Optical Frequency Combs
An optical frequency comb is a light source whose spectrum consists of thousands to millions of discrete, evenly spaced lines. Each line is a laser of known frequency, and the whole set behaves as a ruler laid across the optical spectrum. The comb solved a problem that had blocked precision optics for decades: optical carrier frequencies are hundreds of terahertz, far too fast for any electronic counter, yet the definition of the second and the entire apparatus of frequency metrology rest on counting. A comb bridges that gap by phase-coherently linking optical frequencies to a microwave repetition rate that a counter can handle directly.
The importance of this bridge is hard to overstate. Before combs, measuring an optical frequency against the cesium microwave standard required a harmonic frequency chain: a room full of oscillators, multipliers, and transfer lasers, each locked to the one below, built for a single wavelength and operable by a handful of specialists. A comb replaces that chain with one mode-locked laser on a small optical table, or with a microresonator on a chip. John L. Hall and Theodor W. Hänsch shared half of the 2005 Nobel Prize in Physics for developing the technique. Combs now underpin optical atomic clocks, calibrate astronomical spectrographs, drive broadband molecular spectroscopy, generate low-noise microwaves, and supply the parallel carriers of chip-scale optical computing and communication.
The Comb Equation and Its Two Degrees of Freedom
A mode-locked laser emits a periodic train of ultrashort pulses. In the frequency domain, a periodic train is a set of equally spaced lines. If the pulses repeat at rate frep, the lines are spaced by exactly frep. The spacing alone, however, does not fix where the lines sit. The comb is described by two numbers:
fn = n × frep + fCEO
Here n is a large integer, of order 105 to 106 for a mode-locked laser running at a few hundred megahertz, and fCEO is the carrier-envelope offset frequency, a rigid displacement of the entire comb away from the harmonics of frep. Both frep and fCEO lie in the radio or microwave range, where they can be counted, compared, and servo-controlled with ordinary electronics. Fix those two numbers and every one of the hundreds of thousands of optical lines is known to the same fractional accuracy as the microwave reference. That is the entire trick.
Where the Offset Comes From
Inside the laser cavity, the pulse envelope travels at the group velocity while the carrier wave underneath it travels at the phase velocity. Dispersion makes these differ, so on each round trip the carrier slips relative to the envelope peak by a phase ΔφCE. The offset frequency is that slip expressed as a rate: fCEO = ΔφCE × frep / 2π. A pulse train in which the carrier phase is identical under every envelope has zero offset and lands exactly on harmonics of the repetition rate. In a real laser the offset drifts with pump power, cavity length, temperature, and intracavity intensity, and it must be measured and stabilized rather than assumed.
Why Both Degrees of Freedom Must Be Controlled
Controlling only the repetition rate leaves an unknown common shift on every line. Controlling only the offset leaves the spacing free to breathe. Two independent actuators are therefore required, and they must act on the laser in ways that are not degenerate. A common arrangement uses a piezoelectric transducer on a cavity mirror to change cavity length, and therefore frep, while pump-current modulation changes intracavity intensity and therefore fCEO. Because each actuator perturbs both parameters to some extent, the two servo loops interact, and practical comb control involves either careful choice of actuator pairs with weak cross-coupling or explicit decoupling of the control matrix.
Self-Referencing and the f-2f Interferometer
Measuring frep is simple: send the pulse train to a photodiode and read the beat at the pulse repetition rate. Measuring fCEO is the harder half, and the standard answer is self-referencing, in which the comb measures its own offset by comparison with itself.
The Octave Requirement
Take a low-frequency comb line at fn = n frep + fCEO and double it in a nonlinear crystal, giving 2n frep + 2fCEO. Now take the comb line that already sits near that frequency, f2n = 2n frep + fCEO. Beat the two on a photodetector. The large terms cancel and the difference is exactly fCEO. This is the f-2f interferometer, and it requires the comb to span a full octave, meaning that the highest frequency present is at least twice the lowest.
Few oscillators emit an octave directly. The usual route is spectral broadening in a highly nonlinear fiber, most often a photonic crystal fiber whose small core and engineered zero-dispersion wavelength produce supercontinuum generation. The discovery in 1999 and 2000 that photonic crystal fiber could broaden a titanium-sapphire pulse train to an octave while preserving the comb's phase coherence is what turned self-referencing from a proposal into a laboratory instrument. A related scheme, the 2f-3f interferometer, relaxes the requirement to two-thirds of an octave at the cost of an additional nonlinear step.
Locking Strategies
Once fCEO and frep are available as electrical signals, several architectures follow. In the fully stabilized comb, both are phase-locked to a hydrogen maser or other microwave reference, and the comb inherits that reference's accuracy across the optical spectrum. In the more common metrology arrangement, one comb line is phase-locked to an ultrastable continuous-wave laser held on a high-finesse cavity, and fCEO is locked separately. The comb then inherits the cavity-stabilized laser's short-term linewidth, and the repetition rate becomes an optical-quality microwave signal. This optical frequency division is the basis of the lowest-phase-noise microwave oscillators demonstrated, with absolute phase noise on a 10 GHz carrier reaching levels no room-temperature electronic oscillator approaches.
Offset-free schemes avoid fCEO altogether. Difference-frequency generation between two spectral regions of the same comb cancels the offset in the mixing process, producing an intrinsically offset-free comb in the mid-infrared. Electro-optic combs, generated by driving phase and intensity modulators with a microwave synthesizer, inherit their offset from the seed laser and their spacing from the drive frequency, which makes both parameters directly and independently settable.
Mode-Locked Laser Combs
The mode-locked laser remains the workhorse comb source and the reference against which newer approaches are judged. Mode locking establishes a fixed phase relationship among the cavity's longitudinal modes so that they interfere constructively into a single circulating pulse.
Titanium-Sapphire Systems
Kerr-lens mode-locked titanium-sapphire lasers produce pulses from under ten femtoseconds to tens of femtoseconds near 800 nanometers, with repetition rates typically between 100 MHz and 1 GHz. Their broad gain bandwidth supports the shortest pulses and the widest direct spectra, and the octave-spanning designs can self-reference without external broadening. They offer the lowest intrinsic carrier-envelope phase noise of the common comb sources, which keeps them in service for the most demanding measurements. The costs are a green pump laser, sensitivity to alignment and environment, and free-space optics that require a stable table.
Fiber Combs
Erbium-doped fiber lasers near 1550 nanometers and ytterbium-doped fiber lasers near 1030 nanometers now dominate deployed systems. Mode locking uses nonlinear polarization evolution, a nonlinear amplifying loop mirror, or a semiconductor saturable absorber mirror. An all-fiber cavity has no alignment to lose, tolerates vibration and temperature swings, and reaches the required peak power through an amplifier stage before broadening in highly nonlinear fiber. Erbium combs also sit in the telecommunications band, so they can be distributed over standard optical fiber and use commodity components. These properties have carried fiber combs out of the laboratory and into shipping containers, aircraft, and rocket payloads; a fiber comb operated successfully on a sounding-rocket flight in 2015, and comparable systems now support field campaigns in gas sensing and remote timing.
The trade-off is noise. Fiber cavities are long and store energy in a medium with strong nonlinearity, which raises the free-running carrier-envelope offset linewidth well above that of a titanium-sapphire laser and demands higher servo bandwidth to suppress. Careful dispersion management and low-noise pump diodes narrow the gap considerably.
Diode-Pumped Solid-State and Semiconductor Combs
Diode-pumped solid-state lasers using ytterbium-doped crystals or ceramics, mode-locked with semiconductor saturable absorber mirrors, combine low noise with efficient pumping and reach multi-gigahertz repetition rates in compact cavities. Mode-locked semiconductor lasers and quantum-cascade laser combs push integration further still. The quantum-cascade comb is notable for operating directly in the mid-infrared and terahertz regions on a chip-scale device, where its comb formation relies on four-wave mixing and frequency modulation rather than short-pulse generation.
Microresonator Kerr Combs
A microcomb is generated not by a pulsed laser but by pumping a high-quality-factor microresonator with a single continuous-wave laser. When the pump power inside the resonator exceeds a threshold, the third-order Kerr nonlinearity drives degenerate and cascaded four-wave mixing: pump photons convert into pairs of signal and idler photons on neighboring resonator modes, and those in turn seed further mixing until a comb fills the resonator bandwidth. The line spacing equals the resonator free spectral range, which is set by the round-trip path length and therefore by device geometry.
Platforms and Quality Factors
Silicon nitride is the dominant integrated platform. It is transparent from the visible into the mid-infrared, free of two-photon absorption at telecommunications wavelengths, compatible with silicon foundry processing, and capable of intrinsic quality factors above 107 with careful fabrication and annealing. Crystalline whispering-gallery resonators in magnesium fluoride and calcium fluoride reach far higher quality factors, in the 109 range, and therefore lower threshold power, but they are machined and polished individually rather than lithographically defined. Silica microtoroids and wedge resonators, aluminum nitride, lithium niobate, and tantalum pentoxide each occupy their own corner of the design space, trading nonlinearity, transparency window, and manufacturability.
The dispersion of the resonator matters as much as its quality factor. Comb formation over a broad bandwidth requires anomalous group-velocity dispersion, which in an integrated waveguide is achieved by choosing the cross-sectional geometry so that waveguide dispersion overcomes the normal material dispersion. Dispersion engineering by cross-section is one of the principal reasons integrated platforms have advanced so quickly.
Chaotic Combs Versus Soliton Combs
An early and important discovery was that a spectrally broad microcomb is not automatically a coherent one. Combs formed from multiple independent subcomb families beat against each other and exhibit high-amplitude noise, broad radio-frequency beat notes, and no usable phase relationship. Such modulation-instability combs are useless for metrology.
The resolution came in 2014 with the demonstration of dissipative Kerr solitons in a crystalline microresonator. A dissipative Kerr soliton is a stable pulse that circulates in the resonator, sustained by a double balance: Kerr nonlinearity against anomalous dispersion, and parametric gain from the pump against cavity loss. A single circulating soliton produces a smooth, low-noise, sech-squared spectral envelope and a narrow repetition-rate beat note. The soliton state is what makes a microcomb a real comb.
Reaching and Holding the Soliton State
Solitons exist only on the red-detuned side of the pump resonance, a region that is thermally unstable. As the pump is tuned into resonance, absorbed power heats the resonator and shifts the resonance, so the soliton step is crossed in microseconds and lost. Several access techniques address this. Fast frequency scans sweep through the unstable region faster than the thermal response. Power kicking momentarily drops and restores pump power to arrest the thermal transient. Auxiliary-laser heating uses a second laser in a different mode to hold the resonator temperature constant while the pump is tuned. Self-injection locking, in which light backscattered from the resonator locks a semiconductor pump laser directly, both narrows the pump linewidth and drives the system into a soliton state, and it has produced turnkey integrated modules in which a laser chip is butt-coupled to a resonator chip and solitons form on power-up.
Once running, the number of solitons in the cavity determines the spectral envelope. Multi-soliton states produce interference fringes across the spectrum, so most applications tune to a single-soliton state, sometimes through controlled soliton switching. Related states extend the design space: dark pulses, or platicons, form in normal-dispersion resonators and offer markedly higher pump-to-comb conversion efficiency, and soliton crystals self-organize into regularly spaced multi-soliton arrangements that form more readily, at the cost of an effective line spacing that is a multiple of the resonator free spectral range.
What Microcombs Buy and What They Cost
The gains are size, power, and repetition rate. A microcomb occupies a chip rather than a table, needs milliwatts to hundreds of milliwatts of pump power, and produces line spacings from roughly 10 GHz to 1 THz. Spacings in the tens of gigahertz are exactly what dense wavelength-division multiplexing, microwave photonics, and astronomical spectrograph calibration want, and they are inconveniently high for a bulk mode-locked laser.
The costs are equally real. Large line spacing means fewer lines for a given bandwidth and requires a fast photodetector to observe the repetition rate directly. Reaching an octave for self-referencing needs strong dispersion engineering and often dispersive-wave enhancement at the spectral wings. Pump-to-comb conversion efficiency in bright-soliton states is low, commonly a few percent. And thermal sensitivity means the pump-resonance detuning must be actively held. Fully self-referenced, octave-spanning integrated microcombs have been demonstrated, but they remain harder to build and to keep running than a fiber comb.
Key Specifications
Comparing combs across technologies is easier with a shared vocabulary of parameters.
Repetition Rate
Fiber and titanium-sapphire combs commonly run between 100 MHz and 1 GHz; specialized cavities reach several gigahertz. Microcombs span roughly 10 GHz to 1 THz. Higher repetition rate means more power per line and easier line resolution by a conventional spectrometer, but fewer lines across a given span and a harder photodetection problem.
Spectral Coverage
Direct emission bandwidth is set by the gain medium or resonator dispersion. Coverage is extended by supercontinuum generation, second-harmonic and sum-frequency generation into the visible and ultraviolet, difference-frequency generation and optical parametric oscillation into the mid-infrared, and high-harmonic generation into the extreme ultraviolet. Mid-infrared coverage matters most for spectroscopy because the fundamental vibrational bands of most molecules lie there.
Line Width and Coherence
The optical linewidth of an individual comb tooth reflects the pump or oscillator noise plus the residual servo error. A comb line locked to a cavity-stabilized laser can be sub-hertz. A free-running microcomb line is no narrower than its pump.
Absolute linewidth is not always the figure that matters. Dual-comb and transfer-oscillator schemes depend on mutual coherence, and much of the common-mode noise carried by two lines of the same comb, or by two combs sharing a reference, cancels in the beat. A comb whose individual teeth are kilohertz-wide can still support a measurement limited far below that, which is why a specification should state the relative linewidth over the intended measurement time and not the absolute linewidth alone.
Fractional Frequency Stability and Accuracy
Stability is quoted as an Allan deviation against averaging time, and accuracy as a fractional uncertainty. Independent combs measuring the same optical frequency ratio have been shown to agree at the 10-21 level, and purpose-built optical frequency dividers have been characterized to comparable division uncertainty. The comb's own contribution therefore sits well below the uncertainty of the best optical clocks it is used to measure. In practice, a comb's accuracy is that of whatever reference it is locked to, and the comb transfers it faithfully. Quoting a comb's accuracy in isolation is meaningless; what is measurable, and what these comparisons quantify, is the fidelity of that transfer.
Power per Line and Relative Intensity Noise
Total average power divided across the number of lines determines the signal available per channel. A one-watt comb with 105 lines delivers ten microwatts per line; a microcomb with a hundred lines delivers far more per line from far less total power. Relative intensity noise and line-to-line power flatness govern the achievable signal-to-noise ratio in spectroscopy and the link budget in communication.
Applications in Metrology and Timekeeping
Metrology was the comb's first application and still sets its hardest requirements. Each use below exploits the same property from a different direction: a phase-coherent link between optical and microwave frequencies that runs in both directions.
Optical Atomic Clocks
An optical clock interrogates a narrow atomic transition at hundreds of terahertz with a cavity-stabilized laser. The comb performs the essential division, converting that optical frequency into a countable microwave output while preserving the fractional stability. Without a comb, an optical clock produces exquisite stability that nobody can read. Comb-linked optical clocks now report systematic uncertainties in the 10-18 range and below, and international work toward a redefinition of the SI second in terms of an optical transition depends entirely on comb-based comparisons among laboratories. Comparing distant clocks over fiber links, and eventually the relativistic geodesy that follows from comparing clocks at different gravitational potentials, likewise runs through combs.
Low-Noise Microwave Generation
Optical frequency division inverts the usual direction. Lock a comb line to an ultrastable optical reference, and the repetition rate becomes a microwave signal whose fractional frequency noise is divided down from the optical carrier by the ratio of the frequencies, a factor of order 104 to 105, corresponding to a phase-noise reduction of 80 to 100 decibels. Photodetecting that pulse train yields 10 GHz signals with phase noise below what quartz, sapphire-loaded cavity, or dielectric resonator oscillators achieve at room temperature. Radar with extreme Doppler sensitivity, very-long-baseline interferometry, and high-speed analog-to-digital conversion are the beneficiaries.
Distance Measurement
Comb-based ranging, sometimes called light detection and ranging with combs, uses the comb's simultaneous provision of an absolute optical frequency scale and a precise pulse timing scale. Dual-comb ranging, multi-wavelength interferometry, and time-of-flight cross-correlation together resolve the ambiguity that plagues single-wavelength interferometry, delivering nanometer precision over meters and useful precision over kilometers. Satellite formation flying and large-structure metrology are the driving applications.
Astronomical Spectrograph Calibration
Detecting an Earth-mass planet around a Sun-like star requires resolving a stellar radial-velocity wobble of roughly ten centimeters per second, which corresponds to a fractional wavelength shift near 3 × 10-10. Thorium-argon hollow-cathode lamps, the traditional wavelength reference, have irregular line spacing, uneven intensity, and lines that age. An astro-comb provides uniform, equally spaced lines of known absolute frequency and stable intensity. Because spectrographs cannot resolve a 250 MHz comb, the comb is filtered by Fabry-Perot cavities up to spacings of 10 to 30 GHz, or generated directly at that spacing by a microcomb. Astro-combs are installed on major high-resolution spectrographs and have demonstrated calibration repeatability at the centimeter-per-second level, beyond what stellar surface activity currently allows scientists to exploit.
Applications in Spectroscopy
Spectroscopy uses the comb as a source and not only as a ruler. The attraction is that thousands of precisely known, mutually coherent frequencies illuminate the sample at once, so a broadband spectrum carrying an absolute frequency axis is acquired in a single measurement rather than assembled from a wavelength scan.
Direct Frequency-Comb Spectroscopy
Illuminating a sample with the full comb probes every line at once. Dispersing the transmitted light on a cross-dispersed or virtually imaged phased array spectrometer records thousands of resolution elements in a single frame, with each element carrying an absolute frequency axis fixed by the comb equation rather than by a wavelength calibration. Coupling the comb into an enhancement cavity multiplies the effective path length by the cavity finesse, and cavity-enhanced direct comb spectroscopy reaches trace-gas detection limits in the parts-per-trillion range across broad spectral spans.
Dual-Comb Spectroscopy
Dual-comb spectroscopy is the technique that has spread furthest beyond specialist laboratories. Two mutually coherent combs with slightly different repetition rates, differing by Δfrep, are combined on a single photodetector after one passes through the sample. Each pair of neighboring lines produces a distinct radio-frequency beat, so the optical spectrum maps linearly onto a radio-frequency comb that a digitizer records directly. The result is a Fourier-transform spectrometer with no moving parts, microsecond to millisecond acquisition times, and resolution set by the comb line spacing rather than by a mechanical scan length.
The requirement is mutual coherence between the two combs over the measurement time. Fully locking both combs to common references achieves it at the cost of complexity; free-running dual-comb systems with computational phase correction, and single-cavity dual-comb sources in which both combs share one laser and therefore share most of their noise, have made the technique far more practical. Open-path dual-comb systems now measure methane, carbon dioxide, water vapor, and other species over kilometer-scale atmospheric paths for greenhouse-gas flux and leak-detection studies, and chip-scale dual-microcomb spectrometers have been demonstrated.
Molecular Fingerprinting and Breath Analysis
Mid-infrared combs generated by difference-frequency generation, optical parametric oscillation, or quantum-cascade devices reach the fundamental vibrational bands where molecular absorption cross-sections are strongest. Simultaneous broadband coverage lets a single measurement identify and quantify many species at once, which is what distinguishes comb spectroscopy from a tunable-diode-laser instrument aimed at one line. Applications include industrial process monitoring, combustion diagnostics, semiconductor process gas analysis, and human breath analysis for biomarkers.
Applications in Communication and Computing
A comb is a bank of equally spaced, mutually coherent carriers from one device. In wavelength-division multiplexed transmission, that replaces dozens of independently temperature-controlled distributed-feedback lasers with a single source whose channel grid is defined by physics rather than by individual wavelength lockers, and whose mutual coherence enables joint digital signal processing across channels. Laboratory demonstrations have pushed this hard. In 2022 a single silicon nitride microcomb, operated in the dark-pulse regime, supplied 223 wavelength channels that together carried 1.84 petabits per second over a 37-core fiber. That result combines wavelength and spatial multiplexing in a research setting and is not a deployed system capability, but it establishes that one chip-scale source can feed a channel count that would otherwise demand hundreds of individually controlled lasers.
The same property serves photonic computing. Optical matrix-multiplication engines use each comb line as an independent channel that carries one element of a vector, allowing wavelength-parallel operation through a shared photonic circuit. Microwave photonics uses combs for tunable filtering, arbitrary waveform generation, true-time-delay beamforming, and photonic analog-to-digital conversion, in each case exploiting the comb's simultaneous provision of many precisely known carriers.
Practical Engineering and Troubleshooting
Comb systems fail in characteristic ways, and most problems fall into a few families.
The Offset Beat Is Weak or Absent
A low-contrast fCEO signal usually traces to the nonlinear stage rather than to the detection electronics. Check that the supercontinuum truly spans an octave at usable power at both wings, that the polarization entering the highly nonlinear fiber is correct, and that the two interferometer arms are matched in path length to within the coherence length and overlapped in polarization and spatial mode. Excess pulse energy in the broadening fiber degrades coherence rather than improving it, so more power is not always the fix. A signal-to-noise ratio of at least 30 dB in a 100 kHz resolution bandwidth is the usual working threshold for a reliable phase lock.
Servo Loops Fight Each Other
Because most actuators shift both frep and fCEO, closing one loop can unlock the other. Measure the transfer function of each actuator onto each observable, choose actuator pairs whose cross-coupling is weakest, and separate the loop bandwidths so that the fast loop, usually pump current for the offset, and the slow loop, usually the piezoelectric transducer for the cavity length, do not compete. Piezoelectric mechanical resonances set a hard limit on the slow loop's bandwidth and are a common source of servo oscillation.
Cycle Slips and Silent Errors
A phase lock that occasionally slips a cycle produces data that looks locked but carries frequency errors. Monitor the in-loop error signal, the out-of-loop counter, and the beat-note amplitude continuously, and reject data intervals in which any of them departs from its normal range. For frequency measurements, use a Π-type or Λ-type counter deliberately and understand which one the analysis assumes, since the two weight phase noise differently.
Mode Number Ambiguity
The comb equation gives the frequency of line n only if n is known. Determining it requires an independent coarse measurement, from a wavemeter accurate to better than half the repetition rate, or from repeating the beat measurement at two different repetition rates and solving for the integer that is consistent with both. Errors here produce results wrong by an exact multiple of frep, which is a large and easily missed mistake.
Microcomb Soliton Loss
A microcomb that drops out of the soliton state typically shows a sudden change in transmitted pump power and a collapse of the smooth spectral envelope. Thermal drift moving the pump-resonance detuning out of the existence range is the usual cause. Remedies are active detuning control against the measured pump transmission, auxiliary-laser thermal stabilization, temperature control of the chip, and self-injection-locked designs that recover the state automatically. Backreflection into the pump laser, dust or moisture on a coupling facet, and polarization drift in the input fiber are the other frequent culprits.
Environment
Combs are sensitive to acoustic noise, air currents over free-space paths, temperature gradients, and fiber-length changes in the distribution network. Enclosure, vibration isolation, temperature stabilization, and fiber-noise cancellation on any link longer than a few meters are standard practice rather than refinements.
Selecting a Comb
The choice follows from the application rather than from any ranking of technologies. Absolute optical frequency measurement and clock work require a fully self-referenced comb, which in practice means a fiber or titanium-sapphire system with an f-2f interferometer. Broadband molecular spectroscopy favors a mid-infrared comb, generated by difference-frequency generation or an optical parametric oscillator, at a repetition rate low enough to sample absorption features finely. Wavelength-division multiplexing and microwave photonics favor a microcomb at tens of gigahertz, where line spacing matches the channel grid and per-line power is high. Field deployment favors the fiber comb for its robustness, or an integrated microcomb where size, weight, and power dominate and self-referencing is not required. Where the repetition rate is the product, as in low-noise microwave generation, the comb's own noise floor and the photodetection chain, not the optical bandwidth, set the achievable result.
Trends and Outlook
The clear direction is integration. Heterogeneous and hybrid assembly is bringing pump lasers, silicon nitride resonators, modulators, nonlinear frequency-conversion sections, and photodetectors onto a small number of co-packaged chips, with the goal of a self-referenced comb in a module of a few cubic centimeters. Turnkey soliton generation by self-injection locking removed the most awkward operational barrier, and photonic-damascene and similar processes have raised waveguide quality factors to the point where microwatt-scale threshold powers are realistic.
Parallel efforts extend combs into the ultraviolet and extreme ultraviolet through cavity-enhanced high-harmonic generation, which supports spectroscopy of transitions in highly charged ions and of the low-lying nuclear transition in thorium-229. In 2024 a vacuum-ultraviolet comb, generated by cavity-enhanced harmonic conversion of a comb locked to a strontium optical clock, excited that nuclear transition directly and measured its frequency ratio against the clock. That was the first coherent frequency link between a nuclear and an electronic transition, and it is the enabling step for a nuclear clock. Efficiency work targets the low conversion of bright-soliton microcombs through dark-pulse and coupled-resonator designs. Space qualification continues, driven by optical clocks in orbit, satellite ranging, and greenhouse-gas monitoring. The 27th General Conference on Weights and Measures adopted a roadmap in 2022 whose preferred scenario redefines the SI second in terms of an optical transition at the 2030 conference, conditional on criteria covering the reliability of optical frequency standards and of the frequency links between the laboratories that operate them. Combs sit inside both criteria, which is what moves the technology from a specialist instrument toward a load-bearing element of the international measurement system.
Conclusion
The optical frequency comb converts a spectrum into a ruler. Two radio-frequency numbers, the repetition rate and the carrier-envelope offset, fix the absolute frequency of every line, and self-referencing through an f-2f interferometer measures the second of those numbers using nothing but the comb itself. That structure links optical and microwave frequencies phase-coherently in both directions, which is why one instrument serves optical clocks, low-noise microwave synthesis, broadband spectroscopy, absolute ranging, and astronomical calibration alike.
Two source families now share the field. Mode-locked fiber and titanium-sapphire combs offer the widest spectra, the lowest noise, and mature self-referencing, and they remain the standard for absolute metrology. Microresonator Kerr combs, operating in the bright-soliton regime or in the normal-dispersion dark-pulse states that trade bandwidth for conversion efficiency, offer chip-scale size, large line spacing, and manufacturability, and they are steadily closing the gap in coherence and bandwidth. As integration matures, the comb is completing the same transition that the laser made before it, from a delicate laboratory apparatus into an engineered component that designers specify and deploy.