[time-nuts] Re: Mitigation of lamp oven and cavity oven temperature-induced frequency variation in rubidium atomic clock

[Attila Kinali]

On Wed, 1 Feb 2023 13:33:41 -0500
Mark Kahrs via time-nuts <time-nuts at lists.febo.com> wrote:

> Just published in Review of Scientific Instruments:
> https://doi.org/10.1063/5.0130524

Some small comments on this paper:

One important topic they do not describe is why the light
intensity shifts when the temperature of the lamp or the
cavity changes. This is due to the change in vapor pressure,
ie the number of atoms that evaporate and stay in the gaseous
phase increases with temperature [1]. 

In the lamp this has obviously the effect of having more atoms
that can emit light. Thus higher temperature means more light.
What they do not mention is that this also comes with a shift
in the spectrum which also has an effect on the gas cell frequency
as wel. The effect of this shift would need to be measured and
quantified as well. See [2] for an example that did this work,
though UniNE, like all research groups that work on state of
the art vapor cell Rb standards use lasers, which make
analysing frequency shifts due to light intensity and light
frequency a lot easier as they can be independently tuned.

Similarly, higher temperatures in the cavity mean more atoms
in the vapor cell, means more absorption, thus less light passing
through. In a lot of papers you will read a phrase like "optically
dense" or similar. Which is saying that the amount of rubidium
vapor is so high and the path through the cell so long, that
the amount of light passing through is limited by the absorption
of the rubidium atoms. Any increase in length does not increase
the SNR of the signal any further because of the increased absorption.
The frequency shift in the vapor cell is mostly due to buffer gas
colision. This shift, depending on the gas mixture, has usually
a parabolic shape. I.e. it is possible to choose the mixture such
that the temperature dependend frequency shift is approximately
zero at the desired temperature. A technique that has been known
for at least 30 years (c.f. [3]), see [4] for a more modern
publication on that topic. I do not know why the authors of this
paper chose not to adjust the gas mixture such that they would
get a df/dT null for the vapor cell at their operating temperature,
respectivelly adjusting the operating temperature that such that
they are at the df/dT null. The authors do not discuss why they
do not implement this standard procedure.

It is not clear either  why they chose to implement a C-field
adjustment using analog means (which are hard to adjust and
require manual testing and adjustment for each device) instead
of a digital frequency adjustment in the synthesis chain (see [5]
for a contemporary example of a synthesis chain that is widely
used by many research groups) that can be calibrated in an automated
fashion and can even have non-linear adjustments using a lookup
table. Also using light intensity as detection mechanism is iffy
at best. Not only is light intensity one of the factors that shift
frequency, the intensity that hits the sensor is modulated by two
independent processes with different causes that have different
frequency shift characteristics. I.e. the compensation is limited
by the correlation of the light intensity vs frequency shift
characteristic at calibration time vs during use in the field.
It would have been much better to use independent temperature
sensors at the lamp and vapor cell oven, then do a principle
component analysis of the shifts due to temperature change
of the two ovens and the light intensity. This way all three
factors could have been independently compensated.

			Attila Kinali


[1] "Rubidium 87 D Line Data" by Daniel Adam Steck, 2021, 
Figure 1 on page 25
https://steck.us/alkalidata/rubidium87numbers.pdf

[2] "Long-Term Stability Analysis Toward <1e−14 Level for a
Highly Compact POP Rb Cell Atomic Clock", by Almat, Gharavipour,
Moreno, Gruet, Affolderbach and Mileti, 2020
https://doi.org/10.1109/TUFFC.2019.2940903
https://www.unine.ch/files/live/sites/ltf/files/shared/Publications/2020/2020_Almat_Long-term_stability_TUFFC_67_207.pdf

[3] "On the light shift and buffer gas shift in passive rubidium
frequency standard", by Vanier, Kunski, Aulin, Savard, Tetu and Cyr, 1982
https://doi.org/10.1109/FREQ.1982.200593

[4] "Short and medium term frequency stability of a laser pumped
"rubidium gas-cell frequency standard for satellite navigation",
by Besedina, Berezovskaya, Gevorkyan, Mileti, and Zholnerov, 2006
https://ieeexplore.ieee.org/document/6230986
Also somewhere in https://www.eftf.org/fileadmin/documents/eftf/documents/Proceedings/proceedingsEFTF2006.pdf

[5] "Simple-design ultra-low phase noise microwave frequency
synthesizers for high-performing Cs and Rb vapor-cell atomic clocks",
by François, Calosso, Abdel Hafizz, and Boudot, 2015
http://dx.doi.org/10.1063/1.4929384


-- 
In science if you know what you are doing you should not be doing it.
In engineering if you do not know what you are doing you should not be doing it.
        -- Richard W. Hamming, The Art of Doing Science and Engineering