Jones Calculus

We take one idea from electromagnetism and turn it into an entire engineering language. A monochromatic plane wave travelling along $z$ has an electric field confined to the transverse plane, so its state is fixed by just two complex numbers — the amplitudes and relative phase of the $x$ and $y$ components. The companion article aq-em-jones-vectors-stokes-parameters builds exactly this object from Maxwell's equations and names it the Jones vector $\mathbf{J} = (\tilde E_x, \tilde E_y)^{T}$ ; that two-component field is our starting point, recalled by name and not re-derived. What is ours here is what to do with it. Because Maxwell's equations are linear and passive optical elements act linearly on the field, every polarizer, wave plate, and rotator is a two-by-two complex matrix — the Jones matrix — and a train of elements is simply their ordered product. A question that would be a trigonometric ordeal by the second wave plate becomes one matrix multiplication: the output state of $N$ elements is $\mathbf{J}_{\text{out}} = J_N \cdots J_2 J_1 \mathbf{J}_{\text{in}}$ , with the element the light meets first sitting nearest the input. Malus's law, the quarter-wave plate's linear-to-circular conversion, and a half-wave plate's rotation all fall out as one-line special cases. The power is the economy: reflection, retardance, and rotation reduce to a handful of canonical matrices; rotating an element to a new angle is a similarity transform; and the states an element leaves unchanged are just the eigenvectors of its matrix. The one price is coherence — a Jones vector describes only fully polarized light. The moment phases wander (sunlight, LEDs, multimode lasers) the amplitudes stop being definite and we must move to the Stokes vector and Mueller matrix of the next articles. Within its domain, though, Jones calculus turns all of polarization optics into linear algebra.

Key takeaways

Recalled from electromagnetism: a fully polarized plane wave is a two-component complex Jones vector $\mathbf{J} = (\tilde E_x, \tilde E_y)^{T}$ , built from the transverse field in aq-em-jones-vectors-stokes-parameters; the element calculus below is optics'.
Every passive element is a $2\times2$ complex matrix; a system is the ordered product $J_N \cdots J_1$ , with the first element the light meets nearest the input vector.
Polarizer, retarder, and rotator have compact standard matrices; rotating an element to angle $\theta$ is the similarity transform $R(-\theta)\,J\,R(\theta)$ .
A quarter-wave plate at 45° sends $(1,0)^{T} \to \tfrac{1}{\sqrt2}(1,i)^{T}$ (circular); a half-wave plate reflects linear light about its fast axis, rotating it by $2\theta$ ; a polarizer at $\theta$ reproduces Malus's $\cos^2\theta$ .
The formalism is exact only for fully polarized coherent light; partial polarization and depolarization need Stokes vectors and Mueller matrices.
Key idea: polarization state = a 2-vector, every element = a 2×2 matrix, a system = the ordered product — all of fully-polarized-light optics becomes linear algebra.

The core insight

A fully polarized beam is a two-component complex Jones vector; every passive optical element is a two-by-two complex matrix; a system is the ordered matrix product acting on the input vector. Cascaded polarizers, wave plates, and rotators — and Malus's law, circular-light generation, and polarization rotation with them — all become linear algebra.

Where this lives today

Jones calculus is the daily arithmetic of anyone who designs with polarization. A variable attenuator (rotatable half-wave plate plus polarizing cube), a Lyot filter, a laser-cavity polarization controller, or a liquid-crystal display pixel is analyzed by multiplying its constituent matrices and reading off the output vector. Ellipsometry, the standard thin-film metrology, models a reflecting surface as a diagonal Jones matrix $\mathrm{diag}(r_p, r_s)$ and fits the measured complex ratio $\rho = r_p/r_s = \tan\Psi\,e^{i\Delta}$ to Fresnel coefficients to extract film thickness and index to sub-nanometre precision. The calculus has found a second life in quantum optics. A single photon's polarization is a qubit living in the same two-dimensional complex space, and the Jones matrices of lossless elements are precisely the unitary single-qubit gates: a half-wave plate at angle $\theta$ is a controllable rotation, and quarter- and half-wave plates together generate any $SU(2)$ transformation. Polarization-encoded quantum key distribution, photonic quantum computing, and Bell-inequality tests all set their gates with wave plates whose action is computed exactly as Jones prescribed in 1941 — a striking case of a classical engineering tool becoming the hardware description of a quantum technology.

The physical picture

Think of the polarization state as an arrow in a two-dimensional complex plane: the ratio of the two component magnitudes sets the ellipse's shape, and their relative phase sets its tilt and handedness. Horizontal light is $(1,0)^{T}$ , vertical is $(0,1)^{T}$ , light at 45° is $\tfrac{1}{\sqrt2}(1,1)^{T}$ , and the circular states carry a quarter-cycle phase between the components, $\tfrac{1}{\sqrt2}(1, \pm i)^{T}$ . Every possible fully polarized state is one such vector.

Figure

The Jones vector names a polarization ellipse: horizontal (1,0), vertical (0,1), diagonal (1,1)/√2, and the two circular states (1, ±i)/√2 whose i is a quarter-cycle phase

An optical element rotates and stretches this arrow — a linear operation, hence a matrix. A wave plate does nothing to the amplitudes in its own axis frame; it only delays one component, so its matrix is diagonal with unit-modulus entries $e^{-i\Gamma/2}, e^{+i\Gamma/2}$ . A polarizer projects onto its transmission axis, so its matrix is an outer product $\hat{\mathbf{t}}\hat{\mathbf{t}}^{T}$ . The states an element passes untouched (up to a scalar) are its eigenvectors — the fast and slow axes of a wave plate, the two circular states of a rotator. Diagonalizing the matrix literally is finding an element's eigen-polarizations. Order matters, because matrices do not commute: a quarter-wave plate followed by a polarizer is a different device from a polarizer followed by the plate. And a half-wave plate has a startlingly simple geometric action — it reflects the polarization across its fast axis, so turning the plate turns the output twice as fast.

Figure

A half-wave plate reflects linear light about its fast axis: for horizontal input the output angle is exactly twice the fast-axis angle — a straight line of slope 2, so 20° gives 40° and 45° swings the light to vertical

The core equations

A fully polarized state is two complex numbers; scale away the common brightness and phase and the physical state is two-dimensional. This is the object recalled from the electromagnetic article.

\mathbf{J} = \begin{pmatrix} E_{0x}\,e^{i\phi_x} \\ E_{0y}\,e^{i\phi_y}\end{pmatrix}, \qquad \hat{\mathbf{x}} = \begin{pmatrix}1\\0\end{pmatrix},\ \hat{\mathbf{y}} = \begin{pmatrix}0\\1\end{pmatrix},\ \hat{\mathbf{R}} = \tfrac{1}{\sqrt2}\begin{pmatrix}1\\i\end{pmatrix},\ \hat{\mathbf{L}} = \tfrac{1}{\sqrt2}\begin{pmatrix}1\\-i\end{pmatrix} \qquad (1)

where: $E_{0x}, E_{0y}$ — real component amplitudes; $\phi_y - \phi_x$ — the relative phase that fixes ellipticity and handedness; $\hat{\mathbf{R}}, \hat{\mathbf{L}}$ — right- and left-circular states in the $e^{i(kz-\omega t)}$ convention, matching aq-em-jones-vectors-stokes-parameters. Intensity is $\mathbf{J}^\dagger\mathbf{J}$ ; two states never interfere when $\mathbf{J}_1^\dagger\mathbf{J}_2 = 0$ , and $\hat{\mathbf{R}}^\dagger\hat{\mathbf{L}} = 0$ — the circular states are as good a basis as H and V. A cascade of elements is an ordered product, and the ordering trips everyone up once: the element the light hits first is written last, nearest the input vector.

\mathbf{J}_{\text{out}} = J_N J_{N-1} \cdots J_2 J_1 \, \mathbf{J}_{\text{in}} \qquad (2)

where: $J_k$ — Jones matrix of the $k$ -th element encountered; $J_1$ — the first element, adjacent to $\mathbf{J}_{\text{in}}$ . Because matrix multiplication is non-commutative, swapping the order of two elements generally changes the device. To rotate an element from its own principal frame to an azimuth $\theta$ in the lab, conjugate by the rotation matrix — rotate the field in, apply the element, rotate back.

R(\theta) = \begin{pmatrix}\cos\theta & \sin\theta \\ -\sin\theta & \cos\theta\end{pmatrix}, \qquad J(\theta) = R(-\theta)\,J\,R(\theta) \qquad (3)

where: $R(\theta)$ — the $2\times2$ coordinate rotation; $J$ — the element's matrix in its eigen-frame; $J(\theta)$ — the same element at angle $\theta$ . This one similarity transform generates every off-axis matrix from its diagonal parent. The three canonical elements — retarder, polarizer, rotator — are the whole toolbox. The retarder is diagonal; the polarizer is a projection; the rotator is an ordinary rotation.

J_{\text{ret}}(\Gamma) \propto \begin{pmatrix} e^{-i\Gamma/2} & 0 \\ 0 & e^{+i\Gamma/2}\end{pmatrix}, \quad J_{\text{pol}}(\theta) = \begin{pmatrix}\cos^2\theta & \cos\theta\sin\theta \\ \cos\theta\sin\theta & \sin^2\theta\end{pmatrix}, \quad J_{\text{rot}}(\rho) = \begin{pmatrix}\cos\rho & -\sin\rho \\ \sin\rho & \cos\rho\end{pmatrix} \qquad (4)

where: $\Gamma$ — retardance between slow and fast axes (from the wave-plate article); $J_{\text{pol}}(\theta) = \hat{\mathbf{t}}\hat{\mathbf{t}}^{T}$ — projection onto the transmission axis $\hat{\mathbf{t}} = (\cos\theta, \sin\theta)^{T}$ ; $\rho$ — the angle a circular-birefringent rotator turns the plane (optical activity, Faraday effect). Setting $\Gamma = \pi/2$ or $\pi$ recovers the quarter- and half-wave plates, $\mathrm{diag}(1,i)$ and $\mathrm{diag}(1,-1)$ .

The element calculus, worked from the matrices

Derivation AMalus's law from the polarizer projection

The aim is to show that the familiar $\cos^2$ transmission law of a polarizer is not a separate physical postulate but a one-line consequence of the projection matrix — recovering Malus's Law (the sibling article) from pure linear algebra and confirming that our polarizer matrix is the right object. The strategy is to send horizontally polarized light into a polarizer whose transmission axis makes angle $\theta$ with the horizontal, compute the output Jones vector, and take its squared modulus.

Step 1. Write the input as horizontal, $\mathbf{J}_{\text{in}} = E_0 (1, 0)^{T}$ , and apply the polarizer $J_{\text{pol}}(\theta) = \hat{\mathbf{t}}\hat{\mathbf{t}}^{T}$ . The product is $\mathbf{J}_{\text{out}} = E_0\cos\theta\,(\cos\theta, \sin\theta)^{T}$ . Why: a projection onto $\hat{\mathbf{t}}$ keeps only the component of the input along $\hat{\mathbf{t}}$ , whose length is $E_0\cos\theta$ , and points the result along $\hat{\mathbf{t}}$ — the transmitted light is polarized along the analyzer axis, exactly as observed.

Step 2. Take the intensity, $I(\theta) = \mathbf{J}_{\text{out}}^\dagger \mathbf{J}_{\text{out}} = E_0^2\cos^2\theta\,(\cos^2\theta + \sin^2\theta) = E_0^2\cos^2\theta$ . Why: the transmitted vector has length $E_0\cos\theta$ , so its intensity is that squared — the Pythagorean $\cos^2 + \sin^2 = 1$ cleans up the components — giving Malus's law with $I_0 = E_0^2$ .

Step 3. Read the consequences. At $\theta = 0$ full transmission, at $\theta = 90^\circ$ complete extinction, and crossed polarizers pass nothing — until you slip an element between them, the subject of the next example. Why: the projection is total, so ideal crossed polarizers have infinite extinction ratio; any real leakage is a departure of the true matrix from this ideal. *Check: $J_{\text{pol}}(\theta)\,(1,0)^{T} = \cos\theta\,(\cos\theta,\sin\theta)^{T}$ and $I(\theta) = \cos^2\theta$ , confirming Malus's law symbolically.*

Go deeper · Derivation B · 1 minA quarter-wave plate makes circular light

Derivation BA quarter-wave plate makes circular light

This card proves the headline action of the quarter-wave plate in the Jones language: horizontal light through a quarter-wave plate at 45° emerges circular. It matters because it turns a claim about phase lags into an explicit vector identity, and because it fixes the handedness against the stated time convention — the point where careless sign-keeping produces the wrong circular sense. The method is to rotate the diagonal quarter-wave matrix to 45° and apply it to horizontal input, or equivalently work in the plate frame where the input is already at 45° to the axes.

Step 1. In the plate's frame a quarter-wave plate is $\mathrm{diag}(1, i)$ (up to a global phase), and horizontal light at 45° to its axes is $\tfrac{1}{\sqrt2}(1,1)^{T}$ . The product is $\tfrac{1}{\sqrt2}(1, i)^{T} = \hat{\mathbf{R}}$ . Why: the slow axis is multiplied by $i$ , a quarter-cycle delay, while the input's two equal components are exactly what a circle requires — equal amplitudes in quadrature.

Step 2. Confirm it is circular and not merely elliptical: the two components have equal modulus $1/\sqrt2$ and a relative phase of $90^\circ$ , and the intensity $\mathbf{J}^\dagger\mathbf{J} = \tfrac12(1 + 1) = 1$ is preserved. Why: equal amplitude plus quarter-cycle phase is the definition of circular polarization, and a lossless (unitary) plate cannot change the total intensity.

Step 3. Read the handedness: $\hat{\mathbf{R}} = \tfrac{1}{\sqrt2}(1,i)^{T}$ has a field tip that turns $\hat{\mathbf{x}} \to \hat{\mathbf{y}}$ in the $e^{i(kz-\omega t)}$ convention — the right-circular state of aq-em-jones-vectors-stokes-parameters. Feed $-45^\circ$ linear (or flip the fast axis) and you get $\hat{\mathbf{L}}$ ; send $\hat{\mathbf{R}}$ back through and linear light returns. Why: the quarter-wave plate is its own near-inverse on the linear–circular pair, which is why one plate both makes and analyzes circular light. *Check: $\mathrm{diag}(1,i)\,\tfrac{1}{\sqrt2}(1,1)^{T} = \tfrac{1}{\sqrt2}(1,i)^{T}$ — equal moduli, $90^\circ$ relative phase, unit intensity, confirming circular polarization and energy conservation.*

Go deeper · Derivation C · 1 minThe half-wave plate as a reflection that rotates by twice its angle

Derivation CThe half-wave plate as a reflection that rotates by twice its angle

Here we establish the half-wave plate's defining property — it rotates linear polarization by twice the angle between the input and its fast axis — by showing the plate matrix is literally a mirror reflection. The result is worth a full derivation because the factor of two is the single most useful fact about half-wave plates and because "reflection, not rotation" explains subtleties a naive rotation picture gets wrong. We set the retardance to $\pi$ , rotate the plate to angle $\theta$ , and apply it to general linear input.

Step 1. With $\Gamma = \pi$ the plate is $\mathrm{diag}(1, -1)$ up to a global phase — a mirror across the fast ( $x$ ) axis. Rotate it to angle $\theta$ : $J = R(-\theta)\,\mathrm{diag}(1,-1)\,R(\theta) = \begin{pmatrix}\cos2\theta & \sin2\theta\\ \sin2\theta & -\cos2\theta\end{pmatrix}$ . Why: conjugating a reflection by rotations moves the mirror line to angle $\theta$ ; the $2\theta$ entries are the signature of a reflection about a line at $\theta$ .

Step 2. Apply it to linear light at angle $\alpha$ , $\mathbf{J}_{\text{in}} = (\cos\alpha, \sin\alpha)^{T}$ . The matrix product simplifies with angle-addition identities to $(\cos(2\theta - \alpha), \sin(2\theta - \alpha))^{T}$ — still linear, now at $2\theta - \alpha$ . Why: reflecting a direction $\alpha$ across a line at $\theta$ lands it at $2\theta - \alpha$ ; the output stays fully linear because a half-wave plate has real eigenvalues $\pm1$ .

Step 3. Turn the plate by $\Delta\theta$ and the output swings by $2\Delta\theta$ — twice as fast. Between crossed polarizers, a half-wave plate at $\theta$ transmits $\sin^2 2\theta$ , opening fully at $\theta = 45^\circ$ where horizontal input is rotated to vertical. Why: the mirror line moves at the plate's rate, so the reflected direction moves at twice that rate — the geometric origin of the factor of two, and the basis of continuous lossless attenuation. *Check: $J_{\text{HWP}}(\theta)\,(\cos\alpha,\sin\alpha)^{T} = (\cos(2\theta-\alpha),\sin(2\theta-\alpha))^{T}$ , and crossed-polarizer transmission is $\sin^2 2\theta$ , both confirmed symbolically.*

Limiting cases

Each canonical element is a memorable matrix, and the special cases below are the ones worth knowing cold.

Case	Matrix	Meaning
Ideal polarizer along $x$	$\begin{pmatrix}1 & 0\\ 0 & 0\end{pmatrix}$	transmits $x$ , blocks $y$ (infinite extinction); a real one has a small lower-right entry
Half-wave plate	$\begin{pmatrix}1 & 0\\ 0 & -1\end{pmatrix}$	reflects linear about the fast axis: $\alpha \to -\alpha$ , a $2\alpha$ rotation
Quarter-wave plate	$\begin{pmatrix}1 & 0\\ 0 & i\end{pmatrix}$	45° linear $\leftrightarrow$ circular
Identity / free space	$e^{-i\phi}\,I$	isotropic phase, polarization unchanged
Crossed polarizers	$J_{\perp} J_{\parallel} = 0$	zero transmission — until an element between them rotates the state

Common trap“Just average two Jones vectors to describe unpolarized light”

A Jones vector is a definite field with a fixed relative phase — it can only ever describe fully polarized light. The tempting shortcut "sunlight is half horizontal and half vertical, so write $(1,1)/\sqrt2$ " produces 45° linear light, which sunlight emphatically is not: coherent addition of amplitudes and incoherent mixing of intensities are different physical operations. The 45°-linear beam passes a 45° polarizer completely; sunlight passes any polarizer at one half. Unpolarized and partially polarized light have no Jones vector at all — you must move to the Stokes vector and Mueller matrix, which average intensities, not amplitudes. And even within Jones calculus, mind the convention: $e^{-i\omega t}$ versus $e^{+i\omega t}$ flips every retardance sign and swaps which state you call right-circular, so always record it with your matrices.

Worked examples

1 · Building and checking a circular-light generator (confidence builder)

Horizontal light $(1,0)^{T}$ passes through a quarter-wave plate with its fast axis at 45°. Find the output state, confirm it is circular, and verify no intensity was lost.

Step 1. Rotate the quarter-wave matrix to 45°: $J = R(-45^\circ)\,\mathrm{diag}(1,i)\,R(45^\circ) = \tfrac12\begin{pmatrix}1+i & 1-i\\ 1-i & 1+i\end{pmatrix}$ . Apply it to $(1,0)^{T}$ : $\mathbf{J}_{\text{out}} = \tfrac12(1+i,\,1-i)^{T}$ .

Step 2. Factor out $\tfrac{1+i}{2}$ : $\mathbf{J}_{\text{out}} = \tfrac{1+i}{2}\,(1, -i)^{T} = \tfrac{1+i}{\sqrt2}\cdot\tfrac{1}{\sqrt2}(1,-i)^{T}$ , which is $\hat{\mathbf{L}}$ up to a unit-modulus global phase $(1+i)/\sqrt2 = e^{i\pi/4}$ . The output is circularly polarized.

Step 3. Confirm equal amplitudes and quadrature: both components have modulus $\tfrac{1}{\sqrt2}|1+i|/\sqrt2 = \tfrac{1}{\sqrt2}$ , and their phase difference is $90^\circ$ . Intensity: $\mathbf{J}_{\text{out}}^\dagger\mathbf{J}_{\text{out}} = \tfrac14(|1+i|^2 + |1-i|^2) = \tfrac14(2 + 2) = 1$ , equal to the input — nothing lost.

Step 4. Handedness bookkeeping: working in the plate frame instead, horizontal light is at $45^\circ$ to the axes and $\mathrm{diag}(1,i)\tfrac{1}{\sqrt2}(1,1)^{T} = \tfrac{1}{\sqrt2}(1,i)^{T} = \hat{\mathbf{R}}$ ; the lab-frame calculation gives $\hat{\mathbf{L}}$ because the plate's fast axis sits on the other side of the input. Flip the fast axis to $-45^\circ$ to swap the handedness. Output is circular, intensity preserved; the sense is set by which side of the input the fast axis lies.

2 · Crossed polarizers with a half-wave plate between them (exam level)

Two crossed polarizers (transmission axes horizontal and vertical) block all light. A half-wave plate is inserted between them with its fast axis at angle $\theta$ . Find the transmitted fraction and the angle of maximum transmission.

Step 1. Build the chain right-to-left: input horizontal $(1,0)^{T}$ , then the first polarizer (horizontal, passes it), then the half-wave plate $J_{\text{HWP}}(\theta)$ , then the vertical analyzer $J_{\perp} = \mathrm{diag}(0,1)$ . The output is $\mathbf{J}_{\text{out}} = J_\perp J_{\text{HWP}}(\theta)\,(1,0)^{T}$ .

Step 2. The half-wave plate rotates the horizontal input to $2\theta$ , giving $(\cos 2\theta, \sin 2\theta)^{T}$ ; the vertical analyzer keeps only the $y$ -component, $\sin 2\theta$ . The transmitted intensity is $I(\theta) = \sin^2 2\theta$ .

Step 3. Maximum at $2\theta = 90^\circ$ , i.e. $\theta = 45^\circ$ , where $I = 1$ : the plate has rotated the horizontal input all the way to vertical, so the "crossed" analyzer now passes everything. Minimum (zero) at $\theta = 0^\circ$ and $\theta = 90^\circ$ .

Step 4. This is a sensitive optical switch: a small plate rotation near $\theta = 0$ opens the crossed pair as $\sin^2 2\theta \approx (2\theta)^2$ , a fast quadratic turn-on, which is why half-wave plates between crossed polarizers make excellent nulling and modulation elements. Transmission $\sin^2 2\theta$ , maximal at $\theta = 45^\circ$ .

Physics of the portrait

The calculus is named for the man who, in a burst of eight papers, replaced a generation of polarization trigonometry with two-by-two matrices.

Origins · 1 minjones vector and how this was discovered

jones vector

the two-component state of polarized light · R. Clark Jones, 1941

Robert Clark Jones (1916–2004) was an American physicist at Polaroid and later the Polaroid research laboratories when, in 1941, aged twenty-five, he published the first of a celebrated series of eight papers in the Journal of the Optical Society of America titled "A new calculus for the treatment of optical systems." His insight was to represent the fully polarized field of a plane wave as a two-component complex column — now the Jones vector — and every lossless or lossy optical element as a two-by-two complex matrix, so that a train of polarizers, retarders, and rotators became a single matrix product. The economy was transformative: eigen-polarizations became eigenvectors, cascades became products, and the design of polarization optics became linear algebra. His near-contemporary Hans Mueller built the complementary calculus for partially polarized light, using four-by-four real matrices on the Stokes vector, and the two formalisms have divided polarization science ever since — Jones when the phase is trackable, Mueller when it is not. Jones's own vector turned out to be, in modern language, exactly a single-qubit state, and his matrices exactly the unitary single-qubit gates, so his 1941 bookkeeping is now taught as the classical ancestor of quantum information with photons.

History

1852

— Stokes gives four measurable parameters for partially polarized light

1892

— Poincaré maps polarization states onto a sphere

1941

— R. Clark Jones publishes the first of eight papers defining the two-by-two matrix calculus

1943

— Hans Mueller develops the complementary four-by-four calculus for partially polarized light

1956

— Jones completes the eight-paper series in the Journal of the Optical Society of America

Going further

Texts. Hecht, Optics, 5th ed., §8.13 (Jones vectors and matrices) · Saleh & Teich, Fundamentals of Photonics, 3rd ed., §6.1 · Goldstein, Polarized Light, 3rd ed., Ch. 11 · R. C. Jones, J. Opt. Soc. Am. 31, 488 (1941). The time convention throughout is $e^{i(kz-\omega t)}$ , matching aq-em-jones-vectors-stokes-parameters, from which the two-component field and the $\hat{\mathbf{R}} = (1,i)/\sqrt2$ convention are recalled by name. Key symbolic results: the polarizer projection gives Malus's $\cos^2\theta$ ; the quarter-wave plate maps $(1,1)^{T}/\sqrt2 \to (1,i)^{T}/\sqrt2$ ; the half-wave plate rotates linear light by $2\theta$ ; and crossed polarizers with a half-wave plate between them transmit $\sin^2 2\theta$ . All matrices are standard; the retardances $\Gamma = \pi/2, \pi$ are exact.

Prerequisites. Polarization of Light · Linear Polarization · Circular Polarization · Wave Plates and Retarders. Leads to. Stokes Parameters and the Poincaré Sphere · Mueller Calculus · Malus's Law · Birefringence.

Why is the first element the light meets written last in the matrix product?

Because each matrix acts on the vector to its right. Element 1 transforms the input, $\mathbf{J}_1 = J_1\mathbf{J}_{\text{in}}$ ; element 2 then acts on that, $\mathbf{J}_2 = J_2 J_1 \mathbf{J}_{\text{in}}$ ; and so on, so the product grows leftward and the physically-first element ends up rightmost, adjacent to the input. It reads backwards relative to a left-to-right optical layout, which is a reliable source of sign and ordering errors — the fix is to always build the product starting from the input vector and prepend each new element as the light meets it.

How do I get the Jones matrix of an element rotated to some angle without re-deriving it?

Use the similarity transform $J(\theta) = R(-\theta)\,J\,R(\theta)$ , where $J$ is the element's matrix in its own principal-axis frame and $R(\theta)$ is the standard $2\times2$ rotation. It rotates the field into the element's frame, applies the simple diagonal (or projection) matrix there, and rotates back to the lab. Every off-axis polarizer and wave plate in this article is generated this way from its on-axis parent, so you only ever memorize the diagonal forms and one rotation matrix.

What exactly are the eigen-polarizations of an element, and why do I care?

They are the states the element leaves unchanged except for an overall scalar — the eigenvectors of its Jones matrix. For a wave plate they are the fast and slow linear axes (each just picks up a phase); for a rotator they are the two circular states; for a polarizer, its transmission and blocked axes. They matter because a beam prepared in an eigen-polarization passes through undisturbed, and because writing a general input in the eigen-basis makes the element's action trivial — you multiply each eigen-component by its eigenvalue. Diagonalizing the matrix is literally finding these special states.

Does Jones calculus describe reflection and Fresnel coefficients too?

Yes, for a single reflection off a smooth surface: the surface acts as a diagonal Jones matrix $\mathrm{diag}(r_p, r_s)$ built from the Fresnel amplitude coefficients (the subject of the Fresnel-equations article), one entry for the in-plane $p$ component and one for the perpendicular $s$ component. This is the engine of ellipsometry, which measures the complex ratio $r_p/r_s$ . What Jones calculus cannot do is handle a depolarizing reflection — a rough surface or a scattering sample that scrambles the phase — because that produces partially polarized light, which needs the Mueller description.

Is a Jones vector really the same thing as a quantum qubit?

Structurally, yes. Both are unit vectors in a two-dimensional complex Hilbert space, both have an irrelevant overall phase, and both are acted on by unitary two-by-two matrices — the Jones matrices of lossless elements are exactly single-qubit gates, and the Poincaré sphere is the Bloch sphere. A single photon's polarization is a physical qubit, and wave plates are its gates, which is why polarization is a favourite encoding for photonic quantum information. The one caveat is loss: a polarizer's matrix is not unitary because it discards light, corresponding to a measurement (projection) rather than a reversible gate.

Why does my textbook's quarter-wave matrix read diag(1, −i) instead of diag(1, i)?

Two conventions differ silently between books: the time factor $e^{-i\omega t}$ versus $e^{+i\omega t}$ , and whether the retardance is put on the slow axis as an extra lag or on the fast axis as a lead. Either choice flips the sign of the off-diagonal phase and swaps which circular state you call right versus left — so $\mathrm{diag}(1,i)$ in one convention is $\mathrm{diag}(1,-i)$ in another, describing the same physical plate. This article uses $e^{i(kz-\omega t)}$ with the extra phase on the slow axis, matching the electromagnetic companion. When comparing sources, reconcile the two conventions first; nothing physical is at stake, only the R/L label.