Nonlinear Interferometers (NLIs)

Updated 2 May 2026

Nonlinear Interferometers (NLIs) are systems that replace linear beam splitters with nonlinear elements like parametric amplifiers, enabling quantum-enhanced metrology.
NLIs achieve improved phase sensitivity beyond the shot-noise limit by harnessing nonclassical correlations and amplification through Bogoliubov transformations.
They facilitate advanced quantum state and spectral engineering with applications in imaging, all-optical switching, and multi-parameter sensing.

Nonlinear Interferometers (NLIs) are a general class of interferometric systems in which one or both linear beam splitters of canonical interferometers (such as the Mach–Zehnder or Michelson) are replaced by nonlinear elements—most prominently parametric amplifiers (based on χ² or χ³ processes), four-wave mixing in fibers, or nonlinear optical defects. Such architectures exploit the nonclassical correlations and ampliﬁcation properties of parametric processes to achieve superior phase sensitivity, enable quantum-enhanced metrology, and provide active control over quantum states of light and matter. NLIs embody a family of devices, including SU(1,1) interferometers, nonlinear Sagnac or Michelson variants, crystal superlattices, and devices for active spectral engineering, with broad applications in quantum information, imaging, metrology, and all-optical switching.

1. Fundamental Theory and Key Architectures

A canonical NLI comprises two or more nonlinear mixing stages (e.g., parametric amplifiers), interleaved with phase-encoding and linear or dispersive elements. The prototype is the SU(1,1) interferometer, introduced by Yurke et al., which replaces SU(2) beam splitters with nonlinear beam splitters (NBSs) implemented by parametric down-conversion (PDC) or four-wave mixing (FWM) (Scharwald et al., 2023, Kranias et al., 2023). The input–output relations for the two-mode fields are governed by Bogoliubov transformations:

$\hat{b}_s = u \hat{a}_s + v \hat{a}_i^\dagger, \quad \hat{b}_i = u \hat{a}_i + v \hat{a}_s^\dagger,$

with $u = \cosh{g}$ , $v = \sinh{g}$ , where $g$ is the gain parameter set by the nonlinear interaction strength and pump. After passage through two NBSs separated by a phase element, the output field statistics display phase-sensitive quantum interference.

NLIs are also realized in multi-stage or multi-element configurations (crystal superlattices (Paterova et al., 2019), multi-fiber chains (Ma et al., 2020)), nonlinear Sagnac or Michelson geometries (Lukens et al., 2016, Huang et al., 2012), or folded/“collapsed-arm” designs. Architectures exist which leverage induced coherence or exploit integration with metasurfaces for advanced multiwavelength interferometry (Paterova et al., 2020, Rojas-Santana et al., 2021).

2. Phase Sensitivity and Quantum Limits

NLIs enable phase estimation with sensitivity surpassing the shot-noise limit (SNL) of linear SU(2) interferometers, and, under ideal conditions, can achieve Heisenberg-limited scaling ( $\Delta\phi \sim 1/N$ ) (Scharwald et al., 2023, Kranias et al., 2023). For a balanced, lossless SU(1,1) interferometer (vacuum input), the quantum Fisher information is

$I_Q = 4N_\phi(N_\phi + 1),$

where $N_\phi$ is the mean photon number interrogating the sample, and the corresponding phase variance is

$(\Delta\phi)^2 \geq 1/I_Q \sim 1/N_\phi^2.$

This phase uncertainty is lower than the $1/N_\phi$ scaling of the SNL (Kranias et al., 2023).

However, NLIs’ quantum advantage is intricately dependent on loss and interferometer imbalance. Internal loss (between amplifiers) or mode mismatch degrades sensitivity, setting a critical transmissivity $T_c$ below which quantum advantage is lost; for large $u = \cosh{g}$ 0, $u = \cosh{g}$ 1 (Kranias et al., 2023, Giese et al., 2017). Unbalancing the gains in SU(1,1) schemes can mitigate detection loss sensitivity but does not intrinsically improve quantum scaling (Giese et al., 2017).

In multi-stage NLIs or superlattice chains, phase sensitivity is further enhanced: with $u = \cosh{g}$ 2 nonlinear elements, the fringe width narrows as $u = \cosh{g}$ 3, and the minimum resolvable phase improves proportionally (Paterova et al., 2019). Compensation for multimode effects, such as diffraction and spatial walk-off, is critical when operating at high-gain and high mode number (Scharwald et al., 2023).

3. Spectral and Quantum State Engineering

NLIs provide unique capabilities for in situ quantum-state and spectral engineering. In multi-stage fiber-based NLIs (e.g., N-segment DSF fibers separated by SMF phase shifters), the joint spectral intensity (JSI) of emitted photon pairs exhibits an "island" structure, determined by the multi-slit interference function $u = \cosh{g}$ 4 (Ma et al., 2020). By appropriately binomially weighting segment lengths, secondary maxima ("mini-maxima") are suppressed, yielding isolated, round JSI islands optimal for heralded high-purity photon sources.

These active spectral filtering properties enable direct extraction of factorable two-photon states with high heralding efficiency—of particular value in quantum information processing (Ma et al., 2020). In multi-crystal superlattices (N up to 5), the interference pattern fringe width scales as $u = \cosh{g}$ 5, further enhancing phase resolution and state engineering flexibility (Paterova et al., 2019).

4. Multi-Parameter Quantum Estimation

NLIs are an effective platform for quantum-enhanced multi-parameter estimation. The precision limit for simultaneous displacement estimation in pure Gaussian states is set by the Holevo Cramér–Rao bound (HCRB), which captures the ultimate attainable precision (Zhou et al., 5 Feb 2025). For an SU(1,1) NLI, the HCRB for four quadrature displacement parameters is

$u = \cosh{g}$ 6

where $u = \cosh{g}$ 7 is the squeezing parameter. This limit is achievable via dual homodyne detection; in this scenario, the classic symmetric logarithmic derivative (SLD) bound is not tight at small $u = \cosh{g}$ 8—underscoring that the HCRB must be used for sharp precision analysis. This result substantiates the use of simple measurement strategies (dual homodyne) to saturate optimal quantum performance in NLIs for multi-parameter estimation (Zhou et al., 5 Feb 2025).

5. Nonlinear Interferometers in Sensing and State Control

NLIs have been experimentally realized in both quantum and classical metrological settings. Key demonstrations include:

All-fiber SU(1,1) NLIs using cascaded FWM in highly nonlinear fiber, achieving phase-sensitive interference with 97% visibility and broadband ( $u = \cosh{g}$ 9 over 554 GHz) operation, as well as noise cancellation at dark fringes (Lukens et al., 2018).
Truncated-NLI protocols combining a single parametric amplifier and dual-homodyne detection, yielding quantum noise reduction (up to 3 dB below SQL) in displacement sensing for atomic-force microscopy, with minimized photon backaction (Pooser et al., 2019).
Sagnac–Michelson collapsed-arm NLIs with single-cell geometry, delivering interference visibility up to 99.9% and demonstrating quantum noise cancellation in a passively stable configuration (Lukens et al., 2016).
Nonlinear interferometry-based mid-IR metasurface characterization and optical-coherence-tomography (OCT), exploiting induced-coherence and SU(1,1) architectures for wavelength-decoupled, high-sensitivity imaging using only accessible visible detection hardware (Paterova et al., 2020, Rojas-Santana et al., 2021).

The metrological gain generated by entangled or squeezed states in NLIs has been verified for atomic ensembles and mechanical probes, providing robust enhancement (up to $v = \sinh{g}$ 0 over SQL) even in regimes inaccessible to direct time-reversal protocols (Liu et al., 2021).

6. Practical Considerations: Loss, Seeding, and Robustness

Quantum advantage in NLIs depends vitally on internal and external loss management (Kranias et al., 2023). Internal (between amplifier) transmissivity $v = \sinh{g}$ 1 must exceed a geometry- and photon-number-dependent threshold $v = \sinh{g}$ 2 for nonclassical scaling. The use of coherent-state seeding optimally compensates for internal loss only below $v = \sinh{g}$ 3, after which pure squeezing is preferable. Robustness to detection inefficiency and external (post-amplifier) loss can be fully restored (in principle) by sufficient gain in the final stage and meticulous pump-phase control.

The specific performance benefits and best practices, summarized in (Kranias et al., 2023, Giese et al., 2017, Oglialoro et al., 7 Jan 2026), include:

Balanced SU(1,1) configurations achieve Heisenberg scaling under lossless/internally robust conditions.
For realistic, unbalanced loss or high-gain operation, Mandel-type induced-coherence NLIs with differential detection exhibit the highest phase sensitivity, always attaining (but not surpassing) the shot-noise limit—far more robust than Yurke-type schemes (Oglialoro et al., 7 Jan 2026).
Unbalancing amplifier gains enhances detection-loss tolerance but not intrinsic sensitivity; in practice, stronger sources favor resilience to internal loss.

7. Advanced NLI Concepts: Nonlinear Response and All-Optical Switching

Beyond quantum metrology, NLIs enable nonlinear phase control and all-optical logic. Nonlinear Mach–Zehnder–Fano interferometers, with side-coupled Kerr defects, manifest hybrid resonances enabling low-threshold ( $v = \sinh{g}$ 4100 fJ), all-optical switching with $v = \sinh{g}$ 5 contrast and $v = \sinh{g}$ 6 response (Xu et al., 2012). Nonlinear Sagnac interferometers leverage cross-phase modulation for ultrafast, low-loss optical switching with negligible degradation of entanglement—validated in quantum networking experiments (Huang et al., 2012). In the classical regime, NLIs based on Michelson geometry and embedded Kerr media offer improved length sensitivity, scaling up to $v = \sinh{g}$ 7 with pulse and spatial engineering, beating the classical SQL (Luis et al., 2015, Paltoglou et al., 2015).

NLIs provide a versatile and powerful family of interferometric tools at the interface of quantum optics, nonlinear science, and precision metrology. Through careful design—balancing nonlinear gain, loss management, and mode structure—NLIs can robustly attain or surpass the fundamental quantum limits of measurement, mediate complex quantum-state engineering, and enable implementation of next-generation quantum-enabled sensors, imaging modalities, and information-processing devices.