Phase: A Multidisciplinary Perspective

• Phase is a fundamental concept defined as the angular component of a complex quantity, critical for describing wave behavior, quantum states, and phase transitions.
• Research on phase retrieval reveals sharp transition thresholds in measurement regimes, enabling efficient signal reconstruction through advanced compressed sensing techniques.
• Operator-based frameworks in quantum and classical contexts rigorously define phase, facilitating precise measurements in imaging, MRI, and large-scale computational simulations.

Phase is a fundamental concept with deep implications across mathematics, physics, engineering, and the computational sciences. In its broadest sense, phase refers to the argument or angular component of a complex quantity, governing phenomena as disparate as quantum measurement, wave superposition, material microstructure evolution, algorithmic phase retrieval, and even the structural properties of large-scale simulations. Research on arXiv, spanning from foundational mathematical treatments to application-rich algorithmic innovations, demonstrates that phase is not a unitary notion but a cross-cutting entity whose precise definition, role, and manipulation depend crucially on context.

1. Mathematical and Physical Foundations of Phase

In complex analysis and Fourier theory, the phase of a complex number $z=re^{i\phi}$ is $\phi\in(-\pi,\pi]$. For functions and signals, the phase spectrum $\Phi(\omega)$ pairs with the magnitude $|F(\omega)|$ to fully describe a signal in the Fourier domain. Classic results show that phase carries most of the information responsible for perceptual structure in images, as phase-randomization destroys coherence while magnitude randomization preserves salient features (Zachevsky et al., 2018).

In quantum mechanics, phase appears both at the level of individual amplitudes and in the geometry of Hilbert space. The phase operator for the quantum harmonic oscillator—traditionally elusive—admits a rigorous Hermitian realization with proper trigonometric identities and time evolution, whose eigenstates can be expressed in Gegenbauer polynomial expansions of number states (Ma et al., 2015). For systems endowed with higher symmetry, such as $\mathrm{SU}(3)$, phase operators and phase states can be defined using generalized oscillator algebras; their spectrum and structure under temporal evolution links to discrete Fourier transforms and mutually unbiased bases (Daoud et al., 2011).

In statistical field theory and cosmology, phase refers to distinct thermodynamic states (local minima) of an effective potential, with cosmological phase transitions driven by underlying scalar field configurations changing as temperature varies (Athron et al., 2020).

2. Phase Retrieval: Theory, Algorithms, and Phase Transitions

Phase retrieval is concerned with reconstructing a signal $x$ from intensity-only measurements, typically $|\langle x, \varphi_n \rangle|^2$. A pivotal insight is that uniqueness (injectivity) in phase retrieval exhibits sharp phase transitions as a function of the number of measurements and the dimension of the signal space. For real-valued signals, generic injectivity arises at $N=2M-1$ with $M$ the signal dimension, while in the complex case the conjectured threshold is $N=4M-4$ (Mixon, 2014). Almost-injectivity (generic uniqueness) requires fewer measurements, $\phi\in(-\pi,\pi]$0 (real) or $\phi\in(-\pi,\pi]$1 (complex). These transitions are algebraically and geometrically sharp for generic measurement ensembles.

Recent work in phase-only compressed sensing demonstrates that exact recovery of structured signals is possible from the phase (sign) of complex linear measurements. Surprisingly, the location of the phase transition—the sharp threshold in the number of measurements for successful recovery—is lower than for traditional linear compressed sensing; for $\phi\in(-\pi,\pi]$2-sparse signals, phase-only compressed sensing requires only about $\phi\in(-\pi,\pi]$3 of the measurements needed for the linear case, disproving the prior conjecture of coincident phase transitions (Chen et al., 21 Jan 2025). This is quantified by statistical dimension calculations of descent cones and proved using the Gaussian min-max theorem.

Algorithmically, phase retrieval from modulus or phase-only data often employs projection or basis pursuit formulations, frequently regularized by priors on phase structure. For instance, incorporating local Markovian Gaussian mixture priors for wavelet phases enables more accurate image restoration and phase-retrieval than classical Hybrid Input-Output methods (Zachevsky et al., 2018). A different regularization perspective emphasizes phase sparsity in a suitable basis (e.g., Zernike polynomials), leading to specialized algorithms (e.g., SROP) whose iterations are interpretable as cyclic projections among nonlinear constraint sets—with provable convergence and substantial empirical improvements over unregularized methods (Nguyen et al., 2018).

3. Operator and Measurement Theory: Quantum and Classical Phase

The formalization of phase as an operator in quantum theory faces foundational challenges. For the quantum harmonic oscillator, the Ma–Rhodes phase operator is Hermitian, satisfies trigonometric identities, and carries a complete set of orthonormal phase states constructed from number states (using Gegenbauer polynomials); its dynamical evolution mirrors classical phase shift under the oscillator Hamiltonian, and the classical limit is recovered in coherent states (Ma et al., 2015).

At the level of operator-valued measures (OVMs), the Wigner phase operator gives a natural formalism for the Wigner phase distribution and is equipped with a canonical conjugate, although the latter is not the number operator and the OVM lacks positivity due to Wigner function negativity. By coarse-graining phase-space resolution with a Gaussian filter, the Husimi $\phi\in(-\pi,\pi]$4-function provides a positive phase POVM (Q-phase operator), which can be operationally realized as projective measurements after beam-splitter interference with a strong reference state—a Gelfand–Naimark dilation (Subeesh et al., 2012).

In multi-mode or higher-symmetry systems, such as those governed by $\phi\in(-\pi,\pi]$5 or $\phi\in(-\pi,\pi]$6 algebras, phase operators and vector phase states are constructed via generalized oscillator algebras, with unitary phase operators whose eigenstates encode quantized discrete phases. Physical consequences include temporally stable phase states and inherently algebraic constructions of quadratic discrete Fourier transforms, facilitating mutually unbiased bases central to quantum information (Daoud et al., 2011).

4. Phase in Engineering, Imaging, and Quantum Information

Phase is both a fundamental property to be measured and the target of advanced algorithmic manipulation. In quantitative phase imaging, for example, the goal is to recover the spatially varying optical phase from intensity-only holographic measurements. Traditional techniques require mechanical phase shifters to capture multiple shifted interferograms, but deep neural architectures (DPS-net) can digitally synthesize arbitrary phase-shifted versions from a single input, enabling full-field phase recovery via classic $\phi\in(-\pi,\pi]$7-step algorithms after artificial phase modulation (Zhang et al., 2020).

In MRI, phase-regularized reconstructions must handle the non-convexity arising from phase wraps. The phase-cycling algorithm achieves invariance to phase wraps by randomly cycling wrap locations, enabling robust, joint reconstruction across modalities (partial Fourier, water-fat, flow imaging) without explicit unwrapping and integrating natural application-specific priors as regularizers (Ong et al., 2017).

In quantum information protocols, such as entanglement-based quantum key distribution (QKD), relative phases in distributed Bell states directly impact interference visibility and quantum bit error rate (QBER). Geometric-phase manipulation via Pancharatnam–Berry phase elements (e.g., QHQ sequences of waveplates) enables robust, hardware-agnostic compensation of arbitrary phase shifts either at the source or receiver, restoring entanglement fidelity and suppressing QBER under realistic channel conditions. This technique is extensible to time-bin qubit systems, simplifying quantum communication deployments (Nai et al., 14 Apr 2026).

5. Phase in Computational Physics and Large-Scale Simulation

The notion of phase occupies a foundational role in computational physics via phase-field methods for modeling microstructural evolution. The sharp phase-field method (S-PFM) achieves robust, discretization-independent, rotationally invariant resolution of interfaces by exact construction of the free-energy functional, potential, and gradient terms in real space, yielding numerically exact interface profiles without grid pinning and with rigorous convergence to sharp-interface kinetics (Finel et al., 2018).

Phase-field frameworks have advanced to model complex ferroelectric and antiferroelectric thin-film microstructures, using coupled order parameter fields and grain-resolved three-dimensional models. Such approaches quantitatively capture switching phenomena in HfO$\phi\in(-\pi,\pi]$8-based capacitors, revealing microstructure-assisted switching pathways and phase architecture-driven tuning of coercive fields (Pankaj et al., 15 Feb 2026).

In the context of quantum-ready computational pipelines, the PHASE hierarchical Pauli assembly algorithm enables efficient decomposition of large finite-element stiffness matrices into Pauli operator representations. By leveraging recursive mesh geometry for a hybrid tensorized Pauli decomposition and Fast Walsh–Hadamard transform aggregation, PHASE reduces the exponential scaling exponent of operator assembly, making quantum-compatible simulation of large engineering-scale models computationally tractable (Philo et al., 9 Jun 2026).

AI surrogates for complex scientific simulations are now integrating phase-aware components into their architecture. The PHASE (Physics-Integrated, Heterogeneity-Aware Surrogates) framework for Earth system modeling fuses heterogeneous data-type encoders and physics-based constraints, enabling the replacement of millennial-scale numerical integration with data-driven near-equilibrium prediction, substantially accelerating biogeochemical spin-up workflows (Gao et al., 27 Sep 2025).

In cosmological simulations, phase information embedded in the enormous set of initial random perturbations requires reproducibility and cross-scale consistency. The Panphasia approach encodes all random phases as coefficients of a hierarchical octree of orthogonal real-space functions, enabling unambiguous, multi-resolution phase specification for simulation initial conditions and exact inheritance of phases on nested or zoom-in grids (Jenkins, 2013).

In power-systems engineering, phase selection in three-phase optimal power flow (OPF) for distributed energy resources affects the hosting capacity (HC) and dynamic operating envelope (DOE) computation—optimizing phase allocation with binary variables incurs steep computational complexity but can yield order-of-magnitude improvements in integration potential, highlighting an intrinsic accuracy–scalability tradeoff (Antic et al., 2024).

6. Phase Transitions, Classification, and Applications Beyond Physics

The term phase transition signifies abrupt, threshold-like changes in system properties as a control parameter (e.g., measurement count, temperature, or network configuration) crosses a critical value. In both phase retrieval (Mixon, 2014) and phase-only compressed sensing (Chen et al., 21 Jan 2025), this concept is quantified using injectivity and statistical dimension theory, providing precise demarcation of feasible and unfeasible recovery regimes. In cosmology, phase transitions dictate fundamental events such as the origin of the matter-antimatter asymmetry and the generation of gravitational waves, with specialized tools (e.g., PhaseTracer) designed to map out the temperature evolution of effective potential minima and locate critical and nucleation temperatures (Athron et al., 2020).

Phase, as both a formal and operational construct, is indispensable for quantum metrology, precision imaging, communications, and any domain where coherence, interference, and symmetry breaking play central roles. Its many instantiations—geometric, dynamical, operator-based, or as an emergent property of systems—require a precise mathematical apparatus for analysis, measurement, retrieval, and manipulation. Recent research unambiguously demonstrates that technical mastery of phase, its transitions, and its algorithmic exploitation is not only a unifying theoretical thread but an engine of progress across contemporary science and engineering.