Lorentz Transformations

The postulates have issued their demand: find the transformation between inertial frames under which Maxwell's equations — and in particular the electromagnetic wave operator — keep their form exactly. The previous article showed Galileo's transformation failing that test, a cross term surviving in the wave equation like a seam in what should be smooth. This article finds the transformation that passes. It exists, it is unique once linearity and isotropy are granted, and it is the Lorentz transformation: $ct'=\gamma(ct-\beta x)$ , $x'=\gamma(x-\beta ct)$ , the pair of formulas on which everything relativistic in physics runs. The derivation route chosen here is deliberately the electrodynamic one — the road actually walked between 1887 and 1905. Voigt first wrote a version of these formulas as the substitution leaving the wave equation invariant; Lorentz found them as the transformation under which Maxwell's equations tolerate a moving frame; Poincaré polished them into an exact group and gave them their name — all before, or independently of, Einstein's kinematic derivation from the two postulates (which the companion mechanics article performs, and which this one deliberately does not repeat). Demanding that the d'Alembertian $\partial_x^2-\tfrac{1}{c^2}\partial_t^2$ transform into itself fixes every coefficient, and $\gamma$ emerges from the algebra like a bill coming due. With the transformation in hand, the famous kinematic corollaries — dilation, contraction, the velocity-addition law with its fixed point at $c$ — are read off in lines, and the machinery is staged for the payoff this track cares about: four-vectors, the transformation of $\mathbf{E}$ and $\mathbf{B}$ into each other, and the tensor form of Maxwell's equations in which invariance stops being a theorem and becomes a typeface.

Key takeaways

The standard boost along $x$ : $ct'=\gamma(ct-\beta x)$ , $x'=\gamma(x-\beta ct)$ , $y'=y$ , $z'=z$ , with $\beta=v/c$ , $\gamma=1/\sqrt{1-\beta^2}$ — the unique linear, isotropy-respecting map that leaves the electromagnetic wave operator (equivalently the interval) form-invariant.
This article derives it the electrodynamic way: demand $\partial_x^2-\tfrac{1}{c^2}\partial_t^2=\partial_{x'}^2-\tfrac{1}{c^2}\partial_{t'}^2$ under a general linear map — the requirement Galileo failed — and solve for the coefficients.
Velocity addition follows by differentials: $u'=\dfrac{u-v}{1-uv/c^2}$ , with $u=c$ a fixed point — light stays light, the postulate closing its own loop.
Time dilation and length contraction are one-line read-offs (the full kinematic constructions live in the mechanics companions): $\Delta t=\gamma\Delta\tau$ , $L=L_0/\gamma$ .
Matrix form: $x'^\mu=\Lambda^\mu{}_\nu x^\nu$ with $\Lambda^{\mathsf{T}}\eta\Lambda=\eta$ , $\eta=\mathrm{diag}(+1,-1,-1,-1)$ — the defining condition of the Lorentz group; rapidity $\phi=\mathrm{artanh}\,\beta$ makes boosts hyperbolic rotations with additive angle.
Every four-component object in the chapters ahead — four-current, four-potential, the field tensor — transforms with this same $\Lambda$ : one matrix, all of electrodynamics.
Key idea: the Lorentz transformation is the symmetry group of the wave operator Maxwell built — found inside electrodynamics before relativity existed, and the reason $\mathbf{E}$ and $\mathbf{B}$ must mix.

The core insight

A Lorentz transformation is the unique linear map between inertial coordinates that preserves the Minkowski interval $s^2=(c\Delta t)^2-\Delta x^2-\Delta y^2-\Delta z^2$ ; a boost is a hyperbolic rotation in a time–space plane, parameterized by rapidity $\phi$ with $\tanh\phi=\beta$ .

Where this lives today

The transformation is the most heavily exercised formula in experimental physics. Every collider event reconstruction is Lorentz bookkeeping — lab frame to center-of-momentum and back, detector-frame lifetimes to proper lifetimes ( $b$ -quark tagging lives on $\gamma c\tau$ vertex displacements); every beam-dynamics code pushes particle bunches with (2); and the LHC's data pipelines apply the boost trillions of times per run as unremarkably as addition. Precision tests keep tightening: time dilation verified at $\gamma$ -dependent parts in $10^9$ with stored lithium ions; the boost's isotropy at $10^{-18}$ with optical resonators; and the group structure itself probed wherever Lorentz-violation parameter tables are filled with new zeros. For this track the transformation is the tool about to be used. The next article geometrizes it — interval, causal structure, the arena. Then the harvest: charge–current $(c\rho,\mathbf{J})$ and the potentials $(V/c,\mathbf{A})$ assemble into four-vectors transforming by $\Lambda$ ; the fields, which are derivatives of potentials, must therefore transform with two factors of $\Lambda$ — the antisymmetric field tensor $F^{\mu\nu}$ — and the boost of (2) applied to a pure Coulomb field will manufacture a magnetic field before your eyes, resolving the magnet-and-conductor asymmetry by explicit computation and exposing magnetism as electrostatics seen from a moving chair. That $\mathbf{E}$ – $\mathbf{B}$ mixing is not new physics bolted on: it is (2), applied twice.

Origins · 1 minHenri Poincaré and how this was discovered

Henri Poincaré

1854–1912 · Paris — who named the group, proved it closed, and stopped one step short

Poincaré came at the transformation as the mathematician of the age auditing Lorentz's electron theory. His 1905 papers ("Sur la dynamique de l'électron," the short note and the great Palermo memoir) repaired Lorentz's remaining loose ends: he fixed the transformation's scale factor to unity, proved that the transformations compose — closing them into what he christened "the Lorentz group" — added the charge–current transformation law, and observed that $x^2+y^2+z^2-c^2t^2$ is the invariant, even noting that with $ict$ the group looks like a rotation group in four dimensions (the remark Minkowski would turn into geometry). He had already, in 1898–1904, questioned absolute simultaneity and stated a "principle of relativity" for all physics. And yet he kept the ether all his life, treating local time as a convention and the contraction as a dynamical hypothesis — holding the complete formalism while declining its radical reading. The standard verdict (Pais's): Poincaré possessed the mathematics of relativity; Einstein possessed the physics. The pairing is the century's cleanest case study in the difference — and Poincaré's group-theoretic framing, ironically, became the deepest modern statement of the theory: relativistic physics is representation theory of the group he named.

History

1887

— Voigt publishes wave-equation-preserving substitutions: the boost's first draft, conformal factor and all

1892

— Lorentz introduces contraction and, by 1895, "local time" t − vx/c² as calculational devices

1904

— Lorentz's final transformation equations: Maxwell's equations survive in moving frames

1905

— Poincaré fixes the scale, proves closure, names the Lorentz group; Einstein derives the same formulas from two postulates and means them

1908

— Minkowski geometrizes: boosts are rotations of spacetime (next article's subject)

1938

— Ives–Stilwell measure the transverse Doppler shift: the γ in the time equation, seen directly

1971

— Hafele–Keating fly atomic clocks around the world: dilation with airline tickets

2014

— stored-ion tests confirm time dilation at 2×10⁻⁹ precision at β = 0.34

The physical picture

Think of the transformation as a lock fitted to a known key. The key is the wave operator $\Box=\partial_x^2-\tfrac{1}{c^2}\partial_t^2$ (one spatial dimension suffices for a boost along $x$ ): the object Maxwell's equations hand every frame, the machine whose null solutions travel at $c$ . The postulates say every inertial observer holds the same key — so the coordinate change between two observers must be a motion that leaves the lock unchanged. Galileo's map failed (the cross term); the question is what succeeds. There is a perfect Euclidean warm-up. Rotations are exactly the linear maps leaving $\partial_x^2+\partial_y^2$ (equivalently $x^2+y^2$ ) unchanged — circles map to circles, and the one-parameter family is $x'=x\cos\theta+y\sin\theta$ , etc. Our lock differs by one sign: $(ct)^2-x^2$ instead of $x^2+y^2$ . The level sets are hyperbolas instead of circles, so the invariance group must slide points along hyperbolas — and the trigonometry that parametrizes such sliding is hyperbolic: $\cosh$ and $\sinh$ in place of $\cos$ and $\sin$ , with an "angle" (the rapidity) that adds under composition just as rotation angles do. A boost is a rotation between time and space, performed with the only functions that respect the minus sign. That one flipped sign is the entire difference between geometry and kinematics — and because $\cosh^2\phi-\sinh^2\phi=1$ has no periodicity, you can rotate forever without returning: rapidity is unbounded while $\beta=\tanh\phi$ saturates at 1. The speed limit is the asymptote of a hyperbola.

Figure

One sign, two geometries — rotations preserve circles, boosts slide events along hyperbolas toward the light-cone asymptotes

Who found the key first matters for how you should think about it. Voigt (1887) wrote down wave-equation-preserving substitutions while studying Doppler physics; Lorentz (1892–1904) constructed the transformation so that Maxwell's electron theory would survive in moving frames, discovering "local time" $t-vx/c^2$ as a formal trick he never quite believed was time; Poincaré (1905) fixed the last coefficients, proved the group property, and named the group. Einstein's contribution was not the formulas but their meaning: derived from two postulates about space and time themselves (the mechanics companion's route), they apply to everything, not just electrodynamics — the local time is the time. This article's derivation follows the historical, electrodynamic road because this is the electrodynamics track: the transformation is, first and forever, the symmetry of Maxwell's wave operator.

The math: the boost as the wave operator's symmetry

Set up the demand precisely. A boost along $x$ cannot touch $y,z$ (isotropy about the motion axis, plus a symmetry argument: transverse contraction would let two observers each see the other's meter stick shrink transversally, contradicting a head-on comparison). Homogeneity of space and time forces linearity (a nonlinear map would make identical experiments at different places transform differently). So the general candidate is

ct'=A\,ct+B\,x,\qquad x'=C\,ct+D\,x, \qquad (1)

where: four dimensionless constants $A,B,C,D$ may depend on the relative speed $v$ only. The physics enters as three requirements: the primed origin ( $x'=0$ ) moves at $x=vt$ ; the wave operator is form-invariant (Postulate II doing its work through Maxwell); and the map composes sensibly with its own inverse (Postulate I — no frame is special). Watch all four constants fall.

Derivation AThe boost from invariance of the wave operator

The goal is the Lorentz boost, derived as the linear map under which $\Box=\partial_x^2-\tfrac{1}{c^2}\partial_t^2$ transforms into $\partial_{x'}^2-\tfrac{1}{c^2}\partial_{t'}^2$ with no leftover terms — the exact repair of the Galilean failure exhibited last article, and the route by which the transformation was historically first found. Strategy: chain-rule the operator through the general linear map (1), collect the three second-derivative structures, and demand coefficients $(1,-\tfrac{1}{c^2},0)$ ; the origin-motion condition and the $v\to0$ limit then pin signs and the last constant. (For readers keeping score against the mechanics companion: that article derives the same boost from light-pulse kinematics, and verifies interval invariance afterward; here invariance is the starting demand and the boost is what survives it. Two independent roads, one transformation — that is the whole point.)

Step 1. Chain rule through (1). With $u\equiv ct$ , the map is $u'=Au+Bx$ , $x'=Cu+Dx$ , so

\frac{\partial}{\partial u}=A\frac{\partial}{\partial u'}+C\frac{\partial}{\partial x'},\qquad \frac{\partial}{\partial x}=B\frac{\partial}{\partial u'}+D\frac{\partial}{\partial x'}.

Step 2. Assemble the operator and collect structures:

\frac{\partial^2}{\partial x^2}-\frac{\partial^2}{\partial u^2} =(D^2-C^2)\frac{\partial^2}{\partial x'^2}-(A^2-B^2)\frac{\partial^2}{\partial u'^2}+2(BD-AC)\frac{\partial^2}{\partial u'\partial x'}.

Form invariance demands

A^2-B^2=1,\qquad D^2-C^2=1,\qquad AC=BD.

Why these three conditions are the whole story: the first two normalize the time-time and space-space parts; the third kills the cross term — precisely the term whose survival convicted Galileo. Note the structure: these are the hyperbolic analogues of the orthogonality conditions $A^2+B^2=1$ , etc., that define rotations — same equations, one sign flipped, exactly as the picture section promised.

Step 3. Feed in the physics of relative motion. The primed spatial origin $x'=0$ is the moving observer, at $x=vt=\beta u$ : from $x'=Cu+Dx=0$ along that worldline, $C=-\beta D$ . Substituting into $D^2-C^2=1$ :

D=\frac{1}{\sqrt{1-\beta^2}}\equiv\gamma,\qquad C=-\gamma\beta,

choosing the root that reduces to the identity ( $D\to+1$ ) as $v\to0$ . The condition $AC=BD$ with $A^2-B^2=1$ then forces $A=\gamma$ , $B=-\gamma\beta$ (same root convention).

Step 4. The boost, assembled — and audited:

ct'=\gamma\,(ct-\beta x),\qquad x'=\gamma\,(x-\beta ct),\qquad y'=y,\qquad z'=z.

Audit one: $v\to0$ gives the identity. Audit two: $\beta\ll1$ with $x\sim vt$ scales recovers $x'\approx x-vt$ , $t'\approx t$ — Galileo as the slow shadow. Audit three: the map preserves $u^2-x^2$ by construction (the operator and the quadratic form share their symmetry group), so a light pulse $x=ct$ maps to $x'=ct'$ : Postulate II, closed. Audit four — the group property (Postulate I): composing two boosts must give a boost; it does, and the composition law is the rapidity addition owned by the mechanics companion, recalled below as a statement. Why $\gamma$ was inevitable: the moment the cross term must die while the origin moves, the normalization $D^2-C^2=1$ has no choice but $\gamma$ . The factor that runs all relativistic phenomenology is a hyperbola's protractor reading — algebra, not hypothesis. *Verified: matches the standard boost (Griffiths Eqs. 12.18–12.21, Jackson Eq. 11.16); the operator-invariance route reproduces Voigt's 1887 substitution logic modulo his overall scale (fixed here by the group/identity condition — Voigt's version carried a spurious conformal factor, the one loose end Poincaré's group argument nails).*

The matrix packaging, for the chapters that will live on it. Writing $x^\mu=(ct,x,y,z)$ and the boost as $x'^\mu=\Lambda^\mu{}_\nu x^\nu$ :

\Lambda^\mu{}_\nu=\begin{pmatrix}\gamma&-\gamma\beta&0&0\\-\gamma\beta&\gamma&0&0\\0&0&1&0\\0&0&0&1\end{pmatrix},\qquad \Lambda^{\mathsf{T}}\eta\,\Lambda=\eta,\qquad \eta_{\mu\nu}=\mathrm{diag}(+1,-1,-1,-1), \qquad (2)

where: the condition $\Lambda^{\mathsf{T}}\eta\Lambda=\eta$ — the statement that $\Lambda$ preserves the quadratic form $\eta$ builds — defines the Lorentz group (boosts, rotations, and their products); its derivation from interval invariance, and the geometry of $\eta$ itself, are the next article's and the mechanics companion's business respectively. What matters operationally: det $\Lambda=\gamma^2-\gamma^2\beta^2=1$ , the inverse is the same matrix with $\beta\to-\beta$ , and — the track's coming workhorse — anything with four components transforming by this same $\Lambda$ (the four-current, the four-potential, and, doubly, the field tensor) inherits every kinematic theorem automatically. In rapidity form, $\gamma=\cosh\phi$ , $\gamma\beta=\sinh\phi$ with $\tanh\phi=\beta$ : the boost is literally the hyperbolic rotation of the picture section, and collinear rapidities add, $\phi_{12}=\phi_1+\phi_2$ (derived via matrix composition in the mechanics companion; used freely here).

Go deeper · Derivation B · 1 minVelocity addition by differentials, and the fixed point at c

Derivation BVelocity addition by differentials, and the fixed point at c

The goal is the relativistic velocity-addition law, derived directly from the boost by differentials — the third route in this pair of tracks' constellation (the mechanics companion gets it from rapidity addition and again from an invariant contraction; the differential route is the standard workhorse and the one that makes the fixed-point structure most visible). Strategy: differentiate the boost, form $u'=dx'/dt'$ , and audit the two limits that matter — Galileo below, light speed as a fixed point above.

Step 1. Differentials of the boost (object moving with $u=dx/dt$ in $S$ ; frame $S'$ at speed $v$ ):

dx'=\gamma(dx-v\,dt),\qquad dt'=\gamma\left(dt-\frac{v}{c^2}dx\right).

Step 2. Divide, and cancel the $\gamma$ 's:

u'=\frac{dx'}{dt'}=\frac{dx-v\,dt}{dt-\dfrac{v}{c^2}dx}=\frac{u-v}{1-\dfrac{uv}{c^2}}.

Why the denominator is the relativity of simultaneity in disguise: the $-(v/c^2)dx$ in $dt'$ is the tilt of the moving observer's "now" across the distance $dx$ — the first-order effect Galilean intuition omits. Every departure from naive addition traces to that term.

Step 3. The Galilean audit: $uv/c^2\to0$ gives $u'=u-v$ . For lab speeds the correction is invisible — two 30 m/s cars disagree with Galileo by one part in $10^{14}$ .

Step 4. The fixed point. Set $u=c$ :

u'=\frac{c-v}{1-\dfrac{v}{c}}=\frac{c\,(c-v)}{c-v}=c,

for every $v$ . Light's speed is not approximately preserved — it is a fixed point of the composition law, algebraically exact. And the fixed point is a ceiling: composing any two sub- $c$ speeds, $u',u<c$ , keeps the result below $c$ (clearest in rapidity language, where speeds below $c$ are finite rapidities and $\tanh$ of any finite sum stays below 1). The postulate that demanded the transformation is enforced by the transformation, everywhere, automatically — the loop the 1905 paper closes. Verified: matches Griffiths Eq. 12.31 and Jackson Eq. 11.31; the worked ladder below checks $0.5c\oplus0.5c\oplus0.5c=0.9286c$ against direct rapidity addition — two methods, same digits.

The kinematic read-offs, for completeness and with their ownership labels: a clock fixed in $S'$ ( $\Delta x'=0$ ) has $\Delta t=\gamma\Delta\tau$ — time dilation; a rod at rest in $S'$ measured at one instant in $S$ ( $\Delta t=0$ ) gives $L=L_0/\gamma$ — length contraction. Both are two-line consequences of the boost (invert the appropriate pair, set the appropriate difference to zero), and both are derived, with their full simultaneity bookkeeping and their experimental record, in the mechanics companions — the light-clock construction, the rod protocol, the muon and GPS budgets. This track states them, uses them, and moves toward its own quarry: what the boost does to fields.

Figure

The rapidity dial — speed saturates at c while the additive boost parameter runs forever

Limiting cases

case	result	meaning
$\beta\to0$	$\Lambda\to$ identity; $x'\approx x-vt$ , $t'\approx t$	Galileo is the tangent to the Lorentz group at rest
$\gamma$ expansion	$\gamma=1+\tfrac12\beta^2+\tfrac38\beta^4+\cdots$	leading corrections are second order — why $10^{-8}$ effects hid until 1887
$u=c$ in addition	$u'=c$ exactly	the fixed point: light stays light under every boost
$\beta\to1$	$\gamma\to\infty$ , $\phi\to\infty$	speeds saturate; rapidity does not — the hyperbola's asymptote
lightlike events	$x=ct\Rightarrow x'=ct'$	the cone maps to itself: geometry of Postulate II
composition	$\phi_{12}=\phi_1+\phi_2$ (collinear)	boosts form a group; velocities add through $\tanh$ (mechanics companion)
what transforms with $\Lambda$	$x^\mu$ , $J^\mu$ , $A^\mu$ , $\partial_\mu$ ; $F^{\mu\nu}$ doubly	one matrix runs all of covariant electrodynamics (articles ahead)

Common trapgamma multiplies the whole bracket — and every paradox is a dropped simultaneity term

The two workhorse errors. First, algebraic: applying $\gamma$ to the naive Galilean form — writing $t'=\gamma t$ or $x'=\gamma(x-vt)$ while keeping $t'=t$ — instead of transforming the full pair. The time equation's $-\gamma\beta x/c$ term (Lorentz's "local time") is not decoration: it carries the relativity of simultaneity, it is first order in $\beta$ (larger, for spread-out events, than the famous second-order dilation), and it is the term whose omission generates every classic "paradox." Pole-in-the-barn, the twins, "each sees the other's clock slow — contradiction?": all dissolve the moment events are transformed with both equations and "simultaneous" is treated as frame-tagged. Second, conceptual: reading dilation and contraction as mechanical effects on clocks and rods — something the motion does to the mechanism. They are properties of the coordinate grids, applying identically to atomic clocks, biological aging, and muon decay; the mechanics companion's constructions make this explicit. Rule of thumb: never reason with shortcut formulas when two or more events are spatially separated — transform the events, then read off whatever was asked.

Worked examples

1 · Light stays light — a radar pulse through a moving frame (confidence builder)

Setup. In the lab frame $S$ , a radar pulse leaves the origin at $t=0$ and is detected at $x=299.8$ km, $t=1.000$ ms (i.e. on the light line $x=ct$ ). Transform both events to the frame $S'$ of a spacecraft receding at $v=0.6c$ , and audit the postulate.

Step 1. The machinery: $\beta=0.6$ , $\gamma=1/\sqrt{1-0.36}=1.25$ . Event A $(ct,x)=(0,0)$ maps to $(0,0)$ — shared origin. Event B: $ct=299.8$ km, $x=299.8$ km.

Step 2. Transform event B:

ct'=\gamma(ct-\beta x)=1.25\,(299.8-0.6\times299.8)=1.25\times0.4\times299.8=149.9\ \text{km},

x'=\gamma(x-\beta ct)=1.25\times0.4\times299.8=149.9\ \text{km}.

Step 3. Audit. In $S'$ : $x'=ct'$ exactly — the pulse still travels at $c$ , as the fixed-point algebra guaranteed. Both coordinates shrank by the same factor $0.5$ ; that common factor is no accident: $\sqrt{\dfrac{1-\beta}{1+\beta}}=\sqrt{\dfrac{0.4}{1.6}}=0.5$ is the relativistic Doppler factor, and what we have computed is the elementary fact beneath it — intervals along a light line rescale by $k=\sqrt{(1-\beta)/(1+\beta)}$ under a boost (the " $k$ -calculus" seed; the full Doppler/aberration story for waves is owned by its own article downstream).

Step 4. Contrast the Galilean prediction: $x'=x-vt=119.9$ km with $t'=1$ ms would give speed $0.4c$ . The interval check: $s^2=(ct)^2-x^2=0$ in both frames — null stays null; the invariant the next article builds its geometry on is already doing its work.

The lesson: transform events, audit with the invariant, and the postulates take care of themselves — light cannot be caught by boosting after it.

2 · The rapidity ladder — why three half-lightspeed boosts do not make 1.5c (exam level)

Setup. A staging rocket performs three successive boosts, each of $0.5c$ relative to the frame it just left. Find the final speed relative to the launch frame, both by iterated velocity addition and by rapidity, and expose why the naive $1.5c$ fails.

Step 1. Iterate the addition law. First two stages: $u_2=\dfrac{0.5c+0.5c}{1+0.25}=0.8c$ . Third: $u_3=\dfrac{0.8c+0.5c}{1+0.40}=\dfrac{1.3c}{1.40}=\mathbf{0.9286c}$ .

Step 2. Same ladder in rapidity (the additive variable — additivity derived in the mechanics companion, used here). Each stage contributes $\phi=\mathrm{artanh}(0.5)=0.5493$ ; three stages give $\phi_{\text{tot}}=1.648$ , and

u=c\tanh(1.648)=\mathbf{0.9286c}.

Two routes, same four digits — and the second took one line. Rapidity is to boosts what angle is to rotations: the bookkeeping that makes composition trivial.

Step 3. Where the missing $0.57c$ went. Each successive boost is applied to a frame whose clocks and simultaneity are already tilted relative to launch; the denominators $1+uv/c^2$ are precisely that tilt's toll. In rapidity language nothing is lost at all — rapidities add to $1.648$ without ceiling — but speed is the $\tanh$ of rapidity, and $\tanh$ saturates. A rocket can accumulate unbounded rapidity (and unbounded momentum and energy, as the four-momentum article will price it) while its speed creeps asymptotically toward $c$ .

Step 4. The engineering corollary. Continuous acceleration at constant proper $a$ builds rapidity linearly in proper time, $\phi=a\tau_p/c$ : after one year at $1\,g$ , $\phi\approx1.03$ , $u=0.77c$ ; after five years, $\phi\approx5.2$ , $u=0.9999c$ — never $c$ , with the shortfall shrinking as $2e^{-2\phi}$ . The speed limit is not a force barrier the rocket fights; it is the geometry of $\tanh$ .

The lesson: velocities compose through the hyperbolic tangent of an additive angle — use rapidity for any multi-boost problem and the "paradoxes" of composition become one-line sums.

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., Ch. 12 · Jackson, Classical Electrodynamics, 3rd ed., Ch. 11 · Rindler, Introduction to Special Relativity, 2nd ed., Ch. 2 · Zangwill, Modern Electrodynamics, Ch. 22 · Purcell & Morin, Electricity and Magnetism, 3rd ed., App. G

Provenance. Boost equations and $\gamma$ verified against Griffiths Eqs. 12.18–12.21 and Jackson Eq. 11.16; velocity addition against Griffiths Eq. 12.31/Jackson 11.31; $\det\Lambda=\gamma^2(1-\beta^2)=1$ and interval preservation by direct substitution. A verifier should: (1) re-run the chain-rule derivation and confirm the three coefficient conditions; (2) check the $v\to0$ and Galilean limits; (3) verify $u=c$ is a fixed point and recompute the rapidity ladder $\mathrm{artanh}(0.5)=0.5493$ , $\tanh(1.648)=0.9286$ ; (4) confirm the light-line rescaling factor $\sqrt{(1-\beta)/(1+\beta)}=0.5$ at $\beta=0.6$ . Addendum for this edition: this is the electrodynamics-track treatment — Derivation A (boost as the wave operator's symmetry, the Voigt–Lorentz–Poincaré road) and Derivation B (velocity addition by differentials with the fixed-point audit) are owned here; the postulate-route derivation of the same boost, the interval-invariance verification with the hyperbolic-rotation reading, and the rapidity-additivity composition proof are owned by the mechanics companion Special Relativity Postulates and the Lorentz Transformation (its Derivations A–C) and recalled by name; the kinematic constructions and their experimental record live in Relativistic Kinematics (mechanics); $\Lambda^{\mathsf{T}}\eta\Lambda=\eta$ as a derived group condition belongs to Four-Vectors and Minkowski Spacetime (mechanics) and the geometry to the next article here. Metric signature $(+,-,-,-)$ locked track-wide. ⚠ Slug note: `lorentz-transformations` collides with an old physics-wiki row (recon §c-A, REPLACE at cutover; the aq- prefix protects until then); the mechanics twin uses the singular slug `aq-mech-lorentz-transformation` — different slugs, cross-linked as companions. Worked numbers for this edition (event transform at $\beta=0.6$ : $k=0.5$ both coordinates; ladder $0.5c^{\oplus3}=0.9286c$ two routes; $1\,g$ -year rapidity $1.03\to0.77c$ ; $\gamma$ table spot-checked) recomputed for this edition.

Builds on: Postulates of Special Relativity · Electromagnetic Wave Equation Leads to: Spacetime and Minkowski Geometry · Four-Vectors and the Metric Tensor · Transformation of Electric and Magnetic Fields · The Four-Current Density · The Four-Potential

Voigt had the formulas in 1887. Why do we not call them the Voigt transformations?

Voigt's substitutions preserve the wave equation but with an overall conformal factor — his version multiplies the interval by a $v$ -dependent scale rather than preserving it exactly, and he treated the result as a formal property of the wave equation, useful for Doppler arguments, not as a statement about space and time. He never imposed the group/reciprocity condition that fixes the scale to unity (Derivation A's Step 3–4 root choices), never applied it to electrodynamics at large, and did not pursue priority even when Lorentz later acknowledged the near miss. The naming follows the physics: Lorentz built the transformation into a working theory of electrodynamics in moving frames; Poincaré proved its group structure and named it; Einstein supplied the meaning. Voigt's near-miss is a standing lesson in how much of a discovery is the reading rather than the formula — the same lesson as Liénard's pre-relativistic $\gamma^6$ earlier in this track.

Is deriving the boost from the wave operator circular — using light to define the transformation that then governs light?

Not circular, but the logical structure deserves care. The derivation takes as input: linearity (from homogeneity), isotropy, and the existence of one invariant speed — supplied empirically by electrodynamics via Postulate II. It outputs the unique transformation consistent with that input. Two independent checks break any apparent circle. First, the mechanics companion derives the identical boost from light-pulse kinematics with different intermediate machinery — and a deeper analysis (von Ignatowsky, 1910) shows relativity-plus-homogeneity alone forces a Lorentz-form transformation with some invariant speed, electrodynamics only fixing its value. Second, the output governs far more than light: the same boost is verified on massive-particle kinematics (collider energy-momentum), muon lifetimes, and atomic clocks — systems with no special relation to electromagnetism. Light was the ladder; the structure stands without it, which is also why a hypothetical photon mass would strand light below the invariant speed while leaving the transformation intact.

If each observer sees the other's clock run slow, who is right?

Both — and the question dissolves once "sees the other's clock run slow" is unpacked into events. Comparing clock rates between frames requires one clock to be checked against a sequence of spatially separated clocks in the other frame, and those clocks' synchronization is exactly what the frames disagree about (the $-\gamma\beta x/c$ term). Observer $A$ checks $B$ 's one clock against two of $A$ 's synchronized clocks and finds it slow; $B$ does the mirror-image protocol and finds $A$ 's slow; the two protocols involve different event triples, so no contradiction is available to state. The situation is precisely analogous to two people walking at an angle, each seeing the other "fall behind" along their own direction of march — projection asymmetry, not paradox. It becomes a physical asymmetry only when the protocols are forced onto a common footing, e.g. one clock turns around and returns (the twin scenario): the turnaround breaks the symmetry, and the accounting — owned by the mechanics kinematics article — gives an unambiguous, frame-independent answer.

Why do boosts along different directions fail to commute, and does it matter?

Because hyperbolic rotations in different time-space planes are still rotations in a group where order matters — just as $90°$ rotations about $x$ then $y$ differ from $y$ then $x$ . Two non-collinear boosts compose not into a pure boost but into a boost plus a spatial rotation — the Wigner rotation — and for an orbiting frame the accumulated rotation per period is the Thomas precession. This is not a curiosity: it contributes a factor that halves the naive spin-orbit coupling in hydrogen (the "Thomas half" that puzzled atomic physics in 1926), it enters storage-ring spin dynamics (the physics of muon $g-2$ experiments), and it is the first taste of a structural fact the track will keep meeting: the Lorentz group is non-abelian, and its representation theory — not just the single boost matrix — is what organizes relativistic physics. For collinear compositions (this article's ladder), none of this activates: rapidities along one axis simply add.

Does length contraction mean I would SEE a fast rod as shortened?

Measurement and sight are different operations, and the distinction is a classic trap. Measured length — both ends registered at one instant in your frame — is genuinely $L_0/\gamma$ : that is the coordinate statement, and particle physics uses it (the LHC's proton bunches, and the nuclei in heavy-ion collisions, are dynamically flattened pancakes in the lab frame). Seen length — photons arriving at your eye simultaneously — mixes contraction with light-travel-time differences from the near and far parts of the object, and the composite effect (Terrell–Penrose, 1959) is startling: a receding sphere still presents a circular outline at every speed, and a fast rod appears rotated rather than squashed. Photographs of relativistic objects show geometry sheared by retardation, not the clean $1/\gamma$ . The moral generalizes across the whole track: relativistic statements are about coordinate assignments to events; converting them to optical appearances always requires a second, separate retardation calculation — the same distinction the radiation chapter drew between emitted and received power.

What exactly will transform with this matrix when we reach the electromagnetic payoff?

The full inventory, with its logic: anything whose four components are defined by how coordinates are defined must transform by $\Lambda$ . The four-gradient $\partial_\mu=(\partial_{ct},\nabla)$ — immediately, since it is built from the coordinates. The four-current $J^\mu=(c\rho,\mathbf{J})$ — because charge density and current density are one object's time and space faces (a boosted static charge is a current; this is the four-current article's opening). The four-potential $A^\mu=(V/c,\mathbf{A})$ — because the Lorenz-gauge wave equations $\Box A^\mu=\mu_0J^\mu$ pair an invariant operator with a four-vector source, forcing $A^\mu$ 's vector character (the retarded-potential machinery of the last chapter was secretly covariant all along). And then the crown: the fields are antisymmetrized derivatives of $A^\mu$ , so $F^{\mu\nu}=\partial^\mu A^\nu-\partial^\nu A^\mu$ transforms with two factors of $\Lambda$ — six independent components, which are exactly $\mathbf{E}$ and $\mathbf{B}$ , mixing under boosts by the rules that dissolve the magnet-and-conductor asymmetry quantitatively. One matrix, learned once, runs everything this track does from here to the covariant Maxwell equations.