Polarization Singularities in Wave Fields

• Polarization singularities are specific points or lines in vector fields where the polarization ellipse becomes degenerate and key geometric attributes are undefined.
• They are classified into C-points, L-lines, and V-points using winding numbers and polarization parameters to reveal underlying topological invariants.
• These singularities enable advanced applications in optical manipulation and high-sensitivity sensing by exploiting anomalous scattering and phase vortices.

Polarization singularities are loci in vector wave fields where the local polarization ellipse becomes degenerate and one of its defining geometric attributes ceases to be well defined. In the standard optical language, circular polarization produces C-points in two dimensions or C-lines in three dimensions, whereas linear polarization produces L-lines or L-surfaces; in momentum-space photonics, amplitude zeros are also treated as V-points. Although developed within singular optics, the framework is not restricted to electromagnetic radiation: it applies to sound, water-surface waves, and, in a higher-spin generalization, gravitational waves. Across these settings, the subject is organized by polarization-ellipse geometry, Stokes or Jones descriptions, winding numbers and related indices, and conservation rules governing the creation, motion, splitting, and annihilation of singularities (Liu et al., 2020, Muelas-Hurtado et al., 2022, Rigouzzo et al., 26 Feb 2026).

1. Definitions and classification

At a fixed point of a monochromatic field, the real field traces a polarization ellipse over one period. A polarization singularity occurs when that ellipse becomes degenerate. In two-dimensional sections, the generic singularities are C-points, where the polarization is purely circular and the ellipse orientation is undefined, and L-lines, where the polarization is purely linear and the handedness is undefined. In three dimensions, the corresponding objects are C-lines and L-surfaces for spin-1 vector waves; for spin-2 gravitational waves, C singularities remain lines, whereas L singularities become isolated points because the relevant codimension changes (Chen et al., 2020, Rigouzzo et al., 26 Feb 2026).

In the Stokes formalism, the C-point condition is

$S_1=0,\qquad S_2=0,\qquad |S_3|=S_0,$

while the L-line condition is

$S_3=0,\qquad (S_1,S_2)\neq(0,0).$

In the Jones-vector view for a transverse field $E=(E_x,E_y)^T$, a C-point satisfies

$E_x\pm iE_y=0,$

which encodes equal amplitudes and a phase difference of $\pm 90^\circ$ (Chen et al., 2020). In momentum-space photonics, a V-point is defined by $E_x=E_y=0$, so all Stokes parameters vanish; these objects carry integer topological charge, whereas C-points carry half-integer charge (Wang et al., 2024).

The local morphology of isolated C-points is commonly resolved into lemon, star, and monstar types. Star singularities have index $-\tfrac12$ and three radial lines; lemon singularities have index $+\tfrac12$ and one radial line; monstar singularities also have index $+\tfrac12$ but retain three radial lines. This classification concerns the arrangement of polarization-line directions near the singularity rather than its handedness alone (Jia et al., 2024).

2. Local geometry, quadratic fields, and topological indices

A convenient three-dimensional formulation begins from a complex vector field $$\mathbf V(\mathbf r)=\mathbf P(\mathbf r)+i\,\mathbf Q(\mathbf r)$$. The real field

$S_3=0,\qquad (S_1,S_2)\neq(0,0).$0

traces the local polarization ellipse. The quadratic scalar

$S_3=0,\qquad (S_1,S_2)\neq(0,0).$1

or, in electromagnetic notation, $S_3=0,\qquad (S_1,S_2)\neq(0,0).$2, is central: C singularities occur at its zeros, $S_3=0,\qquad (S_1,S_2)\neq(0,0).$3, which is equivalent to the simultaneous vanishing of its real and imaginary parts. The ellipse normal is encoded by

$S_3=0,\qquad (S_1,S_2)\neq(0,0).$4

or, for electromagnetism,

$S_3=0,\qquad (S_1,S_2)\neq(0,0).$5

L singularities are the zeros of this spin-density-like vector; for spin 1 this yields lines, whereas for spin 2 it yields isolated points (Garcia-Etxarri, 2016, Rigouzzo et al., 26 Feb 2026).

For transverse optical fields, the ellipse azimuth and ellipticity can also be written in terms of Stokes parameters as

$S_3=0,\qquad (S_1,S_2)\neq(0,0).$6

The major-axis orientation is undefined at C singularities because all axes of a circle are equivalent; the minor axis or ellipse normal is undefined at L singularities because the ellipse collapses to a line (Liu et al., 2020).

Topological classification is given by winding numbers. Around a point singularity in a two-dimensional parameter plane, one may define

$S_3=0,\qquad (S_1,S_2)\neq(0,0).$7

where $S_3=0,\qquad (S_1,S_2)\neq(0,0).$8 is the polarization azimuth. Because the polarization ellipse has $S_3=0,\qquad (S_1,S_2)\neq(0,0).$9 symmetry, $E=(E_x,E_y)^T$0 can be half-integer. In the quadratic-field formulation, a small loop linking a C-line measures the winding of $E=(E_x,E_y)^T$1 by an integer multiple of $E=(E_x,E_y)^T$2; for a nondegenerate C-point this produces the familiar half-integer index $E=(E_x,E_y)^T$3 (Chen et al., 2020, Bliokh et al., 2021). Around such loops, the major-axis field can sweep out a polarization Möbius strip with a single half-twist, a geometric manifestation that has been observed in both optical and acoustic settings (Muelas-Hurtado et al., 2022).

3. Creation, splitting, annihilation, and conservation

Polarization singularities do not appear arbitrarily. In several settings they are generated as oppositely charged pairs, move under parameter variation, and annihilate only in charge-compensating events. In a PEC-backed planar resonator with epsilon-near-zero or epsilon near-pole response, an embedded eigenstate in the Hermitian limit supplies a degenerate pole-zero pair on the real-frequency axis, consisting of a $E=(E_x,E_y)^T$4 and a $E=(E_x,E_y)^T$5 charge superposed. When a small non-Hermitian parameter $E=(E_x,E_y)^T$6 is introduced, the degeneracy is lifted and the zero and pole split in opposite directions in the complex-frequency plane:

$E=(E_x,E_y)^T$7

with

$E=(E_x,E_y)^T$8

These singularities therefore emerge as pairs and obey charge conservation under creation and annihilation (Sakotic et al., 2021).

The same conservation principle appears in engineered three-dimensional vector beams. In composite Bessel-like fields, paired C-points of opposite index $E=(E_x,E_y)^T$9 annihilate whenever their transverse displacements become identical, and revive once the offsets reappear. By varying the underlying mode content and transverse shifts, polarization singularity lines can be made to follow arbitrary trajectories, annihilate, revive, and transform at prescribed longitudinal positions (Wand et al., 2022).

A more subtle conservation law arises in non-Hermitian photonic bands with exceptional points. Around loops in $E_x\pm iE_y=0,$0-space that enclose exceptional points, the non-Hermitian Berry phase is $E_x\pm iE_y=0,$1, but the observed far-field polarization charge on a particular loop can be $E_x\pm iE_y=0,$2, $E_x\pm iE_y=0,$3, $E_x\pm iE_y=0,$4, or $E_x\pm iE_y=0,$5, depending on which C-points are enclosed. The invariant Berry phase therefore constrains the globally conserved charge rather than the local partial charge on arbitrary subloops. Extra C-points act as charge compensators when singularities move between bands or across loops (Chen et al., 2020).

A common expectation is that same-charge singularities repel at short range. True two-dimensional random vector waves violate that naive rule: the probability of finding two C-points with the same topological charge at a vanishing distance is enhanced, and the partial correlation $E_x\pm iE_y=0,$6 peaks strongly as $E_x\pm iE_y=0,$7. This exceptional behavior is tied to enforced in-plane transversality and to correlations between left- and right-circular components that are absent in paraxial slices of three-dimensional random fields (Angelis et al., 2018).

4. Real-space optical realizations

One major class of realizations uses structured beams. Polarization-singular beams can be written as coherent superpositions of orthogonally polarized Laguerre–Gauss modes with different orbital charges,

$E_x\pm iE_y=0,$8

A suitably detuned $E_x\pm iE_y=0,$9-plate generates such beams by partial spin-to-orbital angular-momentum conversion. The output contains a central C-point and a surrounding L-line loop of controllable radius, and the transverse polarization pattern may take lemon, star, or spiral form depending on the topological index $\pm 90^\circ$0 (Cardano et al., 2012).

A different mechanism was identified by Chun-Fang Li in the exact angular-spectrum description of vector vortex beams. The divergence-free condition $\pm 90^\circ$1 leaves a gauge-like freedom encoded by the Stratton vector $\pm 90^\circ$2, from which orthonormal polarization bases are constructed as

$\pm 90^\circ$3

These bases are singular when $\pm 90^\circ$4, and the on-axis polarization singularity of the resulting vector vortex beam descends directly from that singularity of the momentum-space basis rather than from a mere superposition of two scalar beams (Li, 2020).

Scattering by subwavelength particles provides another canonical setting. For combined isotropic electric and magnetic dipoles, the scattered field supports L surfaces where the ellipse normal vanishes and C lines where $\pm 90^\circ$5. Garcia-Etxarri showed that these singularities arise naturally in high-index nanoparticles and that, around a C-line, the major-axis field can form a single-twist Möbius strip. The same singular-optics framework yields a derivation of anomalous Kerker conditions: if $\pm 90^\circ$6, the only possible far-field circular solution lies in the exact backscattering direction, which also lies on an L surface, so the amplitude must vanish; similarly, $\pm 90^\circ$7 yields zero forward scattering (Garcia-Etxarri, 2016).

Full-wave calculations on small spheres and tori broaden this picture. In the far field, four C singularities appear and the sum of their polarization topological indices is two, independent of particle shape. In the near field, by contrast, the index sum is no longer fixed by the Poincaré–Hopf theorem because the field is non-transverse and evanescent contributions are significant; near-field C-lines can flip their sign as they propagate outward, with the flip occurring where the polarization-ellipse normal becomes perpendicular to the local tangent of the C-line (Peng et al., 2020).

Closed metallic cavities add a topological constraint from the boundary itself. In spherical cavities, the sum of indices on the surface is $\pm 90^\circ$8, matching the Euler characteristic of the sphere; in toroidal cavities it is $\pm 90^\circ$9, matching the torus. Mirror and cylindrical symmetries determine whether the dominant singular objects are straight V-lines or split C-lines, and polarization Möbius strips arise naturally around isolated C-lines and around symmetry-protected V-lines (Jia et al., 2024).

5. Momentum-space singularities, BICs, and non-Hermitian photonics

In photonic crystal slabs and related open systems, the far-field polarization can be regarded as a vector field over in-plane wavevector space. Bound states in the continuum appear as V-points of this field, while circularly polarized resonances appear as C-points. This momentum-space perspective links polarization singularities to band degeneracies, Berry curvature, and non-Hermitian topology (Liu et al., 2020).

A particularly explicit relation is obtained near non-Hermitian Dirac points. With two orthogonal guided-resonance modes and differential loss, the effective Hamiltonian

$E_x=E_y=0$0

has eigenfrequencies

$E_x=E_y=0$1

When $E_x=E_y=0$2, the system has a Hermitian Dirac point; for $E_x=E_y=0$3, it splits into a pair of exceptional points. Simultaneously, a pair of far-field C-points with opposite handedness is induced, each inheriting the same half-integer charge as the associated exceptional point to leading order through the radiation coupling map $E_x=E_y=0$4 (Wang et al., 2024).

Electrical control of such momentum-space singularities has been demonstrated in highly birefringent planar liquid-crystal microcavities. In a two-mode non-Hermitian Hamiltonian for $E_x=E_y=0$5 polarizations, the Stokes field is reconstructed from the eigenvectors, and C-points are identified where $E_x=E_y=0$6 and $E_x=E_y=0$7. Applying a bias changes the liquid-crystal director angle $E_x=E_y=0$8, which tunes the detuning $E_x=E_y=0$9 and therefore the C-point positions. Experimentally, the two C-points in each branch move outward according to the square-root law

$-\tfrac12$0

in agreement with the two-mode Hamiltonian and transfer-matrix simulations (Oliwa et al., 11 Feb 2025).

An alternative control mechanism uses anisotropy rather than geometry. In an anisotropic grating, off-diagonal entries of the permittivity tensor can shift accidental BICs or split them into C-points without breaking the structural symmetry. In-plane rotation activates $-\tfrac12$1 and shifts accidental BICs in $-\tfrac12$2-space, whereas out-of-plane rotation activates $-\tfrac12$3 or $-\tfrac12$4 and splits an integer-charge V-point into two half-charge C-points. The splitting direction depends on whether the band is TE or TM, and multiple C-point creation and annihilation processes occur while preserving total topological charge (Lei et al., 24 May 2025).

The embedded-eigenstate resonator furnishes a complementary scattering-based picture. There, real-frequency zeros generated by small $-\tfrac12$5 exhibit a $-\tfrac12$6 phase jump over a very narrow linewidth, and the same phase vortices that characterize the complex reflection coefficient map directly into polarization singularities of the reflected field. By coordinating TE and TM zeros in $-\tfrac12$7 space, the reflected state can be swept between near-pure TE, pure TM, and intermediate elliptical states, including linear-to-circular conversions (Sakotic et al., 2021).

6. Extensions beyond optics and physical roles

Polarization singularities are generic features of inhomogeneous vector wave fields of any nature. In acoustics and water-surface waves, the relevant vector quantity is the velocity or displacement field rather than the electric field, but the same quadratic scalar $-\tfrac12$8, the same major and minor axes, and the same C-point and Möbius-strip topology appear. Three-wave acoustic interference fields and nonparaxial acoustic Bessel beams exhibit C-points, Möbius strips, nonzero spin density, and skyrmionic textures, extending singular-optics concepts to longitudinal sound waves (Bliokh et al., 2021, Muelas-Hurtado et al., 2022).

Rigouzzo et al. extended the framework to gravitational waves in TT gauge by replacing the vector phasor with a complex, symmetric, trace-free tensor $-\tfrac12$9. The gravitational analogues

$+\tfrac12$0

play the roles of the quadratic scalar and spin-density vector. In random plane-wave simulations, C singularities form filamentary curves for both electromagnetic and gravitational waves, but L singularities are filamentary lines for spin 1 and isolated points for spin 2. The corresponding C-line length densities converge to

$+\tfrac12$1

and the isolated gravitational-wave L points carry integer Poincaré indices whose creation and annihilation obey index-conservation rules (Rigouzzo et al., 26 Feb 2026).

The practical significance of polarization singularities follows from the extreme field structure that accompanies them. In scattering systems, C-lines can induce complex optical force and torque on nearby dielectric or magnetic particles. For a small particle in the dipole approximation, the time-averaged force and torque are written in terms of local field gradients and spin densities, and the force and torque vary dramatically along electric and magnetic C-lines, providing symmetry-controlled degrees of freedom for on-chip optical manipulation (Peng et al., 2022).

Sensing is another recurrent theme. Near real-frequency zeros produced by embedded-eigenstate splitting, the phase sensitivity to environmental perturbations obeys

$+\tfrac12$2

and can exceed $+\tfrac12$3 near $+\tfrac12$4. Increasing the quality factor enhances sensitivity, whereas tuning close to annihilation conditions can flatten $+\tfrac12$5 while retaining a large phase slope, broadening angular acceptance (Sakotic et al., 2021). More generally, the motion and annihilation of C-points in nanoparticle scattering, cavities, and momentum-space bands provide high-contrast markers for refractive-index changes, geometrical deformations, and polarization-selective mode conversion (Peng et al., 2020).

Across singular optics, metaphotonics, acoustics, and higher-spin wave physics, polarization singularities therefore function as a topological skeleton of vector fields. Their robustness is topological rather than metric, but their observable consequences are sharply physical: abrupt phase vortices, Möbius-strip axis fields, skyrmionic textures, unidirectional guided resonances, anomalous scattering zeros, chiral forces and torques, and extreme phase and polarization sensitivity. This suggests a unified viewpoint in which polarization singularities are not peripheral defects of wave fields but organizing centers for vector-wave topology and its control (Liu et al., 2020).