues 
that needs to be considered as we move forward. We need to learn from 
history and look forward.

BTW, this year it is 50 years since the speed of light was last measured 
at NIST. At the time, it was a major achievement. I have suggested that 
the topic should be revisited, so we could with our advancement in 
technology measure it again, and to see how much we got it wrong back 
then. I reminded NIST and it was brought up with relevant BIPM and UFFC 
parties, but I have not seen any outcome. With some difficultie we could 
setup a Krypton-86 source and measure it with todays technology, and 
with difficultie means also relating it to the conditions as stated at 
the time.

So, no, I was not down there to vote. I have nothing to officially say. 
Hopefully I did provide some inspiration, but I doubt anyone will trace 
it back to me and give me a beer for it. It was a fun mental exercise 
thought, and I want to share that aspect with you.

Cheers,
Magnus