Coherent-Harmonic Dual-Comb Spectroscopy

• The paper demonstrates that coherent-harmonic enhancement in dual-comb spectroscopy increases noise tracking bandwidth by generating multiple phase-structured harmonic centerbursts in a single interferogram.
• It adapts optical frequency comb formation through precise phase modulation to recover spectral modes and significantly boost single-shot SNR, achieving over a 300-fold reduction in averaging time.
• The technique decouples tracking bandwidth from spectral resolution limits, enabling robust free-running correction and accurate FID measurements, as shown in H13C14N experiments.

Searching arXiv for the main paper and closely related dual-comb/coherent-harmonic work to ground the article. arXiv search query: "Coherent-harmonic dual-comb spectroscopy (Long et al., 24 Jul 2025)" Coherent-Harmonic-Enhanced Dual-Comb Spectroscopy (CH-EDCS) is a dual-comb spectroscopic architecture in which coherent-harmonic optical frequency combs are used to increase the information content and effective signal strength of each interferogram without sacrificing the mode spacing that sets spectral resolution. In the free-running formulation reported in 2025, the method deliberately generates multiple phase-structured “harmonic centerbursts” within a single interferogram period, so that self-correction algorithms sample timing and phase noise more densely than in conventional dual-comb interferometry (Long et al., 24 Jul 2025). In the closely related spectral-mode-enhancement formulation, coherent-harmonic pulse trains coherently amplify shared spectral modes, increasing single-shot signal-to-noise ratio (SNR) and reducing the need for long coherent averaging (Long et al., 14 Apr 2025). The method belongs to the broader class of coherent dual-comb techniques in which mutual phase stability is exploited to preserve comb-tooth mapping, recover free-induction decay (FID) information, and, in some implementations, measure the full complex response of a sample (Coddington et al., 2010).

1. Position within coherent dual-comb spectroscopy

Dual-comb spectroscopy (DCS) uses two optical frequency combs with slightly different repetition rates so that optical comb teeth are mapped into a radio-frequency (RF) comb by multiheterodyne detection. In early coherent implementations, two phase-coherent, frequency-stabilized combs were used to measure the fully normalized, complex response of hydrogen cyanide over a 9 THz bandwidth at 220 MHz frequency resolution, yielding 41,000 resolution elements, with an average spectral SNR of 2,500 for both the fractional absorption and the phase and a peak SNR of 4,000 (Coddington et al., 2010). This established the core coherent-DCS paradigm: direct recovery of amplitude and phase, absolute frequency accuracy inherited from the combs, and coherent averaging over many interferograms.

Coherent-harmonic enhancement addresses a different bottleneck. Rather than redefining the optical-to-RF mapping itself, it modifies the temporal and spectral structure of the combs so that the interferogram contains more usable correction features and, in some architectures, intrinsically stronger spectral modes. This places CH-EDCS alongside other coherence-preserving extensions of DCS, rather than outside the standard dual-comb framework.

A related branch is coherent cavity-enhanced dual-comb spectroscopy (CE-DCS), where the probe comb is transmitted through an enhancement cavity and the local oscillator does not pass through the cavity. Fleisher, Long, and Hodges developed the corresponding complex-valued cavity transmission model and used it to analyze amplitude and phase spectra of CO and CO$_2$ with a cavity finesse of $F=19600$ (Fleisher et al., 2018). That work clarified how coherent DCS can be treated as a field measurement. CH-EDCS inherits that same emphasis on preserving coherent phase information, but directs it toward free-running correction bandwidth and per-interferogram SNR.

2. Conventional free-running limitation

Standard free-running dual-comb spectroscopy and interferometry are limited by a sampling constraint in the self-correction stage. In a free-running system, the relative timing jitter, carrier-envelope phase drift, and carrier-frequency drift between the two combs are not removed by active locking; they are estimated from the measured interferogram stream and digitally compensated afterward. In conventional mode-locked dual-comb interferometry, essentially all useful phase information is concentrated into one strong centerburst per interferogram period, while the remainder of the interferogram is comparatively noisy (Long et al., 24 Jul 2025).

Because the self-correction algorithm then receives only one effective sampling point per period $1/\Delta f_s$, where $\Delta f_s$ is the repetition-rate difference, Nyquist reasoning limits the phase-noise tracking bandwidth to about $\Delta f_s/2$. Noise above that limit is undersampled and folds into the estimate, corrupting the correction. This is the central obstacle for free-running DCS: the interferogram refresh rate is often too slow to follow the noise that must be tracked in real time.

The same parameter $\Delta f_s$ also enters the standard trade-off between tracking bandwidth and optical spectral coverage. The 2025 free-running CH-EDCS paper writes the conventional constraint as

$$\Delta \nu \le \frac{f_{rep}^2}{2\Delta f_s} = \frac{f_{rep}^2}{4BW_{tracking}},$$

so increasing tracking bandwidth by increasing $\Delta f_s$ narrows the usable spectral acquisition bandwidth for fixed comb spacing $f_{rep}$ (Long et al., 24 Jul 2025). This coupling is especially damaging in spectroscopy because the scientifically important information is often carried by the FID rather than the centerburst alone. If high-frequency noise is not corrected faithfully, the FID degrades and narrow spectral features broaden or distort.

3. Coherent-harmonic comb formation and harmonic centerbursts

The defining intervention in CH-EDCS is the creation of a coherent-harmonic optical frequency comb (H-OFC). In the spectral-mode-enhancement description, an H-OFC is modeled as a superposition of $m$ sub-harmonic pulse trains. For the illustrative case $F=19600$0, the pulse train is viewed as four interleaved pulse sequences with period $F=19600$1, and their relative carrier-envelope phases are controlled by a temporal phase modulation pattern such as $F=19600$2. The phase of each sub-mode is written as

$F=19600$3

with $F=19600$4 and $F=19600$5 (Long et al., 14 Apr 2025).

The purpose of that phase engineering is to recover spectral modes that would otherwise interfere destructively. In the $F=19600$6 example, the coherent vector sum gives

$F=19600$7

so the total comb retains the same mode spacing while the common mode amplitude is enhanced. In the general case, the line-amplitude enhancement is $F=19600$8, and a quadratic temporal phase modulation satisfying

$F=19600$9

can produce flat coherent enhancement over the comb (Long et al., 14 Apr 2025).

In the free-running CH-EDCS implementation, one comb is deliberately converted into such a coherent-harmonic comb by electro-optic modulation. The paper describes an architecture in which one comb is operated at a higher harmonic repetition structure, for example Comb 2 at $1/\Delta f_s$0, while the other is at $1/\Delta f_s$1. Comb 2 is phase-modulated by an electro-optic phase modulator driven by a periodic waveform with phase sequence $1/\Delta f_s$2 and period $1/\Delta f_s$3. The resulting asynchronous sampling no longer produces only one centerburst per $1/\Delta f_s$4; it produces multiple centerbursts within the same interferogram period, each carrying the prescribed phase structure. These are the “harmonic centerbursts” (Long et al., 24 Jul 2025).

4. Scaling laws: tracking bandwidth and spectral mode enhancement

The central scaling law of free-running CH-EDCS is that the number of effective noise samples per interferogram period is multiplied by the harmonic number $1/\Delta f_s$5. In the ambiguity-function framework emphasized in the 2025 paper, the self-correction algorithm normally estimates timing jitter and carrier-frequency drift by comparing consecutive centerbursts with a reference. Because CH-EDCS generates $1/\Delta f_s$6 centerbursts per fundamental interferogram interval, the noise-tracking speed and bandwidth increase to

$1/\Delta f_s$7

respectively (Long et al., 24 Jul 2025).

The corresponding bandwidth relation becomes

$1/\Delta f_s$8

or equivalently

$1/\Delta f_s$9

The practical implication stated in the paper is that tracking bandwidth is multiplied by $\Delta f_s$0, but the spectral acquisition bandwidth is not forced to shrink in the same way because the system generates multiple centerbursts within one interferogram period instead of increasing the repetition-rate difference itself (Long et al., 24 Jul 2025).

A parallel scaling law governs SNR in the spectral-mode-enhancement framework. The 2025 SNR paper distinguishes two architectures: C-H-DCS, in which one conventional comb beats against one coherent-harmonic comb, and H-H-DCS, in which two coherent-harmonic combs beat against each other. In C-H-DCS, the mode-amplitude gain is about $\Delta f_s$1. In H-H-DCS, the fundamental centerburst amplitude is enhanced by a factor of $\Delta f_s$2, yielding an SNR gain of $\Delta f_s$3 for a single acquisition. The paper states that this is equivalent to coherent averaging over $\Delta f_s$4 conventional interferograms: $\Delta f_s$5 For $\Delta f_s$6, the ideal equivalence is $\Delta f_s$7 (Long et al., 14 Apr 2025).

These two scaling laws address different but related limits. The bandwidth law concerns how rapidly a free-running system can estimate and remove phase noise. The SNR law concerns how much useful signal each interferogram carries before averaging. In both cases, coherent-harmonic structure is used to decouple a desired improvement from a parameter that ordinarily imposes a penalty, namely $\Delta f_s$8 in free-running correction and $\Delta f_s$9 in high-repetition-rate operation.

5. Experimental demonstrations on $\Delta f_s/2$0

The spectral-mode-enhancement proof of concept used two commercial 250 MHz fiber mode-locked combs at 1550 nm. The combs were operated with a 20th-harmonic repetition rate of about 250 MHz while preserving an actual comb spacing of $\Delta f_s/2$1 MHz, and the repetition-rate difference was $\Delta f_s/2$2. The system probed a 10 Torr $\Delta f_s/2$3 gas cell of 16.5 cm path length. The measured interferograms showed the expected progression from one centerburst per period in conventional DCS to repeated centerbursts in C-H-DCS and strong amplitude enhancement in H-H-DCS. The reported outcome was a $\Delta f_s/2$4-fold reduction in averaging time for comparable SNR in conventional DCS, specifically about $\Delta f_s/2$5, together with an $\Delta f_s/2$6-fold single-shot SNR improvement. The paper further reported that a 5 ms H-H-DCS trace gave residuals comparable to a 1.99 s conventional trace at the same 12.5 MHz sampling (Long et al., 14 Apr 2025).

The free-running bandwidth-extension experiment also used $\Delta f_s/2$7, around 1550 nm, but targeted self-correction performance rather than single-shot SNR. Conventional and coherent-harmonic architectures were compared at the same spectral bandwidth, about $\Delta f_s/2$8 at 12 dB, and the same spectral resolution, $\Delta f_s/2$9. The repetition-rate difference was set to $\Delta f_s$0, yielding a 1 ms interferogram period. One optical frequency comb was pulse-picked with an electro-optic intensity modulator to emulate a 12.5 MHz low-repetition-rate comb; in the CH-EDCS configuration the second comb was phase-modulated to create the coherent-harmonic structure, and replacing that phase modulator with another intensity modulator recovered the conventional architecture for comparison (Long et al., 24 Jul 2025).

Cross-ambiguity-function processing extracted timing jitter and carrier-frequency drift over a 50 ms window. In both architectures, carrier-frequency drift was on the order of 1 MHz, consistent with free-running comb linewidths, but the CH-EDCS trace exhibited much denser sampling. In the power spectral density of the estimated jitter and drift, the CH-EDCS system captured noise out to below 10 kHz, including peaks between 0.5 and 1 kHz, whereas the conventional architecture was limited to about 0.5 kHz and showed evidence of high-frequency folding and undersampling. Downsampling the CH-EDCS estimate by a factor of 20 reproduced the conventional power spectral densities, confirming that the reported 20-fold improvement came from the higher sampling rate of the harmonic centerbursts (Long et al., 24 Jul 2025).

After correction and coherent averaging over 1.999 s, both architectures produced corrected interferograms and spectra, but the CH-EDCS interferogram showed a longer FID. In the frequency domain, the corrected spectra retained the same 1 kHz comb-line spacing, and unlike a conventional attempt to raise tracking bandwidth by increasing $\Delta f_s$1 from 1 kHz to 20 kHz, CH-EDCS preserved the same RF spectral bandwidth and avoided spectral narrowing and aliasing near $\Delta f_s$2 Hz or $\Delta f_s$3. The CH-EDCS absorption lines were narrower and deeper, closer to the standard continuous-wave-referenced measurement. The paper attributes the higher SNR in the raw CH-EDCS spectra to the additional centerbursts, which theoretically provide a $\Delta f_s$4-fold mode-amplitude enhancement in the earlier spectral-mode-enhancement framework (Long et al., 24 Jul 2025).

6. Relation to self-correction methods, terminology, and broader architectures

A notable feature of CH-EDCS is that its main intervention is at the hardware level, not in a single correction algorithm. The 2025 free-running paper states that because the method increases the effective centerburst sampling rate without changing spectral resolution, it can improve not only cross-ambiguity-function correction but also other self-correction methods such as STFT, CFD, EKF, and CoCoA. For discrete-burst estimators, more centerbursts directly provide better temporal sampling; for continuous estimators, the added coherent structure improves estimator accuracy and SNR. The same paper also emphasizes that the method is especially relevant for FID correction, because the scientifically important absorption information lies in the FID rather than only in the centerburst (Long et al., 24 Jul 2025).

CH-EDCS is complementary to other strategies that improve coherence at the source level. In reference-free DCS with inbuilt coherence, the two combs share a common comb line through seeded parametric down-conversion, so the carrier-envelope offset difference is essentially eliminated and only small residual timing drifts need numerical correction; that system enabled coherent averaging of nearly 1 million interferograms (Roiz et al., 2024). In a distinct free-running approach, a single-cavity gigahertz dual-comb laser with highly correlated fluctuations permitted long-timescale coherent averaging from interferogram-derived phase information alone, with more than 3 W average power per comb, 78 fs pulse duration, a 1.0327 GHz repetition rate, and a continuously tunable repetition-rate difference up to 27 kHz (Phillips et al., 2022). These source architectures reduce phase instability, whereas CH-EDCS modifies the interferogram structure seen by the estimator.

The term “harmonic” requires careful interpretation. In CH-EDCS proper, the relevant mechanism is coherent-harmonic comb formation and the generation of harmonic centerbursts. By contrast, a 2025 dual-comb platform bridging the mid-infrared and terahertz domains explicitly stated that its “harmonic” aspect was the use of second-harmonic generation to create a near-infrared gate comb and the use of second-order nonlinear processes in GaSe for both intrapulse difference-frequency generation and electro-optic sampling; it was not a harmonic-comb experiment in the sense of generating higher-order harmonics for spectroscopy (Konnov et al., 29 Sep 2025). This distinction matters because “harmonic-enhanced” in dual-comb work can refer either to coherent-harmonic pulse engineering, as in CH-EDCS, or to $\Delta f_s$5-based frequency conversion in a broader nonlinear-optical system.

A broader implication, stated explicitly in the spectral-mode-enhancement study, is that the averaging-time reduction should become even more dramatic with ultra-high-repetition-rate combs such as microresonator combs; that paper estimated that using 22 GHz microresonator combs at 500 MHz spectral resolution could yield more than a $\Delta f_s$6 averaging-time reduction relative to conventional DCS at the same resolution, and it also suggested integrated lithium niobate modulators as a practical route to implementation (Long et al., 14 Apr 2025). This suggests that coherent-harmonic enhancement is not confined to one source class or one wavelength region, but is better understood as a general strategy for redistributing temporal and spectral coherence in dual-comb measurements.