Optical Frequency Combs

A frequency comb takes one fact from Fourier analysis — that a periodic signal has a spectrum of equally spaced lines — and pushes it to a metrological extreme. The periodic signal is the pulse train of a mode-locked laser (owned by aq-optics-mode-locking): identical femtosecond pulses spaced by the cavity round-trip time $T$ , whose Fourier dual is a picket fence of sharp lines spaced by $f_{\mathrm{rep}}=1/T$ . That is ordinary spectroscopy. What makes the comb a ruler is that the entire set of $\sim10^{5}$ – $10^{6}$ optical teeth, each near $5\times10^{14}$ Hz, is pinned by just two radio-frequency numbers: the tooth spacing $f_{\mathrm{rep}}$ and a rigid offset $f_0$ . The problem the comb solves is a counting problem. Optical frequencies ( $\sim500$ THz) oscillate far too fast for any electronic counter, while microwave frequencies ( $\sim10$ GHz) are countable and traceable to the SI second through caesium. Before combs, spanning that five-decade gap took a room-sized harmonic frequency chain dedicated to a single optical line. The comb replaces the chain with a self-consistent grid whose spacing and offset are both measurable in the RF domain: Kerr-lens mode-locking supplies the femtosecond pulse train, and $f$ – $2f$ self-referencing measures $f_0$ using an octave of bandwidth from photonic-crystal fibre. Lock $f_{\mathrm{rep}}$ and $f_0$ and every tooth becomes an absolute optical frequency — the instrument behind optical atomic clocks, hydrogen spectroscopy, astrocombs, and dual-comb spectroscopy.

Key takeaways

A mode-locked laser's pulse train has a spectrum that is a comb of discrete optical modes at $f_n = n\,f_{\mathrm{rep}} + f_0$ , where $f_{\mathrm{rep}}$ is the repetition rate and $f_0$ the carrier-envelope offset.
Two microwave numbers, $f_{\mathrm{rep}}$ and $f_0$ (tens of MHz to a few GHz), fix hundreds of thousands of optical frequencies near $10^{15}$ Hz — a phase-coherent gearbox between optical and microwave.
$f_{\mathrm{rep}}$ is read directly from a photodetector; $f_0$ is measured by self-referencing when the comb spans at least an octave, most commonly with an $f$ – $2f$ interferometer, which compares a frequency-doubled low-frequency comb component with a directly generated high-frequency component.
The offset exists because the carrier-envelope phase slips by $\Delta\phi_{\mathrm{ce}}$ relative to the pulse envelope on each round trip, shifting the comb by $f_0=(\Delta\phi_{\mathrm{ce}}/2\pi)\,f_{\mathrm{rep}}$ , with $\Delta\phi_{\mathrm{ce}}$ defined modulo $2\pi$ .
Combs collapsed optical metrology from a room of apparatus to a table-top box and enabled optical clocks near $10^{-18}$ ; the 2005 Nobel Prize recognised the technique.
Key idea: a mode-locked pulse train is, by Fourier duality, a comb of optical frequencies $f_n=n\,f_{\mathrm{rep}}+f_0$ pinned to two microwave numbers; measuring and locking $f_{\mathrm{rep}}$ and $f_0$ turns the laser into an absolute optical-frequency ruler that phase-coherently links THz optical frequencies to GHz microwaves.

Where this lives today

The comb is the engine room of optical timekeeping. In a modern optical lattice clock, an ultrastable laser is locked to a forbidden transition in neutral strontium or ytterbium — or to a single trapped ion — defining an optical-frequency tick near $10^{15}$ Hz, with state-of-the-art optical clocks reaching fractional uncertainties at or below the $10^{-18}$ level. That tick is useless until it can be counted, and the comb divides it phase-coherently down to a microwave rate an ordinary counter reads. The cold atoms are prepared with laser cooling — the magneto-optical trap, evaporative cooling — and clocks on different species are compared through combs to hunt for drifting fundamental constants and underpin the coming SI redefinition. Beyond clocks, combs have spread into instruments that look nothing like a physics-lab laser. Astrocombs calibrate the spectrographs hunting Earth-mass exoplanets, supplying thousands of absolutely known lines so a stellar Doppler wobble of centimetres per second can be trusted; dual-comb spectrometers map a whole molecular spectrum onto a compressed RF comb with no moving parts; and microresonator "Kerr combs" put an octave of teeth on a photonic chip, pointing toward field-deployable clocks, LIDAR, and terabit communications.

The physical picture

Picture a strobe firing at rate $f_{\mathrm{rep}}$ . Sampling a wave that regularly in time is, in the frequency domain, the same as punching out equally spaced narrow lines: periodicity in time (period $T$ ) forces regularity in frequency (spacing $1/T$ ). The finite duration of each flash — the pulse envelope — sets how many lines light up, so a shorter pulse means a broader comb; a train of $10$ -femtosecond pulses paints a spectrum $10^{5}$ teeth wide. Two subtleties turn a periodic spectrum into a ruler. First, because the pulse envelope travels at the group velocity while the carrier underneath travels at the phase velocity, dispersion slips the carrier under the envelope by a fixed $\Delta\phi_{\mathrm{ce}}$ every round trip, rigidly shifting the whole comb off the harmonics of $f_{\mathrm{rep}}$ by $f_0=\Delta\phi_{\mathrm{ce}}\,f_{\mathrm{rep}}/2\pi$ . Second, that offset can be measured by the comb itself.

Figure

Two faces of one laser output. In time, identical pulses every T = 1/f_rep, with the carrier slipping under the envelope by Δφ_ce each round trip. In frequency, a comb of teeth at f_n = n·f_rep + f_0, spaced by f_rep = 1 GHz and offset from zero by f_0 = 100 MHz.

The genius of self-referencing is that the comb reads its own offset. Take a red-edge tooth $f_n$ , double it to $2f_n=2n\,f_{\mathrm{rep}}+2f_0$ , and beat it against the real blue-edge tooth $f_{2n}=2n\,f_{\mathrm{rep}}+f_0$ . The repetition-rate terms cancel exactly and the beat is simply $f_0$ — provided the comb spans a full octave. To measure an unknown laser you then beat it against the nearest tooth; that beat, $f_{\mathrm{rep}}$ , and $f_0$ are all countable, so the optical frequency drops out once $n$ is fixed by a coarse wavemeter.

Figure

Self-referencing an octave-spanning comb. A red-edge tooth near 1064 nm (f_n) is doubled in a nonlinear crystal to 2f_n near 532 nm, where a genuine tooth f_2n already sits. Their beat is 2f_n − f_2n = f_0, so the comb measures its own offset using nothing but itself.

The core equations

The whole subject rests on one linear formula. Strict periodicity makes the spectrum equally spaced teeth; the carrier slip shifts them off the harmonics of $f_{\mathrm{rep}}$ , giving an arithmetic progression of optical frequencies.

f_n = n\, f_{\mathrm{rep}} + f_0,\qquad n\in\mathbb{Z}^{+} \qquad (1)

where: $f_n$ — frequency of the $n$ -th comb tooth (optical, $\sim10^{15}$ Hz); $n$ — mode index, a large integer $\sim10^{5}$ – $10^{6}$ ; $f_{\mathrm{rep}}$ — repetition rate = tooth spacing (RF, MHz–GHz); $f_0$ — carrier-envelope offset, $0\le f_0 < f_{\mathrm{rep}}$ . Two microwave numbers on the right fix an optical number on the left; the whole technique lives in that mismatch of scales. The tooth spacing is set by the cavity: the pulse re-emerges once per round-trip time $T$ , so $f_{\mathrm{rep}}=1/T$ — geometric, for a linear cavity of length $L$ .

f_{\mathrm{rep}} = \frac{1}{T} = \frac{v_g}{2L} \qquad (2)

where: $T$ — round-trip time; $L$ — cavity length; $v_g$ — intracavity group velocity. Being only a microwave frequency, $f_{\mathrm{rep}}$ is counted directly on a fast photodiode and locked to an RF reference. The offset is the subtle number: the carrier phase advances by a fixed $\Delta\phi_{\mathrm{ce}}$ each round trip, and Fourier-transforming that phase-stepped train shifts every line by the offset.

f_0 = \frac{\Delta\phi_{\mathrm{ce}}}{2\pi}\, f_{\mathrm{rep}} \qquad (3)

where: $\Delta\phi_{\mathrm{ce}}$ — carrier-envelope phase slip per round trip (modulo $2\pi$ ). Physical reading: $f_0$ is the slip as a frequency; zero slip puts the comb through DC, a slip of $\pi$ puts $f_0$ at $f_{\mathrm{rep}}/2$ . That slip has a concrete origin: envelope and carrier travel at the group and phase velocities, accumulating a relative phase each round trip that ties $f_0$ to dispersion.

\Delta\phi_{\mathrm{ce}} = \left(\frac{1}{v_g}-\frac{1}{v_p}\right) L\,\omega_c \ (\mathrm{mod}\ 2\pi) \qquad (4)

where: $v_p$ — carrier phase velocity, $v_g$ — envelope group velocity, $\omega_c$ — carrier angular frequency, and $L$ — one-way cavity length. Since $v_g\ne v_p$ in a dispersive medium, the carrier-envelope phase slips on every round trip, giving a generally nonzero $f_0$ that can drift with intracavity intensity and temperature and therefore must be actively stabilized for precision metrology. Measuring $f_0$ needs no external reference if the comb spans an octave: double a red tooth, beat it against a blue tooth near twice its frequency, and the index terms cancel, leaving the offset.

f_0 = 2f_n - f_{2n} = (2n\, f_{\mathrm{rep}} + 2f_0) - (2n\, f_{\mathrm{rep}} + f_0) \qquad (5)

where: $f_n$ — a red-edge tooth, frequency-doubled; $f_{2n}$ — a real blue-edge tooth at index $2n$ near $2f_n$ . Physical reading: the huge index $n$ vanishes identically, so the beat carries only $f_0$ . With $f_{\mathrm{rep}}$ and $f_0$ locked, an unknown CW laser is measured by its beat against the nearest tooth — every term on the right a counted microwave quantity.

f_{\mathrm{opt}} = n\, f_{\mathrm{rep}} + f_0 \pm f_{\mathrm{beat}} \qquad (6)

where: $f_{\mathrm{opt}}$ — unknown optical frequency; $f_{\mathrm{beat}}$ — beat against the nearest tooth, $|f_{\mathrm{beat}}|<f_{\mathrm{rep}}/2$ ; $\pm$ — which side of the tooth the laser lies on, resolved by dithering $f_{\mathrm{rep}}$ . The index $n$ is fixed separately by a coarse wavemeter good to better than $f_{\mathrm{rep}}/2$ . Finally, the comb is a lossless gearbox for stability: differentiating equation (1) with $f_0$ locked shows fractional fluctuations are preserved across the huge index $n$ .

\frac{\Delta f_{\mathrm{opt}}}{f_{\mathrm{opt}}} \approx \frac{\Delta f_{\mathrm{rep}}}{f_{\mathrm{rep}}} \qquad (\text{locked } f_0) \qquad (7)

where: $\Delta f/f$ — fractional frequency instability. Physical reading: an optical reference stabilises $f_{\mathrm{rep}}$ with its own fractional stability, and run forward, an optical clock's ticks divide down to a countable microwave rate without losing it.

Derivation AFrom a pulse train to a comb of frequencies

The goal is to prove rigorously that a mode-locked pulse train, with the carrier phase advancing by a fixed amount each round trip, has a spectrum of equally spaced lines at $f_n=n\,f_{\mathrm{rep}}+f_0$ . We model the output as a coherent sum of identical envelopes carrying a linearly advancing carrier phase and Fourier-transform it, so that both the spacing $f_{\mathrm{rep}}$ and offset $f_0$ emerge from one calculation, the envelope setting only which teeth carry power.

Step 1. Write the field as $E(t)=\sum_{k}A(t-kT)\,e^{i\omega_c(t-kT)}\,e^{ik(\omega_c T-\Delta\phi_{\mathrm{ce}})}$ : copies of one envelope $A$ , each delayed by $kT$ with an extra phase $k(\omega_c T-\Delta\phi_{\mathrm{ce}})$ . Why: the delay encodes the repetition and the extra phase the carrier slip — the ingredients that become $f_{\mathrm{rep}}$ and $f_0$ .

Step 2. Fourier transform via the delay theorem: $\tilde E(\omega)=\tilde A(\omega-\omega_c)\sum_k e^{-i(\omega-\omega_c)kT}\,e^{ik(\omega_c T-\Delta\phi_{\mathrm{ce}})}$ . Why: each shifted envelope becomes the same spectral shape times a pure phase, factoring the sum into a smooth envelope and a phased geometric series.

Step 3. Collect the geometric sum $\sum_k e^{-ik(\omega T-\Delta\phi_{\mathrm{ce}})}$ , which produces a Dirichlet-kernel comb with principal maxima when its phase is a multiple of $2\pi$ . Why: for many pulses the contributions interfere constructively near these resonances and destructively between them — turning a broad envelope into a picket fence of narrow teeth.

Step 4. Impose $\omega T-\Delta\phi_{\mathrm{ce}}=2\pi n$ and solve: $f_n=\dfrac{\omega}{2\pi}=n\,\dfrac{1}{T}+\dfrac{\Delta\phi_{\mathrm{ce}}}{2\pi T}=n\,f_{\mathrm{rep}}+f_0$ , with $f_0=\Delta\phi_{\mathrm{ce}}\,f_{\mathrm{rep}}/2\pi$ . Why: the spacing $1/T=f_{\mathrm{rep}}$ falls out of the delay and $f_0$ out of the carrier slip — the comb equation, with $\tilde A$ setting only tooth heights.

Go deeper · Derivation B · 1 minSelf-referencing with the f-to-2f interferometer

Derivation BSelf-referencing with the f-to-2f interferometer

The goal is to show how an octave-spanning comb measures its own offset $f_0$ with no external reference, pinning the last unknown in the comb equation. The trick compares the second harmonic of a low tooth against a genuine high tooth near the same frequency, engineered so the index-scaled contributions cancel — the idea that made combs self-contained instruments.

Step 1. Require an octave of optical bandwidth so that a low-frequency tooth $f_n$ and a high-frequency tooth $f_{2n}$ both lie within the comb spectrum and satisfy $f_{2n}=2n\,f_{\mathrm{rep}}+f_0\approx 2f_n$ . Why: without a comb tooth near twice the frequency of another tooth, there is nothing for the frequency-doubled light to beat against; octave spanning, often generated in photonic-crystal fibre, provides the required bandwidth.

Step 2. Frequency-double the red tooth: $2f_n=2n\,f_{\mathrm{rep}}+2f_0$ . Why: second-harmonic generation doubles both terms, so the doubled light carries $2f_0$ while a natural tooth nearby carries only $f_0$ .

Step 3. Identify the natural blue tooth $f_{2n}=2n\,f_{\mathrm{rep}}+f_0$ at index $2n$ . Why: it already exists and shares the exact $2n\,f_{\mathrm{rep}}$ term of the doubled red light, making the subtraction clean.

Step 4. Photodetect the beat: $f_{\mathrm{beat}}=|2f_n-f_{2n}|=f_0$ . Why: the $2n\,f_{\mathrm{rep}}$ terms cancel identically and the enormous index drops out, so one RF beat yields $f_0$ ; counting it and $f_{\mathrm{rep}}$ then fixes every tooth.

Go deeper · Derivation C · 1 minReading an unknown optical frequency off the comb

Derivation CReading an unknown optical frequency off the comb

The goal is to turn a countable RF beat note plus a coarse wavelength reading into an absolute optical frequency traceable to the SI second. The obstacle is a huge unmeasured integer $n$ , and a beat that fixes only the distance to the nearest tooth, not its sign. We resolve the sign by dithering and pin the integer with a modest wavemeter.

Step 1. Overlap the CW laser with the comb on a fast photodiode and record the lowest beat $f_{\mathrm{beat}}=f_{\mathrm{opt}}-(n\,f_{\mathrm{rep}}+f_0)$ , with $|f_{\mathrm{beat}}|<f_{\mathrm{rep}}/2$ . Why: the photodiode is a square-law mixer, reporting the difference between the laser and its neighbouring tooth as a countable RF signal.

Step 2. Resolve the sign by dithering $f_{\mathrm{rep}}$ : if $f_{\mathrm{beat}}$ shifts in the same sense as $n\,f_{\mathrm{rep}}$ the laser lies above the tooth (take $+$ ), otherwise below. Why: the tooth position depends on $n\,f_{\mathrm{rep}}$ , so how the beat tracks a change in $f_{\mathrm{rep}}$ reveals which side the laser sits on.

Step 3. Fix the integer with a coarse wavemeter reading $f_{\mathrm{opt}}^{\mathrm{coarse}}$ good to better than $f_{\mathrm{rep}}/2$ : $n=\mathrm{round}\!\left(\dfrac{f_{\mathrm{opt}}^{\mathrm{coarse}}-f_0}{f_{\mathrm{rep}}}\right)$ . Why: locating the laser to within half a spacing identifies the correct tooth uniquely, turning a five-decade problem into rounding.

Step 4. Assemble $f_{\mathrm{opt}}=n\,f_{\mathrm{rep}}+f_0+f_{\mathrm{beat}}$ , all three RF terms counted against a caesium- or maser-referenced clock. Why: the optical frequency is now expressed entirely in SI-traceable microwave quantities — absolute, with no room-sized frequency chain.

Limiting cases

Case	Result	Meaning
Zero offset ( $f_0=0$ )	$f_n = n\, f_{\mathrm{rep}}$	With the carrier-envelope slip stabilised to zero, teeth land exactly on harmonics of $f_{\mathrm{rep}}$ — a harmonic ruler through DC.
CW / single-mode limit	bandwidth $\to 0$ , one tooth survives	An infinitely long pulse (a CW laser) has vanishing bandwidth, so the comb collapses to a single line; breadth demands short pulses.
Octave-spanning spectrum	$f_{\max}\ge 2f_{\min}$	The condition for standard $f$ – $2f$ self-referencing, reached by supercontinuum in photonic-crystal fibre when the oscillator is too narrow.
Microwave-limit view	$f_{\mathrm{opt}} = n\, f_{\mathrm{rep}} + f_0 + f_{\mathrm{beat}}$	An optical frequency written entirely in countable RF terms — the comb as a coherent divider from $10^{15}$ to $10^{8}$ Hz.
Half-slip ( $\Delta\phi_{\mathrm{ce}}=\pi$ )	$f_0 = f_{\mathrm{rep}}/2$	The offset sits midway between harmonics; the carrier phase flips each pulse.

Worked examples

1 · Reconstructing a strontium clock frequency (concrete)

A self-referenced comb with $f_{\mathrm{rep}}=1{,}000{,}000{,}000$ Hz (1 GHz) and locked offset $f_0=+4{,}000{,}000$ Hz (4 MHz) is beat against a laser on the $^{87}\mathrm{Sr}$ clock transition near 698 nm. A wavemeter locates it near 429.228 THz and the counted beat is $f_{\mathrm{beat}}=+229{,}873$ Hz. Find the absolute optical frequency.

Step 1. Fix the mode index from the coarse reading: $n=\operatorname{round}\!\left[(429{,}228{,}000{,}000{,}000-4{,}000{,}000)/10^9\right]=\operatorname{round}(429{,}227.996)=429{,}228$ . The coarse frequency measurement should locate the laser to better than $\pm f_{\rm rep}/2=\pm500$ MHz to identify the correct comb tooth.

Step 2. Assemble $f_{\rm opt}=n f_{\rm rep}+f_0\pm f_{\rm beat}$ . Here the sign is positive from the beat-sign determination: the index term is $429{,}228\times10^9=429{,}228{,}000{,}000{,}000$ Hz, plus $f_0=4{,}000{,}000$ Hz and $f_{\rm beat}=229{,}873$ Hz.

Step 3. Sum the three RF numbers onto the index term: $429{,}228{,}000{,}000{,}000+4{,}000{,}000+229{,}873=\mathbf{429{,}228{,}004{,}229{,}873\ Hz}$ — the recommended $^{87}\mathrm{Sr}$ value, pinned entirely by three countable microwave numbers. f_opt = 429 228 004 229 873 Hz.

2 · Mode index and tooth spacing of a Ti:sapphire comb

A Ti:sapphire comb runs at $f_{\mathrm{rep}}=1$ GHz with $f_0=0$ . Near $375$ THz (about 799 nm), find the mode index $n$ and the tooth spacing as a wavelength interval.

Step 1. The index is the ratio of optical to repetition frequency: $n=f/f_{\mathrm{rep}}=\dfrac{3.75\times10^{14}}{10^{9}}=\mathbf{375{,}000}$ — a third of a million cavity modes under the pulse spectrum.

Step 2. Convert the 1 GHz spacing to wavelength with $\Delta\lambda=\lambda^{2}\Delta f/c$ , using $\lambda=c/f=799$ nm: $\Delta\lambda=\dfrac{(799\times10^{-9})^{2}(1\times10^{9})}{2.998\times10^{8}}\approx\mathbf{2.1\times10^{-3}\ nm}$ .

Step 3. Adjacent teeth differ by only about $2$ picometres, much finer than the spectral resolution of many conventional grating spectrometers at this wavelength — which is why a comb is typically read through beat notes rather than by resolving individual teeth with a dispersive spectrometer. n = 375 000, Δλ ≈ 2.1×10⁻³ nm per tooth.

3 · Carrier-envelope slip and stability transfer

A fibre comb has $f_{\mathrm{rep}}=100$ MHz and a carrier-envelope slip $\Delta\phi_{\mathrm{ce}}=0.2\pi$ rad per round trip. Find $f_0$ . Then, if one tooth is locked to an optical reference of fractional stability $1\times10^{-15}$ , what fractional stability appears on $f_{\mathrm{rep}}$ ?

Step 1. Apply equation (3): $f_0=\dfrac{\Delta\phi_{\mathrm{ce}}}{2\pi}f_{\mathrm{rep}}=\dfrac{0.2\pi}{2\pi}\times10^{8}=\mathbf{10\ MHz}$ — well within the $100$ MHz spacing.

Step 2. Apply equation (7): with $f_0$ locked, fractional stability is conserved, so $\dfrac{\Delta f_{\mathrm{rep}}}{f_{\mathrm{rep}}}\approx\dfrac{\Delta f_{\mathrm{opt}}}{f_{\mathrm{opt}}}=\mathbf{1\times10^{-15}}$ .

Step 3. The leverage: the optical reference at $\sim4\times10^{14}$ Hz is divided by $n\approx4\times10^{6}$ to $10^{8}$ Hz, yet fractional stability is untouched — exactly how combs synthesise record low-phase-noise microwaves from optical clocks. f_0 = 10 MHz; the microwave inherits 1×10⁻¹⁵.

Physics of the portrait

The frequency comb is where decades of precision-frequency research converged: it turns optical-frequency oscillations into electronically countable radio-frequency signals, enabling direct measurement of optical frequencies and helping make optical atomic clocks possible.

Origins · 1 minTheodor Hänsch and how this was discovered

Theodor Hänsch

b. 1941 · precision laser spectroscopy and the optical frequency comb

Theodor W. Hänsch is a German physicist whose lifelong pursuit of precision spectroscopy of hydrogen helped drive the development of the optical frequency comb. His work focused particularly on the hydrogen 1S–2S transition, a benchmark for testing quantum electrodynamics and determining fundamental constants such as the Rydberg constant. Reading an optical frequency in absolute terms meant bridging five decades to the countable microwave domain, which then required a harmonic frequency chain filling a laboratory. As early as 1978 he noticed that a train of picosecond pulses acts as a comb useful for spectroscopy, decades before the tools existed. When Kerr-lens mode-locked Ti:sapphire lasers and octave-spanning supercontinuum from photonic-crystal fibre arrived at the end of the 1990s, his Garching group showed that a single femtosecond laser could replace the entire chain. In parallel, John L. Hall at JILA in Boulder independently developed the self-referencing and stabilisation techniques — the $f$ – $2f$ interferometer and phase-locked control of $f_{\mathrm{rep}}$ and $f_0$ — that made comb teeth absolute. In 2005 Hall and Hänsch shared the Nobel Prize in Physics for the laser-based precision spectroscopy the comb enabled; it now underpins every optical clock in the world.

History

1978

— Hänsch and coworkers use a picosecond mode-locked laser as a frequency comb for two-photon spectroscopy of sodium

1999

— Udem, Reichert, Holzwarth and Hänsch measure an optical frequency directly with a femtosecond comb

2000

— Ranka, Windeler and Stentz generate octave-spanning supercontinuum in photonic-crystal fibre

2000

— Jones, Diddams, Hall, Cundiff and coworkers realise f–2f self-referencing and stabilise the carrier-envelope offset

2005

— Hall and Hänsch share the Nobel Prize in Physics for the optical frequency-comb technique

Going further

Texts. Foot, Atomic Physics, ch. 8–9 (frequency standards) · Demtröder, Laser Spectroscopy, Vol. 2, ch. 9 (frequency combs and metrology) · Saleh and Teich, Fundamentals of Photonics (ultrafast optics, mode locking) · Cundiff and Ye, "Colloquium: Femtosecond optical frequency combs", Rev. Mod. Phys. 75, 325 (2003) · Hall, Rev. Mod. Phys. 78, 1279 (2006) and Hänsch, Rev. Mod. Phys. 78, 1297 (2006) — the Nobel lectures. The three core relations — $f_n=n\,f_{\mathrm{rep}}+f_0$ , $f_0=\Delta\phi_{\mathrm{ce}}\,f_{\mathrm{rep}}/2\pi$ , and $f_0=2f_n-f_{2n}$ — are standard results of the field. Two quick checks anchor the worked examples: the mode-index calculation gives $3.75\times10^{14}/10^{9}=3.75\times10^{5}$ , and the strontium reconstruction $429{,}228\times10^{9}+4\times10^{6}+229{,}873=429\,228\,004\,229\,873$ Hz matches the recommended value. The $\sim10^{-18}$ clock uncertainty is quoted as an order-of-magnitude figure.

Prerequisites. Fourier analysis of periodic signals · Mode-locked lasers · Optical cavities and dispersion · Nonlinear optics and second-harmonic generation. Leads to. Optical atomic clocks · Precision spectroscopy of hydrogen · Dual-comb spectroscopy · Magneto-optical trap.

Ask the tutor: Ask me how a single femtosecond laser replaced a room-sized frequency chain, and we will build the $f$ – $2f$ trick from scratch.

Is a frequency comb just an ordinary periodic spectrum, or something more

Physically the tooth pattern is nothing exotic: any periodic signal has a line spectrum, and a pulse train spaced by $T$ gives lines spaced by $1/T$ . What elevates the comb to a metrological instrument is coherence and control — the teeth are phase-locked to two RF numbers, $f_{\mathrm{rep}}$ and $f_0$ , that can be counted against the SI second and actively stabilised. Read through beat notes, a comb delivers hundreds of thousands of addressable optical frequencies each known to microwave-counter precision. Structure alone is common; a locked, absolute-frequency structure is the invention.

What actually is the carrier-envelope offset frequency f_0

Inside the laser the envelope travels at the group velocity while the carrier beneath it travels at the phase velocity. Because these differ in a dispersive medium, each round trip the carrier crest slides under the envelope peak by a fixed $\Delta\phi_{\mathrm{ce}}$ . Successive pulses therefore differ, and Fourier-transforming the train shifts every line bodily by $f_0=\Delta\phi_{\mathrm{ce}}\,f_{\mathrm{rep}}/2\pi$ . So $f_0$ is the phase slip re-expressed as a frequency, and measuring it is the whole point of self-referencing.

Why does doubling a red tooth and subtracting a blue tooth give exactly f_0

Because the enormous mode index cancels. A red tooth $f_n=n\,f_{\mathrm{rep}}+f_0$ , doubled, gives $2f_n=2n\,f_{\mathrm{rep}}+2f_0$ — twice the offset. The natural blue tooth at index $2n$ is $f_{2n}=2n\,f_{\mathrm{rep}}+f_0$ — the offset once. Their difference is $2f_n-f_{2n}=f_0$ : the $2n\,f_{\mathrm{rep}}$ terms vanish. It works only because both fields share the same $f_{\mathrm{rep}}$ , and needs an octave so both a tooth $n$ and a tooth $2n$ carry power.

How do you know the integer mode index n if the comb does not measure it

You measure it separately and coarsely. The comb equation hides $n\sim10^{5}$ – $10^{6}$ , and the beat gives only the distance to the nearest tooth, not which tooth. A conventional wavemeter, unable to resolve individual teeth, still locates the laser to within tens of MHz — better than half a spacing — so rounding $(f_{\mathrm{opt}}^{\mathrm{coarse}}-f_0)/f_{\mathrm{rep}}$ returns the correct integer uniquely. If no wavemeter is good enough, remeasure at a slightly different $f_{\mathrm{rep}}$ and solve for $n$ . Miss $n$ by one and the answer is off by a full $f_{\mathrm{rep}}$ .

Why is a shorter pulse needed for a broader comb

The comb width is the pulse bandwidth, and bandwidth is the Fourier partner of pulse duration. A long pulse spans few frequencies, so its spectrum is narrow and few teeth carry power; an infinitely long (CW) pulse collapses the comb to a single line. A few-cycle femtosecond pulse is spectrally broad, lighting up $10^{5}$ teeth or more. The spacing $f_{\mathrm{rep}}$ is fixed by the cavity; duration controls only how many teeth exist — which is why self-referencing wants the shortest pulses and octave-spanning broadening.

How does a comb make an optical clock usable

An optical clock's "pendulum" is an atomic transition near $10^{15}$ Hz, too fast to count electronically. The comb divides it down: lock one tooth to the clock laser and lock $f_0$ , and $f_{\mathrm{rep}}=f_{\mathrm{opt}}/n$ becomes a microwave frequency a counter can read, carrying the clock's ticks. Fractional stability is conserved across the division, so the microwave output inherits the optical precision. Without the comb, an optical clock would be an exquisite oscillator whose ticks nobody could count.