88Sr+ ion-clock live stream
Hi all, you may find our live-stream from the lab amusing:
https://www.youtube.com/watch?v=z9VFbs4FogY
The central bright dot is fluorescence at 422nm from laser cooling a single
trapped 88Sr+ ion. The ion emits about 1e7 photons/s at most and we
currently detect about 500 of those in a 20ms detection window (using a
Hamamatsu PMT module).
The bar-chart shows the clock-transition spectrum at 445 THz (674nm). The
X-axis is a drive-frequency to the last AOM that shifts the clock-laser
frequency to coincide with the ion frequency. The frequency around 75.9MHz
should be doubled to get the real optical frequency-shift (double-pass AOM).
We 'shoot' 100 pulses at each clock-laser frequency towards the ion, and
detect how many times we are able to drive the ion into the 'dark'
clock-state. The clock state is long-lived (400ms or so), and detection is
by turning on the cooling and noticing that the ion is dark. In theory the
ion should blink in the camera-view also, but the exposure-time is now long
enough that there is not much visible blinking.
The height of the bars show that we are able to excite the clock-transition
with about 20-30% probability at the moment.
The clock-transition splits into five symmetric Zeeman pairs, four of which
are observed in this scan range. We have two magnetic shields and
compensating electromagnets to reduce the DC magnetic field to 0.4 uT or
so. This splits the outermost components by about 20kHz (10kHz AOM-range in
the figure, doubled). The line-center is around 75.945 MHz on the AOM. This
corresponds to the clock-transition center, 444 779 044 095 486.5 Hz. Maybe
we should make this more obvious, an AOM number like 75 MHz is not so
impressive...
The peaks are now around 500Hz wide. This is about 1e-12 fractional. There
is room for improvement as the 88Sr+ clock-state natural lifetime of 400ms
only limits linewidth to 4Hz or so... Line-center (the middle of all zeeman
peaks) can be determined more precisely than linewidth.
The control system (ARTIQ) is now just repeating the same scan over and
over, takes around 1h per scan (12kHz range with 25Hz resolution IIRC), and
each new spectrum is plotted with a new color.
Enjoy it while it lasts - this is a live stream and anything may happen
(cooling laser unlocks from Rb-cell, clock-laser unlocks from ULE-cavity,
ion disappears, whatever....). Ion storage times have been in the ~4 days
range previously - so I am hoping it will run nice now for 24h or so.
Final clock-operation will not scan this thoroughly over all the peaks,
just three of them, and just one frequency on the left/right side of each
peak - and then a numerical servo locks on to the center of each peak.
If you run an optical clock I hereby challenge you to live-stream it and
post here!
cheerio,
Anders
Cool !!!
On Dec 6, 2019, at 10:39 AM, Anders Wallin anders.e.e.wallin@gmail.com wrote:
Hi all, you may find our live-stream from the lab amusing:
https://www.youtube.com/watch?v=z9VFbs4FogY
The central bright dot is fluorescence at 422nm from laser cooling a single
trapped 88Sr+ ion. The ion emits about 1e7 photons/s at most and we
currently detect about 500 of those in a 20ms detection window (using a
Hamamatsu PMT module).
The bar-chart shows the clock-transition spectrum at 445 THz (674nm). The
X-axis is a drive-frequency to the last AOM that shifts the clock-laser
frequency to coincide with the ion frequency. The frequency around 75.9MHz
should be doubled to get the real optical frequency-shift (double-pass AOM).
We 'shoot' 100 pulses at each clock-laser frequency towards the ion, and
detect how many times we are able to drive the ion into the 'dark'
clock-state. The clock state is long-lived (400ms or so), and detection is
by turning on the cooling and noticing that the ion is dark. In theory the
ion should blink in the camera-view also, but the exposure-time is now long
enough that there is not much visible blinking.
The height of the bars show that we are able to excite the clock-transition
with about 20-30% probability at the moment.
The clock-transition splits into five symmetric Zeeman pairs, four of which
are observed in this scan range. We have two magnetic shields and
compensating electromagnets to reduce the DC magnetic field to 0.4 uT or
so. This splits the outermost components by about 20kHz (10kHz AOM-range in
the figure, doubled). The line-center is around 75.945 MHz on the AOM. This
corresponds to the clock-transition center, 444 779 044 095 486.5 Hz. Maybe
we should make this more obvious, an AOM number like 75 MHz is not so
impressive...
The peaks are now around 500Hz wide. This is about 1e-12 fractional. There
is room for improvement as the 88Sr+ clock-state natural lifetime of 400ms
only limits linewidth to 4Hz or so... Line-center (the middle of all zeeman
peaks) can be determined more precisely than linewidth.
The control system (ARTIQ) is now just repeating the same scan over and
over, takes around 1h per scan (12kHz range with 25Hz resolution IIRC), and
each new spectrum is plotted with a new color.
Enjoy it while it lasts - this is a live stream and anything may happen
(cooling laser unlocks from Rb-cell, clock-laser unlocks from ULE-cavity,
ion disappears, whatever....). Ion storage times have been in the ~4 days
range previously - so I am hoping it will run nice now for 24h or so.
Final clock-operation will not scan this thoroughly over all the peaks,
just three of them, and just one frequency on the left/right side of each
peak - and then a numerical servo locks on to the center of each peak.
If you run an optical clock I hereby challenge you to live-stream it and
post here!
cheerio,
Anders
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.
On 12/6/19 7:39 AM, Anders Wallin wrote:
Hi all, you may find our live-stream from the lab amusing:
https://www.youtube.com/watch?v=z9VFbs4FogY
The central bright dot is fluorescence at 422nm from laser cooling a single
trapped 88Sr+ ion. The ion emits about 1e7 photons/s at most and we
currently detect about 500 of those in a 20ms detection window (using a
Hamamatsu PMT module).
Very cool. A question about the display - it's a video image of the
fluorescence, and the graph is superimposed on top of it?
I've forwarded this off to some friends who are working Cold Atom Lab, a
revised version of which was just shipped up to ISS yesterday. That's
BECs, a bit different, but subject to all the same issues ions, exciting
them, tuning the laser, etc.
Hi Jim, yes you are right the background is a camera-image (about 30x
magnification, maybe 1um per pixel). We use a microscope-objective, a
narrow 422nm bandpass filter, then an image-intensifier, and a fairly
standard CCD camera that looks at the output of the intensifier. When
everything is working and optimized I think a good quality (high QE) CMOS
or CCD camera alone could work. I think it would be interesting to try e.g.
a budget camera made for astrophotography at some point.
If you have lots of money then an EMCCD camera (30-50k maybe) is a good
solution.
The camera is only used for observation and coarse diagnosis of major
problems. For quantitative measurements we have a beam-splitter that
directs most of the fluorescence towards a pinhole followed by a PMT
photon-counter. The ARTIQ controller can time-stamp each incoming photon
with 1ns resolution (but this creates a lot of data) - mostly we just bin
the counts into some gate time and look at the time-series of counts per
20ms or so.
The live stream is created with OBS. It captures the camera image as
background and can overlay any images/websites etc. that we can imagine.
I create the Zeeman-spectrum barchart in matplotlib as a PNG image with
transparent background - once per minute, from data we store in InfluxDB.
The X-axis labeling was not so good - need to improve that next time.
Somehow indicate that line-center is 445THz and the peaks really are <1e-12
wide (narrow!).
For the impatient, I made a speedup (128x) version of the live-stream:
https://www.youtube.com/watch?v=f3edhdwqXgc
Anders
On Sat, Dec 7, 2019 at 1:48 AM jimlux jimlux@earthlink.net wrote:
The central bright dot is fluorescence at 422nm from laser cooling a
single
trapped 88Sr+ ion. The ion emits about 1e7 photons/s at most and we
currently detect about 500 of those in a 20ms detection window (using a
Hamamatsu PMT module).
Very cool. A question about the display - it's a video image of the
fluorescence, and the graph is superimposed on top of it?
Hello to the group. This is great to watch and see the development of an
exotic clock. I guess I need a bit of help in understanding what I am
seeing. To a point I understand Zeeman images in respect to a cesium
reference at least. But in that case there is a single higher peak thats
the correct one to lock to. In the image thats shared is this a case of
saying the multiple energy level can be seen but no attempt is being made
to choose one peak to work with?
Like the speed up version a bit to impatient.
Thank you for sharing with time-nuts Anders
Paul
WB8TSL
On Sat, Dec 7, 2019 at 7:39 AM Anders Wallin anders.e.e.wallin@gmail.com
wrote:
Hi Jim, yes you are right the background is a camera-image (about 30x
magnification, maybe 1um per pixel). We use a microscope-objective, a
narrow 422nm bandpass filter, then an image-intensifier, and a fairly
standard CCD camera that looks at the output of the intensifier. When
everything is working and optimized I think a good quality (high QE) CMOS
or CCD camera alone could work. I think it would be interesting to try e.g.
a budget camera made for astrophotography at some point.
If you have lots of money then an EMCCD camera (30-50k maybe) is a good
solution.
The camera is only used for observation and coarse diagnosis of major
problems. For quantitative measurements we have a beam-splitter that
directs most of the fluorescence towards a pinhole followed by a PMT
photon-counter. The ARTIQ controller can time-stamp each incoming photon
with 1ns resolution (but this creates a lot of data) - mostly we just bin
the counts into some gate time and look at the time-series of counts per
20ms or so.
The live stream is created with OBS. It captures the camera image as
background and can overlay any images/websites etc. that we can imagine.
I create the Zeeman-spectrum barchart in matplotlib as a PNG image with
transparent background - once per minute, from data we store in InfluxDB.
The X-axis labeling was not so good - need to improve that next time.
Somehow indicate that line-center is 445THz and the peaks really are <1e-12
wide (narrow!).
For the impatient, I made a speedup (128x) version of the live-stream:
https://www.youtube.com/watch?v=f3edhdwqXgc
Anders
On Sat, Dec 7, 2019 at 1:48 AM jimlux jimlux@earthlink.net wrote:
The central bright dot is fluorescence at 422nm from laser cooling a
single
trapped 88Sr+ ion. The ion emits about 1e7 photons/s at most and we
currently detect about 500 of those in a 20ms detection window (using a
Hamamatsu PMT module).
Very cool. A question about the display - it's a video image of the
fluorescence, and the graph is superimposed on top of it?
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe, go to
http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.
In Cs my understanding is that a transition mF=0 to mF=0 transition is used
- so it is insensitive to the magnetic field[1].
There is no magnetic field insensitive component of the 88Sr+ clock
transition (other optical clocks will vary!)
See e.g. around page 10 here for an energy-level diagram:
http://resource.npl.co.uk/docs/networks/time/meeting3/klein.pdf
To measure 88Sr+ line-center (where there is no peak at nonzero B-field!)
the mid-point between a Zeeman pair is a good approximation, but one gets
rid of the electric quadrupole shift by measuring the center of three pairs
of components and calculating line-center from that. The servo-loop will
thus need to probe the left and right side of multiple peaks in sequence.
Our pulse-sequence now does 100 probe-pulses in about 7 seconds. If we
probe left/right side of three pars (twelve frequencies in total) the
line-center can be computed about once per minute. The ultra-stable clock
laser acting as local oscillator needs to maintain stability on its own
during those 1-2 minutes.
The BIPM SRS document has more references
https://www.bipm.org/utils/common/pdf/mep/88Sr+_445THz_2017.pdf
Anders
On Sat, Dec 7, 2019 at 10:04 PM paul swed paulswedb@gmail.com wrote:
Hello to the group. This is great to watch and see the development of an
exotic clock. I guess I need a bit of help in understanding what I am
seeing. To a point I understand Zeeman images in respect to a cesium
reference at least. But in that case there is a single higher peak thats
the correct one to lock to. In the image thats shared is this a case of
saying the multiple energy level can be seen but no attempt is being made
to choose one peak to work with?
Like the speed up version a bit to impatient.
Thank you for sharing with time-nuts Anders
Paul
WB8TSL
On Sat, Dec 7, 2019 at 7:39 AM Anders Wallin anders.e.e.wallin@gmail.com
wrote:
Hi Jim, yes you are right the background is a camera-image (about 30x
magnification, maybe 1um per pixel). We use a microscope-objective, a
narrow 422nm bandpass filter, then an image-intensifier, and a fairly
standard CCD camera that looks at the output of the intensifier. When
everything is working and optimized I think a good quality (high QE) CMOS
or CCD camera alone could work. I think it would be interesting to try
e.g.
a budget camera made for astrophotography at some point.
If you have lots of money then an EMCCD camera (30-50k maybe) is a good
solution.
The camera is only used for observation and coarse diagnosis of major
problems. For quantitative measurements we have a beam-splitter that
directs most of the fluorescence towards a pinhole followed by a PMT
photon-counter. The ARTIQ controller can time-stamp each incoming photon
with 1ns resolution (but this creates a lot of data) - mostly we just bin
the counts into some gate time and look at the time-series of counts per
20ms or so.
The live stream is created with OBS. It captures the camera image as
background and can overlay any images/websites etc. that we can imagine.
I create the Zeeman-spectrum barchart in matplotlib as a PNG image with
transparent background - once per minute, from data we store in InfluxDB.
The X-axis labeling was not so good - need to improve that next time.
Somehow indicate that line-center is 445THz and the peaks really are
<1e-12
wide (narrow!).
For the impatient, I made a speedup (128x) version of the live-stream:
https://www.youtube.com/watch?v=f3edhdwqXgc
Anders
On Sat, Dec 7, 2019 at 1:48 AM jimlux jimlux@earthlink.net wrote:
The central bright dot is fluorescence at 422nm from laser cooling a
single
trapped 88Sr+ ion. The ion emits about 1e7 photons/s at most and we
currently detect about 500 of those in a 20ms detection window
(using a
Hamamatsu PMT module).
Very cool. A question about the display - it's a video image of the
fluorescence, and the graph is superimposed on top of it?
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe, go to
http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe, go to
http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.
On Sat, Dec 7, 2019 at 1:29 PM Anders Wallin anders.e.e.wallin@gmail.com
wrote:
In Cs my understanding is that a transition mF=0 to mF=0 transition is used
- so it is insensitive to the magnetic field[1].
There is no magnetic field insensitive component of the 88Sr+ clock
transition (other optical clocks will vary!)
See e.g. around page 10 here for an energy-level diagram:
http://resource.npl.co.uk/docs/networks/time/meeting3/klein.pdf
To measure 88Sr+ line-center (where there is no peak at nonzero B-field!)
the mid-point between a Zeeman pair is a good approximation, but one gets
rid of the electric quadrupole shift by measuring the center of three pairs
of components and calculating line-center from that. The servo-loop will
thus need to probe the left and right side of multiple peaks in sequence.
Our pulse-sequence now does 100 probe-pulses in about 7 seconds. If we
probe left/right side of three pars (twelve frequencies in total) the
line-center can be computed about once per minute. The ultra-stable clock
laser acting as local oscillator needs to maintain stability on its own
during those 1-2 minutes.
How do you then divide down the optical frequency and shim with the
computed offset?
The servo-loop will
thus need to probe the left and right side of multiple peaks in sequence.
Our pulse-sequence now does 100 probe-pulses in about 7 seconds. If we
probe left/right side of three pars (twelve frequencies in total) the
line-center can be computed about once per minute. The ultra-stable clock
laser acting as local oscillator needs to maintain stability on its own
during those 1-2 minutes.
How do you then divide down the optical frequency and shim with the
computed offset?
In our current setup we lock the frequency comb to a hydrogen maser (100MHz
output), and the comb measures the clock-laser light before the final
"zeeman-AOM" that does the hopping to find the peaks.
We don't produce a physical signal with the line-center frequency, but if
we want to compare the maser to the ion this isn't required. Instead we
take the frequency-comb measurement and add the computed line-center
measurement - this gives us a comparison between the maser and the ion. The
maser in turn is linked to UTC(k), which is linked through GNSS/PPP to UTC.
There are methods to lock a frequency-comb to the ultra-stable clock laser
and get as output microwaves (0.1 to 1 GHz) that are divided down (by 2e6
or so) from the optical carrier - we are not doing this at the moment. For
the ion vs maser comparison this second method would not significantly
improve things (I think).
Anders
Hi,
On 2019-12-07 21:46, Anders Wallin wrote:
In Cs my understanding is that a transition mF=0 to mF=0 transition is used
- so it is insensitive to the magnetic field[1].
There is no magnetic field insensitive component of the 88Sr+ clock
transition (other optical clocks will vary!)
See e.g. around page 10 here for an energy-level diagram:
http://resource.npl.co.uk/docs/networks/time/meeting3/klein.pdf
Let me make a somewhat more accurate model.
The mf values of +3, +2, +1, -1, -2, -3 transitions have a relatively
strong sensitivity to magnetic field, with a strong linear term on the
magnetic field strength, where as the 0 transition has a much weaker
quadratic sensitivity, assuming weak magnetic field which is fair
assumption. This is true for any atomic transition, so it is not unique
to Cesium. Cesium has however the second weakest (of classical neutral
atom microwave frequency, only Thallium being better) magnetic
sensitivity for it's hyper-fine transition. Even with cesium, you can
tweak the frequency of transition with the change of the magnetic field
(B-field in the lingo). All such cesium clocks is really secondary
standards, even if marketing have been boasting their contribution a lot.
Later the +1 and -1 transitions with their much stronger dependence on
the magnetic field, as being measured can be used to servo the magnetic
field to lock the +1 to -1 difference and with that the magnetic field
is stabilized and the offset caused by the magnetic field on the 0
transition used for clock steering and then it starts to approach real
primary clock behavior.
So, insensitive is overstating it, more much less sensitivity than the
other transitions, but the details of them is very useful. At some time
I should re-read it to learn these things deeper.
To measure 88Sr+ line-center (where there is no peak at nonzero B-field!)
the mid-point between a Zeeman pair is a good approximation, but one gets
rid of the electric quadrupole shift by measuring the center of three pairs
of components and calculating line-center from that. The servo-loop will
thus need to probe the left and right side of multiple peaks in sequence.
Our pulse-sequence now does 100 probe-pulses in about 7 seconds. If we
probe left/right side of three pars (twelve frequencies in total) the
line-center can be computed about once per minute. The ultra-stable clock
laser acting as local oscillator needs to maintain stability on its own
during those 1-2 minutes.
The BIPM SRS document has more references
https://www.bipm.org/utils/common/pdf/mep/88Sr+_445THz_2017.pdf
Sounds like lots of fun and would be cool to see in reality. Thanks for
the links.
If one had time and resources... and spare cubic meters in the lab.
Cheers,
Magnus
Magnus,
The mf values of +3, +2, +1, -1, -2, -3 transitions have a relatively
strong sensitivity to magnetic field, with a strong linear term on the
magnetic field strength, where as the 0 transition has a much weaker
quadratic sensitivity, assuming weak magnetic field which is fair
assumption.
You can see this dramatically by turning the cfield adjustment from min
to max:
This is true for any atomic transition, so it is not unique
to Cesium. Cesium has however the second weakest (of classical neutral
atom microwave frequency, only Thallium being better) magnetic
sensitivity for it's hyper-fine transition.
Some late night reading about Thallium:
1957, Kusch
"Precision Atomic Beam Techniques"
https://ieeexplore.ieee.org/document/1536252
kusch1957.pdf
1960, Mockler, Beehler, Snider
"Atomic Beam Frequency Standards"
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.846.7115&rep=rep1&type=pdf
1963, Beehler, Glaze
"Experimental Evaluation of a Thallium Beam Frequency Standard"
https://tf.nist.gov/general/pdf/211.pdf
1966, Beehler, Glaze
"Evaluation of a Thallium Atomic Beam Frequency Standard at the National
Bureau of Standards"
https://tf.nist.gov/general/pdf/9.pdf
1983, Ramsey
"History of Atomic Clocks"
https://tf.nist.gov/general/pdf/1916.pdf
https://nvlpubs.nist.gov/nistpubs/jres/088/jresv88n5p301_A1b.pdf
Even with cesium, you can
tweak the frequency of transition with the change of the magnetic field
(B-field in the lingo). All such cesium clocks is really secondary
standards, even if marketing have been boasting their contribution a lot.
Not sure what make/model cesium clock you're talking about here. Just
because there are knobs to tweak doesn't demote them to secondary.
/tvb
On 12/8/2019 1:45 PM, Magnus Danielson wrote:
Hi,
On 2019-12-07 21:46, Anders Wallin wrote:
In Cs my understanding is that a transition mF=0 to mF=0 transition is used
- so it is insensitive to the magnetic field[1].
There is no magnetic field insensitive component of the 88Sr+ clock
transition (other optical clocks will vary!)
See e.g. around page 10 here for an energy-level diagram:
http://resource.npl.co.uk/docs/networks/time/meeting3/klein.pdf
Let me make a somewhat more accurate model.
The mf values of +3, +2, +1, -1, -2, -3 transitions have a relatively
strong sensitivity to magnetic field, with a strong linear term on the
magnetic field strength, where as the 0 transition has a much weaker
quadratic sensitivity, assuming weak magnetic field which is fair
assumption. This is true for any atomic transition, so it is not unique
to Cesium. Cesium has however the second weakest (of classical neutral
atom microwave frequency, only Thallium being better) magnetic
sensitivity for it's hyper-fine transition. Even with cesium, you can
tweak the frequency of transition with the change of the magnetic field
(B-field in the lingo). All such cesium clocks is really secondary
standards, even if marketing have been boasting their contribution a lot.
Later the +1 and -1 transitions with their much stronger dependence on
the magnetic field, as being measured can be used to servo the magnetic
field to lock the +1 to -1 difference and with that the magnetic field
is stabilized and the offset caused by the magnetic field on the 0
transition used for clock steering and then it starts to approach real
primary clock behavior.
So, insensitive is overstating it, more much less sensitivity than the
other transitions, but the details of them is very useful. At some time
I should re-read it to learn these things deeper.
To measure 88Sr+ line-center (where there is no peak at nonzero B-field!)
the mid-point between a Zeeman pair is a good approximation, but one gets
rid of the electric quadrupole shift by measuring the center of three pairs
of components and calculating line-center from that. The servo-loop will
thus need to probe the left and right side of multiple peaks in sequence.
Our pulse-sequence now does 100 probe-pulses in about 7 seconds. If we
probe left/right side of three pars (twelve frequencies in total) the
line-center can be computed about once per minute. The ultra-stable clock
laser acting as local oscillator needs to maintain stability on its own
during those 1-2 minutes.
The BIPM SRS document has more references
https://www.bipm.org/utils/common/pdf/mep/88Sr+_445THz_2017.pdf
Sounds like lots of fun and would be cool to see in reality. Thanks for
the links.
If one had time and resources... and spare cubic meters in the lab.
Cheers,
Magnus
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.