SU(1,1) Scheme: Theory & Applications

• SU(1,1) scheme is a framework based on a noncompact Lie group that underpins two-mode squeezing, active interferometry, and coherent state constructions.
• It employs bilinear bosonic operators to build interferometric architectures, optimize phase sensitivity, and incorporate non-Gaussian operations like photon subtraction and Kerr nonlinearities.
• The scheme is applied in quantum metrology, hybrid atom–light systems, and phase-space analyses, offering enhanced signal amplification and robustness against losses.

Searching arXiv for recent and foundational work on SU(1,1) schemes, especially interferometric and representation-theoretic formulations. Searching arXiv for "SU(1,1) interferometer number-conserving operation" and related SU(1,1) scheme variants. The expression SU(1,1) scheme denotes a family of constructions organized by the noncompact Lie group SU(1,1). In quantum optics, it most often refers to active interferometric architectures in which passive beam splitters are replaced by parametric amplifiers or two-mode squeezers; in mathematical physics, it also denotes representation-theoretic, coherent-state, recoupling, and phase-space constructions built from positive or holomorphic discrete-series representations. The common structure is an su(1,1) algebra generated by bilinear bosonic operators, a noncompact squeezing geometry, and observables whose statistics encode phase shifts, displacements, tensor-product multiplicities, or quasiprobability distributions (Frascella et al., 2019, Gazeau et al., 4 Apr 2025).

1. Algebraic definition and realizations

The group SU(1,1) may be realized as the set of complex $2\times2$ matrices

$$g=\begin{pmatrix}a & \beta\ \overline\beta & \overline a\end{pmatrix}, \qquad |a|^2-|\beta|^2=1,$$

and its Lie algebra $\mathfrak{su}(1,1)$ is spanned by generators $K_0,K_\pm$ satisfying

$$[\,K_0,K_\pm\,]=\pm K_\pm, \qquad [\,K_+,K_-\,]=-2\,K_0$$

in one standard convention (Gazeau et al., 4 Apr 2025). In two-mode quantum-optical realizations the generators are commonly written as

$$K_+ = a^\dagger b^\dagger,\qquad K_- = a b,\qquad K_0 = \tfrac12(a^\dagger a+b^\dagger b+1),$$

or with closely related normalization conventions (Frascella et al., 2019, Chang et al., 2021). A standard invariant is the Casimir operator

$$\mathcal C =K_0^2-\tfrac12\bigl(K_+K_-+K_-K_+\bigr),$$

which labels irreducible representations (Gazeau et al., 4 Apr 2025).

The same algebra appears in several physically distinct realizations. In the two-boson Schwinger construction, it underlies two-mode squeezing and parametric amplification. In the one-mode realization,

$$K_0=\tfrac12\bigl(a^\dagger a+\tfrac12\bigr),\quad K_+=\tfrac12(a^\dagger)^2,\quad K_-=\tfrac12 a^2,$$

it governs single-mode squeezing and the one-mode version of the SU(1,1) quantum walk (Duan, 2022). For the radial oscillator, states with the same orbital angular momentum furnish a positive-discrete-series representation $D^+(\kappa)$, with basis $|\kappa,n_r\rangle$ and Bargmann index $$g=\begin{pmatrix}a & \beta\ \overline\beta & \overline a\end{pmatrix}, \qquad |a|^2-|\beta|^2=1,$$0 (Rosas-Ortiz et al., 2016).

Holomorphic discrete-series representations provide a canonical analytic model. For each weight $$g=\begin{pmatrix}a & \beta\ \overline\beta & \overline a\end{pmatrix}, \qquad |a|^2-|\beta|^2=1,$$1 with $$g=\begin{pmatrix}a & \beta\ \overline\beta & \overline a\end{pmatrix}, \qquad |a|^2-|\beta|^2=1,$$2, the carrier space is the Fock–Bargmann space on the unit disk $$g=\begin{pmatrix}a & \beta\ \overline\beta & \overline a\end{pmatrix}, \qquad |a|^2-|\beta|^2=1,$$3, with orthonormal basis

$$g=\begin{pmatrix}a & \beta\ \overline\beta & \overline a\end{pmatrix}, \qquad |a|^2-|\beta|^2=1,$$4

and group action

$$g=\begin{pmatrix}a & \beta\ \overline\beta & \overline a\end{pmatrix}, \qquad |a|^2-|\beta|^2=1,$$5

(Gazeau et al., 4 Apr 2025). This analytic realization is the basis for orthogonality relations, character formulas, and tensor-product decompositions.

2. SU(1,1) interferometric architecture

An SU(1,1) interferometer is obtained from a Mach–Zehnder interferometer by replacing each 50:50 beam splitter with a parametric amplifier. In optical implementations the nonlinear “beam splitter” is an optical parametric amplifier (OPA), represented by the two-mode squeeze operator

$$g=\begin{pmatrix}a & \beta\ \overline\beta & \overline a\end{pmatrix}, \qquad |a|^2-|\beta|^2=1,$$6

with Heisenberg-picture transformation

$$g=\begin{pmatrix}a & \beta\ \overline\beta & \overline a\end{pmatrix}, \qquad |a|^2-|\beta|^2=1,$$7

(Kang et al., 2024). In the conventional two-crystal sequence, the first OPA prepares the correlated probe, a phase shift is introduced between the nonlinear elements, and the second OPA either amplifies or de-amplifies the fields from the first pass depending on the relative phase (Frascella et al., 2019).

A standard balanced layout takes coherent-plus-vacuum input,

$$g=\begin{pmatrix}a & \beta\ \overline\beta & \overline a\end{pmatrix}, \qquad |a|^2-|\beta|^2=1,$$8

applies a first OPA $$g=\begin{pmatrix}a & \beta\ \overline\beta & \overline a\end{pmatrix}, \qquad |a|^2-|\beta|^2=1,$$9, a phase shift $\mathfrak{su}(1,1)$0 or its equivalent on the sensing mode, and a second OPA $\mathfrak{su}(1,1)$1 with relative phase $\mathfrak{su}(1,1)$2 and equal gain $\mathfrak{su}(1,1)$3 (Kang et al., 2024, Chang et al., 2021). Vacuum-seeded versions are also standard and give a two-mode squeezed vacuum (Anderson et al., 2017).

The second nonlinear element is not always essential. In the truncated SU(1,1) interferometer, the second NLO is omitted and the internal two-mode squeezed state is interrogated directly by an optimized joint measurement. The truncated layout preserves the same internal probe state but replaces nonlinear recombination by measurement-stage optimization; in bright-seeded operation it can saturate the phase-sensitivity bound set by the quantum Fisher information (Anderson et al., 2017, Gupta et al., 2018). This is one reason the SU(1,1) scheme is not reducible to the slogan “Mach–Zehnder with gain.”

Wide-field operation extends the architecture from a single spatial mode to many angular plane-wave modes. By inserting a lens or spherical mirror between the two crystals, one can image the PDC source region of crystal 1 onto crystal 2, preserve a $\mathfrak{su}(1,1)$4 mrad field, obtain two-dimensional visibility exceeding $\mathfrak{su}(1,1)$5 over the full $\mathfrak{su}(1,1)$6 mrad span, and observe $\mathfrak{su}(1,1)$7 dB quadrature squeezing with an OAM-mode count $\mathfrak{su}(1,1)$8, $\mathfrak{su}(1,1)$9 radial modes, and a total of $K_0,K_\pm$0 spatial modes (Frascella et al., 2019).

3. Readout, phase sensitivity, and ultimate bounds

Several observables are used in SU(1,1) metrology. Intensity detection measures the output total photon number,

$K_0,K_\pm$1

while homodyne schemes measure a final quadrature, often

$K_0,K_\pm$2

or a weighted joint quadrature $K_0,K_\pm$3 (Frascella et al., 2019, Gupta et al., 2018). The standard error-propagation formula is

$K_0,K_\pm$4

or the analogous intensity form $K_0,K_\pm$5 (Kang et al., 2024, Frascella et al., 2019).

The quantum Fisher information sets the detector-independent benchmark. For a pure probe with phase generator $K_0,K_\pm$6, one has

$K_0,K_\pm$7

In many coherent-plus-vacuum SU(1,1) schemes the generator is $K_0,K_\pm$8 or $K_0,K_\pm$9, so $$[\,K_0,K_\pm\,]=\pm K_\pm, \qquad [\,K_+,K_-\,]=-2\,K_0$$0 reduces to four times a number variance (Kang et al., 2024, Chang et al., 2021, Gupta et al., 2018).

Optimized homodyne detection can attain these bounds. For the bright-seeded truncated interferometer, the weighted quadrature observable $$[\,K_0,K_\pm\,]=\pm K_\pm, \qquad [\,K_+,K_-\,]=-2\,K_0$$1 with $$[\,K_0,K_\pm\,]=\pm K_\pm, \qquad [\,K_+,K_-\,]=-2\,K_0$$2 saturates $$[\,K_0,K_\pm\,]=\pm K_\pm, \qquad [\,K_+,K_-\,]=-2\,K_0$$3 exactly in the lossless case; with loss, the optimum becomes

$$[\,K_0,K_\pm\,]=\pm K_\pm, \qquad [\,K_+,K_-\,]=-2\,K_0$$4

(Anderson et al., 2017, Gupta et al., 2018). More generally, proper homodyne detection is nearly optimal for lossy SU(1,1) interferometers (Gao, 2016).

Losses are usually modeled by fictitious beam splitters. Internal losses, introduced before the second OPA or around the phase element, are repeatedly identified as more damaging than external losses because the second OPA re-amplifies any noise introduced before it (Chang et al., 2021, Kang et al., 2024). Mixed-state QFI under loss can be bounded or minimized by Kraus-space constructions, as in the Escher formalism and the $$[\,K_0,K_\pm\,]=\pm K_\pm, \qquad [\,K_+,K_-\,]=-2\,K_0$$5 minimization used for internal-loss channels (Kang et al., 2024, Chang et al., 2021).

4. Internal operations and enhanced SU(1,1) variants

A major development in the modern SU(1,1) literature is the insertion of internal operations between the first OPA and the phase shifter. In one number-conserving scheme, the internal non-Gaussian operation is either photon-addition-then-subtraction,

$$[\,K_0,K_\pm\,]=\pm K_\pm, \qquad [\,K_+,K_-\,]=-2\,K_0$$6

or photon-subtraction-then-addition,

$$[\,K_0,K_\pm\,]=\pm K_\pm, \qquad [\,K_+,K_-\,]=-2\,K_0$$7

with the more general superposition $$[\,K_0,K_\pm\,]=\pm K_\pm, \qquad [\,K_+,K_-\,]=-2\,K_0$$8, $$[\,K_0,K_\pm\,]=\pm K_\pm, \qquad [\,K_+,K_-\,]=-2\,K_0$$9 (Kang et al., 2024). Both operations raise the mean photon number before the second OPA and improve homodyne $$K_+ = a^\dagger b^\dagger,\qquad K_- = a b,\qquad K_0 = \tfrac12(a^\dagger a+b^\dagger b+1),$$0 and $$K_+ = a^\dagger b^\dagger,\qquad K_- = a b,\qquad K_0 = \tfrac12(a^\dagger a+b^\dagger b+1),$$1 in the ideal case. In homodyne-based $$K_+ = a^\dagger b^\dagger,\qquad K_- = a b,\qquad K_0 = \tfrac12(a^\dagger a+b^\dagger b+1),$$2, PS then PA slightly outperforms PA then PS for moderate gains $$K_+ = a^\dagger b^\dagger,\qquad K_- = a b,\qquad K_0 = \tfrac12(a^\dagger a+b^\dagger b+1),$$3 and coherent amplitudes $$K_+ = a^\dagger b^\dagger,\qquad K_- = a b,\qquad K_0 = \tfrac12(a^\dagger a+b^\dagger b+1),$$4, whereas in the lossless QFI the ordering is reversed and PA then PS yields a marginally larger $$K_+ = a^\dagger b^\dagger,\qquad K_- = a b,\qquad K_0 = \tfrac12(a^\dagger a+b^\dagger b+1),$$5. Under internal loss, both operations mitigate degradation, but for moderate to high loss $$K_+ = a^\dagger b^\dagger,\qquad K_- = a b,\qquad K_0 = \tfrac12(a^\dagger a+b^\dagger b+1),$$6 PS then PA delivers the larger $$K_+ = a^\dagger b^\dagger,\qquad K_- = a b,\qquad K_0 = \tfrac12(a^\dagger a+b^\dagger b+1),$$7, lower QCRB, and better homodyne $$K_+ = a^\dagger b^\dagger,\qquad K_- = a b,\qquad K_0 = \tfrac12(a^\dagger a+b^\dagger b+1),$$8 (Kang et al., 2024).

Multi-photon subtraction provides a related non-Gaussian strategy. Internal subtraction $$K_+ = a^\dagger b^\dagger,\qquad K_- = a b,\qquad K_0 = \tfrac12(a^\dagger a+b^\dagger b+1),$$9 improves phase sensitivity, and the performance becomes better by increasing subtraction number. It also efficiently improves robustness against internal photon losses, while asymmetric subtraction exhibits gain- and loss-dependent behavior: subtraction from mode $$\mathcal C =K_0^2-\tfrac12\bigl(K_+K_-+K_-K_+\bigr),$$0 is more beneficial at low $$\mathcal C =K_0^2-\tfrac12\bigl(K_+K_-+K_-K_+\bigr),$$1 or small $$\mathcal C =K_0^2-\tfrac12\bigl(K_+K_-+K_-K_+\bigr),$$2, whereas subtraction from mode $$\mathcal C =K_0^2-\tfrac12\bigl(K_+K_-+K_-K_+\bigr),$$3 becomes best at high gain or high coherent amplitude (Kang et al., 2023).

Other inserted nonlinear resources lead to similar conclusions. A Kerr medium in one arm replaces the linear phase shift by

$$\mathcal C =K_0^2-\tfrac12\bigl(K_+K_-+K_-K_+\bigr),$$4

which yields

$$\mathcal C =K_0^2-\tfrac12\bigl(K_+K_-+K_-K_+\bigr),$$5

In that scheme the Kerr nonlinear case can not only enhance the phase sensitivity and quantum Fisher information, but also significantly suppress the photon losses; internal losses have a greater influence on the phase sensitivity than the external ones (Chang et al., 2021). A single-path local squeezing operation

$$\mathcal C =K_0^2-\tfrac12\bigl(K_+K_-+K_-K_+\bigr),$$6

inside the interferometer gives

$$\mathcal C =K_0^2-\tfrac12\bigl(K_+K_-+K_-K_+\bigr),$$7

and improves robustness against both internal and external photon losses (Kang et al., 2024).

The pumped-up SU(1,1) modification addresses a different limitation. Conventional SU(1,1) interferometry uses only the particles outcoupled to the side modes, which constrains absolute sensitivity when $$\mathcal C =K_0^2-\tfrac12\bigl(K_+K_-+K_-K_+\bigr),$$8. Pumped-up schemes add mode mixing so that all the input particles participate in the phase measurement, surpass the shot-noise limit with respect to the total number of input particles, and are never worse than conventional SU(1,1) interferometry (Szigeti et al., 2016).

5. Experimental extensions and application domains

The SU(1,1) scheme has been extended well beyond single-parameter phase sensing. In quantum dense metrology, an SU(2)-in-SU(1,1) nested interferometer inserts a small-reflection Mach–Zehnder inside an SU(1,1) device so that phase and amplitude modulations can be jointly estimated. With a degenerate SUI and suitable phase-angle control, one can achieve the optimum quantum enhancement in the measurement precision of arbitrary mixture of phase and amplitude modulation, while maintaining tolerance to detection loss (Du et al., 2020).

In atom–light hybrid interferometry, concatenated SU(1,1)–SU(2)–SU(1,1) architectures use nonlinear Raman processes as the active stages and an SU(2) atom–light beam splitter in the middle. For QND measurement of photon number via the AC-Stark effect, the signal-to-noise ratio in a balanced case is improved by a gain factor of the nonlinear Raman process, and the readout-stage gain can be adjusted to reduce the impact due to losses (Jiao et al., 2021).

In surface-plasmon-resonance sensing, an SU(1,1) interferometer embedding an SPR sensor in the sensing arm can estimate Imbert–Fedorov shifts and incidence angle with homodyne detection. The reported incident-angle sensitivity is capable of surpassing the sensitivity limit of $$\mathcal C =K_0^2-\tfrac12\bigl(K_+K_-+K_-K_+\bigr),$$9, and both IF-shift sensitivity and incident-angle sensitivity can breakthrough the shot noise limit, even approaching the QCRB at $$K_0=\tfrac12\bigl(a^\dagger a+\tfrac12\bigr),\quad K_+=\tfrac12(a^\dagger)^2,\quad K_-=\tfrac12 a^2,$$0 and $$K_0=\tfrac12\bigl(a^\dagger a+\tfrac12\bigr),\quad K_+=\tfrac12(a^\dagger)^2,\quad K_-=\tfrac12 a^2,$$1 (Yuetao et al., 2023).

In quantum imaging and multimode metrology, the wide-field SU(1,1) interferometer provides two-dimensional phase-front sensing with sub-shot-noise sensitivity over many spatial modes, and the same multimode structure has been proposed for remote sensing, enhanced sub-shot-noise imaging, and quantum information processing (Frascella et al., 2019). In trapped-ion platforms, red and blue second-sideband driving of orthogonal vibrational modes can synthesize SU(1,1) Perelomov coherent states, separable squeezed states, and SU(2) beam-splitter states, with reversible dynamics proposed as interferometric resources (Alderete et al., 2018).

6. Phase-space, representation-theoretic, and conceptual scope

The SU(1,1) scheme is not restricted to interferometers. For phase-space methods, a bona fide SU(1,1) Wigner function is defined by

$$K_0=\tfrac12\bigl(a^\dagger a+\tfrac12\bigr),\quad K_+=\tfrac12(a^\dagger)^2,\quad K_-=\tfrac12 a^2,$$2

where $$K_0=\tfrac12\bigl(a^\dagger a+\tfrac12\bigr),\quad K_+=\tfrac12(a^\dagger)^2,\quad K_-=\tfrac12 a^2,$$3 is the SU(1,1) parity operator. An optical protocol using a squeezer and photon-number-resolving detectors samples this quasidistribution point by point, without tomographic reconstruction (Fabre et al., 2023).

Representation theory gives the exact structure behind such constructions. The discrete-series characters satisfy closed formulas, and tensor products obey

$$K_0=\tfrac12\bigl(a^\dagger a+\tfrac12\bigr),\quad K_+=\tfrac12(a^\dagger)^2,\quad K_-=\tfrac12 a^2,$$4

with multiplicity one (Gazeau et al., 4 Apr 2025). In the four-fold recoupling problem, connection coefficients between different coupling schemes are bivariate Racah polynomials, and the associated quadratic algebra closes only with an additional shift operator, leading to an extended algebra interpreted through the generic superintegrable system on $$K_0=\tfrac12\bigl(a^\dagger a+\tfrac12\bigr),\quad K_+=\tfrac12(a^\dagger)^2,\quad K_-=\tfrac12 a^2,$$5 (Post, 2015).

SU(1,1) coherent states also support a distinct quantum-walk construction on the hyperboloid or Poincaré disk. Because the coherent states are nonorthogonal, the SU(1,1) walk differs from the idealized Heisenberg–Weyl walk, but the overlap can be reduced by increasing the Bargmann index $$K_0=\tfrac12\bigl(a^\dagger a+\tfrac12\bigr),\quad K_+=\tfrac12(a^\dagger)^2,\quad K_-=\tfrac12 a^2,$$6, especially in the two-mode realization (Duan, 2022). For the radial oscillator, Perelomov and Barut–Girardello coherent states furnish explicit wavefunctions, resolutions of the identity, and squeezing criteria in natural quadratures (Rosas-Ortiz et al., 2016).

A recurrent misconception is that SU(1,1) interferometry is simply a passive interferometer with gain added at the ends. The literature shows a more specific structure: the relevant resource is two-mode squeezing organized by su(1,1), and the information can remain accessible even when the second OPA is removed, provided the measurement is optimized (Anderson et al., 2017, Caves, 2019). A second misconception is that detection inefficiency is always the dominant practical limitation. Several analyses instead identify internal loss as the more severe impairment, because noise introduced before the final active element is re-amplified (Chang et al., 2021, Kang et al., 2024). A broader implication is that “SU(1,1) scheme” properly names a symmetry-based framework whose interferometric, phase-space, and representation-theoretic versions are technically distinct but structurally unified by the same noncompact algebra.