Frequency Phase Transfer (FPT)

• Frequency Phase Transfer (FPT) is a suite of techniques that preserves phase coherence by transferring phase information from one channel to another.
• It is implemented through methods such as passive branching in optical fibers, comb-based transfers, and lower-to-higher frequency referencing in VLBI to improve stability.
• The approach extends coherence times and enhances measurement precision, enabling advances in optical metrology, high-resolution imaging, and precision astrometry.

Searching arXiv for the cited FPT papers and closely related work to ground the article. Searching for "Frequency Phase Transfer" and related terms on arXiv. Frequency Phase Transfer (FPT) denotes a set of phase-coherent transfer and calibration techniques in which phase information measured in one channel, band, wavelength, or propagation path is used to preserve or reconstruct coherence elsewhere. In optical metrology, the term covers delivery of a phase-stable optical carrier through fiber, free space, or an optical frequency comb. In millimeter and sub-millimeter VLBI, it denotes the use of simultaneous lower-frequency phase solutions to calibrate higher-frequency visibility phases. In all of these usages, the central premise is that a dominant phase perturbation is either common or predictably scalable, so that it can be canceled or transferred without destroying the desired phase observable (Xue et al., 2021, Peleg et al., 2019, Rioja et al., 2015).

1. Definition and common mathematical structure

In the cited literature, FPT is not a single protocol but a class of related procedures. Optical-link FPT treats the optical carrier itself as the object whose phase must be preserved across a noisy transmission path. Comb-based FPT transfers the phase stability of a master laser at one wavelength to another wavelength through the comb relation. VLBI FPT uses a lower observing band as a phase reference for a higher band because the dominant atmospheric term is non-dispersive in delay and therefore linear in frequency (Xue et al., 2021, Peleg et al., 2019, Rioja et al., 2015).

Regime	Transferred quantity	Representative relation
Optical fiber or free-space transfer	Optical carrier phase	$\phi(t)=\frac{2\pi \nu}{c} n(t)L(t)$
Optical frequency comb transfer	Comb tooth frequency	$f_n=n f_{\mathrm{rep}}+f_{\mathrm{ceo}}$
mm/sub-mm VLBI	High-band phase calibration	$$\phi_{\mathrm{FPT}}(t)=\phi(\nu_h,t)-R\,\phi(\nu_l,t)$$

For optical path transfer, environmental perturbations drive fluctuations in path length and refractive index, producing

$$\delta \phi(t)=\frac{2\pi \nu}{c}\big[n\,\delta L(t)+L\,\delta n(t)\big],$$

which map to fractional frequency fluctuations through

$y(t)=\frac{1}{2\pi \nu}\frac{d\phi(t)}{dt}.$

The paper on branching optical transfer further emphasizes temperature as a dominant coupling and gives explicit temperature-to-phase relations for short out-of-loop fibers (Xue et al., 2021).

For VLBI, the same linearity appears in a different form. The observed residual visibility phase is decomposed into source, geometric, tropospheric, ionospheric, instrumental, and ambiguity terms, with the non-dispersive components obeying $\phi=2\pi \nu \tau$. If $\nu_l$ and $\nu_h$ are observed simultaneously, then scaling by $R=\nu_h/\nu_l$ cancels the geometric and tropospheric terms to first order, whereas dispersive ionospheric and inter-band instrumental terms remain (Rioja et al., 2015).

For comb-based optical transfer, the relevant structure is the rigid correlation between comb teeth through $f_n=n f_{\mathrm{rep}}+f_{\mathrm{ceo}}$. When the comb is tightly locked to a stable optical reference, the optical phase noise of the reference is inherited by other teeth, allowing a target laser at a different wavelength to be locked to a chosen tooth with a fixed offset (Peleg et al., 2019).

2. Optical-metrology realizations

A fiber-network realization of FPT is given by the passive branching scheme of "Branching optical frequency transfer with enhanced post automatic phase noise cancellation" (Xue et al., 2021). There, a master laser near 193 THz is split into two branches of 50 km and 145 km. Each remote site contains a Faraday mirror, one fiber-pigtailed acousto-optic modulator (AOM), photodetection, and RF processing. The AOM simultaneously provides a branch-specific frequency tag and the optical actuation needed for passive phase cancellation. The single-pass and triple-pass fields are compared locally, yielding a two-way beat from which the remote site imposes the equal-and-opposite phase. No active phase-lock servo is required at the master site. The work also distinguishes “outside-loop” noise, originating in components not sensed by the cancellation loop, from “out-of-loop” residuals in the validation path. By incorporating the remote AOM into the stabilized loop and by actively stabilizing the temperature of the interferometer enclosure, the reported back-to-back stability reaches $f_n=n f_{\mathrm{rep}}+f_{\mathrm{ceo}}$0 and $f_n=n f_{\mathrm{rep}}+f_{\mathrm{ceo}}$1 at $f_n=n f_{\mathrm{rep}}+f_{\mathrm{ceo}}$2. After passive compensation, the 145 km branch reaches $f_n=n f_{\mathrm{rep}}+f_{\mathrm{ceo}}$3 at 1 s and $f_n=n f_{\mathrm{rep}}+f_{\mathrm{ceo}}$4 at $f_n=n f_{\mathrm{rep}}+f_{\mathrm{ceo}}$5, while the 50 km branch reaches $f_n=n f_{\mathrm{rep}}+f_{\mathrm{ceo}}$6 at 1 s and $f_n=n f_{\mathrm{rep}}+f_{\mathrm{ceo}}$7 at $f_n=n f_{\mathrm{rep}}+f_{\mathrm{ceo}}$8 (Xue et al., 2021).

A free-space realization uses the same two-way idea but embeds the open-air path in an imbalanced Michelson interferometer (Gozzard et al., 2018). A 193 THz continuous-wave optical signal is transferred over folded 150 m and 600 m links, with an in-loop heterodyne beat at 240 MHz used to drive an AOM that corrects the transmit phase. Over 600 m, the stabilized link achieves a fractional frequency stability of $f_n=n f_{\mathrm{rep}}+f_{\mathrm{ceo}}$9 at 1 s and $$\phi_{\mathrm{FPT}}(t)=\phi(\nu_h,t)-R\,\phi(\nu_l,t)$$0 at 64 s. The dominant practical limitation is not the phase servo itself but deep fading caused by atmospheric turbulence, beam wander, and angle-of-arrival jitter. The same study estimates that, for a vertical link to low Earth orbit, the 1 s stability should remain below $$\phi_{\mathrm{FPT}}(t)=\phi(\nu_h,t)-R\,\phi(\nu_l,t)$$1, with the main bandwidth limit set by round-trip delay rather than by distributed fiber noise (Gozzard et al., 2018).

Comb-based FPT transfers phase stability across the optical spectrum rather than across a physical path. In "Phase stability transfer across the optical domain using a commercial optical frequency comb" (Peleg et al., 2019), a cavity-stabilized 1560 nm laser is transferred to 674 nm through an erbium-doped fiber comb. The target 674 nm diode laser is locked to a comb tooth with fixed offset,

$$\phi_{\mathrm{FPT}}(t)=\phi(\nu_h,t)-R\,\phi(\nu_l,t)$$2

yielding a fast linewidth of 19 Hz by heterodyne comparison and 16 Hz from Ramsey-MFDD spectroscopy, with fractional instability estimated to be $$\phi_{\mathrm{FPT}}(t)=\phi(\nu_h,t)-R\,\phi(\nu_l,t)$$3 at 1 s (Peleg et al., 2019). A related transfer-oscillator realization suppresses comb-noise contributions directly in the RF combination used for the slave lock, reaching a comb-noise cancellation bandwidth above 1.8 MHz and enabling 99% fidelity for a $$\phi_{\mathrm{FPT}}(t)=\phi(\nu_h,t)-R\,\phi(\nu_l,t)$$4 $$\phi_{\mathrm{FPT}}(t)=\phi(\nu_h,t)-R\,\phi(\nu_l,t)$$5 pulse on a trapped $$\phi_{\mathrm{FPT}}(t)=\phi(\nu_h,t)-R\,\phi(\nu_l,t)$$6 ion without pre-stabilizing the slave diode laser to a cavity (Scharnhorst et al., 2015). A recurring limitation in these comb implementations is that servo bumps and residual RF-path delay mismatch can be transferred along with the desired coherence if they remain inside the target lock bandwidth (Peleg et al., 2019, Scharnhorst et al., 2015).

3. FPT in mm/sub-mm VLBI

In VLBI, FPT is a calibration method rather than a carrier-distribution method. The residual phase on a baseline is written as

$$\phi_{\mathrm{FPT}}(t)=\phi(\nu_h,t)-R\,\phi(\nu_l,t)$$7

and the FPT observable is

$$\phi_{\mathrm{FPT}}(t)=\phi(\nu_h,t)-R\,\phi(\nu_l,t)$$8

Because the neutral troposphere and geometric delay are non-dispersive, their phase scales linearly with $$\phi_{\mathrm{FPT}}(t)=\phi(\nu_h,t)-R\,\phi(\nu_l,t)$$9 and cancels under simultaneous observations. The ionospheric term does not cancel because it is dispersive, approximately $$\delta \phi(t)=\frac{2\pi \nu}{c}\big[n\,\delta L(t)+L\,\delta n(t)\big],$$0. Integer frequency ratios are therefore highly advantageous, since they also eliminate residual $$\delta \phi(t)=\frac{2\pi \nu}{c}\big[n\,\delta L(t)+L\,\delta n(t)\big],$$1 ambiguity terms under consistent wrap tracking (Rioja et al., 2015).

The canonical experimental demonstration is the Korean VLBI Network (KVN) multi-band system operating simultaneously at 22, 43, 87, and 130 GHz (Rioja et al., 2015). Using FPT and then Source Frequency Phase Referencing (SFPR), observations of five AGNs showed that coherence at 130 GHz increased from a few tens of seconds to about twenty minutes with FPT, and to many hours with SFPR. The same work reported the first robust inter-frequency astrometry at 130 GHz and emphasized that lower reference bands with smaller ionospheric residuals, such as 43 GHz for 130 GHz calibration, outperform 22 GHz in practice (Rioja et al., 2015).

Subsequent VLBI work frames FPT as a core capability of simultaneous shared-optical-path (SOP) receivers. The 2023 science overview for multiband VLBI states that KVN routine operation has already demonstrated coherence-time extensions of more than two orders of magnitude and residual phase noise of about $$\delta \phi(t)=\frac{2\pi \nu}{c}\big[n\,\delta L(t)+L\,\delta n(t)\big],$$2 at 129 GHz, with SFPR relative astrometry around $$\delta \phi(t)=\frac{2\pi \nu}{c}\big[n\,\delta L(t)+L\,\delta n(t)\big],$$3. The same review describes emerging 22/43/86 GHz compact triple-band SOP receivers in Europe and other developments in Asia and Australia, aiming at a global FPT network with order-of-magnitude sensitivity and dynamic-range gains at 86 GHz and astrometry at the level of one microsecond of arc (Dodson et al., 2023).

For the ngEHT, FPT is treated as the first step of a broader calibration architecture. Simultaneous observations at 85 or 110 GHz together with 220/255/330/340 GHz permit the high bands to inherit the non-dispersive calibration from the low band. The cited simulations and forecasts state that coherence at 230 GHz without FPT is typically around 10 s, whereas FPT can increase coherence time by more than 100 folds and extend it to hour(s) in simulations. The same studies quote a high-band detection threshold of about 10 mJy set by low-band sensitivity, and astrometric precision around $$\delta \phi(t)=\frac{2\pi \nu}{c}\big[n\,\delta L(t)+L\,\delta n(t)\big],$$4 for a target–calibrator separation of about $$\delta \phi(t)=\frac{2\pi \nu}{c}\big[n\,\delta L(t)+L\,\delta n(t)\big],$$5, improving to about $$\delta \phi(t)=\frac{2\pi \nu}{c}\big[n\,\delta L(t)+L\,\delta n(t)\big],$$6 in an in-beam configuration (Jiang et al., 2022). Complementary ngEHT simulations further argue that an 85 GHz anchor band is optimal because the integer ratios $$\delta \phi(t)=\frac{2\pi \nu}{c}\big[n\,\delta L(t)+L\,\delta n(t)\big],$$7 and $$\delta \phi(t)=\frac{2\pi \nu}{c}\big[n\,\delta L(t)+L\,\delta n(t)\big],$$8 avoid ambiguity problems and are more robust than non-integer pairs such as $$\delta \phi(t)=\frac{2\pi \nu}{c}\big[n\,\delta L(t)+L\,\delta n(t)\big],$$9 (Rioja et al., 2023).

4. Extensions beyond basic FPT

FPT alone removes the fast non-dispersive terms, but it does not generally remove ionospheric or inter-band instrumental residuals. SFPR addresses this by adding a second, inter-source calibration step. In the formulation used for ngEHT and earlier KVN work, FPT is applied on the target first, and then slow nodding to a nearby calibrator removes the remaining dispersive ionospheric and instrumental terms while preserving the inter-frequency astrometric signature. This is why FPT alone is sufficient when the goal is coherence and sensitivity, whereas SFPR is required when astrometry is the goal (Jiang et al., 2022).

A more algebraic extension is FPT-square, introduced for simultaneous 21.5/43/86 GHz KVN data (Zhao et al., 2017). After ordinary FPT leaves an ionospheric residual proportional to $y(t)=\frac{1}{2\pi \nu}\frac{d\phi(t)}{dt}.$0, a second linear combination of two FPT residuals removes the ionosphere by exploiting its $y(t)=\frac{1}{2\pi \nu}\frac{d\phi(t)}{dt}.$1 dependence. In that study, coherence time at 86 GHz was extended beyond 8 hours, with peak-flux recovery remaining above 88% for an 8-hour solution interval. Source-to-source transfer after FPT-square produced 88% recovery for the 10.4° pair 3C 279 → 3C 273 and 97% recovery for the 20.2° pair OJ 287 → 4C 39.25, and the authors further reported all-sky calibration using a small set of calibrators for separations up to about $y(t)=\frac{1}{2\pi \nu}\frac{d\phi(t)}{dt}.$2 (Zhao et al., 2017). The same paper notes a limitation that is often overlooked: FPT-square does not preserve the highest-band astrometry, because the residual source term is a weighted combination of core shifts rather than a single inter-frequency offset (Zhao et al., 2017).

A separate algorithmic issue concerns non-integer frequency ratios. Conventional FPT leaves deterministic jump discontinuities whenever the low-band phase wraps if $y(t)=\frac{1}{2\pi \nu}\frac{d\phi(t)}{dt}.$3 is not an integer. The 2026 paper "Frequency Phase Transfer for Future Millimetre Arrays with Arbitrary Frequency Ratios" derives this explicitly by decomposing $y(t)=\frac{1}{2\pi \nu}\frac{d\phi(t)}{dt}.$4 and showing that the problematic term is

$y(t)=\frac{1}{2\pi \nu}\frac{d\phi(t)}{dt}.$5

so each low-band wrap induces a jump of $y(t)=\frac{1}{2\pi \nu}\frac{d\phi(t)}{dt}.$6 at the high band (Simelane et al., 25 Jun 2026). Its phase-wrap counting (PWC) algorithm tracks the low-band wrap index $y(t)=\frac{1}{2\pi \nu}\frac{d\phi(t)}{dt}.$7 and adds the correction $y(t)=\frac{1}{2\pi \nu}\frac{d\phi(t)}{dt}.$8, yielding a formulation that works for arbitrary $y(t)=\frac{1}{2\pi \nu}\frac{d\phi(t)}{dt}.$9 provided the reference-band phase tracking satisfies the Itoh condition. Implemented in the HITOPS software package, PWC improved coherence on a simulated 86/230 GHz EHT dataset by 143% over raw phases and by 5% over conventional fringe-fit plus self-calibration on the IRAM–NOEMA baseline; across baselines, the improvement over FF+SC ranged from 3% to 50%, and the fractional peak flux recovery improved from $\phi=2\pi \nu \tau$0 to $\phi=2\pi \nu \tau$1 (Simelane et al., 25 Jun 2026).

5. Demonstrated performance and limiting factors

The experimental literature spans optical links, free-space links, and VLBI baselines of continental and Earth-sized scale.

Demonstration	Configuration	Reported outcome
Passive branching optical transfer	50 km and 145 km fiber branches	$\phi=2\pi \nu \tau$2 at 1 s and $\phi=2\pi \nu \tau$3 at 10,000 s on 145 km
Stabilized free-space optical transfer	600 m at 193 THz	$\phi=2\pi \nu \tau$4 at 1 s and $\phi=2\pi \nu \tau$5 at 64 s
Optical-domain comb transfer	1560 nm to 674 nm	19 Hz and 16 Hz linewidth measurements
KVN mm-VLBI FPT	22/43/87/130 GHz	130 GHz coherence increased to about twenty minutes
Earth-sized 3 mm to 1 mm FPT	86 to 215 GHz Spain–Hawai‘i baseline	systematic increase of 215 GHz coherence on all averaging timescales
SOP intercontinental FPT	86.012 to 258.036 GHz, 8,623 km	$\phi=2\pi \nu \tau$6 to about 1 minute and about 0.8 at 5.5 min after FPT

In fiber-transfer experiments, the principal hard limit is propagation delay. For the passive branching system, the residual phase-noise PSD obeys

$\phi=2\pi \nu \tau$7

with effective correction bandwidth

$\phi=2\pi \nu \tau$8

giving about 195 Hz for 145 km and 567 Hz for 50 km. Outside-loop noise is mitigated by bringing more components inside the stabilized path; out-of-loop drift is mitigated separately through temperature stabilization of the comparison interferometer (Xue et al., 2021).

At 1 mm VLBI, the first Earth-sized demonstration used simultaneous 86 and 215 GHz observations in January 2024 on baselines between the IRAM 30 m in Spain and the JCMT and SMA in Hawai‘i. Strong detections were obtained on J0958+6533 and OJ287 at both frequencies, a strong correlation between the interferometric phases at the two frequencies was observed, and FPT systematically increased the 215 GHz coherence on all averaging timescales. The same experiment also demonstrated paired-antenna FPT using co-located JCMT and SMA as a single dual-frequency station (Issaoun et al., 26 Feb 2025).

A subsequent SOP demonstration pushed this to 258.036 GHz with an integer ratio $\phi=2\pi \nu \tau$9 on the 8,623 km APEX–IRAM 30 m baseline (Zhao et al., 14 Jul 2025). After FPT, a representative 258 GHz scan on CTA 102 had coherence factor $\nu_l$0 up to about 1 minute and about 0.8 over the full 5.5 minute scan, whereas without FPT $\nu_l$1 fell below 0.9 at about 10 s and to about 0.6 at about 20 s. The same study reports reductions of phase fluctuation range from 26 rad to 1.2 rad after sunset and a reduction of scan-averaged fringe rate from 13 mHz to 0.6 mHz. It also notes that weaker sources such as PKS B0420−014 and 3C 120 produced no robust 258 GHz fringes with single-band calibration but did after FPT (Zhao et al., 14 Jul 2025).

The limiting factors differ by domain but have a common structure. Free-space optical transfer is limited by scintillation-induced deep fades and cycle slips; the 600 m experiment reports fewer than one cycle slip per hour during the most stable period, but about one per second in rain (Gozzard et al., 2018). Comb-based optical transfer can be limited by servo bumps, grounding-related electrical lines, finite actuator bandwidth, and RF-path delay mismatch; the 1560→674 nm work shows pronounced servo features around 200 kHz, and the transfer-oscillator work identifies path-delay matching as the dominant high-frequency cancellation constraint (Peleg et al., 2019). In VLBI, the dominant practical failure modes are insufficient low-band SNR, residual ionosphere, inter-band instrumental offsets, phase-wrap errors for non-integer $\nu_l$2, and source-structure differences across bands (Rioja et al., 2015, Simelane et al., 25 Jun 2026).

6. Scientific uses and future directions

The scientific motivations for FPT differ across fields but all rely on extended coherence. In optical metrology, passive branching fiber transfer is intended for simultaneous dissemination of ultra-stable optical carriers to multiple independent users within a local area, including optical clock comparisons, timekeeping networks, and radio astronomy arrays requiring phase-synchronized references (Xue et al., 2021). Stabilized free-space transfer targets coherent optical communications, satellite Doppler ranging, optical clock transfer, tests of General Relativity, and other fundamental-physics applications (Gozzard et al., 2018). Comb-based optical-domain transfer enables precision spectroscopy and optical clock operation at a visible wavelength from a telecom-band reference without building an ultra-stable cavity at every wavelength (Peleg et al., 2019).

In VLBI, FPT extends from sensitivity enhancement to precision astrometry. The cited ngEHT studies identify Sgr A* and M87* as primary targets for SFPR-enabled core-shift registration between 86 and 345 GHz, and also discuss weaker SMBH targets such as M84, M104, and IC1459 whose detectability at the highest bands would benefit from low-band-assisted coherence (Jiang et al., 2022). The broader multiband VLBI perspective adds applications in maser astrometry, proper motions and parallaxes, dynamic imaging of black holes, transients, and binary supermassive black holes, and argues that a global 22/43/86 GHz FPT network of at least ten antennas could deliver order-of-magnitude sensitivity and dynamic-range improvements at 86 GHz together with astrometry at the level of about $\nu_l$3 (Dodson et al., 2023). The ngEHT-specific simulations further recommend simultaneous tri-band reception at 85, 230, and 340 GHz, with 85 GHz serving as an integer-ratio anchor for robust high-band calibration (Rioja et al., 2023).

A separate future direction is the use of dedicated instrumental references to support FPT. "A Novel Comb Generator for Frequency Phase Transfer" describes a millimeter-wave comb generator for the Black Hole Explorer mission concept, designed to inject phase-coherent multi-octave tones into 90 and 270 GHz receiver chains so that instrumental delay can be measured and tracked. In simultaneous dual-band tests at the Kitt Peak 12 m telescope, selected tones in the 3 mm and 1 mm bands showed about 99.8–99.9% coherence, directly supporting the use of comb-based instrumental calibration as a complement to astronomical FPT (Montano et al., 27 Nov 2025).

Taken together, these works show that FPT has evolved from a domain-specific calibration trick into a cross-disciplinary phase-coherence strategy. In optical links it suppresses environmentally induced phase noise; in optical comb systems it maps stability across the spectrum; in VLBI it converts simultaneous multi-band reception into longer coherence time, higher sensitivity, and, with SFPR-class extensions, astrometric registration at frequencies where conventional phase referencing is otherwise impractical (Xue et al., 2021, Peleg et al., 2019, Rioja et al., 2015).