This technology enables frequency counters for laser light.[1] Input a red laser, and you can directly measure its frequency in Hertz with 14 digits of precision.
I can't tell from the spec sheet at https://www.thorlabs.com/_sd.cfm?fileName=MENLO_SMART-COMB-S... whether you have to configure the unit at build time to measure lasers near a given wavelength, or whether it really will measure any arbitrary laser you feed to it.
Under "Measurement Wavelength" it says "Choose one in the 630-2000 nm range," and often that type of specification means you have to specify the nominal frequency up front when you order it. Either way, it seems to be a big step forward in commercializing this stuff. The comb hardware I've seen takes up a good chunk of an entire room.
Once optical comb sources become economically available off the shelf, they should be game-changers in several fields from spectroscopy to time/frequency work. Right now everything you can buy is still around 6 figures AFAIK.
Does calibration of an optical frequency comb require that the speed of light be known, either directly or indirectly? By indirectly I mean the case where something you need to use to calibrate your comb depends, directly or indirectly, on knowing the speed of light.
I'm curious because of something the professor did early on in the introduction to optics class I took in college. He picked up a metal ruler and said we were going to measure the speed of light. Everyone laughed (which was fine because he intended it as a joke).
He then set the ruler on a table, directed a laser to reflect at a shallow angle off the ruler onto the blackboard.
The ruler's lines were raised which made it act like a diffraction grating and there was a visible interference pattern on the blackboard.
He then traced the pattern on the blackboard with chalk, turned off the laser, and used the ruler to (1) measure the distance from where it had been to the blackboard, and (2) the spacing of the lines in the diffraction pattern.
From this and the known frequency of the laser and the known spacing of the lines on the ruler the speed of light is an easy calculation.
This was meant as a joke because usually the frequency of light is calculated using methods that depend on knowing the speed of light, so all that was really happening was the he used a rule to very that the frequency calculation of the laser had been done correctly.
But if you could accurately get the frequency without that depending on knowing the speed of light then you could actually measure the speed of light with a ruler.
Frequency combs are optical synthesizers. This means that the frequency of the nth comb mode is exactly n times a radio frequency (+ another radio frequency), where n is a (very large) integer. The speed of light does not matter at all. This is very important since the speed of light (and hence the wavelength) depends on environmental factors such as air pressure and humidity. One way to determine the speed of light (in air) would indeed be to measure the wavelength of a laser whose frequency is calibrated using a frequency comb.
The speed of light in vacuum cannot be measured in SI units since it is the fundamental constants that defines the unit "meter".
I worked closely with some of these devices at a previous employer. Might've even fried one at some point. I'm pretty excited about what other things they can be used for!
Worked on atomic clocks, the frequency comb was used to downsample optical frequency ranges to a range that electronics can reasonably handle. A patent for the system can be found here: https://patents.google.com/patent/US20220390902A1/en
Not sure if it was me or someone else. There are a lot of things that can go wrong, but the thing that would've most likely been my fault is running the device without the seed laser on (or turning off the seed laser while operating).
> seamlessly connected to optical waves that oscillate at 10,000 times higher frequencies
Somehow four orders of magnitude sound too less for the transition from radio to light, but it makes sense. A i9 processor works at ~6 GHz, and light is at the THz range
Microwave communication goes comfortably into the tens of gigahertz range, and visible light is in the hundreds of terahertz. So it is about a factor of 10,000.
"Effective" compared to the state of the art in the bands on either side: in RF/microwave we have very fast arbitrary waveform generators, very nice amplifiers, and well-characterized conductors, and on the optical side we have lasers, lights, fibers, and more. The terahertz gap is so named because it's too high in frequency for our usual RF devices to work well, and too low in frequency for our usual optical devices to work well; terahertz work ends up being a mix of both, taking from either column as needed for a specific application. (You might hear the word "quasi-optical" used in this sense, though I've never heard the dual word "quasi-microwave"!)
We do have terahertz devices - they're just very limited compared to devices in adjacent bands, usually stated in terms of power. But there are a lot of hardworking and talented people working on narrowing the THz gap from all sides. It's a very very hot research area at the moment.
Comparing CPU frequency (tens of thousands individual lanes that have to be skew-matched) to single-path stuff is quite apples to oranges.
Processor frequency is mostly limited by interconnect and cooling (and silicon, but there are other materials available, it's just not worth switching to those unless you solve cooling). Even thunderbolt transceivers go 40GT/lane=20GHz and that's consumer tech. With InP you can go slightly above 100GHz afaik.
There is also another type of comb that is extremely useful: the GKP qubit (introduced in https://arxiv.org/abs/quant-ph/0008040). Its wavefunction is a comb.
I had the privilege to attend a talk by Jun Ye, one of Hall's previous advisees, a few months back about frequency combs. I really felt honored meeting the person who is so tapped into the work being done at JILA. Lots of amazing, mind bending work.
What is the theoretical efficiency of these devices?
Could it be used for example to combine multiple different frequencies of light into one higher frequency to excite a solar cell at exactly the bandgap energy so no energy is wasted?
Perhaps I'm being pedantic, but in the video, when they show multi-spectral light coming in from the left, they show low-frequency light moving faster than high-frequency light. ("Survey says: EEEEEHHHHHNNNNNNK!")
I was also hopeful the video would have actual info on how they work, but no such luck. Just a lot of "Are they cool, or what?".
Don't be too hard on the video or article. I just went through the Frequency_comb Wikipedia article and I'm still none the wiser.
Well, I did get an idea about what the thing actually is: basically a signal consisting of a mixture of frequencies, precisely spaced. Techniques to generate some of the bands include nonlinear mixing. Turns out, light can undergo distortion, so you can get intermodulation distortion to generate colors not present in the inputs.
The unclear part is the details of how the frequency comb is hooked together with the radio frequency domain in a feedback loop to control the comb. I.e. where in the RF domain we have the precise frequency reference we'd like to convey to the optical domain.
The unclear part is the details of how the frequency comb is hooked together with the radio frequency domain in a feedback loop to control the comb. I.e. where in the RF domain we have the precise frequency reference we'd like to convey to the optical domain.
As I understand it, two effects are involved. One is the laser's pulse repetition rate that determines the frequency spacing between adjacent comb lines. This is on the order of hundreds of MHz, so it can be measured with a photodiode detector and phase-locked like any other RF signal.
The other effect is the carrier (light) phase shift that occurs from one pulse to the next. Assuming the pulse rate has been stabilized, nulling out this carrier phase shift is equivalent to stabilizing the laser's frequency. The photodiode can't see the carrier cycles, of course, but if the comb spans at least one octave in frequency, there will be a detectable beatnote between the second harmonic of the fundamental F (which like you say is always present to some extent given various nonlinearities in any real-world system) and the comb line at the beginning of the next octave. Driving this difference frequency to zero stabilizes the actual lightwave carrier.
As far as stabilizing the signal from the photodiode is concerned, that's just a matter of mixing it with a signal from the desired frequency standard to get the difference frequency that you steer to zero by tuning the laser. Some systems care about locking at a specific phase, others are OK with just getting the frequency right.
Disclaimer: treat the above with healthy skepticism, as IANAPhysicist and have never actually had my hands on this sort of hardware. Corrections actively solicited.
(Edit: Actually I like o1-pro's explanation better than mine: https://i.imgur.com/L3b7S8v.png -- although the same disclaimer obviously applies.)
How the pulse rate of the laser determines the frequency between comb lines is unclear. Frequencies are in hundreds of THz; pulsing is way, way slower.
Plus don't you need the fundamental frequency of the comb to follow the radio-frequency references, not just the spacing.
What is clear is that this may indeed be beyond an optical comb appreciation video produced by the NIST for the general public.
Regarding the spacing of the comb lines that's simply a fundamental property of the time and frequency domains. If you take the Fourier transform of a pulse train you will have comb of lines in the frequency domain where the spacing is the inverse of the pulse period. This depends on the size of the optical cavity. The carrier frequency oscillates underneath (at >100 THz frequency as you correctly point out). You can stabilise the offset frequency i.e. the position of the first line, by f-2f sabilization as described by the other poster. We still need to relate this to an absolute frequency and we do that using accurately known gas absorption lines. That way we have a comb of frequency lines at > 100 THz and more than an octave wide, where we know the absolute frequency with very high accuracy.
The fundamental lightwave carrier frequency F is locked by means of the F-2F beatnote. But if I understand it correctly, they normally stabilize the PRF first (which, again, is in the hundreds of MHz). Otherwise, the frequency separation between the second harmonic of F and the comb tooth at 2F will be noisier/driftier.
The business about the PRF determining the comb spacing comes straight out of Fourier. A train of narrow pulses in the time domain is a series of comb lines in the frequency domain, with spacing equal to the PRF. Lots of applications in traditional RF work for this, but femtosecond lasers made it relevant in the optical field as well.
To measure the absolute frequency of the light, as mentioned by cycomanic, one or more well-known quantum transitions is certain to be within range of any octave-bandwidth laser comb. Some of those transitions have line widths in the hundreds of hertz, which gives serious levels of precision if you're working with a THz or PHz carrier. Then you say goodbye to Mr. Fourier and hello to Mssrs Zeeman and Stark.
OK. It seems that the key is the femtosecond capability in the lasers. Even if the pulses are relatively far apart compared to the target frequencies for the comb, if those pulses are super crisp, that will drive energy into that frequency band?
Right. The risetime of the pulses determines how far out the useful comb lines extend (at least that's how it works in the RF world).
Being able to go from IR to UV in one comb spectrum was considered a very worthwhile advance when it happened. Now you can apparently buy a single box that does it, such as the one mentioned in https://news.ycombinator.com/item?id=42890578 . But you have to ask for the price, and usually I find that if I have to ask, I needn't have bothered. :-P
A frequency comb is a pulsed sinusoidal oscillator that satisfies the requirement that the ratio between the repetition frequency of the pulses and the frequency of the sinusoidal signal must be a fixed known value.
The frequency of the pulses must be low enough so that they can be counted with digital counters, while the frequency of the sinusoidal signal may be as high as to reach the ultraviolet light range.
The fixed frequency ratio is achieved by a control loop with feedback, which works in a similar way, but more complex, with a PLL (phase-locked loop).
The control loop of a PLL controls a single variable. For example, a PLL can be made with a VCO (voltage-controlled oscillator), where a voltage determines the frequency of the oscillator, together with a device that provides means to detect whether the ratio between an input frequency and an output frequency deviates from the desired fixed ratio. If a deviation is detected, the voltage at the input of the VCO is increased or decreased, restoring the ratio between the input frequency and the output frequency to the desired value.
Modern PLLs normally use digital frequency dividers in the control loop, which limits their maximum frequency to one that can be counted with digital circuits. Nevertheless, the first PLLs have been implemented many years before the appearance of the first digital counters. Such early PLLs were used for frequency division, not for frequency multiplication, like the modern PLLs. They used inside their control loop a circuit that distorted the sinusoidal output of their controlled oscillator, generating high-order harmonics. A high-order harmonic was selected with a filter and it was compared with the input frequency, detecting any deviation of the PLL from the intended frequency ratio.
The optical frequency combs can be considered as an evolution of the early PLLs that were used for frequency division, and they are necessary for the same reason as the early PLLs. By the time of the first PLLs, pulses could be counted only with mechanical counters, which could not work at the frequencies of quartz crystal oscillators. So a frequency-dividing PLL was used to bridge the gap between what mechanical counters could count and the frequency of the sinusoidal signal produced by a quartz oscillator, exactly like now optical frequency combs bridge the gap between what electronic digital counters can count and the frequencies of optical oscillators.
The main difference between an optical frequency comb and the early PLLs with harmonic generators is that the control loop of a frequency comb is much more complex, because it must control simultaneously 2 distinct output variables in order to keep the frequency ratio at the intended value, not a single output variable, like the control loop of a PLL or most other control loops that are frequently encountered and studied.
The loop control of a PLL needs to control a single variable, because a sinusoidal oscillator has a single degree of freedom, corresponding to its output frequency. The loop control of an optical frequency comb must control 2 variables, because a pulsed sinusoidal oscillator has 2 degrees of freedom, corresponding to the pulse repetition frequency and to the sinusoidal signal frequency.
In PLLs that are used for frequency multiplication of for frequency division, the frequency ratio is provided by an element inserted in the control loop, e.g. a digital frequency divider or a harmonic generator, which has an intrinsic frequency ratio, which does not have to be controlled by the control loop.
This text is so confidently incorrect, that I have to wonder if it is LLM-generated. The control loop for a frequency comb is actually very simple. Generating the carrier-envelope offset frequency signal is the hard (Nobel prize worthy) part.
The word "difficult" may be interpreted subjectively, i.e. what is difficult for one may be easy for another and what is easy for the first may be difficult for the second, but in any case I have not compared the difficulty of the control loop of a frequency comb with the difficulty of another component of a frequency comb, but only with the difficulty of the control loop of a PLL, which is the device replaced by a frequency comb.
There is no doubt that the control loop of a frequency comb is much more difficult to design than the control loop of a PLL or of any other simple regulator with a single degree of freedom.
Perhaps you have designed yourself the control loop for a frequency comb and it is indeed "very simple", but I have seen tons of books and of research papers about optical comb frequencies, which describe a lot of details about other parts of the optical frequency combs, but none of them ventures to present the details of how the control loop actually works.
While I have designed PLLs, I did not have the opportunity to design a frequency comb, so perhaps the additional difficulty is not great, as you claim, but I suspect that this is not true, because if it were easy to describe that in a few words such a description would have existed in one of the many publications about frequency combs.
In any case, when you claim that a text is "confidently incorrect", you should better point to the exact affirmation that is incorrect.
When you just do not agree with some subjective assessment, like whether something is easy or difficult, using the word "incorrect" is itself incorrect.
In this particular case you cannot even disagree with my use of the word "difficult", because if you claim that designing the double control loop of a frequency comb is not more difficult than the single control loop of a PLL, that claim is certainly incorrect.
They may have objected to your assertion in the first paragraph ("....the ratio between the repetition frequency of the pulses and the frequency of the sinusoidal signal must be a fixed known value.")
Admittedly I'm not well-informed on this topic at all, but I haven't run across that exact requirement. You need to know the pulse rep rate, which in practice may just be a matter of triggering the laser from a sufficiently-stable source rather than having to measure it separately -- and you need some way to get a carrier beatnote from a pair of lines, where the F-2F technique is a common approach -- but do you really need a priori knowledge of the PRF:carrier frequency ratio in order to make the whole thing work?
A frequency comb is nothing else but an oscillator that is not producing a continuous sinusoidal wave, but it is producing periodic pulses of sinusoidal waves, and which also satisfies the additional requirement that there must be a precise ratio between the frequency of the sinusoidal signal and the repetition frequency of the pulses.
It is simple to make a pulsed oscillator, but making one where the pulse frequency and the sinusoidal frequency maintain a fixed ratio is not at all simple, especially when the frequency ratio is very large, like what is needed when the pulse frequency must be low enough to drive a digital counter and the sinusoidal frequency must be in the optical range, up to ultraviolet light.
If you have such an oscillator, by tuning the low frequency you can obtain light with an accurately known frequency. Alternatively, by tuning the high frequency to match light with a known frequency, e.g. produced by an optical atomic clock, you can obtain a pulse train with a known frequency, which may be used as a reference frequency, e.g. for a digital clock.
> they show multi-spectral light coming in from the left, they show low-frequency light moving faster than high-frequency light ("Survey says: EEEEEHHHHHNNNNNNK!")
Has Sir never seen a rainbow?
May I kindly refer Sir, to Isaac Newton's prism experiment, as lovingly depicted on the cover of Pink Floyd's 'Dark Side of the Moon' album.
[1] https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=11...