Dual Interferometer Technique

• Dual interferometer technique is a method that uses two parallel interferometric channels to decouple observables and suppress common-mode noise.
• It combines phase outputs from opposing channels to separately extract acceleration and rotational signals, enhancing measurement sensitivity.
• The approach is applied across atom, optical, and fiber interferometry, offering improved noise rejection and extended dynamic range.

Across the cited literature, the expression “dual interferometer technique” denotes a family of interferometric architectures in which two interferometers, or two interferometric channels, are operated in parallel within a shared optical or matter-wave system and then combined algebraically to separate observables, suppress common-mode noise, extend non-ambiguity range, or enable distributed sensing. Its most explicit inertial-sensing form is the use of two oppositely directed light-pulse atom interferometers whose summed and differenced phases isolate acceleration and rotation (Rakholia et al., 2014). Closely related constructions appear in continuous dual-atom-beam sensors, dual-port Michelson readout, two-wavelength heterodyne metrology, dual-comb interferometry, and parallel entangled interferometers (Meng et al., 2022, Lohde et al., 2024, Zhang et al., 2022, Chen et al., 2017, Du et al., 8 Jan 2025).

1. Core interferometric principle

In its canonical atom-interferometric form, the technique starts from the first-order phase of a three-pulse Mach–Zehnder light-pulse interferometer,

$$\Delta \phi = \mathbf{k}_\text{eff} \cdot \left(\mathbf{a} - 2 \mathbf{v} \times \mathbf{\Omega}\right) T^2,$$

with effective Raman wavevector $\mathbf{k}_\text{eff}$, atomic velocity $\mathbf{v}$, acceleration $\mathbf{a}$, rotation $\mathbf{\Omega}$, and pulse spacing $T$. If two ensembles are launched with opposite velocities, $\mathbf{v}_a=+\mathbf{v}$ and $\mathbf{v}_b=-\mathbf{v}$, their phases $\phi_a$ and $\phi_b$ can be recombined as

$\mathbf{k}_\text{eff}$0

The common mode therefore yields acceleration, whereas the differential mode yields the Sagnac-type rotation signal (Rakholia et al., 2014).

A closely related algebra appears in the continuous dual-atom-interferometer sensor based on two continuous cold atomic beams. There, the two outputs are written as

$\mathbf{k}_\text{eff}$1

$\mathbf{k}_\text{eff}$2

so that sum and difference channels decouple acceleration and rotation, with the sum channel used for acceleration feedback and the difference channel used for rotation readout (Meng et al., 2022).

In optical interferometry the same structural idea reappears as quadrature recombination. In dual balanced readout of a Michelson interferometer, the combinations

$\mathbf{k}_\text{eff}$3

isolate arm-specific channels, enabling subtraction of scattered-light contamination from the displacement readout (Lohde et al., 2024). This suggests that “dual interferometer technique” is less a single geometry than a recurring linear-decomposition strategy.

2. Counter-propagating atom interferometers for inertial sensing

A compact realization was demonstrated in a small rectangular quartz cell with inner volume $\mathbf{k}_\text{eff}$4, containing two 3D MOTs located $\mathbf{k}_\text{eff}$5 mm apart. Each MOT loads from Rb vapor at $\mathbf{k}_\text{eff}$6 Torr and holds $\mathbf{k}_\text{eff}$7 atoms at the 60 Hz operating point. The ensembles are launched toward one another at $\mathbf{k}_\text{eff}$8 by differential detuning of the cooling beams, cooled to $\mathbf{k}_\text{eff}$9, interrogated in ballistic flight by a stimulated Raman $\mathbf{v}$0 sequence with $\mathbf{v}$1 ms, detected, and then recaptured in the opposite MOT. With a MOT time constant of about $\mathbf{v}$2 ms and net recapture efficiency of $\mathbf{v}$3–$\mathbf{v}$4, the cycle time is $\mathbf{v}$5 ms, corresponding to 60 Hz, and the architecture supports 50–100 measurements per second (Rakholia et al., 2014).

That instrument reports single-interferometer phase noise of $\mathbf{v}$6 mrad/shot, combined acceleration phase noise $\mathbf{v}$7 mrad/shot, and combined rotation phase noise $\mathbf{v}$8 mrad/shot. Using $\mathbf{v}$9 ms and $\mathbf{a}$0, the extracted sensitivities are approximately $\mathbf{a}$1 for acceleration and $\mathbf{a}$2 for rotation (Rakholia et al., 2014).

Large-area dual-atom-interferometer gyroscopes implement the same phase-separation logic with longer baselines and longer $\mathbf{a}$3. In one $\mathbf{a}$4 system, two symmetric MOTs launch atoms at $\mathbf{a}$5 and $\mathbf{a}$6 with respect to gravity; three pairs of Raman beams separated by 3 cm realize a Mach–Zehnder sequence with $\mathbf{a}$7 ms and enclosed area $\mathbf{a}$8. For a single interferometer the phase is written as

$\mathbf{a}$9

and the dual-interferometer differential phase becomes

$\mathbf{\Omega}$0

so gravity and laser phase terms are nominally common-mode while the Sagnac term is enhanced. The reported rotation sensitivity is $\mathbf{\Omega}$1, and the long-term stability is $\mathbf{\Omega}$2 at $\mathbf{\Omega}$3 s (Yao et al., 2017).

A more recent large-area implementation addresses a deeper systematic: Earth’s rotation breaks the ideal velocity immunity of the Mach–Zehnder geometry. In a dual-atom-interferometer gyroscope with $\mathbf{\Omega}$4 m, $\mathbf{\Omega}$5 ms, and interference area $\mathbf{\Omega}$6, dynamic mirror compensation actively rotates the Raman mirrors during the sequence so that the compensated phase reduces to the pure Sagnac form

$\mathbf{\Omega}$7

After compensation, the phase dependence on atomic velocity is reduced 40-fold, the velocity contribution to scale-factor stability is $\mathbf{\Omega}$8 ppm, the rotation sensitivity is $\mathbf{\Omega}$9, the stability reaches $T$0 at $T$1 s, and the common-mode noise rejection ratio reaches 459 during a seismic event (Gu et al., 28 May 2026).

3. Continuous and closed-loop dual-atom interferometers

A distinct branch replaces pulsed cold-atom cycling by continuous cold atomic beams. In a spatially separated Mach–Zehnder configuration, two continuous cold beams from dual $T$2 MOTs propagate in opposite directions through the same three Raman zones formed by a 3-slit mask. For mean beam velocity $T$3, slit spacing $T$4 mm gives $T$5 ms, and the interferometer area is $T$6. Doppler selection ensures one beam experiences $T$7 and the other $T$8, so acceleration and rotation are decoupled by sum and difference of the dual interferometer signals (Meng et al., 2022).

This continuous device adds a closed-loop layer. One of the $T$9 beams is mounted on a PZT, a sinusoidal phase modulation at 81 Hz is applied, the sum of the demodulated dual-interferometer outputs feeds a PID controller, and the Raman phase is adjusted so that the system remains at mid-fringe. The PID output then provides acceleration, while the difference channel yields rotation with acceleration already compensated. The amplitudes of the sum signal are $\mathbf{v}_a=+\mathbf{v}$0 times that of a single AI signal. The closed-loop dual-AI performance gives $\mathbf{v}_a=+\mathbf{v}$1 short-term and $\mathbf{v}_a=+\mathbf{v}$2 long-term stability for rotation, and after temperature compensation $\mathbf{v}_a=+\mathbf{v}$3 short-term and $\mathbf{v}_a=+\mathbf{v}$4 long-term stability for acceleration (Meng et al., 2022).

Field-compatible dual interferometers depend critically on laser stability. A modular laser system designed for a dual-atom-interferometer gyroscope uses three 1560 nm external-cavity diode lasers, EDFAs, PPLN doubling to 780 nm, and six all-quartz optical modules. At room temperature, laser power fluctuation is under $\mathbf{v}_a=+\mathbf{v}$5, polarization extinction ratio exceeds 30 dB, frequency fluctuation is below 91 kHz, and phase noise reaches $\mathbf{v}_a=+\mathbf{v}$6 at 1 kHz. The associated dual-atom-interferometer gyroscope uses $\mathbf{v}_a=+\mathbf{v}$7, moving-molasses launch velocities of 4 m/s horizontally and 0.49 m/s vertically, pulse durations $\mathbf{v}_a=+\mathbf{v}$8, and interrogation time $\mathbf{v}_a=+\mathbf{v}$9 ms. Fringe contrast reaches $\mathbf{v}_b=-\mathbf{v}$0 at $\mathbf{v}_b=-\mathbf{v}$1C, with measured values of $\mathbf{v}_b=-\mathbf{v}$2 at $\mathbf{v}_b=-\mathbf{v}$3C and $\mathbf{v}_b=-\mathbf{v}$4 at $\mathbf{v}_b=-\mathbf{v}$5C (Sun et al., 2024).

4. Calibration, symmetry, and common-mode rejection

In dual interferometers, suppression of gravity-like terms and technical phase noise is only as good as the symmetry between the two loops. One large-area gyroscope calibrates atomic trajectories by three Raman-based diagnostics: delaying the Raman pulse to measure position along the atomic motion, scanning two-photon detuning to extract vertical velocity from the Doppler shift, and translating a slice behind the Raman beams to map the coordinate perpendicular to the trajectory. Launch speed is controlled by cooling-laser detuning, launch angle by cooling-beam orientation, and initial position by a bias magnetic field. Using this procedure, horizontal position mismatch is reduced to below $\mathbf{v}_b=-\mathbf{v}$6, vertical mismatch to below $\mathbf{v}_b=-\mathbf{v}$7, and velocity along gravity is matched within $\mathbf{v}_b=-\mathbf{v}$8. Optimizing symmetry and overlap raises interference contrast from $\mathbf{v}_b=-\mathbf{v}$9 to more than $\phi_a$0, with final contrasts of $\phi_a$1 and $\phi_a$2 in the two interferometers (Yao et al., 2017).

The compact ensemble-exchange sensor exhibits the same noise logic on a shorter timescale. Because both interferometers are driven by the same Raman pulses and retro-mirror, mirror vibration along $\phi_a$3 and laser phase noise are strongly correlated. In the reported data, the differential rotation channel improves from $\phi_a$4 mrad/shot in each single interferometer to $\phi_a$5 mrad/shot, showing an almost factor-of-2 improvement due to common-mode rejection (Rakholia et al., 2014).

Dynamic compensation in large-area gyroscopes extends this symmetry principle from static alignment to time-dependent Earth-rotation effects. In the compensated $\phi_a$6 dual gyroscope, the contrast maxima of the two interferometers, displaced by about $\phi_a$7 without compensation, coincide after active mirror tilting, and seismic events demonstrate a common-mode rejection ratio of up to 459 (Gu et al., 28 May 2026).

Thermal stability is an equally central concern. In the modular laser system, active module power drift over $\phi_a$8–$\phi_a$9C reaches $\phi_b$0, passive-module fluctuation reaches $\phi_b$1, and the minimum polarization extinction ratios under temperature cycling are 19.8 dB and 23.8 dB for active and passive modules, respectively. The measured contrast slopes, $\phi_b$2 from $\phi_b$3–$\phi_b$4C and $\phi_b$5 from $\phi_b$6–$\phi_b$7C, identify Raman power stability as a dominant residual limitation (Sun et al., 2024).

5. Optical, fiber, comb, and displacement-metrology realizations

Outside atom interferometry, the same dual-channel strategy appears in fiber metrology, laser ranging, self-mixing displacement sensing, and compact inertial sensors. In these systems, duality usually means two wavelengths, two beam paths, or two internal interferometers that share most hardware and are recombined to suppress common-path noise or enlarge measurement range.

Domain	Dual construction	Representative result
Absolute distance metrology	Two-wavelength double heterodyne interferometer with variable synthetic wavelengths	12 GHz wavelength scan, 20 ms scan time, 1.5 m range, better than 1.2 nm accuracy (Kuramoto et al., 2014)
Fiber displacement sensing	1064 nm MIFO + 1055 nm RIFO in a common PM-fiber network	$\phi_b$8 at 1 Hz (Zhang et al., 2022)
Self-mixing displacement and angle sensing	Dual-beam frequency-shifted feedback interferometric system	1 nm linear resolution and 0.02″ angular resolution (Xu et al., 2022)
Compact inertial sensing	Single-element dual-interferometer prism with TM IFO and Ref IFO	Sub-picometer precision above 10 mHz; two devices to 2 mHz (Yang et al., 2020)
Dual-comb interferometry	Two combs or two effective comb states combined in multi-heterodyne detection	Mutual coherence beyond 300 s (Chen et al., 2017)
Single-comb dual-comb interferometry	Repetition-rate switching of one comb plus matched delay	Non-ambiguity range extended from 1.5 cm to 45.7 m (Carlson et al., 2018)
Dual-port Michelson readout	Balanced homodyne detection at symmetric and antisymmetric ports	13.2 dB scattered-light-noise reduction (Lohde et al., 2024)

In the fiber-based two-wavelength heterodyne interferometer, the primary displacement channel at $\phi_b$9 nm and the reference channel at $\mathbf{k}_\text{eff}$00 nm share almost the entire PM-fiber path. The measurement is formed as

$\mathbf{k}_\text{eff}$01

so common-mode path fluctuations cancel. The preliminary test reports $\mathbf{k}_\text{eff}$02 at 0.1 Hz and $\mathbf{k}_\text{eff}$03 at 1 Hz, but imperfect spectral separation introduces periodic errors that must be fitted and subtracted (Zhang et al., 2022).

A closely related dual-beam logic appears in self-mixing metrology. Two spatially separated beams interrogate a non-cooperative target and deliver phases $\mathbf{k}_\text{eff}$04 and $\mathbf{k}_\text{eff}$05, from which

$\mathbf{k}_\text{eff}$06

yield linear and angular displacement. The reported prototype reaches 35 nm and 0.15″ stability over 1 hour, 1 nm and 0.02″ resolution, linearity $\mathbf{k}_\text{eff}$07 over 100 mm, and $\mathbf{k}_\text{eff}$08 over $\mathbf{k}_\text{eff}$09 (Xu et al., 2022).

Single-element dual interferometers miniaturize the same idea. The SEDI inertial sensor places a test-mass interferometer and a reference interferometer inside one heptagonal fused-silica prism, using deep frequency modulation interferometry to extract phase, modulation depth, and arm-length difference from one strongly frequency-modulated laser. Modeling shows sub-picometer precision above 10 mHz in a package of a few cubic inches, and two such devices reach sub-picometer precision down to 2 mHz (Yang et al., 2020).

Comb-based realizations push the dual concept into frequency space. A phase-stable dual-comb interferometer with feed-forward control of the relative carrier-envelope offset frequency achieves mutual coherence over times exceeding 300 s and resolves more than $\mathbf{k}_\text{eff}$10 comb lines across 20 THz (Chen et al., 2017). PHIRE, by contrast, uses a single comb whose repetition rate is rapidly switched between two values and overlapped with itself after a matched delay, preserving dual-comb speed and resolution while extending non-ambiguity range to 45.7 m in absolute distance metrology (Carlson et al., 2018).

6. Quantum, communication, and ultrafast variants, and recurring limitations

In quantum communication, a variable-delay Mach–Zehnder interferometer with 128 actively selectable delays uses a distinct dual strategy: two interferometer arms remain physically fixed, but 128 delay states are calibrated by a phase modulator, DAC, and FPGA-driven lookup table. During QKD, seven Pockels-cell delay gates switch at 10 kHz, while active phase stabilization maintains the visibility of most of the 128 interferometer selections over 96% (Xu et al., 2018).

In ultrafast optics, a dual-functional interferometer integrates an interferometric X-FROG with a radial-shearing Michelson interferometer. A single scan of one interferometer arm simultaneously yields a cross-correlated FROG trace at the beam center and delay-dependent interferograms of the whole beam profile, enabling full spatiotemporal reconstruction. The demonstrated few-cycle pulse has $\mathbf{k}_\text{eff}$11 fs duration at the centroid (Chen et al., 2020).

In quantum metrology, the “quantum twin interferometer” arranges dual pairs of entangled twin beams in a parallel configuration and performs entangled detection so that phase signals add while noise is reduced by inverse correlations. The experiment reports distributed phase sensing with 3 dB quantum noise reduction in phase-sensing power at the level of milliwatts and advances the signal-to-noise ratio record achieved in photon-correlated interferometers by three orders of magnitude (Du et al., 8 Jan 2025).

Across these implementations, the principal limitations recur with striking regularity. Duality only helps when the paired channels remain sufficiently symmetric and sufficiently correlated. In atom interferometers this appears as sensitivity to trajectory mismatch, Raman-beam misalignment, dead time, and residual wavefront, AC-Stark, and temperature effects (Yao et al., 2017, Gu et al., 28 May 2026, Sun et al., 2024). In continuous closed-loop sensors the acceleration-lock bandwidth of the demonstrated implementation is $\mathbf{k}_\text{eff}$12 Hz, so larger navigation bandwidth still requires redesign (Meng et al., 2022). In fiber and Michelson systems the dominant residuals arise from imperfect spectral separation, associated noise couplings, and witness-channel noise (Zhang et al., 2022, Lohde et al., 2024). In entangled optical schemes, loss and mode mismatch limit usable squeezing and therefore absolute quantum advantage (Du et al., 8 Jan 2025).

The broader trajectory is nonetheless clear. Inertial sensing uses dual interferometers to decouple acceleration from rotation and to convert common technical noise into removable common mode. Optical metrology uses them to create synthetic wavelengths, internal references, or dual-port witness channels. Ultrafast and quantum-optical implementations use the same formal structure to combine complementary measurements or entangled resources. The technique’s unifying significance is therefore architectural: by building two interferometric views of the same system and combining them with the correct symmetry, one obtains information that a single interferometer cannot provide at the same stability, dynamic range, or noise floor.