Polarization of Electromagnetic Waves

Of the six numbers on the plane wave's birth certificate, five have been spent — frequency, direction, amplitude, phase. This article opens the last and richest one: the direction of $\mathbf{E}$ within the wavefront plane. Maxwell's equations confine the electric field to a two-dimensional transverse playground and then say nothing more; that leftover freedom is polarization, and its physics runs from the geometry of sunglasses to the angular momentum of single photons. The freedom is larger than one angle. Because the transverse plane holds two independent field components, each with its own amplitude and phase, the tip of $\mathbf{E}$ at a fixed point need not oscillate along a line — it can trace a circle or, in general, an ellipse. Linear, circular, and elliptical polarization are not three phenomena but one two-component superposition read at different relative phases, and complex notation (article 0584) was built for exactly this bookkeeping: a complex polarization vector holds the entire state. The article's second business is what polarization does. A polarizer projects the field onto its axis — Malus's $\cos^2\theta$ law is one line of vector geometry with a deep sting in its tail (the three-polarizer paradox, where inserting an absorber creates light). Reflection, scattering, and stressed materials all imprint polarization, which is why the sky is polarized, why glare lies flat, and why an LCD can gate light with a voltage. The systematic calculus — Jones vectors, Stokes parameters, partially polarized light — is the next article's; here the states themselves, and the physics that filters them.

Key takeaways

Polarization is the surviving transverse freedom: two independent components of $\mathbf{E}$ in the wavefront plane, each with amplitude and phase — four numbers, of which two (overall amplitude and phase) are trivial, leaving a two-parameter family of states.
Relative phase δ between the components decides the shape: δ = 0 or π gives linear polarization; δ = ±π/2 with equal amplitudes gives circular; the general case is an ellipse traced at frequency ω.
Handedness conventions are a genuine mess (optics looks toward the source, particle physics along the momentum); state the rotation explicitly and no one gets hurt.
An ideal polarizer transmits the component along its axis: Malus's law $I = I_0\cos^2\theta$ ; unpolarized light halves.
Inserting a 45° polarizer between crossed polarizers turns 0% transmission into 12.5% — projection is not filtering, and the "paradox" is the measurement-changes-the-state seed that quantum mechanics inherits.
Key idea: a plane wave's E-field owns a two-component transverse freedom — amplitudes plus relative phase — whose shapes are the ellipses, and whose projections are everything polarizers do.

The core insight

Polarization is the two-component transverse freedom of the plane wave: relative amplitude and phase of $E_x$ and $E_y$ select a point on the family line–ellipse–circle, and every polarizer acts by projecting the field onto an axis — Malus's $\cos^2\theta$ in intensity.

Where this lives today

Polarization is quietly load-bearing across display, communication, and sensing technology. Every LCD pixel is the three-polarizer physics of Worked Example 1 run in hardware: liquid-crystal molecules rotate the polarization between crossed polarizers, and a voltage un-twists them — projection bookkeeping, refreshed at 120 Hz on billions of screens (it is also why LCD screens go black through polarized sunglasses at the wrong angle). Photography and LiDAR use polarizing filters to strip Brewster-angle glare (article 0592's physics); polarimetric radar separates rain from hail and reads crop structure; and stress engineers photograph strain fields through crossed polarizers, since stressed transparent parts become weakly birefringent and paint their internal forces as fringes. At the research frontier the transverse freedom has become a quantum and astrophysical resource. Photon polarization is the workhorse qubit — the BB84 quantum-key protocol transmits secrets as randomly-chosen polarization bases, and entangled polarization pairs carried the Bell-test experiments that earned the 2022 Nobel Prize. Astronomy reads magnetic fields from polarized synchrotron light and dust emission; the Event Horizon Telescope's polarization maps trace the magnetic structure feeding M87's black hole; and CMB cosmology's most sought signal — primordial B-mode polarization — is a parts-per-billion polarization pattern that would fingerprint gravitational waves from inflation. The last free number on the plane wave's birth certificate turned out to carry information from everywhere.

The physical picture

Stand at a fixed point and watch the tip of $\mathbf{E}$ in the transverse plane, stroboscopically, over one period. Maxwell's constraints (article 0583) guarantee only that the tip stays in the plane; how it moves there is the wave's polarization state. Three portraits exhaust the possibilities. If the two transverse components oscillate in step (relative phase δ = 0 or π), the tip runs back and forth along a fixed line — linear polarization, the state the previous articles tacitly drew, its one parameter the line's angle ψ. If the components have equal amplitudes and a quarter-cycle offset (δ = ±π/2), the tip's motion is two perpendicular oscillations in quadrature — a circle, traced once per period at angular rate ω: circular polarization, whose one binary parameter is the direction of tracing. Everything else — unequal amplitudes, generic δ — is an ellipse, the family that contains both extremes; its shape and tilt are set by the amplitude ratio and δ together.

Figure

The polarization zoo — one two-component superposition, read at different relative phases: line, circle, and the general ellipse

Where do the states come from physically? Sources imprint them. A straight antenna radiates the linear polarization of its own axis (which is why rooftop aerials are mounted to match the transmitter). An electron circling in a magnetic field radiates circularly along the field axis. A hot filament or the Sun emits no preferred state at all — its light is a statistically churning mixture, "unpolarized," a concept that strictly requires the statistical language deferred to the next article. And matter edits states in transit: reflection near Brewster's angle transmits and reflects the two linear components unequally (article 0592 owns the mechanism), molecular scattering polarizes the blue sky at 90° from the Sun, and anisotropic crystals (the E3 permittivity-tensor article's subject) split the two components onto different speeds — the working principle of every wave plate, deferred to the optics track as hardware but used conceptually below. The freedom has a mechanical payload too. Circularly polarized light carries spin angular momentum — $\pm\hbar$ per photon, torque measurable on a suspended wave plate — but that ledger belongs to the E7 angular-momentum article (EM-15's set) and is recalled here by name only: polarization is where the state lives; the angular-momentum books were already opened there.

The mathematics of the transverse freedom

Fix $\hat{\mathbf{k}} = \hat{\mathbf{z}}$ and write the most general monochromatic plane wave as two transverse components with independent amplitudes and phases:

\mathbf{E}(z,t) = \hat{\mathbf{x}}\,E_1\cos(kz-\omega t) + \hat{\mathbf{y}}\,E_2\cos(kz-\omega t+\delta) \qquad (1)

where: $E_1, E_2 \geq 0$ — component amplitudes; δ — the relative phase, the only phase that matters (an overall phase just resets the clock; article 0583's FAQ). Read it physically: four real numbers describe the transverse state, but one (overall amplitude) sets brightness and one (overall phase) sets timing — the shape of the polarization is the two-parameter family (ratio $E_2/E_1$ , δ). In complex notation the same object is the complex polarization vector $\tilde{\mathbf{E}}_0 = E_1\hat{\mathbf{x}} + E_2e^{i\delta}\hat{\mathbf{y}}$ — one complex 2-component object, the seed the next article grows into the Jones formalism.

Derivation AFrom two components to the ellipse — the complete taxonomy of polarization states

The goal is to classify every state of (1) by the shape the field tip traces, deriving the linear and circular special cases exactly and exhibiting the general ellipse, with handedness made explicit. The strategy: read (1) at a fixed point ( $z = 0$ for convenience) as a parametric curve $(E_x(t), E_y(t))$ and let δ organize the cases. Watch how the quadrature case converts two oscillations into a rotation — the same $i$ -is-rotation fact the complex-notation article flagged — and watch which cases retrace a path (linear) versus orbit it (everything else).

Step 1. In step: δ = 0 (or π). Then $E_y/E_x = E_2/E_1$ (constant, with sign $-$ for δ = π): the tip runs along the fixed line at angle $\psi = \arctan(\pm E_2/E_1)$ , out to amplitude $\sqrt{E_1^2+E_2^2}$ and back. Linear polarization — one angle, no orbit. Why linearity requires phase lock: any δ ≠ 0, π makes $E_y$ peak at a different instant than $E_x$ , and the path opens up into a loop; "linear" is the degenerate ellipse, and it is structurally the statement that the two components are one oscillation viewed in rotated axes.

Step 2. Quadrature and equal: δ = +π/2, $E_1 = E_2 = E_0/\sqrt2$ . At $z = 0$ the phase is $\varphi = -\omega t$ , so $E_x = \tfrac{E_0}{\sqrt2}\cos\omega t$ and $E_y = \tfrac{E_0}{\sqrt2}\cos(-\omega t + \tfrac{\pi}{2}) = \tfrac{E_0}{\sqrt2}\sin\omega t$ :

\mathbf{E}(t) = \frac{E_0}{\sqrt2}\left[\hat{\mathbf{x}}\cos\omega t + \hat{\mathbf{y}}\sin\omega t\right],

whose tip satisfies $E_x^2 + E_y^2 = E_0^2/2$ = const: a circle traced from $+\hat{\mathbf{x}}$ toward $+\hat{\mathbf{y}}$ at angular rate ω. Circular polarization. Flipping δ → −π/2 reverses the tracing. In complex language the two states are $\tilde{\mathbf{E}}_0 \propto (\hat{\mathbf{x}} \pm i\hat{\mathbf{y}})/\sqrt2$ , and they are the natural basis whenever rotations matter (they are the eigenvectors of rotation about $\hat{\mathbf{k}}$ — quantum mechanics will call them helicity states, carrying spin $\pm\hbar$ per photon, the E7 angular-momentum ledger recalled by name).

Step 3. The general case. Eliminate the parameter from (1): with $\varphi = kz - \omega t$ , expand $E_y/E_2 = \cos\varphi\cos\delta - \sin\varphi\sin\delta$ and $\cos\varphi = E_x/E_1$ ; solving and squaring yields

\left(\frac{E_x}{E_1}\right)^2 + \left(\frac{E_y}{E_2}\right)^2 - 2\frac{E_xE_y}{E_1E_2}\cos\delta = \sin^2\delta

— a conic, bounded, hence an ellipse inscribed in the $2E_1\times2E_2$ rectangle, touching it at four points. δ = 0, π collapses it to the diagonal lines (Step 1); δ = ±π/2 aligns the axes with $\hat{\mathbf{x}},\hat{\mathbf{y}}$ (circle iff also $E_1 = E_2$ ); intermediate δ tilts it. Why the tilt and shape are left as read-offs rather than formulas: the tilt angle obeys $\tan2\psi = 2E_1E_2\cos\delta/(E_1^2 - E_2^2)$ and the axis ratio follows from it, but extracting shape parameters from field components is exactly the job the Stokes formalism (next article) mechanizes — deriving the machinery twice would be duplication, and the honest division is: states and their physics here, systematic extraction there.

Step 4. Handedness, stated so it cannot bite. The circle of Step 2 rotates from $+\hat{\mathbf{x}}$ toward $+\hat{\mathbf{y}}$ — counterclockwise as seen by an observer looking back along $-\hat{\mathbf{z}}$ toward the arriving wave... and here lives a two-century convention war. Optics names handedness from the receiver's viewpoint (looking toward the source): counterclockwise = "left circular" in that convention. Particle physics and engineering (IEEE) name it from behind, along the propagation direction: the same wave is "right-handed" (positive helicity — spin parallel to $\mathbf{k}$ ). Identical field, opposite names. The professional habit this derivation recommends: never say only "right circular"; say which way E rotates, viewed from where — or hand over the complex vector $(\hat{\mathbf{x}}+i\hat{\mathbf{y}})/\sqrt2$ , which cannot be misread. Note: the conic is confirmed as an ellipse (not a hyperbola) because its discriminant satisfies $\cos^2\delta - 1 \leq 0$ , so the curve is always bounded; the circular case checks componentwise with the $e^{-i\omega t}$ convention used throughout; and the two handedness conventions (traditional optics and IEEE) name the same physical rotation oppositely, as Step 4 details.

The second half of the subject is projection. An ideal linear polarizer — wire grid for microwaves, aligned dichroic molecules in Polaroid film for light — transmits the field component along its axis $\hat{\mathbf{p}}$ and absorbs the orthogonal one:

\mathbf{E}_{\text{out}} = (\mathbf{E}_{\text{in}}\cdot\hat{\mathbf{p}})\,\hat{\mathbf{p}} \quad\Longrightarrow\quad I_{\text{out}} = I_{\text{in}}\cos^2\theta \qquad (2)

where: θ — angle between the incoming linear polarization and the polarizer axis; intensity goes as the square of the projected amplitude (intensity is quadratic — articles 0585/0587). This is Malus's law (1809), historically the first quantitative polarization statement. Read (2) once more with modern eyes: the polarizer does not select from the beam, it projects the state — the output is a new state aligned with $\hat{\mathbf{p}}$ , at reduced amplitude. That distinction powers the most instructive tabletop experiment in optics, Worked Example 1, and it deserves a careful derivation because three standard results fall out of it at once.

Go deeper · Derivation B · 2 minThe polarizer as projector — Malus's law, the unpolarized half, and the projection staircase

Derivation BThe polarizer as projector — Malus's law, the unpolarized half, and the projection staircase

The goal is three results from one geometric operation: Malus's $I\cos^2\theta$ for linear input, the universal ½ for unpolarized input, and the staircase theorem — that a sequence of N polarizers stepping a total of 90° transmits almost everything as N grows, even though a single 90° step transmits nothing. The strategy is pure projection bookkeeping on amplitudes, squaring for intensities at the end of each stage; the staircase then iterates it. Watch the state-reset at every stage — the property that makes the staircase work and that distinguishes projection from filtering.

Step 1. Malus. Incoming linear light $\mathbf{E} = E_0\hat{\mathbf{n}}$ with $\hat{\mathbf{n}}\cdot\hat{\mathbf{p}} = \cos\theta$ : the transmitted amplitude is the projection $E_0\cos\theta$ , and intensity — quadratic in amplitude per articles 0585/0587 — is

I_{\text{out}} = I_{\text{in}}\cos^2\theta,

with the transmitted light polarized along $\hat{\mathbf{p}}$ : the memory of $\hat{\mathbf{n}}$ is gone except through the amplitude. Why the reset is physical, not notational: the absorbed orthogonal component carried the distinction between input angles; once absorbed, downstream optics can only see $\hat{\mathbf{p}}$ -light of some strength.

Step 2. The unpolarized half. Unpolarized light is a fast random mixture over polarization angles ψ (statistics formalized in 0589); each instantaneous state transmits $\cos^2\psi$ , and the detector sees the average

\langle\cos^2\psi\rangle_{\psi\in[0,2\pi)} = \tfrac12

— one half, at any polarizer orientation, and the output is fully polarized along $\hat{\mathbf{p}}$ . A polarizer is thus also a purifier: it costs unpolarized light 3 dB and hands back a pure state.

Step 3. The staircase. N polarizers at angles $90°/N, 2\cdot90°/N, \ldots, 90°$ after an initial 0° state: each stage sees light polarized along the previous axis, misaligned by only $90°/N$ , so each transmits $\cos^2(90°/N)$ and resets. The total:

T_N = \left[\cos^2\!\left(\frac{90°}{N}\right)\right]^N \xrightarrow{\ N\to\infty\ } 1, \qquad T_N \approx e^{-\pi^2/4N}\ \text{for large } N.

Numbers: $T_2 = 0.25$ , $T_6 = 0.66$ , $T_{45} = 0.95$ — forty-five gentle projections rotate the polarization a full quarter turn while losing 5%, where one abrupt projection loses 100%. Why this matters beyond the parlor trick: adiabatic-following of a slowly rotating axis is exactly how twisted-nematic LCDs steer polarization with negligible loss, and the quantum reader will recognize the Zeno-effect structure — frequent gentle measurements pin and drag a state. *Spot-check: $0.933^6 = 0.660$ and $0.99878^{45} = 0.9466$ , both matching the asymptotic $e^{-\pi^2/4N}$ at $N = 45$ (giving $0.9466$ ); the $N = 2$ case with a 45° insert returns $\tfrac{1}{8}$ , consistent with Worked Example 1.*

Two immediate corollaries of Step 1 for daily use: crossed polarizers (θ = 90°) pass nothing, and any polarizer halves unpolarized light while purifying it.

Figure

Malus's law as vector projection — and the three-polarizer cascade where an absorber creates transmission

Limiting cases

Case	Result	Meaning
δ = 0, π	line at $\psi = \arctan(\pm E_2/E_1)$	linear: the degenerate ellipse
δ = ±π/2, $E_1 = E_2$	circle, traced at ω	circular: rotation eigenstates $(\hat{\mathbf{x}}\pm i\hat{\mathbf{y}})/\sqrt2$
generic (E₁, E₂, δ)	tilted ellipse in the 2E₁×2E₂ box	the whole family; extraction machinery → 0589
E₂ → 0	linear along x̂ regardless of δ	one-component light has no relative phase to speak of
polarizer at θ	$I\cos^2\theta$ , state reset to $\hat{\mathbf{p}}$	Malus: projection, not filtration
unpolarized in	$I/2$ out, any orientation	⟨cos²⟩ over random ψ; statistics → 0589
circular through linear polarizer	$I/2$ , now linear	circle projects equally on every axis
two orthogonal linears, incoherent	no interference, ever	orthogonal states cannot beat — Fresnel–Arago

Common trapHandedness has two opposite dialects — "right circularly polarized" is ambiguous until you name your viewpoint

The same physical wave — E rotating from x̂ toward ŷ while propagating along ẑ — is called LEFT circular by traditional optics (which judges rotation looking from the detector toward the source) and RIGHT circular by particle physics and IEEE engineering (which judge along the propagation direction; this wave has spin parallel to k, positive helicity). Textbooks split roughly down the middle, instrument vendors follow IEEE, and astronomy has flipped conventions historically. The failure modes are concrete: order the wrong quarter-wave plate orientation, wire a satellite feed for the wrong circular sense (a real and expensive mistake — cross-polarized circular feeds lose ~30 dB), or mislabel photon helicity in an atomic-physics experiment. The immunization is cheap: specify the rotation operationally ("E rotates x̂ → ŷ; propagation +ẑ") or give the complex vector $(\hat{\mathbf{x}}+i\hat{\mathbf{y}})/\sqrt2$ with your $e^{-i\omega t}$ convention stated. Never let a bare "RCP" cross a collaboration boundary unchaperoned.

Worked examples

1 · The three-polarizer paradox, with numbers (confidence builder)

Unpolarized light of intensity $I_0$ meets two crossed polarizers (axes at 0° and 90°): transmission zero. Now insert a third polarizer at 45° between them. Find the output.

Step 1. First polarizer (0°): unpolarized → linear along 0°, intensity $I_1 = I_0/2$ .

Step 2. Middle polarizer (45°): Malus with θ = 45°: $I_2 = I_1\cos^245° = I_1/2 = I_0/4$ — and, crucially, the light is now polarized along 45°: projection resets the state.

Step 3. Final polarizer (90°): the incoming state is at 45° to it, not 90°: $I_3 = I_2\cos^245° = I_2/2 = \mathbf{I_0/8 = 12.5\%}$ .

Step 4. Read the paradox. Adding an absorbing element increased transmission from 0 to 12.5% — impossible for classical particles filtered by gates, natural for projections: each polarizer rewrites the state, and 45° is orthogonal to neither neighbor. This is the cleanest classical rehearsal of quantum measurement (the polarizer "measures" the state and the state updates); run with single photons the numbers are identical, with $\cos^2$ becoming a probability. Insert N polarizers stepping 90°/N each and transmission rises toward 100% as $[\cos^2(90°/N)]^N$ — rotating polarization by projection alone, the principle behind LCD twisted-nematic cells.

2 · Polarization link budgets — misaligned and mismatched antennas (exam level)

(a) A vertically polarized FM signal arrives at a receiving dipole tilted 30° from vertical. What fraction of the available power is received, in dB? (b) A GPS satellite transmits circular polarization; a user's antenna is a simple linear dipole. What is the polarization loss? (c) Why does GPS use circular at all?

Step 1. (a) Malus in receive language: the dipole responds to the projected field, power fraction $\cos^230° = 3/4$ . In decibels: $10\log_{10}(4/3) = \mathbf{1.25\ dB}$ loss — modest, which is why hand-held FM radios tolerate casual orientation.

Step 2. (b) Circular light projects equally on every transverse axis (Step 2's circle has no preferred direction): a linear antenna captures exactly half the power. Loss $= 10\log_{10}2 = \mathbf{3.0\ dB}$ , independent of the dipole's rotation about the line of sight.

Step 3. (c) The design logic: a linear space-to-ground link would suffer variable loss (0 to ∞ dB) as the satellite's orientation and the ionosphere's Faraday rotation twist the plane; circular converts that lottery into a flat, budgetable 3 dB against any linear receiver — and dual-circular receivers recover even that. Circular is the rotation-proof currency of space links. Answers: (a) 1.25 dB; (b) 3.0 dB, orientation-independent; (c) immunity to rotation ambiguity.

Origins · 1 minThomas Young and how this was discovered

Thomas Young

1773–1829 · London — who read transversality off polarization when everyone else's waves were longitudinal

Young — physician, Egyptologist (he broke ground on the Rosetta stone), and the "last man who knew everything" — gave wave optics its decisive evidence with the double slit (1801). But his subtlest contribution is this article's subject. Wave theory circa 1810 imagined light as longitudinal pressure waves in an ether, like sound — and polarization then made no sense: a longitudinal oscillation has no transverse direction to remember, yet Malus's reflected light and Iceland spar's double images plainly carried one. In an 1817 letter to Arago, Young proposed the heresy: light's vibrations might be transverse, "like the undulations of a cord" — instantly explaining why two beams polarized at right angles refuse to interfere (the Fresnel–Arago observation; orthogonal displacements cannot cancel). Fresnel, who had reached the same idea, built the full transverse-wave theory on it within four years, deriving the reflection coefficients that article 0592 owns. The transversality Maxwell's equations enforce (Gauss's law, article 0583) was thus read off polarization data four decades before those equations existed — polarization is transversality's fingerprint, and Young was the first to lift the print.

History

1669

— Bartholin describes double refraction in Iceland spar: one beam in, two out

1690

— Huygens finds the two beams differ in a "certain disposition" — polarization observed, unnamed

1808

— Malus sees reflected palace windows extinguish through a calcite crystal: polarization by reflection; his cos² law follows

1817

— Young proposes transverse vibrations to explain polarization and non-interference of orthogonal beams

1821

— Fresnel's fully transverse wave theory; elliptical and circular states classified

1845

— Faraday rotates polarization with a magnet — first hint that light is electromagnetic

1852

— Stokes parameterizes partial polarization (next article's subject)

1929

— Land's Polaroid sheet: mass-produced dichroism, polarization becomes consumer technology

1971

— twisted-nematic LCD gates light by rotating polarization — now on every desk and wrist

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §9.2.2 (polarization vector) · Hecht, Optics, 5th ed., Ch. 8 · Born & Wolf, Principles of Optics, 7th ed., §1.4.3 · Zangwill, Modern Electrodynamics, §16.4. Key self-consistency checks: the ellipse conic follows from eliminating the phase parameter in (1), with degenerate cases δ = 0, π giving lines and δ = ±π/2 giving the axis-aligned circle; the circular state $(\hat{\mathbf{x}}+i\hat{\mathbf{y}})/\sqrt2$ checks componentwise with the $e^{-i\omega t}$ convention; the three-polarizer cascade gives ½ × ½ × ½ = ⅛ and the N-polarizer limit satisfies $[\cos^2(90°/N)]^N \to 1$ ; the link-budget numbers are $\cos^2 30° = 3/4 = 1.25$ dB and circular→linear = 3.01 dB. This article covers the state taxonomy, conventions, and projection physics; systematic state extraction (ellipse tilt/shape formulas, partial polarization, measurement recipes) is deferred to Jones Vectors and Stokes Parameters (0589); polarization by reflection and Brewster's geometry to 0592; wave-plate and birefringence hardware to the optics track (E3 anisotropic-dielectrics article holds the ε-tensor groundwork); spin angular momentum of circular light is recalled from EM-15's Angular Momentum of the Electromagnetic Field (0578), not re-derived here.

Builds on: Monochromatic Plane Waves in Vacuum · The Electromagnetic Wave Equation · Complex Notation for Electromagnetic Waves.

Leads to: Jones Vectors and Stokes Parameters · The Fresnel Equations · Brewster's Angle and Total Internal Reflection · Momentum and Radiation Pressure.

Is unpolarized light a real wave state? What is actually oscillating?

It is not a state of the monochromatic plane wave at all — every strictly monochromatic wave is fully polarized (some definite ellipse). "Unpolarized" describes a statistical stream: light whose polarization ellipse wanders randomly on a timescale (the coherence time, ~femtoseconds for sunlight) far shorter than any detector's average. At each instant there is a definite transverse field; its time-averaged behavior shows no preferred direction. That is why the next article needs two descriptions: Jones vectors for pure states, Stokes parameters for the statistical mixtures — and why "unpolarized" quietly disappears whenever you filter narrowly enough in frequency and time.

Why can't two beams with perpendicular polarizations interfere?

Interference is the cross term in $|\mathbf{E}_1 + \mathbf{E}_2|^2$ , and that cross term is $2\mathbf{E}_1\cdot\mathbf{E}_2$ — a dot product. Orthogonal polarizations have $\mathbf{E}_1\cdot\mathbf{E}_2 = 0$ identically: the total intensity is exactly $I_1 + I_2$ everywhere, no fringes, regardless of coherence. This is the Fresnel–Arago law that forced Young's transversality hypothesis: longitudinal waves (sound) cannot have mutually orthogonal displacement states along the same path, so their interference never switches off this way. The two components don't cancel — but they do combine: coherent orthogonal beams superpose into a single polarization state (generally elliptical), which a subsequent polarizer can then make interfere by projecting both onto one axis — the working principle of polarization interferometry.

What actually happens inside a Polaroid sheet — why does one component die?

The sheet is an array of long, conducting-ish molecules (iodine-doped stretched polymer), all aligned. The field component along the chains drives electrons freely down them — that component does work, is absorbed, and exits attenuated by orders of magnitude. The component across the chains cannot drive significant current and passes. Note the consequence for intuition: the transmission axis is perpendicular to the molecular chains. A wire-grid polarizer for microwaves works identically (and predates Polaroid: Hertz used one), with the same perpendicular rule — the "picket fence lets the aligned wave through" cartoon has the physics exactly backwards.

How do I tell circular polarization from unpolarized light? Both give half through any polarizer.

Add one element with memory of phase: a quarter-wave plate. QWP converts circular light into linear light (it undoes the quadrature: δ = ±π/2 → 0), so QWP + rotating polarizer gives full extinction at some angle — the smoking gun of circularity. Unpolarized light shrugs off the wave plate (its random δ stays random) and still transmits ½ at every angle. This two-element test is the operational heart of the Stokes-parameter measurement recipe (next article): polarizer alone measures linear content; wave plate + polarizer measures circular content; together they fingerprint any state, pure or mixed.

Does polarization carry energy or momentum differently between states?

Energy and linear momentum: no — intensity $I$ and pressure $I/c$ depend on $E_1^2 + E_2^2$ only, blind to δ. Angular momentum: yes, and it is the one mechanical distinction. Circular states carry spin angular momentum $\pm\hbar$ per photon along $\mathbf{k}$ (linear states carry zero on average; ellipses interpolate), torquing absorbers and birefringent plates measurably — Beth's 1936 experiment twisted a suspended wave plate with circularly polarized light. The books for that live in EM-15's angular-momentum article; the polarization state is simply where the field stores the handedness that ledger prices.

Why is the blue sky polarized, and why don't I notice?

Molecular (Rayleigh) scattering re-radiates sunlight from induced dipoles, and dipoles do not radiate along their own axis: light scattered by 90° can only carry the polarization component perpendicular to the scattering plane, so the sky at 90° from the Sun is strongly (up to ~75%) linearly polarized along a great-circle pattern. You don't notice because human vision is polarization-blind (mostly — a faint yellow bowtie, Haidinger's brush, is visible to trained observers against polarized skies). Bees and many insects read the pattern as a sky-wide compass; photographers exploit it with a rotating filter to darken skies at right angles to the Sun; and the same geometry, applied at reflection instead of scattering, becomes Brewster's angle in article 0592.

Is there a longitudinal polarization state hiding anywhere — a third option?

Not for free electromagnetic plane waves: Gauss's law kills the longitudinal component outright (article 0583's derivation), leaving exactly two transverse states per mode — which becomes, in quantum language, the photon's two helicities and no third. The near-misses are instructive. In matter: plasmas support longitudinal charge-density oscillations (plasmons), but those are matter waves dressed in fields, not free radiation. In waveguides (0599–0601): fields develop longitudinal components, but the two-mode counting per propagating mode survives. Hypothetically: a photon with mass would unlock a third, longitudinal polarization — its absence underwrites the measured photon-mass bound of article 0583. Two states is not a convention; it is Gauss's law counted.