Polarization Aberrations and Spatially Varying Polarization

Almost every polarization calculation quietly assumes the state of the light is the same everywhere across the beam — one Jones vector, one point on the Poincaré sphere, valid from axis to edge. Structured light breaks that assumption two ways. The first is a defect: a real lens, mirror, or coating hits marginal rays at steeper angles than paraxial ones, so its Fresnel transmission and phase depend on where the ray crosses the pupil. The Jones matrix becomes a field $J(\rho,\theta)$ across the aperture, and an initially uniform beam emerges with a patterned polarization it did not start with — the polarization aberrations that limit high-numerical-aperture imaging, astronomical polarimetry, and lithography. The second is deliberate and far stranger. One can design beams whose polarization varies across the profile — a cylindrical vector beam, radial polarization pointing outward like a hedgehog, azimuthal circulating like a whirlpool. Here the texture carries something a uniform beam cannot: topology. Assign a Poincaré-sphere point to every place in the beam and the map can wind by an integer — a conserved topological charge, the Poincaré–Hopf index of the on-axis polarization singularity, that no uniform polarizer or wave plate can unwind. A blunt consequence follows: a continuous single-valued transverse field cannot wrap the axis, so the intensity must vanish there. Every cylindrical vector beam is a doughnut whose dark hole is forced by topology, not by the source.

Key takeaways

A polarization aberration is a spatially varying Jones matrix $J(\rho,\theta)$ — retardance and diattenuation maps that a high-NA optical system stamps onto a uniform beam.
A cylindrical vector beam (radial or azimuthal) is linearly polarized at every point, but the linear direction rotates with azimuth — an intrinsic property of the mode, not a defect.
The on-axis polarization singularity carries an integer topological charge — the Poincaré–Hopf index, $+1$ for radial and azimuthal — conserved: no homogeneous element can remove it.
Because a continuous non-zero field cannot wrap the axis, the field must vanish there: the doughnut null is topologically forced, $I(0)=0$ .
A vector beam is a superposition of the two circular polarizations carrying opposite orbital angular momentum — spin and orbit locked together (spin–orbit coupling of light).
Focused at high NA, a radially polarized beam builds a strong on-axis longitudinal field $E_z$ and a sub-Airy focal spot — tighter than any scalar beam at the same aperture.
Key idea: promote the constant Jones vector to a field $\mathbf{E}(\rho,\theta)$ and the constant Jones matrix to $J(\rho,\theta)$ — polarization becomes a texture over the beam, mapped onto the Poincaré sphere, and that map can carry a conserved topological charge that forces a dark core no homogeneous optic can fill.

The core insight

Let polarization vary across the beam. A defect version — the pupil-dependent Jones matrix $J(\rho,\theta)$ — is the polarization aberration that degrades high-NA optics. A designed version — radial or azimuthal polarization — carries a conserved topological charge (the Poincaré–Hopf index of its on-axis singularity) that forces the intensity to vanish on axis and cannot be undone by any spatially uniform element.

Where this lives today

Structured light has moved from curiosity to workhorse. Radially polarized beams give the tightest focal spots at a given aperture, so they turn up wherever the last nanometres of resolution matter: confocal and STED microscopy, tip-enhanced Raman probes that couple to the strong longitudinal field, optical trapping, and laser cutting — where a radial beam stays p-polarized against the walls of a deep kerf all the way around. Vortex and vector beams carry orbital angular momentum in discrete, robust units, an extra alphabet for high-capacity optical and quantum communication, generated and sorted on a chip by q-plates and their Pancharatnam–Berry metasurface descendants. On the aberration side, the same formalism keeps precision instruments honest. Any high-NA system — a lithography projector, a telescope hunting the polarized glow of an exoplanet, a remote-sensing polarimeter — must model and invert its own pupil-dependent Jones matrix, or its measured Stokes vectors are corrupted by the instrument's own retardance and diattenuation. Polarization ray-tracing is now a standard stage of optical design. Structured light and polarization aberration are two faces of one idea: admit that polarization can vary across a beam and you gain a degree of freedom to engineer and a class of errors to control.

The physical picture

Think of the beam cross-section as a canvas and polarization as a colour brushed at each point. A uniform beam is a flat wash; a vector beam is a picture — radial polarization with arrows pointing straight out, azimuthal with them circling. Follow the arrows once around the axis and they turn through a full $360^\circ$ : they wind. That winding is the topological charge, and it is why the centre is special — you cannot comb the arrows to a single direction at the middle, so the amplitude must collapse to zero there. The bright beam is a ring, the middle dark: the hairy-ball obstruction, transplanted from a sphere to the beam's transverse plane.

Figure

Radial (left) and azimuthal (right) polarization as real arrow fields: the line winds once about the axis — a V-point of index +1 where the field vanishes, so I(0) = 0

An imperfect optic does the opposite: it brushes unwanted colour onto a flat wash. A lens strikes its marginal rays at larger angles than its paraxial rays, so Fresnel transmission and phase differ across the pupil, and the beam picks up a position-dependent retardance and diattenuation even with flawless glass. That patterned Jones matrix is the polarization aberration, mapped next.

Figure

A polarization aberration as a pupil map: retardance δ grows as ρ² toward the rim, reaching δ ≈ 0.075 λ ≈ 27° at NA 0.9 — the induced ellipticity that scrambles a uniform input

Decompose a radial beam into left- and right-circular parts and a second secret appears: each is a scalar vortex carrying orbital angular momentum, the two winding oppositely. So the beam's spin (circular polarization) and orbit (phase winding) are locked together — spin–orbit coupling of light. On the Poincaré sphere the radial/azimuthal family lives on the equator; scanning once around the axis walks the equator once — that winding is the charge no homogeneous polarizer can touch.

The core equations

The starting move is to let the Jones algebra acquire a position label. Where uniform optics multiplies a constant Jones vector by a constant matrix, structured light does the same point by point across the pupil, everything now a function of the transverse coordinate.

\mathbf{E}_{\text{out}}(\rho,\theta) = J(\rho,\theta)\,\mathbf{E}_{\text{in}}(\rho,\theta) \qquad (1)

where: $\mathbf{E}_{\text{in}},\mathbf{E}_{\text{out}}$ — transverse Jones vectors at pupil point $(\rho,\theta)$ ; $J(\rho,\theta)$ — the $2\times2$ Jones matrix, now a field across the aperture. Constant $J$ recovers ordinary polarization optics; a varying $J$ is either a designed vector beam or a polarization aberration. A near-perfect optic is close to the identity, so its Jones field expands about $\mathbb{1}$ in the Pauli basis, and the two leading terms are the two physical ways a boundary acts on polarization: splitting amplitudes between eigenpolarizations (diattenuation) or splitting phases (retardance).

J(\rho,\theta) = \mathbb{1} + \varepsilon(\rho,\theta)\,\sigma_1 + i\,\delta(\rho,\theta)\,\sigma_3 + \cdots \qquad (2)

where: $\varepsilon(\rho,\theta)$ — diattenuation map (amplitude-splitting aberration); $\delta(\rho,\theta)$ — retardance map (phase-splitting aberration); $\sigma_1,\sigma_3$ — Pauli matrices spanning small Jones perturbations. For an aplanatic lens both grow from zero as $\rho^2$ , largest for the steep marginal rays — the pattern in the second figure. Now the designed side. Radial and azimuthal polarization are the lowest-order cylindrically symmetric solutions: the local linear direction is $\hat{\mathbf r}$ or $\hat{\boldsymbol\phi}$ , with a radial amplitude that must vanish on axis (by the topological argument below, not by choice).

\mathbf{E}_{\text{rad}} = A(\rho)\big(\cos\theta\,\hat{\mathbf x} + \sin\theta\,\hat{\mathbf y}\big), \qquad \mathbf{E}_{\text{azi}} = A(\rho)\big(-\sin\theta\,\hat{\mathbf x} + \cos\theta\,\hat{\mathbf y}\big) \qquad (3)

where: $A(\rho)$ — the radial (doughnut) amplitude, e.g. $A\propto(\rho/w)\,e^{-\rho^2/w^2}$ , with $A(0)=0$ ; $\theta$ — azimuthal angle. The radial field points along $\hat{\mathbf r}$ , the azimuthal along $\hat{\boldsymbol\phi}$ ; at every point the light is purely linearly polarized — the exotic character is in the arrangement. Rewrite the radial field in the circular basis $\hat{\mathbf e}_\pm = (\hat{\mathbf x}\pm i\hat{\mathbf y})/\sqrt2$ and the spin–orbit structure appears: two circular components of opposite handedness, each dressed by a vortex phase of opposite sign — the bridge between polarization (spin) and spatial phase (orbit).

\mathbf{E}_{\text{rad}} = \tfrac{A(\rho)}{\sqrt2}\big(e^{-i\theta}\,\hat{\mathbf e}_+ + e^{+i\theta}\,\hat{\mathbf e}_-\big) \qquad (4)

where: $\hat{\mathbf e}_\pm$ — circular unit vectors carrying spin $s=\pm1$ ( $\hbar$ per photon); $e^{\mp i\theta}$ — vortex phase carrying orbital angular momentum $\ell=\mp1$ , opposite the spin. Total $j=\ell+s=0$ , but neither piece vanishes — spin and orbit are individually nonzero and interlocked. Finally the topology. Track the orientation angle $\psi$ of the local linear polarization around a loop $C$ encircling the axis; because the pattern returns to itself, the total turning is an integer multiple of $2\pi$ — the conserved charge.

\eta = \frac{1}{2\pi}\oint_C \mathrm{d}\psi \in \mathbb{Z}, \qquad \eta_{\text{radial}} = \eta_{\text{azimuthal}} = +1 \qquad (5)

where: $\eta$ — the Poincaré–Hopf index of the on-axis polarization singularity (a V-point, where the field is purely linear everywhere around it); $\psi$ — orientation of the local linear polarization; $C$ — any small loop enclosing the axis once. The index is a topological invariant: no spatially uniform $J$ can change it, because a homogeneous rotation adds the same constant to $\psi$ everywhere and cannot alter how $\psi$ winds.

The dark core, and how a q-plate makes one

Derivation ASpin-orbit decomposition of a cylindrical vector beam

The goal is to show that a radial (or azimuthal) beam is not a new object but a phase-locked superposition of the two circular polarizations, each carrying orbital angular momentum equal and opposite to its spin — the decomposition that welds the beam's polarization to its spatial phase. The strategy is a basis change: write the Cartesian unit vectors in the circular basis, substitute, and collect the two circular components; the vortex phases fall out of Euler's formula.

Step 1. Invert the circular basis $\hat{\mathbf e}_\pm=(\hat{\mathbf x}\pm i\hat{\mathbf y})/\sqrt2$ to get $\hat{\mathbf x}=(\hat{\mathbf e}_++\hat{\mathbf e}_-)/\sqrt2$ , $\hat{\mathbf y}=(\hat{\mathbf e}_+-\hat{\mathbf e}_-)/(i\sqrt2)$ , and substitute into $\hat{\mathbf r}=\cos\theta\,\hat{\mathbf x}+\sin\theta\,\hat{\mathbf y}$ . Why: the circular states are the spin eigenstates, so whatever multiplies $\hat{\mathbf e}_\pm$ carries the orbital content.

Step 2. Collect each coefficient: the $\hat{\mathbf e}_+$ term multiplies $e^{-i\theta}/\sqrt2$ and the $\hat{\mathbf e}_-$ term multiplies $e^{+i\theta}/\sqrt2$ , giving equation (4). Why: Euler's formula turns the azimuthal dependence into pure vortices $e^{\mp i\theta}$ — one quantum of orbital angular momentum apiece.

Step 3. Read the bookkeeping: the $\hat{\mathbf e}_+$ photon has $s=+1,\ell=-1$ ; the $\hat{\mathbf e}_-$ photon has $s=-1,\ell=+1$ ; both have $j=\ell+s=0$ , yet each carries a full unit of spin and orbit. Why: this is spin–orbit coupling of light — you cannot rotate the pattern without winding the spatial phase. *Verified: the identity $\cos\theta\,\hat{\mathbf x}+\sin\theta\,\hat{\mathbf y}=\tfrac{1}{\sqrt2}(e^{-i\theta}\hat{\mathbf e}_++e^{+i\theta}\hat{\mathbf e}_-)$ is checked symbolically in verify_aq-optics-polarization-aberrations-vector-beams.py.*

Go deeper · Derivation B · 1 minWhy the intensity must vanish on the axis

Derivation BWhy the intensity must vanish on the axis

The goal is to prove the doughnut hole is a theorem, not an artefact: a continuous, single-valued transverse field whose direction winds once around the axis is forced to zero amplitude at the centre, so $I(0)=0$ for every cylindrical vector beam. The strategy is a continuity argument — approach the axis along opposite rays, demand single-valuedness, and read off how fast the amplitude must die.

Step 1. Approach along $\theta=0$ : the radial field points along $+\hat{\mathbf x}$ . Approach the same point along $\theta=\pi$ : it points along $-\hat{\mathbf x}$ . Why: $\hat{\mathbf r}$ reverses across the centre, so the two limits disagree unless the vector is zero — $\hat{\mathbf r}$ is simply undefined at $\rho=0$ .

Step 2. Demand single-valuedness. The only vector equal to both $+A\hat{\mathbf x}$ and $-A\hat{\mathbf x}$ is $\mathbf E(0)=\mathbf 0$ ; the axis is a V-point of index $+1$ where amplitude, not merely direction, is pinned. Why: the field-theoretic hairy-ball obstruction — a nonzero continuous field cannot wrap the axis, so the wrapping is paid for with a zero.

Step 3. Expand near the axis: analyticity forces $A(\rho)=a_1\rho+O(\rho^3)$ , so $I(\rho)\propto|A|^2 = a_1^2\rho^2+\cdots$ rises from zero at least quadratically. Why: a linear amplitude zero gives the annular doughnut intensity — the null is black, not merely dim. *Verified: the sign reversal of $\hat{\mathbf r}$ across the axis and the winding $\oint\mathrm d\psi=2\pi\Rightarrow\eta=+1$ are checked symbolically in verify_aq-optics-polarization-aberrations-vector-beams.py.*

Go deeper · Derivation C · 1 minA q-plate converts spin to orbit

Derivation CA q-plate converts spin to orbit

The goal is to show how a single patterned wave plate makes a cylindrical vector beam: a q-plate is a half-wave retarder whose fast axis winds as $\alpha(\theta)=q\theta$ , and on circularly polarized input it flips the handedness while stamping on an orbital vortex of charge $2q$ . The strategy is direct Jones multiplication — build the space-variant half-wave matrix, apply it to a circular input, and recognise the output.

Step 1. The q-plate matrix is a half-wave plate with fast axis at $q\theta$ : $J_q(\theta)=\begin{pmatrix}\cos 2q\theta & \sin 2q\theta\\ \sin 2q\theta & -\cos 2q\theta\end{pmatrix}$ . Why: rotating $\mathrm{diag}(1,-1)$ to angle $q\theta$ gives this traceless form in $2q\theta$ — the same factor of two by which a plate rotated by $\alpha$ acts at $2\alpha$ .

Step 2. Apply it to $\hat{\mathbf e}_+=(1,i)^{\mathsf T}/\sqrt2$ : the components collapse to $e^{i2q\theta}/\sqrt2$ and $-i\,e^{i2q\theta}/\sqrt2$ . Why: each entry is now a single vortex phase — all the azimuthal dependence has become orbital angular momentum.

Step 3. Recognise $J_q\hat{\mathbf e}_+ = e^{i2q\theta}\,\hat{\mathbf e}_-$ : spin flips $+1\to-1$ ( $\Delta s=-2$ ) while $\ell=+2q$ appears. For $q=\tfrac12$ the output is a charge-one vector beam. Why: total angular momentum is conserved, $\Delta(\ell+s)=0$ for $q=1$ — the standard route to vector and vortex beams, treated as a Pancharatnam–Berry element in the companion geometric-phase article. *Verified: the product $J_q\hat{\mathbf e}_+=e^{i2q\theta}\hat{\mathbf e}_-$ and the bookkeeping $\Delta(\ell+s)=0$ for $q=1$ are checked symbolically in verify_aq-optics-polarization-aberrations-vector-beams.py.*

Limiting cases

Case	Result	Meaning
uniform linear ( $\psi$ constant)	$\eta=0$	trivial texture; scalar polarization optics, no dark core
radial polarization	$\eta=+1$ , $I(0)=0$	outward hedgehog; strong longitudinal $E_z$ at high NA
azimuthal polarization	$\eta=+1$ , $I(0)=0$	circulating whirlpool; purely transverse focus, no on-axis $E_z$
order- $m$ vector beam ( $\psi=m\theta$ )	$\eta=m$	higher charge; $2\|m\|$ petals in a projected intensity
generic elliptic field	$\eta=\pm\tfrac12$ (C-points)	isolated circular points (lemon, star, monstar) — half-integer cousins
$\varepsilon,\delta\to0$ (perfect optic)	$J\to\mathbb 1$	no aberration; the pupil Jones field is one global matrix

Common trap“A radially polarized beam is a kind of circular or elliptical polarization”

It is neither — at every point a radial or azimuthal beam is perfectly linearly polarized, along $\hat{\mathbf r}$ or $\hat{\boldsymbol\phi}$ . Circular polarization needs a $90^\circ$ phase lag between two orthogonal components at a point, and there is none here; only the direction of a linear state varies across the beam. The same warning kills a common fix-it fantasy: you cannot rotate a radial beam into a uniform one with a single wave plate or polarizer, because a homogeneous element adds the same rotation everywhere and cannot change the winding $\eta$ . Only a spatially varying element — a q-plate, a metasurface — can alter the topology, which is why those devices exist.

Worked examples

1 · The Poincaré–Hopf index of a radial beam (confidence builder)

Show that a radially polarized beam carries topological index $\eta=+1$ , and use the index theorem on the sphere to argue the on-axis null is unavoidable.

Step 1. Write the orientation angle. For radial polarization the local linear direction is $\hat{\mathbf r}$ , whose angle is $\psi(\theta)=\theta$ . Circle the axis once, $\theta:0\to2\pi$ , and $\psi$ advances by exactly $2\pi$ . The index is $\eta=\frac{1}{2\pi}\oint\mathrm d\psi=\frac{1}{2\pi}(2\pi)=\mathbf{+1}$ . Azimuthal polarization has $\psi=\theta+\tfrac\pi2$ , the same winding, so also $\eta=+1$ .

Step 2. Contrast: an order- $m$ beam ( $\psi=m\theta$ ) winds $m$ times, $\eta=m$ ; a uniform beam ( $\psi=$ const) has $\eta=0$ and no forced null. No homogeneous $J$ can change the integer — it merely shifts $\psi$ by a constant.

Step 3. Invoke the index theorem. Poincaré–Hopf says the indices of a continuous tangent field on $S^2$ sum to the Euler characteristic, $\sum\eta_i=\chi(S^2)=2$ . Map the transverse plane onto a hemisphere: a single $+1$ V-point must be matched elsewhere (a second $+1$ at infinity closes the budget to $2$ ). No nowhere-zero field of the right winding exists — the defect is forced.

Step 4. Conclude the null. Since the winding cannot be removed and a nonzero field cannot wrap the axis, the amplitude must vanish where the direction is undefined: $I(0)=0$ — a topological invariant, not a property of the laser. $\eta=+1$ ; $\sum\eta_i=\chi(S^2)=2$ ; $I(0)=0$ , forced.

2 · A radially polarized focus beats the scalar spot (exam level)

A radially polarized beam and an ordinary scalar beam are each focused by an aplanatic lens of numerical aperture $\mathrm{NA}=n\sin\alpha=0.9$ (in air). Compare the focal-spot sizes, and explain where the radial beam's advantage comes from.

Step 1. Fix the scalar benchmark. The diffraction-limited scalar (Airy) spot has full width at half maximum $\text{FWHM}\approx0.514\,\lambda/\mathrm{NA}=0.514\lambda/0.9=\mathbf{0.571\,\lambda}$ . Evaluating the full vectorial (Richards–Wolf) diffraction integral for the scalar case reproduces $0.568\lambda$ — the two agree, anchoring the comparison.

Step 2. Build the radial focus. Each converging ray has its $\mathbf E$ in the meridional plane: a transverse part $\propto\cos\alpha$ and a longitudinal part $\propto\sin\alpha$ . On axis the transverse parts cancel across the pupil while the longitudinal parts all point along $+\hat{\mathbf z}$ and add. The on-axis field is purely longitudinal — a component a scalar beam lacks on axis entirely.

Step 3. Size the longitudinal spot. The on-axis longitudinal field is $E_z(\rho)\propto\int_0^\alpha\sqrt{\cos\theta}\,\sin^2\theta\,J_0(k\rho\sin\theta)\,\mathrm d\theta$ ; evaluating it at $\mathrm{NA}=0.9$ gives a central peak of $\text{FWHM}=\mathbf{0.506\,\lambda}$ — about $11\%$ tighter than the scalar Airy spot ( $0.506/0.568=0.89$ ), and roughly $0.79\times$ its area. The transverse $J_1$ term forms a faint surrounding ring, but the bright core is the longitudinal needle.

Step 4. Push it with an annulus. Blocking the pupil centre (inner edge $0.8\,\mathrm{NA}$ ) weights the integral toward the steepest, most-longitudinal rays and sharpens the spot to $\text{FWHM}=\mathbf{0.437\,\lambda}$ — the sub-diffraction needle measured by Dorn, Quabis and Leuchs (about $0.43\lambda$ ), prized in high-resolution microscopy and laser machining. Full aperture: $0.51\lambda$ vs scalar $0.57\lambda$ ; annular: $0.44\lambda$ .

Physics of the portrait

The portrait is a concept, not a person: the integer that refuses to change however you push the light around. Singular optics — the study of where a wave's phase or polarization goes undefined — was founded on the recognition that these singularities carry conserved topological charges, obeying counting rules as rigid as any conservation law.

Origins · 1 mintopological charge and how this was discovered

topological charge

Poincaré–Hopf index of a wave singularity · singular optics

Topological charge entered optics through a 1974 paper by John Nye and Michael Berry, Dislocations in wave trains, which noticed that the phase of any real wavefield must contain dislocations: lines of zero amplitude and undefined phase, around which the phase winds by an integer multiple of $2\pi$ . That integer is the topological charge. For scalar light it is the winding of the phase around an optical vortex; for vector light it is the Poincaré–Hopf index of a polarization singularity — the integer V-points of radial and azimuthal beams, and the half-integer C-points ( $\pm\tfrac12$ : lemon, star, monstar) of generic elliptic fields. The charge is conserved: singularities move, merge, and annihilate in pairs, but the net charge threading any loop is fixed by topology alone. This is why a vortex or vector beam is robust — aberrate it, attenuate it, send it through turbulence, and the charge survives — and it is the organising principle behind optical vortices, orbital angular momentum, geometric-phase flat optics, and structured-light communication.

History

1892

— Poincaré introduces the sphere that now carries polarization states, its equator the linear states and poles the circular ones

1972

— Mushiake, Matsumura and Nakajima generate radially polarized laser modes directly from a cavity

1974

— Nye and Berry identify wavefront dislocations and phase singularities, founding singular optics and the notion of topological charge

2003

— Dorn, Quabis and Leuchs measure a sub-diffraction focal spot from a radially polarized beam, confirming the longitudinal needle

2006

— Marrucci, Manzo and Paparo introduce the liquid-crystal q-plate, a compact spin-to-orbit converter for vector and vortex beams

Going further

Texts. Hecht, Optics, 5th ed., ch. 8 (Polarization) · Saleh & Teich, Fundamentals of Photonics, 3rd ed., ch. 6 (Polarization Optics) · Q. Zhan, "Cylindrical vector beams: from mathematical concepts to applications," Adv. Opt. Photon. 1, 1 (2009) · J. P. McGuire & R. A. Chipman, "Polarization aberrations," Appl. Opt. 33, 5080 (1994) · Novotny & Hecht, Principles of Nano-Optics.

Provenance. The circular-basis identity, the q-plate product $J_q\hat{\mathbf e}_+=e^{i2q\theta}\hat{\mathbf e}_-$ , and the angular-momentum bookkeeping follow Hecht ch. 8 and the q-plate literature (Marrucci et al.); the Poincaré–Hopf index and $\chi(S^2)=2$ are standard singular-optics topology (Nye & Berry). The focal-spot numbers — scalar Airy $0.571\lambda$ at $\mathrm{NA}=0.9$ , radial longitudinal $0.506\lambda$ (full) and $0.437\lambda$ (annular) — are recomputed from the aplanatic Richards–Wolf integral in the verify script, consistent with Zhan (2009) and Dorn, Quabis & Leuchs (2003). No non-standard constants are asserted.

Prerequisites. Polarization of Light · Jones Calculus · Stokes Parameters and the Poincaré Sphere · Circular Polarization · Wave Plates and Retarders. Leads to. Mueller Calculus · Optical Vortices and Orbital Angular Momentum · High-NA Focusing and the Debye–Wolf Integral.

Is a vector beam the same thing as a vortex beam?

They overlap but differ. A scalar vortex beam carries a phase singularity — the phase winds around a dark core — with uniform polarization; its charge is the phase winding $\ell$ . A vector beam carries a polarization singularity, charge the Poincaré–Hopf index $\eta$ . Equation (4) shows the link: a radial beam is two scalar vortices of opposite handedness and opposite $\ell$ , so vector beams are literally built from vortex beams. A general beam can carry both charges at once — full-Poincaré beams that sweep the whole sphere across their profile.

Why is radial polarization better than a plain beam for a tight focus?

Two effects combine. On axis the transverse components cancel across the pupil, so the central peak is built almost entirely from the longitudinal field $E_z$ — which lies along the propagation direction and is not bound by the transverse diffraction limit the same way. And at high NA that component is large, because the steep marginal rays project strongly onto the axis. The result is a needle roughly $11\%$ narrower than the scalar Airy spot, near $0.44\lambda$ with an annular pupil. An azimuthal beam, by contrast, has no longitudinal field on axis and focuses to a hollow spot — useful when you want darkness at the centre.

Can I remove a polarization aberration by cleaning or refocusing the lens?

Not by refocusing — a polarization aberration is a Jones error, not a wavefront error, so shifting focus does nothing. Nor is it a dirt problem: even a spotless lens has one, because it comes from the geometry of refraction. Marginal rays strike each surface at larger angles than paraxial rays, so their Fresnel s- and p-coefficients differ more, producing retardance and diattenuation that grow toward the rim. You reduce it by design (angle-tuned coatings, stress-free mounts, lower NA) and correct the residue by measuring and compensating the pupil Jones matrix — the polarimetry step in any precision instrument.

What exactly is the difference between a V-point and a C-point?

Both are polarization singularities in different fields. A V-point sits in a purely linearly polarized field — like radial or azimuthal light — where the linear orientation is undefined at the centre and winds by an integer, $\eta=+1$ . A C-point sits in a generic elliptic field, at an isolated spot where the light is exactly circular so the ellipse orientation is undefined; because that orientation is defined modulo $\pi$ , a C-point winds by half-integers, $\eta=\pm\tfrac12$ (lemon, star, monstar). Radial and azimuthal beams are the clean integer case; real fields are full of half-integer C-points.

How does a q-plate actually make a radial beam?

A q-plate is a half-wave retarder whose fast axis winds around the centre, $\alpha(\theta)=q\theta$ . On circularly polarized input it does two things at once (Derivation C): it flips the handedness and multiplies the beam by a vortex phase $e^{i2q\theta}$ , converting spin angular momentum into orbital. For $q=\tfrac12$ the output is (after projecting the handedness) exactly a radial or azimuthal beam. Because the phase comes from orientation rather than thickness, a q-plate — a patterned liquid-crystal cell or metasurface — is the same physics as a geometric (Pancharatnam–Berry) phase element.

Is the topological charge really conserved if I aberrate or attenuate the beam?

Yes — that robustness is the whole appeal. The charge is the winding of the polarization around a loop, and a smooth perturbation — defocus, aberration, attenuation, turbulence — deforms the field without tearing it, so the integer cannot jump. Charges change only through singular events: opposite singularities annihilate, or a pair is born together, conserving the net total threading any loop. This is why structured-light communication survives a real channel, and why the on-axis null is hard to fill: erasing it would mean changing an integer, which no continuous process can do.